Download What Drives the Stellar Mass Growth of Early-Type

What Drives the Stellar Mass Growth of
Early-Type Galaxies? Born or made: the
saga continues...
Reinaldo R. de Carvalho (DAS/INPE-MCTI) - (PR)!
Reinaldo R. Rosa (LAC/INPE-MCTI) - (PI)!
Abstract
!
Constraining almost 14 billion years of galaxy evolution from observations of
galaxies as they are seen today is fraught with peril. Here we propose a program
to extend galaxy evolution studies to infrared wavelengths and higher redshifts
(larger lookback times) to consistently investigate galaxies and their environments over a significant time baseline. A major step will be the implementation
of the galaxy photometry tool GALPHAT to give a solid statistical basis for analyzing galaxy structural parameters.
We have already examined in great detail the properties of early-type galaxies
(ETGs) in the nearby Universe. By studying global properties of local ETGs,
such as color gradients, the fundamental plane, stellar populations, and their initial mass function (IMF), we have been able to constrain models of galaxy formation and evolution. We have characterized these galaxies’ environments using a physically meaningful and consistent measure of the host group/cluster
velocity distributions, so that we can begin to separate the effects of environment (nurture) and individual galaxy properties (nature).
We will build upon our expertise and existing toolkit to examine galaxies at an
earlier epoch, providing rigorous, consistent, and minimally biased comparison
samples for evaluating physical scenarios of galaxy formation and assembly.
One fundamental aspect of this proposal is the major investment in the
Bayesian Inference Engine package. This will create the necessary synergy between our team and the two groups of computer scientists participating in this
effort. Projects dealing with massive amounts of data in the near future will require a strong commitment from the two groups
!
!
!
!
!
1
O que Determina o Crescimento da Massa Estelar de Galáxias Elípticas? Intrínseco ou Ambiente: a saga continua...
!
Reinaldo R. de Carvalho (DAS/INPE-MCTI) - (PR)
Reinaldo Rosa (LAC/INPE-MCTI) - (PI)!
Resumo
Estabelecer quase 14 bilhões de anos de evolução a partir da observação de
galáxias como são observadas hoje em dia é uma tarefa repleta de perigos.
Neste projeto propomos um programa para estender estudos de evolução de
galáxias para comprimentos de onda do infravermelho e redshifts mais altos
(maiores tempos no passado) para investigar de forma consistente as propriedades das galáxias e seus ambientes ao longo de um intervalo de tempo
cósmico significativo. Um passo importante será a implementação do programa
de fotometria de galáxias GALPHAT que dará uma base estatística sólida para
a análise dos parâmetros estruturais de galáxias.
Em projeto recentemente concluído examinamos em grande detalhe as propriedades das galáxias do tipo “early” (ETG1) no Universo próximo. Ao estudar
as propriedades globais de ETGs locais, tais como gradientes de cor, plano fundamental, populações estelares, e função de massa inicial (FMI), estabelecemos
importantes vínculos sobre os modelos de formação e evolução de galáxias.
Caracterizamos os ambientes dessas galáxias usando uma medida associada diretamente à distribuição de velocidades das galáxias do grupo/aglomerado, que
nos permite separar os efeitos do ambiente dos efeitos intrínsecos.
Este projeto se beneficiará do conhecimento adquirido nos projetos que realizamos na última década focalizados nos sistemas observados no Universo local
e nas ferramentas de análise de dados que desenvolvemos e que serão vitais no
exame das propriedades das galáxias em redshifts mais altos, fornecendo
amostras de comparação rigorosas, consistentes e minimamente tendenciosas
para avaliar cenários de formação e evolução de galáxias.
Um aspecto fundamental da proposta é o grande investimento no pacote BIE.
Isto irá criar a sinergia necessária entre a nossa equipe e os dois grupos de cientistas da computação que participam deste esforço. Projetos que lidam com
grande quantidade de dados, em um futuro próximo, exigirão um forte compromisso destas duas áreas do conhecimento.
!
1
ETG - sigla do inglês que significa Early-Type Galaxy
2
Contents!
!
1. Definition of the scientific team
04
2. General Introduction
06
!
!
3. Conceptual Context & Scientific Background
!
4. Main Research Topics to be conducted
!
5. GALPHAT Using GPUs
!
!
07
!
11
!
24
!
6. MORFOMETRYKA - Morphological Analysis
!
7. Current Infrastructure and future needs
!
8. Schedule and Future Developments
!
9. Project Budget
!
25
!
28
!
29
!
31
!
!
10. Expected Results
33
!
!
11. Scientific Challenges
34
!
!
12. Public Outreach
34
!
13. Previous Approved Projects at FAPESP
!
14. Administrative Support
!
35
!
37
!
!
15. References
38
!
!
3
1. Definition of the scientific team!
Our team consists of 10 researchers working in São Paulo as staff members of 3
institutes. Outside São Paulo, we have an additional 14 scientists distributed
among 6 institutes. We have a further 12 researchers from abroad, from several
countries, some of whom are already participating in our main activities and
others that are just now entering our collaboration. There are also 8 graduate
students and 3 postdocs actively engaged in the projects described here, bringing the total team size to 47. The proposal is being submitted only by the PR
and PI.
!
Scientists in São Paulo (10)!
Anderson Caproni (UNICSUL)
Diego A.F. Gonçalves (EACH-USP)
Gustavo Lanfranchi (UNICSUL)
Hugo V. Capelato (INPE/MCTI)
Joaquim E.R. Costa (INPE/MCTI)
Luiz Otávio Saraiva Ferreira (DMC/Unicamp)
Oswaldo D. Miranda (INPE/MCTI)
Paulo Kurka (DMC/Unicamp)
Reinaldo R. de Carvalho (INPE/MCTI) PR
Reinaldo R. Rosa (INPE/MCTI) PI
!
Scientists Outside São Paulo (14)!
Adria Ramos de Lyra (IM/UFRRJ)
André L.B. Ribeiro (UESC)
Carlos Eduardo Ribeiro de Mello (IM/UFRRJ)
Esteban Walter Gonzalez Clua (IC/UFF)
Fabricio Ferrari (IMEF/FURG)
Henri Plana (UESC)
João L.K. Moreira (ON/MCTI)
Juliana Mendes Nascente e Silva (IM/UFRRJ)
Marcos de Oliveira Lage Ferreira (IC/UFF)
Mark Eirik Scortegagna Joselli (PUC/Paraná)
Marcelo Panaro Zamith (IM/UFRRJ)
4
Otton Teixeira da Silveira Filho (IC/UFF)
Sandro Rembold (UFSM)
Thiago S. Gonçalves (OV/UFRJ)
!
Scientists Abroad (12)!
Aaron Robotham (ICRAR/Australia)
Alberto G.O. Krone-Martins (Lisbon Univ/Portugal)
Francesco La Barbera (OAC-INAF/Italy)
Gary Mamon (IAP/France)
Ian Bonnell (Univ. of St. Andrews/Scotland)
Ignacio de la Rosa (IAC/Spain)
Ignacio Ferreras (UCL/UK)
José A.F. Pacheco (OCA/France)
Joseph Silk (IAP/France)
Marcelle S. Santos (Fermilab/USA)
Martin Weinberg (UMASS/USA)
Neal Katz (UMASS/USA)
Roy Gal (Univ. Of Hawaii/USA)
!
Students and Postdocs in São Paulo (06)!
Diego Stalder (INPE/MCTI)
Eduardo C. Vasconcellos (INPE/MCT)
Grzegorz Kowal (EACH-USP)
Jaime D. Vargas (DMC/Unicamp)
Marina Trevisan (INPE/MCTI)
Mariana Penna-Lima (INPE/MCTI)
!
Students and Postdocs outside São Paulo (05)!
Graciana B. João (IMEF/FURG)
João Gabriel Felipe Machado Gazolla (IC/UFF)
Leonardo A. Ferreira (IMEF/FURG)
Ramiro D. S. Lopes (IMEF/FURG)
Vanessa O. Gil (IMEF/FURG)
!
5
2. General Introduction!
Galaxies form though collapse of gravitationally unstable density perturbations
which were most likely produced during inflation. The Big Bang (BB) cosmological model with a cosmological constant, known as the ΛCDM2 model, explains reasonably well important observations like the small temperature variation of the Cosmic Microwave Background, the large scale distribution of galaxies, the BB nucleosynthesis abundances, and the acceleration of the expansion of the Universe. Inventorying the total mass-density energy in the Universe, we find that dark matter accounts for ∼26.8% and dominates the gravitational effects on very large scale structures; the cosmological constant is responsible for 68.3% and drives the accelerated expansion of the Universe; and
the remaining 4.9% constitutes the ordinary matter forming planets, stars and
galaxies. Most of the astrophysical complexity is inherent in this elusive component of the Universe, and most of our observations trace exactly this tiny but
fundamental baryonic component.
How does a galaxy form its stars? What determines the total stellar content of a
galaxy? The answers to these seemingly simple questions have eluded astronomers for over half a century. Star formation starting from a cold gas cloud
is an extraordinarily complex problem and probably one of the most difficult in
modern astrophysics. This challenging “closed-box” scenario is further complicated by the fact that most galaxies reside in larger structures, where they interact with both their neighbors and the diffuse material present in groups and
clusters. What can we do to make progress on disentangling the many processes
that effect the stellar mass buildup of galaxies, and understanding which ones
are dominant?
The study of the formation and evolution of galaxies in general requires their
systematic observations over a large redshift range in order to pinpoint the
mechanisms responsible for their observed properties today (z = 0). Ensuring
that the datasets for local and distant galaxies contain the same objects - or
more correctly, today’s galaxies and their actual progenitors - itself requires an
knowledge of the very evolution we are seeking to understand. Early-type galaxies, with their predominantly old stellar populations, provide the simplest systems with which to address these questions.
2
ΛCDM - Cold Dark Matter plus a cosmological constant Λ.
6
3. Conceptual Context & Scientific Background!
The study of the global properties of early-type galaxies (ETG) experienced
significant progress in the past decade thanks to the recent photometric and
spectroscopic surveys in the optical (Sloan Digital Sky Survey - SDSS) and in
the infrared (UKIRT Infrared Deep Sky Survey - UKIDSS). The SDSS (DR73 8423 square degrees) allowed the definition of a sample of elliptical galaxies
with homogeneous data, minimizing the systematics effects that plagued most
samples used in previous works (La Barbera et al. 2010a). The UKIDSS (DR4 1000 square degrees), even with a significantly smaller angular coverage than
SDSS, also yields homogeneous and high-quality data. This unprecedented
combination of optical and infrared permitted for the first time that a statistically significant sample of ETGs were defined over such a broad range in wavelength (La Barbera et al. 2010a,b).
In order to study how galaxies form and evolve it is of fundamental importance
to observe these systems over a large redshift range in order to probe the physical mechanisms responsible for producing galaxies as observed at z = 0. However, it is simpler to observe galaxies in the nearby Universe compared to their
counterparts at high redshift. This simple fact can introduce serious biases in
our interpretations when comparing different samples of galaxies from different
cosmic epochs. For the nearby samples, once homogeneous and high-quality
data became available, the study of ETGs progressed very rapidly. Now, we can
investigate in detail how these systems formed, how their stellar populations
evolved, and how their structural properties are modified by the environments
where they reside (de la Rosa et al. 2011; Swindle et al. 2011; Kormendy &
Bender 2012; La Barbera et al. 2012; Trevisan et al. 2012; Dressler et al. 2013;
Ferreras et al. 2013).
Star formation starting from a cold gas cloud is an extraordinarily complex
problem and probably one of the most difficult in modern astrophysics, requiring everything from simple gravity, through the complex chemistry of atomic
and molecular gas as well as dust, turbulence, through to feedback from stars
and supernovae. One way of studying this topic is through examination of the
Initial Mass Function (IMF), which is defined as the distribution of stellar
masses in a single population at the time of birth. It has been usually considered
3
DR7 - Seventh Data Release
7
a universal function, partly because of the complexities in obtaining proper observational constraints. The traditional approximation by a single power law
(Salpeter 1955) has undergone numerous updates, with more complex functions
that include a significant flattening of the slope towards low-mass stars (Miller
& Scalo 1979; Scalo 1986; Kroupa 2001; Chabrier 2003). For a recent review
on the IMF and its possible variations, see Bastian, Covey & Meyer (2010).
Ferreras et al. (2013) studying a sample of 40,000 ETGs defined in the SPIDER
(Spheroids Panchromatic Investigation in Different Environmental Regions)
project, (La Barbera et al. 2010), find a relation between the slope of the IMF
and the velocity dispersion of the ETG, showing that there is an excess of low
mass stars in the more massive ETGs (see also Conroy & Van Dokkum
2012a,b). Dynamical studies of ETGs reinforce this result (Capellari et al.
2006). The main problem in these works is related to the degeneracy between a
true IMF variation and the abundance ratio of the elements used in the analysis
(see La Barbera et al. 2013). Besides the complexity in characterizing the stellar
population of a galaxy per se, environmental effects are present and in some
cases dominant (Kormendy & Bender 2012). Thus, it is of fundamental importance to investigate the properties of ETGs (including their stellar populations)
as a function of their distance from the center of the nearest group/cluster.
These properties may also vary according to redshift and galaxy mass. The environment manifests itself through three main physical processes. These include
ram-pressure, a process first suggested by Gunn & Gott (1972) whereby a galaxy moving through the hot gas of a cluster of galaxies looses its interstellar
medium (ISM). Abramson et al. (2011) show that the effect is clearly present in
NGC 4330, a spiral galaxy in the Virgo cluster. Recent high-resolution simulation (e.g. Tonnesen & Bryan 2008,2009,2010) together with the morphologydensity relation (Dressler 1980) show that this process may be determinant of a
galaxy’s morphology. Another process, termed “harassment’’, is the collective
effect of several high velocity encounters of a galaxy with many others and
with the cluster global potential. Simulations show that this process may remove mass from the external parts of the galaxies, heat disks, and drive gas into
the center of the galaxy (Moore et al. 1996; Lake et al. 1998). The third process,
“starvation”, looks at the lifetime of a galaxy based on the availability of gas
for star formation. If stars form at a constant rate, with the value observed today, all the gas in a spiral galaxy would be consumed in less than a Hubble
time. Therefore, the lifetime of a spiral galaxy increases if gas is accreted more
8
recently (Larson 1980). A corollary from this idea is that starving a spiral galaxy by removing its gas reservoir transform it into an S0 system. These three
mechanisms operate in such a way that transmutation of a galaxy from a certain
morphology to another may depend on the mass of the galaxy, the mass (and
gas distribution) of the cluster where the galaxy is located, and its distance from
the center of the cluster. It is important to remember that the evolution of the
stellar population of a galaxy depends also on redshift.
All these regulating mechanisms depend on the distance from the center of the
cluster and their effects on the stellar population will depend on how efficiently
the local environment is capable of removing the interstellar gas (den Brok et
al. 2011). This gas loss process affects directly the color gradient (or age, or
metallicity gradient) since it may inhibit star formation. Recent results indicate
that metallicity decreases outwards in an ETG, with the slope being even more
negative from 1 to 8 Re . On the other hand, the external regions of these galaxies present older stellar populations compared to the nuclear region (La Barbera et al. 2012). This combination of older age and less metal rich stellar populations in the outskirts suggest a process where the envelopes of massive ETGs
are made up of small satellite systems which were incorporated via mergers,
and the stars of these satellites were born during the early stages of galaxy formation (La Barbera et al. 2010a, 2012).
Another important aspect determining the global properties of ETGs is related
to the distribution of radial velocities of the cluster galaxies (Einasto et al.
2012a,b), which can characterize the dynamical state of a system (group or
cluster). Theoretical and phenomenological developments suggest that the virialized equilibrium state of a gravitational system is described by a MaxwellBoltzmann distribution function (Hjorth & Williams 2010, Barnes & Williams
2012). In phase-space coordinates this translates to a Gaussian function. N-body
numerical experiments of the relaxation of single isolated gravitational systems
(Merrall & Henrisen, 2003) or that of cosmological halos (Hansen et al, 2005,
2006) also support this conclusion. Einasto et al. (2012a,b), examining a cluster
sample defined from SDSS-DR7, indicate that there is a relation between the
gaussianity of the velocity distribution of cluster galaxies and their global properties. Furthermore, Ribeiro et al. (2013) have shown that a marked segregation
exists in the properties of galaxies located in Gaussian (G) and Non-Gaussian
(NG) groups. In G groups, there is a significant difference between the proper9
ties of galaxies located in the central regions and those in the outskirts, while in
the case of NG groups the difference is null. This suggests that the evolution of
galaxies in NG groups progresses in a slower fashion than in G groups.
It is expected that compact groups (CGs) are the ideal laboratories for studying
environmental effects over the galaxy properties. CGs are characterized by high
galactic density (ρ) and low velocity dispersion (σ). The Hickson sample (Hickson 1982), representative of these systems in the nearby Universe (zmed ∼ 0.03),
was extensively studied in the last 30 years. More recently, with the advent of
high-quality and homogeneous data from SDSS, new CG samples were objectively defined following the same Hickson’s criteria (e.g. McConnachie et al.
2009). They define a sample of ∼ 2300 CGs with zmed ∼ 0.09, only slightly larger than that of Hickson sample. This dataset seems ideal for statistical studies of
CGs, unveiling a plethora of environment effects on the evolution of galaxies
and how tidal interactions can change a galaxy’s morphology.
In summary, to better understand how galaxies evolve it is of crucial importance to first define an unbiased sample covering environments from low-ρ and
low-σ (the field population), up to high-ρ and high-σ (groups/clusters), and including intermediate cases like CGs, with high-ρ and low-σ. Besides, a
panchromatic view of the galaxies in these systems is currently possible thanks
to surveys like SDSS and UKIDSS. As we will describe in the following section, this project intends to incorporate photometric and spectroscopic information for a sample of ETGs up to z ∼ 0.5, allowing investigations of the main
physical processes responsible for not only the transmutation of the different
observed morphological types but also for establishing the stellar populations
that we observed today.
!
!
!
!
!
!
!
!
10
4. Main Research Topics!
The activities proposed in this project follow from the general context presented
in the previous section. The main subject is the evolution of ETGs in different
environments up to z ∼ 0.5, which corresponds to an epoch of 5 billion years
ago. The data used in each specific sub-project is described below and come essentially from the GAMA (Galaxy And Mass Assembly) project, and from observing time obtained with FORS2 at VLT/ESO. For each project we list the
name of the principal investigators leading the topic.
!
4.1) The Initial Mass Function in ETGs (Dr. F. La Barbera, Dr. R.R. de
Carvalho, Dr. I. Ferreras)
An essential component to the theory of galaxy formation is the IMF, which describes the distribution of stellar masses in a single population at the time of
birth. Studying a sample of ∼40,000 ETGs of the SPIDER project, Ferreras et
al. (2013) found a strong correlation between central velocity dispersion and the
slope of the IMF, indicating an excess of low mass stars in massive ETGs. This
means that low mass ETGs are well described by a Kroupa IMF, while massive
ETGs require a bottom-heavier IMF. La Barbera et al. (2013) analyze a variety
of spectral indices, combining gravity-sensitive features with age- and metallicity-sensitive indices, while also considering the effects of non-solar abundance
variations. They conclude that central velocity dispersion, rather than alpha-enhancement, [α/Fe], drives the variation of the IMF. Although the analysis cannot
discriminate between a single power-law (unimodal) IMF and a low-mass (≲0.5
M⊙) tapered (bimodal) IMF, robust constraints can be inferred for the fraction
in low-mass stars at birth. The figure below shows this effect, which corroborates other findings based on dynamical (e.g. Capellari et al. 2006) and stellar
population analyses (Conroy & van Dokkum 2012a.b). In this figure, we show
the variation of the IMF slope – unimodal (top) and bimodal (bottom) distributions – against central velocity dispersion. The shaded region corresponds to the
68% confidence level of the joint PDF4 including spectral fitting and all three
line strengths (TiO1, TiO2 and Na8190). The Salpeter (unimodal) and Kroupa
equivalent (bimodal) cases are shown as horizontal dashed lines. The rightmost
panels give the stellar mass-to-light ratios in the SDSS r band for a 10 Gyr old
4
PDF - Probability Distribution Function
11
stellar population at solar metallicity, exhibiting the large variations one could
expect depending on the choice of IMF.
These recent results expressing the non-universality of the IMF have strong implications for theories of galaxy formation and evolution. However, there is still
great debate on the degeneracy between a true IMF variation and abundance ratio variations. We had a pilot program approved at ESO with X-Shooter to observe two very bright (K ∼ 12) and nearby (z ∼ 0.06) ETGs from the SPIDER
survey, covering the near infrared part of the spectrum. Comparing the line
strengths to predictions of state-of-art stellar population models, we will be
able, for the first time, to distinguish at more than 3σ between a Chabrier and
Salpeter (or even bottom-heavier) IMF, and constrain, at the percent level, the
12
element abundance ratios (e.g. Na and Ca). We show the sensitivity of a testing
set of Na, Ca, and Fe features to IMF (red), age (blue), and leading element
abundances (i.e. [Na/Fe], [Ca/Fe], and [Fe/H], respectively). The sensitivities
are estimated as differences of equivalent widths normalized to their (expected)
uncertainties (given the requested S/N). Notice that all selected features will allow us to detect a bottom-heavy (wrt a Chabrier) IMF at > 4 (up to 10) σ level.
The inset shows the results of simulations where all features are shifted according to their expected errors, and fitted back simultaneously to recover the IMF
slope and Na/Ca/Fe abundances.
!
4.2) The Star Formation History vs. Distance to Cluster Center (Dr.
R.R. de Carvalho, Dr. A. Robotham, Dr. A. Ribeiro)
Several physical processes occurring only in groups and clusters may alter the
evolution of a galaxy, like ram-pressure, harassment, and starvation. The environment also plays an important role in the star formation history of the system. Considering the complex interplay of all these factors, it becomes imperative to define a sample of ETGs covering a large domain in mass, over distinct
environments (field, groups, compact groups, clusters), and over a redshift interval sufficiently large for the evolutionary effects to be significant. We recently embarked on a collaboration with members of the GAMA (http://www.gamasurvey.org/) project and their sample fits the requirements mentioned above.
The groups/clusters in GAMA sample span the redshift range of 0 < z < 0.5 and
the galaxy member catalogs are complete down to r = 19.4. There are 12,200
groups with a total of 37,576 galaxies.
In Trevisan et al. (2012) we investigated the star formation history of visually
classified ETGs, in the local Universe (0.01 < z < 0.025), covering an interval
in stellar mass of 109 M☉ to 1011.5 M☉. We show there is a break around M∗ = 3
x 1010 M☉ in the relation between surface brightness and stellar mass, and a
clear difference in the ages and metallicities of the galaxies with stellar masses
below and above this break point. This result suggests an abrupt transition from
constant and highly efficient star formation in very massive galaxies to a gradually decreasing star formation efficiency in low mass systems. The behavior of
the metallicity-age relation in low mass galaxies is not compatible with accretion of a metal poor gas, suggesting that the relation is driven by “outflow”. The
13
figure below shows the fraction of accumulated stellar mass as a function of
redshift, for three stellar mass intervals. Solid triangles represent the median,
and dashed-line indicates mean values. The 1-σ interval is represented by the
dotted-lines.
In this project we will examine the process of accumulating baryons as a function of the distance from the center of the cluster, its mass, and its redshift, incorporating in the analysis the three specific physical mechanisms described
above, with different efficiencies.
!
4.3) The Star Formation History vs. Galactocentric Radius (Dr. R.R.
de Carvalho, Dr. O. Miranda, Dr. S. Rembold)
As the most massive galactic systems in the nearby Universe, ETGs play a crucial role in our understanding of the formation and assembly of luminous matter. However, there is still no consensus on the way these galaxies form and
evolve. As a natural consequence of the SPIDER project, we started an investigation of the radial behavior of the stellar population of 674 massive ETGs (M∗
≳ 3 x 1010 M⊙). We used data from the optical (SDSS) to the near-infrared
(UKIDSS) to measure color gradients, which are then translated into age and
metallicity gradients through stellar population synthesis models. We find that
metallicity decreases from the center to the outskirts of galaxies, with the slope
increasing (more negative) towards the external regions (Re < R < 8Re). When
we look at these properties as a function of the environment where the galaxy is
located, we find that the age gradient is larger in ETGs inhabiting high density
environments. The figure below shows the gradients in age and metallicity for
three mass intervals and in different environments. These results are of great
14
importance and merit further investigation. Unfortunately, the age-metallicity
degeneracy, especially acute in the analysis of broad-band photometry, precludes definite conclusions on the trends seen in the figure below. We gathered
VLT data with FORS2 for 13 massive ETGs (M∗ ≳ 1011 M⊙). The data will al
low us to measure directly age, metallicity, and [α/Fe] up to galactocentric distances of 4Re. Currently, in the literature, we find mostly data reaching 1-2Re .
!
4.4) The velocity distribution of cluster galaxies and its influence on
galaxy properties (Dr. A. Robotham, Dr. R.R. de Carvalho, Dr. A.
Ribeiro)
In cosmology, initial conditions determine the evolution of a galactic system,
which starts by expanding with the Hubble flow, then decouples from it, reaches the maximum of the expansion, starts collapsing, and eventually virializes
(Gunn & Gott 1972). The details of this general scheme are the main concern of
current research in cosmology and despite many efforts there is no sufficiently
robust methodology to establish the evolutionary stage of a group/cluster (e.g.
Robotham et al. 2008). In ΛCDM, more massive galaxies form from merging
of less massive units (e.g. Naab et al. 2007).
15
Ribeiro et al. (2013) presented a new methodology to quantify the gaussianity
of a velocity distribution. The concept of distance between two distributions
arises naturally in probability theory seen as an application of measure theory
(see Kolmogorov 1933). In this investigation we will use the so-called
Hellinger distance - a stable approximation to the Fisher information metric
(see e.g. Amari 1985). Ribeiro et al. (2013), examining the Berlind group/cluster catalog for the local Universe (z < 0.1), shows a connection between galaxy
properties (age, [Fe/H], eClass, g-r, Rpetro e ⟨µpetro⟩)5 and the gaussianity of the
velocity distribution of the groups. Bright galaxies (Mr ≤ -20.7) located in the
external and internal regions of the groups show no significant differences in
the quantities mentioned above, regardless of whether the groups are G or NG.
However, for galaxies with -20.7 ≤ Mr ≤ -17.9, they find a significant difference in their properties depending on their host groups (G vs. NG), suggesting
strong environmental effects. Additionally, Ribeiro et al. (2013) make a prediction about how gaussianity in galactic systems evolves with redshift (shown in
the figure above). Motivated by these findings we embarked on a collaboration
with the GAMA team to investigate how these effects extend to z = 0.5, tracing
5
eClass is a parameter estimated based on the spectrum of the galaxy that indicates morphology; g-r is the
aperture color; Rpetro is the Petrosian radius; and ⟨μpetro⟩ is the mean surface brightness within Rpetro .
16
the evolutionary history of star formation in group/cluster galaxies and the relation to the dynamical evolution of those group/clusters.
!
4.5) Stellar Populations in Compact Group ETGs (Dr. R.R. de Carvalho, Dr. H. Plana, Dr. I. de la Rosa, Dr. Gary Mamon)
In the last few decades, the study of Hickson’s compact groups (HCG) have
yielded a number of disappointments. Due to their high densities and low velocity dispersions, these systems were considered ideal laboratories for studying
tidal interactions and merging process. Furthermore, the paradigm of interaction-activity points to a significant increase in star formation activity in these
groups (e.g. Barnes & Hernquist 1992). Although a plethora of observations
show signs of interaction among galaxies in HCGs, the star formation rate
(SFR) observed in their constituent galaxies is typical of those in the general
field (e.g. Zepf et al. 1991). With recent massive spectroscopic surveys, large
and homogeneous samples of groups (Hashimoto et al. 1998) and clusters
(Wilman et al. 2005) were studied, showing that high density environments
(central regions of groups and clusters) disfavor star formation. The mechanism
responsible for this truncated star formation is still an open issue. de la Rosa et
17
al. (2007) studying a sample of 22 ETGs in HCGs and 12 ETGs in the field,
found that, on average, ETGs in HCGs are older and more metal poor than
those in the field, reinforcing previous results. They also found that low velocity dispersion ETGs in HCGs exhibit an increase in [Mg/Fe] and decrease in [Z/
H] relative to their counterparts in the field (see figure above). This can be interpreted as evidence for a mechanism for truncation of star formation in ETGs
in HCGs (see figure above). Hydrodynamical simulations confirm this interpretation (Di Matteo et al. 2005). Crossmatching of the McConnachie et al. (2009)
sample with SDSS-DR10 yields a subsample of 70 quartets. Also, we gathered
spectroscopic for extra 35 CGs from TNG6, confirming 20 of them as real quartets. In summary, we have a subsample of 105 CGs covering the redshift range
of 0.05 < z < 0.095 with accurate flux calibrated spectra available. These data
will allow us to investigate in detail the question of how truncated star formation operates in dense systems like CGs.
!
4.6) Star Formation from a theoretical perspective (Dr. D. Falceta
Gonçalves, Dr. I. Bonnell, Dr. R.R. de Carvalho, Dr. O. Miranda, Dr. J.
Silk)
Long before the recent evidence for an IMF that depends on galactic properties,
star formation – even in the most idealized scenarios – was already a long
standing problem. As mentioned before, understanding star formation is key in
modern astrophysics, and has immense implications in several fields, from ISM
dynamics to cosmology. For instance, galaxy evolution and cosmological simulations typically make use of analytical prescriptions of the SFR, stellar feedback, and IMF, which operate as “sub-grid” parameters. Self-consistent treatment of (extra)galactic and stellar dynamics is beyond our current computational limits, and a theoretical refinement of the sub-grid physics of these models is
therefore mandatory.
In order to determine how stars form one must first understand the dynamics of
the interstellar matter. Stars are basically made from dense and cold molecular
gas and a direct relationship between local gas density and SFR should exist
(Kennicutt & Evans 2012). Early type galaxies at z~2, for instance, would have
to contain massive amounts of molecular gas to account for the large SFR observed at these redshifts. However, molecular clouds are very dynamical sys6
TNG - Telescopio Nazionale Galileo
18
tems. These objects may fragment, collapse, and disrupt, resulting in cloud lifetimes that influence the SFR within the cloud. SFR is also influenced by the
rate at which molecular clouds are formed from the diffuse ISM. These
timescales must be understood in terms of both local (stellar feedback, heating
and cooling mechanisms, shocks, turbulence) and global (tidal effects, mass inflow) properties. The figure below shows an example of simulating the fragmentation, formation of bound cores, and stars in turbulent molecular clouds
(Bonnell et al. 2013).
Galactic and cosmological models depend also on the assumed IMFs. Typically
estimated from observational data of the Milky Way and nearby galaxies, the –
largely understood as “universal” – IMF has been recently argued to be dependent on the galaxy mass and its environment. Observations of the cold gas in
nearby galaxies reveal a more or less correlated core mass function (CMF), i.e.
the masses of pre-stellar cores, and the IMF. Equivalent CMF and IMF slopes
have been reported by many authors (Motte et al. 1998; Testi & Sargent; Johstone et al. 2000; Alves et al. 2007; Simpson et al. 2008), resulting in an unexplained “ one-to-one” core-to-star relation, regardless of any efficiency factor.
The origin of dense cores is related to the large scale turbulent ISM. Supersonic
turbulence drives shocks and perturbations that evolve into an ensemble of
small and large structures, such as clumps, filaments, and voids. Statistics of
clumps and cores is then described in terms of the statistics of turbulence itself
(Padoan & Nordlund 2002, Falceta-Goncalves et al. 2013). The further evolution of cores, once these become gravitationally bound, is a bit tricky. Smith et
al. (2009) have investigated the relation between the CMF and the stellar IMF
through numerical simulations. They find that although the CMF does resemble
19
the IMF, the core masses do not map directly into stellar system masses. Further
fragmentation of cores is possible, and a monolithic scenario for the formation
of massive stars remains elusive.
In this project we will address some of these questions such as: i) how molecular clouds form in ETGs, ii) what is the role of internal turbulence and external
sources on the CMF in these objects; iii) how cores evolve into stars; and finally, iv) is it possible to determine the SFR and the IMF based on galactic properties. These will be mostly addressed from analytical methods and by means of
direct numerical simulations in grid and SPH numerical schemes (FalcetaGoncalves et al. 2008,2010; Falceta-Goncalves & Lazarian 2011; Bonnell et al
2009, 2013).
!
4.7) Resolved Galaxies with the Gaia Satellite (Dr. A.G.O. Krone-Martins, Dr. R.R. de Carvalho)
The Gaia satellite is the next cornerstone mission of the European Space
Agency, and is designed to provide the most precise and accurate data about our
Milk Way for years to come (e.g., see Brown, 2013). However, this mission will
also observe tens of millions of stars in other galaxies in the local group, besides millions of unresolved galaxies in low redshifts, and hundreds of thousands of quasars in farther redshifts (e.g. Robin et al., 2012, Krone-Martins et
al., 2013). The satellite launched successfully on 20th December 2013, and although the final catalogue is not expected until 2021, intermediate data releases
will occur during the mission. The final catalogue will be unbiased and complete to G = 20 (V~20-22), and will include positions, proper motions and parallaxes for all these sources, measured at the micro-arc-second level. In addition, spectrophotometric measurements will be obtained between 330nm and
1000nm for all sources. Finally, spectroscopy around 847-874nm for all sources
up to G=17 will provide radial velocities precise to 1-15 km/s. The first intermediate data release is expected to be available at Launch+22 months, and thus
during the development of this FAPESP project. Due to the nature of Gaia’s
data processing, this first release is expected to contain only positions and the G
magnitude of most objects (>90%). The second release is expected to occur at
Launch + 28 months, and will include the first astrometric solutions, integrated
spectrophotometry and mean radial velocities for most sources. Additional data
20
releases are expected at Launch+40 and Launch+65 months, containing additional classification and object parameters.
In this project, we aim mainly at the preparation of the scientific exploitation of
Gaia data for studying nearby galaxies using their resolved stellar population.
Since during this project we expect the first and second releases to take place,
we will focus on extracting maximal information from the first of these intermediate data releases, but will also start preparations to exploit the final catalogue based on our experience with the second data release. Using Gaia’s astrometric and spectrophotometric data from the resolved stellar populations of
nearby galaxies, we will be able to tackle the star formation history of such objects, as well as to reliably estimate their initial mass functions, which are key
points of the overall project. Moreover, for the nearest objects, such as the
Magellanic Clouds, we will likely be in a position to derive their statistical parallaxes as well as large-scale, spatially resolved, kinematics. For other nearby
galaxies, we expect to be able to assess their spatially resolved kinematics, enabling us to derive their rotation curves and thus to directly estimate their gravitational potential.
Ideally, the objects that will be selected for study during this project must be
face-on, since in such cases most of the spatial velocity vector is projected on
the celestial sphere tangential plane – and are thus reflected in proper motions.
Some of the most obvious candidates are M33, M74, M83 and M101. The expected proper motion component of the rotational velocity in these galaxies is
within the expected errors for proper motion measurements of point sources in
the Gaia final data release. For the aforementioned galaxies, these components
should vary between vrot~4.6 uas/yr, for M74 and vrot~45 uas/yr, for M33.
Meanwhile, Gaia proper motion errors for individual point sources are expected
to be between ~7 uas/yr for V~14, up to ~50 uas/yr, for V~20 – naturally, these
figures improve if several point sources are considered for average local proper
motion estimations. Even though the full precision and accuracy of Gaia data
will only be attained at the final catalogue release, the first and second data releases that will be adopted for this project will not only provide important information about the galaxies in the nearby Universe, but will also allow us to
develop methods and strategies necessary to unravel the galaxies in our cosmological neighborhood by 2021.
21
4.8) Chemical Enrichment in ETGs (Dr. G. Lanfranchi, Dr. A. Caproni,
Dr. R.R. de Carvalho)
In studying the global properties of ETGs, the analysis of their chemical evolution can be used to impose several constraints on the general scenario for the
formation and evolution of galaxies, with the advantage that theoretical models
allow one to trace back the evolution of these systems to the epoch of their formation. Additionally, environmental effects can be contrasted against internal
mechanisms in order to explain radial variations of the chemical properties
Generally, chemical evolution models assume that ETGs formed through monolithic collapse of primordial gas clouds, followed by a very intense star formation episode. Then, energy released by SNe explosions is injected and accumulated in the interstellar medium until it exceeds the binding energy of the galaxy
as a whole. When this happens, a galactic wind is developed removing almost
all the gas of the ISM, halting star formation. After that, the galaxy evolves passively (Larson 1975). In an alternative scenario, Pipino and Matteucci (2004),
presented a revised monolithic model which allows the formation of ETGs by a
fast merger of cold gas lumps at high redshift. In both cases, the critical mechanisms controlling star formation are the IMF and the star formation efficiency
(normally the inverse of the star formation timescale). The IMF is assumed to
be constant in time and space and follows a power law similar to the formulation of Salpeter (1955), even though another prescriptions has been tested (Matteucci 1994). In Matteucci (1994), models for ETGs with downsizing in star
formation were adopted, by assuming an increasing efficiency of star formation
with galactic mass. As a consequence, galactic winds occur earlier in the more
massive galaxies. This process was named the “inverse wind” and reproduced
the increase of [<Mg/Fe>] with galactic mass as well as the mass-metallicity relation.
The star formation histories in ETGs, both in clusters and field, have confirmed
a downsizing in star formation and suggested that the formation of these galaxies in the field might have started around 2 Gyr after the ones in clusters. This
suggestion is based on the data on [<α/Fe>] ratios and ages in ETGs (Thomas et
al. 2005). Moreover, multi-zone chemical evolution models predict that in each
ETG the star formation stops in the outer regions before it does in the inner regions, as a result of the galactic winds. An increasing efficiency of star forma22
tion with galactic mass as well as a shorter timescale for the gas assembly with
increasing galactic mass are assumed (Pipino & Matteucci 2004).
The radial variations of the chemical properties of ETGs were also studied using chemical evolution models, considering only internal mechanisms. In this
case, it has been suggested that gradients are a consequence of more prolonged
star formation, and stronger chemical enrichment, in the central regions of the
galaxy. The predicted gradient seems, on average, to be independent of the mass
of the galaxies. Almost all these results were obtained assuming an isolated system, with no interaction with its surroundings. However, ETGs are found especially in groups and their fraction increases in dense environments. Therefore,
the analysis of gradients in ETGs and the comparison with chemical evolution
models may shed some light into the matter of how the environment affects
their evolution.
But even in isolated systems, recent discoveries can alter the standard scenario.
For instance, the IMF seems to be more complex than a single power law with
the same slope in all galaxies, independent of their mass. Recent investigations
(e.g. Ferreras et al. 2013) have shown that more massive ETGs formed a higher
fraction of low mass stars, which could change the predictions from chemical
evolution models. Also, the metallicity decreases in the outer regions of an ETG
whereas the stellar populations get older (La Barbera et al. 2012), and star formation in more massive systems is very efficient and gradually decreases in less
massive galaxies (Trevisan et al. 2012). These and other findings should be
tested by chemical evolution models. We will adopt a detailed chemical evolution code able to predict the abundances of a vast series of elements (from H
and He to heavy elements such as Ba, La, and Eu). The code, which was previously applied to local Dwarf Spheroidal Galaxies (dSph) and Dwarf Irregular
Galaxies (Lanfranchi & Matteucci, 2003, 2004), will be modified to allow investigating large spheroids. In fact, in a certain sense, there are similarities between the evolutionary paths for low and high-mass spheroids: in Lanfranchi &
Matteucci models the local dSph galaxies are formed through an infall of pristine gas, followed by a star formation controlled by its efficiency and by an
IMF. The energy released by SNe explosions during star formation increases the
kinetic energy of the gas until a galactic wind occurs. The main differences between the two scenarios are the intensity of star formation, more than 10 times
lower in dwarfs than in ETGs, and the fact that after the wind the star formation
23
in dwarfs continue in a much lower rate than previously, whereas in massive
ellipticals it is halted by the wind. For local dSph, galactic winds are the main
drivers of the present day chemical properties of these galaxies and there is no
need for invoking external mechanisms (ram-pressure, tidal stripping, etc) to
explain the observed data (Lanfranchi & Matteuci 2007, 2010).
In ETGS it is quite likely that the environment plays an important role. We must
consider a model with multiple zones to study the variation of the chemical
properties of the galaxy with galactic radius and also to take into account gas
inflow from the surroundings. In this vein, the combination of chemical evolution models with results of hydrodynamical simulations will help to constrain
the effects of the environment on the evolution of these systems. This development will lead us directly to the question of whether the mass loss process is
due to an internal mechanism, such as galactic winds, or to the environment.
!
5. Implementing GALPHAT on GPUs (Dr. Reinaldo R. Rosa,
Dr. Esteban Clua, Dr. R.R. de Carvalho)!
In the last five years we have invested considerable time and effort in the development a two-dimensional surface photometry package, 2DPHOT (La Barbera et al. 2008), which was extensively used during the SPIDER project. A
new galaxy image decomposition tool, GALPHAT (GALaxy PHotometric ATtributes), is now available, which is a front-end application of the Bayesian Inference Engine (BIE). BIE is a parallel Markov chain Monte Carlo package,
providing full posterior probability distributions and reliable confidence intervals for all model parameters. This approach was extensively tested using an
ensemble of simulated Sersic model galaxies over a wide range of observational
conditions: the signal-to-noise ratio S/N, the ratio of galaxy size to the PSF and
the image size, and errors in the assumed PSF; and a range of structural parameters: the effective radius re and the Sersic index n. The test results for simulated galaxies demonstrate that, with a careful choice of Markov chain Monte
Carlo algorithms and fast model image generation, GALPHAT is a powerful
analysis tool for reliably inferring morphological parameters for a large number
of galaxies over a wide range of different observational conditions.
24
GALPHAT represents a major forward step in our ability to extract useful information from homogeneous photometric datasets like SDSS and UKIDSS
(and other available imaging data) as compared to using 2DPHOT and GALFIT
(see Yoon et al. 2011). However, three independent developments are necessary
to fully exploit its power: 1) installation on the LNCC (National Laboratory for
Scientific Computing) cluster, where we have 60 nodes devoted to our image
processing project; 2) development of a GPU version, which promises to be
significantly faster than the cluster version; 3) development of an intuitive GUI
which allows processing of individual (or a set of) images in a relatively easy
fashion. This is part of an ongoing development called CosmoBook - an environment designed to handle the complexity of modern astronomical data and is
scalable for the near future. The underlying idea is to provide the underpinnings
for new and more efficient methods for dealing with information complexity,
integrating data from photometric and spectroscopic surveys and simulations.
We plan on using GALPHAT to study the morphology of a K-band magnitude
limited sample of 6000 ETGs from the UKIDSS footprints, and will characterize the luminosity functions of each component, namely, the bulge and disk
separately. The main goal is to search for correlations between model parameters and environment, which is characterized by the gaussianity of the velocity
distribution (see item 3.4). Another important aspect of this project is that using
a GPU cluster may improve by a factor of 700 the processing of the astronomical images, which would make the Bayesian scheme competitive with other
simpler methods.
!
6. MORFOMETRYKA - Morphological Analysis (Dr. Fabricio
Ferrari, Dr. Reinaldo R. Rosa, Dr. Esteban Clua, Dr. R.R. de Carvalho)!
Galaxy morphological properties result from not only the internal formation and
evolution processes but also from interaction with the environment. Galaxies in
groups or clusters may have diverse evolutionary paths compared to isolated
ones, which is clearly reflected in their morphology. Traditionally, galaxy morphology has been addressed visually: an expert examines images of galaxies
and identifies features (or absence of them in the case of ETGs) which distinguish the object as belonging to a specific galaxy class (e.g. Hubble 1926; de
Vaucouleurs 1959; Sandage 1975; van den Bergh 1976). This classification paradigm is strongly subjective, it is prone to errors and cannot be applied to
25
the majority of galaxies present in SDSS or UKIDSS. Thus, it is imperative to
quantify the morphology of galaxy as a measurable quantity that can be coded
in an algorithm.
There have been several attempts to objectively measure galaxy morphology
and to classify them accordingly. One relatively successful system is the concentration, asymmetry, smoothness, Gini and M20 (CASGM) system, presented
in Abraham (1994), Conselice (2000) and Lotz (2004). The concentration (C)
index is the ratio of the circular radii containing 20% and 80% of the total flux
of the galaxy, where these percentages can be defined in order to maximize the
distinction between systems and minimize seeing effects; The asymmetry (A)
coefficient is determined comparing a source image with its rotated counterpart;
the smoothness (S) measures the small scale structures in the galaxy; the Gini
coefficient measures how concentrated light is among the pixels; the M20 coefficient measures the size of the 20% brighter pixels in the galaxy. This basic set
has been enlarged with other quantities such as the Sérsic model parameters
(Sersic index n, effective radius Re, and mean surface brightness within Re,
<µ>e) (Sersic 1968), and the Petrosian radius (Petrosian 1976), among others.
Such a parametrization of galaxy morphology answers two immediate needs.
First, it is possible to reproduce human classification by positioning the galaxies
in the space of these parameters. In such a supervised classification, a set of
previously classified galaxies are used to train a discriminant function that will
assign to each new galaxy a probability of belonging to each class. Even the
simple supervised naive Bayes algorithm provides robust classification, although more sophisticated schemes such as quadratic discriminant functions or
kernel methods improve the classification further, at the expense of numerical
complexity (Bishop 2006). The second reason for establishing a galaxy morphometry system is that we can seek structures in the quantitative morphology
parameter space that may yield clues for the physical reasons for their formation and evolution of galaxies that are not visible in the currently human-based
mode. The Hubble tuning fork classification does not account for all the details
that we can currently measure in galaxy images. Further, it does not hold as we
go deep in space even at redshift z=0.25. So, a new classification procedure is
needed, both to handle the large amount of data becoming available with the
new surveys, and also to help us find the physical processes driving galaxy evolution.
26
Currently we have implemented an algorithm that works on galaxy images and
measures the morphological parameters automatically. The algorithm, called
MORFOMETRYKA (Ferrari, 2014), for a given field image, subtracts the sky
background, locates the main target, measures its center, ellipticity and position
angle; performs aperture photometry and fits a Sersic 2D model; and measures
Petrosian radius, concentration, asymmetry, smoothness and Gini coefficient.
MORFOMETRYKA was used to classify ∼90,000 galaxies belonging to 5,352
groups from the Berlind sample (Ribeiro et al. 2013). We trained the discriminant function from a set of 20,000 previously classified galaxies from Galaxy
Zoo. The figure below shows in different panels the separation between spirals
(in blue) and ellipticals (in red) for the training set. As we can see the parameters Concentration, Asymmetry, and Gini are excellent discriminators of both
morphological types. More details will be presented in Ferrari et al. (2014).
27
7. Current Infrastructure & Future Needs!
Currently, we have access to three main computational facilities in Brazil which
are used for different purposes: 1) LNCC - this system offers high performance
computing and so was used primarily for processing a large number of astronomical images. We recently finished processing all of the UKIDSS images in
K-band with our 2DPHOT package. Making use of 120 processors we were
able to complete the whole job in two months; 2) NAT - two clusters are used
for developing galactic chemical evolution codes: one with 16 processors
(AMD Phenom II 965 Quad core 3.4 GHz, 4GB RAM, 1TB HD); and another
with 96 processors (Core 2 Quad, Q 8400, 2.66 GHz, 2Gb ram, HD SATA
500Gb); 3) LAC - this is the most important facility we have access to and it is
described in the figure below. The LAC Hybrid Computer Cluster has eight
nodes, each one with 2 CPU's and two GPU units. Every node has main memory size of 128 GB DDR3.!
!
CPU configuration
GPU configuration
- Clock Rate: 2,2GHz.
- Instruction cache size L1: 256 Kbytes
- Cache size L2: 2MBytes.
- Cache size L3: 20MBytes
28
1.
2.
3.
4.
- Cores units Up to 2400
- Clock rate Up to 700 MHz
- Memory: 5GB GDRR5
- Memory bandwidth Up to
200Gb/s.
There are two extra systems which will be available for massive processing.
The first, from UNICAMP (University of Campinas - São Paulo/Brazil), is a
cluster with 5 nodes and 20 GPUs in total. The group from UNICAMP has important expertise in developing programs in CUDA to accelerate 2D image processing, an essential part of our project. Second, UFF has a cluster with 600
processors, each with 4GB memory and a GPU cluster with 6 GPUs Teslas
2070 2GB memory (each one with 4 Teslas). This group has a solid background
in developing massively parallel computing using CUDA, which is crucial for
our purposes. The large experience of both groups is the key factor in this
project since our goal is not only to process a large amount of images but also
to prepare for the PetaByte era approaching in the current decade. We are not
requesting any major hardware equipment in this proposal because what we
have access to is enough to not only process large amounts of data but also to
establish a sound benchmark for even larger amounts. We all have adequate infrastructure at our host institutes (i.e. Desktops, printers, servers, PC clusters,
workstations etc) and only a small request will be made.
Our needs, which will be also listed in the proper FAPESP forms, are focused
on the means to reach significant scientific results rather than hardware. We
plan on hiring three postdocs to work on the specific topics described in Section
3. The graduate students listed in item 1 are all involved in the projects listed in
item 3 as well. The major investment we expect from FAPESP is in 1) organizing meetings; 2) supporting travel and other needs for researchers from abroad
to participate in the collaboration via regular visits to INPE; 3) financing participation in international meetings to promote the work we are developing; and 4)
a minimum amount of hardware for specific uses.
!
8. Schedule and Future Developments!
We have already commenced development of a GPU version of GALPHAT, as
this is essential to reduce the processing time per galaxy by a factor of ∼700, elevating our processing power to a level where it becomes feasible to study large
samples of galaxies. To clarify the bottleneck, we consider the processing of the
complete set of galaxies down to r mag of 17.78, in the SDSS-DR7, with spectra (approximately 1,000,000 galaxies). Now, we can only run GALPHAT on
the LNCC cluster (with 60 processors) as the software was developed in C++
under the Linux platform. It takes 3 hours, on average, to process one galaxy,
29
which immediately implies that 1,000,000 galaxies would take a one hundred
years to process, not to mention that this is for just one photometric band, and is
thus completely unfeasible. Initial benchmarks from the UNICAMP group suggest that processing can be accelerated by a factor of ∼700 per GPU, so our
cluster of sixteen GPUs may reduce the processing time for a million objects to
a mere 10 days, which is completely reasonable. We will have access to three
independent GPU clusters, which make the project not only feasible but competitive, especially considering that the Bayesian methodology will provide unbiased results on the global properties of galaxies with strong implications for
the study of how galaxies form and evolve.
Our group is composed of well-established and very active researchers in the
fields of study listed in this project. Thus, we expect a substantial number of
papers per year. In order to increase the synergy between astrophysicists in our
group and the computer scientists collaborating with us, we intend to have regular mini-workshops gathering all participants for discussions. With the perspective of obtaining unprecedented measurements for a large sample of ETGs, our
schedule for technical and scientific developments, scientific meetings and publications is summarized below:
First Year
Second Year
Third Year
Development of a GPU version of
GALPHAT and BIE
Continuing development of BIE
Continuing development of BIE
Organization of two mini-workshops in
Brazil, where all national members will
present their contributions. We expect to
have always one or two members from
abroad participating in these meetings.
Continue with the miniworkshops
Continue with the mini-workshops
Publication of the technical papers
showing the GPU version of GALPHAT.
Also, we anticipate the publication of a
few papers on the subject of our research
We anticipate the publication of
a few papers per year.
We anticipate the publication of a
few papers per year
Adaptation of the chemical evolution code
to elliptical galaxies, making it a multizone model.
First results on chemical
enrichment of ETGs based on
the recent findings.
Combination of the chemical
evolution models with results of
hydrodynamical simulations
Numerical simulations of how molecular
clouds form in ETGs and what is the role
of internal turbulence and external sources
on the core mass function in these objects
Analysis of the simulation of
how cores evolve into stars
Application of numerical
simulations and SPH numerical
schemes to determine the SFR and
the IMF based on galactic
properties
On top of these general developments and organizational items we expect the
following achievements:
30
The first year will be devoted to implementing the GPU version of GALPHAT,
testing against other packages (GALFIT and 2DPHOT) and analyzing the
1,000,000 galaxy sample from SDSS in all four bands, griz.
The second year is devoted to implementing an environment for image analysis
incorporating all tools in a single environment, CosmoBook. This will integrate
photometric and spectroscopic data in a single database. It will be tested within
the group and then released to the international community
The third year will focus on the scalability of the Bayesian Inference Engine
(BIE) in GPU environments so that we can prepare ourselves for processing
data from wide field surveys being prepared for the next ten years. One major
test will be processing data from the GAMA survey through a collaboration
with the GAMA team (Dr. Aaron Robotham).
Over the three-year course of this proposal, we will invest considerable time in
the investigation of how ETGs have accreted their baryons over the past 5 billion years. This will be achieved by studying the GAMA sample, in particular
focusing on how age, metallicity and α/Fe of the central regions of ETGs vary
with the galaxy's mass, redshift and location within the cluster in which they reside. As detailed in Section 4.4, we developed a novel approach for establishing
the environmental properties which promises to be more efficient in tracing the
physical mechanisms influencing galaxy evolution in groups/clusters. Using
smaller samples, but with sufficient spatial resolution and S/N to measure the
stellar population parameters out to 4Re, we will measure the stellar mass assembly as a function of lookback time (see details in Sections 4.2, 4.3, and 4.5).
!
9. Project Budget!
The resources of this project are mainly in the form of visits (air tickets and per
diem), organization of mini workshops, and three postdocs. Below we present a
detailed description for each case. The cost estimates were gathered with the
most common airline companies and we use the per diem rates established by
FAPESP. Estimates are always in US$ and R$ (conversion factor 1US$ = 2.3R
$).
The budget includes the following items:
A) National equipment: One computer system to be acquired in Brazil
31
Workstation Dell Precision T7610 (R$ 12.023,00 or US$ 5.227,39)
Processor - Intel® Xeon® E5-2620 v2 (HT with 6 cores, 2.1GHz, 15MB)
Memory - 16 GB SDRAM DDR3, 1866 MHz
Hard disk - Serial ATA with 1 TB (7.200 rpm)
!
B) International Air tickets for special visits:
Nine international round trips over 3 years. Estimating seven to Europe and two
to USA with average fares quoted from several air companies (Delta, United,
AeroMexico etc) we have a total of ∼R$ 35.500,00 (US$ 15.434,79).
!
C) National Air tickets for the mini workshops and regular visits:
Since we are planing five mini-workshops and five regular meetings and having
collaborators from Rio Grande do Sul, Rio de Janeiro and Bahia, we estimate
that 63 air tickets over the period of thee years. The regular technical visits are
essential to stimulate the synergy with the computer scientists participating in
this project. The grand total for the 3 year period is R$ 44.400,00 (US$
19304,35)
!
D) Mini workshops and regular visits - per diem
The mini workshops will last one week (5 days) each. We intend to cover the
expenses of the collaborators from abroad as well as some of those from Brazil.
Also, we want to cover a substantial part of the cost for bringing collaborators
(researchers, postdocs, and graduate students). We estimate 320 per diem over
the three years, resulting in R$ 102.400,00 (US$ 44.521,74). We would like to
emphasize that during the workshops we will organize a series of lectures on
the subject of Galaxy Formation and Evolution held by the collaborators from
abroad visiting us. This will be of fundamental importance for the graduate students in the state of São Paulo.
!TOTAL BUDGET : R$ 194.323,00 or US$ 73.125,24
!
32
10. Expected Results!
Our project aims to study how ETGs form and evolve through the investigation
of their photometric and spectroscopic properties and environmental dependences. Among the expected fundamental results are: 1) the translation of the
BIE package to a GPU architecture, which will be of paramount importance not
only to our immediate goal (having GALPHAT running on a GPU cluster) but
also to other astrophysical analyses where a Bayesian scheme may be needed;
2) Developing MORFOMETRYKA and GALPHAT for GPUs will allow us to
have an unbiased statistical analysis of a large sample of ETGS. The Bayesian
approach applied to images in the optical and infrared promises to be essential
to study galaxy properties and how they depend on the environment where they
reside; 3) The new spectroscopic data on ETGs being gathered at VLT (FORS2
and X-SHOOTER) is an important factor for the success of this investment.
One of team members, Dr I. de la Rosa from IAC, also has access to GTC
(Gran Telescopio Canarias, the 10.2m reflecting telescope) which increases the
impact of the research we are currently doing. We expect to have five MSc and
two PhD theses completed during the period of this proposal as they are all developing their specific projects within the context of the research proposed here.
!
11. Scientific Challenges!
Apart from the fact that studying the way galaxies evolve through cosmic time
is complex in nature, some of the challenges of this project are related to two
primary aspects. One is the development of the GPU version of BIE, which will
enable GALPHAT to run two or three orders of magnitude faster. Beyond this
specific application, BIE is a full set of libraries for performing Bayesian inference and may be extremely useful for other analyses done during this project
and others as well. We will make BIE available on different platforms, so that
its usage can be enlarged to all segments of our community. The second and
maybe more important aspect of this project is to investigate the reality of the
IMF variation with galaxy mass, a very controversial subject in current literature. The importance of this finding is directly linked to the fact that the implications are enormous. By studying the ETG properties in different environments and especially focusing on the star formation history in these systems we
expect to unravel the intricate process of generating an IMF. The simulations
33
we are planing to do together with the specific study of the stellar population
models promise to be very effective tools to gain understanding of the physics
behind the IMF.
!
12. Public Outreach!
Astronomy is seen as a gateway science to introduce the general public to scientific concepts and techniques, and to encourage students to consider careers
in science, technology, engineering, and mathematics (STEM). Galaxies in particular provide some of the most spectacular images, while simultaneously providing the opportunity to examine everything from the properties of stars, the
large scale structure of the Universe, gas physics, black holes, and supernovae,
to image processing techniques, spectroscopy, and statistics.
Our goal is to implement a multi-faceted outreach program, using the expertise
of one of our foreign collaborators. Dr. Roy Gal at the University of Hawaii Institute for Astronomy is in charge of all of their outreach programs, and he will
share their successful techniques with us, collaborating to develop localized
versions for Brazil.
One program will involve the creation of a public lecture series, featuring talks
aimed at the interested public audience, with speakers drawn from our local
participants and our foreign collaborators. These will be hosted at large facilities (auditoriums with 100s of attendees) and broadly publicized, bringing
knowledge of our research to the community, and raising community awareness
of FAPESP programs.
At the educational level, we will leverage the experience of programs like Galaxy Zoo, our photometric and spectral database, and the Bayesian methodologies, to create activities for students at various grade levels highlighting various
aspects of our research. We will start by examining the learning goals for specific grade levels, and seek ways to address these goals using the resources at
our disposal. For instance, we can at lower grade levels introduce the basic
ideas of galaxies as large collections of stars, bound by gravity, and then build
upon this to study galaxy types, having students do just visual classification, to
adding simpler computer morphological parameters and constructing plots to
search for correlations, to undergraduate and graduate level studies of galaxies
34
and statistics. Where possible, we will work with educators to develop lesson
plans incorporating our data and techniques and then hold workshops to teach
teachers on how to implement these in their classrooms.
!
13. Previous approved projects at FAPESP!
The research done by several members of our team has benefitted very much
from other projects financed by FAPESP in the last five years, and we list them
here to show how all of these independent efforts are being linked to promote a
more efficient use of the resources. Here are the projects:
1 - Project 2012/05142-5, Category: Postdoc Fellowship
Title: Baryons in Dark Matter Halos: Content and Star Formation History in
ETGs, Responsible: Reinaldo R. de Carvalho (Postdoc: Marina Trevisan)
Period: 01/09/2012 a 31/08/2014
2 - Project 2006/57824-1, Category: Research Funding – Young Researcher
Title: Theoretical and Computational Models Applied to Astrophysics
Responsible: Gustavo A. Lanfranchi
Period: 02/01/2007 to 01/31/2011
3 - Project 2006/61377-0, Category: Fellowship – Program Young Researcher
Title: Theoretical and Computational Models Applied to Astrophysics
Responsible: Gustavo A. Lanfranchi
Period: 02/01/2007 a 05/31/2008
4 - Project 2010/17142-4, Category: Postdoc Fellowship
Title: Chemical Evolution of Local Galaxies
Responsible: Gustavo A. Lanfranchi (Postdoc: Monica Midori Marcon Uchida
Sguazzardi)
Period: 03/01/2011 to 02/28/2014
5 - Project 2011/18938-0, Category: Organization of Scientific Meeting
35
Title: NAT - Lectures on Astrophysics I. Collapsing or Colliding Systems: Solving the Galactic Puzzle
Responsible: Gustavo A. Lanfranchi
Period: 02/06/2012 to 02/10/2012
6 - Project 2013/06722-8, Category: Financial Support to Attend Meeting
Title: Chemical Evolution in the Universe, The Next 30 years, Italy
Responsible: Gustavo A. Lanfranchi
Period: 09/16/2013 to 09/20/2013
7 - Project 2009/10102-0, Category: Regular Grant
Title: Study of magnetized collisional and non-collisional plasmas in astrophysics, Responsible: D. Falceta Gonçalves
Period: 10/01/2009 a 09/30/2011
8 - Project 2011/12909-8, Category: Regular Grant
Title: Magnetic fields, turbulence and plasma effects in the intergalactic medium, Responsible: D. Falceta Gonçalves
Period: 10/01/2011 a 09/30/2013
9 - Project 12/21877-5, Category: Postdoc Fellowship
Title: Supermassive Black Holes and the Reionization of the Universe
Responsible: Oswaldo D. Miranda (Eduardo dos Santos Pereira)
Period: 01/01/2013 - 31/12/2014
10 - Project 09/15612-6, Category: Postdoc Fellowship
Title: Signatures of non-singular quantum cosmological models in the cosmic
microwave background radiation
Responsible: Oswaldo D. Miranda (Dennis Fernandes Alves Bessada)
Period: 01/03/2010 - 31/03/2011
36
11 - Project 2006/61378-7, Category: Fellowship – Program Young Researcher
Title: Theoretical and Computational Models Applied to Astrophysics
Responsible: Anderson Caproni
Period: 02/01/2007 a 05/31/2008
12 - Project 2010/03812-8, Category: MSc Fellowship
Title: Modeling of interferometric radio images of the BL Lac OJ 287 through
cross-entropy global optimization method
Responsible: Anderson Caproni (Rafael Teixeira Toffoli)
Period: 09/01/2010 - 08/31/2011
!
14. Administrative Support !
The host institute for this project will be the National Institute for Space Research, which is a well-known research center in Brazil responsible for a significant portion of the investigations being done in space and atmospheric science
in the country. The institute is obviously experienced in supporting and managing large scale projects (including several programs granted by FAPESP). The
Division of Astrophysics in particular will provide secretarial support throughout the project, which will be essential especially for the organization of the
mini workshops being planned.
!
!
15.References!
Abraham, R G.; Valdes, F.; Yee, H. K. C.; van den Bergh, S., ApJ 1994 432 75
Abramson, A. et al. 2011, AJ, 141, 164
Alves, J.; Lombardi, M.; Lada, C. J. 2007, A&A, 462, L17
Amari, S. and Nagaoka, H., 2000, in Methods of Information Geometry, vol.
191, American Mathematical Society and OUP
Amari, S., 1985, Differential-geometrical methods in statistics, LNS, V 28,
Springer
Barnes, J. E., & Hernquist, L. 1992, ARA&A, 30, 705
37
Barnes, E.I, Williams, L.L.R., 2012, ApJ 748, 144
Bastian N., Covey K. R., Meyer M. R., 2010, ARA&A, 48, 339
Bishop, C. (2006). Pattern recognition and machine learning. Berlin: Springer.
ISBN 0-387-31073-8.
Bonnell, I. A.; Dobbs, C. L.; Smith, R. J. 2013, MNRAS, 430,1790
Brown, A., 2013, arXiv:1310.3485.
Cappellari M. et al., 2006, MNRAS, 366, 1126
Chabrier G., 2003, PASP, 115, 763
Conselice, C. J.; Bershady, M. A.; Jangren, A. ApJ. 2000, 529, 886
Conroy C., van Dokkum P., 2012a, ApJ, 747, 69
Conroy C., van Dokkum P., 2012b, ApJ, 760, 71
de Carvalho et al., 2010, JCIS, 1 , 187
de la Rosa, I. G. et al., 2007, AJ, 133, 330
de la Rosa et al., 2011, MNRAS, 418, L74
den Brok, M. et al., 2011, MNRAS, 414, 3052
De Vaucouleurs, G, Handbuch der Physik, 1959, Vol. 53, 275
Di Matteo, T. et al., 2005, Nature, 433, 604
Dresler, A., 1980, ApJ, 236, 351
Dressler, A. et al., 2014, ApJ, in press
Einasto, M. et al., 2012a, A& A, 540, 123
Einasto, M.,et al., 2012b, A& A, 542, 36
Falceta-Gonçalves, D.; Lazarian, A., Kowal, G. 2008, ApJ, 679, 537
Falceta-Gonçalves, D.; Lazarian, A.,Houde, M. 2010, ApJ, 713, 1376
Falceta-Gonçalves, D.; Lazarian, A. 2011, ApJ, 735, 99
Ferrari, F. Et al., 2014, in prep.
Ferreras I. Et al., 2013, MNRAS, 429, L15
Gunn, J.E. & Gott, J.R., 1972, ApJ, 176, 1
Hansen, S.H. et al., 2006, JCAP, 1, 14
Hashimoto, Y., 1998, ApJ, 499, 589
Hjorth, J., & Williams, L.L.R., 2010, ApJ, 722, 851
38
Hubble, E., Astrophys. Journ. 64, 321-369 (1926).
Johnstone, D., et al., 2000, ApJ, 545, 327
Kennicutt, R. C.; Evans, N. J., 2012, ARA&A, 50, 531
Kolmogorov, A.N., 1933, Foundations of the Probability Theory, Chelsea, NY.
Kormendy, J. & Bender, R., 2012, ApJ Sup, 198, 2
Krone-Martins, A., et al., 2013, A&A, 556, A102 (or arXiv:1307.5732).
Kroupa P., 2001, MNRAS, 322, 231
La Barbera et al., 2008, PASP, 120, 681
La Barbera F. et al., 2010a, AJ, 140, 1528
La Barbera F. et al., 2010b, MNRAS, 408, 1313
La Barbera et al., 2012, MNRAS, 426, 2300
La Barbera et al., 2013, MNRAS, in press
Lake, G., Katz, N., & Moore, B., 1998, ApJ, 495, 152
Lanfranchi, G. A., Matteucci, F., 2003, MNRAS, 345, 71
Lanfranchi, G. A., Matteucci, F., 2004, MNRAS, 351, 1338
Lanfranchi, G. A., Matteucci, F., 2007, A&A, 468, 927
Lanfranchi, G. A., Matteucci, F., 2010, A&A 512, A85
Larson, R. B., Tinsley, B. M., & Caldwell, C. N. 1980, ApJ, 237, 692
Lotz, J. M.; Primack, J.; Madau, P., AJ, 2004, 128, 163
McConnachie, A.W. et al., 2009, MNRAS, 395, 255
Motte, F.; Andre, P.; Neri, R. 1998, A&A, 336,150
Moore, B. et al., 1996, Nature, 379, 613
Naab, T., 2007, ApJ, 658, 710
Padoan, P.; Nordlund, Å. 2002, ApJ, 576, 870
Petrosian, V. ApJ. 1976 209, 1
Ribeiro et al., 2013, MNRAS, in press
Robin, A. C., et. al. 2012, A&A, 543, A100
Robotham, A., Phillipps. S. & De Propis, R., 2008, ApJ, 672, 834
Salpeter E. E., 1955, ApJ, 121, 161
39
Sandage, A.R.: Galaxies and the Universe (Sandage, A.R., Sandage. M., Kristian, J., eds.) Stars and Stellar Systems 9, Univ. Chicago (1975) p. 1.
Sérsic, J.-L. 1968, Atlas de Galaxias Australes (Cordoba: Observatorio Astronomico)
Simpson, R. J., Nutter, D., Ward-Thompson, D. 2008, MNRAS, 391, 205
Smith, R. J.; Clark, P. C.; Bonnell, I. A. 2009, MNRAS, 396, 830
Swindle, R. et al.,2011, AJ, 142, 118
Testi, L.; Sargent, A. I. 1998, ApJ, 508, 91
Tonnesen, S. & Bryan, G.L., 2008, ApJ, 684, 9
Tonnesen, S. & Bryan, G.L., 2009, ApJ, 694, 789
Tonnesen, S. & Bryan, G.L., 2010, ApJ, 709, 1203
Trevisan, M. et al. 2012, ApJ, 752, L27
van den Bergh, S.: Astrophys. J. 206 (1976) 883.
Vasconcellos et al. 2012, AJ, 141, 189
Vazdekis A. Et al., 2003, MNRAS, 340, 1317
Zepf, S. E., Whitmore, B. C., & Levison, H. F. 1991, ApJ, 383, 524
Wilman, D.J. et al., 2005, MNRAS, 358, 71
40