Full thesis

Full thesis
Red Galaxies at High Redshift
Red Galaxies at High Redshift
Proefschrift
ter verkrijging van
de graad van Doctor aan de Universiteit Leiden,
op gezag van de Rector Magnificus prof. mr. P.F. van der Heijden,
volgens besluit van het College voor Promoties
te verdedigen op donderdag 27 september 2007
klokke 16.15 uur
door
Stijn Elisabeth Raphaël Wuyts
geboren te Mortsel
in 1980
Promotiecommissie
Promotores:
Prof. dr. M. Franx
Prof. dr. P. G. van Dokkum (Yale University)
Referent:
Dr. S. C. Trager (Rijksuniversiteit Groningen)
Overige leden:
Prof. dr. P. T. de Zeeuw
Prof. dr. K. H. Kuijken
Dr. P. van der Werf
Dr. J. Schaye
Aan mijn ouders
Cover: Acryl on canvas by Imelda Wuyts
Table of contents
vii
Table of contents
Page
Chapter 1. Introduction
1
1.1 Large-scale structure formation . . . . . . . . . . . . . . . . . . . . . . . . .
1
1.2 Galaxy formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1.3 This thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3
Chapter 2. The detailed fundamental plane of two high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 Sample selection and observations . . . . . . . . . . . . . . . .
2.2.2 Basic reduction . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.3 Velocity dispersions . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1 Structural parameters . . . . . . . . . . . . . . . . . . . . . . .
2.3.2 Error in the structural parameters . . . . . . . . . . . . . . . .
2.3.3 Visual and quantitative classifications . . . . . . . . . . . . . .
2.3.4 Transformation to rest-frame magnitude . . . . . . . . . . . . .
2.4 The fundamental plane . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.1 Zero point and scatter . . . . . . . . . . . . . . . . . . . . . . .
2.5 Correlations with other parameters . . . . . . . . . . . . . . . . . . .
2.5.1 The color-magnitude relation . . . . . . . . . . . . . . . . . . .
2.5.2 H β linestrength . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5.3 Location in the cluster . . . . . . . . . . . . . . . . . . . . . . .
2.5.4 Galaxy mass and selection effects . . . . . . . . . . . . . . . . .
2.5.5 Summary of correlations . . . . . . . . . . . . . . . . . . . . . .
2.6 Evolution of M/ L ratio . . . . . . . . . . . . . . . . . . . . . . . . . .
2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
8
9
9
9
10
11
11
12
14
14
16
16
18
18
18
20
20
21
21
23
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength
catalog and total IR properties of distant Ks -selected galaxies
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1 The GOODS Chandra Deep Field South . . . . . . . . . . . . . . . .
3.2.2 The ACS BViz data . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.3 The ISAAC JHKs data . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.4 The IRAC 3.6-8.0 µm data . . . . . . . . . . . . . . . . . . . . . . . .
3.2.5 The MIPS 24 µm data . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Final images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
28
29
29
29
29
29
30
30
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table of contents
viii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
30
31
33
34
35
36
36
37
40
40
41
42
43
43
43
45
46
46
48
52
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
55
56
57
57
57
58
59
60
60
60
63
64
64
65
66
69
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z <
3.5?
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Data, photometry and sample selection . . . . . . . . . . . . . . . . . . . .
5.2.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.2 Photometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.3 Sample selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
77
78
79
79
80
81
3.4
3.5
3.6
3.7
3.8
3.9
3.3.1 Pixel scales and large scale backgrounds . . . . . . .
3.3.2 Image quality and PSF matching . . . . . . . . . . .
3.3.3 Zero points . . . . . . . . . . . . . . . . . . . . . . .
3.3.4 Mosaicing and astrometry . . . . . . . . . . . . . . .
3.3.5 Signal to noise and limiting depths . . . . . . . . . .
Source detection and photometry . . . . . . . . . . . . . .
3.4.1 Ks -band detection . . . . . . . . . . . . . . . . . . . .
3.4.2 Photometry . . . . . . . . . . . . . . . . . . . . . . .
Redshifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5.1 Spectroscopic redshifts . . . . . . . . . . . . . . . . .
3.5.2 Photometric redshifts . . . . . . . . . . . . . . . . . .
Catalog parameters . . . . . . . . . . . . . . . . . . . . . . .
Comparison to the GOODS-MUSIC catalog . . . . . . . . .
3.7.1 Differences in data and strategy . . . . . . . . . . . .
3.7.2 Comparing photometry . . . . . . . . . . . . . . . .
3.7.3 Comparing photometric redshifts . . . . . . . . . . .
Total IR properties of distant Ks -selected galaxies . . . . .
3.8.1 Observed 24 µm flux as function of observed colors
3.8.2 Total IR luminosity as function of rest-frame colors
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter 4. Optical spectroscopy of Distant Red Galaxies
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Sample selection . . . . . . . . . . . . . . . . . . . . .
4.2.1 Pure J − K selected sample . . . . . . . . . . .
4.2.2 DRGs from other surveys . . . . . . . . . . . .
4.3 Observations . . . . . . . . . . . . . . . . . . . . . . .
4.4 Reduction . . . . . . . . . . . . . . . . . . . . . . . . .
4.5 Results from optical spectroscopy of DRGs . . . . . .
4.5.1 Redshift determination, success rate, and bias
4.5.2 Optical spectra . . . . . . . . . . . . . . . . . .
4.5.3 Redshift distribution . . . . . . . . . . . . . . .
4.6 Photometric redshifts . . . . . . . . . . . . . . . . . .
4.6.1 Method and template sets . . . . . . . . . . . .
4.6.2 Quality of photometric redshifts . . . . . . . .
4.7 The nature of low-redshift DRGs . . . . . . . . . . . .
4.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table of contents
ix
5.3 SED modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.4 Rest-frame optical to near-infrared color distribution . . . . . . . . . . . . 84
5.5 Constraints on stellar population properties at 2 < z < 3.5: age, dust and
mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.5.1 Predictions from stellar population synthesis models . . . . . . . . 89
5.5.2 Constraints on mass, dust and age from modeling our observed
galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.6 Stellar mass - optical color relation . . . . . . . . . . . . . . . . . . . . . . . 98
5.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Chapter 6. Recovering Stellar Population Properties and Redshifts from BroadBand Photometry of Simulated Galaxies: Lessons for SED Modeling
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 The simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2.1 Main characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2.2 Extracting photometry from the simulation output . . . . . . . . . .
6.2.3 The colors and SEDs of simulated and observed galaxies . . . . . .
6.3 SED modeling: methodology . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4 Results from SED modeling at fixed redshift . . . . . . . . . . . . . . . . .
6.4.1 Impact of mismatch between true and template SFH . . . . . . . . .
6.4.2 Impact of attenuation . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4.3 Impact of stellar metallicity . . . . . . . . . . . . . . . . . . . . . . .
6.4.4 Impact of AGN contribution . . . . . . . . . . . . . . . . . . . . . . .
6.4.5 Overall performance . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4.6 Lessons for SED modeling . . . . . . . . . . . . . . . . . . . . . . . .
6.5 Results from SED modeling with free redshift . . . . . . . . . . . . . . . .
6.5.1 The photometric redshift code EAZY . . . . . . . . . . . . . . . . . .
6.5.2 Recovering redshifts and stellar population properties from broadband photometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3: comparing observations with merger simulations
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2 Overview of the observations . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2.1 Fields, coverage, and depth . . . . . . . . . . . . . . . . . . . . . . .
7.2.2 Redshifts and rest-frame photometry . . . . . . . . . . . . . . . . .
7.2.3 Stellar masses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2.4 Star formation rates . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.3 Overview of the simulations . . . . . . . . . . . . . . . . . . . . . . . . . .
7.4 Sample selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.5 Methodology for cosmological context . . . . . . . . . . . . . . . . . . . .
7.6 The number density, mass density and mass function of galaxies with
log M > 10.6 at 1.5 < z < 3 . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.7 The color distribution of galaxies with log M > 10.6 at 1.5 < z < 3 . . . .
105
106
107
107
110
111
113
114
114
116
121
122
123
126
127
128
128
130
135
136
137
137
138
138
139
139
140
141
144
146
Table of contents
x
7.8
7.9
7.10
7.11
7.7.1 The U − V color distribution . . . . . . . . . . . . . . . . . . . . . . 146
7.7.2 The V − J color distribution . . . . . . . . . . . . . . . . . . . . . . . 148
7.7.3 U − V versus V − J color-color distribution . . . . . . . . . . . . . . 149
Specific star formation rate as a function of stellar mass . . . . . . . . . . . 151
The abundance of massive galaxies at 1.5 < z < 3: analysis by type . . . . 154
7.9.1 The number and mass density of massive (log M > 10.6) quiescent
red galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
7.9.2 The number and mass density of massive (log M > 10.6) star-forming
galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
7.9.3 The number and mass density of massive (log M > 10.6) galaxies
with SFR/ M > 1 Gyr−1 . . . . . . . . . . . . . . . . . . . . . . . . . 158
7.9.4 The number and mass density of galaxies with M > 1011 M⊙ and
U − V > 1.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
Comments and caveats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
7.10.1 Pair statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
7.10.2 Dependence on stellar population synthesis code . . . . . . . . . . 161
7.10.3 Reproducing dusty red starbursts . . . . . . . . . . . . . . . . . . . 161
7.10.4 Cosmic variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Nederlandse samenvatting
167
Curriculum vitae
175
Nawoord
177
Chapter 1
Introduction
1.1 Large-scale structure formation
S
TANDARD Cold Dark Matter (CDM) theory explains the formation of large-scale
structure from a nearly homogeneous soup to the cosmic web observed today by
gravitational instabilities that grow over time (Press & Schechter 1974). Cold Dark
Matter dominates the mass content of the universe and clumps into so-called halos.
Primordial gas will sink to the bottom of the potential wells outlined by the dark matter
halos, where it can cool and form small units of stars. As time evolves, CDM halos, and
with it their baryonic content, will merge to build up larger structures (White & Rees
1978). Eventually, this results in web-like structure where filaments of young galaxies
and ionized gas are surrounded by large, nearly empty voids. Large clusters of galaxies
reside where the filaments meet. The bottom-up fashion of structure formation stems
from the fact that the initial density fluctuations had most power at small scales.
Observations of the large-scale structure by the Sloan Digital Sky Survey (SDSS,
York et al. 2000; Tegmark et al. 2004) and the 2 degree Field Galaxy Redshift Survey
(2dFGRS, Colless et al. 2001; Sánchez et al. 2006) show a remarkable agreement with
theoretical predictions and state-of-the-art simulations based thereupon (Springel et al.
2005). The evolution of the baryonic content, however, depends on complex dissipational processes that are far from understood.
1.2 Galaxy formation
What is the number of galaxy collisions, if any, needed to build up an M∗ galaxy in
the local universe? When in the history of the universe did most of the hierarchical
build-up take place? What fraction of stars was formed in merger-triggered starbursts
as opposed to quiescent episodes of star formation? What role do supermassive black
holes at the centers of galaxies play in their evolution? How can the wide range in
galaxy colors, from blue to red, be explained? These are all questions that astronomers
are just about to address.
Observed scaling relations between galaxy properties provide stringent constraints
in this quest for evolutionary scenarios. Apart from explaining the local color-magnitude
relation (e.g., Sandage 1972; Bower, Lucey,& Ellis 1992), the Fundamental Plane (Djorgovski & Davis 1987; Dressler et al. 1987) and the correlation between the mass of
supermassive black holes and their hosts (Magorrian et al. 1998), any model for galaxy
formation should also account for the evolution of these scaling relations with redshift
(e.g., Holden et al. 2004; van Dokkum & van der Marel 2007; Peng et al. 2006). Other
1
2
Chapter 1. Introduction
observations that need to be reproduced are the wide range of galaxy colors at low and
high redshift, and their correlation with mass, morphology and environment.
The local universe
In the local universe, the bimodal color distribution of galaxies is well established (e.g.,
Stateva et al. 2001; Blanton et al. 2003; Baldry et al. 2004; Balogh et al. 2004). Blue
galaxies are more abundant in numbers than red galaxies, but since the latter tend to
be more massive, it is the population of red galaxies that contributes most (50% - 75%)
to the total stellar mass in the nearby universe (e.g., Bell et al. 2003). Nearby blue
galaxies generally have a disk-like morphology with spiral arms, whereas red galaxies
show elliptical shapes. The morphological classification of galaxies into spirals and
ellipticals goes back to Hubble (1926). The fraction of red, elliptical galaxies increases
significantly as we probe denser environments (Oemler 1974; Davis & Geller 1976;
Dressler 1980).
The distant universe
Determining if and how the described trends and scaling relations are present at high
redshift, is a challenging task. First of all, it is crucial to obtain unbiased samples of
distant galaxies.
Until the 1990s, radio galaxies and quasars were the only objects known at z > 2
(e.g., Schmidt 1974). Their extremely high luminosities are believed to be powered
by an accreting central supermassive black hole. The searches for ’normal’ distant
galaxies, with a stellar origin of the light, have only been successful since the 1990s.
Brute force spectroscopic surveys of all objects brighter than an optical magnitude limit
in a field are not an efficient means of constructing representative high redshift galaxy
samples. First, because only a small fraction of such a magnitude-limited sample will
lie at high redshift. For example, less than 5% of the I-band limited m I < 24 sample
by Le Fèvre et al. (2005) has a redshift z > 2.5. Second, because distant galaxies with
considerable mass but largely devoid of rest-frame UV emission would be missed by
such surveys.
Over the past decade, several color selection criteria have been designed, often
aided by technological developments, to identify galaxies in the redshift range 1.5 <
z < 3.5. Steidel and collaborators (e.g., Steidel et al. 1996) were the first to efficiently select distant galaxies with relatively unobscured star formation using the state-of-the-art
optical observatories (probing the rest-frame UV light). With the advent of first nearinfrared and then mid-infrared instruments on ground- and space-based telescopes,
new selection methods became possible to select z ∼ 2 galaxies (e.g., Franx et al. 2003;
Daddi et al. 2004; Yan et al. 2004), resulting in samples that were often complementary
to the optically selected objects.
Each of these color-selected samples is subject to its own biases. Small amounts
of unobscured star formation may shift galaxies into or out of the selection window.
Furthermore, due to the range in redshifts and galaxy types, the observed-frame colors used for their selection alone are not uniquely related to the physical properties
of the galaxies. In view of a grand theory of galaxy evolution, it is important to un-
Section 1.3. This thesis
3
derstand the physical conditions such as mass, age and dust extinction for the various color-selected samples. Ideally, we want to go beyond selecting galaxies by color,
and characterize the stellar population properties for a sample of galaxies that is complete above a certain mass limit, since mass is a more fundamental parameter than
(observed-frame) color.
The nature of distant galaxies
The measurement of dynamical masses, used for the study of cluster early-types up to
> 1 (e.g.,
z = 0.83 in Chapter 2 of this thesis, becomes increasingly more difficult above z ∼
van Dokkum & Stanford 2003; Holden et al. 2005). We therefore rely on stellar mass
estimates derived by modeling the broad-band spectral energy distributions (SEDs)
over a wide wavelength range with stellar population synthesis codes (e.g., Bruzual &
Charlot 2003; Maraston 2005).
Likewise, determining spectroscopic redshifts, although requiring less signal-tonoise than the measurement of velocity dispersions, becomes increasingly demanding
in terms of telescope time as we move to higher redshifts and fainter sources. Instead,
the study of mass-limited samples at high redshift relies mostly on photometric redshift estimates.
A reliable interpretation of the multi-wavelength views of distant galaxies in terms
of physical properties such as mass, age, and dust content, is the key to a robust test of
galaxy formation models.
1.3 This thesis
In this thesis, we measure the masses, ages, and dust extinctions of high redshift galaxies, with an emphasis on red galaxies. We aim to establish the accuracy with which
these properties can be determined and identify biases that may occur. Finally, we
use our observations to build a mass-limited sample of 1.5 < z < 3 galaxies and test a
model that explains the formation of red galaxies by collisions of gas-rich disk galaxies,
during which a quasar phase is triggered (Hopkins et al. 2006).
Chapter 2
In Chapter 2, we examine the fundamental plane (FP) of early-type galaxies in two
high-redshift clusters: MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83. In particular, we focus on the zeropoint and scatter of this scaling relation between galaxy
size, velocity dispersion and surface brightness, and its evolution with redshift between z = 0 and z = 0.83. We study the residuals from the FP as a function of color,
mass, H β linestrength and position within the cluster. We find that the residuals
from the FP of MS 2053–04 are correlated with the residuals from the H β − σ relation,
suggesting that stellar populations are playing a role in shaping the FP. Considering
only massive (M > 1011 M⊙ ) early-type galaxies to avoid selection biases in our magnitude limited sample, we conclude that their B-band mass-to-light ratio evolves as
log M/ L B ∼ −0.47z and we find a formation redshift z f orm ∼ 2.95.
4
Chapter 1. Introduction
Chapter 3
Chapter 3 consists of two parts. First, we describe the construction of a B-to-24 µm
multi-wavelength catalog for the GOODS Chandra Deep Field South (CDFS). The catalog contains optical, near-infrared and mid-infrared photometry for sources over an
<
area of 138 arcmin2 down to Kstot
, AB ∼ 24.3 (5σ ). The photometry is based on observations
with the Advanced Camera for Surveys (ACS) on the Hubble Space Telescope (HST),
the ISAAC camera on the Very Large Telescope (VLT) and the IRAC and MIPS instruments on board the Spitzer Space Telescope. We cross-correlate our catalog with the
1 Ms X-ray observations (Giacconni et al. 2002) and with a large database of spectroscopic redshifts in the field. The latter allows us to quantify the reliability of photometric redshift estimates.
Next, we exploit the multi-wavelength data to estimate total IR luminosities for Ks selected galaxies at 1.5 < z < 2.5. We investigate which galaxies have the brightest total
IR luminosities and which galaxies contribute most to the integrated total IR emission.
We conclude that galaxies with red colors in the rest-frame UV, rest-frame optical, and
rest-frame optical-to-NIR are dominating in the IR. However, at the reddest rest-frame
optical colors, there also exists a population of galaxies that is undetected at 24 µm and
therefore has low estimates for the total IR luminosity.
Chapter 4
Chapter 4 describes the optical spectroscopic follow-up of a sample of Distant Red
Galaxies (DRG, Franx et al. 2003) with Kstot
, Vega < 22.5, selected by the simple color criterion J − K > 2.3. All the successful redshift determinations (22% of the targeted
sample) were based on emission lines. With 15 spectroscopic redshifts identified with
only 2 at z < 2, we confirm the efficiency of this simple criterion to select red galaxies
at high redshift. The two lower redshift sources are best fitted by a dusty stellar population. Two other DRGs show CIV in emission, indicative of the presence of an active
galactic nucleus (AGN). We find that the photometric redshift code by Rudnick et al.
(2003) is able to determine redshifts for DRGs (or at least the subclass with emission
lines) to an accuracy of ∆z/(1 + z) ∼ 0.06.
Chapter 5
In Chapter 5, we address the question what new insights the IRAC camera onboard
Spitzer can reveal on the nature of galaxies at 2 < z < 3.5. We approach this question
by modeling the spectral energy distributions of distant galaxies in the Hubble Deep
Field South (HDFS) up to very faint magnitudes (Kstot
, AB = 25), including and excluding
the IRAC photometric datapoints. We find that for blue galaxies in a field where deep
NIR data is already available, the addition of IRAC offers little improvement. For red
galaxies, on the other hand, the uncertainties in the estimated stellar mass decrease
by a factor ∼ 3 by adding IRAC. We caution however that significant systematic uncertainties in stellar mass estimates remain due to the differences between stellar population synthesis codes. Furthermore, IRAC helps to break the degeneracy between
star-forming and quiescent red galaxies. Finally, we find that, as in the local universe,
the most massive galaxies at high redshift are redder than lower mass galaxies, even
Section 1.3. This thesis
5
when allowing for complex star formation histories.
Chapter 6
A problem with addressing the quality of estimated stellar population properties of
observed galaxies is that the ’true answer’ is often not available. Therefore, we test the
standard SED modeling using simulated galaxies for which we know exactly the ages
and masses of its stellar components and the distribution of gas and dust in between.
In order to do this, we derived synthetic broad-band photometry from snapshots of
hydrodynamical merger simulations by Robertson et al. (2006) as they would be observed when placed at redshifts 1.5 < z < 3. The choice for merger simulations allows
us to test the performance of standard SED modeling for different types of galaxies:
from disks to mergers and eventually ellipticals. We discuss the impact of the star formation history, dust distribution, metallicity and AGN activity on the recovered mass,
age, dust reddening and extinction. Systematic underestimates in all these parameters occur during the star-forming episodes. The properties of red quiescent merger
remnants, on the other hand, are recovered very well.
Chapter 7
Multi-wavelength studies of deep fields have revealed a large variety of galaxy types
in the early universe: from relatively unobscured star-forming galaxies to dusty starbursts to quiescent massive red galaxies. Especially the existence of the latter population, whose strongly suppressed star formation has been confirmed by spectroscopic
identifications of their Balmer/4000Å breaks, poses a strong constraint on galaxy formation models.
In the final Chapter of this thesis, we combine the observations of three deep fields
(HDFS, MS 1054–03, and CDFS) to test a model by Hopkins et al. (2006) that aims to
explain the formation of red galaxies. Briefly, the model assumes that every observed
quasar is triggered by the collision between two gas-rich disk galaxies. Starting from
that assumption, it is then possible to translate the observed quasar luminosity function at a range of redshifts (e.g., Ueda et al. 2003; Hasinger, Miyaji,& Schmidt 2005;
Richards et al. 2005) to the demographics of galaxies expected in a particular redshift
interval. Using the same simulations as described in Chapter 6, we compute the colors,
number and mass densities of massive galaxies at 1.5 < z < 3 predicted by this model
and compare it to observed samples subject to identical selection criteria.
We find that post-quasar galaxies have similar colors as red quiescent galaxies.
The observed number and mass densities of quiescent galaxies are consistent with
the model predictions. The model is also able to account for the abundance of starforming galaxies, albeit with large uncertainties. However, the color distribution of
star-forming galaxies is not well reproduced, in particular the observed dusty starbursts have no counterparts in the model predictions. Several possible reasons for this
discrepancy are discussed.
6
Chapter 1. Introduction
References
Baldry, I. K., Glazebrook, K., Brinkmann, J., Ivezić, Ž, Lupton, R. H., Nichol, R. C.,& Szalay, A. S. 2004,
ApJ, 600, 681
Balogh, M. L., Baldry, I. K., Nichol, R., Miller, C., Bower, R.,& Glazebrook, K. 2004, ApJ, 615, L101
Blanton, M. R., et al. 2003, ApJ, 594, 186
Bower, R. G., Lucey, J. R.,& Ellis, R. S. 1992, MNRAS, 283, 1361
Bruzual, G.,& Charlot, S. 2003, MNRAS, 344, 1000
Colless, M., et al. 2001, MNRAS, 328, 1039
Daddi, E., Cimatti, A., Renzini, A., Fontana, A., Mignoli, M., Pozzetti, L., Tozzi, P.,& Zamorani, G. 2004,
ApJ, 617, 746
Davis, M. & Geller, M. J. 1976, ApJ, 208, 13
Djorgovski, S.,& Davis, M. 1987, ApJ, 313, 59
Dressler, A. 1980, ApJ, 236, 351
Dressler, A., Lynden-Bell, D., Burstein, D., Davies, R. L., Faber, S. M., Terlevich, R. J.,& Wegner, G. 1987,
ApJ, 313, 42
Franx, M., et al. 2003, ApJ, 587, L79
Hasinger, G., Miyaji, T.,& Schmidt, M. 2005, A&A, 441, 417
Holden, B. P., Stanford, S. A., Eisenhardt, P.,& Dickinson, M. 2004, AJ, 127, 2484
Holden, B. P., et al. 2005, ApJ, 620, 83
Hopkins, P. F., Hernquist, L., Cox, T. J., Robertson, B.,& Springel, V. 2006, ApJS, 163, 50
Magorrian, J., et al. 1998, AJ, 115, 2285
Maraston, C. 2005, MNRAS, 362, 799
Oemler, A. J. 1974, ApJ, 194, 1
Peng, C. Y., Impey, C. D., Rix, H.-W., Kochanek, C. S., Keeton, C. R., Falco, E. E., Lehár, J.,& McLeod, B.
A. 2006, ApJ, 649, 616
Richards, G. T., et al. 2005, MNRAS, 360, 839
Robertson, B., Cox, T. J., Hernquist, L., Franx, M., Hopkins, P. F., Martini, P.,& Springel, V. 2006, ApJ,
641, 21
Sánchez, A. G., et al. 2006, MNRAS, 366, 189
Sandage, A. 1972, ApJ, 176, 21
Schmidt, M. 1974, ApJ, 193, 505
Springel, V., et al. 2005, Nature, 435, 629
Strateva, I., et al. 2001, AJ, 122, 1861
Tegmark, M. et al. 2004, ApJ, 606, 702
Ueda, Y., Akiyama, M., Ohta, K.,& Miyaji, T. 2003, ApJ, 598, 886
van Dokkum, P. G.,& Stanford, S. A. 2003, ApJ, 585, 78
van Dokkum, P. G.,& van der Marel, R. P. 2007, ApJ, 655, 30
Yan, H., et al. 2004, ApJ, 616, 63
York, D. G., et al. 2000, AJ, 120, 1579
Chapter 2
The detailed fundamental plane of two
high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03
at z = 0.83
Abstract. We study the fundamental plane relation in high-redshift clusters using a sample of 26 galaxies in MS 2053–04 (z = 0.583) and 22 galaxies in MS 1054–
03 (z = 0.83). The zero point and scatter are compared to results for lower redshift clusters in order to trace evolutionary effects. Furthermore, our large sample enables us to investigate correlations between residuals from the fundamental plane and other characteristics of the galaxies, such as color, H β linestrength,
spatial distribution, and mass. The observed scatter of the early-type galaxies
with σ > 100 km s−1 around the fundamental plane is 0.134 and 0.106 in log re
for MS 2053–04 and MS 1054–03 respectively. The residuals from the fundamental plane of MS 2053–04 are correlated with residuals from the H β − σ relation,
suggesting that stellar populations are playing a role in shaping the fundamental
plane. The measured evolution in log M/ L is influenced by selection effects, as
galaxies with lower M/L in the Johnson B-band enter a magnitude-limited sample
more easily. When we select high mass early-type galaxies to avoid this bias, we
find log M/ L B ∼ −0.47z and a formation redshift z f orm ∼ 2.95, similar to earlier
results.
S. Wuyts, P. G. van Dokkum, D. D. Kelson, M. Franx & G. D. Illingworth
The Astrophysical Journal, 605, 677 (2004)
7
8
Chapter 2. The detailed fundamental plane of two high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83
2.1 Introduction
I
N the local universe early-type galaxies follow a tight scaling relation, known as the
Fundamental Plane (FP). The relation between effective radius, central velocity dispersion and surface brightness re ∼ σ α Ieβ , which is a plane in (log re , log σ, log Ie ) space,
was discovered by Djorgovski & Davis (1987) and Dressler et al. (1987). In combination
with the virial theorem
M/ L ∼ σ 2 r−1 Ie−1
(2.1)
the small scatter around the FP implies that, under the assumption of homology, the
M/ L ratios of early-type galaxies are well behaved and scale as
(1+β )/β
M/ L ∼ σ α/β+2r−
.
e
(2.2)
log re = 1.20 log σ − 0.83 log Ie + γ
(2.3)
0.07
M/ L ∼ M0.28 r−
.
e
(2.4)
As the M/ L ratio increases with ageing of a stellar population, the FP is a useful tool in
research on galaxy formation and evolution. Based on a sample of 226 E and S0 galaxies in 10 clusters of galaxies, Jørgensen, Franx,& Kjærgaard (1996, hereafter JFK96)
concluded the local plane in the Johnson B band has the form:
which implies that
Later studies on intermediate and high-redshift clusters of galaxies used the zero-point
shift of the plane to estimate average formation redshifts of the stars in early-type
galaxies (e.g., Bender et al. 1998; van Dokkum et al. 1998, hereafter vD98; Jørgensen et
al. 1999; Kelson et al. 2000c, hereafter K2000; Pahre et al. 2001; van Dokkum & Stanford
2003). The scatter around the plane provides constraints on the spread in galaxy ages.
The slope of the FP (and other scaling relations) constrains systematic age trends
with mass and other parameters. Any evolution of the slope of the FP with redshift
implies that the ages of the stellar populations are correlated with galaxy mass. For
the cluster CL1358+62 at z = 0.33 K2000 finds that the slope of the FP has not evolved
significantly over the past ∼ 4 Gyr. The sample of 5 bright MS 2053–04 galaxies used
by Kelson et al. (1997, hereafter K97) to study the FP at z = 0.583 seemed to agree
with the values of coefficients α and β as given by JFK96. However, the sample was
too small to perform a proper fit. The same conclusions were drawn for MS 1054–
03 at z = 0.83 based on 6 early-type galaxies (vD98). In this chapter, we investigate
the FP of MS 2053–04 and MS 1054–03 early-type galaxies using larger samples spread
over a larger range of distances from the brightest cluster galaxies (BCG). In Sect. 2.2
we discuss the spectroscopy, sample selection and velocity dispersions. Imaging and
measurement of the structural parameters is described in Sect. 2.3. Zero point of the
FP with JFK96 coefficients and scatter around the plane are discussed in Sect. 2.4. In
Sect. 2.5 we study correlations between the residuals from the FP and various other
properties of the galaxies. Finally the conclusions are summarized in Sect. 2.7. VEGA
magnitudes are used throughout this chapter. We use H0 = 70 km s−1 Mpc−1 , ΩΛ = 0.7,
and Ω M = 0.3, but note that our results are independent of the value of the Hubble
constant.
Section 2.2. Spectroscopy
9
2.2 Spectroscopy
2.2.1 Sample selection and observations
All spectra used in the FP analysis were obtained with the LRIS spectrograph (Oke
et al. 1995) on the 10 m W.M. Keck Telescope. The data were obtained in a series of
observing runs from 1996 to 2002. The majority of galaxies in our final sample were
selected on the basis of their spectroscopic redshift and their I- or F814W-band magni< 21 were given
tude. The samples were limited at I ≈ 22 for all runs; galaxies with I ∼
highest priority in the mask designs. The redshift information came from a large Iselected spectroscopic survey of both clusters described in detail in Tran et al. (1999),
van Dokkum et al. (2000), and Tran (2002). For the initial MS 1054–03 observing runs
only limited redshift information was available, and we applied color criteria to select
likely cluster members. Galaxies having ∆(R − I) − 0.25∆(B − R) < −0.4, with ∆(R − I)
and ∆(B − R) colors relative to the central galaxy, were excluded. The color ranges
were chosen such that blue cluster members were unlikely to be excluded. The final
FP sample of MS 1054–03 contains 19 galaxies that were selected with these mild color
constraints. No morphological information was used in the selection process.
For most observations we used the 600 lines mm−1 grating blazed at 7500 Å; some
of our earlier data were taken with the 831 lines mm−1 grating blazed at 8200 Å (see
vD98). The wavelength coverage was typically ∼ 3500 to ∼ 5400 Å in the rest frame. Exposure times ranged from 7500 s to 33400 s and from 10500 s to 22800 s for the MS 2053–
04 and MS 1054–03 galaxies respectively. The instrumental resolution was typically
σinst ∼ 40 − 80 km s−1 and signal-to-noise ratios ranged from 20 to 100 Å−1 in the observedframe (in the continuum).
A total of 43 galaxies (26 early-type) were observed in MS 2053–04. The MS 1054–
03 sample contained 30 galaxies (14 early-type). The morphological classification is
described in Sect. 2.3.3. Early-type galaxies include E, E/S0 and S0 morphologies.
2.2.2 Basic reduction
The spectra were reduced using our own software and standard IRAF software routines (see, e.g., Kelson et al. 2000b). The wavelength calibration was performed using
the night sky emission lines. The typical rms scatter about the fitted dispersion solutions is about 1/15 of a pixel. Since the dispersion is about 1.28 Å pixel−1 for the data
taken with the 600 mm−1 grating (and 0.92 Å pixel−1 for the 831 mm−1 data), the rms
scatter is equivalent to velocity errors smaller than 5 km s−1 .
The flat-fielding accuracy is generally better than a percent, on small scales. The
data have not been accurately flux-calibrated so on large scales, the notion of flatfielding accuracy is not meaningful. There tend to be moderate-scale (k=100 Å−1)
residuals in the flat-fielding that are the result of a mismatch between the fringing
in the flat-fields and the fringing in the data. Such inaccuracies in de-fringing the data
have no effect on the velocity dispersions because the spectra are effectively filtered on
those scales (and larger) in the process of matching the continua of the template and
galaxy spectra.
The subtraction of the sky was performed using standard, published, and welltested methods: for each galaxy the two-dimensional spectra were first rectified, and
10
Chapter 2. The detailed fundamental plane of two high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83
then 1st- or 2nd-order polynomials were fit to the pixels on both sides of the galaxy,
typically excluding the few arcsec where the galaxy is bright. An iterative, clipping
routine was used to reject any remaining bad pixels from the fit, similar to what IRAF
allows the user to do. These methods have been discussed elsewhere at great length
and we choose not to bore the reader with very familiar territory. While more complicated and precise means of sky subtraction are now available (see Kelson 2003), these
data were analyzed before such methods became available, and the accuracy of the sky
subtraction performed in the ”traditional” way is satisfactory for our purposes.
′′
′′
The slit widths varied slightly with each run, ranging from 0 . 90 to 1 . 05. The
′′
extraction apertures for MS 2053–04 and MS 1054–03 were 7 CCD rows (or 1 . 5).
2.2.3 Velocity dispersions
Velocity dispersions were measured with a direct fitting method (Kelson et al. 2000b).
Details of the fitting procedure are given in Kelson et al. (2006). Direct fitting methods
provide several advantages over Fourier-based techniques. Most importantly, pixels
are no longer weighted uniformly in the computation of the fitting metric (in this case
χ2 ). As in Kelson et al. 2000b, we weight each pixel by the inverse of the expected noise
(photon and electronic read noise). Furthermore, the spectra that exhibit strong Balmer
absorption had those features given zero weight in the fitting, in order to minimize the
contribution of those features to template mismatch error, and also to ensure that our
estimates of σ reflected the old, underlying stellar populations. We typically fit the
spectra over a ∼ 1000 Å wavelength range (in the rest frame). We used a range of
template stars, from G5 through K3 and adopted the template star that gave the lowest
mean χ2 . In both clusters, HD102494, a G9IV star, was the “best” template.
Both observations and simulations by Jørgensen et al. (1995b) showed that, at
low S/N, measured velocity dispersions were systematically too large. At the same
S/N, this systematic effect was largest for the galaxies with velocity dispersions below
100 km s−1 . Therefore, we omit all galaxies with σ < 100 km s−1 from our samples. Furthermore, we limit our samples to sources with an error in σ smaller than 15 %. Errors
in the dispersion and velocity are initially determined from the local topology of the
χ2 (σ, V) surface.
Seven MS 2053–04 galaxies with σ > 100 km s−1 were observed both during the 1997
and 2001 observing run. A direct comparison between the derived velocity dispersions
(prior to the aperture correction) is presented in Figure 2.1. Generally, the agreement
is good, with one outlier. All spectra have similar S/N. The rms value of √σ972−σ01 2 is
dσ97 +dσ01
1.23, slightly higher than the expected value of 1. The mean deviation between the two
runs is −2 ± 5%, consistent with zero. We conclude there is no evidence for systematic
effects and conservatively multiply all errors by the factor 1.23.
The final sample consists of 26 galaxies (19 early-type) in MS 2053–04 and 22 galaxies (12 early-type) in MS 1054–03. We applied an aperture correction to a nominal
′′
aperture of D = 3 . 4 at the distance of Coma (see JFK96). The final velocity dispersions have therefore been multiplied by a factor of 1.057 for MS 2053–04 and 1.062 for
MS 1054–03. This correction allows for a fair comparison between clusters at a range
of redshifts. The data are tabulated in Table 2.1.
Section 2.3. Imaging
11
Figure 2.1 — A direct comparison of
velocity dispersions for 7 galaxies in
MS 2053–04. Measured σ values, prior
to aperture correction, for the 2001 run
are plotted against σ values from the
1997 spectra. Formal errors derived
from χ2 (σ, V) are drawn. Based on this
overlap sample, actual error bars are
estimated to be 23% larger.
2.3 Imaging
For both MS 2053–04 and MS 1054–03 large HST WFPC2 mosaics were obtained, each
consisting of 6 pointings. Both clusters were observed in the F606W and F814W filters.
The layout of the MS 2053–04 mosaic is described in Hoekstra et al. (2002). Exposure
times were 3300 s in F606W and 3200 s in F814W per pointing. The MS 1054–03 mosaic
is described in van Dokkum et al. (2000); exposure times were 6500 s for each pointing
√
and in each filter. Interlacing of the images improved the sampling by a factor 2 for
MS 1054–03.
2.3.1 Structural parameters
In this section, we describe the measurement of effective radii re and surface brightnesses at those radii Ie . For the Johnson B passband, Ie in L⊙ pc−2 is related to µe in
mag arcsec−2 as
log Ie = −0.4(µe − 27.0).
′′
′′
(2.5)
We created postage stamps sized 12 . 8 × 12 . 8 for the 26 MS 2053–04 and 22 MS 1054–
03 galaxies and fit 2D r1/n (n = 1, 2, 3, 4) law profiles, convolved with Point Spread Functions (PSF), to the galaxy images. As PSFs depend on the positions of objects on the
CCDs, we used Tiny Tim v6.0 to create an appropriate PSF for each galaxy. Other parameters determining the shape of the PSF are template spectrum (M type star), PSF
size (3 ′′ ), sampling and filter (F814W). The code allows simultaneous fitting of the object of interest and any neighbouring objects. The fits were restricted to radii of 3 ′′ to 5 ′′
around the objects, depending on their size. Image defects were masked in the fit, as
well as neighbouring galaxies not well fitted by r1/4 laws. All other pixels got uniform
weight.
We performed the r1/n fits for Sersic numbers n = 1 (exponential), 2, 3 and 4 (de
Vaucouleurs law). In this chapter, we always use re and Ie based on r1/4 fits to all
galaxies on the postage stamps, even if other Sersic numbers result in a better χ2 of the
fit. Fitting a r1/4 profile resulted in a χ2 < 1.5 for 69% of all galaxies; 86% have χ2 < 2.
Chapter 2. The detailed fundamental plane of two high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83
12
MS2053-04 galaxies
174
S0 416
Sa 937
1652
S0 1667
E 1676
1877
S0/a 1993
E 2232
3155
S0 3549
S0 951
E 977
S0 1372 E/S0 1476
E 1686 E/S0 1688 E/S0 1738
S0/a 2258
S0 2260 E/S0 2345
E 1752
Sab 2497
S0/a 1583 E/S0
S0 1755
E
S0/a 2613 E/S0
Un
Figure 2.2 — 4 ′′ × 4 ′′ images (upper) and residuals (lower) after r1/4 profile fitting for the galaxies in
MS 2053–04. Masked regions are indicated in black. The postage stamps prove that our results for the
early-type galaxies are not suffering from misclassifications or bad profile fitting.
The galaxies in our final samples together with the residuals after profile fitting are
presented in Figure 2.2 (MS 2053–04) and Figure 2.3(MS 1054–03).
2.3.2 Error in the structural parameters
For MS 1054–03 each of the 6 pointings was observed twice, with a shift of 0.5 pixels,
providing a direct way to measure the error in the structural parameters. From fits
Section 2.3. Imaging
13
MS1054-03 galaxies
1192
M/P 1649
4705
Sc 4926
5666
E 2409 E/S0 3058 E/S0 3768 E
S0/a 5280
S0/a 5756 E
5840
S0/a 5298
M/P 6036
S0 5347
3910
Sa 4345 E/S0 4520
E
M/P 5450 E/S0 5529 E/S0 5577
S0/a
M/P 6301 E/S0 6688 E/S0
Figure 2.3 — 4 ′′ × 4 ′′ images (upper) and residuals (lower) after r1/4 profile fitting for the galaxies in
MS 1054–03. Masked regions are indicated in black. The postage stamps prove that our results for the
early-type galaxies are not suffering from misclassifications or bad profile fitting.
to the two independent observations we infer that errors are small (< 4% in re ) for
′′
galaxies larger than 0 . 83. The error in re rises to at most 18.5% for smaller sources.
Though this might seem problematic, we note that this large error does not enter the
FP analysis as the combination re Ie−β enters the FP. Using β = −0.83 from JFK96, the
error in the FP parameter re Ie0.83 is limited to ∼ 2.1% rms, ignoring one outlier with 30%
offset. The fit of a de Vaucouleurs profile to the outlier ID5347 with merger/peculiar
morphology is clearly unstable. The uncertainty in the FP parameter is comparable to
the ∼ 2.5% rms error estimate from Kelson et al. (2000a). The small error in the FP
parameter is an artifact of the slope of the de Vaucouleurs growth curve. Hereafter,
we√use the average of the independent re measurements (which reduces the error by
1/ 2) and the Ie corresponding to this average value, using the empirical result that
re Ie0.66 is the most stable combination of the structural parameters. In general we can
conclude from the error analysis that random errors are small. Therefore the scatter
found around the FP will not be due to errors in the photometry.
14
Chapter 2. The detailed fundamental plane of two high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83
Figure 2.4 — Visual classification histogram for different best fitting Sersic numbers. Early-type morphologies
generally correspond to Sersic numbers
3 and 4. A different greyscale is used
for MS 2053–04 and MS 1054–03.
2.3.3 Visual and quantitative classifications
Galaxies were visually classified by P.G. van Dokkum, M. Franx,& D. Fabricant using
the procedure as described in Fabricant, Franx,& van Dokkum (2000). In this chapter, we consider cluster members with early-type morphology (E, E/S0, S0). Late-type
morphologies are also plotted, but are not included in fitting procedures unless mentionned otherwise. A quantitative alternative to the classification by eye is based on
the Sersic number that results in the smallest χ2 of the profile fit.
Figure 2.4 shows histograms of the visual classifications for the galaxies with best
fitting Sersic number 1, 2, 3 and 4. As could be expected, early-type morphologies
generally correspond to Sersic numbers 3 and 4. We note that the difference in χ2
between n = 3 and 4 is often too small to consider them as a seperate class of objects. We
conclude that the visual and quantitative classifications divide the sample in roughly
the same bulge and disk dominated classes.
2.3.4 Transformation to rest-frame magnitude
In order to compare the FP of MS 2053–04 and MS 1054–03 with the FP of Coma at
z = 0.023 we have to transform the effective radius to units of kpc. Furthermore, for
a meaningful comparison it is necessary to study all data in a common photometric
band in the rest frame of the galaxies. The observed F606W and F814W filters straddle
the redshifted Bz band for 0.5 ≤ z ≤ 0.8. Therefore, we transform the observed surface
brightnesses to rest-frame B. The procedure is described in van Dokkum & Franx
(1996). It involves an interpolation between F606W and F814W, and is different from
applying a “K-correction”. For z = 0.583 we find
µ Bz = µ F814W + 0.42(F606W − F814W) + 0.84,
(2.6)
µ Bz = µ F814W + 0.01(F606W − F814W) + 1.13.
(2.7)
and for z = 0.83
MS 2053–04
ID
∆R. A .a
[ ′′ ]
∆Deca
[ ′′ ]
log re
[kpc]
174
416
937
951
977
1372
1476
1583
1652
1667
1676
1686
1688
1738
1752
1755
1877
1993
2232
2258
2260
2345
2497
2613
3155
3549
9.34
29.26
-30.58
-31.74
-72.15
-39.50
-38.26
-24.76
-9.65
0
47.76
14.28
-9.00
-23.00
-1.54
-1.14
19.96
49.67
-82.50
-2.50
46.61
16.82
15.46
-23.80
-90.37
-29.45
-206.60
-150.90
-119.80
-118.00
-132.10
-72.32
-54.80
-41.87
-27.37
0
0.89
-12.29
-22.49
-17.84
-9.86
-3.41
15.62
43.54
13.16
45.70
67.51
67.38
78.88
71.80
98.46
165.70
0.588
0.833
-0.223
0.288
0.261
0.153
0.505
0.289
0.475
1.103
0.415
0.408
0.392
0.667
0.353
0.482
0.683
0.540
0.664
0.264
0.490
0.601
0.451
0.584
0.376
0.307
a Coordinates
b Surface
µ Bz b
[mag ′′ −2 ]
21.19
22.92
18.20
20.88
19.92
20.88
21.33
21.04
21.94
22.65
21.56
21.58
21.78
22.04
21.08
21.48
22.07
20.86
22.01
20.77
21.80
21.99
21.71
22.39
21.09
20.70
MS 1054–03
σ
[km / s]
225 ± 13
144 ± 14
127 ± 11
151 ± 13
249 ± 11
164 ± 14
143 ± 16
143 ± 10
181 ± 13
292 ± 10
125 ± 14
129 ± 15
138 ± 13
124 ± 13
151 ± 17
234 ± 23
169 ± 14
212 ± 8
141 ± 12
134 ± 18
140 ± 14
152 ± 19
244 ± 14
171 ± 25
203 ± 18
155 ± 19
F814W T
[mag]
19.68
20.48
20.70
20.81
20.03
21.55
20.04
21.04
21.12
18.56
20.76
20.95
21.31
20.01
20.88
20.23
20.17
19.40
20.16
20.93
20.81
20.58
21.13
21.07
20.64
20.45
Type
ID
∆R. A .c
[ ′′ ]
∆Decc
[ ′′ ]
log re
[kpc]
S0
Sa
S0
E
S0
E/S0
S0/a
E/S0
S0
E
E
E/S0
E/S0
E
S0
E
S0/a
E
S0/a
S0
E/S0
Sab
S0/a
E/S0
S0
Un
1192
1649
2409
3058
3768
3910
4345
4520
4705
4926
5280
5298
5347
5450
5529
5577
5666
5756
5840
6036
6301
6688
136.10
122.30
84.18
52.73
38.86
32.00
21.51
0
6.10
-4.07
-20.46
-21.74
-34.24
-46.30
-37.79
-43.54
-58.07
-59.26
-52.17
-63.37
-70.57
-91.54
-58.31
-16.89
-24.01
-9.72
-0.38
-11.84
-13.22
0
8.34
-3.68
20.82
19.24
56.37
-3.51
31.29
48.16
-83.77
97.86
23.03
-29.09
25.38
-41.79
0.518
0.565
0.574
1.227
0.507
0.457
0.643
1.141
1.006
0.387
0.548
0.550
0.795
0.866
0.549
0.501
0.731
0.551
0.090
0.559
0.565
0.458
µ Bz b
[mag ′′ −2 ]
20.51
20.80
21.26
22.97
20.78
20.67
20.98
22.29
22.22
20.43
21.06
21.38
21.54
21.72
20.99
20.75
21.20
20.91
18.38
21.17
21.03
20.50
σ
[km / s]
138 ± 15
243 ± 28
287 ± 33
303 ± 33
222 ± 24
295 ± 42
336 ± 34
322 ± 30
253 ± 36
310 ± 38
259 ± 31
284 ± 39
254 ± 24
234 ± 26
182 ± 23
305 ± 40
286 ± 23
232 ± 27
212 ± 26
254 ± 22
249 ± 24
274 ± 37
F814W T
[mag]
20.24
20.69
21.36
19.83
21.04
21.23
20.55
19.48
20.61
21.32
21.17
21.52
20.58
19.98
21.01
21.05
20.50
20.93
20.66
21.18
20.99
20.82
Type
Section 2.3. Imaging
Table 2.1. FP sample
M/P
E
E/S0
E/S0
E
Sa
E/S0
E
Sc
S0/a
S0/a
S0
M/P
E/S0
E/S0
S0/a
S0/a
E
M/P
M/P
E/S0
E/S0
with respect to the BCG of MS 2053–04: ID1667 at (20:56:21.4; -04:37:50.8) (J2000).
brightnesses µ Bz are corrected for galactic extinction and cosmological dimming.
c Coordinates
with respect to the BCG of MS 1054–03: ID4520 at (10:56:59.9; -03:37:37.3) (J2000).
15
16
Chapter 2. The detailed fundamental plane of two high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83
Figure 2.5 — The fundamental plane of clusters MS 2053–04 (z=0.583) and MS 1054–03 (z=0.83). The
Coma FP is drawn for reference. Typical error bars are plotted in the upper left corner. Cluster galaxies
in the higher redshift clusters follow the FP scaling relation, but with an offset with respect to Coma.
Galaxies with early-type morphologies show a larger scatter than in the local universe.
The F606W − F814W colors in (2.6) and (2.7) are obtained using SExtractor (Bertin
′′
& Arnouts 1996) with fixed apertures of 0 . 7 diameter. Extinction corrections for both
fixed aperture colors and surface brightnesses were derived from Schlegel et al. (1998).
Another correction compensates for cosmological dimming ∝ (1 + z)4 . The final samples are summarized in Table 2.1. Included are the coordinates with respect to the
BCG, aperture corrected σ , re , µ Bz corrected for galactic extinction and cosmological
dimming, total F814W T magnitudes and morphological classifications.
2.4 The fundamental plane
In this section we discuss the FP relation in MS 2053–04 and MS 1054–03. For determination of the zero point and scatter we adopt the slope (α, β ) = (1.20, −0.83) that JFK96
found for the local B-band FP. In Figure 2.5 we show the edge-on view of the FP for
both clusters. Different symbols indicate different morphological types. For comparison the Coma FP is drawn as well. The galaxies in MS 2053–04 and MS 1054–03 follow
a similar FP, but with an offset with respect to Coma.
2.4.1 Zero point and scatter
2.4.1.1 MS 2053–04
Using the biweight mean (Beers et al. 1990) we fit a zero point of the FP to the earlytype galaxies in MS 2053–04. Under the assumption of homology, the zero-point shift of
the FP traces the mean evolution of the galaxy M/ L ratio. The observed zero-point off-
Section 2.4. The fundamental plane
17
Table 2.2. MS 2053–04 and MS 1054–03 zero point and scatter around FP. Earlier
results from K97 and vD98 are also in this Table.
Sample
MS 2053–04 early-type
MS 2053–04 Sersic34
MS 2053–04 K97
MS 1054–03 early-type
MS 1054–03 Sersic34
MS 1054–03 vD98
# objects
∆ log(M/ L B )
(biweight mean)
19
23
5
12
15
6
−0.365 ± 0.037
−0.382 ± 0.028
−0.280 ± 0.036
−0.405 ± 0.037
−0.368 ± 0.027
−0.393 ± 0.040
∆ log(M/ L B )
(median)
−0.404 ± 0.037
−0.404 ± 0.028
−0.257 ± 0.036
−0.418 ± 0.037
−0.392 ± 0.027
−0.405 ± 0.040
Scatter in log re
(biweight)
0.134 ± 0.034
0.111 ± 0.024
0.058 ± 0.018
0.106 ± 0.023
0.086 ± 0.026
0.049 ± 0.018
set is ∆ log(M/ L B ) = −0.365 ± 0.037, larger than ∆ log(M/ L B ) = −0.280 ± 0.036 found
by K97 based on older data for a sample of 5 bright galaxies. We will return to this
issue in Sect. 2.5.
We find a biweight scatter for the early-type population in MS 2053–04 as large as
0.134 ± 0.034 in log re , with the error derived from bootstrapping. This is significantly
larger than the observed scatter of 0.071 around the B-band FP of local clusters (JFK96).
After subtraction of the measurement uncertainties in quadrature, we find the intrinsic
scatter for the early-type galaxies to be 0.124 ± 0.037. We conclude that measurement
uncertainties cannot account for all of the enhanced scatter. Not only is the scatter
larger than in the local universe, it also exceeds the previous result of 0.058 ± 0.018 obtained by K97. Our new measurements for 4 early-type galaxies that were also in the
K97 sample give a scatter of 0.116 ± 0.035 (as opposed to 0.050 ± 0.018 for the original
K97 data on these 4 galaxies). A larger sample and new data on previously studied
objects leads to the conclusion that the early-type galaxies in MS 2053–04 show a considerably larger spread around the FP than early-type galaxies in the local universe.
We next analyzed the scatter of the bulge-dominated systems selected by Sersic
index (3 and 4), which is a more objective method of classifying. The scatter drops by
about 20% with respect to the classification by eye, to 0.111 ± 0.024. Using both visual
and quantitative classifications, we find that bulge-dominated systems in MS 2053–04
are less tightly confined to a plane than in the local universe by a factor 1.5 to 2.
Zero-point shifts and scatters from biweight statistics are summarized in Table 2.2.
Median zero-point shifts are given for comparison.
2.4.1.2 MS 1054–03
For the early-type galaxies in MS 1054–03 the zero-point offset agrees well with the
previous result from vD98 (∆ log(M/ L B ) = −0.405 ± 0.037 for the new sample and
∆ log(M/ L B ) = −0.393 ± 0.040 from vD98). The observed scatter is 0.106 ± 0.024, and
the intrinsic scatter is 0.086 ± 0.028, consistent with the scatter in local clusters (JFK96).
Using new measurements of 5 MS 1054–03 early-type galaxies studied by vD98, we
obtain a biweight scatter of 0.062 ± 0.018, consistent with 0.047 ± 0.024 for the original
vD98 data on these objects. As for MS 2053–04, the scatter decreases by roughly 20% if
we select bulge-dominated systems by Sersic index (3 and 4) instead of by eye.
18
Chapter 2. The detailed fundamental plane of two high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83
2.5 Correlations with other parameters
One of the striking effects we found in Sect. 2.4.1 is the large FP scatter for MS 2053–04.
Here we investigate the cause of this enhanced scatter. In the case of a stellar population effect such as variations in age, metallicities or dust content, we expect the residual
from the FP to correlate with color and linestrength indices. We also discuss the residual from the FP as a function of environment and investigate the dependence on galaxy
mass. Hereafter, we again adopt the JFK96 slope and we express residuals from the FP
as deviations in log(M/ L B ). A positive residual means the galaxy has a higher M/ L
than the FP prediction based on its re and σ . The approach of measuring offsets along
the surface brightness axis is physically intuitive since -if we ignore mergers- a galaxy
only moves along this axis during its lifetime. For consistency with our FP analysis,
we adopt locally determined slopes for all other considered scaling relations as well.
In Figure 2.6 the residual from the FP is plotted against various properties for both
clusters. Different symbols refer to different morphological classes. For each panel,
the probability that a random sample has the same Spearman rank order correlation
coefficient as the early-type galaxies in our sample, is printed in the corner.
2.5.1 The color-magnitude relation
The most straightforward interpretation for the large scatter is a stellar population effect. For age, metallicity, and dust trends, we expect lower M/ L ratios than the FP
prediction to correlate with bluer colors than the CMR prediction. We refer the reader
to Tran et al. (2003) for the color-magnitude relation of MS 2053–04 and to van Dokkum
et al. (2000) for the MS 1054–03 CMR. Similar to our FP analysis, we assume that the
slope of the relation does not evolve and adopt the slope measured in the Coma cluster
by Bower, Lucey,& Ellis (1992b). After conversion of the (F606W,F814W) photometry
to Johnson U and V (using the same procedure as K2000), we fit a CMR zero point to
the early-type galaxies in our samples. A positive residual ∆CMR in Figure 2.6a and
Figure 2.6b corresponds to a redder color of the galaxy than the CMR prediction based
on its total V magnitude. The Spearman rank order correlation coefficient points in the
expected direction for a stellar population effect. However, the level of significance is
insufficient to confirm that galaxies with lower M/ L than the FP prediction are bluer
and the higher (more evolved) ones are redder than the CMR prediction.
2.5.2
H β linestrength
Apart from M/ L ratio (residual from FP) and color (residual from CMR), strengths of
absorption lines are a valuable tool for tracing stellar population effects. In this section
we discuss the H β index (expressed in Å, see Trager et al. 1998). H β is an age-sensitive
parameter with only a minor contribution due to metallicity. For spectral reduction and
derivation of linestrengths we refer the reader to Kelson et al. (2006).
The H β − σ relation for early-type galaxies in the Coma cluster was derived from
the H β G − σ relation presented by Jørgensen (1999). The H β G index is related to the
Lick/IDS H β index as H β G = 0.866H β + 0.485 (Jørgensen 1997). Assuming a nonevolving slope of the H β − σ scaling relation, we fit a zero point to the relation in
MS 2053–04 and MS 1054–03.
Section 2.5. Correlations with other parameters
19
Figure 2.6 — Residual from the fundamental plane ∆ log(M/ L B ) plotted against residual from the colormagnitude relation ∆CMR, residual from the H β − σ relation, distance from BCG and galaxy mass.
Different symbols refer to different morphological types. Error bars for colors and angular distances
are assumed to be smaller than the symbol sizes. For the early-type galaxies, the p-values for statistical
significance from the Spearman rank order correlation test are printed in the lower right corners. In
panels g) and h) the magnitude limit I ≈ 21.15 at which serious incompleteness due to uncertainties in
the σ measurements sets in, is indicated with the solid line.
20
Chapter 2. The detailed fundamental plane of two high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83
The residuals from this relation (positive means stronger H β ) are plotted against the
FP residual in Figure 2.6c and Figure 2.6d. Only for MS 2053–04 are the error bars
small enough to draw robust conclusions. A correlation is present, with confidence
level 95.8%. H β absorption lines are stronger for younger stellar populations, and the
correlation with ∆ log(M/ L B ) could confirm that the scatter in the FP is not random
noise, but determined by age variations among the galaxies.
2.5.3 Location in the cluster
Here we consider if the enhanced scatter reported in Sect. 2.4.1 is related to environment. Clusters of galaxies are not isolated systems, as infall of galaxies from the field
occurs. FP studies of field early-type galaxies indicate that their stellar populations
may be somewhat younger than those of their counterparts in clusters (van Dokkum
et al. 2001; Treu et al. 2002; van de Ven, van Dokkum,& Franx 2003; Rusin et al. 2003;
van Dokkum & Ellis 2003; Gebhardt et al. 2003). In the cluster CL1358+62 van Dokkum
et al. (1998) reported evidence for disky galaxies to be systematically bluer and to show
a larger scatter in the CMR at larger radii (with the sample ranging to ∼ 1Mpc). As earlier studies of both clusters (K97; vD98) were based on one pointing, the larger scatter
we find could be explained if residuals from the FP increase with distance from the
BCG. Therefore, we might expect to see a gradient in age and larger scatter around
the FP going to a larger range in cluster radii. For both clusters our samples extend
to roughly 1Mpc from the BCG. Figure 2.6e and Figure 2.6f do not show a significant
correlation.
2.5.4 Galaxy mass and selection effects
Finally we try to explain the range of FP residuals as a function of galaxy mass. The
mass M in solar units of a galaxy is calculated as follows (see JFK96):
log M = 2 log σ + log re + 6.07
(2.8)
Figure 2.6g and Figure 2.6h show residual from the FP against log M. Only if we were
to include the 6 galaxies with σ < 100 km s−1 , the trend of lower mass galaxies to have
lower M/ L ratios than predicted by the FP with JFK96 coefficients is significant at the
95% level.
For a good understanding of Figure 2.6g and Figure 2.6h, we need to take into
account that our FP samples are magnitude-limited, and not mass-limited. The FP can
be rewritten as
0.07
M/ L ∼ M0.28 r−
.
e
(2.9)
In the following, we ignore the dependence on re . Hence the residual is given by
∆M/ L =
M0.72
Mobs / Lobs
= obs .
(M/ L)FP
Lobs
(2.10)
For a fixed luminosity we expect all galaxies to fall on a line in a plot of ∆ log M/ L
versus log M. In Figure 2.6g and Figure 2.6h this line is drawn for I=21.15. At this
Section 2.6. Evolution of M/ L ratio
21
magnitude serious incompleteness due to uncertainties in the σ measurements sets in.
The lowest mass galaxies in Figure 2.6a and Figure 2.6b lie close to the line representing
the magnitude limit. Low mass galaxies that lie on or above the FP would be too faint
to allow for accurate dispersion measurements and would not enter the FP samples.
Hence the few remaining low mass galaxies in our samples are brighter than their FP
prediction based on re and σ .
Selection effects are clearly less relevant at the high mass end, and hence we determine the offset and scatter of the MS 2053–04 FP seperately for the subsample of earlytype galaxies with M > 1011 M⊙ (following van Dokkum & Stanford 2003). For this
subsample of 8 early-type galaxies we derive a zero-point shift with respect to Coma
of ∆ log(M/ L B ) = −0.288 ± 0.056 and a scatter of 0.132 ± 0.039 in log re . As expected,
this is slightly different from the zero-point shift of ∆ log(M/ L B ) = −0.365 ± 0.037 for
the full early-type sample. The zero-point offset is now in good agreement with K97,
but the scatter is still larger.
If we apply a mass cut for the MS 1054–03 early-type galaxies of M > 1011.5 M⊙ , we
find a shift in zero point of ∆ log(M/ L B ) = −0.311 ± 0.051 compared to ∆ log(M/ L B ) =
−0.405 ± 0.037 for the full early-type sample. The scatter of the high mass subsample
(0.104 ± 0.030 in log re ) is similar to that of the full early-type sample (0.106 ± 0.023 in
log re ).
2.5.5 Summary of correlations
We did not find evidence that FP residuals are related to environment. Instead, stellar
population effects are playing a role in shaping the FP and selection effects in our
magnitude-limited samples need to be taken into account.
First, early-type galaxies with stronger H β absorption also tend to have lower M/ L
ratios than predicted by their re and σ . This correlation supports the interpretation
of FP scatter as a measure of age variation among early-type galaxies. The larger FP
scatter in MS 2053–04 therefore reflects a larger spread in relative ages than in the local
universe. Comparison of FP residuals with residuals from the CMR cannot confirm or
rule out the presence of such a stellar population effect.
Second, the fact that we do not see galaxies with low masses that lie on or above
the FP, may be entirely explained by selection effects. Only low mass galaxies that are
brightened by some recent star formation enter the FP sample. Their older counterparts are fainter than the magnitude limit for the dispersion measurements.
2.6 Evolution of M/ L ratio
Correlations with residual from the H β − σ relation show that differences in FP zero
point can be explained by age differences of the stellar population. In this section
we study the evolution of the M/ L ratio as a function of redshift and use this to estimate the mean formation redshift of cluster early-type galaxies. In Figure 2.7a we
show the evolution of the M/ L ratio with respect to Coma. Crosses refer to measured
zero-point shifts from the literature: Coma at z = 0.023 (JFK96), CL1358+62 at z = 0.33
(K2000), CL0024+16 at z = 0.39 (van Dokkum,& Franx 1996), MS 2053–04 at z = 0.583
22
Chapter 2. The detailed fundamental plane of two high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83
Figure 2.7 — Evolution of the M/ L ratio with redshift. The cross symbols are results from the literature,
namely Coma at z = 0.023 (JFK96), CL1358+62 at z = 0.33 (K2000), CL0024+16 at z = 0.39 (van Dokkum
& Franx 1996), MS 2053–04 at z = 0.583 (K97), MS 1054–03 at z = 0.83 (vD98) and J0848+4453 at z = 1.27
(van Dokkum & Stanford 2003). Boxes denote the shift in M/ L ratio based on our larger samples for
MS 2053–04 and MS 1054–03. Single burst models for z f orm = 2 and ∞, assuming a Salpeter IMF (x =
2.35) and a range of metallicities, are drawn with solid and dotted curves resp. Panel (a) is based on
all early-type galaxies, panel (b) shows results for the subsample of massive early-type galaxies (M >
1011 M⊙ for clusters up to z = 0.583 and M > 1011.5 M⊙ for the two higher redshift clusters).
(K97), MS 1054–03 at z = 0.83 (vD98) and J0848+4453 at z = 1.27 (van Dokkum,& Stanford 2003). The new results for MS 2053–04 and MS 1054–03, based on our larger samples, are plotted with boxes. Single burst evolutionary models for a formation redshift
z f orm = 2 and z f orm = ∞ are drawn with a solid and dotted curve respectively. They
represent a galaxy that is fixed in mass and whose luminosity evolves as:
L(t) ∼ 1/(t − t f orm )κ
(2.11)
Here t f orm represents the age of the universe at the moment the stars were formed.
κ depends on the slope of the IMF, passband and metallicity. For a Salpeter (1955) IMF
and −0.5 < [Fe/ H] < 0.5, the models of Bruzual & Charlot (2003), Vazdekis et al. (1996)
and Worthey (1998) give 0.86 < κ B < 1.00. Using the new offsets for the two higher z
clusters, the single burst model for a Salpeter IMF and solar metallicity favoured by the
+0.28
least square method has z f orm ∼ 2.26−
0.20 . The 1σ confidence level was derived from
2
the difference in χ between the model and the overall minimum, ∆χ2 = χ2 − χ2min , to
which Gaussian confidence levels were assigned (e.g., Press et al. 1992). As we showed
in Sect. 2.5.4, the point of MS 2053–04 deviates from the earlier result (K97) since at low
mass only galaxies with low M/L enter the magnitude-limited sample.
Section 2.7. Summary
23
To avoid this bias, we apply a mass cut of M > 1011 M⊙ to all clusters up to z = 0.583.
A mass cut of M > 1011.5 M⊙ was applied to MS 1054–03 and J0848+4453 since at these
higher redshifts the selection effect sets in at a higher galaxy mass. For J0848+4453
two galaxies are left after omitting the one low mass outlier. As the biweight location
estimator is robust against outliers, the zero-point shift only changes slightly for this
cluster. We obtain the evolution of the M/ L ratio as presented in Figure 2.7b. Now the
zero point of the MS 2053–04 FP follows the trend seen for the other clusters in Figure
2.7. If we constrain the analysis of the evolution in M/ L ratio to massive early-type
galaxies, a simple linear fit gives log M/ L B ∼ −0.47z, agreeing well with earlier determinations based on smaller samples or lower redshift (see, e.g., vD98). The formation
+0.81
redshift favoured by a least squares method is z f orm ∼ 2.95−
0.46 , slightly higher than
the mean formation redshift for all early-type galaxies. A similar formation epoch for
early-type galaxies in clusters was found by Kelson et al. (2001) based on the evolution
of Balmer absorption-line strengths with redshift. It is remarkable how earlier studies
of clusters based on smaller samples (see, e.g., K97; vD98) gave similar results.
2.7 Summary
We used visual and quantitative classifications to select bulge-dominated systems in
our MS 2053–04 and MS 1054–03 samples. For MS 2053–04 we find a zero-point offset with respect to Coma of ∆ log(M/ L B ) = −0.365 ± 0.037, larger than determined
earlier on the basis of a smaller sample (K97). The scatter around the MS 2053–04
FP is 0.134 ± 0.034 in log re , enhanced with respect to both K97 and the scatter in local clusters. The FP zero point of MS 1054–03 (∆ log(M/ L B ) = −0.405 ± 0.037) agrees
well with the earlier result of ∆ log(M/ L B ) = −0.393 ± 0.040 by vD98. The scatter of
0.106 ± 0.024 in MS 1054–03 is larger than reported by vD98 for a smaller sample of
MS 1054–03 early-type galaxies. However, taking into account measurement uncertainties, the scatter is consistent with that in local clusters (JFK96). Late-type galaxies
also follow the FP scaling relation and show a similar scatter around the FP as the
early-type galaxies. Adding the late-type galaxies to the early-type sample results in
a scatter of 0.136 ± 0.029 for MS 2053–04 and 0.117 ± 0.031 for MS 1054–03. The larger
samples presented in this chapter allow us to study correlations with other properties
of the early-type galaxies. We do not find evidence that the formation history depends
on environment in the cluster. No significant correlation of FP residuals with CMR
residuals was found. The presence of a correlation between FP residuals and residuals
from the H β − σ relation indicates that stellar population effects are playing a role. Assuming non-evolving slopes for all scaling relations, we find that galaxies with lower
M/ L than the FP prediction tend to show stronger H β than predicted based on the
H β − σ relation. Finally, we show that the lack of low mass galaxies on or above the
FP may be entirely due to selection effects. To avoid a bias induced by the magnitude
limit of our sample, we focus on the high mass end, where selection effects are less
relevant. Applying a mass cut at M > 1011 M⊙ to all 4 considered clusters below z ∼ 0.6
and at M > 1011.5 M⊙ to the 2 higher redshift clusters at z = 0.83 and z = 1.27, increases
+0.81
+0.28
the best fitting formation redshift from z f orm = 2.26−
0.20 to z f orm = 2.95−0.46 . The mass
11
cut at M = 10 M⊙ is well below the typical mass of early-type galaxies: galaxies with
24
Chapter 2. The detailed fundamental plane of two high-redshift clusters:
MS 2053–04 at z = 0.58 and MS 1054–03 at z = 0.83
M = 1011 M⊙ have dispersions of ∼ 168 km s−1 which is significantly lower than the σ∗
dispersion of early-type galaxies which is 228 ± 14 km s−1 (Kochanek 1994).
The implication of this work is that selection effects need to be taken into account, especially if the scatter is high. The scatter in MS 2053–04 is slightly higher than at low
redshift; the scatter in the field at high redshift seems to be even higher (e.g., Gebhardt
et al. 2003; van Dokkum & Ellis 2003). Hence those studies are likely to suffer from
much more significant selection effects.
Acknowledgments
We thank Arjen van der Wel for useful discussions on the fundamental plane. Based
on data obtained at the W. M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California and
the National Aeronautics and Space Administration. The Observatory was made possible by the generous financial support of the W. M. Keck Foundation. The authors
wish to recognize and acknowledge the very significant cultural role and reverence
that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from
this mountain.
References
Beers, T. C., Flynn, K.,& Gebhardt, K. 1990, AJ, 100, 32
Bender, R., Saglia, R. P., Ziegler, B., Belloni, P., Greggio, L., Hopp, U.,& Bruzual, G. 1998, ApJ, 493, 529
Bertin, E.,& Arnouts, S. 1996, A&AS, 117, 393
Bower, R. G., Lucey, J. R.,& Ellis, R. S. 1992b, MNRAS, 254, 589
Bruzual, G.,& Charlot, S. 2003, MNRAS, 344, 1000
Djorgovski, S.,& Davis, M. 1987, ApJ, 313, 59
Dressler, A., Lynden-Bell, D., Burstein, D., Davies, R. L., Faber, S.M., Terlevich, R. J.,& Wegner, G. 1987,
ApJ, 313, 42
Fabricant, D., Franx, M.,& van Dokkum, P. G. 2000, ApJ, 539, 577
Gebhardt, K., et al. 2003, ApJ, in press
Hoekstra, H., Franx, M., Kuijken, K., van Dokkum, P. G. 2002, MNRAS, 333, 911
Jørgensen, I., Franx, M.,& Kjærgaard, P. 1995b, MNRAS, 276, 1341
Jørgensen, I., Franx, M.,& Kjærgaard, P. 1996, MNRAS, 280, 167 (JFK96)
Jørgensen, I. 1997, MNRAS, 288, 161
Jørgensen, I. 1999, MNRAS, 306, 607
Jørgensen, I., Franx, M., Hjorth, J.,& van Dokkum, P. G. 1999, MNRAS, 308, 833
Kelson, D. D., van Dokkum, P. G., Franx, M., Illingworth, G. D.,& Fabricant, D. 1997, ApJ, 478, L13 (K97)
Kelson, D. D., Illingworth, G. D., Franx, M.,& van Dokkum, P. G. 2000a, ApJ, 531, 137
Kelson, D. D., Illingworth, G. D., Franx, M.,& van Dokkum, P. G. 2000b, ApJ, 531, 159
Kelson, D. D., Illingworth, G. D., Franx, M.,& van Dokkum, P. G. 2000c, ApJ, 531, 184 (K2000)
Kelson, D. D., Illingworth, G. D., Franx, M.,& van Dokkum, P. G. 2001, ApJ, 552, L17
Kelson, D. D. 2003, PASP, 115, 688
Kelson, D. D., Illingworth, G. D., Franx, M.,& van Dokkum, P. G. 2006, ApJ, 653, 159
Kochanek, C. S. 1994, ApJ, 436, 56
Oke, J. B., et al. 1995, PASP, 107, 375
Press, W. H., Teukolsky, S. A., Vetterling, W. T.,& Flannery, B. P. 1992, Numerical Recipes. Cambridge
Univ. Press, Cambridge
Rusin, D., et al. 2003, ApJ, 587, 143
Section 2.7. Summary
25
Salpeter, E. 1955, ApJ, 121 161
Schlegel, D. J., Finkbeiner, D. P.,& Davis, M. 1998, ApJ, 500, 522
Trager, S. C., Worthey, G., Faber, S. M., Burstein, D.,& Gonzalez, J. J. 1998, ApJS, 116, 1
Tran, K. V., Kelson, D. D., van Dokkum, P. G., Franx, M., Illingworth, G. D.,& Magee, D. 1999, ApJ, 522,
39
Tran, K. V. 2002, PhD thesis, Univ. California at Santa Cruz
Tran, K. V., et al. 2003, submitted to ApJ
Treu, T., Stiavelli, M., Casertano, S., Møller, P.,& Bertin, G. 2002, ApJ, 564, L13
van de Ven, G., van Dokkum, P.G.,& Franx, M. 2003, MNRAS, 344, 924
van Dokkum, P. G.,& Franx, M. 1996, MNRAS, 281, 985
van Dokkum, P. G., Franx, M., Kelson, D. D., Illingworth, G. D., Fisher, D.,& Fabricant, D. 1998, ApJ,
500, 714
van Dokkum, P. G., Franx, M., Kelson, D. D.,& Illingworth, G. D. 1998, ApJ, 504, L17 (vD98)
van Dokkum, P. G., Franx, M., Fabricant, D., Illingworth, G. D.,& Kelson, D. D. 2000, ApJ, 541, 95
van Dokkum, P. G., Franx, M., Kelson, D. D.,& Illingworth, G. D. 2001, ApJ, 553, 39
van Dokkum, P. G.,& Stanford, S. A. 2003, ApJ, 585, 78
van Dokkum, P. G.,& Ellis, R.S. 2003, ApJ, 592, 53
Vazdekis, A., et al. 1996, ApJS, 106, 307
Worthey, G. 1998, PASP, 110, 888W
Chapter 3
B-to-24 µm photometry of the
GOODS-CDFS:
multi-wavelength catalog and
total IR properties of distant Ks-selected
galaxies
Abstract. We present a Ks -selected catalog for the Chandra Deep Field South
(CDFS) containing photometry in B435 , V606 , i775 , z850 , J, H, Ks , [3.6 µm], [4.5 µm],
[5.8 µm], [8.0 µm], and the MIPS [24 µm] band. The imaging has a typical Kstot
, AB
limit of 24.3 mag (5σ ) and coverage over 113 arcmin2 in all bands and 138 arcmin2
in all bands but H. We cross-correlate our catalog with the 1 Ms X-ray catalog
by Giacconi et al. (2002) and with all available spectroscopic redshifts to date.
We find systematic differences due to aperture corrections in a comparison with
the ’z + Ks ’-selected GOODS-MUSIC catalog that covers ∼90% of the field. We
exploit the B-to-24 µm photometry to determine which Ks -selected galaxies at
1.5 < z < 2.5 have the brightest total IR luminosities and which galaxies contribute
most to the integrated total IR emission. The answer to both questions is that red
galaxies are dominating in the IR. This is true no matter whether color is defined
in the rest-frame UV, optical, or optical-to-NIR. We do find however that among
the reddest galaxies in the rest-frame optical, there is a population of sources with
only little mid-IR emission.
S. Wuyts, N. M. Förster Schreiber, M. Franx, I. Labbé, G. Rudnick & P. G. van Dokkum
27
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
28
IR properties of distant Ks -selected galaxies
3.1 Introduction
S
the original Hubble Deep Field (Williams et al. 1996), deep multi-wavelength
observations of blank fields have revolutionized our understanding of the highredshift universe. Especially the epoch around z ∼ 2 is of great interest since it is then
that the cosmic star formation rate density was peaking (Hopkins & Beacom 2006). At
z ∼ 2, the observed optical probes the redshifted rest-frame UV emission of young O
and B stars, making it a good tracer for relatively unobscured star formation. NearInfrared (NIR) observations of distant galaxies, such as undertaken by the FIRES survey in the Hubble Deep Field South (HDFS, Labbé et al. 2003, hereafter L03) and the
MS 1054–03 field (Förster Schreiber et al. 2006, hereafter FS06), show relatively small
variations in the mass-to-light ratio. Selecting galaxies in the Ks -band (e.g., L03; FS06)
thus provides a good probe of the massive galaxy content at high redshift.
In the presence of dust, large amounts of rest-frame UV emission can be absorbed
and re-emitted in the Far-Infrared (FIR). Dust corrections of the UV luminosities of
such systems involve large uncertainties. Direct observations of the dust emission are
therefore crucial to get a better census of the bolometric energy output. Unfortunately,
current submillimeter observations (e.g., Smail et al. 1997) are only sensitive enough to
detect the most luminous dust-enshrouded starbursts. In order to study the bolometric
properties of typical galaxies at z ∼ 2, infrared luminosities have been derived from the
observed 24 µm flux by means of IR spectral energy distribution (SED) templates (e.g.,
Papovich et al. 2005; Reddy et al. 2006). Despite the extra model uncertainty involved,
this approach adds complementary information to the shorter wavelength studies of
high-redshift galaxies.
In this chapter, we present a Ks -band selected multi-wavelength catalog for the
GOODS-CDFS, comprising ACS BViz, ISAAC JHKs, IRAC 3.6-8.0 µm and MIPS 24
µm imaging. We adopt a similar format as for the FIRES catalogs of the HDFS and
MS 1054–03. This allows the user to exploit the combined photometry of the CDFS,
MS 1054–03, and the HDFS in a straightforward manner. The fields are complementary
tot
tot
in depth (5σ for point sources Kstot
, AB = 24.3, Ks , AB = 25.0, and Ks , AB = 25.6 respectively)
and area (138 arcmin2 , 24 arcmin2 , and 5 arcmin2 respectively).
An analysis of the space density and colors of massive galaxies at 2 < z < 3 (van
Dokkum et al. 2006), of the rest-frame optical luminosity density and stellar mass
density up to z ∼ 3 (Rudnick et al. 2006), and of the rest-frame luminosity functions
of galaxies at 2 < z < 3.5 (Marchesini et al. 2006) were partly based on the presented
catalog for the GOODS-CDFS presented here.
After describing the catalog construction, we particularly address the questions
which Ks -selected galaxies at 1.5 < z < 2.5 are brightest at 24 µm, which galaxies have
the largest total infrared luminosity L I R (≡ L(8 − 1000 µm)) and contribute most to
the total integrated IR luminosity emitted by Ks -selected galaxies. We address this
question by studying the IR emission as function of color defined in three wavelength
regimes: the rest-frame UV, optical, and optical-to-NIR.
An overview of the observations is presented in §4.3. §3.3 describes the construction of the final mosaics. Source detection and photometry is discussed in §3.4. Next,
we present our photometric redshifts (z phot ) and cross-correlation with the available
INCE
Section 3.2. Observations
29
spectroscopic surveys in §3.5. §3.6 summarizes the catalog content. A photometric
comparison for the wavelength bands in common with the GOODS-MUSIC catalog by
Grazian et al. (2006a) and a z phot comparison with the same authors is discussed in §3.7.
Results on 24 µm properties and total infrared luminosities of Ks -selected galaxies at
1.5 < z < 2.5 are discussed in §3.8. §7.11 summarizes the chapter.
AB magnitudes are used throughout this chapter.
3.2 Observations
3.2.1 The GOODS Chandra Deep Field South
Centered on (α, δ ) = (03:32:30, -27:48:30), the CDFS (Giaconni et al. 2000) has been
targeted by most of today’s major telescope facilities, both in imaging mode over the
whole spectral range and in spectroscopic mode. In this section, we describe the public
GOODS-South dataset that we used to build a Ks -band selected catalog containing
homogeneous colors from the optical to 24 µm.
3.2.2 The ACS BViz data
During 5 epochs of observations, the ACS camera on HST acquired imaging of the
GOODS-South field in 4 filter bands: F435W, F606W, F775W, and F850LP (hereafter referred to as B,V,i, and z). Exposure times amounted to 7.2, 6, 6, and 12 ks respectively.
The mosaics (version v1.0; Giavalisco et al. 2004), were drizzled onto a pixelscale of
0.′′ 03 pixel−1 . From the 150 arcmin2 area that is well covered by the Ks -band detection image, 138 arcmin2 is well exposed with ACS. We restrict our analysis of the IR
properties of the distant galaxy population in §3.8 to this overlap region.
3.2.3 The ISAAC JHKs data
We use the ESO/GOODS data release v1.51 to complement the optical observations
with NIR imaging by the Very Large Telescope (VLT). For a full description of the
dataset, we refer the reader to Vandame et al. (in preparation). Briefly, the v1.5 data
release consists of 24 fully reduced VLT/ISAAC fields in the J and Ks bands and 19
fields in the H-band, each with a 2.′′ 5x2.′′ 5 FOV and 0.′′ 15 pixel−1 scale. The ISAAC data
were reduced using the ESO/MVM image processing pipeline (v1.9, see Vandame 2002
for the description of an earlier version). Exposure times varied from field to field,
with typical exposures of 11.5 ks, 15 ks, and 18 ks in J, H, and Ks -band respectively,
and respective ranges between ISAAC fields of 7-18 ks, 7-22 ks, and 13-27 ks. The
variations in depth resulting from the unequal exposure times are discussed in §3.3.5.
A total area of 113 arcmin2 is well exposed in all optical and NIR filter bands. Without
a restriction on the H-band, the covered area increases to 138 arcmin2 .
3.2.4 The IRAC 3.6-8.0 µm data
As a Spitzer Space Telescope Legacy Program, superdeep images of the GOODS-South
field were taken with the Infrared Array Camera (IRAC, Fazio et al. 2004) on-board
1
http://www.eso.org/science/goods/releases/20050930/
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
30
IR properties of distant Ks -selected galaxies
Spitzer. Over 2 epochs the whole field was covered in the 3.6 µm, 4.5 µm, 5.8 µm, and
8.0 µm bands. For each epoch, exposure times per channel per sky pointing amounted
to 23 hours. With the telescope orientation being rotated by 180 degrees between the
two epochs, the second epoch IRAC channel 1 and 3 observations targeted the area
covered by IRAC channel 2 and 4 during the first epoch, and vice versa. An overlap
region of roughly 40 arcmin2 , including the Hubble Ultra Deep Field (Beckwith et al.
2003), got twice the exposure time. We use the data releases DR2 and DR3 for the
second and first epoch respectively. Images were released on a 0.′′ 60 pixel−1 scale. A
full description of the observations and reduction will be presented by Dickinson et al.
(in preparation).
3.2.5 The MIPS 24 µm data
The GOODS-South field was observed at 24 µm with the Multiband Imaging Photometry for Spitzer (MIPS, Rieke et al. 2004) on-board Spitzer, closely overlapping the
IRAC fields with a position angle that is rotated with respect to the IRAC observations
by approximately 3 degrees. The MIPS campaign led to a nearly uniform exposure
time of 10 hours. We use the version v0.30 reduced images, released on a 1.′′ 20 pixel−1
scale, based on the Spitzer Science Center (SSC) Basic Calibrated Data (BCD) pipeline
(version S11.0.2).
3.3 Final images
In this section, we describe the image quality of the publicly released data products, the
subsequent steps undertaken to obtain the final mosaics from which the photometric
catalog is extracted, and the limiting depths reached at all wavelengths.
3.3.1 Pixel scales and large scale backgrounds
First, we converted the ACS images to the 0.′′ 15 pixel−1 scale of the Ks -band detection
image, using the IRAF blkavg task with flux conservation. All optical and NIR photometry was performed on this pixel scale using the SExtractor software version 2.2.2
(Bertin & Arnouts 1996) (see §3.4).
A source fitting algorithm developed by Labbé et al. (in preparation), especially
suited for heavily confused images for which a higher resolution prior (in this case
the Ks -band image) is available, was used to extract photometry from the IRAC and
MIPS images. The algorithm requires a higher resolution image than provided by the
IRAC and MIPS images. However, the native Ks -band pixel scale makes it more computationally expensive without benefit in accuracy. A version of the Ks -band mosaic
was therefore produced on a 0.′′ 3 pixel−1 scale. We registered the publicly released
IRAC and MIPS images onto this Ks -band image using the WCS information in the
image header in combination with a minor additional shift, again forcing flux conservation. We note that the source fitting algorithm takes care of residual shifts. Since the
programme does not take into account large scale background variations, these were
removed a priori by subtracting SExtractor background images produced with large
background mesh settings.
Section 3.3. Final images
31
3.3.2 Image quality and PSF matching
3.3.2.1 Optical-to-NIR wavelengths
In order to obtain consistent color measurements, we match all optical and NIR images to a common resolution, namely that of the field with the broadest point spread
function (PSF). In this section, we describe the selection of stars used to build the PSFs,
the construction of the PSFs for the ACS B-, V-, i-, and z-mosaics and for each of the
ISAAC fields in the J-, H-, and Ks -band, the construction of the convolving kernels,
and the quality of the PSF matching.
First, we compiled a list of bright, isolated, unsaturated stars. Initially, well covered
objects with (J − Ks ) AB < 0.04 and Kstot
, AB < 22.86 mag were selected from a preliminary
catalog of the CDFS. For ISAAC fields where the number of J − Ks selected stars was
low, we complemented the sample with stars from the EIS stellar catalog (Groenewegen et al. 2002). During a first iteration, the list was cleaned from galaxy-like objects,
stars with neighbors within 3” radius, stars too close to the edge of an image and objects that were not identified in the ACS r1.1z catalog (Giavalisco et al. 2004) or with a
FWHM in the z-band larger than 0.′′ 13. Measurements of the FWHM were performed
by fitting Moffat profiles to the stars using the imexam task in IRAF. We excluded stars
based on a 2σ clipping of the measured FWHMs. Finally, we inspect by eye the radial
profiles and curves of growth, produced with the IRAF tasks radprof and phot respectively. For each ISAAC field the PSF was determined, and the same stars were used
to build the J-, H-, and Ks -band PSF. The number of stars ranged from 3 to 5 stars per
ISAAC field, with the exception for field f30, for which only 1 good star was available.
The numbers of stars used to build the ACS PSFs were 31, 45, 49 and 53 for the B, V, i,
and z mosaics respectively.
Next, we computed PSF images per ISAAC field by averaging the registered and
flux-normalized images of the selected stars. The flux was normalized within 1.′′ 5
diameter apertures rather than the total aperture to optimize the signal-to-noise and
avoid contributions from residual neighboring sources. Any neighbors in the ISAAC
images of good stars, sufficiently far not to bias the FWHMs and PSFs, were masked
while averaging. This method was preferred over taking the median, since only a
handful of good stars per ISAAC field were available to build a PSF. PSFs for the ACS
mosaics were constructed from a large enough number of stars to average out any
influence of faint neighboring sources without masking.
The ACS PSFs in the B-, V-, i-, and z-band, as measured using Mofat profiles on the
5x5 blocked ACS mosaics, had a FWHM of 0.′′ 22, 0.′′ 22, 0.′′ 21, and 0.′′ 22 respectively. The
seeing FWHM of the NIR ISAAC observations varied from 0.′′ 35 to 0.′′ 65, with median
values of 0.′′ 47, 0.′′ 48, and 0.′′ 47 in J, H, and Ks respectively. Figure 3.1 illustrates the
distribution of FWHMs of the natural PSFs for the individual ISAAC fields. In all of
the considered bands and fields, the FWHM of the individual stars were within ≈10%
of that measured on the final PSF. We adopted the 0.′′ 65 H-band PSF of ISAAC field f15
as target to which all higher resolution images were matched.
We computed the kernel for convolution for each ISAAC field and band separately,
using the Lucy-Richardson deconvolution algorithm. The ratio of the growth curve of
the convolved PSF over that of the target PSF is a good measure for the PSF matching
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
32
IR properties of distant Ks -selected galaxies
Figure 3.1 — Distributions of seeing
FWHM for the ISAAC J, H, and Ks
observations. Moffat profiles were fitted to the PSFs that were built from
bright, isolated, unsaturated stars for
each field and band separately.
accuracy. In order to minimize the discrepancies between both growth curves, we
performed the deconvolution using a series of sizes for the postage stamp images of
the PSFs, from 1.′′ 7 to 5.′′ 9 on a side. The kernel corresponding to the box size that
gave the curve of growth ratio closest to unity was adopted for the final convolution.
Overall, the ratio of growth curves deviated by at most 5.1% from unity for apertures
between 1” (≈1.5FWHM of the PSF of the smoothed field maps) and 6” (the reference
aperture for photometric calibration), with an average of 0.54% ± 0.90%. Flux is well
conserved during the convolution process, with an average deviation of 0.37% and
maximum discrepancy of 4.2% for one of the fields.
The construction of convolving kernels for the ACS mosaics required an extra step.
The kernels obtained from deconvolution with the IRAF Lucy task had significant
noise in the outer parts, leading to noise spikes around bright point-like sources in the
convolved ACS mosaics. To remove these artifacts, we modeled the ACS-to-ISAAC
kernels by fitting isophotes using the IRAF tasks ellipse and bmodel, and used the
modeled kernels for the convolution. This is possible because the kernels are otherwise well behaved and very azimutally symmetric.
Because of the different basic shapes of the ACS and ISAAC PSFs, an excellent
matching over the relevant radii is more difficult than in general among ISAAC fields.
Nevertheless, the offsets of the growth curve ratios between 1” and 6” are limited to
below 4.7%, with an average of 1.58% ± 1.32%. The average over all stars of the ratio
of the flux measured in the convolved and that measured in the natural image showed
a flux conserving accuracy of 0.7% or better for all ACS bands.
3.3.2.2 MIR wavelengths: IRAC and MIPS 24 µm
The instrumental PSF at Mid-Infrared wavelengths is significantly broader than that of
our Ks -band detection image. The FWHM measured on the average image of bright,
Section 3.3. Final images
Table 3.1 — H-band zero points in the AB system derived from the
NIR stellar locus
33
Field
03
04
05
08
09
10
11
13
14
15
16
19
20
21
22
23
24
25n
26n
H-band zero point
25.99
26.02
26.07
25.89
25.92
25.94
25.93
26.02
25.82
26.07
25.97
25.86
25.89
25.97
26.03
26.07
25.95
25.94
26.09
isolated stars in the IRAC images amounts to 1.′′ 6, 1.′′ 7, 1.′′ 9, and 2.′′ 0 for the 3.6 µm,
4.5 µm, 5.8 µm, and 8.0 µm bands respectively. The MIPS 24 µm beam even has a
FWHM as large as 6”. Since confusion and blending effects are unavoidable in deep
observations at this resolution, we decide not to degrade the optical and NIR images
to the MIR resolution. Instead, we construct PSFs and convolving kernels similarly as
described in §3.3.2.1, but apply them using a source fitting algorithm that makes fully
use of the higher resolution information in the Ks -band detection image (see §3.4.2.2).
3.3.3 Zero points
The zero-point calibrations for all bands but the H-band were taken from the respective
GOODS data release. In the case of the NIR ISAAC observations, the publicly released
zero points were based on SOFI images of the EIS-DEEP and DPS infrared surveys
conducted over the same region (Vandame et al. 2001), which themselves were photometrically calibrated using standard stars from Persson et al. (1998). That procedure
yielded zero points with rms scatters ranging between 0.01 and 0.06 mag in the J-band,
0.01 and 0.08 mag in the Ks -band and up to 0.17 mag in the H-band.
To improve on the H-band calibration, we make use of stellar photometry in the
FIRES HDFS (L03) and MS 1054–03 (FS06) fields, for which H-band zero points were
determined to a ∼ 0.03 mag accuracy. For each of the 19 ISAAC fields with H-band
coverage, we measured the mean offset of the stars used for PSF matching along the
J − H axis of a J − Ks versus J − H color-color diagram with respect to the stellar locus
in the FIRES fields. Assuming the J- and Ks -bands are well calibrated, this immediately provides us with the H-band zero point corrections to be applied. We list the
derived H-band zero points in Table 3.1. Zero-point corrections ranged from -0.18 mag
to 0.09 mag, with a median correction over all fields of -0.03 mag. After applying
the zero-point correction, the median absolute deviation in J − H color of individual
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
34
IR properties of distant Ks -selected galaxies
Figure 3.2 — Map of residual shifts
of compact sources in the Ks -band
mosaic with respect to the reference
ACS i-band mosaic. 2σ -clipped reference sources used for the alignment are indicated in black. Grey
vectors represent the residual shift
of the 2σ outliers.
stars around the stellar locus is 0.03 mag, similar as measured for the FIRES HDFS and
MS 1054–03 fields.
3.3.4 Mosaicing and astrometry
Here, we describe the combination of the smoothed ISAAC NIR fields and the astrometric precision of the final mosaics. The 5x5 blocked and smoothed ACS i-band
mosaic was adopted as astrometric reference image. The astrometric solution for the
ACS data itself was based on a cross-identification of sources with deep ground-based
WFI data that on its turn was astrometrically matched to stellar positions in the Guide
Star Catalog 2 (GSC2, STScI 2001). The final solution had a clipped rms deviation of
< 0.′′ 01 in ACS-to-ACS and 0.′′ 12 in ACS-to-ground difference.
∼
The smoothed ISAAC fields were registered onto the smoothed ACS i-band mosaic
by applying simple x- and y- shifts without further distortion corrections. In each
ISAAC field, we measured the shift with respect to the ACS i-band mosaic for stars
and compact sources using the imexam task in IRAF. A 2σ clipped sample of reference
sources typically consisted of 15-20 objects per ISAAC field. The difference between
the shifts implied by individual reference sources and the final astrometric solution
had a standard deviation of less than 0.′′ 6 in all NIR bands. A map of residual shifts for
the Ks -band mosaic with respect to the convolved ACS i-band mosaic is presented in
Figure 3.2.
First we applied the fractional pixel shift for each ISAAC field in each band using
the IRAF imshift task with a cubic spline interpolation. Next, we summed the integer
pixel shifted fields applying an identical weighing scheme as described by FS06 to
optimize the S/N for point sources, namely:
w pix =
wnorm
(rms1.5FW HM)2
(3.1)
Section 3.3. Final images
35
where the weight factor for a given pixel w pix equals its value in the normalized weight
map wnorm , scaled with the square of the rms noise measured within an aperture of
1.5FWHM diameter.
We chose not to combine the 2 epochs of IRAC observations into one mosaic because the 180 degrees difference in position angle would lead to a different PSF shape
in the overlap region than in either of both single epoch areas, demanding the use of
a different convolving kernel over different parts of the field. Instead, we treat each
of the IRAC epochs independently, providing an empirical quality check of the photometry in the overlap region. The registration of each of the IRAC images (epoch 1
and 2) onto the 2x2 blocked Ks -band image has a positional accuracy of better than 0.′′ 4,
as measured from offsets between bright star positions on IRAC and Ks -band images.
The positional accuracy for the MIPS images is of the order of 0.′′ 3 rms. We note that
minor positional offsets between the Ks and IRAC/MIPS image are solved for by the
source fitting algorithm applied to IRAC and MIPS photometry (see §3.4.2.2).
3.3.5 Signal to noise and limiting depths
We analyzed the noise properties of the optical-to-24 µm imaging following the same
approach as for the FIRES HDFS (L03) and MS 1054–03 (FS06) data. Briefly, the technique uses aperture photometry on empty parts of the image to quantify the rms of
background pixels within the considered aperture size. For each convolved ISAAC
field in each band, between 200 and 400 non-overlapping apertures were randomly
placed at a safe distance from the nearest segmentation pixels in a SExtractor segmentation map. For a given aperture size, the distribution of empty aperture fluxes is
well-fitted by a Gaussian, as illustrated in Figure 3.3(a). We applied a 5σ clipping in
determining the background rms. Panel (b) of Figure 3.3 shows that a simple linear
scaling of the measured background rms σ (N) = N σ̄ , where N is the linear size of
the aperture and σ̄ is the pixel-to-pixel rms, would lead to underestimated flux uncertainties. The reason is that correlations between neighboring pixels were introduced
during the reduction and PSF matching. We model the background rms as a function
of aperture size with a polynomial of the form
σi (N) =
N σ̄ (ai + bi N)
√
wi
(3.2)
where i refers to the considered band and field, and the weight term wi is derived
from the weight map of the respective field. Figure 3.3(b) illustrates the variations in
depth for the different ISAAC fields, originating from variable integration times and
observing conditions, and reflected in the range of flux uncertainties for objects with
similar color aperture in the final catalog. For example, the upper two curves in the
ISAAC Ks panel correspond to fields f03 and f04 that had the lowest integration time.
For the ACS mosaics, we used the same empty apertures as for the NIR, provided
they were within the ACS FOV. Every object below the Ks -band detection threshold,
even though detectable in the ACS imaging, contributes to the background noise and
photometric uncertainties of Ks -band detected sources. If we were to restrict our empty
aperture analysis to apertures that contain neither Ks -band nor ACS segmentation pixels, the background rms estimates for the ACS mosaics would decrease by 3 to 9%. In
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
36
IR properties of distant Ks -selected galaxies
Figure 3.3 — The background rms derived from the distribution of fluxes within empty apertures. (a)
Distribution of empty aperture fluxes within a 1”, 2”, and 3” aperture diameter on the Ks -band image of
ISAAC field f15. The distribution is well described by a Gaussian with an increasing width for increasing
aperture size. (b) Background rms as derived from flux measurements within empty apertures versus
aperture size for the ACS bands and the J, H, and Ks ISAAC fields. Solid lines represent the functional
form from Eq. 3.2 fit to the observed rms noise values. Dashed lines indicate a linear extrapolation of the
pixel-to-pixel rms. Correlations between pixels introduce a stronger than linear scaling with aperture
size.
Figure 3.3(b), we scaled the background rms measured on the ACS and ISAAC images
to the flux corresponding to AB = 26.
To characterize the noise for each object, we applied the noise as measured with an
aperture of the same size as that used for the photometry.
3.4 Source detection and photometry
3.4.1 Ks -band detection
We aimed to construct a catalog that is especially suited to extract stellar mass-limited
samples from (e.g., van Dokkum et al. 2006). Although the rest-frame NIR, probed
by IRAC, is a better tracer for stellar mass than the rest-frame optical, the downside
is its coarser resolution, leading to severe confusion. Therefore, we decided to detect
sources in the observed Ks -band.
We used the SExtractor v2.2.2 source extraction software by Bertin & Arnouts (1996)
to detect sources with at least 1 pixel above a surface brightness threshold of µ(Ks, AB) =
24.6 mag arcsec−2 , corresponding to ≈ 5σ of the rms background for a typical Ks -band
Section 3.4. Source detection and photometry
37
field. Setting the threshold to the same number of ADUs across the image instead of
adopting a S/ N criterion was favored, since in the latter case the varying noise properties in the Ks -band mosaic would lead to different limiting magnitudes and limiting
surface brightnesses from one field to the other. We smoothed the detection map with
a gaussian filter of FWHM = 0.′′ 65, the size of the PSF in the detection image. This
procedure optimizes the detection of point sources.
The resulting catalog contains 6308 sources, 5687 of which have a weight in Ks
above 30% of the median weight, which is above ∼ 10% of the maximum weight that
is reached in one of the overlap regions between ISAAC fields. Running SExtractor
with identical parameters on the inverse detection map, we obtain a total of 43 spurious sources in the area with more than 30% of the median weight. Only one of these
has S/ NKs > 5. The estimated fraction of false detections increases from < 1% to < 3%
(< 8%) as we lower the weight criterion from 30% to 20% (10%) of the median weight
in the Ks -band mosaic. The estimated fraction of false detections with S/ NKs > 5 stays
below 0.6% in the area with more than 10% of the median weight.
SExtractor flagged 12% of the detected sources as blended and/or biased. These
sources were treated separately in doing the photometry.
3.4.2 Photometry
3.4.2.1 Optical and NIR photometry
We performed the photometry on the convolved B-to-Ks mosaics using SExtractor in
dual image mode, with the Ks -band mosaic as detection map. We derive the color
and total aperture from the detection image. The same apertures were used in each
band. We follow L03 and FS06 in defining the color aperture based on the Ks -band
isophotal aperture, more precisely on the equivalent circularized isophotal diameter
diso = 2(Aiso /π )1/2, where Aiso is the area of the isophotal aperture. For isolated sources,
we apply

′′
′′

 APER(ISO), 1. 0 < diso < 2. 0
APER(1.′′0), diso ≤ 1.′′ 0
APER(COLOR) =


APER(2.′′0), diso ≥ 2.′′ 0
(3.3)
where APER(ISO) refers to the isophotal aperture defined by the surface brightness
detection threshold. Blended sources (indicated with SExtractor flag “blended” or “biased”) were treated separately,

′′
′′

 APER(diso/s), 1. 0 < diso /s < 2. 0
APER(1.′′0), diso /s ≤ 1.′′ 0
APER(COLOR) =


APER(2.′′0), diso /s ≥ 2.′′ 0
(3.4)
where the reduction factor s for the aperture sizes is introduced to minimize contamination by close neighbors. We adopt the optimal value of s = 1.4 that was determined
from experimentation by L03 and FS06.
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
38
IR properties of distant Ks -selected galaxies
The motivation for the tailored isophotal apertures defined in Eq. 3.3 and Eq. 3.4 is
that it maximizes the S/ N of the flux measurement. The minimum diameter of 1.′′ 0 corresponds to 1.5 × FWHM of the PSF-matched mosaics. The maximum diameter of 2.′′ 0
was adopted to avoid flux from neighboring sources and avoid the large uncertainties
corresponding to large isophotal apertures.
SExtractor’s “MAG AUTO” was used to derive the total flux of the Ks -band detected objects, unless the source was blended, in which case the total aperture was set
to the color aperture:
(
APER(AUTO), isolated sources
APER(TOTAL) =
(3.5)
APER(COLOR), blended sources
Finally, an aperture correction was applied to compute the total integrated flux. The
correction factor equaled the ratio of the total flux of a star inside a 3” radius to its flux
inside a radius rtot, where rtot = (Atot /π )1/2 is the radius of a circle with the same area
as the total aperture.
Flux uncertainties in both color and total aperture were derived from Eq. 3.2. The
quoted uncertainties thus take into account both the aperture size used for the flux
measurement and the limiting depth in the respective region of the mosaic.
3.4.2.2 IRAC and MIPS 24 µm photometry
The photometry of Ks -band detected objects in the Spitzer IRAC and MIPS 24 µm imaging of the CDFS was performed by IL. For an in-depth discussion of the source fitting
algorithm used, and simulations of its performance, we refer the reader to Labbé et
al. (in preparation). A short description with illustration was also presented by Wuyts
et al. (2007). Briefly, the information on position and extent of sources based on the
higher resolution Ks -band segmentation map was used to model the lower resolution
3.6 µm to 24 µm images. Each source was extracted separately from the Ks -band image and, under the assumption of negligible morphological K-corrections, convolved
to the IRAC or the MIPS resolution as needed. A fit to the IRAC/MIPS image was then
made for all sources simultaneously, where the fluxes of the objects were left as free parameters. Next, we subtracted the modeled light of neighboring objects and measured
the flux on the cleaned IRAC/MIPS maps within a fixed aperture, 3” for the IRAC
bands and 6” for the MIPS 24 µm band. Using growth curves of the IRAC and Ks -band
PSFs, we then scaled the photometry to the same color apertures that were used for the
optical and NIR photometry, allowing a straightforward computation of colors over a
B-to-8 µm wavelength baseline. An aperture correction based on the growth curve of
the 24 µm PSF was applied to scale the 24 µm flux measurents to the integrated 24 µm
flux.
Uncertainties in the measured fluxes in the 3.6 µm to 24 µm wavelength bands have
a contribution from the background rms (see §3.3.5) and from the residual contamination of the subtracted neighbors. Here, we follow an empirical approach to validate
the size of the uncertainties in the IRAC photometry. We exploit the overlap region between the 2 independent observation epochs of the CDFS with the IRAC instrument.
The position angle was rotated over 180 degrees, causing the PSF to have a different
Section 3.4. Source detection and photometry
39
Figure 3.4 — Comparison between IRAC observations from epoch 1 and epoch 2 for Ks -band detected
sources in the overlap region between the 2 epochs. The large panels show a good correspondence
between the 2 independent photometric measurements, with a slight zero-point
q drift of 0.03 mag in the
3.6 µm band. The inset panels shows the distribution of ( f epoch1 − f epoch2)/
err2epoch1 + err2epoch2, where
a minimum relative error of 3% was assumed to account for relative zero-point uncertainties over the
field. The standard deviation of the distribution is of order unity, meaning that estimated flux errors
account well for the empirically determined uncertainties.
orientation with respect to the positions of neighboring sources. In Figure 3.4, we show
the difference between the IRAC magnitude measured during epoch 1 and epoch 2.
The rms ranges from 5% in the 4.5 µm band to 10% in the 8.0 µm band for sources with
an AB magnitude brighter than 22. The largest systematic offset was measured for the
3.6 µm band, where a zero-point drift of 0.03 mag was measured
between the 2 epochs.
q
2
In the inset panels the distribution of ( f epoch1 − f epoch2)/ errepoch1 + err2epoch2 is plotted.
The distribution is well described by a gaussian. For well estimated errors the expected
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
40
IR properties of distant Ks -selected galaxies
Table 3.2. Spectroscopic redshifts for Ks -band detected objects
Survey
High quality flags
Numbera
FORS2 (v2.0)
K20 (Mignoli et al. 2005)
VVDS (v1.0, Le Fèvre et al. 2004)
CXO (Szokoly et al. 2004)
Norman et al. (2002)
Croom et al. (2001)
van der Wel et al. (2004)
Cristiani et al. (2000)
Strolger et al. (2004)
Daddi et al. (2004)
IMAGES (Ravikumar et al. 2006)
LCIRS (Doherty et al. 2005)
Wuyts et al. (Chapter 4)
Kriek et al. (2007)
A
1
4,3
3,2,1
all
all
all
all
all
all
1
3
all
all
324
263
247
92
1
20
21
3
7
7
107
3
7
2
a The
numbers are non-redundant. For objects targeted during
multiple surveys, the redshift with the highest quality flag was
adopted.
standard deviation of the distribution is 1. We adopted a minimum relative uncertainty
in the flux of 3% to account for zero-point variations over the field. This is particularly relevant for the 3.6 µm and 4.5 µm band, where the sources
q are detected with a
high signal-to-noise. The standard deviation of ( f epoch1 − f epoch2)/ err2epoch1 + err2epoch2 in
these bands is smaller than 1. Adopting a more conservative minimum relative uncertainty would only decrease this value, suggesting that zero-point variations within
the field are limited to the few percent level. In the less sensitive 5.8 µm and 8.0 µm
bands, where the minimum
relative uncertainty is not reached, we find a distribution
q
of ( f epoch1 − f epoch2)/ err2epoch1 + err2epoch2 with a standard deviation of nearly unity, confirming empirically the validity of our estimated uncertainties.
3.5 Redshifts
3.5.1 Spectroscopic redshifts
The CDFS-GOODS area has been targeted intensively by various spectroscopic surveys, listed in Table 3.2. The combined sample of spectroscopic redshifts forms a heterogeneous family of objects, with selection criteria varying from pure I-band (VVDS,
Le Fèvre et al. 2004), Ks -band (Mignoli et al. 2005) or X-ray (Szokoly et al. 2004) flux
limits to various color criteria (e.g., Doherty et al. 2005; Chapter 4). It is therefore
impossible to build a combined spectroscopic sample that is complete in any sense.
Rather, we aim to provide a list of trustworthy spectroscopic redshifts that are reli-
Section 3.5. Redshifts
41
Figure 3.5 — Comparison between photometric and spectroscopic redshifts for 814 Ks -band detected
sources with reliable zspec identification and coverage in all wavelength bands. (a) A direct comparison
with 68% confidence intervals determined from Monte Carlo simulations. (b) The distribution of ∆z/(1 +
z). 5% of the sources fall outside the plotted range.
ably cross-identified with a Ks -band detection in our catalog. To do so, we apply a
conservative quality cut based on the quality flags that come with each of the spectroscopic catalogs, and assign the redshift to the nearest Ks -band selected object within a
radius of 1.′′ 2. The quality flags and number of sources included in our reliable list of
cross-correlated spectroscopic redshifts are summarized in Table 3.2. We mark these
sources with a “zsp qual” flag of 1 in our catalog. For completeness, other spectroscopic redshifts for Ks -band detected objects are also listed in our catalog, marked with
a “zsp qual” flag lower than 1, together with the original quality flag from the respective survey. We proceed to use only the 1104 spectroscopic redshifts with zsp qual =
1.
3.5.2 Photometric redshifts
Together with the observed photometry, we release a list of photometric redshifts computed with the algorithm described by Rudnick et al. (2001; 2003). The algorithm fits a
linear combination of template spectra to the optical-to-NIR spectral energy distribution. The template set consisted of 10 Single Stellar Population (SSP) templates with a
Salpeter (1955) initial mass function and solar metallicity from the Bruzual & Charlot
(2003) stellar population synthesis code, with ages logarithmically spaced between 50
Myr and 10 Gyr. We allowed each of the templates to be attenuated according to the
Calzetti et al. (2000) law by E(B − V) = 0.0, 0.1, 0.3, or 0.6.
The accuracy of the photometric redshifts z phot is quantified by a comparison to
the spectroscopic redshifts with zsp qual = 1. Figure 3.5(a) shows the correspondence
between z phot and zspec for all 814 sources with Kstot
, AB < 24.3 that are covered by all
bands and for which a reliable spectroscopic redshift is available. The uncertainties
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
42
IR properties of distant Ks -selected galaxies
on z phot are derived from Monte Carlo simulations and indicate the 68% confidence
intervals. Figure 3.5(b) presents the distribution of ∆z/(1 + z), which is commonly
used to determine the accuracy of photometric redshifts. We find a median ∆z/(1 + z)
of 0.001 and a normalized median absolute deviation (equal to the rms for a Gaussian
distribution) of σ NMAD = 0.053. 5% of the objects with spectroscopic redshift have
|∆z|/(1 + z) > 5σ NMAD and fall outside the plotted range of Figure 3.5(b). Considering
the 95 spectroscopically confirmed sources with a cross-identification within 2.′′ 0 in the
1Ms X-ray catalog by Giacconi et al. (2002), we find a scatter of σ NMAD = 0.064. It
is reassuring that despite the lack of AGN spectrum in our template set, the overall
performance of our photometric redshift code for AGN candidates remains good. We
do note however that, independent of redshift, the fraction of catastrophic outliers
(|∆z|/(1 + z) > 5σ NMAD) is 2.5 times larger for the AGN candidates than for the total
sample of spectroscopically confirmed sources.
3.6 Catalog parameters
Here we describe the entries of our Ks -band selected catalog of the GOODS-CDFS.
The format is similar to the FIRES catalogs of the HDFS (L03) and MS 1054–03 (FS06),
making a straightforward combination of all three fields possible for the user.
• ID– Unique identification number
• x, y– Pixel position of the object, based on the Ks -band detection map. The pixelscale is 0.′′ 15 pixel−1 .
• RA, DEC– Right ascension and declination coordinates for equinox J2000.0.
• [band ] colf– Flux in microjanskies measured within the color aperture (§ 3.4.2.1).
The bandpasses are B,V,i,z,J,H,Ks, [3.6 µm], [4.5 µm], [5.8 µm], and [8.0 µm].
• [band ] colfe– Uncertainty in the [band ] colf flux measurement, derived from the
noise analysis (§ 3.3.5). The units are microjanskies.
• Ks totf– Total Ks -band flux in microjanskies, measured within the total aperture
and scaled by the aperture correction (§ 3.4.2.1). Total fluxes in other bandpasses
can be calculated by [band ] tot f = [band ] col f × (Ks tot f / Ks col f ).
• Ks totfe– Uncertainty associated with Ks totf, also in microjanskies.
• [24 µm] totf– Total MIPS 24 µm-band flux in microjanskies, measured within a
6” diameter circular aperture and then aperture corrected (§ 3.4.2.2).
• [band ]w– Effective weight in the bandpass [band ], normalized to the median effective weight of all sources in that band.
• ap col– Aperture diameter in arcsec within which [band ] colf was measured. In
cases where the color aperture was the isophotal aperture defined by the surface
brightness threshold of µ(Ks, AB) = 24.6 mag arcsec−2 , ap col is the diameter in
arcsec of a circular aperture with equal area.
• ap tot– Aperture diameter in arcsec used for measuring the total Ks -band flux.
When the isophotal or SExtractor’s “MAG AUTO” aperture was used, this entry
contains the equivalent circularized diameter corresponding to that aperture.
• f deblend1– Flag equal to 1 when the source was deblended somewhere in the
process (SExtractor’s “blend”).
• f deblend2– Flag equal to 1 when the photometry is affected by a neighboring
Section 3.7. Comparison to the GOODS-MUSIC catalog
43
source (SExtractor’s “bias”).
• Kr50– Half light radius in arcsec, measured on the Ks -band image (SExtractor’s
flux radius scaled to arcsec).
• Keps– Ellipticity of the isophotal area, measured on the Ks -band image.
• Kposang– Position angle of the isophotal area, measured on the Ks -band image.
• zph best– Best estimate of the photometric redshift (§ 4.6).
• zph low, zph high– Lower and upper edge of the 68% confidence interval around
zph best.
• zsp– Spectroscopic redshift (set to -99 when no spectroscopic information is available).
• zsp qual– Quality flag from 0 to 1 assigned to the spectroscopic redshift. Only
zsp qual=1 entries are considered reliable.
• zsp source– Spectroscopic survey from which zsp was taken (Tab. 3.2).
• zsp qual orig– Original quality flag for zsp from the respective spectroscopic survey.
• XID– Identification number from the 1Ms X-ray catalog by Giacconi et al. (2002),
set to -99 when no cross-identification within 2” was found. Note that we accounted for the ∼ 1.′′ 3 systematic offset in the Giacconi et al. (2002) X-ray centroids, as pointed out before by Roche et al. (2003).
3.7 Comparison to the GOODS-MUSIC catalog
3.7.1 Differences in data and strategy
Recently, Grazian et al. (2006a) presented a multicolor catalog for the GOODS-CDFS
field, referred to as the GOODS-MUSIC catalog. The clustering evolution of distant
red galaxies was quantified based on this sample (Grazian et al. 2006b), as was the
contribution of various color-selected samples of distant galaxies to the stellar mass
density (Grazian et al. 2007). Despite the overlap in public data used to compile the
GOODS-MUSIC and our catalog, there are a number of marked differences.
First, our catalog is purely Ks -band selected. Since the Ks -band magnitude is a
good proxy for stellar mass, this makes it ideally suited to extract mass-limited samples from. The GOODS-MUSIC sample on the other hand is to first order z-band selected (at the ACS resolution), with an addition of the remaining Ks -band sources that
are detected in a map with masked z-band detections. Although valuable in its own
respect, this makes it less trivial to understand the completeness of the sample.
Second, we based our catalog on the ESO/GOODS data release v1.5, consisting of
3 extra ISAAC pointings in J and Ks , and 7 more in H with respect to the v1.0 release
used by Grazian et al. (2006a).
Finally, we include MIPS 24 µm measurements, enabling us to constrain the total IR
luminosity of the Ks -band selected galaxies. Before doing so, we compare the photometry in common between both catalogs, and the photometric redshifts derived from
it.
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
44
IR properties of distant Ks -selected galaxies
Figure 3.6 — A direct comparison of total magnitudes for sources with S/ N > 10 in the B-to-8.0 µm
bandpasses in common between GOODS-MUSIC and our catalog. Sources that are blended in the Ks band image are plotted as empty symbols. On the right side of each panel, a histogram shows the
distribution of offsets. We find an overall good correspondence in the optical and NIR bands, with
offsets of roughly 6% due to aperture corrections. Larger aperture corrections for the IRAC bands lead
to offsets up to 0.4 mag.
Section 3.7. Comparison to the GOODS-MUSIC catalog
45
3.7.2 Comparing photometry
We cross-correlated the two catalogs using a search radius of 1.′′ 2 and in Figure 3.6
present a comparison of the B-to-8.0 µm total magnitudes for objects with S/ N > 10 in
the Ks -band and the band under consideration. Objects that are marked by SExtractor
as blended in the Ks -band, are indicated with empty symbols. The overall correspondence in the B-to-Ks bands is good, and offsets can be well understood from the differences in the applied photometric method. We measure a typical median offset for nonblended sources in the optical and NIR bands of magtot, SW − magtot, MUSIC = −0.06, and
a scatter of σ NMAD < 0.2. Grazian et al. (2006a) based their total magnitudes on SExtractor’s “MAG AUTO” parameter for the z-band detections and on the “MAG BEST”
for the remaining Ks -band detections that were not detected in the z-band. Grazian
et al. (2006a) did not apply an aperture correction based on the stellar growth curve
to correct for the flux lost because it fell outside the “MAG AUTO” or “MAG BEST”
aperture. The lack of aperture correction explains at least part of the systematic offset. Sources marked as blended in our Ks -band detection map typically are brighter by
0.2 - 0.4 mag in the MUSIC catalog. This can be explained by the contamination from
neighboring sources within the “MAG AUTO” aperture, which we avoid by using the
isophotal aperture in combination with an aperture correction for blended sources.
For the IRAC photometry, the discrepancies are larger, ranging from 0.16 mag in
the 4.5 µm band to 0.42 mag at 8.0 µm. Again, aperture corrections (or the lack thereof)
are most likely responsible for the offset. Using simple aperture photometry of isolated
stars, we find that the GOODS-MUSIC IRAC magnitudes account for the light within
an aperture of 2.′′ 0 − 2.′′ 5. From the growth curves of our constructed IRAC PSFs, we
derive that the correction factors needed to account for the light outside this aperture
are consistent with the measured offsets between the GOODS-MUSIC and our IRAC
magnitudes. Grazian et al. (2006a) apparently did not apply this aperture correction.
We stress that, since the aperture correction for the IRAC photometry is considerably
larger than for the optical and NIR bands, this not only affects the estimate of total
magnitudes and its derived properties such as stellar mass, but also the optical-to-MIR
and NIR-to-MIR colors. For example, our z − [3.6 µm], z − [4.5 µm], z − [5.8 µm],
and z − [8.0 µm] colors are redder than the GOODS-MUSIC colors by 0.23, 0.11, 0.17,
and 0.37 mag in the median respectively. Similarly, our Ks − [3.6 µm], Ks − [4.5 µm],
Ks − [5.8 µm], and Ks − [8.0 µm] colors are redder in the median by 0.29, 0.16, 0.22, and
0.42 respectively. The scatter in the color differences with respect to GOODS-MUSIC
typically amounts to 1.5 times the size of the median offset.
3.7.3 Comparing photometric redshifts
Finally, we compare the photometric redshifts presented in §4.6 with those derived
by Grazian et al. (2006a). The numbers quoted in §4.6 and by Grazian et al. (2006a)
cannot directly be compared since new spectroscopic redshifts were added, and objects that showed evidence for the presence of an AGN in their optical spectrum were
rejected from the GOODS-MUSIC photometric redshift analysis. Nevertheless, when
comparing the performance of the z phot estimates for a set of 569 non-AGN with reliable zspec and coverage in all bands in both catalogs, we find a scatter in ∆z/(1 + z)
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
46
IR properties of distant Ks -selected galaxies
that is nearly 1.5 times smaller for GOODS-MUSIC (σ NMAD = 0.037) than for our best
estimates (σ NMAD = 0.054). The median ∆z/(1 + z) is 0.009 and -0.003 for the GOODSMUSIC and our z phot estimates respectively.
Two facts could attribute to the better performance by GOODS-MUSIC. First, an
observed U to rest-frame 5.5 µm wavelength baseline was used by GOODS-MUSIC
to estimate z phot, whereas our estimate was based on B-to-Ks photometry. A second
possible reason, is the difference in template sets. GOODS-MUSIC used PEGASE 2.0
models (Fioc & Rocca-Volmerange 1997), whereas our estimates were based on synthetic models by Bruzual & Charlot (2003). Comparing the photometric redshifts for
all our Ks -band detected objects with a cross-identification within 1.′′ 2 in the GOODSMUSIC sample, we find that ∆z/(1 + z) has a median of 0 and σ NMAD of 0.073.
We conclude that there is an overall reasonable agreement between both catalogs
with differences that can be understood from the applied method. We therefore proceed with strengthened confidence to exploit our catalog to analyze the colors and total
IR energy output of distant galaxies.
3.8 Total IR properties of distant Ks -selected galaxies
With the catalog at hand, we aim to answer the following simple questions: Which
<
Ks -selected (S/ NKs > 5, Kstot
, AB ∼ 24.3) galaxies at 1.5 < z < 2.5 have the brightest total
IR luminosities, and which contribute most to the integrated total IR luminosity? The
answer will be either red or blue galaxies, with the color defined in the rest-frame UV,
optical or NIR wavelength regime. We focus on the 1.5 < z < 2.5 interval, since at
those redshifts the observed 24 µm broadly correlates with the total IR luminosity.
3.8.1 Observed 24 µm flux as function of observed colors
We approach the questions raised above by first studying the correlation between
purely observational properties: the 24 µm flux as proxy for IR luminosity and the
observed B − V, J − Ks , and Ks − [4.5 µm] colors as proxy for the rest-frame UV, optical and optical-to-NIR color respectively. Unless the redshift dependence of the conversion from 24 µm to total IR luminosity and of the conversion from observed to
rest-frame colors are conspiring, any trend in the directly observable properties should
be a signpost for correlations in the rest-frame properties, whose derivation involves
significant systematic uncertainties.
Since a large number of Ks -selected galaxies at 1.5 < z < 2.5 remains undetected
in the 24 µm observations ([24 µm]tot ∼ 16 µJy; 5σ ), we divide our galaxies in bins of
similar color. Each bin contains 80 objects. To start, we leave the origin of the 24 µm
emission (dust heated by AGN or star formation) as an open question. We note however that excluding X-ray detected sources each bin would contain 76 objects, and applying such a selection would not affect the results of our stacking analysis. The mean
and median stacked 24 µm flux densities of the galaxies in each bin are significantly
detected, and plotted in Figure 3.7 versus the observed B − V, J − Ks , and Ks − [4.5 µm]
color. The mean stack counts has a contribution from all the galaxies in the color bin.
Section 3.8. Total IR properties of distant Ks -selected galaxies
47
Figure 3.7 — Stacked 24 µm flux densities as
function of observed B − V, J − Ks , and Ks −
[4.5 µm] color for galaxies at 1.5 < z < 2.5
with S/ N > 5 in the Ks -band (corresponding
<
to Kstot
, AB ∼ 24.3). Filled circles represent the median for each equal-number bin. Empty boxes
represent the mean stacked flux. Light-grey
and dark-grey polygons indicate the central
68% and 50% of the distribution within each
bin. We find a trend of increasing [24 µm]tot
with redder observed-frame color that gets
progressively stronger as we consider colors
measured at longer wavelengths.
The
median stack is lower since it does not capture the brightest sources, but has the advantage that it is more robust against any residual contamination from bright neighbors. The error
√bars on the mean flux measurement indicate the errors in the mean
(σ ([24 µm]tot)/ N), whereas√the error bars on the median flux measurement are computed as σ NMAD ([24 µm]tot)/ N. Furthermore, the light-grey and dark-grey polygons
show the range containing 68% and 50% of the binned galaxies. Each color bin contains galaxies with a large spread in 24 µm fluxes. In most bins, at least 16% and often
more than 25% of the galaxies are individually undetected at 24 µm.
Figure 3.7 shows that the galaxies in the bluest B − V bins are the faintest 24 µm
sources. However, the stacked [24 µm]tot flux is not uniformly increasing over the
whole observed optical color range. Considering colors measured at longer wavelengths, we do find a highly significant increase in the stacked [24 µm]tot flux over the
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
48
IR properties of distant Ks -selected galaxies
entire J − Ks and Ks − [4.5 µm]tot color range. The trend is strongest in the observed
Ks − [4.5 µm]tot color, where we find an increase in [24 µm]tot of factor ∼ 30 over a color
range of ∼ 1.5 mag. Since the bins contain an equal number of objects, it is trivial to see
that not only the reddest galaxies in J − Ks and Ks − [4.5 µm] are brightest at 24 µm,
they also contribute the most to the total 24 µm emission integrated over all distant
Ks -selected galaxies.
3.8.2 Total IR luminosity as function of rest-frame colors
Although the trend of more 24 µm emission for galaxies with a redder observed color
is highly significant for J − Ks and Ks − [4.5 µm], it could still be contaminated or,
alternatively, driven by redshift dependencies within the 1.5 < z < 2.5 redshift interval
under consideration. Now, we will attempt to remove possible redshift dependencies
by converting both axes to a rest-frame equivalent. Moreover, instead of converting
the measured flux density at 24 µm to a rest-frame flux density at 24 µm/(1 + z), we
use it as a probe to determine the total IR luminosity L I R ≡ L(8 − 1000 µm). Since
this conversion assumes that the 24 µm emission originates from dust heated by star
formation, we further reject all X-ray detections from our sample to rule out AGN
candidates.
In the following, we first describe the derivation of rest-frame UV to NIR colors.
Next, we explain the method to estimate the total IR luminosity. Finally, we repeat the
stacking analysis using the derived rest-frame properties.
3.8.2.1 UV slope and rest-frame colors
For each of the galaxies in our sample, we modeled the spectral energy distribution
(SED) using the stellar population synthesis code by Bruzual & Charlot (2003). We
used an identical approach as Wuyts et al. (2007), assuming a Salpeter IMF and solar
metallicity, and fitting three star formation histories: a single stellar population without
dust, an exponentially declining model with e-folding time of 300 Myr and allowed
dust attenuation in the range A V = 0 − 4, and a constant star formation model with
the same freedom in attenuation. We characterize the rest-frame UV part of each SED
by fitting the functional form Fλ ∼ λβ to the best-fitting template, using the rest-frame
UV bins defined by Calzetti, Kinney,& Storchi-Bergmann (1994). The robustness of this
technique is discussed by van Dokkum et al. (2006).
The rest-frame (U − V)rest and (V − J)rest colors were determined by interpolation
between the directly observed bands using templates as a guide. For an in-depth discussion of the algorithm, we refer the reader to Rudnick et al. (2001; 2003).
3.8.2.2 Converting 24 µm flux to total IR luminosity
At redshifts 1.5 < z < 2.5, the 24 µm fluxes trace the rest-frame 7.7 µm emission from
polycyclic aromatic hydrocarbons (PAHs). To convert this MIR emission to a total IR
luminosity L I R ≡ L(8 − 1000 µm), we use the infrared spectral energy distributions of
star-forming galaxies provided by Dale & Helou (2002). The template set allows us to
quantify the IR/MIR flux ratio for different heating levels of the interstellar environment, parameterized by dM(U) ∼ U −α dU where M(U) represents the dust mass heated
Section 3.8. Total IR properties of distant Ks -selected galaxies
49
by an intensity U of the interstellar radiation field.
We computed the total infrared luminosity L I R,α for each object for all Dale & Helou
(2002) templates within the reasonable range from α = 1 for active galaxies to α = 2.5
for quiescent galaxies. Following Geach et al. (2006), the mean of all L I R,α was adopted
as best estimate for the IR luminosity, and the 0.9 dex variation from L I R,α=2.5 to L I R,α=1
was taken as a measure for the systematic uncertainty in the conversion.
Apart from the random photometric error and systematic template uncertainty, uncertainties in the photometric redshift contribute to the total error budget. For each
galaxy, we calculated the spread in L I R caused by variations of z phot within the 68%
confidence interval. Although the uncertainty in photometric redshift is partly random (propagating from photometric uncertainties in the SED), we treat it as purely
systematic, originating from template mismatches. This means
√ the error bars related
to z phot on the stacked L I R measurements do not scale with 1/ N. Instead, they range
from the stacked L I R based on the lowest L I R,individual estimates allowed for each object within its z phot uncertainty, to the stacked L I R based on the maximum L I R,individual
allowed for each object.
3.8.2.3
L I R versus rest-frame color
Having determined the rest-frame colors and total IR luminosities of Ks -selected nonAGN at 1.5 < z < 2.5, we now proceed to investigate which Ks -selected galaxies contribute most to the IR emission. Again, we make use of a stacking procedure to enhance
the robustness of our results. In Figure 3.8, we plot the mean and median stacked
total IR luminosities of our sample divided into color bins versus the rest-frame UV
slope β , the rest-frame optical color (U − V)rest, and the rest-frame optical-to-NIR color
(V − J)rest. The black error bars indicate the errors in the mean and median. With dotted
error bars, we show the systematic variations allowed within the photometric redshift
uncertainties. Finally, the error bar in the bottom right corner represents the range
from quiescent to active galaxy templates by Dale & Helou (2002). Clearly, systematic
uncertainties are dominating the error budget in this analysis.
As in Figure 3.7, we find a large range in IR properties within each bin, illustrated
by the light-grey and dark-grey polygons that mark the central 68% and 50% of the
distribution of L I R of individual objects respectively. Nevertheless, a general trend is
visible of redder colors corresponding to larger IR luminosities. No matter which part
of the spectral energy distribution is used to define red or blue galaxies, the redder
2/3 of the galaxies in our sample have stacked IR properties in the LIRG (L I R = 1011 −
1012 L⊙ ) regime. The trend seems to flatten at the reddest UV slopes, and possibly the
contribution to the total IR luminosity even drops for the reddest (U − V)rest bin. In
(V − J)rest on the other hand, the increase in L I R with reddening color continues over
the entire color range, reaching ULIRG luminosities (L I R > 1012 L⊙ ) for the stack of
the reddest bin. While Luminous (1011 L⊙ < L I R < 1012 L⊙ ) and Ultraluminous (L I R >
1012 L⊙ ) Infrared Galaxies, (U)LIRGS (Sanders & Mirabel 1996), are locally very rare,
they are found to be increasingly more common toward higher redshifts (e.g., Caputi
et al. 2006). Considering the IR luminosities of individual objects in our high-redshift
sample, we find that ULIRGs make up 14% of our Ks -selected sample with Kstot
, AB < 24.3.
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
50
IR properties of distant Ks -selected galaxies
Figure 3.8 — Stacked total IR luminosities as
function of rest-frame UV slope, (U − V)rest,
and (V − J)rest for non-AGN at 1.5 < z < 2.5
<
with S/ N > 5 in the Ks -band (Kstot
, AB ∼ 24.3).
Filled circles represent the median for each
equal-number bin. Empty boxes represent the
mean stacked flux. Light-grey and dark-grey
polygons indicate the central 68% and 50% of
the distribution within each bin. The systematic uncertainty induced by template uncertainties in the conversion to L I R is indicated in
the bottom right corner. The dotted error bars
indicate the variation in the mean and median
stacked L I R by systematic variations in z phot .
We find a trend of increasing L I R with redder
rest-frame color. Since the bins contain equal
numbers of objects, this also means that red
galaxies contribute most to the integrated IR
emission of distant Ks -selected galaxies.
The fraction of ULIRGs increases to 37% when only considering massive (M >
10 M⊙ ) galaxies, with the masses derived from SED modeling as described by Wuyts
et al. (2007). A similar fraction was found by Daddi et al. (2007). Although the IR/MIR
conversion factor varies by nearly an order of magnitude between the use of quiescent
(α = 2.5) or active (α = 1) galaxy templates, we stress that we conservatively use a
mean over all α for each of the galaxies. Using a lower α for brighter galaxies would
only increase the correlation between stacked total IR luminosity and rest-frame color.
We thus conclude that amongst distant Ks -selected galaxies that show no sign of
AGN at X-ray wavelengths, the redder galaxies have on average larger total IR luminosities. Given that each bin in Figure 3.8 contains an equal number of objects, it is also
clear that red galaxies in our sample contribute more to the integrated total IR lumi11
Section 3.8. Total IR properties of distant Ks -selected galaxies
51
nosity than blue galaxies. We argue that this trend cannot be explained by systematic
errors, and tested that this conclusion is robust against the precise choice of redshift
interval by varying the lower edge between redshift 1 and 2 and the upper edge between 2 and 3. Likewise, we verified that none of our conclusions critically depend
on the number of color bins. Dividing the sample in two equal-number bins according to β , (U − V)rest, and (V − J)rest, we find that the integrated total IR luminosity of
the red half is larger than that of the blue half by a factor 5.3, 2.9, and 4.4 respectively.
Imposing a brighter cut in the Ks -band magnitude weakens the correlation of L I R with
(V − J)rest, and causes the stacked L I R of the reddest UV slopes and (U − V)rest colors to
drop. Adopting a Kstot
, AB < 22.86 cut, as is the case for the NIR-selected sample studied
by Reddy et al. (2006), we find that the ratio of the stacked L I R of the red and the blue
half of the galaxies in β , (U − V)rest, and (V − J)rest amounts to a factor of 2.7, 0.9, and
2.3 respectively. Finally, we note that, although X-ray detections were excluded from
our analysis to validate the use of IR templates for starforming galaxies, the stacking
results remain nearly unaffected when we treat them as normal galaxies.
It is tempting to elaborate on the physical interpretation in terms of star formation
rate (SFR), age, and dust content of the galaxies in our sample implied by the presented results. Colors in different wavelength regimes are to a greater or lesser extent
determined by these physical parameters. The UV slope and (V − J)rest color are both
sensitive tracers of dust attenuation (Meurer et al. 1999 and Wuyts et al. 2007 respectively). The (U − V)rest color on its behalf is primarily a tracer of stellar age and to
lesser extent reddened by dust. The fact that in the reddest (U − V)rest bin the total IR
luminosity drops again might therefore indicate an increasing contribution from low
L I R galaxies with little dust-obscured star formation. In combination with the fact that
rest-frame optically selected galaxies often have faint UV luminosities and thus little
unobscured star formation (Förster Schreiber et al. 2004), this suggests that part of the
galaxies making up the reddest (U − V)rest bin have a low overall SFR (unobscured +
obscured). This is consistent with the spectral evidence for galaxies with quenched star
formation at z ∼ 2 (Kriek et al. 2006). A similar conclusion was drawn by Reddy et al.
(2006), who found for a sample of galaxies at similar redshifts selected by optical and
NIR color criteria that the IR luminosity of 24 µm detected sources increased toward
redder observed z − K, but that at the reddest z − K color a population without 24
µm detection exists that satisfies the Distant Red Galaxies (Franx et al. 2003) and/or
BzK/PE (Daddi et al. 2004) color criteria. We note that in samples with a brighter
magnitude limit, e.g., Kstot
, AB < 22.86 as for the NIR-selected galaxies studied by Reddy
et al. (2006), the population without 24 µm detection is more prominently present
among the red galaxies than in our deeper Ks -selected sample, while the overall fraction of galaxies without 24 µm detection is lower by a factor 1.5 in the Kstot
, AB < 22.86
sample. Based on X-ray stacking in the GOODS-North field, although also probing
only to Kstot
, AB < 22.86, Reddy et al. (2005) found a similar turnover in inferred SFR
at z − K > 3. Although the observed z − K color is redshift dependent and at z ∼ 2
spans a somewhat broader wavelength range than (U − V)rest, both colors probe the
Balmer/4000Åbreak and the observed turnover therefore likely share the same origin. The strength of the Balmer/4000Åbreak correlates with age, and is to a slightly
lesser degree also dependent on metallicity and dust (see, e.g., BC03; Worthey 1994;
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
52
IR properties of distant Ks -selected galaxies
MacArthur 2005). An in-depth analysis of the mix between dust-obscured starforming
systems and evolved red galaxies would require a careful SED modeling, estimating
the SFR based on different wavelength tracers from X-ray over UV to the infrared,
and a treatment of each object on an individual basis to assess their relative contribution. Such a study is clearly beyond the scope of this chapter, and will be presented by
Labbé et al. (in preparation) based on the combined sample of galaxies in the CDFS
(this chapter), MS 1054–03 (FS06) and the HDFS (L03).
3.9 Summary
We present a Ks -band selected catalog for the GOODS-CDFS, containing consistent
photometry in the B,V,i,z,J,H,Ks,[3.6 µm], [4.5 µm], [5.8 µm], [8.0 µm], and [24 µm]
bands. Together with the photometry, we release a list of photometric redshifts with
a scatter in ∆z/(1 + z) of 0.053, a cross-correlation with all available spectroscopic redshifts to date, and a cross-correlation with the 1Ms X-ray catalog by Giacconi et al.
(2002). After a description of the catalog construction, we discuss the differences with
the GOODS-MUSIC ’z + Ks ’-selected catalog by Grazian et al. (2006a). Finally, the
catalog, which has been used to estimate stellar mass densities (Rudnick et al. 2006),
construct luminosity functions (Marchesini et al. 2006) and study the predominance
of red galaxies at the high mass end (van Dokkum et al. 2006), is exploited to answer
the following question: Which distant Ks -band selected galaxies are brightest and contribute most to the total IR luminosity?
First, we compared the stacked 24 µm fluxes of galaxies at 1.5 < z < 2.5 with Kstot
, AB <
24.3 split in observed color bins. Overall, a large spread in IR properties is found in
each color bin. Nevertheless, stacking the fluxes within each bin reveals a clear trend
with color. Both in the observed B − V, J − Ks , and Ks − [4.5 µm] colors, the lowest
mean and median [24 µm]tot fluxes are found for the bluest color bin. In J − Ks and
Ks − [4.5µm], the emission at 24 µm continues to rise toward redder colors.
Second, we use our photometric redshifts to convert the observed spectral energy
distributions to rest-frame colors and translate the observed 24 µm flux to the total
IR luminosity L I R ≡ L(8 − 1000 µm). In this procedure, all AGN candidates, selected
by their X-ray detection, were rejected from the sample. Removing the redshift dependence and extrapolating from MIR to total IR goes at the cost of systematic uncertainties. We carefully measured the systematic contribution to the total error budget
from uncertainties in z phot and from our lack of knowledge about which IR template
SED matches best the spectral shape of the objects in our sample. Doing so, we find
a continuous increase in L I R with (V − J)rest. An increasing L I R is also measured with
UV slope β , flattening at the largest β . The rising trend of the stacked L I R luminosity
toward redder (U − V)rest seems to reverse in the reddest color bin. The large range
of total IR properties in this bin suggests a mixture of galaxies with large amounts of
dust emission (LIRGs up to ULIRGs) and objects devoid of it. We note that, if we were
to apply a different translation from MIR to total IR luminosity than simply averaging
over the conversion factors derived from all reasonable templates, the observed trend
would only increase. This is e.g. the case when an SED template corresponding to
a larger heating intensity of the interstellar radiation field is used for objects with a
Section 3.9. Summary
53
larger rest-frame IR luminosity density, as done by Papovich et al. (2006). Since our
stacking analysis divides our Ks -band selected sample in bins containing equal numbers of objects, it is immediately clear that not only do red galaxies have on average the
largest total IR luminosities, it is also true that they form the dominant contribution to
the overall total IR luminosity emitted by Ks -selected galaxies at 1.5 < z < 2.5.
Acknowledgments
This research was supported by grants from the Netherlands Foundation for Research
(NWO), the Leids Kerkhoven-Bosscha Fonds, and the Lorentz Center.
References
Bertin, E.,& Arnouts, S. 1996, A&AS, 117, 393
Bruzual, G.,& Charlot, S. 2003, MNRAS, 344, 1000 (BC03)
Calzetti, D., Kinney, A. L.,& Storchi-Bergmann, T. 1994, ApJ, 429, 582
Calzetti, D., et al. 2000, ApJ, 533, 682
Caputi, K. I., Dole, H., Lagache, G., McLure, R. J., Dunlop, J. S., Puget, J.-L., Le Floc’h, E.,& PérezGonzaléz, P. G. 2006, A&A, 454, 143
Cristiani, S., et al. 2000, A&A, 359, 489
Croom, S. M., Warren, S. J.,& Glazebrook, K. 2001, MNRAS, 328, 150
Daddi, E., Cimatti, A., Renzini, A., Fontana, A., Mignoli, M., Pozzetti, L., Tozzi, P.,& Zamorani, G. 2004,
ApJ, 617, 746
Daddi, E., et al. 2007, submitted to ApJ (astro-ph/07052831)
Dale, D. A.,& Helou, G. 2002, ApJ, 576, 159
Doherty, M., Bunker, A. J., Ellis, R. S.,& McCarthy, P. J. 2005, MNRAS, 361, 525
Fazio, G. G., et al. 2004, ApJS, 154, 10
Fioc, M.,& Rocca-Volmerange, B. 1997, A&A, 326, 950
Förster Schreiber, N. M., et al. 2004, ApJ, 616, 40
Förster Schreiber, N. M., et al. 2006, AJ, 131, 1891
Franx, M., et al. 2003, ApJ, 587, L79
Geach, J. E., et al. 2006, ApJ, 649, 661
Giacconi, R., et al. 2000, A&AS, 197, 9001
Giacconi, R., et al. 2002, ApJS, 139, 369
Giavalisco, M.,& the GOODS Team 2004, ApJ, 600, L93
Grazian, A., et al. 2006a, A&A, 449, 951
Grazian, A., et al. 2006b, A&A, 453, 507
Grazian, A., et al. 2007, A&A, 465, 393
Groenewegen, M. A. T., et al. 2002, A&A, 392, 741
Hopkins, A. M.,& Beacom, J. F. 2006, ApJ, 651, 142
Kriek, M., et al. 2006, ApJ, 645, 44
Kriek, M., et al. 2007, ApJ, astro-ph/0611724
Kriek, M., et al. 2007, submitted to ApJ
Labbé, I., et al. 2003, AJ, 125, 1107
Le Fèvre, O., et al. 2004, A&A, 428, 1043
MacArthur, L. A. 2005, ApJ, 623, 795
Marchesini, D., et al. 2006, 656, 42
Meurer, G., Heckman, T. M.,& Calzetti, D. 1999, ApJ, 521, 64
Mignoli, M., Cimatti, A., Zamorani, G., et al. 2005, A&A, 437, 883
Norman, C., et al. 2002, ApJ, 571, 218
Papovich, C., Dickinson, M., Giavalisco, M., Conselice, C. J.,& Ferguson, H. C. 2005, ApJ, 631, 101
Persson, S. E., Murphy, D. C., Krzeminski, W., Roth, M.,& Rieke, M. J. 1998, AJ, 116, 2475
Chapter 3. B-to-24 µm photometry of the GOODS-CDFS: multi-wavelength catalog and total
54
IR properties of distant Ks -selected galaxies
Ravikumar, C. D., et al. 2007, A&A, 465, 1099
Reddy, N. A., et al. 2005, ApJ, 633, 748
Reddy, N. A., et al. 2006, ApJ, 644, 792
Rieke, G. H., et al. 2004, ApJS, 154, 25
Roche, N. D., Dunlop, J.,& Almaini, O. 2003, MNRAS, 346, 803
Rudnick, G., et al. 2001, AJ, 122, 2205
Rudnick, G., et al. 2003, ApJ, 599, 847
Rudnick, G., et al. 2006, ApJ, 650, 624
Salpeter, E. E. 1955, ApJ, 121, 161
Sanders, D. B.,& Mirabel, I. F. 1996, ARA&A, 34, 749
Smail, I., Ivison, R. J.,& Blain, A. W. 1997, ApJ, 490, L5
Strolger, L.-G., et al. 2004, ApJ, 613, 200
STScI, 2001, VizieR Online Data Catalog, 1271
Szokoly, G. P., et al. 2004, ApJS, 155, 271
Vandame, B. 2002, SPIE, 4847, 123
Vandame, B., et al. 2001, astro-ph/0102300
van der Wel, A., Franx, M., van Dokkum, P. G., Rix, H.-W., Illingworth, G. D.,& Rosati, P. 2005, ApJ, 631,
145
van Dokkum, P. G., et al. 2006, ApJ, 638, 59
Williams, R. E., et al. 1996, AJ, 112, 1335
Worthey, G. 1994, ApJS, 95, 107
Wuyts, S., et al. 2007, ApJ, 655, 51
Chapter 4
Optical spectroscopy of Distant Red
Galaxies
Abstract. We present optical spectroscopic follow-up of a sample of Distant Red
Galaxies (DRGs) with Kstot
, Vega < 22.5, selected by (J − K)Vega > 2.3, in the Hubble
Deep Field South (HDFS), the MS 1054–03 field, and the Chandra Deep Field South
(CDFS). Spectroscopic redshifts were obtained for 15 DRGs. Redshifts were measured for an additional 11 objects satisfying the DRG criterion by other surveys in
the CDFS. Only 2 out of 15 DRGs are located at z < 2, confirming the high efficiency
to select high-redshift sources. We use the sample of spectroscopically confirmed
DRGs to establish the high quality (∆z/(1 + z) ∼ 0.06) of photometric redshifts in
the considered deep fields. Photometric redshifts based on a semi-empirical and
an entirely synthetic template set are discussed. The combination of spectroscopic
and photometric redshifts is used to analyze the distinct intrinsic and observed
properties of DRGs at z < 2 and z > 2. In our photometric sample to Kstot
, Vega < 22.5,
low-redshift DRGs are brighter in Ks than high-redshift DRGs by 0.7 mag, and
more extincted by 2 mag in A V .
S. Wuyts, N. M. Förster Schreiber, M. Franx, G. D. Illingworth,
I. Labbé, G. Rudnick & P. G. van Dokkum
55
Chapter 4. Optical spectroscopy of Distant Red Galaxies
56
4.1 Introduction
S
of the history of star formation and mass assembly in galaxies requires samples of galaxies over a range of lookback times. Since large spectroscopic surveys
of purely magnitude-limited samples (e.g., VVDS, Le Fèvre et al. 2004) become progressively less efficient at probing higher redshifts, a variety of photometric criteria
have been developed to efficiently select distant galaxies. The application of one or
combination of several of these criteria should allow us to construct samples that are
representative for the whole galaxy population at the considered redshift. The Lymanbreak technique (Steidel & Hamilton 1993) was the first to be routineously used, identifying relatively unobscured, actively star-forming galaxies at z ∼ 3 based on their
rest-frame UV colors. Similar criteria were designed to probe star-forming galaxies at
z ∼ 2.3 and z ∼ 1.7, referred to as BX and BM galaxies respectively (Adelberger et al.
2004). Finally, the advent of near-infrared (NIR) instruments on 8-10m class telescopes
encouraged the study of NIR-selected galaxies at high redshift. The NIR flux is less affected by dust obscuration and small amounts of recent star formation and is therefore
a better tracer of stellar mass than the optical fluxes. The two most commonly used
color criteria in the NIR to probe distant galaxies are based on the BzK bands (Daddi et
al. 2004, identifying galaxies at z > 1.4) and J − K color (Franx et al. 2003, designed to
select red galaxies at z > 2). The latter class of galaxies, so-called Distant Red Galaxies
(DRGs), are characterized by the simple color criterion J − K > 2.3. They are found to
<
be massive (M∗ ∼ 1011 M⊙ for Kstot
, Vega ∼ 21.5) systems (van Dokkum et al. 2004; Förster
Schreiber et al. 2004) and range from dusty star-forming to quiescent types (Labbé et
al. 2005; Kriek et al. 2006; Wuyts et al. 2007).
In all of the surveys mentioned above, spectroscopic confirmation is indispensable.
The high-redshift nature of a color-selected population can only be directly verified
by measuring redshifts from their spectra. Apart from establishing the redshift range
probed, the presence of emission and/or absorption lines provides valuable information on the nature of the galaxies. Moreover, having a spectroscopic redshift reduces
the number of free parameters in Spectral Energy Distribution (SED) modeling by one.
Finally, the availability of spectroscopic redshifts allows us to address the quality of
photometric redshift estimates, on which many analyses of the high-redshift galaxy
population rely.
Large samples of optically selected galaxies have been spectroscopically confirmed
and their stellar populations, metallicity and kinematics such as large-scale outflows
have been studied extensively based on the obtained optical and NIR spectra (e.g.,
Steidel et al. 1996; Shapley et al. 2003; Erb et al. 2006). The samples of NIR-selected
distant galaxies with spectroscopic confirmation to date are considerably smaller, the
reason being twofold. First, their faint nature in the rest-frame UV makes optical spectroscopic follow-up challinging. Second, NIR spectroscopic follow-up (e.g., Kriek et
al. 2006) is time-consuming due to the lack of NIR Multi-object spectrographs and the
> 1 µm.
brightness of the night sky at λ ∼
In this chapter, we report on optical spectroscopic follow-up of DRGs, extending
initial results by van Dokkum et al. (2003, hereafter vD03). The sample is defined in
§4.2. In §4.3, we give an overview of the observations, followed by a description of
TUDIES
Section 4.2. Sample selection
57
the data reduction in §4.4. Success rate and bias are discussed in §4.5.1. §4.5.3 presents
the spectroscopic redshift distribution and §4.6 discusses the quality of photometric
redshifts. In §4.7 we consider how the observed broad-band properties of DRGs at z <
2 differ from their high-redshift counterparts. Finally, §7.11 summarizes the chapter.
Vega magnitudes are used throughout this chapter.
4.2 Sample selection
4.2.1 Pure J − K selected sample
During 9 observing runs from February 2002 to November 2003 we obtained optical
spectra for NIR-selected galaxies in the following three fields: HDFS, MS 1054–03, and
CDFS-GOODS. Very deep Js and Ks imaging of the 2.5’x2.5’ HDFS (Labbé et al. 2003)
and the 5’x5’ field around cluster MS 1054–03 (Förster Schreiber et al. 2006) were obtained as part of the FIRES survey (Franx et al. 2000). A Ks -band selected photometric
catalog containing 10’x15’ BVizJHKs imaging of the CDFS-GOODS (Dickinson 2001) is
presented in Chapter 3.
Sources for optical spectroscopy were selected with the simple color criterion J −
K > 2.3 (DRGs) and, with lower priority, galaxies with I − H > 3.0 and J − K < 2.3
were placed in the masks. The masks were usually shared with other high-redshift
candidates and bright fillers. Finally, 11 sources selected by their flux excess in a
narrow-band filter centered at 4190 Å were placed in one of the masks targeting the
MS 1054–03 field. In some rare cases, targets were selected with J − K > 2.3 in an older
catalog, and have J − K < 2.3 in the final catalog. This explains why objects #1195 and
#1458 from vD03 are not part of the DRG sample presented in this chapter.
A total of 64 DRGs was placed in the spectroscopic masks, all of them having Ks,tot <
22.5. Figure 4.1 illustrates their location (large symbols) in a V606 − K versus V606,tot colormagnitude diagram with respect to all DRGs with Ks,tot < 22.5 (small symbols) in the
three fields. The figure demonstrates that the DRGs selected for optical spectroscopic
follow-up span the whole 5 magnitudes in V606 − K color occupied by the total DRG
sample. Furthermore, they exhibit a similar range of V606,tot magnitudes, with a median
V606,tot of 26.3.
4.2.2 DRGs from other surveys
The CDFS-GOODS field is likely the most heavily studied deep field on the sky. Several spectroscopic surveys have been conducted, each with their own selection criteria, resulting in a vast database of spectroscopic redshifts from nearby to the most
distant currently attainable. We cross-correlated our Ks -band selected catalog for the
CDFS field with an up-to-date list of reliable redshifts, most of which were provided
by GOODS-FORS2 (v2.0, Vanzella et al. 2006), the K20 survey (Mignoli et al. 2005),
the VVDS survey (Le Fèvre et al. 2004), and the CXO survey (Szokoly et al. 2004). For
each DRG with a matching object within a (reasonably large) search radius of 1.′′ 2, we
checked both reliability of redshift identification and cross-correlation by eye, resulting
in a list of 11 additional DRGs with spectroscopic confirmation (see Table 4.2).
Since different photometric criteria were applied to select these objects (e.g., an Xray selection for the CXO survey), the spectroscopically confirmed DRGs in the lit-
58
Chapter 4. Optical spectroscopy of Distant Red Galaxies
Figure 4.1 — Sample selection for the
spectroscopic survey of DRGs. The
location of all DRGs with Ks,tot <
22.5 in the HDFS, MS 1054–03, and
CDFS fields is plotted with small circles in the V606 − Ks versus V606,tot colormagnitude diagram. Large circles represent DRGs observed during the spectroscopic campaign described in this
chapter, with filled black symbols indicating the successful redshift determinations. Filled grey circles are DRGs
in the CDFS for which a spectroscopic
redshift is available from the literature.
Lines of constant Ks,tot = 22.5 (the magnitude limit of our sample; solid) and
Ks,tot = 20 (dashed) are plotted to guide
the eye. The sample targeted by our
survey shows a representative range in
V606 − Ks and in V606,tot. The success
rate is biased toward DRGs that are
bright in the Ks -band.
erature are not necessarily representative for the whole population of galaxies with
J − K > 2.3. We therefore decide to mark them throughout the chapter as having spectroscopic redshifts, but treat them as a seperate class, i.e., they are not taken into account to compute the fraction of z < 2 interlopers or to estimate the AGN fraction
based on the optical spectra.
4.3 Observations
A variety of optical spectrographs on 8-10m class telescopes was used to identify redshifts of the optically very faint DRGs: the Low Resolution Imaging Spectrograph
(LRIS, Oke et al. 1995) and DEIMOS (Faber et al. 2003) on the W.M. Keck Telescope,
FORS2 (Nicklas et al. 1997) on VLT and GMOS (Hook et al. 2003) on Gemini South.
An overview of the spectroscopic observations is presented in Table 4.3.
Specifications for the February 2002 run, targeting the MS 1054–03 field with LRIS,
are described by vD03. During the other LRIS runs, the 400 lines mm−1 grism (3400
Å blaze) was used on the blue arm and the 400 lines mm−1 grating (8500 Å blaze)
on the red arm. The D680 dichroic was used in January 2003, whereas in March and
November 2003 the D560 dichroic was inserted. The total exposure time with LRIS,
spread over 2 masks in MS 1054–03 and one in CDFS, amounted to 30.5 ks. Series of
3 or 4 exposures (typically 1800 s each), dithered in 2 ′′ steps along the slit, enabled a
more efficient sky subtraction.
In January 2003, DEIMOS was pointed on MS 1054–03 using a 600 lines mm−1 grism
in conjunction with the gg495 order-blocking filter. The exposure time was 18 ks. Two
other masks, containing a handful of J − K > 2.3 objects as fillers, were exposed for
36.24 ks altogether. For the latter the grism was blazed at 7700 Å and the og550 filter
was inserted. Similar to the LRIS observations, we dithered along the slit.
Section 4.4. Reduction
59
FORS2 observations with the grism GRIS 300V, partly in combination with filter
gg375, took place in September 2002, December 2002, March 2003 and October 2003.
A total of 88.37 ks exposure time was spread over masks in the HDFS, MS 1054–03
and the CDFS. The same dithering strategy as for the LRIS spectroscopy was used. In
September 2003 the GMOS spectrograph on Gemini South was targeted on the HDFS.
In order to allow for smaller slit lengths and consequently a larger number of objects
in the mask, no dithering was applied along the slit. Instead, a 600 lines mm−1 grating
was blazed at 4500 Å during half of the exposures and at 4530 Å during the second
half. For all DRGs we obtained 28.8 ks total exposures. One red galaxy was exposed
for an additional 9.6 ks as a filler in a mask with optically brighter objects. Using the
described instrument settings, we obtained spectra for a total of 64 DRGs. No slits
containing DRGs were lost due to failures in the reduction process or other technical
problems. Exposure times per object varied from a minimum of 7.9 ks to a maximum of
75.34 ks. In the course of the 9 observing runs seeing conditions were highly variable,
ranging from 0.′′ 5 to 2.′′ 0, with a typical value of 1.′′ 0. The 1 to 1.1 ′′ wide slits gave a
typical resolution of 7.5 Å, 3.6 Å, 10.5 Å and 4.6 Å (FWHM) for LRIS, DEIMOS, FORS2,
and GMOS respectively.
4.4 Reduction
Multi-object spectroscopic data obtained by LRIS, DEIMOS, FORS2 and GMOS generally undergo the same reduction steps. For a detailed description of the standard LRIS
reduction process, we refer the reader to van Dokkum & Stanford (2003). Briefly, the
observations were divided in sessions of four dithered exposures. We used standard
IRAF tasks to subtract the bias and apply the flatfielding and fringe correction to each
of the slit exposures. Next, cosmic rays were cleaned and skylines subtracted. The
wavelength calibration was based on arc lamp images, and we used the location of a
bright skyline to apply a zero-point correction. Finally, the 4 reduced slit exposures
were aligned, averaged, and the s-distortion was removed.
The part of the slit where the target object (and possibly a second object) is located,
needs to be masked during several reduction steps. It is of great importance that the
correct part of the slit is masked. As the NIR-selected galaxies are extremely faint in the
optical, it is impossible to measure their positions in the slit on the raw science frames.
We determined the object position in the slit from the mask design and verified the
predicted position for bright filler objects on the raw science frames. The maskwidth
was set to ∼ 1.′′ 9.
In the case of the GMOS run, where no dithering was applied, the use of 2 gratings
blazed at 4500 Å and 4530 Å helped to distinguish hot pixels (at fixed CCD position)
from real spectral features (at fixed wavelength). Nevertheless, the lack of dithering
resulted in a lower quality of the spectra. Ten out of 64 DRGs targeted by our survey
were only placed in the GMOS masks.
60
Chapter 4. Optical spectroscopy of Distant Red Galaxies
4.5 Results from optical spectroscopy of DRGs
4.5.1 Redshift determination, success rate, and bias
Given the faint median V606,tot magnitude of 26.3 for all targeted and 25.8 for all successfully targeted DRGs, it comes as no surprise that continua, if detected, have a too
low signal-to-noise ratio to allow for redshift identifications based on absorption lines.
Therefore, all spectroscopic redshifts for DRGs in our sample are based on emission
lines. In cases where only a single emission line was detected, the presence of a break
(lower continuum on the blue side of the spectral feature) and absence of Hβ and
[OIII]5007 at the expected wavelength if the emission line were [OII]3727 was used to
distinguish Lyα from [OII]3727 as identification.
Out of 64 galaxies satisfying the DRG criterion without further selection bias, the
optical spectroscopic follow-up resulted in 14 redshift identifications (a success rate of
22%). Furthermore, NIR spectroscopy with NIRSPEC (McLean et al. 1998) on the W.
M. Keck Telescope presented by van Dokkum et al. (2004) provided a redshift for one
targeted DRG that did not show emission lines in its optical spectrum. The 15 redshifts
for purely J − K selected DRGs are listed in Table 4.4. Spectroscopic redshifts obtained
for non-DRGs during our spectroscopic campaign are listed in Table 4.1.
We investigate a possible bias of the subsample of DRGs with a successful redshift
determination in Figure 4.1. The 15 spectroscopically confirmed galaxies that were selected purely on the basis of their red (J − K > 2.3) color are plotted with large filled
circles. The other DRGs targeted by our survey are marked with large empty circles.
With smaller circles, we plot all other (small empty circles) DRGs with Ks,tot < 22.5 in the
observed fields and the subsample for which a redshift was obtained by other spectroscopic surveys (small grey circles). The successful targets in our spectroscopic campaign
of DRGs are biased toward brighter magnitudes in both V606 and Ks with respect to
both the whole spectroscopically observed sample and the complete sample of DRGs
in the three considered fields. One could expect a bias toward brighter magnitudes
based on signal-to-noise arguments. However, the possible presence of emission lines
makes the relation between success rate and broad-band flux less direct. A redshift
may be more easily obtained from a faint emission line spectrum than from a brighter
absorption spectrum. We discuss the spectral types in §4.5.2. Remarkably, Figure 4.1
suggests a larger dependence of the success rate on the Ks,tot magnitude than on the
V606,tot magnitude, even though the spectra were obtained in the optical. Out of the 10
(20) brightest targeted DRGs in Ks,tot, a redshift was successfully derived from the optical spectra for 60% (45%) of them. Considering the brightest 10 (20) targets in V606,tot ,
the success rates drop to 50% (25%). As noted before, all redshifts were based on the
presence of emission lines. Although caution should be taken due to small number
statistics and variable seeing conditions between the observing runs, this might hint
toward an increasing prevalence of DRGs with Lyα emission with brighter Ks -band
flux.
4.5.2 Optical spectra
Figure 4.2 presents the 1D spectra of our successful redshift identifications. Since all
spectroscopic redshifts for DRGs in our sample are based on emission lines, we should
Section 4.5. Results from optical spectroscopy of DRGs
61
Figure 4.2 — 1D optical spectra of DRGs observed in our survey with successful redshift identification.
The presented spectra of DRGs at z > 2 show Lyα in emission, possibly in combination with other lines.
Two interlopers at z < 2 were identified by the presence of [OII]3727 in emission, with the continuum
extending blueward of the emission line. Inset for object H-66 is a part of the GMOS 2D spectrum,
showing a smaller feature close to the Lyα emission from the target. Galaxies C-1787 and C-2659 show
evidence of AGN activity in their optical spectra. Interstellar absorption lines are detected in C-5442.
62
Chapter 4. Optical spectroscopy of Distant Red Galaxies
keep in mind that we are likely dealing with a biased representation of the whole population of galaxies with J − K > 2.3. Inverting the success rate, we can place a conservative upper limit of 78% on the fraction of DRGs without emission lines.
Galaxies M-203 and M-508 show [OII]3727 in emission at z < 2. All other spectra
presented in Figure 4.2 feature Lyα in emission, possibly in combination with interstellar absorption lines (C-5442) or confirmed by NV, SiIV, CIV and other emission lines
indicating the presence of an AGN (C-1787, C-2659). The presence of Lyα indicates
that at least a quarter of the DRGs must host regions of star formation that are not
heavily obscured, complementary to an old underlying or dusty young population
that according to SED modeling (e.g., Labbé et al. 2005; Wuyts et al. 2007) is responsible for their red rest-frame optical color. Differences between the rest-frame UV and
rest-frame optical morphologies of DRGs also indicate that these galaxies do not have
homogeneous stellar populations (Toft et al. 2005).
As illustrated by the inset 2D GMOS spectrum of H-66, a smaller feature is visible
near the Lyα emission line of the target, offset from H-66 in the spatial direction by
0.′′ 35 and in the wavelength direction by 13.7 Å. The large dispersion of GMOS allows
for an accurate measurement of the emission line centers: 5330.8 Å (H-66) and 5317.1
Å (serendipitous object). Interpreting both lines as Lyα at identical cosmological distance, the shift in wavelength corresponds to a relative velocity of ∆vr = 771 km s−1 .
At z = 3.385 the projected spatial offset corresponds to 2.6 kpc.
Lyα at 4781 Å was detected in both the LRIS and 2 FORS2 spectra of M-1061. However, the spectrum is offset by 1.′′ 5 from the predicted position in the slit as calculated
from the center of the K-band flux. An identical offset is measured between the centers
of flux on the B- and Ks -band images. Whether the optical and NIR light correspond to
different parts of the same galaxy, or come from physically unrelated sources, remains
uncertain. NIR spectroscopy could confirm the redshift of the DRG unambiguously if
Hα is detected at 2.5811 µm. At z = 2.933 the offset of 1.′′ 5 corresponds to 11.6 kpc. We
verified that our results would not be affected by excluding M-1061 from our spectroscopic redshift sample.
C-1787 was also observed by Norman et al. (2002). These authors find that at z =
3.7, C-1787 is the most distant type-2 QSO known to date, showing a bright X-ray
counterpart in the 1 Ms Chandra imaging of the CDFS. The detection of OVI, Lyα , NV,
SiIV, NIV, CIV, HeII, and CIII in our FORS2 spectrum of the source confirms its nature.
Interpreting a detection of CIV in emission as evidence for an AGN, we find active
nuclei in 13% of the DRGs with spectroscopic redshifts. Under the assumption that all
DRGs without redshift identification lack emission lines in their spectra, the estimated
(unobscured) AGN fraction among the observed DRGs could be as low as ∼ 3%. For
comparison, 4 out of 28 (14%) of our spectroscopically observed DRGs in the CDFS
have a X-ray detection in the 1Ms Chandra exposure on that field (Giacconi et al. 2002).
The X-ray detected fraction among all DRGs with Kstot
, Vega < 22.5 in the CDFS amounts
to 9%. The estimated AGN fraction based on our optical spectroscopy is surprisingly
low compared to the AGN fraction of 20 - 30% implied by recent multi-wavelength
studies by Reddy et al. (2005), Papovich et al. (2006), and Daddi et al. (2007). This
might imply a prevalence of obscured AGN.
Section 4.5. Results from optical spectroscopy of DRGs
63
Figure 4.3 — Spectroscopic redshift
histogram of DRGs in the HDFS,
MS 1054–03, and the CDFS. Redshifts
obtained for purely J − K > 2.3 selected
galaxies are presented in black. Additional spectroscopic redshifts of objects
from other surveys (with their own selection criteria) satisfying J − K > 2.3
are indicated in dark grey. The hatched
and light-grey regions mark the range
in redshifts where [OII]3727 falls redward and Lyα falls blueward of the
sensitive part of the FORS2 and LRIS
detectors respectively.
4.5.3 Redshift distribution
We next discuss the distribution of spectroscopic redshifts obtained for DRGs. Three
questions need to be addressed. How efficient is the DRG selection criterion to isolate
galaxies at z > 2, for which it was designed? What is the typical redshift of DRGs?
And to what range of redshifts are they confined?
The solid histogram in Figure 4.3 shows the redshift distribution of spectroscopically confirmed DRGs from our purely J − K selected sample. The vertical bar (light
grey) at 1.68 < z < 1.88 marks the region in redshift space where spectroscopic confirmation with LRIS is complicated because [OII]3727 lies redward of the covered wavelength range while Lyα has not entered the blue sensitive region of the detector yet.
The corresponding region for the FORS2 spectrograph, whose sensitivity in the blue
reaches down to ∼ 4000 Å, is indicated with the shaded area. Two out of 15 sources
(13%) are located below z = 2, at z = 1.580 and z = 1.189. The median of the purely
J − K selected DRGs lies at z = 2.7 with a distribution ranging to z = 3.7.
Considering the DRGs whose redshifts were obtained as part of other surveys, we
find that all those with a X-ray detection (Szokoly et al. 2004) lie above z = 2. Crosscorrelation with the K20 survey (Ks < 20 selected), the VLT/FORS2 survey (z850 < 25
and i775 − z850 selected) and NIR spectroscopy of Ks -selected galaxies by Kriek et al.
(in preparation) added 6 extra low-redshift (0.6 < z < 1.7) interlopers. Combining
the spectroscopic redshifts from our and other surveys, we find that DRGs at z < 2
have a median Ks -band magnitude that is 1 magnitude brighter than those at z > 2, a
difference at the 10σ level. No significant offset in V606,tot is measured. Our result is
in qualitative agreement with Conselice et al. (2007) who studied a sample of bright
NIR-selected DRGs. Using a combination of photometric redshifts and spectroscopic
redshifts from the DEEP2 survey, the latter reaching to z = 1.4, they conclude that at
64
Chapter 4. Optical spectroscopy of Distant Red Galaxies
the bright end (Kstot
, Vega < 20.5) 64% of all DRGs are located at z < 2. Quadri et al. (2007)
also found that their (photometric) redshift distribution of DRGs shifts toward lower
redshift when imposing a brighter Ks -band cut.
We note that the two low-redshift interlopers from our survey are the faintest in Ks
of all spectroscopically confirmed z < 2 DRGs. The suggested Ks -band dependence of
the success rate to identify redshifts (see §4.5.1) is thus not trivially related to a redshift
dependence of the success rate.
4.6 Photometric redshifts
In order to better address the observed and intrinsic properties, and fraction of lowredshift (z < 2) DRGs, we will complement the spectroscopic sample presented above
with photometric redshift estimates for the remaining DRGs in the HDFS, MS 1054–03,
and the CDFS. We first present the method and templates used to estimate redshifts
from broad-band photometry. Next, we analyse the quality of the photometric redshifts by comparison to the available spectroscopic redshifts. In this chapter, we restrict
ourselves mainly to the quality and distribution of photometric redshifts of DRGs. For
an in-depth discussion of the z phot quality for the whole galaxy population, template
mismatch etc., we refer the reader to Förster Schreiber et al. (in preparation).
4.6.1 Method and template sets
Using the algorithm developed by Rudnick et al. (2001, 2003), updated photometric
redshifts (z phot ) were derived for all Ks -band selected sources in the HDFS, MS 1054–03,
and the CDFS, presented in detail by Förster Schreiber et al. (in preparation). Briefly, a
linear combination of empirical and/or synthetic templates is fit to the spectral energy
distribution of each galaxy. The broad-band photometry used in deriving the photometric redshifts consisted of U300 B450 V606 I814 Js HKs for the HDFS, UBVV606I814 Js HKs
for MS 1054–03, and B435 V606 i775 z850 JHKs for the CDFS. Uncertainties in z phot are estimated from Monte Carlo simulations, accounting for photometric uncertainties and
template mismatch.
We present results obtained with 2 sets of templates, which we refer to as the
CWW++ (used previously by e.g., Rudnick et al. 2006; Quadri et al. 2007; Marchesini
et al. 2007) and the BC03 template set (Förster Schreiber et al. in preparation).
First, the CWW++ template set consists of 8 templates: the empirical E, Sbc, Scd
and Im templates from Coleman, Wu, & Weedman (1980), the two least reddened starburst templates from Kinney et al. (1996) and a 1 Gyr and 10 Myr Bruzual & Charlot
(2003; hereafter BC03) single stellar population (SSP) with a Salpeter (1955) initial mass
function. The BC03 stellar population synthesis code also provided extensions into the
IR for the empirical templates. The empirical templates inherently include some intrinsic reddening, but only a small amount for the Sbc and Scd templates which were
constructed from nearby face-on spirals, and reaching up to E(B − V) ≤ 0.21 for the
SB2 template from Kinney et al. (1996).
The second template set, fed to the same algorithm, consists of synthetic templates
only. Ten SSP templates with ages evenly spaced in log time between 50 Myr and 10
Gyr were selected from the stellar population synthesis code by BC03. Each of the
Section 4.6. Photometric redshifts
65
Figure 4.4 — Direct comparison between photometric and spectroscopic redshifts for all sources with
Ks,tot < 22.5 in the HDFS, MS 1054–03, and CDFS fields for which a reliable spectroscopic redshift is
available. Distant Red Galaxies are highlighted in black. Large symbols denote redshifts obtained during our spectroscopic survey. (a) z phot based on the CWW++ template set. (b) z phot based on the BC03
template set.
templates was allowed to have E(B − V)= 0.0, 0.1, 0.3, or 0.6, applying a Calzetti et al.
(2000) attenuation law. The BC03 template set thus effectively contains 40 templates
and allows for a larger degree of reddening than the CWW+ template set.
4.6.2 Quality of photometric redshifts
We quantify the performance of the photometric redshift code by Rudnick et al. (2003)
by a direct comparison with the available spectroscopic redshifts (see Figure 4.4). DRGs
are marked in red, with large symbols representing objects targeted by our spectroscopic survey. Galaxies with J − K < 2.3 are plotted in black. Their spectroscopic
redshifts are compiled from the literature on the 3 fields, carefully cross-correlating
galaxies from the spectroscopic surveys to objects in the Ks -band selected catalogs and
conservatively limiting ourselves to high quality flags.
Ideally, one algorithm and set of templates provides simultaneously accurate redshift estimates for galaxies of different types and at a range of cosmological distances.
Here, we focus on the z phot quality of DRGs, but place it in context by comparing the
(z
−z
)
spec
phot
distribution of ∆z/(1 + z) = (1
for DRGs to that of the whole population of galax+ zspec )
ies and the subsample at z > 2.
The results for the CWW++ and BC03 z phot estimates are quantified with 3 statistical measures in Table 4.5: the median of ∆z/(1 + z) quantifies systematic offsets, the
normalized median absolute deviation σ NMAD (equal to the rms for a gaussian distribution) is a measure of scatter robust against outliers. The mean absolute deviation
(MAD) is sensitive to catastrophic outliers.
66
Chapter 4. Optical spectroscopy of Distant Red Galaxies
We find a tight correlation between z phot and zspec for the DRGs, characterised by a
0.05 < σ NMAD < 0.07 for the two template sets and without serious catastrophic outliers. The scatter for the DRGs is marginally smaller for the BC03 than for the CWW++
template set. We note that whereas the CWW++ template set systematically overpredicts the z phot of DRGs by 0.03, the BC03 template set underpredicts by the same
amount. It is reassuring that, despite the lack of AGN templates, both template sets
perform equally well for those DRGs with an X-ray detection as for the others. This
might mean that the optical-to-NIR SEDs of these DRGs with an X-ray detection is
dominated by stellar light, and that the AGN is obscured.
Considering all 1090 galaxies with spectroscopic redshifts, 91% (95%) of which lie
below z = 1.5 (2), we find that the BC03 template set removes the systematic underprediction of redshifts below z = 2 that was present in CWW++. Furthermore, the
scatter is reduced to σ NMAD = 0.058, a similar high quality as that for DRGs. At redshifts above 2, we note that the nature of catastrophic outliers is different for the z phot
based on the CWW++ template set than for those based on the BC03 templates. On the
one hand, the BC03 template set reduces the contamination by low-redshift galaxies
mistakenly placed at high redshift. Such catastrophic outliers will lead to artificially
boosted stellar mass and rest-frame luminosity estimates, and may does have a critical impact on studies of the bright end of the high-redshift galaxy population. On the
other hand, a larger number of sources at z > 2 will be placed at z < 0.5, leading to an
underprediction of the number density at high redshift. The MAD values in Table 4.5
reflect this effect.
We conclude that a similar high quality of photometric redshifts is reached for the
spectroscopically confirmed DRGs as for the total galaxy population. However, as
noted earlier, the subsample of DRGs with spectroscopic confirmation is biased toward sources with emission lines. NIR multi-object spectrographs that will come online during the following years will be able to establish the z phot accuracy for the DRG
sample as a whole in a time-efficient manner, targeting either rest-frame optical emission lines (e.g., Kriek et al. 2007) or Balmer/4000 Å breaks in the continuum (Kriek et
al. 2006). We proceed by using the z phot based on the BC03 template set.
4.7 The nature of low-redshift DRGs
Having established confidence in the z phot estimates for DRGs, we can now revisit the
question how efficient the DRG selection criterion is at selecting high-redshift galaxies,
and how the low-redshift DRGs stand out with respect to their high-redshift counterparts. To this purpose, we plot the J − K color of all galaxies with Kstot
, Vega < 22.5 in
the considered fields versus z phot (empty symbols), or zspec (filled symbols) when available (Figure 4.5). The efficiency of the J − K > 2.3 criterion in selecting galaxies above
z = 2 is found to be 68% using the BC03 template set. The efficiency progressively
increases with redder J − K color. Only 9% of the galaxies with J − K > 2.9 was assigned a redshift below z = 2. Less than half of the DRGs at z < 2 have a J − K color
that is consistent at the 1σ level with being photometrically scattered into the DRG
selection window, making it unlikely that all of the low-redshift interlopers are due to
photometric uncertainties.
Section 4.7. The nature of low-redshift DRGs
Figure 4.5 — J − K versus
redshift for all sources with
Ks,tot < 22.5 in the HDFS,
MS 1054–03, and CDFS fields.
Filled symbols are used for
spectroscopic redshifts. For
other sources the photometric redshift estimate based on
the BC03 template set is plotted. Large symbols represent galaxies selected for our
spectroscopic follow-up. Objects above the horizontal line
marking J − K = 2.3 satisfy
the DRG criterion. Selecting galaxies based on their
red J − K color is an efficient
means to find z > 2 galaxies.
Figure 4.6 —
Observed
Ks -band magnitude versus
redshift for all DRGs with
Ks,tot < 22.5 in the HDFS,
MS 1054–03, and CDFS fields.
Filled circles are used for
DRGs with spectroscopic
redshifts. For other DRGs
(empty circles) the photometric redshift estimate based on
the BC03 template set is plotted. Large symbols represent
galaxies in our spectroscopic
survey. Low-redshift DRGs
reach to brighter Ks,tot than
high-redshift DRGs.
67
68
Chapter 4. Optical spectroscopy of Distant Red Galaxies
Figure 4.7 — Top panel: Rest-frame
broad-band SEDs, normalized to the
rest-frame I-band flux, of all lowredshift (z < 2) DRGs to Ks,tot < 22.5
in the HDFS, MS 1054–03, and CDFS
fields. Bottom panel: High-redshift (z >
2) DRGs to the same magnitude limit.
Upper limits indicate the 1σ confidence
levels. Low-redshift DRGs have a red
SED shape from the rest-frame UV
to the rest-frame J-band, whereas the
SEDs of high-redshift DRGs show a
wide range in rest-frame UV slopes and
are on average declining redward of
the rest-frame V-band.
We now proceed to examine the nature of DRGs at z < 2. First, we consider the
observed Ks -band magnitude of DRGs as a function of redshift (Figure 4.6). Apart from
the spectroscopically confirmed redshifts from our (large filled circles) and other (small
filled circles) surveys, we plot the other DRGs (empty circles) in the considered fields
using their photometric redshift estimates. Both the spectroscopic and the photometric
sample of DRGs show a correlation between Ks -band magnitude and redshift. In our
sample to Ks,tot < 22.5, we find a median Ks,tot = 20.5 for z < 2 DRGs, compared to a
median Ks,tot = 21.2 for z > 2 DRGs. Consequently, the fraction of low-redshift (z < 2)
DRGs increases toward brighter Ks -band magnitudes, consistent with Quadri et al.
(2007).
In order to investigate the difference in intrinsic properties between low- and highredshift DRGs, we plot their rest-frame SEDs, normalized to the rest-frame I-band flux,
in Figure 4.7. Although satisfying the same observed color criterion (J − K > 2.3), the
populations at low- and high redshift show a marked difference in rest-frame SED
shapes. The low-redshift DRGs show low flux levels in the UV and a positive slope of
the SED at the rest-frame I-band. The high-redshift DRGs instead show a wide range
in rest-frame UV slopes and have SEDs with a declining slope at the rest-frame I-band
(see also Förster Schreiber et al. 2004).
An interpretation of the difference in rest-frame SED shapes is provided by modeling of the optical-to-MIR SEDs using the Bruzual & Charlot (2003) stellar population
synthesis code following the procedure described by Wuyts et al. (2007), keeping the
redshift fixed to that derived with the BC03 template set. A maximal visual extinction
of A V = 4 magnitudes was allowed during the fit, adopting a Calzetti et al. (2000) attenuation law. Figure 4.8 shows that this artificial upper limit is only reached for DRGs
with zbest, BC03 < 2. Although DRGs at z > 2 with several magnitudes of extinction in
the V-band do exist, a trend of A V with redshift is significant at the 99.9% level, both
Section 4.8. Summary
69
Figure 4.8 — Best-fitted A V versus
redshift (z phot, BC03 or zspec when available) for all DRGs with Ks,tot < 22.5
in the HDFS, MS 1054–03, and CDFS
fields.
Spectroscopic redshifts are
marked with filled symbols. Large
symbols indicate galaxies that were
part of our spectroscopic follow-up of
DRGs. The dust content of DRGs decreases with increasing redshift.
for the total sample and the subsample with spectroscopic redshifts. The median dust
extinction of z < 2 DRGs is A V = 2.8, compared to a median value of A V = 0.8 for the
z > 2 DRGs to the same Ks,tot < 22.5 limit. We note that more than 85% of the DRGs at
z < 2 would also be picked up by the I − H > 3 selection criterion for Extremely Red
Objects (EROs, McCarthy et al. 2001). This fraction drops to about 60% for the DRGs at
higher redshifts. Based on Keck spectroscopy of I − H > 3 selected EROs, Doherty et
al. (2005) inferred a dominant old stellar population for 75% of the ERO sample, being
responsible for their red color. Based on our SED modeling we conclude that, with
the additional constraint of J − K > 2.3, one preferentially selects those EROs whose
large dust content is responsible for the red slope of the SED over a large wavelength
range. The fact that the BC03 template set allows for SED shapes that are more heavily
affected by dust obscuration explains the larger fraction of DRGs placed at z < 2.
4.8 Summary
In this chapter, we presented optical spectroscopic follow-up for a sample of Distant
Red Galaxies with Kstot
, Vega < 25 in the fields HDFS, MS 1054–03, and CDFS. Redshifts
were identified for a total of 15 of the observed DRGs. An additional 11 DRGs, though
not necessarily representative for that population, are spectroscopically confirmed by
other surveys in the CDFS.
Using 8-10m class telescopes under varying seeing conditions, we obtain a modest
success rate of 22% only, increasing toward brighter V606,tot and especially Ks,tot magnitude. Emission line spectra are more easily identified, meaning that the spectroscopic
70
Chapter 4. Optical spectroscopy of Distant Red Galaxies
sample is biased toward those sources with at least some unobscured radiating gas
present. Apart from Lyα , interstellar absorption lines are detected in one and emission
lines typical for AGN activity in two of the high-redshift DRGs. With only 2 objects at
z < 2 in the purely J − K selected sample, we confirm that the DRG criterion J − K > 2.3
is an efficient means to isolate galaxies at z > 2, with their redshift distribution peaking
around z ∼ 2.7.
We use the total sample of 26 spectroscopically confirmed DRGs to address the
quality of the photometric redshift code developed by Rudnick et al. (2001, 2003). We
quantified the deviation between z phot and zspec, ∆z/(1 + z), using two sets of templates.
The semi-empirical CWW++ template set was used for several analyses in the literature (e.g., Rudnick et al. 2006; Marchesini et al. 2007). Furthermore, a new synthetic
template set is presented, based on models from Bruzual & Charlot (2003) and allowing
for a larger impact of dust on the spectral energy distribution (up to E(B − V) = 0.6).
Although both template sets give significantly different results for the galaxy population as a whole, the σ NMAD (∆z/(1 + z)) for the DRGs has an equally small value of
0.05-0.07 (depending on the restriction to z > 2 DRGs or not) for both, similar in quality to what is measured for all 1090 galaxies spanning the entire redshift range with
spectroscopic confirmation in the considered deep fields.
Including DRGs with photometric redshifts, we find that the median of the predicted redshift distribution is z = 2.4, and the efficiency to select galaxies at z > 2
is 68%, for the CWW++ and BC03 template sets respectively. DRGs at redshifts below z = 2 are significantly more extincted by dust than those at higher redshifts. In
observed properties, they are generally characterized by having brighter Ks,tot magnitudes (0.7 mag brighter in the median than z > 2 DRGs to the same Ks,tot < 22.5 limit),
and J − K colors close to J − K = 2.3. SED modeling implies a median dust extinction
for z < 2 DRGs that is as high as A V = 2.8.
Acknowledgments
SW would like to thank the Yale astronomy department for its hospitality during several working visits. This research was supported by grants from the Netherlands Foundation for Research (NWO), the Leids Kerkhoven-Bosscha Fonds, and the Lorentz Center. Based on observations carried out at the European Southern Observatory, Paranal,
Chile. Based on observations obtained at the Gemini Observatory, which is operated
by the Association of Universities for Research in Astronomy, Inc., under a cooperative agreement with the NSF on behalf of the Gemini partnership. Also based on data
obtained at the W. M. Keck Observatory, which is operated as a scientific partnership
among the California Institute of Technology, the University of California and the National Aeronautics and Space Administration. The Observatory was made possible by
the generous financial support of the W. M. Keck Foundation. The authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are
most fortunate to have the opportunity to conduct observations from this mountain.
Section 4.8. Summary
71
Table 4.1. Spectroscopic redshifts for non-DRGs obtained during our spectroscopic
survey
IDa
ra
dec
zspec
Remarkb
H-92
H-228
H-245
H-257
H-290
H-294
H-408
H-470
H-565
H-620
H-657
H-806
HHHHHHM-147
M-161
M-266
M-303
M-383
M-450
M-713
M-897
M-972
M-1132
M-1155
M-1272
M-1396
M-1450
M-1459
M-1637
M-1728
MMMMMMMMMMC-2363
C-2472
C-2484
C-3358
338.22568
338.21679
338.22862
338.21121
338.26335
338.27042
338.24993
338.22038
338.22220
338.23714
338.20360
338.20579
338.25705
338.27145
338.27145
338.28201
338.25686
338.28486
164.23573
164.24502
164.22595
164.21742
164.22318
164.20416
164.24837
164.24914
164.21320
164.27260
164.22757
164.27786
164.24016
164.24319
164.25297
164.23843
164.26288
164.23486
164.21390
164.19865
164.22060
164.23906
164.27251
164.21590
164.22655
164.22023
164.22426
53.082743
53.093660
53.092048
53.178065
-60.569154
-60.561796
-60.561701
-60.557914
-60.558267
-60.558536
-60.551115
-60.554717
-60.544237
-60.536690
-60.531616
-60.540609
-60.590965
-60.577366
-60.579903
-60.587112
-60.59766
-60.57794
-3.6498842
-3.6475178
-3.6422003
-3.6400908
-3.6365197
-3.6339978
-3.6252800
-3.6203344
-3.6176475
-3.6095794
-3.6094061
-3.6050289
-3.6010686
-3.5979289
-3.5974653
-3.5876183
-3.5815978
-3.5825150
-3.5891633
-3.6408465
-3.6178541
-3.5812418
-3.5855079
-3.6068938
-3.6836915
-3.6792324
-3.6761484
-27.831706
-27.826402
-27.827811
-27.792739
2.412
3.295
2.676
2.027
2.025
2.365
1.228
1.284
1.114
1.558
2.793
2.789
0.695
0.439
0.844
0.344
2.899
3.190
1.265
1.859
2.005
2.486
2.123
0.346
1.700
2.973
2.448
1.060
1.622
0.829
2.514
0.622
2.081
1.300
2.93200
2.428
2.436
2.428
2.422
2.280
0.559
0.119
0.261
1.086
0.577
0.246
0.732
0.731
1.427
I814 − H = 1.84; Js − Ks = 1.54
I814 − H = 1.35; Js − Ks = 1.69
I814 − H = 0.97; Js − Ks = 1.24
I814 − H = 2.19; Js − Ks = 1.49
I814 − H = 2.11; Js − Ks = 1.37
I814 − H = 1.64; Js − Ks = 1.78
I814 − H = 1.80; Js − Ks = 1.29
I814 − H = 2.94; Js − Ks = 2.01
I814 − H = 2.39; Js − Ks = 1.81
I814 − H = 1.58; Js − Ks = 1.26
I814 − H = 2.09; Js − Ks = 1.91
I814 − H = 1.20; Js − Ks = 1.15
LBG candidate
LBG candidate
I814 − H = 2.45; Js − Ks = 1.55
I814 − H = 2.21; Js − Ks = 1.87
I814 − H = 1.57; Js − Ks = 1.08
I814 − H = 2.03; Js − Ks = 1.25
I814 − H = 2.30; Js − Ks = 1.62
no I814 coverage; Js − Ks = 1.85
I814 − H = 3.67; Js − Ks = 1.75
I814 − H = 1.13; Js − Ks = 1.31
I814 − H = 2.01; Js − Ks = 1.82
I814 − H = 3.23; Js − Ks = 2.14
I814 − H = 3.59; Js − Ks = 1.90
I814 − H = 1.31; Js − Ks = 1.12
I814 − H = 2.13; Js − Ks = 1.65
I814 − H = 1.23; Js − Ks = 1.08
I814 − H = 3.92; Js − Ks = 2.22
I814 − H = 3.10; Js − Ks = 2.24
I814 − H = 1.63; Js − Ks = 1.42
NB4190
NB4190
NB4190
NB4190
NB4190
NB4190
NB4190
I775 − H = 1.91; J − Ks = 1.15
I775 − H = 3.11; J − Ks = 2.18
I775 − H = 1.33; J − Ks = 0.96
I775 − H = 3.54; J − Ks = 2.17
a H- stands for HDFS, M- for MS 1054–03, and C- for CDFS. Objects without ID
number are either located outside the area covered by the Ks -selected catalog or are
not detected in Ks .
b Objects
with a narrow-band flux excess at 4190 Å are indicated with NB4190.
Chapter 4. Optical spectroscopy of Distant Red Galaxies
72
Table 4.2. Spectroscopic redshifts for DRGs from cross-correlation with other
surveys in the CDFS
ID
ra
dec
zspec
Sourcea
C-1553
C-1957
C-2482
C-2855
C-3129
C-3968
C-4712
C-5177
C-5605
C-5842
C-6132
53.0784636
53.1988252
53.2021505
53.1652224
53.0446457
53.1729054
53.0632815
53.1070458
53.1205657
53.0362490
53.1169241
-27.8598817
-27.8438850
-27.8263119
-27.8140093
-27.8019901
-27.7444701
-27.6996566
-27.7181950
-27.7365600
-27.7522039
-27.7684461
3.660
1.612
1.120
3.064
0.654
1.296
2.402
2.291
3.368
1.294
1.109
CXO
Kriek et al.
VLT/FORS2
CXO
K20
VLT/FORS2
CXO
CXO
MUSYC IMACS
K20
K20
a NIR spectroscopy from Kriek
et al. and optical IMACS spectroscopy by the MUSYC survey from private communication.
Date
Telescope
Instrument
Field
Total exposure time
s
February 2002
Keck
LRIS
MS 1054–03
72000
September 2002
December 2002
January 2003
VLT
VLT
Keck
FORS2
FORS2
LRIS
HDFS
CDFS
MS 1054–03
19800
29700
6800
DEIMOS
MS 1054–03
March 2003
Keck
LRIS
MS 1054–03
18000
36240
14400
March 2003
September 2003
October 2003
VLT
Gemini-South
VLT
FORS2
GMOS
FORS2
November 2003
Keck
LRIS
MS 1054–03
HDFS
CDFS
HDFS
CDFS
14400
38400
24470
16200
9300
Instrument settings
Seeing
“
D680 dichroic
blue: 300 line mm−1
red: 400/8500 Å and 600/1 µm grating
GRIS 300V, filter gg375
GRIS 300V
D680 dichroic
blue: 400/3400 Å grism
red: 400/8500 Å grating
mask1: 600/7300 Å grism, filter gg495
mask2,3: 600/7700 Å grism, filter og550
D560 dichroic
blue: 400/3400 Å grism
red: 400/8500 Å grating
GRIS 300V, filter gg375
B600/4500 Å and B600/4530 Å grating
GRIS 300V
GRIS 300V
D560 dichroic
blue: 400/3400 Å grism
red: 400/8500 Å grating
0.8 - 1.5
Section 4.8. Summary
Table 4.3. Spectroscopic observing runs
0.8 - 2.0
1.0 - 2.3
0.7 - 0.8
0.8 - 1.0
0.7 - 1.4
0.9 - 1.1
0.6 - 0.9
0.9 - 1.4
0.5 - 2.0
0.65 - 1.8
0.7 - 1.5
73
Chapter 4. Optical spectroscopy of Distant Red Galaxies
74
Table 4.4. Spectroscopic redshifts from our spectroscopic follow-up of DRGs
IDa
ra
dec
zspec
Remark
H-66
M-140
M-203
M-508
M-903
M-1061
M-1319
M-1383
M-1734
C-1787
C-2659
C-3119
C-3726
C-5442
C-5900
338.2713649
164.2106125
164.2078833
164.2299500
164.1998917
164.2394875
164.2775375
164.2603167
164.2233917
53.1243363
53.1488159
53.1231066
53.0550864
53.1177728
53.1080817
-60.5703250
-3.6508417
-3.6463678
-3.6315592
-3.6207567
-3.6131875
-3.6010592
-3.6006669
-3.5811008
-27.8516408
-27.8211517
-27.8033550
-27.7785031
-27.7342424
-27.7539822
3.385
2.705
1.580
1.189
2.603
2.933
2.424
2.423
2.699
3.700
2.582
2.349
3.521
3.256
2.728
has close companion at 2.6 kpc
optical and NIR flux offset by 1.′′ 5
redshift from NIR spectroscopy
also analysed by Norman et al. (2002)
-
a H-
stands for HDFS, M- for MS 1054–03, and C- for CDFS.
Table 4.5. Quality of photometric redshifts: statistical measures of ∆z/(1 + z)
CWW++ template set
BC03 template set
Sample
Median
σ NMAD
MAD
Median
σ NMAD
MAD
DRGs
DRGs zspec > 2
All
All zspec > 2
All z phot > 2
0.031
0.014
-0.033
0.021
0.034
0.068
0.051
0.079
0.061
0.070
0.077
0.061
0.122
0.098
0.544
-0.033
-0.033
0.006
-0.052
-0.027
0.056
0.055
0.058
0.076
0.069
0.060
0.046
0.104
0.145
0.392
Section 4.8. Summary
75
References
Adelberger, K. L., Steidel, C. C., Shapley, A. E., Hunt, M. P., Erb, D. K., Reddy, N. A.,& Pettini, M. 2004,
ApJ, 607, 226
Bruzual, G.,& Charlot, S. 2003, MNRAS, 344, 1000 (BC03)
Calzetti, D., et al. 2000, ApJ, 533, 682
Coleman, G. D., Wu, C.-C.,& Weedman, D. W. 1980, ApJS, 43, 393
Conselice, C. J., et al. 2007, ApJ, 660, 55
Daddi, E., Cimatti, A., Renzini, A., Fontana, A., Mignoli, M., Pozzetti, L., Tozzi, P.,& Zamorani, G. 2004,
ApJ, 617, 746
Daddi, E., et al. 2007, astro-ph/07052832
Dickinson, M.,& the GOODS Legacy Team 2001, A&AS, 198, 2501
Doherty, M., Bunker, A. J., Ellis, R. S.,& McCarthy, P. J. 2005, MNRAS, 361, 525
Erb, D. K., Shapley, A. E., Pettini, M., Steidel, C. C., Reddy, N. A.,& Adelberger, K. L. 2006, ApJ, 644, 813
Faber, S. M., et al. 2003, in Iye M., Moorwood, A. F. M., eds, Proc. SPIE, Vol. 4841, Instrument Design
and Performance for Optical/Infrared Ground-Based Telescopes. p. 1657
Förster Schreiber, N. M., et al. 2004, ApJ, 616, 40
Förster Schreiber, N. M., et al. 2006, AJ, 131, 1891
Franx, M., et al. 2000, The Messenger, 99, 20
Franx, M., et al. 2003, ApJ, 587, L79
Giacconi, R., et al. 2002, ApJS, 139, 369
Hook, I., et al. 2003, SPIE, 4841, 1645
Kinney, A. L., et al. 1996, ApJ, 467, 38
Kriek, M., et al. 2006, ApJ, 645, 44
Kriek, M., et al. 2007, ApJ, astro-ph/0611724
Labbé, I., et al. 2003, AJ, 125, 1107
Labbé, I., et al. 2005, ApJ, 624, L81
Le Fèvre, ), O., et al. 2004, A&A, 428, 1043
Marchesini, D., et al. 2007, ApJ, 656, 42
McCarthy, P. J., et al. 2001, ApJ, 560, 131
McLean, I. S., et al. 1998, Proc. SPIE, 3354, 566
Mignoli, M., et al. 2005, A&A, 437, 883
Nicklas, H., Seifert, W., Boehnhardt, H., Kiesewetter-Koebinger, S.& Rupprecht, G. 1997, SPIE, 2871,
1222
Norman, C., et al. 2002, ApJ, 571, 218
Oke, J. B., et al. 1995, PASP, 107, 375
Papovich, C., et al. 2006, ApJ, 640, 29
Quadri, R., et al. 2007, ApJ, 654, 138
Reddy, N. A., Erb, D. K., Steidel, C. C., Shapley, A. E., Adelberger, K. L.,& Pettini, M. 2005, ApJ, 633, 748
Rudnick, G., et al. 2001, AJ, 122, 2205
Rudnick, G., et al. 2003, ApJ, 599, 847
Rudnick, G., et al. 2006, ApJ, 650, 624
Salpeter, E. E. 1955, ApJ, 121, 161
Shapley, A. E., Steidel, C. C., Pettini, M.,& Adelberger, K. L. 2003, ApJ, 588, 65
Steidel, C. C.,& Hamilton, D. 1993, AJ, 105, 2017
Steidel, C. C., Giavalisco, M., Pettini, M., Dickinson, M.,& Adelberger, K. L. 1996, ApJ, 462, L17
Szokoly, G. P., et al. 2004, ApJS, 155, 271
Toft, S., van Dokkum, P. G., Franx, M., Thompson, R. I., Illingworth, G. D., Bouwens, R. J.,& Kriek, M.
2005, ApJ, 624, 9
van Dokkum, P. G., et al. 2003, ApJ, 585, 78
van Dokkum, P. G., et al. 2003, ApJ, 587, L83
van Dokkum, P. G., et al. 2004, ApJ, 611, 703
Vanzella, E., et al. 2006, A&A, 454, 423
Wuyts, S., et al. 2007, ApJ, 655, 51
Chapter 5
What do we learn from IRAC
observations of galaxies at 2 < z < 3.5?
Abstract. We analyze very deep HST, VLT and Spitzer photometry of galaxies at
2 < z < 3.5 in the Hubble Deep Field South. The sample is selected from the deepest public K-band imaging currently available. We show that the rest-frame U − V
vs V − J color-color diagram is a powerful diagnostic of the stellar populations of
distant galaxies. Galaxies with red rest-frame U − V colors are generally red in
rest-frame V − J as well. However, at a given U − V color a range in V − J colors exists, and we show that this allows us to distinguish young, dusty galaxies
from old, passively evolving galaxies. We quantify the effects of IRAC photometry
on estimates of masses, ages, and the dust content of z > 2 galaxies. The estimated distributions of these properties do not change significantly when adding
IRAC data to the UBV I JHK photometry. However, for individual galaxies the addition of IRAC can improve the constraints on the stellar populations, especially
for red galaxies: uncertainties in stellar mass decrease by a factor of 2.7 for red
[(U − V)rest > 1] galaxies, but only by a factor of 1.3 for blue [(U − V)rest < 1] galaxies. We find a similar color-dependence of the improvement for estimates of age
and dust extinction. In addition, the improvement from adding IRAC depends on
the availability of full NIR JHK coverage; if only K-band were available, the mass
uncertainties of blue galaxies would decrease by a more substantial factor 1.9. Finally, we find that a trend of galaxy color with stellar mass is already present at
z > 2. The most massive galaxies at high redshift have red rest-frame U − V colors
compared to lower mass galaxies even when allowing for complex star formation
histories.
S. Wuyts, I. Labbé, M. Franx, G. Rudnick, P. G. van Dokkum, G. G. Fazio,
N. M. Förster Schreiber, J. Huang, A. F. M. Moorwood, H.-W. Rix,
H. Röttgering & P. van der Werf
The Astrophysical Journal, 655, 51 (2007)
77
78
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
5.1 Introduction
T
WO of the major challenges in observational cosmology are understanding the history of star formation in galaxies, and understanding the assembly of mass through
cosmic time. In the local universe elaborate surveys mapped the diversity of nearby
galaxies (e.g., Blanton et al. 2003) and characterized the dependence of their colors
(Baldry et al. 2004) and star formation (Kauffmann et al. 2003) on galaxy mass. The
study of their progenitors at z> 2 is important since it is believed that at this epoch the
∼
most massive galaxies formed their stars (Glazebrook et al. 2004; van der Wel et al.
2005; Rudnick et al. 2006).
The first method to efficiently identify distant galaxies was the Lyman-break technique (Steidel et al. 1996). Large samples have been spectroscopically confirmed (Steidel et al. 1999). Their stellar populations have been characterized by means of broadband photometry (e.g., Papovich, Dickinson, & Ferguson 2001, Shapley et al. 2005),
optical spectroscopy (e.g., Shapley et al. 2003) and near-infrared (NIR) spectroscopy
(Erb et al. 2003, 2006). Lyman break galaxies (LBGs) have spectral energy distributions
similar to nearby starburst galaxies.
In recent years, new selection criteria provided evidence for a variety in color space
among high-redshift galaxies as rich as in the local universe. Among the newly discovered populations are submm galaxies (e.g., Smail et al. 2004), “IRAC Extremely Red
Objects” (IEROs; Yan et al. 2004), “BzK” objects (Daddi et al. 2004) and distant red
galaxies (DRGs; Franx et al. 2003). The latter are selected by the simple color criterion (J − K)Vega > 2.3. Their rest-frame UV-to-optical SEDs resemble those of normal
nearby galaxies of type Sbc-Scd (Förster Schreiber et al. 2004). NIR spectroscopy of
DRGs (Kriek et al. 2006) and extension of the broad-band photometry to mid-infrared
wavelengths (Labbé et al. 2005) suggests that evolved stellar populations exist already
at 2 < z < 3.5. Rudnick et al. (2006) showed that DRGs contribute significantly to the
mass density in rest-frame optically luminous galaxies. van Dokkum et al. (2006) studied a stellar mass-limited sample of galaxies with M > 1011 M⊙ and found that DRGs,
rather than LBGs, are the dominant population at the high mass end at 2 < z < 3.
In this chapter, we exploit the 3-8 µm imaging of the Hubble Deep Field South by
Spitzer’s Infrared Array Camera (IRAC; Fazio et al. 2004) to extend the SED analysis
of distant galaxies to the rest-frame NIR and constrain their stellar masses and stellar
populations. Our sample is complete to Ktot, AB = 25. No color selection criteria are
applied. The depth of our imaging allows us to probe down to stellar masses of a few
109 M⊙ . We investigate whether IRAC helps to study the diversity of galaxies at high
redshift and if the addition of IRAC improves the constraints on stellar mass, age and
dust content. Finally, we investigate the dependence of galaxy color on stellar mass.
The chapter is structured as follows. In §5.2 we describe the data, IRAC photometry and sample definition. §5.3 explains the modeling of spectral energy distributions
(SEDs). The rest-frame optical to NIR color distribution of our K-selected sample is discussed in §5.4. §5.5 provides an in-depth discussion of the constraints that IRAC places
on estimates of age, dust extinction and stellar mass. First wavelength and model dependence are discussed from a theoretical perspective. Next we discuss results from
applying the models to our U-to-8 µm spectral energy distributions. In §5.6 we investi-
Section 5.2. Data, photometry and sample selection
79
Table 5.1. Characteristics of the IRAC observations
Filter
(µ)
Exposure time
(hr)
FWHM
(′′ )
Limiting depth
(5σ , 3′′ diameter aperture)
Positional Accuracya
(′′ )
3.6
4.5
5.8
8.0
3.76
3.76
3.76
3.64
1.95
1.90
2.10
2.15
25.6
25.6
23.4
23.3
0.09
0.15
0.14
0.11
a The
rms difference between bright star positions in IRAC and K-band image.
Table 5.2. Characteristics of the optical-to-NIR observations (see L03)
Instrument/Telescope
Filter
Exposure time
(hr)
FWHM
(′′ )
Limiting depth
(5σ , 0.′′ 7 diameter aperture)
WFPC2/HST
WFPC2/HST
WFPC2/HST
WFPC2/HST
ISAAC/VLT
ISAAC/VLT
ISAAC/VLT
F300W
F450W
F606W
F814W
Js
H
Ks
36.8
28.3
27.0
31.2
33.6
32.3
35.6
0.16
0.14
0.13
0.14
0.45
0.48
0.46
27.8
28.6
28.9
28.3
26.9
26.4
26.4
gate the rest-frame optical colors of high-redshift galaxies as a function of stellar mass.
Finally, the conclusions are summarized in §5.7.
Throughout this chapter we adopt a cosmology with H0 = 70 km s−1 Mpc−1 , Ωm =
0.3, and ΩΛ = 0.7.
5.2 Data, photometry and sample selection
5.2.1 Data
Observations of the HDFS/WFPC2 field were obtained with the IRAC camera (Fazio
et al. 2004) on the Spitzer Space Telescope (Werner et al. 2004) in June 2004 and June 2005
(GTO program 214). A 5′ × 5′ field of view was covered by the 4 broadband filters at
3.6, 4.5, 5.8 and 8 microns. The data, reduction and photometry will be described in detail by I. Labbé et al. (in preparation). Briefly, we started with the Basic Calibrated Data
(BCD) as provided by the Spitzer Science Center pipeline. We applied a series of procedures to reject cosmic rays and remove artifacts such as column pulldown, muxbleed,
and the “first frame effect” (Hora et al. 2004). Finally, the frames were registered to and
projected on a 2x2 blocked (0.′′ 2396 pixel scale) version of an existing ISAAC K-band
image (Labbé et al. 2003, hereafter L03)1 , and average-combined. Characteristics such
as exposure time, FWHM, limiting depth (5σ , 3” diameter aperture) and positional ac1
NIR data from the FIRES
http://www.strw.leidenuniv.nl/˜fires
survey
of
the
HDFS
is
publicly
available
from
80
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
Figure 5.1 — Postage stamps (9.′′ 8 ×
9.′′ 8) illustrating the deblending procedure for IRAC photometry. Confusion by nearby neighbors in the original
3.6 µm image (a) is reduced using the
higher resolution K-band image (b) and
its SExtractor segmentation map (c). A
model 3.6 µm image (d) is created using information on position and extent
of the galaxies from the K-band image.
The model of the nearby neighbors (e)
is subtracted from the original image to
obtain a cleaned 3.6 µm image (f).
curacy in each of the 4 IRAC bands are summarized in Table 5.1. A summary of the
optical-to-NIR observations by L03 is provided in Table 5.2. All magnitudes quoted in
this chapter are in the AB system.
5.2.2 Photometry
In this section we describe the steps to combine the IRAC data and optical-to-NIR data
(L03) into one consistent K-band selected photometric catalog. In this chapter we limit
ourselves to the 2.5′ × 2.5′ field where very deep K-band data is available from L03. The
main challenge in doing IRAC photometry is a proper treatment of source confusion
and PSF matching of the data. Integrating for nearly 4 hours with IRAC at 3.6 µm and
4.5 µm reaches a depth only 1 mag shallower than 36 hours of ISAAC K-band imaging
(10σ limit Ktot, AB = 25), but the IRAC images have a 4 times broader PSF causing many
sources to be blended. Information on the position and extent of K-band detected
objects was used to fit and subtract the fluxes of neighbouring sources. Each K-band
detected source was isolated using the SExtractor “segmentation map” and convolved
individually to the considered IRAC PSF. Next, all convolved sources were fitted to the
IRAC image, leaving only their fluxes as free parameters. We subsequently subtract the
best-fit fluxes of all neighboring sources to remove the contamination. An illustration
of this measurement technique is presented in Fig 5.1. The resulting cleaned IRAC
images are matched to the broadest PSF (of the 8 µm image). We measured fluxes on
the cleaned, PSF-matched images within a fixed 4.′′ 4 diameter circular aperture. The
aperture size is a compromise between quality of PSF matching (within 3% as derived
from dividing growthcurves) and adding too much noise. Finally, we applied for each
source an aperture correction to scale the IRAC fluxes to the “color” apertures defined
for the K-band catalog by L03. The correction factor is the ratio of the original K-band
flux in the color aperture and the K-band flux in the 8 µm PSF matched image within a
4.′′ 4 diameter aperture. Photometric errors were calculated by taking the rms of fluxes
in 4.′′ 4 diameter apertures on empty places in the IRAC image. The end product is a
photometric catalog with consistent photometry from optical to MIR wavelengths with
11 filters (UBV I JHK+IRAC).
Section 5.3. SED modeling
81
5.2.3 Sample selection
From the catalog described in §5.2.2 we selected all galaxies, well covered by all 11
filters, that have S/ N > 10 in the K-band. The sample reaches to a limiting total Kband magnitude Ktot, AB = 25.
Since spectroscopic redshifts are only available for 63 out of 274 objects, we mostly
rely on photometric redshift estimates to select high-redshift galaxies and compute
rest-frame colors and luminosities. The photometric redshifts and derived rest-frame
photometry were calculated as follows. We used an algorithm developed by Rudnick
et al. (2001, 2003) to fit a nonnegative linear combination of galaxy templates to the
spectral energy distribution of each galaxy. The template set consisted of empirical
E, Sbc, Scd and Im templates from Coleman, Wu,& Weedman (1980), the two least
reddened starburst templates from Kinney et al. (1996) and two Bruzual & Charlot
(2003; hereafter BC03) single stellar populations (SSP) with a Salpeter (1955) stellar initial mass function (IMF), aged 1 Gyr and 10 Myr respectively. The empirical templates
were extended into the IR using the BC03 stellar population synthesis code. The derived photometric redshifts show a good agreement with the available spectroscopic
redshifts. The average value of | zspec − z phot |/(1 + zspec) is 0.06, 0.09 and 0.08 for galaxies
at 0 < z < 1, 1 < z < 2 and 2 < z < 3.5 respectively.
Once the redshift was derived, we calculated rest-frame luminosities and colors by
interpolating between observed bands using the best-fit templates as a guide. For a
detailed description, we refer the reader to Rudnick et al. (2003).
The K-band selected sample contains 121 sources at 0 < z < 1, 72 at 1 < z < 2 and
75 at 2 < z < 3.5. The K+IRAC photometry of the galaxies at 2 < z < 3.5 is provided
in Table 5.3. In §5.4 we study the color-distribution of galaxies with LV > 5 × 109 L⊙
over the whole redshift range. From that point on we focus on the high-redshift bin.
Two commonly color-selected populations at z > 2 are highlighted where they are of
interest. LBGs are selected from the WFPC2 imaging using the criteria of Madau et al.
(1996). DRGs are selected by the simple color criterion (J − K) AB > 1.34 (Franx et al.
2003).
5.3 SED modeling
To study physical characteristics of the galaxies such as stellar mass, stellar age and
amount of dust extinction, we make use of the evolutionary synthesis code developed
by BC03. We fitted the synthetic spectra to our observed SEDs using the publicly available HYPERZ stellar population fitting code, version 1.1 (Bolzonella et al. 2000). Redshifts were fixed to the z phot measurement (see §5.2.3, Rudnick et al. 2003) or zspec when
available. A minimum error of 0.08 mag was adopted to avoid the problem of data
points with the largest errors being effectively ignored in the SED fits. We fitted three
distinct star formation histories: a single stellar population (SSP) without dust, a constant star formation (CSF) history with dust (A V varying from 0 to 4 in steps of 0.2)
and an exponentially declining star formation history with an e-folding timescale of
300 Myr (τ300 ) and identical range of A V values. The exponentially declining model
allows for quiescent systems that underwent a period of enhanced star formation in
their past.
82
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
Table 5.3. K+IRAC photometry of HDFS galaxies at 2 < z < 3.5
Objecta
62
66
92
96
114
116
130
133
143
158
f K,tot b
f K,col
f 3.6µm,col
f4.5µm,col
f 5.8µm,col
f8.0µm,col
1.90 ± 0.33
6.03 ± 0.58
3.17 ± 0.13
4.55 ± 0.62
1.44 ± 0.25
1.72 ± 0.27
1.83 ± 0.26
2.44 ± 0.27
4.47 ± 0.15
1.42 ± 0.23
1.30 ± 0.10
4.96 ± 0.22
2.11 ± 0.08
3.45 ± 0.19
0.99 ± 0.09
1.26 ± 0.09
1.71 ± 0.10
1.94 ± 0.10
3.56 ± 0.12
0.79 ± 0.07
1.85 ± 0.19
8.82 ± 0.22
2.29 ± 0.19
4.45 ± 0.21
1.06 ± 0.18
1.75 ± 0.20
2.03 ± 0.20
3.10 ± 0.20
6.16 ± 0.21
0.81 ± 0.16
1.57 ± 0.21
11.15 ± 0.23
2.31 ± 0.20
4.33 ± 0.23
1.25 ± 0.20
2.26 ± 0.21
1.53 ± 0.22
2.64 ± 0.22
6.12 ± 0.23
1.00 ± 0.18
−2.36 ± 1.46
14.95 ± 1.63
0.45 ± 1.40
2.46 ± 1.62
−0.09 ± 1.38
1.77 ± 1.49
4.50 ± 1.51
3.19 ± 1.52
4.75 ± 1.60
−0.89 ± 1.24
−3.48 ± 1.73
20.31 ± 1.93
−1.85 ± 1.66
4.72 ± 1.93
−4.15 ± 1.64
0.52 ± 1.76
2.98 ± 1.79
0.53 ± 1.81
1.18 ± 1.90
1.93 ± 1.47
Note.– Table 5.3 is published in its entirety in the electronic edition of the Astrophysical Journal.
A portion is shown here for guidance regarding its form and content.
a Object
identification number corresponds to that of the U-to-K catalog by Labbé et al.
(2003).
b Fluxes
in total (tot) and color (col) aperture are scaled to an AB zero point of 25, i.e.,
mag AB = 25 − 2.5 log f .
Table 5.4. Modeling results for HDFS galaxies at 2 < z < 3.5
Objecta
62
66
92
96
114
116
130
133
143
158
z
SFH
log(M∗ )
(M⊙ )
AV
log(Agew )
(Gyr)
+0.04
2.72−
0.04
+0.00
3.38−
0.00
+0.28
2.66−
0.08
+0.08
2.06−
0.02
+0.16
2.98−0.38
+0.14
3.14−
0.10
+0.04
2.16−0.12
+0.02
2.04−
0.28
+0.04
2.16−0.12
+0.14
2.08−
0.18
CSF
τ300
SSP
CSF
τ300
CSF
SSP
τ300
τ300
CSF
+0.15
9.69−
0.05
+0.14
11.04−
0.00
+0.26
9.75−
0.00
+0.01
10.02−
0.12
+0.11
9.75−
0.13
+0.02
10.18−
0.14
+0.22
9.44−
0.01
+0.12
9.68−
0.12
+0.07
10.12−
0.06
+0.14
9.55−
0.09
+0 . 0
0.4−
0.2
+0 . 0
1.6−
0.6
+0 . 4
0.0−
0.0
+0 . 0
0.2−
0.0
+0 . 2
0.2−0.2
+0 . 0
0.4−
0.2
+0 . 4
0.0−0.0
+0 . 2
0.6−
0.2
+0 . 0
0.8−0.2
+0 . 0
0.2−
0.2
+0.45
−0.75−
0.05
+0.37
−0.50−
0.00
+0.54
−1.09−
0.00
+0.00
−0.30−
0.31
+0.12
−0.50−0.12
+0.00
−0.10−
0.39
+0.47
−0.84−0.00
+0.43
−0.73−
0.07
+0.18
−0.67−0.06
+0.33
−0.20−
0.35
Note.– Table 5.4 is published in its entirety in the electronic edition of the Astrophysical Journal.
A portion is shown here for guidance regarding its form and content.
a Object
identification number corresponds to that of the Uto-K catalog by Labbé et al. (2003).
Section 5.3. SED modeling
83
Figure 5.2 — The U-to-8 µm spectral energy distributions of a subset of galaxies occupying different
locations in (U − V)rest vs (V − J)rest color-color space. Each row shows observed and BC03 model SEDs
for galaxies with redshifts ranging from z ∼ 0.7 to z ∼ 3. A broad range of galaxy types is present at
all redshifts. Galaxies with blue (U − V)rest colors (top row) have young ages and a modest amount of
dust obscuration. Objects with red (U − V)rest colors that are on the blue side of the (V − J)rest color
distribution (middle row) are best fit by old stellar populations with little dust obscuration. The bottom
row shows examples of galaxies with red optical and red optical-to-NIR colors. They are consistent with
young stellar populations with a large dust reddening.
Förster Schreiber et al. (2004) showed that the estimated extinction values do not
vary monotonically with the e-folding timescale τ , but reach a minimum around 300
Myr. Including the τ300 model thus ensures that the allowed star formation histories
encompass the whole region of parameter space that would be occupied when fitting
models with different values of τ . For each of the star formation histories (SFHs),
we constrained the time elapsed since the onset of star formation to a minimum of
50 Myr, avoiding fit results with improbable young ages. The age of the universe at
the observed redshift was set as an upper limit to the ages. Furthermore, we assume
a Salpeter (1955) IMF with lower and upper mass cut-offs 0.1M⊙ and 100M⊙ , solar
metallicity and we adopt a Calzetti et al. (2000) extinction law. For each object the star
formation history resulting in the lowest χ2 of the fit was selected and corresponding
model quantities such as age, mass and dust extinction were adopted as the best-fit
value. We calculated the mass-weighted age for each galaxy by integrating over the
different ages of SSPs that build up the SFH, weighting with their mass fraction. We
use this measure since it is more robust with respect to degeneracies in SFH than the
time passed since the onset of star formation; it describes the age of the bulk of the stars.
See Table 5.4 for a summary of the results of our SED modeling for the subsample of
84
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
Figure 5.3 — Rest-Frame U − V versus
V − J color-color diagram of all galaxies with L V > 5 × 109 L⊙ . SDSS+2MASS
galaxies (small grey dots) are plotted as a
local reference. Greyscale coding refers
to the redshift bin. Galaxies with red
U − V colors are also red in V − J. Compared to the local SDSS galaxies the
high-redshift color distribution extends
to bluer U − V colors (where Lymanbreak galaxies are located) and for the
same U − V color to redder V − J colors.
galaxies at 2 < z < 3.5. In Figure 5.2 we show example U-to-8 µm SEDs with best-fit
BC03 models of galaxies over the whole redshift range, illustrating that at all epochs a
large variety of galaxy types is present.
We fitted all objects in our sample twice, once with and once without IRAC photometry. We repeated the SED modeling with the same parameter settings using the
models by Maraston (2005; hereafter M05). The results are discussed in §5.5.2.2. Variations in modeled parameters due to a different metallicity are addressed in §5.5.2.3.
The effects of adopting a different extinction law are discussed in §5.5.2.4. Unless noted
otherwise, we refer to stellar mass, mass-weighted age and dust extinction values derived from the U-to-8 µm SEDs with BC03 models.
5.4 Rest-frame optical to near-infrared color distribution
At redshifts above 1 all rest-frame NIR bands have shifted redward of observed K, and
mid-infrared photometry is needed to compute rest-frame NIR fluxes from interpolation between observed bands. It has only been with the advent of IRAC on the Spitzer
Space Telescope that the rest-frame NIR opened up for the study of high-redshift galaxies. As the 3.6 µm and 4.5 µm images are much deeper than the 5.8 µm and 8.0 µm
images (see Table 5.1), we focus on the rest-frame J band (Jrest).
Several studies have focussed on the optical to NIR colors and inferred stellar populations of particular color-selected samples (e.g., Shapley et al. 2005, Labbé et al. 2005).
In this section we take advantage of the multiwavelength data and the very deep Kband selection to study the rest-frame optical to NIR colors of all galaxies up to z = 3.5
without color bias. For the first time we can therefore investigate what range in optical
to NIR colors high-redshift galaxies occupy, how their optical to NIR colors relate to
pure optical colors, and what this tells us about the nature of their stellar populations.
In Figure 5.3 we present a color-color diagram of (U − V)rest versus (V − J)rest for the
Section 5.4. Rest-frame optical to near-infrared color distribution
85
redshift bins 0 < z < 1, 1 < z < 2 and 2 < z < 3.5. A clear correlation of (U − V)rest with
(V − J)rest is observed at all redshifts. The (U − V)rest color samples the Balmer/4000 Å
break. The large wavelength range spanned by (U − V)rest and (V − J)rest together is
useful to probe reddening by dust.
To study how the color distribution compares to that in the local universe, we indicate the colors of galaxies in the low-redshift New York University Value-Added
Galaxy Catalog (NYU VAGC; Blanton et al. 2005) with small grey dots. The low-z
NYU VAGC is a sample of nearly 50000 galaxies at 0.0033 < z < 0.05 extracted from
the Sloan Digital Sky Survey (SDSS data release 4; Adelman-McCarthy et al. 2006).
It is designed to serve as a reliable reference for the local galaxy population and contains matches to the Two Micron All Sky Survey Point Source Catalog and Extended
Source Catalog (2MASS; Cutri et al. 2000). Only the subsample of 20180 sources that
are detected in the 2MASS J-band are plotted in Figure 5.3. This results effectively
in a reduction of the blue peak of the bimodal U − V distribution. We only show
those galaxies (both for the local sample and for our sample of HDFS galaxies) with a
rest-frame V-band luminosity LV > 5 × 109 L⊙ . At this luminosity the distribution of
low-z NYU VAGC galaxies with SDSS and 2MASS detections starts falling off. From
the much deeper HDFS imaging the luminosity cut weeds out low- to intermediateredshift dwarf galaxies.
The same trend of optically red galaxies being red in optical to NIR wavelengths
that we found for galaxies up to z = 3.5 is observed in the local universe. However,
there are two notable differences in the color distribution between distant and local
galaxies. First, a population of luminous high-redshift galaxies with very blue (U −
V)rest and (V − J)rest exists without an abundant counterpart in the local universe. The
2MASS observations are not deep enough to probe very blue V − J colors, but we can
ascertain that 95% of all low-z NYU VAGC sources with LV > 5 × 109 L⊙ lie in the range
0.73 < U − V < 2.24. About half of the blue galaxies at z > 2 with (U − V)rest < 0.73 and
LV > 5 × 109 L⊙ satisfy the Lyman-break criterion. Their stellar populations have been
extensively studied (e.g., Papovich et al. 2001; Shapley et al. 2001; among many others)
and their blue SEDs (see e.g., object #242 and #807 in Figure 5.2) are found to be well
described by relatively unobscured star formation. The rest-frame optical bluing with
increasing redshift of galaxies down to a fixed LV is thoroughly discussed by Rudnick
et al. (2003).
A second notable difference with respect to the color distribution of nearby galaxies
is present at (U − V)rest > 1, where most local galaxies reside. Our sample of HDFS
galaxies has a median offset with respect to the SDSS+2MASS galaxies of 0.22 ± 0.04
mag toward redder (V − J)rest at a given (U − V)rest. Furthermore, the spread in (V −
J)rest is larger, extending from colors similar to that of local galaxies to (V − J)rest colors
up to a magnitude redder. The larger spread in (V − J)rest colors at a given (U − V)rest is
not caused by photometric uncertainties. After subtraction in quadrature of the scatter
expected from measurement errors (0.05 mag) we obtain an intrinsic scatter of 0.3 mag,
significantly larger than that for SDSS+2MASS galaxies (0.19 mag) at a 4.5σ level.
In order to understand the nature of galaxies with similar or redder (V − J)rest than
the bulk of nearby galaxies, we make use of stellar population synthesis models by
BC03. In Figure 5.4 we draw age tracks for three different dust-free star formation
86
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
Figure 5.4 — Rest-Frame U − V versus
V − J color-color diagram of all galaxies with L V > 5 × 109 L⊙ . SDSS+2MASS
galaxies (small grey dots) are plotted as
a local reference. The dust vector indicates an extinction of A V = 1 mag.
Color evolution tracks of (unreddened)
Bruzual & Charlot (2003) models are
overplotted: a simple stellar population (SSP, solid line), an exponentially
declining (τ300 , dotted line) and constant
(CSF, dashed line) star formation model.
Galaxies are greyscale-coded by best-fit
mass-weighted age. The tracks show
an increase toward redder U − V and
slightly redder V − J with age. At
a given U − V color redder than 1
galaxies that are red in V − J have the
youngest best-fit mass-weighted ages.
histories in the (U − V)rest vs (V − J)rest color-color diagram. The solid line represents
a single stellar population (SSP), the dashed line a continuous star formation model
(CSF) and the dotted line an exponentially declining star formation model with an efolding timescale of 300 Myr (τ300 ). All star formation histories show an evolution to
redder (U − V)rest and (V − J)rest with age. The τ300 model first has similar colors as
a CSF model and eventually moves to the same region in color space as an evolved
SSP, namely where the red peak of the SDSS bimodal U − V distribution is located. In
the absence of dust a population with a constant star formation history only reaches
U − V = 1 in a Hubble time.
We now investigate how the location in this color plane is related to stellar populations. Using the best-fit model parameters (see §5.3) we plot the mass-weighted ages
for the galaxies with LV > 5 × 109 L⊙ with greyscale-coding on Figure 5.4. Galaxies with
blue optical colors are indeed found to be young, the median mass-weighted age for
galaxies at (U − V)rest < 1 being 250 Myr. At (U − V)rest > 1 galaxies with a wide range
of stellar ages are found. The oldest stellar populations show the bluest (V − J)rest colors at a given (U − V)rest. Over the whole redshift range galaxies are present that have
red optical colors and whose SEDs are consistent with evolved stellar populations and
low dust content. According to their best-fit model, three of them started forming stars
less than 0.5 Gyr after the big bang and already at z > 2.5 have star formation rates
less than a percent of the past-averaged value. We note that in the Chandra Deep Field
South Papovich et al. (2006) find a number density of passively evolving galaxies at
high redshift that is nearly an order of magnitude lower than in the HDFS, possibly
owing to the fact that the HDFS observations probe to fainter K-band magnitudes. The
red (V − J)rest side of the color distribution is made up of galaxies that are best fitted
by young stellar populations. Since the age tracks alone cannot explain the presence of
galaxies with such red SEDs from the optical throughout the NIR, we investigate the
Section 5.4. Rest-frame optical to near-infrared color distribution
87
Figure 5.5 — Rest-Frame U − V versus
V − J color-color diagram of all galaxies with L V > 5 × 109 L⊙ . SDSS+2MASS
galaxies (small grey dots) are plotted as
a local reference. The vector indicates a
dust extinction of A V = 1 mag. Galaxies are greyscale-coded by best-fit A V .
The presence of dust moves galaxies to
redder U − V and V − J colors. Galaxies falling redward in V − J of the distribution of local galaxies are best described by dusty stellar populations.
role of dust in shaping the galaxy color distribution.
Figure 5.5 shows again the (U − V)rest versus (V − J)rest color-color diagram, now
greyscale-coded by best-fit dust extinction, expressed in A V . The arrow indicates an
A V of 1 magnitude using a Calzetti et al. (2000) extinction law. It is immediately apparent that the optical to NIR color-color diagram is a useful diagnostic for distinguishing
stellar populations with various amounts of dust extinction. At the bluest (U − V)rest
colors there is little evidence for dust obscuration. The degree of dust extinction increases as we move along the dust vector to redder colors.
Independent constraints on dust-enshrouded activity in distant galaxies can be derived from MIPS 24 µm imaging (Webb et al. 2006; Papovich et al. 2006). The midinfrared emission is usually thought to be powered by a dusty starburst in which PAH
features are produced or by an active galactic nucleus (AGN). Of the area with very
deep U-to-8 µm in the HDFS 95% is covered by a 1 hr MIPS pointing. We performed
the same photometric procedure to reduce confusion as for the IRAC photometry (see
§5.2.2). Fluxes were measured within a 6” diameter aperture and then scaled to total
using the growthcurve of the 24 µm PSF.
In Figure 5.6 we plot the (U − V)rest versus (V − J)rest color-color diagram of all
objects in the redshift interval 1.5 < z < 3.5 with LV > 5 × 109 L⊙ that are covered by
MIPS (empty circles). At these redshifts, strong PAH features, if present, move through
the MIPS 24 µm passband. Six sources have a MIPS 24 µm detection above 28 µJy
(3σ ). Their 24 µm flux is indicated by the filled circles. Object #767 is well detected
with F24µm = 95 µJy. As noted by Labbé et al. (2005) its SED shows an 8µm excess
with respect to the best-fitting template. The combination of 8 µm excess and 24 µm
detection suggests that this galaxy hosts an AGN whose power law SED dominates
throughout the mid-infrared. All other 24 µm detections are located in the part of the
diagram where our U-to-8 µm SED modeling found dusty young populations. None
88
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
Figure 5.6 — Rest-Frame U − V versus
V − J color-color diagram of galaxies at
1.5 < z < 3 with L V > 5 × 109 L⊙ with
MIPS 24 µm coverage. SDSS+2MASS
galaxies (small grey dots) are plotted as
a local reference. Filled circles represent MIPS 24 µm detections above a
28 µJy (3σ ) threshold. #767 is detected
at 24 µm and has an excess of 8 µm
flux compared to the best-fitting template SED, suggesting the presence of
an obscured AGN. All other 24 µm detections lie in the U − V, V − J region
populated by galaxies with dusty stellar populations. Assuming the 24 micron flux originates from PAH emission
produced by dust-enshrouded star formation, the MIPS observations confirm
the diagnostic power of this color combination.
of the blue relatively unobscured star-forming galaxies or red evolved galaxies show
evidence of PAH emission from the observed 24 µm flux. There are various reasons
why not all star-forming dusty galaxies have a 24 µm detection. The density of the UV
radiation field exciting the PAHs may vary among galaxies. Furthermore, the narrow
PAH features with respect to the width of the 24 µm passband make the 24 µm flux
very sensitive to redshift. Overall, MIPS observations agree well with SED modeling
and rest-frame optical-NIR color characterization.
We conclude that over the whole redshift range from z = 0 to z = 3.5 a trend is
visible of galaxies with redder optical colors showing redder optical to NIR colors.
However, at a given optical color, a spread in optical to NIR colors is observed that
is larger than for nearby galaxies. At (U − V)rest > 1 evolved galaxies with little dust
extinction are found at the bluest (V − J)rest. Dusty young star-forming galaxies occupy the reddest (V − J)rest colors. This is once more illustrated by the SEDs of galaxies
with (U − V)rest > 1 presented in Figure 5.7. The top row shows SEDs of objects at the
blue side of the (V − J)rest color distribution. The bottom panels show SEDs of galaxies matched in (U − V)rest, but with comparatively redder (V − J)rest colors. The latter
galaxies have comparatively younger ages and a larger dust content. Since this distinction could not be made on the basis of (U − V)rest color alone, the addition of IRAC 3.6
- 8 µm photometry to our U-to-K SEDs proves very valuable for the understanding of
stellar populations at high redshift.
We verified that no substantial changes occur to the rest-frame optical-to-NIR color
distribution and its interpretation in terms of age and dust content of the galaxies when
we derive photometric redshifts by running HYPERZ with redshift as free parameter
instead of using the algorithm developed by Rudnick et al. (2003; see §5.2.3).
Section 5.5. Constraints on stellar population properties at 2 < z < 3.5: age, dust and mass89
Figure 5.7 — Comparison of galaxies with similar (U − V)rest color but different (V − J)rest color. The top
row shows the galaxies with blue (V − J)rest colors, and the bottom row shows galaxies with matching
(U − V)rest color but much redder (V − J)rest color. The systematic difference in the SEDs of the two rows
is striking. Fits indicate old bursts of star formation with little dust in the top row, and dusty young
galaxies in the bottom row. This demonstrates the power of (V − J)rest in separating these classes. Note
that the U-band photometry for objects #224 and #249 deviates by more than 2σ from the predicted
U-band flux of the best-fit template.
5.5 Constraints on stellar population properties at 2 < z < 3.5: age,
dust and mass
We now proceed to analyze in more detail the constraints that IRAC places on the stellar populations of the subsample of galaxies at 2 < z < 3.5 (75 galaxies). In particular
we will focus on stellar mass, which likely plays a key role in galaxy evolution at all
redshifts (e.g., Kauffmann et al. 2003; Bundy et al. 2005; Drory et al. 2005; Rudnick et
al. 2006). Fortunately, estimates of stellar mass from modeling the broad-band SEDs
are generally more robust than estimates of dust content and stellar age (Bell& de Jong
2001; Shapley et al. 2001; Papovich et al. 2001; Förster Schreiber et al. 2004). Nevertheless, translating colors to mass-to-light ratios and subsequently stellar masses requires
a good understanding of the effects of age and dust.
5.5.1 Predictions from stellar population synthesis models
5.5.1.1 Wavelength dependence: optical versus near-infrared
In its simplest form the stellar mass of a galaxy can be estimated from one color (see,
e.g., Bell & de Jong 2001). To illustrate this process, we present the evolutionary track
of a dust-free BC03 model in a M/ LV versus U − V diagram (Fig. 5.8). Up to 2.5 Gyr
90
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
Figure 5.8 — Evolutionary track of a
two-component stellar population in
the M/ L V versus U − V plane. Filled
circles mark age steps of 100 Myr. Open
circles represent 10 Myr age steps. The
dust vector, indicating an extinction of
A V = 1 mag, lies parallel to the age
track. The histogram represents the
color distribution of galaxies at 2 < z <
3.5, with DRGs highlighted in solid
light-grey, LBGs in solid dark-grey. The
age track starts as an exponentially declining star formation model (τ = 300
Myr, BC03). At 2.5 Gyr a new burst
of star formation is introduced, lasting
100 Myr and contributing 20% to the
mass. Translating U − V into M/ L V
assuming one-component models can
lead to underestimates of M/ L V and
thus stellar mass. The possible underestimate is largest for blue galaxies.
after the onset of star formation (Fig. 5.8, top right corner) the track represents a onecomponent population with a star formation history that is exponentially declining,
with an e-folding timescale of 300 Myr. For most of the galaxies in our sample this was
the best-fitting star formation history. For more extreme star formation histories such
as an SSP or CSF the process of estimating M/ L values follows similar arguments. The
filled circles on Figure 5.8 represent age steps of 100 Myr. As the stellar population
ages, its V-band luminosity fades with only a small decrease in stellar mass from mass
loss, moving the galaxy up in M/ LV . Simultaneously the U − V color reddens as the
hot early-type stars with short lifetimes die. The dust vector indicating a reddening
of A V = 1 mag runs parallel to the age track of the one-component model. Ironically,
the mass estimate benefits greatly from this degeneracy between age and dust in the
optical. Under the assumption of a monotonic star formation history (U − V)rest can
uniquely be translated to M/ LV , regardless of the precise role of dust or age. Only a
normalization with LV is needed to derive the stellar mass. A similar relation was used
by Rudnick et al. (2003) to translate the integrated (U − V)rest color of high-redshift
galaxies into a global M/ LV and stellar mass density ρ∗ . They found that the conversion to mass-to-light ratio is more robust from the (U − V)rest color than from the
(U − B)rest or (B − V)rest color.
What if the actual star formation history is more complex? What effect does it
have on the derived stellar mass? There is ample evidence from local fossil records
(e.g., Trager et al. 2000; Lancon et al. 2001; Freeman & Bland-Hawthorn 2002; Förster
Schreiber et al. 2003; Angeretti et al. 2005) and high-redshift studies (e.g., Papovich
et al. 2001; Ferguson et al. 2002; Papovich et al. 2005) that galaxies of various types
have complex and diverse star formation histories, often with multiple or recurrent
episodes of intense star formation . Such a scenario is also predicted by cold dark matter models (e.g., Somerville, Primack,& Faber 2001; Nagamine et al. 2005; De Lucia
Section 5.5. Constraints on stellar population properties at 2 < z < 3.5: age, dust and mass91
Figure 5.9 — Evolutionary track of a
two-component stellar population in
the M/ L J versus V − J plane. A τ300
model from BC03 is shown. At 2.5
Gyr a 100 Myr burst is added , contributing 20% to the mass. Age marks
represent 100 Myr (filled circles) and 10
Myr (open circles) respectively. The histogram shows the color distribution of
our sample at 2 < z < 3.5, with DRGs in
solid light-grey and LBGs in solid darkgrey. For blue galaxies the V − J color
is insensitive to M/ L J , further complicated by the dust vector (A V = 1 mag)
that lies nearly orthogonal to the age
track meaning blue galaxies can have
a range of masses for the same V − J
color. On the other hand, the introduction of a second burst only causes
a small offset in M/ L J from the singlecomponent track, showing that the inclusion of a rest-frame NIR band reduces the uncertainties in stellar M/ L
caused by poor knowledge of the star
formation history.
et al. 2005). In order to address this question qualitatively, we consider the case of a
two-component population. At t = 2.5 Gyr we added a burst of star formation to the
τ300 model, lasting 100 Myr and contributing 20% to the mass. To follow the evolution of the two-component population closely, we mark 10 Myr timesteps with open
circles. Over a timespan of only 10 Myr the galaxy color shifts by 1.6 mag toward the
blue, while the M/ LV value stays well above the one-component M/ LV corresponding to that color. As the newly formed stars grow older, the galaxy moves toward the
upper right corner of the diagram again. The offset of M/ LV with respect to the onecomponent model is a decreasing function of U − V. This means that if a bursty star
formation is mistakenly fit with a one-component model the mass and mass-to-light
ratio are underestimated more for blue than for red galaxies, confirming what Shapley
et al. (2005) found for a sample of star-forming galaxies at z > 2.
The histogram at the bottom of Figure 5.8 indicates the (U − V)rest color distribution
of galaxies in the HDFS at 2 < z < 3.5. The population of Lyman-break galaxies (LBGs)
is marked in blue, Distant Red Galaxies (DRGs) in red. The possible underestimate in
mass-to-light ratio and thus mass is largest for blue galaxies, up to a factor of 3 for
(U − V)rest = 0.2, the bluest color reached by this two-component model. For DRGs
only a modest amount of mass can be hidden under the glare of a young burst of star
formation. The exact error that bursts cause depends on the form of the bursty star
formation history (see, e.g., Fig. 6 in Rudnick et al. 2003 for a different example).
We can now test whether rest-frame NIR photometry, as provided by IRAC, improves the constraints on the SED-based stellar mass estimates of high-redshift galaxies. Labbé et al. (2005) found that the range in M/ L K for DRGs and LBGs together is as
92
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
large as a factor 6, meaning that a Spitzer 8 µm-selected sample would be very different from a mass-selected sample. However, if a similar relation between mass-to-light
ratio and color exists in the rest-frame NIR as in the rest-frame optical, this does not
mean that the stellar mass estimate is uncertain by the same amount (a factor of 6).
Here we consider whether the mass-to-light ratio can robustly be derived from a given
rest-frame NIR color. We discuss only the rest-frame J-band but note that the results for
rest-frame K are similar. In Figure 5.9 we repeat the same exercise of drawing a M/ L
versus color evolutionary track for the rest-frame NIR. The burst that we superposed
on the τ300 model after 2.5 Gyr is again contributing 20% to the mass over a period of
100 Myr. Note that the scale is identical to that of Figure 5.8. The (V − J)rest histogram
of sources at 2 < z < 3.5 is derived from observed near- to mid-infrared wavelengths.
During the first gigayear, the V − J color hardly changes whereas M/ L J does by a factor of 7. As an immediate consequence, the translation of V − J into M/ L J is highly
uncertain for the blue galaxies in our sample and the additional IRAC observations do
not improve the constraints on the mass-to-light ratio. The situation is further complicated by the effect of dust. V − J is a lot more sensitive to dust than M/ L J , illustrated
by the dust vector of A V = 1 mag. The effects of dust and age no longer conspire to give
robust mass estimates at a given V − J color. At redder V − J the situation improves as
the slope of the age track flattens. Here the inclusion of a rest-frame NIR color clearly
reduces the uncertainty in stellar M/ L that stems from the poor knowledge of the star
formation history. The loop toward bluer colors is a magnitude smaller in size and we
see no large offsets in M/ L between the one- and two-component modeling.
We have discussed the different behavior of dust and age in simplified one- and
two-component models and have investigated the improvements expected from the
inclusion of the rest-frame NIR with respect to the rest-frame optical. While additional
rest-frame NIR data can lead to better M/ L estimates, in particular for redder galaxies
(U − V > 1; V − J > 0.4), it is clear that we need to take advantage of the full U-to-8
µm SED information to derive reliable estimates of stellar mass, stellar age and dust
content.
5.5.1.2 Model dependence: Bruzual & Charlot vs Maraston
It is important to note that different stellar population synthesis models do not paint
a consistent picture of evolution in the rest-frame NIR. To illustrate, we compare BC03
models to M05 models under the same assumption of Salpeter initial mass function
and solar metallicity.
Whereas the age track in a M/ LV versus U − V diagram behaves similarly for M05
and BC03, the NIR evolution of a τ300 model looks very different (see Fig. 5.10). The
grey dashed line represents the age track of a BC03 τ300 model with superposed burst at
2.5 Gyr as described in §5.5.1.1. In black we overplot the age track of a two-component
model with identical parameters by M05. In the 0.2 − 2 Gyr age range the two models look strikingly different. At the same V − J color the M05 model predicts M/ L J
values that are up to a factor 2.5 smaller than those of the BC03 model. The offset
between M/ L J as predicted from one- and two-component modeling is also larger by
a similar factor. The BC03 and M05 models differ in several aspects: the stellar evolutionary tracks adopted to construct the isochrones, the synthesis technique and the
Section 5.5. Constraints on stellar population properties at 2 < z < 3.5: age, dust and mass93
Figure 5.10 — Evolutionary track of
two-component stellar populations in
the M/ L J vs V − J plane based on
BC03 (grey dashed line) and M05 (black
solid line) models. For ages between
0.2 and 2 Gyr, the M05 model predicts much lower M/ L J values than
the BC03 model. The underestimate of
M/ L J as derived from one-component
modeling is therefore much more severe for the M05 model than for the
BC03 model, and the inclusion of restframe NIR data does not necessarily
improve constraints on stellar M/ L.
treatment of the thermally pulsating Asymptotic Giant Branch (TP-AGB) phase. The
Padova stellar tracks (Fagotto et al. 1994) used by BC03 include a certain amount of
convective-core overshooting whereas the Frascati tracks (Cassisi et al. 1997) do not.
The two stellar evolutionary models also differ for the temperature distribution of the
red giant branch phase. The higher NIR luminosity originates mainly from a different
implementation of the Thermally Pulsating Asymptotic Giant Branch (TP-AGB) phase
(M05). Following the fuel consumption approach, M05 finds that this phase in stellar
evolution has a substantial impact on the NIR luminosity at ages between 0.2 and 2
Gyr. BC03 follow the isochrone synthesis approach, characterizing properties of the
stellar population per mass bin. The latter method leads to smaller luminosity contributions by TP-AGB stars. We refer the reader to recent studies from M05, van der
Wel et al. (2006) and Maraston et al. (2006) for discussions of the model differences in
greater detail.
For our purpose it is sufficient to state that a given V − J color corresponds to
younger ages, lower mass-to-light ratios and thus lower masses for the M05 model
than for the BC03 model. Most importantly, we note that for M05 models inclusion
of NIR data does not reduce stellar mass uncertainties caused by the unknown star
formation history.
5.5.2 Constraints on mass, dust and age from modeling our observed galaxies
5.5.2.1 Wavelength dependence: optical versus near-infrared
Having investigated the qualitative relationship between M/ L and the rest-frame opticalto-NIR color in §5.5.1.1, we now quantify the effect of inclusion of IRAC MIR photometry on the stellar population constraints of galaxies at 2 < z < 3.5. Our goal is to
investigate whether and how the addition of IRAC imaging changes our best estimate
of the stellar population properties and their confidence intervals.
94
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
Figure 5.11 — Top row: Comparison of best-fit stellar masses, dust extinctions and mass-weighted ages
for galaxies at 2 < z < 3.5 when fit with IRAC photometry or without. The error bars are based on Monte
Carlo simulations given the photometric errors. Bottom row: Corresponding histograms with IRAC
photometry (filled) or without (dashed). No significant change in the overall distributions is observed,
but the best-fit properties of individual galaxies may change substantially.
First we compare the distribution of stellar mass, dust content and mass-weighted
stellar age as fit with or without IRAC. The top row of Figure 5.11 shows a direct comparison of the inferred model parameters with or without IRAC photometry for all
galaxies at 2 < z < 3.5. The filled histogram in the bottom row of Figure 5.11 shows
the distribution of mass, dust extinction and age derived from the full U-to-8 µm SED.
The dotted line indicates the distribution of best-fitting parameters from modeling the
U-to-K photometry. Both the median and the width of the distribution stays the same
for all three parameters. Defining the difference between mass, mass-weighted age,
and A V as ∆ log(M) = log(MwithI RAC) − log(MnoI RAC), ∆A V = A V,withI RAC − A V,noI RAC, and
∆ log(agew ) = log(agew,withI RAC) − log(agew,noI RAC) we find a median and normalized median absolute deviation (equal to the rms for a gaussian distribution) [x̂, σ NMAD(x)] of
(−0.007 ± 0.009, 0.07), (0.00 ± 0.03, 0.30), and (0.00 ± 0.02, 0.16) respectively. The average and standard deviation [hxi; σ (x)] of ∆ log(M), ∆A V and ∆ log(agew ) are (−0.04 ±
0.02; 0.13), (−0.08 ± 0.04; 0.36) and (−0.02 ± 0.03; 0.28) respectively. Thus the differences for the galaxy sample as a whole after including IRAC are very small. The results for stellar mass are similar to what Shapley et al. (2005) found for a more specific
sample of optically selected star-forming galaxies at z ∼ 2.
Having determined that the overall distribution of best-fit age, dust content and
Section 5.5. Constraints on stellar population properties at 2 < z < 3.5: age, dust and mass95
stellar mass does not change after including IRAC, the question remains whether IRAC
helps to improve the constraints on the stellar population characeristics for individual
galaxies. We address this question using the measure σnoI RAC /σwithI RAC, defined as
the ratio of confidence intervals without and with IRAC. The 1σ confidence intervals,
representing random uncertainties propagating from photometric errors, are derived
from Monte Carlo simulations. For each galaxy SED we create 100 mock SEDs where
the flux-point in each band is randomly drawn from a Gaussian with the measured flux
as the mean, and its error as the standard deviation. Next, each SED was fitted with
the same fitting procedure as the observed version. As we want to isolate the effect
of including IRAC observations on the confidence intervals we fix the redshift to z phot
(or zspec where available). In calculating σnoI RAC /σwithI RAC we measure the confidence
interval in log-space for stellar mass and mass-weighted age and in magnitude for
A V . Furthermore we set a lower limit to the confidence intervals to account for the
discreteness of our models, i.e., age and A V steps.
Figure 5.12 shows the values of σnoI RAC /σwithI RAC for mass, age and dust content
as a function of rest-frame (U − V)rest color for the galaxies at 2 < z < 3.5. We divide
the sample into a blue and red bin and indicate the median reduction of confidence
intervals for each bin with dashed lines. The separation between blue and red is chosen
to be (U − V)rest = 1, corresponding to the oberved (J − K)Vega > 2.3 color cut for Distant
Red Galaxies at the median redshift of our sample z = 2.66. We find that the typical
improvement of confidence intervals is dependent on galaxy color for all considered
stellar population parameters. For red galaxies, the reduction amounts to a median
factor of 2.7, 1.7 and 2.9 in the case of stellar mass, A V and age respectively. For blue
galaxies the reduction of the mass confidence interval is only a factor 1.3, though with
a large scatter, while for A V and age no median reduction is found. With the color
tracks of the stellar population models in mind (see Figs. 5.8-5.9) this color dependence
should come as no surprise. We demonstrated in §5.5.1.1 that for blue galaxies optical
to NIR colors are degenerate with the mass-to-light ratio. Hence, the IRAC bands of
blue galaxies contribute little information about their mass.
For a sample of (generally blue) optically selected star-forming galaxies at z ∼ 2
Shapley et al. (2005) found a reduction in stellar mass uncertainties by a factor 1.5 − 2
due to the addition of IRAC photometry, which seems like a contradiction. However,
the distribution of observed R − K color of their galaxies extends toward redder colors
than the Lyman-break galaxies (LBGs) in our sample, which may partly explain the
larger improvement than we find for blue LBGs. Another important difference is that
Shapley et al. (2005) lacked J and H images and hence did not probe rest-frame U − B
or U − V for their galaxies. It is possible that the lack of the near infrared J and H
bands in Shapley et al. (2005) is the main reason for the discrepancy. We simulated
this effect by omitting J and H and repeating the Monte Carlo simulations with and
without IRAC. The median reduction of the 1σ mass confidence interval now increases
to a factor 1.9 when including IRAC.
We conclude that, in the presence of very deep observed J, H, and K photometry,
inclusion of mid-infrared data places little extra constraints on the stellar populations
of blue galaxies. However, for galaxies redder than (U − V)rest = 1, IRAC reduces the
confidence interval by a substantial factor 2.5 − 3.
96
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
Figure 5.12 — Tightening of the confidence interval around best-fit stellar mass, age and dust extinction
as a function of rest-frame U − V color for galaxies at 2 < z < 3.5. The median improvement after
including the IRAC photometry is a factor 2.7 for red galaxies [(U − V)rest > 1], significantly larger than
the factor 1.3 for blue galaxies [(U − V)rest < 1]. A similar color-dependence is found for constraints on
age and dust extinction.
Table 5.5. Differences between stellar population properties derived from BC03 and
M05
median(∆ BC03 M05 )
log(M∗ )
AV
log(Agew )
0.14 ± 0.01
−0.20 ± 0.00
0.29 ± 0.02
σ NMAD (∆ BC03 M05 ) mean(∆ BC03 M05) σ (∆ BC03 M05 )
0.06
0.00
0.17
0.15 ± 0.01
−0.18 ± 0.02
0.34 ± 0.02
0.08
0.19
0.18
5.5.2.2 Model dependence: Bruzual & Charlot vs Maraston
In §5.5.1.2 we pointed out strong differences in the rest-frame optical-to-NIR colors
between the BC03 and M05 models. In this paragraph we quantify how our results
change if we use M05 models. The median and normalized median absolute deviation, average and standard deviation of the differences between BC03 fits and M05 fits
[∆ log(M) BC03 M05, ∆A V, BC03 M05 and ∆ log(agew ) BC03 M05] are summarized in Table 5.5.
As expected, the BC03 models predict older ages and thus higher stellar masses
than the M05 models for our z = 2 − 3.5 galaxies. The estimated mass for the M05
models is systematically lower by a factor 1.4. Maraston et al. (2006) found a similar
discrepancy for a sample of 7 galaxies in the Hubble Ultra Deep Field that satisfy the
BzK criterion (Daddi et al. 2004) for z > 1.4 passively evolving galaxies. Apart from a
systematic shift a scatter of 0.1 in dex is found in ∆ log(M) BC03 M05, meaning the choice
of stellar population synthesis model introduces a considerable systematic uncertainty.
It is of great importance to test whether ∆ log(M) BC03 M05 = log(MBC03) − log(MM05)
correlates with redshift, color or stellar mass, since such dependencies, if present,
could bias studies of galaxy evolution or trends with mass. In Figure 5.13 we plot
∆ log(M) BC03 M05 versus redshift, (U − V)rest color, and stellar mass (the latter derived
from BC03 models). We show galaxies with LV > 5 × 109 L⊙ at 1 < z < 2 (open symbols) and at 2 < z < 3.5 (filled symbols); no evidence for a redshift dependence is found.
Section 5.5. Constraints on stellar population properties at 2 < z < 3.5: age, dust and mass97
Figure 5.13 — Difference between best-fit stellar mass as derived from BC03 and M05 models as a function of redshift, (U − V)rest color and BC03 stellar mass for galaxies with L V > 5 × 109 L⊙ at 1 < z < 2
(open symbols) and 2 < z < 3.5 (filled symbols). The stellar masses derived from BC03 models are systematically higher than those derived from M05 models by a factor 1.4. The scatter in log(MBC03 / M M05 ) is
0.1 in dex. No significant dependence of log(MBC03 / M M05 ) on redshift, (U − V)rest color or stellar mass is
found. The bias introduced by the choice of stellar population synthesis model amounts to a maximum
of 15% over the whole (U − V)rest color or stellar mass range of our sample.
For the (U − V)rest (middle panel) and stellar mass (right) panel, the p-values for statistical significance from the Spearman rank order correlation test are also larger than
0.05, meaning no significant correlation is found. Fitting a line to the points in the
∆ log(M) BC03 M05 versus (U − V)rest diagram, a difference of 0.06 dex in ∆ log(M) BC03 M05
is found over the 2 mag range in (U − V)rest color spanned by the galaxies in our sample. Even if a trend of increasing ∆ log(M) BC03 M05 with redder (U − V)rest color is real,
it only introduces a small bias of the order of 15%. A similar conclusion can be drawn
for the dependence on stellar mass.
5.5.2.3 Metallicity dependence
We test how variations from solar metallicity affect the estimates of stellar mass, massweighted age and dust extinction. We study the effect of a different metal abundance
by fitting BC03 templates with metallicity Z = 0.2 Z⊙ to the observed SEDs, leaving the
extinction law to Calzetti et al. (2000). NIR spectroscopy of DRGs (van Dokkum et al.
2004) and LBGs (Erb et al. 2006a) indicates that a range of Z = 0.2 − 1 Z⊙ is appropriate
for galaxies at 2 < z < 3.5. Furthermore, at metallicities below Z = 0.2 Z⊙ the tracks
and spectral libraries used to build the BC03 templates become more uncertain by lack
of observational constraints. Decreasing the metallicity from Z = Z⊙ to Z = 0.2 Z⊙
lowers the estimated stellar masses of galaxies at 2 < z < 3.5 by 0.1 dex, leads to a massweighted age that is typically lower by 0.2 dex, and is compensated by an average
increase in A V of 0.2 mag. The fact that age estimates are more strongly affected than
estimates of stellar mass when changing the assumed metallicity was demonstrated in
detail by Worthey (1994). While absolute values of ages and dust extinctions may be
biased as just described, the relative age and dust trends within the galaxy population
as discussed in §5.4 based on the standard SED modeling (see §5.3) are robust.
98
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
5.5.2.4 Dependence on extinction law
The Calzetti et al. (2000) extinction law was empirically derived from observations
of local starburst galaxies. We quantify the variations in stellar population properties
due to the adopted extinction law by comparing our modeling results with a Calzetti
et al. (2000) law to those obtained with reddening laws from Fitzpatrick (1986) for
the Large Magellanic Cloud (LMC) and Prévot et al. (1984) for the Small Magellanic
Cloud (SMC), leaving the metallicity to solar. Stellar masses, mass-weighted ages and
A V values of galaxies at 2 < z < 3.5 derived with the LMC law models are similar
to those obtained with the Calzetti et al. (2000) law. The SMC law, which rises more
steeply toward shorter wavelengths in the near-UV, gives similar mass estimates, A V
values that are on average smaller by 0.3 mag and mass-weighted stellar ages that are
older by 0.23 dex, with the ages of the oldest galaxies being limited by the age of the
universe constraint. As for metallicity, we conclude that using a different extinction
law has a larger impact on the age estimates than on estimates of stellar mass.
5.6 Stellar mass - optical color relation
In this section we study the relation between the rest-frame optical color of highredshift galaxies and their stellar mass. We start with a model-independent approach
in Figure 5.14, plotting rest-frame (U − V)rest versus rest-frame Jrest magnitude for all
galaxies at 2 < z < 3.5. The emission of low mass long-lived stars that make up the
bulk of the mass in a galaxy peaks in the rest-frame NIR. Jrest is therefore expected to
be a reasonably good tracer of stellar mass. The galaxies that satisfy the DRG selection criterion (light-grey circles) are found at redder (U − V)rest than the Lyman-break
galaxies (dark-grey circles). The reddest (U − V)rest colors are found at the brightest
Jrest magnitudes. Note however that the observed trend is partially driven by the Kband selection of our sample. The line on Figure 5.14 indicates at which magnitude
a galaxy with identical colors to our observed galaxies would fall out of the sample.
Even if we only consider galaxies brighter than the limiting Jrest = −21.5 to which we
are complete over the whole (U − V)rest color range, we find that galaxies redder than
(U − V)rest = 1 are 1 mag brighter than galaxies with (U − V)rest < 1, significant at the
3σ level. Studying a sample without color bias (as advocated by van Dokkum et al.
2006) proves crucial to pick up the trend of (U − V)rest with Jrest. We note that Meurer
et al. (1999) found that LBGs with higher rest-frame UV luminosities tend to have redder rest-frame UV colors, illustrating that, while trends of color with luminosity are
most notable in samples without color bias, they are still present in at least some color
selected samples.
If Jrest is a reasonable tracer of stellar mass, we expect to see a similar or stronger
trend of (U − V)rest with the stellar mass. This is shown in Figure 5.15. The plotted mass
is derived from one-component SED modeling of the U-to-8 µm SED as described in
§5.3. The typical error bar is indicated in the bottom left corner. The depth of our K
detection band allows us to probe stellar masses from 3 × 1011 M⊙ down to 2 × 109 M⊙ .
A correlation of (U − V)rest with stellar mass is clearly visible. The most massive galaxies have a red optical color. LBGs and other blue galaxies at high redshift contain
typically 5 times less stellar mass than the DRGs in our sample. Again the K-band
Section 5.6. Stellar mass - optical color relation
Figure 5.14 — Rest-Frame U − V color
versus absolute J magnitude for galaxies at 2 < z < 3.5. Lyman-break galaxies are plotted with dark-grey circles.
DRGs (light-grey circles) populate the
red side of the U − V color distribution.
Black symbols denote those objects that
do not meet either criteria. The solid
line marks the K-band selection of our
sample. The dust vector indicates an
extinction of A V = 1 mag. The most luminous galaxies in the rest-frame NIR
have redder rest-frame optical colors
than fainter galaxies.
Figure 5.15 — Rest-Frame U − V color
versus stellar mass for galaxies at 2 <
z < 3.5. DRGs are marked with lightgrey circles, LBGs with dark-grey circles. The K-band selection of our sample is indicated by the solid line. The
dust vector indicates an extinction of
A V = 1 mag. Low-mass galaxies with
red colors might exist, but would not
enter the sample. The most massive
galaxies have redder U − V colors than
less massive galaxies; notice the striking absence of massive blue galaxies.
99
100
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
Figure 5.16 — Rest-Frame U − V color versus stellar mass for galaxies at 2 < z < 3.5 with complex
star formation histories. For each object a vector starts at the best-fit mass from one-component SED
modeling. The vector ends at an upper limit for the stellar mass as obtained by fitting two-component
models. The two-component model is composed of a maximally old stellar population with a second
burst of star formation during the last 100 Myr superposed. Empty symbols refer to objects for which
χ2red,two−component − χ2red,one−component > 2 for all considered burst fractions. The mass that is plotted here
corresponds to the burstfraction that gave the lowest χ2red. The typical amount of mass that can be
hidden under the glare of a young secondary burst is on average larger for blue than for red galaxies.
Nevertheless, even allowing for more complex star formation histories than one-component models a
lack of massive blue galaxies remains visible.
selection of our sample (solid line) limits our ability to detect faint red galaxies. Therefore we can not exclude the presence of low-mass red galaxies. The lack of massive
blue galaxies seems to be real. Rigopoulou et al. (2006) find a co-moving density of
Φ = (1.6 ± 0.5) × 10−5 Mpc−3 for LBGs with M > 1011 M⊙ at an average redshift hzi ≃ 3,
consistent with the absence of such massive but rare LBGs in our sample.
However, the lack of massive blue galaxies could be an artifact of our choice of
simple star formation histories. As demonstrated in §5.5.1.1 a severe underestimate of
the stellar mass is possible when the true star formation history is more complex than
that of the modeled one-component stellar population. When a young burst of star
formation is superposed on a maximally old population, its blue light will dominate
the (U − V)rest color and the mass from the underlying population will be hidden. In
order to constrain the possible underestimate in mass, we fit two-component models
to our SEDs. Erb et al. (2006b) describe a procedure to achieve this in two steps, where
first a maximally old population is fit to the K(+IRAC) data and subsequently a young
population is fit to the (primarily UV) residual. However, this procedure does not
guarantee a good fit in the χ2 sense. Instead, we decided to perform a simultaneous
fit of both old and young components. We constructed template SEDs consisting of
Section 5.7. Summary
101
a maximally old single stellar population with a recent burst of star formation that
started 100 Myr ago and lasted till the moment of observation superposed. We made
templates where the mass fraction created in the burst is 2x with x going from -6 to 2 in
steps of 1. We assume that the same reddening by dust applies to the old and the young
population, with A V ranging from 0 to 3 in steps of 0.2. Without this assumption, one
could in principle hide an infinite amount of mass in an old population as long as
an optically thick medium is shielding it from our sight. However, such a scenario
is physically implausible. Since we are interested in an upper limit on the mass, as
opposed to the most likely value, we do not search for the least-squares solution over
all x. Instead we perform the fit for every burst fraction and select the highest mass
that still has ∆χ2red = χ2red,two−component − χ2red,min,one−component < 2.
Fitting the two-component models to the U-to- 8 µm SEDs of our galaxy sample at
2 < z < 3.5, we indeed see that a higher stellar mass is allowed when more complex
star formation histories are adopted (Fig. 5.16). The upper bound on stellar mass that
we derive from this particular two-burst model is in the median a factor 1.7 higher than
the one-component estimate for galaxies redder than (U − V)rest = 1. For blue galaxies
the median increase is a factor 2.1. Despite the fact that more mass can be hidden in
blue galaxies, a trend of optical color with stellar mass remains visible. We performed
a Mann-Whitney U-test to compare the (U − V)rest colors of galaxies with different
stellar mass. We conservatively adopted the one-component stellar mass for galaxies
with (U − V)rest > 1 and the two-component upper limit for objects with (U − V)rest < 1.
To avoid selection effects we only consider galaxies more massive than M = 1010 M⊙ .
Dividing them in two mass bins with an equal number of objects the Mann-Whitney
U-test (Walpole & Myers 1985) confirms at a 99% significance level that the mean of
the (U − V)rest distributions differs. Applying the same two-component models to the
U-to-K SEDs (omitting IRAC), the median upper mass estimate increases to a factor
2.3 above the one-component estimate for red objects and a factor 3.7 for blue objects.
We conclude that, as expected from §5.5.1.1, more mass can be hidden in blue than
in red galaxies, but this effect is insufficient to remove the trend of stellar mass with
color. Furthermore, the amount of mass that can be hidden is constrained by addition
of IRAC photometry.
The color dependence that we derive for the amount of mass that can be hidden in
an underlying old population confirms findings from Shapley et al. (2005) based on
a sample of star-forming galaxies at z ∼ 2. The predominance of distant red galaxies
at the high-mass end was illustrated recently by van Dokkum et al. (2006) using a
mass-selected sample of galaxies at 2 < z < 3 with M > 1011 M⊙ . Only with very deep
imaging such as that of the HDFS analyzed in this chapter it is possible to probe down
to lower masses and prove that the most massive galaxies have red (U − V)rest colors
compared to lower mass galaxies.
5.7 Summary
We investigated the rest-frame optical to NIR color distribution of galaxies up to 2 <
z < 3.5 in the Hubble Deep Field South. At all redshifts, galaxies with redder (U − V)rest
tend to have redder (V − J)rest, as is the case in the local universe. At (U − V)rest colors
102
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
comparable to that of local galaxies, the color distribution of distant galaxies extends
to redder (V − J)rest. At (U − V)rest > 1 the population of galaxies at the red (V − J)rest
end is well described by dust-enshrouded star-forming models, whereas galaxies with
(V − J)rest similar to that of local galaxies are consistent with old passively evolving
systems. We conclude that (U − V)rest alone allows us to isolate blue relatively unobscured star-forming galaxies, but addition of (V − J)rest is necessary to distinguish
young dusty from old passively evolved systems. At redshifts above z = 1, this means
IRAC observations are crucial in understanding the wide variety in stellar populations.
We note that our analysis is not subject to uncertainties due to field-to-field variations,
but surveys over much larger areas are needed to study the relative contributions of
galaxies with different stellar populations.
We analyzed the constraints that IRAC places on stellar mass, stellar age and dust
content of galaxies at 2 < z < 3.5. No evidence is found for systematic offsets when
determining the stellar population characteristics with or without IRAC. However, the
ratio of confidence intervals on stellar mass, mass-weighted age and dust extinction
is typically reduced by a factor 2.7, 2.9 and 1.7 respectively for red [(U − V)rest > 1]
galaxies. In general, IRAC does not provide stronger constraints for blue galaxies [(U −
V)rest < 1] when very deep NIR imaging is available (as is the case for the HDFS).
We caution that, in characterizing the stellar populations using M05 models, we
find stellar masses that are typically a factor 1.4 lower than for BC03 models with a
scatter of 0.1 in dex.
A trend of brighter Jrest with redder (U − V)rest is observed for galaxies at 2 < z < 3.5,
where the NIR luminosity serves as a (imperfect but model-independent) tracer for
stellar mass. Plotting (U − V)rest versus modeled stellar mass, we arrive at a similar
conclusion: the most massive galaxies in our sample have red rest-frame optical colors. A possible concern is that this trend with mass is caused by our simplistic choice
of star formation histories. When we allow for more complex star formation histories,
more mass can be hidden than in the case of a one-component stellar population and
the amount depends on the color of the galaxy. We used two-component stellar populations, consisting of a maximally old population with a young population superposed,
to set an upper bound on the stellar mass present. Even though relatively more mass
can be hidden in blue galaxies compared to red galaxies, under the assumption of an
equal dust reddening of the young and old component, a trend of (U − V)rest increasing
to redder colors with stellar mass remains visible.
Acknowledgments
This research was supported by a grant from the Netherlands Foundation for Research
(NWO), and the Leids Kerkhoven-Bosscha Fonds. The authors would like to thank
Carnegie Observatories and the Lorentz Center for its hospitality during working visits and workshops. Support from NASA LTSA grant NNG04GE12G is gratefully acknowledged.
Section 5.7. Summary
103
References
Adelman-McCarthy, J. K. et al. 2006, ApJS, 162, 38
Angeretti, L., Tosi, M., Greggio, L., Sabbi, E., Aloisi, A.,& Leitherer, Claus 2005, AJ, 129, 2203
Baldry, I. K., Glazebrook, K., Brinkmann, J., Ivezić, Z̆., Lupton, R. H., Nichol, R. C.,& Szalay, A. S. 2004,
ApJ, 600, 681
Bell, E. F.,& de Jong, R. S. 2001, ApJ, 550, 212
Blanton, M. R., et al. 2003, ApJ, 594, 186
Blanton, M. R., et al. 2005, AJ, 129, 2562
Bolzonella, M., Miralles, J.-M.,& Pelló, R. 2000, A&A, 363, 476
Bruzual, G.,& Charlot, S. 2003, MNRAS, 344, 1000 (BC03)
Bundy, K., Ellis, R. S.,& Conselice, C. J. 2005, ApJ, 625, 621
Calzetti, D., et al. 2000, ApJ, 533, 682
Cassisi, S., Castellani, M.,& Castellani, V. 1997, A&A, 317, 108
Coleman, G. D., Wu, C.-C.,& Weedman, D. W. 1980, ApJS, 43, 393
Cutri, R. M., et al. 2000, The 2MASS Explanatory Supplement (Pasadena: IPAC),
http://www.ipac.caltech.edu/2mass/releases/allsky/doc/explsup.html
Daddi, E., Cimatti, A., Renzini, A., Fontana, A., Mignoli, M., Pozzetti, L., Tozzi, P.,& Zamorani, G. 2004,
ApJ, 617, 746
De Lucia, G., Springel, V., White, S. D. M., Croton, D.,& Kauffmann, G. K. 2006, MNRAS, 366, 499
Drory, N., Salvato, M., Gabasch, A., Bender, R., Hopp, U., Feulner, G.,& Pannella, M. 2005, ApJ, 619, 131
Erb, D. K., et al. 2003, ApJ, 591, 101
Erb, D. K., Shapley, A. E., Pettini, M., Steidel, C. C., Reddy, N. A.,& Adelberger, K. L. 2006a, ApJ, 644,
813
Erb, D. K., Steidel, C. C., Shapley, A. E., Pettini, M., Reddy, N. A.,& Adelberger, K. L. 2006b, ApJ, 646,
107
Fagotto, F., Bressan, A., Bertelli, G.,& Chiosi, C. 1994, A&AS, 104, 365
Fazio, G. G., et al. 2004, ApJS, 154, 10
Ferguson, H. C., Dickinson, M., Papovich, C. 2002, ApJ, 569, 65
Fitzpatrick, E. L. 1986, AJ, 92, 1068
Förster Schreiber, N. M., et al. 2003, ApJ, 599, 193
Förster Schreiber, N. M., et al. 2004, ApJ, 616, 40
Franx, M., et al. 2003, ApJ, 587, L79
Freeman, K.,& Bland-Hawthorn, J., 2002, ARA&A, 40, 487
Glazebrook, K., et al. 2004, Nature, 430, 181
Hora, J. L., et al. 2004, Proc. SPIE, 5487, 77
Kauffmann, G., et al. 2003, MNRAS, 341, 54
Kriek, M., et al. 2006, ApJ, 645, 44
Kinney, A. L., et al. 1996, ApJ, 467, 38
Labbé, I., et al. 2003, AJ, 125, 1107
Labbé, I., et al. 2005, ApJ, 624, L81
Lancon, A., et al. 2001, ApJ, 552, 150
Madau, P., et al. 1996, MNRAS, 283, 1388
Maraston, C. 2005, MNRAS, 362, 799
Maraston, C., et al. 2006, ApJ, 652, 85
Meurer, G. R., Heckman, T. M.,& Calzetti, D. 1999, ApJ, 521, 64
Nagamine, K., Cen, R., Hernquist, L., Ostriker, J. P.,& Springel, V. 2005, ApJ, 627, 608
Papovich, C., Dickinson, M., & Ferguson, H. C. 2001, ApJ, 559, 620
Papovich, C., Dickinson, M., Giavalisco, M., Conselice, C. J.,& Ferguson, H. C. 2005, ApJ, 631, 101
Papovich, C., et al. 2006, ApJ, 640, 29
Prévot, M. L., Lequeux, J., Prévot, L., Maurice, E.,& Rocca-Volmerange, B. 1984, A&A, 132, 389
Rigopoulou, D., et al. 2006, ApJ, 648, 81
Rudnick, G., et al. 2001, AJ, 122, 2205
Rudnick, G., et al. 2003, ApJ, 599, 847
Rudnick, G., et al. 2006, ApJ, 650, 624
104
Chapter 5. What do we learn from IRAC observations of galaxies at 2 < z < 3.5?
Salpeter, E. E. 1955, ApJ, 121, 161
Shapley, A. E., et al. 2001, ApJ, 562, 95
Shapley, A. E., et al. 2003, ApJ, 588, 65
Shapley, A. E., et al. 2005, ApJ, 626, 698
Smail, I., Chapman, S. C., Blain, A. W.,& Ivison, R. J. 2004, ApJ, 616,71
Somerville, R. S., Primack, J. R.,& Faber, S. M. 2001, MNRAS, 320, 504
Steidel, C. C., et al. 1996, ApJ, 462, L17
Steidel, C. C., et al. 1999, ApJ, 519, 1
Trager, S. C., Faber, S. M., Worthey, G.,& González, J. J. 2000, AJ, 120, 165
van der Wel, A., Franx, M., van Dokkum, P. G., Rix, H.-W., Illingworth, G. D.,& Rosati, P. 2005, ApJ, 631,
145
van der Wel, A., et al. 2006, ApJ, 652, 97
van Dokkum, P. G., et al. 2006, ApJ, 638, 59
Walpole, R. E.,& Myers, R. H. 1985, Probability and Statistics for Engineers and Scientists, 3rd edition
(Macmillan Publishing Company)
Webb, T. M., et al. 2006, ApJ, 636, 17
Werner, M. W., et al. 2004, ApJS, 154, 1
Worthey, G. 1994, ApJS, 95, 107
Yan, H., et al. 2004, ApJ, 616,63
Chapter 6
Recovering Stellar Population Properties
and Redshifts from Broad-Band
Photometry of Simulated Galaxies:
Lessons for SED Modeling
Abstract. We present a detailed analysis of our ability to determine stellar masses,
ages, reddening and extinction values of high-redshift galaxies by modeling broadband SEDs with stellar population synthesis. In order to do so, we computed synthetic optical-to-NIR SEDs for model galaxies taken from hydrodynamical merger
simulations placed at redshifts 1.5 ≤ z ≤ 2.9. Viewed under different angles and
during different evolutionary phases, the simulations represent a wide variety of
galaxy types (disks, mergers, spheroids). The broad-band SEDs were then fed to
a standard SED modeling procedure and resulting stellar population parameters
were compared to their true values. We specifically analyze how well the SED
modeling reproduces masses, ages, and extinction. Disk galaxies generally show a
decent median correspondence between the true and estimated mass and age, al+0.06
+0.26
beit with a significant scatter (∆ log M = −0.05−
0.13 , ∆ log agew = −0.04−0.27 ). Dur+0.09
ing the merger itself, we find larger offsets: ∆ log M = −0.11−
0.14 and ∆ log agew =
+0.34
−0.11−0.25 . E(B − V) values are generally recovered well, but the estimated total visual absorption A V is consistently too low, increasingly so for larger optical depths
+0.42
(∆A V = −0.48−
0.45 in the merger regime). The masses, ages, E(B − V), and A V of
simulated ellipticals are very well reproduced.
We discuss possible biases in SED modeling results caused by mismatch between
the true and template star formation history, dust distribution, metallicity variations and AGN contribution. Mismatch between the real and template star formation history, as is the case during the merging event, drives the age, and consequently mass estimate, down with respect to the true age and mass. However, the
larger optical depth toward young stars during this phase reduces the effect considerably. Finally, we tested the photometric redshift code EAZY on the simulated
galaxies placed at high redshift. We find a small scatter in ∆z/(1 + z) of 0.030 to
0.054, depending on the template set used.
S. Wuyts, T. J. Cox, N. M. Förster Schreiber, M. Franx,
P. F. Hopkins, L. Hernquist, B. Robertson & P. G. van Dokkum
105
106
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
6.1 Introduction
U
NDERSTANDING the growth and aging of galaxies over cosmic time requires reliable estimates of their mass, formation epoch and star formation history. With the
current generation of telescopes, stellar velocity dispersion measurements can probe
the gravitational potential in which the baryonic galaxy content resides out to z ∼ 1.3
(van Dokkum & Stanford 2003; Holden et al. 2005). Beyond this redshift, gas velocity
dispersions can be measured from emission lines, but do not always trace the potential due to outflows (Franx et al. 1997; Pettini et al. 1998, 2001; Shapley et al. 2003),
and would lead to biased samples missing quiescent galaxies lacking emission lines in
their spectra (Kriek et al. 2006). For these reasons, most studies of high-redshift galaxies have used stellar mass estimates derived by modeling of the broad-band stellar
energy distribution to characterize the mass.
Since age estimates from H α equivalent widths (van Dokkum et al. 2004; Erb et
al. 2006c) or Balmer/4000Å break strengths (Kriek et al. 2006) are very demanding in
terms of telescope time and only attainable for the brightest galaxies, stellar ages as
well are commonly derived from broad-band photometry.
Over the past few years, SED modeling has been proven extremely valuable in
characterizing the galaxy population in the early universe (e.g. Papovich et al. 2001;
Shapley et al. 2001, 2005; Förster Schreiber et al. 2004). Nevertheless, a number of
assumptions are required for the limited number of datapoints (11 passbands in our
case, but often less) to lead to a single solution in terms of physical properties such as
stellar mass, stellar age, dust extinction, and often redshift.
First, the star formation history (SFH) is generally modelled by a simple functional
form: a single burst, constant star formation, or an exponentially declining model. In
reality, high-redshift galaxies show evidence of more complex SFHs, often with brief
recurrent episodes of star formation (e.g. Papovich et al. 2001; Ferguson et al. 2002;
Papovich et al. 2005). Second, we use the approximation of a single foreground screen
of dust in accounting for the attenuation, even though in reality the dust will be distributed in between the stars. Third, we fit solar metallicity models. Although consistent with the current metallicity estimates from near-infrared (NIR) spectroscopy of
high-redshift galaxies (van Dokkum et al. 2004; Erb et al. 2006a), it must be kept in
mind that these measurements are currently limited to the bright end of the galaxy
population. Fourth, SED modeling generally assumes a purely stellar origin of the
light, while observational evidence for a substantial fraction of low luminosity AGN
at high redshift has been accumulating (van Dokkum et al. 2004; Reddy et al. 2005;
Papovich et al. 2006; Kriek et al. 2006; Daddi et al. 2007). They may contribute to the
optical SEDs.
Finally, one adopts a certain attenuation law, initial mass function (IMF), and stellar
population synthesis code. Their appropriateness at low and high redshifts is much
debated.
In this chapter, we address the impact of the first four assumptions (related to SFH,
dust attenuation, metallicity, and AGN) using hydrodynamical simulations of merging
galaxies (see Robertson et al. 2006; Cox et al. 2006). The SPH simulations follow the
star formation on a physical basis, resulting in more complex SFHs than are allowed
Section 6.2. The simulations
107
in typical SED modeling. They keep track of the distribution and metallicity of gas
and stellar particles, allowing a determination of the line-of-sight dependent extinction
toward each stellar particle seperately and a knowledge of the stellar metallicity as a
function of time. Here, we apply the same SED modeling that we use for observed
galaxies to broad-band photometry extracted from the simulation outputs, and study
how well the mass, age, and dust content of the simulated galaxies can be recovered.
The reason we use merger simulations for this exercise is threefold. First, galaxy
mergers are believed to play an important role in galaxy evolution (see, e.g., Holmberg
1941; Zwicky 1956; Toomre & Toomre 1972; Toomre 1977), increasingly so at high redshift (see, e.g., Glazebrook et al. 1995; Driver, Windhorst,& Griffiths 1995; Abraham
et al. 1996). Moreover, along their evolutionary path they are visible as vastly different galaxy types, allowing to test the recovery of stellar population parameters under
a wide range of conditions: gas-rich star-forming disks, dust-obscured mergers, and
quiescent spheroids. Finally, in Chapter 7 we will compare predictions of the color
distribution and mass density of high-redshift galaxies derived from these simulations
with the observed galaxy population in deep fields. A good understanding of what it
is we measure with SED modeling is crucial in order to compare identical mass-limited
samples of observed and simulated galaxies.
We start with a description of the simulations in §6.2. Next, we explain the methodology of our SED modeling in §6.3. §6.4 discusses how well we can measure stellar
population properties when a spectroscopic redshift is available. §6.5 repeats the analysis, now leaving the redshift as an extra free parameter (i.e., fitting for the photometric
redshift). Finally, we summarize the results in §7.11.
6.2 The simulations
6.2.1 Main characteristics
The simulations on which we test our SED modeling were performed by Robertson et
al. (2006). We refer the reader to that paper for a detailed description of the simulations.
Briefly, the simulations were performed with the parallel TreeSPH code GADGET-2
(Springel 2005). The code uses an entropy-conserving formulation of smoothed particle hydrodynamics (Springel & Hernquist 2002), and includes gas cooling, a multiphase model for the interstellar medium (ISM) to describe star formation and supernova feedback (Springel & Hernquist 2003), and a prescription for supermassive black
hole growth and feedback (Springel et al. 2005b).
At the start, each simulation consists of 120000 dark matter particles, 80000 gas
particles, and 80000 stellar particles. They represent two stable, co-planar disk galaxies,
each embedded in an extended dark matter halo with Hernquist (1990) profile. We
have realisations where the disks start with a gas fraction of 40% and 80%. Stellar
masses at the start of the simulation varied from 7.0 × 109 M⊙ to 2.3 × 1011 M⊙ per
disk galaxy. For a given virial velocity, the halo concentration, virial mass and virial
radius were scaled following Robertson et al. (2006) to approximate the structure of
disk galaxies at redshift z = 3. In practice, this means that the mass- and redshift-
108
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
dependent halo concentration measured by Bullock et al. (2001) was adopted:
−0.13
Mvir
(1 + z)−1 ,
Cvir (Mvir , z) ≈ 9
Mcoll ,0
(6.1)
where Mcoll ,0 ∼ 8 × 1012 h−1 M⊙ is the linear collapse mass at z=0, and that the following
scaling relations were used for the virial mass and virial radius of the progenitors:
3
Vvir
10GH(z)
Vvir
Rvir =
,
10H(z)
Mvir =
(6.2)
(6.3)
where Vvir is the virial velocity and H(z) is the Hubble parameter.
We set the ages of the stars existing at the start of the simulation such as to represent a constant star formation history prior to the start of the simulation at a star
formation rate (SFR) equal to that calculated in the first phases of the simulation. The
corresponding stellar metallicities were then set according to the closed box model:
Z(t) = − y ln[ f gas (t)], where Z(t) is the metallicity of a stellar particle formed at time t,
the yield y=0.02 and f gas (t) is the gas fraction of the system at the considered time. Similarly, the gas at the start of the simulation was assigned a uniform metallicity Zgas (t S ) =
− y ln[ f gas (t S )] where t S represents the start of the simulation, and f gas (t S ) = 0.4 or 0.8
respectively for our 2 gas fraction runs. The closed box model represents an upper
limit on the allowed enrichment by heavy elements, which in reality may be reduced
by outflows or infall of metal-poor gas (Edmunds 1990). The fact that we consider 2 gas
fractions guarantees a wide range of progenitor types, with ages of a few 100 Myr and
Zgas = 0.004 for f gas = 0.8 to typical stellar ages of a Gyr and nearly solar gas metallicity
for f gas = 0.4.
The overall timespan covered by each simulation was 2 Gyr. Figure 6.1(a) illustrates a typical star formation history of one of the merger simulations. Figure 6.1(b)
illustrates the build-up of stellar mass and Figure 6.1(c) presents the accretion history
onto the black hole(s). We draw the time axis relative to the actual moment of merging,
defined as the timestep when the two black hole particles become one, coinciding with
the peak in the accretion history. Cross symbols indicate the snapshots, separated by
70 Myr, when all physical information was stored to disk.
As time progresses, the orderly rotation and star formation in the disks is disturbed
by each others gravitational pull. The star formation history shows a first, but rather
shallow, bump during the first passage of the disks. Next, gravitational torques enable
the gas to loose angular momentum and flow to the centers where it triggers a starburst
(Larson & Tinsley 1978; Noguchi 1988; Hernquist 1989; Barnes & Hernquist 1991, 1996;
Mihos & Hernquist 1994, 1996). Meanwhile, part of the inflowing gas is fed to the
central supermassive black holes (SMBHs). Once the SMBHs grow massive enough,
they produce a luminous quasar (Sanders et al. 1988a,b; Hernquist 1989; Sanders &
Mirabel 1996; Genzel et al. 1998) whose feedback halts subsequent star formation (Di
Matteo et al. 2005; Springel et al 2005a), leaving a red spheroid galaxy as remnant
(Robertson et al. 2006; Cox et al. 2006).
Section 6.2. The simulations
109
Figure 6.1 — Evolution of a typical merger simulation. (a) The star formation history, (b) the mass
build-up, (c) the accretion rate history onto the black hole(s), (d) the evolution of the intrinsic (i.e., unattenuated) V-band luminosity, and (e) the binned distribution of effective visual extinctions (attenuated
minus intrinsic V-band magnitude) corresponding to different viewing angles. A darker intensity indicates a larger number of viewing angles. The solid line represents the median evolution which peaks at
the moment of actual merging. The dotted lines indicate the interval containing the central 68% of the
viewing angles. The cross symbols in panels (a), (b) and (d) mark the sampling of snapshots when the
full physical information of all SPH particles was stored to disk. After a first bump in the star formation
rate during the first passage of the progenitors, a peak in star formation is reached for a brief period during which several hundreds of solar masses of gas are converted into stars. The typical extinction for a
random line of sight is peaking around the same time. Shortly after, the accretion onto the supermassive
black hole is maximal, coinciding with the merger between the two progenitor black holes.
110
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
6.2.2 Extracting photometry from the simulation output
The evolutionary path as outlined in §6.2.1 is followed by the GADGET-2 code at a
fine time resolution (∆t ∼ 104 yr). At sparser timesteps (70 Myr apart), the positions,
masses, ages, and metallicities of all particles were stored. It is from these simulation
snapshots that we derive the observed SEDs of the merger as a function of time.
The light a virtual observer would receive from the simulated merger, is composed
of stellar and AGN emission, the latter only contributing significantly during a brief
period of time. We ignore any contribution from emission lines produced by the gas
content of the galaxies, possibly contributing on the order of 0.1 mag in the optical. Furthermore, we account for attenuation by interstellar dust and Lyman forest attenuation
by the intervening medium between the redshifted galaxy and the observer following
Madau (1995). The combination of these steps, described in this section, leads to observables that are similar to the real observations that we model with stellar population
synthesis codes.
First, we focus on the computation of intrinsic (i.e., unattenuated) magnitudes from
the stellar component. Each of the stellar particles is treated as a single stellar population characterized by its mass, age, and metallicity. We choose to use the Salpeter
(1955) IMF, as was done in previous observational work (e.g. Förster Schreiber et al.
2004; Wuyts et al. 2007). We then interpolate the corresponding luminosity for each
stellar particle from a grid of SSP templates with different ages and metallicities from
the stellar population synthesis code by Bruzual & Charlot (2003, hereafter BC03). Figure 6.1(d) illustrates the evolution of the intrinsic rest-frame V-band luminosity for one
of the simulations.
For the AGN emission, we scale a template SED by the bolometric black hole luminosity given by the simulation. The template SED was derived from the optically
blue (i.e., unreddened) quasar sample by Richards et al. (2006) with locally attenuated
light being reprocessed as an IR bump longward of λ > 1 µm. A full discussion of the
AGN template is presented by Hopkins, Richards,& Hernquist (2007). In most of our
analysis, we will consider the stellar light only. §6.4.4 addresses the impact AGN can
have on the outcome of SED modeling during the brief period when its contribution to
the total light is significant.
Galaxies, certainly in their actively star-forming phases, are not devoid of gas and
dust. It is therefore crucial to account for the obscuring and reddening effect dust
has on the stellar and AGN emission. We compute the optical depth along the line
of sight toward each stellar particle. To do so, we compute the local gas density on
a fine grid derived from the SPH formalism and the particle distribution (Hopkins
et al. 2005a) and integrate out from each particle along the line of sight to large distance. The simulations are based on the GADGET multi-phase ISM model developed
by Springel & Hernquist (2003). This model calculates the local mass fraction in the
hot (T = 105 − 107 K, diffuse, partially ionized) and cold (T = 103 K, molecular and
HI cloud core) phases of dense gas, assuming pressure equilibrium between the two
phases. Following Hopkins et al. (2005b), the attenuation along the line of sight is
then derived from the density of the hot-phase component only. The assumption that
most of the lines of sight only pass through the hot-phase component provides effec-
Section 6.2. The simulations
111
tively a lower limit on the optical depths. We use a gas-to-dust ratio equal to that of
the Milky Way, (A B / N HI ) MW = 8.47 × 10−22 cm2, with a linear scaling factor accounting
for gas metallicities deviating from solar: A B / N HI = (Z/0.02)(A B / N HI ) MW . As default,
we adopt the Calzetti et al. (2000) attenuation law for the wavelength dependence of
the optical depth. Changes in the synthetic photometry when adopting a SMC-like
or Milky Way-like attenuation law from Pei (1992) will be discussed in due time. The
computation of optical depths was repeated for 30 viewing angles, uniformly spaced
in solid angle d cos θ dφ. Figure 6.1(e) presents the distribution of effective visual extinction values (attenuated minus intrinsic V-band magnitude) as a function of time
since the merger. The extinction varies in the following way. In the early stages typical
extinction values are modest, with the exception for a few lines-of-sights were the disks
are seen edge-on. The overall extinction along all lines-of-sight reaches a peak during
the merger-triggered starburst and drops to very low values after star formation has
ceased.
Finally, in computing the observer-frame apparent magnitudes, we redshift the attenuated SED and convolve it with the same set of filtercurves that we have observations for in the Chandra Deep Field South (CDFS; Chapter 3). Here, we apply the
depression factors D A (z) and D B(z) given by Madau (1995) for the Lyman forest attenuation of the continuum between Lyα and Lyβ and between Lyβ and the Lyman limit
respectively. The flux blueward of the Lyman limit (λ L = 912Å) was set to 0, as is done
by the HYPERZ code (v1.1, Bolzonella et al. 2000) that we use for SED modeling.
In practice, it is computationally more convenient to interpolate the apparent magnitudes in a given passband for all stellar particles on a precompiled grid of BC03
apparent magnitudes at the redshift of interest. The internal dust attenuation is then
applied using the value of the Calzetti et al. (2000) attenuation law at the effective
wavelength for that passband. We tested that this method, as opposed to attenuating
the full resolution BC03 spectrum and then convolving with the filtercurve, leads to
photometric differences of at most a few percent.
We note that we never attempt to separate the light into the contribution from the
two progenitors. Instead, we always study the total photometry, as if the merging
system were unresolved.
6.2.3 The colors and SEDs of simulated and observed galaxies
Prior to analyzing the performance of our SED modeling procedure, it is important to
confirm that the simulated galaxies have spectral shapes resembling those of real highredshift galaxies in observed deep fields, thus validating their role as test objects. To
this end, we indicate the binned color distribution of simulated galaxies, viewed from
different angles and during different phases of their evolution, in a rest-frame U − V
versus V − J color-color diagram. Labbé et al. (2005) first introduced the observedframe equivalent of this diagram to illustrate the wide range of galaxy types at high
redshift ranging from blue, relatively unobscured star-forming systems to dusty starbursts to quiescent red galaxies. Plus symbols show the location of observed galaxies
in the HDFS (Labbé et al. 2003), MS 1054–03 (Förster Schreiber et al. 2006), and the
CDFS (Chapter 3) selected by their photometric redshift (or spectroscopic when available) to lie in the same redshift range (1.5 < z < 3.0). We also applied a stellar mass
112
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
Figure 6.2 — Rest-frame U −
V versus V − J color-color diagram showing the binned color
distribution of the simulations
seen under different viewing angles and at different epochs.
Overplotted (plus symbols) are
the rest-frame colors of observed galaxies with M∗ > 1.4 ×
1010 M⊙ at 1.5 < z < 3 in
the HDFS, MS 1054–03, and the
CDFS. Observed galaxies with
matching colors are found for
all simulated galaxies. The reddest observed sources in U − V
and V − J are not reproduced
by the considered set of simulations. Rest-frame SEDs for
sources in regions 1-6 are displayed in Figure 6.3.
Figure 6.3 — Rest-frame SEDs of simulated galaxies in regions 1-6 of Figure 6.2. A darker intensity of
the binned representation indicates a larger density of simulated galaxies with that flux level. In each
panel, the central 68% interval of the age distribution of simulated galaxies in the respective region is
given. Overplotted (black dots) are the rest-frame broad-band SEDs of observed 1.5 < z < 3 galaxies with
M∗ > 1.4 × 1010 M⊙ in the HDFS, MS 1054–03, and the CDFS. A general agreement between observed
and simulated spectral shapes is observed, also outside the U-to-J range where the correspondence was
not imposed by selection.
Section 6.3. SED modeling: methodology
113
cut at M∗ > 1.4 × 1010 M⊙ for the observed sample; the lowest initial stellar mass for
the considered set of simulations. Here, we do not attempt to statistically compare
the two samples. The abundances of different types of galaxies as predicted from the
simulations will be addressed in Chapter 7. For our current purpose of analyzing the
effects from star formation history, dust, metallicity and AGN on SED modeling, it is
sufficient to note that there is a large overlap between the color-color space spanned
by the simulated and observed galaxies. However, the observed distribution extends
to redder colors by a few 0.1 mag, both in U − V and in V − J. Given the one-sided
nature of the different color spread, it is unlikely that this can be attributed to photometric uncertainties alone. Therefore, we caution that our results may not necessarily
be extrapolated to the reddest galaxies present in observed samples.
To ascertain that observed and simulated galaxies with similar U − V and V − J colors have similar SEDs over the whole spectral range, Figure 6.3 presents the rest-frame
SEDs of objects in region 1-6 of Figure 6.2. Again, the binned distribution represents
the simulations, with the grayscale indicating a larger number of objects. Overplotted
with black dots is the broad-band photometry of our observed sample within the same
region of color-color space, placed at the respective rest-frame wavelength. The SEDs
are normalized to the rest-frame V-band. By selection, the observed and simulated
photometry matches well at rest-frame U and J. In between the UV J filters, and outside the U-to-J range, no correspondence was imposed. The fact that the UV spectral
shape and the NIR tail of the observed and simulated SEDs show a general agreement,
is encouraging. We conclude that the simulated photometry can be adopted as a realistic input to our SED modeling procedure. The results of our analysis will be applicable
to observed galaxies with similar colors.
6.3 SED modeling: methodology
We characterize physical parameters such as stellar mass, stellar age, and dust attenuation by matching the observed-frame broad-band photometry to synthetic templates
from the stellar population synthesis code by BC03. We use the HYPERZ stellar population fitting code, version 1.1 (Bolzonella et al. 2000) and fit the SED twice: first fixing
the redshift to the true value (for which we computed the simulated photometry), next
adopting a photometric redshift estimate obtained from the EAZY version 0.5 photometric redshift code (Brammer et al. in preparation). In each case, the full B-to-8
µm SED, sampled with identical passbands as available for the GOODS-CDFS (B435 ,
V606 , i775 , z850 , J, H, Ks , [3.6 µm], [4.5 µm], [5.8 µm], [8.0 µm]), was fed to HYPERZ.
Random photometric uncertainties were assigned as to mimic real observations in the
CDFS, and fluxes in each band were perturbed accordingly. Precisely, for each of the
5400 SEDs corresponding to a simulated galaxy observed during a certain phase of its
evolution, placed at a certain redshift, and observed along a certain line-of-sight, we
compute 5 realizations of the SED by introducing a gaussian perturbation in all bands
with the amplitude derived from the depth of GOODS-CDFS observations in the respective bands. A minimum error of 0.08 mag was adopted for all bands, preventing
small errors from dominating the fit.
As in Wuyts et al. (2007), we selected the least χ2 solution out of three possible
114
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
star formation histories: a single stellar population (SSP) without dust, a constant star
formation (CSF) history with dust (A V varying from 0 to 4 in steps of 0.2), and an
exponentially declining star formation history with an e-folding timescale of 300 Myr
(τ 300) and identical range of A V values. Ages were constrained to be larger than 50
Myr, to prevent imporbably young ages, and smaller than the age of the universe at
the observed redshift. We used a Calzetti et al. (2000) attenuation law, and assumed
solar metallicity and a Salpeter (1955) IMF with lower and upper mass cut-offs 0.1M⊙
and 100M⊙ .
When refering to the age derived from SED modeling, we mean the age obtained
by integrating over the different ages of SSPs that build up the best-fit SFH, weighted
with their mass fraction. This measure aims to quantify the age of the bulk of the
stars. For an SSP, it equals the time passed since the single burst. For a CSF history,
it is essentially half the time passed since the onset of star formation. The τ 300 SFH
represents an intermediate case.
6.4 Results from SED modeling at fixed redshift
In order to isolate effects from star formation history (§6.4.1), dust attenuation (§6.4.2),
metallicity variations (§6.4.3), and AGN contribution (§6.4.4), we computed the photometry for each snapshot with and without attenuation, with and without AGN contribution, and using solar metallicity, or the metallicity as computed by the simulation
for each stellar particle. To each of these sets of SEDs, we applied the modeling described in §6.3. In §6.4.1 - §6.4.4, we build up the analysis step by step adding one
aspect at a time. The overall performance of the standard SED modeling applied to the
’full’ photometry, taking into account the effects of both attenuation, metallicity, and
AGN contribution as realistically as possible, is discussed in §6.4.5.
6.4.1 Impact of mismatch between true and template SFH
The contribution of massive O and B stars makes young stellar populations brighter
than older stellar populations, giving them more weight in the integrated SED. Consequently, the light-weighted stellar age will be younger than the mass-weighted stellar
age. This is always the case, but provided we have a template representing the correct SFH, it is possible to account for this effect and still find the correct age of the
bulk of the stars. Our three allowed SFHs are an SSP, where all stars formed in a single burst, a model with SFR ∝ e−t/τ with τ = 300 Myr, and a constant star formation
history. These are standard choices in analyses of distant galaxies. However, they do
not encompass a star formation history where the rate of star formation was lower
in the past than it is now, as is the case during first passage and during the actual
merger-triggered
starburst
(see Figure 6.1). In general, fitting a template SFH that has
dSFR dSFR
< dt true, the older population will be lost to some degree under the
dt
template
glare of newly formed stars, leading to an underestimate of the age. Since one tends to
count the young light only, mass will be underestimated as well. For the same reason,
models allowing for a secondary burst of star formation on top of an older stellar population were found to reveal larger total stellar masses, in particular for blue objects
(Papovich et al. 2006; Erb et al. 2006b; Wuyts et al. 2007).
Section 6.4. Results from SED modeling at fixed redshift
115
Figure 6.4 — Impact of star formation history. The difference between estimated and true (a) mass and
(b) mass-weighted age as a function of time for all simulations, with the SED modeling performed on
the intrinsic (i.e., unattenuated) stellar photometry with all stars set to solar metallicity. The solid line
indicates the median. The dotted lines contain the central 68% of the distribution. Deviations from 0
(negative indicating an underestimate) are due to mismatch between the actual star formation history
and the histories allowed in our SED modeling (SSP/CSF+dust/τ 300+dust). Maximal underestimates
of mass and age are reached during the merger itself. A secondary minimum is reached during first
passage of the progenitors, 0.3 to 0.4 Gyr before.
We demonstrate that the underestimate of mass and age takes place by considering the performance of our SED modeling procedure as applied on intrinsic stellar photometry with all stars set to solar metallicity. Here, we define ∆log(agew ) as
log(agew,recovered) − log(agew,simulation). Hereafter, similar definitions will be used to quantify the offset in mass, reddening and extinction, always indicating an underestimate
with a negative value of ∆. Figure 6.4 shows ∆log(agew ) as a function of time with respect to the merger between the supermassive black holes. We bin the distribution of
points for different initial conditions, timesteps and lines-of-sight. Darker intensities
represent a higher density in the bin. Empty boxes contain less than 1% of the total
number of SEDs at that timestep. The solid line represents the median of the distribution and the dotted curves mark the central 68% interval. During the first snapshot,
when the star formation history matches (by construction) our CSF template, we find
no systematic offset and a low scatter, purely resulting from photometric uncertainties.
Soon after, we start to underestimate the age and mass, with minima coinciding with
the moment of first passage (500 Myr before the actual merger) and that of the actual
merger-triggered starburst. It is precisely at these moments that the real SFH deviates
most from the allowed template SFHs. During the starburst phase itself, the median
offset of true mass-weighted age versus recovered age exceeds 0.5 dex, with a large
scatter due to differences in the SFH for different initial conditions. For example, the
ratio of SFR at first passage over SFR during the central starburst increases with gas
fraction. After all activity has quieted down, the derived ages and masses lie within
0.1 dex of their true value.
116
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
Figure 6.5 — The effective reddening (attenuated minus intrinsic B − V
color) versus total absorption in the Vband for all timesteps, viewing angles
and initial conditions. The intensity
of the binned distribution indicates the
number of simulations in the respective part of the diagram. A ratio of total to selective absorption R V = 4.05 as
by Calzetti et al. (2000) is plotted with
the thick dashed line. The dot-dashed
curve indicates a toy model where the
distribution of A V values is uniform between 0 and a maximum value, and all
emitting sources are identical. Stellar
particles individually have R V = 4.05,
but in the case of a non-uniform dust
distribution the sum of all stellar particles has an effective R V > 4.05.
6.4.2 Impact of attenuation
As described in §6.3, we use the approach of a foreground screen to account for the
attenuation by dust in our SED modeling. Figure 6.1(e), illustrating the range of effective visual extinction values (attenuated minus intrinsic V-band magnitude) for a
random simulation depending on the viewing angle, proves that such a representation
is not valid. Here we address the impact that a non-uniform distribution of the dust
will have when modeled by a foreground screen.
First, we consider a situation where the optical depth to the stellar particles is not
constant, but the variations are uncorrelated with the intrinsic properties of the stellar
particles. Such a scenario is by construction the case at the start of the simulation. For
each stellar particle individually the ratio of total to selective absorption,
RV =
AV
= 4.05,
E(B − V)
(6.4)
was taken from Calzetti et al. (2000). Since less extincted regions are also less reddened
and have a larger weight in the integrated SED, the effective extinction A V,e f f ≡ VAtt −
VInt and effective reddening E(B − V)e f f ≡ (B − V) Att − (B − V) Int of the galaxy as a
whole will not be related by the same factor 4.05 as for the individual particles. Instead,
the overall reddening for a given A V will be smaller than predicted by Calzetti (i.e., the
extinction is greyer). This is illustrated in Figure 6.5 where the dashed line represents
the A V = 4.05 × E(B − V) scaling by Calzetti et al. (2000) and the dotted line represents
a toy model with a uniform distribution of A V values between 0 and A Vmax to stellar
particles that all emit at identical intrinsic luminosities:
A V,e f f
10−0.4AVmax − 1
= −2.5 log
−0.4AVmaxln(10)
(6.5)
Section 6.4. Results from SED modeling at fixed redshift
117
Figure 6.6 — Effective extinction curves of simulated galaxies with A V,e f f > 1 for different input attenuation laws: (a) the Calzetti et al. (2000) law, MW-like reddening from Pei (1992), and (c) SMC-like
reddening from Pei (1992). The black curve indicates the median over all snapshots and viewing angles
with A V,e f f > 1. The light grey polygon indicates the central 68% interval. The Calzetti, MW, and SMC
attenuation laws are plotted in grey. In all cases, the effective extinction of simulated galaxies with large
A V,e f f is greyer than the Calzetti et al. (2000) law that is used in standard SED modeling. The offset is
smallest when each stellar particle is attenuated according to the SMC-like law.
E(B − V)e f f = 2.5 log
"
10−0.4AVmax − 1
1
10−0.4(1+ 4.05 ) AVmax − 1
1 + 4.105
#
.
(6.6)
Since the Calzetti et al. (2000) attenuation law was derived empirically for galaxies
as a whole, it is arguably not the appropriate law to apply to the individual stellar
particles, i.e., the smallest stellar populations that our simulation can resolve, typically
105 − 106 M⊙ . We investigated the changes in photometry when adopting a MW and
SMC-like reddening curve by Pei (1992), which were derived in a more bottom-up
fashion from the physics of interstellar dust grains. Again, we scaled the optical depth
with the metallicity along the line of sight. For the SMC reddening curve, the resulting
colors become redder by up to 0.05, 0.1, and 0.2 mag in rest-frame B − V, U − V, and
V − J respectively. The MW-like attenuation law is also less grey than Calzetti, thus
producing slightly redder colors, though less so than for the SMC law. The effective
extinction curve, expressed as E(BA−λ V) as a function of wavelength, of snapshots and
viewing angles with large optical depths (A V,e f f > 1) is presented for different input
attenuation laws in Figure 6.6.
λ
Not only does non-uniform extinction change the reddening ( dA
) at a given A V ,
dλ
d2 A λ
it also affects the dependence of the reddening on wavelength ( dλ2 ). For extinction
that is uncorrelated to the properties of the emitting sources, this gives the dust vector in the U − V versus V − J color-color diagram a shallower slope, i.e., for a given
reddening in V − J, the reddening in U − V is smaller than predicted by the Calzetti
2
et al. (2000) law. The consequence of a different ddλA2λ than Calzetti is clarified in Figure
6.7. The solid line represents the evolutionary track of a stellar population following a
CSF history. The track starts 50 Myr after the onset of star formation and ends 2 Gyr
later. Suppose different parts of a galaxy all contain a 1 Gyr old CSF population whose
118
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
Figure 6.7 — Rest-frame U − V versus
V − J color-color diagram illustrating
the effect of a non-uniform distribution
of A V values that is uncorrelated with
the intrinsic properties of the emitting
sources. The black curve indicates a
CSF population with age between 50
Myr and 2 Gyr. Suppose an intrinsic population (box symbol) is reddened
by such a dust distribution to the location in color-color space of the filled
circle. Under the assumption of a uniform foreground screen of dust, the observed colors will then be traced back
along the Calzetti et al. (2000) reddening vector (dashed black line), resulting
in an artificially young age (star symbol).
intrinsic location in color-color space is marked by the empty grey box. A distribution
of dust as described by the above mentioned toy model will redden the galaxy along
the dotted line. Interpreting the observed colors (filled circle) as a CSF population attenuated by a foreground screen according to the Calzetti et al. (2000) law, will lead to
a best-fit age (star symbol) that is too young and reddening that is too large.
Since in our simulations the ages of the stellar particles (that are each treated as
SSPs) present at the start of the simulation were drawn randomly from a uniform distribution, the system has a CSF history in the earliest snapshots without a correlation
between the optical depth and intrinsic light of the stellar particles. Therefore, it comes
as no surprise that, when looking at the attenuated stellar photometry in Figure 6.8 (for
now all stars still set to solar metallicity), the central 68% interval in ∆ log agew reaches
to more negative values (to -0.5 dex) during the earliest phases than was the case for
the unattenuated photometry (Figure 6.4). The estimated reddening is slightly larger
than the true value, but nevertheless the use of Eq. 6.4 still causes an underestimated
A V , as can be understood from Figure 6.5. The systematic underestimate in age and
A V combined cause the evaluation of the stellar mass during the first snapshots, when
template mismatch due to the SFH is still negligible, to be too small by ∼ 0.12 dex.
After a few 100 Myr after the beginning of the simulation however, Figure 6.8 reveals an improved recovery of the mass-weighted stellar age compared to that obtained by SED modeling of the intrinsic light (Figure 6.4). Clearly, the assumption of a
non-uniform dust distribution that is uncorrelated with the intrinsic properties of the
emitting sources breaks down.
Figure 6.9 demonstrates the occurence of preferential extinction toward young star
forming regions in one of our simulations. The three panels indicate the binned distribution of the metallicity-scaled hydrogen column density measured along various
lines-of-sight versus the age of the stellar particle to which the column density was
computed for the 3 epochs marked in the star formation history panel of Figure 6.1.
Section 6.4. Results from SED modeling at fixed redshift
119
Figure 6.8 — Effect of extinction. The difference between estimated and true (a) stellar mass, (b) massweighted age, (c) effective reddening, and (d) effective visual extinction as a function of time since
the merger. The SED modeling was performed on the attenuated stellar photometry with all stars set
to solar metallicity. The solid line indicates the median and dotted lines comprise the central 68% of
the distribution. Ages are still underestimated for the first 0.8 Gyr of the evolution, but to a lesser
degree than estimates based on the intrinsic light. Added to the underestimated A V , this leads to a
characterization of the stellar mass that is too low by 0.1 - 0.15 dex.
The vertical arrow indicates the start of the simulation. All stellar ages older than this
value (cut off for illustrational purposes) were set by hand as explained in §6.2.1. As we
already pointed out in the A V history panel of Figure 6.1, the typical column densities
are higher during the merger (panel b) than before (a) or after (c). Moreover, Figure 6.9
shows that the ratio of column densities toward ongoing star formation over column
densities toward older populations reaches a maximum during the merger (b). Using
sticky particle simulations of dusty starburst mergers, Bekki & Shioya (2001) found
a similar age-dependent extinction, confirming that this is a generic feature of merging systems and not determined by the method used to model dissipative processes.
120
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
Figure 6.9 — Distribution of hydrogen column densities, linearly scaled with the
metallicity of the gas along different linesof-sight to the stellar particles versus the
age of the respective stellar particle. The relation between column density and stellar
age is plotted for 3 snapshots: before (2),
during (9), and after (16) the merger (see
Figure 6.1). The solid and dotted lines indicate the median and 68% interval of the
distribution respectively. Darker intensity
means a larger number of stars is present
with that age. All stellar ages rightward
of the arrow correspond to initial stars and
were set by hand. The optical depth, which
is proportional to the metallicity-scaled gas
density, is larger toward newly formed stars
during the merger-triggered starburst. The
signature of this age-dependent extinction
weakens during more quiescent episodes of
star formation.
Poggianti & Wu (2000) inferred age-dependent extinction during a starburst to explain
the nature of so-called e(a) galaxies: galaxies with [OII] in emission and strong Balmer
absorption lines, frequently associated with merger morphologies.
From a physical perspective, it is expected that during the merging process hydrodynamical and gravitational forces channel gas and dust to the central regions where
it triggers a starburst. Once started, supernovae going of on a few 107 yr timescale
further increase the dust content of the regions where newly formed stars reside. The
fact that the distribution of younger (and thus intrinsically bluer) stellar populations
does not trace that of the older populations of stars and that it is intimately correlated
with the dust distribution leads to an overestimate in age. In analogy to Figure 6.7, the
2
effective ddλA2λ is such that the galaxy is reddened along a steeper vector in the U − V
versus V − J diagram than is the case for the Calzetti curve. Since an observer will
mistakenly model the galaxy with an intrinsically redder template, the reddening by
dust E(B − V) will be underestimated. Although a given total absorption corresponds
to a stronger reddening in the presence of age-dependent extinction compared to uncorrelated non-uniform extinction, the Calzetti et al. (2000) relation between E(B − V)
and A V given by Eq. 6.4 can still be considered as an upper limit. Therefore, the total absorption will be underestimated. This is illustrated in Figure 6.5 where we plot
the binned distribution of true E(B − V) versus true A V for all of our simulation snapshots, viewed under a range of viewing angles. Finally, the derived stellar mass owes
its more robust character to the compensating effects of systematic offsets in age and
absorption.
The effect of the larger extinction toward young stars will in practice be superposed
on the effect of mismatch between template and true SFH, that prevents us from fully
accounting for the difference between light- and mass-weighted stellar age (see §6.4.1).
Section 6.4. Results from SED modeling at fixed redshift
121
Figure 6.10 — Rest-frame V-band
light-weighted age versus massweighted age for an initially 40% gas
fraction simulation. The boxes mark
the mean age weighted with the attenuated V-band light. Darker intensities
indicate a larger number of viewing
angles. The solid and dotted curves
mark the median and central 68%
interval respectively. The dashed line
indicates the mean age weighted with
the intrinsic V-band light (no attenuation). The attenuated light-weighted
age is a better approximation of the
mass-weighted age than the intrinsic
light-weighted age, increasingly so for
younger stellar populations. Larger
optical depths to young than to old
stars are responsible for this effect.
Figure 6.10 illustrates how an increased extinction toward young stars reduces the difference between the light-weighted and mass-weighted measure of age. We conclude
that the SED modeling on galaxies with solar metallicity stars and dust distributed in
between still underestimates the age, but adding dust has improved our best guess to
an overall median offset of -0.04 dex (compare Figure 6.8(b) to Figure 6.4(b)). Similar conclusions were drawn by Bell & de Jong (2001) who examine the reddening and
dimming effects of dust and its impact on estimating stellar mass-to-light ratios.
6.4.3 Impact of stellar metallicity
So far, we tested our SED modeling on synthetic photometry that was computed assuming a solar metallicity for all emitting sources. In reality, stars with a range of
metallicities will be present, reflecting the level of enrichment in the gas at the epoch
of their formation. Before we repeat our analysis now setting the stellar metallicities
to their appropriate value calculated by the GADGET-2 code, we anticipate the effect
using the diagnostic U − V versus V − J color-color diagram in Figure 6.11.
The tracks represent exponentially declining SFHs for metallicities of Z=0.008 (grey)
and 0.02 (solar, in black). Both evolutionary tracks are drawn from 50 Myr to 2 Gyr after the onset of star formation. The classic age-metallicity degeneracy states that the
optical broad-band colors of a young stellar population are nearly indistinguishable
from that of an older, more metal-poor population (O’Connell 1986). For the τ 300 star
formation history drawn here, this effect gets only notable at later times: 2 Gyr after
the onset of star formation the sub-solar metallicity track has the same U − V color as
a solar metallicity population that started forming stars 1.8 Gyr ago. On the one hand,
the addition of dust will complicate the age-metallicity degeneracy. On the other hand,
the addition of NIR photometry helps to separate the evolutionary tracks for different
metallicities. A galaxy whose attenuated light has colors marked by the filled circle
may correspond with one of the intrinsic colors indicated by the grey boxes depend-
122
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
Figure 6.11 — Rest-frame U − V versus V − J color-color diagram illustrating the effect of fitting solar metallicity templates to stellar populations
of sub-solar metallicity.
The black
and grey curves represent evolutionary tracks for an exponentially declining star formation history with efolding time of 300 Myr for solar (Z =
0.02) and sub-solar (Z = 0.008) metallicity respectively, each starting at 50
Myr. Stellar population with intrinsic colors indicated by the grey boxes
will be reddened to the location in
color-space marked by the filled circle in the case of (a) non-uniform ageindependent extinction, (b) extinction
by a uniform foreground screen, and (c)
age-dependent extinction. In all three
cases, the assumption of solar metallicity and Calzetti attenuation will lead to
the conclusion that the stellar population formed its first stars 0.5 Gyr ago.
This is an underestimate (a) or overestimate (b, c) respectively. The reddening
is always underestimated.
ing on the kind of extinction: (a) for non-uniform age-independent extinction, (b) for
a foreground screen of dust, and (c) for age-dependent extinction. In case (a), the assumption of Calzetti attenuation and solar metallicity in our SED modeling leads to a
recovered evolutionary stage that is too young, marked with the star symbol on the
solar metallicity track. In case (b) and (c), the same recovered evolutionary stage is
too old. In all cases, the determination of the reddening will be too low, as will consequently be the case for the A V and the stellar mass, and increasingly so for lower
metallicities. Obviously, the effects described will again be superposed on the previously discussed effects of star formation history and dust. It is also noteworthy that
following the enrichment by heavy elements reduces the effect of age-dependent extinction. Young stellar populations are still intrinsically bluer than old populations,
but to a lesser degree since they have formed at later times from gas that was more
enriched.
In our recovery analysis of stellar population properties, we find that at metallicities
of a quarter solar and below, the age is overestimated by 0 to 0.5 dex (central 68% interval of ∆ log agew ). However, the underestimate in reddening and therefore extinction
for these low-metallicity galaxies is such that the mass estimate (which is dependent
on both age and A V ) stays within ±0.1 dex of its true value for 68% of the cases.
6.4.4 Impact of AGN contribution
Since the merger simulations described in this chapter take into account the role of
supermassive black holes on its environment (see e.g. Di Matteo et al. 2005; Springel
Section 6.4. Results from SED modeling at fixed redshift
123
Figure 6.12 — Attenuated spectrum of
a simulated merger placed at z = 2.1
during the peak of AGN activity. The
total attenuated light (black solid line) is
decomposed into a contribution from
stars (dotted line) and AGN (dashed
line). An observer who samples the
total attenuated light with an identical set of broad-band filters as available for GOODS-CDFS and models the
SED using stellar population synthesis
only, will find as best-fitting model the
spectrum in grey. Its age is too young
by ∆ log agew = −0.4. The reddening
E(B − V) and absorption A V are overestimated by 0.1 and 0.4 mag respectively.
The opposite sign of offsets in age and
A V leads to a mass recovery that is only
0.05 dex below its true value.
et al. 2005b), it is straightforward to include its contribution to the integrated galaxy
SED. We note that only during a timespan of the order of a Salpeter time, a few 107
to 108 year, the AGN emission amounts to a significant fraction of the stellar emission.
Admittedly, the peak of AGN activity can be missed by the time sampling of our snapshots. Nevertheless, the current dataset provides a useful insight on its impact on the
SED modeling.
We illustrate the typical behavior in Figure 6.12 showing the photometry computed
at the time of merging when the accretion onto the supermassive black hole is maximal.
Here, the solid black curve represents the light received by an observer. We break down
the attenuated SED in a stellar (dotted line) and AGN contribution (dashed line). Finally,
the best-fit model (in this case an exponentially declining star formation history that
started 0.8 Gyr ago) is plotted in grey. Although resulting in a low χ2reduced ∼ 1, the SED
modeling is mislead by a degeneracy between the stellar+AGN light and the stellar
light of a younger population obscured by large columns of dust. The addition of
AGN light, when exceeding 10% of the total emission, adds another -0.1 to -0.15 dex to
∆ log agew , +0.05 to +0.1 mag to ∆E(B − V), and +0.3 to +0.5 mag to ∆A V . These cases
typically show a larger χ2reduced (70% have χ2reduced > 5).
6.4.5 Overall performance
Our analysis was performed on synthetic photometry of galaxies placed at redshifts
z = 1.5 to z = 2.9. The results do not show a trend with redshift. This indicates that
the details of the filterset play no role. Our study only focuses on continuum shape
and in the presence of spectral lines, a higher wavelength sampling in the form of
spectroscopic studies will obviously provide valuable extra information. The trends
described in this section are all systematic and cannot be attributed to signal-to-noise
effects (e.g. more extincted galaxies at the highest redshifts being fainter and therefore
less well recovered). We tested this in two ways: first by omitting the perturbation of
124
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
Figure 6.13 — Overall performance of the SED modeling. Recovered versus true (a) ratio of current to
final stellar mass, (b) mass-weighted stellar age, (c) effective reddening (i.e., attenuated minus intrinsic
B − V color), and (d) effective visual extinction (i.e., attenuated minus intrinsic V-band magnitude).
The SED modeling was performed on the total (stellar+AGN) attenuated photometry. The solid line
indicates the median and dotted lines comprise the central 68% of the distribution. The total visual
extinction A V is the least constrained of the four studied parameters. In particular for heavily extincted
galaxies the A V is greatly underestimated.
the synthetic fluxes, second by applying a conservative cut in the observed Ks -band
magnitude: Ks,obs < 23.6, corresponding to S/ NKs > 10. In both cases, the same trends
described in this Section are still present.
The combined effects of mismatch in SFH, attenuation by dust, metallicity variations and AGN activity on our ability to characterize the mass, age, reddening and extinction of a galaxy is summarized in Figure 6.13. Figure 6.14 presents the performance
of the SED modeling on the full photometry (including dust, metallicity variations, and
Section 6.4. Results from SED modeling at fixed redshift
125
Figure 6.14 — Overall performance of the SED modeling. The difference between estimated and true
(a) stellar mass, (b) mass-weighted age, (c) effective reddening, and (d) effective visual extinction as
a function of time since the merger. The SED modeling was performed on the total (stellar+AGN)
attenuated photometry. The solid line indicates the median and dotted lines comprise the central 68%
of the distribution. The properties of merger remnants are well reproduced. The results for star-forming
galaxies, especially for those in the phase of merging, show underestimates in both age, extinction, and
mass.
AGN) in a different manner, as a function of time since the merger. Figure 6.13(a) compares the recovered and true stellar mass, normalized to the final stellar mass of the
simulation. At low M/ M f inal ratios, i.e., at the start of the high gas fraction simulations, the mass estimates agree well with the true values. The largest systematic underestimates occur at intermediate M/ M f inal , during the merger-triggered star-forming
phases of the simulation. Finally, the correspondence is best at log(M/ M f inal ) ∼ 0,
where the merger remnants reside. The same scenario is visible in Figure 6.14(a). We
quantify the performance of the SED modeling separately for galaxies in the ’disk’,
’merger’, and ’spheroid’ regime by averaging the solid (median ∆ log M as a function
126
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
of time) and dotted (central 68% interval) lines over the respective time interval in+0.06
+0.09
dicated in Figure 6.14(a). We find ∆ log Mdisk = −0.05−
0.13 , ∆ log Mmerger = −0.11−0.14 ,
+0.04
and ∆ log Mspheroid = −0.01−
0.09 . The errors indicate the range around the typical offset
comprising 68% of the simulations.
Quantifying the quality of age estimates (Figure 6.13(b), Figure 6.14(b)), we find
+0.10
+0.34
+0.26
∆ log agew,disk = −0.04−
0.27 , ∆ log age w,merger = −0.11−0.25, ∆ log age w,spheroid = −0.03−0.10 .
Again, the underestimate and scatter is largest for the phases of merger-triggered star
formation. The overestimate of the youngest ages shown in Figure 6.13(b) concerns the
80% gas fraction simulations in their earliest phases, when the metallicity is low and
the results from §6.4.3 apply.
The reddening (Figure 6.13(c), Figure 6.14(c)) is overall well reproduced: ∆E(B −
+0.12
+0.09
+0.08
V)disk = −0.02−
0.07 , ∆E(B − V)merger = −0.01−0.09, and ∆E(B − V)spheroid = −0.04−0.04 . Only
at the highest reddening levels, the agreement deteriorates. The latter correspond to
the times when and viewing angles under which the effect of increased extinction toward young stars is maximal (see §6.4.2).
As opposed to the reddening, however, the extinction (Figure 6.13(d), Figure 6.14(d))
shows large systematic underestimates, in particular during the star-forming (disk and
merger) phases. Using an RV of 4.05 to translate the selective absorption E(B − V)
into a total visual absorption A V results in an underestimate over the whole range
of A V values, particularly during the highly obscured phases. The sum of emitting
sources that are each attenuated according to Calzetti et al. (2000) does not follow
that same reddening law. An observer is limited by the light that he/she receives.
+0.42
+0.27
These results are quantified as follows: ∆A V,disk = −0.35−
0.29, ∆A V,merger = −0.48−0.45 ,
+0.25
and ∆A V,spheroid = −0.27−0.21.
6.4.6 Lessons for SED modeling
How can the modeling of real high-redshift galaxies benefit from our analysis of merger
simulations? In order to answer this question, we need to translate the mechanisms described above in terms of the physical properties of stars and dust into their combined
effect as a function of observables, such as color.
Since the results of our SED modeling showed no trend with redshift for which the
synthetic photometry was computed, we can describe the quality of recovering stellar
population properties in terms of rest-frame colors. Although not true observables,
their computation as described by Rudnick et al. (2003) suffers from only a minor
template dependence compared to parameters such as mass, age, and dust content. In
Figure 6.15, we present the performance of our SED modeling as a function of location
in the rest-frame U − V versus V − J color-color diagram. Simulations with initial
gas fractions of 40% and 80% are shown separately. The SED modeling was based
on the full (stellar+AGN) photometry in both cases. Downward triangles indicate a
median value of ∆parameter that is negative for simulations with the respective colors.
Upward triangles represent an overprediction of the true value. Lighter symbols are
used for a better correspondence between the true and modeled parameter value. The
different initial stellar ages and lower metallicities for f gas = 80% runs explain why
they extend to bluer U − V colors. At blue optical colors (U − V < 1), the attenuation is
seriously underestimated. An overestimate of the age by a similar order of magnitude
Section 6.5. Results from SED modeling with free redshift
127
Figure 6.15 — Median quality of recovered stellar population properties for simulations with a gas fraction of 40% and 80% as a function of rest-frame optical and optical-to-NIR color. Downward triangles
indicate underestimates with respect to the true value. Upward triangles mark overpredictions. Galaxies in regions with white triangles are characterized most accurately. Simulations with f gas = 80% reach
bluer colors in (U − V)rest than those with f gas = 40% since their initial stars were set to younger ages (see
§6.2.1). At red optical colors [(U − V)rest > 1], galaxies with relatively blue (V − J)rest colors are better
recovered than those at the red (V − J)rest end. The former are older, less obscured systems, while the
latter have a young and dusty nature. At blue optical colors [(U − V)rest < 1], large systematics in the
determination of age and A V occur. Their opposite signs cancel out in the derivation of stellar mass.
(at U − V < 0.65 mostly due to the lower metallicity and at 0.65 < U − V < 1 due to
age-dependent extinction) leads to a relatively robust estimate of the mass.
At red optical colors (U − V > 1), sources with relatively blue V − J colors are generally better modeled than those with the reddest V − J colors in our sample. In the
1 < U − V < 1.8 color regime, the effects of dust attenuation play an important role.
Here we find objects that are heavily extincted during the merger-triggered starburst,
but also disk galaxies seen edge-on during the earliest phases of the simulation. All
sources with V − J > 1.5 in our sample belong the latter category. Since we did not
impose an age gradient for the initial stars, the dust distribution in thoses cases is nonuniform but uncorrelated with intrinsic color of the stellar particles. This explains why
they have the strongest overestimate in E(B − V). The upper part of the color-color distribution (U − V > 1.8) is where each galaxy ends up by aging without further inflow
of gas. In such a system, the A V values are modest to low and the lack of a template
exactly matching the SFH is less problematic as the epoch of major star formation lies
further back in time.
6.5 Results from SED modeling with free redshift
In practice, complete spectroscopic surveys of mass-limited samples of high-redshift
galaxies are rare. Consequently, we often are not able to fix the redshift in the SED
128
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
modeling procedure to its exact true value. Over the past few years, several codes
have improved on estimating the redshift based on the broad-band SED, by experimenting with different template sets and fitting algorithms. In this section, we will use
the new photometric redshift code EAZY (Brammer et al. in preparation) to establish
the quality of photometric redshift (z phot) estimates derived from our synthetic photometry and analyze the impact of z phot uncertainties on the derived stellar population
properties.
6.5.1 The photometric redshift code EAZY
Here, we summarize the main characteristics of the EAZY photometric redshift code.
A full description of the algorithm and template sets will be presented by Brammer et
al. (in preparation). We test our ability to recover the redshifts using two template sets.
The first template set is based on the same BC03 code from which the broad-band photometry of our simulations was derived (§ 6.2.2). This set consists of 10 SSP templates
with a Salpeter (1955) IMF, solar metallicity, and ages logarithmically spaced between
50 Myr and 10 Gyr. Each template is allowed to have an attenuation of A V =0.0, 0.1, 0.3,
or 0.6, applied according to the Calzetti et al. (2000) law. We fit a non-negative superposition of what are effectively 40 different templates to the B-to-Ks band photometry,
using a template error function that effectively downweights the rest-frame UV portion of the templates in the fit. The maximum likelihood solution is then adopted as
best z phot estimate.
The second setup relies on a set of hundreds of PEGASE 2.0 (Fioc & Rocca-Volmerange
1997) templates, closely matching those described by Grazian et al. (2006). The template set includes Calzetti et al. (2000) reddened CSF models to account for the presence of dusty star-forming sources. Since it is completely independent from the photometry of our simulations, it provides a check of robustness against choice of stellar
population synthesis code. Given the large number of templates, we now match each
template individually to the B-to-Ks band photometry, using the same template error
function as for the BC03 templates.
EAZY allows for the use of a magnitude prior function, constructed from observed
or simulated number counts as a function of apparent magnitude and redshift. However, since we shift each simulation over the entire redshift range 1.5 ≤ z ≤ 2.9, our
methodology does not allow to test this feature.
6.5.2 Recovering redshifts and stellar population properties from broad-band photometry
6.5.2.1 Recovering redshifts
In Figure 6.16, we compare the photometric redshifts obtained with EAZY+BC03 and
EAZY+Pegase with the input redshifts for which the synthetic broad-band SEDs were
computed. The measure commonly used to quantify the photometric redshift quality is
∆z/(1 + z). Its normalized median absolute deviation is 0.030 using the BC03 template
set and 0.054 using the Pegase template set. Even when using the Pegase templates,
whose population synthesis is largely independent of the input photometry, the performance of EAZY is very good and competitive with that of other codes presented in
Section 6.5. Results from SED modeling with free redshift
129
Figure 6.16 — Comparison of photometric redshifts z phot by EAZY versus true redshift for a template
set based on (a) the BC03 and (b) the Pegase stellar population synthesis code. The correspondence is
good in both cases (σ N MAD (∆z/(1 + z)) < 0.06), without significant systematic offsets. The lower scatter
for EAZY+BC03 with respect to EAZY+Pegase is likely due to the fact that the input photometry is
essentially a function of BC03 single stellar population templates.
the literature.
We find a correlation between ∆z/(1 + z) and the time since (or before) the merger,
with opposite sign for the two template sets. For EAZY+BC03, the median ∆z/(1 + z)
increases from -0.01 at the start of the simulation to +0.01 at the end 2 Gyr later. For
EAZY+Pegase, the median ∆z/(1 + z) drops from +0.02 to -0.02 over the same timespan. The correlation is with time relative to the merger (i.e., phase in star formation
history) and not the stellar age. We can tell this by considering separately the 40% and
80% gas fraction simulations, which differ by their initial conditions in mass-weighted
age, but show a similar dependence of the z phot accuracy on time relative to the merger.
No trend with A V was found except for the reflection of the correlation with evolutionary time, i.e., the attenuation history is correlated with the star formation history (see
Figure 6.1).
This exercise offers a valuable complementary test to the empirical comparison with
spectroscopic samples of high-redshift galaxies. The latter are direct measurements
and therefore insensitive to our knowledge of stellar tracks and population synthesis.
On the other hand, spectroscopic samples of high-redshift galaxies often suffer from
selection biases, especially against galaxies lacking emission lines.
6.5.2.2 Impact of z phot uncertainties
Having quantified the quality of photometric redshifts, we now repeat the SED modeling fixing the redshift to its best-fit value (EAZY+BC03). The same mechanisms as
discussed in §6.4 are still influencing the recovery of physical parameters. Given the
130
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
partially random nature of the z phot uncertainties, it comes as no suprise that the central
68% interval broadens with respect to the SED modeling at fixed redshift. Averaged
over time, the broadening amounts to 25%, 10%, 11%, and 15% in ∆ log M, ∆ log agew ,
∆E(B − V)e f f , and ∆A V,e f f respectively.
The systematic part of the z phot uncertainty translates into additional but small systematic offsets in the stellar population properties. Qualitatively, when a source is
mistakenly placed at higher redshift, a larger mass estimate and lower dust reddening are required to match the observed SED. In numbers, the median of log M f ixz −
log MEAZY+ BC03 evolves from -0.015 to +0.01 over the timespan of the simulation. A V, f ixz −
A V, EAZY+ BC03 evolves from ∼ 0 to -0.1 mag, and E(B − V) f ixz − E(B − V)EAZY+ BC03 from
+0.01 to -0.02 mag. We find no systematic propagation of the z phot uncertainty in the
stellar age. Note that the impact of z phot uncertainties is highly dependent on the template set used, e.g., the systematic dependence on time has the opposite sign when
using Pegase templates.
6.6 Summary
We analyzed the performance of a simple SED modeling procedure applied to synthetic optical-to-NIR broad-band SEDs of merger simulations placed at redshifts z=1.5,
1.7, ..., 2.9. First, we modeled the SEDs assuming the redshift was known. The masses,
ages, E(B − V), and A V of simulated ellipticals are very well reproduced, with an av+0.08
+0.10
+0.04
erage value of ∆ log M = −0.01−
0.09 , ∆ log age w = −0.03−0.10 , ∆E(B − V) = −0.04−0.04 ,
+0.25
and ∆A V = −0.27−
0.21 . Here the errors indicate the central 68% interval of the distribution of ∆parameter values of all the simulations (different masses, gas fractions,
viewing angles) in the spheroid regime. In earlier, actively star-forming, phases, the
scatter in recovered stellar population properties with respect to the true value increases, and larger systematic underestimates of age, mass, and extinction occur. This
is particularly the case for the simulation snapshots of phases with merger-triggered
star formation, where we find the following offsets and scatter (averaged over the
+0.34
+0.09
merger regime indicated in Figure 6.1): ∆ log M = −0.11−
0.14 , ∆ log age w = −0.11−0.25 ,
+0.09
+0.42
∆E(B − V) = −0.01−0.09, and ∆A V = −0.48−0.45. The SED modeling performs better
on regular star-forming disks than on galaxies during the merging event. Compared
to spheroids however, the results of the SED modeling on disks show a larger scat+0.06
+0.26
ter and larger systematic underestimates: ∆ log M = −0.05−
0.13, ∆ log age w = −0.04−0.27 ,
+0.27
+0.12
∆E(B − V) = −0.02−
0.07 , and ∆A V = −0.35−0.29 .
By adding the effects of dust attenuation, metallicity variations and AGN step by
step to the basic intrinsic photometry, we were able to disentangle the different mechanisms at play and their impact on the estimation of the mass, age, reddening, and
extinction of the galaxy. The qualitative impact on the SED modeling results by different aspects of the galaxy content is summarized in Table 6.1.
A mismatch between the real SFH and the allowed template SFHs leads to an inability to account for the difference between light-weighted and mass-weighted properties
such as stellar age and mass. If the optical depth toward intrinsically bluer emitting
sources is larger than to intrinsically redder stellar populations, the netto effect of the
age and mass underestimate due to mismatch between true and template SFH is less
Section 6.6. Summary
131
severe. We find proof of such an increased extinction toward younger stars during
the merger-triggered starburst. Applying the Calzetti et al. (2000) reddening law toward each stellar particle, we find that the overall reddening for a given A V is less than
predicted by the Calzetti et al. (2000) law, particularly when the optical depth is uncorrelated with the intrinsic colors of the sources it is hiding. In the latter case, the dust
vector has a shallower slope in the U − V versus V − J color-color diagram than the
Calzetti et al. (2000) vector. In the case of larger optical depths toward young (blue)
stellar populations, there is relatively more reddening in U − V for a given reddening
in V − J. Applying a MW or SMC-like attenuation law to the individual stellar particles in the simulation increases the reddening, but the effective extinction is still greyer
than the Calzetti et al. (2000) law.
All other properties remaining the same, the effect of applying our SED modeling
to stellar populations with sub-solar metallicities is that one would underpredict the
reddening. For the young ages where such sub-solar metallicities could be expected,
interpreting the light as coming from a solar-metallicity population will lead to an
overestimate of the stellar age.
Finally, our SED modeling is based on purely stellar emission. During the brief
period when the AGN contribution is significant, the addition of its light will make the
galaxy look younger and dustier.
We next repeated our analysis adopting the best-fit photometric redshift estimate.
Using the photometric redshift code EAZY (Brammer et al. in preparation) in combination with a set of BC03 templates or Pegase templates, we obtain a median normalized
absolute deviation σ NMAD (∆z/(1 + z)) = 0.030 or 0.054 respectively. The random uncertainty in z phot boosts the scatter in the quality measures ∆ log M, ∆ log agew , ∆E(B − V),
and ∆A V , by 10-25% in the case of EAZY+BC03. A slight dependence on time with
respect to the merger (opposite in sign for the two template sets) propagates into systematic differences in the estimated stellar mass, on the 6% level for EAZY+BC03 between the start and end of the simulation 2 Gyr later. Offsets in reddening and visual
extinction A V are anti-correlated with ∆z/(1 + z).
Acknowledgments
We thank Eric Bell, Yuexing Li and Sukanya Chakrabarti for useful discussions in the
process of writing this chapter. The hospitality of the Institute for Theory and Computation during several working visits is gratefully acknowledged. This work was
supported by the Netherlans Foundation for Research (NWO), the Leids KerkhovenBosscha Fonds, and the Lorentz Center.
132
Effect
SFH mismatch (higher SFR than in the past)
Non-uniform extinction (uncorrelated with intrinsic colors of emitters)
Age-dependent extinction (more extinction toward young stars)
Z < Z⊙
AGN
on age
on E(B − V) on A V
on mass
+
+
-
+
+
+
-a
+b
+c
+
a
The overestimated age and underestimated A V compete in the determination of stellar mass. Since in practise we
find this effect to be most outspoken in phases where the SFH mismatch is largest, the stellar mass will effectively
be underestimated.
b
Here too, the systematic offsets in age and A V have an opposite sign. From comparison of mass estimates based
on the full attenuated photometry and its equivalent with all stars fixed to Z⊙ , we find the effective ∆ log M is
positive.
c
Similar to (b), we compared the results from SED modeling with/out AGN contribution and find that in the
median the addition of AGN light increases the mass estimate slightly.
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
Table 6.1. Qualitative summary of systematic trends in recovering stellar population
properties.
Section 6.6. Summary
133
References
Barnes, J. E.,& Hernquist, L. 1996, ApJ, 471, 115
Barnes, J. E.,& Hernquist, L. 1991, ApJ, 370, L65
Bell, E. F.,& de Jong, R. S. 2001, ApJ, 550, 212
Bolzonella, M., Miralles, J.-M.,& Pelló, R. 2000, A&A, 363, 476
Bruzual, G.,& Charlot, S. 2003, MNRAS, 344, 1000 (BC03)
Bullock, J. S., Kolatt, T. S., Sigad, Y., Somerville, R. S., Kravtsov, A. V., Klypin, A. A., Primack, J. R.,&
Dekel, A. 2001, MNRAS, 321, 559
Calzetti, D., et al. 2000, ApJ, 533, 682
Cox, T. J., Dutta, S. N., Di Matteo, T., Hernquist, L., Hopkins, P. F., Robertson, B.,& Springel, V. 2006,
ApJ, 650, 791
Daddi, E., et al. 2007, astro-ph/07052832
Di Matteo, T., Springel, V.,& Hernquist, L. 2005, Nature, 433, 604
Driver, S. P., Windhorst, R. A.,& Griffiths, R. E. 1995, ApJ, 453, 48
Edmunds, M. G. 1990, MNRAS, 246, 678
Erb, D. K., Shapley, A. E., Pettini, M., Steidel, C. C., Reddy, N. A.,& Adelberger, K. L. 2006a, ApJ, 644,
813
Erb, D. K., Steidel, C. C., Shapley, A. E., Pettini, M., Reddy, N. A.,& Adelberger, K. L. 2006b, ApJ, 646,
107
Erb, D. K., Steidel, C. C., Shapley, A. E., Pettini, M., Reddy, N. A.,& Adelberger, K. L. 2006c, ApJ, 647,
128
Ferguson, H. C., Dickinson, M., Papovich, C. 2002, ApJ, 569, 65
Fioc, M.,& Rocca-Volmerange, B. 1997, A&A, 326, 950
Förster Schreiber, N. M., et al. 2004, ApJ, 616, 40
Förster Schreiber, N. M., et al. 2006, AJ, 131, 1891
Franx, M., Illingworth, G. D., Kelson, D. D., van Dokkum, P. G.,& Tran, K. 1997, ApJ, 486, L75
Genzel, R., et al. 1998, ApJ, 498, 579
Glazebrook, K., Ellis, R. S., Santiago, B.,& Griffiths, R. 1995, MNRAS, 275, L19
Grazian, A., et al. 2006a, A&A, 449, 951
Hernquist, L. 1989, Nature, 340, 687
Holden, B. P., et al. 2005, ApJ, 620, 83
Holmberg, E. 1941, ApJ, 94, 385
Hopkins, P. F., Hernquist, L., Martini, P., Cox, T. J., Robertson, B., Di Matteo, T.,& Springel, V. 2005a, ApJ,
625, L71
Hopkins, P. F., Hernquist, L., Cox, T. J., Di Matteo, T., Martini, P., Robertson, B.,& Springel, V. 2005b, ApJ,
630, 705
Hopkins, P. F., Richards, G. T.,& Hernquist, L. 2007, ApJ, 654, 731
Kriek, M., et al. 2006, ApJ, 645, 44
Kriek, M., et al. 2006, astro-ph/0611724
Labbé, I., et al. 2003, AJ, 125, 1107
Larson, R. B.,& Tinsley, B. M. 1978, ApJ, 219, 46
Mihos, J. C.,& Hernquist, L. 1994, ApJ, 431, L9
Mihos, J. C.,& Hernquist, L. 1996, ApJ, 464, 641
Noguchi, M. 1988, A&A, 203, 259
O’Connell, R. W. 1986, in Stellar Populations, ed. C. Norman, A. Renzini,& M. Tosi (Cambridge: Cambridge Univ. Press), 213
Panuzzo, P., Granato, G. L., Buat, V., Inoue, A. K., Silva, L., Iglesias-Páramo, J.,& Bressan, A. 2007,
MNRAS, 375, 640
Papovich, C., Dickinson, M., & Ferguson, H. C. 2001, ApJ, 559, 620
Papovich, C., Dickinson, M., Giavalisco, M., Conselice, C. J.,& Ferguson, H. C. 2005, ApJ, 631, 101
Papovich, C., et al. 2006, ApJ, 640, 92
Pei, Y. C. 1992, ApJ, 395, 130
Pettini, M., Kellogg, M., Steidel, C. C., Dickinson, M., Adelberger, K. L.,& Giavalisco, M. 1998, ApJ, 508,
539
134
Chapter 6. Recovering Stellar Population Properties and Redshifts from Broad-Band
Photometry of Simulated Galaxies: Lessons for SED Modeling
Pettini, M., et al. 2001, ApJ, 554, 981
Reddy, N. A., Erb, D. K., Steidel, C. C., Shapley, A. E., Adelberger, K. L,& Pettini, M. 2005, ApJ, 633, 748
Richards, G. T., et al. 2006, ApJS, 166, 470
Robertson, B., Cox, T. J., Hernquist, L., Franx, M., Hopkins, P. F., Martini, P.,& Springel, V. 2006, ApJ,
641, 21
Sanders, D. B., Soifer, B. T., Elias, J. H., Madore, B. F., Matthews, K., Neugebauer, G.,& Scoville, N. Z.
1988a, ApJ, 325, 74
Sanders, D. B., Soifer, B. T., Elias, J. H., Neugebauer, G.,& Matthews, K. 1988b, ApJ, 328, L35
Sanders, D. B.,& Mirabel, I. F. 1996, ARA&A, 34, 749
Shapley, A. E., Steidel, C. C., Adelberger, K. L., Dickinson, M., Giavalisco, M.,& Pettini, M. 2001, ApJ,
562, 95
Shapley, A. E., Steidel, C. C., Pettini, M.,& Adelberger, K. L. 2003, ApJ, 588, 65
Shapley, A. E., Steidel, C. C., Erb, D. K., Reddy, N. A., Adelberger, K. L., Pettini, M., Barmby, P.,&
Jiasheng, H. 2005, ApJ, 626, 698
Springel, V.,& Hernquist, L. 2003, MNRAS, 339, 289
Springel, V., Di Matteo, T.,& Hernquist, L. 2005a, ApJ, 620, L79
Springel, V., Di Matteo, T.,& Hernquist, L. 2005b, MNRAS, 361, 776
Toomre, A.,& Toomre, J. 1972, ApJ, 178, 623
Toomre, A. 1977, in Evolution of Galaxies and Stellar Populations, 401, Yale Univ. Obs: New Haven
van Dokkum, P. G.,& Stanford, S. A. 2003, ApJ, 585, 78
van Dokkum, P. G., et al. 2004, ApJ, 611, 703
Wuyts, S., et al. 2007, ApJ, 655, 51
Zwicky, F. 1956, Ergebnisse der Exakten Naturwissenschaften, 29, 344
Chapter 7
Color distributions, number and mass
densities of massive galaxies at
1.5 < z < 3: comparing observations
with merger simulations
Abstract. We present a comparison between the observed color distribution, number and mass density of massive galaxies at 1.5 < z < 3 and a model by Hopkins et al. that relates the quasar and galaxy population on the basis of gas-rich
mergers. In order to test the hypothesis that quiescent red galaxies are formed
after a gas-rich merger involving quasar activity, we confront photometry of massive (M > 4 × 1010 M⊙ ) galaxies extracted from the FIRES, GOODS-South, and
MUSYC surveys, together spanning an area of 430 arcmin2 , with synthetic photometry from hydrodynamical merger simulations. The merger simulations are
placed in a cosmological context using the observed quasar luminosity function.
We find that the synthetic U − V and V − J colors of galaxies that had a quasar
phase in their past match the colors of observed galaxies that are best characterized by a quiescent stellar population. The model predicts a number density of
quiescent red galaxies with M > 4 × 1010 M⊙ of 3.6 ± 0.6 × 10−4 Mpc−3 at z ∼ 1.9
+0 . 8
and 1.0 ± 0.2 × 10−4 Mpc−3 at z ∼ 2.6, while the observations amount to 2.3−
0.6 ×
+0 . 6
−
4
−
3
−
4
−
3
10 Mpc at z ∼ 1.9 and 1.3−0.4 × 10 Mpc . The corresponding mass densities
are 4.4 ± 0.6 × 107 M⊙ Mpc−3 at z ∼ 1.9 and 1.3 ± 0.3 × 107 M⊙ Mpc−3 at z ∼ 2.6 for
+1 . 0
7
−3 at z ∼ 1.9 and 2.0+0.9 × 107 M Mpc−3
the model against 2.9−
⊙
0.7 × 10 M⊙ Mpc
−0 . 6
for the observations. Hence, the data are consistent with the models in which every quiescent massive galaxy underwent a quasar phase in the past. The merger
model also predicts a large number and mass density of galaxies undergoing star
formation driven by the merger. We find that the number and mass density is
consistent with the observations of star-forming galaxies. However, their colors do
not match those of observed star-forming galaxies. In particular, the colors of dusty
red galaxies are not reproduced by the simulations. Several possible origins of this
discrepancy are discussed, ranging from the method to compute the model predictions to the validity of the model assumptions to physical mechanisms such as
a large-scale wind producing a foreground screen of obscuring material and thus
leading to more efficient reddening
S. Wuyts, T. J. Cox, N. M. Förster Schreiber, M. Franx, L. Hernquist,
P. F. Hopkins, I. Labbé, B. Robertson, G. Rudnick, S. Toft & P. G. van Dokkum
135
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
136
comparing observations with merger simulations
7.1 Introduction
I
N recent years, deep near- and mid-infrared observations have revealed significant
populations of red galaxies at redshifts z ∼ 2 and above (Franx et al. 2003; Daddi
et al. 2004; Yan et al. 2004). The population of Distant Red Galaxies (DRGs), selected
by the simple observed color criterion J − K > 2.3, makes up 66% in number and 73%
in mass of the 2 < z < 3 galaxy population at the high mass end (M > 1011 M⊙ , van
Dokkum et al. 2006). Probing to lower masses, Wuyts et al. (2007) found that the lower
mass galaxies at redshifts 2 < z < 3.5 have bluer rest-frame U − V colors compared
to the most massive galaxies. A substantial fraction of the massive red galaxies at
high redshift are best characterized by a quiescent stellar population on the basis of
their broad-band SEDs (Labbé et al. 2005; Wuyts et al. 2007) and the presence of a
Balmer/4000Å break and absence of emission lines in their rest-frame optical spectra
(Kriek et al. 2006).
Any satisfying theory of galaxy formation has to account for the presence and abundance of these massive red galaxies in the early universe, a condition that was by no
means met by the state-of-the-art hierarchical galaxy formation models at the time of
their discovery (Somerville 2004).
In the meantime, merger scenarios involving AGN activity have been invoked by
semi-analytic models (Granato et al. 2004; Croton et al. 2006; Bower et al. 2006; De Lucia & Blaizot 2007) and hydrodynamical simulations (Springel et al. 2005a; Di Matteo
et al. 2005) to explain simultaneously the mass build-up of galaxies and the shutdown
of star formation. Such an evolutionary scenario predicts an obscured (and thus red)
star-burst phase and ends with a quiescent (and thus red) remnant galaxy (e.g., Hopkins et al. 2006a). Observational support for the connection between dust-enshrouded
starbursts, merging, and AGN activity from samples of nearby Ultra-Luminous Infrared Galaxies (ULIRGs) dates from as early as Sanders et al. (1988). Furthermore, the
observed relation between the supermassive black hole (SMBH) mass and the mass
(Magorrian et al. 1998) or the velocity dispersion (Ferrarese & Merritt 2000; Gebhardt
et al. 2000) of their host suggests that black hole and galaxy growth are intimately
connected. This scaling relation can be reproduced by merger simulations with implemented AGN feedback (Robertson et al. 2006).
Motivated by the observed and simulated correlations between the properties of
SMBHs and their hosts, Hopkins et al. (2006b) used the observed quasar luminosity
function to derive the galaxy merger rate as a function of mass. This chapter uses
the merger rate function derived by this model in combination with hydrodynamical
SPH simulations to predict the color distribution, number and mass density of massive
galaxies in the redshift range 1.5 < z < 3 under the assumption that each galaxy once
had or will undergo a quasar phase. We discuss the systematics involved and compare
the results to mass-limited samples in the same redshift interval, extracted from the
multi-wavelength surveys FIRES (Franx et al. 2000; Labbé et al. 2003; Förster Schreiber
et al. 2006), GOODS-South (Giavalisco et al. 2004; Chapter 3), and MUSYC (Quadri et
al. 2006).
We give an overview of the observations and simulations in §7.2 and §7.3 respectively. Next, the sample selection is explained in §7.4. §7.5 addresses the methodology
Section 7.2. Overview of the observations
137
to place the binary merger simulations in a cosmological context. We compare the
predicted abundance of massive galaxies by the model to the observations in §7.6. The
optical and optical-to-NIR color distribution of observed and simulated massive galaxies will be addressed in §7.7, followed by a discussion of their specific star formation
rates (§7.8) and of the number and mass density of quiescent and star-forming massive
galaxies in §7.9. We briefly compare observed and modeled pair statistics and address a
few caveats on the observational and modeling results in §7.10. Finally, we summarize
results in §7.11.
We work in the AB magnitude system throughout the chapter and adopt a H0 =
70 km s−1 Mpc−1 , Ω M = 0.3, ΩΛ = 0.7 cosmology.
7.2 Overview of the observations
7.2.1 Fields, coverage, and depth
We combine Ks -band selected catalogs of three different surveys: FIRES, GOODSSouth, and MUSYC. The reduction and photometry of the FIRES observations of the
Hubble Deep Field South (HDFS) is presented by Labbé et al. (2003) and was later
augmented with IRAC data. The field reaches a Ks -band depth of 25.6 mag (AB, 5σ for
point sources) and covers 5 arcmin2 . It was exposed in the WFPC2 U300 , B450 , V606 , I814
passbands, the ISAAC Js , H, and Ks bands, and the 4 IRAC channels. Following similar
procedures, a Ks -band selected catalog for the FIRES MS 1054–03 field was constructed
by Förster Schreiber et al. (2006). The field, covering 24 arcmin2 , has a Ks -band depth of
25 mag (AB, 5σ for point sources). The catalog comprises FORS1 U, B, and V, WFPC2
V606 , and I814 , ISAAC J, H, and Ks , and IRAC 3.6 µm - 8.0 µm photometry.
Over a significantly larger area (113 arcmin2 ), but to a shallower depth, a Ks -band
selected catalog was constructed based on the publicly available GOODS-South data
(Chapter 3). The variations in exposure time and observing conditions between the
different ISAAC pointings lead to an inhomogeneous depth over the whole GOODSSouth field. The 90% completeness level in the Ks -band mosaic is reached at an AB
magnitude of Ktot, AB = 23.7. The photometry was performed in an identical way to
that of the FIRES fields, allowing a straightforward combination of the three fields. The
included passbands are the ACS B435 , V606 , i775 , and z850 bands, the ISAAC J, H, and Ks
bands, and the 4 IRAC channels. We also use the ultradeep MIPS 24 µm (20 µJy, 5σ )
imaging of the GOODS-South field. As for the IRAC bands, we used the information
on position and extent of the sources from the higher resolution Ks -band image to
reduce confusion effects on the 24 µm photometry (Labbé et al. in preparation).
Finally, we complement the FIRES and GOODS-South imaging with optical-to-NIR
observations of the MUSYC HDFS1, HDFS2, and 1030 fields for parts of our analysis.
The Ks -band selected catalogs are presented by Quadri et al. (2006). Together, the
MUSYC fields span an area of 291 arcmin2 . They reach the 90% completeness level
at Ktot, AB = 22.7. Given the current lack of IRAC data for the MUSYC fields and their
shallower depth, they will only be used in the analysis of the rest-frame U − V color
distribution of the most massive (M > 1011 M⊙ ) high-redshift galaxies.
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
138
comparing observations with merger simulations
7.2.2 Redshifts and rest-frame photometry
Despite the large number of spectroscopic campaigns in the GOODS-South and FIRES
fields, the fraction of Ks -selected 1.5 < z < 3 galaxies that is spectroscopically confirmed is only 7%. The fraction drops to 3% when the MUSYC fields are included.
Therefore, a reliable estimate of the photometric redshift is crucial in defining robust
samples of massive high-redshift galaxies.
Förster Schreiber et al. (in preparation) used the algorithm developed by Rudnick
et al. (2001, 2003) to fit a nonnegative linear combination of galaxy templates to the
optical-to-NIR spectral energy distribution of each galaxy. The template set used for
the FIRES and GOODS-South fields consisted of 10 Single Stellar Population (SSP)
templates with a Salpeter (1955) initial mass function and solar metallicity from the
Bruzual & Charlot (2003, hereafter BC03) stellar population synthesis code, with ages
logarithmically spaced between 50 Myr and 10 Gyr. Each of the templates was allowed
to be attenuated according to the Calzetti et al. (2000) law by E(B − V) = 0.0, 0.1, 0.3, or
0.6.
The uncertainties in the photometric redshifts were determined from Monte Carlo
simulations. For each galaxy, a set of 100 mock SEDs was created by perturbing each
flux point according to its formal error bar, and repeating the z phot computation. The
lower and upper error on z phot comprise the central 68% of the Monte Carlo distribution.
We tested the quality of the photometric redshifts in two ways. First we compare
them to the available spectroscopic redshifts inthe 1.5 <
z < 3 interval, resulting in a
z
−z
spec
= 0.10. The quality measure
normalized median absolute deviation σ NMAD phot
1+ zspec
σ NMAD remains the same when the spectroscopic redshifts in the MUSYC fields are
included or excluded. Second we tested how well we could recover the redshift from
synthetic broad-band photometry of simulated SPH galaxies placed at redshifts 1.5 to
3. We found that the considered template set performed very well (σ NMAD(∆z/(1 +
z)) = 0.03). The scatter in the comparison to spectroscopically confirmed galaxies is
larger than that derived from the simulations. This is likely due to the fact that the
synthetic photometry is based on the same stellar population synthesis code as the
template set used to recover the redshifts. Therefore, the second test only studies the
impact of an unknown star formation history, dust and metallicity distribution on the
derived z phot.
We computed the rest-frame photometry by interpolating between observed bands
using the best-fit templates as a guide. Uncertainties in the rest-frame colors were derived from the same Monte Carlo simulations mentioned above, and comprise both
a contribution from photometric uncertainties and from z phot uncertainties. For a detailed description, we refer the reader to Rudnick et al. (2003).
7.2.3 Stellar masses
Förster Schreiber et al. (in preparation) derived stellar masses of the observed galaxies
following the procedure described by Wuyts et al. (2007). Briefly, we fit BC03 templates
to the optical-to-8 µm SED with the HYPERZ stellar population fitting code, version 1.1
(Bolzonella et al. 2000). We allow the following star formation histories: a single stellar
Section 7.3. Overview of the simulations
139
population (SSP) without dust, a constant star formation history (CSF) with dust, and
an exponentially declining star formation history with an e-folding timescale of 300
Myr (τ300 ) with dust. The allowed A V values ranged from 0 to 4 in step of 0.2, and the
attenuation law applied was taken from Calzetti et al. (2000). We constrain the time
since the onset of star formation to lie between 50 Myr and the age of the universe
at the respective redshift. Finally, a Salpeter (1955) IMF was assumed with lower and
upper mass cut-offs of 0.1 M⊙ and 100 M⊙ . For consistency, the same IMF was adopted
by the simulations.
7.2.4 Star formation rates
We derived estimates of the total (unobscured plus obscured) star formation rate of
the observed galaxies by adding the UV and IR light, scaled by the calibrations for the
local universe (Kennicutt 1998):
SFR [M⊙ yr−1 ] = 1.8 × 10−10 (L I R + 3.3L2800 ) / L⊙
(7.1)
where the rest-frame luminosity L2800 ≡ ν Lν (2800 Å) was derived from the observed
photometry with the algorithm by Rudnick et al. (2003). The total IR luminosity
L I R ≡ L(8 − 1000 µm) was derived from the observed 24 µm flux density in combination with the photometric redshift estimate (spectroscopic when available) following
the prescription of Dale & Helou (2002). As best estimate, we adopt the mean conversion factor of all Dale & Helou (2002) IR spectral energy distributions within the range
α = 1 − 2.5, where α parameterizes the heating intensity level from active (α = 1) to
quiescent (α = 2.5) galaxies. The variation from L I R,α=2.5 to L I R,α=1 is 0.9 dex in the
redshift interval 1.5 < z < 3. Where relevant, we indicate this systematic uncertainty
in the conversion from 24 µm to L I R and eventually star formation rate in the plots.
7.3 Overview of the simulations
We use a set of smoothed particle hydrodynamics (SPH, Lucy 1977; Gingold & Monaghan 1977) simulations performed by Robertson et al. (2006) of co-planar, equalmass, gas-rich ( f gas = 0.8) mergers over a range of galaxy masses. A description of
the GADGET-2 code used to run the simulations is given by Springel (2005b). Springel
& Hernquist (2003) describe the prescriptions for star formation and supernova feedback. The interplay between the supermassive black hole(s) and the environment is
discussed by Springel et al. (2005b). We refer the reader to Robertson et al. (2006) for
specifications on this particular set of simulations and an explanation of how the progenitors were scaled to approximate the structure of disk galaxies at redshift z = 3. The
photometry of the snapshots was derived in post-processing as described in Chapter
6.
Briefly, the total attenuated spectral energy distribution (SED) for a given snapshot
consisting of N stellar particles is computed as follows:
N
L Att,tot(λ) = ∑
i =1
mi · L Int (agei , Zi , λ)
Zi,los
· σ (λ )
· exp − NHi,los ·
Z⊙
(7.2)
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
140
comparing observations with merger simulations
Figure 7.1 — The relation between stellar mass and observed total Ks magnitude for galaxies in the
FIRES and GOODS-South fields at (a) 1.5 < z < 2.25 and (b) 2.25 < z < 3. The solid lines show the
adopted log M > 10.6 (FIRES+GOODS-South) and log M > 11 (FIRES+GOODS-South+MUSYC) mass
limits. The dotted line indicates the photometric limit of the GOODS-South imaging. The dashed line
indicates the approximate limit for the MUSYC fields. There are few galaxies with log M > 10.6 and
Ks,tot > 23.7, or log M > 11 and Ks,tot > 22.7. The largest incompleteness correction is needed for the
highest redshift bin in the MUSYC fields. A third of the log M > 11 galaxies would be undetected by
MUSYC, as estimated from the deeper FIRES+GOODS fields.
where mi , agei , and Zi are, respectively, the mass, age, and metallicity of stellar particle
i that is treated as a single stellar population. L Int is the intrinsic (unattenuated) SED
interpolated from a grid of templates from a stellar population synthesis code. Here,
we use SSP templates from BC03 as default. Results obtained when using a grid of
Maraston (2005, hereafter M05) SSP templates for different ages and metallicities will
be addressed as well. Parameters in Eq. 7.2 that are dependent on the line of sight are
subscripted with “los”. To each stellar particle, the column density of hydrogen and
the average metallicity along the line of sight was computed for 100 viewing angles,
uniformly spaced on a sphere. The optical depth is proportional to this metallicityscaled column density, with the wavelength dependence adopted from an attenuation
law (parameterized by the cross section σ (λ)). We use the Calzetti et al. (2000) reddening curve unless mentioned otherwise. The change in predicted colors when adopting
the SMC-like attenuation law from Pei (1992) will be discussed as well.
7.4 Sample selection
Our aim is to compare the color distribution, number and mass density of mass-limited
samples of observed and simulated galaxies. We choose the mass-limit such that the
observed sample is reasonably complete in the considered redshift interval, even for
the field with the shallowest Ks -band depth from which the sample was drawn. In
order to optimally exploit the range in area and depth of the considered surveys, we
define two mass-limited samples and divide each in two redshift bins: 1.5 < z < 2.25
Section 7.5. Methodology for cosmological context
141
and 2.25 < z < 3, probing a similar comoving volume. The first sample contains galaxies more massive than log M = 10.6 (M ≃ 4 × 1010 M⊙ ) in the FIRES and GOODSSouth fields. It contains 152 and 84 objects in the low- and high-redshift bin respectively. We present the sample in Figure 7.1, where we plot the stellar mass of all FIRES
and GOODS-South sources that are detected above the 5σ level in the respective redshift bin against their total observed Ks -band magnitude. The stellar mass correlates
with the Ks -band magnitude, but a scatter of an order of magnitude is present due to
the range in redshifts and spectral types of the galaxies. The 90% completeness limit
(Ks,tot = 23.7) for the GOODS-South field , which is shallower than the FIRES fields, is
indicated with the dotted line. At 1.5 < z < 2.25, no massive (log M > 10.6) galaxies
fainter than Ks,tot = 23.7 are found in the FIRES fields and deeper parts of the GOODSSouth mosaic. The lowest Ks -band signal-to-noise ratio in the massive galaxy sample
is S/ NKs ≃ 12, strongly suggesting that no incompleteness correction is needed to compute the number and mass density in the 1.5 < z < 2.25 redshift bin. In the 2.25 < z < 3
redshift bin, we find four well-detected massive (log M > 10.6) galaxies fainter than
the 90% completeness limit of GOODS-South. Three out of 4 have 6.4 < S/ NKs < 10,
whereas all other massive galaxies are detected above the 10σ level. Evaluating the
fraction of massive galaxies fainter than Ks,tot = 23.7 in the area that is sufficiently deep
to detect these sources, we estimate the completeness in the high-redshift bin to be
∼ 95%.
In order to reduce the uncertainty from cosmic variance in the derived number
and mass densities, we also compose a sample including the MUSYC fields, increasing
the sampled area by roughly a factor of 3. The shallower depth forces us to restrict
the mass limit to M > 1011 M⊙ . We derive the completeness in the two redshift intervals using the deeper FIRES and GOODS-South fields in Figure 7.1. The dashed line
marks the approximate depth (90% completeness) for the MUSYC fields. 1.5% of the
1.5 < z < 2.25 galaxies with log M > 11 in the deeper FIRES and GOODS-South fields
are fainter than this limit. For the 2.25 < z < 3 bin, the fraction of massive galaxies
that would be missed by MUSYC increases to 33%. In our analysis, we will mention
both the directly measured number and mass densities and those obtained after the
incompleteness correction.
7.5 Methodology for cosmological context
To date, hydrodynamical simulations including a self-consistent treatment of SMBH
growth have only been run with adequate resolution on binary merger systems (Springel
et al. 2005a; Di Matteo et al. 2005; Robertson et al. 2006; Cox et al. 2006) or as zoom-in
on overdense regions of cosmological N-body simulations at very high redshift z ∼ 6
(Li et al. 2006). In order to confront observations of 1.5 < z < 3 galaxies with the hydrodynamical simulations, we place the binary mergers in a cosmological context using
the observed quasar luminosity function following the prescription by Hopkins et al.
(2006b).
Briefly, the conversion from quasar demographics to galaxy demographics goes as
follows. From a large set of binary merger simulations, Hopkins et al. (2006a) dedt(L , L peak )
spent by a
termined the distribution of quasar lifetimes, describing the time d log(L)
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
142
comparing observations with merger simulations
Figure 7.2 — The birth rate of
spheroids (in greyscale) as a function
of redshift and final stellar mass as
derived from the observed quasar
luminosity function. The meaning of
the time scale arrow and the open circle
is described in the text. The model
by Hopkins et al. (2006b) assumes
that this birth rate equals the merger
rate of galaxies. The birth rate (i.e.,
merger rate) reaches a maximum of
4.5 × 10−4 log M−1 Mpc−3 Gyr−1 at
z ∼ 2. As time evolves, the peak of
the merger rate function shifts toward
lower mass galaxies.
quasar of peak luminosity L peak in the luminosity interval d log(L). The observed quasar
luminosity function simply corresponds to the convolution of this differential quasar
lifetime with the birth rate ṅ(L peak ) of quasars with peak luminosity L peak :
Φ(L) =
Z
dt(L, L peak)
ṅ(L peak ) d log L peak
d log(L)
(7.3)
Using a compilation of observed quasar luminosity functions in the hard X-rays
(Ueda et al. 2003), soft X-rays (Hasinger, Miyaji,& Schmidt 2005), and optical (Richards
et al. 2005), Eq. 7.3 was then de-convolved to solve for ṅ(L peak ). The relation between
peak luminosity of the quasar and the final black hole mass, derived from the same
simulations, was then adopted to calculate the birth rate of black holes of a certain
final mass ṅ(MBH ). This function was on its turn converted to a birth rate of spheroids
ṅ(Msph ) as a function of their final stellar mass using the SMBH-host connection M BH =
2.5
0.0012 (1+zz )2.5 Msph (Hopkins et al. 2007).
1+( 1.775 )
The model by Hopkins et al. (2006b) assumes that the birth rate of spheroids equals
the major merger rate of galaxies. The resulting merger rate as a function of stellar mass is displayed with greyscales in Figure 7.2 (darker meaning a higher merger
rate). Its redshift-dependence was derived by considering observed quasar luminosity
functions at a range of redshifts. The peak of the merger rate at z ∼ 2 has a value of
4.5 × 10−4 log M−1 Mpc−3 Gyr−1 . A clear trend is visible of mergers occuring in lower
mass systems as we proceed in time (i.e., to lower redshifts) after this peak. If mergers
are responsible for a significant part of the growth in stellar mass, this trend explains at
least qualitatively the observed downsizing of star formation over cosmic time (Cowie
et al. 1996).
To evaluate the post-merger (i.e., post-quasar, since the merging event triggers
Section 7.5. Methodology for cosmological context
143
Figure 7.3 — Typical evolution of a
merger simulation: (a) star formation
history, (b) history of the mass buildup, and (c) evolution of the rest-frame
V-band luminosity. The dotted line indicates when the peak in quasar luminosity is reached. For a detailed
description of the time evolution in
these and other parameters (e.g., accretion rate, quasar luminosity, extinction)
we refer the reader to Hopkins et al.
(2006a) and Chapter 6.
quasar activity in the simulations) galaxy population at z ∼ 2, we integrate the merger
rate function from z = ∞ to 2 and over the whole stellar mass range. For example,
when the integration reaches (M∗, f inal = 1011 M⊙ ; z = 3), marked by the circle in Figure 7.2, we compute the photometry of a merger simulation with a final stellar mass
of 1011 M⊙ at 1.1 Gyr after the peak of quasar luminosity (the time elapsed between
z = 3 and z = 2). As explained in §7.3, we compute the synthetic photometry along 100
lines-of-sight, uniformly spaced on a sphere. The number density of galaxies at z = 2
with colors corresponding to the 100 lines-of-sight is then scaled according to the value
of the merger rate function at (M∗, f inal = 1011 M⊙ ; z = 3). Finally, a mass cut is applied
to guarantee an identical selection of observed and simulated galaxies.
In order to predict the abundance and properties of galaxies at z ∼ 2 that have yet
to reach their peak in quasar luminosity or did not even start merging at the evaluated epoch, one can in principle integrate the merger rate function down to lower
and lower redshifts. How far one integrates beyond the evaluated redshift is a rather
arbitrary choice. We caution that counting galaxies long before they will contribute
to the quasar luminosity function will lead to large uncertainties given their unconstrained pre-merger history. The typical evolution of a merger simulation is illustrated
in Figure 7.3 where we plot the star formation rate, stellar mass, and rest-frame V-band
luminosity as a function of time since the peak in quasar luminosity. We decide to integrate 700 Myr beyond the evaluated redshift, thus counting both the galaxies that
are undergoing a merger-induced nuclear starburst (sometime between 0 and 200 Myr
before the quasar phase) and those with star formation triggered by the first passage
(sometime between 200 and 700 Myr before the quasar phase). Hereafter, we will refer to all galaxies in an evolutionary stage between 0 and 700 Myr before the quasar
phase as merging galaxies. Such a prediction only counts those galaxies that will later
merge and produce a quasar. Apart from predicting the abundance and properties of
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
144
comparing observations with merger simulations
Figure 7.4 — The number and mass
density of observed (filled symbols;
FIRES + GOODS-S) and modeled
(empty symbols) galaxies with log M >
10.6 as a function of redshift. The cross
symbols indicate the observed number
and mass density after a correction for
incompleteness. The black error bar
represents the Poisson shot noise solely.
The grey error bar accounts for uncertainties in redshift and mass, and
a (dominating) contribution from cosmic variance. We find that both the
predicted number and mass densities
agree within the error bars with the observed values.
the post-quasar population, we will thus be able to constrain how much of the massive
star-forming galaxies can be accounted for by merger-induced star formation.
Provided the assumption of a one-to-one correspondence between quasars and major mergers is valid, the formal uncertainty in the merger rate function presented in
Figure 7.2 originates mostly from the weakly constrained faint end of the observed
quasar luminosity function, where one can assume a pure luminosity evolution or also
a slope evolution. At the bright end, and therefore for our massive galaxy samples, the
predictions are robust, as will be indicated in due time.
7.6 The number density, mass density and mass function of galaxies
with log M > 10.6 at 1.5 < z < 3
Before analyzing the observed and modeled massive galaxy sample as a function of
color and galaxy type, we consider the overall abundance of galaxies above log M >
10.6. We computed the model number and mass density by integrating the merger rate
function to 700 Myr beyond the evaluated redshift, i.e., including galaxies up to 700
Myr before the quasar phase. The number and mass densities of galaxies with log M >
10.6 predicted by the model (empty symbols) are compared against the abundance of
observed galaxies (filled symbols) above the same mass limit in Figure 7.4. The results
are listed in Table 7.1. The spread of the empty circles indicates the freedom allowed by
the model due to the weakly constrained faint end of the quasar luminosity function.
The cross symbols represent the observed number and mass density after applying
a 0% and 5% correction for incompleteness in the low- and high-redshift bin respectively. We considered three sources of error in the observations: Poisson shot noise,
cosmic variance and selection uncertainties stemming from uncertainties in the redshift and the mass of individual galaxies. The black error bars in Figure 7.4 indicate
Section 7.6. The number density, mass density and mass function of galaxies with
log M > 10.6 at 1.5 < z < 3
145
the contribution from Poisson noise, ranging from 8 to 10%. We are more severely limited by cosmic variance. We follow the method outlined by Somerville et al. (2004)
to calculate the cosmic variance as predicted from cold dark matter theory for a population with unknown clustering as a function of its number density and the probed
comoving volume of the sample. The resulting contribution to the error budget is 28%
for the 1.5 < z < 2.25 and 29% for the 2.25 < z < 3 redshift bin. Finally, the uncertainties in the individual redshift and mass determinations propagate into the number
and mass density of quiescent red galaxies. We estimate the contribution to the total
error budget from Monte Carlo simulations. We constructed 1000 mock catalogs for
the FIRES and GOODS-South fields by perturbing the redshift, rest-frame colors, and
stellar masses so that 68% of the perturbed values lie within the formal 1σ lower and
upper limits. The uncertainties in the photometric redshift and the rest-frame colors
were derived as explained in §7.2.2. For the mass estimates, we adopt a lower error bar
of -0.1 dex and upper error bar of +0.04 dex for all merger remnants. This corresponds
to the quality with which stellar masses were recovered from synthetic photometry of
simulated merger remnants when applying the same SED modeling procedure as we
use for our observations (Chapter 6). The median recovered mass was only 0.01 dex
lower than the true stellar mass of the simulated merger remnants, suggesting that
systematic offsets are negligible for this type of galaxies. For star-forming galaxies that
have yet to reach their quasar phase we found a typical mass underestimate of -0.1
dex, with the central 68% interval of ∆ log M ≡ log Mrecovered − log Mtrue ranging from
-0.25 to 0 dex. We should keep in mind however that the input photometry for this test
and the templates used to recover the masses are based on the same stellar population
synthesis code. It has been noted by several authors (Maraston et al. 2006; van der Wel
et al. 2006; Wuyts et al. 2007) that the use of M05 templates instead of BC03 templates
leads to stellar mass estimates that are lower by a factor 1.5.
After we constructed the 1000 mock catalogs, we apply the same sample selection
(redshift interval, log M > 10.6) and compute the number and mass density for each of
them. The lower and upper limits comprising 68% of the distribution of mock number
and mass densities were added in quadrature to the uncertainty from Poisson shot
noise and cosmic variance, shown with the grey error bar in Figure 7.4. The uncertainty
in the number density propagating from redshift and mass uncertainties for individual
objects amounts to 5% and 10% for the low and high-redshift bin. The contribution to
the uncertainty in the mass density is 6% and 14% for the low- and high-redshift bin
respectively. We conclude that, even with the 142 arcmin2 area of our combined deep
fields, cosmic variance is still the limiting factor for the determination of the number
and mass density of quiescent red galaxies.
Figure 7.4 shows that the model number and mass density for the population of
massive (log M > 10.6) galaxies as a whole agrees within the error bars with the observations. Plotting the mass function for the observations (black histogram) and the
model (dark-grey polygon) in Figure 7.5, we find that the comparable abundance of observed and modeled galaxies still holds when studied as a function of galaxy mass.
With lighter grey polygons, we illustrate the model prediction when including only
galaxies up to 200 Myr before the merger (t > -200 Myr) or only merger remnants
(t > t QSO). The width of the polygons reflects the uncertainty in the merger rate func-
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
146
comparing observations with merger simulations
Figure 7.5 — The mass function of observed (black histogram;
FIRES + GOODS-S) and modeled (grey polygons) galaxies with
log M > 10.6 at redshift 1.5 <
z < 2.25 (top panel) and 2.25 <
z < 3.0 (bottom panel). Merger
remnants alone (t > t QSO) cannot account for the total population of observed galaxies above
the same mass limit. Integrating
the merger rate function to include galaxies up to 700 Myr before the quasar phase results in
a mass function that is consistent
with the observations.
tion. We conclude that merger remnants alone cannot account for the entire observed
massive galaxy population. However, including galaxies with merger-triggered star
formation, the mass function predicted by the model is in good agreement with the
observations. This results strengthens the idea that the model fairly reflects reality and
encourages a more detailed investigation of the properties of observed and simulated
massive galaxies.
7.7 The color distribution of galaxies with log M > 10.6 at 1.5 < z < 3
7.7.1 The U − V color distribution
First, we consider the optical color distribution of our sample of FIRES and GOODSSouth galaxies with M > 4 × 1010 M⊙ . A histogram of their rest-frame U − V colors is
plotted with a solid line in Figure 7.6(a) and Figure 7.6(b) for the low- and high-redshift
bin respectively. No corrections for incompleteness were applied here, but we remind
the reader that those are negligible for the low-redshift bin and of the order of 5% only
for the high-redshift bin. The total number of massive galaxies is 152 and 85 in the lowand high-redshift bin respectively. They span a broad U − V color range. In both cases,
the median color is U − V = 1.5 and 68% of the galaxies in each redshift bin lie within
the 1.1 < U − V < 1.9 interval.
It is interesting to consider whether the descendants and progenitors of quasars
(or rather quasar hosts) above the same mass limit show colors that are similar and
Section 7.7. The color distribution of galaxies with log M > 10.6 at 1.5 < z < 3
147
Figure 7.6 — The rest-frame U − V color distribution of observed galaxies with masses above log M =
10.6 in the FIRES and GOODS-South fields (solid line) for the redshift intervals (a) 1.5 < z < 2.25 and (b)
2.25 < z < 3. With filled histograms, we overplot the predicted U − V color distribution of merging and
post-quasar galaxies, scaled to the same solid angle as the observations. The light grey top of the model
histogram indicates the uncertainty in the merger rate function used to place the simulated mergers
in a cosmological context. For a given redshift interval, the model predictions in the three panels give
an indication of the uncertainty in the synthetic photometry induced by the choice of attenuation law
(Calzetti et al. 2000 versus the SMC curve from Pei 1992) and the choice of stellar population synthesis
code (BC03 versus M05). Overall, the predicted color distribution coincides with that of the observed
massive galaxy sample, with roughly equal numbers. The model distribution in the high-redshift bin
(b) shows a slight excess at blue U − V colors. The red tail of the observed color distribution is not
reproduced by the modeled merger and post-quasar population.
come in numbers comparable to those of the observed massive galaxy sample. In this
section, we focus mainly on the first question, but note in passing that we show the
predicted color distribution scaled to the same solid angle as probed by the FIRES and
GOODS-South observations. The filled grey histograms show the synthetic photometry of merger simulations in either their post-quasar phase or in a phase of at most
700 Myr before their peak in quasar luminosity. The numbers at each color are derived
from the observed quasar luminosity function by integrating the merger rate function
from z = ∞ to 700 Myr beyond the evaluated redshift as described in §7.5. The colors of
different evolutionary phases will be discussed seperately in due time. The difference
between the dark and light grey histogram reflects the uncertainty in the merger rate
function, itself due to uncertainties in the observed quasar luminosity function. Apart
from uncertainties in the merger rate function, uncertainties in the synthetic photometry for a given simulation snapshot contribute to the total error budget of the model
predictions. To translate the simulated properties such as age, mass, and metallicity
of the stellar particles to observables, we make use of a stellar population synthesis
code to compute the intrinsic colors and assume an attenuation law to calculate the
dimming and reddening by dust. We investigate the dependence on attenuation law
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
148
comparing observations with merger simulations
empirically by computing the synthetic photometry using a Calzetti et al. (2000) reddening curve and the SMC-like reddening curve from Pei (1992). We note that the
synthetic colors derived with the Milky Way-like attenuation curve by Pei (1992) lie in
between those produced by the two reddening curves considerd here. This is demonstrated in Chapter 6. Similarly, we test the dependence on adopted stellar population
synthesis templates empirically by computing the synthetic photometry based on a
grid of BC03 single stellar populations (SSPs) and based on a grid of SSPs by M05.
We note that the choice of attenuation law has a minor effect only on the U − V
color. The use of M05 templates gives the simulated galaxies a slightly redder color.
Overall, the same conclusion can be drawn independent of the way the model color
distribution was computed. Namely, the simulated galaxies with log M > 10.6 span
a color range that reaches from the bluest observed U − V colors to U − V ∼ 2. At
1.5 < z < 2.25, the color distribution resembles remarkably well that of the bulk of
the observed massive galaxies, both in shape and numbers. At 2.25 < z < 3, the predicted model colors show a slight excess at blue U − V colors. In both redshift bins,
the modeled color distribution does not reach the reddest U − V colors of observed
galaxies above the same mass limit. The good overall correspondence between the observed and modeled optical color distributions gives a first indication that the number
of massive post-quasar galaxies plus the number of galaxies in the process of merging
at 1.5 < z < 3 as expected from the observed quasar luminosity function may account
for a large fraction of the observed massive galaxy population at 1.5 < z < 3.
7.7.2 The V − J color distribution
Turning to longer wavelengths, we now compare the V − J colors predicted for mergers and merger remnants (i.e., post-quasars) with masses above log M = 10.6 to the
color distribution of observed galaxies in the same redshift interval and above the same
mass limit (Figure 7.7).
Again, the color distribution of our observed massive galaxy sample has a large
range of colors, reaching from V − J = 0.5 to V − J = 2.5 and peaking centrally at
V − J = 1.3 and 1.2 for the low- and high-redshift bin respectively. The central 68%
interval is 1.0 < V − J < 1.8 and 0.8 < V − J < 1.8 for the low- and high-redshift bin
respectively.
As for the U − V color distribution, we find that the adopted attenuation law has
only a minor influence on the color distribution, reaching at most shifts of 0.2 mag
toward redder V − J colors when the SMC-like reddening curve from Pei (1992) is
used instead of the Calzetti et al. (2000) attenuation law. Comparing the model V − J
color distribution derived from BC03 or M05 templates immediately shows that the
predictive power of the merger model is strongly hampered by the uncertainties in
the rest-frame NIR wavelength regime that today’s stellar population synthesis codes
are facing. In the low- and high-redshift bin, the median V − J color of the model
distribution is 0.4 and 0.5 mag redder when using M05 than when using BC03. One of
the main differences between the BC03 and M05 templates is the treatment of thermally
pulsating AGB stars. Using the fuel consumption approach as M05 does instead of
the isochrone synthesis approach that BC03 follow, one finds significantly larger NIR
luminosities for SSPs at ages between 0.2 and 2 Gyr. For an in-depth discussion of
Section 7.7. The color distribution of galaxies with log M > 10.6 at 1.5 < z < 3
149
Figure 7.7 — The rest-frame V − J color distribution of observed galaxies with masses above log M =
10.6 in the FIRES and GOODS-South fields (solid line) for the redshift intervals (a) 1.5 < z < 2.25 and (b)
2.25 < z < 3. With filled histograms, we overplot the predicted V − J color distribution of merging and
post-quasar galaxies, scaled to the same solid angle as the observations. The light grey top of the model
histogram indicates the uncertainty in the merger rate function used to place the simulated mergers
in a cosmological context. For a given redshift interval, the model predictions in the three panels give
an indication of the uncertainty in the synthetic photometry induced by the choice of attenuation law
(Calzetti et al. 2000 versus the SMC curve from Pei 1992) and the choice of stellar population synthesis
code (BC03 versus M05). The model V − J color distribution is weakly constrained by the uncertainties
at NIR wavelengths in the stellar population synthesis codes. Nevertheless, we can conclude that there
exist massive galaxies with redder V − J colors than those of modeled merging and post-quasar galaxies.
the differences between the two codes, we refer the reader to Maraston (2005) and
Maraston et al. (2006). It is worth stressing that, irrespective of whether the BC03 or
M05 stellar population synthesis code is used, the red (V − J > 1.8) tail of the observed
distribution has no counterparts in the modeled color distribution of merging and postmerger galaxies. Conversely, an excess of galaxies is found at blue (V − J ∼ 0.9) or
intermediate (V − J ∼ 1.4) optical-to-NIR colors for the BC03 and M05 model color
distributions respectively.
7.7.3 U − V versus V − J color-color distribution
7.7.3.1 Quiescent red galaxies
Recently, a diagnostic color-color diagram of observer-frame I − K versus K-[4.5 µm]
has been proposed by Labbé et al. (2005) to distinguish three basic types of z > 2 galaxies. The rest-frame equivalent of this diagram, U − V versus V − J, was presented by
Wuyts et al. (2007), allowing a comparison of galaxies over a wider redshift range.
First, there are galaxies with relatively unobscured star formation, such as Lyman break
galaxies (Steidel et al. 2003) and their lower redshift BX/BM analogs (Adelberger et al.
2004). Their young ages and low reddening values result in blue colors, both in the
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
150
comparing observations with merger simulations
Figure 7.8 — Model U − V versus V − J color-color distribution of simulated galaxies with log M > 10.6
that have had a merger and quasar phase in their past (greyscales), with a darker intensity indicating a
larger number of post-quasars. Observed galaxies above the same mass limit in the FIRES and GOODSSouth fields are overplotted. Empty symbols mark the galaxies that satisfy the quiescent galaxy criterion, whose selection window is marked by the grey wedge. A notable difference between the synthetic
photometry derived using the BC03 and M05 stellar population synthesis code is the redder V − J color
in the latter case. Recognizing this uncertainty in the model prediction, we can still conclude that the
predicted color distribution of post-quasars roughly coincides with that of quiescent red galaxies.
rest-frame optical and in the rest-frame optical-to-NIR. Second, there is a population
of star-forming galaxies with much redder colors, due to the presence of dust. Their
intrinsic (unobscured) colors are similar to those of Lyman break galaxies, but they
are driven along the dust vector toward redder U − V and redder V − J colors. Finally, a population of galaxies with red U − V colors is present at z ∼ 2 whose SED is
well matched by that of a passive or quiescently star-forming galaxy at an older age.
Their V − J colors are relatively blue compared to those of dusty starbursts at the same
optical color.
Labbé et al. (in preparation) designed a color criterion to select the quiescent red
galaxies based on their rest-frame U, V, and J photometry. The selection window is
defined as follows:
Section 7.8. Specific star formation rate as a function of stellar mass
U − V > 1.3 & V − J < 1.8 & U − V > 0.6(V − J) + 0.5
151
(7.4)
The validity of this selection criterion was confirmed by the fact that quiescent z ∼
2 galaxies with a prominent Balmer/4000Å break in their rest-frame optical spectra
mostly lie within the wedge. Conversely, MIPS detected galaxies at z ∼ 2, suggesting
dust-enshrouded star formation, tend to lie redward of the wedge. We draw the wedge
in Figure 7.8 and indicate the location of all galaxies with log M > 10.6 in the FIRES
and GOODS-South fields in the color-color diagram. Empty circles mark the objects
that satisfy Eq. 7.4.
We also present a binned representation of the model color-color distribution of
post-quasar galaxies only in Figure 7.8. The panels correspond to the 1.5 < z < 2.25
and 2.25 < z < 3 redshift bins, and model photometry derived from BC03 and M05
templates respectively. The color-color distribution computed with the SMC-like reddening curve from Pei (1992) instead of the Calzetti et al. (2000) law is not plotted, but
looks very similar.
We conclude that in all realizations of the synthetic photometry, the predicted colorcolor distribution of the post-quasar population coincides more or less with the region
of color-color space selected by the quiescent galaxy criterion.
7.7.3.2 Star-forming galaxies
A significant fraction (∼ 50%) of the observed massive galaxy population at 1.5 < z < 3
has colors located outside the quiescent red galaxy wedge. These objects reach from
blue U − V colors typical for Lyman break galaxies, which are known to host relatively
unobscured star formation, up to the redder optical and optical-to-NIR colors from
galaxies that are believed to host heavily obscured star formation. Here, we investigate
whether the predicted color-color distribution for merging galaxies that will undergo
a quasar phase in less than 700 Myr can reproduce the color range of observed starforming galaxies. Figure 7.9 compares the model prediction (greyscales) to the observed
massive galaxy colors (empty circles for star-forming galaxies).
As could be anticipated from §7.7.2, the model photometry does not reproduce the
colors of observed dusty star-forming galaxies (U − V > 1.3 and outside the quiescent
red galaxy wedge).
At bluer U − V, the synthetic photometry based on M05 templates gives a decent
match to the observations, whereas the BC03 colors in combination with a Calzetti et
al. (2000) attenuation law are offset by a few 0.1 mag toward bluer V − J.
7.8 Specific star formation rate as a function of stellar mass
So far, we have compared the synthetic colors of merging and post-quasar galaxies
with those of observed star-forming and quiescent galaxies respectively. The separation between star-forming and quiescent galaxies for our observed galaxies was based
on their broad-band optical-to-NIR colors. As an independent check, we now use the
UV + 24 µm derived star formation rates to compare the observed distribution of specific star formation rates as a function stellar mass with the distribution predicted by
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
152
comparing observations with merger simulations
Figure 7.9 — Model U − V versus V − J color-color distribution of simulated galaxies with log M > 10.6
that will undergo a quasar phase in less than 700 Myr (greyscales), with a darker intensity indicating a
larger number of galaxies. Observed galaxies above the same mass limit in the FIRES and GOODS-South
fields are overplotted. Empty symbols mark the galaxies that fall outside the quiescent galaxy criterion
(grey wedge). A notable difference between the synthetic photometry derived using the BC03 and M05
stellar population synthesis code is the redder V − J color in the latter case. (a) and (b) The model colors
based on BC03 are a poor match to the observed star-forming galaxies (empty symbols). The V − J colors
fall blueward of the observed distribution, and only the lower half of the observed U − V distribution
of star-forming galaxies is reproduced. (c) and (d) The model colors based on M05 give a better match
in the blue U − V regime, but objects with V − J >2 are missing.
∼
the merger model. The specific star formation rate is defined as the ratio of the star formation rate over the stellar mass. It equals the inverse of a mass-doubling time in the
case of constant star formation. Here, we limit our sample to the GOODS-South field,
where the 24 µm imaging is sufficiently deep (20 µJy, 5σ ) to obtain useful constraints
on the star formation rates.
Figure 7.10 shows the binned model distribution in greyscales and overplotted are
the observed massive galaxies that fall inside (empty symbols) and outside (filled symbols) the quiescent red galaxy wedge. Upper limits are drawn for objects that were
undetected by MIPS. Cross symbols mark those objects that are detected in the 1 Ms
Chandra X-ray exposure (Giacconi et al. 2001). We caution that the 24 µm flux of these
Section 7.8. Specific star formation rate as a function of stellar mass
153
Figure 7.10 — Specific star formation rate as a function of stellar mass for massive galaxies at 1.5 < z < 3
in the GOODS-South field with colors falling inside (empty circles) or outside (filled circles) the selection
window for quiescent red galaxies. Cross symbols indicate which sources are detected in X-rays. The
vertical error bar indicates the systematic error in SFR/ M. The model predictions are plotted with
greyscales. The top and side panels show the mass and SFR/ M distribution, with the black histogram
representing the observed sample, and the greyscaled curves showing the model predictions for postquasars and merging galaxies up to 700, 200, and 0 Myr before the quasar phase. When integrating
down to 700 Myr before the quasar phase, the predicted number density of galaxies with SFR/ M > 1
Gyr−1 is 1.6 times smaller than observed, possibly due to AGN contribution to the 24 µm emission from
which the observed SFR were derived.
objects could have an AGN contribution. Moreover, Daddi et al. (2007b) recently found
tot
that a significant fraction (20-30% to KVega
< 22, and up to ∼ 50 − 60% for M ∼ 1011 M⊙ )
of star-forming galaxies that are not individually detected in the X-rays show evidence
for heavily obscured AGN by the presence of a mid-IR flux excess. The vertical error
bar indicates the systematic uncertainty in the conversion from 24 µm flux to the obscured part of the star formation rate. The top and side panels show the distribution
of masses and specific star formation rates separately. With lighter polygons, we illustrate how the predicted distribution changes when integrating the merger rate function
only to the evaluated redshift or 200 Myr past the evaluated redshift. The latter case
includes the nuclear starburst phase, but not earlier star-forming phases.
We conclude that at 1.5 < z < 2.25 the broad-band color criterion is efficient in
distinguishing quiescent from star-forming galaxies with high specific star formation
rates. In the higher redshift bin, we are more limited by upper limits on the 24 µm flux.
The bulk of broad-band selected quiescent galaxies shows smaller specific star formation rates than their counterparts outside the broad-band selection window, although
some reach values above SFR/ M = 1 Gyr−1 .
As in the observations, in particular at 1.5 < z < 2.25, there is a slight hint that
the most heavily star-forming objects reside primarily at the lower masses within our
mass-limited sample. Papovich et al. (2006) and Reddy et al. (2006) find that the
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
154
comparing observations with merger simulations
specific star formation rate is inversely proportional to mass, implying that the ongoing
star formation at z ∼ 2 contributes more significantly to the mass buildup of low-mass
galaxies than to high-mass galaxies.
The predicted abundance of merger-triggered nuclear starbursts, occuring between
0 and 200 Myr before the quasar phase, seems to be insufficient to account for all observed massive galaxies with high specific star formation rates (SFR/ M > 1 Gyr−1 ).
However, when we include earlier phases of star formation induced by the merging
event (up to 700 Myr before the quasar phase), we find that the observed number density of galaxies with SFR/ M > 1 Gyr−1 is only a factor 1.6 larger than predicted by the
model. Such an offset might be expected from possible AGN contributions to the 24
µm emission from which the star formation rates were derived (see, e.g., Daddi et al.
2007b).
7.9 The abundance of massive galaxies at 1.5 < z < 3: analysis by
type
We now proceed to quantify the observed and modeled number and mass densities
of massive galaxies at 1.5 < z < 3. As before, the model prediction was derived by
integrating the merger rate function to include all galaxies that once contributed to the
observed quasar luminosity function or will do so in less than 700 Myr. From this,
we extracted 4 samples using the tools discussed in §7.7.3 and §7.8: galaxies above
log M > 10.6 with broad-band colors satisfying the quiescent red galaxy criterion (Eq.
7.4, §7.9.1), galaxies above log M > 10.6 that do not satisfy Eq. 7.4 (§7.9.2), galaxies with
log M > 10.6 and SFR/ M > 1 Gyr−1 (§7.9.3), and finally a sample of galaxies more
massive than 1011 M⊙ with red (U − V > 1.3) optical colors (§7.9.4). The last sample
allows us to include the larger area MUSYC survey in the comparison, for which no
IRAC or MIPS imaging is currently available. In each case, we impose an identical
selection criterion on the observed sample of galaxies.
7.9.1 The number and mass density of massive (log M > 10.6) quiescent red galaxies
Having established the similarity in colors of the model post-quasar population and
the observed quiescent red galaxy population above a same mass limit, we now turn
to a comparison of their number and mass densities. Our aim is to constrain the fraction (in number and mass) of massive quiescent red galaxies at redshifts 1.5 < z < 3
that descendants of merger-triggered quasars can account for. In order to do this, we
selected the observed and modeled galaxies with log M > 10.6 that lie inside the wedge
defined by Eq. 7.4 and compute the number and mass density for the probed comoving
volume of ∼ 3.5 × 105 Mpc3 in each redshift bin. The resulting number and mass densities are plotted as a function of central redshift of the redshift bin in Figure 7.11(a). The
filled circles indicate the observed number and mass density of quiescent red galaxies
above log M = 10.6. Their values and corresponding uncertainties are listed in Table
7.1.
Section 7.9. The abundance of massive galaxies at 1.5 < z < 3: analysis by type
155
Figure 7.11 — The number and mass density of observed (filled symbols) and modeled (empty symbols)
massive galaxies as a function of redshift above the same mass limit and satisfying the same selection
criterion. The cross symbols indicate the observed number and mass density after a correction for incompleteness (which is negligible except for the MUSYC fields). The black error bar represents the
Poisson shot noise solely. The grey error bar accounts for uncertainties in redshift, mass, and rest-frame
colors and a (mostly dominating) contribution from cosmic variance. The dashed error bar in panel
(c) reflects the systematic uncertainty in the SFR. We consider 4 samples: (a) Quiescent red galaxies
with log M > 10.6 in FIRES+GOODS-S, (b) Star-forming (non-quiescent) galaxies with log M > 10.6 in
FIRES+GOODS-S, (c) galaxies with SFR/ M > 1 Gyr−1 with log M > 10.6 in GOODS-S, and (d) galaxies
with U − V > 1.3 and log M > 11 in FIRES+GOODS-S+MUSYC. The model predictions were derived by
integrating the merger rate function to 700 Myr beyond the evaluated redshift. Changing this value only
alters the predictions for panels (b) and (c). We find that both the predicted number and mass densities
agree within the error bars with the observed values.
Observationsa
Type
All
All
Quiescent
Quiescent
Star-forming
Star-forming
SFR/ M > 1 Gyr−1
SFR/ M > 1 Gyr−1
U − V > 1.3
U − V > 1.3
Mass limit
M⊙
4 × 1010
4 × 1010
4 × 1010
4 × 1010
4 × 1010
4 × 1010
4 × 1010
4 × 1010
1011
1011
Redshift
1.5 < z < 2.25
2.25 < z < 3
1.5 < z < 2.25
2.25 < z < 3
1.5 < z < 2.25
2.25 < z < 3
1.5 < z < 2.25
2.25 < z < 3
1.5 < z < 2.25
2.25 < z < 3
Model Predictionb
n
10−4 Mpc−3
ρ∗
107 M⊙ Mpc−3
n
10−4 Mpc−3
ρ∗
107 M⊙ Mpc−3
+1 . 3
4 . 4−
1.0
+1 . 0
2 . 5−
0.6
+0 . 8
2 . 3−
0.6
+0 . 6
1 . 3−
0.4
+0 . 9
2 . 2−
0.7
+0 . 6
1 . 2−
0.4
+0 . 6
1 . 8−
0.5
+0 . 5
1 . 2−
0.4
+0 . 9
2 . 6−
0.6
+0 . 3
2 . 1−
0.3
+1 . 6
5 . 1−
1.2
+1 . 5
3 . 2−
0.8
+1 . 0
2 . 9−
0.7
+0 . 9
2 . 0−
0.6
+0 . 9
2 . 2−
0.8
+0 . 8
1 . 3−
0.6
+0 . 6
1 . 7−
0.5
+0 . 5
1 . 3−
0.4
+4 . 6
5 . 6−
2.2
+0 . 8
5 . 4−
1.0
4.8 − 5.6
1.8 − 2.2
3.0 − 4.1
0.8 − 1.2
1.3 − 2.1
0.8 − 1.3
1.0 − 1.2
0.7 − 0.8
1.6 − 2.4
0.4 − 0.7
5.6 − 6.3
2.1 − 2.4
3.8 − 5.0
1.0 − 1.5
1.2 − 2.0
0.8 − 1.2
0.9 − 1.0
0.6 − 0.7
2.9 − 3.9
0.8 − 1.2
a
The error bars in the observed densities account for Poisson noise, cosmic variance, and the uncertainties in
redshift, rest-frame color and mass of the individual galaxies. They do not account for the systematic dependence
on the stellar population synthesis code used to derive the stellar masses, nor was the systematic uncertainty in the
conversion from 24 µm to SFR (of the order of 1 dex) included in the results for the sample selected on SFR/ M.
b
The range in model densities indicates a crude estimate of the size of uncertainties in the merger rate function
and the dependence on choice of attenuation law and stellar population synthesis code to compute the synthetic
photometry.
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
156
comparing observations with merger simulations
Table 7.1. Number and mass densities for massive galaxies
Section 7.9. The abundance of massive galaxies at 1.5 < z < 3: analysis by type
157
The cross symbols represent the observed number and mass density after applying a
0% and 5% correction for incompleteness in the low- and high-redshift bin respectively.
As in §7.6, the black error bars account for Poisson shot noise. The grey error bars also
include selection uncertainties stemming from uncertainties in the redshift, mass, and
rest-frame colors of individual galaxies, and a dominating contribution from cosmic
variance.
The empty symbols on Figure 7.11(a) indicate the predicted number and mass density of galaxies with log M > 10.6 at 1.5 < z < 2.25 and 2.25 < z < 3 whose synthetic
photometry places them within the selection wedge for quiescent red galaxies. 95%
of these modeled galaxies are in a post-quasar phase of their evolution. The different
empty circles represent predictions derived with the BC03 and M05 stellar population
synthesis codes, with the Calzetti et al. (2000) attenuation law and the SMC-like attenuation law from Pei (1992). Their spread gives a crude indication of the freedom
allowed by the model. It also takes into account the uncertainty in the merger rate
function used to place the binary merger simulations in a cosmological context.
We find that in both redshift bins, the observed number and mass density of massive quiescent red galaxies agrees within the error bars with the predicted number and
mass density of simulated galaxies satisfying the same selection criterion. In other
words, assuming a one-to-one correspondence between quasars and gas-rich mergers,
the model by Hopkins et al. (2006b) predicts an abundance of merger remnants (i.e.,
post-quasar galaxies) that is similar to the observed abundance of quiescent red galaxies. The model predicts an increase by a factor 3.5 in the number and mass density for
massive post-quasar galaxies in the 1 Gyr that passed between z = 2.6 and z = 1.9. The
observed sample seems to suggest less evolution (a factor 1.8 in number density and 1.5
in mass density), but is formally consistent with both the factor 3.5 and a non-evolving
number and mass density over the considered redshift range.
7.9.2 The number and mass density of massive (log M > 10.6) star-forming galaxies
Following identical procedures as outlined above, we analyze the number and mass
density of massive galaxies with colors outside the quiescent red galaxy wedge in Figure 7.11(b). Again, we used a Monte Carlo simulation to determine how many galaxies
moved into or out of the selection window when perturbing their properties within the
error bars. We took into account the fact that the mass estimates of star-forming galaxies are less robust than for quiescent galaxies. In Chapter 6 we found a typical mass
underestimate of -0.1 dex for star-forming galaxies, with the central 68% interval of
∆ log M ≡ log Mrecovered − log Mtrue ranging from -0.25 to 0 dex.
We find a similar number density of massive quiescent and massive star-forming
galaxies in the observed fields. The mass density of the observed star-forming galaxies
is lower than that for the quiescent ones above the same mass limit by a factor 1.3 at
z = 1.9 and 1.5 at z = 2.6. We find that the abundances of star-forming galaxies, both in
number and mass, as predicted by the merger model agree within the error bars with
the observed values. The ratio of quiescent to star-forming galaxies as predicted by
the model amounts to 2 (1) for the number density and 2.8 (1.2) for the mass density in
the low (high) redshift bin. A more robust model prediction of the pre-quasar number
and mass density would also require a careful simulation of the evolutionary stages
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
158
comparing observations with merger simulations
during which and viewing angles under which the binary merger would be detected
as two separate galaxies, thus contributing twice to the number density, but with half
the mass and therefore possibly dropping out of the mass-limited sample.
7.9.3 The number and mass density of massive (log M > 10.6) galaxies with SFR/M > 1
Gyr−1
Selecting galaxies with specific star formation rates above SFR/ M > 1 Gyr−1 , we find
an observed number density of 1.8 × 10−4 Mpc−3 and 1.2 × 10−4 Mpc−3 at 1.5 < z <
2.25 and 2.25 < z < 3 respectively (Figure 7.11(c)). Since we interpreted all the 24 µm
emission as dust re-emission from star formation, the true number density can be lower
depending on the contribution from AGN (see, e.g., Reddy et al. 2005; Papovich et al.
2006; Daddi et al. 2007b). The merger model predicts an abundance of galaxies with
high specific star formation rates that is a factor 1.6 smaller than observed. Given the
possible AGN contribution to the 24 µm emission and the large systematic uncertainty
in the conversion from 24 µm to the dust-obscured contribution to the star formation
rate (dashed line in Figure 7.11(c)), the model and observational results are formally
consistent.
7.9.4 The number and mass density of galaxies with M > 1011 M⊙ and U − V > 1.3
In order to reduce the effect of cosmic variance, we now include the MUSYC fields
in our analysis, increasing the area by a factor 3 and reducing the cosmic variance
with a similar factor. This goes at the cost of depth (the 90% completeness limit for
the MUSYC fields is 1 magnitude shallower than for GOODS-South) and wavelength
coverage (no IRAC photometry is currently available for the MUSYC fields). Consequently, we are restricted to a sample limited at M > 1011 M⊙ , even then requiring a
33% correction for incompleteness in the 2.25 < z < 3 bin. Moreover, the lack of IRAC
observations prevents us from selecting galaxies that fall inside the selection window
for quiescent red galaxies. We therefore compare the number and mass density of all
M > 1011 M⊙ galaxies with U − V > 1.3 (the lower edge of the quiescent red galaxy
wedge), knowing from §7.7 that a significant fraction of the observed galaxies will fall
outside the quiescent red galaxy wedge.
Figure 7.11(d) shows their number and mass density as a function of redshift for the
combined FIRES, GOODS-South, and MUSYC surveys (black symbols). Poisson noise
is negligible for this sample. The grey error bars again account for cosmic variance
and the uncertainties in redshift, rest-frame color, and mass of the individual galaxies making up the sample. For the MUSYC survey, we assumed an increase in the
mass uncertainty by a factor of 3 with respect to the FIRES and GOODS-South samples due to the lack of IRAC photometry (see Wuyts et al. 2007). In contrast to the
FIRES + GOODS-South sample discussed above, the total error budget is not always
dominated by cosmic variance. The large uncertainty in the mass density at z ∼ 1.9
for example is mostly attributed to errors in the properties of individual galaxies as
derived from the mock catalogs. Seperately, we indicate the results obtained from the
deeper FIRES and GOODS-South surveys (grey symbols, dominated in area by GOODSSouth). We conclude that the GOODS-South field is significantly underdense in terms
Section 7.10. Comments and caveats
159
of the highest mass (M > 1011 M⊙ ) galaxies, in particular in the 2.25 < z < 3 redshift
bin. Our analysis shows an agreement between the observed abundance (in number
and mass) of red (U − V > 1.3) massive (M > 1011 M⊙ ) galaxies at 1.5 < z < 3 on the
one hand, and the model prediction of galaxies satisfying the same criteria that either
had a merger-triggered quasar phase in their past lifetime or will undergo such a phase
within 700 Myr on the other hand. In fact, the vast majority of massive red galaxies
predicted by the merger model are post-quasar galaxies (see §7.7.3). At 2.25 < z < 3,
the predicted massive post-quasar population can account for at least 25% in number
and 20% in mass of the observed massive red galaxy population.
Assuming that the ratio of massive red galaxies (M > 1011 M⊙ & U − V > 1.3)
falling inside and outside the quiescent red galaxy selection window in the MUSYC
fields equals that in the FIRES and GOODS-South fields, we find that the merger model
can account for 70% of the number density and 30% of the mass density of massive
quiescent red galaxies at 2.25 < z < 3 and is consistent with forming all of the massive
quiescent red galaxies at 1.5 < z < 2.25 through mergers involving AGN activity. A
recent discussion on the importance of AGN activity at high redshift is provided by
Daddi et al. (2007b).
7.10 Comments and caveats
In this Section, we list a number of caveats, and indicate prospects for improvements
on both the model and observational side. We list a number of possible reasons for
the discrepancy between the synthetic and observed colors of massive star-forming
galaxies in §7.10.3.
7.10.1 Pair statistics
In this chapter, we used the integrated properties (color, mass, SFR) for the merging
pair to predict number and mass densities for different samples of massive galaxies.
In other words, we made the approximation that the merging pair is never detected as
two separate galaxies. As suggested in §7.9.2, a more detailed modeling should take
into account that, depending on the orbital configuration and the viewing angle, this
may not always be the case.
The post-quasar predictions are robust, since by that time the two progenitors have
formed one galaxy. If at earlier times some of the progenitor pairs are resolved as two
separate objects, this will decrease the mass density of massive galaxies since galaxies
will drop out of the sample. The effect on the number density is less trivial. On the
one hand, galaxies will drop out of the mass-limited sample. On the other hand, some
merging pairs will contribute twice. We leave such an extension of the merger model
for later work.
Here, we focus on an additional test of the merger model allowed by the fact that
some of the pairs will be resolved into two objects. If a large fraction of the massive
galaxy population at 1.5 < z < 3 is indeed related to merging events, as our analysis
suggests, we expect to see an excess in the pair statistics with respect to a random
distribution of galaxies on the sky.
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
160
comparing observations with merger simulations
Figure 7.12 — Relative galaxy density
(δ ) as a function of massive (log M >
10.6) 1.5 < z < 3 galaxy to z > 1.3
galaxy separation (Rs ) in the GOODSSouth field.
The distribution predicted by the merger model is indicated
with the dashed line. At separations
smaller than 1.′′ 5 (dotted line) an increasing number of galaxy pairs, if present,
will be missed because they would be
detected as a single object. We find a
clear excess at small pair separations
(Rs < 8′′ ), as predicted by the merger
model. A weak pair excess is also visible when only considering the distribution of separations between massive
1.5 < z < 3 galaxies (inset panel), but the
excess is much below the prediction.
We present the distribution of galaxy-galaxy separations in the GOODS-South field
in Figure 7.12 (solid histogram). We decide not to include the other fields, to prevent
differences in depth from influencing the pair excess signal. The main panel shows
the results from a cross-correlation of our massive (log M > 10.6) galaxy sample at
1.5 < z < 3 with the sample of all galaxies above z > 1.3 in the GOODS-South field,
thus avoiding the risk of losing pair members that by a typical photometric redshift
error were placed at some lower redshift. For each massive galaxy at 1.5 < z < 3, we
measure the distance to all z > 1.3 galaxies. We compute the statistic
j
∑ Ni (Rs )
δ (Rs ) =
i =1
π ((Rs + ǫ)2 + (Rs − ǫ)2 )
(7.5)
where j is the total number of objects in our massive galaxy sample and Ni (Rs )
is the number of z > 1.3 galaxies that lie between a distance Rs − ǫ and Rs + ǫ from
galaxy i. For a random uniform distribution of galaxies, δ (Rs ) will be flat. Figure 7.12
shows that for our sample of massive galaxies at 1.5 < z < 3, this is clearly not the
case. An excess of pairs at Rs < 8′′ is visible, also when we consider the distribution of
separations between members of the massive galaxy sample at 1.5 < z < 3 only (inset
panel).
From the simulations, we measured the physical separations between the 2 merging
galaxies and computed the distribution of separation angles in arcseconds on the sky
using the merger rate function (see §7.5). Adding the mean value of δ as measured
in the interval 30′′ < Rs < 80′′ , we obtain a model prediction (dashed line) that is in
qualitative agreement with the cross-correlation results, but larger than the weak pair
Section 7.10. Comments and caveats
161
excess seen in the auto-correlation. Admittedly, the predicted distribution is subject to
the orbital configuration set at the start of the simulation, an effect that is not explored
in this Chapter.
7.10.2 Dependence on stellar population synthesis code
As pointed out in §7.9.1, the predicted rest-frame NIR luminosities for a single stellar
population of a given mass are brighter for the M05 than for the BC03 stellar population synthesis code. Consequently, the mass estimates for observed galaxies with
ages between 0.2 and 2 Gyr are lower by about a factor 1.5 when modeled with M05
instead of BC03 templates. We indicated the resulting systematic uncertainties in the
number and mass densities of the modeled post-quasar galaxy population. This systematic uncertainty is also present in the estimates of number and mass density for
the observations, for which so far we only included the formal uncertainties in mass
obtained with BC03 SED modeling. To quantify the impact on the derived number and
mass densities, we simply divide all stellar masses by 1.5 and repeat the selection procedure and the calculation of the densities. Using this crude approach, we find number
densities for all samples discussed in this chapter to be 50-80% and mass densities to
be 40-60% of the value obtained with BC03 masses. Although a significant source of
uncertainty, we find that for most samples discussed in this Chapter (except for the
quiescent red galaxy sample at z ∼ 1.9) adopting the M05 mass estimates would only
improve the agreement between data and model. Finally, deviations from a Salpeter
(1955) IMF would change our results on number and mass densities of massive galaxies, but in a similar manner for the observations and the simulations, thus keeping
intact the relative trends.
7.10.3 Reproducing dusty red starbursts
As discussed in §7.7.3.2, the synthetic colors during the star-forming phases of the
merger do not reproduce the red optical and optical-to-NIR colors observed for dusty
starbursts. If the lack of model colors redder than V − J > 1.8 could be fixed, the
statistics based on the merger model, when compared to the observed abundances of
massive galaxies, strengthen the idea that the model gives a valid representation of
galaxy evolution in the real universe. Here we list a number of possible origins for the
discrepancy in colors. Future investigations along these lines will help to further test
the merger model.
7.10.3.1 Simulating the observing procedure
First, it is possible that the colors of observed and modeled galaxies are in fact in agreement, but that a discrepancy was found because we did not simulate the whole observing procedure. The observed colors are measured on PSF-matched images within
apertures of size 1” to 2”, whereas the synthetic colors were based on integrated photometry of all stellar particles, irrespective of their location. The presence of a color
gradient with redder emission in the central regions of the galaxies could therefore
induce an offset in colors in the observed direction.
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
162
comparing observations with merger simulations
7.10.3.2 Stellar population synthesis
Second, the discrepancy in colors could be real, but due to an incorrect modeling of
the stellar populations, rather than invalid assumptions at the basis of the model (i.e.,
the one-to-one correspondence between quasars and gas-rich mergers). Apart from
the choice of stellar population synthesis code (see §7.10.2), the synthetic photometry depends on the attenuation law applied to each of the stellar particles. We note
however that the use of a Milky Way-like attenuation law from Pei (1992) leads to colors intermediate between those based on the Calzetti et al. (2000) and SMC-like (Pei
1992) attenuation laws presented in this Chapter. An attenuation law that is less grey
than that of the SMC would be required to reproduce the red colors of dusty starburst
galaxies.
Another stellar population parameter influencing the synthetic colors is the metallicity of the gas and the stars. In this Chapter, we adopted initial gas metallicities derived from the closed box model (Talbot & Arnett 1971) for the 80% gas fraction ( f gas )
at the start of the simulation:
Zinit = − y ln( f gas )
(7.6)
where y = 0.02 is the yield. The simulation keeps track of the subsequent evolution in
the gas metallicity, and stellar metallicities are based on the metallicity of the gas out
of which they form. If the gas was pre-enriched, this would boost the optical depths
and redden the colors. Evidence of high (∼ Z⊙ ) metallicities of massive high-redshift
galaxies with red colors is given by van Dokkum et al. (2004). Repeating the postprocessing of simulation snapshots with 1Z⊙ added to the gas and stellar metallicities,
we obtain colors that are 0.2 to 0.5 mag redder in U − V and 0.3 to 1.1 mag redder in
V − J. We note however that in V − J the largest increase occurs for blue galaxies and
the color distribution based on BC03 does not reach beyond V − J ∼ 2.
7.10.3.3 Merger parameters
Third, the discrepancy in colors might imply that the simulations are not characteristic
for the merger activity occuring in the real universe. Hopkins et al. (2006b) confirmed
the robustness of the model for quasar lifetimes and the derived merger rate function
against changes in various parameters of the merging galaxies, such as gas fraction,
orbital parameters and changes in the mass ratio of the progenitors (considering 1:1,
2:1, 3:1, and 5:1 mass ratios). All of the simulations used in this work are equal-mass
gas-rich ( f gas = 0.8) mergers. A smaller initial gas fraction would mean that a larger
fraction of the stars was already formed before the merger-triggered starburst. This
leads to slightly redder U − V colors. Considering earlier stages in the evolutionary
scenario, Dasyra et al. (2006) find for a population of local ULIRGs that still have
2 distinct nuclei that the typical mass fraction is 1.5:1, close to equal-mass mergers.
In order to extend the model predictions to lower masses, a detailed study of minor
merger simulations will be required to determine the minimum mass ratio required to
trigger a (low-luminosity) quasar phase.
Section 7.11. Summary
163
7.10.3.4 Evolutionary history
Fourth, it is possible that dusty starburst galaxies are not triggered by mergers, but had
a different evolutionary history. Daddi et al. (2007a) make this claim based on the long
star formation timescales of ULIRGs at high redshift.
7.10.3.5 Mass loss and intergalactic environment
Fifth, gas replenishment from mass loss and infall of gas from the intergalactic environment could change the optical depths and thus the reddening factors. The simulations
only take into account a small amount of mass loss: 10% of the gas mass converted
into stars is instantaneously returned to the interstellar medium, accounting for shortlived stars that die as supernovae (Springel & Hernquist 2003). The total fraction of
the mass lost by a single stellar population with Salpeter (1955) IMF amounts to ∼ 70%
(BC03) and is even higher for the top-heavy IMFs of Kroupa (2001) or Chabrier (2003).
Furthermore, the simulations do not allow for infall of primordial gas at later times.
Consequently, they cannot prove that descendants of galaxies that once showed up in
the quasar luminosity function and after the shutdown of star formation reached red
colors, will remain quiescent forever. Small amounts of newly accreted gas triggering
star formation may be enough to shift a post-quasar galaxy outside the quiescent region of color-color space defined by Eq. 7.4, thus dropping their contribution to the
observed galaxy population of massive quiescent red galaxies. Cosmological simulations at sufficient resolution might resolve this problem. At the very least, it would be
interesting to test the behavior of simulated merger remnants hosting a supermassive
black hole when a small but continuous gas supply is applied.
7.10.3.6 Dust distribution
Finally, the distribution of dust in the simulated galaxies might not reflect reality. A
more efficient reddening would be obtained if a foreground screen of obscuring material were present. One possible mechanism that could produce such a configuration is
a large-scale wind. The GADGET-2 code (Springel 2005b) used to run the simulations
in principle allows for such a phenomenon, but an investigation of the velocity field
of the gas in the simulations is required to check whether such a wind is effectively
taking place.
7.10.4 Cosmic variance
From the observational side, cosmic variance is the dominant source of uncertainty
for the determination of the number and mass density of massive quiescent galaxies.
IRAC photometry over a MUSYC area and to a similar depth as FIRES and GOODSSouth would be required to further constrain the fraction of massive quiescent galaxies
that post-quasar galaxies can account for.
7.11 Summary
We confronted the model by Hopkins et al. (2006b) with observations of massive
galaxies at 1.5 < z < 3. The model translates the observed quasar luminosity function
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
164
comparing observations with merger simulations
into the abundance of massive merging galaxies and merger remnants. We derived
the synthetic photometry for these systems from a set of binary merger SPH simulations with a range of masses, and including stellar and AGN feedback. We extracted
mass-limited samples of 1.5 < z < 3 galaxies with M > 4 × 1010 M⊙ and M > 1011 M⊙
from the FIRES+GOODS-South and FIRES+GOODS-South+MUSYC surveys respectively. We tested the model by comparing the predicted number and mass densities,
and the U − V and V − J color distributions with our observations of massive galaxies
at 1.5 < z < 3.
We find that the overall number density of galaxies with M > 4 × 1010 M⊙ in the
+1 . 0
+1 . 3
−4
Mpc−3 at z ∼ 1.9 and n = 2.5−
FIRES and GOODS-South fields (n = 4.4−
0.6 ×
1.0 × 10
−4
−3
10 Mpc at z ∼ 2.6) is in good agreement with the model prediction (n = 4.8 −
5.6 × 10−4 Mpc−3 at z ∼ 1.9 and n = 1.8 − 2.2 × 10−4 Mpc−3 at z ∼ 2.6). Likewise,
+1 . 6
7
−3
the results obtained for the mass density are consistent: ρ∗ = 5.1−
1.2 × 10 M⊙ Mpc
+1 . 5
7
−3
at z ∼ 1.9 and ρ∗ = 3.2−
at z ∼ 2.6 for the observations and ρ∗ =
0.8 × 10 M⊙ Mpc
7
−3
5.6 − 6.3 × 10 M⊙ Mpc at z ∼ 1.9 and ρ∗ = 2.1 − 2.4 × 107 M⊙ Mpc−3 at z ∼ 2.6.
Separating massive galaxies by type, we find that the model photometry of the
post-quasar population coincides with the region of U − V versus V − J color-color
space that was defined by Labbé et al. (in preparation) to select quiescent red galaxies.
The observed and modeled number and mass densities of massive (M > 4 × 1010 M⊙ )
quiescent red galaxies at 1.5 < z < 3 are consistent within the error bars (see Table 7.1).
We added the MUSYC survey to our sample, increasing the area by a factor of 3,
but by lack of IRAC data loosing the ability to break at least partially the age-dust
degeneracy. As pointed out earlier by van Dokkum et al. (2006), the GOODS-South
field is underdense in the 1.5 < z < 3 redshift interval. Based on the FIRES+GOODSSouth+MUSYC sample, we find that more than 25% of the z ∼ 2.6 galaxies with M >
1011 M⊙ and U − V > 1.3 and essentially all of the z ∼ 1.9 galaxies satisfying the same
criteria can be accounted for by the predicted post-quasar population. The fraction of
observed quiescent galaxies at z ∼ 2.6 that the model can account for increases if we
use Maraston (2005) models to derive stellar masses for our observed galaxy sample.
Although less constrained, the predicted abundances of galaxies with star formation triggered by merging and galaxies with SFR/ M > 1 Gyr−1 during the merging
phases are also consistent with the observations. However, the predicted color distribution of star-forming galaxies does not match the observations. In particular the
colors of red (V − J > 1.8) dusty starburst galaxies are not reproduced. We suggest a
number of explanations for the lack of dusty red starburst galaxies in the model predictions. Possible reasons are an incomplete simulation of the observing procedure, differences in stellar population properties or merger characteristics between the observed
and simulated galaxies, a different history for dusty starbursts than a merger-triggered
scenario, additional gas (and dust) from the intergalactic environment or mass loss,
and a different distribution of the dust, e.g., caused by the presence of large-scale outflows.
Finally, we find a pair excess at small angular scales, further strengthening the hypothesis that mergers play a key role in galaxy evolution.
Section 7.11. Summary
165
Acknowledgments
We would like to thank the Institute for Theory and Computation for its hospitality
during several working visits, and the Leids Kerkhoven-Bosscha Fonds for the generous travel support.
References
Adelberger, K. L., Steidel, C. C., Shapley, A. E., Hunt, M. P., Erb, D. K., Reddy, N. A.,& Pettini, M. 2004,
ApJ, 607, 226
Bolzonella, M., Miralles, J.-M.,& Pelló, R. 2000, A&A, 363, 476
Bower, R. G., Benson, A. J., Malbon, R., Helly, J. C., Frenk, C. S., Baugh, C. M., Cole, S.,& Lacey, C. G.
2006, MNRAS, 370, 654
Calzetti, D., et al. 2000, ApJ, 533, 682
Chabrier, G. 2003, ApJ, 586, L133
Cowie, L. L., Songaila, A., Hu, E. M.,& Cohen, J. G. 1996, AJ, 112, 839
Cox, T. J., Dutta, S. N., Di Matteo, T., Hernquist, L., Hopkins, P. F., Robertson, B.,& Springel, V. 2006,
ApJ, 650, 791
Croton, D. J. et al. 2006, MNRAS, 367, 864
Daddi, E., Cimatti, A., Renzini, A., Fontana, A., Mignoli, M., Pozzetti, L., Tozzi, P.,& Zamorani, G. 2004,
ApJ, 617, 746
Daddi, E., et al. 2007a, astro-ph/07052831
Daddi, E., et al. 2007b, astro-ph/07052832
Dasyra, K. M., Tacconi, L. J., Davies, R. I., Genzel, R., Lutz, D., Naab, T., Burkert, A., Veilleux, S.,&
Sanders, D. B. 2006, ApJ, 638, 745
De Lucia, G.,& Blaizot, J. 2007, MNRAS, 375, 2
Di Matteo, T., Springel, V.,& Hernquist, L. 2005, Nature, 433, 604
Erb, D. K., Steidel, C. C., Shapley, A. E., Pettini, M., Reddy, N. A.,& Adelberger, K. L. 2006, ApJ, 646, 107
Ferrarese, L.,& Merritt, D. 2000, ApJ, 539, L9
Förster Schreiber, N. M., et al. 2006, AJ, 131, 1891
Franx, M., et al. 2000, The Messenger 99, pp. 20-22
Franx, M., et al. 2003, ApJ, 587, L79
Gebhardt, K., et al. 2000, ApJ, 539, L13
Giavalisco, M.,& the GOODS Team 2004, ApJ, 600, L93
Gingold, R. A.,& Monaghan, J. J. 1977, MNRAS, 181, 375
Granato G. L., De Zotti, G., Silva, L., Bressan, A.,& Danese, L. 2004, ApJ, 600, 580
Hasinger, G., Miyaji, T.,& Schmidt, M. 2005, A&A, 441, 417
Hopkins, P. F., Hernquist, L., Cox, T. J., Di Matteo, T., Robertson, B.,& Springel, V. 2006a, ApJS, 163, 1
Hopkins, P. F., Hernquist, L., Cox, T. J., Robertson, B.,& Springel, V. 2006b, ApJS, 163, 50
Hopkins, P. F., Hernquist, L., Cox, T. J., Robertson, B.,& Krause, E. 2007, astro-ph/0701351
Kriek, M., et al. 2006, ApJ, 649, 71
Kroupa, P. 2001, MNRAS, 322, 231
Labbé, I., et al. 2003, AJ, 125, 1107
Labbé, I., et al. 2005, ApJ, 624, L81
Li, Y., et al. 2006, astro-ph/0608190
Lucy, L. B. 1977, AJ, 82, 1013
Magorrian, J., et al. 1998, AJ, 115, 2285
Maraston, C. 2005, MNRAS, 362, 799
Maraston, C., Daddi, E., Renzini, A., Cimatti, A., Dickinson, M., Papovich, C., Pasquali, A.,& Pirzkal, N.
2006, ApJ, 652, 85
Papovich, C., et al. 2006, ApJ, 640, 92
Pei, Y. C. 1992, ApJ, 395, 130
Pettini, M., Kellogg, M., Steidel, C. C., Dickinson, M., Adelberger, K. L.,& Giavalisco, M. 1998, ApJ, 508,
539
Chapter 7. Color distributions, number and mass densities of massive galaxies at 1.5 < z < 3:
166
comparing observations with merger simulations
Pettini, M., et al. 2001, ApJ, 554, 981
Quadri, R., et al. 2007, ApJ, 654, 138
Reddy, N. A., Erb, D. K., Steidel, C. C., Shapley, A. E., Adelberger, K. L.,& Pettini, M. 2005, ApJ, 633, 748
Reddy, N. A., et al. 2006, ApJ, 644, 792
Richards, G. T., et al. 2005, MNRAS, 360, 839
Robertson, B., Cox, T. J., Hernquist, L., Franx, M., Hopkins, P. F., Martini, P.,& Springel, V. 2006, ApJ,
641, 21
Rudnick, G., et al. 2001, AJ, 122, 2205
Rudnick, G., et al. 2003, ApJ, 599, 847
Sanders, D. B., Soifer, B. T., Elias, J. H., Madore, B. F., Matthews, K., Neugebauer, G.,& Scoville, N. Z.
1988, ApJ, 325, 74
Shapley, A. E., Steidel, C. C., Pettini, M.,& Adelberger, K. L. 2003, ApJ, 588, 65
Somerville, R. S. 2004, in Multiwavelength mapping of galaxy formation and evolution, ed. R. Bender,&
A. Renzini (Berlin: Springer) (astro-ph/0401570)
Somerville, R. S., Lee, K., Ferguson, H. C., Gardner, J. P., Moustakas, L. A.,& Giavalisco, M. 2004, ApJ,
L171
Springel, V.,& Hernquist, L. 2003, MNRAS, 339, 289
Springel, V., Di Matteo, T.,& Hernquist, L. 2005a, ApJ, 620, L79
Springel, V., Di Matteo, T.,& Hernquist, L. 2005b, MNRAS, 361, 776
Steidel, C. C., Adelberger, K. L., Shapley, A. E., Pettini, M., Dickinson, M.,& Giavalisco, M. 2003, ApJ,
592, 728
Talbot, R. J,& Arnett, W. D., 1971, ApJ, 170, 409
Ueda, Y., Akiyama, M., Ohta, K.,& Miyaji, T. 2003, ApJ, 598, 886
van der Wel, A., Franx, M., Wuyts, S., van Dokkum, P. G., Huang, J., Rix, H.-W.,& Illingworth, G. D.
2006, ApJ, 652, 97
van Dokkum, P. G., et al. 2004, ApJ, 611, 703
van Dokkum, P. G., Kriek, M., Rodgers, B., Franx, M.,& Puxley, P. 2005, ApJ, 622, L13
van Dokkum, P. G., et al. 2006, ApJ, 638, 59
Wuyts, S., et al. 2007, ApJ, 655, 51
Yan, H. et al. 2004, ApJ, 616, 63
Nederlandse samenvatting
Van toen tot nu
D
en een half miljard jaar geleden bestond het heelal uit een saaie, bijna volledig homogene oersoep van deeltjes. Bijna, want op die gladde verdeling van
energie en materie kwamen minuscule rimpelingen (kleiner dan 0.01%) voor. Hoe onbeduidend ze aanvankelijk ook waren, in deze fluctuaties lag de kiem van alle structuur die vandaag in het heelal aanwezig is: sterrenstelsels, sterren, planeten, inclusief
wijzelf. Gebieden in het vroege heelal waar de dichtheid iets groter was dan gemiddeld oefenden een sterkere zwaartekrachtswerking uit op hun omgeving en waren
daardoor in staat meer materiaal naar zich toe te trekken. Op die wijze groeiden kleine
overdichtheden uit tot grote aantrekkingspolen in een kosmisch web. De hoofdrolspeler in de geschiedenis van structuurvorming is de zogenaamde donkere materie.
Hoewel tot op heden niet rechtstreeks gedetecteerd, blijkt uit indirecte waarnemingen
dat deze mysterieuze vorm van materie de totale massa aan zichtbare materie, waaruit
sterren, planeten en wijzelf zijn opgebouwd, met een factor 6 overtreft. Aangetrokken
tot de concentraties van donkere materie, verzamelde de zichtbare materie, die in het
vroege heelal voornamelijk uit waterstof en helium bestond, zich in de knooppunten
van het kosmisch web. Daar kon het gas afkoelen en sterren vormen. Door middel
van kernfusie in het binnenste van deze sterren werden zwaardere chemische elementen dan waterstof en helium geproduceerd. Aan het einde van hun levensloop geven
sterren een gedeelte van dit materiaal terug aan het interstellaire medium in de vorm
van sterwinden of spectaculaire explosies. Uit het interstellaire gas dat op die manier
verrijkt werd met zwaardere elementen ontstond vervolgens een nieuwe generatie van
sterren, die de cyclus voortzetten.
Vier en een half miljard jaar geleden, toen deze cyclus aan de derde generatie toe
was, ontstond een ster die wij Zon noemen. Eerdere generaties van sterren hadden
voldoende zware elementen geproduceerd om in een schijf om de Zon de Aarde en
enkele andere planeten te laten vormen. In de tijd die volgde, zou de aardkorst het toneel vormen van een merkwaardig schouwspel waarin het ooit door sterren gevormde
materiaal zich in vele structuren ontwikkelde. Een van die structuren, die ons nauw
aan het hart ligt, noemen we Mens.
ERTIEN
Van nu tot toen
De gemeenschap die zich professioneel bezighoudt met het optekenen van de kosmische geschiedenis, of een aspect ervan, heeft wereldwijd de omvang van een dorp.
Hoewel dit astronomendorp al vele eeuwen zoniet millennia bestaat, berust de zonet
geschetste geschiedenis grotendeels op ontdekkingen van de afgelopen 100 jaar en resten er ook nu nog vele open vragen. Eén wijk van dit dorp bestudeert de vorming en
167
168
Nederlandse samenvatting
evolutie van sterrenstelsels. Sterrenstelsels zijn verzamelingen van miljarden sterren,
ingebed in een enorme wolk of ’halo’ van donkere materie. Ze kunnen beschouwd
worden als de bouwstenen van het heelal. Sommige sterrenstelsels behoren op hun
beurt weer tot grote groepen of ’clusters’ van sterrenstelsels. Onze Zon maakt deel
uit van de Melkweg, een spiraalsterrenstelsel waar nog steeds gas in nieuwe sterren
wordt omgezet. In de jaren ’20 ontdekte de Amerikaanse astronoom Edwin Hubble
dat er naast de Melkweg nog talloze andere sterrenstelsels bestonden. Deze stonden
op zulke gigantische afstanden dat het licht er miljoenen jaren over had gedaan om de
Aarde te bereiken. Bovendien constateerde hij dat sterrenstelsels die verder van ons af
gelegen zijn zich sneller van ons verwijderen. De ontdekking van de uitdijing van het
heelal was daarmee een feit.
Inmiddels hebben de bouw van grotere telescopen en de ontwikkeling van gevoeligere detectoren het mogelijk gemaakt sterrenstelsels tot meer dan 10 miljard lichtjaar
(een terugkijktijd van meer dan 10 miljard jaar) op te sporen. Aangezien ieder sterrenstelsel afzonderlijk op tijdschalen van honderdduizenden tot zelfs miljarden jaren
evolueert, kunnen we hun levensloop niet ’live’ volgen. In plaats daarvan beschikken we over een momentopname. Door de momentopnamen van sterrenstelsels op
verschillende afstanden (en dus terugkijktijden) met elkaar te vergelijken, trachten
sterrenkundigen een typische levensloop van een sterrenstelsel te traceren. De sleutelvragen zijn daarbij: Wanneer zijn de sterren in sterrenstelsels gevormd? en Hoe
en wanneer werd de massa waaruit een sterrenstelsel is opgebouwd bij elkaar gebracht? De huidige theorieën over vorming en evolutie van sterrenstelsels beschrijven
een hiërarchisch scenario waarbij aanvankelijk kleine groeperingen van gas en sterren
in de loop der tijd samensmolten tot de grote sterrenstelsels die in het huidige heelal
voorkomen. Hoeveel van dergelijke botsingen nodig waren, wanneer ze plaatsvonden,
en of de meeste sterren toen al waren gevormd of niet, zijn vooralsnog onbeantwoorde
vragen.
De kleuren van sterrenstelsels
Al spoedig na de ontdekking dat sommige lichtbronnen aan de hemel geen sterren
of gasnevels binnen onze Melkweg waren, maar afzonderlijke sterrenstelsels op veel
grotere afstanden, begon men de waargenomen sterrenstelsels te ordenen naar kleur,
vorm en omgeving. In het lokale heelal treffen we een tweedeling aan tussen blauwe
spiraalstelsels (ook wel laat-type stelsels genoemd) die een schijfvorm hebben, en rode
elliptische stelsels (vroeg-type) met de vorm van een bol of rugbybal. Ook valt op dat
de fractie elliptische stelsels aanzienlijk hoger is in clusters dan in een omgeving met
lagere dichtheid aan sterrenstelsels. De verschillende eigenschappen van deze twee
soorten sterrenstelsels duiden erop dat ofwel hun vormingsgeschiedenis verschilt, ofwel ze zich in een verschillende fase van hun evolutie bevinden. Bekend is dat de
meest massieve sterren het helderst zijn en een kortere levensduur hebben dan minder massieve sterren. Ook zijn de zwaarste sterren het heetst en hebben ze daardoor
een blauwe kleur. Het is daarom een logische conclusie dat de sterpopulatie van spiraalstelsels door de band genomen jonger is dan die van elliptische sterrenstelsels. De
blauwe kleur van spiraalstelsels duidt er immers op dat de meest massieve sterren nog
169
Figuur 1 — Fragment van het Chandra Deep Field South, een stukje hemel waarvan
we de waarnemingen in dit proefschrift analyseren. De afmeting van dit gebiedje aan
de hemel bedraagt slechts een twintigste van de breedte van een vinger op gestrekte
armlengte. De diepe waarnemingen tonen sterrenstelsels met een grote verscheidenheid aan vormen en helderheden. Ook alle kleine vlekjes op de foto zijn sterrenstelsels
(Bron: NASA, ESA, M. Giavalisco (STScI) en het GOODS Team).
in leven zijn. Omgekeerd danken elliptische stelsels hun rode kleur aan het feit dat de
meest massieve sterren reeds zijn opgebrand en het licht wordt gedomineerd door de
zwakkere, rode sterren. De kleur van sterrenstelsels bevat dus belangrijke informatie
over hun leeftijd.
Het vinden, bestuderen en interpreteren van sterrenstelsels in het jonge heelal is
om tal van redenen een stuk gecompliceerder. In de eerste plaats ontvangen we minder licht van sterrenstelsels die zich op grote afstand bevinden. Vaak wordt daarom
gekozen om met grote telescopen slechts een klein stukje van de hemel waar te nemen, maar dan met bijzonder lange belichtingstijden (tot tientallen uren). Zo beslaan
de waarnemingen waarop dit proefschrift zich concentreert slechts een miljoenste van
de gehele hemelbol. Zoals te zien is in Figuur 7.11, telt zelfs een klein deel ervan duizenden sterrenstelsels. De lessen die getrokken worden over de vormingsgeschiedenis
van sterrenstelsels berusten op de aanname dat het betreffende stukje hemel representatief is voor het gehele heelal.
Een belangrijk aspect om rekening mee te houden is het feit dat de grote verwijderingssnelheid van ver weg gelegen sterrenstelsels tot een verschuiving van het licht
naar langere golflengten (rodere kleuren) leidt. Dit fenomeen wordt ’roodverschuiving’ genoemd en sterrenkundigen maken er dankbaar gebruik van om afstanden
170
Nederlandse samenvatting
tot sterrenstelsels mee te bepalen (zo ook in dit proefschrift). Het licht van een sterrenstelsel op 10 miljard lichtjaar afstand dat optische telescopen ontvangen, is zelfs
dermate roodverschoven dat het oorspronkelijk als UV-licht door het betreffende stelsel werd uitgezonden. Een robuuste vergelijking van momentopnamen uit het nabije
en verre heelal vereist daarom dat we de ver weg gelegen stelsels ook met nabijeinfraroodcamera’s waarnemen om zo het intrinsiek optische licht in kaart te brengen.
Tenslotte wordt de kleur van een sterrenstelsel, zelfs na correctie voor roodverschuiving, niet louter door de leeftijd van haar sterren bepaald. Stof, mits in voldoende
mate aanwezig tussen de sterren, kan sterlicht absorberen en doet dit met een hogere
efficiëntie bij blauwere golflengten dan bij rodere golflengten. Ook de aanwezigheid
van een supermassief zwart gat in het centrum van een sterrenstelsel kan, mits het
gevoed wordt met voldoende materiaal, een kleurverandering veroorzaken. Sterrenstelsels die tot de laatste categorie behoren, worden actieve stelsels genoemd. Actieve
sterrenstelsels komen voor in allerlei soorten. Met name de zogenaamde ’quasars’ komen in dit proefschrift aan bod.
In het afgelopen decennium is duidelijk geworden dat ook in het vroege heelal
reeds een grote diversiteit aan sterrenstelsels voorkwam, in verschillende vormen en
gewichten, van blauwe tot rode kleuren. Met name aan de ontdekking van rode sterrenstelsels in het vroege heelal hebben Leidse sterrenkundigen een grote bijdrage geleverd. Het is in navolging van dit werk dat dit proefschrift is geschreven. De doelstelling van dit proefschrift is om de helderheid en de kleuren van ver weg gelegen
sterrenstelsels, in het bijzonder die met rode kleuren, te interpreteren in termen van
fysische grootheden: massa, leeftijd, hoeveelheid stof. Net zo belangrijk is het te
weten met welke nauwkeurigheid we deze schattingen kunnen maken. Tenslotte
gebruiken we de waargenomen kleuren en afgeleide grootheden om een model te
toetsen dat de vorming van rode sterrenstelsels tracht te verklaren aan de hand van
botsingen tussen sterrenstelsels die een quasar-fase teweegbrengen.
Dit proefschrift
Bij mijn onderzoek naar de aard van rode sterrenstelsels op hoge roodverschuiving
maakte ik gebruik van waarnemingen, modellen van stellaire populaties en hydrodynamische simulaties. De waarnemingen bestonden uit optische opnamen door de
Hubble ruimtetelescoop, nabije-infraroodopnamen met behulp van de ISAAC camera
op de Europese Very Large Telescope (VLT), en mid-infraroodopnamen door de Spitzer ruimtetelescoop. Verder werden spectroscopische waarnemingen uitgevoerd op
enkele 8-10m klasse telescopen: VLT, Gemini South en Keck. Een spectroscopische
waarneming houdt in dat de verdeling van de lichtintensiteit over verschillende golflengten wordt gemeten. Dit laat toe om erg nauwkeurig de roodverschuiving, en dus
afstand van een sterrenstelsel, te bepalen.
In hoofdstuk 2 meten we de grootte van sterrenstelsels die zich in clusters tot op
7 miljard lichtjaar afstand bevinden. Onze aandacht gaat daarbij in het bijzonder naar
stelsels met een vroeg-type (elliptische of lensvormige) morfologie. In combinatie met
spectroscopische waarnemingen, die informatie bevatten over de variatie in snelheden waarmee sterren in een sterrenstelsels bewegen, leiden we de massa af van deze
171
stelsels. Door de eigenschappen van clusterstelsels met eenzelfde massa op verschillende terugkijktijden met elkaar te vergelijken, schatten we dat de sterren in massieve
vroeg-type clusterstelsels reeds 11 miljard jaar geleden werden gevormd. Dat is slechts
2 miljard jaar na de oerknal. Stelsels waarvoor deze methode een iets jongere leeftijd
dan gemiddeld oplevert, hebben een relatief blauwere kleur, in overeenstemming met
de eerder beschreven relatie tussen kleur en leeftijd.
Kennelijk speelt veel van de stervormingsactiviteit zich vroeger in de geschiedenis
van het heelal af. Om die interessante periode 10 miljard jaar geleden te bestuderen,
maken we in hoofdstuk 3 een catalogus met nauwkeurige helderheden en kleuren
van sterrenstelsels in het Chandra Deep Field South, een stukje hemel dat met nagenoeg het gehele arsenaal aan telescopen op aarde en in de ruimte is waargenomen.
De catalogus bestaat uit metingen met 12 kleurfilters in het optische en infrarode deel
van het spectrum. Met behulp van al deze kleurinformatie schatten we de roodverschuiving tot alle objecten. Een afstandsschatting op basis van kleuren wordt ook wel
fotometrische roodverschuiving genoemd. Vergelijking met een uitgebreide database
van (uiterst precieze) spectroscopische roodverschuivingen leert dat de afstandsschattingen betrouwbaar zijn. Voor de sterrenstelsels op 10 miljard lichtjaar die we uit de
catalogus selecteerden, schatten we vervolgens de energie output die ze bij alle infrarode golflengten tesamen (golflengte 0.008 tot 1 millimeter) uitzenden. Indien het licht
een puur stellaire oorsprong heeft, is dit een maat van de hoeveelheid stervorming die
schuilgaat achter stofwolken. Wanneer stof licht absorbeert, warmt het immers op en
zendt vervolgens de opgenomen energie weer uit bij infrarode golflengten. Activiteit
rond een supermassief zwart gat in het centrum van een sterrenstelsel kan ook voor
opwarming van stof zorgen, maar deze actieve sterrenstelsels verraden hun aard vaak
door röntgenstraling. We vinden dat de som van alle totale infrarood emissie uitgezonden door sterrenstelsels op 10 miljard lichtjaar gedomineerd wordt door bronnen
met een rode kleur in het UV, optisch en nabije infrarood deel van het spectrum. Onder de stelsels met een rode optische kleur bevinden er zich echter ook objecten die
slechts een geringe hoeveelheid infrarood licht uitzenden. Gezien hun rode optische
kleur en gebrek aan re-emissie door stof, lijkt het erop dat deze stelsels op het moment
van waarneming nauwelijks enige stervorming vertonen, noch open en bloot, noch
afgeschermd door stof.
Hoofdstuk 4 beschrijft hoe we voor 15 rode sterrenstelsels de spectroscopische
roodverschuiving bepalen. Aangezien deze stelsels nauwelijks UV-licht uitzenden,
vraagt het waarnemen van hun optische spectra het uiterste van zelfs de grootste telescopen op aarde. We vinden dat het kleurcriterium waarmee ze geselecteerd werden
(gebaseerd op waarneming in slechts 2 filterbanden), efficient is om ver weg gelegen
rode sterrenstelsels te selecteren. Slechts 2 van de 15 sterrenstelsels bevinden zich op
meer nabij gelegen afstanden. Hun kleuren worden het best geı̈nterpreteerd als afkomstig van erg stoffige stervormende stelsels. Twee andere spectra vertonen kenmerken van activiteit rond een centraal zwart gat in het sterrenstelsel. We vinden dat
de spectroscopische roodverschuivingen voor rode ver weg gelegen sterrenstelsels in
goede overeenstemming is met de geschatte fotometrische roodverschuivingen. Deze
vaststelling is van essentieel belang omdat we voor het merendeel van het onderzoek
aangewezen zijn op de fotometrische methode. Fouten in de afstandsbepaling zouden
172
Nederlandse samenvatting
leiden tot het verkeerd inschatten van zowel intrinsieke helderheden als kleuren. Op
hun beurt zou dit een nauwkeurige bepaling van massa’s en leeftijden van sterrenstelsels verhinderen.
In Hoofdstuk 5 richten we onze aandacht op fysieke eigenschappen als massa, leeftijd en stofgehalte van sterrenstelsels op 10 miljard lichtjaar. We maken in deze analyse
gebruik van modellen voor de kleurevolutie van stellaire populaties. In het bijzonder staan we stil bij de vraag welke extra informatie mid-infrarood waarnemingen
met de IRAC camera aan boord van de Spitzer ruimtetelescoop levert. We vinden
dat voor sterrenstelsels met blauwe optische kleuren de IRAC waarnemingen weinig
meerwaarde opleveren. Voor rode sterrenstelsels treedt er geen globale verschuiving
van de massaverdeling op, maar de onzekerheden op de geschatte grootheden voor
individuele sterrenstelsels nemen wel af met een factor 3. Bovendien stellen de IRAC
waarnemingen ons in staat een onderscheid te maken tussen sterrenstelsels waarvan
de rode kleur veroorzaakt wordt door stervorming die schuilgaat achter grote hoeveelheden stof, en stelsels die hun rode kleur danken aan een volwassen sterpopulatie. We
concluderen ook dat, net als in het lokale heelal, de meest massieve sterrenstelsels in
het vroege heelal een rodere kleur hebben dan hun minder massieve tegenhangers. Het
bestaan van dergelijke massieve sterrenstelsels die zo vroeg in de geschiedenis van het
heelal al over een relatief oude sterpopulatie beschikken, betekent een uitdaging voor
de theoretische modellen.
Hoofstuk 6 bouwt voort op de vraag hoe nauwkeurig we fysieke eigenschappen
van ver weg gelegen sterrenstelsels kunnen bepalen. Echte waarnemingen stellen ons
maar in beperkte mate in staat om deze vraag te beantwoorden, omdat het antwoord
simpelweg niet vaak voorhanden is. We benaderen de vraag daarom vanuit een andere invalshoek en betrekken computersimulaties van sterrenstelsels in de vergelijking.
Deze simulaties worden gedraaid met een computerprogramma waarin allerlei fysische wetten, van zwaartekracht en vloeistofdynamica tot stervorming en het voeden
van een centraal zwart gat, zijn ingeprogrammeerd. We kiezen ervoor simulaties van
botsende sterrenstelsels te bestuderen omdat de stelsels voor, tijdens en na de botsing
verschillende gedaanten en kleuren aannemen. Bovendien groeit het vermoeden dat
zulke gebeurtenissen een belangrijke rol spelen in de evolutie van sterrenstelsels. Gegeven een beginconditie, in ons geval twee schijfstelsels die op het punt staan te botsen,
rekent het programma de tijdsevolutie van het systeem uit. Op die wijze kan voor iedere willekeurige fase in de simulatie de totale massa aan sterren, hun gemiddelde
leeftijd en de hoeveelheid licht dat geabsorbeerd wordt door stof berekend worden.
Tevens berekenden we de kleuren zoals we ze zouden zien als we de gesimuleerde
sterrenstelsels vanaf 10 miljard lichtjaar zouden waarnemen. Op basis van deze synthetische fotometrie schatten we vervolgens de massa, leeftijd en absorptie door stof
alsof het echte waarnemingen waren. Vergelijking met de echte waarden leert dat de
eigenschappen van rode elliptische, ’volwassen’ sterrenstelsels (de eindfase van de simulatie) goed gereproduceerd kunnen worden. Dit in tegenstelling tot eerdere fases in
de evolutie wanneer er nog actief sterren worden gevormd. In dergelijke omstandigheden vormen systematische onderschattingen van de massa en leeftijd met een factor
1.5 geen uitzondering. Het onderscheid tussen veel en erg veel absorptie door stof is
nauwelijks te maken.
173
In hoofdstuk 7 tenslotte gebruiken we onze kennis van afstanden tot en fysieke eigenschappen van ver weg gelegen sterrenstelsels om een theoretisch model te toetsen
dat een essentiële rol toeschrijft aan quasars in de evolutie van sterrenstelsels en met
name de vorming van rode sterrenstelsels. Zoals eerder vermeld, zijn quasars een soort
actieve sterrenstelsels. Hoog-energetische processen zorgen ervoor dat deze objecten
tot de grenzen van het waarneembaar heelal betrekkelijk helder oplichten. De huidige consensus is dat een dergelijke gigantische hoeveelheid energie bij de aanvoer van
materiaal naar het centrale supermassieve zwarte gat van een sterrenstelsel vrijkomt.
Simulaties van botsende sterrenstelsels tonen aan dat tijdens de botsing voldoende
materiaal naar het centrum wordt gevoerd om zowel een grote hoeveelheid sterren
te vormen als een zwart gat te voeden. Wanneer de quasar actief wordt, verhindert
die volgens het model nieuwe stervorming. Aannemend dat iedere quasar die in het
heelal wordt waargenomen correspondeert met een botsing tussen twee gasrijke sterrenstelsels, vertaalt het model vervolgens de waargenomen hoeveelheid quasars naar
het aantal botsende stelsels in de loop van de tijd. Wij toetsen twee voorspellingen van
dit model door vergelijking met onze waarnemingen. Ten eerste het aantal massieve sterrenstelsels (in aantal per volume en massa per volume, en voor ’volwassen’ en
stervormende stelsels afzonderlijk). We vinden hierbij een opmerkelijk goede overeenkomst. De resultaten zijn consistent met de stelling dat ieder rood ’volwassen’ stelsel
ooit een quasar fase heeft ondergaan. Ten tweede vergelijken we de kleurverdeling
van waargenomen en gesimuleerde stelsels. De voorspelde kleuren voor ’volwassen’
stelsels zijn in overeenstemming. Dit geldt echter niet voor de kleurverdeling van stervormende stelsels. De rode kleuren van stervormende stelsels met veel stof worden
niet gereproduceerd door het model.
Curriculum vitae
O
14 december 1980 ben ik geboren te Mortsel (België). Toen ik als 14-jarige besefte dat veel van de pioniersgeest van de ruimtevaart in 1972 op de maan was
achtergelaten, besloot ik niet astronaut maar astronoom te worden. Wij leefden toen
nog in het enige zonnestelsel, het heelal dijde nog niet versneld uit en voorbij 10 miljard lichtjaar leefden slechts een handvol quasars en radiostelsels. Zelf volgde ik de
richting Latijn-Wiskunde aan het Sint-Jan Berchmanscollege te West-Malle.
In september 1998 stak ik de rivieren over en begon de opleiding sterrenkunde aan
de Universiteit Leiden. Tijdens mijn studie volgde ik te Castel Gandolfo (Vaticaanstad)
een zomerschool over sterrestanten en onderzocht ik tijdens een zomerproject aan
Caltech onder begeleiding van Prof. dr. Pieter van Dokkum de vormingsgeschiedenis
van vroeg-type sterrenstelsels in clusters. In december 2002 slaagde ik cum laude voor
het doctoraalexamen sterrenkunde.
In januari 2003 luidde een waarneemsessie met de 10 meter Keck telescoop mijn
promotietijd in. Dit proefschrift beschrijft de resultaten van het onder begeleiding
van Prof. dr. Marijn Franx en Prof. dr. Pieter van Dokkum verrichte onderzoek. Ik
nam deel aan conferenties in Leiden, Cambridge, Vlieland, Marseille, Agios Nikolaos
(Kreta) en Durham. Verder presenteerde ik mijn onderzoeksresultaten in Pasadena
(SSC en Carnegie Observatories), Tucson (NOAO), New Haven (Yale), Cambridge
(CfA), Baltimore (STScI) en Heidelberg (MPIA). Een belangrijk deel van mijn onderzoek
werd uitgevoerd tijdens werkbezoeken aan Prof. dr. Pieter van Dokkum (Yale), Dr. Ivo
Labbé (Carnegie Observatories) en Prof. dr. Lars Hernquist (CfA). Tijdens de NOVA
herfstschool (Dwingeloo) en Novicosmo zomerschool (Novigrad, Kroatië) kreeg ik de
gelegenheid mijn sterrenkundige kennis te verbreden. Met plezier heb ik geassisteerd
bij het Sterrenkundig Practicum en het college Sterren. Tevens gaf ik een tiental lezingen
voor niet-sterrenkundig publiek in heel Nederland.
Na mijn promotie zal ik mijn onderzoek naar de evolutie van sterrenstelsels voortzetten als Keck fellow aan het Harvard-Smithsonian Center for Astrophysics (Cambridge, USA).
P
175
Nawoord
D
sterrenstelsels die ik bestudeerde, gaan zonder naam, slechts voorzien van een
identificatienummer, door het leven. Op deze plek wil ik graag de regels omtrent
nomenclatuur doorbreken en de verre lichtbronnen opdragen aan al wie van dichtbij
heeft geholpen bij de totstandkoming van dit proefschrift.
E
De Sterrewacht bood me iedere dag een fijne en stimulerende werkomgeving, waar
dankzij de inzet van het secretariaat en de computergroep alles probleemloos verliep.
In het bijzonder wil ik mijn groepsgenoten in binnen- en buitenland bedanken voor
alle hulp, boeiende discussies en motiverende gesprekken: Ivo, Natascha, Ned, Arjen,
Maaike, Mariska, Rik, Ryan, Phil en TJ, ik heb veel van jullie geleerd. Verscheidene
werkbezoeken aan Yale, Carnegie en CfA werden financieel mogelijk gemaakt door
het Leids Kerkhoven-Bosscha Fonds. De mooie herinneringen aan die periodes in het
buitenland heb ik te danken aan de gastvrijheid en het vrolijke gezelschap van Milan,
Michelle, Kamson en de Juan gang: Tamara, Ben en Anton, binnenkort zien we elkaar
vaker.
Eenmaal uitgereisd en terug op de Sterrewacht kon ik altijd een beroep doen op lotgenoten Dominic, Isa, Simon, Liesbeth, Martijn en Roderik om het kantinevoedsel te
trotseren en te mijmeren over het sterrenkundige leven.
Minder wetenschappelijk, maar net zo essentieel, was de steun van familie en vrienden
buiten de wetenschap bij het realiseren van dit proefschrift. In de eerste plaats mijn
ouders, die me van jongsaf hebben gesteund in het nastreven van mijn dromen zonder
ooit verwachtingen op te leggen. Tante Imelda wil ik bedanken voor haar prachtige
schilderij op de omslag.
Tenslotte het Leidse leven; SKC H1 leerde me driemaal per week dat niets te serieus
genomen moet worden. De Volleyrd Lopers leverden een unieke combinatie van puur
sportgevoel en hechte vriendschap, een gezonde geest én een gezond lichaam. Ilona,
Ruud, Diana, het is goed te weten dat je altijd bij iemand terecht kan. Femke, dankzij
jou weet ik dat de wereld groter is dan het heelal alleen. Bedankt!
177
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement