STAR FORMATION AND AGING AT COSMIC NOON:

STAR FORMATION AND AGING AT COSMIC NOON:
STAR FORMATION AND AGING
AT COSMIC NOON:
the spectral evolution of galaxies from z=2
STAR FORMATION AND AGING
AT COSMIC NOON:
the spectral evolution of galaxies
from z = 2
Proefschrift
ter verkrijging van
de graad van Doctor aan de Universiteit Leiden,
op gezag van Rector Magnificus prof.mr. C.J.J.M. Stolker,
volgens besluit van het College voor Promoties
te verdedigen op dinsdag 8 september 2015
klokke 10:00 uur
door
Mattia Fumagalli
geboren te Lecco, Italië
in 1986
Promotiecommissie
Promotores:
Overige leden:
Prof. dr. M. Franx
Prof. dr. P. G. van Dokkum
Prof. dr. H. J. A. Röttgering
Prof. dr. K. H. Kuijken
Prof. dr. J. Schaye
Prof. dr. M. Kriek
dr. K. I. Caputi
dr. I. Labbé
Yale University
University of California at Berkeley
Rijksuniversiteit Groningen
Cover: 3D-HST stacks of spectra of star-forming galaxies pictured as mountain ranges. Peaks fade to the horizon as redshift increases.
Designed by Mattia Fumagalli and Marco Vedoá
To Anna
Two roads diverged in a wood, and I
I took the one less traveled by
(Robert Frost)
Contents
1 Introduction
1.1 The Birth of Extragalactic Astronomy . . . . . .
1.2 The Striking Diversity of Galaxies . . . . . . . .
1.3 Measuring Star Formation through cosmic time
1.4 Issues at high redshifts . . . . . . . . . . . . . . .
1.5 This Thesis . . . . . . . . . . . . . . . . . . . . . .
2
3
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
1
3
6
7
Hα Equivalent Widths from the 3D-HST survey
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 3D-HST . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2 SDSS . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.3 VVDS . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.4 High redshift data . . . . . . . . . . . . . . . . . . . .
2.3 The EW(Hα) - mass relation . . . . . . . . . . . . . . . . . . .
2.4 The Evolution of EW(Hα) with redshift . . . . . . . . . . . .
2.5 The sSFR(Hα) - mass relation and its evolution with redshift
2.6 Linking the characteristic SFH of galaxies and EW(Hα) . . .
2.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13
14
15
15
15
16
16
16
18
18
21
22
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
25
26
26
29
29
30
31
31
33
36
36
36
38
38
40
43
43
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
How dead are dead galaxies?
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Sample selection and motivations of the study . . . . . . . . . . . .
3.3.1 Selection of Quiescent Galaxies . . . . . . . . . . . . . . . . .
3.3.2 Spectra and SEDs of the sample . . . . . . . . . . . . . . . .
3.3.3 SFRs from SED fitting and expectations from gas recycling
3.3.4 How much star formation could be hidden? . . . . . . . . .
3.4 Measuring Obscured Star-Formation Rates of Quiescent Galaxies .
3.5 Other possible contributions to LIR . . . . . . . . . . . . . . . . . . .
3.5.1 AGN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5.2 Circumstellar dust . . . . . . . . . . . . . . . . . . . . . . . .
3.5.3 Cirrus dust . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.A Appendix A: Photometry . . . . . . . . . . . . . . . . . . . . . . . .
3.B Appendix B: Field-to-field variation . . . . . . . . . . . . . . . . . .
i
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
5
Stacked spectra
4.1 Introduction . . . . . . . . . . .
4.2 Data . . . . . . . . . . . . . . . .
4.2.1 The 3D-HST survey . .
4.2.2 Sample Selection . . . .
4.3 Methods . . . . . . . . . . . . .
4.3.1 Stacking . . . . . . . . .
4.3.2 Model fitting . . . . . .
4.4 Quiescent Galaxies . . . . . . .
4.4.1 Quality of fits . . . . . .
4.4.2 Determination of Ages .
4.5 Star Forming Galaxies . . . . .
4.5.1 Quality of fits . . . . . .
4.5.2 Determination of Ages .
4.6 Discussion . . . . . . . . . . . .
4.6.1 Differences among SPSs
4.6.2 Evolution of Ages . . .
4.6.3 Hα in quiescent galaxies
4.7 Conclusions . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Decreasing Hα for redder star-forming galaxies: influence of dust and star
formation rates
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Data and Sample Selection . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1 The 3D-HST survey . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.2 Sample selection and the UVJ diagram . . . . . . . . . . . . . .
5.3 EW(Hα): trend with color . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.1 Separating star-forming and quiescent galaxies with the UVJ
selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.2 Color dependence of EW(Hα) for star-forming galaxies . . . .
5.4 Specific Star Formation Rates of star-forming galaxies: trend with color
5.5 Dust absorption of star-forming galaxies along the UVJ diagram . . .
5.5.1 Absorption in Hα . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.2 Absorption of the continuum . . . . . . . . . . . . . . . . . . . .
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
50
51
51
53
55
55
56
57
57
59
60
60
62
63
63
63
67
68
71
72
73
73
74
75
75
77
77
80
80
83
87
Samenvatting
91
Publications
97
Curriculum Vitae
103
Acknowledgments
105
ii
1
Introduction
1.1
The Birth of Extragalactic Astronomy
At the beginning of the XX century, our perception of the size and the structure
of the Universe dramatically changed. If we had to set a symbolic date for that
paradigm shift, we should go back to April the 26th, 1920.
On that day, two influential astronomers of the time, Harlow Shapley and Heber
Curtis, debated the nature of spiral galaxies and the size of the Universe in front of a
crowded auditorium at the Smithsonian Museum of Natural History, in Washington
DC. Shapley argued in favor of the Milky Way, the faint stripe of stars visible in the
sky during a dark clear night, as the entirety of the Universe. He believed that spiral
nebulae, such as the ones classified by Messier and Herschel in the XVIII century,
were part of our own galaxy. Curtis instead thought that Andromeda and the other
nebulae were separate galaxies, or island universes (as Immanuel Kant had defined
them one hundred years before).
Both scholars were backing their claims with different observations available at
the time. However, the main support for Shapley’s theory, i.e. the observation of
the rotation of the Pinwheel Galaxy (which would have implied a distance smaller
than the radius of the Milky Way disk) by Adriaan van Maanen, was soon shown
to be incorrect. Observations by Edwin Hubble in the next years finally settled the
debate. In 1922 Edwin Hubble measured the periods of Cepheids (a type of variable
stars) in the outskirts of the Andromeda Nebula. Thanks to the work of Henrietta
Swan Leavitt (1912), Cepheid stars were known to have a tight relation between
their luminosity and the period of their variability. Hubble’s observations showed
incontrovertibly that Andromeda was in fact a separate island Universe, far outside
the Milky Way.
It was again Hubble, a few years later (1927), who found a rough proportionality
between the distance of galaxies and their receding velocity: since then, astronomers
started to realize that the Universe was expanding. The observations by Edwin
Hubble marked the start of modern observational cosmology; it is not by chance
that the most ambitious space telescope orbiting the Earth is named after him.
1.2
The Striking Diversity of Galaxies
Even though the large diversity in morphology of nebulae was identified since the
XVIII century, the most common classification scheme for galaxies in use today is
1
Chapter 1
2
due, again, to Edwin Hubble (1926). Hubble noticed that galaxies could be roughly
separated in two classes: elliptical galaxies, consisting of a round or flattened smooth
distribution of light, and spiral galaxies, consisting of a flat disc with spiral structures extending from a central concentration of light (known as the bulge).
Hubble referred to elliptical and lenticular galaxies as "early-type", and spirals as
"late-types" with no intent of this nomenclature to be an evolutionary path, contrary
to popular belief (Hubble, 1927):
The nomenclature, it is emphasized, refers to position in the sequence,
and temporal connotations are made at one’s peril. The entire classification is purely empirical and without prejudice to theories of evolution.
The definition morphology of galaxies have been made quantitative by Sérsic in
the Sixties, who proposed to fit the surface brightness profile of a galaxy (i.e. how
the intensity of light varies from the center) with a parametric function of the form:
Σ(r) = Σe × exp −bn
"
r
re
#!
1/n
−1
where re is the radius within which the galaxy emits half of his brightness and Σe is
the surface brightness at re . The value of n determines how concentrated the profile
is, with particular cases of n = 1 corresponding to a disk-like profile, and n = 4
corresponding to a bulge-like profile.
Subsequent studies have shown that morphology of present-day galaxies is tightly
correlated to other properties such as mass, color, and environment.
In general, elliptical galaxies have redder colors than spirals (Strateva et al., 2001;
Blanton et al., 2003; Driver et al., 2006), reflecting the fact that the light of elliptical
galaxies tends to be dominated by old stars, while spirals tend to be actively forming
new stars. Higher-density environments tend to be dominated by early-type galaxies
(Dressler, 1980, Blanton et al., 2005), and the most massive galaxies tend to be earlytype as well (Kauffmann et al., 2003b).
In the local Universe, this dichotomy can be interpreted as being primarily an
effect of mass (even though further studies such as Franx et al. 2008 hint that velocity dispersion might be a more fundamental parameter to describe this transition). Kauffmann et al. (2003b) have shown that the distinction between evolved
early-type, red, quiescent objects, and late-type, blue, star-forming galaxies occurs
at M∗ = 3 × 1010 M .
A similar bimodality in colors, star-formation rates, and morphology of galaxies
has been observed all the way to z ∼ 2 (e.g. Labbé et al. 2005, Kriek et al. 2006,
Szomoru et al. 2012). However, while in the local Universe massive (M∗ > 1011 M )
galaxies constitute a substantially uniform population of red-and-dead objects, at
z ∼ 1.5 a high fraction (60%, compared to 10% in SDSS) of them is found to be
star-forming, blue, and disk-like (Figure 1.1, van Dokkum et al. 2011). The fraction
of quiescent galaxies increases towards lower redshifts while star-forming galaxies
tend to dominate the galaxy counts at progressively lower mass (Brammer et al.
2011, Muzzin et al. 2014). Understanding the origin and the evolution of the galaxy
Introduction
3
bimodality, and the process that quenches galaxies are perhaps the most fundamental questions of extragalactic astrophysics.
Figure 1.1: The diversity of massive galaxies at 1.0 < z < 1.5 (from van Dokkum et al. 2011),
in a mass selected sample (logM∗ /M > 11) from the 3D-HST survey. The relation between
EW(Hα), morphology (parametrized by the Sersic index) and color of galaxies shows that the
high redshift population is made up of a group of quiescent, red, elliptical galaxies with low
star-formation rates, complemented by blue, spiral-like, star-forming objects.
1.3
Measuring Star Formation through cosmic time
One of the most fundamental parameter describing a galaxy is the star formation
rate (SFR), defined as the solar masses formed as stars per unit time. As the youngest
stellar population emit the bulk of their energy in the rest-frame ultraviolet (λ <
3000 Å), the most direct way to measure SFRs consists in integrating the light of
galaxies at those wavelengths. However, since stars form within clouds of gas and
dust, the light they emit is at least partially attenuated and therefore any measure
of SFR from the UV might be light might be severely underestimated.
One can either correct for the dust absorption, by for instance comparing the
observed UV spectrum with the theoretical slope one would expect the spectrum
to have (e.g. Meurer et al. 1995, Bouwens et al. 2009), or consider the additional
contribution of the light absorbed in the UV and re-emitted at longer wavelengths
(Bell et al. 2005, among others). Infrared measurements are however challenging
4
Chapter 1
as well, and SFRs in the infrared are often inferred from observations at a single
wavelength, from which the total IR luminosity is extrapolated under assumptions
on the overall IR spectral shape.
Since young massive stars produce copious amounts of ionizing photons that
ionize the surrounding gas, Hydrogen recombination lines (including the Balmer
lines at optical wavelengths) represent the most traditional SFR indicator (Kennicutt 1998). The relation between the ionizing photon rate and the intensity of an
hydrogen recombination line is dictated by quantum mechanics (e.g. Osterbrock &
Ferland, 2006). Dust absorption affects optical Balmer lines to a lesser extent that
UV - light, but dust corrections are still necessary. A commonly used technique
consists in estimating the dust absorption in a system by comparing the ratio of the
intensities of two emission lines (generally Hα and Hβ) to that expected by quantum
mechanics in the absence of dust.
More indirect SFR indicators are based on radio and X-ray emission, respectively
based on the acceleration of cosmic rays in supernovae explosions and the number of
high-mass X-ray binaries, both correlated with the presence of young stars. At these
wavelengths however, active galactic nuclei often dominate the emission, making
SFR measurements more uncertain.
A comprehensive measurement of the SFRs at large lookback times has necessarily to rely on different tracers at different redshifts. Various measurements (among
others: Madau et al. 1996; Lilly et al. 1996; Bouwens et al. 2007; Karim et al.
2011; Sobral et al. 2013; Madau & Dickinson 2014) of the star formation rate density
(SFRD) at different cosmic times give an indication of the star formation activity of
the universe (from z=8 to 0). Even though large uncertainties in the determination
of the SFRD still exist, the global picture is well established: the SFRD rises from the
Big Bang to a peak at z ∼ 2 (“Cosmic Noon”), and afterwards falls by a factor of
approximately 10 to the current value (Figure 1.2). Measuring the time evolution of
the SFRD has implications for the reionization of the Universe, the cosmic chemical
evolution, the transformation of gas into stars and the buildup of stellar mass.
In order to understand the physical processes driving the evolution of the Universe, one would ideally want to go beyond the global description of the SFRD
and trace galaxy evolution on a galaxy-by-galaxy base, by connecting star formation
rates with other physical properties of galaxies. In that respect, with the building of
large statistical samples (SDSS) it became possible to establish that, for star-forming
galaxies, mass and star-formation rates are tightly correlated, with most of the objects having a linear (or slightly sublinear) relation between log M∗ /M - log SFR,
with a relatively small scatter (0.2 dex, Brinchmann et al. 2004).
In the meanwhile, infrared telescopes (IRAS, Spitzer) had already led to the identification of a different class of star-forming galaxies, with infrared luminosities and
star-formation rates 100 to 1000 times higher than those of the Milky Way (Lonsdale
et al. 2006, and references therein). Those rare objects have been dubbed (U)LIRG,
i.e. (Ultra)Luminous InfraRed Galaxies, and turned out to be mainly products of
merging or interacting galaxies inducing huge bursts of star formation (e.g. Armus,
Heckman & Miley, 1987).
Looking back in the past, ULIRG-like star-formation rates appeared to be more
Introduction
5
Figure 1.2: Evolution of the cosmic star formation density, from Madau & Dickinson (2014).
The star formation rate of the Universe reached a peak around z ∼ 2 and has declined by a
factor of 10 since then. Determinations based on infrared measurements are shown in red,
determinations based on UV measurements in blue/green/magenta.
Figure 1.3: From Whitaker et al. 2012, the SFR-mass sequence for star-forming galaxies out to
z = 2.5. At each redshift, more massive galaxies have higher SFRs than those of lower mass
galaxies, with a non-linear slope. The normalization of the sequence increases towards higher
redshifts.
Chapter 1
6
prevalent (e.g. Lilly et al. 1996, Cowie et al. 1996). However subsequent studies
showed that z ∼ 1 star-forming galaxies, despite the similar high star-formation
rates, had nothing in common with local (U)LIRGS, while instead resembled relaxed
disks, with less than 20% of them being interacting systems (Zheng et al. 2004, Bell
et al. 2005). Subsequent studies showed that a relation between stellar mass and starformation rates, similar to that seen in the local Universe, was present for galaxies
at high redshift too (Noeske et al. 2007, Elbaz et al. 2011, and others).
Since star-formation is thought to be regulated by the balance between the accretion rate of cold gas onto the galaxy and some feedback process (e.g., Dutton
et al. 2010; Bouche et al. 2010), the star-forming main sequence may be a natural
consequence of “cold mode accretion” (e.g., Birnboim & Dekel 2003), as the SFR is
approximately a steady function of time and yields a relatively tight relationship
between SFR and M∗ .
The normalization of the star-forming main sequence increases towards higher
redshifts (Karim et al 2010, Whitaker et al. 2012, Figure 1.3), with a slope that is
generally steeper than that predicted by semi-analytical models of galaxy formation
(e.g. Guo et al. 2010). The fact that the main sequence shifts towards lower values
as the Universe gets older reflects a gradual decline of the average star-formation
in most individual galaxies, as gas gets gradually exhausted, accompanied by an
increase in the fraction of quenched galaxies (e.g. Muzzin et al. 2014).
1.4
Issues at high redshifts
Despite the invaluable technological advances of instruments and telescopes in the
last twenty years, starting with the building of 8-10m class telescopes and the new
generation of space telescopes (such as Hubble and Spitzer), measurements of highredshift galaxies are still extremely challenging, and our knowledge of those systems
is nowhere near to that we have of local galaxies.
In the first place, galaxies become fainter as their distances increase. Spectroscopy of high-redshift galaxies is therefore prohibitively time-consuming for all
but the brightest sources. The most fundamental problem related to the absence of
spectroscopy is that the redshift determination is uncertain. High-redshift surveys
use multiband-photometry to obtain a spectral energy distribution (SED) of galaxies, and fit those with a set of modeled SEDs in order to derive redshifts and other
physical properties of galaxies. Even in extragalactic fields covered by 20-30 different photometric bands spanning from the UV to the NIR photometric redshifts are
hardly more precise than δz/(1 + z) = 3% once compared to spectroscopic redshifts
of generally bright objects with emission lines (Skelton et al. 2014). Estimation of
galaxy masses from photometry are affected by systematic uncertainties of the order
of 0.3dex or more (e.g. Muzzin et al. 2009, Dahlen et al. 2013, Pacifici et al. 2014).
A second fundamental problem for observations at high redshift is that their light
is strongly redshifted. Light from old stars (representing the bulk of mass in most
galaxies) is redshifted into the infrared at redshifts higher than z ∼ 0.5. This causes
problems because of the inefficiency of infrared detectors, and the atmospheric absorption in those bands caused by water vapor. Even in the most ideal places for
Introduction
7
infrared ground-based telescopes (dry locations at high altitude), the transparency
of the Earth’s atmosphere is limited except in a few infrared wavelength windows.
Our knowledge of ages, star-formation rates, and metallicities of galaxies relies often on spectral indicators at rest-frame optical wavelengths (such as D4000,
Balmer and metal lines), which are challenging and time-consuming to measure at
high redshift when they shift in the infrared. For instance, a well-calibrated standard indicator of the SFR is the already mentioned Hα luminosity (Kennicutt, 1998).
As a consequence of its shift into the near-IR at redshifts higher than z ∼ 0.5 (8
billion years ago), studies of the evolution of star-formation rates covering a wide
redshift range use diverse SFR indicators (such as UV, IR, [OII], SED fitting), relying
on a set of assumptions and inter-calibrations. For each indicator, accessing fluxes
correspondent to SFR < 10 − 20M /yr becomes challenging, if not impossible, for
individual sources at z > 0.5. The identification of samples of galaxies with low
star formation at high redshift is therefore generally based on their rest-frame colors
only, by selecting galaxies whose optical and near-IR light is dominated by an old
stellar population.
An additional bias induced by ground-based spectroscopy is that samples for
spectroscopy are generally optimized for observations in the atmospheric windows,
and they are consist generally in blue star-forming objects selected on the basis
of their rest-frame UV emission (Steidel et al. 2004, and others), while continuum
observations are available for limited samples of bright objects (Bezanson et al. 2013,
van de Sande et al. 2013). The absence of bias-free and mass-complete samples of
measurements of physical properties of galaxies such as ages and metallicities limits
our understanding of the assembly history and the evolution of galaxies.
1.5
This Thesis
This thesis addresses several of the issues described in the previous section. In
particular, we take advantage of a novel set of observations taken with the Wide
Field Camera 3 (WFC3) grism onboard Hubble Space Telescope (HST), in the context
of the 3D-HST survey (Figure 1.4, Brammer et al. 2012), in order to investigate the
evolution of star-formation rates, emission line contributions and stellar population
properties of both star-forming and quiescent galaxies, in mass selected samples at
0.5 < z < 2. 3D-HST provides rest-frame optical spectra for a sample of ∼ 10000
galaxies at 1 < z < 3.5, the epoch when 60% of all star formation took place, the first
galaxies stopped forming stars, and the structural regularity that we see in galaxies
today must have emerged. Such a wide-field near-IR spectroscopic survey would be
currently infeasible from the ground, since it proves a larger cosmic volumes thanks
to the broad range of redshifts covered by the WFC3 grism, and targets every object
in the field of view.
In Chapter 2, we combine the first available data from the 3D-HST survey (40 %
of the entire survey) with those of ground-based surveys at lower redshift in order
to evaluate the evolution of EW(Hα), the equivalent width of Hα. Since EW(Hα)
8
Chapter 1
Figure 1.4: The 3D-HST survey provides spectra for all galaxies in a particular field with the
WFC3/IR grism. The panels on the left show 50x28 arcsec cutouts of the F140W and G141
observations within the GOODS-South field, with wavelength increasing towards the right
on the grism panel. Galaxy spectra are extracted in 2D and 1D (bottom right) and used in
combination with the full SED of the objects (top center) in order to determine a redshift
measurement which is greatly improved to that from photometry alone: the top-right panel
shows the probability distribution of the redshift determined from the photometry alone (grey
region), and that determined with the addition of grism data (black region), compared to a
spectroscopic redshift (vertical line). Image from Brammer et al. (2012).
is defined as the ratio of the Hα luminosity to the underlying stellar luminosity,
it represents a measure of the current to past star formation, and it is therefore
a model-independent, directly observed proxy for the specific star formation rate
(sSFR=SFR/M). We find that at each redshift EW(Hα) goes down with mass, and that
at fixed mass the EW(Hα) grows towards higher redshifts as EW(Hα) = (1 + z)1.8 .
This evolution is independent of stellar mass, and it is steeper than that predicted by
models of galaxy evolution. We moreover predict the evolution of EW(Hα) at higher
redshift, finding that the contribution of emission lines to the total light of galaxies
continues to increase at z = 4 − 8, with important consequences for spectroscopy
and photometry of sources that will be accessed with James Webb Space Telescope.
In Chapter 3 we investigate the SFRs of galaxies selected as quiescent on the basis of their optical and near-IR spectral energy distributions, which indicate an old
stellar population. Spectral energy distribution fits for optically selected quiescent
galaxies indicate SFRs even lower than those expected from gas recycling, assuming
that the mass loss from evolved stars refuels star formation. However, optical and
near-IR SED fitting can miss star formation if it is hidden behind high dust obscuration, and its ionizing radiation is reemitted in the mid-infrared. We therefore select
spectroscopically confirmed quiescent galaxies in the 3D-HST survey, and measure
Introduction
9
their dust-obscured SFRs with stacks of mid-infrared fluxes from Spitzer-24µm, in
five redshift bins centered on z = 0.5, 0.9, 1.2, 1.7, 2.2. We show that, at each redshift, SFRs of quiescent galaxies are 20-40 times lower than those of star-forming
galaxies at the same redshift, indicating that quenching is very efficient even in the
young Universe where typical SFRs on the main sequence reach hundreds of solar masses per year. The true SFRs of quiescent galaxies might be even lower than
that, as we show that mid-infrared fluxes can be due also to processes uncorrelated
with present star formation, such as dust heating by old stellar populations and
circumstellar dust.
Chapter 4 focuses on the spectra of star-forming and quiescent galaxies from
z=0.5 to z=2 in more detail, in order to determine their stellar ages. We stack spectra of quiescent and star-forming galaxies (selected on the basis of a rest-frame
color-color technique), and fit them with commonly used stellar population synthesis models. We find that stellar population models fit the observations well at
wavelengths lower than 6500Å, while they show systematic differences from the observed spectra at redder wavelengths. We show that quiescent galaxies have little
emission line contribution, and those are consistent with SFR measurements from
mid-infrared. The ages of quiescent galaxies implied by the models differ according
to the model in use, but on average quiescent galaxies are young, i.e. younger than
half of the age of the Universe at each redshift. For star-forming galaxies the inferred
ages depend strongly on the assumed stellar population model and star-formation
history.
In Chapter 5 we take advantage of the full 3D-HST data to analyze how the
EW(Hα) depends on galaxy properties and in particular on the optical/near-IR spectral energy distribution shape of the galaxy, in the redshift range where Hα can be
observed with the HST/WFC3 grism (0.7 < z < 1.5). We demonstrate that galaxies
with strong and weak Hα are well separated in a rest-frame color-color diagram. For
star-forming galaxies, we investigate how Hα varies as a function of the rest-frame
colors of the galaxy and how it relates to the specific star formation rate, measured
from the ultraviolet and mid-infrared emission. At a fixed mass, red star-forming
galaxies have lower EW(Hα) than blue star-forming galaxies. We also show that, at
fixed mass, the median specific star formation rates of galaxies decreases towards
redder U-V colors, and that the dust absorption increases towards redder colors. We
show that the overall variation of EW(Hα) as a function of color can be explained by
the combined effect of lower specific star formation rate and higher dust absorption
for galaxies with redder colors.
Bibliography
Armus, L., Heckman, T., & Miley, G. 1987, AJ, 94, 831
Bell, E. F., Papovich, C., Wolf, C., et al. 2005, ApJ, 625, 23
10
Chapter 1
Bezanson, R., van Dokkum, P., van de Sande, J., Franx, M., & Kriek, M. 2013, ApJL,
764, L8
Birnboim, Y., & Dekel, A. 2003, MNRAS, 345, 349
Blanton, M. R., Hogg, D. W., Bahcall, N. A., et al. 2003, ApJ, 594, 186
Blanton, M. R., Eisenstein, D., Hogg, D. W., Schlegel, D. J., & Brinkmann, J. 2005,
ApJ, 629, 143
Bouché, N., Dekel, A., Genzel, R., et al. 2010, ApJ, 718, 1001
Bouwens, R. J., Illingworth, G. D., Franx, M., & Ford, H. 2007, ApJ, 670, 928
Bouwens, R. J., Illingworth, G. D., Franx, M., et al. 2009, ApJ, 705, 936
Brammer, G. B., Whitaker, K. E., van Dokkum, P. G., et al. 2011, ApJ, 739, 24
Brinchmann, J., Charlot, S., White, S. D. M., et al. 2004, MNRAS, 351, 1151
Cowie, L. L., Songaila, A., Hu, E. M., & Cohen, J. G. 1996, AJ, 112, 839
Dahlen, T., Mobasher, B., Faber, S. M., et al. 2013, ApJ, 775, 93
Dressler, A. 1980, ApJ, 236, 351
Driver, S. P., Allen, P. D., Graham, A. W., et al. 2006, MNRAS, 368, 414
Dutton, A. A., van den Bosch, F. C., & Dekel, A. 2010, MNRAS, 405, 1690
Elbaz, D., Dickinson, M., Hwang, H. S., et al. 2011, A&A, 533, A119
Franx, M., van Dokkum, P. G., Schreiber, N. M. F., et al. 2008, ApJ, 688, 770
Hubble, E. P. 1926, ApJ, 64, 321
Hubble, E. P. 1927, The Observatory, 50, 276
Karim, A., Schinnerer, E., Martínez-Sansigre, A., et al. 2011, ApJ, 730, 61
Kennicutt, R. C., Jr. 1998, ApJ, 498, 541
Kriek, M., van Dokkum, P. G., Franx, M., et al. 2006, ApJ, 645, 44
Kauffmann, G., Heckman, T. M., White, S. D. M., et al. 2003, MNRAS, 341, 54
Labbé, I., Huang, J., Franx, M., et al. 2005, ApJL, 624, L81
Lilly, S. J., Le Fevre, O., Hammer, F., & Crampton, D. 1996, ApJL, 460, L1
Lonsdale, C. J., Farrah, D., & Smith, H. E. 2006, Astrophysics Update 2, 285
Madau, P., Ferguson, H. C., Dickinson, M. E., et al. 1996, MNRAS, 283, 1388
Madau, P., & Dickinson, M. 2014, ARA&A, 52, 415
Introduction
Meurer, G. R., Heckman, T. M., Leitherer, C., et al. 1995, AJ, 110, 2665
Muzzin, A., van Dokkum, P., Franx, M., et al. 2009, ApJL, 706, L188
Muzzin, A., van der Burg, R. F. J., McGee, S. L., et al. 2014, ApJ, 796, 65
Noeske, K. G., Weiner, B. J., Faber, S. M., et al. 2007, ApJL, 660, L43
Pacifici, C., da Cunha, E., Charlot, S., et al. 2015, MNRAS, 447, 786
Skelton, R. E., Whitaker, K. E., Momcheva, I. G., et al. 2014, ApJS, 214, 24
Sobral, D., Smail, I., Best, P. N., et al. 2013, MNRAS, 428, 1128
Steidel, C. C., Shapley, A. E., Pettini, M., et al. 2004, ApJ, 604, 534
Strateva, I., Ivezić, Ž., Knapp, G. R., et al. 2001, AJ, 122, 1861
Szomoru, D., Franx, M., Bouwens, R. J., et al. 2011, ApJL, 735, L22
van de Sande, J., Kriek, M., Franx, M., et al. 2013, ApJ, 771, 85
van Dokkum, P. G., Brammer, G., Fumagalli, M., et al. 2011, ApJL, 743, L15
Zheng, X. Z., Hammer, F., Flores, H., Assémat, F., & Pelat, D. 2004, A&A, 421, 847
11
12
Chapter 1
2
Hα Equivalent Widths from the
3D-HST survey: evolution with
redshift and dependence on stellar
mass
We investigate the evolution of the Hα equivalent width, EW(Hα), with redshift and
its dependence on stellar mass, using the first data from the 3D-HST survey, a large
spectroscopic Treasury program with the HST-WFC3. Combining our Hα measurements of 854 galaxies at 0.8 < z < 1.5 with those of ground based surveys at lower
and higher redshift, we can consistently determine the evolution of the EW(Hα) distribution from z=0 to z=2.2. We find that at all masses the characteristic EW(Hα) is
decreasing towards the present epoch, and that at each redshift the EW(Hα) is lower
for high-mass galaxies. We find EW(Hα) ∼ (1 + z)1.8 with little mass dependence.
Qualitatively, this measurement is a model-independent confirmation of the evolution of star forming galaxies with redshift. A quantitative conversion of EW(Hα)
to sSFR (specific star-formation rate) is model dependent, because of differential
reddening corrections between the continuum and the Balmer lines. The observed
EW(Hα) can be reproduced with the characteristic evolutionary history for galaxies,
whose star formation rises with cosmic time to z ∼ 2.5 and then decreases to z =
0. This implies that EW(Hα) rises to 400 Å at z = 8. The sSFR evolves faster than
EW(Hα), as the mass-to-light ratio also evolves with redshift. We find that the sSFR
evolves as (1 + z)3.2 , nearly independent of mass, consistent with previous reddening insensitive estimates. We confirm previous results that the observed slope of
the sSFR-z relation is steeper than the one predicted by models, but models and
observations agree in finding little mass dependence.
Mattia Fumagalli; Shannon G. Patel; Marijn Franx; Gabriel Brammer; et al.
The Astrophysical Journal Letters, Volume 757, Issue 2, L22, 2012
13
2. Hα Equivalent Widths from the 3D-HST survey
14
2.1
Introduction
Several studies have combined different star formation indicators in order to study
the evolution of star-forming galaxies (SFGs) with redshift. At a given redshift low
mass galaxies typically form more stars per unit mass (i.e., specific star-formation
rate, sSFR) than more massive galaxies (Juneau et al. 2005, Zheng et al. 2007, Damen
et al. 2009). In addition the sSFR of galaxies with the same mass increases at higher
redshift. However, semi-analytical models and observations are at odds with regards
to the rate of decline of the sSFR towards low-redshift (Damen et al. 2009, Guo et al.
2010).
One of the main observational caveats is that most of the studies covering a wide
redshift range use diverse SFR indicators (such as UV, IR, [OII], Hα, SED fitting).
This is a consequence of the fact that it is difficult to use the same indicator over
a wide range of redshifts. One therefore has to rely on various conversion factors,
often intercalibrated at z = 0 and re-applied at higher redshift.
A well-calibrated standard indicator of the SFR is the Hα luminosity (Kennicutt,
1998). However, Hα is shifted into the infrared at z > 0.5, and it is difficult to
measure due to the limitations of ground-based near-IR spectroscopy. Comparing
measures of Hα at different redshifts has therefore been a challenge. Most of the
Hα studies at high redshift are based on narrow-band photometry (e.g. the HiZELS
survey, Geach et al. 2008).
The 3D-HST survey (Brammer et al., 2012) provides a large sample of rest-frame
optical spectra with the WFC3 grism, which includes the Hα emission in the redshift
range 0.8 < z < 1.5. Taking advantage of the first data from the survey (45% of the
final survey products) we investigate for the first time the star formation history
(SFH) of the Universe with Hα spectroscopy, using a consistent SFR indicator over a
wide redshift range.
We evaluate the dependence of the Hα Equivalent Width, EW(Hα), on stellar
mass (M∗ ) and redshift (up to z ∼ 2), comparing the 3D-HST data with other surveys
in mass selected samples with M∗ > 1010 M . Since EW(Hα) is defined as the ratio
of the Hα luminosity to the underlying stellar continuum, it represents a measure
of the current to past average star formation. It is therefore a model independent,
directly observed proxy for sSFR.
We also derive SFRs from the Hα fluxes. We evaluate the mean sSFR in stellar
mass bins and study its evolution with redshift. The slope of the sSFR-z relation
in different mass bins indicates how fast the star formation is quenched in galaxies of various masses. Finally, we compare our findings to other studies (both observations and models), discussing the physical implications and reasons for any
disagreements.
2. Hα Equivalent Widths from the 3D-HST survey
2.2
15
Data
2.2.1
3D-HST
We select the sample from the first available 3D-HST data. They include 25 pointings in the COSMOS field, 6 in GOODS-South, 12 in AEGIS, 28 in GOODS-North1 .
Spectra have been extracted with the aXe code (Kummel et al., 2009). Redshifts have
been measured via the combined photometric and spectroscopic information using a
modified version of the EAZY code (Brammer, van Dokkum, Coppi, 2008), as shown
in Brammer et al. (2012). Stellar masses were determined using the FAST code by
Kriek et al. (2009), using Bruzual & Charlot (2003) models and assuming a Chabrier
(2003) IMF. The FAST fitting procedure relies on photometry from the NMBS catalogue (Whitaker et al. 2011) for the COSMOS and AEGIS fields, the MODS catalog
for the GOODS-N (Kajisawa et al. 2009) and the FIREWORKS catalogue for GOODSS (Wuyts et al. 2008).
The mass completeness limit is log (M∗ /M ) > 10 at z=1.5 (Wake et al. 2011,
Kajisawa et al. 2010). We select objects with 0.8 < z < 1.5; this resulted in a sample
of 2121 galaxies with log (M∗ /M ) > 10.
In slitless spectroscopy, spectra can be contaminated by overlapping spectra of
neighboring galaxies. The aXe package provides a quantitative estimate of the contamination as a function of wavelength, which can be subtracted from the spectra.
We conservatively use spectra where the average contribution of contaminants is
lower than 10% and for which more than 75% of the spectrum falls on the detector.
After this selection we have 854 objects in the redshift range 0.8 < z < 1.5 (40% of
the objects). The final sample is not biased with respect to the mass relative to the
parent sample.
Line fluxes and EWs were measured as follows.2 We fit the 1D spectra with a
gaussian profile plus a linear continuum in the region where Hα is expected to lie.
We subtract the continuum from the fit and measure the residual flux within 3σ from
the line center of the gaussian. Errors are evaluated including the contribution from
the error on the continuum. We distinguish between detections and non-detections
of Hα with a S/N threshold of 3. The typical 3σ detection limit corresponds to
SFR = 2.8M yr−1 at z=1.5 (Equation 2.2).
Due to the low resolution of the WFC3 grism, the Hα and [NII] lines are blended.
In this work EW(Hα) therefore includes the contribution from [NII]. For the other
datasets, which have higher spectral resolution, we combine Hα and [NII] for consistency with 3D-HST.
2.2.2
SDSS
We retrieve masses and EW(Hα) for the SDSS galaxies from the MPA-JHU catalogue
of the SDSS-DR7. Masses are computed based on fits to the photometry, following
Kauffmann et al. (2003) and Salim et al. (2007). At redshift 0.03 < z < 0.06, for
masses higher than M∗ = 1010 M , the SDSS sample is spectroscopically complete
1 From
program GO-11600 (PI: B. Weiner)
the entire paper the quoted EWs are rest-frame values.
2 Through
2. Hα Equivalent Widths from the 3D-HST survey
16
in stellar mass (Jarle Brinchmann, private communication). We consider as detections only measurements greater than 3Å, as the ones with EW < 3Å are affected by
uncertainties in the stellar continuum subtraction (Jarle Brinchmann, private communication).
In the redshift range 0.03 < z < 0.06, the spectroscopic fiber of SDSS does not
cover the entirety of most galaxies. As a consequence sSFR are evaluated with emission line fluxes and masses from the fiber alone.
2.2.3
VVDS
The VIMOS VLT Deep Survey (VVDS, Le Févre et al. 2005) is a wide optical selected
survey of distant galaxies. Hα is covered by the VIMOS spectrograph at 0.0 < z <
0.4.
Lamareille et al. (2009) released a catalog of 20,000 galaxies with line measurements, complete down to M∗ = 109.5 M at z = 0.5 . Masses are retrieved from
the VVDS catalog; they have been computed though a Bayesian approach based on
photometry (equivalent to Kauffmann et al. 2003 and Tremonti et al. 2004), and
are relative to a Chabrier IMF. We select a sample with redshift 0.2 < z < 0.4 and
M∗ > 1010.0 M containing 741 objects, of which 477 (64 %) have an Hα measurement with S/N > 3. The percentage of Hα detected objects drops to 32% at masses
M∗ > 1011.0 M .
2.2.4
High redshift data
Erb et al. (2006) published EW(Hα) for galaxies selected with the BX criterion (Steidel et al. 2004), targeting SFGs at redshift 2.0 < z < 2.5. We evaluate the completeness of the sample as follows. From the FIREWORKS catalogue (Wuyts et al. 2008)
we reconstruct the BX selection and evaluate the fraction of objects with spectroscopically confirmed redshift 2.0 < z < 2.5 that fall in the BX selection. Percentages
are 44%, 32% and 27% for mass limited samples with log(M∗ /M ) = 10.0-10.5,
log(M∗ /M ) = 10.5-11, log(M∗ /M ) > 11.0.
2.3
The EW(Hα) - mass relation
We first study how EW(Hα) depends on stellar mass in each available data set. The
3D-HST sample has been divided in two redshift bins, 0.8 < z < 1.1 and 1.1 < z <
1.5. The results are shown in Figure 2.1. At each redshift, highest mass galaxies have
lower EW(Hα). Note however, that there is a large scatter in the relation. We quantify
the trend in the following way: we determine the average EW(Hα) in three 0.5 dex
wide mass bins (10.0 < log( M∗ /M ) < 10.5, 10.5<log( M∗ /M )<11, log(M∗ /M ) >
11.0) and evaluated its error through bootstrapping the sample. The mean EW(Hα)
in a given mass bin is obtained in two ways: (1) using only detected, highly SFGs
(blue lines in Figure 2.1) and (2) using all galaxies but assigning EW(Hα)=0 to the
objects detected in Hα with S/N < 3 (red lines in Figure 2.1). For the z = 2.2 data,
2. Hα Equivalent Widths from the 3D-HST survey
EW(Hα+[NII])
1000
0.2<z<0.4 (VVDS)
0.8<z<1.1 (3D-HST)
1000
100
100
10
10
1
1000
1
1.1<z<1.5 (3D-HST)
10.0
2.0<z<2.6 (Erb+06)
100
100
10
10
1
10.5
11.0
log(M*/MO•)
11.5
EW(Hα+[NII])
EW(Hα+[NII])
z=0 (SDSS)
17
1
10.0
10.5
11.0
log(M*/MO•)
11.5
10.0
10.5
11.0
log(M*/MO•)
11.5
Figure 2.1: EW(Hα) against mass, for different redshift samples. Vertical lines represent the
limiting mass of the analysis. Black symbols are objects with Hα detection with S/N > 3 and
red arrows represent upper limits. The green diagonal lines represent the detection limit of
the 3D-HST data. Blue solid lines represent the mean EW(Hα) of detected SFGs, in 0.5 dex
mass bins. Red solid lines represent the mean EW(Hα) of all galaxies, assuming EW(Hα)=0
for non-detected objects. Errors of the means are computed with a bootstrap approach. At
each redshift higher mass galaxies have lower EW(Hα) than less massive objects.
we use the FIREWORKS catalog to establish the fraction of galaxies excluded by the
BX selection and therefore give an estimation of EW(Hα) for all galaxies.
Using either method we find that an EW(Hα)-mass relation is in place at each
redshift, not just for strongly star forming objects but also for the entire galaxy
population. Galaxies in the lowest mass bin (10.0 < log(M∗ /M ) < 10.5 ) have on
average an EW(Hα) which is 5 times higher than galaxies in the highest mass bin
(log( M∗ /M ) > 11.0).
We discuss the evolution of the EW(Hα)-mass relation with redshift in the next
section of the Letter. However, it is immediately evident from Figure 2.1 that the 3DHST survey targets galaxies with EW(Hα) typical of an intermediate regime between
what is seen at z=0 and what is seen at higher redshift. In other words the EW(Hα)mass relation seems to rigidly shift towards higher EW(Hα) at higher redshifts.
A considerable fraction of detected galaxies in 3D-HST have EW ( Hα) > 30Å,
while in SDSS similar objects are extremely rare: 3.8%, 1.4% and 0.4% for increasing
mass samples. This study can be seen as an extension of the findings of van Dokkum
et al. (2011), who reported that massive galaxies at z > 1 show a wider range
of EW(Hα) compared to galaxies in the local Universe. Following this trend with
redshift, in the z > 2 bin we find typical EW(Hα) of 150 Å for SFGs. Such high
18
2. Hα Equivalent Widths from the 3D-HST survey
values of EW(Hα) represent just 11% of the 3D-HST sample (14.6%, 8.1% and 4.7%
respectively for increasing mass bins).
2.4
The Evolution of EW(Hα) with redshift
The evolution of EW(Hα) with redshift can be seen as an observational (i.e. modelindependent) proxy for the sSFR-z relation. Figure 2.2 (top panels) shows the redshift evolution of the average EW(Hα) in different mass bins for the detected SFGs
(top left panel) and for all galaxies (top right panel). A substantial increase of the
EW(Hα) is seen at higher redshifts in both samples. We therefore infer that evolution of the SFR is not a byproduct of selection effects from different SFR indicators.
At 0.8 < z < 1.5, a galaxy has on average an EW(Hα) that is 3-4 times higher than
that of an object of comparable mass in the local universe. For each mass bin we
parametrize the redshift evolution of the EW(Hα) as follows:
EW ( Hα)(z) ∼ A × (1 + z) p
(2.1)
The coefficient p has an average value of 1.8, with little dependence on mass (best
fit values are listed in Table 2.1). As can be deduced from Table 2.1 there may be
a weak mass dependence such that the relations steepen with mass; however, the
difference between the slopes at the lowest and highest mass bin is not statistically
significant.
This indicates that the decrease of EW(Hα) happens at the same rate for all galaxies irrespectively of their masses. As seen in the right panel of Figure 2.2, the addition of non-SFGs amounts to a negative vertical shift in the EW(Hα) but not to a
change in the slope of the relation.
An uncertainty is the effect of dust on the EW(Hα). Without more measurements
we cannot state what effect dust has, and in literature there is disagreement on the
relative extinction suffered by the nebular emission lines and the underlying stellar
continuum (Calzetti et al. 2000, Erb et al. 2006, Wuyts et al. 2011). However, the
data motivated model described in Section 6 suggests that dust has a mild effect on
the EW(Hα).
2.5
The sSFR(Hα) - mass relation and its evolution with
redshift
The EW(Hα) has the advantage that it is a direct observable, but it is more difficult
to interpret than a more physical quantity like the specific star formation rate. The
latter can only be derived with a proper extinction correction for Hα. We lack this
information, as we do not have a proper Balmer decrement measurement. In the
following the briefly explore the specific star formation rate (sSFR) evolution implied
by assuming no extinction and later discuss the effect of a dust correction to the
measured slopes of the sSFR-z relation.
2. Hα Equivalent Widths from the 3D-HST survey
0.0
log(1+z)
0.2 0.3 0.4
0.1
0.5
19
0.6
0.0
10.0 < logM/MO• < 10.5
10.5 < logM/MO• < 11.0
11.0 < logM/MO•
0.6
EW(Hα+[NII])
100
Detected SFGs
ALL
10
Erb+06
3DHST (T.W.)
VVDS
SDSS
1
0
0
0.2
0.3
0.4
0
0.6
0.5
0.0
0.7
0
1
0
1 1.5 2
Redshift
log(1+z)
0.2 0.3 0.4
0.1
Erb+06
3DHST (T.W.)
VVDS
SDSS
0.5
0
3
0.6
-9
-9
-9
-9
-9
-9
-9
10.0 < logM/MO• < 10.5
10.5 < logM/MO• < 11.0
11.0 < logM/MO•
-9
-9
-9
-9
-9
-9
0.3
0.4
-9
-9
-9
-9
-9
-9
-9
-9
-10
-10
-10
-10
-10
-10
-10
-10
-10
-10
-10
-10
-10
-10
0.6
0.7
0
0
1 1.5 2
Redshift
log(1+z)
0.2 0.3 0.4
0.1
10.0 < logM/MO• < 10.5
10.5 < logM/MO• < 11.0
11.0 < logM/MO•
-9
-9
-9
-9
0.2
0.5
0.0
-9
-9
0
0
3
0.5
0.6
-8
-8
A(Hα)=1
-8
-8
-8
-8
-9
-9
-9
-9
-9
-9
-10
-10
-10
-10
-10
-10
-10
-11
-11
-11
-11
Detected SFGs
-11
-11
-11
-11
-11
-11
-11
Erb+06
3DHST (T.W.)
VVDS
SDSS
-11
-11
-11
-11
-11
-12
-12
-12
log sSFR(Hα)
-10
)
α
-10
log sSFR(H
log sSFR(Hα)
-9
-9
-10
-9
-9
-10
-10
-9
-10
-9
-10
-9
-10
-10
-10
-10
-10
-10
-10
-10
-10
-11
-10
-10
-10
-11
-10
ALL
-11
-11
-11
-11
-11
-11
-11
-11
-11
-11
-11
-11
-11
-11
-12
-12
-12
-12
-12
Erb+06
3DHST (T.W.)
VVDS
SDSS
-12
-10
-10
-11
-11
-11
-11
-11
-11
-11
-11
-11
-11
-12
-12
-12
-12
-12
-12
0
0
0
0.2
0.3
0.4
0.6
0.5
0.7
0
0
1 1.5 2
Redshift
-12
3
-12
-12
-12
-12
0
0
0
0.2
0.3
0.4
0.6
0.5
0.7
0
0
1 1.5 2
Redshift
3
Figure 2.2: Evolution of EW(Hα) (top) and sSFR(Hα) (bottom) with redshift, in different mass
bins, for SFGs (left) and all objects (right). Errors on the average EWs have been evaluating through bootstrapping. Dotted lines are the best fit power laws EW (z) ∼ (1 + z) p . At
fixed mass the average EW(Hα) and sSFR(Hα) increase with redshift, with a power law of
EW ( Hα) ∼ (1 + z)1.8 and sSFR( Hα) ∼ (1 + z)3.3 with little mass dependence. The effect of a
luminosity dependent dust correction (Garn et al. 2010) correction is shown by the right axis.
The effect of A(Hα)=1 is shown by the black arrow.
log sSFR(Hα) [DUST CORR]
10
-9
0.5
10.0 < logM/MO• < 10.5
10.5 < logM/MO• < 11.0
11.0 < logM/MO•
100
EW(Hα+[NII])
log(1+z)
0.2 0.3 0.4
0.1
2. Hα Equivalent Widths from the 3D-HST survey
20
-9
-9
-8
-9
-8
-9
-9
-9
Erb+06
-8
-9
-9
-9
A(Ha)=1
-10
-10
3DHST
-9
-9
-9
-9
log sSFR(Hα)
-9
-9
-10
-9
-10
-9
-10
z=2.3
-10
-10
-10
-10
VVDS
-10
-10
-10
z=1.2
-11
-10
-10
-10
-10
-10
-11
-11
SDSS
-10
z=0.9
-10
-10
-11
-11
-11
-11
-11
-11
-11
-11
z=0.3
z=0
-11
-12
-11
-11
-11
-11
-11
-11
-11
-11
-12
-12
-12
9.5
log sSFR(Hα) [DUST CORR]
-8
-9
-12
-12
10.0 10.5 11.0 11.5 12.0
log(M*/MO•)
Figure 2.3: Mean values of sSFR(Hα) in 0.5 dex mass bins at various redshift for SDSS, VVDS,
3D-HST and from Erb et al. (2006). At each redshift more massive galaxies have less sSFR(Hα)
than less massive ones. The effect of a luminosity dependent dust correction (Garn et al. 2010)
correction is shown by the right axis. The effect of A(Hα)=1 is shown by the black arrow.
The SFR is derived from the Hα flux3 using Kennicutt (1998):
SFR( Hα)[ M yr −1 ] = 7.9 × 10−42 × L( Hα)[erg/s] × 10−0.24
(2.2)
where the 10−0.24 factor accounts for a conversion to the Chabrier IMF, from Salpeter
(as in Muzzin et al. 2010).
Figure 2.3 shows the mean value of sSFR(Hα) in different stellar mass bins, at
different redshifts. In each redshift bin higher mass galaxies have lower sSFR(Hα).
In Figure 2.2 (bottom panels) we show the redshift evolution of the average
sSFR(Hα) in different mass bins for detected SFGs (bottom left panel) and for all
galaxies (bottom right panel). A typical galaxy at z=1.5 has as sSFR(Hα) 15-20 times
higher than a galaxy of the same mass at z=0. In each mass bin we fit the evolution
of the sSFR in redshift with a power law:
sSFR(z) ∼ B M × (1 + z)n
(2.3)
obtaining a value of n = 3.2 ± 0.1. As can be deduced from Table 2.1 there may be
a weak mass dependence such that the relations steepen with mass; however, the
slopes at the lowest and highest mass bin are not statistically different.
3 We
assume a [NII]/(Hα+[NII]) ratio of 0.25 for the 3D-HST sources.
2. Hα Equivalent Widths from the 3D-HST survey
21
Table 2.1: Slopes of EW-z and sSFR-z Relation
log( M∗ /M )
10.0-10.5
10.5-11.0
11.0-11.5
EW(det)
1.79 ± 0.18
1.89 ± 0.20
2.21 ± 0.22
EW(all)
1.52 ± 0.21
1.75 ± 0.13
2.12 ± 0.43
sSFR(det)
3.32 ± 0.08
3.18 ± 0.09
3.50 ± 0.12
sSFR(all)
3.06 ± 0.13
3.11 ± 0.18
3.45 ± 0.26
The sSFR-z relation is steeper than the EW-z relation, because of the additional
evolution of the M/L ratio:
EW/sSFR ∼ L( Hα)/L R × M∗ /(K ∗ L( Hα)) ∼ M/L R
(2.4)
where K is the conversion factor in Equation 2.2 and LR is the R-band luminosity.
Implementing a luminosity dependent dust correction for Hα (Garn et al. 2010,
shown with the right axis in Figure 2.2, bottom right) would increase the value
of n to 3.7 ± 0.1. However, several studies (Sobral et al. 2012, Dominguez et al.
2012, Momcheva et al., submitted) have indicated that a better indicator for the
Hα extinction at different redshifts is the stellar mass, and that the Hα extinction
depends strongly on mass but little on redshift (at constant mass). The Garn & Best
2010 relation gives median A(Hα) of 1, 1.5 and 1.7 mag for the increasing mass bins
in this study. A mass-dependent dust correction impacts the normalization of the
sSFR but not the slope n.
Our implied evolution of the sSFR compares well to results from literature. For
example, Damen et al. (2009) found n = 4 ± 1 based on UV + IR inferred sSFRs,
and Karim et al. (2011) found n = 3.50 ± 0.02 for SFGs and n = 4.29 ± 0.03 for all
galaxies, based on stacked radio imaging.
All results indicate an evolution which is steeper with redshift than semianalytical models (Guo & White 2008, Guo et al. 2011), who find slopes close to n=2.5. All
studies find that the slope does not depend on the stellar mass out so z=2. In short,
our results are consistent with previous determinations.
2.6
Linking the characteristic SFH of galaxies and EW(Hα)
We compare the observed evolution of EW(Hα) to what might be expected from
other observations. We construct the typical SFH of a galaxy with mass ∼ 1011 M
at z=0. As a starting point, we assume that the cumulative number density remains
constant with redshift (similar to van Dokkum et al. 2010, Papovich et al. 2011, Patel
et al. 2012). We use the mass functions of Marchesini et al. (2009) and Papovich et
al. (2011), and we show the resulting mass evolution in Figure 2.4c. We determine
the SFR at these masses from Damen et al. (2009), Papovich et al. (2011) and Smit et
al. (2012), and we fit a simple curve to these values (indicated by the curve in Figure
2.4a). This evolutionary history reproduces the mass evolution well (Figure 2.4c).
2. Hα Equivalent Widths from the 3D-HST survey
22
Figure 2.4: Comparison of observed EW(Hα) with predictions from a simple observational
supported model, at different redshifts. (a) Input SFH, and Hα Luminosity (b) Luminosities
at 6563 Å, from the Bruzual & Charlot 2003 code. (c) Mass growth. (d) Evolution of EW(Hα)
with redshift. (e) Evolution of sSFR with redshift. Data points are mean EW(Hα)/sSFR(Hα)
of observed galaxies with mass in a 0.3 dex bin around the typical mass of the model at a
given redshift.
Next we calculate the implied EW(Hα): L(Hα) is derived using Equation 2.2,
and Bruzual & Charlot 2003 models are used to calculate the stellar continuum, assuming solar metallicity (Figure 2.4b). The predicted EW(Hα) rises monotonically
to high redshift, reaching 400Å at z=8 (Figure 2.4d). The predicted EW(Hα) corresponds surprisingly well to the observed EW(Hα) Even the z=4 detections by Shim
et al. (2011) based on broadband IRAC photometry are consistent within the errors.
Apparently, our simple method produces a robust prediction of the evolution of
EW(Hα). We note that the implied sSFR (Figure 2.4e) is higher than expected from
straight measurement of L(Hα), consistently with significant dust extinction. One
magnitude of extinction for Hα is needed to reconcile this discrepancy.
It is remarkable that our prediction worked well for EW(Hα): the average evolution of galaxies was derived from the evolution of the mass function and SFR,
which carry significant (systematic) uncertainties when derived from observations;
whereas the EW(Hα) is a direct observable.
2.7
Conclusions
We have used the 3D-HST survey to measure the evolution of the EW(Hα) from
z=0 to z=2. We show that the EW(Hα) evolves strongly with redshift, at a constant
mass, like (1 + z)1.8 . The evolution is independent of stellar mass. The equivalent width goes down with mass (at constant redshift). The increase with redshift
demonstrates the strong evolution of star forming galaxies, using a consistent and
completely model independent indicator. We explore briefly the implied sSFR evolution, ignoring dust extinction. We find that the evolution with redshift is strong
(sSFR ∼ (1 + z)3.2 ). This stronger evolution is expected as the mass-to-light ratio
2. Hα Equivalent Widths from the 3D-HST survey
23
of galaxies evolves with time, and this enters the correction from EW to sSFR. The
increase with redshift is faster that predicted by semi-analytical models (e.g., Guo &
White 2008), consistent with earlier results.
We construct the characteristic SFH of a 1011 M galaxy. This simple history
reproduces the observed evolution of the EW(Hα) to z=2.5, and even to z=4. It
implies that the EW(Hα) continue to increase to higher redshifts, up to 400 Å at z=8.
This has a significant impact for the photometry and spectroscopy of these high
redshift sources.
The study can be expanded in the future when the entire 3D-HST survey will
be available, doubling the sample and including the ACS grism. In addition to
increased statistics, the ACS grism will allow evaluation of the Balmer decrement
and therefore a precise dust corrected evaluation of SFR. Moreover, a statistically
significant Hα sample at z ∼ 1 will be central to understand the composition, the
scatter and the physical origin of the so called ’star-forming-main sequence’.
We thank the referee for providing valuable comments, and Jarle Brinchmann,
David Sobral and Simone Weinmann for useful discussions. We acknowledge funding from ERC grant HIGHZ no. 227749.
Bibliography
Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008, ApJ, 686, 1503
Brammer, G., van Dokkum, P., Franx, M., et al. 2012, arXiv:1204.2829
Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000
Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682
Chabrier, G. 2003, ApJL, 586, L133
Damen, M., Förster Schreiber, N. M., Franx, M., et al. 2009, ApJ, 705, 617
Damen, M., Labbé, I., van Dokkum, P. G., et al. 2011, ApJ, 727, 1
Domínguez, A., Siana, B., Henry, A. L., et al. 2012, arXiv:1206.1867
Erb, D. K., Steidel, C. C., Shapley, A. E., et al. 2006, ApJ, 647, 128
Garn, T., & Best, P. N. 2010, MNRAS, 409, 421
Garn, T., Sobral, D., Best, P. N., et al. 2010, MNRAS, 402, 2017
Geach, J. E., Smail, I., Best, P. N., et al. 2008, MNRAS, 388, 1473
Guo, Q., & White, S. D. M. 2008, MNRAS, 384, 2
Guo, Q., White, S., Li, C., & Boylan-Kolchin, M. 2010, MNRAS, 404, 1111
24
2. Hα Equivalent Widths from the 3D-HST survey
Guo, Q., White, S., Boylan-Kolchin, M., et al. 2011, MNRAS, 413, 101
Juneau, S., Glazebrook, K., Crampton, D., et al. 2005, ApJL, 619, L135
Kajisawa, M., Ichikawa, T., Tanaka, I., et al. 2009, ApJ, 702, 1393
Kajisawa, M., Ichikawa, T., Yamada, T., et al. 2010, ApJ, 723, 129
Karim, A., Schinnerer, E., Martínez-Sansigre, A., et al. 2011, ApJ, 730, 61
Kauffmann, G., Heckman, T. M., White, S. D. M., et al. 2003, MNRAS, 341, 33
Kennicutt, R. C., Jr. 1998, ApJ, 498, 541
Kriek, M., van Dokkum, P. G., Labbé, I., et al. 2009, ApJ, 700, 221
Kümmel, M., Walsh, J. R., Pirzkal, N., Kuntschner, H., & Pasquali, A. 2009, PASP,
121, 59
Lamareille, F., Brinchmann, J., Contini, T., et al. 2009, A&A, 495, 53
Le Fèvre, O., Vettolani, G., Garilli, B., et al. 2005, A&A, 439, 845
Marchesini, D., van Dokkum, P. G., Förster Schreiber, N. M., et al. 2009, ApJ, 701,
1765
Momcheva, I., Lee, J. C., Ly, C., et al. 2012, arXiv:1207.5479
Muzzin, A., van Dokkum, P., Kriek, M., et al. 2010, ApJ, 725, 742
Papovich, C., Finkelstein, S. L., Ferguson, H. C., Lotz, J. M., & Giavalisco, M. 2011,
MNRAS, 412, 1123
Patel, S. G., van Dokkum, P. G., Franx, M., et al. 2012, arXiv:1208.0341
Salim, S., Rich, R. M., Charlot, S., et al. 2007, ApJS, 173, 267
Shim, H., Chary, R.-R., Dickinson, M., et al. 2011, ApJ, 738, 69
Smit, R., Bouwens, R. J., Franx, M., et al. 2012, arXiv:1204.3626
Sobral, D., Best, P. N., Matsuda, Y., et al. 2012, MNRAS, 420, 1926
Steidel, C. C., Shapley, A. E., Pettini, M., et al. 2004, ApJ, 604, 534
Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al. 2004, ApJ, 613, 898
van Dokkum, P. G., Whitaker, K. E., Brammer, G., et al. 2010, ApJ, 709, 1018
van Dokkum, P. G., Brammer, G., Fumagalli, M., et al. 2011, ApJL, 743, L15
Wake, D. A., Whitaker, K. E., Labbé, I., et al. 2011, ApJ, 728, 46
Whitaker, K. E., Labbé, I., van Dokkum, P. G., et al. 2011, ApJ, 735, 86
Wuyts, S., Labbé, I., Schreiber, N. M. F., et al. 2008, ApJ, 682, 985
Wuyts, S., Förster Schreiber, N. M., Lutz, D., et al. 2011, ApJ, 738, 106
Zheng, X. Z., Bell, E. F., Papovich, C., et al. 2007, ApJL, 661, L41
3
How dead are dead galaxies?
Mid-Infrared fluxes of quiescent
galaxies at redshift 0.3 < z < 2.5:
implications for star-formation
rates and dust heating
We investigate star-formation rates (SFR) of quiescent galaxies at high redshift (0.3 <
z < 2.5) using 3D-HST WFC3 grism spectroscopy and Spitzer mid-infrared data. We
select quiescent galaxies on the basis of the widely used UVJ color-color criteria.
Spectral energy distribution (SED) fitting (rest-frame optical and near-IR) indicates
very low star-formation rates for quiescent galaxies (sSFR ∼ 10−12 yr−1 ). However,
SED fitting can miss star formation if it is hidden behind high dust obscuration and
ionizing radiation is re-emitted in the mid-infrared. It is therefore fundamental to
measure the dust-obscured SFRs with a mid-IR indicator. We stack the MIPS-24µm
images of quiescent objects in five redshift bins centered on z = 0.5, 0.9, 1.2, 1.7,
2.2 and perform aperture photometry. Including direct 24µm detections, we find
sSFR ∼ 10−11.9 ×(1 + z)4 yr −1 . These values are higher than those indicated by
SED fitting, but at each redshift they are 20-40 times lower than those of typical
star-forming galaxies. The true SFRs of quiescent galaxies might be even lower, as
we show that the mid-IR fluxes can be due to processes unrelated to ongoing star
formation, such as cirrus dust heated by old stellar populations and circumstellar
dust. Our measurements show that star-formation quenching is very efficient at
every redshift. The measured SFR values are at z > 1.5 marginally consistent with
the ones expected from gas recycling (assuming that mass loss from evolved stars
refuels star formation) and well below that at lower redshifts.
Mattia Fumagalli; Ivo Labbé; Shannon G. Patel; Marijn Franx; et al.
The Astrophysical Journal, Volume 796, Issue 1, article id. 35, 11 pp., 2014
25
Chapter 3
26
3.1
Introduction
A bimodal distribution in galaxy properties (star-formation rate, size, morphology)
has been observed in the local Universe (e.g. Kauffmann et al. 2003). This bimodality
is made of blue, predominantly late-type galaxies, whose emission is dominated
by young stellar populations and experiencing significant level of star formation,
complemented by red, early-type (elliptical or S0) galaxies dominated by an old
stellar population with little or absent star formation.
The bimodality has been observed all the way to z ∼ 2 (Labbé et al. 2005, Kriek
et al. 2006, Ilbert et al. 2010, Brammer et al. 2011, Whitaker et al. 2013).
Specific star-formation rates (sSFR) from spectral energy distribution (SED) fitting and equivalent widths from emission lines indicate for quiescent galaxies very
low values (log10 sSFR · yr < −12) even at high redshift (Ciambur et al. 2013, Kriek
et al. 2006, Whitaker et al. 2013), suggesting that these galaxies really are dead.
These levels of star formation are much lower than expected. Even if the galaxy
would have stopped accreting new gas from the intergalactic medium, some gas
should always become available again for star formation due to gas recycled from
evolved stages of stellar evolution (e.g. Leitner & Kravtsov, 2010). If the low levels of
star formation are confirmed, it could have important implications for gas recycling
and the effectiveness of quenching at high redshift. Alternatively, it is possible that
amounts of star formation have been overlooked in previous studies because of
heavy obscuration by dust.
To address this question, in this paper we determine the obscured SFRs of quiescent galaxies up to redshift z ∼ 2.5 using their 24µm emission. In Section 2,
we discuss the data. In Section 3, we describe the selection of QGs, and compare
their SFRs from optical and near-infrared (IR) SED fitting to the values expected
from the recycling of gas from mass loss. We additionally evaluate how much obscured star formation might be hidden in our selection: this proves the need of
looking at a mid-IR indicator for SFR. In Section 4, we stack 24µm thumbnails of
QGs in order to measure their obscured SFR. We evaluate possible contributions to
the mid-IR fluxes of QGs in Section 5. We discuss our findings in Section 6 and
summarize them in Section 7. Through the paper we assume a standard cosmology
with H0 = 70km s−1 Mpc−1 , ΩM = 0.30, and ΩΛ = 0.70 (Komatsu et al. 2011).
3.2
Data
The 3D-HST Survey (van Dokkum et al. 2011; Brammer et al. 2012) is a 600 arcmin2
survey using the Hubble Space Telescope (HST) to obtain complete, unbiased lowresolution near-IR spectra for thousands of galaxies. (Cycles 18 and 19, PI: van
Dokkum).
It targets five fields (COSMOS, GOODS-S, GOODS-N1 , AEGIS, UDS) where a
wealth of ancillary multi-wavelength data are available (U band to 24µm); they are
crucial for interpreting spectra that often contain a single emission line, if any. The
1 GOODS-N
has been taken as part of program GO-11600 (PI: B. Weiner) and integrated into 3D-HST
How dead are dead galaxies?
27
0.3<z<0.7
2.0
1.5
2.0
1.5
1.5
1.0
1.0
U-V
1.0
0.7<z<1.1
2.0
0.5
1.0
1.5
0.5
1.5<z<2.0
2.0
1.0
1.1<z<1.5
1.5
0.5
2.0<z<2.5
2.0
1.0
1.5
QG(24µm)
2.0
QG(no24µm)
1.5
1.5
1.5
1.0
1.0
SFG
(dusty)
SFG(blue)
1.0
0.5
1.0
1.5
0.5
0.5
1.0
1.5
0.5
1.0
1.0
V-J
1.5
0.5
1.5
0.5
1.0
1.0
1.5
1.5
Figure 3.1:
UVJ selection in different redshift bins, for mass-selected samples
(log10 ( M? /M ) > 10.3). The Whitaker et al. (2012) boundary divides (solid black line) quiescent and star-forming galaxies. SFGs are subdivided into dusty (U − V > 1.5, purple dots)
and unobscured (U − V < 1.5, blue dots). QGs are color coded according to the presence of a
24µm detection. We notice that 24µm-detected galaxies do not preferentially lie in a particular
locus of the UVJ diagram.
3D-HST photometric catalogue is described in detail in Skelton et al. (2014). It
contains ∼ 170000 sources, detected on a noise-equalized combination of the F125W,
F140W and F160W images. The completeness of 3D-HST as a function of magnitude
is evaluated by comparing the number of detections in the catalog to those in a
deeper image of GOODS-S: the two of them deviate at magnitudes fainter than
F160 = 25 mag.
The WFC3 grism spectra have been extracted with a custom pipeline, described
in Momcheva et al. (2014, in prep). Redshifts have been measured via the combined
photometric and spectroscopic information using a modified version of the EAZY
code (Brammer et al. 2008). The precision of redshifts is proven to be σ( 1dz
+z ) = 0.3%
(Brammer et al. 2012, Momcheva et al. 2013).
Accurate redshifts allow the derivation of accurate rest-frame fluxes: we interpolate rest-frame filters from the observed SED with the Inter-rest code (Taylor et
al., 2009), based on the algorithm by Rudnick et al (2003). Stellar masses have been
determined using the FAST code by Kriek et al. (2009), using Bruzual & Charlot
(2003) models, and assuming exponentially declining star-formation histories (with
28
Chapter 3
e-folding times log10 (τ/yr ) ranging from 107 to 1011 yr), solar metallicity and a
Chabrier (2003) IMF. The 3D-HST catalogs are evaluated to be 90% complete in stellar mass down to log10 ( M? /M ) > 9.4 at z < 2.5 (Tal et al. 2014).
In this paper we restrict the analysis to the GOODS-N and GOODS-S fields,
for which very deep Spitzer-MIPS (S24µm = 10µJy, 3σ) data are available (Dickinson et al., 2003), necessary for inferring low levels of SF. The MIPS 24µm beam
has a FWHM of 6 arcsec, therefore confusion and blending effects are unavoidable
in deep observations at this resolution. We perform photometry using a sourcefitting algorithm (Labbé et al. 2006, Wuyts et al. 2007) that takes advantage of
the higher resolution information contained in the F160W images, as described in
the Appendix. This method produces a model PSF for the image with the broader
native PSF (MIPS-24µm in our case), which is then used to estimate the flux distribution of each source identified in the detection image (F160W) segmentation map
output by SExtractor. For each individual object, the flux from neighboring sources
(closer than 10 arcsec) is fitted and subtracted, allowing for a reliable aperture flux
measurement of individual objects.
Total IR luminosities (LIR = L(8 − 1000µm)) were derived from the observed
24µm fluxes, on the basis of a single template that is the average of Dale & Helou
(2002) templates with 1 < α < 2.5, following Wuyts et al. (2008; see also Franx et
al. 2008, Muzzin et al. 2010), and in good agreement with recent Herschel/PACS
measurements by Wuyts et al. (2011). SFRs are determined from the IR emission
as in Bell et al. (2005) for a Chabrier IMF: SFR(IR) = 0.98 × 10−10 LIR ( L ).2 This
quantity accounts properly just for obscured SF.
We derive SFRs without using data at wavelengths longer than 24µm because
photometry from the PACS and SPIRE instruments on Herschel in the GOODS fields
is not as deep as that from MIPS-24µm. We evaluate the potential for detecting low
SFRs with the Herschel instruments, by using the PACS and SPIRE detection limits
(Elbaz et al. 2011) to SFRs: we extrapolate total IR luminosities from monochromatic
fluxes, and convert them to SFR as described above. We find that at z = 1 the PACS100µm and 160µm photometry is able to detect SFRs higher than ∼ 2M /yr (1σ),
and SPIRE-250 higher than ∼ 5M /yr (1σ), while at the same redshift MIPS-24 µm
is one order of magnitude deeper (∼ 0.3M /yr). The same conclusion holds true at
z = 2, with detection limits for MIPS-24µm, PACS-100µm, PACS-160µm and SPIRE250µm being respectively ∼ 2M /yr, ∼ 20M /yr, ∼ 10M /yr, ∼ 20M /yr (all 1σ
limits).
On the 3D-HST GOODS fields extremely deep X-ray data are also available, 4Ms
in CDF-South (see Xue et al. 2011), and 2Ms in CDF-North (see Alexander et al.
2003), that we use for identifying bright AGNs.
2 The Bell et al.(2005) relation properly applies to starbursts of continuous star formation, with recent
star-formation timescale of ∼ 108 yr and solar metallicity.
How dead are dead galaxies?
29
0.7 < z < 1.1
4
Hα
1.1 < z < 1.5
Fλ + arbitrary
3
1.5 < z < 2.0
2
[OII]
[OIII]
2.0 < z < 2.5
1
0
3000
SFG (blue)
SFG (dusty)
QG (24 µm)
QG (no 24 µm)
4000
5000
6000
7000
Rest Frame Wavelength (Å)
8000
Figure 3.2: Stacked 3D-HST spectra for mass-selected (log10 ( M? /M ) > 10.3) galaxies in
different redshift bins. In each redshift bin, blue means blue SFGs (U − V < 1.5), purple
dust-reddened SFGs (U − V > 1.5), green QGs with a 24µm detection, red QGs without a
24µm detection. Vertical dashed lines show the position of Hα, [OIII] and [OII].
3.3
3.3.1
Sample selection and motivations of the study
Selection of Quiescent Galaxies
In order to select quiescent galaxies (QGs) we use a color-color technique (Figure
3.1), specifically rest-frame U − V versus rest-frame V − J (hereafter: UVJ diagram).
This technique has been widely used to distinguish QGs from SFGs, including the
heavily reddened SFGs (Labbé et al. 2005; Wuyts et al. 2007; Williams et al. 2009;
Brammer et al. 2009; Whitaker et al. 2010; Patel et al. 2012; Bell et al. 2012; Gobat
et al. 2013). QGs are identified using the criteria (U − V ) > 0.8 × (V − J ) + 0.7,
U − V > 1.3 and V − J < 1.5, as in Whitaker et al. (2012) 3 . Effectively, this
3 We test the stability of the selection by shifting the box by ±0.05 magnitude, which does not affect
the analysis.
Chapter 3
30
selection targets galaxies whose optical and near-IR light is dominated by an old
stellar population. We select galaxies more massive than log10 ( M? /M ) > 10.3 and
divide the sample in five redshift bins, centered on z = 0.5, 0.9, 1.2, 1.7, 2.2. The
sample is mass complete at > 97.5% level even at the highest redshift (z < 2.5) we
consider (Tal et al. 2014). At each redshift the QG sample consists of at least 60
galaxies (Table 3.1).
1.4
SFG (blue)
SFG (dusty)
QG (no 24µm)
QG (with 24µm)
1.2
1.0
0.8
F / F(6000A)
0.6
0.4
0.2
0.0
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
10000
10000
RF Wavelength
10000
Figure 3.3: Composite SEDs for mass-selected samples (log10 ( M? /M ) > 10.3) of starforming galaxies (divided into blue and dusty) and quiescent galaxies (divided according
to the presence of a 24µm detection) at redshift 0.3 < z < 2.5. Light lines indicate the scatter
in the stacks. In the bottom-right panel we overplot the four composite SEDs, showing that
quiescent galaxies with and without 24µm detection have very similar optical and near-IR
SED shapes, while star-forming galaxies and dusty star-forming galaxies are clearly distinguishable.
3.3.2
Spectra and SEDs of the sample
In Figure 3.2 we show stacked optical spectra of QGs and SFGs from 3DHST in
mass-selected samples (log10 ( M? /M ) > 10.3). SFGs are subdivided into blue SFGs
(U − V < 1.5) and dust-reddened SFGs (U − V > 1.5).
QGs are subdivided according to the presence of a MIPS 24µm detection. As
noted by other authors (e.g. Brammer et al. 2009, Barro et al. 2013), approximately
25% of the optically-selected QGs have a 24µm detection, which is in apparent contrast with the red optical colors and the SEDs. We also notice that 24µm-detected
How dead are dead galaxies?
31
QGs do not lie preferentially in any locus of the UVJ diagram (green dots in Figure
3.1).
The spectra in Figure 3.2 clearly show that the UVJ selection is efficient in dividing the two populations; the SFG selection includes the heavily dust-reddened SFGs,
that despite red U − V colors, show spectral features (Hα, D4000) characteristic of
SFGs. It is also noteworthy to see that QGs with 24µm detections have some Hα
and [OIII] (cfr. Whitaker et al. 2013), that indicate the presence of low level star
formation and/or nuclear activity.
Figure 3.3 shows composite SEDs (following the methodology of Kriek et al.
2011) for SFGs (divided into blue and dusty) and QGs (divided according to the
presence of a 24µm detection). The SED shapes of star-forming galaxies, dusty starforming galaxies and quiescent galaxies are clearly different. The rest-frame optical
and near-IR SEDs of QGs with and without 24µm detection are instead very similar.
3.3.3
SFRs from SED fitting and expectations from gas recycling
We first analyze the sSFR from the SED fits to the UV-optical and near-IR photometry
(see Section 2). The values are shown against redshift for the quiescent galaxies in
Figure 3.4 (gray dots and black line). The median value is sSFR = 1.7 × 10−12 yr−1 ,
and the correlation with redshift is weak. These values compare well with those of
Ciambur et al (2013), who used a similar method. No significative difference in the
SED-derived SFRs is seen if we split the quiescent population between galaxies with
and without a 24µm detection.
The low sSFRs can be compared to the stellar mass loss from evolved stellar
populations (Parriott & Bregmann, 2008; Leitner & Kravtsov, 2010). Green dots
in Figure 3.4 represent for the QG sample the sSFR expected from stellar mass loss,
assuming that 100% of the gas expelled from old stars is recycled into star formation.
Mass loss is computed directly from Mgas of the BC03 models at the best fit age of
the galaxy, given the best fit τ model (see Section 2, Data). The expected sSFR from
gas recycling is 2 − 4 × 10−11 yr−1 , with a weak redshift dependence. It overpredicts
the sSFR from optical and near-IR SED fitting by more than one order of magnitude.
The discrepancy between the two values at each redshift tells us that one of the
following options must hold true:
• a mechanism able to prevent the cooling of gas expelled from old stars and
therefore the fueling of new star formation exists, or
• SFRs from optical and near-IR best fitting are underestimated (and a lot of star
formation shows up in the mid-IR).
In the rest of the paper we test the latter possibility measuring SFRs in the mid-IR,
in order to prove the former.
3.3.4
How much star formation could be hidden?
We evaluate how much star formation a galaxy can hide (with high dust obscuration), while still retaining red optical-NIR colors. We stack the rest-frame SEDs of
Chapter 3
32
QGs (log M > 10.3)
Expected SFR from
recycling of Mass Loss
sSFR (yr-1)
10-9
SFG
SFR from SED fit
(opt, near-IR)
tH
10-10
10-11
10-12
0
1
2
3
Redshift
Figure 3.4: sSFR in different redshift bins (gray filled dots) and expected sSFR from recycling
of gas from the mass loss of evolved stars (green open dots), as determined from FAST best
fits to the SEDs of quiescent galaxies. Solid lines represent mean values in different redshift
bins. The mass loss is computed from Mgas in BC03 models. It overpredicts the SFR by a
factor of 20 at each redshift. sSFRs of star-forming galaxies on the main sequence (cyan) and
the Hubble time (dashed gray) are shown as references.
2.2
stacks
z ∼ 2.2
2.0
U-V
1.8
1.6
z ∼ 1.7
z ∼ 1.3
z ∼ 0.9
1.4
1.2
1.0
SFG(dusty)
QG
QG+30%SFG
0.6 0.8 1.0 1.2 1.4 1.6 1.8
V-J
Figure 3.5: The UVJ position of stacks of QGs (red) and dusty SFGs (purple) at different redshifts is shown. Black arrows show the tracks obtained summing a variable fraction (normalized at 6000 Å, FSFG ) of the dusty SFGs SEDs to the QGs SEDs. Orange dots show composite
SEDs on the UVJ boundary, corresponding to FSFG = 30 %.
How dead are dead galaxies?
33
QGs and dusty SFGs in different redshift bins; to each QG SED we add a variable
fraction (FSFG , normalized in light at 6000Å) of the dusty SFG SED. Figure 3.5 shows
the position in the UVJ diagram of the QGs stacks (red), the dusty SFGs stacks (purple) and the SED with FSFG = 30% (orange), on the UVJ separation border. Adding
a 30% dusty SFG SED to our typical QG SED would keep such a galaxy as quiescent
under our selection criteria despite the non-negligible contribution of obscured star
formation. Since the SFR of the average SFG evolves with redshift, FSFG = 30% corresponds to sSFR ∼ 8 × 10−11 yr−1 at z = 0.5 and sSFR ∼ 3 × 10−10 yr−1 at z = 2.2.
This shows that with high dust content, a red (optical and near-IR) galaxy can hide
a significant amount of SF. It is therefore necessary to measure SFR from MIR indicators in order to evaluate the SFRs of QGs. There is also a potential for entirely
obscured populations with AV 5, which are known to exist the the centers of local
dusty star-bursting galaxies (e.g. Arp 220, Sturm et al. 1996).
3.4
Measuring Obscured Star-Formation Rates of Quiescent Galaxies
In this Section we discuss the SFRs determined from the IR emission with the
methodology described in Section 2. In Figure 3.6 we plot the relation between
stellar mass and SFR for galaxies in the mass-selected sample. As already noticed
by various authors using a variety of SFR indicators (e.g. Noeske et al. 2007, Damen
et al. 2009, Whitaker et al. 2012), SFRs and masses of SFGs are correlated (light-blue
dots), with a scatter of approximately 0.3 dex. The vast majority of the QGs lies in
this plane below the ’star-forming main sequence’. Most of the QGs are undetected
in the MIPS 24µm image at 3σ (red dots), while some of them (approximately 25 %,
Table 3.1) have a 24µm detection (green dots), placing them in the Mass-SFR plane
between SFGs and the detection limit.
In order to measure the SFRs for QGs, we stack 24µm thumbnails. We emphasize
that in this step we stack cleaned images (20” wide), where neighboring sources identified in the high resolution F160W band have been subtracted with the technique
described in the Appendix.
Summing original 24µm thumbnails would lead to a stack with a very poorly
constrained background, raised by the presence of neighboring objects. Since the
goal of this paper is to measure very low SFRs with accuracy, it is fundamental to
perform photometry on a stacked image with small uncertainty on the background
(as shown in Figure 3.7).
We perform an average-stacking4 in different redshift bins, for two samples: all
QGs and only non-24µm-detected QGs. Photometry on the stack is performed
within an aperture of 6 arcsec diameter, similar to the size of the 24µm FWHM.
To measure the total 24µm flux, we create a MIPS growth curve from several bright,
isolated, and unsaturated point sources within each field. These square postage
stamps are 20 arcsec wide, and we derive an aperture correction of a factor of 2.57
from r =3 arcsec to r = 10 arcsec. To convert to total flux, we include an additional
4 Using
instead median stacks does not modify the conclusions of the paper.
Chapter 3
34
0.3<z<0.7
100
SFR(IR) [Msun/yr]
10
1
0.7<z<1.1
100
100
10
10
1
1
10.0
10.5
11.0
11.5
10.0
10.5
1.5<z<2.0
100
1.1<z<1.5
11.0
11.5
2.0<z<2.5
100
SFGs
X-rays AGN
QGs (with 24µm)
100
QGs (no 24µm)
stack QGs
10
1
10
stack QGs (no 24µm)
10
0.0 0.2 0.4 0.6 0.8 1.0
legend
1
1
10.0
10.5
11.0
11.5
10.0
10.5
11.0
11.5
10.0 10.5 11.0 11.5 10.0 10.5 11.0 11.5
log10 (M*/Msun)
Figure 3.6: Mass-SFR(IR) diagram for galaxies in the 5 redshift bins analyzed in the paper.
SFRs are computed assuming that all of the 24µm flux is due to reprocessed UV photons from
HII regions. Filled symbols denote galaxies with log( M? /M ) > 10.3. Galaxies are divided
in quiescent (QGs) and star-forming (SFGs) according to the box defined in Whitaker et al.
(2012): light-blue dots represent SFGs, green dots QGs detected at 24µm (S/N > 3), red dots
QGs not detected at 24µm. Sources with S/N < 1 are shown as arrows at the 1σ level. Orange
stars represent X-ray detected QGs in the CDF-S 4Ms catalog and CDF-N 2Ms catalog. The
large red dots show the SFRs obtained stacking thumbnails of individually undetected QGs
(red), and all QGs (yellow), in mass-selected samples of log10 ( M? /M ) > 10.3. Errors on the
stacks are computed through bootstrapping of the sample.
aperture correction for the 22% of the flux that falls outside 10 arcsec, derived from
the MIPS handbook. 5
We obtain mostly clear detections with signal-to-noise of 3–5, and fluxes F24µm
=2 − 5µJy, corresponding to SFR ∼ 0.5-5 M /yr. We summarize the measured
stacked fluxes in Table 3.1. Errors on the stacks are measured through bootstrapping of the sources. Errors on the stacks are measured through bootstrapping, as
follows. Each sample of QGs is resampled 1000 times. We stack the individual
24µm images of galaxies belonging to each resampling and perform photometry on
the new stacked images. The dispersion of the flux values in the resampled stacks
5 Since the calibration for MIPS refers to an object with T = 10000K, we color-correct fluxes by dividing
them by a factor of 0.967 (MIPS Handbook, Table 4.17). This way flux densities correspond to those of
sources with a flat spectrum.
How dead are dead galaxies?
35
Profile of Stacks
3
+
2
+
Stack
+
+
+ ... +
=
24µm Clean Thumbnails
Stack
Stack
0.006
Flux [uJy/px]
24µm Thumbnails
0.004
0.002
Stack
(Clean Thumbs)
1
+
0
0
+
2
QGs (0.7 < z < 1.1)
+
+
4
+ ... +
6
=
0.000
8
0
2
4
6
8
R [arcsec]
10
Figure 3.7: Stacking procedure for quiescent galaxies. Left: we show on the top row a random
sample of 24µm postage stamps (20” wide) from the QGs sample at redshift 0.7 < z <
1.1 and the resulting stack. The same postage stamps after the neighboring sources have
been subtracted (with the PSF-matching technique described in Section 2) are shown in the
bottom row, along with the resulting mean stack. Right: Profile of the mean stacks (lines) and
individual pixel values (dots) at their distance from the center. The ’normal’ stack has a high,
poorly constrained, background level (artificially raised by neighboring sources).
gives the uncertainty on the flux measurement.
In Figure 3.6 we overplot with large yellow dots the SFR obtained from all QGs
(big yellow dots) and non-24µm-detected galaxies (big red dots), representative of
the deadest fraction of the galaxy population (this definition of ’quiescent galaxy’ is
the same as in Bell et al. 2012).
Despite the different sample selection (all QGs or just QGs not detected at 24µm),
it is evident that at each redshift the average QGs has a SFR which is at least ∼
20 − 40 times lower than the ones on the ’star-forming sequence’.
In Figure 3.8 (left panel) we show the redshift evolution of SFRs of SFGs and
QGs. We plot sSFR since it is more mildly dependent on stellar mass than SFR itself.
As noted by previous studies (e.g. Damen et al. 2010, Whitaker et al. 2012, Karim
et al. 2012, Fumagalli et al. 2012), the evolution of sSFR in redshift for star-forming
galaxies is well fit by a power law (1 + z)n where n ∼ 3 − 4. At each redshift QGs
have a sSFR at least 20 times lower than SFGs. The evolution with redshift of sSFR
of QGs seems to resemble the evolution of SFGs. At redshift z ∼ 2, QGs form 10
times more stars than at redshift z ∼ 0.5.
In Figure 3.9 we show for our samples the ratio of the average SFR of QGs to
the average SFR of SFGs at the same redshift. For quiescent galaxies undetected at
24µm, the mean value of the ratio is hSFRQG i/hSFRSFG i = 1/(45 ± 11) while for
the entire sample it is hSFRQG i/hSFRSFG i = 1/(22 ± 7). This confirms that at each
redshift quenching of star formation is very efficient.
For QGs the SFRs inferred from the IR emission are generally an order of magnitude larger than those inferred from stellar population modeling (black dashed
line in Figure 3.8, left panel). At the highest redshifts they are similar to the values
predicted by the recycling of mass loss (green dashed line in Figure 3.8, left panel),
while at redshift lower than 1.5 they are significantly lower than those.
12
Chapter 3
36
3.5
Other possible contributions to LIR
Strictly should be that the IR-inferred SFRs for QGs are upper limits, because of
contributions of AGNs, AGB stars and dust heating from old stellar populations
to the IR fluxes. We treat each of these components separately in the following
Subsections and compare their contributions to LIR with the observed stacked values
of LIR in the Discussion Section.
3.5.1
AGN
We evaluate the possible contribution of AGNs by stacking X-ray thumbnails (from
the CDF-S 4Ms and CDF-N 2Ms) of the QGs in different redshift bins. Mullaney
et al. 2011 demonstrates (Equation 4) the existence of a linear relation between
the X-ray luminosity LX and LIR for a sample of local AGNs. After subtracting
individually-detected X-ray point sources (marked with orange stars in Figure 3.6),
we obtain marginal detections (2 − 3σ) ranging from LX ∼ 3.8 × 1040 erg/s 6 in the
lowest redshift bin to LX ∼ 2.0 × 1041 erg/s in the highest redshift bin. Converting
the obtained X-ray luminosities to IR luminosities with the Mullaney relation, we
obtain the gray line in Figure 3.8 (right panel). It lies three orders of magnitude
below the observed L IR /M? 7 . Olsen et al. (2013) suggest that at redshift z ∼ 2
most QGs host a low-luminosity AGN, comparing SFR inferred from IR and X-ray.
They find, for QG at 1.5 < z < 2.5 of the same mass of that of our sample, a mean
luminosity of LX < 2.5 × 1041 , consistent with our study. Even though most of QGs
host a (low-luminosity) AGN, we find that those weak AGN can not account for
the MIR emission of the galaxies. Other studies (Donley et al. 2008; Kartaltepe et
al. 2010) have also already pointed out that systems with 24µm flux dominated by
AGNs are not the dominant population at low LIR , such as QGs.
3.5.2
Circumstellar dust
AGB stars are known to evolve embedded in a circumstellar dusty envelope (e.g.
Bressan et al. 1998, Lancon & Mouhcine 2002, Piovan et al. 2003). They are the
dominant source of the rest-frame K-band luminosity between 0.1 and 1.5 Gyr of age
(Kelson & Holden 2010) and significantly contribute to MIR emission, but their dust
contribution is not included in classical optical-near infrared SED fitting (Bruzual
& Charlot 2003, Maraston 2005). We evaluate the contribution to LIR with the new
Charlot & Bruzual 2010 model (CB2010) of an SSP with solar metallicity (private
communication). Given galaxy ages from the FAST best fits (see Section 2, and
Whitaker et al. 2013, in press), for each galaxy in our QG sample we estimate the
observed 24µm flux from the CB2010 model and convert it to LIR with the same
relation of Wuyts et al.(2008) (Figure 3.8, right panel, orange line).
luminosities are evaluated assuming a power law spectrum with Γ = 1.8
high fraction of Compton-thick AGNs in the sample would originate a higher IR luminosity inferred from X-ray stacks. The percentage of Compton-thick AGNs is however poorly constrained at high
redshift (e.g. Akylas et al. 2012).
6 X-ray
7A
How dead are dead galaxies?
IR
sSFR (yr-1)
10-9
t-1
H
IR
10-10
Mass Loss
IR
10-11
[ L sun / Msun ]
logM>10.30
SFGs
QG
QG(no24µm)
1
log10 L(IR) / M
10-8
37
-1
0
-2
-9
SFGs
Dust Heating
(SED)
TP-AGB
(CB2010)
QGs
QGs
(no24µm)
0
-12
AGN
-3
1
2
-11
SFR (SED)
SED fits
10-12
-10
3
-13
0
1
Redshift
2
3
Redshift
<SFRQG> / <SFRSFG>
Figure 3.8: Left: Evolution of sSFR(IR) with redshift in a log10 ( M? /M ) > 10.3 mass-selected
sample. light-blue dots indicate mean values for SFGs, while yellow and red points are
stacked values of non-24µm-detected QGs (red), and all QGs (yellow). At any redshift the
average QG has a sSFR 20 times lower than the star-forming sequence. The evolution of sSFR
of QGs resembles the one of SFGs. As in Figure 3.4, we indicate with a black line the sSFR
from SED fitting and with a green line the expected sSFR from mass loss. At high redshift, the
sSFR(IR) of QGs is comparable to the mass-loss. Right: Comparison of observed and modeled
LIR /M? . Values from the stacks of quiescent galaxies are represented by dotted yellow and
red lines. SFGs mean values (light-blue) are put as a reference. Expected contributions to
LIR for the QG samples from models described in Section 3.5 are drawn with solid lines
(gray: AGN, orange: circumstellar dust, black: SFR from best fits, green: cirrus dust heating).
Circumstellar dust and cirrus dust can account for most of the observed LIR .
1/10
QGs
QG(no24µm)
1/100
0
1
2
3
Redshift
Figure 3.9: Ratio of the average SFR of QGs to the average SFR of SFGs at the same redshift. Red dots represent QGs which are individually undetected at 24µm, while yellow dots
represent all QGs. For the two samples, the average ratio is respectively 1/(45 ± 11) and
1/(22 ± 7). These ratios are possibly even lower because for QGs IR inferred SFRs can be
significantly contaminated by other sources of dust heating (Section 5).
Chapter 3
38
3.5.3
Cirrus dust
Another possible contribution to LIR is dust heating from old stellar populations.
Salim et al. (2009) concludes that, for a sample of 24µm-detected galaxies in the
DEEP2 survey (0.2 < z < 1.0), the bulk of IR emission in red (NUV − r) galaxies
comes from the heating of diffuse cirrus dust by old stellar populations, rather than
by dust heating in star-forming regions. We test if this holds true for the galaxies
in our sample as follows. Given the stellar population parameters from the FAST
best-fit to the SEDs (age, τ, AV ), we evaluate the luminosity absorbed at λ < 1µm
by integrating the difference between the unattenuated and the attenuated synthetic
SED, and assume it is re-emitted in the IR (see Charlot & Fall 2000, Da Cunha et al.
2008).
We then compare the model LIR predicted by the attenuated SED with the best
fit SFR. If LIR originates in dust associated with star-forming regions, we expect
the ratio LIR /SFR to be ∼ 9.8 × 109 L /M (Bell et al., 2005). Figure 3.10 shows
that SFGs (blue) are consistent with this prediction. On the other hand, for QGs
(red points) LIR is systematically higher than the expectations from dust heating in
star-forming regions. This indicates that in QGs a significant contribution to LIR
comes from dust heated by old stellar populations. Inferring SFR from LIR (and
therefore from 24µm fluxes) overestimates the real SFR of QGs. For each galaxy in
the QG sample we estimate the expected LIR luminosity from circumstellar dust,
and compute the mean value in different redshift bins (Figure 3.8, right panel, green
line).
3.6
Discussion
As we have seen above, various processes other than star formation can contribute
to the observed mid-IR flux. We next discuss the impact on the derived SFRs. Moreover, we put constraints on the mass growth of QGs implied by the measured SFRs
and on their size growth implied by the stellar mass loss.
In Figure 3.8 (right panel) we show the approximate evolution of LIR /M? , for
data (dashed lines) and models (thick solid lines). We saw earlier that observations
of LIR are based on the extrapolation of the single band 24µm to L(IR) assuming a
template for dust heating by star-forming regions (Section 2). Model predictions estimate that the AGN contribution (gray line) to the LIR is negligible for our sample,
while the model expectation for LIR from cirrus dust (green) and circumstellar dust
(orange) is comparable to the observed values from stacking. We note that qualitatively both of them decrease towards lower redshift, respectively because of higher
AV and younger stellar ages at higher redshifts (which leads to more absorbed optical light re-emitted in the IR in the younger Universe) and because of the aging of
galaxies (which leads to lower contribution of AGB stars in the SED).
If SFRs from SED fitting are correct, their contribution to LIR (black line in Figure
3.8, right panel) would be 1 dex lower than the observed LIR , while dust heated by
old stellar population can account for the most of the observed luminosities.
All the measured values from 24µm stacks must therefore be considered as upper
How dead are dead galaxies?
39
LOG L(IR)/SFR
12
11
10
QG
9 SFG
Bell et al. 2005
8
-13
-12 -11 -10
sSFR (SED)
-9
Figure 3.10: Model predictions of LIR /SFR, for QGs (red) and SFGs (blue). LIR is reconstructed
assuming that the light absorbed by dust at UV-optical wavelengths is re-emitted in the IR
(Section 3.5). For SFGs the ratio is comparable to the Bell et al. (2005) relation (black line),
while for QGs LIR are systematically higher than the expectations from SFR, meaning that for
QGs most of dust heating comes from old stellar populations.
limits to the SFR. At each redshift, the mean QG has a SFR at least ∼20-40 times
lower than that of a SFG at the same redshift. These SFRs are significantly higher
than estimates based on optical and near-IR model fits (see Section 3 and Ciambur,
Kauffman & Wuyts 2013).
In order to evaluate the growth of a QG via star formation we integrate the
sSFR(IR)-z trend of Figure 3.8 (left). This leads to estimate that the maximum growth
of a QG via star formation is 20% from redshift 2 to 0. Some authors (e.g. van
Dokkum et al. 2010, Patel et al. 2012) have inferred that a present-day 1011.2 M
galaxy has to grow 60% of its mass from redshift z ∼ 1.75 to z ∼ 0. We show that
star formation can not be responsible for the entire stellar growth of QGs, while
other mechanisms must be in place, such as minor merging (see, among others,
Hopkins 2009). The limit we compute on the mass increase via star formation is
more stringent than that computed by Pérez-González et al. (2008), who estimates
that massive spheroid-like galaxies may have doubled (at the most) their stellar mass
from redshift 2 to 0.
The SFRs expected from stellar mass loss are probably much higher than the real
SFRs of QGs, meaning that star formation from mass loss is inefficient. If mass loss
from evolved stars is not converted into stars and gas is expelled from the galaxy,
an interesting consequence is that the potential of the system becomes shallower
and the system expands (Zhao et al. 2002, Murray et al. 2010). In brief (following
Damjanov et al., 2009), if a system loses a fraction δM/M of its mass in a time
Chapter 3
40
longer than a dynamical timescale, it will expand its radius by a factor of δR/R ∼
(1 − δM/M)−1 . The modeled mass losses for galaxies in our sample (Figure 3.4)
integrated over the redshift range 0 to 2 give δM/M ≈ 0.4, which leads to δR/R ≈
0.6. The observed size growth of quiescent galaxies from redshift 2 to 0 amounts to
a factor of 2-3 (e.g. Williams et al. 2010, Newman et al. 2012, Whitaker et al. 2012),
therefore mass loss can not be its unique cause but only one of the concurrent ones
(see also Damjanov et al., 2009). We note that the assumed mass loss depends on
the absolute ages of galaxies at each redshift, which are very uncertain.
A possible caveat in the study comes from the fact that the 24µm-to-L(IR) conversion relies on a single infrared template (Wuyts et al. 2008), while the underlying
SED for high-redshift quiescent galaxies is unknown. As explained in Section 2
(Data), measuring the low fluxes of quiescent galaxies at Herschel wavelengths is
extremely challenging. Available observationally motivated far-IR SEDs of galaxies
at high redshift are based on bright Herschel sources (e.g. Elbaz et al. 2011, Magdis
et al. 2014) and refer to galaxies on or above the star-forming main sequence. We
evaluate the possible bias introduced by our synthetic template by comparing the
L(IR) integrated under the Elbaz et al. 2011 SED for main-sequence galaxies to that
inferred from a simulated 24µm observation of that SED, at different redshifts. We
obtain that at z > 1.5 the inferred L(IR) is a factor of 2 higher than the integrated
L(IR), while at z < 1.5 the bias is lower than 50%.
The possibility that galaxies below the main-sequence have different infrared
SED shapes must however be taken into account (see also Utomo et al. 2014, Hayward et al. 2014). We compute the systematic uncertainty on L(IR) arising from the
unknown underlying SED as follows. We compare the 24µm-to-LIR conversion of
Wuyts et al. 2008, obtained by averaging a vast array of infrared templates from
Dale & Helou 2002, with those obtained by using each single Dale & Helou 2002
template for different values of the ionization parameter α. The dispersion on the
values is 0.3 dex, which we consider the systematic uncertainty on the conversion.
We conclude that the possible biases and uncertainties induced by inferring L(IR)
from a single band and a single template amount to a factor of 2, and do not affect
the conclusions of the paper.
3.7
Conclusions
We select quiescent galaxies at redshift 0.3 < z < 2.5 in the 3D-HST survey from
their rest-frame optical and near-IR colors. Fitting their UV to near-IR photometry with stellar population models, we find very low star-formation rates (sSFR ∼
10−12 yr−1 ). These values are much lower than the stellar mass loss rates predicted
by the same models. This suggests that the star formation is either missed because
it is dust obscured, or that the gas from stellar mass loss is expelled from the galaxy
or prevented from refueling star formation.
We put upper limits on the obscured star-formation rate of quiescent galaxies by
stacking 24µm images. Including direct 24µm detections, we find that sSFR(IR) ≤
10−11.9 × (1 + z)4 yr−1 . At each redshift the sSFR of quiescent galaxies is ∼ 20-40
times lower than the typical value on the main sequence of star-forming galaxies.
How dead are dead galaxies?
41
SFRs of quiescent galaxies are possibly even lower than this, because the IR luminosity can also be due to other sources, such as the presence of AGB dust enshrouded
stars and dust heating from older stellar populations. Stacks of longer wavelength
data (such as from Herschel) are necessary for constraining the dust temperature and
therefore distinguishing between the different contributions to LIR , however a large
sample may be necessary to achieve adequate S/N (e.g. Viero et al. 2013). We show
nevertheless that dust heating from old stellar populations can account for most of
the observed LIR .
The observed SFR(IR) are therefore upper limits to the real SFR, which are possibly one order of magnitude lower. This means that there must be a mechanism
that not only shuts down star formation, but also keeps the galaxy dead for a long
period of time, preventing the ejected gas from cooling and forming new stars. If
gas from mass-loss is expelled from galaxies, we predict that it is responsible for a
growth in stellar radii of 60% from redshift 2 to 0.
We acknowledge funding from ERC grant HIGHZ no. 227749. This work is
based on observations taken by the 3D-HST Treasury Program (GO 12177 and 12328)
with the NASA/ESA HST, which is operated by the Association of Universities for
Research in Astronomy, Inc., under NASA contract NAS5-26555.
Chapter 3
42
Redshift
0.3 - 0.7
0.7 - 1.1
1.1 - 1.5
1.5 - 2.0
2.0 - 2.5
NQG
97
154
84
72
35
F (24µm)QG
7.7 ± 0.5 µJy
6.6 ± 0.7 µJy
9.3 ± 1.8 µJy
6.8 ± 1.8 µJy
5.7 ± 1.8 µJy
NQG,no24µm
67
108
58
51
25
Table 3.1: Properties of Stacks
SFR( IR)QG
0.4 ± 0.1 M /yr
1.2 ± 0.1 M /yr
3.7 ± 0.7 M /yr
4.6 ± 1.3 M /yr
8.8 ± 2.2 M /yr
F (24µm)QG,no24µm
3.9 ± 1.3 µJy
4.1 ± 0.7 µJy
4.4 ± 1.8 µJy
3.0 ± 1.6 µJy
3.2 ± 1.3 µJy
SFR( IR)QG,no24µm
0.2 ± 0.1 M /yr
0.5 ± 0.1 M /yr
1.8 ± 0.6 M /yr
2.0 ± 1.0 M /yr
3.8 ± 1.5 M /yr
For different redshift bins: number of galaxies in the quiescent sample (QG) and quiescent sample without 24µm detection
(QG,no24µm), along with their stacked 24m fluxes, and the implied SFR from IR emission.
How dead are dead galaxies?
3.A
43
Appendix A: Photometry
The MIPS-24µm beam has a FWHM of 6 arcsec, therefore confusion and blending effects are unavoidable in deep observations at this resolution. We use a source-fitting
algorithm designed to extract photometry from IRAC and MIPS images (see, e.g.,
Labbé et al. 2006; Wuyts et al. 2007). The information on position and extent of the
sources based on the higher resolution F160W segmentation map is used to model
the lower resolution MIPS-24µm images. Local convolution kernels are constructed
using bright, isolated, and unsaturated sources in the F160W and MIPS-24µm, derived by fitting a series of Gaussian-weighted Hermite functions to the Fourier transform of the sources. Each source is extracted separately from the F160W image and,
under the assumption of negligible morphological K-corrections, convolved to the
MIPS-24µm resolution using the local kernel coefficients. All sources in each MIPS24µm image are fit simultaneously, with the flux left as the only free parameter. The
modeled light of neighboring sources (closer than 10 arcsec) is subtracted, thereby
leaving a "clean" MIPS-24µm image to perform aperture photometry and stacking
of faint sources. The technique is illustrated in Figure A1 and A2, respectively for a
bright and a faint source.
3.B
Appendix B: Field-to-field variation
The paper is built on data from the GOODS-North and GOODS-South fields. The
two fields feature a similar large number of optical-near-IR observations included in
the 3D-HST photometric catalog, and data quality in the 3D-HST fields is uniform
(see Skelton et al., 2014). The depths of MIPS-24µm data are similar in the two fields
(Dickinson et al. 2003). We show in Figure B1 the main result of the paper - the
evolution of sSFRs of QGs - once the data are stacked separately in the two fields.
Differences and errors are consistent with lower statistics.
Chapter 3
44
Figure B1: The process of modeling and deblending 24µm fluxes for objects identified in
the F160W detection image. Panel 1 shows the original 24µm cutout for an object in the
catalog. Panel 2 and 3 show the matching F160W detection image and segmentation map
from SExtractor. The bottom row shows the modeled 24µm flux for all objects in the region
(Panel 4), the residual image with all modeled fluxes removed (Panel 5), and the flux for the
central object alone (Panel 6).
Figure B2: Same as Figure A1, but for a faint object in the catalog.
How dead are dead galaxies?
45
QG(no24µm)
sSFR
10-10
10-11
ALL
GOODS-N
GOODS-S
0.5
1.0
1.5
z
2.0
2.5
Figure B3: Evolution of sSFR of QGs without an individual 24µm detection, for different fields
and in the entire sample. Red/black dots are stacked values from GOODS-North/GOODSSouth, and large blue dots are values from the combined sample. Errors are computed bootstrapping the sample. Mean redshifts have been shifted of ± 0.05 for clarity. Differences and
errors are consistent with lower statistics.
Bibliography
Akylas, A., Georgakakis, A., Georgantopoulos, I., Brightman, M., & Nandra, K. 2012,
A&A, 546, A98
Alexander, D. M., Bauer, F. E., Brandt, W. N., et al. 2003, AJ, 126, 539
Barro, G., Faber, S. M., Pérez-González, P. G., et al. 2013, ApJ, 765, 104
Bell, E. F., Papovich, C., Wolf, C., et al. 2005, ApJ, 625, 23
Bell, E. F., van der Wel, A., Papovich, C., et al. 2012, ApJ, 753, 167
Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008, ApJ, 686, 1503
Brammer, G. B., Whitaker, K. E., van Dokkum, P. G., et al. 2009, ApJL, 706, L173
Brammer, G. B., Whitaker, K. E., van Dokkum, P. G., et al. 2011, ApJ, 739, 24
Brammer, G., van Dokkum, P., Franx, M., et al. 2012, arXiv:1204.2829
Parriott, J. R., & Bregman, J. N. 2008, ApJ, 681, 1215
Bressan, A., Granato, G. L., & Silva, L. 1998, A&A, 332, 135
Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000
Chabrier, G. 2003, ApJL, 586, L133
46
Chapter 3
Charlot, S., & Fall, S. M. 2000, ApJ, 539, 718
Ciambur, B. C., Kauffmann, G., & Wuyts, S. 2013, MNRAS, 432, 2488
da Cunha, E., Charlot, S., & Elbaz, D. 2008, MNRAS, 388, 1595
Damen, M., Labbé, I., van Dokkum, P. G., et al. 2011, ApJ, 727, 1
Damjanov, I., McCarthy, P. J., Abraham, R. G., et al. 2009, ApJ, 695, 101
Dickinson, M., & GOODS Team 2004, Bulletin of the American Astronomical Society,
36, #163.01
Donley, J. L., Rieke, G. H., Pérez-González, P. G., & Barro, G. 2008, ApJ, 687, 111
Elbaz, D., Dickinson, M., Hwang, H. S., et al. 2011, A&A, 533, A119
Franx, M., van Dokkum, P. G., Schreiber, N. M. F., et al. 2008, ApJ, 688, 770
Fumagalli, M., Patel, S. G., Franx, M., et al. 2012, ApJL, 757, L22
Gobat, R., Strazzullo, V., Daddi, E., et al. 2013, arXiv:1305.3576
Hopkins, P. F., Bundy, K., Murray, N., et al. 2009, MNRAS, 398, 898
Ilbert, O., Salvato, M., Le Floc’h, E., et al. 2010, ApJ, 709, 644
Karim, A., Schinnerer, E., Martínez-Sansigre, A., et al. 2011, ApJ, 730, 61
Kartaltepe, J. S., Sanders, D. B., Le Floc’h, E., et al. 2010, ApJ, 709, 572
Kauffmann, G., Heckman, T. M., White, S. D. M., et al. 2003, MNRAS, 341, 54
Kelson, D. D., & Holden, B. P. 2010, ApJL, 713, L28
Kennicutt, R. C., Jr. 1998, ApJ, 498, 541
Komatsu, E., Smith, K. M., Dunkley, J., et al. 2011, ApJS, 192, 18
Kriek, M., van Dokkum, P. G., Franx, M., et al. 2006, ApJL, 649, L71
Kriek, M., van Dokkum, P. G., Labbé, I., et al. 2009, ApJ, 700, 221
Kriek, M., van Dokkum, P. G., Whitaker, K. E., et al. 2011, ApJ, 743, 168
Labbé, I., Huang, J., Franx, M., et al. 2005, ApJL, 624, L81
Labbé, I., Bouwens, R., Illingworth, G. D., & Franx, M. 2006, ApJL, 649, L67
Lançon, A., & Mouhcine, M. 2002, A&A, 393, 167
Leitner, S. N., & Kravtsov, A. V. 2011, ApJ, 734, 48
Maraston, C. 2005, MNRAS, 362, 799
Mullaney, J. R., Alexander, D. M., Goulding, A. D., & Hickox, R. C. 2011, MNRAS,
414, 1082
How dead are dead galaxies?
47
Muzzin, A., van Dokkum, P., Kriek, M., et al. 2010, ApJ, 725, 742
Murray, N., Quataert, E., & Thompson, T. A. 2010, ApJ, 709, 191
Newman, A. B., Ellis, R. S., Bundy, K., & Treu, T. 2012, ApJ, 746, 162
Noeske, K. G., Weiner, B. J., Faber, S. M., et al. 2007, ApJL, 660, L43
Olsen, K. P., Rasmussen, J., Toft, S., & Zirm, A. W. 2013, ApJ, 764, 4
Patel, S. G., Holden, B. P., Kelson, D. D., et al. 2012, ApJL, 748, L27
Pérez-González, P. G., Trujillo, I., Barro, G., et al. 2008, ApJ, 687, 50
Piovan, L., Tantalo, R., & Chiosi, C. 2003, A&A, 408, 559
Rudnick, G., Rix, H.-W., Franx, M., et al. 2003, ApJ, 599, 847
Salim, S., Dickinson, M., Michael Rich, R., et al. 2009, ApJ, 700, 161
Skelton, R. E., Whitaker, K. E., Momcheva, I. G., et al. 2014, arXiv:1403.3689
Sturm, E., Lutz, D., Genzel, R., et al. 1996, A&A, 315, L133
Taylor, E. N., Franx, M., van Dokkum, P. G., et al. 2009, ApJS, 183, 295
van Dokkum, P. G., Whitaker, K. E., Brammer, G., et al. 2010, ApJ, 709, 1018
Viero, M. P., Moncelsi, L., Quadri, R. F., et al. 2013, arXiv:1304.0446
Xue, Y. Q., Luo, B., Brandt, W. N., et al. 2011, ApJS, 195, 10
Whitaker, K. E., van Dokkum, P. G., Brammer, G., et al. 2010, ApJ, 719, 1715
Whitaker, K. E., van Dokkum, P. G., Brammer, G., & Franx, M. 2012, ApJL, 754, L29
Whitaker, K. E., van Dokkum, P. G., Brammer, G., et al. 2013, ApJL, 770, L39
Williams, R. J., Quadri, R. F., Franx, M., van Dokkum, P., & Labbé, I. 2009, ApJ, 691,
1879
Wuyts, S., Labbé, I., Franx, M., et al. 2007, ApJ, 655, 51
Wuyts, S., Labbé, I., Schreiber, N. M. F., et al. 2008, ApJ, 682, 985
Wuyts, S., Förster Schreiber, N. M., Lutz, D., et al. 2011, ApJ, 738, 106
48
Chapter 3
4
Ages of massive galaxies at
0.5 < z < 2.0 from 3D-HST rest-frame
optical spectroscopy
We present low-resolution near-infrared stacked spectra from the 3D-HST survey
up to z = 2.0 and fit them with commonly used stellar population synthesis models
(BC03, FSPS10 and CKC14). The accuracy of the grism redshifts, in combination
with stacking techniques, allows the unambiguous detection of many emission and
absorption features, and thus a first systematic exploration of the rest-frame optical
spectra of galaxies up to z = 2. For a quantitative analysis, we select massive galaxies (log(M∗ /M ) > 10.8), we divide them into quiescent and star-forming via a
rest-frame color-color technique, and we median-stack the samples in 3 redshift bins
between z = 0.5 and z = 2.0. We find that stellar population models fit the observations well at wavelengths below 6500Å rest-frame, but show systematic residuals at
redder wavelengths. The CKC14 model generally provides the best fits (evaluated
with a χ2 statistics) for quiescent galaxies, while BC03 performs the best for starforming galaxies. The stellar ages of quiescent galaxies implied by the models vary
from 4 Gyr at z ∼ 0.75 to 1.5 Gyr at z ∼ 1.75. On average the stellar ages are half
the age of the Universe at these redshifts. We show that the inferred evolution of
ages of quiescent galaxies is in agreement with fundamental plane measurements,
assuming an 8 Gyr age for local galaxies. For star-forming galaxies the inferred
ages depend strongly on the stellar population model and the shape of the assumed
star-formation history. We finally notice that our low-resolution data is not able to
constrain the metallicity of galaxies.
Mattia Fumagalli; Marijn Franx; Pieter van Dokkum; Katherine Whitaker; et al.
Submitted to the Astrophysical Journal
49
Chapter 4
50
4.1
Introduction
In recent years, multi-wavelength surveys at high redshift have revealed a significant evolution of galaxies from redshift z ∼ 2 to the present epoch. The emerging
picture is based on a few key observations. First, the star formation rates (SFRs) of
galaxies have declined by a factor of 10 in the last 10 billion years. Different observational techniques agree that this trend is largely independent of mass (Damen et
al. 2009, Karim et al. 2011, Fumagalli et al. 2012). This decline is accompanied by
the evolution of the mass function, that once split into star-forming and quiescent
population reveals a differential behavior for the two categories: while the number of massive star-forming galaxies remains constant or even declines, the number
density of massive quiescent galaxies grows by 0.5-1.0 dex from z ∼2 (Muzzin et al.
2013, Ilbert et al. 2013). An immediate consequence is that the quiescent fraction
at the massive end becomes increasingly larger at lower redshifts (Bell et al. 2007;
Bundy et al. 2006, Brammer et al 2011). While at lower redshift massive galaxies (log(M∗ /M ) > 11) are dominated by a homogenous group of quiescent, red,
early-type objects (Djorgovski & Davis 1987; Blanton et al. 2003; Kauffmann et al.
2003a), at redshift z∼1 the population shows a large diversity of colors, structural
parameters and SFRs (Abraham et al. 2004, van Dokkum et al., 2011).
An additional insight into the assembly history of galaxies is given by their stellar population parameters, namely their age and metallicity. In the local universe
the light-weighted ages and metallicities (both stellar and gaseous) have been shown
to correlate tightly with mass (e.g. Tremonti et al. 2004, Gallazzi et al. 2006). While
chemical properties of gas in star-forming objects have been traced up to z ∼ 3 by
emission line studies of Lyman break galaxies (e.g. Erb et al. 2006, Moustakas et
al. 2011), studies of stellar population parameters at high redshift have been proven
challenging, since they require deep spectroscopy in order to trace the rest-frame
continuum. Recent works by Gallazzi et al. 2014 and Choi et al. 2014 push stellar
population analysis to redshifts of z ∼ 0.7, where the absorption lines commonly
used for metallicity and age determinations (Balmer lines, Mg, Na, etc) fall at the
edge of optical spectrographs. At higher redshifts, the optical rest-frame shifts to
the infrared, where observations from the ground are notoriously challenging. Determinations of stellar population parameters at z > 1.5 are limited to a few bright
galaxies (Kriek et al 2009, van de Sande et al. 2012, Onodera eta al. 2012, Bezanson
et al. 2013, Onodera et al. 2014, Mendel et al. 2015) or composite spectra (Whitaker
et al. 2013)
In this paper we present observations of galaxies at 0.5 < z < 2.0 obtained
with the low-resolution Wide Field Camera 3 (WFC3) grism onboard Hubble Space
Telescope (HST). These spectra cover the observed wavelengths 11000 < Å < 16000,
which correspond to the optical rest-frame for the targeted redshift range. We divide
galaxies into quiescent and star-forming, stack their spectra in mass selected samples, and fit them with models from commonly used stellar population synthesis
(SPS) codes.
The goal of the paper is two-fold. In the first place we test the accuracy of SPS
models at the observed redshifts and wavelengths. Second, we determine constraints
Ages of massive galaxies at 0.5 < z < 2.0
N = 86
2.0 N = 105
51
N = 85
N = 121
N = 77
N = 97
U-V
1.5
1.0
0.5
0.5<z<1.0
0.0
0.0
0.5
1.0
1.5
1.0<z<1.5
2.0 0.0
0.5
1.0
1.5
V-J
1.5<z<2.0
2.0 0.0
0.5
1.0
1.5
2.0
Figure 4.1: The selection of quiescent (red) and star-forming (blue) galaxies more massive
than log(M∗ /M ) > 10.8 in the UVJ diagram. Grey dots represent all galaxies from 3D-HST
in the same redshift bin, including those with a significantly lower mass.
on the stellar ages of galaxies in mass-selected samples, at previously unexplored
redshifts.
We note that we apply and test the models in a relatively new regime, both
in terms of redshifts and in terms of spectral resolution. Most model tests have
been done either at very low spectral resolution (i.e., broad-band and medium-band
imaging, with R up to ∼ 8), or at moderate to high spectral resolution (R >∼ 5000).
Here we apply the models to spectra with R = 50 − 100, intermediate between
imaging and typical ground-based spectroscopy.
4.2
4.2.1
Data
The 3D-HST survey
The 3D-HST program (van Dokkum et al. 2011; Brammer et al. 2012) is a 625
arcmin2 survey using HST to obtain low-resolution near-IR spectra for a complete
and unbiased sample of thousands of galaxies. (Cycles 18 and 19, PI: van Dokkum).
It observes the AEGIS, COSMOS, GOODS-S and UDS fields with the HST/WFC3
G141 grism over 248 orbits, and it incorporates similar, publicly-available data, in
the GOODS-N field (GO:11600; PI:Weiner). These fields coincide with the area covered by CANDELS (Grogin et al. 2011; Koekemoer et al. 2011) and have a wealth
of publicly available imaging data (U band to 24µm). The 3D-HST photometric catalogue is described in Skelton et al. (2014), and it constitutes a fundamental step
in interpreting the spectra that often contain only a single emission line, if any, by
providing a photometric redshift prior to the redshift fitting.
The WFC3 grism spectra have been extracted with a custom pipeline, described
in Momcheva et al. (2015, in prep). Redshifts have been measured via the combined
photometric and spectroscopic information using a modified version of the EAZY
code (Brammer et al. 2008). The precision of redshifts is shown to be σ( 1dz
+z ) = 0.3%
Chapter 4
52
Redshift
2.0
1.5
40
30
1.0 20
0.7 10
10
20
30
40
60
50
40
30
20
10
0 11000
0
10
20
30
40
50
60
0.5 0
0
40 30 20 10 0 11000
N
2.0
1.5
1.0
0.7
0
0.5
60 40 20
N
12000
12000
14000
15000
Wavelength
14000
15000
Wavelength
GRISM redshifts
D4000
13000
13000
Mg
16000
OIII
16000
Na
TiO
TiO
TiO
17000
Na
Hα
TiO
TiO
TiO
17000
Redshift
Redshift
2.0
1.5
40
30
1.0 20
0.7 10
10
20
30
40
60
50
40
30
20
10
10
20
30
40
50
60
0.5 0
0
40 30 20 10 0 11000
N
2.0
1.5
1.0
0.7
0
0
0.5
50 40 30 20 10 0 11000
N
12000
12000
14000
15000
Wavelength
14000
15000
Wavelength
16000
Mg
16000
Mg
Photometric redshifts
D4000
13000
D4000
13000
Na
TiO
TiO
TiO
17000
Na
TiO
TiO
TiO
17000
Figure 4.2: Observed spectra of massive (log(M∗ /M ) > 10.8) galaxies sorted by redshift, and divided into quiescent (top) and star-forming (bottom).
Galaxies are stacked in a narrow, approximately logarithmic redshift spacing. In the left column grism redshifts are used, while in the right column
we take advantage of photometric redshifts only: this demonstrates the quality of grism redshifts and the necessity of high precision in redshift
evaluation for stacking galaxies together. The most prominent features in emission and absorption are marked respectively in green and red. No
significant emission line is seen in the quiescent sample.
Redshift
Ages of massive galaxies at 0.5 < z < 2.0
53
(Brammer et al. 2012, Momcheva et al. 2015 in prep; see Kriek et al. 2015 for a
comparison based on the new MOSFIRE redshifts from the MOSDEF survey).
Stellar masses have been determined using the FAST code by Kriek et al. (2009),
using Bruzual & Charlot (2003) models, and assuming exponentially declining SFHs,
solar metallicity, a Chabrier (2003) IMF, and a Calzetti (2000) dust law.
4.2.2
Sample Selection
We separate quiescent galaxies from star-forming galaxies using a color-color technique, specifically rest-frame U-V versus rest-frame V-J (hereafter: UVJ diagram). It
has been noted in the past that selecting QGs and SFGs based on a single color is
not reliable, because heavily reddened SFGs can be as red as QGs (among others:
Williams et al. 2009). Adding information from a second color (V-J) makes it possible to empirically distinguish between galaxies that are red in U-V because of an old
stellar population featuring strong Balmer/D4000 breaks (which are relatively blue
in V-J) from galaxies that are instead red in U-V because of dust (and therefore are
red in V-J too).
The UVJ diagram has been widely used in a variety of high redshift studies (e.g.,
Wuyts et al. 2007; Williams et al. 2009; Bell et al. 2012; Gobat et al. 2013), it has been
shown to correspond closely to the traditional morphological classes of early-type
and late-type galaxies up to at least z ∼ 1 (Patel et al. 2012) and is able to select
dead galaxies with low mid-infrared fluxes (Fumagalli et al. 2014).
Effectively, QGs are identified with the criteria (U − V ) > 0.8 × (V − J ) + 0.7,
U − V > 1.3 and V − J < 1.5 (as in Whitaker et al. 2014). The separating lines are
chosen with the main criteria being that they lie roughly between the two modes of
the population seen in Figure 4.1.
We select galaxies more massive than log(M∗ /M ) > 10.8. In order to achieve
a sample of high-quality spectra, we exclude spectra contaminated by neighboring
objects for more than 10% of their total flux, with a wavelength coverage lower than
80% of the full regime of 1.1 to 1.7 µm, and with a fraction of bad pixels higher
than 10%. All of these quantities are listed in the 3D-HST catalogs. The final sample
contains 572 galaxies between redshift 0.5 and 2.0. Figure 4.1 shows the selection of
massive galaxies, divided into SFGs and QGs, in three redshift bins, superposed on
the entire population of galaxies from 3D-HST at the same redshift.
Figure 4.2 (left) shows all the spectra in the sample in observed wavelength sorted
by redshift and divided into QGs (top) and SFGs (bottom). Spectra are stacked in 50
redshift bins with a roughly exponential spacing. We show the number of galaxies
in each bin in the histograms on the right. We see emission and absorption lines
being shifted in the wavelength direction, and entering and exiting the observed
range at different redshifts. For instance, the Hα line enters the wavelength range of
the WFC3 grism at z ∼ 0.7 and exits at z ∼ 1.5.
Chapter 4
54
Quiescent Galaxies 10.80 < logM < 11.50
1.4
1.5<z<2.0
Flux / Continuum (+ constant)
N = 77
1.2
1.0<z<1.5
N = 85
0.5<z<1.0
1.0
N = 86
Hd
Hg+G
OIII
Hb Mg
Na
SII
OI Ha
TiO
TiO
TiO
0.8
4000
6000
Rest-Frame Wavelength
8000
10000
Figure 4.3: Rest-frame stacks of quiescent galaxies with log(M∗ /M ) > 10.8, in three redshift
bins. The stacks are continuum-subtracted. Many absorption bands are visible, while no
obvious emission lines are seen.
Star-Forming Galaxies 10.80 < logM < 11.50
1.4
1.5<z<2.0
Flux / Continuum (+ constant)
N = 100
1.2
1.0<z<1.5
N = 126
0.5<z<1.0
1.0
N = 113
Hd
Hg+G
OIII
Hb Mg
Na
SII
OI Ha
TiO
TiO
TiO
0.8
4000
6000
Rest-Frame Wavelength
8000
10000
Figure 4.4: Rest-frame stacks of star-forming galaxies with log(M∗ /M ) > 10.8, in three
redshift bins. The stacks are continuum-subtracted. Both emission and absorption lines are
visible.
Ages of massive galaxies at 0.5 < z < 2.0
55
We notice that the subdivision into SFGs and QGs corresponds well to a selection
on the presence of emission lines. In the QGs sample (Figure 4.2, left) there are no
obvious emission lines visible, while at different redshifts we observe deep absorption bands (CaII, Mg, Na, TiO). The SFGs sample (Figure 4.2, right) features strong
emission lines, such as the already mentioned Hα, and Hβ and [OIII] at higher redshift. Significantly, some absorption bands are detectable also in the SFG sample.
This experiment proves the quality of 3D-HST grism redshifts. As a comparison,
we show in the right column of Figure 4.2 the effect of lower quality redshifts on
the stacking procedure, by using photometric redshifts instead of grism redshifts.
Even though the photometric redshifts provided in the 3D-HST photometric catalogs (Skelton et al. 2014) reach an excellent absolute deviation from spectroscopic
redshifts of just σ = 1 − 2% (depending on the field), this level of precision is not
good enough to resolve spectral features. In the right column of Figure 4.2, emission
and absorption lines are smoothed across every redshift element; the lines do not
line up well in the redshift space and they are almost indistinguishable from the
continuum at z > 1.5. This experiment demonstrates that stacking grism spectra
requires the sub-percent precision in redshift achieved with the 3D-HST z grism .
4.3
Methods
4.3.1
Stacking
In individual galaxies in our sample, spectral features are often too weak to be used
for reliable measurements of stellar population parameters. We therefore achieve
the necessary signal-to-noise ratio by stacking spectra in 3 redshift bins, and in the
two populations of SFGs and QGs, as follows. We shift the spectra to rest-frame
and fit the continuum in each spectrum with a third order polynomial. In this
process we mask regions around known strong emission lines. We normalize the
spectra by dividing them by the best-fit polynomial. We next determine the median
flux of the normalized rest-frame spectra in a grid of 20Å. Errors on the stacks are
evaluated via bootstrapping: we perform 100 realizations of each sample by drawing
random galaxies from the original sample (repetitions are possible) and we perform
the stacking analysis on each resampling. The uncertainty in the flux measurement
of each wavelength bin is given by the dispersion of the flux values in the resampled
stacks.
The composite spectra are shown in Figure 4.3 (for QGs) and Figure 4.4 (for
SFGs). Each individual stack is made from the sum of at least 75 galaxies. As the observations cover a constant range of observed wavelength (11000Å < λ < 16000Å), we
probe different rest-frame wavelength regimes at different redshifts. The strongest
features are the emission lines of Hα and [OIII](λ=5007Å) with a peak strength of
15% over the normalized continuum. The absorption lines have depths of 5% or less.
The features are weak due to the limited resolution of the spectra.
Chapter 4
56
4.3.2
Model fitting
We compare the stacked spectra with predictions from stellar-population synthesis
models (SPS hereafter). Our goal is to test whether different SPS codes can reproduce the absorption line properties of high redshift galaxies, and to infer stellar ages
for these galaxies. We use models from Bruzual & Charlot (2003, BC03), Conroy &
Gunn (2010, FSPS10), and Conroy et al. (in prep, CKC14)
We use the most standard settings for each SPS code. The BC03 models are based
on the Padova stellar evolution tracks and isochrones (Bertelli et al. 1994); they use
the STELIB empirical stellar library (Le Borgne et al. 2003) for wavelengths between
3200Å < λ < 9500Å and the BaSeL library of theoretical spectra elsewhere.
The FSPS10 models are based on a more updated version of the Padova stellar
evolution tracks and isochrones (Marigo et al. 2008); they use the MiLeS empirical
stellar library for wavelengths between 3500Å < λ < 7500Å and the BaSeL library
of theoretical spectra library elsewhere.
We have also considered a new, high resolution theoretical spectral library (CKC14,
Conroy et al. in prep). This library is based on the Kurucz suite stellar atmosphere
and spectral synthesis routes (ATLAS12 and SYNTHE) and the latest set of atomic
and molecular line lists. The line lists include both lab and predicted lines, the latter being particularly important for accurately modeling the broadband SED shape.
The grid was computed assuming the Asplund (2009) solar abundance scale and a
constant microturbulent velocity of 2 km/s.
For each observed stack, we perform a least-squares minimization using the three
different models, to find the best-fit age of the stack. In order to compare the high
resolution models with the low resolution stacks, we need to downgrade the models
to 3D-HST resolution. The dispersion of the G141 grism is 46Å/pixel (R ∼130 in the
raw data) with a raw pixel scale of 0.12 arcsec, sampled with 0.06 arsec pixels; as the
spectra have high spatial resolution and low spectral resolution (see Brammer et al.
2012), the line width almost exclusively reflects the size of the galaxy in the dispersion direction (Nelson et al. 2012, 2013). We simulate the expected morphological
broadening by convolving the high-resolution model spectra with the object morphology in the HF140W continuum image collapsed in the spatial direction. Model
spectra for each galaxy in the sample are continuum-divided and stacked with the
same procedure we use for observed spectra. In the fitting of models to data, we
allow an additional 3rd order polynomial continuum component with free parameters.
In summary, for each sample of galaxies, we create mock 3D-HST stacks based
on three stellar population models (BC03, FSPS10, CKC14) with three different starformation histories (single stellar burst, constant star formation, and an exponentially declining model with τ = 1 Gyr), with a spacing in ages of 0.1 Gyr.
Our choice of performing the stacking and the fitting analysis on continuumdivided spectra is motivated by the goal of measuring ages of galaxies without being
influenced by the slope of the continuum, which is degenerate in dust and age.
Aging stellar populations have redder broadband colors, however dust reddening
has a similar effect. For instance, the difference in g − r color corresponding to 1
Ages of massive galaxies at 0.5 < z < 2.0
57
Gyr of passive aging within the BC03 models, can also be caused by 0.5 mag of dust
reddening following the Calzetti et al. (2000) dust law.
In this study we do not treat galaxies hosting an active galactic nucleus (AGN)
separately. We test the influence of AGNs by selecting all sources falling in the
IRAC color-color selection presented in Donley et al. (2012). The IRAC selected
AGNs count for less than 5% of each sample of QGs/SFGs at different redshifts.
The conclusions of the paper do not change when these sources are removed from
the stacks.
4.4
Quiescent Galaxies
We fit the stacks of QGs with SSPs from three SPS models (BC03, FSPS10, CKC14)
assuming stellar metallicity and a Chabrier IMF. In this section, we will discuss separately the quality of the fits for different SPS models, and the stellar ages determined
from the best fits.
4.4.1
Quality of fits
Figure 4.5 shows for each of the models (BC03, FSPS10, CKC14) the best fits to the
QG stack for the lower redshift bin (0.5 < z < 1.0). The grey shaded area represents
the area around Hα, masked in the fitting. With BC03 (Figure 4.5, red) the best fit
QG, 0.5 < z < 1.0
Flux / Continuum
1.15
BC03, age = 3.8 Gyr
CKC14, age = 4.0 Gyr
FSPS10, age = 2.4 Gyr
1.10
1.05
1.00
0.95
3D-HST stack, QG, 10.8 < logM < 11.5
0.90
Residuals
0.85
0.15
BC03 χ2 = 11.46
CKC14 χ2 = 5.85
FSPS10 χ2 = 8.46
0.10
0.05
0.00
-0.05
6000
7000
8000
Wavelength
9000
10000
Figure 4.5: Best fits to the stack of quiescent galaxies at 0.5< z <1.0, log(M∗ /M ) > 10.8
(purple), with BC03 (red), FSPS10 (green), and CKC14 (blue) SSPs. Errors on stacks are
computed through bootstrapping on the sample. The grey area represents the wavelength
region around Hα masked in the fitting process. A comparison of residuals from different
models is shown in the bottom panel. The CKC14 models provide the lowest χ2 ; its best-fit
age is 4.0 Gyr.
Chapter 4
58
QG, 1.0 < z < 1.5
Flux / Continuum
1.15
BC03 SSP, age = 3.8 Gyr
CKC14 SSP, age = 2.0 Gyr
FSPS10 SSP, age = 1.4 Gyr
1.10
1.05
1.00
0.95
3D-HST stack, QG, 10.8 < logM < 11.5
0.90
0.85
0.15
BC03 χ2 = 3.36
CKC14 χ2 = 2.85
FSPS10 χ2 = 3.63
Residuals
0.10
0.05
0.00
-0.05
5000
5500
6000
6500
Wavelength
7000
7500
Figure 4.6: Best fits to the stack of quiescent galaxies at 1.0< z <1.5, log(M∗ /M ) > 10.8,
with the same color coding as Figure 4.5. The χ2 of different models are comparable. The age
determinations span a wide range from 1.4 to 3.8 Gyr, according to the model in use.
QG, 1.5 < z < 2.0
Flux / Continuum
1.15
BCO3 SSP, age = 1.4 Gyr
CKC14 SSP, age = 1.2 Gyr
FSPS10 SSP, age = 1.4 Gyr
1.10
1.05
1.00
0.95
3D-HST stack, QG, 10.8 < logM < 11.5
0.90
Residuals
0.85
0.15
BC03 χ2 = 2.57
CKC14 χ2 = 3.35
FSPS10 χ2 = 2.62
0.10
0.05
0.00
-0.05
4000
4500
5000
Wavelength
5500
6000
Figure 4.7: Best fits to the stack of quiescent galaxies at 1.5< z <2.0, log(M∗ /M ) > 10.8,
with the same color coding as Figure 4.5. A comparison of residuals from different models
is shown in the bottom panel. Residuals from different models are comparable, and the age
determination converges to values of 1.2-1.4 Gyr.
is very poor at wavelengths higher than 7500 Å, as shown by the residuals in the
lower panel. Moreover, the χ2red value of the best fit is high (11.46). Using FSPS10
(green), the best fit also has significant (> 3%) residuals at the reddest wavelengths
(> 8000Å) and around the > 7000Å regime, where the first TiO band lies. The χ2red
Ages of massive galaxies at 0.5 < z < 2.0
59
value of the best fit is still high (8.5). Finally, using the latest CKC14 models (Figure
4.5, bottom left) the best fit converges with a lower χ2red = 5.8, and residuals are
below 2% consistently over the entire wavelength range. We compare residuals from
different SPS models (data-model) in the bottom panel of Figure 4.5. All best-fits
have positive residuals in the 7000Å region (up to 4% for FSPS10), underestimating
the fluxes at those wavelengths. At wavelengths higher than 8000Å the BC03 models
have positive residuals, while FSPS10/CKC14 have negative ones.
Moving to higher redshifts, the quality of fits with different SPS models is comparable. In the intermediate (1.0 < z < 1.5, Figure 4.6) and in the high redshift bin
(1.5 < z < 2.0, Figure 4.7) we examine, all χ2red range from 2.8 to 3.3. Residuals in
these redshift bins are comparable among different models, and tend to be smaller
than a few percent. We evaluate residuals corresponding to the Hα line in quiescent
galaxies in Section 4.6.3.
4.4.2
Determination of Ages
The stellar ages of quiescent galaxies implied by the best-fit models vary according
to the used SSP. In order to evaluate the uncertainty of the age measurement we
bootstrap the sample 100 times and repeat the fitting analysis on the bootstrapped
realizations of the stack.
At the lowest redshift bin (0.5 < z < 1.0, Figure 4.5) we obtain a stellar age of 3.8
±0.6 Gyr with BC03, a younger age (2.4 ±0.4 Gyr) with FSPS10 and again 4.0 ±0.2
Gyr with CKC14 (which is the model with the lowest residuals). A similar wide
range of age determinations is obtained for the intermediate redshift bin (Figure
4.6) ranging from 1.4 ±0.1 Gyr for FSPS10, to 2.0 ±0.3 Gyr for CKC14 to 3.8 ±0.8
Gyr for BC03.
In the highest redshift bin (1.5 < z < 2.0, Figure 4.7) all the age determinations
are between 1.2 and 1.4 Gyr. This value is consistent with Whitaker et al. (2013),
Flux / Continuum
1.15
CKC14
CKC14
CKC14
1.10
Z = 2 ZO•
Z = ZO•
Z = 0.5 ZO•
1.05
1.00
0.95
0.90
3D-HST stack, 0.5 < z < 1.0
0.85
0.15
Residuals
age = 1.2 Gyr
age = 4.0 Gyr
age = 2.0 Gyr
χ2red = 8.08
χ2red = 5.65
χ2red = 4.63
0.10
0.05
0.00
-0.05
6000
7000
8000
Wavelength
9000
10000
Figure 4.8: Best fits to the stack of quiescent galaxies at 0.5< z <1.0, log(M∗ /M ) > 10.8,
with the CKC14 models and varying metallicity. A comparison of residuals from different
models is shown in the bottom panel.
Chapter 4
60
who studied a sample of galaxies at slightly different masses and redshifts (10.3 <
log(M∗ /M ) < 11.5, 1.4 < z < 2.2) obtaining an age of 1.25 Gyr computed with
the Vakzdekis models. We also agree with Mendel et al. (2015), who investigate the
stellar population of 25 massive galaxies with VLT-KMOS, deriving a mean age of
1.08 0.13
−0.08 Gyr.
Our study relies on the assumption that QGs already have a solar metallicity at
high redshift. This assumption is supported by the study of Gallazzi et al. (2014),
who studied 40 quiescent galaxies at 0.65 < z < 0.75 with IMACS spectra, obtaining
a mass-metallicity relation consistent with that at z = 0 from SDSS. We explore the
effect of changing metallicity in Figure 4.8 on the stack with the highest signal-tonoise ratio. We use the stellar population model that gave the lowest χ2 with the
standard solar metallicity (CKC14), and vary the metallicity to twice solar and half
solar. The best fit is not significantly improved in either case and both the stellar ages
are lower than in the solar case. A more accurate determination of the metallicity
with higher resolution spectra is needed to measure ages more accurately.
4.5
Star Forming Galaxies
We fit SFGs with the same set of models (BC03, FSPS10, CKC14) with two different
star formation histories: a model with constant star formation (CSF), and one with
an exponentially declining SFR in the form of SFR(t) ∼ exp(−t/τ ), with τ = 1Gyr.
4.5.1
Quality of fits
Figure 4.9 summarizes the best fits to the SFG sample at 0.5 < z < 1.0 obtained with
these combinations of models, letting the age t vary. We again mask a 400Å wide
region around Hα in the fit. According the χ2 statistics, the models assuming an
SFG, 0.5 < z < 1.0
SFG, 0.5 < z < 1.0
1.10
1.05
1.00
0.95
3D-HST stack, SFG, 10.8 < logM < 11.5
0.90
0.85
0.15
Residuals
1.15
Flux / Continuum
BC03 CSF, age = 0.2 Gyr
CKC14 CSF, age = 5.0 Gyr
FSPS10 CSF, age = 1.2 Gyr
BC03 χ2 = 4.20
CKC14 χ2 = 7.11
FSPS10 χ2 = 8.80
0.10
0.05
0.00
BC03 tau 1Gyr, age = 0.6 Gyr
CKC14 tau 1Gyr, age = 5.0 Gyr
FSPS10 tau 1Gyr, age = 1.4 Gyr
1.10
1.05
1.00
0.95
3D-HST stack, SFG, 10.8 < logM < 11.5
0.90
0.85
0.15
Residuals
Flux / Continuum
1.15
BC03 χ2 = 3.71
CKC14 χ2 = 7.06
FSPS10 χ2 = 5.59
0.10
0.05
0.00
-0.05
-0.05
6000
7000
8000
Wavelength
9000
6000
7000
8000
Wavelength
9000
Figure 4.9: Best fits to the stack of star-forming galaxies at 0.5< z <1.0, log(M∗ /M ) >
10.8, with BC03 (red), FSPS10 (green) and CKC14 (blue) models. The grey area represents
the wavelength region around Hα masked in the fitting process. On the left models with
a constant star formation rate are used; on the right, exponentially declining models with
τ = 1Gyr. BC03 models provide the best fits, according to a χ2 statistic. Best-fit ages vary
significantely among SPS models and assumed SFHs.
Ages of massive galaxies at 0.5 < z < 2.0
61
SFG, 1.0 < z < 1.5
SFG, 1.0 < z < 1.5
1.05
1.00
0.95
3D-HST stack, SFG, 10.8 < logM < 11.5
0.90
0.85
0.15
BC03 χ2 = 4.71
CKC14 χ2 = 2.39
FSPS10 χ2 = 5.37
0.10
Residuals
Flux / Continuum
BC03 tau 1Gyr, age = 2.4 Gyr
CKC14 tau 1Gyr, age = 2.4 Gyr
FSPS10 tau 1Gyr, age = 2.4 Gyr
1.10
1.15
0.05
0.00
BC03 tau 1Gyr, age = 3.8 Gyr
CKC14 tau 1Gyr, age = 2.2 Gyr
FSPS10 tau 1Gyr, age = 4.0 Gyr
1.10
1.05
1.00
0.95
3D-HST stack, SFG, 10.8 < logM < 11.5
0.90
0.85
0.15
BC03 χ2 = 2.80
CKC14 χ2 = 2.41
FSPS10 χ2 = 2.21
0.10
Residuals
Flux / Continuum
1.15
0.05
0.00
-0.05
-0.05
5000
5500
6000
6500
Wavelength
7000
7500
5000
8000
5500
6000
6500
Wavelength
7000
7500
8000
Figure 4.10: Best fits to the stack of star-forming galaxies at 1.0< z <1.5, log(M∗ /M ) > 10.8,
with BC03 (red), FSPS10 (green) and CKC14 (blue) models. The grey area represents the
wavelength region around [OIII] masked in the fitting process. On the left models with
a constant star formation rate are used; on the right, exponentially declining models with
τ = 1Gyr. Best fit ages vary from 2 to 4 Gyr.
SFG, 1.5 < z < 2.0
SFG, 1.5 < z < 2.0
1.10
1.05
1.00
0.95
3D-HST stack, SFG, 10.8 < logM < 11.5
0.90
0.85
0.15
Residuals
1.15
Flux / Continuum
BC03 CSF, age = 0.2 Gyr
CKC14 CSF, age = 0.2 Gyr
FSPS10 CSF, age = 0.2 Gyr
BC03 χ2 = 1.93
CKC14 χ2 = 2.39
FSPS10 χ2 = 2.21
0.10
0.05
0.00
BC03 tau 1Gyr, age = 0.2 Gyr
CKC14 tau 1Gyr, age = 0.2 Gyr
FSPS10 tau 1Gyr, age = 0.2 Gyr
1.10
1.05
1.00
0.95
3D-HST stack, SFG, 10.8 < logM < 11.5
0.90
0.85
0.15
Residuals
Flux / Continuum
1.15
BC03 χ2 = 2.01
CKC14 χ2 = 2.41
FSPS10 χ2 = 2.21
0.10
0.05
0.00
-0.05
-0.05
4000
4500
5000
Wavelength
5500
6000
4000
4500
5000
Wavelength
5500
6000
Figure 4.11: Best fits to the stack of star-forming galaxies at 1.5< z <2.0, log(M∗ /M ) > 10.8,
with BC03 (red), FSPS10 (green) and CKC14 (blue) models. The grey area represents the
wavelength region around [OIII] masked in the fitting process. On the left models with
a constant star formation rate are used; on the right, exponentially declining models with
τ = 1Gyr. Different SPS models have similar residuals and converge to very young ages.
exponentially declining SFH provide marginally beter fits than models assuming a
CSF.
Figures 4.10 and 4.11 show the best fits to the SFG samples at 1.0 < z < 1.5 and
1.5 < z < 2.0 obtained with the same combination of models. We mask 400Å wide
regions around the expected strongest emission lines ([OIII], Hα). In the highest
redshift stack, we see the biggest residual corresponding to the wavelength of the
Hγ line (λ = 4341Å). For these redshifts, FSPS10 and CKC14 model have a lower
best-fit χ2 than that of BC03. However the models fit almost equally well.
Star-forming galaxies of these masses are in fact known to follow declining starformation histories at redshifts lower than 1.5 (e.g. Pacifici et al. 2012). As in the case
of QGs, different models have qualitatively and quantitatively different residuals. In
this case however BC03 is the model with the smallest residuals, especially at the
longest wavelengths (> 8500Å).
Chapter 4
62
0.10
Residuals
0.05
0.00
-0.05
CKC14 CSF Z=0.190
CKC14 tau1Gyr Z=0.190
CKC14 CSF Z=0.096
CKC14 tau1Gyr Z=0.096
-0.10
6000
7000
8000
Wavelength
9000
Figure 4.12: For star-forming galaxies at redshift 0.5 < z < 1.0, we compare residuals to the
best-fits obtained by varying metallicity and star formation history. The shape of residuals
changes less than by changing the assumed stellar population model (compare with Figure
4.9). The maximum difference in the χ2 statistics for best fits with varying metallicity is ∆χ2 =
0.2, negligible in comparison to the difference obtained by varying the stellar population
model (∆χ2 = 5).
The effect of varying metallicity is explored in Figure 4.12 and compared to that
of varying stellar population models. We perform this test on SFGs at the lowest
redshift bin (0.5 < z < 1.0), where the signal-to-noise ratio is high. Figure 4.12 shows
a comparison between residuals from best-fit CKC14 models with solar metallicity
models (Z=0.190) and models with half of the solar metallicity (Z=0.096), for a 1
Gyr τ model SFH, and a Constant SFR. We notice that in this case residuals do not
vary significantely. We conclude that the difference between different SPS models is
greater than that obtained by using the same SPS code, with different metallicities
and/or SFHs.
4.5.2
Determination of Ages
For star-forming galaxies at the lowest redshift we examine (0.5 < z < 1.0, Figure
4.9), the overall best-fit (χ2red = 3.71) is obtained with a young (0.6 Gyr of age) stellar
population with the BC03 τ model. We also obtain a young age (0.2 Gyr) when
assuming a constant star formation history for the same stellar population model.
The age determinations from FSPS10 and CKC14 indicate instead an older age, from
1 to 5 Gyr (Figure 4.9). We notice that for star-forming galaxies we cannot exclude
any particular age range, since the stellar ages inferred from different models vary
greatly.
At intermediate redshift (1.0 < z < 1.5, Figure 4.10), we obtain ages around 2-3
Gyrs with different models (the typical error on each inferred age for star-forming
galaxies is 1 Gyr). At the highest redshifts (1.5 < z < 2.0, Figure 4.11), all bestfits (with different stellar population models and different star formation histories)
converge to the lowest age value.
For galaxies with active star formation, the intrinsic strength of features does not
Ages of massive galaxies at 0.5 < z < 2.0
63
vary significantely with time since the spectra are dominated by light from young
stars that been constantly forming. With the current signal-to-noise, we therefore
cannot draw any conclusion on the ages of SFGs at the redshift under consideration.
4.6
4.6.1
Discussion
Differences among SPSs
In order to investigate the origin of the qualitative difference in the best fits described
in Section 4.4, we compare model spectra from different SPS codes. We show model
SSPs from the BC03, FSPS10 and CKC14 codes in Figure 4.13, for different ages. The
strengths of the absorption lines vary among models, for every age. We notice in
particular that at wavelengths higher than 7500Å different models predict different
absorption bands at different wavelengths. We quantify the spread in the models
at different wavelengths by computing the mean difference between every possible
combination of models at the same age (Figure 4.13, bottom). This value is lower
than 1% at wavelength between ∼ 4500Å and 6500Å, and is larger otherwise. In
particular the region with wavelengths greater than 8000Å has a large discrepancy
between BC03 on one side, and FSPS10 and CKC14 on the other. This explains why
determinations of ages from 3D-HST at higher redshifts are more stable between
different models than those at lower redshift. For our lowest redshift sample (0.5 <
z < 1.0) we observe the region of the spectrum where discrepancies among models
are the largest, while at high redshift we observe rest-frame wavelengths where
models are more similar to each others.
4.6.2
Evolution of Ages
We investigate the evolution of ages of quiescent galaxies in a mass limited sample.
In Fig. 4.14 (left) we show the ages obtained by fitting 3D-HST stacks with different
stellar population synthesis models. Even though the model dependent spread in
ages is large, we observe that QGs are younger at higher redshift and that, at each
redshift, QGs are not maximally old; instead their age is smaller than half of the age
of the Universe at the same redshift. We compare to data in a similar mass range
by Gallazzi et al. (2014) at z ∼ 0.6 and by Whitaker et al. (2013, who also uses
spectra from 3D-HST) at 1.4 < z < 2.2, obtaining good agreement. The young ages
of quiescent galaxies can be naturally explained by the addition of newly quenched
galaxies to the sample. The conclusion is enforced at lower redshift by Choi et al.
(2014) who find that quiescent galaxies at 0.2 < z < 0.7 tend to be younger than half
of the age of the Universe at those redshifts. The age value from the highest redshift
mapped by Choi et al. (2014) is consistent with the youngest determination from
our sample (the best fit from the FSPS10 models).
Chapter 4
64
Hd Hg+G Hb Mg
Na
Ha
TiO
TiO
CaII
Age = 0.5 Gyr
1.0
BC03
CKC14
FSPS10
Flux / Continuum (+ constant)
Age = 1.0 Gyr
BC03
CKC14
FSPS10
0.8
Age = 4.0 Gyr
BC03
CKC14
FSPS10
Age = 9.0 Gyr
0.6
BC03
CKC14
FSPS10
<|modelI-modelJ|>
Hd Hg+G Hb Mg
Na
Ha
TiO
TiO
CaII
0.4
4000
5000
6000
7000
Wavelength
8000
9000
10000
0.05
0.04
0.03
0.02
0.01
0.00
4000
5000
6000
7000
Wavelength
8000
9000
10000
Figure 4.13: Top: SSPs with solar metallicities from BC03 (red), FSPS10 (green), CKC14 (blue).
Models are convolved to 3D-HST resolution (see Section 4.3). Bottom: The purple line shows
the average absolute difference between models at different wavelengths, for every combination of models of the same age: the difference among models is the biggest at the longest
wavelengths and in the D4000 region.
Ages of massive galaxies at 0.5 < z < 2.0
65
Quiescent Galaxies
Quiescent Galaxies
8
/L
∆(M
8
)
Age (Gyr)
e
of
e
Un
i
ve
rs
e
4
2
0
0.0
BC03
FSPS10
CKC14
FAST
Mendel+15
Choi+14
Gallazzi+14
Whitaker+13
0.5
Ag
6
th
Age (Gyr)
Ag
6
e
of
th
e
Un
ive
rs
e
4
2
1.0
Redshift
1.5
2.0
0
0.0
CKC14 (T.W.)
Choi+14
Gallazzi+14
0.5
1.0
Redshift
1.5
2.0
Figure 4.14: Left: Evolution of ages of massive quiescent galaxies (log(M∗ /M ) > 10.8) with
redshift. Open circles represent values measured from 3D-HST. Different colors represent
different stellar population synthesis models (red: BC03, green: FSPS10, blue: CKC14) used
for the determination of ages. As a comparison, values inferred from photometry (black
diamonds) and from the literature selected in a similar mass range (grey symbols) are plotted.
Right: Comparison between the ages determined from 3D-HST spectra and the literature
(blue) to the evolution of ages predicted from the evolution of the mass-to-light ratio inferred
from the fundamental plane.
Comparison with photometry
Ages of galaxies can also be inferred from photometry only. In 3D-HST stellar
masses, star-formation rates, ages and dust extinction are estimated with the FAST
code (Kriek et al. 2009), assuming exponentially declining star formation histories
with a minimum e-folding time of log10 (τ/yr ) = 7, a minimum age of 40 Myr,
0 < AV < 4 mag and the Calzetti et al. (2000) dust attenuation law (see Skelton et
al. 2014). The output from the FAST code is an age defined as the time since the
onset of SF, that is not necessary equivalent to a light-weighted age.
For each galaxy we therefore compute a light-weighted age (tlum ) following the
definition:
tlum =
∑i SFR(ti ) × VSSP (t − ti ) × (t − ti ) × ∆t
∑i SFR(ti ) × VSSP (t − ti ) × ∆t
where:
• t is the time of observation (equivalent to the age of FAST)
• SFR(ti ) is the star formation rate at time ti . In the case of the FAST fits SFR(ti )
has the functional shape of a τ model SFR(ti ) ∼ exp(−ti /τ )
• VSSP (t − ti ) is the V-band flux of 1 M element formed at ti and observed at
time t.
• ∆t is the time-step we divide the SFH in (we use ∆t = 50 Myr).
Chapter 4
66
We notice that for an SSP tlum (t) = t.
Figure 4.15 shows the relation between the time from the onset of star-formation t
and tlum for a range of models: an SSP, a CSF model, and an exponentially declining
model with τ = 1Gyr. As expected, tlum for an SSP (red line) is well approximated
by t. For a CSF (blue line) tlum is always smaller than t, with a increasing difference at later times. This effect is naturally explained by the fact that younger stars
are brighter than older stars: VSSP peaks at 10 Myr, and declines afterwards (in
other words the mass-to-light ratio M/LV increases for older stellar populations,
see among others Bruzual & Charlot 2003, Fig 1 to 5). The τ model (purple line) has
an intermediate behavior, being similar to the CSF for t → 0, and parallel to the SSP
for large t.
SFR
For each quiescent galaxy in the sample, we infer tlum from the best fit to its
photometry, and find the average value in the three redshift bins 0.5 < z < 1.0,
1.0 < z < 1.5 and 1.5 < z < 2.0. Figure 4.14 (left) shows how the quantity compares
to the ages measured from the spectra (Section 4.4) with different sets of models.
10
8
6
4
2
0
0
10
10.0
1.0
2
4
6
8
10
0.1
0.1
10.0
SSP
10.0
1.0
Time since onset of SF (Gyr)
10.0
SSP
8 τ = 1Gyr
Light-weighted Age (Gyr)
1.0
τ = 1Gyr
CSF
CSF
6
1.0
4
2
0
0
2
4
6
8
Time since onset of SF (Gyr)
10
0.1
0.1
Figure 4.15: Light-weighted ages (tlum ) at different times from the onset of star-formation,
in linear (left) and logarithmic scale (right). Three different SFH are shown: single stellar
population (SSP, red), constant star formation (CSF, blue), and an exponentially declining τmodel with τ = 1Gyr. For a SSP the light weighted age corresponds to the time from the
burst. For a CSF tlum is always lower than the time from the onset of star-formation, with
approximately tlum ∼ t/2. A τ model has an intermediate behaviour, being asymptotically
similar to the CSF at t → 0 and to the 1-1 slope at later times.
Ages of massive galaxies at 0.5 < z < 2.0
67
For each redshift bin ages derived from photometry tend to be comparable to the
lowest values obtained with the spectral fitting.
Comparison with fundamental plane studies
Another technique commonly used to constrain ages of high redshift quiescent
galaxies is provided by the fundamental plane (Djorgovski & Davis 1987, hereafter:
FP). In particular, the FP is a model-independent tool for measuring the mass to light
ratio (M/L). The offset between the M/L of high redshift galaxies and that of local
galaxies can therefore be used to estimate the age of their stellar populations (Franx
1993, van Dokkum & Franx 1996, van der Wel et al. 2004, Treu et al. 2005). Since the
luminosity of an SSP evolves with time as L ∼ tk , (with the parameter k derived from
stellar population models), the relation between the evolution of M/L and lightweighted ages can therefore be approximated as ∆ln(M/L) ∼ −k∆ln(t). Measurements of the evolution of M/L up to z ∼ 1 agree with values ∆ln(M/LB ) ∼ −1 × z
(van Dokkum & Stanford 2003, Wuyts et al. 2004, Holden et al. 2005). Given a value
of k=-0.98 (from BC03 models in the B band, with a Chabrier IMF), we derive the
following relation between the local light-weighted ages and those at high redshift:
tlum (z) = tlum (0) × e−z/0.98 . The observed evolution of the M/L predicts that the
ages of quiescent galaxies at z ∼ 1 are 2.8 times younger than those at z = 0, and 4
times younger at z ∼ 1.5. Figure 4.14 shows the age evolution predicted from M/L
measurements. We use 8 Gyr as the age of galaxies at z = 0, as measured from SDSS
spectra by Gallazzi et al. (2004). The agreement between the FP prediction and
measurements from 3D-HST spectra is excellent. Only the spectral measurement
with the BC03 models at z ∼ 1.25 and the one with the FSPS10 models at z ∼ 0.75
significantly deviate from the FP prediction.
4.6.3
Hα in quiescent galaxies
At redshifts lower than 1.5 we can quantify the Hα emission1 in QGs from the residuals to the best fits (Figure 4.5 and 4.6). We subtract the best-fit model with the CKC14
SPS from the stacks, and fit residuals with a Gaussian centered at the Hα wavelength
(λ =6563Å). Since the stacks are continuum subtracted, this is effectively a direct
measurement of EW(Hα+[NII]). At the lowest redshifts (0.5 < z < 1.0), we do not
obtain a significant detection, with EW(Hα+[NII])=0.5 ± 0.3Å, while at 1.0 < z < 1.5
we robustly detect the emission line, measuring EW(Hα+[NII]) = 5.5 ± 0.8Å. In the
same mass/redshift regime typical SFGs have EW(Hα+[NII]) ∼ 60Å (Fumagalli et
al. 2012). This shows that the Hα emission in QGs is quenched by a factor of ∼ 10.
To estimate the Hα fluxes, we multiply the EW(Hα+[NII]) by the median continuum flux of galaxies in the stack, and assume a 0.25 ratio for [NII]/(Hα+[NII]).
We finally estimate SFR(Hα) with the Kennicutt (1998) relation, obtaining that QGs
have SFR(Hα) = 0.46 ± 0.06M /yr at 1.0 < z < 1.5 and 0.10 ± 0.05M /yr at
0.5 < z < 1.0, assuming no dust absorption. In Fumagalli et al. (2014) we reported
1 Given the WFC3 grism resolution Hα and [NII] are inevitably blended, therefore we will refer to
measurements of Hα+[NII]
Chapter 4
68
for QGs higher SFR inferred from mid-infrared emissions, up to 3.7 ± 0.7M /yr at
1.1 < z < 1.5. Reconciling the measurements of SFR(Hα) and SFR(IR) would require
a significant dust extinction for the Hα line (A Hα ∼ 3). Both estimates are however
affected by possible contaminations of other physical processes which can contribute
to the observed fluxes. A variety of studies (Fumagalli et al. 2014, Utomo et al. 2014,
Hayward et al. 2014) suggests that SFR inferred from IR are overestimated because
of the contribution of dust heating by old stars and/or TP-AGB stars to the MIR
fluxes. SFRs measured from Hα are instead contaminated by potential AGN or
LINER emission and affected by dust extinction. Our combined multiwavelength
findings agree however in indicating that SFRs of QGs are very low, they are negligible in comparison to those of SFGs at the same redshift, and they are potentially
consistent with 0.
4.7
Conclusions
We select massive galaxies from the 3D-HST survey and divide them into quiescent
and star-forming according to their rest-frame optical and near-infrared colors. We
stack their low-resolution spectra from 3D-HST in three redshift bins, and fit them
with models from three stellar population synthesis codes, in order to infer the mean
stellar ages of the sample.
For quiescent galaxies, we show that the new CKC14 code provides more accurate fits to the data. Other codes do not reproduce the observed features at the
reddest optical wavelengths.
For star-forming galaxies, we are not able to put significant constraints on the
stellar ages of the samples.
Even though we infer different stellar ages from different models, stellar ages of
quiescent galaxies appear to be overall younger than half of the age of the Universe,
confirming the trends found at lower redshift by Choi et al. (2014) and Gallazzi et
al. (2014). The evolution of stellar ages is moreover in accordance with the expected
evolution from fundamental plane studies.
We thank Jesse van de Sande, Charlie Conroy and Adam Muzzin for the useful
discussions. We acknowledge funding from ERC grant HIGHZ no. 227749. This
work is based on observations taken by the 3D-HST Treasury Program (GO 12177
and 12328) with the NASA/ESA HST, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555.
Bibliography
Bell, E. F., Zheng, X. Z., Papovich, C., et al. 2007, ApJ, 663, 834
Bell, E. F., van der Wel, A., Papovich, C., et al. 2012, ApJ, 753, 167
Ages of massive galaxies at 0.5 < z < 2.0
69
Bezanson, R., van Dokkum, P., van de Sande, J., Franx, M., & Kriek, M. 2013, ApJL,
764, L8
Blanton, M. R., Hogg, D. W., Bahcall, N. A., et al. 2003, ApJ, 594, 186
Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008, ApJ, 686, 1503
Brammer, G. B., Whitaker, K. E., van Dokkum, P. G., et al. 2011, ApJ, 739, 24
Brammer, G. B., van Dokkum, P. G., Franx, M., et al. 2012, ApJS, 200, 13
Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000
Bundy, K., Ellis, R. S., Conselice, C. J., et al. 2006, ApJ, 651, 120
Chabrier, G. 2003, ApJL, 586, L133
Choi, J., Conroy, C., Moustakas, J., et al. 2014, ApJ, 792, 95
Damen, M., Labbé, I., Franx, M., et al. 2009, ApJ, 690, 937
Damen, M., Labbé, I., van Dokkum, P. G., et al. 2011, ApJ, 727, 1
Djorgovski, S., & Davis, M. 1987, ApJ, 313, 59
Donley, J. L., Koekemoer, A. M., Brusa, M., et al. 2012, ApJ, 748, 142
Erb, D. K., Shapley, A. E., Pettini, M., et al. 2006, ApJ, 644, 813
Fumagalli, M., Patel, S. G., Franx, M., et al. 2012, ApJL, 757, L22
Fumagalli, M., Labbé, I., Patel, S. G., et al. 2014, ApJ, 796, 35
Gallazzi, A., Charlot, S., Brinchmann, J., & White, S. D. M. 2006, MNRAS, 370, 1106
Gallazzi, A., Bell, E. F., Zibetti, S., Brinchmann, J., & Kelson, D. D. 2014, ApJ, 788, 72
Gobat, R., Strazzullo, V., Daddi, E., et al. 2013, ApJ, 776, 9
Grogin, N. A., Kocevski, D. D., Faber, S. M., et al. 2011, ApJS, 197, 35
Hayward, C. C., Lanz, L., Ashby, M. L. N., et al. 2014, MNRAS, 445, 1598
Ilbert, O., McCracken, H. J., Le Fèvre, O., et al. 2013, A&A, 556, AA55
Karim, A., Schinnerer, E., Martínez-Sansigre, A., et al. 2011, ApJ, 730, 61
Kashino, D., Silverman, J. D., Rodighiero, G., et al. 2013, ApJL, 777, LL8
Kauffmann, G., Heckman, T. M., White, S. D. M., et al. 2003, MNRAS, 341, 33
Kennicutt, R. C., Jr. 1998, ARA&A, 36, 189
Koekemoer, A. M., Faber, S. M., Ferguson, H. C., et al. 2011, ApJS, 197, 36
Kriek, M., van Dokkum, P. G., Labbé, I., et al. 2009, ApJ, 700, 221
70
Chapter 4
Kriek, M., Shapley, A. E., Reddy, N. A., et al. 2015, ApJS, 218, 15
Mendel, J. T., Saglia, R. P., Bender, R., et al. 2015, ApJL, 804, L4
Moustakas, J., Zaritsky, D., Brown, M., et al. 2011, arXiv:1112.3300
Muzzin, A., Marchesini, D., Stefanon, M., et al. 2013, ApJ, 777, 18
Nelson, E. J., van Dokkum, P. G., Brammer, G., et al. 2012, ApJL, 747, L28
Nelson, E. J., van Dokkum, P. G., Momcheva, I., et al. 2013, ApJL, 763, L16
Onodera, M., Renzini, A., Carollo, M., et al. 2012, ApJ, 755, 26
Onodera, M., Carollo, C. M., Renzini, A., et al. 2014, arXiv:1411.5023
Patel, S. G., Holden, B. P., Kelson, D. D., et al. 2012, ApJL, 748, L27
Price, S. H., Kriek, M., Brammer, G. B., et al. 2014, ApJ, 788, 86
Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al. 2004, ApJ, 613, 898
Treu, T., Ellis, R. S., Liao, T. X., et al. 2005, ApJ, 633, 174
Utomo, D., Kriek, M., Labbé, I., Conroy, C., & Fumagalli, M. 2014, ApJL, 783, LL30
van de Sande, J., Kriek, M., Franx, M., et al. 2013, ApJ, 771, 85
van Dokkum, P. G., Whitaker, K. E., Brammer, G., et al. 2010, ApJ, 709, 1018
van Dokkum, P. G., Brammer, G., Fumagalli, M., et al. 2011, ApJL, 743, LL15
Williams, R. J., Quadri, R. F., Franx, M., van Dokkum, P., & Labbé, I. 2009, ApJ, 691,
1879
Wuyts, S., Labbé, I., Franx, M., et al. 2007, ApJ, 655, 51
5
Decreasing Hα for redder
star-forming galaxies: influence of
dust and star formation rates
We analyze a large sample of 7000 galaxies from the 3D-HST survey in the redshift
range 0.7 < z < 1.5, where Hα falls into the wavelength coverage of the WFC3
G141 grism. We divide galaxies onto quiescent and star-forming on the basis of
the widely used UVJ rest-frame color-color criterion. We demonstrate that galaxies
with strong and weak Hα are well separated in the UVJ diagram. The Hα line is
detected (S/N > 3) in ∼ 85% of the star-forming sample, while less than 20% of
UVJ selected quiescent galaxies have an Hα detection, with an average EW(Hα) of
< 5 Å. For star-forming galaxies, we investigate how Hα varies as a function of the
rest frame colors of the galaxy and how it relates to the specific star formation rate
(sSFR), measured from the UV and mid-IR emission. We find that, at a fixed mass,
red star-forming galaxies have lower EW(Hα) than blue star-forming galaxies, with
a decrease of 0.5 dex in EW(Hα) per magnitude in U-V color. We also show that the
median sSFR(UV+IR) of galaxies decreases towards redder U-V colors. In addition,
the median IRX ratio (log LIR / LUV ) increases towards redder colors. This result
demonstrates that the redder colors of red star-forming galaxies, compared to those
of similar-mass blue star-forming galaxies are due to both their higher dust content
and their lower sSFR. We show that the ratio L(Hα) / SFR varies systematically with
both U-V color and stellar mass. The systematic variation is approximately 1 dex
with stellar mass, with an additional variation of 0.5 dex with color after the mass
dependence has been removed. We show that the overall variation of EW(Hα) as a
function of color can be explained by the combined effect of lower sSFR and higher
dust absorption for galaxies with redder colors
Mattia Fumagalli; Marijn Franx; Ivo Labbé; Pieter G. van Dokkum; et al.
Submitted to the Astrophysical Journal
71
Chapter 5
72
5.1
Introduction
In studies of galaxy evolution, populations of star-forming and quiescent objects
are generally tracked through cosmic time based on their rest-frame colors (among
others: Faber et al. 2007, Muzzin et al. 2013). This technique is supported by the
observed bimodality in galaxy colors: in a color-luminosity (or more fundamentally
color-mass) diagram, the distribution of galaxies consists of early-type objects residing on a sequence at red colors well separated from a cloud of blue late-type
galaxies. This bimodality is observed in both the local Universe (e.g. Blanton et al
2003), and up to z ∼ 2.5-3 (e.g. Brammer et al. 2009).
A selection based on a single rest-frame color can however result in a mixed
set of both passive, dead galaxies, and dust-reddened star-forming galaxies, whose
optical colors can be as red as (or redder than) those of purely quiescent galaxies
(Maller et al. 2009). In order to better distinguish star-forming (SFGs) and quiescent
galaxies (QGs), the usage of a color-color diagram, typically rest-frame U-V versus
rest-frame V-J, has emerged (UVJ diagram, Labbé et al. 2005, Williams et al. 2009).
In this color-color space, the old stellar populations of QGs produce red U-V colors
and relatively blue V-J colors, while the reddest SFGs are red in both U-V and V-J.1
The selection of QGs in the UVJ diagram has been used to identify quenched
objects up to z = 4 (Straatman et al. 2014), and corresponds up to at least z ∼ 2.0
to a selection of dead galaxies with low mid-IR fluxes (Fumagalli et al. 2014) and
an old stellar population (Whitaker et al. 2013, Fumagalli et al. 2015, Mendel et al.
2015). Moreover, it is roughly equivalent to a morphological selection of early-type
galaxies, at least up to intermediate redshifts (Patel et al. 2012).
In the UVJ space, the two sequences are basically parallel and both feature a significant color spread. An identical reddening effect, parallel to the sequences, can
be obtained by increasing age, dust, or metallicity. The three effects are notoriously
difficult to disentangle: the difference in rest-frame U-V and V-J colors corresponding to 1 Gyr of passive evolution with the Bruzual & Charlot (2003) models is almost
identical in magnitude and direction to that obtained by increasing the metallicity
from log(Z)=0.02 (Solar) to log(Z)=0.05, or adding 0.5 mag of dust reddening following the Calzetti et al. (2000) dust law.
For QGs, the color spread is interpreted as an effect of mainly age, with blue QGs
featuring a post starburst-like spectral energy distribution (SED), while redder QGs
have SEDs consistent with older stellar ages (Whitaker et al. 2012). Recent works
based on spectroscopy (Whitaker et al. 2013, Mendel et al. 2015) confirmed that the
bluest UVJ selected QGs have strong Balmer lines and a light-weighted age of ∼ 1
Gyr, while the reddest quiescent galaxies are dominated by metal lines and have a
relatively older mean age.
The color spread of SFGs, spanning 2 magnitudes in both U-V and V-J, is instead
generally interpreted as driven by different levels of dust absorption (Williams et al.
2009). At intermediate redshifts Patel et al. (2012) showed that the variation in [OII]
luminosity of star-forming galaxies is consistent with that predicted by models of
dust absorption through an inclined disk, as most of the highly inclined spirals are
1 An
alternative is to correct the observed U-V colors for reddening; see Brammer et al. (2009).
Decreasing Hα for redder star-forming galaxies
73
identified at the reddest rest-frame colors.
In this paper we use the observed Hα to investigate the systematic variation of
SFGs as a function of their rest-frame color. We focus on galaxies in the redshift
window where Hα can be observed with the Hubble Space Telescope / Wide Field
Camera 3 (HST/WFC3) grism (0.7 < z < 1.5), in the context of the 3D-HST survey,
a large spectroscopic survey aimed at obtaining unbiased, mass-selected, samples of
rest-frame spectra at high redshift.
After introducing the data in Section 5.2, we analyze the dependence of EW(Hα)
on color in Section 5.3. In Section 5.4 and 5.5 we investigate its origin by analyzing
color trends for sSFR and dust absorption. We further confirm in Section 5.5 that
emission lines are more absorbed than the underlying continuum.
In the entire paper, we assume a Chabrier (2003) IMF and refer measurements
to a ΛCDM cosmology with ΩM = 0.3, ΩΛ = 0.7, and H0 = 70 km/(s Mpc). All
magnitudes are given in the AB system.
5.2
5.2.1
Data and Sample Selection
The 3D-HST survey
The 3D-HST program (van Dokkum et al. 2011; Brammer et al. 2012) is a 600
arcmin2 wide survey using WFC3 grism onboard Hubble Space Telescope (HST) to
obtain low-resolution near-IR spectra for a complete, unbiased sample of thousands
of galaxies (Cycles 18 and 19, PI: van Dokkum). It has an integration time of 248 orbits and it observes four fields with a wealth of publicly available imaging data from
the U band to 24µm (AEGIS, COSMOS, GOODS-S and UDS), and it incorporates
similar, publicly-available data, in the GOODS-N field (GO:11600; PI:Weiner). Extensive photometric data in these fields have been described in Skelton et al. (2014).
The WFC3 grism spectra have been extracted with the pipeline described in
Momcheva et al. (2015, in prep). Redshifts have been measured via the combined
photometric and spectroscopic information using a modified version of the EAZY
code (Brammer et al. 2008). The precision of redshifts is shown to be σ( 1dz
+z ) = 0.3%
(Brammer et al. 2012, Momcheva et al. 2015 in prep.). In grism spectroscopy the
width of spectral features reflects almost exclusively the size of the galaxy in the dispersion direction (see Nelson et al. 2012, 2013). Emission lines are therefore fit using
a line shape determined by the object profile (Brammer et al. 2012, Momcheva et al.
2015 in prep.). Given the low spectral resolution of 3D-HST, the Hα and [NII ] lines
are unavoidably blended together. For the ease of reading we refer to EW(Hα+[NII ])
simply as EW(Hα) through the entire paper.
Stellar masses have been determined using the FAST code by Kriek et al. (2009),
using Bruzual & Charlot (2003) models, and assuming exponentially declining SFHs,
solar metallicity, a Chabrier (2003) IMF, and a Calzetti (2000) dust law.
SFR measurements are described in Whitaker et al. (2014). In brief, SFRs are
determined by adding the rest-frame UV light from young stars to the IR luminosity
(Gordon et al. 2000). Total IR luminosities (L( IR) = L(8 − 1000µm)) are estimated
from the observed Spitzer/MIPS 24µm fluxes, using a conversion based on a single
Chapter 5
74
template which is the average of Dale & Helou (2002) templates with 1 < α < 2.5,
following Wuyts et al. (2008), Franx et al. (2008), and Muzzin et al. (2010). The
rest-frame UV luminosity is estimated by using the 2800Å rest-frame luminosity
multiplied by 1.5 to account for the UV spectral shape of a 100 Myr old population
with constant SFR (as in Bell et al. 2005): L(UV)=1.5 νLν (2800Å). In the 3D-HST
data, rest frame luminosities such as νLν (2800Å) are estimated from the observed
photometry by interpolating over the best fit templates (Brammer et al. 2011).
Assuming a Chabrier (2003) IMF, SFRs are then evaluated as
SFR[ M /yr ] = 1.09 × 10−10 ( L( IR) + 2.2L(UV ))[ L ]
(5.1)
where L(IR) is the bolometric IR luminosity, L(UV) = 1.5 νLν (2800Å) the 1216 −
3000Å luminosity, and the factor of 2.2 accounts for the unobscured starlight of
young stars emitted redward of 1216Å and blueward of 3000Å (Bell et al. 2005).
5.2.2
Sample selection and the UVJ diagram
The 3D-HST G141 grism spectroscopy covers the Hα rest-frame wavelength at 0.7 <
z < 1.5. We therefore select all galaxies in this redshift range with grism coverage
(which amounts to 75% of F160W-selected galaxies in the CANDELS sample).
The 3D-HST team has done a visual inspection of all the objects in the catalogs
(see Momcheva et al. 2015). Each object has been seen by two classifiers. Each one
independently marked objects with bad spectra, such as ones with a non-subtracted
star contaminating the object, a fake emission line caused by the contamination subtraction, or numerical effects causing blank blobs in the 2D spectrum; an uncertain
category was also available for objects with no clear classification. We exclude objects classified by at least one person as bad or flagged by both as uncertain.
After this selection we are left with 7100 objects in the redshift range 0.7 < z < 1.5
(80% of the original sample). We test possible biases between the original sample
and the final one with a Kolmogorov-Smirnov test on their distribution in mass,
redshifts, F140W magnitudes and U-V color. We find that the two samples are drawn
from the same distribution with respect to the mass (at the 97% level), redshift (at
the 95% level), magnitude (at the 99% level), and color (at the 95% level).
We divide galaxies into star-forming (SFGs) and quiescent (QGs) using a colorcolor technique (Labbé et al. 2005), specifically rest-frame U-V versus rest-frame V-J
(hereafter: UVJ diagram). This method has been shown to be effective in selecting
star-forming galaxies, including the red dust-reddened ones, from galaxies that are
red because their light is dominated by an old stellar population (Williams et al.
2009). Effectively, SFGs are identified with the criteria (U − V ) > 0.8 × (V − J ) + 0.7,
U − V > 1.3 and V − J < 1.5 (as in Whitaker et al. 2014).
Decreasing Hα for redder star-forming galaxies
75
fraction of Hα emitters
9.5 < logM < 10.0
10.0 < logM < 10.5
2.0
1.5
100%
80%
60%
40%
20%
0%
1.0
U-V
0.5
100%
80%
60%
40%
20%
0%
10.5 < logM < 11.0
11.0 < logM < 11.5
2.0
1.5
100%
80%
60%
40%
20%
0%
1.0
0.5
0.5
1.0
1.5
100%
80%
60%
40%
20%
0%
2.0
0.5
1.0
1.5
2.0
V-J
Figure 5.1: UVJ diagram for 3D-HST galaxies at 0.7 < z < 1.5 in different mass bins. The
selection box divides quiescent and star-forming galaxies. Each 0.1×0.1 mag bin has an area
proportional to the number of galaxies in the region, and it is color-coded by its percentage
of Hα detections (S/N > 3). Star-forming galaxies have generally a high fraction (>80%) of
Hα emitters, while less than 20% quiescent galaxies have an Hα detections. The fraction of
Hα detections in star-forming galaxies decreases towards the reddest colors.
5.3
EW(Hα): trend with color
5.3.1
Separating star-forming and quiescent galaxies with the UVJ
selection
We investigate first how the Hα emission varies as a function of the spectral energy
distribution shape of galaxies, as tracked by a galaxy’s position in the UVJ diagram.
In Figure 5.1 we show, for different mass bins, the UVJ diagram divided into colorcolor bins of size 0.1 mag, where each bin is color-coded by the fraction of galaxies
with an Hα detection (defined by signal-to-noise > 3). For QGs the fraction of Hα
emitters is generally low (< 20%), while for SFGs it is, for most color-color bins,
Chapter 5
76
median log EW(Hα)
9.5 < logM < 10.0
10.0 < logM < 10.5
2.0
1.5
3.00
2.25
1.50
0.75
0.00
1.0
U-V
0.5
3.00
2.25
1.50
0.75
0.00
10.5 < logM < 11.0
11.0 < logM < 11.5
2.0
1.5
3.00
2.25
1.50
0.75
0.00
1.0
0.5
0.5
1.0
1.5
2.0
3.00
2.25
1.50
0.75
0.00
V-J
0.5
1.0
1.5
2.0
Figure 5.2: Dependence of EW(Hα) on U-V and V-J colors. Each 0.1x0.1mag bin is color-coded
by median log EW(Hα), for different mass bins. The selection box is used to divide quiescent
and star-forming galaxies. Quiescent galaxies have EW(Hα) of a few Å; star-forming galaxies
span a range in EW(Hα) from 10 to a few hundreds. At the same mass red star-forming
galaxies tend to have lower EW(Hα) than that of blue star-forming galaxies.
higher than 80%. We notice that the fraction of SFGs that are Hα emitters decreases
towards the reddest colors (to about 60%), in both U-V and V-J. This decrease could
be explained by galaxies with Hα emission below the detection limit in the reddest
part of the sample.
This is also seen in the color dependence of EW(Hα): in Figure 5.2 we populate
the UVJ diagram with the median log EW(Hα) in each 0.1×0.1 mag color-color bin,
for different mass bins. The EW(Hα) of SFGs spans a range from ∼ 20 to ∼ 200
(in agreement with Fumagalli et al. 2012). In each mass range, the EW(Hα) of
SFGs depends on the colors: galaxies with bluer colors (in both U-V and V-J) have
a higher EW(Hα) than that of red objects. In the mass ranges with the largest color
spread, the difference in EW(Hα) from the bluest to the reddest galaxies amounts to
Decreasing Hα for redder star-forming galaxies
77
a factor of 10. In contrast, the median EW(Hα) of QGs is low, generally lower than
a few Angstroms, in accordance with Fumagalli et al. (2015), who stacked massive
(logM∗ /M > 10.8) QGs, subtracted the best fit stellar continuum and evaluated
the residual Hα emission.
5.3.2
Color dependence of EW(Hα) for star-forming galaxies
We next investigate how the EW(Hα) varies as a function of color, for star-forming
galaxies. Figure 5.3 shows, for SFGs only, the U-V color dependence of EW(Hα), in
two redshift bins and four mass bins. Blue SFGs have on average higher EW(Hα)
than red SFGs. We quantify the relation between EW(Hα) and U-V from a linear
least squares fitting to the data shown in Fig. 5.3:
logEW ( Hα) = (2.56 ± 0.06) − (0.47 ± 0.05) × (U − V )
(5.2)
We see no significative dependence in the fit by dividing the sample in two redshift
bins (Figure 5.3). At 0.7 < z < 1.0, the best fit values of the EW ( Hα) = a + b × (U −
V ) equation are a = 2.47 ± 0.09 and b = −0.49 ± 0.08, while at 1.0 < z < 1.5 they
are a = 2.49 ± 0.07 and b = −0.43 ± 0.07. We find that the biweight scatter at fixed
color is approximately constant (0.25 dex).
Galaxies with a higher mass tend to have lower EW(Hα) than that of lower mass
galaxies (see also Figure 5.2), in agreement with Fumagalli et al. (2012). We therefore evaluate if the trend with color might be only a consequence of the mass trend
by examining narrow mass bins, in Figure 5.3 (right). We observe the same trend
between color and EW(Hα) in 0.5 dex wide mass bins. The median trends in narrow
mass bins fall perfectly on top of the one observed for the entire sample and described in Equation 5.2. In other words, galaxies with different masses but the same
color have on average the same EW(Hα).
The observed trend with EW(Hα) can be explained by either changes in the SFH
or differences in the extinction. In the following Sections we investigate if the variation in the EW(Hα) can be due to differences in the sSFR (Section 5.4) or to differences in the dust extinction (Section 5.5).
The sequence of star-forming galaxies in the UVJ diagram has been generally
linked to mainly variations in the dust absorption (Williams et al. 2009, Patel et al.
2012, Kriek et al. 2015). This study will test this hypothesis by analyzing differences
between red and blue SFGs with independent Hα data, and investigating the link
between variations in the EW(Hα) and those in sSFR and dust extinction.
5.4
Specific Star Formation Rates of star-forming galaxies: trend with color
Since EW(Hα) is defined as the ratio of the Hα flux to that of the underlying stellar
continuum, it represents a measure of the the current to past average star formation.
Chapter 5
78
3.0
3.0
0.7 < z < 1.1
0.7 < z < 1.1
1.0 All
0.5 9.0 < log M / MO• < 9.5
3.0
0.7 < z < 1.1
2.5
All
9.5 < log M / MO• < 10.0
2.5
0.7 < z < 1.1
2.5
2.0
1.5
2.0
1σ
1.5
0.7 < z < 1.1
2.0
9.0 < log M / MO• < 11.5
Median
Best Fit
1.5
0.5
3.0
1.1 < z < 1.5
2.5
log EW(Hα)
log EW(Hα)
1.0
1.0 All
All
0.5 10.0 < log M / MO• < 10.5 10.5 < log M / MO• < 11.0
3.0
1.1 < z < 1.5
1.1 < z < 1.5
2.5
2.0
1.5
2.0
1.0 All
0.5 9.0 < log M / MO• < 9.5
3.0
1.1 < z < 1.5
2.5
1σ
1.5
All
9.5 < log M / MO• < 10.0
1.1 < z < 1.5
2.0
1.0
9.0 < log M / MO• < 11.5
Median
Best Fit
1.5
1.0 All
All
0.5 10.0 < log M / MO• < 10.5 10.5 < log M / MO• < 11.0
0.5
0.0
0.5
1.0
1.5
U-V
2.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
U-V
Figure 5.3: Rest Frame EW(Hα) vs U-V for star-forming galaxies. Left: data are divided in
two redshift bins (top and bottom rows). Red star-forming galaxies have lower EW(Hα) than
that of blue galaxies. Blue solid lines represent the median values through the samples, while
blue dotted lines show the best fit from a linear least-squares minimization. Red dotted lines
represent the median 1σ errors through the samples. Right: star-forming galaxies are divided
into 0.5 dex wide mass bins, showing that the EW(Hα)-color trend holds also in narrow
mass bins. Solid (purple, red, green, orange) lines represent the median values through the
samples, and always lie on top of the median for all galaxies in the same redshift bin (blue).
Star-forming galaxies with different mass but same color have on average the same EW(Hα).
It is related to the sSFR, assuming no extra dust absorption towards HII regions:
EW ( Hα) =
L( Hα)
K −1 × SFR
M
∼
= sSFR ×
× K −1
L( R)
L
(
R
)
Lλ (6563Å)
M× M
(5.3)
where K is the Kennicutt (1998) conversion factor from Hα luminosity to SFR
and L(MR) is the mass-to-light ratio in the R-band.
In Figure 5.4 (left) we show the dependence of sSFR on the U-V color of SFGs.
We notice again that that blue galaxies have on average higher star formation rates
than redder galaxies. We evaluate the decline of sSFR with color as a factor of 4 (0.6
dex) per magnitude of U-V color. The scatter in sSFR is large (0.45 dex). After we
subtract the trend with color, it is reduced to 0.3 dex. At fixed color, the scatter in
sSFR increases towards redder colors: from σ = 0.2 dex at U-V∼0.5 to σ = 0.4 dex
at U-V∼1.5 .
Decreasing Hα for redder star-forming galaxies
79
-8.0
-8.0
0.7 < z < 1.1
0.7 < z < 1.1
-8.5
0.7 < z < 1.1
-8.5
-9.0
-9.0
-10.0
-9.5
-10.5
-11.0
-8.0
-9.5
All
9.0 < log M / MO• < 9.5
0.7 < z < 1.1
All
9.5 < log M / MO• < 10.0
0.7 < z < 1.1
-8.5
-10.0
-9.0
-9.5
-10.5
-11.0
-8.0
1.1 < z < 1.5
-8.5
log (sSFR / y-1)
log (sSFR / y-1)
9.0 < log M / MO• < 11.5
-10.0
-10.5
-11.0
-8.0
All
All
10.0 < log M / MO• < 10.5 10.5 < log M / MO• < 11.0
1.1 < z < 1.5
1.1 < z < 1.5
-8.5
-9.0
-9.5
-9.0
-10.0
-9.5
-10.5
-11.0
-8.0
All
9.0 < log M / MO• < 9.5
1.1 < z < 1.5
All
9.5 < log M / MO• < 10.0
1.1 < z < 1.5
-8.5
-10.0
-9.0
-9.5
9.0 < log M / MO• < 11.5
-10.5
-10.0
-11.0
0.0
All
-10.5 All
10.0 < log M / MO• < 10.5 10.5 < log M / MO• < 11.0
-11.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
U-V
0.5
1.0
1.5
U-V
2.0
Figure 5.4: Rest Frame sSFR (UV+IR) vs U-V for star-forming galaxies in the 3D-HST catalog.
Left: data are divided in two redshift bins (top and bottom rows). Solid blue lines represent
median values through the samples in each redshift bin. Red star-forming galaxies have
lower sSFR than that of blue star-forming objects. Right: star-forming galaxies are divided
into 0.5 dex wide mass bins, showing that the relation holds true also in narrower mass bins.
Solid (purple, red, green, orange) lines represent the median values through the samples, and
always lie on top of the median for all galaxies in the same redshift bin (blue). Star-forming
galaxies with different mass but same color have on average the same sSFR.
We notice that recent studies (Utomo et al. 2014, Hayward et al. 2014, Fumagalli
et al. 2014) have indicated that SFR(UV+IR) systematically overestimates the SFR
for low sSFRs, as cirrus dust heated by evolved stellar populations significantly
contributes to the mid-IR emission. Utomo et al. (2014) evaluates that sSFR(IR+UV)
is a reliable measure of sSFR only for sSFR > 10−10 yr−1 , and it overestimates sSFR
by a factor of 0.5 at sSFR ∼ 10−10.5 yr−1 . All but a fraction of the reddest galaxies
(U-V > 1.7) in our sample have sSFR > 10−10 yr−1 . If the sSFRs of some of the
reddest objects are overestimated, the decline of sSFR with color is possibly even
steeper than 0.6 dex per magnitude of U-V color.
It is moreover known that masses and SFRs are correlated, with most authors
finding a non-linear relation for the ’star-forming main sequence’ SFR ∼ Mα with α
ranging from 0.6 to 0.9 (e.g. Whitaker et al. 2010, Karim et al. 2012), resulting in a
negative correlation between M and sSFR. In Figure 5.4 (right) we show, for 0.5 dex
wide mass bins, the relation between color and sSFR: we find that also in narrow
Chapter 5
80
mass bins redder galaxies have on average lower sSFR than bluer galaxies, and that
the scatter in the relations increases at the reddest colors. At a fixed color, SFGs of
different masses have on average the same sSFR.
The relation between sSFR and EW(Hα) depends on the assumed SFH of the
galaxy, as shown by the dependence of Equation 5.3 on the mass-to-light ratio, which
is different for different SFHs.
We evaluate the relation between sSFR and EW(Hα) from the models as follows:
for a set of SFHs (constant SFR, and exponentially declining models with τ = 0.3,
1.0, and 3.0 Gyr), L(Hα) is derived from the Kennicutt (1998) law, while BC03 models are used to calculate the stellar continuum, assuming solar metallicity and no
dust. As the sSFR in the models decreases, EW(Hα) gets smaller, with a different
steepness in the EW(Hα)-sSFR relation according to the model in use. For a CSF, a
variation of ∆sSFR = 1 dex corresponds to a factor of 2 (0.3 dex) in ∆EW(Hα). For
an exponentially declining model with a short τ=300 Myr, a variation of ∆sSFR =1
dex corresponds to a factor of 5 (0.7 dex) in ∆EW(Hα).
We measure the observed variation in sSFR as ∆sSFR = 0.67 dex per magnitude
of U-V color (Figure 5.4). Based on variations in sSFR alone, we predict the EW(Hα)
to vary of a factor of ∼ 1.4-2.4 (0.15-0.38 dex), respectively under the assumption of
a CSF or a τ=300 Myr model.
The observed decline of EW(Hα) with color amounts to a factor of 3 per magnitude of U-V color (0.5 dex, Figure 5.2 and Equation 5.2). We therefore conclude that
the variation in sSFR alone is unable to explain the full variation of EW(Hα).
5.5
5.5.1
Dust absorption of star-forming galaxies along the
UVJ diagram
Absorption in Hα
We quantify the dust absorption in Hα by comparing the Hα luminosity2 to the
SFR (inferred from UV+IR, Section 2). In the absence of dust, we expect the two
quantities to follow the Kennicutt (1998) relation:
SFR[ M /yr ] = 7.9 × 10−42 × L( Hα)[erg/s] × 10−0.2
(5.4)
where the 10−0.2 factor accounts for a conversion to the Chabrier IMF, from a
Salpeter IMF (as in Marchesini et al. 2009).
In Figure 5.5 we plot the ratio of L(Hα) to SFR against the masses of galaxies. At
each mass, the ratio is lower than the expectation from the Kennicutt law (shown in
green). The ratio decreases towards higher masses, with a decrease of ∼ 0.4 dex per
each dex in mass, as quantified by a least-square fit to the data:
η = log
L( Hα)
= (45.1 ± 0.2) − (0.41 ± 0.02) × log( M∗ /M )
SFR
(5.5)
2 We assume that [N ] contributes to 15% of the measured Hα emission line flux, as in Wuyts et al.
II
2013, such that the [NII ]/(Hff + [NII ]) ratio equals 0.15.
Decreasing Hα for redder star-forming galaxies
43
81
0.7 < z < 1.1
Best Fit
1.1 < z < 1.5
Best Fit
K98
K98
41
40
0.0
GB10
GB10
1σ
1σ
1.0
A(Hα)
log L(Hα) / SFR (UV+IR)
42
2.0
3.0
4.0
39
38
9
10
log Mass
11
9
10
log Mass
11
Figure 5.5: Ratio of Hα luminosities to SFR as a function of mass for star-forming galaxies,
in two different redshifts. Galaxies with higher mass have lower L(Hα)/SFR than that of
low mass objects, indicating a higher absorption in the Hα line. The expected value from the
Kennicutt (1998) law is shown in green. The right axis shows the effect of different magnitudes
of absorption in Hα. The Garn & Best relation between mass and dust obscuration at z=0 is
shown in orange, while the best fit to the 3D-HST data is shown in blue.
indicating that SFGs with a higher mass have higher dust obscuration towards Hα
than that of lower mass objects. We do not see any significant redshift dependence
in the relation: after we split the sample in two redshift bins (Figure 5.5) and fit them
with a linear model, we find slopes and y-intercept values compatible at the 1σ level
with those in Equation 5.5.
This trend is consistent with previous studies finding that A(Hα) increases towards higher stellar masses, both in the local Universe (Garn & Best 2010) and at
higher redshift (Sobral et al. 2012, Kashino et al. 2013, Price et al. 2014), with no
significative redshift dependence in the A(Hα)-mass relation. We show the effect
of the Garn & Best (2010) relation for galaxies at z ∼ 0.1 onto our measurements
of L(Hα)/SFR at 0.7 < z < 1.5 in Figure 5.5 (orange lines), demonstrating a good
agreement with our data.
We further investigate the dependence of L(Hα) / SFR on color. In Figure 5.6
we show how the quantity varies in the UVJ diagram, in different mass bins. In
each mass bin, red SFGs have lower η than blue SFGs, indicating that Hα is more
absorbed in those systems.
In Figure 5.7 we show how η relates to the U-V color of the galaxy. We find that
on average galaxies with red (U-V=1.5) color have η ≈ 40.5, while for blue objects
(U-V = 0.7) η ≈ 41. In Figure 5.7 (right) we divide the sample in narrow mass-bins,
showing that the color-η trend persists also at fixed mass. We notice that galaxies
of the same color have, on average, the same L(Hα)/SFR ratio despite differences in
Chapter 5
82
log L(Hα) / SFR
9.5 < logM < 10.0
10.0 < logM < 10.5
2.0
1.5
41.50
40.88
40.25
39.62
39.00
1.0
U-V
0.5
41.50
40.88
40.25
39.62
39.00
10.5 < logM < 11.0
11.0 < logM < 11.5
2.0
1.5
41.50
40.88
40.25
39.62
39.00
1.0
0.5
0.5
1.0
1.5
2.0
41.50
40.88
40.25
39.62
39.00
V-J
0.5
1.0
1.5
2.0
Figure 5.6: UVJ diagram for 3D-HST galaxies at 0.7 < z < 1.5 in different mass bins, colorcoded by median ratio of Hα to SFR. The selection box divides quiescent and star-forming
galaxies. Red SFGs have lower L(Hα)/SFR ratios, indicating a higher dust extinction in Hα.
mass.
Since redder galaxies tend to be more massive than blue galaxies, we investigate
if the trend with color can be explained by the mass dependence only. We subtract
the trend of Equation 5.5 from the observed L(Hα) / SFR, and look at the residual
trend with color in Figure 5.8. We still observe a decrease of ≈ 0.5 dex in L(Hα) /
SFR per each dex in U-V color.
The influence of dust on EW(Hα) depends however on the relative absorption
between the stellar continuum and the nebular emission (Calzetti et al. 2000), which
is a matter of debate at high-redshift (Erb et al. 2006, Forster-Schreiber et al. 2010,
Reddy et al. 2010, Kashino et al. 2013, Price et al. 2014, Reddy et al. 2015). We
therefore need to also evaluate the attenuation in the stellar continuum.
Decreasing Hα for redder star-forming galaxies
83
0.7 < z < 1.1
0.7 < z < 1.1
41.5
0.7 < z < 1.1
41.5
41.0
K+98 (Chabrier IMF)
40.5
41.0
40.0
39.5
40.5
9.0 < log M / MO• < 9.5
9.5 < log M / MO• < 10.0
All
All
0.7 < z < 1.1
0.7 < z < 1.1
41.5
41.0
40.0
9.0 < log M / MO• < 11.5
1.1 < z < 1.5
41.5
K+98 (Chabrier IMF)
η = log10 L(Hα) / SFR
η = log10 L(Hα) / SFR
40.5
39.5
40.0
39.5
10.0 < log M / MO• < 10.5 10.5 < log M / MO• < 11.0
All
All
1.1 < z < 1.5
1.1 < z < 1.5
41.5
41.0
40.5
41.0
40.0
39.5
40.5
9.0 < log M / MO• < 9.5
9.5 < log M / MO• < 10.0
All
All
1.1 < z < 1.5
1.1 < z < 1.5
41.5
41.0
40.0
40.5
39.5
40.0
9.0 < log M / MO• < 11.5
39.5
0.0
0.5
1.0
1.5
2.0
U-V
10.0 < log M / MO• < 10.5 10.5 < log M / MO• < 11.0
All
All
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
U-V
Figure 5.7: Rest Frame U-V vs log L(Hα)/SFR for star-forming galaxies. Left: data are divided
in two redshift bins (top and bottom rows). Solid blue lines represent median values through
the samples in each redshift bin. Red star-forming galaxies have lower L(Hα)/SFR than that
of blue star-forming objects, indicating more dust absorption. Right: star-forming galaxies
are divided into 0.5 dex wide mass bins, showing that the relation holds true also in narrower
mass bins. Solid (purple, red, green, orange) lines represent the median values through
the samples, and always lie on top of the median for all galaxies in the same redshift bin
(blue). Star-forming galaxies with different mass but same color have on average the same
L(Hα)/SFR
5.5.2
Absorption of the continuum
The dust obscuration of the continuum can be estimated from the ratio of infrared
emission to ultraviolet emission, defined as IRX = log L(IR) / L(UV). In Figure 5.9,
we show how IRX varies in the U-V and V-J color-color space: the UVJ diagram
is divided in 0.1×0.1 bins and color-coded by median IRX of the galaxies in each
color-color bin. For QGs, IRX is relatively low. SFGs span instead a wide range of
IRX values, from ∼ 0.3 to ∼ 2.5. Redder galaxies have on average higher IRX than
bluer galaxies (in accordance to Whitaker et al. 2012), which implies that they have
higher dust absorption in the continuum. Comparing different mass bins with each
other, we notice that IRX values of galaxies in the same region of the UVJ diagram
have similar IRX values in spite of the different masses.
The IRX essentially measures the attenuation in the UV. By assuming a stellar
Chapter 5
84
1.0
0.7 < z < 1.1
0.7 < z < 1.1
0.5
0.0
-0.5
log L(Hα) / SFR (subtracting mass dependence)
9.0 < log M / MO• < 9.5
-1.0
1.0
9.5 < logM < 10.0
2.0
1.5
0.50
0.25
0.00
-0.25
-0.50
1.0
U-V
0.5
10.0 < logM < 10.5
0.50
0.25
0.00
-0.25
-0.50
10.5 < logM < 11.0
2.0
log (L(Hα) / SFR) - log (LHa / SFR)Mass
9.0 < logM < 9.5
0.7 < z < 1.1
0.7 < z < 1.1
0.5
0.0
-0.5
10.0 < log M / MO• < 10.5 10.5 < log M / MO• < 11.0
-1.0
1.0
1.1 < z < 1.5
1.1 < z < 1.5
0.5
0.0
-0.5
9.0 < log M / MO• < 9.5
-1.0
1.0
1.5
9.5 < log M / MO• < 10.0
1.1 < z < 1.5
9.5 < log M / MO• < 10.0
1.1 < z < 1.5
0.5
0.50
0.25
0.00
-0.25
-0.50
1.0
0.5
0.0
0.5
1.0
1.5
2.0 0.0
V-J
0.50
0.25
0.00
-0.25
-0.50
0.5
1.0
1.5
2.0
0.0
-0.5
10.0 < log M / MO• < 10.5 10.5 < log M / MO• < 11.0
-1.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
U-V
Figure 5.8: Ratio of L(Hα) to SFR, after its mass dependence (Figure 5.5 and Equation 5.5)
has been removed. Left: UVJ diagram color coded by the residual trend in L(Hα)/SFR (L(Hα)/SFR )Mass . Right: U-V color versus L(Hα)/SFR - (L(Hα)/SFR )Mass . At the same mass,
redder sources have a lower ratio of Hα to SFR than bluer galaxies.
population model and an attenuation curve, we can estimate the implied attenuation
of the stellar continuum around Hα.We assume a 100 Myr old stellar population
from the BC03 stellar population models, with constant SFR and a Chabrier (2003)
IMF, and we redden its spectrum assuming a Calzetti (2000) dust law, and varying
A(V) from 0 to 6, with a 0.1 step. For each value of dust reddening, we compute
the flux absorbed from 0 to 1 µm and assume it is reemitted in the infrared (as
in Charlot & Fall 2000)3 . This provides an estimate of LIR . We measure in the
models the luminosity at 2800 Å and convert it to L(UV) as in Bell et al. (2005):
L(UV) = 1.5 ∗ L(2800). We finally compute the expected IRX = log L(IR)/L(UV) for
each value of A(6563).
For each galaxy in the sample we estimate the absorption in the continuum at
6563Å (A6563 ) from it IRX value, and compare it to its absorption towards the Hα
line, evaluated in the previous Section via the L(Hα)/SFR ratio (Figure 5.10). We
observe that the Hα emission line is more absorbed than its underlying continuum.
We find A Hα = (1.63 ± 0.14)× A6563 , in agreement with Price et al. (2014), who evaluated the absorption around star-forming regions of 3D-HST galaxies at z ∼ 1.4 via
the Balmer decrement and the continuum AV via SED fitting, and reported a value
of 1.86 ± 0.40. This evaluation depends on the assumption that dust obscuration of
3 As
the model galaxy is young, the absorbed light is dominated by that of young, blue, stars.
Decreasing Hα for redder star-forming galaxies
85
IRX = log L(IR) / L(UV)
9.5 < logM < 10.0
10.0 < logM < 10.5
2.0
1.5
0.00
0.62
1.25
1.88
2.50
1.0
U-V
0.5
0.00
0.62
1.25
1.88
2.50
10.5 < logM < 11.0
11.0 < logM < 11.5
2.0
1.5
0.00
0.62
1.25
1.88
2.50
1.0
0.5
0.5
1.0
1.5
2.0
0.00
0.62
1.25
1.88
2.50
V-J
0.5
1.0
1.5
2.0
Figure 5.9: UVJ diagram for 3D-HST galaxies at 0.7 < z < 1.5 in different mass bins, colorcoded by median IRX (logLIR /LUV ). The selection box divides quiescent and star-forming
galaxies. Red SFGs have a higher dust attenuation than blue SFGs. The areas of squares are
proportional to the number of galaxies in that 0.1x0.1 mag color-color bin.
high-redshift galaxies follows a Calzetti law, which is currently matter of debate (see
Kriek & Conroy 2013, Reddy et al. 2015).
The relation between the stellar continuum and nebular attenuation has consequences on EW(Hα). If continuum and emission lines were extinguished by the
same amount of light (as in Reddy et al. 2010 or Erb et al. 2006) EW(Hα) would be
fundamentally a dust-free measurement. However, the present study (in agreement
with Kashino et al. 2013, Price et al. 2014) shows that EW(Hα) retains a mild dependence on dust, with 1 mag of extinction in the continuum originating 0.6 mag of
extinction for EW(Hα).
Given the estimates of dust absorption in the continuum and in the Hα line, we
can finally evaluate the effect of dust onto the EW(Hα). We estimate that, per each
magnitude in U-V color, Hα gets on average 2.5 times more absorbed (0.4 dex, Figure
Chapter 5
86
5.7). Using the differential extinction between emission lines and continuum computed in this Section, ∆L(Hα) = 0.4 dex translates into a variation of ∆EW(Hα) = 0.15
dex per magnitude in U-V color; this is the variation of EW(Hα) as a consequence of
dust absorption only.
We check whether the full range of EW(Hα), which amounts to a decline of 0.47
dex per each magnitude in U-V (Section 5.3), can be explained by the contributions
of different dust absorption and different sSFRs (0.15-0.38 dex per magnitude in UV as computed in Section 5.4). The impact of ∆sSFR onto ∆EW(Hα), even though
model dependent, tends to be higher than that of different dust attenuations.
We find that the observed ∆EW(Hα) can be reproduced by summing the effect of
dust (0.15 dex) to that of sSFR at the highest end of its determination (∼ 0.35 dex),
i.e. by assuming exponentially declining SFHs with short τ (∼ 300 Myr). This is
justified since, from best fits to the photometry (Section 2), we obtain τ values with
a median value of 400 Myr4 , with no significant trend with mass.
In summary, we show that the variation of EW(Hα) with color can be explained
by the combined effect of different dust extinctions and different sSFRs, under the
assumption that SFGs have SFHs with short τ.
4
A(Hα)
3
2
1
1-1
best fit
0
0
1
2
A(6563A, cont)
3
4
Figure 5.10: For each star-forming galaxy in the sample, relation between absorption in the
continuum at 6563 Å (as computed in Section 5.5) and that for the Hα emission line. The
Hα emission line is more absorbed than the light in the underlying continuum. The red line
represents the best fit to the data computed with a least-square minimization, corresponding
to a slope of 1.63.
4 The
8.3.
mean value for τ is 1 Gyr. The mean and median values for log(τ/yr) are respectively 8.6 and
Decreasing Hα for redder star-forming galaxies
5.6
87
Conclusions
We have analyzed a large sample of 7000 galaxies at redshift 0.7 < z < 1.5 from
the 3D-HST survey, divided in star-forming and quiescent on the basis of the widely
used UVJ rest-frame color-color criterion. We investigate how the Hα emission varies
as a function of the rest-frame colors of galaxies, and investigate the origin of its
variation. The main conclusions of the study are the following:
• We confirm the effectivenes of the UVJ selection technique by demonstrating
that galaxies with strong and weak Hα are well separated in the UVJ diagram:
Hα is detected (S/N > 3) in ∼ 85% of the UVJ-selected star-forming galaxies,
while less than 20% of the UVJ-selected quiescent galaxies have an Hα detection. While the EW(Hα) of star forming galaxies spans a range from ∼ 20 to ∼
200, the median EW(Hα) of quiescent galaxies is lower than 5 Å.
• For star forming galaxies, EW(Hα) has a tight relation with the U-V color of
the galaxy. The average EW(Hα) of red star-forming galaxies is 0.7 dex lower
than that of blue star-forming galaxies, at a fixed mass. The trend persists in
narrow mass bins, and galaxies of different masses but the same U-V color
have on average the same EW(Hα).
• At a fixed mass, the median IRX ratio (log LIR /LUV ) increases towards redder
U-V colors (in accordance with Whitaker et al. 2012), while the median sSFR
(UV+IR) of star-forming galaxies decreases towards redder colors.
• The ratio L(Hα) / SFR of star-forming galaxies varies as a function of both U-V
color and stellar mass, indicating higher A(Hα) attenuation for the most massive and for the reddest galaxies. The systematic variation is approximately 1
dex with stellar mass, with an additional variation of 0.5 dex with color after
the mass dependence has been removed.
• The trend of EW(Hα) with U-V color is explained by a combination of both
a lower specific star-formation rate and a higher dust absorption in red starforming galaxies, compared to those of blue star-forming galaxies.
• By determining the extinction in Hα and that in the continuum (under the
assumption of a Calzetti dust law) we confirm previous studies indicating that
the Hα emission line is more absorbed than its underlying continuum.
We acknowledge funding from ERC grant HIGHZ no. 227749. This work is based on
observations taken by the 3D-HST Treasury Program (GO 12177 and 12328) with the
NASA/ESA HST, which is operated by the Association of Universities for Research
in Astronomy, Inc., under NASA contract NAS5-26555.
88
Chapter 5
Bibliography
Bell, E. F., Papovich, C., Wolf, C., et al. 2005, ApJ, 625, 23
Bell, E. F., van der Wel, A., Papovich, C., et al. 2012, ApJ, 753, 167
Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008, ApJ, 686, 1503
Brammer, G. B., Whitaker, K. E., van Dokkum, P. G., et al. 2009, ApJL, 706, L173
Brammer, G. B., Whitaker, K. E., van Dokkum, P. G., et al. 2011, ApJ, 739, 24
Brammer, G. B., van Dokkum, P. G., Franx, M., et al. 2012, ApJS, 200, 13
Brinchmann, J., Charlot, S., White, S. D. M., et al. 2004, MNRAS, 351, 1151
Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000
Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682
Chabrier, G. 2003, ApJL, 586, L133
Charlot, S., & Fall, S. M. 2000, ApJ, 539, 718
Daddi, E., Cimatti, A., Renzini, A., et al. 2004, ApJ, 617, 746
Damen, M., Förster Schreiber, N. M., Franx, M., et al. 2009, ApJ, 705, 617
Erb, D. K., Shapley, A. E., Pettini, M., et al. 2006, ApJ, 644, 813
Faber, S. M., Willmer, C. N. A., Wolf, C., et al. 2007, ApJ, 665, 265
Franx, M., van Dokkum, P. G., Schreiber, N. M. F., et al. 2008, ApJ, 688, 770
Fumagalli, M., Patel, S. G., Franx, M., et al. 2012, ApJL, 757, L22
Fumagalli, M., Labbé, I., Patel, S. G., et al. 2014, ApJ, 796, 35
Garn, T., & Best, P. N. 2010, MNRAS, 409, 421
Geach, J. E., Smail, I., Best, P. N., et al. 2008, MNRAS, 388, 1473
Gobat, R., Strazzullo, V., Daddi, E., et al. 2013, arXiv:1305.3576
Gordon, K. D., Clayton, G. C., Witt, A. N., & Misselt, K. A. 2000, ApJ, 533, 236
Hayward, C. C., Lanz, L., Ashby, M. L. N., et al. 2014, MNRAS, 445, 1598
Karim, A., Schinnerer, E., Martínez-Sansigre, A., et al. 2011, ApJ, 730, 61
Kashino, D., Silverman, J. D., Rodighiero, G., et al. 2013, ApJL, 777, LL8
Kennicutt, R. C., Jr. 1998, ARA&A, 36, 189
Kriek, M., van Dokkum, P. G., Labbé, I., et al. 2009, ApJ, 700, 221
Decreasing Hα for redder star-forming galaxies
89
Kriek, M., & Conroy, C. 2013, ApJL, 775, L16
Kriek, M., Shapley, A. E., Reddy, N. A., et al. 2015, ApJS, 218, 15
Labbé, I., Huang, J., Franx, M., et al. 2005, ApJL, 624, L81
Marchesini, D., van Dokkum, P. G., Förster Schreiber, N. M., et al. 2009, ApJ, 701,
1765
Maller, A. H., Berlind, A. A., Blanton, M. R., & Hogg, D. W. 2009, ApJ, 691, 394
Mendel, J. T., Saglia, R. P., Bender, R., et al. 2015, ApJL, 804, L4
Muzzin, A., van Dokkum, P., Kriek, M., et al. 2010, ApJ, 725, 742
Muzzin, A., Marchesini, D., Stefanon, M., et al. 2013, ApJ, 777, 18
Nelson, E. J., van Dokkum, P. G., Brammer, G., et al. 2012, ApJL, 747, L28
Nelson, E. J., van Dokkum, P. G., Momcheva, I., et al. 2013, ApJL, 763, L16
Noeske, K. G., Weiner, B. J., Faber, S. M., et al. 2007, ApJL, 660, L43
Patel, S. G., Holden, B. P., Kelson, D. D., et al. 2012, ApJL, 748, L27
Price, S. H., Kriek, M., Brammer, G. B., et al. 2014, ApJ, 788, 86
Reddy, N. A., Erb, D. K., Pettini, M., Steidel, C. C., & Shapley, A. E. 2010, ApJ, 712,
1070
Reddy, N. A., Kriek, M., Shapley, A. E., et al. 2015, arXiv:1504.02782
Shim, H., Chary, R.-R., Dickinson, M., et al. 2011, ApJ, 738, 69
Smit, R., Bouwens, R. J., Franx, M., et al. 2012, ApJ, 756, 14
Sobral, D., Best, P. N., Geach, J. E., et al. 2010, MNRAS, 404, 1551
Speagle, J. S., Steinhardt, C. L., Capak, P. L., & Silverman, J. D. 2014, ApJS, 214, 15
Stark, D. P., Schenker, M. A., Ellis, R., et al. 2013, ApJ, 763, 129
Straatman, C. M. S., Labbé, I., Spitler, L. R., et al. 2014, ApJL, 783, L14
Utomo, D., Kriek, M., Labbé, I., Conroy, C., & Fumagalli, M. 2014, ApJL, 783, L30
van Dokkum, P. G., Brammer, G., Fumagalli, M., et al. 2011, ApJL, 743, L15
Whitaker, K. E., van Dokkum, P. G., Brammer, G., & Franx, M. 2012, ApJL, 754, L29
Whitaker, K. E., van Dokkum, P. G., Brammer, G., et al. 2013, ApJL, 770, L39
Whitaker, K. E., Franx, M., Leja, J., et al. 2014, ApJ, 795, 104
Williams, R. J., Quadri, R. F., Franx, M., van Dokkum, P., & Labbé, I. 2009, ApJ, 691,
1879
Wuyts, S., Labbé, I., Schreiber, N. M. F., et al. 2008, ApJ, 682, 985
Wuyts, S., Labbé, I., Franx, M., et al. 2007, ApJ, 655, 51
90
Chapter 5
Samenvatting
Aan het begin van de negentiende eeuw, was het nog onduidelijk of de Melkweg,
de zwakke streep van sterren die zichtbaar is aan de hemel op een donkere, heldere
nacht, het enige sterrenstelsel in het heelal is. Op 26 april 1920, bediscussieerden
twee invloedrijke astronomen van de tijd, Harlow Shapley en Heber Curtis, hun
tegengestelde standpunten voor een publiek in het ’Smithsonian Museum of Natural
History’ in Washington DC. Shapley betoogde dat de Melkweg het gehele heelal is
en dat spiraalnevels deel uit maken van ons eigen melkwegstelsel. Aan de andere
kant dacht Curtis dat Andromeda en de andere nevels afzonderlijke sterrenstelsels
waren oftewel eiland universums (zoals Immanuel Kant ze genoemd heeft honderd
jaar eerder).
De standpunten van beide geleerden werden gesteund door verschillende waarnemingen die destijds beschikbaar waren. Echter, de belangrijkste observationele ondersteuning van Shapley’s theorie bleek al snel onjuist te zijn. Tegelijkertijd bleek
uit nieuwe waarnemingen van Edwin Hubble dat het Andromeda sterrenstelsel een
apart eiland universum was, ver buiten de Melkweg. Een paar jaar later in 1927, was
het weer Hubble die een ruwe evenredigheid vond tussen de afstand tot sterrenstelsels en hun terugwijkende snelheid: hierna begonnen astronomen te beseffen dat
het heelal uitdijt. De waarnemingen van Edwin Hubble markeerden het begin van
de moderne observationele kosmologie; het is niet bij toeval geweest dat de meest
ambitieuze ruimtetelescoop, die in een baan om de aarde zit, vernoemd is naar hem.
De opmerkelijke diversiteit van sterrenstelsels
Sinds de vroegste waarnemingen is er een grote diversiteit in de morfologie van sterrenstelsels geconstateerd. Het meest gebruikte indelingssysteem vandaag de dag, is
voorgesteld door (wederom) Edwin Hubble. Hubble merkte op dat sterrenstelsels
ruwweg kunnen worden ingedeeld in twee klassen: elliptische stelsels bestaande
uit een ronde of elliptische verdeling van licht, en spiraalstelsels die bestaan uit een
platte schijf met spiraalvormige structuur en in het centrum een concentratie van
licht (beter bekend als de bulge).
Daaropvolgende studies hebben aangetoond dat de morfologie van huidige stelsels
nauw gecorreleerd is met andere eigenschappen, zoals de massa, de kleur en de
omgeving van de sterrenstelsels. In het algemeen hebben elliptische sterrenstelsels
rodere kleuren dan spiraal stelsels, omdat hun licht wordt gedomineerd door rode
oude sterren. Aan de andere kant worden in spiraal stelsels actief nieuwe, jonge sterren gevormd, en daarom lijken ze blauw. Zeer weinig sterrenstelsels liggen tussen
die twee categorieën.
Een soortgelijke bimodaliteit in kleur, stervormingssnelheid, en morfologie van
sterrenstelsels is ook waargenomen als men terug kijkt in de tijd naar het vroege
heelal (m.a.w. bij hoge roodverschuiving).
91
92
Samenvatting
Voor sterrenstelsels met veel stervorming, zijn de massa en de stervormingssnelheid (d.w.z. het aantal nieuwe sterren gevormd in een jaar) strak gecorreleerd. Meer
massieve sterrenstelsels vormen meer sterren per jaar. Als we terug kijken in de tijd,
zien we dezelfde relatie voor sterrenstelsels op hoge roodverschuiving. De absolute
stervormingssnelheid neemt toe met roodverschuiving: d.w.z. sterrenstelsels van
dezelfde massa hadden in het verleden een hogere stervormingssnelheid.
Problemen bij waarnemingen op hoge
roodverschuiving
Ondanks de onvoorstelbare technologische vooruitgang van telescopen en instrumenten in de laatste twintig jaar, zijn metingen van sterrenstelsels op hoge roodverschuiving nog steeds erg uitdagend. Onze kennis van deze stelsels komt niet
in de buurt van onze kennis van de lokale melkwegstelsels. Ten eerste komt dit
doordat het licht van sterrenstelsels zwakker wordt naar mate de afstand tot het
stelsel toeneemt. Om de fysische parameters van sterrenstelsels nauwkeuriger te
meten, verdelen astronomen het licht in meerdere frequenties met een spectrograaf.
Een spectrograaf kun je vergelijken met een prisma dat wit licht splitst in verschillende kleuren. Met uitzondering van de helderste bronnen, is spectroscopie van
verre, zwakke, sterrenstelsels een buitengewoon tijdrovend proces. Het meest fundamentele probleem als de spectroscopische methode niet toegepast kan worden, is
dat afstandsbepalingen erg onzeker worden.
Een tweede fundamentele probleem voor waarnemingen op hoge roodverschuiving is dat het licht sterk roodverschoven is door de uitdijing van het heelal. Dit
fenomeen is verwant aan de frequentie van geluid dat naar een lagere toonhoogte
gaat als de bron van ons af beweegt. Licht van oude sterren (die het merendeel
van de totale massa beslaan in de meeste sterrenstelsels) wordt verplaatst naar het
infrarood voor sterrenstelsels die op een afstand staan van meer dan een paar miljard lichtjaar. Op die golflengtes is de atmosfeer niet transparant. Waarnemingen
met telescopen vanaf de grond (in plaats van telescopen op satellieten) zijn alleen
mogelijk in bepaalde bereiken van golflengtes.
Onze kennis van leeftijden, stervormingssnelheden en metaalgehaltes van sterrenstelsels is vaak gebaseerd op spectrale indicatoren uit optische golflengtes. Het
is uitdagend en tijdrovend om deze te meten op hoge roodverschuiving wanneer ze
verschoven zijn naar het infrarood. Bijvoorbeeld, een goed gekalibreerde standaard
indicator van de stervormingssnelheid is de waterstof emissie lijn: alleen jonge sterren hebben licht met genoeg energie om een belangrijke bijdrage leveren aan de
ionisatie van het waterstof in hun directe omgeving: waterstof lijnen bieden een bijna directe meting van de stervormingssnelheid. Als gevolg van de verschuiving van
het licht naar het (nabije) infrarood voor stelsels die ouder zijn dan 8 miljard jaar,
maken studies van de evolutie van stervormingssnelheden met een breed roodverschuivingsbereik gebruik van een reeks van indicatoren voor de stervormingssnelheid die gebaseerd zijn op een verschillende veronderstellingen en inter-kalibraties.
Voor elke indicator, een flux die overeenkomt met een stervormingssnelheid van
Samenvatting
93
10-20 zonsmassa’s per jaar is een uitdaging, zo niet onmogelijk voor afzonderlijke
bronnen.
De identificatie van collecties van sterrenstelsels met lage stervormingssnelheden
op hoge roodverschuiving is daarom in het algemeen uitsluitend gebaseerd op de
kleuren van het stelsel. Dit wordt gedaan door het selecteren van sterrenstelsels
waarvan het optische en nabij-infrarode licht wordt gedomineerd door een oude
sterren.
Een extra complicatie voor spectroscopie van de grond is dat spectroscopische
collecties geoptimaliseerd zijn voor waarnemingen in specifieke bandbreedtes. Over
het algemeen bestaan ze daarom slechts uit blauwe objecten die veel stervorming
bevatten. Aan de andere kant kunnen waarnemingen die niet gebaseerd zijn op
emissie lijnen, maar op het continuum van het licht, slechts toegepast worden op
kleine collecties van heldere objecten. Deze biases beperken ons kennis van de fysische eigenschappen van sterrenstelsels zoals leeftijd en metaalgehalte. Bovendien
beperken ze ons begrip van de vormingsgeschiedenis en de evolutie van sterrenstelsels.
Dit proefschrift
Dit proefschrift richt zich op een aantal van de eerder beschreven problemen. In
het bijzonder maken wij gebruik van een nieuwe reeks waarnemingen met de Wide
Field Camera 3 (WFC3) grism (tralie prisma) aan boord van de Hubble Space Telescope (HST), in het kader van het 3D-HST onderzoek. Dit onderzoek richt zich op
de evolutie van de stervormingssnelheden, de sterke van emissie lijnen en eigenschappen van groepen sterren van zowel stervormings- als passieve sterrenstelsels,
in collecties van sterrenstelsels tot 10 miljard jaar geleden, toen de stervormingssnelheid van het heelal op zijn piek was en de structurele regelmaat die we vandaag
de dag zien in sterrenstelsels ontstaan moet zijn. Omdat de WFC3 grism gelijktijdig
alle objecten in een veld waarneemt, zijn de waarneminge niet beïnvloed door de biases voor spectroscopie vanaf de grond, en bieden ze een mogelijkheid om afstanden
te meten voor zowel stervormingsgebieden als passieve sterrenstelsels.
In hoofdstuk 2, combineren we de eerste gegevens van het 3D-HST onderzoek
met die van waarnemingen vanaf de grond op lagere roodverschuiving, om de evolutie van de helderste waterstof lijn (Hα) te bestuderen. In het bijzonder meten
we de equivalente breedte van de Hα-lijn (EW(Hα)). Dat is de verhouding van
de helderheid van Hα en de onderliggende stellaire helderheid. De equivalente
breedte is een maat voor de verhouding tussen de huidige stervormingssnelheid
en die in het verleden. We vinden dat op elk cosmisch tijdstip EW(Hα) afneemt
met de massa van het sterrenstelsel, en dat bij constante massa EW(Hα) toeneemt
met roodverschuiving (m.a.w. verder terug in de tijd). Deze evolutie is onafhankelijk van de massa in sterren, en het hangt sterker af van roodverschuiving dat tot
nu toe aangenomen werd en voorspeld werd door theoretische modellen van de
evolutie van sterrenstelsels. Daarnaast voorspellen we de evolutie van EW(Hα) op
94
Samenvatting
hoge roodverschuiving; bijdrage van emissie lijnen voor sterrenstelsels op roodverschuivingen van z = 4 − 8 blijft toenemen. Dit heeft belangrijke consequenties voor
spectroscopie en fotometrie van bronnen die waargenomen zullen worden met de
James Webb Space Telescope.
In hoofdstuk 3, onderzoeken we de stervormingssnelheden van passieve sterrenstelsels die geselecteerd zijn op basis van hun optische en nabij infrarode spectrale
energie verdeling, die wijzen op een oude stellaire populatie. Voorgaande metingen van de spectrale energie verdeling voor optisch geselecteerde passieve stelsels,
wijzen op nog lagere stervormingsnelheden dan verwacht op basis van het hergebruiken van gas (onder de aanname dat sterren gevormd worden uit massa verloren
door geëvolueerde sterren). Echter, deze meting mist stervorming als deze verborgen wordt door door extinctie door stof. Hierdoor wordt de optische straling niet
waargenomen, maar verschuift de straling naar het mid-infrarood. In hoofdstuk
3, selecteren we daarom spectroscopisch bevestigde passieve sterrenstelsels in het
3D-HST onderzoek, en meten hun extinctie door stof stervormingssnelheid door
gestapelde mid-infrarode fluxes van Spitzer-24µm voor verschillende cosmische
tijdstippen. We tonen aan dat, voor elk kosmisch tijdstip, de stervormingssnelheid
van passieve sterrenstelsels 20-40 keer lager ligt dan voor sterrenstelsels die nog actief sterren vormen op dezelfde roodverschuiving. Dit geeft aan dat het uitdoven
van stervorming erg efficient is, zelfs in het jonge heelal waar de typische stervormingssnelheden honderden zonsmassa’s per jaar bereiken, in vergelijking met
1-4 zonsmassa per jaar voor de Melkweg. De daadwerkelijke stervormingssnelheid
van passieve sterrenstelsels kan zelfs nog lager zijn, omdat we laten zien dat de
mid-infrarode flux ook van andere processen dan stervorming kan afkomen, zoals
bv. circumstellair stof and opwarming van stof door oude stellaire populaties.
Hoofdstuk 4 richt zich op de spectra van passieve en actieve sterrenstelsels op
roodverschuivingen van z = 0.5 tot z = 2 met als doel de stellaire leeftijden te bepalen.
We stapelen spectra van stervormings- and passieve stelsels, die geselecteerd zijn op
basis van een kleur-kleur techniek, en vergelijken ze met veelgebruikte modellen
van synthetische stellaire populaties. We vinden dat stellaire populatie modellen
goed passen bij waarnemingen op golflengtes lager dan 6500Å, terwijl ze systematische afwijkingen op rodere golflengtes. Wij bevestigen dat passieve sterrenstelsels
zwakke emissie lijnen hebben. Dit is in overeenstemming met metingen van de stervormingssnelheden in het mid-infrarood. De leeftijden van passieve sterrenstelsels
hangen af van het gebruikte model, maar in het algemeen zijn passieve sterrenstelsels jong, m.a.w. jonger dan de helft van de leeftijd van het heelal op elk kosmisch tijdstip. Voor stervormings-stelsels hangt de gemeten leeftijd sterk af van het
gebruikte stellaire populatie model en de evolutie van de stervorming.
In hoofdstuk 5 maken we gebruik van de volledige 3D-HST data om te analyseren hoe waterstof emissie afhangt van de eigenschappen van sterrenstelsels, in het
bijzonder de optische en nabij infrarode spectrale energie verdeling van het licht, in
het bereik van roodverschuiving waar Hα kan worden waargenomen met de HST
Samenvatting
95
en WFC3 grism (0.7 <z <1.5). We laten zien dat sterrenstelsels met sterke en zwakke
Hα goed van elkaar te scheiden zijn in een kleur-kleur diagram. Voor sterrenstelsels met actieve stervorming, onderzoeken we hoe Hα afhangt van de kleuren
van het sterrenstelsel en de specifieke stervormingssnelheid (stervormingssnelheid
per totale stellaire massa), gemeten in het ultraviolet en mid-infrarood. Bij constante
massa van het sterrenstelsel, hebben rode stervormings stelsels een lagere EW(Hα)
dan blauwe stervormings stelsels. We tonen ook aan dat bij constante massa, de
mediaan van de specifieke stervormingssnelheid afneemt als de U-V kleuren roder
worden, en dat absorptie door stof toeneemt voor rodere kleuren. We laten zien dat
in het algemeen de kleur-afhankelijkheid van EW(Hα) verklaard kan worden door
zowel een lagere specifieke stervormingssnelheid als een hogere mate van absorptie
door stof in sterrenstelsels met rodere kleuren.
96
Samenvatting
Publications
1. Hα Equivalent Widths from the 3D-HST Survey: Evolution with Redshift and Dependence on Stellar Mass
Fumagalli, Mattia; Patel, Shannon G.; Franx, Marijn; Brammer, Gabriel; van
Dokkum, Pieter; da Cunha, Elisabete; Kriek, Mariska; Lundgren, Britt; Momcheva, Ivelina; Rix, Hans-Walter; Schmidt, Kasper B.; Skelton, Rosalind E.;
Whitaker, Katherine E.;Labbe, Ivo; Nelson, Erica
The Astrophysical Journal Letters, Volume 757, Issue 2, article id. L22, 6 pp.
(2012)
2. How dead are dead Galaxies? Mid-Infrared fluxes of quiescent galaxies at 0.3 < z <
2.5: implications for star formation rates and dust heating
Fumagalli, Mattia; Labbe, Ivo; Patel, Shannon G.; Franx, Marijn: Van Dokkum,
Pieter, Brammer, Gabriel; da Cunha, Elisabete; Kriek, Mariska; Lundgren, Britt;
Momcheva, Ivelina; Rix, Hans-Walter; Schmidt, Kasper B.; Skelton, Rosalind
E.; Whitaker, Katherine E.; Nelson, Erica June, Ryan F. Quadri
The Astrophysical Journal, Volume 796, Issue 1, article id. 35, pp. (2014).
3. Ages of massive galaxies at 0.5 < z < 2.0 from 3D-HST rest-frame optical spectroscopy
Fumagalli, Mattia; Franx, Marijn; van Dokkum, Pieter; Whitaker, Katherine
E.; Skelton, Rosalind E.; Brammer, Gabriel; Nelson, Erica; Maseda, Micheal;
Momcheva, Ivelina; Labbé, Ivo; Lundgren, Brit; Rix, Hans-Walter
Submitted to The Astrophysical Journal
4. Decreasing Hα for redder star-forming galaxies: influence of dust and star formation
rates
Fumagalli, Mattia; Franx, Marijn; Labbé, Ivo; van Dokkum, Pieter; Kriek,
Mariska; Skelton, Rosalind E.; Whitaker, Katherine E.; Brammer, Gabriel; Momcheva, Momcheva
Submitted to The Astrophysical Journal
5. The structural evolution of Milky Way like star forming galaxies since z ∼ 1.3
Patel, Shannon G.; Fumagalli, Mattia; Franx, Marijn; van Dokkum, Pieter G.;
van der Wel, Arjen; Leja, Joel; Labbé, Ivo; Brammer, Gabriel; Skelton, Rosalind E.; Momcheva, Ivelina; Whitaker, Katherine E.; Lundgren, Britt; Muzzin,
Adam; Quadri, Ryan F.; Nelson, Erica June; Wake, David A.; Rix, Hans-Walter
The Astrophysical Journal Volume 778 Number 2 ApJ 778 115
6. First Results from the 3D-HST Survey: The Striking Diversity of Massive Galaxies
at z > 1
van Dokkum, Pieter G.; Brammer, Gabriel; Fumagalli, Mattia; Nelson, Erica;
Franx, Marijn; Rix, Hans-Walter; Kriek, Mariska; Skelton, Rosalind E.; Patel,
Shannon; Schmidt, Kasper B.; Bezanson, Rachel; Bian, Fuyan; da Cunha, Elisabete; Erb, Dawn K.; Fan, Xiaohui; Förster Schreiber, Natascha; Illingworth,
97
98
Publications
Garth D.; Labbé, Ivo; Lundgren, Britt; Magee, Dan; Marchesini, Danilo; McCarthy, Patrick; Muzzin, Adam; Quadri, Ryan; Steidel, Charles C.; Tal, Tomer;
Wake, David; Whitaker, Katherine E.; Williams, Anna
The Astrophysical Journal Letters, Volume 743, Issue 1, article id. L15, 6 pp.
(2011).
7. Spatially Resolved Hα Maps and Sizes of 57 Strongly Star-forming Galaxies at z = 1
from 3D-HST: Evidence for Rapid Inside-out Assembly of Disk Galaxies
Nelson, Erica June; van Dokkum, Pieter G.; Brammer, Gabriel; Förster Schreiber,
Natascha; Franx, Marijn; Fumagalli, Mattia; Patel, Shannon; Rix, Hans-Walter;
Skelton, Rosalind E.; Bezanson, Rachel; Da Cunha, Elisabete; Kriek, Mariska;
Labbe, Ivo; Lundgren, Britt; Quadri, Ryan; Schmidt, Kasper B.
The Astrophysical Journal Letters, Volume 747, Issue 2, article id. L28, 6 pp.
(2012).
8. 3D-HST: A Wide-field Grism Spectroscopic Survey with the Hubble Space Telescope
Brammer, Gabriel B.; van Dokkum, Pieter G.; Franx, Marijn; Fumagalli, Mattia; Patel, Shannon; Rix, Hans-Walter; Skelton, Rosalind E.; Kriek, Mariska;
Nelson, Erica; Schmidt, Kasper B.; Bezanson, Rachel; da Cunha, Elisabete;
Erb, Dawn K.; Fan, Xiaohui; Förster Schreiber, Natascha; Illingworth, Garth
D.; Labbé, Ivo; Leja, Joel; Lundgren, Britt; Magee, Dan; Marchesini, Danilo;
McCarthy, Patrick; Momcheva, Ivelina; Muzzin, Adam; Quadri, Ryan; Steidel,
Charles C.; Tal, Tomer; Wake, David; Whitaker, Katherine E.; Williams, Anna
The Astrophysical Journal Supplement, Volume 200, Issue 2, article id. 13, 19
pp. (2012).
9. 3D-HST Grism Spectroscopy of a Gravitationally Lensed, Low-metallicity Starburst
Galaxy at z = 1.847
Brammer, Gabriel B.; Sanchez-Janssen, Rubén; Labbé, Ivo; da Cunha, Elisabete;
Erb, Dawn K.; Franx, Marijn; Fumagalli, Mattia; Lundgren, Britt; Marchesini,
Danilo; Momcheva, Ivelina; Nelson, Erica; Patel, Shannon; Quadri, Ryan; Rix,
Hans-Walter; Skelton, Rosalind E.; Schmidt, Kasper B.; van der Wel, Arjen; van
Dokkum, Pieter G.; Wake, David A.; Whitaker, Katherine E.
The Astrophysical Journal Letters, Volume 758, Issue 1, article id. L17, 7 pp.
(2012).
10. Large-scale Star-formation-driven Outflows at 1 < z < 2 in the 3D-HST Survey
Lundgren, Britt F.; Brammer, Gabriel; van Dokkum, Pieter; Bezanson, Rachel;
Franx, Marijn; Fumagalli, Mattia; Momcheva, Ivelina; Nelson, Erica; Skelton, Rosalind E.; Wake, David; Whitaker, Katherine; da Cunha, Elizabete; Erb,
Dawn K.; Fan, Xiaohui; Kriek, Mariska; Labbé, Ivo; Marchesini, Danilo; Patel,
Shannon; Rix, Hans Walter; Schmidt, Kasper; van der Wel, Arjen
The Astrophysical Journal, Volume 760, Issue 1, article id. 49, 13 pp. (2012).
11. The Radial Distribution of Star Formation in Galaxies at z 1 from the 3D-HST Survey
Nelson, Erica June; van Dokkum, Pieter G.; Momcheva, Ivelina; Brammer,
Publications
99
Gabriel; Lundgren, Britt; Skelton, Rosalind E.; Whitaker, Katherine E.; Da
Cunha, Elisabete; Förster Schreiber, Natascha; Franx, Marijn; Fumagalli, Mattia; Kriek, Mariska; Labbe, Ivo; Leja, Joel; Patel, Shannon; Rix, Hans-Walter;
Schmidt, Kasper B.; van der Wel, Arjen; Wuyts, Stijn
The Astrophysical Journal Letters, Volume 763, Issue 1, article id. L16, 6 pp.
(2013).
12. Quiescent Galaxies in the 3D-HST Survey: Spectroscopic Confirmation of a Large
Number of Galaxies with Relatively Old Stellar Populations at z ∼ 2
Whitaker, Katherine E.; van Dokkum, Pieter G.; Brammer, Gabriel; Momcheva,
Ivelina G.; Skelton, Rosalind; Franx, Marijn; Kriek, Mariska; Labbé, Ivo; Fumagalli, Mattia; Lundgren, Britt F.; Nelson, Erica J.; Patel, Shannon G.; Rix,
Hans-Walter
The Astrophysical Journal Letters, Volume 770, Issue 2, article id. L39, 6 pp.
(2013).
13. The spatial extent and distribution of star formation in 3D-HST mergers at z ∼ 1.5
Schmidt, Kasper B.; Rix, Hans-Walter; da Cunha, Elisabete; Brammer, Gabriel
B.; Cox, Thomas J.; van Dokkum, Pieter; Förster Schreiber, Natascha M.; Franx,
Marijn; Fumagalli, Mattia; Jonsson, Patrik; Lundgren, Britt; Maseda, Michael
V.; Momcheva, Ivelina; Nelson, Erica J.; Skelton, Rosalind E.; van der Wel,
Arjen; Whitaker, Katherine E.
Monthly Notices of the Royal Astronomical Society, Volume 432, Issue 1, p.285300
14. The Assembly of Milky-Way-like Galaxies Since z 2.5
van Dokkum, Pieter G.; Leja, Joel; Nelson, Erica June; Patel, Shannon; Skelton, Rosalind E.; Momcheva, Ivelina; Brammer, Gabriel; Whitaker, Katherine E.; Lundgren, Britt; Fumagalli, Mattia; Conroy, Charlie; Förster Schreiber,
Natascha; Franx, Marijn; Kriek, Mariska; Labbé, Ivo; Marchesini, Danilo; Rix,
Hans-Walter; van der Wel, Arjen; Wuyts, Stijn
The Astrophysical Journal Letters, Volume 771, Issue 2, article id. L35, 7 pp.
(2013).
15. Confirmation of Small Dynamical and Stellar Masses for Extreme Emission Line
Galaxies at z 2
Maseda, Michael V.; van der Wel, Arjen; da Cunha, Elisabete; Rix, Hans-Walter;
Pacifici, Camilla; Momcheva, Ivelina; Brammer, Gabriel B.; Franx, Marijn; van
Dokkum, Pieter; Bell, Eric F.; Fumagalli, Mattia; Grogin, Norman A.; Kocevski, Dale D.; Koekemoer, Anton M.; Lundgren, Britt F.; Marchesini, Danilo;
Nelson, Erica J.; Patel, Shannon G.; Skelton, Rosalind E.; Straughn, Amber N.;
Trump, Jonathan R.; Weiner, Benjamin J.; Whitaker, Katherine E.; Wuyts, Stijn
The Astrophysical Journal Letters, Volume 778, Issue 1, article id. L22, 5 pp.
(2013).
16. A CANDELS-3D-HST synergy: Resolved Star Formation Patterns at 0.7 < z < 1.5
Wuyts, Stijn; Förster Schreiber, Natascha M.; Nelson, Erica J.; van Dokkum,
100
Publications
Pieter G.; Brammer, Gabe; Chang, Yu-Yen; Faber, Sandra M.; Ferguson, Henry
C.; Franx, Marijn; Fumagalli, Mattia; Genzel, Reinhard; Grogin, Norman A.;
Kocevski, Dale D.; Koekemoer, Anton M.; Lundgren, Britt; Lutz, Dieter; McGrath, Elizabeth J.; Momcheva, Ivelina; Rosario, David; Skelton, Rosalind E.;
Tacconi, Linda J.; van der Wel, Arjen; Whitaker, Katherine E.
The Astrophysical Journal, Volume 779, Issue 2, article id. 135, 16 pp. (2013).
17. Simultaneous Modeling of the Stellar and Dust Emission in Distant Galaxies: Implications for Star Formation Rate Measurements
Utomo, Dyas; Kriek, Mariska; Labbé, Ivo; Conroy, Charlie; Fumagalli, Mattia
The Astrophysical Journal Letters, Volume 783, Issue 2, article id. L30, 6 pp.
(2014).
18. 3D-HST+CANDELS: The Evolution of the Galaxy Size-Mass Distribution since z =
3
van der Wel, A.; Franx, M.; van Dokkum, P. G.; Skelton, R. E.; Momcheva, I.
G.; Whitaker, K. E.; Brammer, G. B.; Bell, E. F.; Rix, H.-W.; Wuyts, S.; Ferguson,
H. C.; Holden, B. P.; Barro, G.; Koekemoer, A. M.; Chang, Yu-Yen; McGrath, E.
J.; Haussler, B.; Dekel, A.; Behroozi, P.; Fumagalli, M.; Leja, J.; Lundgren, B. F.;
Maseda, M. V.; Nelson, E. J.; Wake, D. A.; Patel, S. G.; Labbé, I.; Faber, S. M.;
Grogin, N. A.; Kocevski, D. D.
The Astrophysical Journal, Volume 788, Issue 1, article id. 28, 19 pp. (2014).
19. Direct Measurements of Dust Attenuation in z ∼ 1.5 Star-forming Galaxies from 3DHST: Implications for Dust Geometry and Star Formation Rates
Price, Sedona H.; Kriek, Mariska; Brammer, Gabriel B.; Conroy, Charlie; Förster
Schreiber, Natascha M.; Franx, Marijn; Fumagalli, Mattia; Lundgren, Britt;
Momcheva, Ivelina; Nelson, Erica J.; Skelton, Rosalind E.; van Dokkum, Pieter
G.; Whitaker, Katherine E.; Wuyts, Stijn
The Astrophysical Journal, Volume 788, Issue 1, article id. 86, 13 pp. (2014).
20. The Nature of Extreme Emission Line Galaxies at z = 1-2: Kinematics and Metallicities
from Near-infrared Spectroscopy
Maseda, Michael V.; van der Wel, Arjen; Rix, Hans-Walter; da Cunha, Elisabete;
Pacifici, Camilla; Momcheva, Ivelina; Brammer, Gabriel B.; Meidt, Sharon E.;
Franx, Marijn; van Dokkum, Pieter; Fumagalli, Mattia; Bell, Eric F.; Ferguson,
Henry C.; Förster-Schreiber, Natascha M.; Koekemoer, Anton M.; Koo, David
C.; Lundgren, Britt F.; Marchesini, Danilo; Nelson, Erica J.; Patel, Shannon
G.; Skelton, Rosalind E.; Straughn, Amber N.; Trump, Jonathan R.; Whitaker,
Katherine E.
The Astrophysical Journal, Volume 791, Issue 1, article id. 17, 17 pp. (2014).
21. Dense Cores in Galaxies Out to z = 2.5 in SDSS, UltraVISTA, and the Five 3DHST/CANDELS Fields
van Dokkum, Pieter G.; Bezanson, Rachel; van der Wel, Arjen; Nelson, Erica
June; Momcheva, Ivelina; Skelton, Rosalind E.; Whitaker, Katherine E.; Brammer, Gabriel; Conroy, Charlie; Förster Schreiber, Natascha M.; Fumagalli, Mattia; Kriek, Mariska; Labbé, Ivo; Leja, Joel; Marchesini, Danilo; Muzzin, Adam;
Publications
101
Oesch, Pascal; Wuyts, Stijn
The Astrophysical Journal, Volume 791, Issue 1, article id. 45, 18 pp. (2014).
22. 3D-HST WFC3-selected Photometric Catalogs in the Five CANDELS/3D-HST Fields:
Photometry, Photometric Redshifts, and Stellar Masses
Skelton, Rosalind E.; Whitaker, Katherine E.; Momcheva, Ivelina G.; Brammer, Gabriel B.; van Dokkum, Pieter G.; Labbé, Ivo; Franx, Marijn; van der
Wel, Arjen; Bezanson, Rachel; Da Cunha, Elisabete; Fumagalli, Mattia; Förster
Schreiber, Natascha; Kriek, Mariska; Leja, Joel; Lundgren, Britt F.; Magee,
Daniel; Marchesini, Danilo; Maseda, Michael V.; Nelson, Erica J.; Oesch, Pascal; Pacifici, Camilla; Patel, Shannon G.; Price, Sedona; Rix, Hans-Walter; Tal,
Tomer; Wake, David A.; Wuyts, Stijn
The Astrophysical Journal Supplement, Volume 214, Issue 2, article id. 24, 49
pp. (2014).
23. Constraining the Low-mass Slope of the Star Formation Sequence at 0.5 < z < 2.5
Whitaker, Katherine E.; Franx, Marijn; Leja, Joel; van Dokkum, Pieter G.;
Henry, Alaina; Skelton, Rosalind E.; Fumagalli, Mattia; Momcheva, Ivelina
G.; Brammer, Gabriel B.; Labbé, Ivo; Nelson, Erica J.; Rigby, Jane R.
The Astrophysical Journal, Volume 795, Issue 2, article id. 104, 20 pp. (2014).
24. On the importance of using appropriate spectral models to derive physical properties
of galaxies at 0.7 < z < 2.8
Pacifici, Camilla; da Cunha, Elisabete; Charlot, Stéphane; Rix, Hans-Walter;
Fumagalli, Mattia; Wel, Arjen van der; Franx, Marijn; Maseda, Michael V.;
van Dokkum, Pieter G.; Brammer, Gabriel B.; Momcheva, Ivelina; Skelton,
Rosalind E.; Whitaker, Katherine; Leja, Joel; Lundgren, Britt; Kassin, Susan A.;
Yi, Sukyoung K.
Monthly Notices of the Royal Astronomical Society, Volume 447, Issue 1, p.786805
102
Publications
Curriculum Vitae
I was born on June the 12th, 1986, in Lecco, on the shores of Lake Como in Northern
Italy. I grew up in the small town of Morbegno, in the Valtellina valley on the Alps,
where I attended elementary, middle, and high school, all within 500m from my
house. In high school, at the Scientific Lyceum “Pierluigi Nervi”, I became fond of
natural and physical sciences.
In the fall of 2005, I moved to Milan to pursue a Bachelor Degree in Physics at
the University of Milano-Bicocca. I graduated from my Bachelor’s with a research
project on the calibration of the CMS-ECAL detector at the Large Hadron Collider,
partially completed during a Summer School at CERN, Geneva, under the guidance
of Prof. Marco Paganoni and Dr. Pietro Govoni. Despite the fact that I later changed
my field of research, it was during that summer that I became passionate about the
academic research environment and decided I wanted to continue with a research
based education. In October 2008 I enrolled in the Master’s in Astrophysics and
Space Physics at the University of Milano-Bicocca. During that time I was lucky
enough to join observing trips at the 150cm telescope of Loiano (Italy) and to the
2.1m of San Pedro Martir (Mexico), which made me passionate about observational
Astronomy. I graduated with a research project on the environmental effects on
galaxies in the local Universe, with Prof. Giuseppe Gavazzi as my supervisor.
In 2010 I was offered a PhD position at Leiden Observatory working with Prof.
Marijn Franx and Prof. Pieter van Dokkum (Yale University) to study high-redshift
galaxies with the novel spectroscopic data from the Wide Field Grism onboard Hubble Space Telescope. During my PhD I enjoyed numerous visits to Yale University,
New Haven (Connecticut, US), and as a member of the 3D-HST collaboration I attended meetings of the consortium in New Haven, Leiden, Heidelberg, and Puerto
Rico; I presented my work at conferences in Aix-en-Provence (France), Turku (Finland), Beijing (China) and Ameland (the Netherlands), and gave talks / colloquia at
the Carnegie Observatories in Los Angeles, University of California at Santa Barbara,
University of California at Irvine, Space Telescope Institute in Baltimore, and Goddard Space Flight Center in Greenbelt (Maryland). In 2012 I attended the Jerusalem
Cosmology Winter School in Israel.
Besides my research work, during the years at the Sterrewacht I served as a
teaching assistant to the Stellar Dynamics MSc Course taught by Prof. Vincent Icke,
and I have participated in UNAWE (Universe Awareness), an outreach program
funded by UNESCO and the International Astronomy Union, as a coordinator for
its Student Ambassador Program.
In January-March 2015 I took a hiatus from my research to join The Data Incubator, a highly selective training program in data science based in New York City, in
preparation for a transition from academia to the data science world.
103
104
Acknowledgments
It’s a sign of mediocrity when you demonstrate gratitude with moderation.
(Roberto Benigni)
Writing a thesis is on one hand a solitary process, but such a long effort would
not be possible without family, friends and colleagues, that more or less directly
teach us lessons that contribute to the positive outcome of a PhD.
I was very lucky in going through this process at Leiden Observatory, which provided great support in every daily matter. Thanks to the administrative staff and the
computer guys I could think 100% only about my research, without worrying about
anything else. Thanks to Jeanne, Alexandra, Evelijn, Els, Debbie, Anita, Liesbeth,
Eric, David, and Aart. A special word for Xander: without your support in the last
year this work would have been completed in an even longer timescale - thanks for
listening and advising, your help has been invaluable.
I feel privileged for having worked in a research group full of smart, inquisitive,
and fun people. Shannon and Adam, I learned so much from you guys I don’t even
know where to start. You taught me about galaxies, life in academia and life in
general. You listened when I needed scientific and personal help. Thanks guys, and
keep the wolfpack alive! Jesse, you have been a great friend, and we shared laughs,
frustrations, and adventures from China to Finland. Daniel, you inspired me a lot
with your precision for details and with your career path. Moein, I immensely
admired your calmness and your artistic skills. Allison, I am glad you brought back
a necessary female touch to the group; thanks for all the laughs and for the more
serious moments. I could have not chosen a better “academic little sister” - and I
wish you the very best for the next years to come! Joanna and Simone, I wish we
shared more time at the Sterrewacht, but I don’t forget your voices, wise and sweet
at the same time.
My thesis would have not been possible without the 3D-HST collaboration, and I
owe scientifically a lot to many members, particularly Ivo, Iva, Gabe, Kate, and Ros.
Visits to New Haven, and 3D-HST meetings around the world, added to my PhD
opportunities for discussing insightful science and developing friendships. Iva, you
are a great inspiration for your enthusiasm in everything you do, and I am thankful
to you and James for having always been incredibly welcoming in New Haven. Joel,
conferences and 3D-HST meetings would have not been the same without you - I
love recalling our roamings around Puerto Rico (credits to Micheal too), Israel and
Jordan (+Renske and Monica), and China (+Rachel and Jesse). Rachel and Erica,
it was always a great pleasure to see you in the Netherlands and in the US, and
I hope there will be more chances in the future. Micheal, you are definitely one
105
106
of the coolest astronomers I know and I really enjoyed the time spent together in
Heidelberg and in your soon-to-be new home Leiden. During my visits in New
Haven I always found a warm room and a nice breakfast at David and Bob’s thanks to you guys for making me always feel at home.
At the Sterrewacht I was lucky to share an office with a bunch of incredible individuals that gave a pleasant atmosphere to my day-to-day life even in the hardest
moments. First and foremost: Benjamin Darwin Oppenheimer, mostly known as
benopp - Ben, thanks for making office 441 a very light place to be, for your friendship and the hilarity you can get out of every situation. Berenice, I am glad we spent
all this journey together from the first to the last moment. Silvia, thanks for always
being calm and available, and for your wisdom - thanks for translating my Samenvatting as well! Gabriela, thanks for your friendship, for all the laughs together, and
for fostering my interest in Peruvian music.
During these years I shared lunches, coffee breaks, and borrels with many people
that always provided topics for interesting and stimulating conversations: Renske,
Rob, Nicola, Alessandra, Silvia, Gilles, Mason, Cameron, Aayush, Eva, Nico, Cristobal, Alex South, Alex North, Marijke, Sebastiaan, Steven, Caroline, Bram, Emanuele,
Lorrie, Sylvia, Thanja, Marcello, Massimo and David. Thanks everybody!
In my early days at the Observatory I also felt very welcomed by the “old guard”,
Rafa, Sergio, Olivera, Jeanette, Stefania, Edith, Elisabetta, Ernst, Bernadetta, and
Olmo. Thanks to all of you for providing an example for the years to come.
With some colleagues I found friendship that will last forever, even though we
are incresingly more and more scattered around the world. Matteo and Irene, I
remember like yesterday the day we celebrated our moving in together, and I could
have not had better housemates to start this adventure here. Thanks for having
bravely endured me as a housemate, and also for keeping the good time running
afterwards. Marco, life at the Sterrewacht would have been very different without
you. Thanks for the bike rides, the coffees, the chats about life and work, and for
forcing me to go to the gym. Bernard, despite our fierce football rivalry we never
argued a single time. You are such a cool guy I still cannot believe you support that
team. Monica, thanks for your camaraderie, we shared a lot of our thoughts in these
years and I always felt my feelings were in good hands when communicated to you.
Andra, I wish you didn’t travel that much during these years so we could spend
more time together, but the memories of the time together are all great. Marissa,
when I got here I didn’t know you but you treated me like I knew you since forever.
Thanks for being the center of so many activities, and for your sincere enthusiasm
for everything around us. Tiff, I am so glad you are one of the first people I met
here and we mantained our friendship through the years. You have been a great
source of support, and your help with the graduation and visa processes has been
invaluable.
I am thankful I could share artistic and athletic enterprises with some Sterrewacht people. With Andra (voice), Francisco (guitar), Simon (accordion) and Matteo (drums), I played the bass in The Redshift District, a rock-polka-punk band that
deserved more gigs than the ones it had - thanks guys for that short but incredibly
fun experience! A mention is deserved to the football crew that in the last 3 years
107
managed to achieve a decent success at the Sterrewacht tournament: Sebastiaan,
Marco, Rob, Aayush, Clement, Allison, Francisco, Pedro, Bram, and Eric.
My friends outside the Sterrewacht also made Leiden such a special place to be.
Thanks Hester, Simon, Lucia, for making Dutch class so fun. Thanks to Raphaela,
Nick, Anna, Bart, Niels, Sonja, Mart, Mark, Ugo, Ursula, Kristin, Sandro, and Gleb
for the dinners and festivities around town. Thanks Fabrizio for being my football
buddy, and thanks Agnese for being so patient with this small religion of ours. Sara,
even though we attended the same University in Italy we only met here and I am
really glad we did, because your joy is contagious and your accent reminds me of
home like nobody else. Tom and Alex West, our knowledge started as “boyfriend
of”, but I am happy I got to know you better through the years.
During these years away from home, I didn’t lose touch with the best friends
from my university years, sharing our post-university adventures, and providing
always good (g)chats. Franz, thanks for being at the same time so inspirational and
so much fun. Vedo, thanks for keeping me in the loop with your always interesting projects, and thanks for the consulting on the design of the cover. Dani, you
have been my long-standing housemate, thanks for being one of my best contacts
in our valley. Giulio, the synthesis of our friendship stands in the 7621 km we travelled together through Asia without killing each other, and instead integrating our
strengths for the best possible outcome.
Thanks to my parents, Vera and Maurizio, for their unconditional support during
these years. Having a single child that moves abroad is not easy for anybody, but
you have been always encouraging and understanding of my choices.
The greatest “thank you” has to go to my better half, my partner in crime, my
voice of reason, my biggest source of joy, my greatest support through these years.
Annie, you participated in this journey more than anybody else. I cannot state how
immensely thankful I am for all the times you calmed me down, listened to me, and
cheered me up. This thesis is yours too. I always felt that the Alps and the Rockies
forged us akin, and after we met in the flattest place on Earth we have walked so
much side by side that I just cannot wait for our next step in life together.
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