Optical spectroscopy and initial mass function of $z=0.4$ red galaxies

0.36 MB·

by  Baitian Tang

Mon.Not.R.Astron.Soc.000,1–12() Printed16January2017 (MNLATEXstylefilev2.2)
z = 0.4
Optical spectroscopy and initial mass function of
red galaxies
7 Baitian Tang1,2⋆ and Guy Worthey1†
1 1Department of Physics and Astronomy, Washington State University,Pullman, WA 99163-2814, USA
0 2Departamento de Astronom´ıa, Casilla 160-C, Universidad de Concepci´on, Concepci´on, Chile
2

n
a
J
Accepted .Received;inoriginalform2014

3
1

ABSTRACT
]
A
Spectralabsorptionfeaturescanbeusedtoconstrainthestellarinitialmassfunc-
G
tion (IMF) in the integrated light of galaxies. Spectral indices used at low redshift
. are in the far red, and therefore increasingly hard to detect at higher and higher
h
p redshifts as they pass out of atmospheric transmission and CCD detector wavelength
- windows. We employ IMF-sensitive indices at bluer wavelengths.We stack spectra of
o red,quiescentgalaxiesaroundz =0.4,fromthe DEEP2GalaxyRedshift Survey.The
r z =0.4redgalaxieshave2Gyraverageagessothattheycannotbepassivelyevolving
t
s precursors of nearby galaxies. They are slightly enhanced in C and Na, and slightly
a
depressedin Ti. Split by luminosity, the fainter half appears to be older,a resultthat
[
shouldbe checkedwith largersamples in the future. We uncoverno evidence for IMF

1 evolution between z = 0.4 and now, but we highlight the importance of sample se-
v lection, finding that an SDSS sample culled to select archetypal elliptical galaxies at
3
z∼0isoffsettowarda morebottomheavyIMF.Othersamples,including ourDEEP2
8
sample,showanoffsettowardamorespiralgalaxy-likeIMF.Allsamplesconfirmthat
6
the reddest galaxies look bottom heavy compared with bluer ones. Sample selection
3
alsoinfluencesage-colortrends:red,luminousgalaxiesalwayslookoldandmetal-rich,
0
. but the bluer ones can be more metal-poor, the same abundance, or more metal-rich,
1
depending on how they are selected.
0
7 Key words: galaxies: abundances — galaxies: evolution — galaxies: elliptical and
1 lenticular, cD — galaxies: luminosity function, mass function
:
v
i
X
r
a 1 INTRODUCTION Salpeter(1955)describedtheIMFwithapower-lawdis-
tribution, ξ(m) ∝ m−α, and adopted an IMF slope (α) of
Thestellarinitialmassfunction[IMF;ξ(m)]iskeyinmany
2.35forsolarneighbourhoodstarsnearthemassofthesun.
active research fields, such as early universestudies, galaxy
Consideringstarsofveryhighandverylowmass,thepower
evolution, star cluster evolution, and star formation. The
lawdoesnothold,andtheGalacticIMFisnowconsideredto
IMFregulatesthenumberdistributionofstellarpopulations
peakatafewtenthsofonesolarmass(Miller & Scalo1979;
asafunctionofmass,dN =ξ(m)dm,leadingtoimpactson
Scalo 1986; Kroupa2001;Chabrier 2003).In terms of func-
theluminosity function, integrated mass to light (M/L) ra-
tional forms to model the IMF, oft-cited examples are the
tio,andnumberofstellarremnants.DirectIMFderivations
Kroupa(2001)seriesofpowerlawsandtheChabrier(2003)
are limited by observational capabilities and uncertainties log-normal distribution for m<1 M⊙1 as representativeof
concerning the stellar mass-luminosity relation, stellar evo-
thelog-normalprobabilitydensityfunctionofturbulentgas.
lution,dynamicalevolution,binaryfraction,countingstatis-
Because both models are calibrated by observational data,
tics, and other factors. To make the IMF even harder to
they show a similar distribution between 0.2 and 0.8 M⊙
study, there is the distinct possibility that the IMF might
(Bastian et al. 2010; Greggio & Renzini 2012; Offneret al.
vary in different galactic environments (Larson 1998, 2005;
2014).A“steeper”IMFwithmorelowmassstarsistermed
Marks et al. 2012; Hopkins2013; Chabrier et al. 2014).
⋆ E-mail:[email protected] (BT)
† E-mail:[email protected] (GW) 1 apowerlawfunctionform&1M⊙
(cid:13)c RAS
2 Baitian Tang and Guy Worthey
bottom heavy and a “shallower” IMF with more high mass have an IMF similar to a spiral galaxy. Since that conclu-
stars is termed top heavy. sion seems incorrect, it may be that elliptical galaxies are
Extending local studies based on counting individual also subject to star formation events periodically (gas rich
stars to external galaxies, recent studies explore the uni- mergersorgasaccretion)butstarformationdoesnotlinger,
versality of the IMF using integrated light and also dy- but is rapidly quenched. In this variant of the hierarchical
namical models. Gravity-sensitive integrated spectral fea- formationscenario,ellipticalgalaxiesmaydeveloptheirown
turessuchasthegiant-sensitiveCaiitriplet andthedwarf- characteristicIMF,andalso,thatIMFmayevolveovercos-
sensitive Nai and FeH Wing-Ford band may indicate that mic time; early ellipticals will have a spiral-like IMF, but
systematicIMFvariationexistsasafunctionofstellarveloc- as time and mass-buildup goes on and they become more
ity dispersion (Cenarro et al. 2003; van Dokkum & Conroy idiosyncratic, the elliptical galaxies may develop a distinct
2010; Conroy & van Dokkum 2012b). This trend was fit by elliptical-flavoured IMF due to the altered star formation
Ferreras et al.(2013)andSpiniello et al.(2014),butthesin- environment.
gle power law IMFslope (α) of theformer studyis afactor There is some theoretical support for expecting an
oftwogreaterthanthelatterone.Minimizingtheuncertain- evolving IMF.
ties arising from the ingredients of stellar population (SP)
modelsappearscrucialtofutureIMFstudy(Spiniello et al.
2015). Another popular approach for studying unresolved 1.1 IMF theories
stellar systems is using dynamical models with a dark
Logically,thesteeperIMFsofETGswereeitherimposedat
matter halo involved. The IMF is estimated by compar-
an early creation epoch or haveevolved overtime. Present-
ing the M/L ratios of the dynamical models and the stel-
dayETGsareunlikelytohavesufferedbuildupsofonly low-
lar population models (Augeret al. 2010; Cappellari et al.
mass stars in the last third of the universe. We see no evi-
2012;Duttonet al.2012;Posacki et al.2015).Forexample,
dence of such an odd star formation mode ongoing, c.f. the
Cappellari et al. (2012) concluded an universal IMF is in-
case study of giant elliptical galaxy NGC 5128 (Neff et al.
consistent with early-type galaxies (ETGs), although this
2015), caught in a gas accretion event, which emits plenty
kinematicresultisunabletodistinguishbetweenmorestel-
of UV light, indicating that massive stars are forming.
lar remnants (from a top heavy IMF) and relatively more
Historically, the simple Jeans mass model and the tur-
low mass stars (from a bottom heavy IMF).
bulentJeansmassmodelwereinfluential.ThesimpleJeans
Simple chemical evolution of a time-independent,
mass model hypothesizes that the peak mass of the IMF
bottom-heavy IMF suggests a small number of stellar rem-
is simply a reflection of the mean Jeans mass. For exam-
nants and also implies lower metallicities for the fossil
ple, the Larson model (Larson 1998, 2005) showed that the
stars.Inempiricalrebuttal,observationssuggestsuper-solar JeansmassvarieseitherasT3/2ρ−1/2 orasT2P−1/2,where
metallicities in elliptical galaxies e.g., Trager et al. 2000a,b;
T is temperature, ρ is density, and P is pressure. In the
Tang et al. 2009). Also, the number of stellar remnants is
early Universe, the high cosmic background temperature,
probably not small. Kim et al. (2009) suggested the num-
low metallicity, and intense radiation from young stars and
ber of low-mass X-ray binaries in three nearby elliptical
core-collapse supernovainevitablyincreasetheproto-stellar
galaxies with mass about 1011 M⊙ is similar to that of the cloud temperature. But the relation between T and ρ (or
Milky Way. Peacock et al. (2014) found a constant num-
P)isstillneededtoestimatetheJeansmass.Larson(2005)
ber2 of black holes and neutron stars among eight different
noted the T −ρ relation drawn from the observational and
mass ETGs3. To reconcile these facts with a bottom-heavy
theoretical results of that time:
IMF, Weidneret al. (2013a,b) simulated galaxy evolution
with a time-dependent IMF, in which the IMF slope steep- T =4.4ρ−0.27K, ρ<10−18gcm−3 (1)
18
ensas thestar formation rate decreases gradually (Also see
T =4.4ρ+0.07K, ρ>10−18gcm−3 (2)
Gargiulo et al.2015forasemi-analyticalmodelwithsimilar 18
IMF slope assumption). whereρ18isthedensityinunitsof10−18 gcm−3.Therefore,
IMF evolution, if it occurs in nature, may potentially Larsonsuggestedtheearlyuniversefavorsatop-heavyIMF,
bedetected through studiesof distant galaxies, and topos- where a low density T −ρ relation was assumed.
siblydetect suchisthemotivation forthepresentpaper.In TheturbulentJeansmassmodellinksthecharacteristic
the formation scenario that giant elliptical galaxies formed mass of stars to the galactic-scale processes responsible for
very eary in cosmic history and have been passively evolv- setting the characteristic temperatures and linewidth-size
ing ever since, the IMF cannot change over time, though relations of molecular clouds (Krumholz 2014). The most
the IMF for elliptical galaxies can be very different from representative models are given by Chabrier et al. (2014)
spiralgalaxies.Inthehierarchicalgalaxyformationpicture, and Hopkins (2013). These models are widely cited, partly
elliptical galaxies are the end result of merger trees that due to the consistency between their model prediction and
start with irregular and spiral galaxies as basic ingredients, recent observations: a bottom-heavy IMF for massive early
and theelliptical galaxy should bethesum of its parts and typegalaxy.
However,aspointedoutbyKrumholz(2014),themass
oftheIMFpeakintheHopkinsmodelcan beexpressedas:
2 ScaledbytheamountofK-bandstellarlightcovered c4
3 To connect black holes and other stellar remnants to the low- Mpeak ≈ QGs2Σ (3)
mass IMF posits a very coherent and simplefunctional form for
theIMF. Innature, thehighmassendof theIMFmaybeinde- where cs is the sound speed, Q≈1 is the Toomre stability
pendentofthelowmassIMF. parameter for the disk, and Σ is the gas surface density. cs
(cid:13)c RAS,MNRAS000,1–12
Optical spectroscopy and initial mass function of z = 0.4 red galaxies 3
and Σ are both large for high surface density star forma- µmincreasethedifficultyofmeasuringthesefeaturesindis-
tion, in which elliptical galaxies are assumed to form. The tantobjects.ItmotivatesustolookforIMF-sensitiveindices
parameters may or maynot implya bottom-heavyIMF for in a more accessible band. For example, the Na D, TiO1,
ETGs. andTiO2 indicesfromtheLick/IDSsystem(Worthey et al.
1994; Trager et al. 1998) are known to be IMF-sensitive.
In addition, several optical IMF-sensitive indices published
1.2 IMFs observed at different redshifts bySpiniello et al. (2014) andLa Barbera et al. (2013) have
given us more options in index selection (Table 1).
IfETGsperiodicallyformstars,buttheepisodesarerapidly
Theseindicesarealsosensitivetopopulationage,over-
quenched,thenmostETGswillbeobservedinthequiescent
all heavy element content, and altered abundance ratios in
phasesoftheirstarformationdutycycle,andyetasagroup
the elements that give rise to the spectral features them-
theIMFmayevolveovertime.Inthatcase,lookbackstudies
selves, such as Na, Ti, and Ca. Fortunately, element ratios
may uncoverthe drift in IMF.
seemfairlywellconstrained,judgingbythegoodagreement
There is further research. Shetty& Cappellari (2014)
onelementmixtureinrecentpapers(Johansson et al.2012;
derivedthemass/light(M/L)ratiosof68fieldgalaxiesinthe
Conroy et al.2014;Worthey et al.2014b).Additionally,the
redshiftrangeof0.7−0.9withbothdynamicalmodellingand
element sensitivity problem may be eased if multiple IMF-
stellar population modelling. The comparison of (M/L)
dyn sensitiveindiceswithdifferentelementsensitivitiesareused.
and (M/L) implies a Salpeter IMF, which is also pos-
pop Observationally, spectra from the DEEP2 redshift sur-
sessed by nearby galaxies with similar masses. Meanwhile,
vey (Newman et al. 2013), which targeted galaxies over a
Mart´ın-Navarro et al.(2015b)studiedtheTiO2 indicesofa broad span of redshifts in the spectral range 6500 ˚A−9100
sample of 49 massive quiescent galaxies at 0.9 < z < 1.5. ˚Ausing KeckObservatory,offer themselvesasan appropri-
The heaviest galaxies (M∗ > 1011.0 M⊙) show a bottom- ate data set. The combination of spectral observations plus
heavy IMF and lighter galaxies (1010.5 <M∗ < 1011.0 M⊙) new spectral indicators may allow the measurement of the
donot.TheyalsoconcludedthattheIMFofmassivegalax-
IMF overcosmic time.
ies has remained unchanged for thelast ∼8 Gyr.
This paper is organized as follows: The procedure of
Luminosity evolution may also betray the IMF slope
stacking DEEP2 spectra is illustrated in §2. We compare
in that a top-heavy galaxy fades more rapidly than a
themeasured indices from these composite spectra with lo-
galaxy with the standard IMF, since the present-day lu-
cal measurements and two different models in §4. The im-
minosity of ETG mainly comes from old stellar popula-
plications on IMF evolution are discussed in §5, and then a
tions (∼ 1 M⊙). According to Tinsley & Gunn (1976), the brief summary of theresults are given in §5.1.
luminosity of an old stellar population is proportional to
t−1.6+0.3α. Therefore, a shallower IMF implies a more dra-
maticluminositychangeoverafixedamountoftime.Inthat
2 SPECTRAL REDUCTION
spirit, vanDokkum (2008) compared the luminosity evolu-
tion(∆log(M/LB))tocolourevolution(∆(U−V))formas- 2.1 Sample Selection
sive galaxies in clusters at 0.02<z <0.83. The luminosity
evolution oftheseobservedgalaxies isfaster thanthetrend The DEEP2 Galaxy Redshift Survey utilizes the DEIMOS
predicted by the Maraston (2005) models with a standard multi-object spectrograph (Faberet al. 2003) on the Keck
IMF.Thus,theIMFsinthesegalaxiesmightbetop-heavy4. II telescope. Most of the spectra cover 6500−9100 ˚A, with
More supportforashallow IMFathigh redshiftcomes spectral resolution R = λ/∆λ ∼ 6000 (Newman et al.
from Dav´e (2008), who successfully brought the observed 2013).Theoptical IMF-sensitiveindicesaredefinedaround
and predicted M∗–SFR relation into broad agreement by 4500−6500 ˚A,listed in Table1.Inordertomatchobserved
modellingthecharacteristicmassMˆ asafunctionofredshift: andemittedspectralwavelengthrangesforthisindexset,we
Mˆ=0.5(1+z)2 M⊙, where z<2. chose galaxies around z =0.4, corresponding to a lookback
Thisrichvarietyofresultsisfascinating,yetambiguous. time of about 4.3 Gyr.
Indetail,weselected galaxies6 with0.3 6z 60.5from
the Data Release 4 (DR4) redshift catalog, and made sure
1.3 Observational strategy these galaxies had photometric measurements by matching
them with theDR4photometric catalogs. The photometric
Integrated light spectral features that are sensitive to dataweretakenwiththeCFH12Kcameraonboardthe3.6-
stellar surface gravity are often used for studies of meter Canada-France-Hawaii Telescope (Coil et al. 2004).
nearby galaxy IMFs. M-type giants and dwarfs emit The chosen redshift range balances thenumberof available
most of their light and have important diagnostic fea- IMF-sensitiveindices,thesamplesize,andtheredshiftsim-
tures at red wavelengths5 (van Dokkum& Conroy 2010; ilarityofoursample.Notethatz<0.7galaxieswererejected
Conroy & van Dokkum2012b;Smith et al.2012).However, duringtheobservationinthreeoutofthefourDEEP2fields
at cosmic distances, the inevitable redshift of spectral lines (excepttheExtendedGrothStrip,DEEP2field1),andthus
andtherelativelylowqualityofspectraobtainablebeyond1 galaxies at 0.3 < z < 0.5 are less numerous than other
higher-redshiftsamples(e.g.,Schiavon et al.2006).Next,to
minimize the contribution of late type galaxies (LTGs), we
4 Greggio&Renzini(2012)pointsoutthattheluminosityevo-
lutionat0<z<1.5canonlyconstraintheIMFbetween 1M⊙
and1.4M⊙. 6 Confirmedbyexaminingthe “CLASS”parameter inthecata-
5 e.g.,Caiiλ8600, Naiλ8190, FeHWing-Fordbandλ9900 log.
(cid:13)c RAS,MNRAS000,1–12
4 Baitian Tang and Guy Worthey
Table 1.OpticalIMF-sensitiveIndices
Index Units Blue Pseudo-continuum Central Feature Red Pseudo-continuum Source
bTiO mag 4742.750-4756.500 4758.500-4800.000 4827.875-4847.875 2
aTiO mag 5420.000-5442.000 5445.000-5600.000 5630.000-5655.000 2
NaD ˚A 5860.625-5875.625 5876.875-5909.375 5922.125-5948.125 1
TiO1 mag 5816.625-5849.125 5936.625-5994.125 6038.625-6103.625 1
TiO2 mag 6066.625-6141.625 6189.625-6272.125 6372.625-6415.125 1
TiO2 mag 6066.625-6141.625 6189.625-6272.125 6442.000-6455.000 3
SDSS
CaH1 mag 6342.125-6356.500 6357.500-6401.750 6408.500-6429.750 2
1 – Worthey et al. (1994); 2 – Spiniello et al. (2014); 3 – La Barbera et al. (2013)
selected red galaxies using the galaxy colour-magnitude di-
agram(CMD).Weplotthe(B−R)vs.RCMDinFigure1.
302galaxieshave(B−R)colourredderthan1.8magandR
2.5
bandmagnitudebrighterthanR=22mag.Theredshiftsof
these galaxies are reasonably well determined: 260 galaxies
have quality code 4 (99.5% reliability rate), and 42 galax- 2.0
ies havequality code3(95% reliability rate, Newman et al. R
m
2013). 1.5
−
AccordingtoWeiner et al.(2005),atz61,LTGscom- B
m
prise about 25% of the red population. Since we lack mor- 1.0
phology information, we cannot exclude LTGs on the ba-
sic of isophotal structure. However, we may use [Oiii] and 0.5
Hα emission lines to eliminate galaxies with strong emis-
sion lines (Weiner et al. 2005; Schiavon et al. 2006) and
therefore ongoing star formation or active galactic nuclear 20 21 22 23 24
m
activity. Based on the wavelength coverage of the dered- R
shifted spectra, [Oiii]EWs7 can bemeasured in 282 galax-
Figure1.Galaxycolourmagnitudediagram.Darkerblueregions
ies.Amongthose,7galaxies have[Oiii]EW<−5˚A8.After
havegreaternumberdensity.Theredsequenceisthepeakinthe
[Oiii]selection,275galaxiesareleftinthesample(SampleI). upper left, the blue cloud is the peak at lower right, and the
Similarly, 5 out of 77 galaxies with Hα EW measurements greenvalleyisthelowdensitygapbetween.Theredgalaxiesare
have Hα EW< −5 ˚A, and we define thus an Hα selected selectedbymR <22mag(verticalredline),andmB−mR >1.8
sample (Sample II) consisting of 72 galaxies. mag(horizontal redline). Amagnitude cutforsubsamplingwas
With a large galaxy sample at z ∼ 0.9 and a selec- madeatmR=20.5mag(verticaldotted redline).
tion criteria of [Oii]EW>−5 ˚A, Schiavon et al. (2006) es-
timated theLTG proportion of their sample is at most 5%.
2.2 Composite Spectra and Index Measurements
Thoughthereissimilaritybetweenourselectionmethodand
theirs,oursamplesizeissmallerandlesscomplete,thuswe We retrieved the one dimensional spectra from the DEEP2
estimatetheLTGportion ofoursampleis5−25%,between Data Release 4 website9. The reduced two dimensional
the predictions of Schiavon et al. (2006) and Weiner et al. spectra obtained from DEIMOS are flat-corrected and
(2005). wavelength-calibrated. The pipeline also takes care of the
Besides a few bright, red LTGs in the sam- sky subtraction and cosmic ray rejection. One dimensional
ple, there may also be some low power AGN. The spectra are extracted from each of the slitlets using an op-
Baldwin−Phillips−Terlevich (BPT) diagnostic diagram timal extraction method (Horne 1986) that assumes a con-
(Baldwin et al. 1981) suggested that [Oiii] EW< −5 ˚A, or stant Gaussian profile at all wavelengths, which implies the
HαEW<−5˚A,whichisusedasoneofoursampleselection spectral extraction region is the whole visible galaxy. Flux
criteria,cannoteffectivelyeliminatetheAGNcontributions. calibration isnotattemptedalong theprocess, andtheflux
unit is DEIMOS counts perhour(e−/hour).
To stack the spectra we used pipeline programs devel-
opedbytheDEEP2team(Cooper et al.2012).Wemodified
the “coadd spectra” program to meet our (mild) need for
flux correction. In measuring spectral indices, scalings and
even linear corrections do not affect the result. However,
if the response curve correction is more complicated than
7 DefinedinGonz´alez(1993) linear, there is a mild effect, and so we included the flux
8 Weineretal.(2005)showedthemedianerrorinrest-frameEW correction.Wedividedeachspectrumbythethroughtputof
is6.2˚A.Schiavonetal.(2006)suggested−5˚AfortheEWlimit,
though they use [Oii] due to a different rest-frame wavelength
range. 9 http://deep.ps.uci.edu/DR4/spectra.html
(cid:13)c RAS,MNRAS000,1–12
Optical spectroscopy and initial mass function of z = 0.4 red galaxies 5
ods in Serven & Worthey (2010), that is, via measurement
of an Hα index, then using Mg b to estimate a continuum
SingleSpectrum slope correction. The corrections were substantial since the
1.0 stacked Hα indices were generally in emission, from 0.5 ˚A
set forthelow-luminositysubsampleupto1.1˚Aforhighlumi-
Off nosity.
y
call 0.8 We also subdivide the sample by luminosity at mR =
erti SampleI 20.5mag.Cross-correlation indicatesa15%smallervelocity
V
dispersion in the faint subsample and a 5% larger velocity
F,λ 0.6
dispersioninthebrightsubsamplecomparedtotheaverage.
e,
ativ These differences were propagated through the smoothing
Rel 0.4 and index measurement procudures.
SampleII
4800 4900 5000 5100 5200 5300
3 MODELS AND LOCAL OBSERVABLES FOR
RestWavelength(A˚)
COMPARISON
Figure 2. A swath of DEEP2 spectra beforeand after stacking
toillustrateS/Nimprovement.SampleIstacks 275spectra,and Stellar population model information and additional obser-
SampleIIstacks 72spectra toachieve higher S/N atcost ofthe vational material is needed for a fair comparison of IMF
lossofindividualgalaxyidentities. indicators.
Worthey models: Models (Worthey 1994;
Trager et al. 1998) that start with empirical stellar li-
DEIMOS in spectroscopic mode for the gold 1200-line/mm braries, then use synthetic spectra (Lee et al. 2009) to
grating10. Next, each individual spectrum is shifted to the gauge the effects of detailed chemical composition were
rest-frameandnormalizedbydividingthemedianspectrum. employed,with a few ongoing improvements.
Thecompositespectraareachievedbycoaddingthenormal- For this version, the isochrones of Bertelli et al. (2008,
izedspectra,wheretheinversevarianceofeachpixelisused 2009) with the thermally-pulsing asymptotic giant branch
as weight (see Figure 2). (TP-AGB) treatment described in Marigo et al. (2008)
We estimated the velocity dispersions (σ) of the com- were employed. This set of isochrones has a low mass
positespectraintwoways;theFaber& Jackson(1976)σ−L limit of 0.15 M⊙. In philosophy similar to Poole et al.
scalingrelation andcross-correlation. Forluminosity,weK- (2010), indices were measured from four stellar spec-
correctedtheRbandmagnitudetoBbandmagnitudeusing tral libraries (Valdeset al. 2004; Worthey et al. 2014a;
the code of Blanton & Roweis (2007). The velocity disper- S´anchez-Bla´zquezet al.2006;Rayneret al.2009),alltrans-
sion was calculated by adopting the Faber-Jackson relation formed to a common 200 km s−1 spectral resolution. Mul-
presented in Whitmore & Kirshner (1981). The average σ tivariate polynomials were fit over five overlapping temper-
of our 302 red galaxy pool is about 235±40 km s−1. For ature zones as a function of θ = 5040/T , log g, and
eff eff
thelatter estimate, we cross-correlated the composite spec- [Fe/H], then smoothed and summarized in a lookup table.
tra with model stellar spectrum templates, with the width Tocomputetheintegratedpropertiesofthefinalmodel,the
of thecross-correlation function fitted with a Gaussian and isochronesplusanIMFgivethenumbersofstarsineachbin
treated as in Tonry & Davis (1979). Composite spectra of of the isochrone. The stellar index was found (and any op-
Sample I and Sample II show σ ∼241 km s−1, and ∼231 tional chemical element tweaksimposed), then decomposed
km s−1, respectively. The velocity dispersions determined into“index”and“continuum”fluxes,whichwereseparately
by thesetwo methods agree. added,then,after summation, re-formed intoan index rep-
We broadened the spectra to 300 km s−1 (σ2 = resenting the integrated value. To transform from 200 km
broaden
σ3200−σs2ample) , and measured spectral indices and associ- s−1 to 300 km s−1 small additive corrections for each in-
ated errors propagated from the errors on each flux pint. dexwereestimated byGaussian-broadeninghighresolution
In terms of studying indices rather than the spectra di- syntheticcomposite spectra.
rectly,wechoseindicesbecausemoretheoreticalpredictions For this work, we updated the models with additional
are available, systematic fluxing issues are minimized, and IMF options. We calculate SSP models with the Kroupa
analysis is more secure and direct. Our index table con- (2001) IMF to represent LTGs and two power law IMF
sistsofLick-styleindicespresentedinWortheyet al.(1994); slopes: α =2.35, 3.0. For each IMF slope, our SSP mod-
Trager et al. (1998); Serven et al. (2005), and Table 1. The elsaregivenattheagesof5,10Gyr,with[M/H]={−0.33,
measuredindicesanderrorsareshownasred(SampleI)and 0, 0.37}.
lightgrey(SampleII)filledcircleswitherrorbarsinFigure CvD12: We also employ the Stellar Population Syn-
4. The errors plotted are pixel bypixel measurement errors thesis (SPS) presented in Conroy & van Dokkum (2012a)
propagatedthroughtotheindices(anddonotincludeacon- (CvD12). CvD12 models scaled-solar old stellar popula-
tributionfromvelocitydispersionuncertaintythatmodestly tions (>3 Gyr) with empirical spectral libraries: MILES
affects narrower indices). (S´anchez-Bla´zquezet al. 2006) and IRTF (Cushing et al.
TheHβ indexwascorrected foremission viathemeth- 2005; Rayneret al. 2009). Within the limitation of near-
solar abundance, they also probe individual abundance
variationsandαelementenhancementswithsyntheticspec-
10 http://www.ucolick.org/∼ripisc/results.html tral libraries as well as provide IMF variation options. For
(cid:13)c RAS,MNRAS000,1–12
6 Baitian Tang and Guy Worthey
our illustrations, we adopt the spectra presented in CvD12 4 RESULTS
from http://scholar.harvard.edu/cconroy/sps-models.
Figures 3 and 4intercompare index valuesfrom themodels
These spectra are theSPS model output at four ages (3, 5,
and the observations. The [MgFe]11 is chosen as the x-axis
7, and 9 Gyr) with four different types of IMFs (α = 3.5,
variable since it tracks mostly [M/H] rather than [α/H] or
α = 3.0, Salpeter, and Chabrier). We measured indices
after we broadened thespectra to300 km s−1. [Fe/H],thoughitretainsconsiderable sensitivitytopopula-
tion age.
Local galaxies:Asstellar population modelsstill suf- ExaminingFig.3,thetop-leftpanelshowsage-sensitive
fer from uncertainties (Charlot et al. 1996; Conroy et al. Hβ.Themodelgridshowtwoisochronsat 2,5,and10Gyr
2009;Tang et al. 2014),local galaxies are usefulfor empiri- with dotsat [M/H] =−0.33, 0.0, and 0.37. Grid extrapola-
cal comparison. Since oursample galaxies consists of a ma- tionsareapproximatelylinear.VariableIMFmodelsarenot
jority of ETGs and a few LTGs, templates of both galaxy shown in Fig. 3,and theIMFis apower law with low mass
types are desired. Note that spectra from all sources were cutoffof0.15M⊙.TheDEEP2galaxiesaremarkedwithcir-
broadened to 300 km s−1 for purposes of comparison. clesanderrorbars.Thetwosamplegrandaverages,marked
withlarger circles, areflankedbysmaller circlesthat repre-
sentasplitofthesampleatR=20.5.Thisisapproximately
half by number, although the fainter half has larger errors.
The Graves passive galaxies (open black diamonds repre-
sentingaseriesofbinsinvelocitydispersion),Dobospassive
galaxies (open triangles representing three color bins), and
(i) Wewerekindlyprovidedwiththeupdatedmeanspec-
Dobos RED galaxies (open stars representing AGN activ-
tra of non-star forming, non-LINER galaxies by G. Graves.
ityincreasingfromsmallsymbolstolargesymbols)arealso
Galaxies from SDSS DR7 were selected with the follow-
plotted.
ing criterion: (1) 0.025 < z < 0.06; (2) No detectable Hα
In terms of velocity dispersions, the Dobos galaxies
or [Oii] 3727 emission (Peek & Graves 2010); (3) Median
locate about the 3rd (from the left) Graves bin, while
S/N>5 per ˚A. Then, the galaxies were divided into six
the DEEP2 galaxies locate between the 5th and 6th. The
bins in logσ: 1.86−2.00, 2.00−2.09, 2.09−2.18, 2.18−2.27,
DEEP2 galaxies should be younger by ∼ 4.3 Gyr due to
2.27−2.36,2.36−2.50;(4)furthercullingifthegalaxieswere
lookback time, all else being equal. The Hβ emission cor-
farfromthefundamentalplane.SeeConroy et al.(2014)for
rections for the active galaxies are more uncertain because
amoredetaileddescriptionofthesample,wherethelastve-
thecorrections(Serven& Worthey2010)werebuiltforstar
locitydispersionbinofoursamplearesub-dividedintotwo.
formation scenarios, not AGN.
Finally, Hβ was corrected via Serven & Worthey (2010).
The trends apparent in the Hβ diagram are that more
(ii) Dobos et al. (2012) classified the galaxies of SDSS massive/redder galaxies appear older and more metal rich
DR7 by both colour and nuclear activity. We select the relativetothemodelgrid,atrendseensincetheinventionof
twomostrelevantsubsamples:1)Thepassivegalaxies(with this diagnostic diagram (Faberet al. 1992). There appears
no Hα detection) split by color into red, green, and blue to be only one contradiction apparent in the Hβ diagram,
sub-samples, and 2) a red sample, sub-divided into five which is that the Dobos sequences might be expected to
smaller samples based on nuclearactivity from low tohigh: cross the Graves sequence at the third diamond (matching
(Their RED 1, RED 2, RED 3, RED 4, and RED 5 sam- velocity dispersions), not converge at the 6th, as observed.
ples). In order to avoid Malmquist bias, each sub-sample However, that is only a contradiction if thesamples are ex-
isconstrainedbyredshiftandabsolutemagnitudetoensure pectedtobeequivalent.Ifwepositthatsampleselectioncan
volume-limitedsampling.Thedetailedredshiftandabsolute have an influence, then what we see are diagnostics of the
magnitude ranges can befound in their Table 2. For exam- sampleselection method.PointstonoteintheHβ diagram:
ple, thepassive samples are selected inside0.03<z <0.14,
−20.5<M <−21.5.Notethatthissampleismoredistant • ETGsformanarrowsequencein[M/H]butrangeover
r
than the Graves sample. Applying the Faber-Jackson law a large range of average age.
to thissample, we infer velocity dispersions around 142 km • TheLTGtemplatehadalargenebularemissioncorrec-
s−1 with narrow variance, so we adopt this value for pur- tion, but within that uncertainty appears to be of interme-
poses of correcting to 300 km s−1 for all the subsamples. diateagebutafewtenthsmoremetalpoorthantheETGs.
Hβ corrections were applied similar to the DEEP2 sample. • The three samples (Graves nonLINER,Dobos Passive,
Note that the Dobos passive sample’s selection criteria are Dobos Red) tilt progressively such that the bluer Graves
themostsimilartoourDEEP2galaxies.Thispointbecomes galaxiesaremoremetalpoor,thebluerDobosPassivegalax-
important when we discuss IMFevolution. ies are on an isometallicity track, and thebluer Dobos Red
galaxies (theleast AGN active) are more metal rich.
(iii) We also retrieved the Sb galaxy template from
• The DEEP2 samples lie slightly more metal rich from
Kinney et al. (1996). This optical template is a combina-
the convergent red end of the other samples, but much
tion of two Sb galaxies, NGC 210 and NGC 7083, whose
younger. The youthening effect is much stronger than pas-
spectra were obtained at the CTIO 1 m telescope with the
siveevolution.
two-dimensional Frutti detector. The CTIO spectra covers
3200−10000 ˚A with a resolution of 8 ˚A, which was then
smoothedto300kms−1withasmoothingkernelthatvaries
withwavelength.Hβ correctionswereappliedsimilartothe 11 [MgFe] ≡ pMgb∗(Fe5270+Fe5335)/2, from Gonz´alez
DEEP2 sample. (1993)
(cid:13)c RAS,MNRAS000,1–12
Optical spectroscopy and initial mass function of z = 0.4 red galaxies 7
3.5
SampleI 0.37
SampleII 7
3.0
6
2Gyr
2.5
βH e4668 5 0.0
F
2.0
5Gyr 4
[M/H]=-0.33
1.5 3
10Gyr
SDSSnonLINER
4.0 SDSSRed
3.0 SDSSPassive
LTG
b 3.5 >e 2.5
F
g <
M
3.0
2.0
2.5
1.5
2.0 2.5 3.0 3.5 2.0 2.5 3.0 3.5
[MgFe] [MgFe]
Figure 3.Index planes that are notsensitive to theIMF. They show expected ageand abundance trends forthe mostpart. Our SSP
modelsaregivenatage=2,5,and10Gyrwith[M/H]=−0.33,0,and0.37(dashedtracks).TheobservablesaretheGravesnon-LINER
averagesbinnedbyvelocitydispersion(opendiamonds),DobosPassivesplitintothreecolorbins(openmagentatriangles),DobosRED
splitbynuclearactivity(blackfive-pointedstars;thesmallestsymbolhastheleastnuclear activity,thelargestthemost),KinneyLTG
(cyan filledtriangle),[Oiii]selected SampleI(redfilledcircles),andHαselected SampleII(greyfilledcircles).EachSampleI/II grand
averageisindicatedbyalargersymbol,andthesmallersymbolsindicateasplitinthesampleatmagnitudemr =20.5.
MovingtotheotherthreepanelsofFig.3,thesearedi- empirically, except for the least-diagnostic Na D panel, the
agrams typically usedtotrackabundanceratio changesbe- Gravessequenceisstronger-linedinallIMFdiagnostics.The
cause they are age-metallicity degenerate (Worthey 1998). empirical points seem to split into two families, with the
The usual conclusion is that smaller ETGs resemble LTGs Gravessampledefiningone,andalltheothers,includingthe
in their abundance ratios, and that these ratios are scaled- LTG template, Dobos samples and DEEP2 ETGs, defining
solar.Fromthere,themassiveETGsshowenhancedMgand theother.ThattheDobossamplesandDEEP2ETGsshare
other light elements, while Fe-peak elements are depressed the same shallow IMF indicates little cosmic evolution in
relativetotheaverage.Thesetrendsareconfirmedhere.Itis IMF overthelast 4 Gyr.
statistically significant that,except for thereddest few, the For disentangling possible age effects, perhapsthebest
DobossamplesareabitmoreFe-rich,Mg-poor, alsolielow diagramsaretheTiO2andTiO2-SDSSindices,becausethe
inFe4668,whichismostlyduetocarbon.Thismustbesam- models begin to develop strong AGBs at younger ages so
pleselection.Mentallyallowingfortheagedifference,which that the models never dip weak enough to reach the ob-
weakensall themetallic featureindices, theDEEP2 sample served DEEP2 or Dobos indices unless the IMF is shallow.
shares the typicalabundancepattern of massive ETGs. For disentangling abundance ratio effects, we look for the
Fig.4showsIMF-sensitiveindices.Inbothsetsofmod- variousTiOdiagrams tobeechoedintheMg4780 diagram.
els,ayoungpopulationagedecreasessensitivytoIMF.The Themorphologyisthesame,andweconcludethatIMFmay
two sets of models agree qualitatively, though the present play a role in the indexstrengths.
model set is always shifted a bit more strong-lined than TheNaDindexhasasmallamountofIMFsensitivity,
CvD12. In all the various indices pictured, shallow IMFs
imply weaker feature strengths12. Looking at the diagrams
SDSS indices. This is because the IMF effects in young, metal-
poor populations andold, metal-richpopulations aredominated
12 Exceptionstothisgeneralbehaviourisfoundintheyoungest by stars from different stellar phases. Readers are referred to
and most metal-poor populations for TiO1, TiO2, and TiO2- Tang&Worthey(2015)formoredetails.
(cid:13)c RAS,MNRAS000,1–12
8 Baitian Tang and Guy Worthey
0.045
5.0 ThisWorkSal3.0 0.10
ThisWorkKroupa 0.040
4.5 0.09
CvD12Sal3.0
0.035
4.0 CvD12Chabrier 0.08
0.030
D 1 2
Na3.5 TiO0.025 TiO0.07
3.0 0.06
0.020
2.5 0.015 0.05
2.0 0.010 0.04
0.025 10Gyr slope3 0.09 1.0
0.08
0.020 3Gyr S 0.8
O Kroupa DS0.07 80
bTi0.015 O2S0.06 Mg47 0.6
Ti
0.05
0.010 0.4
0.04
0.005 0.03
2.0 2.5 3.0 3.5 2.0 2.5 3.0 3.5 2.0 2.5 3.0 3.5
[MgFe] [MgFe] [MgFe]
Figure 4.Index planes that are sensitive to the IMF. Mostsymbols areas in Fig.3. Theselection of models is altered, however. Our
SSP models are shown for ages 3 and 10 Gyr, and for a bottom-heavy α = 3.0 power-law IMF (solid blue lines) and a Kroupa IMF
(dashed blue lines) that is similar to the Chabrier IMF. In addition, CvD12 models are given with lines connecting ages 3 and 9 Gyr
at solar metallicity. An α = 3.0 power-law IMF (solid green line) and a Chabrier IMF (dotted green line) give the IMF sensitivity. If
both sets of models agreed perfectly, the CvD12 models would nearly connect the middle two dots on the metallicity-sensitive model
isochrones.
butithasalargeamountofsensitivitytoabundanceratios. most (Blades & Morton 1983; Bica & Alloin 1986). Milky
It may also, especially in the cases of the LTG template Way Na absorption is not an issue due to the redshifts of
and the DEEP2 sample, suffer from galaxy self-absorption the galaxies. The most serious uncontrolled systematic ef-
ifneutralNaispresentintheinterstellarmedium,sinceboth fect is likely to be spectrophotometric integrity. If spectral
thecomponentofNaDareresonancelines.TheGravesand response curvaturesthe same order as theindex widths oc-
Dobos sequences tilt such that more massive ETGs have a cur, and are not averaged away, they will produce spurious
steeper IMF than thelightweight ones. unastrophysicaldriftsintheindices.Thefluxingprocedures
in DEEP2 and the averaging in SDSS samples will mini-
mizethis.TheindicesmostpronetofluxingerrorsareTiO1,
TiO2, and TiO2-SDSS due to their long wavelength spans.
5 DISCUSSION, SUMMARY, AND Indeed, it is in these indices that Sample I and Sample II
CONCLUSION divergestrongly, and that is probably nocoincidence.
RandomuncertaintiesareshowninFigs.3and4exceptthat Systematic effects in the models are likely to be more
therandomuncertaintiesaresmallerthanthepointsymbols severe. Fluxing should not be a problem, but velocity dis-
for all the SDSS averages and could not be estimated with persioncorrectionsmustbemade.Furthermore,besidessim-
certainty for the LTG template, but are plausibly of order plified,parameterizedIMFs,themodelsarebasedonstellar
0.1 ˚A or 0.005 mag. evolutionary models which, while admirable in many ways,
Systematic uncertainties are a bigger worry. Velocity arealsopronetouncontrolleddriftsinstellartemperatures,
dispersion corrections are important for Hβ, Mg b, <Fe>, luminosities, andlifetimes (Charlot et al. 1996).Werecom-
Na D, and Mg4780, but a much lesser concern for Fe4668 mendeyeingthemodels onlyin adifferentialsense,looking
and the four TiO indices. Corrections were applied to all forthevectorsofage,[M/H],andIMFinthevariousindex-
indices,however,sotheprimaryuncertaintyisthedifficulty index diagrams.
of knowing the appropriate velocity dispersion to assign to A few addtional considerations deserve a few words.
each averaged bin, weighted by how close the galaxy bin is Tang & Worthey (2015) studied two effects that might en-
to the target 300 km s−1, because the closer the galaxy is tangle with the IMF slope determination, namely the IMF
tothattarget,thesmallerthecorrection.Wejudgethatthe Low Mass Cut-Off(LMCO), and AGBcontribution effects.
onlydatapointsthatmightsufferasignificantly skewedre- Thedegeneraciesbetweenslope,LMCO,andAGBstarcon-
sultarethelow-luminosity subsamplesfromDEEP2.There tributionsarereal, buttestingindicates theyaretoosubtle
are additional systematics for theHβ diagram dueto emis- to affect theappearance of Fig. 4.
sion corrections, as discussed above, and a small worry for In ETGs, [Ti/Fe] ≈ 0 (Johansson et al. 2012;
extraNaD self-absorption, afew tenthsof an Angstrom at Conroy et al.2014;Worthey et al.2014b).CouldTibeeven
(cid:13)c RAS,MNRAS000,1–12
Optical spectroscopy and initial mass function of z = 0.4 red galaxies 9
more underabundant in our galaxies? Additional indices • The SDSS sample that is culled to be near the fun-
Ti4553 and Ti5000 (not illustrated and not sensitive to the damental plane of ETGs and thus is probably the purest
IMF) lie slightly lower than solar for our sample galaxies. in terms of ETG fraction forms a separate sequence that
ThusthesmallvaluesofTi-related indicesmaybepartially isoffsettowardabottom-heavyIMF.TheotherSDSSsam-
dueto thelow Ti abundances.However,sub-solar Tiabun- plesthatlikelycontainmoreLTGsgroupwiththesimilarly-
dances can only explain indices containing Ti. The Mg4780 selectedDEEP2samplesandtheLTGtemplategalaxyitself
indexisnotTi-sensitive,yetindicatesashallowerIMFeven for a sequence at a more bottom-light location.
withoutaccountingforMg-enhancement.Mg4780isdefined • Split by luminosity, the low-luminosity half of the
in Servenet al. (2005), and shown to be IMF-sensitive in DEEP2redgalaxysampleisslightlymoremetal-poor,which
La Barbera et al. (2013). Our models accomodate varying isexpectedifmetallicitydrivesthecolor-magnituderelation,
Mg and Ti separately, but testing with abundance-altered butalsoolderonaverage,whichisaninterestingnewresult
models fails to indicate any easy way to relieve theappear- that should be confirmed when more and better spectra of
ance that shallow IMFs are required in Fig. 4. high redshift galaxies are available.
Since the SDSS spectra are limited to the galaxy cen- • The DEEP2 samples appear to be slightly enhanced
ter by the fixed fiber size (3 arcsec), the relatively nearby in carbon (Fe4668 index) and sodium (Na D index) and
(0.025<z <0.06) Graves stack may be biased towards the quitepossiblyslightlydeficientintitanium(thevariousTiO
stellarpopulationsinthegalaxycenter.TheDobossamples indices) compared to theirzero-redshift cohorts.
are less affected by the aperture effect, since the sampling • TheDEEP2samples areaboutafactoroftwoyounger
galaxies are on average further away (0.03 < z < 0.14 for than would be inferred if they were the passively evolving
thepassive samples). For the DEEP2 samples, thelong-slit precursors of the nearby strong-lined galaxies. That is, in
spectroscopyandtheoptimalextractionmethodensurethat Fig. 3,galaxies at theDEEP2 velocity dispersion haveages
theDEEP2spectraarefreefrom apertureeffect.According around 8 Gyr, the DEEP2 galaxies have ages less than 2
to the recent work of the MaNGA team on stellar popu- Gyr,and thelookback time is roughly 4 Gyr.
lation gradients (Zhenget al. 2016; Goddard et al. 2016), • SincetheDEEP2samples meshwith similarly-selected
no or slightly negative metallicity gradients are found in nearby galaxies (the Dobos samples), we do not find evi-
a large sample of nearby galaxies, though the metallicity denceofcosmic evolution in IMFoverthelast 4Gyr.How-
gradients may be stronger in more narrowly selected mas- ever,ourfidelityatdetectingIMFvariationsislow,andsuch
siveETGs(van Dokkumet al.2016).Therefore,theGraves evolution could exist at a modest level.
stack measurements with larger σ may be slightly affected
by the metallicity gradient. The correcting vectors should
point to the lower-left corner of the panels in Figure 4, in-
5.2 Future Improvements
dicated by our SSP models (blue lines). Furthermore, the
IMF gradients suggested by Mart´ın-Navarro et al. (2015a); The spectroscopic study of IMF cosmic evolution can be
van Dokkumet al. (2016) may also affect the the Graves improved in several ways:
stack measurements with larger σ, thus the aperture effect
may magnify theIMF-related indices. (i) Studyinggalaxiesathigherredshiftwouldpresumably
accentuatetrendsseenatlessextremelookbacktimes.Given
thatDEEP2GalaxyRedshiftSurveyisdesignedforgalaxies
atz∼1,thewavelengthrangeweseeisthenear-ultraviolet.
5.1 Summary FindingIMF-sensitiveindicesintheultraviolet, however,is
formidable,ifnotimpossible, sincethediagnosticcoolstars
With these caveats in mind, the firm conclusions of exam-
emit only a small percentage of theirlight there.
ining Figs. 3 and 4 are
(ii) Alternatively, keeping to modest redshifts of 0.3 <
• Sample selection strongly drives every diagnostic dia- z < 0.5, a larger sample would increase the S/N ratio of
gram. the composite spectra and allow subdivision of samples to
• The brightest, reddest galaxies are the oldest on aver- better characterize any IMF evolution. Morphological clas-
age. From there, however, sample selection drives a newly sification,fundamentalplanesubselections,andAGNinfor-
seen trend for age. In the sample selected by ETG funda- mation would improvethe certainty of what sort of objects
mentalplane,bluergalaxiesaremoremetal-poor.Inasam- are under study. SDSS-BOSS (Ross et al. 2016) could be
ple composed purely of non-AGN, non-star-forming galax- mined,for example.
ies, bluer galaxies are the same metallicity as red ones but (iii) Spectroscopic study of the IMF in the J band
much younger. In a sample with detectable AGN activity, has recently become possible, and has been successful
the bluer galaxies are more metal rich, and that partially (Smith et al. 2015). At z=0.4, the Nai and Caii features
anticorrelates with AGN activity; the least-active galaxies would fall into the J-band, therefore these classical feature
are both younger and more metal rich. lines can beused for future cosmic evolution study.
• To explain all the index drifts in comparison to the (iv) Targeting individual distant galaxies at high S/N
models, IMF variations seems to be required, along with would yield information on galaxy to galaxy cosmic scatter
age and abundance variations. In all SDSS averages there inIMFparameters,aswellasageandabundances,perhaps
is trend that the strongest lined galaxies appear to have a revealingawholenewlevelofdetailaboutgalaxyevolution.
more bottom heavy IMF, while the weaker lined galaxies It may be feasible to observe a sample of bright galaxies in
have a spiral-like bottom light IMF, in accord with many a cluster with a multiplexed instrument (e.g., VLT/KMOS
recent studies. or Keck/MOSFIRE)in a reasonable time.
(cid:13)c RAS,MNRAS000,1–12
10 Baitian Tang and Guy Worthey
5.3 Conclusion Bastian N., Covey K. R., Meyer M. R., 2010, Annual Re-
view of Astronomy and Astrophysics,48, 339
Recent research on nearby galaxies suggests variable IMFs
Bertelli G., Girardi L., Marigo P., Nasi E., 2008, Astron.
and questions the universality of the IMF seen around the
Astrophys.,484, 815
solar neighbourhood.Inparticular, massiveelliptical galax-
Bertelli G., Nasi E., Girardi L., Marigo P., 2009, Astron.
ies appear to have a bottom-heavy IMF in comparison to
Astrophys.,508, 355
low-massellipticalorspiralgalaxies.Forhigh-redshiftgalax-
Bica E., Alloin D., 1986, Astron. Astrophys.,166, 83
ies, the red IMF-sensitive indices shift out of the CCD
Blades J. C., Morton D. C., 1983, Monthly Notices Royal
wavelength range. In this paper, we apply a set of bluer
Astron. Soc., 204, 317
IMF-sensitiveindicestothestudiesofintermediate-redshift
Blanton M. R.,Roweis S., 2007, Astron. J., 133, 734
(0.3 < z < 0.5) galaxies. Red galaxies are selected from
CappellariM.,McDermidR.M.,AlataloK.,BlitzL.,Bois
the DEEP2 Galaxy Redshift Survey and stacked. Spectral
M.,BournaudF.,BureauM.,CrockerA.F.,DaviesR.L.,
indices measured from the composite spectra are compared
Davis T. A., de Zeeuw P. T., Duc P.-A., Emsellem E.,
with two sets of models and also local galaxies.
KhochfarS.,Krajnovi´cD.,KuntschnerH.,LablancheP.-
Weconfirmrecentworkthatstrong-lined(red,massive)
Y., Morganti R., Naab T., Oosterloo T., Sarzi M., Scott
galaxies appear to have a bottom-heavy IMF compared to
N.,SerraP.,WeijmansA.-M.,YoungL.M.,2012,Nature,
weak-linedETGsandLTGs.Thereisanaffirmingadditional
484, 485
trendthatthelocalsamplethatcullsoutfundamentalplane
CenarroA.J.,GorgasJ.,VazdekisA.,CardielN.,Peletier
outliersandthusselectswhatwethinkofasellipticalgalax-
R.F.,2003,MonthlyNoticesRoyalAstron.Soc.,339,L12
ies thebest is offset from samples that are selected in more
Chabrier G., 2003, Pub.Astron. Soc. Pacific, 115, 763
inclusive ways, in the sense that the pure-E sample is off-
ChabrierG.,HennebelleP.,CharlotS.,2014,Astrophys J.,
settowardamorebottom-heavyIMF.TheDEEP2galaxies
796, 75
do not appear that bottom-heavy, joining local LTGs and
Charlot S., Worthey G., Bressan A., 1996, Astrophys J.,
similarly-selected SDSS averages. There is no evidence for
457, 625
evolution in theIMF over thelast 4 Gyr, at least with cur-
Coil A. L., Newman J. A., Kaiser N., DavisM., Ma C.-P.,
rent data and tools.
KocevskiD. D., Koo D. C., 2004, Astrophys J., 617, 765
Intermsofagesandabundancesforlocalgalaxies,sam-
Conroy C., Graves G. J., van Dokkum P. G., 2014, Astro-
ple selection drives a fascinating trend in which all of the
phys J., 780, 33
reddest galaxies converge at metal-rich and old, but bluer
ConroyC.,GunnJ.E.,WhiteM.,2009,Astrophys J.,699,
galaxiesdonotagree:fundamentalplaneculledbluegalaxies
486
liemoremetal poorandabityounger,zero-nebulaselected
Conroy C., van DokkumP., 2012a, Astrophys J., 747, 69
bluegalaxies lie at thesame abundance but muchyounger,
Conroy C., van Dokkum P. G., 2012b, Astrophys J., 760,
and thebluest binsof red galaxies with detectableAGN lie
71
much youngerand more metal rich.
CooperM.C.,NewmanJ.A.,DavisM.,FinkbeinerD.P.,
The DEEP2 red galaxies, if split in half by luminos-
Gerke B. F., 2012, spec2d: DEEP2 DEIMOS Spectral
ity, show that the faint half is more metal poor, and, sur-
Pipeline. Astrophysics Source Code Library
prisingly, older. This trend may be driven by small number
CushingM.C.,RaynerJ.T.,VaccaW.D.,2005,Astrophys
statistics.TheDEEP2redgalaxiesarealsoquiteyoung,less
J.,623, 1115
than 2 Gyr, as compared to 8 Gyr for local galaxies at the
Dav´e R., 2008, Monthly Notices Royal Astron. Soc., 385,
samevelocitydispersion,effectivelyrulingoutapassiveevo-
147
lutionhypothesis.TheDEEP2redgalaxiesarelikelyslightly
Dobos L., Csabai I., Yip C.-W., Budava´ri T., Wild V.,
enhanced in C and Na and slightly deficient in Ti.
Szalay A. S., 2012, Monthly Notices Royal Astron. Soc.,
420, 1217
DuttonA.A.,MendelJ.T.,SimardL.,2012,Monthly No-
6 ACKNOWLEDGEMENT
tices Royal Astron. Soc., 422, L33
B.T.wouldliketothankY.ChenforherhelpontheDEEP2 Faber S.M., Jackson R. E., 1976, Astrophys J., 204, 668
spectral reduction and J. Newman and R. Yan for useful FaberS.M., Phillips A.C., Kibrick R.I.,Alcott B., Allen
technical advice. We thank the referee, Russell Smith, for S.L.,BurrousJ.,CantrallT.,ClarkeD.,CoilA.L.,Cow-
insightful comments. Support to G. W. for program AR- ley D. J., Davis M., Deich W. T. S., Dietsch K., Gilmore
13900 was provided by NASA through a grant from the D.K.,HarperC.A.,HilyardD.F.,Lewis J.P.,McVeigh
SpaceTelescope ScienceInstitute,which is operated bythe M., Newman J., Osborne J., Schiavon R., Stover R. J.,
Association ofUniversitiesforResearch inAstronomy,Inc., Tucker D., Wallace V., Wei M., Wirth G., Wright C. A.,
underNASAcontract NAS5-26555. 2003, in Society of Photo-Optical Instrumentation Engi-
neers(SPIE)ConferenceSeries,Vol.4841,InstrumentDe-
sign andPerformance forOptical/Infrared Ground-based
Telescopes, Iye M., Moorwood A. F. M., eds., pp. 1657–
REFERENCES
1669
AugerM.W.,TreuT.,GavazziR.,BoltonA.S.,Koopmans Faber S. M., Worthey G., Gonzales J. J., 1992, in IAU
L. V. E., Marshall P. J., 2010, Astrophys J. Letters, 721, Symposium,Vol.149,TheStellarPopulationsofGalaxies,
L163 Barbuy B., Renzini A., eds., p. 255
Baldwin J. A., Phillips M. M., Terlevich R., 1981, Pub. FerrerasI.,LaBarberaF.,delaRosaI.G.,VazdekisA.,de
Astron.Soc. Pacific, 93, 5 Carvalho R. R., Falc´on-Barroso J., Ricciardelli E., 2013,
(cid:13)c RAS,MNRAS000,1–12