Download The Evolution of Star Formation Activity in . Cory R. Wagner

The Evolution of Star Formation Activity in
Cluster Galaxies Over 0.15 < z < 1.5
by
Cory R. Wagner
A thesis submitted to the
Department of Physics, Engineering Physics & Astronomy
in conformity with the requirements for
the degree of Master of Science
Queen’s University
Kingston, Ontario, Canada
September 2014
Copyright © Cory R. Wagner, 2014
Abstract
In this thesis, we explore 7.5 billion years of evolution in cluster galaxy star formation activity using a sample of 11 high-redshift (1 < z < 1.5) clusters from the
IRAC Shallow Cluster Survey, and 25 low-redshift (0.15 < z < 1) clusters from The
Cluster Lensing And Supernova survey with Hubble. We compare cluster galaxy star
formation to that of the field over 0.15 < z < 1.5 using ∼8000 galaxies from the
UltraVISTA survey. Mid-infrared star formation rates are measured using Spitzer
24 µm data for isolated high-redshift galaxies. We calculate rest-frame ultraviolet
star formation rates for low-redshift cluster members using Hubble Space Telescope
observations. Using publically available mid-infrared and ultraviolet data for our field
sample, we empirically derive scaling relations to adjust low-redshift cluster galaxy
ultraviolet star formation rates to mid-infrared levels. We classify cluster galaxy morphology by visual inspection, and use quantitatively measured morphologies for field
galaxies. Cluster late-type galaxies at z > 1 show enhanced star formation activity
relative to the field, and account for nearly 90% of the overall star formation activity in high-redshift clusters. While high-redshift early-type galaxies are substantially
quenched relative to cluster late-types, they still contribute ∼13% of the total cluster
star formation activity. With early-type fractions increasing from 34 to 56% from
z ∼ 1.5 → 1.16, we find that new cluster early-type galaxies are likely being formed
i
around z ∼ 1.4. The fraction of early-type galaxies that are star-forming drops from
29 to 11% over this period, yet their specific star formation rates are roughly constant.
These factors suggest that the events that created these new galaxies, possibly mergers, were both recent and gas-rich. With typical coverages of 50% of z < 1 cluster
virial radii, we can only probe the cores of low-redshift clusters. We find that in this
regime, the star formation activity of cluster galaxies is quenched relative to the field.
We compare the mean star formation rate of cluster galaxies to the results of Alberts
et al. (2014), who fit the mean star formation rate evolution over 0.3 < z < 1.5,
and measured star formation rates by stacking 250 µm Herschel images. We find
excellent agreement between the Herschel-based fit and both our Spitzer-derived and
ultraviolet→infrared star formation rates.
ii
Statement of Co-Authorship
The research presented in this thesis was done under the supervision of Stéphane
Courteau (Queen’s University), and Mark Brodwin (University of Missouri-Kansas
City). All the work presented here was done by the author (Cory R. Wagner) except
where explicitly stated otherwise.
Chapter 2 contains a version of a paper submitted to The Astrophysical Journal
entitled: “Star Formation in High-Redshift Cluster Ellipticals” by Cory R. Wagner,
Mark Brodwin, Gregory F. Snyder, Anthony H. Gonzalez, S. A. Stanford, Stacey
Alberts, Alexandra Pope, Daniel Stern, Gregory R. Zeimann, Ranga-Ram Chary,
Arjun Dey, Peter R. M. Eisenhardt, Conor L. Mancone, and John Moustakas. I am
the lead author on this paper, and performed all the analysis, wrote the manuscript,
and created all the figures and tables. All observations and reductions were performed
by others. Gregory F. Snyder provided the red-sequence data from Snyder et al.
(2012). Mark Brodwin provided IRAC Shallow Cluster Survey data, and star forming
fractions from Brodwin et al. (2013). Stacey Alberts provided mean star formation
rates from Alberts et al. (2014), and attempted to derive 250 µm star formation rates
for the IRAC Shallow Cluster Survey sample in this paper. S. A. Stanford, Anthony
H. Gonzalez, and Daniel Stern assisted with morphological classifications on a subset
of ISCS member galaxies.
iii
Acknowledgments
Stéphane Courteau, for his encouragement.
Mark Brodwin, for a great project.
Mike McDonald, for his insights during the latter half of this thesis.
Dan McIntosh, for introducing me to the world universe of astrophysics.
My thesis defense committee (Stéphane Courteau, David Hanes, Danilo Marchesini,
Alastair McLean) for their time and effort in reading through this thesis, and providing valuable comments that enhanced the quality of the work herein.
This work is based in part on observations made with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of
Technology under a contract with NASA. Support for this work was provided by
NASA through an award issued by JPL/Caltech. Support for HST programs 10496,
11002, 11597, and 11663 were provided by NASA through a grant from the Space
Telescope Science Institute, which is operated by the Association of Universities for
Research in Astronomy, Inc., under NASA contract NAS 5-26555. This work is based
in part on observations obtained with the Chandra X-ray Observatory, under contract
SV4-74018, A31 with the Smithsonian Astrophysical Observatory which operates the
Chandra X-ray Observatory for NASA.
iv
Contents
Abstract
i
Statement of Co-Authorship
iii
Acknowledgments
iv
Table of Contents
v
List of Tables
viii
List of Figures
ix
List of Common Acronyms
xii
Chapter 1:
Introduction
1.1 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Do Early-Type Galaxies Share in the Wealth of Star Formation Activity in High-Redshift Clusters? . . . . . . . . . . . . . . . . . . . . . .
1.3 Low-Redshift: Where Have All The Star Formers Gone? . . . . . . .
Chapter 2:
2.1
2.2
2.3
Paper I: Star Formation in High-Redshift
lipticals
Paper I Organization . . . . . . . . . . . . . . . . . . .
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 IRAC Shallow Cluster Survey . . . . . . . . . .
2.2.2 HST Data . . . . . . . . . . . . . . . . . . . . .
2.2.3 Mid-Infrared Data . . . . . . . . . . . . . . . .
2.2.4 Chandra X-ray Data . . . . . . . . . . . . . . .
Galaxy Selection Method . . . . . . . . . . . . . . . . .
2.3.1 Identification of Cluster Members . . . . . . . .
2.3.2 Rejection of AGNs . . . . . . . . . . . . . . . .
2.3.3 Stellar Masses and Mass Limit . . . . . . . . . .
v
1
1
2
4
Cluster El.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6
6
7
7
7
9
10
10
11
11
12
.
.
.
.
.
.
.
.
.
.
.
.
.
12
13
16
17
17
18
20
23
25
26
28
30
34
.
.
.
.
.
.
.
37
37
38
39
39
42
43
47
.
.
.
.
.
48
49
49
50
52
53
Chapter 5:
Infrared-Ultraviolet Star Formation Rate Comparison
5.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Empirical Fits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
56
60
Chapter 6:
Analysis
6.1 Projected Radius . . . . . . . . . . . . . . . . . . . . .
6.2 Galaxy Morphology Versus Projected Radius . . . . . .
6.3 Star Formation Rate Versus Projected Radius . . . . .
6.4 Mean Star Formation Rate Versus Projected Radius . .
6.5 Specific Star Formation Rate Versus Projected Radius .
6.6 Galaxy Morphology Versus Redshift . . . . . . . . . . .
63
63
63
65
67
70
72
2.4
2.5
2.6
2.3.4 Isolation . . . . . . . . . . . . . . . . . . . . . .
2.3.5 Morphology . . . . . . . . . . . . . . . . . . . .
2.3.6 Comparison Field Sample Selection . . . . . . .
Analysis . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.1 Star Formation Rate vs. Radius . . . . . . . . .
2.4.2 Mean Star Formation Rate . . . . . . . . . . . .
2.4.3 Fraction of Star-forming Galaxies . . . . . . . .
2.4.4 Specific Star Formation Rate . . . . . . . . . .
Discussion . . . . . . . . . . . . . . . . . . . . . . . . .
2.5.1 Comparison to Brodwin et al. (2013) . . . . . .
2.5.2 Comparison to Alberts et al. (2014) . . . . . . .
2.5.3 Star Formation in High-Redshift Cluster ETGs
Conclusions . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
Chapter 3:
Cluster Galaxy Data and Sample Selection
3.1 High-Redshift ISCS Clusters . . . . . . . . . . . . . . . .
3.2 Low-Redshift CLASH Clusters . . . . . . . . . . . . . . .
3.2.1 Redshifts . . . . . . . . . . . . . . . . . . . . . .
3.2.2 Stellar Masses . . . . . . . . . . . . . . . . . . . .
3.2.3 Star Formation Rates . . . . . . . . . . . . . . . .
3.2.4 Cluster Membership . . . . . . . . . . . . . . . .
3.2.5 Galaxy Morphology . . . . . . . . . . . . . . . . .
Chapter 4:
Field Galaxy Data and Sample Selection
4.1 Redshifts . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Star Formation Rates . . . . . . . . . . . . . . . . . .
4.3 Stellar Masses . . . . . . . . . . . . . . . . . . . . . .
4.4 Galaxy Morphology . . . . . . . . . . . . . . . . . . .
4.5 Stellar Masses of Final Galaxy Samples . . . . . . . .
vi
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6.7
6.8
Mean Star Formation Rate Versus Redshift . . . . . . . . . . . . . . .
Specific Star Formation Rate Versus Redshift . . . . . . . . . . . . .
75
79
Chapter 7:
Summary and Conclusions
7.1 What’s Next? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
86
Bibliography
88
vii
List of Tables
2.1
ISCS Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8
3.1
CLASH Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
5.1
Coefficients of z < 0.1 fits of IR-derived SFR verus UV-derived SFR .
59
6.1
Comparison Cluster ETG Fractions From the Literature . . . . . . .
73
viii
List of Figures
2.1
500 ×500 cutouts of five isolated high-redshift cluster ETGs . . . . . . .
14
2.2
SFR versus cluster-centric radius for high-redshift ETGs and LTGs .
17
2.3
Mean SFR versus cluster-centric radius for high-redshift cluster ETGs
and LTGs, and mean SFR of high-redshift field ETGs . . . . . . . . .
2.4
19
Fraction of star-forming galaxies versus projected radius for high-redshift
cluster ETGs, and low-redshift cluster galaxies, and fraction of starforming high-redshift field galaxies . . . . . . . . . . . . . . . . . . .
2.5
21
Specific SFR versus radius for high-redshift cluster ETGs and lowredshift cluster galaxies, and sSFR of high-redshift field ETGs . . . .
24
2.6
Fraction of star-forming ISCS cluster galaxies versus redshift . . . . .
27
2.7
Mean SFR versus redshift for ISCS cluster galaxies . . . . . . . . . .
29
2.8
Upper panel: Specific SFR as a function of redshift for ISCS cluster
ETGs, and field ETGs and LTGs. Lower panel: ETG fraction versus
redshift for ISCS clusters . . . . . . . . . . . . . . . . . . . . . . . . .
3.1
32
Redshifted rest-frame UV window, and rest-frame g- and i-band filters.
Best-fit observed-frame HST filters. Upper panel: z = 0.348 Lower
panel: z = 0.686 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix
41
3.2
Distribution of number of observed filters, and number of filters with
5 σ detections, for spectroscopically confirmed CLASH members. . . .
44
3.3
Spectroscopic and photometric redshift distributions for CLASH clusters. 46
4.1
Comparison between stellar masses derived with SED fitting and colormass-to-light ratios for UltraVISTA galaxies. . . . . . . . . . . . . . .
4.2
51
Stellar mass and median stellar mass versus redshift for CLASH and
ISCS cluster galaxies, and UltraVISTA field galaxies, of varying morphologies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1
54
IR-derived SFR versus UV-derived SFR for z < 0.1 galaxies from UltraVISTA, Rosa-González et al. (2002), and Iglesias-Páramo et al. (2004) 58
5.2
IR-derived SFRs versus UV-derived SFRs for 0.15 < z < 1 COSMOS/UltraVISTA field galaxies . . . . . . . . . . . . . . . . . . . . .
5.3
Linear least squares fits to IR-derived SFR versus UV-derived SFR for
0.15 < z < 1 UltraVISTA field galaxies, binned by redshift . . . . . .
6.1
60
61
ETG fractions versus cluster-centric radius for CLASH and ISCS clusters. ETG fractions for 0.15 < z < 1 and 1 < z < 1.5 UltraVISTA
field galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
64
6.2
SFR versus cluster-centric radius for CLASH ETGs and LTGs . . . .
66
6.3
Mean SFR versus cluster-centric radius for CLASH and ISCS cluster
galaxies, of varying morphologies. Mean SFR of 0.15 < z < 1 and
1 < z < 1.5 UltraVISTA field galaxies, of varying morphologies . . . .
x
68
6.4
Specific SFR versus cluster-centric radius for CLASH and ISCS galaxies, of varying morphologies. Specific SFR for 0.15 < z < 1 and
1 < z < 1.5 UltraVISTA field galaxies, of varying morphologies . . . .
6.5
ETG fraction versus redshift for CLASH and ISCS clusters, UltraVISTA field, and comparison clusters from the literature . . . . . . . .
6.6
72
Mean SFR versus redshift for CLASH and ISCS cluster galaxies, and
UltraVISTA field galaxies . . . . . . . . . . . . . . . . . . . . . . . .
6.7
71
76
Mean SFR versus redshift for CLASH, ISCS, and UltraVISTA galaxies
of all morphologies, compared to fits of field and cluster mean SFR
versus redshift from Alberts et al. (2014) . . . . . . . . . . . . . . . .
6.8
Specific SFR versus redshift for CLASH and ISCS cluster galaxies, and
UltraVISTA field galaxies . . . . . . . . . . . . . . . . . . . . . . . .
6.9
78
80
Specific SFR versus redshift for CLASH, ISCS, and UltraVISTA galaxies of all morphologies, compared to fits of field and cluster sSFR versus
redshift from Alberts et al. (2014), and a fit of field sSFR versus redshift from Elbaz et al. (2011) . . . . . . . . . . . . . . . . . . . . . . .
xi
81
List of Common Acronyms
ΛCDM:
Lambda Cold Dark Matter
ACS:
Advanced Camera for Surveys
AGN:
Active galactic nucleus
CLASH:
Cluster Lensing And Supernova survey with Hubble
CMLR:
Color-mass-to-light relation
CMR:
Color-magnitude relation
COSMOS: Cosmic Evolution Survey
ETG:
Early-type galaxy
HST:
Hubble Space Telescope
IMF:
Initial mass function
IR:
Infrared
IRAC:
Infrared Array Camera
ISCS:
IRAC Shallow Cluster Survey
xii
LTG:
Late-type galaxy
MIPS:
Multiband Imaging Photometer for Spitzer
MIR:
Mid-infrared flux
M/L:
Mass-to-light
NDWFS:
NOAO Deep Wide-Field Survey
NIR:
Near-infrared
NOAO:
National Optical Astronomy Observatory
SDWFS:
Spitzer Deep, Wide-Field Survey
SED:
Spectral energy distribution
SFH:
Star-formation history
SFR:
Star formation rate
S/N:
Signal-to-noise
sSFR:
Specific star formation rate
TP-AGB: Thermally pulsing asymptotic giant branch
UV:
Ultraviolet
WFC3:
Wide Field Camera 3
WFPC2:
Wide Field Planetary Camera 2
WMAP7: Seven-Year Wilkinson Microwave Anisotropy Probe
xiii
Chapter 1
Introduction
1.1
Thesis Organization
This thesis is organized into two parts. The first (Chapter 2) is based on a manuscript
(Paper I), “Star Formation in High-Redshift Cluster Ellipticals,” submitted to The
Astrophysical Journal. It focuses solely on the 1 < z < 1.5 cluster regime, 7.8 to
9.4 Gyr ago. Part II (Chapter 3 through Chapter 6) forms the basis of what will
become Paper II, and with cluster observations spanning 0.15 < z < 1.5, we are able
to extend the morphology driven investigation described in Chapter 2 down to ∼2
Gyr ago. For clarity, and to avoid significant overlap in the material covered, we
absorb the majority of Paper I’s introduction here into Chapter 1 of this thesis. We
also note that the references in Paper I will be contained in the bibliography at the
end of this thesis.
Throughout this work, we use AB magnitudes unless otherwise indicated. We
adopt a WMAP7 cosmology (Komatsu et al. 2011), with (ΩΛ , ΩM , h) = (0.728,
0.272, 0.704).
1
1.2. DO EARLY-TYPE GALAXIES SHARE IN THE WEALTH OF
STAR FORMATION ACTIVITY IN HIGH-REDSHIFT CLUSTERS?
1.2
Do Early-Type Galaxies Share in the Wealth of Star Formation Activity in High-Redshift Clusters?
In the local Universe galaxy clusters are primarily populated by quiescent, early-type
galaxies (ETGs) with little ongoing star formation and evolved stellar populations
(Oemler 1974; Dressler 1980; Caldwell et al. 1993; Gómez et al. 2003; Bressan et al.
2006; Clemens et al. 2009; Edwards & Fadda 2011). Studies of cluster galaxy populations to z . 1 find that the evolution in the color and scatter of cluster red-sequences
is consistent with simple passive evolution models in which the bulk of galaxies’ stars
formed in a short, high-redshift starburst (Bower et al. 1992; Aragon-Salamanca et al.
1993; Stanford et al. 1998; Kodama 1999; Blakeslee et al. 2006; Mei et al. 2006, 2009;
Eisenhardt et al. 2008; Muzzin et al. 2008). However, ΛCDM predicts a more extended, hierarchical formation history. For instance, simulations by De Lucia et al.
(2006) find that only ∼50% of massive elliptical galaxies would have formed the bulk
(80%) of their stellar mass by z ∼ 1.5.
Recently, infrared (IR) measurements of the z > 1 cluster population have revealed
substantial dust-obscured star formation activity (Hilton et al. 2010; Tran et al. 2010;
Santos et al. 2013; Brodwin et al. 2013, hereafter B13). Alberts et al. (2014, hereafter A14) found that dust-obscured star formation in cluster galaxies increases with
lookback time from z = 0.3 → 1.5.
Several recent studies of the high-redshift (z > 1) clusters have been conducted
using the IRAC Shallow Cluster Survey (ISCS; Eisenhardt et al. 2008). Mancone
et al. (2010, 2012) measured the rest-frame near-infrared (NIR) luminosity function
evolution, and found that while it matched what would be expected from passive evolution up to z ∼ 1.3, it disagreed with such a model at z & 1.3, which they suggested
2
1.2. DO EARLY-TYPE GALAXIES SHARE IN THE WEALTH OF
STAR FORMATION ACTIVITY IN HIGH-REDSHIFT CLUSTERS?
as evidence for a significant epoch of galaxy assembly via merging (see, however, Andreon 2013; Wylezalek et al. 2014). In the hierarchical evolution framework, massive
ETGs are formed as the result of major mergers (Negroponte & White 1983; Barnes
1988; Naab & Burkert 2003; Cox et al. 2006), and these mergers can cause bursts of
star formation (Sanders et al. 1988; Barnes & Hernquist 1996; Hopkins et al. 2008)
and fuel an active galactic nucleus (AGN Springel et al. 2005) that quenches star
formation on the order of a few 100 Myr (Di Matteo et al. 2005; Hopkins et al. 2006).
Snyder et al. (2012, hereafter S12) found that red-sequence members have roughly
constant stellar ages across 1.0 < z < 1.5—indicating that star formation must be
ongoing—and bluer and more stochastic colors at 1.0 < z < 1.3 than would be expected of a passive population. Zeimann et al. (2013), B13, and A14 measured high,
and consistent, Hα, 24 µm, and 250 µm star formation rates (SFRs), respectively, in
these clusters.
With these high SFRs, it is clear that members in 1 < z < 1.5 clusters have
not exhausted their supplies of cold gas, which, in combination with the evidence
of significant ongoing merger activity, suggests that a substantial number of these
mergers are gas-rich. If gas-rich, major mergers are common in clusters then the
ETGs formed in these mergers would be expected to have high SFRs, at least for a
short time after their formation. Conversely, ETGs observed several hundred Myr
post-merger would likely appear to be recently quenched.
In Chapter 2, we use high-resolution Hubble Space Telescope (HST ) images of
ISCS galaxy clusters at 1 < z < 1.5 to identify isolated, early-type members, and
then measure their dust-inferred SFRs using Spitzer 24 µm data. The goal of Chapter
2 is to test whether the high SFRs seen in high-redshift clusters are merely due to the
3
1.3. LOW-REDSHIFT: WHERE HAVE ALL THE STAR FORMERS
GONE?
morphological mix—a result of the elevated late-type galaxy (LTG) fraction relative
to local clusters—or whether significant star formation is present in the early-types
as well.
1.3
Low-Redshift: Where Have All The Star Formers Gone?
A number of studies in the z . 1 regime show that the star formation of cluster galaxies decreases with age of the Universe (Couch & Sharples 1987; Saintonge et al. 2008;
Finn et al. 2008; Webb et al. 2013), while cluster ETG fractions increase (Stanford
et al. 1998; Poggianti et al. 2009). However, the majority of studies that investigate
cluster star formation tend to look at the gross star formation activity, and do not
attempt to quantify it for different morphologies.
With the goal of our high-redshift sample in Paper I well established—to quantify
the star formation activity of cluster ETGs—and the relative dearth of similar such
studies at z < 1, we propose to extend this study of morphology-dependent star
formation down to z = 0.15. With cluster ETG fractions increasing as the cluster
star formation activity decreases, we expect to find a large population of quenched
early-types. What would be surprising is cluster LTG star formation activity at the
level of ETGs or lower.
In Chapter 3, we describe the publicly available Cluster Lensing and Supernova
survey with Hubble (CLASH) data set, and discuss how we calculate stellar masses,
and ultraviolet (UV) star formation rates, and how we determine cluster membership,
and galaxy morphology. With the overall decline in cluster star formation over the
last ∼ 9 billion years, it is clearly necessary to be able to compare cluster galaxies at
different times. However, by studying cluster galaxies in the first place, we are taking
4
1.3. LOW-REDSHIFT: WHERE HAVE ALL THE STAR FORMERS
GONE?
advantage of a population that is (likely) evolving more quickly than galaxies in lowdensity environments, so it would be beneficial to be able to compare galaxies in
different environments. To that end, we describe in Chapter 4, a comparison sample
of field galaxies drawn from the UltraVISTA survey. We then discuss our empirical
SF RUV → SF RIR relations, derived using the UltraVISTA galaxies, in Chapter 5.
We use these relations to correct the CLASH UV SFRs to be directly comparable to
24 µm SFRs from ISCS and UltraVISTA.
In Chapter 6, we explore galaxy morphologies, mean SFRs, and specific SFRs,
first as a function of cluster-centric radius, then as a function of redshift.
5
Chapter 2
Paper I: Star Formation in
High-Redshift Cluster Ellipticals
2.1
Paper I Organization
Aside from some slight alterations to the sectioning depth, §2.2 to §2.6 are reproduced
here as they were submitted in Paper I. As noted above, the majority of Paper I’s
introduction section appears in Chapter 1.
In §2.2 we summarize the ISCS cluster sample, multi-wavelength data sets, and
red-sequence catalogs of S12 which form the basis of this work. In §2.3 we describe
our criteria for selecting isolated, early-type cluster members, and present our measurements of their star formation activity in §2.4. We discuss our results in §2.5 and
present our conclusions in §2.6.
6
2.2. DATA
2.2
Data
2.2.1
IRAC Shallow Cluster Survey
In the ISCS, Eisenhardt et al. (2008) identified candidate galaxy clusters in a 7.25 deg2
area of the Boötes field of the NOAO Deep Wide-Field Survey (NDWFS; Jannuzi &
Dey 1999), over a redshift range of 0.1 < z < 2, using imaging from the IRAC Shallow
Survey (Eisenhardt et al. 2004). Using accurate photometric redshifts from Brodwin
et al. (2006), a wavelet algorithm was used to identify clusters from the 4.5 µmselected galaxies as three-dimensional overdensities. The cluster centers were taken
to be the peaks in the wavelet detection maps. Three more epochs were subsequently
obtained in all IRAC bands as part of the Spitzer Deep, Wide-Field Survey (SDWFS;
Ashby et al. 2009), which increased the photometric depth of the IRAC images by
a factor of two. The deeper SDWFS data were used to improve the photometric
redshift accuracy for all galaxies, as well as to extend the catalog to lower flux limits.
The SDWFS catalog is 80% complete at 18.1 mag in the 4.5µm band (Ashby et al.
2009).
In this work we focus on 11 spectroscopically confirmed 1 < z < 1.5 ISCS clusters
selected by S12 for follow-up HST observations, as described below. We list these
clusters, along with their positions and spectroscopic redshifts, in Table 2.1.
2.2.2
HST Data
A subset of ISCS clusters spanning 1 < z < 1.5 were imaged with HST in the NIR
and optical with instrument and filter combinations chosen to bracket the 4000 Å
break. NIR data were acquired with the Wide Field Camera 3 (WFC3; Kimble et al.
2008) at F160W. In the optical, observations were either taken with the Advanced
7
2.2. DATA
Table 2.1: ISCS Clusters
ISCS Cluster
Name
R.A.
(J2000)
Dec.
(J2000)
zspec
Visually
Selected ETGs
Visually
Selected LTGs
J1429.2+3357
J1432.4+3332
J1426.1+3403
J1426.5+3339
J1434.5+3427
J1429.3+3437
J1432.6+3436
J1433.8+3325
J1434.7+3519
J1438.1+3414
J1432.4+3250
14:29:15.16
14:32:29.18
14:26:09.51
14:26:30.42
14:34:30.44
14:29:18.51
14:32:38.38
14:33:51.13
14:34:46.33
14:38:08.71
14:32:24.16
33:57:08.5
33:32:36.0
34:03:41.1
33:39:33.2
34:27:12.3
34:37:25.8
34:36:49.0
33:25:51.1
35:19:33.5
34:14:19.2
32:50:03.7
1.059
1.112
1.136
1.163
1.238
1.262
1.349
1.369
1.372
1.414
1.487
8
3
5
8
2
6
3
5
1
4
4
5
4
7
5
4
3
3
7
7
7
6
Total Number
49
58
Camera for Surveys (ACS; Ford et al. 1998) in filters F775W, F814W or F850LP, or
with the Wide Field Planetary Camera 2 (WFPC2; Holtzman et al. 1995) at F814W.
The reader is referred to S12 for a more detailed description of these data.
We use the all-HST color-magnitude relations (CMRs) of 11 distant, spectroscopically confirmed ISCS clusters presented in S12. That work isolated the CMRs by
subtracting a passively evolving Coma CMR model with a fixed rest-frame slope.
Galaxies brighter than an evolving magnitude limit of H ∗ (z) + 1.5 with color offsets,
∆, within −0.25 < ∆ < 0.75 of the model CMR were identified as red-sequence
galaxies. To reduce the effect of outliers, they removed galaxies that were more than
two median absolute deviations in color from their measured ∆ zero point.
We independently calculate the F160W completeness via Monte Carlo methods.
We insert 100 artificial stars into each WFC3 cluster image in steps of 0.05 magnitudes
from 23 to 25.5, ensuring a minimum separation of 5 pixels between input stars. We
8
2.2. DATA
repeat this procedure 10 times providing 1000 input sources per magnitude bin, per
cluster. We run SExtractor (ver. 2.8.6, Bertin & Arnouts 1996) on each image to
generate source positions and mag auto magnitudes. We match the SExtractor
positions with our input positions in each magnitude bin, counting the source as
matched if it lies within 2 pixels of the input position of the artificial star. For each
matched source, we calculate the magnitude difference between the input source and
mag auto output, and find that the absolute mean ∆mag is . 0.2 up to a magnitude
of 24.9. We perform a least squares fit to the fraction of matched sources and find a
90% completeness limit of 24.1 mag, consistent with that found in S12. We confirm
that all galaxies in our final cluster sample (§2.3) have an F160W magnitude brighter
than our 90% completeness.
2.2.3
Mid-Infrared Data
The high-redshift clusters studied in this work were also imaged at 24 µm with the
Multiband Imaging Photometer for Spitzer (MIPS; Rieke et al. 2004). The exposure
times, which increased with redshift from 12 to 48 min, were designed to produce
similar sensitivities in IR luminosity for all clusters. Following the method of Magnelli
et al. (2009), MIPS source catalogs were generated by using the positions of objects
in the higher-resolution IRAC images as priors. This method produces 24 µm flux
measurements (or limits) for all IRAC galaxies. For consistency with B13, we infer
total IR luminosities for these sources using templates from Chary & Elbaz (2001),
and convert these to SFRs using the Murphy et al. (2011) relation. The 1 σ depth of
our SFRs is ∼ 13 M yr−1 .
While the Chary & Elbaz (2001) templates typically overestimate LIR at z > 1.5
9
2.3. GALAXY SELECTION METHOD
by a factor of ∼2–8 (Murphy et al. 2009; Nordon et al. 2010; Rodighiero et al. 2010),
it provides an estimate that is accurate to 40% up to z ∼ 1.5 (Marcillac et al. 2006;
Murphy et al. 2009; Elbaz et al. 2010).
In determining their total LIR to SFR calibration, Murphy et al. (2011) assumed
that the entire Balmer continuum is absorbed and reradiated by optically thin dust.
They also assumed a solar metallicity, and continuous star formation over a timescale
of ∼100 Myr. The relation was defined using a Kroupa (2001) IMF, which has a
similar normalization to the Chabrier (2003) IMF we use to calculate our stellar
masses.
2.2.4
Chandra X-ray Data
Murray et al. (2005) obtained X-ray imaging of the NDWFS field with Chandra to
depths of 5–15 ks. Follow-up Cycle 10 Chandra observations brought the exposure
time to a uniform depth of 40 ks for the clusters in the present sample.
2.3
Galaxy Selection Method
We first build our cluster member sample by selecting galaxies that are robust spectroscopic or red-sequence members (§2.3.1). We remove galaxies that likely harbor
AGNs (§2.3.2), and galaxies that fall below our uniform stellar mass cut (§2.3.3).
In order to robustly measure SFRs for cluster members, we select galaxies that are
free from potential 24 µm contamination due to nearby neighbors, based on visual
inspection of optical and IR images (§2.3.4). We then visually classify our isolated
cluster members, separating them into ETGs and LTGs (§2.3.5). Finally, we describe
our selection of a comparison sample of high-redshift field galaxies in §2.3.6.
10
2.3. GALAXY SELECTION METHOD
2.3.1
Identification of Cluster Members
We select the 996 galaxies from the 11 high-redshift ISCS clusters studied in S12 for
which we have optical/NIR (HST ) and 24 µm (Spitzer) images. We match these
galaxies to the 4590 SDWFS catalog positions for which we have SFR (§2.2.3), and
stellar mass (§2.3.3) estimates. We choose galaxies where the separation between the
Spitzer and HST positions is ≤ 200 . The matched sample consists of 465 galaxies.
The F160W-selected galaxies without matches are undetected in the SDWFS IRAC
imaging, and hence they fall below the uniform IRAC-based stellar mass cut we
impose below.
To identify the subset of galaxies that are robust cluster members, we only retain
for the final catalog objects with either high-quality spectroscopic redshifts consistent
with membership (Eisenhardt et al. 2008; B13; Zeimann et al. 2013), or those that
are red-sequence members based on HST photometry, as described in §2.2.2. We cut
219 non-members, reducing our sample to 246 galaxies.
2.3.2
Rejection of AGNs
The presence of an AGN can affect the mid-infrared (MIR) flux, potentially leading to
an incorrect estimate of the SFR. While only 1% of local cluster galaxies show AGN
signatures (Dressler et al. 1985), the surface density of AGNs increases with redshift
(Galametz et al. 2009; Martini et al. 2013), making the effect more prominent in our
redshift range of interest.
Following B13, we remove AGNs identified via either X-ray or MIR techniques.
Galaxies whose counterparts in our 40 ks Chandra images were point sources with
hard X-ray luminosities brighter than LX,H > 1043 erg s−1 were removed as likely
11
2.3. GALAXY SELECTION METHOD
AGN. Similarly, objects with signal-to-noise (S/N) ≥ 5 in all IRAC bands that fall
in the AGN wedge from Stern et al. (2005), which are reddened in the MIR due
to heating of their dust by AGN, were also removed. These cuts removed a total
12 objects (∼5% of cluster members), bringing our sample to 234 position-matched
cluster members apparently free of significant AGN contamination. The resulting
SFRs will necessarily be lower limits due to these exclusions.
2.3.3
Stellar Masses and Mass Limit
We measure stellar masses for our galaxies using the Bayesian spectral energy distribution (SED) fitting code, iSEDfit (Moustakas et al. 2013), which infers galaxies’
physical properties by fitting population synthesis models to their broadband spectral
energy distributions. In this work, we use population synthesis models from Bruzual
& Charlot (2003), which are based on the Padova 1994 stellar evolutionary tracks
(Girardi et al. 1996), the stelib empirical stellar library (Le Borgne et al. 2003), and
the Chabrier (2003) IMF.
While all of the cluster members, and indeed all the galaxies from our initial
sample, are brighter than our 90% F160W completeness limit of 24.1 mag, we impose
the 80% IRAC-based completeness limit from B13. We remove the four galaxies that
fall below this limit, resulting in 230 cluster members with log(M? /M ) > 10.
2.3.4
Isolation
Because of its broad point spread function (∼600 , Rieke et al. 2004), a single source in
the 24 µm band of MIPS can be comprised of multiple distinct physical sources. Due
to the difficulty in deconvolving a multiple-object SFR into its constituent SFRs, we
12
2.3. GALAXY SELECTION METHOD
choose to limit our work to isolated objects. To that end, we visually inspect 24 µm
Spitzer images and the available optical HST images of the remaining 230 cluster
members, removing from our final sample those for which we are not able to rule out
significant contributions to the MIPS flux from other nearby objects. Through these
inspections, we further reduce our sample by 123 galaxies, resulting in 107 isolated
cluster members with no NIR or X-ray luminous AGNs.
This cut on isolation removes interacting and merging members, thus likely lowering our total measured SFRs. It was necessary, however, to obtain robust SFR
measurements for morphological early-types, a key goal of this work. The significant star formation activity seen amongst the isolated red-sequence galaxies (§2.4) is
therefore a lower limit. We do not attempt to correct the SFR to the total value, but
simply note that the sense of the correction—to higher cluster SFRs for both earlyand late-type members—serves to strengthen our conclusions.
2.3.5
Morphology
We visually inspect the optical and NIR HST images of the 107 galaxies in our sample,
classifying those consistent with smooth elliptical and S0 shapes as ETGs. Galaxies
which exhibit either late-type signatures, or disturbed or irregular morphologies, are
collectively classified as LTGs. By requiring the ETGs to have smooth early-type
profiles with no signs of interaction, we are removing from this sample galaxies with
merger signatures, which again biases our sample against ETGs with potentially
higher SFRs.
S12 performed visual inspections of their sample using F160W images, assigning
morphologies and recording the local environments. Additionally, Sérsic indices (ns )
13
2.3. GALAXY SELECTION METHOD
Figure 2.1: 500 ×500 cutouts of five isolated cluster ETGs, with HST filter listed on
each image. Listed below each pair of images is the galaxy name, spectroscopic
cluster redshift, SFR, and stellar mass.
14
2.3. GALAXY SELECTION METHOD
were measured in the F160W filter for all S12 galaxies using Galfit (Peng et al.
2010) and Galapagos (Häußler et al. 2011), with ETGs defined as having ns > 2.5.
These Sérsic index measurements will be described in more detail by C. Mancone et
al. (2014, in preparation).
As a test of the robustness of our visual morphological classifications, we compare
with the independent S12 classifications. The primary difference between these two
morphological catalogs is that our classification takes advantage of the higher resolution ACS and WFPC2 HST images, in addition to using the F160W images. After
applying isolation and AGN cuts to the visually classified S12 sample consistent with
those in the present work, we find agreement for ∼85% of the ETG sample. Where
the visual morphologies differed, the cause tended to be late-type features (typically
disks with spiral structure) that were clearly visible in the high resolution optical
images but not apparent at F160W.
We also test the Sérsic indices measured from the F160W images, finding that
this quantitative measure disagrees with ∼30% of our visual classifications. This
high discrepancy level is in line with the 30-40% sample contamination reported by
Mei et al. (2012) for selecting morphology using Sérsic indices.
For this work we choose to use visual morphologies, though we have run the
analysis both ways and have verified that none of the major qualitative results depend
on this choice. Our final sample of visually classified ETGs (LTGs) contains 49 (58)
galaxies, and the breakdown by cluster is listed in Table 2.1. In Figure 2.1, we show
500 ×500 optical (left) and NIR (right) cutouts of isolated cluster ETGs of varying SFR
and redshift. Below each galaxy, we list its name, spectroscopic cluster redshift, SFR,
and stellar mass. At these redshifts, 500 corresponds to ∼41–43 kpc.
15
2.3. GALAXY SELECTION METHOD
2.3.6
Comparison Field Sample Selection
We select our comparison sample of 1.0 < z < 1.5 field galaxies from UltraVISTA
(McCracken et al. 2012; Muzzin et al. 2013), a deep Ks -selected survey covering 1.62
deg2 of the COSMOS field (Scoville et al. 2007). The publicly available UltraVISTA
survey1 has photometry in 30 bands, used to calculate photometric redshifts and infer
stellar masses, and also includes Spitzer photometry in the 24 µm band of MIPS and
all four channels of IRAC (Muzzin et al. 2013).
We select only isolated galaxies by removing photometric catalog members with a
neighbor within 600 . We also exclude likely AGNs using IRAC photometry to identify
galaxies that fall into the Stern et al. (2005) wedge. Muzzin et al. (2013) calculated
stellar masses using Bruzual & Charlot (2003) models, and assuming a Chabrier
(2003) IMF. We impose on our field sample the z ∼ 1 UltraVISTA 100% mass completeness cut of log(M? /M ) > 9.74. Following the method in §2.2.3, we infer total
IR luminosities using the Chary & Elbaz (2001) templates, and calculate SFRs with
the Murphy et al. (2011) relation. We use the updated, higher-redshift version of the
morphological catalog of Cassata et al. (2007)2 to separate ETGs and LTGs. Our
final comparison sample contains 91 ETGs and 2109 LTGs.
1
http://www.strw.leidenuniv.nl/galaxyevolution/ULTRAVISTA/Ultravista/Data_
Products_Download.html
2
Publicly available COSMOS data sets, including the morphological catalog used in this work,
are located at http://irsa.ipac.caltech.edu/data/COSMOS/datasets.html
16
2.4. ANALYSIS
Figure 2.2: SFR versus cluster-centric radius for cluster ETGs (red filled circles),
and LTGs (blue open squares). The horizontal dashed line is the 1 σ SFR detection
level (13 M yr−1 ), and the error bar represents the 40% systematic uncertainty in
the SFRs. SFRs of isolated cluster galaxies—of all morphological types—are weakly
correlated with projected radius.
2.4
Analysis
2.4.1
Star Formation Rate vs. Radius
All the galaxies in our sample are robustly detected in optical and IRAC imaging,
and our MIPS 24 µm fluxes are measured for all sources using these positional priors.
The resulting SFRs are thus physically meaningful down to very low significances,
albeit with large uncertainties.
In Figure 2.2 we plot SFR versus projected cluster-centric radius for cluster ETGs
17
2.4. ANALYSIS
(red filled circles) and LTGs (blue open squares). Galaxies plotted below the horizontal dashed line have SFRs below our 1 σ depth of 13 M yr−1 . The large error
bar shows the systematic error in the SFR, which we take to be 40%, based on a
comparison between 24 µm and Herschel SFR measurements over z = 0–1.5 (Elbaz
et al. 2010).
On their own, the ETG and LTG samples show little-to-no radial dependence
in the SFRs. However, when considering all cluster galaxies, we do find a weak
correlation (Spearman’s rs = 0.30 ± 0.06 at the 99% confidence level) between SFR
and cluster-centric radius.
As can be seen in Figure 2.2, we are largely limited to radii less than 0.75 Mpc due
to the small footprint of WFC3. We are able to probe beyond 1 Mpc in J1432.4+3250
due to two adjacent pointings on this cluster. We have verified our results are unchanged if we limit our analysis to the well-sampled region below 0.75 Mpc.
2.4.2
Mean Star Formation Rate
To explore the effect of environment on the SFRs of both ETGs and LTGs, we show
in Figure 2.3 the mean SFR, hSF Ri, as a function of projected cluster-centric radius.
ETGs are plotted as red circles and LTGs as blue squares. We separate the galaxies
into three non-overlapping annuli, selecting the radial bins such that the S/N in each
is approximately equal. From inner to outer, the bin sizes are 200, 175, and 800
kpc, respectively. The errors in each bin are calculated from the quadrature sum of
bootstrap resampling (1000 samples, with replacement) and simple Poisson errors.
We also show the mean SFR of our comparison sample of field ETGs with the red
horizontal line. The 68% error in the mean is given as the shaded region.
18
2.4. ANALYSIS
Figure 2.3: Mean SFR versus cluster-centric radius for ETGs (red circles) and LTGs
(blue squares), with bin widths depicted by the horizontal error bars. In each bin, the
median SFR of non-detections is assigned to each undetected galaxy and included in
the mean. The vertical error bars show the quadrature sum of bootstrap resampling
and Poisson error. The solid horizontal red line (shaded region) is the mean SFR
(error) for field ETGs. On average, cluster ETGs are forming stars at a rate almost
15% that of cluster LTGs, and 32% that of field ETGs.
In computing the mean SFRs, objects with individual SFRs below 13 M yr−1
were assigned the median value of all such objects in the bin. This is the catalog-space
equivalent of median stacking; from inner to outer annuli, the median SFRs of these
ETGs (LTGs) are 1.5, 2.6, and 3.0 (3.6, 2.8, and 2.7) M yr−1 . We have verified that
none of our main results change even in the extreme case of setting the SFRs of all
such < 1 σ SFRs to zero.
We find mean SFRs of 5.1 ± 2.4, 5.6 ± 2.1, and 12.0 ± 5.3 M yr−1 for our
cluster ETG sample from the inner to outer annuli, respectively. With mean SFRs
19
2.4. ANALYSIS
of 41.4 ± 17.3, 49.4 ± 21.1, and 55.2 ± 14.3 M yr−1 over the same range, the LTGs
have mean SFRs ∼5 to 9 times higher at all radii. The ETGs and—to a lesser
extent—the LTGs show some decrease in mean SFR at small radii, although this is
not statistically significant given the large errors. Averaging over all radii we find a
mean SFR of 7.4 ± 2.0 M yr−1 for cluster ETGs, which is a factor of 3.1 lower than
the mean SFR of field ETGs (hSF Ri = 23.2 ± 5.0 M yr−1 ). While the uncertainty
in our SFRs is too large to determine any radial dependence, we do find that cluster
ETGs show suppressed star formation activity relative to field ETGs. We note that
even with 91 galaxies in the sample, the mean SFR of field ETGs is heavily affected
by one galaxy with SFR > 300 M yr−1 . However, the mean SFR without this object
(19.6 ± 3.0 M yr−1 ) would still be more than a factor of two higher than that of
cluster ETGs.
These results show that although cluster ETGs have SFRs that are fairly quenched
relative to both their field analogs and the remainder of the cluster population, they
still contribute 12.7% of the vigorous star formation observed in these clusters.
2.4.3
Fraction of Star-forming Galaxies
In Figure 2.4, we plot fSF , the fraction of star-forming, visually selected cluster ETGs
(red circles) as a function of cluster-centric radius. We conservatively limit this measurement to members with SFRs of at least 26 M yr−1 , above our 2 σ detection level.
We use only two annuli due to the relatively small size of our sample, and use the
binomial error in the fraction as our total error. The radial bins and error ranges are
shown by the horizontal and vertical error bars, respectively. The horizontal red line
shows the fraction of star-forming field ETGs, with the binomial error in the fraction
20
2.4. ANALYSIS
Figure 2.4: Fraction of star-forming galaxies versus projected radius for high-redshift
cluster ETGs (red circles) and low-redshift cluster galaxies from C11 (gray squares
and error bars). For cluster ETGs, the size of the horizontal error bars represents the
bin widths, and the vertical error bars represent the binomial error in our fractions.
The lower axis corresponds to the data from this work; the upper from C11. The
lower gray curve is a least squares fit to all six points from Figure 4 of C11 (extending
to 3Rproj /R200 ), while the upper gray curve is the same fit, shifted up by a factor
of 11.6. The red horizontal line (shaded region) shows the fraction (error) for the
comparison high-redshift field ETGs. High-redshift cluster ETGs have star-forming
fractions at least an order of magnitude higher than local cluster galaxies, and two
times lower than field ETGs in the same redshift range.
shown by the shaded region. The gray points and error bars show the fraction of local
(z . 0.1) star-forming (LIR > 4.7 × 1010 L ) cluster galaxies versus projected radius,
from Chung et al. (2011, hereafter C11), who studied 69 low-redshift clusters with
21
2.4. ANALYSIS
total dynamical masses in the range ∼ (1 − 7) × 1014 M , determined by using caustic infall patterns (Rines & Diaferio 2006), and selecting only galaxies brighter than
Mr = −20.3. The lower x-axis corresponds to the projected cluster-centric radius for
this work, while the upper x-axis corresponds to the projected R200 -normalized radius
from C11. Based on the X-ray, weak lensing and dynamical masses that have been
measured for a subset of the Boötes clusters (Brodwin et al. 2011; Jee et al. 2011),
as well as on a clustering analysis of the full ISCS sample (Brodwin et al. 2007), our
z > 1 ISCS clusters have halo masses in the range ∼(0.8 − 2) × 1014 M , and virial
radii of ∼1 Mpc. Therefore the upper and lower axes in Figure 2.4 are approximately
equivalent. While the median mass of the C11 clusters is larger than that of the
Boötes clusters, the latter will grow in mass in the ∼8–9 Gyr to the present epoch.
+6.6%
Averaging over all radii, we find that 12.2%−4.7%
of isolated cluster ETGs are star
forming. It is clear that the star-forming fraction is substantially higher at z > 1.0
than it is locally. We quantify this difference by first fitting a least squares curve to
all six C11 points (while only the first three points are shown in Fig. 2.4, the C11
measurements extend to 3Rproj /R200 ), then using χ2 minimization to determine that
a simple scaling factor of 11.6 provides an excellent fit to our data. In both radial
bins, we find that the fraction of star-forming ETGs in our sample is approximately
an order of magnitude higher than for local cluster galaxies of all types.
Due to the very low redshift of the clusters in C11 (z . 0.1), their minimum
cutoff for star-forming galaxies (LIR > 4.7 × 1010 L ) is ∼ 3 times lower than our 2
σ level of ∼1.5 × 1011 L . Using the published LIR values from C11, we find that six
of the 109 cluster galaxies (within 3Rproj /R200 ) considered star-forming in C11 have
LIR > 1.5 × 1011 L , which is only ∼0.1% of their cluster member sample, a factor
22
2.4. ANALYSIS
of ∼18 lower than with their SFR cut, and ∼100 times lower than our cluster ETG
fraction.
Within the errors, we do not have enough evidence to determine whether there
is a radial trend in our ETGs, though the shifted radial profile of C11, indicated by
the upper curve, is clearly consistent with our data. Also, the fSF of field ETGs
(0.28+0.06
−0.05 ) is 2.3 times greater than that of cluster ETGs, implying an environmental
dependence on ETG star formation.
2.4.4
Specific Star Formation Rate
We next explore the specific star formation rate (sSFR), defined as the sum of the
SFRs divided by the sum of the stellar masses, in each radius bin. In normalizing
the SFR of a galaxy to its mass, the sSFR allows us to explore the relative efficiency
with which it converts its cold gas into stars.
In Figure 2.5 we plot the sSFR versus radius for both our cluster ETGs (red
circles and error bars) and for the low-redshift, star-forming, cluster galaxies of C11
(gray points). The radial binning and error ranges for our points are calculated as
in Figure 2.3. We show the sSFR and similarly calculated error of field ETGs as the
red horizontal line and shaded region, respectively.
As in §2.4.3, we fit a best-fit curve to the C11 points, then determine the shift in
amplitude required to match our cluster ETG points. We find that scaling the curve
by a factor of 102.08 , also shown in the figure, provides a good fit to our high-redshift
sSFR measurements. High-redshift cluster ETGs are forming stars at a rate 120 times
higher than local cluster galaxies.
C11 calculated sSFR for star-forming (LIR > 4.7 × 1010 L ) galaxies, while we
23
2.4. ANALYSIS
Figure 2.5: Specific SFR versus radius for high-redshift cluster ETGs (red filled circles) and low-redshift cluster galaxies from C11 (gray filled squares), with the same
x-axes as Figure 2.4 (§2.4.3). The vertical error bars represent the bootstrapping and
Poisson error in our sSFRs, and the horizontal error bars show the size of each bin.
The lower gray curve is a least squares fit to the C11 points (all six points from their
Fig. 3, extending to 3Rproj /R200 ), while the upper gray curve is the same fit shifted
up by a factor of 120. The red horizontal line (shaded region) shows the sSFR (error) for the comparison high-redshift field ETGs. The sSFR of high-redshift cluster
ETGs is 19 times lower than similar redshift field ETGs, yet more than two orders
of magnitude larger than low-redshift cluster galaxies of all types. While we place no
lower limit on the LIR of high-redshift cluster ETGs, C11 only measured sSFR for
galaxies with LIR > 4.7 × 1010 L (their star-forming cut; see §2.4.3). The sSFR of
C11’s cluster galaxies would be lower if no cut was imposed, making the factor of 120
we find here a lower limit.
place no such constraint on either high-redshift ETG sample plotted in Figure 2.5.
Additionally, the morphological mix of the low redshift cluster galaxies to which
24
2.5. DISCUSSION
we are comparing is unclear. Although clusters in the local Universe are primarily
inhabited by early-type, “red and dead” galaxies (Oemler 1974; Dressler 1980), the
star-forming subset detected by C11 may be preferentially drawn from the small
fraction of late-type members or from recently accreted field galaxies. Correcting for
such LTG contamination, and removing the LIR limit, would lower the sSFR in the
low redshift sample, and hence make the evolution over this redshift range even more
dramatic.
Averaging across all radii, the cluster ETGs have an sSFR of 0.11 ± 0.04 Gyr−1 ,
which is a factor of 19.0 lower than the sSFR of the field ETG sample (2.02 ± 0.42
Gyr−1 ). While we cannot definitively determine whether there is a radial trend in
cluster ETG sSFR, this drop relative to ETGs in the field is further evidence for the
environmental dependence of the star formation of ETGs, as shown in the previous
two sections.
What is not immediately clear is how much of this offset between cluster and field
ETGs is due to the difference in stellar mass—both the distribution, and minimum
cut applied in §2.3.3—and how much is due to the actual SFRs. To quantify this,
we find that the mean stellar mass of cluster ETGs is 7.0 × 1010 M , a factor of 6
times larger than the mean stellar mass of field ETGs (1.2×1010 M ). This, however,
only accounts for 31% of the offset between the field and cluster, implying that the
remaining ∼70% is due to the quenching of cluster ETGs relative to the field.
2.5
Discussion
Two related studies (B13, and A14) identified high levels of star formation activity in
a superset of the ISCS clusters studied in this work. However, these studies could not
25
2.5. DISCUSSION
isolate the ETGs due to a lack of morphological information. In the present work,
which benefits from the high-resolution imaging of HST, we are able to expand upon
their results by analyzing the star formation properties of isolated, massive ETGs,
and comparing them to high-redshift ETGs in the field and to galaxies in low-redshift
clusters.
2.5.1
Comparison to Brodwin et al. (2013)
B13 measured the star formation activity in 16 ISCS galaxy clusters at 1.0 < z < 1.5,
including the 11 clusters studied in this work. Their large sample enabled them to
finely bin their data as functions of both redshift and radius. The morphology and
isolation cuts in the present work result in a relatively small sample size that precludes a similar analysis, but instead permits an investigation of physically interesting
subsamples of these cluster members.
Both B13 and this work used the same 24 µm Spitzer photometry, so the measurements of star formation activity should be consistent, despite the sample size
difference. To test this, we first compared, as a function of redshift, the fraction of
isolated star-forming galaxies of all types (by combining ETGs and LTGs) from this
work to the fraction of star-forming galaxies from B13 within 1 Mpc. We used their
redshift binning (1 < z < 1.2, 1.2 < z < 1.37, and 1.37 < z < 1.5) and adopted their
(S/N ≥ 4) flux limit, corresponding to SFR & 47 M yr−1 . Our results are in good
agreement with those reported in B13.
Probing to lower star formation rates, we plot in Figure 2.6 the fraction of isolated
star-forming cluster galaxies down to our full 2 σ SFR ≥ 26 M yr−1 limit. We bin the
galaxies (1 < z < 1.16, 1.16 < z < 1.35, and 1.35 < z < 1.5; shown by the horizontal
26
2.5. DISCUSSION
Figure 2.6: Fraction of isolated, high-redshift star-forming (SFR ≥ 26 M yr−1 )
cluster galaxies (see legend for symbol types) as a function of redshift. Vertical error
bars show the binomial error in the fractions, while the horizontal error bars show
the redshift binning for all three sets of cluster galaxies. The significant drop in
star-forming fraction from 61.0% to 26.5% exhibited by all cluster galaxies (green
triangles) at z ∼ 1.4 is in good agreement with the redshift of transition—away from
vigorous star formation—observed by B13.
error bars) such that the binomial error in each redshift bin is approximately equal
for galaxies of all morphological type (green filled triangles). ETGs are plotted as red
circles and LTGs as blue squares, slightly offset to the left and right, respectively.
61.0% of all isolated cluster galaxies are star-forming at 1.35 < z < 1.5, followed
by a decrease to 26.5% in the middle redshift bin. This result is in good agreement
with the transition redshift of z ∼ 1.4 found by B13 between the era of vigorous
star formation in high-redshift clusters and the quenched epoch at later times. The
LTG population experiences a similar trend, with a decrease from 77.8% to 46.7%,
27
2.5. DISCUSSION
over z ∼ 1.4 → 1.25, while the ETG sample shows a milder downward trend in starforming fraction at z ∼ 1.4 (from 28.6% to 10.5%), though it is formally consistent
with being constant within the errors.
2.5.2
Comparison to Alberts et al. (2014)
A14 explored 274 ISCS clusters from z = 0.3 to 1.5, including ∼100 over the same
redshift range as our observations. By stacking 250 µm Herschel data, they were
able to probe to mean LIR values almost an order of magnitude lower than our 1 σ
detection limit. Although our sample is a subset of the A14 sample, the measurement
techniques—24 µm detections versus stacking at 250 µm—are relatively independent.
Here we compare some of our results with those reported in A14.
In order to compare the two SFR measurements for our morphologically selected
sample, we first attempted to directly measure stacked 250 µm fluxes for our ETG
and LTG samples. However, with the relatively small sample size, and source contamination due to the large beam size (18.00 1, Swinyard et al. 2010), the S/N was too
low to permit this measurement.
In Figure 2.7 we compare mean SFR as a function of redshift, derived from 24
µm and 250 µm measurements. To be consistent with the radial selection in A14, we
only plot galaxies with cluster-centric radii < 1 Mpc. We plot our cluster galaxies as
the filled points, using the same binning as in Figure 2.6. The filled green triangles
represent our isolated cluster galaxies of all morphologies, while the open triangles
show the mean SFR of galaxies of all types from A14. Despite the order of magnitude
difference in observed wavelength, and the very different measurement methodologies,
the results are in excellent agreement for galaxies of all morphological types.
28
2.5. DISCUSSION
Figure 2.7: Mean SFR versus redshift for our isolated cluster galaxies (filled symbols,
with the same binning as in Figure 2.6) and for cluster galaxies from A14 (open
symbols), with all galaxies having cluster-centric radius < 1 Mpc. Our errors are
calculated as the quadrature sum of bootstrapping and simple Poisson errors. The two
different measurement methods used for calculating these SFRs—24 µm detections
versus stacking at 250 µm—show consistent results.
While unable to visually or quantitatively determine morphologies, A14 matched
each galaxy in their sample against seven Polletta et al. (2007) templates representing
different morphologies. Using this template fitting as a proxy for color, A14 selected
galaxies that were best fit by late-type templates as “blue” (star-forming) galaxies.
We plot the mean SFR of these galaxies as the light blue open squares, and compare
them to our LTGs (blue filled squares). We again find that there is excellent agreement
between the 250 µm-derived SFRs in A14 and the 24 µm-derived SFRs in this work.
29
2.5. DISCUSSION
2.5.3
Star Formation in High-Redshift Cluster ETGs
Low-redshift ETGs, particularly those in clusters, have quiescent, old stellar populations. A key issue in the evolution of galaxy populations in clusters is determining
the nature of the star formation history of these “red-and-dead” galaxies. Specifically,
when did present-day massive cluster ETGs experience their last major burst of star
formation? Determining the epoch during which these galaxies experienced such a
burst can provide constraints on when cluster galaxies experienced their last phase
of gas-rich major merging.
We find that the fraction of star-forming galaxies is an order of magnitude larger
for our 1 < z < 1.5 cluster ETGs than for local cluster galaxies of all morphologies
(C11). Measurements of star-forming fractions in nearby clusters, such as those by
C11, necessarily include large contributions from late-type galaxies as ETGs typically
have SFRs below the survey limits. The increase we find is therefore a lower limit to
the evolution between nearby and z > 1 cluster ETGs.
Even more striking is the comparison of the mass-normalized star formation rates
between these two galaxy populations. With sSFRs more than two orders of magnitude higher than cluster galaxies in the local universe, our high-redshift cluster ETGs
have significantly more ongoing star formation activity per unit stellar mass. This
dramatic evolution would be even more extreme if C11 had measured the sSFRs of
all galaxies above a fixed mass limit, rather than just those above their star formation
detection limit.
An interesting result stemming from the comparisons in §2.5.1 and §2.5.2 is that
despite the decline in the fraction of star-forming (SFR ≥ 26 M yr−1 ) cluster ETGs
30
2.5. DISCUSSION
at z ∼ 1.4 seen in Figure 2.6, this population has mean SFRs that are roughly constant across this period. Specifically, although we see that the star-forming fraction
drops ∼18 percentage points (albeit with very large scatter) at this redshift, their
mean SFRs (shown in Figure 2.7) remain relatively constant from z ∼ 1.4 → 1.25.
One potential conclusion from this is that while a significant quantity of early-type
galaxies are being quenched, there must be some mechanism that is enhancing the
star formation activity of the remaining star-forming ETGs.
If the above results are not a product of environment, we would expect to see
ETGs in both the cluster and the field have a similar lack of SFR evolution. To
determine whether this is the case, we plot in the upper panel of Figure 2.8 the sSFR
for cluster (red filled circles) and field (pink open circles) ETGs, and field LTGs
(light blue open squares) as a function of redshift. We choose to plot sSFR as it best
represents the efficiency of star formation activity. For our cluster ETGs, we use the
same binning as in Figures 2.6 and 2.7, while we bin the field galaxies such that the
S/N of the ETGs is approximately equal in each bin. Errors are calculated as the
quadrature sum of bootstrap resampling and simple Poisson errors.
Not only do field ETGs show a steady decreasing trend in sSFR from z = 1.5 → 1,
but also their sSFRs drop by a factor of 4.5 over this redshift range, in line with the
factor of 4.0 decrease we find for field LTGs. Cluster ETGs, on the other hand, show
little-to-no evolution between the two highest redshift bins, followed by a factor of
3.5 decrease at z ∼ 1.08, however, with a very large scatter. The lack of evolution in
the cluster ETG sSFR between the two highest redshift bins may suggest that any
mechanism operating to sustain their SFRs is dependent on environment.
In the lower panel of Figure 2.8 we plot the fraction of isolated cluster galaxies
31
2.5. DISCUSSION
Figure 2.8: Upper panel: Specific SFR as a function of redshift for ETGs in clusters
(red filled circles) and in the field (pink open circles), and for field LTGs (light blue
open squares). Errors are calculated as the quadrature sum of bootstrap resampling
and simple Poisson errors, as in §2.4.4. Field galaxies show similar sSFR evolution
over our redshift range, with ETGs (LTGs) dropping by a factor of 4.5 (4.0). Cluster
ETGs show little-to-no evolution at z ∼ 1.4, and are quenched relative to their field
counterparts at all epochs. Lower panel: Fraction of cluster galaxies that are ETGs
as a function of redshift.
that we classified as ETGs, using the same redshift bins as above. We find that the
fraction of ETGs increases from 34% to 56% from z ∼ 1.4 → 1.25, which suggests
that new ETGs are being formed during this period.
32
2.5. DISCUSSION
It should be noted that there are processes other than major merging that may
potentially play a role in forming new ETGs, or shaping existing early-types (e.g.,
Kaviraj et al. 2013). Violent disk instability can cause late-type systems to lose their
disks through turbulence, forming compact gas-rich “blue nuggets” with gas inflows
similar to those generated by wet mergers (Dekel & Burkert 2014). However, the cold
streams (Kereš et al. 2005; Dekel & Birnboim 2006) that feed disks in this model are
only important in regions with low galaxy density (Kereš et al. 2005), and likely not
a significant factor in the hot halo environment of ISCS clusters.
Strazzullo et al. (2010) found a large fraction of quenched, compact ETGs in the
z = 1.39 cluster XMMU J2235, suggesting that minor—and likely dry—mergers can
increase the size of such galaxies over later epochs, without drastically altering their
star formation activity. However, XMMU J2235 is a very massive cluster (∼7 × 1014
M , Jee et al. 2009; Rosati et al. 2009), where major merger activity has likely ceased,
and a factor of at least a few times more massive than the clusters studied in this
work. As such, the mechanism suggested by Strazzullo et al. (2010) is not likely
currently playing a role in ISCS clusters—especially at z & 1.16—when considering
the star formation activity shown above.
From z ∼ 1.4 → 1.25, the fraction of cluster ETGs that are star-forming drops
from 28.6% to 10.5%. It is likely that a substantial number of cluster ETGs formed
before z = 1.5, and that they make up a significant portion of this large subset of
quenched ETGs that we find in ISCS clusters. However, from 1.35 < z < 1.5 to 1.16 <
z < 1.35, there is a 21.8 percentage point increase in the fraction of cluster galaxies
that are morphologically early-type; some portion of the remaining star-forming ETGs
in this epoch are likely recent byproducts of major mergers. Furthermore, cluster ETG
33
2.6. CONCLUSIONS
mean and specific SFRs are roughly constant over this period. These ETGs have not
yet had sufficient time for their star formation to be quenched, implying that their
progenitors’ mergers occurred relatively recently.
Moving into the lower redshift bin (1 < z < 1.16), the star-forming ETG fraction
falls to 0%, and their sSFR and mean SFR are both quenched (by factors of 3.5 and
4.0, respectively), while the ETG fraction remains relatively constant. Given our
assumptions above about gas-rich major mergers, the dearth of star-forming galaxies,
and overall lack of star-formation activity seen in this epoch, at 1 < z < 1.16, suggests
we are seeing the quenching of ETGs due to post-merger AGN activity.
2.6
Conclusions
We have used a sample of 11 high-redshift (1.0 < z < 1.5), IR-selected ISCS galaxy
clusters to investigate the star formation properties of isolated, early-type galaxies.
After conservatively removing AGNs through X-ray and IR criteria, we visually inspected our sample using high-resolution HST imaging, separating our galaxy sample
into two coarse morphological bins: ETGs and LTGs. We used deep 24 µm imaging
from Spitzer to measure the obscured SFRs, excluding galaxies for which we could
not rule out contamination from nearby neighbors.
We compared the star formation of the cluster ETG sample with low-redshift
cluster galaxies, finding an order of magnitude larger fraction of star-forming galaxies,
and a greater than two order of magnitude larger sSFR for our high-redshift cluster
ETGs. Averaging across our entire cluster ETG sample, we find that 12.2% are still
experiencing relatively enhanced (SFR > 26 M yr−1 ) star-formation activity.
By comparing the mean SFR of ETGs with LTGs in ISCS clusters, we found that
34
2.6. CONCLUSIONS
despite their enhanced star formation relative to low-redshift cluster galaxies, highredshift cluster early-types have substantially less star formation activity relative to
the rest of the isolated cluster population. However, averaging across all cluster radii,
ETGs still contribute 12.7% of the significant star formation activity observed in these
clusters.
Due to our relatively small sample size, we were unable to detect the radial dependence in star formation activity reported by B13. However, we found that our
cluster ETGs are quenched relative to a comparison sample of field ETGs, by a factor
of 3.1 in mean SFR, and 19.0 in sSFR. Even when “correcting” the sSFR of cluster
ETGs for their factor 6 higher mean stellar masses, we find that it only accounts for
∼30% of the difference between the field ETG sSFR and the significantly quenched
cluster ETG sSFR.
We then used the conservative IR luminosity cut from B13 to compare the fraction
of star-forming galaxies, fSF , with their results, finding that our measurements in
isolated galaxies agreed with the B13 measurements of all cluster galaxies in these
z > 1 clusters. We also found that our mean SFR measurements correlated well with
those of A14, who measured SFR using stacked 250 µm Herschel flux.
We used our 2 σ SFR detection limit (26 M yr−1 ) to explore the fSF evolution
from z ∼ 1.5 → 1.0. We considered cluster galaxies of all morphologies and found
that while 61.0% are star-forming at z ∼ 1.4, the fraction drops to 26.5% by z ∼ 1.25.
This drop of almost 35 percentage points suggests that the epoch of enhanced star
formation in these clusters is ending around z ∼ 1.4, a finding consistent with that
first reported by B13.
Only 28.6% of our ETGs are star forming at 1.35 < z < 1.5, and the fraction drops
35
2.6. CONCLUSIONS
to 10.5% by 1.16 < z < 1.35. While the fraction of star-forming cluster ETGs drops
18.1 percentage points over z ∼ 1.4 → 1.25, their mean and specific SFRs are largely
unchanged over this period. With a corresponding increase of 21.8 percentage points
in the fraction of ETGs over this period, these results are consistent with a scenario
where major gas-rich mergers form new early-type galaxies, temporarily enhancing
their star-formation activity.
A number of recent studies of the ISCS cluster population have presented lines of
evidence supporting the role of mergers in building the stellar mass in these clusters.
Specifically, the NIR luminosity function evolution disagrees with a passive evolution
model at z & 1.3 (Mancone et al. 2010, 2012), galaxies are experiencing substantial
star formation (A14, B13, Zeimann et al. 2013), young galaxies are continuously
migrating on the cluster red sequence (S12), and an increase in AGN activity has
been observed (Galametz et al. 2009; Martini et al. 2013). This work helps to solidify
the implications of all these studies by showing that massive ETGs in clusters likely
formed in gas-rich mergers.
36
Chapter 3
Cluster Galaxy Data and Sample
Selection
In this section we will briefly review the data and sample selection methodology for
our high-redshift cluster sample, which we introduced in Chapter 2. We will then
discuss the low-redshift CLASH cluster observations and sample construction.
3.1
High-Redshift ISCS Clusters
Our high-redshift cluster sample is based the IRAC Shallow Cluster Survey (ISCS;
Eisenhardt et al. 2008), which identified potential 0.1 < z < 2 galaxy clusters in 7.25
deg2 of the Boötes field of the NOAO Deep Wide-Field Survey (Jannuzi & Dey 1999).
A wavelet algorithm was used, with accurate photometric redshifts from Brodwin
et al. (2006), to identify three-dimensional overdensities of 4.5 µm-selected galaxies,
where the cluster centers were found from the peaks in the wavelet detection maps.
We use 11 spectroscopically confirmed clusters over 1 < z < 1.5 selected by Snyder
37
3.2. LOW-REDSHIFT CLASH CLUSTERS
et al. (2012) for follow-up HST observations, and studied by Wagner et al. (2014,
Chapter 2 of this thesis). These clusters, their positions, and their spectroscopic
redshifts, are listed in Table 2.1.
IRAC observations were used as positional priors to match the MIPS fluxes to
HST catalog sources. We used the Chary & Elbaz (2001) templates to convert the
measured 24 µm fluxes to total IR luminosities, then used the Murphy et al. (2011)
relation
LIR
SF RIR
= 3.88 × 10−44
−1
M yr
erg s−1
!
(3.1)
to calculate SFRs.
We removed likely AGNs by identifying high-luminosity X-ray point sources, or
based on IRAC colors from Stern et al. (2005). Stellar masses were estimated by fitting
galaxies SEDs to population synthesis models using iSEDfit (Moustakas et al. 2013),
using Bruzual & Charlot (2003) models, and the Chabrier (2003) IMF. We impose
the uniform 80% mass completeness limit of log (M? /M ) > 10.
We determined cluster membership by spectroscopic redshift if available; if not,
we used red sequence membership as determine by Snyder et al. (2012). Finally, we
visually classified our high-redshift cluster members to determine both isolation and
morphology. Galaxies with nearby neighbors were flagged and removed from the final
sample, as we could not reliably measure their SFRs.
3.2
Low-Redshift CLASH Clusters
For our low-redshift (0.15 < z < 1) cluster sample, we use the publicly available1
CLASH survey (Postman et al. 2012), which has observations of 25 clusters in 16
1
https://archive.stsci.edu/prepds/clash/
38
3.2. LOW-REDSHIFT CLASH CLUSTERS
HST bands. For each CLASH cluster, the available dataset includes equatorial coordinates, SExtractor (Bertin & Arnouts 1996) stellarity, AB magnitudes in 16 HST
filters, and photometric redshift estimates. In Table 3.1, we list the CLASH clusters,
their central coordinates derived from X-ray images, and their spectroscopic redshifts
(Postman et al. 2012). From our final sample, we remove objects that are likely
stars by cutting on SExtractor’s stellarity, only taking objects with stellarity
≤ 0.05.
3.2.1
Redshifts
From the literature, we have spectroscopic redshift measurements for five of the 25
CLASH clusters. Specifically, we take redshifts for Abell 383 from Geller et al. (2014),
Abell 611 from Lemze et al. (2013), and for Abell 1423, Abell 2261, and RXJ2129
from Rines et al. (2013).
Our photometric redshift estimates were derived by Postman et al. (2012) using
BPZ (Benítez 2000; Benítez et al. 2004; Coe et al. 2006), a χ2 minimization template
fitting software package.
3.2.2
Stellar Masses
We estimate stellar masses using an i-band mass-to-light (M/L) ratio, derived using
the color-M/L relation (CMLR)
log (M? /Li ) = 1.032 × (g − i) − 0.963,
39
(3.2)
3.2. LOW-REDSHIFT CLASH CLUSTERS
Cluster
Name
Abell 383
Abell 209
Abell 1423
Abell 2261
RXJ2129.7+0005
Abell 611
MS2137-2353
RXJ2248.7-4431
MACS1931.8-2635
MACS1115.9-0129
RXJ1532.9+3021
MACS1720.3+3536
MACS0416.1-2403
MACS0429.6-0253
MACS1206.2-0847
MACS0329.7-0211
RXJ1347.5-1145
MACS1311.0+0310
MACS1149.6+2223
MACS1423.8+2404
MACS0717.5+3745
MACS2129.4-0741
MACS0647.8+7015
MACS0744.9+3927
CLJ1226.9+3332
Table 3.1:
R.A.
(J2000)
02:48:03.36
01:31:52.57
11:57:17.26
17:22:27.25
21:29:39.94
08:00:56.83
21:40:15.18
22:48:44.29
19:31:49.66
11:15:52.05
15:32:53.78
17:20:16.95
04:16:09.39
04:29:36.00
12:06:12.28
03:29:41.68
13:47:30.59
13:11:01.67
11:49:35.86
14:23:47.76
07:17:31.65
21:29:25.32
06:47:50.03
07:44:52.80
12:26:58.37
CLASH Clusters
Dec.
zspec
(J2000)
-03:31:44.7 0.189
-13:36:38.8 0.209
+33:36:37.4 0.214
+32:07:58.6 0.224
+00:05:18.8 0.234
+36:03:24.1 0.288
-23:39:40.7 0.315
-44:31:48.4 0.348
-26:34:34.0 0.352
+01:29:56.6 0.353
+30:20:58.7 0.363
+35:36:23.6 0.391
-24:04:03.9 0.396
-02:53:09.6 0.399
-08:48:02.4 0.440
-02:11:47.7 0.450
-11:45:10.1 0.451
-03:10:39.5 0.494
+22:23:55.0 0.544
+24:04:40.5 0.545
+37:45:18.5 0.548
-07:41:26.1 0.570
+70:14:49.7 0.591
+39:27:24.4 0.686
+33:32:47.4 0.890
Restframe
g
i
F475W F850LP
F475W F850LP
F625W F850LP
F625W F850LP
F625W F850LP
F606W F105W
F625W F105W
F625W F105W
F625W F105W
F625W F105W
F625W F105W
F625W F105W
F625W F105W
F625W F105W
F625W F105W
F625W F105W
F625W F105W
F775W F105W
F775W F125W
F775W F125W
F775W F125W
F775W F125W
F775W F125W
F814W F125W
F105W F140W
from Zibetti et al. (2009). The Zibetti et al. (2009) CMLRs were derived using a
Chabrier (2003) initial mass function, and exponential star formation history models with starbursts, using revised 2007 versions of the Bruzual & Charlot (2003)
stellar population synthesis models; these models were updated to include the thermally pulsing asymptotic giant branch (TP-AGB) stellar evolutionary phase based
on isochrones from Marigo & Girardi (2007) and Marigo et al. (2008).
40
3.2. LOW-REDSHIFT CLASH CLUSTERS
Figure 3.1: 1250–2500 Å UV window (vertical purple lines) from Kennicutt (1998),
and filter response curves for g and i band (black curves), shifted to the cluster
redshift listed in each panel. The best-fit HST filters are shown at their observed
frame wavelengths.
In order to calculate the (g − i) color and i-band luminosity required in Equation
3.2, we first determine, for each cluster’s redshift, the HST filters that best match
restframe g- and i-bands. In Figure 3.1, we show two examples of our filter choices,
for clusters RXJ2248.7-4431 and MACS0744.9+3927, at redshifts of z = 0.348 and
z = 0.686, respectively. We redshift the g and i filter curves to each cluster’s redshift,
then select by eye the observed frame HST filters that best match the shifted filters.
41
3.2. LOW-REDSHIFT CLASH CLUSTERS
We list these best matches2 in Table 3.1.
We calculate, for each CLASH object, its restframe (g − i) color, and M/L ratio. To calculate an object’s stellar mass, we must turn its i-band magnitude into a
luminosity. To do this, we use the relation
L
= 4π [D (z)]2 10(−(m+48.6)/2.5) × f req
L
(3.3)
where m is the magnitude, f req is the frequency of the observed filter in Hz, and
D (z) is the luminosity distance of the cluster redshift, which we calculate using the
analytical approximation of Wickramasinghe & Ukwatta (2010). Finally, we estimate
each object’s stellar mass using its i-band M/L ratio and i-band luminosity. This
results in stellar mass estimates for 42483 CLASH objects.
3.2.3
Star Formation Rates
For each cluster’s redshift, we determine which WFC3 broadband filter best respresents the restframe UV, which we take to be the 1250–2500 Å UV window suggested
by Kennicutt (1998). Similar to the method in §3.2.2 for selecting the best HST filters
for g and i band, we shift the UV window to each cluster’s redshift, and determine
by eye which filter is best contained within the window. In Figure 3.1, we show the
redshifted UV window, and the HST filter that fits best, at the redshift of the cluster listed in each panel. For MACS0744.9+3927 (z = 0.686) and CLJ1226.9+3332
(z = 0.89) we select F275W as our UV proxy. For the remaining 23 clusters, we use
F225W.
2
Although F625W is the best g-band proxy for Abell 611, there are no observations in that filter
for this cluster, and we use F606W instead.
42
3.2. LOW-REDSHIFT CLASH CLUSTERS
Using Equation 3.3, we calculate UV luminosities, converting from solar units to
erg s−1 , then use the relation
SF RUV
L
= 4.42 × 10−44
−1
M yr
erg s−1
!
(3.4)
from Murphy et al. (2011) to calculate star formation rates. This results in SFRs for
29876 CLASH objects.
3.2.4
Cluster Membership
For the clusters listed in §3.2.1, for which we have spectroscopic redshifts—except
for Abell 611—we select cluster members using the membership flag in the published
galaxy redshift tables. For Abell 611, no membership flag exists, so we select as
members galaxies with |vcluster − vgalaxy | ≤ 4000 km s−1 , following the velocity cut in
Lemze et al. (2013). These selections result in 121 spectroscopic members across the
five CLASH clusters. However, we note that our final sample has fewer than this
number due to the filter detection requirements in §3.2.2 and §3.2.3.
As all 25 CLASH clusters have photometric redshifts estimates, we use these to
select the remainder of our z < 1 cluster member sample. To better constrain the
accuracy of our photometric redshift-based membership selection, we first use the five
CLASH clusters with spectroscopic redshifts to determine how conservative we must
be in our photometric selection. Each galaxy’s photometric redshift is published with
P (z), an odds parameter, defined as the probability that zphot is contained within
2 × 0.02 (1 + z) (Benítez et al. 2004). Our final sample only contains galaxies with
P (z) > 0.9, which are considered the most reliable (Postman et al. 2012).
43
3.2. LOW-REDSHIFT CLASH CLUSTERS
Figure 3.2: Distribution of spectroscopically confirmed galaxies based on the total
number of filters observed (left panel), and based on the number of filters with 5 σ
detections (right panel). For both panels, the blue filled histograms show the galaxies
with zphot within zspec ± 2 × 0.02 (1 + zspec ), while the red outlined histogram is for
galaxies outside this range.
Additionally, we use this probability to classify the acceptable range for photometric redshifts. If a galaxy’s photometric redshift is within ±2 × 0.02 (1 + zspec ), we
consider it to be correct; otherwise, the galaxy is removed from our sample (see Figure
3.2). We use this same criteria for selecting cluster members for our final sample.
With our correct and incorrect sets defined, we now plot in Figure 3.2 the total
number of filters used to observe each galaxy (left panel), and the number of filters
with 5 σ detections for each galaxy (right panel). Galaxies considered to be genuine
members are plotted as the blue filled histograms, while the remainder are plotted
with the red outlined histograms.
While we find a fairly even distribution for the galaxies with incorrect photometric
44
3.2. LOW-REDSHIFT CLASH CLUSTERS
redshift, more importantly, we find that there is a mimimum limit, in both number
of filters observed, and in number of 5 σ filters, for the correct galaxies. Since there is
a large peak at 7 filters in both panels, we choose this as a minimum cut for number
of filters observed.
In summary, in addition to the spectroscopic members identified above, we select
photometric members that satisfy the following criteria: P (z) > 0.9; zphot that is
within zcluster ± 0.04 (1 + zcluster ); and ≥ 7 filters comprising its HST observations.
We subject our membership criteria on 76546 CLASH objects that meet the stellarity≤ 0.05 cut, and in Figure 3.3, we plot the distribution of photometric and
spectroscopic redshifts for the 2005 total CLASH members. The 121 zspec members
are plotted as the red filled histograms, while the 1884 zphot members are represented
by the blue open histograms. Each cluster’s name and redshift are listed, and the
latter is shown by the downward arrow.
We must combine the 2005 cluster members with the 29876 and 42483 objects
for which we could calculate SFRs and stellar masses, respectively. While we determine cluster membership based on spectroscopic and photometric redshifts, we
independently calculated SFRs and stellar masses based on one restframe UV and
two restframe optical bands, respectively. As we must combine our set of cluster
members with our independently calculated sets of SFRs and stellar masses, which
we were unable to do for all objects, we caution that our final sample will contain
fewer than the 2005 cluster members we find here.
45
3.2. LOW-REDSHIFT CLASH CLUSTERS
Figure 3.3: Distribution of spectroscopic (red histogram) and photometric (blue open
histogram) CLASH members. In each panel, we list the shortened cluster name,
spectroscopic redshift (in parentheses), the number of photometric members, and
spectroscopic members, if any. Arrows show the cluster redshift.
46
3.2. LOW-REDSHIFT CLASH CLUSTERS
3.2.5
Galaxy Morphology
For consistency with our high-redshift cluster sample, we choose to visually classify
the morphology of CLASH galaxies. As there are more than 90000 objects in the
entire CLASH catalog, we first reduce this number to a manageable size. We match
our list of cluster members (2005 galaxies) to our list of stellar mass (42483 objects)
and SFR calculations (29876 objects). This results in an overlap of 1061 CLASH
cluster member galaxies with good stellar mass and SFR estimates.
Following the method we described in §2.3.5, we simultaneously visually inspect
two optical and one NIR HST images, choosing one each of F606W/F625W, F77W/F814W,
and F125W/F140W/F160W. We classify galaxies that have smooth elliptical/S0
shapes at ETGs (735/1061 = 69.3%). Galaxies that show clear spiral structure, or
irregular or distrubed morphologies are collectively classified as LTGs (244/1061 =
23.0%). Galaxies for which we cannot definitively determine morphology are marked
as not classified (82/1061 = 7.7%). We do not include unclassifiable galaxies in the
forthcoming analysis, unless we specify that we are selecting galaxies of all types.
47
Chapter 4
Field Galaxy Data and Sample
Selection
We use the publicly available1 UltraVISTA survey (McCracken et al. 2012; Muzzin
et al. 2013) as the basis of our field comparison sample. UltraVISTA is a deep Ks selected survey covering 1.62 deg2 of the COSMOS field (Scoville et al. 2007), and
has photometry in 30 bands, used to calculate photometric redshifts and infer stellar
masses, and also includes Spitzer photometry in the 24 µm band of MIPS, and all four
bands of IRAC (Muzzin et al. 2013). We select galaxies brighter than the UltraVISTA
calculated 90% completeness limit of Ks,tot = 23.4 AB. The catalog has also been
pruned of stellar objects, contamination from bright stars, and contamination from
nearby saturated objects. Our final UltraVISTA field sample contains 8015 galaxies
over 0.15 < z < 1.5, of which 6194 are at z < 1.
1
http://www.strw.leidenuniv.nl/galaxyevolution/ULTRAVISTA/Ultravista/Data_
Products_Download.html
48
4.1. REDSHIFTS
4.1
Redshifts
Photometric redshifts in UltraVISTA were derived by Muzzin et al. (2013) using
EAZY (Brammer et al. 2008), a template fitting software package. A small subset of
our final sample (N < 1000) has spectroscopic redshifts from the zCOSMOS surveys
(Lilly et al. 2007, 2009).
4.2
Star Formation Rates
The UltraVISTA dataset provides rest-frame UV luminosities. We use Equation 3.4
to convert these luminosities into SFRs.
Following the process in §2.2.3, for every UltraVISTA galaxy which has a 24 µm
MIPS flux, we use Chary & Elbaz (2001) templates to infer total IR luminosities,
then convert these to SFRs using Equation 3.1. As noted in §2.3.4, the 24 µm band
of MIPS has a large point spread function (∼600 , Rieke et al. 2004), making it difficult
to deconvolve IR flux from multiple objects in close proximity. While the effects of
this blending have been significantly reduced in the available photometry (Muzzin
et al. 2013), to be consistent with our ISCS sample construction, we choose to study
isolated galaxies, calculating SFRs only for objects in the UltraVISTA photometric
catalog that have no neighbors within 600 .
Since AGNs can contribute to the MIR flux of a galaxy (as noted in §2.3.2), we use
the IRAC color selection from Stern et al. (2005) to remove from our sample objects
likely harboring an AGN. While AGNs may only account for . 5% of all galaxies (see
§2.3.2), we conservatively remove objects that do not have measured fluxes in all four
IRAC channels (i.e. objects for which we cannot use the Stern et al. (2005) wedge).
49
4.3. STELLAR MASSES
4.3
Stellar Masses
Muzzin et al. (2013) estimated stellar masses by using FAST (Kriek et al. 2009) to
fit galaxy SEDs to template SEDs. The templates were generated with Bruzual &
Charlot (2003) models, using a Chabrier (2003) IMF, and assuming an exponentially
declining SFH. Muzzin et al. (2013) calculated an evolving 95% mass completeness
limit, which we show in the upper panel of Figure 4.2 with the stepped black line.
We cut from our sample any UltraVISTA galaxies that fall below this level.
While stellar masses for both CLASH and UltraVISTA are derived using a Chabrier
(2003) IMF and Bruzual & Charlot (2003) models, the estimation methods—CMLRs
for CLASH galaxies, and SED fitting for UltraVISTA galaxies—differ. We now aim to
test the consistency of using stellar masses derived with these two different methods
by using the CMLR in Equation 3.2 to calculate stellar masses for our UltraVISTA
sample.
Following the procedure in §3.2.2, we first select restframe g- and i-band proxies,
using the best available UltraVISTA broadband optical and NIR filters. We then
calculate, using Equation 3.2, stellar masses for the 8436 isolated UltraVISTA galaxies
with no AGN across 0.15 < z < 1.5.
In the upper panel of Figure 4.1, we plot the stellar masses derived by both the
Muzzin et al. (2013) SED fitting (x-axis) and by using the CMLR from Zibetti et al.
(2009, y-axis). To test how well the stellar masses derived by these two methods
match, we fit the data using a linear least squares approach, plotting the fit (1 σ
uncertainty) with the dark green line (light green shaded region). For comparison,
we show a one-to-one relation by the black dashed line. Within the uncertainty,
stellar masses derived by these two methods are consistent with being the same. In
50
4.3. STELLAR MASSES
Figure 4.1: Upper panel: Stellar masses derived using Equation 3.2 versus stellar
masses derived by SED fitting from Muzzin et al. (2013), shown by the red points.
The dark green line and light green shaded region show a linear least squares fit
to the data, and the 1 σ uncertainty of the fit, respectively. The dashed black line
shows a one-to-one relation. Lower panel: Ratio of SED-derived stellar masses to
CMLR-derived stellar masses as a function of SED-derived stellar mass.
the lower panel, we plot the ratio of SED-derived to CMLR-derived stellar masses as
a function of SED-derived stellar masses. We find that 90.0% of the stellar masses
51
4.4. GALAXY MORPHOLOGY
calculated with these two methods are within a factor of 2.3 of each other.
With this good agreement, we choose to use the existing SED-derived stellar
masses from Muzzin et al. (2013). We have verified that none of our qualitative
results from Chapter 6 depend on this choice.
4.4
Galaxy Morphology
We morphologically classify all our cluster galaxies using visual methods, and while
it is ideal to have a consistent classification method for all three samples in this
work, as we mention above, the UltraVISTA field sample contains ∼8000 galaxies,
so performing visual classification on this set would be incredibly time intensive. We
estimate that it takes approximately an hour to classify 50 low-redshift galaxies, or 20
high-redshift galaxies, so for the ∼6200 z < 1 and ∼1800 z > 1 UltraVISTA galaxies,
visual classifications would take on the order of 200 hours. Instead, as we noted
in §2.3.6, we use the updated, higher-redshift version of the morphological catalog
of Cassata et al. (2007) to separate ETGs and LTGs. Cassata et al. (2007) use a
non-parametric automatic technique to separate galaxies into early types, described
as being elliptical or S0, and late types, which includes all spirals, irregulars, and
merging galaxies. Their method is an extension of Cassata et al. (2005), which used
concentration, C, asymmetry, A, and clumpiness, S (Conselice 2003; Abraham et al.
2003; Lotz et al. 2004). Cassata et al. (2007) includes Gini, a concentration parameter,
which measures how fairly the light is distributed amongst a galaxy’s pixels in the
image, and M20 , a concentration parameter that measures the moment of the brightest
20% of the galaxy flux. Cassata et al. (2005) found, using the C-A-S parameter space,
a locus for typical ETGs. With the two new parameters, this locus has been updated,
52
4.5. STELLAR MASSES OF FINAL GALAXY SAMPLES
and trained against a control sample of 211 galaxies.
To test how well the Cassata et al. (2007) method matches with our visual classifications, we randomly select 300 UltraVISTA galaxies, 100 each in redshift slices
of 0.15 < z < 0.5, 0.5 < z < 1, and 1 < z < 1.5. With no knowledge of the automated technique’s classifications, we visually classify all 300 using the same method
as with our low- and high-redshift cluster galaxies. We find that our overall agreement of 86.3% (259/300) is quite good. The 41 galaxies where the two methods do
not agree include 16 galaxies that we deem visually unclassifiable (16/300 = 5.3%),
which is similar to the unclassifiable fraction of our CLASH galaxies. When considering redshift, we find that we agree with the Cassata et al. (2007) classifications 84%,
94%, and 81% of the time with increasing redshift. With no discernable trend with
redshift, we conclude that our visual classifications are robust against morphological
k-correction.
4.5
Stellar Masses of Final Galaxy Samples
Now that we have calculated stellar masses for all three of our galaxy samples, we plot,
in the upper panel of Figure 4.2, the stellar mass versus redshift for our cluster and
field samples. Cluster ETGs (LTGs) are represented by the red (dark blue) points,
with the two samples separated by the vertical dot-dashed line at z = 1. Field
ETGs and LTGs are shown with the pink and light blue points, respectively. The
blue dashed line is the 80% mass completeness limit for the ISCS galaxies, while the
black stepped line is the evolving UltraVISTA 95% mass completeness limit, binned
in intervals of ∆z = 0.1.
In the lower panel, we plot the median stellar mass as a function of redshift for
53
4.5. STELLAR MASSES OF FINAL GALAXY SAMPLES
Figure 4.2: Upper panel: Stellar mass versus redshift for CLASH and ISCS galaxies,
separated by the vertical dot-dashed line at z = 1, and UltraVISTA field galaxies.
Cluster ETGs are plotted as red points, while LTGs are shown with the dark blue
points. Field ETGs (LTGs) are represented by pink (light blue) points. The UltraVISTA evolving 95% mass completeness limit is shown, binned in intervals of ∆z = 0.1,
by the stepped horizontal line. The 80% mass completeness limit for ISCS is shown
by the dashed blue line. Lower panel: Median stellar mass versus redshift. CLASH
(ISCS) ETGs and LTGs are represented by the filled (open) red circles and blue
squares, respectively, while galaxies of all morphologies are plotted with filled (open)
green triangles. UltraVISTA field galaxies are shown by the filled grey points, with
ETGs (LTGs) as circles (squares), and galaxies of all morphologies as triangles.
all CLASH galaxies, and ISCS and UltraVISTA galaxies above their respective mass
completeness cuts shown in the upper panel. CLASH ETGs and LTGs are plotted
with the filled red circles (blue squares), while galaxise of all type are shown with
54
4.5. STELLAR MASSES OF FINAL GALAXY SAMPLES
the filled green triangles. ISCS galaxies are plotted with open symbols of the same
color and shape. Field ETGs (LTGs) are shown by the grey circles (squares), and
galaxies of all types are represented by the grey triangles. All median mass errors in
this plot are calculated as the quadrature sum of bootstrap resampling and simple
Poisson uncertainty.
55
Chapter 5
Infrared-Ultraviolet Star
Formation Rate Comparison
The available SFRs for our high-redshift cluster sample are derived using 24 µm
Spitzer detections. With only UV-derived SFRs for our CLASH cluster sample, in
this section, we aim to derive a correction factor that will allow us to directly compare
the star-formation activity in both our high- and low-redshift galaxy samples. As
UltraVISTA galaxies have both UV- and 24 µm-derived SFRs, and span the entire
redshift range of our cluster samples, they provides an excellent test bed to compare
the two measurements.
5.1
Motivation
Specific comparisons of the IR and UV SFRs in the literature are rather scant. In
Figure 5.1, we compare the IR-derived SFR versus UV-derived SFRs based on two
56
5.1. MOTIVATION
low-redshift (z . 0.1) studies. The green circles represent 30 galaxies from RosaGonzález et al. (2002), and the blue circles represent 49 galaxies from Iglesias-Páramo
et al. (2004), while the red stars show 82 z < 0.1 UltraVISTA galaxies. These latter
galaxies are not included in 8015 sample galaxies listed in §4.2 above, as we do no
include them in any further analysis beyond this section.
−1
Rosa-González et al. (2002) provide UV fluxes in units of erg s−1 cm−2 Å , and
IR fluxes in erg s−1 cm−2 Hz−1 . Given these units, we calculate SFRs for the RosaGonzález et al. (2002) sample, taking the redshift1 and flux for each galaxy in their
Table 1, and converting the fluxes into luminosities, using the relation
L = 4π [D (z)]2 fλ × Cλ
(5.1)
where D (z) is the luminosity distance, fλ is the flux, and Cλ is a constant which
depends on whether we are calculating IR or UV luminosities. In the UV, we have
Cλ = 1875 Å
(5.2)
as it is the midpoint of our UV window. In the IR, we have
CIR = 1.7 ×
c
60 µm
(5.3)
where c is the speed of light, and the scaling factor of 1.7 is to convert L60µm into
total IR luminosity (Chapman et al. 2000).
1
We use the NASA/IPAC Extragalactic Database (NED) to retrieve redshifts. NED is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the
National Aeronautics and Space Administration.
57
5.1. MOTIVATION
Iglesias-Páramo et al. (2004) provide their UV and IR observations in AB magnitudes, so we use Equation 3.3 to turn them into luminosities, converting from solar
units to erg s−1 . They also provide galaxy distances in physical units, which we use
instead of D (z) in Equation 3.3. Finally, we convert the IR luminosities into SFRs
using Equation 3.1, and the UV luminosities into SFRs using Equation 3.4.
Figure 5.1: IR-derived SFR versus UV-derived SFR for z < 0.1 galaxies from UltraVISTA (red stars), Rosa-González et al. (2002, green circles), and Iglesias-Páramo
et al. (2004, light blue circles).
In Figure 5.1, we plot log (SF RIR /M yr−1 ) versus log (SF RUV /M yr−1 ) for our
z < 0.1 UltraVISTA galaxies (red stars), and the galaxies from Rosa-González et al.
(2002, green circles) and Iglesias-Páramo et al. (2004, light blue circles). For each of
58
5.1. MOTIVATION
the three galaxy samples, we fit the data, in log space, with a simple linear relation
log SF RIR /M yr−1 = a × log SF RUV /M yr−1 + b ± σIR
(5.4)
where a and b are the slope and y-intercept of the line, respectively, and σIR is the
1 σ uncertainty in log (SF RIR /M yr−1 ). We plot the three fits with the same color
as their respective points. We refrain from plotting the uncertainty ranges of our
three samples for clarity. We combine the galaxies of all three sets, fitting them with
Equation 5.4, which we plot as the solid black line, with the shaded grey region as the
1 σ uncertainty. We find that while there is substantial overlap in the data points for
the three samples, the fits have a wide range in slope and y-intercept. We list their
coefficients and 1 σ uncertainties in Table 5.1.
Table 5.1: Results of log (SF RIR /M yr−1 ) = a × log (SF RUV /M yr−1 ) + b ± σIR
linear least squares fit to z < 0.1 galaxy sample
Sample
Ngal
UltraVISTA z < 0.1
82
Rosa-González et al. (2002)
30
Iglesias-Páramo et al. (2004) 49
All
161
a
b
σ
1.29 0.30 0.50
1.02 1.01 0.71
0.74 0.44 0.59
1.00 0.41 0.70
With the large sample-to-sample scatter, it is likely unwise to choose only one fit
in this redshift slice to use as a SF RUV → SF RIR correction. Instead, given that the
fit to all 161 galaxies encompasses each individual fit, it seems apparent that choosing
as large a sample as possible is the most beneficial. We use this as motivation for the
remainder of this chapter.
59
5.2. EMPIRICAL FITS
5.2
Empirical Fits
We first fit Equation 5.4 for all 0.15 < z < 1 UltraVISTA galaxies, resulting in a slope
and y-intercept of 0.66 and 0.79, respectively, with σIR = 0.47. While the y-intercept
is with the range of the first three samples (Table 5.1), the slope is lower than all
three, and also significantly lower than the fit to z < 0.1 UltraVISTA galaxies. We
take this as suggestive that there is a potential redshift dependence between UV and
IR SFRs.
Figure 5.2: IR-derived SFRs versus UV-derived SFRs for 0.15 < z < 1 COSMOS/UltraVISTA field galaxies. Points are color coded based on the redshift bin
in which they lie (see legend).
To determine whether this is the case, we show in Figure 5.2 log (SF RIR /M yr−1 )
versus log (SF RUV /M yr−1 ) for all 0.15 < z < 1 UltraVISTA galaxies, separating
them into the redshift bins shown in the bottom right corner. We notice immediately
60
5.2. EMPIRICAL FITS
that as redshift increases, the mimimum values of both quantities increase.
Figure 5.3: IR-derived versus UV-derived SFRs for UltraVISTA galaxies, split into
six non-overlapping redshift bins. For each redshift slice, we fit Equation 5.4 to the
data. In each panel, we show the fit and 1 σ uncertainty with a black line and grey
shaded region, respectively.
61
5.2. EMPIRICAL FITS
Given this result, in Figure 5.3, we show log (SF RIR /M yr−1 ) as a function of
log (SF RUV /M yr−1 ) for each redshift slice individually, with the same color scheme
as in Figure 5.2. We fit Equation 5.4 for each redshift slice, and plot this as the solid
black line in each panel, with the grey shaded region showing the 1 σ uncertainty.
In each panel, we list the redshift slice, and the slope, y-intercept, and uncertainty
of the fit. While there is potentially a downward trend in a with increasing redshift,
we see a steadily increasing b from z = 0.15 → 1. We consider these coefficients to
be our correction factors for SF RUV → SF RIR . Rewriting Equation 5.4 yields an
alternative form, in linear space
SF RIR /M yr−1 = 10b±σIR SF RUV /M yr−1
62
a
(5.5)
Chapter 6
Analysis
6.1
Projected Radius
Given HST ’s small footprint, we only have coverage of the inner ∼650 (∼800) kpc
of our CLASH (ISCS) clusters, so our analysis is largely limited to the cores of these
clusters. While we can only probe out to rproj ' 650 kpc in our CLASH clusters,
we choose to include ISCS galaxies at 650 < rproj < 800 kpc, due to the small ISCS
cluster sample. We verify that our qualitative results do not change by including
these galaxies.
6.2
Galaxy Morphology Versus Projected Radius
While we must select isolated galaxies when measuring IR SFRs (see §2.3.4 and
§4.2), we are under no such constraint when determining morphology through visual
classification. Thus, despite not including some cluster galaxies when investigating
star formation activity, we choose to include all ISCS cluster galaxies, regardless of
isolation, when exploring morphology. This increases our ISCS sample size from 104
63
6.2. GALAXY MORPHOLOGY VERSUS PROJECTED RADIUS
to 191 galaxies, which helps to reduce the random error inherent in a small sample
size.
Figure 6.1: Fraction of galaxies classified as early-type versus cluster-centric radius
for CLASH (red filled circles) and ISCS (red open circles). The vertical error bars
show the binomial error in the fraction, while the horizontal error bars show the
sizes of the rproj annuli. The fractions (binomial confidence) for 0.15 < z < 1 and
1 < z < 1.5 UltraVISTA field galaxies are shown by the red horizontal lines (pink
shaded regions).
In Figure 6.1 we plot the fraction of galaxies classified as early-type as a function
of projected cluster radius for CLASH and ISCS galaxies, as red filled circles and
red open circles, respectively. The vertical error bars show the binomial error in the
fractions, and we note that for some of the CLASH points these errors are small
enough to be covered by the point itself.
We radially bin our ISCS galaxies into two non-overlapping annuli, with the bin
64
6.3. STAR FORMATION RATE VERSUS PROJECTED RADIUS
size shown by the horizontal error bars. Our choice for this binning will be explained
in §6.4. We use four non-overlapping bins for our CLASH galaxies, with the inner
three bins all being 125 kpc in size. The outermost CLASH bin extends from 375
to 650. We plot the early-type fractions for low- (0.15 < z < 1) and high-redshift
(1 < z < 1.5) UltraVISTA field galaxies as the red horizontal lines, with the pink
shaded regions showing the binomial errors in the fractions.
We find that both the ETG fractions of both CLASH and ISCS clusters decrease
with increasing projected radius, dropping by 37% and 29%, respectively, over the
range probed. At all radii probed, CLASH clusters consistently have higher ETG
fractions than our high-redshift cluster sample, with fractions a factor of ∼1.4 to 1.8
times higher.
In the outermost bin of both the CLASH and ISCS clusters, there is a ∼30–45
percentage point difference between the ETG fraction in clusters and in the field,
over the same redshift ranges. These large differences are further evidence that we
are looking mainly at the cores of these clusters, as we noted in §6.1. Indeed, virial
radii have been measured for 19 of the 25 CLASH clusters (Merten et al. 2014); with
an average virial radius of 1.2 Mpc, we have typical coverage of only the inner 50%.
For ISCS clusters, which have virial radii on the order of 1 Mpc (see reference in
§2.4.3), our sampling covers the inner 80%.
6.3
Star Formation Rate Versus Projected Radius
As is well known, star formation activity in clusters tends to increase with projected
cluster radius. In Figure 6.2 we plot log(SFR/M yr−1 ) versus projected cluster radius
for CLASH cluster members. We show the ETGs (LTGs) in the upper (lower) panel.
65
6.3. STAR FORMATION RATE VERSUS PROJECTED RADIUS
Figure 6.2: Star formation rate versus radius for CLASH ETGs (upper panel; red
points) and CLASH LTGs (lower panel; blue points). The solid black lines show
linear least squares fits to the data in each panel. The pink (light blue) shaded
regions show the 1 σ uncertainty in the ETG (LTG) fit.
To determine if there exists a relation between star formation activity and projected
radius, we first perform a Spearman’s rank test on CLASH cluster members, between
log (SFR/M yr−1 ) and projected radius, rproj . We find correlation coefficients of
rs = 0.19 ± 0.02 and rs = 0.33 ± 0.04, respectively, for ETGs and LTGs, both at the
100% confidence level. Thus only a mild correlation between star formation activity
66
6.4. MEAN STAR FORMATION RATE VERSUS PROJECTED
RADIUS
and rproj is found. We quantify this correlation through a linear least squares fit to
log SFR/M yr−1 = a × rproj + b
(6.1)
for each morphological subset, where a and b are the slope and y-intercept, respectively. In each panel, the fit is represented by the solid black line, while the shaded
pink (light blue) regions show the 1 σ uncertainty in the ETG (LTG) fit. We find
that the slope of Equation 6.1 is approximately the same (m ' 0.001) for both ETGs
and LTGs. While these results may suggest a radial dependence in star formation
activity, the uncertainties are too large to draw any certain conclusions, in agreement
with the low correlation coefficients found above.
6.4
Mean Star Formation Rate Versus Projected Radius
The large uncertainties that we found when fitting individual SFRs may be reduced
by binning our data, as we did in §2.4. In Figure 6.3, we show mean SFR as a function
of cluster-centric radius for our low- and high-redshift cluster samples, using filled and
open symbols, respectively, for each subset, with the colors and styles shown in the
figure legend. Due to the relatively large scatter in ISCS SFRs, we separate these
galaxies into two radial annuli, such that each bin has approximately equal signal-tonoise, with the bins shown by the horizontal error bars on the ISCS LTG points. The
vertical error bars show the error estimated by 5000 iterations of bootstrap resampling
the galaxies in each bin, added in quadrature with a simple Poisson error. With the
larger sample size of CLASH galaxies, we are not as constrained when determining
the size of our bins. However, in order to both compare with the ISCS clusters, and
have the ability to detect a potential radial trend in the data, we choose our inner
67
6.4. MEAN STAR FORMATION RATE VERSUS PROJECTED
RADIUS
three bins to each be 125 kpc in size, while the outer bin extends from 375 to 650
kpc. We plot the mean SFR of our comparison low- and high-redshift field samples to
the right of the vertical dashed line. We estimate errors for this sample in the same
manner as with our cluster galaxies.
Figure 6.3: Mean SFR versus projected cluster radius for CLASH (filled symbols)
and ISCS (open symbols) galaxies, left of the vertical dashed line. The mean SFRs
for 0.15 < z < 1 (filled symbols) and 1 < z < 1.5 (open symbols) field galaxies
are shown to the right of the vertical dashed line. The horizontal error bars on the
ISCS and CLASH LTG points show the size of the radial bins for the high- and lowredshift cluster samples, respectively. The errors in this plot are the quadrature sum
of bootstrap resampling and simple Poisson errors.
We find that the mean SFR of CLASH galaxies is relatively flat across 0 < rproj <
375 kpc, but we observe a noticeable increase in the outermost bin. Given the errors,
we cannot determine if this rise is statistically significant for ETGs, but for LTGs,
the rise is at least 0.7 M yr−1 between the two outer bins, a 23% increase. When
68
6.4. MEAN STAR FORMATION RATE VERSUS PROJECTED
RADIUS
considering all morphologies, the rise is similar, however with an increase of 15%,
it is noticeably tempered by the large number of ETGs (59%) that make up the
375 < rproj < 650 kpc annulus.
A suprising result—given the large differences between mean SFR for ETGs and
LTGs in ISCS clusters—is that in the inner three annuli, mean SFRs for CLASH ETGs
and LTGs are indistinguishable within the errors, and it is not until the outermost
bin that they become statistically distinct. We are combining galaxies over a very
large swath of cosmic time (∼6 Gyr), so as a simple test, we perform our analysis
in Figure 6.3, cutting our sample at z = 0.5. However, aside from overall changes
in mean SFR with time—as we will discuss in §6.7—we observe roughly the same
radial behaviour both above and below this redshift. We surmise that instead of our
averaging over 0.15 < z < 1 causing variations in ETG and LTG mean SFRs, we are
likely observing the quenching of LTGs as they travel through the cluster, particularly
through the cores. This idea is reinforced by Koopmann & Kenney (2004), who found
that spiral galaxies in the Virgo cluster have truncated H i disks, relative to spirals in
the field, and suggest that interactions with the cluster medium are likely responsible
for removing the outer gas from these cluster spirals.
With observations of the cluster outskirts, we would be able to more accurately
constrain the relative mean SFRs, and determine whether, and at what radial extent,
they increase to field levels. However, with the data available to us, we find that at all
projected radii probed, CLASH galaxies, regardless of morphology, have mean SFRs
lower than that of the field. Even with the increase in mean SFR at rproj > 375 kpc,
cluster ETGs have mean SFRs a factor of at least 1.2 times lower than field ETGs.
Cluster LTGs have mean SFRs a factor of at least 1.9 times lower than field LTGs.
69
6.5. SPECIFIC STAR FORMATION RATE VERSUS PROJECTED
RADIUS
We will postpone the main discussion on the evolution of cluster star formation
activity until §6.7 and §6.8, but we do note here that similar to CLASH galaxies,
ISCS ETGs are quenched relative to the field at all cluster radii, with mean SFRs
a factor of 2.0 to 3.0 times lower than that of the 1 < z < 1.5 UltraVISTA ETGs.
This result is similar to that found in §2.4.2, where we compared ISCS galaxies to the
ETG subset of UltraVISTA field galaxies at 1 < z < 1.5. Unlike high-redshift ETGs,
and all low-redshift cluster galaxies, ISCS LTGs show star formation activity that is
either at the level of the field, or higher. With our inclusion of disturbed/irregular
galaxies in the LTG morphological type, these enhanced SFRs could be due to either
very recent merger activity, or even ongoing mergers that are sufficiently progressed
to appear as a single object.
6.5
Specific Star Formation Rate Versus Projected Radius
Specific star formation rate is the SFR per unit stellar mass, which we calculate by
dividing the sum of the SFRs by the sum of the stellar masses. We now turn to the
sSFR of our field and cluster samples, as it allows us to examine how efficiently these
galaxies are converting their cold gas into new stars.
In Figure 6.4, we plot the sSFR for the same samples, using the same radial
binning, symbols, and error calculation, as in Figure 6.3. While the mean SFRs
of CLASH ETGs and LTGs were nearly indistinguishable at all projected radii, we
find that the mass normalized star formation activity of these two subsets are quite
different, at all but the most central annulus. Here, the error of the LTG subset is
large enough that the two morphological types are formally consistent with having
the same sSFR. These errors also preclude us from making any conclusions on any
70
6.5. SPECIFIC STAR FORMATION RATE VERSUS PROJECTED
RADIUS
Figure 6.4: Specific SFR versus projected cluster radius for CLASH (filled symbols)
and ISCS (open symbols) galaxies, left of the vertical dashed line. We use the same
radial binning as in Figure 6.3. The sSFRs for 0.15 < z < 1 (filled symbols) and
1 < z < 1.5 (open symbols) field galaxies are shown to the right of the vertical
dashed line.
sort of radial sSFR trend for LTGs. Within the errors, we cannot determine any
form of trend in for CLASH ETGs either, at least in the inner three annuli. The
375 < rproj < 650 kpc bin, however, shows a clear increase above the middle two bins.
As we found with mean SFR, the sSFRs of CLASH galaxies are quenched relative
to the field at all projected radii. CLASH LTGs have sSFRs a factor of at least 9.4
lower that the sSFR of field LTGs, while for CLASH ETGs, that factor is at least
4.3, relative to field ETGs.
71
6.6. GALAXY MORPHOLOGY VERSUS REDSHIFT
6.6
Galaxy Morphology Versus Redshift
In Figure 6.5 we show the fraction of galaxies classified as early-type as a function
of redshift for CLASH (red filled circles), ISCS (red open circles), and UltraVISTA
(pink filled circles) samples. The vertical error bars show the binomial error in the
fractions. For galaxies below (above) z = 1, we use the same redshift bins for the
CLASH (ISCS) and UltraVISTA samples. For all samples, the bins are shown by the
horizontal error bars on the UltraVISTA points.
Figure 6.5: Fraction of galaxies classified as early-type versus redshift for CLASH (red
filled circles) and ISCS (open red circles) clusters, and the UltraVISTA field (filled
pink circles). Comparison ETG fractions from the literature, spanning 0.33 < z <
1.46, are plotted with the symbols shown in the legend.
We include comparison ETG fractions for galaxy clusters from the literature,
spanning the vast majority (0.33 < z < 1.46) of the redshift range covered by our
72
6.6. GALAXY MORPHOLOGY VERSUS REDSHIFT
Table 6.1: Comparison Cluster ETG Fractions From the Literature
Cluster
Name
Cluster
Redshift
fETG
Ngal
Reference
Cl 1358+62
0.33
0.71+0.04
−0.04
138
Fabricant et al. (2000)
A370 2
0.37
0.48+0.07
−0.07
71
Dressler et al. (1997)
Cl 1446+26
0.37
0.57+0.05
−0.05
107
Dressler et al. (1997)
Cl 0024+16
0.39
0.61+0.04
−0.04
170
Dressler et al. (1997)
Cl 0939+47
0.41
0.55+0.05
−0.05
124
Dressler et al. (1997)
Cl 0939+47 2
0.41
0.42+0.07
−0.06
72
Dressler et al. (1997)
0.42
0.58+0.06
−0.06
93
Dressler et al. (1997)
3C 295
0.46
0.72+0.05
−0.06
87
Dressler et al. (1997)
Cl 0412-65
0.51
0.54+0.06
−0.06
91
Dressler et al. (1997)
Cl 1601+42
0.54
0.45+0.06
−0.05
100
Dressler et al. (1997)
Cl 0016+16
0.55
0.73+0.03
−0.04
193
Dressler et al. (1997)
Cl 0054-27
0.56
0.52+0.05
−0.05
119
Dressler et al. (1997)
CLJ1324+3011
0.76
0.55+0.17
−0.14
22
Lubin et al. (2002)
MS1054-0321
0.83
0.78 ± 0.11
130
Postman et al. (2005)
RXJ0152.7-1357
0.84
0.72 ± 0.11
125
Postman et al. (2005)
Cl 1604+4304
0.90
0.57 ± 0.11
124
Postman et al. (2005)
Cl 1604+4321
0.92
0.39 ± 0.10
150
Postman et al. (2005)
J1229+0151
0.98
0.80+0.06
−0.08
46
Cerulo et al. (2014)
RDCS J0910+5422
1.10
0.38 ± 0.10
146
Postman et al. (2005)
RDCS J1252-2927
1.24
0.68 ± 0.14
67
Postman et al. (2005)
RX J0849+4452
1.27
0.30 ± 0.09
214
Postman et al. (2005)
XMMXCS J2215.9−1738
1.46
0.62 ± 0.17
39
Hilton et al. (2009)
Cl 0303+17
CLASH and ISCS clusters. In Table 6.1 we list the additional clusters used in Figure
6.5, their redshifts, ETG fractions (fETG ), number of galaxies in the sample (Ngal ),
and the reference for the morphological data. For all comparison samples, we use
73
6.6. GALAXY MORPHOLOGY VERSUS REDSHIFT
classifications that match our own as closely as possible. For MS 1054-0321, RX
J0152-1357, CL 1604+4304, CL 1604+4321, RDCS J0910+5422, RDCS J1252-2927,
and RX J0849+4452, we use the Postman et al. (2005) fractions of galaxies classified
as either elliptical or S0, and their errors, which are the quadrature sum of counting
statistics and classification uncertainty. For XMMXCS J2215.9−1738 we use the
Hilton et al. (2009) fraction of galaxies classified as either elliptical or lenticular (S0),
and their error, which is the quadrature sum of Poisson, classification, and cluster
membership uncertainties. For the remainder of the classifications, we determine
the fraction of ETGs by adding the number of galaxies classified as elliptical with
those classified as S0/bulge-dominated, then dividing this sum by the total number
of galaxies in the sample. For these fractions, the errors listed in Table 6.1 are our
calculated binomial errors. Despite the large cluster-to-cluster scatter, we find good
agreement between our morphological fractions and those from the literature.
We note that the majority of the 0.37 < z < 0.56 ETG fractions from Dressler
et al. (1997) fall below our points. We surmise that this may be a selection effect, as
the CLASH clusters are quite morphologically evolved, most having been selected to
be massive (Postman et al. 2012). To see if it is the case that there are significant
differences in total mass for these two samples, we compare the CLASH and Dressler
et al. (1997) clusters for which we could find total mass estimates. 19 of the 25
CLASH clusters have total mass estimates (Merten et al. 2014), with a median of
6.0 × 1013 M . A literature search reveals total mass estimates for five of the Dressler
et al. (1997) clusters. Four cluster have total mass estimates from Comerford et al.
(2006): A370 2 (1.1 × 1014 M ), Cl 0939+4713 (2.7 × 1013 M ), Cl 0016+1609
(3.3 × 1013 M ), and Cl 0054-27 (2.6 × 1013 M ). Comerford et al. (2006) listed two
74
6.7. MEAN STAR FORMATION RATE VERSUS REDSHIFT
or three mass estimates for each cluster, so we take the median value in each case.
The remaining mass estimate is for Cl 00214+16 (1.1×1014 M ; Broadhurst et al.
2000). The median total mass of these five clusters is 3.3 × 1013 M , almost a factor
of two smaller than the median total mass of CLASH clusters.
As others have found (see references in Table 6.1), the vast majority of low- to
moderate-redshift cluster galaxies are morphologically early-type. In our clusters, we
find that from z = 1.5 → 0.15, the ETG fraction rises from 0.40±0.05 to 0.80 ± 0.03,
a 100% increase. In the field, the fraction increases from 0.05±0.01 to 0.31+0.02
−0.01 . With
6 and 12% increases at z = 1.25 and z = 0.36, respectively, we find that the bulk of
the morphological evolution in the field occurs over z = 1 → 0.57. Despite this, at all
redshifts we study, field ETG fractions are 3 to 10 times lower than those of cluster
ETGs.
While field galaxies apparently undergo the bulk of their morphological evolution
at moderate redshift (0.6 . z . 1), we find that there are no increases larger than
15% between bins for z < 1 clusters. This more gentle increase is in agreement with
the concept that cluster populations evolved at earlier cosmic times. Additionally,
the 64% increase over z = 1.5 → 1 further suggests that this is an epoch of enchanced
galaxy formation and evolution, as suggested in Chapter 2.
6.7
Mean Star Formation Rate Versus Redshift
Figure 6.6 shows the mean SFR for our low- and high-redshift cluster samples, with
the filled (open) red circles, blue squares, and green triangles representing the CLASH
(ISCS) ETGs, LTGs, and galaxies of all morphologies, respectively. We show our field
sample, over 0.15 < z < 1.5, as the filled grey circles. We calculate errors as in Figure
75
6.7. MEAN STAR FORMATION RATE VERSUS REDSHIFT
6.3, and use the same redshift binning as in Figure 6.5, except for 1 < z < 1.5, which
we combine into one bin.
Figure 6.6: Mean SFR versus redshift for ETGs, LTGs, and galaxies of all morphologies in CLASH and ISCS clusters, and the UltraVISTA field.
We find that all cluster galaxies are quenched relative to the field at z < 1, and
even at 1 < z < 1.5, cluster ETGs have mean SFRs lower than that of field ETGs.
Cluster LTGs, however, show a significant increase in star formation activity relative
to the field in this high-redshift range. By combining both z > 1 bins, and averaging
across all high-redshift galaxies—thus driving down the random error—we find the
mean SFR of cluster LTGs is 51.2 ± 10.6 M yr−1 , a factor of 1.8 higher than field
galaxies of the same morphology. Even if we cut ISCS LTGs with rproj > 500 kpc—to
match the ∼50% virial radius coverage of CLASH clusters—we find that these ‘core’
LTGs have a mean SFR (44.8 ± 11.7 M yr−1 ) only 12.5% lower than that of the
76
6.7. MEAN STAR FORMATION RATE VERSUS REDSHIFT
full high-redshift LTG sample, and still a factor of 1.6 higher than the 1 < z < 1.5
field LTG mean SFR. This relative SFR enchancement brings up the mean SFRs of
all cluster galaxies to field levels over this redshift range. This relatively high mean
SFR, which we previously showed in Figure 2.7, is made even more drastic when
compared with mean cluster SFRs at z < 1, further providing evidence for the epoch
of enhanced star formation activity at z ∼ 1.4, as proposed by B13.
In Figure 6.7, we now plot the mean SFR as a function of redshift for galaxies of
all morphology for CLASH (filled green triangles) and ISCS (open green triangles)
galaxies. As mentioned in §2.5.2, A14 studied 274 clusters over 0.3 < z < 1.5, measuring IR-derived SFRs through 250 µm stacking. In so doing, they fit the evolution
of the mean SFR of both cluster members and field galaxies, using a function of the
form
y = βeαt
(6.2)
where t is cosmic time in Gyr, and y = hSF Ri. A14 fit to both the cluster ‘core’
(rproj < 500 kpc), and the ‘core+outskirts’ (rproj < 1000 kpc). Because we are limited
to the inner 650 (800) kpc in our CLASH (ISCS) clusters, we choose to use the A14
‘core’ fit, and limit ourselves to galaxies in the inner 0.5 Mpc of our clusters. With this
cut, our low- and high-redshift cluster samples have 1032 and 80 galaxies, respectively.
Despite this, by combining LTGs and ETGs, we are still able to finely bin our cluster
data, with 6 (4) evenly-sized redshift slices for our z < 1 (z > 1) cluster samples. We
use the same redshift binning for our field galaxies, plotting their mean SFRs as the
grey triangles.
77
6.7. MEAN STAR FORMATION RATE VERSUS REDSHIFT
Figure 6.7: Mean SFR vs redshift for cluster (green triangles) and field (grey triangles)
of all types. The solid (dashed) black curve is the mean SFR vs. redshift fit (1 σ
uncertainty) for core (rproj < 0.5 Mpc) cluster galaxies over 0.3 < z < 1.5 from A14.
The solid grey curve and light grey shaded region is the fit and 1 σ uncertainty,
respectively, for field galaxies over the same redshift range from A14.
We show the cluster fit and its 1 σ uncertainty,
hSF Ri = (810 ± 400) e(−0.66±0.08)t ,
(6.3)
as the solid and dashed curves, respectively. The dark grey curve shows the fit to the
A14 field sample, with its 1 σ uncertainty plotted as the light grey shaded band.
With our higher-resolution bins, we find a relatively steady increase in mean SFR
with redshift for both cluster and field galaxies over 0.15 < z < 1.5. At all redshifts,
we see excellent agreement between the A14 fit of stacked Herschel mean SFRs and
78
6.8. SPECIFIC STAR FORMATION RATE VERSUS REDSHIFT
both our low-redshift UV→IR-corrected mean SFRs, and our high-redshift 24 µmderived mean SFRs.
The mean SFRs of the UltraVISTA field galaxies show moderate agreement to
the A14 field fit, with only slight disagreement in four redshift bins. A14, however,
fit the field sample with two versions of Equation 6.2, allowing for a break at z = 0.8.
The lower mean SFRs between 0.8 . z . 1.1 may indeed be a similar such break,
only with a larger magnitude in the UltraVISTA sample. The UltraVISTA galaxies
also appear to have a similar field-cluster crossover at z ∼ 1.25.
6.8
Specific Star Formation Rate Versus Redshift
As in §6.1, we now turn to specific SFRs, in this case to investigate the cosmic
evolution of the star formation efficiency of our cluster galaxies.
In Figure 6.8 we plot the mass normalized SFR versus redshift for our cluster and
field samples, with the same symbol types, and binning, as in Figure 6.6. In fact, the
large uncertainties in sSFR, specifically for cluster LTGs, drove our decision to use
only four z < 1 bins.
As with their mean SFRs, CLASH cluster galaxies have quenched sSFRs relative
to the field—of the respective morphological type—across 0.15 < z < 1. We similarly
see a large enhancement in the star formation efficiency of LTGs at z > 1. Despite
having a mean SFR a factor of ∼2 greater than that of the field, high-redshift LTGs
have median stellar masses 2.5 times higher than z > 1 field galaxies (see Figure 4.2),
which accounts for the sSFR that reaches, but does not exceed, the level of the field.
Across 0.15 < z < 1 CLASH LTGs have median stellar masses 4 to 10 times
lower than CLASH ETGs (see Figure 4.2). So despite seeing little-to-no separation
79
6.8. SPECIFIC STAR FORMATION RATE VERSUS REDSHIFT
Figure 6.8: Specific SFR versus redshift for ETGs, LTGs, and galaxies of all morphologies in CLASH and ISCS clusters, and the UltraVISTA field. Redshift bins, and
symbols are the same as in Figure 6.7.
amongst the mean SFRs of early- and late-types, we find LTG sSFRs in the lower
three redshift bins that are 3 to 5 times higher than that of ETGs.
Over this same period of z ∼ 0.75 → 0.25, CLASH galaxies show little-to-no
evolution regardless of morphology. Within the errors, we cannot determine whether
there is a radial trend for LTGs, and ETGs only drop by a factor of 1.6. However,
from z ∼ 1.25 to ∼ 0.75, both morphological types experience significant decreases
in their sSFR, with ETGs dropping nearly an order of magnitude (9.4), and LTGs
falling by a factor of 15.6.
Figure 6.9 shows sSFR versus redshift for galaxies of all morphologies in the cluster
(green triangles) and field (grey triangles), using the same redshift binning as in Figure
6.7. A14 used Equation 6.2, letting y = sSF R, to fit the evolution of their measured
80
6.8. SPECIFIC STAR FORMATION RATE VERSUS REDSHIFT
Figure 6.9: Specific SFR versus redshift redshift for CLASH and ISCS cluster galaxies
of all morphologies (filled and open green triangles, respectively), and UltraVISTA
field galaxies (grey triangles). The solid (dashed) black curve is the sSFR vs. redshift
fit (1 σ uncertainty) for core (rproj < 0.5 Mpc) cluster galaxies over 0.3 < z < 1.5 from
A14. The solid grey curve and light grey shaded region is the fit and 1 σ uncertainty,
respectively, for field galaxies over the same redshift range from A14. The gold shaded
region shows the main sequence fit to the sSFR redshift evolution from Elbaz et al.
(2011).
cluster (rproj < 0.5 Mpc) and field sSFRs. We show these fits using the same style as
we did in Figure 6.7.
Elbaz et al. (2011) measured the sSFR redshift evolution of galaxies observed in
the northern and southern fields of the Great Observatories Origins Deep Survey.
Their best fit sSFR evolution is described by
h
i
sSF R Gyr−1 = 26 × t−2.2
81
(6.4)
6.8. SPECIFIC STAR FORMATION RATE VERSUS REDSHIFT
where t is cosmic time. They classify galaxies with sSFR greater than twice Equation 6.4 to be starbursts, and galaxies below half Equation 6.4 to be galaxies with
‘significantly lower’ star formation. This gives a range of
h
i
13 × t−2.2 ≤ sSF R Gyr−1 ≤ 52 × t−2.2
(6.5)
which defines, according to Elbaz et al. (2011), the ‘main sequence’ of galaxies. We
plot the main sequence of Equation 6.5 as the gold shaded region in Figure 6.9.
We find that our UltraVISTA sSFRs lie in excellent agreement with the Elbaz
et al. (2011) main sequence, while they are almost uniformly higher than the field
sSFR fit from A14.
A14 imposed a mass cut of M? = 1.3 × 1010 M across their entire field sample,
while we adopt an evolving mass cut that ranges from M? = 1.0 × 1010 M at z = 1.5
down to M? = 2.3 × 108 M at z = 0.15, as shown in the top panel of Figure 4.2. Our
inclusion of lower mass galaxies mostly accounts for the somewhat enhanced sSFR
values of our field sample relative to A14’s field fit. By imposing stellar mass cut of
M? = 1.3×1010 M , we find that the sSFR of our field sample agrees within the errors
with the A14 field sSFR fit in all but three redshift bins, a moderate improvement
over the results with the UltraVISTA evolving stellar mass cut.
We find that from z = 1.5 to z ∼ 0.35 cluster galaxies exhibit a steady quenching,
with good agreement to the A14 cluster core sSFR fit down to z ∼ 0.9. In the lowestredshift bin, we see a slight enhancement in the sSFR of cluster galaxies relative to
the sSFR at z ∼ 0.35. However, we suggest that the lack of completeness in high-mass
CLASH galaxies in this redshift bin (see Figure 4.2) is the cause of this apparent rise
in sSFR.
82
Chapter 7
Summary and Conclusions
We have used a sample of 11 high-redshift (1 < z < 1.5) ISCS clusters, 25 low-redshift
(0.15 < z < 1.0) CLASH clusters, and 8015 low- and high-redshift UltraVISTA field
galaxies to explore the evolution of cluster star formation activity over 0.15 < z < 1.5.
We selected CLASH members by either spectroscopic redshifts, if available, or
with photometric redshifts. We similarly preferentially used spectroscopic redshifts
to select ISCS members; red-sequence membership was used for galaxies with no
spectroscopic redshift.
We separated our cluster galaxies into late- and early-type morphologies by visually inspecting them in optical and NIR HST images.
We used the updated
CAS/Gini/M20 classifications from Cassata et al. (2007) to determine field galaxy
morphologies, using the same two coarse morphological bins. We visually inspected a
random subset of 300 UltraVISTA galaxies, and found good agreement (86%) between
the two methods.
We used 24 µm imaging to measure the dust-obscured SFRs for our ISCS and
83
UltraVISTA samples, excluding galaxies that had nearby neighbors, or likely harbored an AGN. We also used rest-frame UV photometry to calculate UV SFRs for
UltraVISTA and CLASH galaxies. We used the two sets of SFRs from UltraVISTA
and performed a linear least squares fit to the IR vs UV SFRs, binned in redshift.
We used the fit as an empirical calibration to correct the CLASH UV SFRs, which
allowed them to be directly compared to the ISCS and UltraVISTA IR-derived SFRs.
We used SED template fitting to estimate stellar masses for our ISCS sample. The
same methodology was used to calculate stellar masses in the UltraVISTA catalog.
We derived stellar masses for our CLASH sample by using a Zibetti et al. (2009)
CMLR, taking advantage of the 16 available HST filters to measure rest-frame (g − i)
colors, and calculate i-band luminosities, across the entire CLASH redshift range.
We began our analysis in Chapter 2 by focusing on the star formation activity
of 1 < z < 1.5 cluster and field galaxies. By comparing the mean SFR of ISCS
ETGs and LTGs, and UltraVISTA ETGs, we found that high-redshift cluster ETGs
are significantly quenched, relative to both field early types, and cluster late types.
Despite their relatively quenched SFRs, ETGs still account for 12.7% of the total SFR
observed in ISCS clusters. We found that the fraction of ISCS ETGs that are star
forming (SF R > 26 M yr−1 ) decreases from 28.6% at 1.35 < z < 1.5 to 10.5% by
1.16 < z < 1.35, as more of the early-type cluster population becomes quenched. Over
the same period, however, their mean and specific SFRs remain relatively unchanged,
which we suggested implies that some mechanism must be acting to increase the
SFRs of the remaining star-forming ETGs. Not only are ISCS ETGs experiencing
the same overall star formation activity during this epoch, but also the fraction of
cluster galaxies that are morphologically early type increases from 34.1% to 55.9%.
84
We concluded that these results were consistent with a scenario where new ETGs
are being created through major merging, and that the mergers must be fairly recent
and gas-rich to account for the enhanced star formation activity of the star-forming
subset.
Because of the small footprint of HST, we were only able to observe the inner
∼ 650 kpc of CLASH clusters. Based on virial radius estimates by Merten et al.
(2014), which give an average virial radius of 1.2 Mpc for the 19 CLASH clusters
they studied, we are only considering the inner 50% of low-redshift clusters. At all
radii that we can probe, we find that all z < 1 cluster galaxies have mean SFRs that
are quenched relative to the field. Furthermore, at cluster-centric radii ≤ 375 kpc,
the mean SFR of LTGs is indistinguishable, within the errors, from that of ETGs.
We suggest that the LTGs interior to ∼ 400 kpc are likely being quenched due to
environmental processes. In the outermost radial bin, the mean LTG SFR rises above
that of CLASH ETGs, and also closer to the field value. However, the mean SFR of
the CLASH LTGs in this range is still a factor of 1.9 lower than that of LTGs in the
field. This further emphasizes that we are observing well into the cluster cores.
We investigated the morphology evolution of cluster galaxies, and found steadily
increasing cluster ETG fractions with time, rising from 40% at 1.25 < z < 1.5 to
80% at 0.15 < z < 0.36, with the majority of that increase coming over z = 1.5 → 1,
a period of evidently major growth in the ISCS early-type population. We also
compared our cluster ETG fractions against 22 clusters drawn from the literature,
and found generally good agreement, albeit with a fair amount of scatter in the
individual cluster fractions.
As our ISCS and CLASH clusters were building up their ETG populations, their
85
7.1. WHAT’S NEXT?
star formation activity was strongly declining. From z ∼ 1.5 → 0.15, the mean SFR of
cluster galaxies of all morphologies drops from 63.3±7.5 M yr−1 to 0.5±0.1 M yr−1 .
We found that this result matched quite well with the A14 fit in Equation 6.3. We
previously showed in §2.5.2, for ISCS galaxies, that the 24 µm-derived SFRs agreed
quite well with the 250 µm-derived SFRs from A14. We have now additionally found
that rest-frame UV-derived SFRs, corrected by an empirical SF RIR versus SF RUV
relation also show excellent agreement to the A14 fit.
7.1
What’s Next?
• To augment our analysis of high-redshift ISCS galaxies, we would like to attempt
to deconvolve 24 µm SFRs for non-isolated galaxies. We propose using PyGFit
(Mancone et al. 2013), a program designed to extract SEDs from crowded images
by first generating models from higher-resolution images (e.g. ACS), then fitting
them to the lower resolution images (e.g. MIPS).
• We would like to find archival MIPS imaging, for as many of the CLASH clusters
as possible, in order to calculate true 24 µm SFRs. This would allow us to
be more confident in our low-redshift SFR measurements. We would also be
able to test how accurate our empirical SF RUV → SF RIR relations (§5.1) are
when correcting the UV SFRs of our CLASH galaxies. Specifically, given the
vastly different morphological makeup of the field, we could determine if these
relations, calculated using field galaxies, are truly applicable to cluster galaxies.
• We would like to take advantage of the 16 bands of HST observations of CLASH
galaxies to perform full SED fits to determine stellar masses.
86
7.1. WHAT’S NEXT?
• As we noted in §6.1, we only have coverage of the inner ∼650 kpc of our CLASH
clusters, or approximately 50% of their virial radii (§6.2). To be able to get
a census of the entire cluster populations, high-resolution observations of the
outskirts of CLASH clusters would be ideal.
87
BIBLIOGRAPHY
Bibliography
Abraham, R. G., van den Bergh, S., & Nair, P. 2003, ApJ, 588, 218
Alberts, S., Pope, A., Brodwin, M., et al. 2014, MNRAS, 437, 437
Andreon, S. 2013, A&A, 554, A79
Aragon-Salamanca, A., Ellis, R. S., Couch, W. J., & Carter, D. 1993, MNRAS, 262,
764
Ashby, M. L. N., Stern, D., Brodwin, M., et al. 2009, ApJ, 701, 428
Barnes, J. E. 1988, ApJ, 331, 699
Barnes, J. E., & Hernquist, L. 1996, ApJ, 471, 115
Benítez, N. 2000, ApJ, 536, 571
Benítez, N., Ford, H., Bouwens, R., et al. 2004, ApJS, 150, 1
Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393
Blakeslee, J. P., Holden, B. P., Franx, M., et al. 2006, ApJ, 644, 30
Bower, R. G., Lucey, J. R., & Ellis, R. S. 1992, MNRAS, 254, 601
88
BIBLIOGRAPHY
Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008, ApJ, 686, 1503
Bressan, A., Panuzzo, P., Buson, L., et al. 2006, ApJ, 639, L55
Broadhurst, T., Huang, X., Frye, B., & Ellis, R. 2000, ApJ, 534, L15
Brodwin, M., Gonzalez, A. H., Moustakas, L. A., et al. 2007, ApJ, 671, L93
Brodwin, M., Brown, M. J. I., Ashby, M. L. N., et al. 2006, ApJ, 651, 791
Brodwin, M., Stern, D., Vikhlinin, A., et al. 2011, ApJ, 732, 33
Brodwin, M., Stanford, S. A., Gonzalez, A. H., et al. 2013, ApJ, 779, 138
Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000
Caldwell, N., Rose, J. A., Sharples, R. M., Ellis, R. S., & Bower, R. G. 1993, AJ, 106,
473
Cassata, P., Cimatti, A., Franceschini, A., et al. 2005, MNRAS, 357, 903
Cassata, P., Guzzo, L., Franceschini, A., et al. 2007, ApJS, 172, 270
Cerulo, P., Couch, W. J., Lidman, C., et al. 2014, MNRAS, 439, 2790
Chabrier, G. 2003, PASP, 115, 763
Chapman, S. C., Scott, D., Steidel, C. C., et al. 2000, MNRAS, 319, 318
Chary, R., & Elbaz, D. 2001, ApJ, 556, 562
Chung, S. M., Eisenhardt, P. R., Gonzalez, A. H., et al. 2011, ApJ, 743, 34
Clemens, M. S., Bressan, A., Panuzzo, P., et al. 2009, MNRAS, 392, 982
89
BIBLIOGRAPHY
Coe, D., Benítez, N., Sánchez, S. F., et al. 2006, AJ, 132, 926
Comerford, J. M., Meneghetti, M., Bartelmann, M., & Schirmer, M. 2006, ApJ, 642,
39
Conselice, C. J. 2003, ApJS, 147, 1
Couch, W. J., & Sharples, R. M. 1987, MNRAS, 229, 423
Cox, T. J., Dutta, S. N., Di Matteo, T., et al. 2006, ApJ, 650, 791
De Lucia, G., Springel, V., White, S. D. M., Croton, D., & Kauffmann, G. 2006,
MNRAS, 366, 499
Dekel, A., & Birnboim, Y. 2006, MNRAS, 368, 2
Dekel, A., & Burkert, A. 2014, MNRAS, 438, 1870
Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604
Dressler, A. 1980, ApJ, 236, 351
Dressler, A., Thompson, I. B., & Shectman, S. A. 1985, ApJ, 288, 481
Dressler, A., Oemler, Jr., A., Couch, W. J., et al. 1997, ApJ, 490, 577
Edwards, L. O. V., & Fadda, D. 2011, AJ, 142, 148
Eisenhardt, P. R., Stern, D., Brodwin, M., et al. 2004, ApJS, 154, 48
Eisenhardt, P. R. M., Brodwin, M., Gonzalez, A. H., et al. 2008, ApJ, 684, 905
Elbaz, D., Hwang, H. S., Magnelli, B., et al. 2010, A&A, 518, L29
90
BIBLIOGRAPHY
Elbaz, D., Dickinson, M., Hwang, H. S., et al. 2011, A&A, 533, A119
Fabricant, D., Franx, M., & van Dokkum, P. 2000, ApJ, 539, 577
Finn, R. A., Balogh, M. L., Zaritsky, D., Miller, C. J., & Nichol, R. C. 2008, ApJ,
679, 279
Ford, H. C., Bartko, F., Bely, P. Y., et al. 1998, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 3356, Society of Photo-Optical
Instrumentation Engineers (SPIE) Conference Series, ed. P. Y. Bely & J. B. Breckinridge, 234–248
Galametz, A., Stern, D., Eisenhardt, P. R. M., et al. 2009, ApJ, 694, 1309
Geller, M. J., Hwang, H. S., Diaferio, A., et al. 2014, ApJ, 783, 52
Girardi, L., Bressan, A., Chiosi, C., Bertelli, G., & Nasi, E. 1996, A&AS, 117, 113
Gómez, P. L., Nichol, R. C., Miller, C. J., et al. 2003, ApJ, 584, 210
Häußler, B., Barden, M., Bamford, S. P., & Rojas, A. 2011, in Astronomical Society
of the Pacific Conference Series, Vol. 442, Astronomical Data Analysis Software
and Systems XX, ed. I. N. Evans, A. Accomazzi, D. J. Mink, & A. H. Rots, 155
Hilton, M., Stanford, S. A., Stott, J. P., et al. 2009, ApJ, 697, 436
Hilton, M., Lloyd-Davies, E., Stanford, S. A., et al. 2010, ApJ, 718, 133
Holtzman, J. A., Hester, J. J., Casertano, S., et al. 1995, PASP, 107, 156
Hopkins, P. F., Hernquist, L., Cox, T. J., Dutta, S. N., & Rothberg, B. 2008, ApJ,
679, 156
91
BIBLIOGRAPHY
Hopkins, P. F., Hernquist, L., Cox, T. J., Robertson, B., & Springel, V. 2006, ApJS,
163, 50
Iglesias-Páramo, J., Buat, V., Donas, J., Boselli, A., & Milliard, B. 2004, A&A, 419,
109
Jannuzi, B. T., & Dey, A. 1999, in ASP Conf. Ser., Vol. 191, Photometric Redshifts
and the Detection of High Redshift Galaxies, ed. R. Weymann, L. Storrie-Lombardi,
M. Sawicki, & R. Brunner (San Francisco, CA: ASP), 111
Jee, M. J., Rosati, P., Ford, H. C., et al. 2009, ApJ, 704, 672
Jee, M. J., Dawson, K. S., Hoekstra, H., et al. 2011, ApJ, 737, 59
Kaviraj, S., Cohen, S., Ellis, R. S., et al. 2013, MNRAS, 428, 925
Kennicutt, Jr., R. C. 1998, ARA&A, 36, 189
Kereš, D., Katz, N., Weinberg, D. H., & Davé, R. 2005, MNRAS, 363, 2
Kimble, R. A., MacKenty, J. W., O’Connell, R. W., & Townsend, J. A. 2008, in SPIE
Conf. Ser. 70101E, Vol. 7010
Kodama, T. 1999, in Astronomical Society of the Pacific Conference Series, Vol. 163,
Star Formation in Early Type Galaxies, ed. P. Carral & J. Cepa, 250
Komatsu, E., Smith, K. M., Dunkley, J., et al. 2011, ApJS, 192, 18
Koopmann, R. A., & Kenney, J. D. P. 2004, ApJ, 613, 851
Kriek, M., van Dokkum, P. G., Labbé, I., et al. 2009, ApJ, 700, 221
92
BIBLIOGRAPHY
Kroupa, P. 2001, MNRAS, 322, 231
Le Borgne, J.-F., Bruzual, G., Pelló, R., et al. 2003, A&A, 402, 433
Lemze, D., Postman, M., Genel, S., et al. 2013, ApJ, 776, 91
Lilly, S. J., Le Fèvre, O., Renzini, A., et al. 2007, ApJS, 172, 70
Lilly, S. J., Le Brun, V., Maier, C., et al. 2009, ApJS, 184, 218
Lotz, J. M., Primack, J., & Madau, P. 2004, AJ, 128, 163
Lubin, L. M., Oke, J. B., & Postman, M. 2002, AJ, 124, 1905
Magnelli, B., Elbaz, D., Chary, R. R., et al. 2009, A&A, 496, 57
Mancone, C. L., Gonzalez, A. H., Brodwin, M., et al. 2010, ApJ, 720, 284
Mancone, C. L., Gonzalez, A. H., Moustakas, L. A., & Price, A. 2013, PASP, 125,
1514
Mancone, C. L., Baker, T., Gonzalez, A. H., et al. 2012, ApJ, 761, 141
Marcillac, D., Elbaz, D., Chary, R. R., et al. 2006, A&A, 451, 57
Marigo, P., & Girardi, L. 2007, A&A, 469, 239
Marigo, P., Girardi, L., Bressan, A., et al. 2008, A&A, 482, 883
Martini, P., Miller, E. D., Brodwin, M., et al. 2013, ApJ, 768, 1
McCracken, H. J., Milvang-Jensen, B., Dunlop, J., et al. 2012, A&A, 544, 156
Mei, S., Holden, B. P., Blakeslee, J. P., et al. 2006, ApJ, 644, 759
93
BIBLIOGRAPHY
—. 2009, ApJ, 690, 42
Mei, S., Stanford, S. A., Holden, B. P., et al. 2012, ApJ, 754, 141
Merten, J., Meneghetti, M., Postman, M., et al. 2014, ApJ, submitted
Moustakas, J., Coil, A. L., Aird, J., et al. 2013, ApJ, 767, 50
Murphy, E. J., Chary, R.-R., Alexander, D. M., et al. 2009, ApJ, 698, 1380
Murphy, E. J., Condon, J. J., Schinnerer, E., et al. 2011, ApJ, 737, 67
Murray, S. S., Kenter, A., Forman, W. R., et al. 2005, ApJS, 161, 1
Muzzin, A., Wilson, G., Lacy, M., Yee, H. K. C., & Stanford, S. A. 2008, ApJ, 686,
966
Muzzin, A., Marchesini, D., Stefanon, M., et al. 2013, ApJS, 206, 8
Naab, T., & Burkert, A. 2003, ApJ, 597, 893
Negroponte, J., & White, S. D. M. 1983, MNRAS, 205, 1009
Nordon, R., Lutz, D., Shao, L., et al. 2010, A&A, 518, L24
Oemler, A. 1974, ApJ, 194, 1
Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H.-W. 2010, AJ, 139, 2097
Poggianti, B. M., Fasano, G., Bettoni, D., et al. 2009, ApJ, 697, L137
Polletta, M., Tajer, M., Maraschi, L., et al. 2007, ApJ, 663, 81
Postman, M., Franx, M., Cross, N. J. G., et al. 2005, ApJ, 623, 721
94
BIBLIOGRAPHY
Postman, M., Coe, D., Benítez, N., et al. 2012, ApJS, 199, 25
Rieke, G. H., Young, E. T., Engelbracht, C. W., et al. 2004, ApJS, 154, 25
Rines, K., & Diaferio, A. 2006, AJ, 132, 1275
Rines, K., Geller, M. J., Diaferio, A., & Kurtz, M. J. 2013, ApJ, 767, 15
Rodighiero, G., Cimatti, A., Gruppioni, C., et al. 2010, A&A, 518, L25
Rosa-González, D., Terlevich, E., & Terlevich, R. 2002, MNRAS, 332, 283
Rosati, P., Tozzi, P., Gobat, R., et al. 2009, A&A, 508, 583
Saintonge, A., Tran, K.-V. H., & Holden, B. P. 2008, ApJ, 685, L113
Sanders, D. B., Soifer, B. T., Elias, J. H., et al. 1988, ApJ, 325, 74
Santos, J. S., Altieri, B., Popesso, P., et al. 2013, MNRAS, 433, 1287
Scoville, N., Aussel, H., Brusa, M., et al. 2007, ApJS, 172, 1
Snyder, G. F., Brodwin, M., Mancone, C. M., et al. 2012, ApJ, 756, 114
Springel, V., White, S. D. M., Jenkins, A., et al. 2005, Nature, 435, 629
Stanford, S. A., Eisenhardt, P. R., & Dickinson, M. 1998, ApJ, 492, 461
Stern, D., Eisenhardt, P., Gorjian, V., et al. 2005, ApJ, 631, 163
Strazzullo, V., Rosati, P., Pannella, M., et al. 2010, A&A, 524, A17
Swinyard, B. M., Ade, P., Baluteau, J.-P., et al. 2010, A&A, 518, L4
Tran, K.-V. H., Papovich, C., Saintonge, A., et al. 2010, ApJ, 719, L126
95
BIBLIOGRAPHY
Wagner, C., Brodwin, M., Snyder, G., et al. 2014, ApJ, submitted
Webb, T. M. A., O’Donnell, D., Yee, H. K. C., et al. 2013, AJ, 146, 84
Wickramasinghe, T., & Ukwatta, T. N. 2010, MNRAS, 406, 548
Wylezalek, D., Vernet, J., De Breuck, C., et al. 2014, ApJ, 786, 17
Zeimann, G. R., Stanford, S. A., Brodwin, M., et al. 2013, ApJ, 779, 137
Zibetti, S., Charlot, S., & Rix, H.-W. 2009, MNRAS, 400, 1181
96