Caught in the act: gas and stellar velocity dispersions in a fast quenching compact star-forming galaxy at z~1.7
G. Barro, S. M. Faber, A. Dekel, C. Pacifici, P. G. Perez-Gonzalez, E. Toloba, D. C. Koo, J. R. Trump, S. Inoue, Y. Guo, F. Liu, J. R. Primack, A. M. Koekemoer, G. Brammer, A. Cava, N. Cardiel, D. Ceverino, C. M. Eliche, J. J. Fang, S. L. Finkelstein, D. D. Kocevski, R. C. Livermore, E. McGrath
aa r X i v : . [ a s t r o - ph . GA ] M a r Submitted to the Astrophysical Journal
Preprint typeset using L A TEX style emulateapj v. 5/2/11
CAUGHT IN THE ACT: GAS AND STELLAR VELOCITY DISPERSIONS IN A FAST QUENCHINGCOMPACT STAR-FORMING GALAXY AT z ∼ . Guillermo Barro , Sandra M. Faber , Avishai Dekel , Camilla Pacifici , Pablo G. P´erez-Gonz´alez , ElisaToloba , David C. Koo , Jonathan R. Trump , Shigeki Inoue , Yicheng Guo , Fengshan Liu , Joel R. Primack ,Anton M. Koekemoer , Gabriel Brammer , Antonio Cava , Nicolas Cardiel , Daniel Ceverino , CarmenEliche , Jerome J. Fang , Steven L. Finkelstein , Dale D. Kocevski , Rachael C. Livermore , ElizabethMcGrath Submitted to the Astrophysical Journal
ABSTRACTWe present Keck-I MOSFIRE spectroscopy in the Y and H bands of GDN-8231, a massive, compact,star-forming galaxy (SFG) at a redshift z ∼ .
7. Its spectrum reveals both H α and [N ii ] emissionlines and strong Balmer absorption lines. The H α and Spitzer
MIPS 24 µ m fluxes are both weak,thus indicating a low star formation rate of SFR . −
10 M ⊙ y − . This, added to a relativelyyoung age of ∼
700 Myr measured from the absorption lines, provides the first direct evidence for adistant galaxy being caught in the act of rapidly shutting down its star formation. Such quenchingallows GDN-8231 to become a compact, quiescent galaxy, similar to 3 other galaxies in our sample, by z ∼ .
5. Moreover, the color profile of GDN-8231 shows a bluer center, consistent with the predictionsof recent simulations for an early phase of inside-out quenching. Its line-of-sight velocity dispersion forthe gas, σ gas LOS = 127 ±
32 km s − , is nearly 40% smaller than that of its stars, σ ⋆ LOS = 215 ±
35 km s − .High-resolution hydro-simulations of galaxies explain such apparently colder gas kinematics of up to afactor of ∼ . Subject headings: galaxies: photometry — galaxies: high-redshift INTRODUCTION
The formation and structural evolution of the first qui-escent galaxies at z & ⊙ ) > z & University of California, Santa Cruz The Hebrew University Yonsei University Observatory Universidad Complutense de Madrid Pennsylvania State University Hubble Fellow Shenyang Normal University Santa Cruz Institute for Particle Physics Space Telescope Science Institute Observatoire de Geneve Centro de Astrobiologia, CSIC-INTA The University of Texas at Austin Colby College centrally concentrated luminosity profiles and high S´ersicindices. The similar properties and number densities sug-gest that these compact star-forming galaxies (SFGs)are the immediate progenitors of the similarly com-pact first quiescent galaxies (Barro et al. 2013). Com-pact SFGs are typically found in a dusty star-formingphase characterized by bright far-IR and sub-mm detec-tions, and relatively normal star formation rates (SFRs)similar to those of other star-forming galaxies of thesame mass and redshift, in what is usually referred toas the star-forming main sequence (Noeske et al. 2007;Elbaz et al. 2007; Salim et al. 2007; Pannella et al. 2009;Magdis et al. 2010; Wuyts et al. 2011a; Elbaz et al.2011; Rodighiero et al. 2010; Whitaker et al. 2012b;Pannella et al. 2014). However, they have radicallydifferent morphologies suggesting that their compactnature is the result of strongly dissipative trans-formation processes, such as mergers (Hopkins et al.2006; Naab et al. 2007; Wuyts et al. 2010; Wellons et al.2014) or accretion-driven disk instabilities (Dekel et al.2009; Ceverino et al. 2010; Dekel & Burkert 2014;Zolotov et al. 2014) that funnel a large fraction of theirgas reservoirs into the center, rapidly building up a densestellar core.Additional evidence in support of the evolutionaryconnection between compact SFGs and quiescent galax-ies came recently when NIR spectroscopy of a sam-ple of compact SFGs revealed emission line widths of σ = 200 −
300 km s − (Barro et al. 2014b; Nelson et al.2014), in good agreement with the observed stellar ve-locity dispersions of compact quiescent galaxies of simi-lar stellar mass (Newman et al. 2010; van de Sande et al.2013; Bezanson et al. 2013; Belli et al. 2014b). However,these measurements are based on different dynamicaltracers, and measured on disjoint populations. Thereforethe implied evolutionary connection between them is in-direct. Some caveats to this evolutionary sequence are,for example: 1) the broad emission lines may be drivenby shocks and outflows rather than the gravitational po-tential (Newman et al. 2012b; Genzel et al. 2014) , and2) if compact SFGs do not quench immediately, their cur-rent dynamics may have little to do with their eventualtransition into quiescent galaxies.A way forward to address these issues is to study theproperties of both the gas and the stars on the samegalaxies. These kinematic properties can be used to testwhether the gas and the stars are probing the same gravi-tational potential, and the simultaneous measurement ofemission and absorption lines can be used to estimatethe age and star-formation history (SFH) of their stellarpopulations. However, these kind of measurements, arestill observationally challenging, particularly for galaxiesat z & .
5. Firstly, absorption line measurements (e.g.,the Balmer or the Ca HK lines) require NIR spectro-graphs and long ( > −
10 m classtelescopes, and probing the stronger emission lines (e.g.,H αλ ii ] λ iii ] λ βλ z & . ii ]. The [O ii ] λ α or[O iii ] ranges have so far been unsuccessful.In this paper, we present Y and H band spectroscopyof a massive (log(M/M ⊙ ) = 10 . z = 1 .
674 obtained with the MOSFIRE multi-object in-frared spectrograph (McLean et al. 2010; McLean et al.2012) on the Keck-I telescope. The spectra of GDN-8231 show emission and absorption lines which sug-gest that the galaxy is quenching rapidly. We modelthese lines to estimate the kinematics of the gas andthe stars. In addition, we combine the spectra withoptical medium-band photometry from the SHARDSsurvey (P´erez-Gonz´alez et al. 2013) and Hubble SpaceTelescope (
HST ) WFC3 grism spectroscopy in G102(GO13420; PI=Barro) and G141 (GO11600; PI=Weiner)to obtain detailed spectral energy distributions (SEDs)for the quenching galaxy and 3 older quiescent galaxies,observed as a part of the same MOSFIRE survey. Fromthe modeling of their SEDs, we estimate their stellar ages and SFHs, and analyze the implications for the evolu-tionary path from compact SFG to compact quiescentgalaxy.Throughout the paper, we adopt a flat cosmology withΩ M =0.3, Ω Λ =0.7 and H = 70 km s − Mpc − and wequote magnitudes in the AB system. DATA
Target Selection
We select spectroscopic targets from the CAN-DELS (Grogin et al. 2011; Koekemoer et al. 2011)WFC3/F160W ( H band) multi-wavelength catalogs inGOODS-N (Barro et al. in prep.). This catalog in-cludes deep, UV-to-NIR ground-based observations inseveral medium and broad bands, HST photometry in9 different bands and
Spitzer / Herschel observations inthe mid-to-far IR. The source extraction and mergedmulti-wavelength spectral energy distributions (SEDs)are measured following the procedures described inGalametz et al. (2013) and Guo et al. (2013). In addi-tion, GOODS-N includes complementary
HST /WFC3observations in both G102 and G141 grisms, allow-ing for continuous wavelength coverage from 0 . <λ < . µ m with a resolution better than R= 130(e.g., Brammer et al. 2012). The grism data are re-duced using the threedhst pipeline. The pipelinehandles the combination and reduction of the ditheredexposures, and the extraction of the individual spec-tra. These are background subtracted and corrected forcontamination from the overlapping spectra of nearbysources (Brammer et al. 2012; Momcheva et al. in prep).The grism data are joined at shorter wavelengths bythe 25 optical medium-bands of the SHARDS survey( R ∼
50; P´erez-Gonz´alez et al. 2013). Together, thesedatasets provide remarkable spectral resolution on agalaxy-by-galaxy basis, that is uniquely suited for SED-fitting analysis (see § transition stage between thestar-forming main sequence and the quiescent, red se-quence using − . < log(sSFR/Gyr − ) < − .
25. Thiscorresponds to galaxies approximately ∼ σ below themain sequence at z ∼ .
75 (Whitaker et al. 2012b). Forgalaxies with log(M/M ⊙ ) >
10, this threshold roughlycorresponds to a SFR &
10 M ⊙ y − , which is the5 σ detection level with MOSFIRE in ∼
1h exposures(e.g., Trump et al. 2013; Kriek et al. 2014). We alsoimpose the compactness criterion of Barro et al. (2013),
M/r . = 10 . M ⊙ kpc − . , to select galaxies with sim-ilar structural and morphological properties as the qui-escent population at that redshift. Figure 1 illustratesthat the sSFR criterion is consistent with the U − V vs. http://code.google.com/p/threedhst V−J U − V S t a r − f o r m i n g Q u i e s c e n t Q u e n c h i n g MgUV λ [ µ m] SHARDS H β +[OIII] G141 λ [ µ m] H δ [OII] G102 λ [ µ m] λ [ µ m] F λ [ C G S ] G DN − Broad−band
GDN−8231z = 1.673 Y MOSFIRESHARDS G G H ACS WFC3
Fig. 1.—
Left panel:
UVJ diagram for galaxies at 1 . < z < . ⊙ ) >
10 in the GOODS-N field. The colorshighlight different populations of star-forming galaxies (blue; log(sSFR/Gyr − ) > − . − ) < − .
75) and quenching, transition, galaxies (green; − . < log(sSFR/Gyr − ) < − .
75) identified according to their UV+IR SFRs. Thelarge markers show the quenching, compact SFG (GDN-8231; dark green) and the 3 quenched galaxies (GDN-2617, 17360, 12632; orange tored) observed with MOSFIRE. The
HST color images ( zJH ) of the 4 galaxies are shown at the bottom. The location of GDN-8231 in theUVJ is consistent with the observed spectral and photometric properties, indicating that it is a weakly star-forming galaxy transitioning to aquiescent phase. The quiescent galaxies fall within the selection region for recently quenched galaxies (left of the dashed line; Whitaker et al.2012a).
Right panel:
Color images (ACS and WFC3), and composite SED of GDN-8231. The grey circles show the (low-resolution) broad-band photometry, the cyan markers show the SHARDS medium-band data (R ∼ HST /WFC3 G102 and G141 grism spectra. The spectral regions probed by Y and H band MOSFIRE spectra are indicatedin red. The green and orange lines show the best-fit stellar population templates from Pacifici et al. (2012) at a resolution of R= 50 forGDN-8231 and the quiescent galaxy GDN-17360. The 3 sub-panels on the right show the zoom-in around the SHARDS, G102 and G141data highlighting spectral features in Mg UV , [O ii ], H δ and H β . V − J (hereafter UVJ) rest-frame color criterion that hasbeen shown to be very successful in identifying quiescentgalaxies, breaking the age/dust degeneracy (Wuyts et al.2007; Williams et al. 2010; Whitaker et al. 2011). Thespread of the transition sample along the wedge of theUVJ quiescent region is primarily driven by differencesin the dust reddening and in the stellar population ages,which suggests a wide diversity of extinction levels andSFHs among quenching galaxies (e.g., Wild et al. 2014).We select the most promising candidates for spectro-scopic follow-up by prioritizing bright galaxies with lowdust reddening to maximize the signal-to-noise (S/N)ratio. This additional restriction preferentially selectsgalaxies with bluer UVJ colors, near the so-called post-starburst region of the UVJ diagram (left of the dashedline in Figure 1; Whitaker et al. 2011; Bezanson et al.2013). This region encompasses colors that are typi-cal of recently quenched galaxies, with low extinctionlevels and young stellar ages of . ⊙ ) . .
8) than older quiescent galaxies withredder colors (Newman et al. 2013; Barro et al. 2014a).The redshift is restricted to the range 1 . < z < . α emission line in the H band and sev-eral other Balmer lines around the 4000 ˚A break in the Y band. The observed mask contains 1 candidate that isa transitioning , compact SFG (GDN-8231; green circle)and 3 recently quenched galaxies (GDN-12632, GDN-17360 and GDN-2617; orange to red circles). GDN-8231is relatively massive (log(M/M ⊙ ) = 10 .
7) and presentslow levels of star-formation (SFR = 10 M ⊙ y − ) evi-denced by a weak detection in MIPS 24 µ m ( f (24) =40 µ Jy). The low sSFR and mild extinction ( A V = 0 . Spectroscopic data
Data were collected on the nights of 2014 April 17 andMay 11 using the MOSFIRE instrument (McLean et al.2010, 2012) on the Keck-I telescope. The sky conditionswere clear and the median seeing ranged from 0 . ′′ − . ′′ Y (0 . < λ < . µ m) and H bands (1 . < λ < . µ m), with individual exposure times of 180 s and120 s, for a total 5.5 h and 2 h, respectively. We used2-point dithers separated by 1 . ′′ . ′′ . ′′ R = 3200 ( ∼ . ANALYSIS
Kinematic measurements and line ratios
The 4 galaxies are NIR bright (
H <
22 mag) andpresent clear continuum detections (S/N >
5) in boththe Y and H band spectra. GDN-8231 exhibits multipleBalmer absorption lines in the Y band spectrum, fromH δ to H10, and it is the only galaxy showing H α and[N ii ] emission lines in the H band spectrum (Figure 2). H δ H ε + CaII HCaII KH8H9H10 σ * = ±
35 km/s
GDN−8231 λ [Angstroms] F λ [ a r b i t r a r y un i t s ] H α [NII] λ [Angstroms] F λ [ a r b i t r a r y un i t s ] σ H α = 120 ±
34 km/s σ [NII] = 127 ±
32 km/s
Fig. 2.—
Left panel:
MOSFIRE Y band spectrum of the quenching, compact SFG (GDN-8231) with detected absorption lines. Thebest-fit model used to determine the stellar velocity dispersion and the absorption indices is shown in red. Right panel:
MOSFIRE H bandspectrum of GDN-8231 showing the spectral range around the H α and [N ii ] lines. The regions contaminated by sky lines are shown ingray. The blue Gaussians show the best fit to the H α ( σ LOS = 97 km s − ) and [N ii ] ( σ LOS = 127 km s − ) lines using the average continuumlevel. The red Gaussians show the best-fit to the H α line ( σ LOS = 120 km s − ) correcting for the H α stellar absorption as determined fromthe fit to the Y band spectrum. The other 3 quiescent galaxies also have clear Balmerabsorption lines in the Y band but present no emissionlines in the H band.We measure the line-of-sight (LOS) stellar velocity dis-persion using the penalized pixel-fitting software pPXF(Cappellari & Emsellem 2004). This software fits thegalaxy spectrum with a model created as a linear com-bination of the stellar templates that best reproduce thegalaxy spectrum allowing different weights for each tem-plates. The stellar templates used are the stars from theMILES stellar library (S´anchez-Bl´azquez et al. 2006b;Cenarro et al. 2007). Before fitting the galaxy spec-trum, we mask the regions contaminated by telluric at-mospheric bands and pixels with strong residuals fromthe sky lines, and convolve it with a Gaussian functionwhose width is the quadratic difference between the res-olution of the MILES stellar library (FWHM = 2 . . σ ⋆ LOS = 215 ±
35 km s − (seeTable 1 for the quiescent galaxies).We measure the LOS gas velocity dispersion, σ gas LOS , ofGDN-8231 by fitting a Gaussian profile to the emissionlines, measuring its FWHM, and subtracting the instru-mental broadening in quadrature from the FWHM. Thevelocity dispersion is then the corrected FWHM dividedby 2.355. We fit the H α and [N ii ] lines independently.As shown in Figure 2, [N ii ] is detected at higher S/Nratio because the H α line is partially contaminated bya skyline, and it appears to be self-absorbed (i.e., thecontinuum emission is affected by Balmer absorption). As a result, if we fit the lines using the same contin-uum level the velocity dispersion inferred from H α issmaller than that from [N ii ], σ gas LOS = 90 ±
18 km s − and127 ±
32 km s − , respectively. Although it is possiblefor Balmer and forbidden lines to have different widths ifthey originate in different regions, the most likely expla-nation for such a large difference is the self-absorptionin H α . To account for that effect, we use the best-fitstellar template to the absorption spectra to establishthe continuum level for the H α emission, and we recom-pute the fit. With this method, the inferred velocitydispersion is σ gas LOS = 120 ±
34 km s − , consistent with the[N ii ] result. Although the two measurements agree afteraccounting for Balmer absorption, we adopt the higher-S/N [N ii ] measurement as the more reliable tracer ofgas velocity dispersion.The spectra are not flux calibrated. However, wecan use the equivalent width of H α corrected for stel-lar absorption (EW(H α ) corr = 8 . ± .
7) and the con-tinuum flux, inferred from the SED modeling, to cal-culate the H α line flux and SFR. Using the empiricalrelation from Kennicutt (1998) and the attenuation in-ferred from SED-fitting (A V = 0 . − ⊙ y − , depending on the extra neb-ular extinction with respect to the continuum (A Hα =2 . − .
86 A cont ; e.g., Calzetti et al. 2000; Price et al.2014). These values are consistent with the estimate fromMIPS 24 µ m data. The H α line flux corrected for stellarabsorption also allows us to estimate the intrinsic valueof the line ratio [N ii ]/H α = 1 .
2. Even without measur-ing the [O iii ]/H β line ratio to fully constrain the ioniza-tion diagnostic diagram (i.e., BPT; Baldwin et al. 1981),an [N ii ]/H α ratio of the order of unity already suggeststhat the nebular emission is at least partially poweredby an AGN. The galaxy, however, is not detected inthe 2Ms X-ray data (Alexander et al. 2003) which im-plies that the AGN is relatively weak, perhaps shut-ting down along with the star formation in the galaxy.This result, together with the large fraction of AGNs( ∼ z = 2 − z ∼ Luminosity profile and Morphology
The MOSFIRE spectra of GDN-8231 are spatially un-resolved and thus do not provide additional insightson the kinematic profile of the gas and the stars, orthe spatial distribution of star formation. The high-resolution
HST imaging, however can be used to de-termine the overall structural properties and the radialdistribution of star-formation in the galaxy. Figure 3shows the surface brightness and color profiles of GDN-8231 computed from the fitting of the HST-based SEDsmeasured at each radius. The profile shows a positivecolor gradient of d ( N U V − V ) /dr = 6 · − mag/kpcthat is indicated by a bluer center with a 20% smaller r e in the rest-frame NUV with respect to the V band( r e,NUV /r e,V = 0 . =9 . ⊙ kpc − ), which seems to be a pre-requisite forquiescent galaxies (Cheung et al. 2012; Fang et al. 2013;van Dokkum et al. 2014). Such interpretation is an ex-cellent match for the predictions of simulations that de-scribe the formation of compact galaxies as the resultof strongly dissipative gas-rich events, such as merg-ers and/or disk instabilities (e.g., Dekel & Burkert 2014;Zolotov et al. 2014; Wellons et al. 2014). These pro-cesses are characterized by a wet-inflow (i.e., gas-inflowrate > SFR) that builds-up the central gas density,thereby enhancing the SFR at the center and growinga dense stellar core. The weakening of such inflow marksthe onset of inside-out quenching due to gas depletion.In that onset phase, the SFR is still high and thus thecolor gradient is still positive, as observed in GDN-8231.However, as inside-out quenching progresses, the SFRprofile flattens, and the color gradient turns negative, asobserved in compact quiescent galaxies (Guo et al. 2012; ? ; Szomoru et al. 2012).GDN-8231 has high-S´ersic values consistent withother compact star-forming and quiescent galaxies(Barro et al. 2013). Its visual appearance in the H band is smooth and spheroidal. However, the ACS im-ages, that probe the rest-frame UV, show small irregularpatches, perhaps indicative of its prior, dissipative com-paction event. Adopting the assumptions of Miller et al.(2011), we infer a viewing-angle inclination of i = 42 ◦ from an axis-ratio value of b/a = 0 . Spectral indices and SED fitting
One of the main obstacles to estimating stellarpopulation properties from the analysis of SEDs isthat at the typical resolution of broad-band surveys
GDN−8231 NU V − V F W H W H M F775W rest−NUVn = 5.2
F160W rest−Vn = 3.9 radius [kpc] µ [ m ag a r cs e c − ] NU V − V [ r e s t − f r a m e ] r e,V r e,NUV Fig. 3.—
Rest-frame surface brightness profile and color profile ofGDN-8231. The rest-frame profiles are computed by interpolatingat each radius the best-fit SED derived from the observed surfacebrightness profile in 9 HST bands measured with IRAF/ellipse (seeLiu et al. 2013 and Liu et al. in prep for more details). At z =1 .
67, the rest-frame NUV and V bands roughly correspond withthe observed i and H bands, respectively. The arrows indicatethe effective radius in those bands from the best-fit S´ersic profilesobtained using GALFIT. The grey line indicates the PSF half widthat half-maximum (HWHM) in the H band. The insets show theimages of GDN-8231 in the i (PSF-matched to H band) and Hbands. The black circle has radius of 1 ′′ ( ∼ . (FWHM ∼ µ m/(1+z); R ∼
6) the most relevant con-tinuum and emission/absorption line features are usu-ally diluted, which results in large uncertainties in in-ferred properties (Muzzin et al. 2009; Conroy & Gunn2010). One way to circumvent this problem is byusing higher spectral resolution data to obtain moreaccurate measurements of line strengths and spectralindices, which are key indicators of stellar age, andpresent/past star-formation activity (e.g., Kelson et al.2001; Kauffmann et al. 2003).From the MOSFIRE Y band spectra we measurethe H δ A Lick index (Worthey & Ottaviani 1997) usingPPXF to estimate the best-fit value and uncertainty.This index has been traditionally used as an age and SFHindicator (Trager et al. 2000; S´anchez-Bl´azquez et al.2006a). From the H band spectra we measure theH α Equivalent Width, EW(H α ), which is sensitive tothe instantaneous (last few Myr) SFR. Only GDN-8231presents H α emission, the other 3 quiescent galaxieshave strong continuum detections that place reliable up-per limits on EW(H α ). In addition to the MOSFIREspectra, we use the SHARDS medium-band data andthe HST grism data to probe for additional absorp-tion features and continuum breaks. The right panelof Figure 1 illustrates the high spectral resolution ( ∼ − × better than broad-band filters) of the mergedmedium-band/grism SED, revealing Mg ii absorption at2800 ˚A in SHARDS and Balmer absorption lines andthe 4000 ˚A break in G102 and G141. We quantify thelater using the D4000 index (e.g., Balogh et al. 1999; M ax i m a ll y O l d S t a r f o r m i n g M S S A M S F H s S l o w Q u e n c h i ng F as t Q u e n c h i ng τ − model ( τ = 100 Myr ) Age of the Universe [Gyr] A ge [ G y r ] M ax i m a ll y O l d M ax i m a ll y O l d M ax i m a ll y O l d M ax i m a ll y O l d M ax i m a ll y O l d M ax i m a ll y O l d M ax i m a ll y O l d Whitaker et al. (2014b) log (M ⋆ /M ⊙ ) = 10.6 - 10.8 τ − model ( τ = 100 Myr ) S t a r f o r m i ng M S S A M S F H s S l o w Q u e n c h i ng F as t Q u e n c h i ng l og ( s S F R ) [ G y r − ] Age of the Universe [Gyr]2 3 4 5 6−3−2−1012 4 3 2.5 2 1.7 1.4 1.2 1.0 0.8redshift
Fig. 4.—
Evolutionary tracks in sSFR (left) and luminosity-weighted age (right) vs. age of the Universe for different SFHs. Theblue shaded region depicts the star-forming main sequence determined from the average SFH of SFGs drawn from the model library ofPacifici et al. (2012). This region agrees well with the observational results of Whitaker et al. (2014) for SFGs of intermediate mass. Theblack lines illustrate the evolution of 2 galaxies that have a secular growth from z ∼ z = 1 .
7, followed by either fast (solid) or slow(dashed) quenching of star formation. In a fast quenching galaxy the luminosity-weighed age grows linearly with time (passive evolution).However, in a slow quenching (or main sequence) galaxy the luminosity-weighed age increases more slowly (i.e., the slope is < τ model can describe either a fast or slow quenching. However, a short τ (gray line) would provide unrealistic results for galaxies describedby a two-phase SFH (e.g., a main sequence + fast quenching). Kauffmann et al. 2003). The G102 grism shows also aweak [O ii ] emission line (EW 15 ± iii ] emission in G141. The later isexpected to have also low EW .
10 ˚A, and thus canbe partially hidden by the H β absorption, as hintedin the stacked spectra of recently quenched galaxies inWhitaker et al. (2013).The spectral indices are usually analyzed by compar-ing measurements to a grid of values derived from stellarpopulation synthesis models. Here we follow a slightlydifferent approach to combine information from boththe indices and the overall UV-to-NIR SED by usingof the SED-fitting code of Pacifici et al. (2012, here-after P12). The code performs a simultaneous fitting ofthe low and high resolution data and includes priors onthe EW of emission and absorption lines to obtain bet-ter constraints on the SFR and SFH of the galaxy (seealso Pacifici et al. 2014, 2015 and Barro et al. 2014a formore details). The galaxy templates are computed fromnon-parametric SFHs adapted from semi-analytic mod-els (SAMs) and include both the stellar continuum andthe nebular emission. The SAM SFHs provide a richerparameter space featuring short-timescale variations ofthe SFR (burst and truncations) that are missing in ex-ponentially declining (exp[-t/ τ ]) or delayed (t × exp[-t/ τ ])models (see also § RESULTS
Stellar ages and SFHs
The stellar age and SFH are not independent proper-ties and thus, the chosen parametrization of the latter(single burst, N-bursts, τ models, SAMs, etc) often leadsto strong degeneracies in the age and the characteris-tic timescale(s) of star formation (e.g., age- τ ; Conroy 2013). These degeneracies, however, can be reduced us-ing higher spectral-resolution datasets to probe featuresthat are sensitive to different star-formation and quench-ing timescales (Kriek et al. 2011; P´erez-Gonz´alez et al.2013; Belli et al. 2015).Figure 4 illustrates the implications of assuming dif-ferent SFHs to estimate the age of the galaxies. Theblue shaded region shows the evolution of the sSFR andluminosity-weighted age vs. time for galaxies in the star-forming main sequence, as described by the SAM SFHs inthe model library of Pacifici et al. (2012). Those galaxiesare thought to be growing in a relatively smooth, secu-lar mode (Elbaz et al. 2007; Rodighiero et al. 2010) inwhich gas inflow and SFR have reached a steady-statephase (e.g., Dekel et al. 2013). The main sequence pic-tured by SAM SFHs follows closely the observational re-sults (e.g., Whitaker et al. 2014) which are also in goodagreement with the predictions of other empirical models(e.g., Peng et al. 2010; Behroozi et al. 2013). In the mainsequence paradigm, quenching can be interpreted as thedeparture from a stable growth phase, which can be ei-ther fast (e.g., SFR truncation; solid line) or slow (shal-lower slope; dashed line). After fast quenching, age in-creases linearly with time (i.e., passive evolution), whilefor slower quenching the slope is shallower, and the ageincreases slowly due to slowly declining star-formation(right panel of Figure 4). In τ models, the timescale con-trols the quenching time, and it can be tuned to describefast or slow quenching (grey line in Figure 4). However,as one parameter characterizes the whole SFH, the pre-diction is not realistic for galaxies that are described by 2or more distinct phases, such as a secular growth followedby fast quenching.Panels a) and b) of Figure 5 show index-index diagramssensitive to the SFH and quenching time. The colored redshift d) redshift redshift d) redshift redshift d) redshift redshift d) redshift S t a r f o r m i n g M S S l o w Q u e n c h i ng W13 − OldW13 − Young
Age of the Universe [Gyr] A ge [ G y r ] redshift d) S t a r f o r m i ng M S S l o w Q u e n c h i ng l og ( s S F R ) [ G y r − ] Age of the Universe (Gyr) c)c)c)c)c) redshift S l o w Q u e n c h i ng S t a r f o r m i ng M S D4000 H δ A b)b)b)b)b) [ ˚A ] S t a r f o r m i ng M S S l o w Q u e n c h i ng D4000 E W ( H α ) [ ˚A ] a) [ ˚A ] a) [ ˚A ] a) [ ˚A ] a) [ ˚A ] a) GDN−2617GDN−17360GDN−12632
GDN−8231
Fig. 5.—
Panels a) & b):
Index-index diagrams for the galaxies in the sample. The circles show the values measured in the spectra. Thecolors are the same as in Figure 1. The colored lines depict the evolutionary tracks for the best-fit SAM SFHs. The tracks start at the onsetof quenching (colored squares) and continue for ∼ α andH δ indices are sensitive to recent changes in the SFR, while D4000 is a good tracer of the SFH averaged over longer timescales. In a fastquenching, the EW(H α ) drops abruptly while H δ A increases due to the appearance of prominent Balmer absorption lines. In contrast, aslow quenching galaxy has higher EW(H α ) and weaker H δ A for the same value of the D4000 index. Panels c) & d):
Same as Figure 4 forthe galaxies in our sample. The grey stars show the age of the oldest and youngest quiescent galaxies at z ∼ circles indicate the values measured in the MOSFIREspectra of the 4 galaxies, and the lines show the evolu-tionary tracks of their best-fit SAM SFHs. The D4000index is a good tracer of the average age of the stellarpopulations, while the EW(H α ) and H δ A are more sen-sitive to recent star-formation. For slow quenching, theEW(H α ) in emission decreases slowly while the Balmerabsorption is relatively small and constant with time,H δ A ∼ .
8. In contrast, fast quenching causes a rapiddecline in the H α emission followed by an absorptionplateau with EW(H α ) ∼ − δ A absorp-tion. This relatively short-lived phase ( ∼ α emission and strong absorptionin other Balmer lines, typical of A-stars, is known as apost-starburst phase (pSB; e.g., Wild et al. 2010). The spectral properties of GDN-8231 agree well withthe sSFR selection criteria indicating that the galaxyis on a fast quenching path. Besides the H α emis-sion, which indicates the presence of weak, ongoing star-formation, the SED fit, the D4000 index and the H δ A val-ues suggest that the emission is rapidly declining. In fact,given the short quenching time, the chances of finding agalaxy in this phase are small, which makes GDN-8231quite unique. The 3 quiescent galaxies in our sampleare further along in their evolution, as indicated by theirstronger D4000 ∼ .
6. However their indices and best-fit SFHs are also consistent with fast quenching. Thesimilar SFHs are another indication that GDN-8231 isevolutionarily linked to the quenched galaxies. For a def-inition of quenching time, t q , as the elapsed time fromhaving EW(H α ) values consistent with those of a mainsequence galaxy, to EW(H α ) ∼ ∼ × higher SFR than the main sequence). Amore practical definition of t q as the elapsed time fromlog(sSFR/Gyr − )= 0 to -1 (tracks in panel c)), resultsin similar values of t q = 300 −
800 Myr.The fast quenching scenario for all 4 galaxiesagrees well with recent works that find strong Balmerabsorptions on similarly young quiescent galaxiesat z ∼ . . < z < . α ) values consistent with gradually declin-ing SFHs. Assuming the difference is not caused bythe slightly different redshift range, a plausible expla-nation for this discrepancy is that there are indeed dif-ferent evolutionary quenching tracks (e.g., Martin et al.2007; Gon¸calves et al. 2012; Schawinski et al. 2014 orBelli et al. 2015), and current spectroscopic samples at z & . w = 700 Myrfor GDN-8231 to t w = 1 . − . z form &
6, ahalf-mass assembly by z ∼ z ∼
2. Interestingly, the latter is similar to the z form of the galaxies inferred from τ models, which sug-gests a rapid build up ( τ <
100 Myr). This, however,is most likely a limitation of single-parameter τ mod-els, which are biased towards short values of τ in or-der to reproduce the strong pSB features in the SED.Secular SFHs with rise and decay timescales of severalhundred Myr appear to be more realistic, and are pre-dicted by physically motivated models (e.g., Peng et al.2010; Behroozi et al. 2013; Gladders et al. 2013). How-ever, obtaining a direct, reliable measurement of the for-mation timescale would require additional measurementssuch as the metallicity [Z/H] or the α element abun-dance [ α /Fe], which trace the evolution of the metals(e.g., Conroy 2013; Onodera et al. 2014).The stars in panel d) of Figure 5 show the age of theoldest, t w = 1 . w = 0 . z ∼ z = 1 .
5. This implies that as new quenchinggalaxies join the red sequence the age spread in the qui-escent population will increase by roughly 1 Gyr from z = 2 to z = 1 .
5. Interestingly, if all the galaxies inour sample follow a pure passive evolution since z = 1 . z = 0 . w ∼ . − log(M ) [M ] ⊙ ⋆ l og ( σ L O S ) [ k m / s ] Compact SFGsCompact quiescent
Emission (gas)Absorption (stars)
Em. & Ab.GDN−8231
COS−10289HydroSim−V12 (i=90 ° )HydroSim−V12 (i=45 ° ) GDN−2617GDN−17360GDN−12632
Fig. 6.— σ LOS vs. M ⋆ for different galaxy samples. The point upand down triangles show emission (gas) and absorption (stars) linemeasurements, respectively. The blue triangles depict 13 compactSFGs in Barro et al. (2014a), and one galaxy from Nelson et al.(2014). The red triangles show a compilation of quiescent galax-ies (van de Sande et al. 2013; Bezanson et al. 2013; Belli et al.2014a,b) at z & .
5. The overlapping distributions for compactSFGs and quiescent galaxies suggests that both populations havesimilar kinematic properties. This is supported by the agreementin the gas and stellar dispersions of COS-10289 (Belli et al. 2014b,and Barro et al. 2014b for the stellar and gas kinematics). How-ever, GDN-8231 (green) has ∼
40% lower dispersion in the gasthan in the stars. A possible explanation is that the gas has colderkinematics than the stars ( v φ /σ r > σ gas LOS with respect to σ ⋆ LOS for viewing-angles closer to face-on (magenta). for the oldest quiescent galaxies at that redshift. Thiscould be an indication that some quiescent galaxies re-tain low levels of star formation, or perhaps experienceminor wet mergers that rejuvenate star formation.In summary, the 4 galaxies in our sample have rel-atively young ages of t w . Kinematic properties
Figure 6 shows the M ⋆ − σ LOS relation for a com-pilation of compact SFGs from Barro et al. (2014a)and quiescent galaxies at z & . V12 − STARS r/r e V φ , σ r , σ L O S [ k m s − ] σ r ( d i s p e r s i on ) V φ ( r o t a t i on ) σ starLOS (i = 90 ° ) σ starLOS (i = 45 ° ) x [kpc] y [ k p c ] −8 −1 1 8−8−118 σ r (dispersion) V φ ( r o t a t i on ) V12 − GAS r/r e V φ , σ r , σ L O S [ k m s − ] σ gasLOS (i = 90 ° ) σ gasLOS (i = 45 ° ) x [kpc] y [ k p c ] −8 −1 1 8−8−118 Fig. 7.—
Stellar (left) and gas (right) kinematic profiles for the simulated galaxy V12 at z = 2 . σ LOS profiles for a line-of-sight inclination of i = 90 ◦ (edge-on)and i = 45 ◦ . The triangles show the integrated, mass-weighted values at the r = r e . The line-of-sight dispersion can be written as σ = β (sin i v φ ) + σ r , where β depends on the inclination and density profile of the galaxy. This implies that for a rotation dominatedcomponent the value of σ LOS depends more strongly on projection effects. In V12 the gas has higher rotation than the stars ( v φ /σ r ∼ . v φ /σ r ∼
5) and therefore, the ratio of velocity dispersions can be as high as σ ⋆ LOS / σ gas LOS ∼ . i = 45 ◦ . Theinsets in the upper-right show the 10 ×
10 kpc face-on density maps for the stars and the gas. The black circle has r = r e . by showing agreement between its gas and stellar kine-matics. However, the low S/N ratio of the emission linemeasurement and the presence of an X-ray AGN withinthe galaxy place caveats on the interpretation of the kine-matic properties of the gas.Quite surprisingly, the better-constrained measure-ments for GDN-8231 (green triangles in Figure 6) showthat the gas has a lower velocity dispersion than the starsby a factor of σ ⋆ LOS / σ gas LOS = 1 . ± .
5. Naively, we ex-pected a value ∼ σ gas LOS caused by strong feedback pro-cesses (e.g., shocks or outflows; Diamond-Stanic et al.2012; Genzel et al. 2014). In turn, the smaller valuesof σ gas LOS suggest that the gas is in dynamical equilibriumand that it has colder kinematic properties than the stars(i.e., higher v φ /σ r in the gas). If that is indeed the case,the lower dispersion in the gas could be the result of:1) widespread star-formation activity in a disk observedat low inclination (i.e., close to edge-on), or 2) a cen-trally concentrated star-forming region probing, on av-erage, lower values of the rotational velocity v φ , whichgrows inside-out. As described in § i = 42 ◦ , and 20% smaller effective ra-dius in the rest-frame NUV, which can lead to the smallervalues of σ gas LOS compared to σ ⋆ LOS .In order to provide a better intuition, and quantify howmuch projection effects and/or a concentrated SFR pro-file affect the measurement of σ LOS , we study the kine-matic profiles of the gas and the stars in V12, a high-resolution, hydrodynamic galaxy simulation ( ∼
25 pcgrid) drawn from the sample of Zolotov et al. (2014) andCeverino et al. (2014). As described in Zolotov et al.(2014), these galaxies have similar stellar and structuralevolution as the compact SFGs and therefore, V12 pro-vides a excellent proxy for the analysis of the kinematicproperties of GDN-8231. Figure 7 shows that the starsin V12 have comparable rotation and dispersion whereas the gas is rotation dominated (stellar v φ /σ r ∼ . v φ /σ r ∼ σ gas LOS is more sensitive toprojection effects and shows smaller values for low in-clinations. For example, the integrated, mass-weighted σ gas LOS at r = r e show ratios of σ ⋆ LOS / σ gas LOS ∼ . i = 90 ◦ ) and an intermediate inclination( i = 45 ◦ ), respectively (see values in Figure 6). There-fore, we conclude that a large fraction of the observed ra-tio of velocity dispersions in GDN-8231 can be accountedfor by projection effects.On the other hand, given the relatively flat σ gas LOS pro-file of V12 (right panel of Figure 7), the change in theintegrated value of σ gas LOS ( r e ) would be small for a gas den-sity ( ∝ SFR) profile more centrally concentrated than thestellar profile (i.e., r e ,⋆ > r e , gas ). Using the ratio of effec-tive radii in GDN-8231 ( r e,NUV /r e,V = 0 .
8) as an exam-ple of different gas-to-stellar concentrations, we find thatintegrating σ gas LOS only up r = 0 . r e decreases its valueby less than 5% for the edge-on case. Nevertheless, theeffect of the gas density profile in σ gas LOS can be larger inGDN-8231 if: 1) it had a slowly-increasing rotation curveand a flat σ r in the center, and/or 2) the emission lineregion, traced by H α , had an even smaller r e than theNUV luminosity profile. The latter is more plausiblegiven the quenching nature of GDN-8231 and the shorterstar-formation timescales probed by H α .Lastly, note that Figure 7 shows only the radial com-ponent of the intrinsic dispersion ( σ r ) for V12. In thecase of anisotropic dispersion, for example due to strongcollimated winds in the center of the galaxy ( σ z ≫ σ r ),the observed value of σ gas LOS for a face-on inclination wouldbe much larger than that of σ ⋆ LOS . Dynamical mass
We estimate the dynamical mass of GDN-8231 from σ ⋆ LOS which, as described above, is less sensitive to projec-0tion effects. Following the virial equation: M dyn ( r < r e ) = K σ r e G (1)where K depends on the galaxy’s mass distribution,the inclination, and velocity field. We use K = 2 . σ ⋆ LOS tothe value in a circular aperture of radius r e , σ e =1 . × σ ⋆ LOS (Cappellari et al. 2006; van de Sande et al.2011). The inferred dynamical mass log(M dyn /M ⊙ ) =11 . M dyn /M ⋆ ) = 0 . − .
16 for recently quenched qui-escent galaxies z & . CONCLUSIONS
We present Keck-I MOSFIRE NIR spectroscopy ofGDN-8231, a massive, compact SFG galaxy at z ∼ . Y and H band spectra reveal strongBalmer absorption lines and H α and [N ii ] in emission.The emission and absorption lines yield spectral indicesand the kinematics of the gas and the stars.The spectral indices, SED-modeling, and the compar-ison to 3 compact quiescent galaxies at similar redshift,indicate that GDN-8231 was caught in a rare, earlystage of fast quenching. Still relatively young, with aluminosity-weighted age of 700 ±
250 Myr, GDN-8231will mature to become a compact quiescent galaxy byredshift z = 1 .
5. The rapid truncation of the SFR isevidenced by the low EW(H α ) and weak MIPS 24 µ mflux. The color profile is bluer in the center, which isconsistent with the predictions of recent simulations foran early stage of inside-out quenching. The line ratio of[N ii ]/H α ∼ z ∼
6) onset of star-formation, a secular build-up in the star-forming main sequence, forming 50% of itsstellar mass before z = 3, and fast quenching at z ∼ . σ gas LOS =147 ±
32 km s − , is 40% smaller than that of the stars, σ ⋆ LOS = 215 ±
35 km s − . This difference can be explainedif the gas has colder kinematics (rotation dominated)than the stars, and therefore: a) σ gas LOS is smaller if theviewing-angle is low (close to face-on), and b) σ gas LOS issmaller if the emission-line (star-forming) region is con-centrated at the center of the galaxy, and thus probeslow values of v φ . These options are consistent with thefindings of state-of-the-art galaxy simulations which pre-dict that the gas in compact SFGs reside in rotatingdisks (Zolotov et al. 2014). In the simulations, stars have σ ⋆ LOS up to 1.5 × larger depending on the projection of thegas disk. A clear prediction of these models is that thecompact quiescent descendants should retain some rota-tion from its disky progenitors.GDN-8231 stresses the need for larger samples of com-pact SFGs with emission and absorption line kinemat-ics to quantify the effects predicted in the simulations.Those samples would allow us to study the dependenceof σ ⋆ LOS / σ gas LOS with the viewing-angle and star-formationactivity. Similarly, high spatial resolution imaging in thesub-millimeter with ALMA, or in the NIR with adaptiveoptics, can provide direct measurements of the size andlocation of the star-forming regions and, in some cases,resolved kinematics for the ionized gas.
ACKNOWLEDGMENTS
Support for Program number HST-GO-12060 was pro-vided by NASA through a grant from the Space Tele-scope Science Institute, which is operated by the Associ-ation of Universities for Research in Astronomy, Incorpo-rated, under NASA contract NAS5-26555. GB acknowl-edges support from NSF grant AST-08-08133. PGP-Gand MCEM acknowledge support from grant AYA2012-31277. JRT acknowledges support from NASA throughHubble Fellowship grant
REFERENCESAlexander, D. M., Bauer, F. E., Brandt, W. N., et al. 2003, AJ,126, 539Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981, PASP, 93, 5Balogh, M. L., Morris, S. L., Yee, H. K. C., Carlberg, R. G., &Ellingson, E. 1999, ApJ, 527, 54Barro, G., Faber, S. M., P´erez-Gonz´alez, P. G., et al. 2013, ApJ,765, 104—. 2014a, ApJ, 791, 52Barro, G., Trump, J. R., Koo, D. C., et al. 2014b, ArXiv e-prints Bedregal, A. G., Scarlata, C., Henry, A. L., et al. 2013, ArXive-printsBehroozi, P. S., Wechsler, R. H., & Conroy, C. 2013, ApJ, 770, 57Belli, S., Newman, A. B., & Ellis, R. S. 2014a, ApJ, 783, 117—. 2015, ApJ, 799, 206Belli, S., Newman, A. B., Ellis, R. S., & Konidaris, N. P. 2014b,ApJ, 788, L29Bezanson, R., van Dokkum, P., van de Sande, J., Franx, M., &Kriek, M. 2013, ApJ, 764, L8 TABLE 1Stellar and spectroscopic properties of the MOSFIRE sample
ID R.A. DEC z spec F160W log M ⋆ r e n σ ⋆ σ gas EW(H α ) H δ A D4000 Age(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)8231 189.0656 62.1987 1.674 21.40 10.75 2.48 3.95 215 ±
35 127 ±
32 5.9 +1 . − . +0 . − . ± ± a ±
34 - -0.4 +0 . − . +1 . − . ± ± ±
29 - -0.7 +0 . − . +1 . − . ± ± ±
110 - 0.0 +0 . − . +0 . − . ± ± Note . —(a) Observed in Newman et al. (2010); Belli et al. (2014a), LRIS optical spectroscopy with σ = 174km s − .(1) General ID in the CANDELS H -band selected catalog in GOODS-N (Barro et al. in prep.) catalog.(2,3) R.A and Declination J2000.(4) Spectroscopic redshift.(5) H band (F160W) magnitude.(6) Stellar mass determined from SED fitting using Bruzual & Charlot (2003) and a Chabrier (2003) IMF, see § § § − ) of the stars measured with PPXF (Cappellari & Emsellem 2004), see § − ) of the gas measured from the emission line width (FWHM), see § α (˚A) measured in the H band MOSFIRE spectra, see § δ A (˚A) measured in the Y band MOSFIRE spectra with PPXF, see § HST grism spectroscopy, see § Binney, J. & Tremaine, S. 2008, Galactic Dynamics: SecondEdition (Princeton University Press)Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008, ApJ,686, 1503Brammer, G. B., van Dokkum, P. G., Franx, M., et al. 2012,ApJS, 200, 13Bruzual, G. & Charlot, S. 2003, MNRAS, 344, 1000Buitrago, F., Trujillo, I., Conselice, C. J., et al. 2008, ArXive-printsCalzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682Cappellari, M., Bacon, R., Bureau, M., et al. 2006, MNRAS, 366,1126Cappellari, M. & Emsellem, E. 2004, PASP, 116, 138Carollo, C. M., Bschorr, T. J., Renzini, A., et al. 2013, ApJ, 773,112Cassata, P., Giavalisco, M., Guo, Y., et al. 2011, ApJ, 743, 96Cenarro, A. J., Peletier, R. F., S´anchez-Bl´azquez, P., et al. 2007,MNRAS, 374, 664Ceverino, D., Dekel, A., & Bournaud, F. 2010, MNRAS, 404, 2151Ceverino, D., Klypin, A., Klimek, E. S., et al. 2014, MNRAS, 442,1545Chabrier, G. 2003, PASP, 115, 763Cheung, E., Faber, S. M., Koo, D. C., et al. 2012, ArXiv e-printsChoi, J., Conroy, C., Moustakas, J., et al. 2014, ApJ, 792, 95Cimatti, A., Cassata, P., Pozzetti, L., et al. 2008, A&A, 482, 21Conroy, C. 2013, ARA&A, 51, 393Conroy, C. & Gunn, J. E. 2010, ApJ, 712, 833Daddi, E., Renzini, A., Pirzkal, N., et al. 2005, ApJ, 626, 680Damjanov, I., McCarthy, P. J., Abraham, R. G., et al. 2009, ApJ,695, 101Dekel, A., Birnboim, Y., Engel, G., et al. 2009, Nature, 457, 451Dekel, A. & Burkert, A. 2014, MNRAS, 438, 1870Dekel, A., Zolotov, A., Tweed, D., et al. 2013, MNRAS, 435, 999Diamond-Stanic, A. M., Moustakas, J., Tremonti, C. A., et al.2012, ApJ, 755, L26Elbaz, D., Daddi, E., Le Borgne, D., et al. 2007, A&A, 468, 33Elbaz, D., Dickinson, M., Hwang, H. S., et al. 2011, A&A, 533,A119Falc´on-Barroso, J., S´anchez-Bl´azquez, P., Vazdekis, A., et al.2011, A&A, 532, A95Fang, J. J., Faber, S. M., Koo, D. C., & Dekel, A. 2013, ApJ, 776,63Galametz, A., Grazian, A., Fontana, A., et al. 2013, ApJS, 206, 10Genzel, R., F¨orster Schreiber, N. M., Rosario, D., et al. 2014,ArXiv e-printsGladders, M. D., Oemler, A., Dressler, A., et al. 2013, ApJ, 770,64Gon¸calves, T. S., Martin, D. C., Men´endez-Delmestre, K., Wyder,T. K., & Koekemoer, A. 2012, ApJ, 759, 67 Grogin, N. A., Kocevski, D. D., Faber, S. M., et al. 2011, ApJS,197, 35Guo, Y., Ferguson, H. C., Giavalisco, M., et al. 2013, ApJS, 207,24Guo, Y., Giavalisco, M., Cassata, P., et al. 2012, ApJ, 749, 149Hern´an-Caballero, A., Alonso-Herrero, A., P´erez-Gonz´alez, P. G.,et al. 2014, MNRAS, 443, 3538Hopkins, P. F., Hernquist, L., Cox, T. J., et al. 2006, ApJS, 163, 1Kauffmann, G., Heckman, T. M., Tremonti, C., et al. 2003,MNRAS, 346, 1055Kelson, D. D., Illingworth, G. D., Franx, M., & van Dokkum,P. G. 2001, ApJ, 552, L17Kennicutt, Jr., R. C. 1998, ARA&A, 36, 189Koekemoer, A. M., Faber, S. M., Ferguson, H. C., et al. 2011,ApJS, 197, 36Kriek, M., Shapley, A. E., Reddy, N. A., et al. 2014, ArXive-printsKriek, M., van Dokkum, P. G., Labb´e, I., et al. 2009, ApJ, 700,221Kriek, M., van Dokkum, P. G., Whitaker, K. E., et al. 2011, ApJ,743, 168Liu, F. S., Guo, Y., Koo, D. C., et al. 2013, ApJ, 769, 147Magdis, G. E., Rigopoulou, D., Huang, J.-S., & Fazio, G. G.2010, MNRAS, 401, 1521Martin, D. C., Wyder, T. K., Schiminovich, D., et al. 2007, ApJS,173, 342McLean, I. S., Steidel, C. C., Epps, H., et al. 2010, in Society ofPhoto-Optical Instrumentation Engineers (SPIE) ConferenceSeries, Vol. 7735, Society of Photo-Optical InstrumentationEngineers (SPIE) Conference SeriesMcLean, I. S., Steidel, C. C., Epps, H. W., et al. 2012, in Societyof Photo-Optical Instrumentation Engineers (SPIE) ConferenceSeries, Vol. 8446, Society of Photo-Optical InstrumentationEngineers (SPIE) Conference SeriesMiller, S. H., Bundy, K., Sullivan, M., Ellis, R. S., & Treu, T.2011, ApJ, 741, 115Muzzin, A., Marchesini, D., van Dokkum, P. G., et al. 2009, ApJ,701, 1839Naab, T., Johansson, P. H., Ostriker, J. P., & Efstathiou, G.2007, ApJ, 658, 710Nelson, E., van Dokkum, P., Franx, M., et al. 2014, Nature, 513,394Newman, A. B., Ellis, R. S., Andreon, S., et al. 2013, ArXive-printsNewman, A. B., Ellis, R. S., Bundy, K., & Treu, T. 2012a, ApJ,746, 162Newman, A. B., Ellis, R. S., Treu, T., & Bundy, K. 2010, ApJ,717, L1032