Observational Evidence For AGN Feedback in Early-Type Galaxies

Mon. Not. R. Astron. Soc. 382, 1415–1431 (2007) doi:10.1111/j.1365-2966.2007.12487.
Observational evidence for AGN feedback in early-type galaxies
Kevin Schawinski,1 Daniel Thomas,1,2 Marc Sarzi,1,3 Claudia Maraston,1,2

Sugata Kaviraj,1 Seok-Joo Joo,4 Sukyoung K. Yi4 and Joseph Silk1
1 Department of Physics, University of Oxford, Oxford OX1 3RH
2 Instituteof Cosmology & Gravitation, University of Portsmouth, Portsmouth PO1 2EG
3 Centre for Astrophysics Research, Science & Technology Research Institute, University of Hertfordshire, Hatfield AL 10 9AB
4 Centre for Space Astrophysics, Yonsei University, Seoul 120-749, Korea
Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

Accepted 2007 September 11. Received 2007 August 31; in original form 2007 August 1
ABSTRACT
A major amendment in recent models of hierarchical galaxy formation is the inclusion of so-
called active galactic nucleus (AGN) feedback. The energy input from an active central massive
black hole is invoked to suppress star formation in early-type galaxies at later epochs. A major
problem is that this process is poorly understood, and compelling observational evidence for
its mere existence is still missing. In search for signatures of AGN feedback, we have compiled
a sample of 16 000 early-type galaxies in the redshift range 0.05 < z < 0.1 from the Sloan
Digital Sky Survey (SDSS) data base (MOSES: Morphologically Selected Ellipticals in SDSS).
Key in our approach is the use of a purely morphological selection criterion through visual
inspection which produces a sample that is not biased against recent star formation and nuclear
activity. Based on the nebular emission-line characteristics we separate between star formation
activity, black hole activity, the composite of the two and quiescence. We find that emission is
mostly LINER (low ionization nuclear emission line region) like in high-mass galaxies (σ >
200 km s−1 ) and roughly evenly distributed between star formation and AGN at intermediate
and low (σ < 100 km s−1 ) masses. The objects with emission (∼20 per cent) are offset from
the red sequence and form a well-defined pattern in the colour–mass diagram. Star-forming
early-types inhabit the blue cloud, while early-types with AGN are located considerably closer
to and almost on the red sequence. Star formation–AGN composites are found right between
these two extremes. We further derive galaxy star formation histories using a novel method that
combines multiwavelength photometry from near-ultraviolet (UV) to near-infrared (IR) and
stellar absorption indices. We find that in those objects deviating from the red sequence star
formation occurred several 100 Myr in the past involving 1–10 per cent of the total stellar mass.
We identify an evolutionary sequence from star formation via nuclear activity to quiescence.
This transition process lasts about 1 Gyr, and the peak AGN phase occurs roughly half a Gyr
after the starburst. The most likely interpretation is that star formation is suppressed by nuclear
activity in these objects before they settle on the red sequence. This is empirical evidence for
the occurrence of AGN feedback in early-type galaxies at recent epochs.
Key words: galaxies: active – galaxies: elliptical and lenticular, cD – galaxies: evolution –
galaxies: formation.
and structure formation theory. In every generation of theoretical

1 INTRODUCTION
models, early-type galaxies have helped provide the insights into
The formation of early-type galaxies is one of the major unsolved what physics was still missing from them and they continue to do
problems of astrophysics. Despite their apparent simplicity, they so. The vast majority of early-type galaxies appear to have formed
have time and again refused to fit into the bigger picture and thus, most of their stars at high redshift on very short time-scales as
they have been the most difficult and illuminating test for galaxy suggested by the stellar population properties of nearby samples
(e.g. Trager et al. 2000; Nelan et al. 2005; Thomas et al. 2005;
Bernardi et al. 2006; Jimenez et al. 2006) as well as the redshift
E-mail: kevins@astro.ox.ac.uk evolution of luminosity and mass functions of massive galaxies

C 2007 The Authors. Journal compilation
C 2007 RAS
1416 K. Schawinski et al.
(e.g. Cimatti, Daddi & Renzini 2006; Wake et al. 2006; Rodighiero for the mere existence of AGN feedback in early-type galaxies is
et al. 2007b). still missing.
Until very recently, the predictions from semi-analytic models of The aim of this paper is to tackle the key question of the role of
hierarchical galaxy formation were in severe conflict with the ob- AGN feedback observationally by looking for its signatures in early-
servationally derived formation ages and time-scales of early-type type galaxies at recent epochs in the nearby universe. Crucial for this
galaxies. These models take the merger trees of dark matter haloes in purpose is the compilation of a sample of early-type galaxies with
a cold dark matter (CDM) universe either by following them in a selection method that does not bias against AGN and/or SF. The
a cosmological N-body simulation (Kauffmann et al. 1999) or gen- sample further needs to be large, because the potentially short dura-
erating them using the extended Press–Schechter formalism (White tion of the AGN feedback phase makes such objects relatively rare.
& Frenk 1991). They then calculate the more complex fate of bary- To this end we have constructed a sample of nearly 16 000 early-type
onic matter inside them with prescriptions for star formation (SF; galaxies from the Sloan Digital Sky Survey (SDSS) in the redshift
Kennicutt 1998), radiative cooling of gas (Sutherland & Dopita range 0.05 z 0.1. These early-types are part of a sample com-

1993; Springel et al. 2001) and feedback from supernovae prising 50 000 early- and late-type galaxies in that redshift range
(Kauffmann et al. 1999). Early semi-analytic models such as that compiled in a project called Morphologically Selected Ellipticals
of Kauffmann & Charlot (1998) predicted that the most massive in SDSS (MOSES; Schawinski et al., in preparation). The major
galaxies in the universe harbour the youngest stellar populations novelty is the purely morphological selection mechanism through
and assemble the last, contrary to what has been deduced from ob- visual inspection of all 50 000 galaxies. Through emission line diag-
servations. For a critical example, these models struggle to match nostics we identify active early-type galaxies with SF activity and/or
the observed chemical abundance ratios and their trends with galaxy AGN. From the stellar populations we derive SF histories for these
mass (Thomas 1999; Nagashima et al. 2005). objects, which allows us to explore the possible interaction between
The reason for this failure in the early models is that SF con- the formation histories of the galaxies and their current SF/AGN
tinues unchecked in massive haloes where supernova feedback is activities.
insufficient (Dekel & Silk 1986). Benson et al. (2003) showed that The paper is organised as follows. In Sections 2 and 3, we outline
in massive galaxies, an additional, more powerful heating source the sample selection, present the emission-line diagnostics and the
than supernovae is needed to heat the cold gas that fuels SF and stellar population analysis. We place the sample galaxies subdivided
therefore stop it. This additional heating source is now generally in quiescent, AGN and star forming objects on the colour–σ relation
identified as feedback from active galactic nuclei (AGN) where jets and discuss first tentative conclusions in Section 4. A more quantita-
and outflows from the central supermassive black hole are presumed tive discussion is provided in Section 5 based on the two-component
to heat and expel most residual gas via a wind thus preventing fur- SF histories derived. Conclusions in the context of AGN feedback
ther SF. This AGN feedback has been discussed by a number of in galaxy formation are discussed in Sections 6 and 7.
authors (Ciotti & Ostriker 1997; Silk & Rees 1998; Binney 2004; Throughout this paper, we assume a standard CDM cosmol-
Silk 2005; Springel, Di Matteo & Hernquist 2005a) and is now ogy consistent with the WMAP (Wilkinson Microwave Anisotropy
part of many galaxy formation models (Granato et al. 2001, 2004; Probe) results (Bennett et al. 2003) with m = 0.3, = 0.7 and
Kaviraj et al. 2005; Kawata & Gibson 2005; Springel et al. 2005b; H 0 = 70 h70 km−1 s−1 . All magnitudes are in the AB system.
Bower et al. 2006; Cattaneo et al. 2006, 2007; Croton et al. 2006;
De Lucia et al. 2006; Cavaliere & Menci 2007). The phenomeno-
2 DATA S A M P L E
logical model of Schawinski et al. (2006a) is specifically motivated
by Galaxy Evolution Explorer (GALEX) observations of early-type The sample utilized here is part of a project called MOSES. Key is
galaxies. Di Matteo, Springel & Hernquist (2005) and Springel et al. that all 48 023 galaxies in this sample have been visually inspected
(2005a) have also investigated the effect of AGN feedback in hydro and classified into early- and late-type morphology by eye, yielding
simulations of mergers and have now placed them in large-scale a sample of 15 729 early-type galaxies. A detailed description of the
simulations probing cosmological volumes (Di Matteo et al. 2007; catalogue can be found in a companion paper (Schawinski et al., in
Sijacki et al. 2007). Note that alternatives have been suggested based preparation), and we refer the reader to this paper for any details. In
on the processes of cooling and shock heating of baryonic matter the following, we provide a brief summary with focus on the aspects
in dark matter haloes (Birnboim, Dekel & Neistein 2007; Hopkins that are most relevant for the present work.
et al. 2007; Khochfar & Ostriker 2007).
AGN feedback may come in two different modes at different
2.1 Sample selection
epochs during cosmic time. In an early phase at high redshifts the
AGN may truncate and quench SF leading to the formation of the We use the SDSS Data Release 4 (DR4) as our main sample (York
dominant old population in massive galaxies. This so-called ‘quasar et al. 2000; Stoughton et al. 2002; Adelman-McCarthy et al. 2006).
mode’ (Croton et al. 2006) is most likely accompanied by vigorous The SDSS is a survey of the northern sky providing us with photom-
black hole growth and leads to very strong kinetic feedback that etry in the five filters u,g,r,i and z (Fukugita et al. 1996; Gunn et al.
quenches strong starbursts. The other face of AGN feedback can be 1998). We selected objects from the redshift slice 0.05 < z < 0.10
thought of as a ‘maintenance’ mode. It turns on at more recent epochs with the magnitude limit r < 16.8. These cuts are mostly motivated
such that further significant episodes of SF are prevented (Best et al. by the limit out to which a visual inspection of galaxy morphology
2005, 2006; Fabian et al. 2005, 2006). The coupling between the is reliable (see Schawinski et al. 2006b).
black hole accretion and the gas heating is poorly constrained by the
observation. The latter mechanism has only recently been included
2.2 Morphological classification
in semi-analytic galaxy formation models. While such prescriptions
are successful at curing some key issues, their major problem is that The selection criteria used to identify early-type galaxies are crucial.
the physics of this process is still poorly understood and essentially Selection criteria based on galaxy properties such as colour, spectral
unconstrained by observations. Compelling observational evidence features or concentration index rather than morphology are at risk

C 2007 RAS, MNRAS 382, 1415–1431
Evidence for AGN feedback in early-type galaxies 1417
to be biased against early-type galaxies with residual SF or AGN
2.3 Photometry
activity. As these are the objects most relevant for the present study,
we decided for a radical, purely morphological selection. We retrieved SDSS model magnitudes (modelMag) for all five filters
The first catalogues of early-type galaxies in the SDSS used for u,g,r,i and z from the SDSS DR4 data base. In order to improve
extensive study are those of Eisenstein et al. (2001) and Bernardi the wavelength coverage of the galaxies in our sample, we match
et al. (2003a,b,c,d). The Bernardi et al. (2003a) catalogue is based them to a number of other surveys, namely GALEX, 2MASS (Two
primarily in the likelihood of the surface-brightness profile resem- Micron All Sky Survey) and SWIRE (Spitzer Wide Area Infrared
bling a de Vaucouleurs law and a high concentration index. This Extragalactic Survey).
is a very efficient way of selecting large numbers of spheroids, but
can fail to reject spiral galaxies with prominent bulges. In partic- 2.3.1 Matching to GALEX
ular, spirals with face-on faint discs and edge-on discs can pass
these criteria. Furthermore, Bernardi et al. (2003a) reject galax- The GALEX (Martin et al. 2005) is conducting several surveys of

ies with a principal component analysis (PCA) classification num- the ultraviolet sky in two filters: far-UV (FUV; 1344–1786 Å) and
ber atypical of red, passive early types (Connolly & Szalay 1999). near-UV (NUV; 1771–2831 Å). The Medium Imaging Survey from
This means that early types with strong emission lines or that are the General Release 3 (MIS/GR3) is covering a portion of the sky
dominated by young stellar populations are rejected despite having coinciding with the SDSS to a limit of 23 AB in NUV and FUV. We
early-type morphology. For an example, the early-type galaxies dis- match our galaxies to MIS within 4 arcsec to find counterparts. GR3
cussed by Fukugita et al. (2004) would not be included. Recently, only covers a fraction of SDSS DR4 and so this matching gives us
Fukugita et al. (2007) have compiled an catalogue of morphologi- detections for 13.4 per cent of galaxies.
cal early-type galaxies by visual inspection of 2658 objects from the
SDSS. 2.3.2 Matching to 2MASS
For the MOSES sample we carried out manual inspection of mul-
ticolour images from SDSS by eye for all 48 023 galaxies, in order The 2MASS (Skrutskie et al. 2006) provides us with J, H and
to avoid any potential bias introduced by selecting by colours or Ks band imaging for our galaxies selected in SDSS. We match our
spectral features. We define as early-type galaxies all objects earlier galaxies to the Extended Source Catalogue (XSC) within 4 arcsec,
than and including S0 galaxies and reject galaxies from our sam- giving us counterparts for 87.1 per cent. The 2MASS photometry is
ple if they show distinct spiral arms or a disc. We include in this done in the Vega system, so in order to keep consistency with the
selection objects with clear tidal features or other signs of morpho- rest of the MOSES sample, we convert the magnitudes from Vega
logical disturbance. In the paper presenting the MOSES catalogue to AB following Blanton et al. (2003b, 2005).
(Schawinski et al., in preparation) we present a number of example
images from SDSS to illustrate the effectiveness of our visual in- 2.3.3 Matching to Spitzer-SWIRE
spection. Out of the 48,023 galaxies in our volume-limited sample,
we identify 15 729 early-type galaxies. Fig. 1 illustrates the differ- The SWIRE (Lonsdale et al. 2003) is one of the Spitzer Legacy
ence between our selection and that of Bernardi et al. (2003a) in Surveys and imaged nearly 50 deg2 using both the Infrared Array
the colour–σ diagram. It can be seen clearly that our morphological Camera (IRAC; 3.6, 4.5, 5.8, 8.0 μm) and Multiband Imaging Pho-
selection yields a significantly higher fraction of early-type galaxies tometer for Spitzer (MIPS; 24, 70, 160 μm) instruments. We cross-
with emission lines and/or blue colours. Less than 5 per cent of the match the MOSES sample to the public source catalogues for ELAIS
early-type galaxies in the catalogue of Bernardi et al. (2003a) show N1, N2 and Lockman Hole fields. Only a small fraction of our sam-
emission lines, while over 18 per cent of ours do. ple (0.3 per cent) has counterparts at 3.6 μm.
2.4 Spectroscopy
SDSS has performed follow-up spectroscopy using 3-arcsec fibres
yielding a reasonably complete sample to r < 17.77 (Blanton et al.
2003a). The spectra cover the range of 3800 to 9200 Å at a resolving
power of 1800 and for the MOSES sample have a typical signal-to-
noise ratio (S/N) of about 30. We retrieve the SDSS optical spectra
for all our galaxies.
2.4.1 Gas and stellar kinematics

We measure the kinematics of the stars and the emission-line fluxes
of our galaxies from their SDSS spectra following the methods of
Cappellari & Emsellem (2004) and Sarzi et al. (2006), respectively.1
We separate the contribution of the stellar continuum and of the
ionized-gas emission to the observed galaxy spectrum by fitting
simultaneously stellar population templates (from Tremonti et al.
2004) and Gaussian emission-line templates to the data. In these fits
Figure 1. Optical colour–magnitude relation for the early-type catalogue of
Bernardi et al. (2003a) and this work. Early-type galaxies are divided into
‘quiescent’ and ‘active’ depending on whether emission lines are detected. 1We made use of the corresponding PPXF and GANDALF IDL (Interactive Data
Our purely morphologically selected sample contains a significantly larger Language) codes adapted for dealing with SDSS data. Both softwares can
fraction of active early-type galaxies. be retrieved at http://www.strw.leidenuniv.nl/sauron/

C 2007 RAS, MNRAS 382, 1415–1431

Figure 2. The BPT line diagnostic diagrams for the early-type galaxies of the MOSES sample. Each galaxy is coloured by its optical u − r colour (see colour
bar in left-hand panel). Point sizes scale with the galaxy velocity dispersion as an indicator of mass (see legend in left-hand panel). In each diagram, we indicate
the demarcation lines used in our classification scheme. The dashed line in the [N II]/Hα diagram (left-hand panel) is the empirical SF line of Kauffmann et al.
(2003) (labelled Ka03), while the solid curve (in all three panels) is the theoretical maximum starburst model from Kewley et al. (2001) (labelled Ke01). The
galaxies in between these two lines are SF–AGN composites or ‘transition region objects’. Galaxies below the Ka03 line are dominated by SF. The division
line between Seyferts and LINERs is shown as the straight line following Kewley et al. (2006) for the middle and right-hand panel and our own definition (see
text) for the left-hand panel.
we account also for the impact of diffuse dust in the galaxy of dust in Seyfert AGN and LINERs (low ionization nuclear emission line
the emission-line regions, which allows to obtain a decrement on the regions). The diagrams also contain a transition region, where the
strength of the Balmer lines that is at least what expected by recom- emission lines indicate a blend of SF and AGN activity.
bination theory. From the fit to the stellar continuum and absorption We use the four optical line ratios [O III] λ5007/Hβ, [N II]
features, we measure the line-of-sight velocity dispersions. From λ6583/Hα, [S II] λ6717/Hα and [O I] λ6300/Hα. We follow the S/N
subtraction of the emission-line spectrum from the observed one, criterion of Kauffmann et al. (2003) and classify all galaxies that
we get the clean absorption line spectrum free from emission-line have a S/N > 3 detection of Hα, Hβ, [O III] λ5007 and [N II] λ6583.
contamination. We then use this cleaned spectrum to measure the The two low ionization species [S II] and [O I] are used when de-
stellar absorption indices. The physical constraints on the emission tected, as they are usually weaker than the other four.
from high-order Balmer lines ensures the strength of the correspond-
ing absorption features is correctly estimated, which is crucial for
constraining the ages of stellar populations. 2.5.1 Diagnostic diagrams
The corresponding three diagnostic diagrams are shown in Fig. 2.
Each galaxy is coloured by its optical u − r colour. Point sizes
2.4.2 Lick absorption line indices
scale with the galaxy velocity dispersion as an indicator of mass. In
On each spectrum we measure the 25 standard Lick absorption line each diagram, we indicate the demarcation lines used in our clas-
indices (Worthey et al. 1994; Worthey & Ottaviani 1997) following sification scheme. In the left-hand column, we show the [N II]/Hα
the most recent index definitions of Trager et al. (1998). For this pur- diagram used to separate star-forming objects (blue) by means of
pose the spectral resolution is reduced to the wavelength-dependent the demarkation line by Kauffmann et al. (2003, dashed line). We
Lick resolution (Worthey & Ottaviani 1997). The measurements verified that the somewhat more restrictive separation between star
are then corrected for velocity dispersion broadening. The correc- forming and AGN suggested by Stasińska et al. (2006) does not
tion factor is evaluated using the best-fitting stellar template and alter the results of this work. The remaining objects are divided in
velocity dispersion obtained previously. Errors are determined by composite transition region objects (purple) and AGN using the the-
Monte Carlo simulations on each spectrum individually based on oretical maximum starburst model from Kewley et al. (2001, solid
the S/N provided by the SDSS. Possible discrepancies between the lines).
flux-calibrated SDSS spectral system and the Lick system are negli- The AGN are then further subclassified into Seyferts (green) and
gible, as shown in a detailed analysis of Lick standard stars observed LINERs (red) by the straight solid lines. The more indicative low
with SDSS (Carson 2007). ionization species (middle and right-hand panel) are used where
possible. The best separation of Seyferts and LINERs is achieved
using [O I] (right-hand panel of Fig. 2), which is used wherever
2.5 Emission-line diagnostics
[O I] is detected with S/N > 3. The [O I] line is generally weaker
Emission-line diagnostic diagrams are a powerful way to probe the then any of the other emission lines used, so where it is not avail-
nature of the dominant ionizing source in galaxies (Baldwin, Phillips able, we use [S II]. To separate the two types of AGN, we use the
& Terlevich 1981, hereafter BPT). They allow the separation of lines defined by Kewley et al. (2006). If both these lines are not de-
galaxies into those dominated by ongoing SF and non-stellar pro- tected we adopt the original [N II] diagram for AGN classification.
cesses; and with sufficient information can further split those into Based on the location of those galaxies classified by [O I] or [S II]

C 2007 RAS, MNRAS 382, 1415–1431

Figure 3. A sample of morphological early-type galaxies from the MOSES sample sorted by their emission line classification. The cut outs are 40 × 40 h−1 kpc
in size.
on the [N II] diagram, we define a new empirical Seyfert–LINER 2.5.3 Star formation rates
separation:
For those galaxies classified as star forming, we derive SF rates
log([O III]/Hβ) = 1.05 log([N II]/Hα) + 0.45. (1) from the extinction-corrected Hα line luminosity according to the
conversion of Kennicutt (1998). All SF rates for our galaxies quoted
in this paper are based on Hα unless otherwise noted. We cannot
2.5.2 Classification results derive reliable SF rates for any of the other emission line classes,
as an unknown fraction of the Hα emission line is of non-stellar
Following the emission-line classification we divide our early-type
origin.
galaxy sample into five classes:
(i) quiescent (orange); 3 S T E L L A R P O P U L AT I O N A N A LY S I S
(ii) star forming (blue);
(iii) transition region (purple); The spectrum of a galaxy contains a record of its SF history over the
(iv) Seyfert (green); age of the universe. Its stellar populations are the combination of
(v) LINER (red). all the episodes of SF that a galaxy and its progenitors underwent.
Reconstructing this SF history or at least recovering it partially is
Most objects in the quiescent early-type galaxy class are virtually a challenging task that requires excellent observational data, stellar
free of emission lines. A small subfraction exhibits weak [O III] population models for the interpretation and a statistically sound
emission, but lack the other emission lines necessary for the BPT method for the recovery.
classification. We verified that a further separation of this subclass For the problem at hand in this paper, we turn to the parametric
does not affect the results of this work. The quiescent objects make approach. It has been used with great success by Ferreras & Silk
up the bulk of the early-type population (82 per cent). The early-type (2000) and by Kaviraj et al. (2007) to use the immense discriminat-
galaxies with significant emission lines (S/N > 3 in the four main ing power of rest-frame NUV and FUV photometry to test whether
BPT lines) represent the remaining fifth of the early-type sample. individual early-type galaxies have undergone episodes of recent
We call these active early-type galaxies, and separate them into star residual SF or not.
forming, transition objects, Seyferts and LINERs. The results of
this classification are summarized in Table 1. For further details
we refer the reader to the paper describing the MOSES catalogue 3.1 Two-component star formation histories
(Schawinski et al., in preparation). We show some example images The aim of this work is the mapping of current SF/AGN activities in
of morphological early-type galaxies in Fig. 3. galaxies with their SF histories. Of particular relevance for us is the
epoch and the strength (hence mass fraction) of the last significant
starburst that occurred in a galaxy. Since the most recent episode
Table 1. Emission line classification results. of SF dominates the spectrum at all wavelengths, it is the easiest to
recover (see Fig. 4). Following the approach of Kaviraj et al. (2007),
Classification Number Fraction Colour
galaxies
we therefore fit two-component SF histories to the galaxy data.
The old component is modelled as an instantaneous starburst of
Early-type galaxies 15 729 100 per cent variable age and metallicity, and represents the formation of the bulk
Quiescent 12 828 81.5 per cent Orange stellar population of the galaxy before the most recent episode of SF.
Star forming 671 4.3 per cent Blue We allow this old component to vary between a to of 1 and 15 Gyr.
Transition region 1 087 6.9 per cent Purple By marginalizing over this parameter later on, we are summing all
Seyfert 242 1.5 per cent Green possible linear combinations of single starbursts and so take into
LINER 901 5.7 per cent Red account all possible SF histories before the most recent starburst.

C 2007 RAS, MNRAS 382, 1415–1431
Harris & Harris 2000, 2002; Sarajedini & Jablonka 2005). This
metallicity distribution is modelled with two Gaussians. The bulk
metallicity is a Gaussian with a mean metallicity μZ and a spread
of σ = 0.22 dex, while the tail is parametrized as a Gaussian shifted
by 0.75 dex to lower metallicity with the much larger spread of σ =
1.5 dex. Both components are such metallicity composites, and we
vary μZ to cover a mean mass-weighted metallicity Z(μZ ) ranging
between ∼1/200 and ∼3.5 Z .
3.3 Dust extinction

Internal dust extinction causes a degeneracy with age and mass that

needs to be taken into account. To describe dust attenuation we adopt
the extinction law by Calzetti et al. (2000). This choice is mainly
motivated by the fact that the active early-type galaxies in our sample
are undergoing significant episodes of SF, so that an extinction law
derived from a sample of star-forming galaxies seems reasonable.
Figure 4. The pace of stellar evolution: we show the SED of a solar metal- In separate runs we have tested a number of other extinction laws
licity M05 SSP from 0.01 Myr to 15 Gyr. The different ages are colour coded (Milky Way Allen 1976; LMC Fitzpatrick 1986; SMC Prevot et al.
and we indicate the filters we use by their filter transmission curves of the 1984; Bouchet et al. 1985) and verified that in general, the Calzetti
data we use in this paper at the bottom. The broad-band colours of a galaxy et al. (2000) law provides the best fits. The dust parameter E(B −
will always be dominated by the youngest stellar population formed in the V) is varied between 0.0 and 0.8.
most recent episode of SF. Any older population will be small by compari-
son even with respect to a relatively small young mass fraction. This feature
of stellar evolution allows us to effectively split the SF history into two 3.4 Combining photometry and spectroscopy
components until about 1.5–2 Gyr when the pace of stellar evolution slows
down and populations of different ages up to 15 Gyr become similar and the A crucial novelty of our approach is the combination of photometric
differences are on the order of the effects of dust and metallicity. and spectroscopic data, which increases the power of breaking the
degeneracies between age, mass, metallicity and dust content. The
reason why this helps is that the degeneracies are different because
On top of this, we add a young component of a certain age ty of the dissimilar non-linear responses of broad-band colours and
and mass fraction fy . It turns out that this young burst is poorly stellar absorption indices to these parameters. Like cosmologists
modelled by an instantaneous starburst, so that we choose to rep- who combine the very degenerate contours of one experiment with
resent it by an exponentially declining SF rate with an e-folding another to find a well-constrained set of parameters in the intersec-
time of 100 Myr. The choice of 100 Myr is motivated by a test tion, we harness two very different observables that couple to the
on a small subsample, where we left the e-folding time τ as an fundamental physical parameters in very different fashions.
additional free parameter. In this test, we recovered a typical τ of It should be noted that absorption line indices are not signifi-
∼100 Myr for early-type galaxies with young components, so we set cantly affected by dust (MacArthur 2005). A further cornerstone in
τ to 100 Myr a posteriori. The parameter ty stands for the look-back our fitting algorithm for the identification of intermediate-age popu-
time to the beginning of the starburst, fy is the total mass fraction lations is the inclusion of near-infrared (IR) photometry and stellar
formed in the recent burst extrapolating the 100 Myr e-folding SF population models (Maraston 2005) that adequately describe the
history to infinity. Since virtually all ages we recover for the starburst thermally pulsing asymptotic giant branch (TP-AGB) phase. This
population are at least ∼100 Myr, the difference between the mass is crucial as the TP-AGB provides most of the near-IR light of a
actually formed by today and the total mass that will form is small. stellar population between 0.5 and 2 Gyr (Maraston 1998).
This parametric quantification of the age and mass fraction of In a nutshell, we perform a comparison of the observed photomet-
the secondary population allows us to derive a recent SF history ric spectral energy distribution (SED) (of as many as are available
against the background of any possible SF histories in the distant of FUV, NUV, u, g, r, i, z, J, H, K) to a library of SEDs of over five
past, represented by a simple stellar population (SSP), whose effect million two-component SF histories from Maraston (2005) and com-
we can account for by marginalization. This approach is different pute the appropriate model SED parametrized by the five parameters
from purely non-parametric approaches in the literature that make described above (ty , fy , to , μz and E(B − V)) compute the associated
no a priori assumptions on the SF history to be recovered (e.g. Panter, χ 2 statistic to obtain a probability distribution function for those five
Heavens & Jimenez 2003; Panter et al. 2007). parameters. An intrinsic error for our photometry across all filters of
0.015 mag is adopted to deal with the uncertainties in the absolute
photometric calibration of SDSS.
3.2 Internal metallicity distribution
Simultaneously, we compute the associated luminosity-weighted
Besides assigning an average metallicity to each model, we addi- SSP ages and metallicities for each model. We derive luminosity-
tionally consider the internal distribution of metallicities necessarily weighted SSP ages and metallicities by χ 2 fitting the Thomas,
present in each galaxy. This is important, as a tail of low metallicity Maraston & Bender (2003) and Thomas, Maraston & Korn (2004)
stars can have a significant effect in particular on the ultraviolet part models to the 25 Lick absorption line indices and compute a second
of the spectrum (Maraston & Thomas 2000). We construct an ana- probability distribution through the comparison of these quantities
lytic form of the typical metallicity distribution from observations with the V–luminosity weighted ages and metallicities of the pho-
of the bulge of M31 and the closest early-type galaxy NGC 5128 tometric comparison. These two probability distributions from the
where individual stars can be resolved (Harris, Harris & Poole 1999; SED and absorption-line fitting are then convolved to the final one.

C 2007 RAS, MNRAS 382, 1415–1431
In order to estimate the parameters of interest, we marginal- in panel (d). This plot quantifies strength and epoch of most recent
ize the resulting five-dimensional probability distribution function episode of SF. In panels (e) and (f), we show the projected, pro-
to obtain the two-dimensional distribution in the recent starburst jected probability distributions for the parameters ty and fy . In the
age versus mass-fraction plane (ty –fy ). We integrate over this sur- main analysis we will stack these distributions for various classes
face to obtain confidence levels constraining the age and mass- of galaxies. In panels (g) and (h), we show dust extinction E(B −
fraction of the last significant episode of SF in the galaxy. These V) and the age of the old component.
probability distributions are stacked to get typical values for the The two example galaxies show very different SF histories. The
subpopulations. active galaxy has a very young luminosity-weighted age, which
is well consistent with the increased UV flux. The two-component
SED fit yields an old component of 3 Gyr and places the most recent
3.5 Example galaxies starburst at an age of 400 Myr with a mass fraction of 7 per cent. Its
In this section we present two example fits to galaxies with very dust content is relatively high (E(B − V) = 0.1) in agreement with

different properties and SF histories. The first galaxy is an active the current SF activity. The low luminosity-weighted age of the ‘old’
early type in the transition region of the emission-line classification component implies that this object had several rejuvenation events
which suggests that the object is dominated both by ongoing SF and in the past.
an AGN. Its Hα SF rate is 2.1 M yr−1 and its UV-optical colours The emission line free galaxy, instead, has an old luminosity-
place it squarely into the blue cloud (u − r = 2.3, NUV − r = 3.2). weighted age derived from the spectroscopic data. The two-
The second galaxy is a typical passive quiescent early-type galaxy component fit yields basically two old components with ages around
with no detected emission lines. Its UV-optical colours place it on 10 Gyr. We can conclude that this galaxy clearly has not undergone
the red sequence (u − r = 2.8, NUV − r = 6.1). The availabil- a significant starburst in the recent past, while the details of its SF
ity of GALEX, NUV and FUV photometry generally improves the history in the more distant past are not resolved. The 99 per cent
confidence contours, but does not shift the best fit. confidence intervals exclude any sort of significant SF activity in
The fit results are summarized in Figs 5 and 6. In panel (a) we the recent past. The very low dust attenuation value fits well into
show the fit to the photometry: in black are the observed broad-band this picture.
colours with the errors magnified by a factor of 3 to make them We also note that while only a fraction of the galaxies in our sam-
visible. To guide the eye, we show the filter transmission curves of ple have GALEX, their fitted SF histories do not differ significantly
the various filters used from GALEX far-UV to 2MASS K band. from those without.
The best-fitting model is shown in green and the separate old and
young components are red and blue, respectively. We also show the 4 RED SEQUENCE AND BLUE CLOUD
SDSS colour images.
In panels (b) and (c) we show the probability distributions In this section, we discuss the position of the various galaxy classes
for the luminosity-weighted age and metallicity derived from the defined through the emission-line diagnostics in the colour–mass
absorption-line indices. The final result in the ty –fy plane is shown relation. This simple comparison allows us to draw first tentative
Figure 5. An example of a galaxy where we can resolve and quantify the last significant episode of SF. This object is a star-forming early-type galaxy. For a
detailed description, see Section 3.5.

C 2007 RAS, MNRAS 382, 1415–1431

Figure 6. An example of a galaxy where we cannot quantify the first significant episode of SF, but can place strong limits on the recent SF history. For a
detailed description, see Section 3.5.
conclusions that are independent of our stellar population analysis in a green valley between the blue cloud and the red sequence has
method. Quantitative results of the detailed analysis will be pre- been discussed in other recent work in the literature (Faber et al.
sented in the subsequent section. 2007; Graves et al. 2007; Kauffmann et al. 2007; Martin et al. 2007;
Nandra et al. 2007; Salim et al. 2007; Wild et al. 2007).
4.1 Active galaxies in the colour–mass relationship

4.2 Mass dependence
In Fig. 7 we show u − r colour as a function of galaxy velocity
dispersion σ for the galaxies in our sample. The morphological This tentative sequence is convolved with a clear mass dependence
late-type galaxies are shown as grey points, the early-type galaxies of the emission-line classification. The blue star-forming objects are
are coloured points subdivided in the various emission-line classes: found at the low-mass end, while the majority of early-type galaxies
quiescent (i.e. no emission lines, orange), LINER (red), Seyfert classified as LINERs lie at the high-mass end on the red sequence.
(green), transition (purple) and star forming (blue). The quiescent Transition objects and Seyferts reside between these two extremes
early-type galaxies (see most right-hand panel) form the well-known also in their galaxy mass distributions.
red sequence, while the late-type galaxies in the sample define the The presence of a significant fraction of low-mass early-type
so-called blue cloud (grey points). galaxies showing signs of ongoing SF of a few M per year (see
The first key result of this work is that the active galaxies are off- Fig. 8) is a striking result, which is mostly the merit of the pure
set from the red sequence and exhibit well-defined patterns across morphological selection strategy. Still, this result is not surpris-
the colour–σ diagram. Star-forming objects inhabit the blue cloud. ing. The GALEX UV satellite has found evidence for minor re-
While this may seem like an obvious result, note that the broad-band cent SF in early-type galaxies (Yi et al. 2005), and the SAURON
SED and the emission lines probe very different physical compo- (Spectrographic Areal Unit for Research on Optical Nebulae) sur-
nents of a galaxy. While the former traces the stellar populations vey of 48 nearby elliptical and lenticular galaxies has shown that
and therefore the past, fossil record, the emission lines provide in- many of them have (often extended) emission lines and signs
formation on the current state of the ionized gas in the galaxy. It of recent SF as well as CO emissions from molecular gas and
is very reassuring that objects with ongoing SF show significant atomic hydrogen indicating sometimes substantial reservoirs of
rejuvenation also in their stellar populations. gas (Morganti et al. 2006; Sarzi et al. 2006; Combes, Young &
More importantly, those objects not dominated by SF but by AGN Bureau 2007).
activity, Seyferts and LINERs, are located considerably closer to In general, low-mass early-type galaxies are well known to show
and almost on the red sequence. Transition objects, instead, which young stellar ages, emission lines and even ongoing SF. This down-
are composites of SF and AGN activity are found right between sizing of galaxy formation, wherein the most massive galaxies
these two extremes. There is a sequence between the blue cloud harbour the oldest stellar populations is now generally accepted.
and the red sequence from star forming via transition region and Today’s early-type galaxies in particular show a clear correlation
AGN to quiescence. In more general terms, the presence of AGN between mass and age (Caldwell, Rose & Concannon 2003; Nelan

C 2007 RAS, MNRAS 382, 1415–1431

Figure 7. The colour–σ relation for our sample. In each panel, morphological late-types are grey, quiescent early-types are orange and the various active
early-types are blue (star forming), purple (transition region), green (Seyfert) and red (LINER). The mass tracer is the stellar velocity dispersion.
et al. 2005; Thomas et al. 2005; Bernardi et al. 2006; Jimenez et al. the blue star forming and the red sequence objects. This indicates
2006). that AGN may have a decisive role to play in this transformation
Fig. 7 shows that our sample comprises a large, statistical sub- process. Fig. 7 suggests that residual SF and nuclear activity are
sample of active early-type galaxies in the intermediate-mass and connected in early-type galaxies. It is possible that the AGN actu-
low-mass regimes (σ 200 km s−1 ) allowing us to study the in- ally suppress the SF activity, and we are witnessing AGN feedback
terplay between SF, nuclear activity and quiescence in early-type in action. Our sample represents the ideal laboratory to probe these
galaxies. We note that many of the high-mass transition region ob- processes.
jects are most likely misclassified LINERs. They do not have any It should be noticed that the massive early-type galaxies in our
star forming or Seyfert counterparts at the same mass. sample do not appear to be part of such transformation process. At
the high-mass end only quiescent and LINER objects are found.
4.3 AGN feedback in action? What needs to be shown is twofold. First, solid evidence for the
At low and intermediate masses we can identify a sequence at a given presence of an evolutionary sequence for the low/intermediate-mass
mass from SF via transition region and AGN activity to quiescence, objects is required. To this end, we have derived detailed SF histories
which is echoed in a sequence from blue to red stellar populations. using the method presented in Section 3 focusing on the epochs and
Taking u − r colour as a proxy for the stellar age and hence using the mass fractions of the most recent starburst. This allows us to inves-
stars as cosmic clock, this is highly suggestive of an evolutionary tigate whether the galaxies of the various subclasses in our sample
sequence. We are observing the act of galaxy transformation from truly belong to the same object type at different evolutionary stages.
the star forming blue cloud via the green valley to the passive red Second, the interaction between SF and nuclear activity along this
sequence in action. sequence must be investigated, with the aim to understand whether
We interpret the deviation from the red sequence as rejuvenation and how the two processes are related. These are addressed in the
due to a recent episode of SF. On their way back to the red sequence, following sections.
galaxies appear to pass through a phase of significant nuclear activ-
ity, as transition region objects and Seyferts have colours between 5 T H E E VO L U T I O N A RY S E Q U E N C E
Following the discussion of the previous section we divide our sam-
ple in three velocity dispersion bins (σ 100 km s−1 , 100 < σ
200 km s−1 , σ > 200 km s−1 ) representing low-mass, intermediate-
mass and massive early-type galaxies. Within each σ bin we main-
tain the five subclasses: star forming, transition object, Seyfert,
LINER and quiescent. Within each of the 15 resulting subclasses
we stack the probability distributions for the SF histories obtained
in the analysis of Section 3. We marginalize over all parameters ex-
cept the mass fraction fy and age ty of the starburst component. The
resulting contours of the probability distributions in the ty –fy plane
are shown in Fig. 9. The panels are analogues to panel (d) of Figs 5
and 6. The five columns show the various emission-line classes, the
rows are the three σ bins. A bar in each panel indicates the relative
fraction of objects in this particular subclass.
5.1 Evidence for the time sequence

Figure 8. Histogram of the SF rate and specific SF rate of the star-forming
early-type galaxies. The SF rate is based on the luminosity of the Hα line The results shown in Fig. 9 allow one immediate conclusion. The
and is converted to a rate using the relation of Kennicutt (1998). SF histories derived for the low-mass and intermediate-mass objects

C 2007 RAS, MNRAS 382, 1415–1431

Figure 9. The recent SF histories of early-type galaxies by mass and emission line class. In the horizontal direction, we split our sample into three velocity
dispersion bins. Vertically, we split the galaxies in each σ bin by their activity type from star forming to quiescent (cf. Fig. 7). The contours are 25 and
50 per cent probability contours, respectively. Note that the area of these contours can be misleading; only a small fraction of early-type galaxies is in classes
other than the quiescent bin. In order to guide the eye, we indicate the fraction of galaxies in each mass bin (low, intermediate and high) that are in each emission
line class. For example, in the high-mass regime, only a tiny fraction of early-types are classified as star forming (most likely misclassifications), while the
almost 90 per cent are quiescent; nevertheless, the contours occupy similar areas on the ty –fy plane.
in our sample galaxies reveal clear evidence for the occurrence of a diate mass bin containing galaxies with 100 σ < 200 km s−1 . In
starburst between 100 Myr and 1 Gyr in the past involving a typical the highest bin for objects with velocity dispersion above 200 km s−1
mass fraction of 1 to 10 per cent. This reinforces the interpretation the evolutionary sequence is not detected. At the high-mass end, the
of deviations from the red sequence in the colour–σ diagram (see vast majority of galaxies is quiescent, and no recent starburst within
Fig. 7) to be caused by recent rejuvenation events. Key is the epochs the past few Gyrs has occurred. The small fraction of active galaxies
and mass fractions of this rejuvenation for the various emission-line is by far dominated by LINER-like emission, with some transition
classes. While the starburst ages clearly increase from star forming region objects that are in fact closer to the properties of LINERs
via AGN dominated to quiescent in a continuous sequence from (see Section 6.5.3).
∼100 Myr to beyond 1 Gyr, the mass fractions remain the same.
This points to the various emission line classes having undergone
the equivalent rejuvenation event in the past, and the only feature
5.2 Time-scales along the sequence
differentiating the various classes is the time elapsed since that SF
episode. Hence, the objects of the various emission classes in the For the low- and intermediate-mass objects (σ < 200 km s−1 ) we
panels along one row in Fig. 9 are identical in terms of mass, mor- can now determine the relative and absolute timing of this transfor-
phology and SF history, except that the epoch of the last starburst mation process. We marginalize over the mass fraction and compute
varies. This means we have resolved an evolutionary sequence, and the stacked probability distributions of starburst ages for the various
what we are seeing is the process of rejuvenation in early-type galax- emission-line classes. The result is shown in Fig. 10 for the low-
ies at different evolutionary stages. This time sequence includes a and intermediate-mass early-type galaxies in the left- and right-
phase of AGN activity along the way. hand panels, while in Table 2, we give the figures in Myrs. In both
The fraction of objects taking part in this transformation is mass bins, we can see the shift in the age of the last episode of
strongly dependent on mass. In the lowest mass bin about one third SF in a sequence as could already be seen from Fig. 9. The star-
of the objects shows emission lines and is caught in the process of forming objects are clearly the ones most recently rejuvenated with
this transformation. This fraction drops to 13 per cent in the interme- starburst ages around 150–300 Myr, with the latter increasing along

C 2007 RAS, MNRAS 382, 1415–1431

Figure 10. The time sequence in the low- and intermediate-mass bin: in the top of each panel, we show the stacked, normalized probability distribution
functions for the time elapsed since the start of the starburst ty . In the bottom, we indicate where on the ty axis 50 per cent of the probability lie. We give the
numbers for this in Table 2.
Table 2. The time sequence: the age of the last episode of SF as a function of mass and emission line class.
Mass regime Velocity dispersion Star formation Transition region Seyfert AGN LINERs Quiescent
(km s−1 ) (Myr) (Myr) (Myr) (Myr) (Myr)
Low mass 0–100 145–280 195–550 270–800 470–1800 550–2700

Intermediate mass 100–200 140–350 260–1200 350–1800 700–4000 950–4000
Note. The time intervals given in this table correspond to the 50 per cent probability intervals from Fig. 10.
the sequence from transition region through Seyfert and LINER to

5.3 Lifetimes and number fractions
quiescent.
A global interpretation yields the following picture: the most com- An important consistency check is the comparison of the lengths of
mon starburst age ty for the transition region objects is around 300 to the various phases as derived from the stellar population ages with
500 Myr. Seyfert activity appears further 100 to 300 Myr later with the number of objects found in each phase. If the transition region
starburst ages peaking at about 600 Myr. This phase is followed by and Seyfert phases are equally long, the number of objects found in
LINER and quiescent with separations of a few 100 Myr, respec- these two classes must be the same. From Fig. 9 it can be seen that
tively. The widths of the distributions increase along this sequence this is not the case. We observe three to four times more transition
from about 200 Myr for the star-forming objects to 400–600 Myr region objects than Seyferts. However, an unknown fraction of the
in the transition region and Seyfert objects. The phases involving objects in the Seyfert phase must be Seyfert type I rather than type
nuclear activity, transition and Seyfert, appear to be of comparable II depending on the viewing angle from which the nucleus is seen.
length in time and must be a factor of 2 to 3 longer than the star Indeed, broad emission line objects like type I Seyferts are not clas-
forming one. Note that the widths of LINERs and quiescent objects sified as ‘galaxies’ by the SDSS spectroscopic pipeline and hence
are much less well determined as they extend into the regime of are not included in our sample. Thus, it is plausible that the osten-
old ages beyond 1 Gyr, in which the age resolution is significantly sible lack of Seyfert AGN in our sample stems from this geometric
reduced. effect, which puts object numbers and phase lifetimes in very good
The transition region and the Seyfert phases seem to overlap con- agreement. It should also be noted that our sample is not volume
siderably, which suggests that objects at this point along the time limited.
sequence may switch back and forth between transition region (a
mixture of AGN and SF) and Seyfert phase (pure AGN). If this
is the case, the Seyfert phase represents episodes of enhanced ac- 5.4 Dependence on mass
cretion on to the black hole between phases with lower levels of The comparison between the lifetimes of the star forming and AGN
nuclear activity together with low residual SF. The SF we see in the phases is more complex, as an additional mass dependence comes
transition region objects may even be induced by the (potentially) into play. The fractions of the active early-type galaxy population in
recurrent Seyfert state as the jets and outflows pile up gas before a given mass bin that are classified as transition region and Seyfert
finally pushing it out (Silk & Rees 1998; Silk 2005). objects decrease by about a factor of 2.5 from the low- to the
Toward the end of the sequence, the transition from the high- intermediate-mass bin, while the number of star-forming objects
ionization Seyfert phase to quiescence passes through a phase of decreases by a factor of 7. This results in a factor of 3.5 more tran-
nuclear activity characterized by low ionization (LINER). In total, sition region objects than star forming ones at intermediate masses,
the transformation process from blue cloud to red sequence passing while their fractions are comparable at low masses. As the tran-
through nuclear activity takes about 800 Myr to 1 Gyr. sition region phase lasts about two to three times longer than the

C 2007 RAS, MNRAS 382, 1415–1431
star forming one in both mass regimes this suggests that not all As ought to be expected, dust content decreases along the sequence
low-mass objects go through the time sequence. Hence SF is not discussed above with highest dust attenuation in the star-forming
followed by nuclear activity in all cases, and the true proportions objects (E(B − V) ∼ 0.1) and the lowest in the quiescent ones
between star forming and transition region objects for the SF–AGN (E(B − V) 0.05). This internal consistency further strengthens
sequence may be observed only at intermediate mass. This implies the conclusion that we are indeed observing a time sequence and
the presence of a tail of young quiescent objects in the low-mass are witnessing the transformation of early-type galaxies from SF
bin, which is indeed indicated by the probability distributions in via AGN to quiescence. In order to mimic a time sequence in the
Fig. 10. observed u − r colours with a variation in the amount of dust rather
Still, the ratio transition region to star-forming objects seems than in age, the trend in E(B − V) would have to be the opposite of
slightly too high at intermediate masses. The reason for this may what we observe.
be that some fraction of transition objects are misclassified and are
actually LINERs (see Section 6.5.3). This would also explain why
6 DISCUSSION

transition objects and Seyfert appear less well separated at interme-
diate mass. With Fig. 7 in Section 4 we show that early-type galaxies of dif-
Within these uncertainties, the fraction of galaxies that experi- ferent emission-line classes occupy different loci in the colour–σ
ences residual SF, and transitions from starburst to quiescence via diagram. Star forming early types are the bluest deviating most from
nuclear activity drops by about a factor of 2 to 4 between the low- the red sequence. Once some AGN contribution appears in the op-
mass and the intermediate-mass bins. The highest mass bin (σ > tical emission lines, the colours dramatically redden. In Section 5,
200 km s−1 ) concludes this trend, as no such sequence is detected for we demonstrate quantitatively that there is indeed strong evidence
these masses. This is very well supported by the fact that the appar- for the existence of an evolutionary sequence of early-type galax-
ent dichotomy in the starburst age distribution within the quiescent ies moving from the blue cloud to the red sequence (Salim et al.
early-types galaxies (left-hand panel of Fig. 10) is considerably 2007). We show that on this path the galaxies go through a number
weaker at intermediate masses (right-hand panel), and disappears of phases starting at SF followed by a transition region of SF and
entirely at the high-mass end (see Fig. 9). AGN, a Seyfert AGN phase and a LINER phase before finally set-
Finally, it is interesting to note that, while the vast majority of tling to total quiescence and passive evolution on the red sequence.
LINERs detected in the low-mass bin appears to take part in the This process is currently building the low-mass end of the red se-
evolutionary sequence, the presence of a second independent LINER quence in the low redshift universe. In this section we discuss the
population seems to emerge at intermediate masses. Fig. 10 reveals astrophysical implications of this finding, the role of LINER-like
a clear dichotomy with a fair fraction of LINER-like objects that early-type galaxies, possible trigger mechanisms of this sequence
does not show signs of recent rejuvenation. This trend becomes and follow-up observational campaigns in the context of the time
even more evident at the highest masses, where the largest fraction sequence we have derived.
of LINERs is detected (8 per cent), which contain only old stellar
populations (see Fig. 9). 6.1 The starburst–AGN sequence
A clear implication of our results is the existence of a link between
5.5 Dust properties along the sequence residual starburst and nuclear activity at recent epochs in low- and
intermediate-mass early-type galaxies. Whatever triggers the star-
Fig. 11 shows the stacked, normalized probability distribution func-
burst, it leads also to an AGN, and the star-forming phase reaches
tions for the dust parameter E(B − V) obtained through our SED
quiescence via an AGN phase. We see the starburst fading while
fitting algorithm for the low- and intermediate-mass bins. The dif-
the AGN rises. This leads to two possible scenarios depending on
ferent line styles are the various emission-line classes (see labels).
whether the two processes, SF and AGN, are connected.
6.1.1 AGN feedback or black hole growth in action?

If the fading of SF and nuclear activity are linked, then we are
observing AGN feedback in action. The appearance of the AGN in
the transition region and Seyfert phases are concurrent with the end
of SF, so it is plausible that these two phenomena are connected such
that the rise of the AGN suppresses and terminates SF by heating
and expulsion of the remaining gas.
Alternatively, it is possible that starburst and AGN actually do not
interact with each other, even though they may be triggered by the
same global event. In this case SF fades simply owing to supernova
feedback and/or the exhaustion of cold gas. Most interesting in the
framework of this scenario is our finding that the AGN rises while SF
declines. This shows that in this case AGN activity must be delayed
substantially with respect to the starburst by several hundred Myr
Figure 11. This plot shows the stacked, normalized probability distribution
and may actually be triggered by SF. This delay might be the result
functions for the dust parameter E(B − V) for the low- and medium-mass bin. of slow AGB winds fuelling the black hole (Davies et al. 2007).
Star-forming early-types are the dustiest with the transition region, Seyfert This implies that the early-type galaxy or its progenitor hosting the
early-types being less dusty and the LINERs and quiescent early-types con- starburst is just in the process of building up its central supermassive
taining even less dust. black hole.

C 2007 RAS, MNRAS 382, 1415–1431
6.1.2 Gas depletion time-scale: in favour of AGN feedback our selection criteria, hence they most likely do not have early-type
morphology.
In order to distinguish these two scenarios, we compare the ob-
served time-scale for the fading of SF with simple theoretical
estimates based on the Schmidt–Kennicutt law (Schmidt 1959;
6.3.1 Prodigious star formation in the recent past
Kennicutt 1998). The latter relates the SF rate with the ratio of gas
mass surface density and the dynamical time-scale of the system As we model the starburst as composite stellar population with ex-
scaled by an efficiency . We assume that the dynamical time-scale ponentially declining SF histories, these progenitors must have had
of SF is on typical galaxy time-scales of ∼0.1–0.5 Gyr, as observed significantly higher SF rates in the past. The typical star-forming
in gas discs in elliptical galaxies (Young 2002). We can exclude early-type galaxy has a mass of about 5 × 1010 M and has formed
shorter dynamical time-scales caused by a concentration of the star- about 5 per cent of its stellar mass in the current burst (i.e. 2.5 ×
burst activity to small radii, as the mean petrosian half-light radii of 109 M ) over the last 100 Myr. This corresponds to an average SF
the u-band light in the star-forming early-types (2.25 arcsec) imply rate of 25 M yr−1 , resulting in an infrared luminosity LIR of 1.25 ×

that SF in these objects is extended over several kpc. Then to deplete 1011 L , well above the threshold for so-called luminous infrared
any gas reservoir would take on the order of many Gyr. galaxies (LIRGs). The SF rate was undoubtedly higher in the past
This time-scale is significantly longer than the fading time-scale as the best-fitting decay time-scale τ is around 100 Myr. Such an
deduced here for our sample galaxies. In our modelling analysis exponentially decaying SF rate can easily boost the past LIR into the
we find that the recent rejuvenation event is best modelled by a category of ultraluminous infrared galaxies (ULIRGs).
exponentially declining starburst with e-folding time-scale around We find that the most luminous star-forming early-type galaxy de-
100 Myr. Thus, the gas reservoir cannot get depleted solely via self- tected by the Spitzer Space Telescope in the SDSS–SWIRE overlap
regulated SF. An additional heating source is required. The fact that has LFIR = 5.7–7.4 × 1010 L (Sanders & Mirabel 1996) implying a
we find an AGN phase subsequent to the starburst strongly suggests SF rate of 13 M yr−1 based on the relationship of Gao & Solomon
that the AGN is (at least partly) responsible for the suppression of (2004). This galaxy is thus just slightly below the cut defining a
SF by depleting the gas reservoir accelerating the transition from LIRG of LIR = 1011 L at the time it is observed.
the blue cloud to the red sequence. A more detailed investigation of We conclude that the progenitors of the objects we observe on the
this process is beyond the scope of this paper. It will be explored in SF–AGN sequence were most likely ULIRGs. This fits well with the
more detail in a forthcoming work (Kaviraj et al., in preparation). evidence that the local ULIRG population most likely will evolve
This argument strongly supports the AGN feedback interpretation into intermediate-mass early-type galaxies (Genzel et al. 2001).
of the time sequence. This result has important consequences for Dasyra et al. (2006a,b) and Veilleux et al. (2006) have analysed
our understanding of galaxy formation. It is empirical proof that Hubble Space Telescope (HST) H-band imaging and Very Large
AGN feedback, the existence of which is now postulated in most Telescope (VLT) spectra of low redshift ULRIGs and quasi-stellar
recent models of hierarchical galaxy formation (see Introduction objects (QSOs) and concluded that while most of them are ongoing
and references therein), occurs in early-type galaxies at late stages mergers, their kinematic, structural and photometric properties are
of their evolution. Beyond this important conclusion, our sample consistent with them becoming elliptical galaxies and settling on the
that is resolving the actual sequence of AGN feedback provides fundamental plane. Most relevant here is that they find their stellar
further insight into this process. We find its overall time-scale to velocity dispersions to have typical values of around 150 km s−1 in
be of the order of a Gyr. Moreover, we observe a substantial delay overlap with the typical velocity dispersions of our active galax-
between fading of SF and rise of the AGN by a few hundred Myr, ies that mark the transformation process from blue cloud to red
which might reflect the time-scale required for the black hole to build sequence.
up the accretion disc. This implies that AGN feedback is not well The appearance of an AGN phase several hundred Myr after the
approximated with an immediate truncation as implicitly assumed U/LIRG phase may explain why most local ULIRGs do not appear
in current hydrodynamical simulations (Springel et al. 2005a). to contain a luminous AGN (Genzel et al. 1998).
6.2 Obscured AGN? 6.3.2 Merger as main trigger

A key result is that we see the AGN rising while SF fades. As The origin of ULIRG activity is thought to be a galaxy merger
absolute O[ III] luminosities increase along the sequence, it can be (e.g. Genzel et al. 2001). This implies that we are possibly see-
excluded that the AGN is simply out-shined by the starburst during ing a slightly more advanced stage of a gas-rich merger where the
the star-forming phase. However, it is still possible that the AGN morphology has already settled into a spheroid, and the infrared
is not observed during the starburst, because it is highly obscured luminosity has already decayed below the LIRG threshold.
in the beginning. In this case, SF and AGN would co-exist from Because of our sample selection, all objects discussed here are
the beginning, and while SF fades, the AGN remains active. This morphologically early-type galaxies without spiral or disc structure
does not invalidate but rather strengthen the interpretation of AGN as far as detectable in the SDSS images. Elliptically shaped galaxies
feedback occurring in our sample galaxies. with minor morphological perturbations such as tidal features or dust
lanes are not excluded. This selection should in principle enable us
to investigate the role of morphological disturbance, hence merger
6.3 Progenitors and trigger mechanisms
and galaxy interaction activity, along the sequence. However, the
The youngest early-type galaxies on the evolutionary sequence we fraction of disturbed early-type galaxies in our sample is quite small.
resolve are the star forming ones with starburst ages around 100– We have not found any compelling trend between morphological
200 Myr. The progenitors of these objects within the sequence must irregularities and SF/nuclear activity, most likely because SDSS
have stellar populations younger than 200 Myr. The fact that we do imaging does not provide the necessary sensitivity in terms of depth
not observe these objects implies that they are excluded through and spatial resolution. Deeper imaging, preferably at higher spatial

C 2007 RAS, MNRAS 382, 1415–1431

Figure 12. In the left-hand part of this figure, we show the histograms for the L[O III] λ5007 luminosities of our AGN and transition region objects. [O III] is a
tracer of the accretion rate (Heckman et al. 2004) and so is an upper limit on the accretion rate of the transition region objects. In the left-hand part, we show the
histogram of the quantity L[O III]/σ 4 , which is proportional to the Eddington ratio (Kewley et al. 2006). In each part, we go along the sequence from early times
to late. First, we show the transition region objects, then the Seyferts and then the LINERs. We split the LINERs into two; while the low- and intermediate-mass
LINERs and the high-mass LINERs have similar [O III] luminosities, their Eddington ratios are very different. The low/intermediate-mass LINERs form the
low luminosity and high Eddington ratio end of the Seyfert distribution, as would be expected from their place in the time sequence. The high-mass LINERs
have Eddington ratios an order of magnitude below their lower mass counterparts. Along the sequence, both absolute L[O III] and Eddington ratio rise, peak
and then decline over the time domain covered by the time sequence.
resolution, is clearly needed to identify signatures of recent merger Eddington ratios, instead, might be responsible (or at least part of)
activity in our sample objects (see e.g. van Dokkum 2005; Bell et al. the rigorous AGN feedback process in massive early-type galax-
2006). ies that avoid rejuvenation in the first place. Fig. 12 illustrates the
extended rise and decline in both accretion rate and Eddington ra-
tio over the time domain covered by the time sequence. While the
6.4 Eddington ratios along the sequence transition region to Seyfert evolution brings with it an increase in
accretion, the progression to a LINER is accompanied by a decline.
In the following we discuss the nature of accretion on to the super-
massive black hole along the AGN feedback sequence. In Fig. 12,
we show the distributions of [O III] luminosity and the quantity 6.5 The role of LINERs
L[O III]/σ 4 for Seyfert, low-mass and high-mass LINERs. L[O III]/σ 4
has been introduced by Kewley et al. (2006) as a tracer of the Ed- Fig. 9 shows that the fraction of LINERs observed increases with
dington ratio assuming that [O III] luminosity scales with the AGN increasing galaxy mass, while the fraction of galaxies undergoing
bolometric luminosity (Heckman et al. 2004), and that σ 4 provides the SF–AGN sequence decreases. Hence, LINERs appear to come
an estimate of the central supermassive black hole mass (Ferrarese & in two flavours. Some are part of the evolutionary sequence of low-
Merritt 2000; Gebhardt et al. 2000). Based on this quantity Kewley and intermediate-mass galaxies, but the majority is found in the
et al. (2006) show that the fundamental difference between Seyferts massive, quiescent early-type galaxies of our sample.
and LINERs in the SDSS data base is the accretion efficiency, with It should be emphasized that the LINER-like emission observed
the LINERs being at lower Eddington ratios than the Seyferts. Since here is most likely caused by nuclear activity. The extended LINER-
the transition region objects [O III] line luminosities have some con- like emission not associated with an AGN found by Sarzi et al.
tribution from SF, we can treat them as upper limits on both the (2006) in very nearby early-type galaxies is significantly weaker
accretion rate and efficiency. (typically with total L[O III] <106 L ; cf. Fig. 12) and would remain
We find that the Seyfert early types in our sample are high- undetected in our sample.
luminosity AGN with high L[O III] luminosities and high L[O III]/σ 4
ratios, indicating that they are strongly accreting systems relatively
6.5.1 LINERs in low and intermediate masses
close to the Eddington limit. The LINERs, instead, have lower
L[O III] luminosities and lower Eddington ratios. There is an in- As discussed in the previous sections, the early-type galaxies with
teresting trend with mass, however. The massive LINERs exhibit LINER emission appear on the evolutionary sequence of low- and
the lowest accretion rates, while the low- and intermediate-mass intermediate-mass objects between Seyfert-like nuclear activity and
objects are found between massive LINERs and Seyferts. Hence quiescence. This means that early-type galaxies transition from a
our low-mass and high-mass LINER early-type galaxies are objects high-accretion regime characterized by Seyfert emission to a low
in different states of nuclear activity. The former represent probably accretion regime characterized by LINER emission (Kewley et al.
the end stage of the SF–AGN sequence. If the Seyfert phase is a 2006) as they proceed to fade on to the red sequence. This can be
high-accretion phase during which the AGN feedback process oc- interpreted as the AGN running out of fuel. This is well consistent
curs, then the transition to the LINER phase is the point at which with the study of Yan et al. (2006), who find that many post-starburst
the supply of gas has been largely used up by SF and/or accretion galaxies show strong [O II] emission, but no Hα, indicating no fur-
on to the black hole. The massive LINERs characterized by low ther ongoing SF, which they associate with LINERs.

C 2007 RAS, MNRAS 382, 1415–1431
can disentangle contributions from polycyclic aromatic hydrocar-
6.5.2 LINERs in massive galaxies bon (PAH) lines to any hotter dust components due to an AGN.
Already in the intermediate-mass bin, most LINERs appear to be Finally, deep optical imaging would be very valuable and help to
associated with stellar quiescence (Fig. 9), and none of the LINERs determine the role of mergers and interactions along the AGN feed-
in the highest mass bin are connected with rejuvenation. This might back sequence.
be AGN feedback in its most rigorous form and is consistent with Observations at the observatory of the Institut de Radioas-
the fact that the efficiency of AGN feedback must increase with in- tronomie Milimetrique (IRAM) and at the Very Large Array (VLA)
creasing galaxy mass in the models in order to reproduce observed to address part of these questions have been taken in 2007 June and
UV colours (Schawinski et al. 2006a). While late SF seems to be July.
terminated by nuclear activity in low-mass early-type galaxies, the
AGN in massive galaxies may prevent SF in the first place. These 7 CONCLUSIONS
massive galaxies most likely have experienced strong AGN feed-

back at high accretion nuclear activity at earlier epochs. Indeed, We study the relationship between SF and nuclear activity in early-
observational evidence for the existence of this process has recently type galaxies at recent epochs. The principle aim is to test whether
been found (Daddi et al. 2007; Kriek et al. 2007; Rodighiero et al. AGN feedback occurs in these objects, as invoked in most recent
2007a). models of hierarchical galaxy formation.
This LINER-like activity may play the crucial role in maintaining For this purpose, we analyse a magnitude-limited sample of
an atmosphere of hot X-ray emitting gas which can shield the galaxy 48 023 galaxies drawn from the SDSS DR4 in the redshift range
from gas accretion and cooling and so prevent any further SF as of 0.05 to 0.1 limited to r < 16.8. We visually inspected all galaxies
suggested by Binney (2004). Such hot gas may then trigger extended in this sample identifying 15 729 morphological early-type galax-
LINER-like emission via thermal conduction of self-irradiation as ies. This paper is the first of a project called MOSES. A detailed
has been seen in massive nearby elliptical galaxies (see e.g. Sarzi description of the catalogue can be found in a companion paper
et al. 2007 and references therein), although it is likely that the (Schawinski et al., in preparation). Most importantly, the MOSES
majority of the LINER emission in our sample is associated with sample is not biased against SF and nuclear activity thanks to the
(more luminous) nuclear activity. purely morphological selection criterion. The SDSS photometry is
matched to a number of other surveys, namely GALEX, 2MASS and
SWIRE, yielding a wavelength coverage from the far-UV to mid-IR.
6.5.3 Misclassified LINERs Re-analysing the SDSS spectra, we measure emission-line fluxes,
stellar absorption line indices and velocity dispersions. The contri-
We note that some transition region early types are found in the butions of the stellar continuum and of the ionized-gas emission to
highest mass bin, which show no evidence for a SF episode (see the observed galaxy spectrum are separated by fitting simultane-
Fig. 9). These transition region objects, being the result of a combi- ously stellar population templates.
nation of SF and AGN activity, have no corresponding star forming 18.4 per cent of the early-type galaxies in our sample show
or Seyferts at the same mass and do not seem to take part in the emission lines, and we employ emission-line ratio (BPT) diagrams
AGN–SF time sequence. Instead, their properties are far more sim- to classify them into star forming, AGN–SF composites, Seyfert
ilar to those of high-mass LINERs. The most likely explanation AGN and LINERs. The fraction of emission-line galaxies increases
for this apparent contradiction is that the extreme starburst line of with decreasing galaxy mass to 40 per cent at the low-mass end
Kewley et al. (2001) led to the misclassification of some LINERs (σ < 100 km s−1 ). The emission-line classes are roughly evenly
as transition region objects. An improved classification scheme in- distributed between SF and AGN at intermediate (100 < σ
cluding the low-ionization lines may help here but is beyond the 200 km s−1 ) and low (σ 100 km s−1 ) masses. At high masses (σ >
scope of this paper. 200 km s−1 ), instead, only LINER-like emission is detected. The Hα
SF rates of those objects classified as currently star-forming range
from 0.1 to 25 M yr−1 .
6.6 Follow-up observations
We find that the objects with emission are offset from the red
The sample presented in this paper is the ideal laboratory for sequence and form a well-defined pattern in the colour–σ diagram.
more detailed follow-up studies of the link between SF and nu- Star-forming early-types inhabit the blue cloud, while early-types
clear activity in early-type galaxies. Multiwavelength data will with AGN (Seyfert and LINER) are located considerably closer to
tighten constraints on the formation histories, the SF activities, dust and almost on the red sequence. SF–AGN composites are found
contents, central black hole activities and accretion rates of our right between these two extremes. There is a sequence between the
sample galaxies. This will provide deeper insight into the physics blue cloud and the red sequence from star forming via transition
involved. region and Seyfert AGN and LINER to quiescence.
By observing CO molecular transitions the fate of the molecular To analyse this transition in more detail, we have developed a
gas can be explored. Such observations clarify whether it is really method for deriving SF histories with particular focus on quantify-
the feedback from AGN that quenches SF, or whether this decline ing the epoch and mass fraction of the last significant episode of SF.
in SF is caused by the fuel supply being exhausted by SF itself. Novel about our approach is to use a combination of UV-optical-
High-resolution radio-continuum observations at different frequen- NIR photometry and stellar absorption indices as observational
cies provide important clues on the relative contributions of SF and constraints.
the AGN along the sequence. The relative importance and interac- We find that low- and intermediate-mass early-type galaxies with
tion between AGN and SF in the galaxies along our timeline can be emission lines have experienced an episode of SF less than ∼1 Gyr
further addressed by the analysis of the full SED in the mid-infrared ago involving 1–10 per cent of their total stellar mass. This re-
wavelength regime. Mid-infrared imaging helps to pin down the cent starburst is best modelled by a steeply declining exponential
relative amounts of cold, warm and hot dust, while spectroscopy with e-folding time-scale around 100 Myr. Most importantly, this

C 2007 RAS, MNRAS 382, 1415–1431
time-scale and the mass fraction involved is the same for all the emis- University, the United States Naval Observatory and the University
sion line early-type galaxies whether they are star forming, transition of Washington.
region, Seyferts or LINERs. They are offset in time, however, with This publication makes use of data products from the 2MASS,
the star forming being the youngest, and the Seyfert/LINERs being which is a joint project of the University of Massachusetts and the In-
the oldest. Quiescent galaxies (objects without emission lines) have frared Processing and Analysis Centre/California Institute of Tech-
starburst ages larger than 1 Gyr, which sets the time-scale for the nology, funded by the National Aeronautics and Space Administra-
transition process we observe. The duration of the various phases tion and the National Science Foundation.
along this sequence agree reasonably well with the number fractions
observed, once geometric effects of Seyfert activity are taken into
account. Dust content decreases along the sequence with highest
REFERENCES
dust attenuation in the star-forming objects (E(B − V) ∼ 0.1) and
the lowest in the quiescent ones (E(B − V) 0.05). Adelman-McCarthy J. K. et al., 2006, ApJS, 162, 38

These results lead to the conclusion that we are observing an evo- Allen C. W., 1976, Astrophysical Quantities, 3rd edn. Athlone Press, London
lutionary sequence and are witnessing the transformation of early- Baldwin J. A., Phillips M. M., Terlevich R., 1981, PASP, 93, 5 (BPT)
type galaxies from SF via AGN to quiescence. Along this sequence, Bell E. F. et al., 2006, ApJ, 640, 241
nuclear activity rises while SF fades. We discuss that gas exhaustion Bennett C. L. et al., 2003, ApJS, 148, 1
Benson A. J., Bower R. G., Frenk C. S., Lacey C. G., Baugh C. M., Cole S.,
or self-regulation of SF alone does not suffice to explain this fading,
2003, ApJ, 599, 38
which leads to the conclusion that we have detected a sequence in Bernardi M. et al., 2003a, AJ, 125, 1817
which nuclear activity suppresses SF. We are seeing AGN feedback Bernardi M. et al., 2003b, AJ, 125, 1849
in action, while low- and intermediate-mass early-type galaxies tran- Bernardi M. et al., 2003c, AJ, 125, 1866
sition from the blue cloud through the green valley, characterized by Bernardi M. et al., 2003d, AJ, 125, 1882
nuclear activity, on to the red sequence. The high-mass galaxies in Bernardi M., Nichol R. C., Sheth R. K., Miller C. J., Brinkmann J., 2006,
our sample, instead, did not go through this transition in the recent AJ, 131, 1288
past. These objects have only LINER-like emission and are not as- Best P. N., Kauffmann G., Heckman T. M., Brinchmann J., Charlot S., Ivezić
sociated with rejuvenation. This indicates that the AGN in massive Ž., White S. D. M., 2005, MNRAS, 362, 25
galaxies may prevent SF in the first place, which would be AGN Best P. N., Kaiser C. R., Heckman T. M., Kauffmann G., 2006, MNRAS,
368, L67
feedback in its most rigorous form.
Binney J., 2004, MNRAS, 347, 1093
To conclude, we have found empirical evidence for the occur- Birnboim Y., Dekel A., Neistein E., 2007, MNRAS, 380, 339
rence of AGN feedback in early-type galaxies at recent epochs. The Blanton M. R., Lin H., Lupton R. H., Maley F. M., Young N., Zehavi I.,
galaxy sample presented here is the ideal laboratory for testing more Loveday J., 2003a, AJ, 125, 2276
detailed physics of this process. A first important hint from this work Blanton M. R. et al., 2003b, AJ, 125, 2348
is that there is a substantial delay between fading of SF and rise of Blanton M. R. et al., 2005, AJ, 129, 2562
the AGN by a few hundred Myr. This implies that AGN feedback is Bouchet P., Lequeux J., Maurice E., Prevot L., Prevot-Burnichon M. L.,
not well approximated with an immediate truncation as implicitly 1985, A&A, 149, 330
assumed in current models. Bower R. G., Benson A. J., Malbon R., Helly J. C., Frenk C. S., Baugh
C. M., Cole S., Lacey C. G., 2006, MNRAS, 370, 645
Caldwell N., Rose J. A., Concannon K. D., 2003, AJ, 125, 2891
Calzetti D., Armus L., Bohlin R. C., Kinney A. L., Koornneef J., Storchi-
AC K N OW L E D G M E N T S Bergmann T., 2000, ApJ, 533, 682
Cappellari M., Emsellem E., 2004, PASP, 116, 138
We would like to thank the referee for the thorough and prompt Carson D., 2007, PhD thesis, Univ. Portsmouth
report. We thank Andi Burkert, Stephen Justham, Sadegh Khochfar, Cattaneo A., Dekel A., Devriendt J., Guiderdoni B., Blaizot J.,
Chris Lintott, Lance Miller, Dimitra Rigopouplou, Clive Tadhunter 2006, MNRAS, 370, 1651
and Roberto Trotta for many stimulating discussions. We would also Cattaneo A. et al., 2007, MNRAS, 377, 63
like to thank Stephen Justham and Philipp Podsiadlowski for letting Cavaliere A., Menci N., 2007, ApJ, 664, 47
us use their computing resources. Cimatti A., Daddi E., Renzini A., 2006, A&A, 453, L29
This project has been partly supported by grant BMBF-LPD Ciotti L., Ostriker J. P., 1997, ApJ, 487, L105
Combes F., Young L. M., Bureau M., 2007, MNRAS, 337, 1795
9901/8-111 of the Deutsche Akademie der Naturforscher Leopold-
Connolly A. J., Szalay A. S., 1999, AJ, 117, 2052
ina. SK acknowledges a Leverhulme Early-Career Fellowship, a Croton D. J. et al., 2006, MNRAS, 365, 11
BIPAC fellowship and a Research Fellowship from Worcester Col- Daddi E. et al., 2007, preprint (arXiv:0705.2831)
lege, Oxford. S-JJ was supported by the Korea Research Foundation Dasyra K. M. et al., 2006a, ApJ, 638, 745
Grant funded by the Korean Government (MOEHRD) (KRF-2005- Dasyra K. M. et al., 2006b, ApJ, 651, 835
213-C00017). CM acknowledges the grant MEIF-CT-2005-011566 Davies R., Mueller Sanchez F., Genzel R., Tacconi L., Hicks E., Friedrich
of the Training and Mobility of Researchers Programme financed S., Sternberg A., 2007, ApJ, in press (arXiv:0704.1374)
by the European Community. Dekel A., Silk J., 1986, ApJ, 303, 39
The SDSS is managed by the Astrophysical Research Consor- De Lucia G., Springel V., White S. D. M., Croton D., Kauffmann G., 2006,
tium (ARC) for the Participating Institutions. The Participating In- MNRAS, 366, 499
Di Matteo T., Springel V., Hernquist L., 2005, Nat, 433, 604
stitutions are the University of Chicago, Fermilab, the Institute for
Di Matteo T., Colberg J., Springel V., Hernquist L., Sijacki D., 2007, preprint
Advanced Study, the Japan Participation Group, the Johns Hopkins (arXiv:0705.2269)
University, the Korean Scientist Group, Los Alamos National Lab- Eisenstein D. J. et al., 2001, AJ, 122, 2267
oratory, the Max-Planck-Institute for Astronomy (MPIA), the Max- Faber S. M. et al., 2007, ApJ, 665, 265
Planck-Institute for Astrophysics (MPA), New Mexico State Univer- Fabian A. C., Sanders J. S., Taylor G. B., Allen S. W., 2005, MNRAS, 360,
sity, University of Pittsburgh, University of Portsmouth, Princeton L20

C 2007 RAS, MNRAS 382, 1415–1431
Fabian A. C., Sanders J. S., Taylor G. B., Allen S. W., Crawford C. S., Nandra K. et al., 2007, ApJ, 660, L11
Johnstone R. M., Iwasawa K., 2006, MNRAS, 366, 417 Nelan J. E., Smith R. J., Hudson M. J., Wegner G. A., Lucey J. R., Moore
Ferrarese L., Merritt D., 2000, ApJ, 539, L9 S. A. W., Quinney S. J., Suntzeff N. B., 2005, ApJ, 632, 137
Ferreras I., Silk J., 2000, ApJ, 541, L37 Panter B., Heavens A. F., Jimenez R., 2003, MNRAS, 343, 1145
Fitzpatrick E. L., 1986, AJ, 92, 1068 Panter B., Jimenez R., Heavens A. F., Charlot S., 2007, MNRAS, 378, 1550
Fukugita M., Ichikawa T., Gunn J. E., Doi M., Shimasaku K., Schneider Prevot M. L., Lequeux J., Prevot L., Maurice E., Rocca-Volmerange B.,
D. P., 1996, AJ, 111, 1748 1984, A&A, 132, 389
Fukugita M., Nakamura O., Turner E. L., Helmboldt J., Nichol R. C., 2004, Rodighiero G. et al., 2007a, MNRAS, 376, 416
ApJ, 601, L127 Rodighiero G., Cimatti A., Franceschini A., Brusa M., Fritz J., Bolzonella
Fukugita M. et al., 2007, AJ, 134, 579 M., 2007b, A&A, 470, 21
Gao Y., Solomon P. M., 2004, ApJ, 606, 271 Salim S. et al., 2007, ApJS, in press (arXiv:0704.3611)
Gebhardt K. et al., 2000, ApJ, 539, L13 Sanders D. B., Mirabel I. F., 1996, ARA&A, 34, 749
Genzel R. et al., 1998, ApJ, 498, 579 Sarajedini A., Jablonka P., 2005, AJ, 130, 1627

Genzel R., Tacconi L. J., Rigopoulou D., Lutz D., Tecza M., 2001, ApJ, 563, Sarzi M. et al., 2006, MNRAS, 366, 1151
527 Sarzi M. et al., 2007, New Astron. Rev., 51, 18
Granato G. L., Silva L., Monaco P., Panuzzo P., Salucci P., De Zotti G., Schawinski K. et al., 2006a, Nat, 442, 888
Danese L., 2001, MNRAS, 324, 757 Schawinski K. et al., 2006b, ApJS, in press (astro-ph/0601036)
Granato G. L., De Zotti G., Silva L., Bressan A., Danese L., 2004, ApJ, 600, Schmidt M., 1959, ApJ, 129, 243
580 Sijacki D., Springel V., Di Matteo T., Hernquist L., 2007, MNRAS, 380, 877
Graves G. J., Faber S. M., Schiavon R. P., Yan R., 2007, preprint Silk J., 2005, MNRAS, 364, 1337
(arXiv:0707.1523) Silk J., Rees M. J., 1998, A&A, 331, L1
Gunn J. E. et al., 1998, AJ, 116, 3040 Skrutskie M. F. et al., 2006, AJ, 131, 1163
Harris G. L. H., Harris W. E., 2000, AJ, 120, 2423 Springel V., White S. D. M., Tormen G., Kauffmann G., 2001, MNRAS,
Harris W. E., Harris G. L. H., 2002, AJ, 123, 3108 328, 726
Harris G. L. H., Harris W. E., Poole G. B., 1999, AJ, 117, 855 Springel V., Di Matteo T., Hernquist L., 2005a, MNRAS, 361, 776
Heckman T. M., Kauffmann G., Brinchmann J., Charlot S., Tremonti C., Springel V. et al., 2005b, Nat, 435, 629
White S. D. M., 2004, ApJ, 613, 109 Stasińska G., Cid Fernandes R., Mateus A., Sodré L., Asari N. V., 2006,
Hopkins P. F., Cox T. J., Keres D., Hernquist L., 2007, preprint MNRAS, 371, 972
(arXiv:0707.1523) Stoughton C. et al., 2002, AJ, 123, 485
Jimenez R., Bernardi M., Haiman Z., Panter B., Heavens A. F., 2006, preprint Sutherland R. S., Dopita M. A., 1993, ApJS, 88, 253
(astro-ph/0670724) Thomas D., 1999, MNRAS, 306, 655
Kauffmann G., Charlot S., 1998, MNRAS, 294, 705 Thomas D., Maraston C., Bender R., 2003, MNRAS, 339, 897
Kauffmann G., Colberg J. M., Diaferio A., White S. D. M., 1999, MNRAS, Thomas D., Maraston C., Korn A., 2004, MNRAS, 351, L19
303, 188 Thomas D., Maraston C., Bender R., Mendes de Oliveira C., 2005, ApJ, 621,
Kauffmann G. et al., 2003, MNRAS, 346, 1055 673
Kauffmann G. et al., 2007, ApJS, in press (astro-ph/0609436) Trager S. C., Worthey G., Faber S. M., Burstein D., Gonzalez J. J., 1998,
Kaviraj S., Devriendt J. E. G., Ferreras I., Yi S. K., 2005, MNRAS, 360, 60 ApJS, 116, 1
Kaviraj S. et al., 2007, ApJS, in press (astro-ph/0601029) Trager S. C., Faber S. M., Worthey G., González J. J., 2000, AJ, 120, 165
Kawata D., Gibson B. K., 2005, MNRAS, 358, L16 Tremonti C. A. et al., 2004, ApJ, 613, 898
Kennicutt R. C., Jr, 1998, ApJ, 498, 541 van Dokkum P. G., 2005, AJ, 130, 2647
Kewley L. J., Dopita M. A., Sutherland R. S., Heisler C. A., Trevena J., Veilleux S. et al., 2006, ApJ, 643, 707
2001, ApJ, 556, 121 Wake D. A. et al., 2006, MNRAS, 372, 537
Kewley L. J., Groves B., Kauffmann G., Heckman T., 2006, MNRAS, 372, White S. D. M., Frenk C. S., 1991, ApJ, 379, 52
961 Wild V., Kauffmann G., Heckman T., Charlot S., Lemson G., Brinchmann
Khochfar S., Ostriker J. P., 2007, preprint (arXiv:0704.2418) J., Reichard T., Pasquali A., 2007, MNRAS, 381, 543
Kriek M. et al., 2007, ApJ, in press (astro-ph/0611724) Worthey G., Ottaviani D. L., 1997, ApJS, 111, 377
Lonsdale C. J. et al., 2003, PASP, 115, 897 Worthey G., Faber S. M., Gonzalez J. J., Burstein D., 1994, ApJS, 94, 687
MacArthur L. A., 2005, ApJ, 623, 795 Yan R., Newman J. A., Faber S. M., Konidaris N., Koo D., Davis M., 2006,
Maraston C., 1998, MNRAS, 300, 872 ApJ, 648, 281
Maraston C., 2005, MNRAS, 362, 799 Yi S. K. et al., 2005, ApJ, 619, L111
Maraston C., Thomas D., 2000, ApJ, 541, 126 York D. G. et al., 2000, AJ, 120, 1579
Martin D. C. et al., 2005, ApJ, 619, L1 Young L. M., 2002, AJ, 124, 788
Martin D. C. et al., 2007, in press (astro-ph/0703281)
Morganti R. et al., 2006, MNRAS, 371, 157
Nagashima M., Lacey C. G., Okamoto T., Baugh C. M., Frenk C. S., Cole
S., 2005, MNRAS, 363, L31 This paper has been typeset from a TEX/LATEX file prepared by the author.

C 2007 RAS, MNRAS 382, 1415–1431

Observational Evidence For AGN Feedback in Early-Type Galaxies

Uploaded by

Copyright:

Available Formats

Observational Evidence For AGN Feedback in Early-Type Galaxies

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Observational Evidence For AGN Feedback in Early-Type Galaxies

Uploaded by

Copyright:

Available Formats

Mon. Not. R. Astron. Soc. 382, 1415–1431 (2007) doi:10.1111/j.1365-2966.2007.12487.

Observational evidence for AGN feedback in early-type galaxies

Kevin Schawinski,1 Daniel Thomas,1,2 Marc Sarzi,1,3 Claudia Maraston,1,2

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

and structure formation theory. In every generation of theoretical

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

2.4.1 Gas and stellar kinematics

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

3.3 Dust extinction

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

4.1 Active galaxies in the colour–mass relationship

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

5.1 Evidence for the time sequence

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

Low mass 0–100 145–280 195–550 270–800 470–1800 550–2700

the sequence from transition region through Seyfert and LINER to

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

6.1.1 AGN feedback or black hole growth in action?

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

6.2 Obscured AGN? 6.3.2 Merger as main trigger

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

Downloaded from https://academic.oup.com/mnras/article-abstract/382/4/1415/1141702 by guest on 20 April 2020

You might also like