From Starburst to Quiescence: Testing AGN feedback in Rapidly Quenching Post-Starburst Galaxies
Hassen M. Yesuf, S. M. Faber, Jonathan R. Trump, David C. Koo, Jerome J. Fang, F. S. Liu, Vivienne Wild, Christopher C. Hayward
aa r X i v : . [ a s t r o - ph . GA ] J u l Accepted to ApJ: July 7, 2014
Preprint typeset using L A TEX style emulateapj v. 08/13/06
FROM STARBURST TO QUIESCENCE: TESTING AGN FEEDBACK IN RAPIDLY QUENCHINGPOST-STARBURST GALAXIES.
Hassen M. Yesuf , S. M. Faber , Jonathan R. Trump , David C. Koo , Jerome J. Fang , F. S. Liu , VivienneWild , Christopher C. Hayward Accepted to ApJ:
July 7, 2014
ABSTRACTPost-starbursts are galaxies in transition from the blue cloud to the red sequence. Although they arerare today, integrated over time they may be an important pathway to the red sequence. This workuses SDSS, GALEX, and WISE observations to identify the evolutionary sequence from starburststo fully quenched post-starbursts in the narrow mass range log M ( M ⊙ ) = 10 . − .
7, and identifies“transiting” post-starbursts which are intermediate between these two populations. In this massrange, ∼ .
3% of galaxies are starbursts, ∼ .
1% are quenched post-starbursts, and ∼ .
5% are thetransiting types in between. The transiting post-starbursts have stellar properties that are predictedfor fast-quenching starbursts and morphological characteristics that are already typical of early-typegalaxies. The AGN fraction, as estimated from optical line ratios, of these post-starbursts is about 3times higher ( & ± & ±
100 Myr), in agreement with previous studies. The time delay is inferred bycomparing the broad-band near NUV-to-optical photometry with stellar population synthesis models.We also find that starbursts and post-starbursts are significantly more dust-obscured than normalstar-forming galaxies in the same mass range. About 20% of the starbursts and 15% of the transitingpost-starbursts can be classified as the “Dust-Obscured Galaxies” (DOGs), with near-UV to mid-IRflux ratio of & primary role in the original quenching ofstarbursts but may be responsible for quenching later low-level star formation by removing gas anddust during the post-starburst phase. Subject headings: galaxies: active, galaxies: evolution, galaxies: formation, galaxies: starburst, galax-ies: stellar content, galaxies: structure INTRODUCTION
Galaxies show bimodality in their colors, morpholo-gies, and star formation rates both locally and athigh redshift (e.g., Strateva et al. 2001; Baldry et al.2004; Bell et al. 2004; Brammer et al. 2009). It isthought that star formation quenching causes “bluecloud” galaxies to migrate to the “red sequence”(Bell et al. 2004; Faber et al. 2007). A wide varietyof quenching mechanisms have been proposed to ex-plain the observed bi-modal galaxy properties (e.g.,Di Matteo et al. 2005; Kereˇs et al. 2005; Croton et al.2006; Dekel & Birnboim 2006; Hopkins et al. 2006;Somerville et al. 2008; Martig et al. 2009). These mech-anisms quench star formation by heating up gas in thegalaxy (halo), stabilizing it against collapse, or (rapidly)using it up or expelling it from the galaxy. These quench-ing mechanisms can be classified broadly into two modes:fast and slow. The slow mode occurs when star formationgradually fades, probably due to simple gas exhaustionover timescales longer than & Department of Astronomy & Astrophysics, University of Cali-fornia, Santa Cruz, CA 95064 USA Department of Astronomy & Astrophysics, Penn State, 515ADavey Lab, University Park, PA 16802 CA 95064 USA College of Physical Science and Technology, Shenyang NormalUniversity, Shenyang 110034, China School of Physics and Astronomy, University of St Andrews,North Haugh, St Andrews, KY16 9SS U.K. Heidelberger Institut f¨ur Theoretische Studien, Schloss–Wolfsbrunnenweg 35, 69118 Heidelberg, Germany quire any special triggering event such as mergers (e.g.,Noeske et al. 2007; Fang et al. 2013). On the other hand,rapid quenching is often identified with a triggering eventassociated with a merger-induced starburst and the re-sulting feedback (from either the starburst or from anassociated AGN) that rapidly removes or exhausts thegas (e.g., Sanders et al. 1988; Hopkins et al. 2006). Thiswork focuses on rapidly quenching or recently quenchedgalaxies.(Quenched) post-starburst galaxies, also known asK+A or E+A galaxies (e.g., Dressler & Gunn 1983;Zabludoff et al. 1996; Quintero et al. 2004), offer aunique view into galaxy evolution because they are be-lieved to be recently quenched starbursts rapidly transi-tioning from the blue cloud to the red sequence. Theymay contain lingering signatures of a quenching processimprinted on their spectral and structural properties.The fact that these galaxies have unusually large A-starpopulations but lack younger stars has been interpreted The term “K+A” refers to a galaxy with significant popula-tions of both (old) K stars and (young) A stars, indicative of rapidlyquenched recent star formation. The term “post-starburst” tradi-tionally refers to a K+A galaxy that was necessarily preceded by astarburst. We use the terms K+A and post-starburst interchange-ably. We often use the term K+A in a general sense when we referto related past studies. We avoid the term “E+A” which refers toa quenched galaxy with early-type morphology and a young stellarpopulation, because we show that many starburst galaxies alreadyhave compact and early-type morphologies before quenching intopost-starbursts.
Hassen Yesuf, et al.as evidence for recently quenched starbursts.Theoretically, post-starburst galaxies might be theend-product of galaxy mergers (Hopkins et al. 2006,2008; Bekki et al. 2001, 2005; Snyder et al. 2011). Ingas-rich model mergers, tidal torques channel gas togalaxy centers and power intense nuclear starbursts(Barnes & Hernquist 1991, 1996). The gas channeledto the centers may also lead to the onset of ob-scured nuclear AGN activity (Di Matteo et al. 2005;Hopkins et al. 2006). At the end of the starburst, af-ter gas has been exhausted by the starburst itself and/orexpelled by stellar feedback, the leftover gas and dust ob-scuring the active galactic nucleus (AGN) are cleared outdue to feedback from the AGN (Springel et al. 2005a,b;Hopkins et al. 2006; Kaviraj et al. 2007; Hopkins et al.2008; Snyder et al. 2011; Cen 2012). Consequently, starformation and further black hole growth are halted.Then the galaxies pass through the quenched post-starburst phase before they passively age and become“red and dead”. This work aims to test this hypothesisin detail.Using galaxy merger simulations, Snyder et al. (2011)have constrained the typical K+A life-time of merger-induced post-starbursts to be . . − . ∼ Finding a more complete sample of post-starbursts
Post-starbursts are rare galaxies, especially at low red-shift (Wild et al. 2009). They comprise . z ∼ . ′′ aperture and may not be representative of a galaxy as awhole. They are also not sufficient to identify heavilydust-obscured post-starbursts, which happen to overlapwith normal galaxies in their spectral indices. To iden-tify such objects we find that mid-IR colors are usefulaugmentation.Finally, one of the methods used in this work is toidentify common members of the starburst sequence in anarrow mass slice around log M ( M ⊙ ) ∼ .
5. This valuecorresponds to the transition mass in the color-mass di-agram where both star-forming and quiescent galaxiesare presently observed (Kauffmann et al. 2003b). A nar-row mass window approximately captures galaxies onthe starburst evolutionary sequence because the quench-ing timescale ( . . .
05 Gyr − (Hopkins et al. 2010)). Therefore, the precursors of post-starbursts are unlikely to grow in mass via multiple merg-ers in the time it takes them to quench. The burstfraction (amount of new stars formed) in a merger istypically <
20% of the total mass (Norton et al. 2001;Balogh et al. 2005; Kaviraj et al. 2007; Wild et al. 2009;Swinbank et al. 2012). Hence, it also does not signifi-cantly contribute to the mass increase on the starburstsequence.Even though a galaxy can be linked to its immediateprogenitors by its mass, mass by itself is not a sufficientpredictor of galaxy properties. Recent studies show thatstructural parameters that combine mass and radius (i.e.,stellar mass surface density, µ ∗ or velocity dispersion, σ )are better tracers of galaxy quenching (Kauffmann et al.2006; Franx et al. 2008; Cheung et al. 2012; Wake et al.rom Starburst to Quiescence 32012; Fang et al. 2013). This work will examine whether µ ∗ and σ of starbursts and post-starbursts are indicativeof quenching in these galaxies. Finding all AGN, including a population of highlyobscured AGN.
Most previous studies on post-starburst galaxies ex-cluded strong AGN because they adopted a restrictivedefinition of post-starbursts as galaxies with weak orno emission lines. Wild et al. (2010) attempted to im-prove this by defining post-starbursts using spectral in-dices only, bypassing the need for (weak) emission linerequirements. However, heavily dust-obscured post-starbursts and broad-line AGN are still excluded or miss-ing from their sample. This work attempts to includedust-obscured post-starbursts as part of the starburstsequence. It also constrains the star formation ratesof broad-line AGN and investigates whether they arepreferential to a specific stage in the merger sequence(e.g., Hopkins et al. 2006). We use GALEX and WISEphotometery in our selection criteria of obscured post-starbursts, and we use the WISE 12 µ m luminosity as aproxy for star formation rates (upper limits) of broad-line AGN. In a follow up paper, we plan to do furtherstudy on the star formation rates of broad-line AGN us-ing far-infrared data.Past studies of quenched post-starburst galaxies hintthat AGN are more common in these galaxies thanin normal galaxies (Yan et al. 2006; Georgakakis et al.2008; Brown et al. 2009). However, these past stud-ies were explicitly biased against strong AGN (Seyferts)since they excluded emission-line galaxies from theirpost-starburst samples. These studies also cannot ex-clude the possibility that the weak AGN signatures intheir post-starbursts are from “LINER-like” emissionunrelated to AGN activity (Cid Fernandes et al. 2011;Yan & Blanton 2012; Singh et al. 2013). Regardless,Yan et al. (2006) have found that 95% of their K+Agalaxies have LINER-like line ratios. Using a sampleof 44 K+A galaxies at z ∼ .
8, Georgakakis et al. (2008)have found a higher fraction of X-ray sources in post-starbursts ( ∼ ∼ z ∼ . ∼ erg s − .To improve on these previously incomplete estimatesof the AGN fraction in post-starbursts, we assemble alarge and less biased sample of post-starbursts which in-cludes emission-line galaxies to robustly identify AGN.This enables us to estimate the AGN fraction in transit-ing post-starbursts for the first time. We will infer therelationship between AGN and recent quenching in post-starbursts from the AGN fraction, and the time intervalbetween the peak of starburst to the peak of AGN activ-ity. If the AGN fraction is low, it indicates that AGN andquenching of starbursts are likely not related. A signifi-cant AGN delay might indicate a non-causal or secondaryrelationship (e.g., a common fueling mechanism or lateradditional quenching) between starbursts and AGN evenif AGN are more common in post-starbursts than in nor- mal galaxies.The rest of the paper is structured as follows. Section 2describes the multi-wavelength data. Section 3 presentsthe sample selection. Starbursts and the different classesof post-starbursts are defined in this section. Section 4investigates the AGN properties of post-starbursts. Sec-tion 5 presents the bulge properties of post-starburstsas an independent check on our sample selection. Sec-tion 6 presents a discussion on the importance of post-starbursts in the build-up of the red sequence. Thissection also summarizes the main results of the paper.Throughout the paper, an (Ω m , Ω Λ , h ) = (0 . , . , . DATA AND MEASUREMENTS
In the first three subsections, we will briefly describethe SDSS, GALEX and WISE data used. In the latersubsections, we will describe the dust correction, galaxystructural parameters and stellar population modelingemployed in the following sections.
SDSS
The Sloan Digital Sky Survey (York et al. 2000, SDSS)is a large photometric and spectroscopic survey. It hasmapped out about a third of the celestial sphere withits five filter band-passes, ugriz (Fukugita et al. 1996).The parent sample ( § ugriz pho-tometery using the Bayesian methodology to calculatethe likelihood of each model star formation history (SFH)given the data (Kauffmann et al. 2003a). The mass es-timate assumes that the SFH is approximated by a sumof discrete bursts and uses templates over a wide rangein age and metallicity. Thus, there should be no concernover systematic differences between the stellar mass esti-mates of starbursts, post-starbursts and normal galaxies.In addition, the masses of these galaxies are dominatedby their old pre-burst stellar populations as the contri-bution to the total mass from newly formed stars in aburst is only 3 −
20% of the total mass (Norton et al.2001; Balogh et al. 2005; Kaviraj et al. 2007; Wild et al.2009; Swinbank et al. 2012).DR8 also provides spectral indices and emission linemeasurements (Tremonti et al. 2004; Aihara et al. 2011).To measure the nebular emission lines of a galaxy, thecontinuum is modeled as a non-negative linear combina-tion of single stellar population (SSP) template spectragenerated using the Bruzual & Charlot (2003) (hereafterBC03) population synthesis code, and the best fittingmodel is subtracted from the galaxy spectrum.
GALEX
We use UV data from the Galaxy Evolution Explorer(GALEX, Martin et al. 2005) to to exploit the greatersensitivity of its near-UV ( m NUV < .
8) band to re-cent star formation. The near-UV (1771 - 2831 ˚A) imag-ing data have a spatial resolution of 6-8 ′′ and 1 ′′ as- Hassen Yesuf, et al.trometry. The data come from the cross-matched cat-alog between GALEX GR6 against SDSS DR7. Thiscatalog is available through the GALEX CASJobs in-terface . At fainter UV magnitudes, GALEX loses redgalaxies because they drop below the GALEX detectionthreshold. About 82% ( ∼ m r < . . < z < . ′′ . Adopting brighter r-band limitgives higher completeness ( & ′′ . Although theGALEX photometery for post-starbursts with multiplematches may not be accurate, we do not exclude themlest we systematically exclude merging systems. About90% of these post-starbursts are significantly dust ob-scured compared to normal galaxies. The exclusion ofthese post-starbursts does not significantly alter any ofour main results. WISE
The Wide-field Infrared Survey Explorer (WISE,Wright et al. 2010) performed an all-sky survey withphotometery in the 3 . µ m, 4 . µ m, 12 µ m, and 22 µ mbands. We used the Infrared Science Archive (IRSA) to match SDSS galaxies with the closest WISE sourceswithin a 5 ′′ radius. About 99 (92)% of SDSS galax-ies with 5 ′′ (2 ′′ ) GALEX matches have correspondingmatches in WISE. We use WISE data to study obscuredstar formation and AGN properties of post-starburstgalaxies. Dust correction
The main purpose of the dust correction is to re-duce the number of dusty obscured emission-line galax-ies, which otherwise masquerade as post-starbursts. Weuse the Balmer decrements, H α/ H β , with the physi-cally motivated two-component dust attenuation modelof Charlot & Fall (2000) to correct for attenuation of thenebular emission lines by dust. In the two-componentmodel, the diffuse dust accounts for 40% of the opticaldepth at 5500 ˚A while the denser birth-cloud dust ac-counts for the other 60% (Wild et al. 2011b). The op-tical depth of the dust is assumed to be a power-law ofthe form τ λ ∝ λ − . for the diffuse dust and τ λ ∝ λ − . for the birth-cloud dust. We adopt this model becauseit has a physical basis and is broadly consistent with ob-servations (Wild et al. 2011b).In addition, we correct the continuum fluxes (i.e., inte-grated magnitudes) using the empirical relationship be-tween the emission line and continuum optical depthsfound in Wild et al. (2011a) and their empirical stellarattenuation curve. They found that Balmer emissionlines experience two to four times more attenuation thanthe continuum at 5500 ˚A. We apply the dust correction http://galex.stsci.edu/casjobs/ http://irsa.ipac.caltech.edu/Missions/wise.html only on galaxies whose H α and H β lines are well mea-sured (with signal-to-noise ratio (SNR) > α/ H β = 2 .
86 for H II regions (Osterbrock 1989) and H α/ H β = 3 . ′′ fiber and do not reflect the galaxy-wide values, as thereare dust gradients across galaxies (Mu˜noz-Mateos et al.2009; Wild et al. 2011a). We make an approximate cor-rection for this effect following Wild et al. (2011a).We also correct for Galactic extinction of optical fluxesusing the catalog values provided in SDSS DR8 and ofthe NUV fluxes assuming a ratio A NUV /E ( B − V ) = 8 . A NUV is the NUV Galacticextinction and E ( B − V ) is the B − V color excess.More details on the dust correction can be found inAppendix A, where it is shown that our post-starburstselection does not significantly depend on the detailedassumptions of the dust correction described above. Forinstance, using single foreground screen model for dustdistribution Calzetti et al. (2000), we recover 85% ofPSBs selected using the two-component dust attenuationmodel. However, the single-component model identifies ∼ −
25% more PSB candidates, which may also bedusty contaminants. Throughout the paper, the sub-script ‘dc’ on a given quantity denotes dust-correction.For example, W H α, dc denotes a dust-corrected H α equiv-alent width (W H α ). K-correction
In addition to the dust correction, all galaxy magni-tudes and colors used in this work are k-corrected to z =0 using the public kcorrect IDL code (Blanton & Roweis2007). The GALEX NUV magnitude and the five SDSS ugriz magnitudes are used in estimating the k-correction.
Structural parameters
This section describes three structural parameters usedto study the relationship between star formation quench-ing and bulge growth.The stellar surface mass density is defined as the ra-tio of half the total stellar mass to the half-light Pet-rosian z band area, µ ∗ = M ∗ / πR , z . Kauffmann et al.(2006) found that µ ∗ is inversely proportional to the con-sumption time of the accreted gas from a galaxy halo(i.e, the burst decline time). They suggested that a highstellar surface mass density may be connected to bulgeformation through a nuclear starburst and quenching ofstar formation. However, Fang et al. (2013) showed thatthe mass surface densities as defined above may exag-gerate structural differences between blue and red galax-ies because they use a light-profile based radius as op-posed to mass-profile based radius. We use µ ∗ as definedabove only to show that starbursts and post-starburstsare both bulge-dominated galaxies, unlike most normalstar-forming galaxies.The velocity dispersion, σ , corrected to 1/8 of the effec-tive radius, r e , is estimated from the velocity dispersionmeasured within the 1.5 ′′ radius fiber, σ . , using the re-lation: σ = σ . (8 × . ′′ /r e ) . (Cappellari et al. 2006).rom Starburst to Quiescence 5 σ . is measured by the SDSS idlspec2d pipeline usingbroadened stellar PCA templates (Aihara et al. 2011).For r e , we use the weighted average of the circularized r -band radi of the de Vaucouleurs profile ( r e , dev ) and ex-ponential profile ( r e , exp ) : r e = f dev × r e , dev p b/a + (1 − f dev ) × r e , exp p b/a , where f dev is a coefficient that char-acterizes a galaxy image as a linear combination of a deVaucouleurs profile and an exponential profile (availablein the SDSS catalog).The color gradient, ∇ color , is defined as the differencebetween the g − r galaxy-wide color and the 2 ′′ g − r aperture color. The 2 ′′ aperture magnitudes are availablein SDSS DR8. The global galaxy colors are derived frommodel magnitudes by fitting the galaxy light with eitherde Vaucouleurs or exponential profile.Previous studies have used color gradients definedbased on 3 ′′ apertures (Roche et al. 2009; Bernardi et al.2011). We define ∇ color using the 2 ′′ aperture instead tobetter probe galaxy centers. For instance, about 90%of galaxies in the parent sample have half-light r -bandareas that are twice the 2 ′′ aperture areas at the corre-sponding redshifts. In comparison, only ∼
60% of thegalaxies have half-light areas that are twice the 3 ′′ aper-ture areas. We note that this is the only time we use aquantity measured within a 2 ′′ aperture.Positive ∇ color means blue-centered (young bulge),negative ∇ color means red-centered (old bulge) and ∇ color ∼ Stellar population modeling
To illustrate how a starburst evolves in some of ourdiagrams, we overplot Bruzual & Charlot (2003) modeltracks on these diagrams. To do so, we model SFHs of apost-starburst as a superposition of an old stellar popula-tion initially starting to form at time t = 0 and followinga delayed exponential SFH of the form ψ ∝ t exp( − t/τ )with e-folding time τ = 1 Gyr (cf. Kriek et al. 2011)plus a young stellar population formed in a recent burstat t = 12 . z ∼ .
1) with exponentially decliningSFH, ψ ∝ exp( − t/τ ) and τ = 0 . bf ∼ − τ = 0 .
05 Gyr or τ = 0 . Galaxy merger simulation
To further justify our selection of dust-obscured post-starburst galaxies, we use results from the M2M2 simula-tion presented in Lanz et al. (2014) and Hayward et al.(2014), which is an equal-mass merger of two disk galax-ies. Each disk galaxy is composed of a dark mat-ter halo, gaseous and stellar exponential disks, and a bulge. The progenitor galaxies each have a stellar massof 1 . × M ⊙ and a gas mass of 3 . × M ⊙ . SeeLanz et al. (2014) for full details of the specific simula-tion used.The merger was simulated using the smoothed-particlehydrodynamics code Gadget-3 (Springel et al. 2005a).The simulation includes models for star formationand stellar feedback (Springel & Hernquist 2003) andblack hole accretion and AGN feedback (Springel et al.2005b). In post-processing, the three-dimensional dustradiative transfer code
Sunrise (Jonsson et al. 2006;Jonsson et al. 2010) was used to calculate synthetic UV–mm SEDs of the simulated merger at various timesthroughout the merger.
Sunrise uses the stellar andAGN particles from the
Gadget-3 simulation as sourcesof radiation and calculates the effects of dust absorption,scattering, and re-emission as the radiation propagatesthrough the dusty ISM of the simulated galaxies.
Sun-rise calculates SEDs and images from arbitrary viewingangles. For clarity, we show only results from a singleviewing angle in this work. See Jonsson et al. (2010) andHayward et al. (2011) for further details of the
Sunrise calculations. SAMPLE SELECTION
This section presents the parent sample, and details onhow starbursts and post-starbursts are selected from thissample.
The Parent Sample
The basic sample selection is shown Figure 1. The sam-ple consists of a SDSS/GALEX/WISE-matched volume-limited sample (0 . < z < .
1) in a narrow stellar massrange of log M ( M ⊙ ) = 10 . − .
7. We call this sam-ple of ∼ ,
000 galaxies the parent sample . The chosenmass range roughly corresponds to the transition massin the color-mass diagram (Figure 1b) from lower-massstar-forming blue galaxies to higher-mass quiescent redgalaxies (Kauffmann et al. 2003b). We located the cen-ter of the mass bin on the lower end of the transitionmass because post-starbursts are preferentially found insmaller-mass galaxies, unlike slowly transitioning galax-ies which dominate at higher masses (see also Wong et al.2012). Moreover, restricting the redshift to be less than0.1 ensures higher GALEX completeness of the parentsample to red-sequence galaxies. As discussed in §
1, thestarburst-to-post-starburst evolution is followed in thenarrow mass-slice because mass likely does not increasesignificantly more than a factor of 2 along the starburstsequence.As schematically outlined in Figure 2, the next threesubsections describe in detail the selection of starburstsand post-starbursts from the parent sample. Starburstsare selected to have H α emission equivalent width above175 ˚A. The selection of post-starbursts generalizes theconventional definition to encompass both quenchingand quenched objects. The conventional post-starbursts,which are characterized by weak or no emission lines butstrong Balmer absorption lines, are termed as “QuenchedPost-starbursts (QPSBs)” in this paper. Transiting post-starburst (TPSB) galaxies, which precede quenched post-starbursts but come after the starbursts, are selected intwo ways. The first selection is based on the distinctiveevolutionary path that starbursts and post-starbursts fol- Hassen Yesuf, et al. Fig. 1.—
Panel a): Redshift versus stellar mass. The black points are galaxies in the SDSS-GALEX-WISE-matched catalog. Panelb): Dust-corrected
NUV − g color versus stellar mass for galaxies in redshift range 0 . < z < .
1. This study uses a volume-limitedparent sample of galaxies in a narrow mass slice around the transition mass between the blue-cloud and the red-sequence. The hatchedregions in both panels define the parent sample (log M ( M ⊙ ) = 10 . − . . < z < . Fig. 2.—
Schematic outline of our starburst and post-starburst selection. Identifying a reasonably complete transiting post-starburstpopulation between the starbursts and the quenched post-starbursts is the major new aspect of this paper. The TPSBs are identified bycombining their GALEX and/or WISE photometery with with their optical photometery and spectral indices. low in the 3D parameter space defined by dust-corrected
N U V − g color, H δ equivalent width and the 4000 ˚Abreak. Objects in this first class are called “Fading Post-starbursts (FPSBs)”. They are clearly offset from nor-mal galaxy locus in the parameter space that definesthem. The second selection of TPSBs uses GALEX andWISE photometery to identify dust-obscured transitingpost-starbursts, which are simply referred as “ObscuredPost-starbursts (OPSBs)”. We will later show that bothclasses of transiting post-starbursts have similar proper-ties (e.g., morphology) and that OPSBs generally pre- cede the FPSBs.The above discussion has isolated four classes of SBand PSB galaxies, we now proceed to explain how thefour classes are selected. Starbursts
A starburst has been defined in at least three ways(Knapen & James 2009). The definition we adopt con-siders a starburst as a galaxy with a temporarily highercurrent SFR than its past average by a factor of 2 − α (W H α ). Galaxies with ratios of cur-rent to past average SFR greater than two or three have(dust-extincted) W H α & −
110 ˚A (Lee et al. 2009;McQuinn et al. 2010).We define a starburst as a galaxy with dust-corrected W H α, dc >
175 ˚A . This threshold correspondsto & σ deviation from the mean W H α, dc distri-bution of star-forming galaxies in the parent sam-ple (star-forming galaxies are objects below the max-imum starburst boundary of Kewley et al. (2001) inthe BPT diagram). We note that a starburst withW H α & −
110 ˚A and nebular extinction A V = 1 willhave W H α, dc & −
175 ˚A if the continuum is extinctedless than the gas by a factor of two, as observed in star-bursts (Calzetti et al. 2000). Our starbursts have a me-dian A V of 2.3 and (fiber) SFR of about 10 M ⊙ yr − (spe-cific SFR of about 10 − yr − ). For comparison, the typ-ical SFR of a normal star-forming galaxy in the parentsample is about 1 M ⊙ yr − . Quenched Post-Starbursts (QPSB)
The conventional post-starburst galaxies are charac-terized as having no detectable or weak current star for-mation, but with significant star formation in their re-cent past ( < α and/or [O II] emission lines. The episodeof significant recent star formation is inferred from thepresence of strong Balmer absorption lines ( Hδ & H α, dc in emission versus W H δ in absorption, after theemission line infill correction. This diagram is used todefine QPSBs, which will help us motivate and explorehow such a conventional definition of post-starburst canbe improved on to include emission line galaxies (AGNor star-forming post-starbursts).The grey points in the figure represent all galaxies inthe parent sample. For the majority of galaxies, W H α, dc and W H δ are well-correlated with some scatter. Nor-mal star-forming galaxies form an elongated concentra-tion above W H α &
10 ˚A and W H δ & H α, dc . H δ . H α, dc < H δ > H δ (SNR >
3) and W H α (contam-inants with bad H α equivalent width measurements due The adjective “normal” is used throughout the paper to de-scribe galaxies that have not undergone a large burst ( > < to spectral gaps around H α are excluded). QPSBs aredenoted by (red) squares and are found in the lower-right corner of Figure 3. The (blue) stars in the top-right corner represent the starburst galaxies selected inthe previous subsection (W H α, dc >
175 ˚A). A large gapexists between starbursts and QPSBs, which must con-tain many transiting objects if the basic picture of ag-ing starbursts in this paper is correct. Identifying thesetransiting post-starbursts is the next goal of this paper.The (green) circles and the (brown) Xs represent the twotypes of transiting post-starbursts that are found in thenext subsection.
Transiting Post-starbursts (TPSB)
This subsection will describe our two ways of identify-ing TPSBs. Because of its similarity to that of previousworks, the selection of FPSBs is described first for con-venience, but FPSBs actually come after the OPSBs intime.
Fading Post-starbursts (FPSB)
Figure 3 illustrates the point made in the introduc-tion that it is difficult to identify the transiting post-starburst population without additional constraints be-cause they mostly overlap in this figure with the normal(non-bursty) star-forming galaxy sequence. However, itis possible to find these transiting objects by using othercombinations of colors and spectral indices. For this pur-pose, Figure 4 shows 2D projections of the 3D parame-ter space defined by the 4000 ˚A break, D n (4000) dc , W H δ ,and ( N U V − g ) dc . D n (4000) dc and W H δ are often used to distinguish re-cent star formation histories dominated by bursts fromthose that are more continuous (Kauffmann et al. 2003a;Martin et al. 2007; Wild et al. 2007). D n (4000), probesthe mean temperature of the stars responsible for thecontinuum and is a good indicator of mean stellar popula-tion age (Bruzual A. 1983; Kennicutt 1998; Balogh et al.1999). It is also much less sensitive, but not impervious,to dust effects (MacArthur 2005). We correct for (pos-sibly small) dust effects on D n (4000) using the averageattenuation in the narrow wavelength range in which itis defined.The ( N U V − g ) dc color is sensitive to young massivestars and as a result it evolves rapidly in rapidly quench-ing galaxies. It provides an additional lever arm thatcan be used to cleanly separate galaxies that are rapidlyquenching from the general star-forming population. Thefact that ( N U V − g ) dc color is an integrated galaxy-widephotometric quantity also makes it complementary to D n (4000) dc and W H δ , which are spectroscopic quantitiesmeasured within a 3 ′′ aperture and therefore may not berepresentative values of the entire galaxy. We select thefading post-starbursts as galaxies that are outliers fromnormal galaxies in ( N U V − g ) dc and W H δ at a given D n (4000) dc .As shown in Figure 4a, D n (4000) dc and W H δ are well-correlated for normal galaxies with smooth SFHs. Galax-ies with bursty histories are found off the main relation,as shown by the (orange and navy) curved BC03 modeltracks, which represent bursty SFHs. The (magenta)dashed curve across the main sequence denotes the fourthorder polynomial fit to the normal data (see Appendix Cfor more information). Hassen Yesuf, et al. Fig. 3.—
The dust-corrected H α emission equivalent width (W H α, dc ) against H δ A absorption strength(W H δ ) for the galaxies in theparent sample (grey), starbursts (blue stars), fading post-starbursts (FPSBs; green circles), obscured post-starbursts (OPSBs; brown Xs),and quenched post-starbursts (QPSBs; red squares). This diagram defines starbursts and QPSBs only, as galaxies lying above the upperhorizontal line and below the lower horizontal line; FPSBs and OPBS are defined by the next figures. The definition for each class is givenin § α EW for star-forminggalaxies with well-measured emission lines (shaded sky-blue histogram) and the parent sample. As shown by the blue curve, the distributionof H α EW for star-forming galaxies is well fit by a log-normal distribution with µ = 1 . σ = 0 .
35. We define starbursts as objects withW H α, dc >
175 ˚A, which is more than 2 σ from the mean. Figure 4b plots dust-corrected (
N U V − g ) dc color ver-sus D n (4000) dc . In this figure, two clouds of points arevisible for normal galaxies, the blue-cloud of young star-forming galaxies to the upper left, and the quenchedold and red galaxies to the lower right. The (magenta)dashed curve across the two clouds again denotes thefourth order polynomial fit to the normal data (see Ap-pendix C). Galaxies with bursty star formation historiesdeviate off the main relation to the lower left.In both Figures 4 a & b, the starbursts and thequenched post-starbursts are located at the extrema ofthe burst tracks. SBs are found at the tip of the blue-cloud, with very blue ( N U V − g ) dc color, low D n (4000) dc and relatively high W H δ . Likewise, QPSBs are also lo-cated off the main relation for normal galaxies, with veryred ( N U V − g ) dc color, intermediate D n (4000) dc and rel-atively high W H δ . The FPSBs are located in the inter-mediate region between starbursts and quenched post-starbursts in both figures. Hence, both ( N U V − g ) dc and W H δ are useful to identify this population.FPSBs are selected quantitatively as objects that aremore than 2 σ outliers from normal galaxies in ( N U V − g ) dc and W H δ . This is illustrated in Figure 4c, which de-picts the difference in ( N U V − g ) dc color, ∆( N U V − g ),and the difference in H δ equivalent width, ∆(H δ ), fromthe polynomial fit values at a given D n (4000). TheFPSBs are indicated by the (green) circles. The (pur-ple) ellipse encloses most normal galaxies at the 2 σ level. Thus, FPSBs are selected to be well outsidethe normal galaxy locus (defined by the purple ellipse)with well-measured ∆ quantities (∆(H δ ) /σ (H δ ) > N U V − g ) /σ ( N U V − g ) > , the σ s denoting themeasurement errors of H δ and N U V − g ). This method of selecting post-starbursts recovers almost all of thequenched post-starbursts from Figure 3 and identifiesmany FPSBs ( N ∼ N U V − g ) dc color asan additional selection criterion, specifically by requir-ing ∆( N U V − g ) >
0, a large number of contaminants( N ∼
50) are removed. A significant number of thesecontaminant galaxies show color gradient (have red cen-ters but blue outer parts) and are (edge-on) disk galaxies.In Figure 4d we plot (
N U V − g ) dc color versus W H δ ,a variant of Figure 3 in which W H α, dc is replaced by( N U V − g ) dc color. The overall trends of this figureand Figure 3 are similar. We previously used ∆( N U V − g ) and ∆(H δ ) in our selection because a starburst willcause these deviations in this diagram. By construction,the FPSBs do not overlap with normal galaxies in thisdiagram. There is also minimum overlap in Figure 4a,and the overlap in Figure 4b is a projection effect. Theselection in 3D space cleanly separates FPSBs because itremoves contaminants that are offset from normal galaxyin Figure 4a but not Figure 4d. Perhaps the offset ofthese contaminants in Figure 4a is due to the fiber effectof SDSS spectra.The selection of the TPSBs employed so far only iden-tifies objects that are significantly offset from the normalgalaxy locus and therefore misses a subset that overlapswith the normal galaxies (or those whose colors and spec-tral indices are not well-measured). This is evident fromthe small gaps between the SBs and the FPSBs in Fig-ure 3 & 4. The next subsection will describe how someof these missing objects are identified. Obscured Post-starbursts (OPSB) rom Starburst to Quiescence 9
Fig. 4.—
The co-joined plots show the relation among dust-corrected D n (4000), W H δ and dust-corrected NUV − g color for galaxies inthe parent sample (grey), starbursts (blue stars), obscured post-starburst (OPSB; brown Xs), fading post-starbursts (FPSB; green circles),and quenched post-starbursts (QPSB; red squares). The (magenta) dashed curves in panel a and panel b are the polynomial fits to themain galaxy sequence (these are fits to the data, not burst models). The (dark blue) dash-dotted curve and the solid (orange) curve areBC03 burst tracks with star formation timescales, τ = 0 . b f ) 3% or 20%. Panel c shows the differencein W H δ and ( NUV − g ) dc from the polynomial fit at a given D n (4000). The FPSBs are selected if they are found in the upper rightcorner and outside the (purple) ellipse, which encloses normal galaxies at the > σ level. The typical errors in each panel for transitingpost-starbursts are shown as green crosses. As discussed in the introduction, we aim to testthe merger-driven evolutionary framework for post-starbursts. Theoretically, it is thought that majormergers naturally result in highly dust-obscuredgalaxies (Hopkins et al. 2006; Jonsson et al. 2006;Chakrabarti et al. 2008; Narayanan et al. 2010;Hayward et al. 2012). Since PSBs are believed tobe the results of such mergers (Hopkins et al. 2006,2008; Bekki et al. 2001, 2005; Snyder et al. 2011), itis plausible that they exist in dust-obscured phase asthey quench (Poggianti & Wu 2000; Bekki et al. 2001;Shioya et al. 2001). Thus, we search for dust-obscuredobjects in our sample that likely bridge the gap betweenSBs and FPSBs. These objects have similar spectralindices and near UV colors as normal galaxies andtherefore could not be identified in the previous section.Figure 5 plots the flux density ratio between WISE12 µ m and GALEX NUV, f µ m / f . µ m versus the ra-tio of WISE 4 . µ m to WISE 3 . µ m. The f µ m / f . µ m ratio roughly quantifies the amount of obscured ver-sus unobscured star formation (Narayanan et al. 2010;Hwang & Geller 2013). We consider galaxies withf µ m / f . µ m >
200 as significantly dust-obscured (cf. Narayanan et al. 2010). Local DOGs havef µ m / f . µ m >
892 (Hwang & Geller 2013). Accordingto our definition, 69% of the starbursts, 45% the FPSBsand the 20% QPSBs are significantly dust-obscured.Likewise, 20% of the starbursts and 8% of the FPSBsare classified as DOGs. In comparison, only about 13%of galaxies in the parent sample are significantly dust-obscured and only 0.8% are DOGs.The fact that starbursts and post-starbursts selectedthus far are significantly more dust-obscured than normalgalaxies provides further motivation to select the secondclass of transiting post-starbursts using Figure 5. Wedefine the obscured post-starbursts (OPSBs) as galaxieswith W H δ > µ m / f . µ m >
200 and f . µ m / f . µ m > .
85 (the median value for SB is 0.8). Note that 20% ofthe OPSBs are DOGs.As further confirmation of the OPSB selection, Fig-ure 5b shows how a simulated major merger evolves inthe f . µ m / f . µ m vs. f µ m / f . µ m plot. The inset in thisfigure shows the time evolution of the star formation rateand AGN luminosity near the time of coalescence of thegalaxies (at ∼ .
13 Gyr). As the galaxies coalesce, a0 Hassen Yesuf, et al.
Fig. 5.—
Panel (a): The flux density ratio between WISE 12 µ m and GALEX NUV, f µ m / f . µ m , versus the ratio of WISE 4 . µ mto WISE 3 . µ m. The f µ m / f . µ m ratio quantifies the amount of obscured star formation versus unobscured star formation while thef . µ m / f . µ m ratio quantifies hot dust emission from an AGN or a starburst. We define obscured post-starbursts (OPSBs) as galaxiesin the upper right box and with W H δ > rom Starburst to Quiescence 11strong starburst is induced. Simultaneously, the AGNluminosity increases rapidly as the black hole particlesaccrete gas. Because most of the gas in the galaxiesis consumed or heated (by shocks and AGN feedback)during the starburst, the star formation rate rapidly de-creases. The AGN continues to accrete for ∼
100 Myrafter star formation is terminated because the gas in-flow rate needed to sustain the black hole accretion is . . M ⊙ yr − , which is orders of magnitude less thanthe star formation rate during the starburst. Dynami-cal effects can also cause a delay between the maximain the star formation rate and black hole accretion rate(Hopkins 2012). Note that gas consumption, not AGNfeedback, is the dominant cause for the termination ofthe simulated starburst. The effect of the AGN feedbackin the simulation is to further reduce the post-starburststar formation rate and expel the remaining gas and dustin the nuclear region (Hayward et al. 2013; Snyder et al.2011). AGN AND THEIR CONNECTION TOPOST-STARBURSTS
Having identified plausible candidate galaxies on theevolutionary pathway from starburst to quenched post-starbursts, we now explore the possible connection be-tween AGN activity and quenching in these objects.The tight correlation between masses of galactic cen-ter super-massive black holes (SMBH) and proper-ties of host galaxy bulges (e.g., Magorrian et al. 1998;Ferrarese & Merritt 2000; Tremaine et al. 2002) implythat galaxy evolution and SMBH accretion occur in along history of coupled growth and regulation (but seeKormendy & Ho (2013) for a contrarian perspective onco-evolution). Many semi-analytical models and the-oretical simulations require AGN feedback to quenchstar formation and correctly predict the observed colorbi-modality of galaxies and the shape of the galaxyluminosity function (e.g., Kauffmann & Haehnelt 2000;Croton et al. 2006; Hopkins et al. 2006; Somerville et al.2008; Gabor et al. 2011).The rapid quenching of post-starburst galaxies makesthem the ideal test-bed for AGN feedback models (e.g.,Hopkins et al. 2006; Snyder et al. 2011; Cen 2012). Withour samples spanning the whole post-starburst evolution-ary path, we quantify the fraction of AGN hosts amongpost-starbursts and their properties (stellar populationage, AGN strength, dust properties, etc). These quan-tities may help us infer whether AGN are primarily re-sponsible for quenching starbursts or not.
Optical AGN diagnostics
In Figure 6a, we show the BPT diagnostic using the[O
III]/H β and [N II]/H α line ratios (Baldwin et al. 1981;Veilleux & Osterbrock 1987). The position of an ob-ject in this diagram depends on its nebular metallic-ity and the hardness of its radiation field. Thus, theBPT diagram distinguishes between emission lines fromH II regions and AGN. AGN-dominated galaxies havelarger [O III]/H β and [N II]/H α ratios and occupy the up-per right of the diagram, while the softer ionization ofH II regions means star-forming galaxies occupy the lowerleft.The dashed (magenta) curve demarcates the theoreti-cal boundary for extreme starbursts, and galaxies above this curve probably host AGN (Kewley et al. 2001).The solid (orange) curve demarcates the empirical lowerboundary for AGN (Kauffmann et al. 2003c). Objectsbelow this curve are likely “pure” star-forming galaxies.Galaxies between the boundaries of extreme starburstsand “pure” star formation are thought to be mostly com-posites of star formation and AGN, although some haveargued that unusual ionization in H II regions can leadto starbursts without AGN lying in the composite region(e.g., Brinchmann et al. 2008). Similarly, galaxies in theAGN region may also have some star formation contri-bution, but their ionization state is dominated by theAGN.The starburst galaxies are distributed over the star-forming and composite regions (25%) of the diagramand only 3% are AGN. On the other hand, almost all(93%) of the quenched post-starburst galaxies with well-measured emission lines lie in the AGN region of theBPT diagram (cf. Yan et al. 2006). This might indicateweak AGN, although there is some evidence that photo-ionization in weak emission-line galaxies such as QPSBscan also be produced by shocks or post-asymptotic gi-ant branch stars (Ho 2008; Cid Fernandes et al. 2011;Yan & Blanton 2012; Singh et al. 2013). For instance,Cid Fernandes et al. (2011) have found that the ioniza-tion in galaxies with (dust-extincted) W H α < II]/H α ) > − . H α > II]/H α ) > − . ≤ W H α ≤ H α, dc < A and accordingly they arenot LINERs, but they are LINER-like (objects abovethe starburst boundary of (Kewley et al. 2001) and withW H α, dc < A ).The OPSBs and FPSBs bridge the starbursts andQPSBs. This is consistent with our evolutionary pathfrom starburst to transiting to quenched post-starburstgalaxies, with star formation decreasing along the se-quence as AGN emerge. 53% of the FPSBs and 37%OPBSs are AGN while about 16% FPSBs and 49% ofOPSBs are composite. Therefore, about 36% and 35%of transiting PSBs are AGN and composites respectively.In comparison, only 10% and 32% of normal galaxies inthe parent sample with W H α, dc > H α, dc > ≤ W H α, dc ≤ A delay between AGN and starbursts
In this subsection, we quantify the time delay betweenthe starburst and AGN phase.Figure 7 shows the distribution of (
N U V − g ) dc colorand D n (4000) dc (i.e, observable proxy for age) of star-bursts and AGN in transiting post-starbursts. The2 Hassen Yesuf, et al. Parent Sample Starburst OPSB FPSB QPSB020406080100 P e r c e n t a g e ( % )
52 72 14 32 032 25 49 16 77 3 37 36 03 0 0 17 07 0 0 0 93 star-formingComposite ?SeyfertLINERLINER-Like
Fig. 6.—
Panel (a) shows the BPT emission-line ratio AGN diagnostic for the parent sample, starbursts and post-starbursts whose emissionlines are detected with SNR >
3. The (magenta) dashed curve denotes the theoretical boundary for extreme starbursts (Kewley et al.2001) while the solid (orange) curve denotes the empirical boundary of pure star-forming galaxies (Kauffmann et al. 2003c). The diagramshows that the quenched post-starbursts have LINER-like emission while transiting post-starbursts have both star formation and AGN-dominated emission line ratios. The latter smoothly bridge the starbursts and the quenched post-starbursts. Panel (b) shows the percentageof AGN in starbursts, post-starbursts and galaxies in the parent sample. LINERs are objects in AGN region of the BPT diagram with3 ˚A ≤ W H α ≤ H α < ( N U V − g ) dc color and D n (4000) dc of TPSBs are sig-nificantly offset to higher values (older age) compareto values of starburst. The Kolmogorov-Smirnov test(K-S test) indicates that the null hypothesis that the( N U V − g ) dc color and D n (4000) dc of starbursts andTPSBs come from the same distribution (i.e, the twopopulation are coeval) can be rejected at α < .
001 sig-nificance level.Furthermore, Figure 8 shows the z band-normalizedmedian and quartile SEDs of galaxies evolving from thestarburst to quenched post-starburst phase. We over-plot BC03 models with SFR timescale, τ , of 100 Myrand burst mass-fraction b f of 20% at different ages inorder to indicate the time after the second burst. Thisballpark estimate shows that the median age of OPSBsis about 400-500 Myr and there is &
200 Myr gap be-tween the median age of starbursts and and the AGNhosts among TPSB. Because of the burst mass-age de-generacy, the ages of the post-starbursts depends on thedecay timescale ( τ ) assumed. As shown in Appendix B,models tracks with τ = 0 . − . τ outside this range are excluded since they would notproduce the observed population of post-starburst galax-ies (cf. Wild et al. 2010). Therefore, in agreement withthe findings of several recent observational works (e.g,Davies et al. 2007; Bennert et al. 2008; Schawinski et al.2009; Wild et al. 2010), the time delay might range be-tween 100 −
400 Myr depending on the assumed τ .The significance of this time delay is that it stronglysuggests that AGN do not directly quench starbursts.Recent theoretical works are converging to a view that,in merger-fueled post-starburst evolution, AGN may playa secondary or limited role in quenching (Croton et al.2006; Wild et al. 2009; Snyder et al. 2011; Cen 2012;Hayward et al. 2013). In other words, a post-starburstresults from exhaustion of a bulk of its gas supply in astarburst and/or from its expulsion by stellar feedback;AGN feedback mainly reheats or ejects the remaining gas that would otherwise fuel low-level star formationover the next few billion years.In particular, Cen (2012) proposed a new evolution-ary model of galaxies and their SMBH. In this model,starbursts and AGN are not coeval and AGN do notquench starbursts. They argued that the main SMBHgrowth occurs in the post-starburst phase, fueled by recy-cled gas (cf. Scoville & Norman 1988; Ciotti & Ostriker2007; Wild et al. 2010; Hopkins 2012) from aging starsin a self-regulated fashion on a timescale that is sub-stantially longer than 100 Myr. Our analysis supportsthe Cen (2012) model in that AGN are more frequentin post-starbursts and they appear significantly delayedfrom the starbursts phase. But as we will show later, wedo not find observational support for the model’s predic-tion that a substantial ( ×
10) black hole growth occursin the post-starburst phase compared to the starburstphase.
Dust properties of AGN host
In Figure 5, we showed that more than two-thirdsof starbursts and more than a third of FPSBs aresignificantly more dust-obscured compared to normalstar-forming galaxies. We also identified heavily dust-obscured PSBs that precede the FPSBs. Therefore, ourfinding that quenched post-starbursts were once heavilydust-obscured, and that some dust-obscured AGN arelikely post-starbursts, is consistent with the later removalof obscuring gas and dust by AGN feedback. However,beyond this consistency, there is no clear observationalevidence yet that AGN clear away the remaining gasand dust in post-starburst galaxies (e.g., Tremonti et al.2007; Coil et al. 2011). Therefore, future study of post-starburst with strong AGN identified in this work, mayprovide further clues on the (secondary) role of AGN andits relationship with its host galaxy.Figure 9 shows the distribution of V-band nebularattenuation A V for normal star-forming galaxies, star-bursts and transiting post-starbursts. SBs and AGN inrom Starburst to Quiescence 13 Fig. 7.—
The distributions of (
NUV − g ) dc color and D n (4000) dc for starburst galaxies and transiting PSB with AGN. The offsetbetween the peaks indicates that the two population are not coeval, with peak AGN activity appearing considerably after the peak starformation activity. TPSB have higher dust attenuation (A V = 2 . ± . V = 2 . ± . V = 1 . ± . V distribution of SBs or AGNTPSBs come from that of normal star-forming galaxiescan be rejected at α < .
001 significance while A V dis-tribution of SBs and AGN TPSBs are similar only at α . .
05. This observation is consistent with a removalof dust by AGN feedback.So far we have shown: 1) Starbursts and post-starbursts are likely more dust-obscured than normalstar-forming galaxies. The starburst to quenched post-starbursts evolutionary sequence is a decreasing dustsequence. 2) AGN are about three times more com-mon in transiting post-starbursts than in normal galax-ies. However, we found, similar to previous works, thatthere a significant time delay between starburst and thepeak of AGN activity in both obscured and fading post-starbursts.
Broad-Line AGN (BLAGN)
Special techniques are often required to disentangleAGN and galaxy emission in BLAGN host galaxies.Trump et al. (2013) have recently used SDSS aperturephotometry and z band concentration index to disentan-gle the light of broad-line AGN and their host galax-ies. By doing so, they have assembled a large sample ofBLAGN with host galaxy colors and stellar mass mea-surements.The selection criteria of post-starbursts discussed inprevious subsections will not identify post-starburstsgalaxies hosting BLAGN because their NUV fluxes andspectral indices are rendered immeasurable by the brightAGN. Nevertheless, to constrain how BLAGN fit in ourstarburst sequence, we select a subset of broad-line AGNfrom Trump et al. (2013) that have similar stellar massand redshift range as the parent sample. The properties of these objects are discussed in Appendix D . THE BULGE PROPERTIES OF POST-STARBURSTS ANDITS NECESSITY FOR QUENCHING
The overall aim of this section is to provide a com-plimentary check on our sample selection by showingthat the starbursts and post-starbursts are both bulge-dominated, unlike most normal star-forming galaxies.We show that their morphology is consistent with that ofgalaxies transitioning between blue and red galaxies. In afuture paper we will present other structural parametersthat better discriminate between post-starbursts and theslowly quenching normal galaxies.Figure 10a shows the relationship between the stellarmass surface density, µ ∗ , and the dust-corrected ( N U V − g ) dc color. NUV-optical color and µ ∗ are known totrace gas consumption time and the change in SFH thattakes place as galaxies transition from disc-dominatedto bulge-dominated systems (Kauffmann et al. 2006;Catinella et al. 2010). Comparisons between stellarsurface mass density and bulge-to-total ratio by Wildet al.(in preparation) shows that galaxies with µ ∗ > . × M ⊙ kpc − are classical bulge-dominated galax-ies while ones with 1 . × M ⊙ kpc − < µ ∗ < . × M ⊙ kpc − are pseudo-bulges. About 67% (95%)of starbursts and 65% (91%) of post-starburst have µ ∗ > × (3 × ) M ⊙ kpc − . In comparison, onlyabout 30% (68%) of normal star-forming galaxies have µ ∗ > × (3 × )M ⊙ kpc − . K-S test also indi-cates that the distribution µ ∗ for starbursts and post-starbursts are significantly different from normal star-forming galaxies (they are drawn from same distribu-tion at α . . Fig. 8.—
The z band-normalized median and quartile fluxes at the effective wavelengths of the NUV, u, g, r, i, z bands (the flux ratiosare dust-corrected). The (cyan, orange and magenta) overplotted spectra are Bruzual & Charlot (2003) burst models with SFR timescale τ = 0 . b f = 20% of different ages, as indicated on each panel. The lowest (black) spectrum in each panel isthat of a 12.5 Gyr old galaxy (before the burst). The model spectra are not actual fits to the data but are chosen to be approximatelyconsistent with data. Galaxies follow an age sequence from starbursts (panel a) to obscured PSBs (panel b) to fading PSBs (panel c) toquiescent PSBs (panel d). It is also notable that the SEDs of transiting PSBs hosting AGN (panels e and f) are significantly older thanthe starbursts, indicating a ∼
200 Myr delay between a starburst and the appearance of an AGN. This indicates that the AGN is not theprimary source of quenching in starbursts.
Similarly, Figure 10b shows (
N U V − g ) dc color as afunction of the velocity dispersion, σ . The velocity dis-persion is the best correlated parameter with galaxy colorand star formation history (Wake et al. 2012; Fang et al.2013). The M BH − σ relation (Magorrian et al. 1998) alsomeans that velocity dispersion is a tracer of black holemass: the upper x -axis in Figure 10b shows the inferredblack hole mass using the Tremaine et al. (2002) relation.The general galaxy population forms the blue cloud atlower σ (median σ = 108 km s − ) and the red-sequenceat higher σ (median σ = 160 km s − ). As expectedfor quenching/recently quenched galaxies, the starburstsand the three post-starbursts classes are located in thetransition region between the blue cloud and the red se-quence, at intermediate velocity dispersion ( σ ∼ −
140 km s − ).The SB to QPSB sequence is offset as a whole fromthe normal SFR galaxies by about a factor of two inblack hole mass. However, from starburst to transitingto quenched post-starbursts, there is little or no blackhole growth along the evolutionary sequence. This ob-servation does not not support the prediction of substan-tial ( ×
10) black hole growth in the post-starburst phase compared to the starburst phase (Cen 2012).In summary, we have shown that the three post-starburst classes are bulge-dominated unlike most nor-mal star-forming galaxies. The fact both SBs and PSBshave similar morphology is independent evidence thatthese two populations are linked. Similarly, the factFPSBs and OPSBs have structural properties that arefully consistent with each other supports that they areobjects in the same category despite their different selec-tion criteria. DISCUSSION AND CONCLUSION
Building the red-sequence through post-starbursts
The quenching process happens in both slow andfast-mode (e.g., Cheung et al. 2012; Barro et al. 2013;Dekel & Burkert 2013; Fang et al. 2013). We attemptto constrain the transit time and the fraction of galax-ies evolving through the two modes of quenching usingsimple crude estimates. Assuming that the starburstsare triggered by mergers or by some other phenomenonthat has a redshift dependence and using our thoroughand fairly complete post-starbursts sample today, onecan constrain how many of each kind of product evolvedrom Starburst to Quiescence 15
Fig. 9.—
The distribution of V-band nebular attenuation A V for normal star-forming galaxies, starbursts and AGN in transitingpost-starbursts.. The continuum attenuation are approximatelyhalf the nebular attenuation. The AGN hosted by TPSBs are sig-nificantly less dusty than the starbursts, consistent with the re-moval of dust by AGN feedback. to the red sequence through the two quenching modes inthe past 10 Gyr.The number of galaxies in the parent sample is ∼ ,
000 of which ∼ ,
400 galaxies are located in theblue cloud, ∼ ,
700 galaxies are in the red-sequence and ∼ ,
900 galaxies are in the green valley (see Figure 1b).If half of galaxies currently on the red sequence had un-der gone a dry major merger since z ∼ . ,
000 red-sequence galaxies must have been in theparent sample since z ∼ ∼
600 Myr for star for-mation timescale of τ = 0 . z ∼ ∝ (1 + z ) − (Kartaltepe et al. 2007;Hopkins et al. 2010; Lotz et al. 2011). In this case, thetransit rate through post-starbursts integrated to z ∼ − z ∼ z & &
25% of the red-sequence galaxies in the parent sample (over the past 10Gyr) descended from post-starbursts, we can constrainthe transit time across green valley for slowly quench-ing galaxies. Excluding the ∼ . ,
000 red-sequence galaxies must have gonethrough the slow mode of quenching over the past 10billion years. Assuming a constant transit time acrossthe green valley (Faber et al. 2007), the fact that wecurrently observe ∼ ,
900 slowly fading normal galax-ies in green valley implies that the transit time throughGV for the slow track is & ∼ ∼ ∼ < z < z ∼
2. They estimate that more than 65% ofthese galaxies are disk dominated. At a similar redshift,Kocevski et al. (2012) found that moderate luminosity,X-ray-selected AGN do not exhibit a significant excess of6 Hassen Yesuf, et al.
Fig. 10.—
The median and upper/lower quartile values of (
NUV − g ) dc color versus stellar mass surface density in Panel (a) and( NUV − g ) dc color versus velocity dispersion in panel (b) are plotted. Starbursts and post-starbursts have similar morphology and they areoffset from normal star-forming galaxies in mass surface density and velocity dispersion (i.e, have more prominent bulges). If we assumethat velocity dispersion correlates with black hole mass following the M BH − σ relation from Tremaine et al. (2002), then there is littleblack hole growth from SBs to QPSBs (in contrast with the prediction of Cen (2012)). distorted morphologies relative to a mass-matched con-trol sample. About half of the AGN reside in galaxieswith discernible disks. The observed high disk fractionin AGN hosts is hard to reconcile with the merger pictureof AGN fueling discussed in § tentatively find thatstarbursts are more disturbed than normal star-forminggalaxies (the disturbance could be due to major or minormergers). We visually classified about 30% the starburstsas merging or disturbed galaxies (they show either tidaltails or strong asymmetries or have close companions). Incontrast, only about 3 % of 200 randomly selected normalstar-forming galaxies show merger signatures. Likewise,according to the Galaxy Zoo classification (Lintott et al.2011), which rather tend to be conservative in callingsomething a merger, about 10% of the starbursts have aweighted merger fraction f m > . f m is calculated by takingthe ratio of the number of merger classifications for agiven galaxy to the total number of classifications for thatgalaxy multiplied by a weighting factor that measures thequality of the classifiers that have classified the galaxy.Darg et al. (2010) have shown that almost all galaxies with f m > . f m > .
4. Per-haps the merger signature are washed out because of thesubstantial time lag between the starburst and the PSB(AGN) phases. Galaxy merger simulations estimate thatmajor merger signatures have a timescale of 200-400 Myr(Lotz et al. 2010). Our estimated age of the transitingpost-starburst phase ( &
300 Myr) or the time delay be-tween starbursts and AGN ( & ±
100 Myr) is in accordwith the timescale for the disappearance of merger sig-natures. The color gradient and metallicity of starburstsand PSB are also consistent with the merger origin ofthese galaxies (see Appendix E & F).The above tentative result supports that the fast-track,in local universe, is triggered by merger starbursts, whosesignatures are washed out in the post-starbursts phase.We have also shown that velocity dispersion and globalmass surface density are high, presumably from merg-ers, leaving post-starburst remnants which are smaller,more compact, and with high stellar surface mass densitythan non-bursty star-forming galaxies. However, despitetheir high velocity dispersion and global mass surfacedensity, the post-starbursts still overlap in morphologywith many slowly quenching galaxies. Future work willexplore better morphological discriminants between thefast and slow mode (Yesuf et al., in preparation).Deep high resolution studies of handful of K+Agalaxies and post-starburst quasars however find sig-nificant morphological disturbances in these objects(e.g., Canalizo & Stockton 2001; Bennert et al. 2008;Yang et al. 2008; Cales et al. 2011). Galaxies we clas-sified as undisturbed using the SDSS images may havefaint tidal features visible in deeper images. Therefore,deep high resolution studies with more robust measure-ments of merger signatures in transiting post-starburstgalaxies will be useful to test merger origin of post-starbursts and to understand the AGN triggering mech-rom Starburst to Quiescence 17anism in post-starbursts.
Conclusion and summary
The unique spectral properties of quenched post-starburst galaxies hint that these objects are recentlyquenched starbursts. We investigated this inferred re-lationship in detail by directly tracing them back tothe starbursts through a newly identified population of“transiting” post-starbursts in the midst of quenching.We showed that dust-obscured post-starbursts comprisethe majority of the transiting post-starburst population.With our new sample of post-starbursts, we studiedthe connection between quenching and AGN in post-starbursts. We found that AGN are more commonlyhosted by post-starbursts than by normal galaxies. Post-starburst AGN hosts make up & ±
8% of transitingpost-starbursts. Despite the high frequency of AGN inpost-starbursts, we found that the clear presence of AGNis significantly delayed from the peak of starbursts by & ±
100 Myr.As long as the AGN appearance is delayed, our re-sults are generally consistent with “merger hypothesis” ofpost-starbursts (Hopkins et al. 2006; Snyder et al. 2011;Cen 2012), where mergers between gas-rich galaxies drivenuclear inflows of gas thereby leading to nuclear star-bursts, bulge formation, AGN activity, and eventually toquenched post-starbursts. In support of the merger hy-pothesis, we tentatively find that the starbursts are rel-atively metal-poor at earlier stages, exhibit clear mergersignatures, and have shallower color gradients and promi-nent young bulges. On the other hand, consistent withthe time delay, merger signatures disappear after thestarburst phase and that our three post-starburst classesalso have shallower color gradients and prominent youngbulges.We also showed that starbursts and transiting post-starbursts are significantly more dust-obscured than nor-mal galaxies and quenched post-starbursts. The fact thatstarbursts and post-starbursts evolve through a heavilydust-obscured phase which also seems to coincide withAGN activity, is consistent with later removal of dust byAGN feedback. We therefore conclude that AGN maynot primarily
APPENDIX
A: DETAILS OF DUST CORRECTION
Methods of dust correction
We correct for dust effects on emission-line luminosities ([O
II] , H α , and etc), GALEX and SDSS colors, the H α equivalent width, W H α , and D n (4000). For emission line extinction curve, we use eqn. A1 and for the continuumextinction curve, we use eqn. A2 (Charlot & Fall 2000; Wild et al. 2011a,b). In this section, continuum quantities willbe denoted by ‘ ∗ ’ superscript. Q λ = (1 − µ ) ( λ/ − . + µ ( λ/ − . where µ = 0 . Q ∗ λ = 1 /N [( λ/λ b ) n s + ( λ/λ b ) n s + ( λ/λ b ) n s + ( λ/λ b ) n s ] − /n (A2) Q ∗ λ is composed of four power-law functions with exponents s [1 − smoothly joined with a smoothness parameter n = 20. The power-law exponents vary with both axis ratio, b/a , and fiber specific star formation rate, ψ s , accordingto linear functions given in Wild et al. (2011a) eqn. 22-25. The λ b [1 − are related to the position of the three breakpoints at 0.2175 µ m, 0.3 µ m and 0.8 µ m and the power-law exponents according to eqn. 19-21 in Wild et al. (2011a). N is the normalization, defined such that Q ∗ λ is unity at 5500 ˚ A .The line optical depth is given by eqn. A3 and uses the expression of τ V in eqn. A4. τ λ = τ V Q λ (A3) τ V = 0 . × . / ( Q − Q ) × log (cid:16) H α/ H β × (H α/ H β ) − (cid:17) (A4)We require SNR > σ on H α and H β lines. We assume dust-free (H α/ H β ) df = 2 .
86 for star-forming galaxies(Osterbrock 1989) and (H α/ H β ) df = 3 . II] flux is given by: f OII , dc = f OII × . × . × τ (A5)To correct for galaxy fluxes (colors) we use the ratios of τ V /τ ∗ V in eqn. 13-16 Wild et al. (2011a), which are foundto vary strongly with galaxy properties such as axis-ratio and specific SFR. We use the ratios that give maximalstellar extinction. In other words, min { τ V /τ ∗ V ( ψ s ) , τ V /τ ∗ V ( b/a ) } . We prefer maximal stellar extinction because theoptical depth ratios in Wild et al. (2011a) are generally smaller but asymptote to 2.08, the measured values in starbursts(Calzetti et al. 2000). In estimating the optical depth ratios, we use star formation rates calculated from dust-correctedH α using the conversion factor of Kennicutt (1998). The SFRs will be overestimated if there is a significant contributionfrom AGN to the H α emission line. We used the optical depth ratios estimated from axis-ratio (i.e, inclination) onlyas a check, and the possible over-estimation of SFR due to AGN does not significantly affect our results. It should benoted that we do not purposely use the star formation rate measurements provided in SDSS DR8 which are derivedfrom photometry for AGN, because they may be systematically underestimated for dusty galaxies (including AGN,Wild et al. 2011a). τ ∗ λ = Q ∗ λ × τ ∗ V = Q ∗ λ × ( τ ∗ V /τ V ) × τ V (A6) A ∗ λ = 1 . τ ∗ λ (A7)Because we are correcting for the global galaxy colors, while our estimate of τ is based on fiber quantities, we willapproximately correct for gradient (aperture bias) in τ V by dividing A ∗ λ with a correction factor f ∇ = 1 . − . f ∇ = . R fib /R ≥ .
05 if R fib /R < ψ ≤ − . . R fib /R ≥ − . < log ψ < − . .
15 if R fib /R < − . < log ψ < − . . R fib /R ≥ ψ > − . .
25 if R fib /R < ψ > − . f ∇ = . R fib /R ≥ .
05 if R fib /R < ψ ≤ − . .
05 if R fib /R ≥ − . < log ψ < − . . R fib /R ≥ ψ ≥ − . .
15 if R fib /R < − . < log ψ < − . . R fib /R < ψ ≥ − . N U V − g ) dc is given by eqn. A10 below. We use the effective wavelengths of SDSS bands given inFukugita et al. (1996). ( N U V − g ) dc = ( N U V − g ) − ( A ∗ − A ∗ ) /f ∇ (A10)We correct the W H α using the following equation: W H α, dc = W H α × . × . × τ . × . × τ ∗ = W H α × . × . × ( τ − τ ∗ ) (A11) D n (4000) is defined as a flux ratio of a narrow continuum range red-ward of 4000 ˚A break (4000 − − D n (4000) dc ≈ D n (4000) × . × ( A ∗ red − A ∗ blue ) . where A ∗ red = ( A ∗ + A ∗ + A ∗ ) / . A ∗ blue =( A ∗ + A ∗ + A ∗ ) / . N U V − g color versus D n (4000)diagram for different dust-correction assumptions. In Panel a and b, we show the version of the figure in which N U V − g color versus D n (4000) are corrected using the Calzetti et al. (2000) extinction curve with the ratio of excess B − V colors of gas to stars is, E ∗ ( B − V ) /E ( B − V ) = 0 .
44 and 1.0. Note that, as described in § E ∗ ( B − V ) /E ( B − V )ratios. Since the selection of FPSBs explicitly depends on the dust correction, we show in panel a to b, the alternativeselection of this class for the the given dust-correction prescription adopted in each panel. In panel a 136 FPSBs andin panel b 126 FPSBs are identified. About 15-25% of the FPSBs are previously ( §
3) unidentified but about 85% ofthe FPSBs identified in section three are also identified in panel a to b. The Calzetti et al. (2000) curve lacks the2175 ˚A bump and Wild et al. (2011a) find typically 0 . − . §
3, suggeststhat the details of dust-correction are not important for the selection of these objects. Furthermore, the figure alsoshows (in orange square) the subset starbursts with (dust-extincted) W H α > α is important to identify majority of dust-extincted starbursts.The AGN fraction for FPSBs selected in panel a and pane b is 45% and 48% respectively. In comparison, theAGN fraction for FPSBs selected in the main text ( § The color-color diagram: the intrinsic colors of obscured post-starbursts
Moreover, in this subsection we aim to show that our dust-correction works and our starburst evolutionary path isplausible. Figure A2 shows the
U V gz diagram (
N U V − g vs. g − z ), a variant of the widely used U V J diagram ingalaxy evolution studies (e.g., Wuyts et al. 2007; Williams et al. 2009; Whitaker et al. 2012). In these diagrams dustystar-forming, non-dusty star-forming and quiescent galaxies are well separated. Star-forming galaxies form a diagonaltrack which extends from blue to red colors. The red end of this track is populated by dusty galaxies. The quiescentgalaxies form a separate clump above the dusty star-forming galaxies. We show the
U V gz diagram before and afterthe dust correction.After dust-correction, blue star-forming and red quiescent galaxies are cleanly separated in the
U V gz diagram.The starbursts are significantly bluer after the dust correction and they lie well off the blue cloud to the lower left.The dust correction is difficult for the quenched post-starbursts because of their weak emission lines. However theirlocation in the upper right corner of the
U V gz diagram is consistent with little or no dust extinction(cf. Balogh et al.2005; Kaviraj et al. 2007; Brown et al. 2009; Chilingarian & Zolotukhin 2012; Whitaker et al. 2012). We also showedin Figure 5 that about 80% of quenched post-starbursts do not show significantly dust-obscured star formation ( havef µ m / f . µ m < g − z colors of PSBs get redder from OPSBs to FPSBs to QPSB, suggesting adecreasing dust sequence we have seen in previous diagrams. The plausible arrangement of SBs and PSBs in color-colorspace is also further evidence that the dust corrections work. B: DETAILS OF STELLAR POPULATION MODELING
We modeled SFHs of a post-starburst as a superposition of an old stellar population initially formed at time t = 0following a delayed exponential SFH of a form ψ ∝ t exp( − t/τ ) with e-folding time, τ = 1 Gyr (cf. Kriek et al.2011) and a young stellar population formed in a recent burst at t = 12 . z ∼ .
1) with exponentially decliningSFH, ψ ∝ exp( − t/τ ), of τ = 0 . Fig. A1.—
Panel (a): Dust-corrected
NUV − g color versus D (4000). The FPSBs are selected assuming the Calzetti extinction curvewith E ∗ ( B − V ) /E ( B − V ) = 0 .
44. In panel (b) the FPSB selection assumes the Calzetti extinction curve with E ∗ ( B − V ) /E ( B − V ) = 1 . rom Starburst to Quiescence 21 Fig. A2.—
Panel (a) : Rest-frame
NUV − g vs. g − z color-color diagram, using observed (not dust-corrected) magnitudes. Panel(b) : Dust-corrected rest-frame NUV − g vs. g − z color-color diagram. The rapid quenching/strong burst model tracks nicely describethe sequence of starburst to transiting post-starburst to classical quenched post-starburst. The dust-obscured galaxies are also consistentwith the transiting post-starbursts along this track. We use this with our other evidence (in Figures 4 and 8) to infer that, like the fadingpost-starburst population, the dust-obscured post-starbursts represent an intermediate phase from starbursts to post-starbursts. Fig. B1.—
Panel (a): dust-corrected rest-frame
NUV − g and g − z color-color diagram. Overplotted are burst modeled tracks of τ = 0 . , . . g − z color .These diagrams exclude burst models outside τ = 0 . − . degeneracy, the ages of the post-starbursts depends on the decay timescale τ we assumed. In this section, we quantifythe effect of using different decay timescales ( τ = 0 .
05 or τ = 0 .
2) instead of our fiducial value of τ = 0 . N U V − g and g − z color-color diagram to show that the starburststo quenched post-starbursts evolution can alternatively be modeled with τ = 0 .
05 Gyr and burst fraction b f = 10%or τ = 0 . b f = 30%. Likewise, Figure B1b plots fiber specific SFR against dust-correctedfiber g − z color to make a similar point. Thus, models with τ outside the range 0.05-0.2 are excluded since they dorom Starburst to Quiescence 23 Fig. B2.—
The z band normalized median and quartile fluxes at the effective wavelengths of the NUV, u, g, r, i, z bands. This figure isthe similar to Figure 8 but overplots Bruzual & Charlot (2003) burst models with different SFR timescale τ and burst fraction b f . Topfigure: the overplotted spectra are τ = 0 .
05 Gyr and b f = 10% of different ages, as indicated on each panel. Bottom figure uses τ = 0 . b f = 30% instead. The main point of the figure is that starbursts and AGN are not coeval, they are separated at least by about100-400 Myr. Fig. D1.—
The WISE color-color plot which can reliably identify luminous (obscured and unobscured) AGN. Starbursts (blue star),fading post-starbursts (green circles), obscured post-starbursts (brown Xs), quenched post-starbursts (red squares) and broad-line AGN(orange diamonds) are overplotted on the figure. The f µ m / f . µ m flux ratio is sensitive to PAH emission and is a first order starformation indicator while f . µ m / f . µ m is sensitive to hot dust emission from AGN. The dashed wedge denotes the Mateos et al. (2012)AGN selection criteria while the horizontal dash-dotted line demarcates that of Ashby et al. (2009). The latter is more complete but lesspure. For comparison, we overplot broad line AGN of comparable mass and redshift studied by Trump et al. (2013). The figure confirmsthat almost all starbursts and strongly star-forming transiting PSBs do not host strong or obscured AGN. Some transiting post-starburstsshow hot dust emission from AGN. The BLAGN have lower f µ m / f . µ m ratios (less star-forming) than obscured post-starbursts buthigher flux ratios than quenched post-starbursts, suggesting that they may come after the obscured AGN phase (Hopkins et al. 2006), ifthey are related to post-starbursts. not produce the observed population of post-starburst galaxies. The fiber specific star formation rates are estimatedfrom dust-corrected H α ( § z band normalizedmedian and quartile SEDs of these galaxies. Accordingly, the time lag between the starburst and AGN phase may bebetween 100 and 400 Myr. C: DETAILS OF POST-STARBURST SELECTION
The following equations specify the fourth order polynomial fits to the data of main sequence galaxies in Figure 4.W H δ = 23 . − . × x + 3 . × x + 0 . × x − . × x (C1)( N U V − g ) dc = 7 . − . × x + 8 . × x − . × x + 0 . × x (C2)where x = D n (4000) dc D: MORE AGN PROPERTIES OF POST-STARBURSTS
WISE AGN diagnostic
An AGN has a spectral energy distribution (SED) that rises from ∼ − µ m due to hot dust emission from itsdusty torus (Nenkova et al. 2008), while a starburst has a composite stellar spectrum which peaks at 1 . µ m anddeclines over the range from ∼ − µ m. Mid-IR color-color diagnostic diagrams use this idea to distinguish AGNdominated galaxies from starburst dominated ones (e.g., Lacy et al. 2004; Stern et al. 2005; Donley et al. 2012). TheIR color-color diagrams select only luminous AGN, and do not detect weak AGN (Donley et al. 2012).Figure D1 plots the WISE color-color diagram: log(f µ m / f . µ m ) versus log(f . µ m / f . µ m ) (Wright et al. 2010;Izotov et al. 2011; Assef et al. 2012; Lake et al. 2012; Stern et al. 2012). This figure is similar to Figure 5 and itis presented here for completeness. Stellar populations younger than 0.6 Gyr dominate 12 µ m emission and, as a re-sult, [4 . µ m] − [12 µ m] color is known to correlate well with SFR (Donoso et al. 2012). Normal galaxies in the parentrom Starburst to Quiescence 25 Fig. D2.—
The flux density ratio between WISE 12 µ m and GALEX NUV, f µ m / f . µ m , versus the WISE 12 µ m luminosity. The12 µ m luminosity is used as a proxy for the (obscured) star formation rate ( an upper limit in AGN, see Donoso et al. 2012). f µ m / f . µ m ratio may indicate the amount of dust-obscuration. The histograms on the right show the distribution of 12 µ m luminosities for BLAGNand OPSBs respectively. Even with some contribution of AGN to the 12 µ m luminosity, BLAGN are generally less star-forming thanOPSBs (and majority of BLAGN are also likely less obscured). Therefore, BLAGN do not play a primary role quenching starbursts. Theirproperties in this diagram are consistent with the idea that BLAGN come after the obscured AGN phase (Hopkins et al. 2006). sample form a tight and elongated bi-modal sequence with some vertical scatter at f µ m / f . µ m &
1. Generally, thestarbursts are located at the right-most high-SFR end of the sequence, while quenched post-starbursts typically havelower f µ m / f . µ m ratios like quiescent galaxies. The transiting post-starbursts mostly have intermediate f µ m / f . µ m ratios between SBs and QPSBs. The arrangement of SBs, FPSBs and QPSBs in decreasing order of redness due todust is another independent confirmation for the consistency of our evolutionary sequence. As expected from theirselection, the obscured post-starbursts are found in between the FPSBs and SBs.The f . µ m / f . µ m ratio may indicate emission from hot dust, ionized gas or stars. The simple f . µ m / f . µ m > . ∼
50% reliability (Ashby et al. 2009; Stern et al.2012). The dashed wedge, which is calibrated by X-ray-selected AGN, identifies a highly complete and reliable sampleof luminous (hard X-ray luminosity, L −
10 kev > erg s − ) AGN (Mateos et al. 2012). For a reference, we alsooverplot broad-line AGN in the similar mass and redshift range.According to the Mateos et al. (2012) classification, only 7% of the FPSBs and 21% of the OPSBs show clear AGNsignatures in WISE, and the majority of these galaxies are already classified as Seyferts by the optical emission linediagnostics. This indicates that most of AGN found in the transiting post-starbursts, including the ones in compositeregions of the BPT diagram, must be weak ( L −
10 kev < erg s − ) if their presence is hidden by dilution from stellaremission. The star formation rates of broad-line AGN
Note that our post-starburst selection does not apply to broad-line AGN (BLAGN) hosts because many of theindicators that we have used to characterize the main evolutionary PSB sequence are diluted or polluted by the strongAGN signature in BLAGN (e.g., optical-UV colors and optical spectral signatures). Therefore, we cannot directlyconstrain the role of broad-line AGN in our post-starburst evolutionary sequence. However, in the following analysiswe use 12 µ m luminosities of BLAGN hosts to infer upper limits on their star formation rates (Chary & Elbaz 2001)and argue that their exclusion from the parent sample is not a problem. Their inferred star formation rates suggestthat they either come after the obscured post-starburst AGN (e.g., Hopkins et al. 2006) or they are not part of ourevolutionary sequence at all.The fact that BLAGN seem older than SBs and OPSBs was already suggested by their intermediate [4 . µ m] − [12 µ m]color in Figure D1. However, some AGN are known to exhibit suppressed aromatic features short-ward of 11 . µ m(Smith et al. 2007; Diamond-Stanic & Rieke 2010), suggesting that the [4 . µ m] − [12 µ m] color might underestimatethe SFR.Figure D2 shows 12 µ m luminosity against the flux density ratio of WISE 12 µ m to GALEX NUV, f µ m / f . µ m .6 Hassen Yesuf, et al.The 12 µ m luminosity is dominated by stellar populations younger than 0.6 Gyr in star-forming galaxies and in type II AGN (Donoso et al. 2012). The f µ m / f . µ m ratio roughly quantifies the ratio of obscured to unobscured SFR instar-forming galaxies and in type II AGN. It is not clear what f µ m / f . µ m ratio exactly means for BLAGN becausewe do not know how much of their NUV and IR flux comes from stars and how much from the AGN. For this reason,we place more emphasis on the 12 µ m luminosity as an upper limit on star formation.The general galaxy population shows a bi-modality in 12 µ m luminosity, reflecting the global bi-modality in starformation rates. As expected, the starbursts have higher 12 µ m luminosity than normal star-forming galaxies whilequenched post-starbursts have intermediate 12 µ m luminosity between quiescent and star-forming galaxies. Mostobscured post-starbursts have comparable 12 µ m luminosity to that of starbursts. This, at face value, is inconsistentwith the fact their SFRs as indicated by their H α and N U V − g colors are lower than those of starbursts (Figure 3 &4). However the excess mid-IR emission in OPSBs may be due to additional dust heating from their intermediate age( ∼ . µ m luminosity but they are clearly offset to the right from normal star-forming galaxies, that is, they are more dust-obscured. On the other hand, the BLAGN have similar 12 µ m luminositiesto those of FPSBs but most of them coincide with normal blue star-forming galaxies, i.e, they are less obscured.As the histograms of 12 µ m luminosities appended to the right of the plot shows, obscured post-starbursts are onaverage more luminous than BLAGN in 12 µ m. K-S test indicates that distribution of 12 µ m luminosities of OPSBsand BLAGN are significantly different ( D = 0 . , p ks = 5 . × − , i.e, α < . . − . µ m] − [12 µ m] color and 12 µ m luminosity and are likely less dust-obscured is consistent with the expectation that AGN might quench or prevent low-level star formation in post-starburstgalaxies by removing leftover gas and dust after the starburst events (Hopkins et al. 2006). Similarly, Zakamska et al.(2008) have shown that type II quasar hosts have increased star formation than type I quasar hosts, thereby supportingthe suggestion that obscured quasars come before naked quasars.We conclude that broad-line AGN are unlikely to play a primary role in the initial quenching of starbursts and theirexclusion from our post-starburst sample does not affect our main conclusions. Future work to robustly constrainthe instantaneous star formation rate of local BLAGN hosts would be very useful to understand whether BLAGN areassociated with quenching of starbursts or low star-forming post-starburst galaxies. E: FLAT COLOR GRADIENTS OF STARBURSTS & POST-STARBURSTS
Normal star-forming galaxies have red centers. A major merger likely alters or erases a color gradient of a pre-mergernormal galaxy by inducing star formation at the center or throughout the galaxy. Since the truncation of the starburstis abrupt, post-starburst galaxies should still carry the imprint of their merger origin by having flatter or more positivecolor gradients than normal star-forming galaxies.Figure E1a shows the color gradient, ∇ color , versus ( N U V − g ) dc color while Figure E1b plots dust-corrected N U V − g and g − z diagram color-coded by the color gradient. Starbursts and their descendants (transiting and quenched post-starbursts) have much shallower color gradients than the bulk of normal star-forming galaxies.As expected, red quiescent galaxies have flat color gradients and are red throughout but blue galaxies show aninteresting regularity in their color gradients: blue galaxies with bluer in ( N U V − g ) dc color (or lower H α ) havenegative color gradients (i.e, show large reddening in their centers) while blue galaxies with redder ( N U V − g ) dc colors(or higher H α ) have flat color gradients. In other words, galaxies with red centers have most of their star formationin their outer blue disks and have small quiescent bulges. On the other hand, galaxies which are blue throughout areeither experiencing nuclear bursts (have star formation rates above the average) or their nuclear bursts are abruptlyterminated (have redder N U V − g colors). They have young bulges.The fact that both starbursts and post-starbursts have flat color gradients (young blue bulges) suggests that starformation is uniformly distributed throughout these galaxies, remaining so even as the burst quenches throughout thegalaxy. It also suggests that the centers of these galaxies must have been unreddened from the typical red centers ofdisk galaxies, perhaps by gas inflow during a merger. Likewise, the flat color gradients of obscured post-starburstssuggests that the dust-obscuration in these objects is likely a galaxy wide phenomenon. Previous studies have alsoshown that A stars in K+A galaxies are widespread and are not confined to their nuclear regions (Kauffmann et al.2003c; Swinbank et al. 2005; Goto et al. 2008; Pracy et al. 2009; Swinbank et al. 2012). F: METALICITY OF POST-STARBURSTS
Figure F1 depicts the distribution of dust-corrected ([N
II]/[O II] ) dc ratios for normal star-forming galaxies, starburstsand post-starbursts. This ratio is a very reliable metallicity diagnostic and it is not very sensitive to the ionizationlevel (Kewley & Dopita 2002). Panel a compares the ([N II]/[O II] ) dc ratios of transiting PSBs with those of quenchedPSBs and starbursts. The ([N II]/[O II] ) dc ratio increases as the starbursts progressively evolve to the transiting androm Starburst to Quiescence 27 Fig. E1.—
Panel a: the median and upper/lower quartile values of (
NUV − g ) dc color versus color gradient, ∇ color . Panel b: NUVgzdiagram color-coded by color gradient. The contours represent the number density of all galaxies in the parent sample. Where the numberof galaxies within a bin is more than 15, we color code by the median values of the color gradient or the concentration index. Otherwise,the individual values for the galaxy is used. quenched post-starbursts. The starbursts have ([N II]/[O II] ) dc ratios that correspond to 1 − . Z ⊙ solar metallicityrange (Kewley & Dopita 2002) while the QPSBs have ratios corresponding to 2 − Z ⊙ solar metallicity, although their([N II]/[O II] ) dc ratio might not be well measured because their emission lines are weak. The TPSBs have intermediatemetallicity between starbursts and QPSBs. K-S test indicates that the distributions of metallicity of starbursts andTPSBs are significantly different ( D = 0 . , p KS = 2 . × − , i.e, α < . D = 0 . , p KS = 2 . × − , i.e, α < . D n (4000) dc < .
1) have evenlower metallicity than the AGN hosts in TPSBs (cf. Groves et al. 2006) and normal star-forming galaxies. K-S testalso indicates that the distributions of metallicity of younger starbursts (or all SBs) are significantly different ( at α < .
001 level) from AGN in TPSBs or normal star-forming galaxies. The transiently lower metallicity of youngerstarbursts is consistent with metal poor gas inflows during merger-induced starbursts (Barnes & Hernquist 1991, 1996)while the higher metallicity in post-starburst AGN is consistent with time delay between AGN and starbursts (e.g.,Wild et al. 2010; Hopkins 2012; Cen 2012).8 Hassen Yesuf, et al.
Fig. F1.—
Panel (a): The distribution dust-corrected [N
II]/[O II] ratios for the starbursts, transiting post-starburst and quenchedpost-starbursts. The metallicity increases from the starbursts to the quenched post-starbursts. Panel (b):The distribution dust-corrected[N
II]/[O II] ratios for the young starbursts (D n (4000) < . rom Starburst to Quiescence 29 REFERENCESAihara, H., et al. 2011, ApJS, 193, 29Ashby, M. L. N., et al. 2009, ApJ, 701, 428Assef, R. J., et al. 2012, ArXiv e-printsBaldry, I. K., Glazebrook, K., Brinkmann, J., Ivezi´c, ˇZ., Lupton,R. H., Nichol, R. C., & Szalay, A. S. 2004, ApJ, 600, 681Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981, PASP, 93, 5Balogh, M. L., Miller, C., Nichol, R., Zabludoff, A., & Goto, T.2005, MNRAS, 360, 587Balogh, M. L., Morris, S. L., Yee, H. K. C., Carlberg, R. G., &Ellingson, E. 1999, ApJ, 527, 54Barnes, J. E., & Hernquist, L. 1996, ApJ, 471, 115Barnes, J. E., & Hernquist, L. E. 1991, ApJ, 370, L65Barro, G., et al. 2013, ApJ, 765, 104Bekki, K., Couch, W. J., Shioya, Y., & Vazdekis, A. 2005, MNRAS,359, 949Bekki, K., Shioya, Y., & Couch, W. J. 2001, ApJ, 547, L17Bell, E. F., et al. 2004, ApJ, 608, 752—. 2006, ApJ, 640, 241—. 2012, ApJ, 753, 167Bennert, N., Canalizo, G., Jungwiert, B., Stockton, A., Schweizer,F., Peng, C. Y., & Lacy, M. 2008, ApJ, 677, 846Bernardi, M., Roche, N., Shankar, F., & Sheth, R. K. 2011,MNRAS, 412, L6Blake, C., et al. 2004, MNRAS, 355, 713Blanton, M. R., & Roweis, S. 2007, AJ, 133, 734Bournaud, F., et al. 2008, A&A, 486, 741Brammer, G. B., et al. 2009, ApJ, 706, L173Brinchmann, J., Charlot, S., White, S. D. M., Tremonti, C.,Kauffmann, G., Heckman, T., & Brinkmann, J. 2004, MNRAS,351, 1151Brinchmann, J., Pettini, M., & Charlot, S. 2008, MNRAS, 385, 769Brown, M. J. I., et al. 2009, ApJ, 703, 150Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000Bruzual A., G. 1983, ApJ, 273, 105Bundy, K., et al. 2010, ApJ, 719, 1969Cales, S. L., et al. 2011, ApJ, 741, 106Calzetti, D., Armus, L., Bohlin, R. C., Kinney, A. L., Koornneef,J., & Storchi-Bergmann, T. 2000, ApJ, 533, 682Canalizo, G., & Stockton, A. 2001, ApJ, 555, 719—. 2013, ArXiv e-printsCappellari, M., et al. 2006, MNRAS, 366, 1126Catinella, B., et al. 2010, MNRAS, 403, 683Cen, R. 2012, ApJ, 755, 28Chabrier, G. 2003, PASP, 115, 763Chakrabarti, S., Fenner, Y., Cox, T. J., Hernquist, L., & Whitney,B. A. 2008, ApJ, 688, 972Charlot, S., & Fall, S. M. 2000, ApJ, 539, 718Chary, R., & Elbaz, D. 2001, ApJ, 556, 562Cheng, J. Y., Faber, S. M., Simard, L., Graves, G. J., Lopez, E. D.,Yan, R., & Cooper, M. C. 2011, MNRAS, 412, 727Cheung, E., et al. 2012, ApJ, 760, 131Chilingarian, I. V., & Zolotukhin, I. Y. 2012, MNRAS, 419, 1727Cid Fernandes, R., Stasi´nska, G., Mateus, A., & Vale Asari, N.2011, MNRAS, 413, 1687Ciotti, L., & Ostriker, J. P. 2007, ApJ, 665, 1038Cisternas, M., et al. 2011, ApJ, 726, 57Coil, A. L., Weiner, B. J., Holz, D. E., Cooper, M. C., Yan, R., &Aird, J. 2011, ApJ, 743, 46Croton, D. J., et al. 2006, MNRAS, 365, 11Darg, D. W., et al. 2010, MNRAS, 401, 1043Davies, R. I., M¨uller S´anchez, F., Genzel, R., Tacconi, L. J., Hicks,E. K. S., Friedrich, S., & Sternberg, A. 2007, ApJ, 671, 1388Dekel, A., & Birnboim, Y. 2006, MNRAS, 368, 2Dekel, A., & Burkert, A. 2013, ArXiv e-printsDekel, A., et al. 2009, Nature, 457, 451Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433,604Diamond-Stanic, A. M., & Rieke, G. H. 2010, ApJ, 724, 140Donley, J. L., et al. 2012, ApJ, 748, 142Donoso, E., et al. 2012, ApJ, 748, 80Dressler, A., & Gunn, J. E. 1983, ApJ, 270, 7Ellison, S. L., Patton, D. R., Mendel, J. T., & Scudder, J. M. 2011,MNRAS, 418, 2043Faber, S. M., et al. 2007, ApJ, 665, 265 Falkenberg, M. A., Kotulla, R., & Fritze, U. 2009, MNRAS, 397,1940Fang, J. J., Faber, S. M., Koo, D. C., & Dekel, A. 2013, ArXive-printsFang, J. J., Faber, S. M., Salim, S., Graves, G. J., & Rich, R. M.2012, ApJ, 761, 23Ferrarese, L., & Merritt, D. 2000, ApJ, 539, L9Franx, M., van Dokkum, P. G., Schreiber, N. M. F., Wuyts, S.,Labb´e, I., & Toft, S. 2008, ApJ, 688, 770Fukugita, M., Ichikawa, T., Gunn, J. E., Doi, M., Shimasaku, K.,& Schneider, D. P. 1996, AJ, 111, 1748Gabor, J. M., Dav´e, R., Oppenheimer, B. D., & Finlator, K. 2011,MNRAS, 417, 2676Georgakakis, A., et al. 2008, MNRAS, 385, 2049Goto, T. 2007, MNRAS, 377, 1222Goto, T., Kawai, A., Shimono, A., Sugai, H., Yagi, M., & Hattori,T. 2008, MNRAS, 386, 1355Groves, B. A., Heckman, T. M., & Kauffmann, G. 2006, MNRAS,371, 1559Hayward, C. C., Jonsson, P., Kereˇs, D., Magnelli, B., Hernquist,L., & Cox, T. J. 2012, MNRAS, 424, 951Hayward, C. C., Kereˇs, D., Jonsson, P., Narayanan, D., Cox, T. J.,& Hernquist, L. 2011, ApJ, 743, 159Hayward, C. C., Torrey, P., Springel, V., Hernquist, L., &Vogelsberger, M. 2013, arXiv:1309.2942Hayward, C. C., et al. 2014, arXiv:1402.0006Ho, L. C. 2008, ARA&A, 46, 475Hopkins, P. F. 2012, MNRAS, 420, L8Hopkins, P. F., Hernquist, L., Cox, T. J., Di Matteo, T., Robertson,B., & Springel, V. 2006, ApJS, 163, 1Hopkins, P. F., Hernquist, L., Cox, T. J., & Kereˇs, D. 2008, ApJS,175, 356Hopkins, P. F., et al. 2010, ApJ, 715, 202Hwang, H. S., & Geller, M. J. 2013, ArXiv e-printsIzotov, Y. I., Guseva, N. G., Fricke, K. J., & Henkel, C. 2011, A&A,536, L7Jahnke, K., Kuhlbrodt, B., & Wisotzki, L. 2004, MNRAS, 352, 399Jahnke, K., & Macci`o, A. V. 2011, ApJ, 734, 92Jonsson, P., Cox, T. J., Primack, J. R., & Somerville, R. S. 2006,ApJ, 637, 255Jonsson, P., Groves, B. A., & Cox, T. J. 2010, MNRAS, 403, 17Kartaltepe, J. S., et al. 2007, ApJS, 172, 320Kauffmann, G., & Haehnelt, M. 2000, MNRAS, 311, 576Kauffmann, G., Heckman, T. M., De Lucia, G., Brinchmann, J.,Charlot, S., Tremonti, C., White, S. D. M., & Brinkmann, J.2006, MNRAS, 367, 1394Kauffmann, G., et al. 2003a, MNRAS, 341, 33—. 2003b, MNRAS, 341, 54—. 2003c, MNRAS, 346, 1055Kaviraj, S., Kirkby, L. A., Silk, J., & Sarzi, M. 2007, MNRAS, 382,960Kelson, D. D., & Holden, B. P. 2010, ApJ, 713, L28Kennicutt, R. C., Lee, J. C., Akiyama, S., Funes, J. G., & Sakai,S. 2005, in American Institute of Physics Conference Series,Vol. 783, The Evolution of Starbursts, ed. S. H¨uttmeister,E. Manthey, D. Bomans, & K. Weis, 3–16Kennicutt, Jr., R. C. 1998, ARA&A, 36, 189Kereˇs, D., Katz, N., Weinberg, D. H., & Dav´e, R. 2005, MNRAS,363, 2Kewley, L. J., & Dopita, M. A. 2002, ApJS, 142, 35Kewley, L. J., Dopita, M. A., Sutherland, R. S., Heisler, C. A., &Trevena, J. 2001, ApJ, 556, 121Knapen, J. H., & James, P. A. 2009, ApJ, 698, 1437Kocevski, D. D., et al. 2012, ApJ, 744, 148Kormendy, J., & Ho, L. C. 2013, ArXiv e-printsKriek, M., van Dokkum, P. G., Whitaker, K. E., Labb´e, I., Franx,M., & Brammer, G. B. 2011, ApJ, 743, 168Lacy, M., et al. 2004, ApJS, 154, 166Lake, S. E., Wright, E. L., Petty, S., Assef, R. J., Jarrett, T. H.,Stanford, S. A., Stern, D., & Tsai, C.-W. 2012, AJ, 143, 7Lanz, L., Hayward, C. C., Zezas, A., Smith, H. A., Ashby,M. L. N., Brassington, N., Fazio, G. G., & Hernquist, L. 2014,arXiv:1402.51510 Hassen Yesuf, et al.