The Outer Stellar Populations and Environments of Unusually HI-rich Galaxies

2.3 MB·English

by  Guinevere Kauffmann

Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed7January2015 (MNLATEXstylefilev2.2)
The Outer Stellar Populations and Environments of
Unusually HI-rich Galaxies
Guinevere Kauffmann
⋆ Max-Planck Institut fu¨r Astrophysik, 85741 Garching, Germany
5
1
0
7January2015
2

n
a ABSTRACT
J We investigate the nature of HI-rich galaxies from the ALFALFA and GASS surveys,

5 which are defined as galaxies in the top 10th percentile in atomic gas fraction at a
given stellar mass. We analyze outer (R > 1.5Re) stellar populations for a subset of
] face-onsystemsusingopticalg−rversusr−z colour/colourdiagrams.Theresultsare
A
compared with those from control samples that are defined without regardto atomic
G gas content, but are matched in redshift, stellar mass and structural parameters. HI-
. rich early-type (C > 2.6) and late-type (C < 2.6) galaxies are studied separately.
h
When compared to the control sample, the outer stellar populations of the majority
p
of HI-rich early-type galaxies are shifted in the colour/colour plane along a locus
-
o consistent with younger stellar ages, but similar metallicities. The outer colours of
r HI-rich late-type galaxies are much bluer in r−z than the HI-rich early types, and
t
s we infer that they have outer disks which are both younger and more metal-poor.
a We then proceed to analyze the galaxy environments of HI-rich galaxies on scales
[
comparabletothe expectedvirialradiioftheirdarkmatterhalos(R<500kpc).Low

1 mass(logM∗ <10.5)HI-richearly-typegalaxieshavegalaxyenvironmentsthatdiffer
v significantlyfromthe controlsample. HI-richearly types aremore likely to be central
7 rather than satellite systems. Their satellites are also less massive and have younger
1 stellarpopulations.Similar,butweakereffectsarefoundforHI-richlate-typegalaxies
0
of the same mass. In addition, we find that the satellites of HI-rich late-types exhibit
1
a greater tendency to align along the major axis of the primary. No environmental
0
. differences are found for massive (logM∗ >10.5) HI-rich galaxies, regardless of type.
1
0 Key words: galaxies: formation; galaxies: ISM; galaxies: stellar content
5
1
:
v
i 1 INTRODUCTION gain significant insight into the way in which galaxies grow
X
in mass at thepresent day.
r Recent targeted HI surveys of representative samples of
a galaxiesselectedbystellarmass(Catinellaetal2010;2013) BycomparingthecolourprofilesoflargesamplesofHI-
rich galaxies drawn from the wide-field ALFALFA survey
havedemonstratedthatL∗ galaxieslikeourownMilkyWay
(Giovanellietal2005)tothoseofnormalspirals,welearned
span a broad range in atomic gas mass fraction. The me-
thatgas-richgalaxieshaveunusuallyblueouterdisks(Wang
dianHImassfraction forgalaxies withstellar massesof4-5
×1010M⊙ is ∼ 0.1, but 10% of galaxies in this mass range et al 2011). Follow-up spectroscopy demonstrated that Hα
emission was strong in these regions and that the ionized
have gas fractions that are more than three times higher
gas metallicities frequently exhibited strong drops beyond
(Kauffmann et al 2012; Lemonias et al 2013).
theoptical radius(Moran et al2010; 2012). Analysis of the
Galaxies thataresignificantly moregas-rich thanaver-
ratio of star formation rate to stellar mass surface densi-
ageareextremelyinterestingforfurtherstudy.Iftheatomic
ties as a function of radius demonstrated that most of the
gas is able to reach high enough densities and metallicities
present-day growth in HI-rich galaxies occurs in their out-
to become shielded from the ambient UV radiation field, it
skirts. These results have recently been confirmed by inde-
will form H2 and from there, molecular clouds and stars.
pendentstudies (Huanget al 2014).
Atomicgasmaythusberegardedasa“reservoir”forfuture
starformation inagalaxy.BystudyingHI-richsystems,we Wethenundertookafollow-upprogram tomaptheHI
in a small sample of 25 unusually HI-rich galaxies and 25
“control” galaxies with normal HI content using the West-
⋆ E-mail:[email protected] erborkSynthesisRadioTelescope(WSRT;Wangetal2013,
(cid:13)c 0000RAS
2 G.Kauffmann
2014). The incidence of galaxies with disturbed HI mor- The most recent work on colour gradients in nearby
phologies was slightly lower in the HI-rich sample than in galaxies hasinvolvedstatistical studiesof muchlarger sam-
the control sample, strongly arguing against a scenario in ples. The Sloan Digital Sky Survey (York et al 2000) pro-
which the extra HI gas had been recently added by a sig- vided very uniform 5-band u,g,r,i,z photometry for mil-
nificant merging event. The main distinguishing character- lions of galaxies across a quarter of the sky and has served
istic of theHI-richgalaxies was thevery large radial extent as the underlying data set for most of these. Tortora et
of the HI in comparison to the optical light. In addition, al (2010) derived g−i colour gradients for 50,000 galaxies
therewastentativeevidencethattheouterHIgaswasmore with redshifts z < 0.05. They found stronger gradients for
clumpy/irregularinHI-richgalaxiesthaninthecontrolsam- lowmassgalaxies thanhigh massgalaxies. Suhetal(2010)
ple. studied g−r colour gradients for a large sample of early-
In this paper, we extend our work on HI-rich galaxies type galaxies. They found that the majority of early-types
by studying their outer stellar populations in more detail, have flat colour profiles. 11 percent of the sample exhib-
as well as their environments on scales of 20-500 kpc. We itednegativecolourgradients(i.e.bluercolours ontheout-
extract the largest available sample by combining the lat- side),while4percentwere“bluecore”systemswithpositive
est public release from the ALFALFA survey (the alpha.40 gradients. Gonzalez-Perez, Castander & Kauffmann (2011)
catalog; Hayneset al 2011) with thefinal data release from showed that if galaxies were split into early and late-type
the GASS survey (Catinella et al 2013). As in previous pa- classesatar-bandconcentrationindexon2.6,therewasno
pers, we focus our attention on galaxies with stellar masses longer a significant dependence of colour gradient on stel-
greaterthan1010M⊙.Wedividethissampleintotwoclasses: lar mass for each type (i.e. the trend with mass found by
a) Galaxies that are structurally “early-type” in that they Tortora et al was driven by an increasing early-type frac-
have have concentrated light profiles, indicative of galaxies tion among more massive galaxies). The SDSS survey has
with large bulge-to-disk ratios, b) Galaxies that are struc- also been used to study the colour gradients of rare galaxy
turally “late-type”in that theyhaveextendedlight profiles types.Roche,Bernardi&Hyde(2010) comparedthecolour
indicative of disk-dominated systems. The stellar popula- profiles of brightest cluster galaxies with those of ordinary
tionsandenvironmentsofthe10-25% mostHI-richsystems ellipticals and SOs and found that they were flatter on av-
arethencomparedwithgalaxiesofthesamemassandstruc- erage. We note that most studies of colour gradients using
turaltypedrawnfromthefullSloanDigitalSkySurveydata SDSShavebeenconfinedtoasinglecolour(mostofteng−r
release 7 (DR7;Abazajian et al 2009). or g−i).
The paper is divided into two parts. In the first part,
weaddressthequestionofwhatcanbelearnedaboutstellar
populations using available SDSS photometry. We discuss 2.2 Two-colour profiles of HI-rich galaxies
previous work on the colour profiles of disk galaxies and compared to the underlying population
ellipticals, andthen show howunusually HI-richgalaxies of
Infigures1-4, we show aseries of r−z versusg−r colour-
both typescompare with theparent samples. In thesecond
colour diagrams. In each figure, the the 6 panels show the
part of the paper, we examine whether HI-rich galaxies are
colours of galaxies at a fixed radius normalized to the ra-
more likely to be central or satellite galaxies in their halos.
dius R , defined as the radius containing half the total r-
Wealsoanalyzethestellarmasses,starformationratesand e
bandluminosityofthegalaxy.Theradiiplottedrangefrom
orientations of the galaxies in the vicinity of the HI-rich
one tenth R to 2.5 R . The SDSS photometry is not deep
systems and analyze whether any systematic difference is e e
enough to extract r−z colours at radii larger than this for
seen with respect to control samples.
individualgalaxies.
Figure 1 shows results for early-type galaxies in the
mass range 10.0 < logM∗ < 10.3, while Figure 2 shows
resultsforlate-typegalaxiesinthesamemassrange.Wedi-
2 STELLAR POPULATIONS
vide galaxies into early/late-type using a simple cut on the
2.1 Colour profiles of spiral and elliptical galaxies: r-band concentration index C of 2.6, where C is defined as
background the ratio of the radii enclosing 90% and 50% of the total
r−band light from thegalaxy. Shimasaku et al (2001) have
Theexistenceofcolourgradientsinbothspiralandelliptical
shown that this cut can separate galaxies that are visually
galaxies has been known for more than 50 years (e.g. Tifft
classified as E/S0/Sa from later-type spirals and irregulars
1963).Inthe1970’s,astronomersbegantoimagegalaxiesat
with reasonable accuracy. Figures 3 and 4 are the same as
near-IRwavelengthsforthefirsttime.Oneoftheimportant
Figures1and2,exceptthattheyshowgalaxiesinthemass
motivationsfornear-IRsurveyswastobreakthedegeneracy
range 10.3<logM∗ <10.6.
of the UBV colour-system in distinguishing whether age or
Ineachfigure,thegrey-scalecontoursshowthelocation
metallicity effects drive basic scaling relations, for example of the general population of galaxies in r−z versus g−r
the relation between galaxy colour and velocity dispersion.
colour/colour space. Thesecontourplotsareconstructed as
Strom et al (1976) used the V-K radial colour profiles of
follows:
5 elliptical and S0 galaxies to derive metallicity gradients
and compared them with the model predictions of Larson (i) From the data release 7 (DR7) of the Sloan Digital
& Tinsley (1974). Aaronson, Huchra & Mould( 1979) and Sky Survey, we select all spectroscopically-targeted galax-
Terndrup et al (1994) emphasized the importance of dust ies with redshifts in the range 0.015 < z < 0.06 and
extinctionwheninterpretingradialcolourtrendsingalactic withstellarmassesgreaterthan1010M⊙.Thestellarmasses
disks. are taken from the the MPA/JHU value-added catalogue
(cid:13)c 0000RAS,MNRAS000,000–000
Unusually HI-rich Galaxies 3
(http://www.mpa-garching.mpg.de/SDSS/DR7/). We dis- Early type (C>2.6) Late type (C<2.6)
card galaxies that are highly inclined with axial ratios
b/a < 0.6 in order to minimize the effects of dust extinc-
tion on our analysis.
(ii) TheSDSS photometric pipeline extracts azimuthally
averagedradialsurfacebrightnessprofilesforalltheobjects
in the survey. In the catalogs, this is given as the average
surface brightness in a series of annuli with fixed angular
dimension, which we then interpolate onto a fixed grid in
units of R/R . We discard the outer annuli in which the
e
errors on theg−r or r−z colours are greater than 0.2. 1
(iii) Thecontoursindicatethestellarmass-weightedfrac-
tionofthegalaxypopulationwithr−zandg−rcoloursina
Figure 5. The 50th (green symbols), 75th (blue symbols) and
givenrange.Webinther−zversusg−rcolour/colourplane
intocellsofsize0.05×0.05. EachgalaxyisweightedbyM∗, a9r0ethpl(octytaendsaysmabfoulsn)ctpieornceonfttilheesloofgtahreithdmistroifbuthtieonsteolflaMr(mHaIs)s/Mfo∗r
its mass, and 1/Vmax, the inverseof thevolume overwhich early-typegalaxies withC >2.6(left)andforlate-type galaxies
it is found in our sample. Our chosen contours are spaced withC<2.6(right).
a factor 2.5 apart in mass fraction, i.e. the lowest (black)
contour corresponds to a factor 244 lower mass-weighted
fraction than thehighest (white) contour. coloursgreaterthan0.65intheirveryouterregions,whereas
a fair numberexist in thelow mass sample.
The main two-colour trends can be summarized as fol-
Cyan symbols in Figures 1-4 show the locations of HI-
lows:
rich galaxies in r−z versus g−r colour-colour space. Our
On average, early-type galaxies exhibit weak gradients
inbothg−randr−z colours.Thepeakofthedistribution HI-richsampleisconstructedbycombiningthefinaldatare-
shifts from g−r = 0.87 at 0.1R to g−r = 0.7 at 2.5R , leaseoftheGALEXAreciboSDSSSurvey(GASS;Catinella
e e
while r−z changes from 0.7 to 0.6. As noted by Suh et al et al 2013) with the alpha.40 HI Source Catalog from the
Arecibo Legacy Fast ALFA Survey (Haynes et al 2011).
(2010), a minority of early-type galaxies have significantly
GASS is designed to reach a limiting HI mass fraction of
bluerouterregions.Inourplots,itcanbeseenthattheblue
tailismuchmoreprominentinr−z thanitising−r,but between 1.5 and 3% for galaxies with M∗ > 1010M⊙ and
thus can be used to define our HI-rich sample. In Figure 5,
as we will show later, at least part of this is due to larger
errors on ther−z colours at large radii. we plot plot 50th, 75th, and 90th percentiles of the distri-
Comparison of Figures 1 and 3 shows that more mas- bution of M(HI)/M∗ as a function of stellar mass for early-
typegalaxieswithconcentrationindexC >2.6andforlate-
sive early type galaxies exhibit less scatter in their central
coloursandhaveweakercolourgradients.Theshift ing−r type galaxies with C < 2.6. This plot is constructed using
colourfrom0.1Reto2.5Reis0.1for1010M⊙ellipticalscom- theGASSDR3“representativesample”, whichsamples the
pared to 0.17 for ellipticals that are 4 times more massive. M(HI)/M∗ distribution in each stellar mass in an unbiased
However,r−z colour gradients donot appearto varywith way (see Catinella 2010; 2013 for details). Note that more
than half of early-type galaxies with masses greater than
stellar mass.
As seen in Figure 2, the central colours of late-type 1011M⊙ are not detected in HI. For these galaxies, theme-
galaxies are spread much more widely than those of early- dianvalueofM(HI)/M∗ isplottedatthevalueoftheupper
limit, correpsonding to a HI mass fraction of 0.015. This
type galaxies. As radius increases, the mass weighted frac-
tionofgalaxiesintheverybluepartofther−zversusg−r is an over-estimate of the true median value. This will not
concernushere,becausewedefineHI-richgalaxiesasthose
colour plane increases. Interestingly, however, the location
with HI mass fractions that lie above the 90th percentile
of the peak of the colour-colour distribution remains very
pointsshown asfilledcyancirclesinthetwopanels.Ascan
nearly fixedwith radius.
beseen,the90th percentilepointsspangas fractions in the
Comparison of Figures 2 and 4 show that massive
range 0.1-0.3 for early-type galaxies and 0.3-1 for late-type
late-type galaxies exhibit significantly smaller dispersion in
galaxies.
colour in their central regions than their less massive coun-
terparts. However, g−r and r−z colour gradients appear It isimmediately clearthat HI-richgalaxies arealmost
always displaced towards the blue region of colour-space
to be stronger for more massive late type galaxies. At radii
R>1.5R ,theg−rcoloursoflowmasslate-typesexhibita when compared to the underlying population. A closer ex-
e
amination of Figures 1-4 reveals, however, that the nature
peak intherange0.7-0.8, i.e. at relatively redcolours. This
of the displacement depends on radius, on galaxy mass,
red peak is not found in the high mass late-type sample.
Likewise, thereare nomassive late-typegalaxies with r−z andonmorphological type.HI-richearly-typegalaxieswith
low masses have bluer g−r colours in their inner regions
(R<R50),butthisisnotseen forHI-richearly-typegalax-
ies with high masses. Both low and high mass HI-rich late-
1 Thiscorrespondstoa20%erroronthecolour,whichmayseem
typegalaxiesbecomesignificantlybluerthantheunderlying
ratherlarge.Wenotethatwillmostlybeconcernedwithrelative
comparisonsbetweendifferentsub-samplesatfixedradiusinthis population in g −r at R > 0.5R50. The r −z colours of
section.Insection2.4,whereweattempttointerpretthecolours bothearly-typeandlate-typeHI-richgalaxieswithlowstel-
in terms of physical parameters, we will create higher S/N esti- larmasses arenotnoticeably displacedfrom theunderlying
matesofoutercolours. population at R < R50. At larger radii, there is increasing
(cid:13)c 0000RAS,MNRAS000,000–000
4 G.Kauffmann
Figure 1. Thedistributionofthestellarpopulations ofgalaxiesinther−z versusg−r colour/colour planein6radialbins.Results
areshownearly-typeforgalaxiesinthestellarmassrange10<logM∗<10.3andwithconcentrationindexR90/R50>2.6Thecontour
plotsshowthestellarmassand1/Vmax weightedfractionaldistributionofallgalaxiesintheSDSSDR7.Thecontour levelsarespaced
a factor of 2.5apart and range from0.05 (white) to 0.0002 (black). The cyan diamonds show galaxies fromthe ALFALFA and COLD
GASSsurveyswhichareintheupper10thpercentileinHImassfraction.Thedashedandsolidredlinesareplacedinthesamelocation
ineachpaneltoguidetheeye.
Figure2. AsinFigure1,exceptforgalaxiesinthestellarmassrange10<logM∗<10.3andwithconcentrationindexR90/R50<2.6
(cid:13)c 0000RAS,MNRAS000,000–000
Unusually HI-rich Galaxies 5
Figure3. AsinFigure1,exceptforgalaxiesinthestellarmassrange10.6<logM∗<10.9andwithconcentrationindexR90/R50>2.6
Figure4. AsinFigure1,exceptforgalaxiesinthestellarmassrange10.6<logM∗<10.9andwithconcentrationindexR90/R50<2.6
scatter in r−z colour towards the blue. In section 2.4, we galaxies with high stellar masses are displaced to slightly
will come back to whether this effect is real or simply a re- redder values in their central regions. Once again, we find
flectionofincreasingphotometricerrorsinthelowersurface large scatter to bluerr−z colours at large radii.
brightnessouterregions ofthegalaxy.Finally,wenotethat
the r−z colours of both early-type and late-type HI-rich
(cid:13)c 0000RAS,MNRAS000,000–000
6 G.Kauffmann
2.3 Interpretation of r−z versus g−r
colour/colour diagrams in terms of stellar
population parameters
Thecombinationofopticalandoptical/infrared colourshas
been used a way of estimating stellar ages and metallicities
in both spiral and elliptical galaxies (Bell & De Jong 2000;
MacArthur et al 2004; Tortora et al 2010). These methods
areonlylikelytoyieldaccurateanswersifthestarformation
histories of galaxies have been smooth rather than bursty,
and if there is very little dust. These limitations are illus-
trated in detail in Figure 6, where we show the predicted
locations of galaxies with different star formation histories
in the r−z versusg−r colour-colour plane using thepop-
ulation synthesis models of Bruzual & Charlot (2003).
The coloured tracks indicate the location of galaxies
that have had smooth star formation histories of the form
SFR ∝ e−t/τ, where τ ranges from 0.1 to 100 Gyr. Three
differentlook-backtimesforstarformationtocommenceare
adopted:13,6and3Gyr.Differentcoloursindicatedifferent
metallicities: magentacorresponds to1.5 solar, green to0.5
solarandblueto0.1solar.Ascanbeseen,modelsofdiffering
age,butthesamemetallicity,occupyverynarrowlociinthe
Figure6. Thepredictedlocationsofgalaxieswithdifferentstar
r−z versus g−r plane. Under the assumption of smooth
formationhistoriesinther−z versusg−r colour-colourplane.
starformation histories andnodust,galaxies with different
The coloured tracks indicate the location of galaxies that have
stellarmetallicitiescanbeseparatedquiteeasilyiftheirr−z had smooth star formation histories. Different colours indicate
colours can bemeasured accurately enough. differentmetallicities:magentacorrespondsto1.5solar,greento
0.5solarandblueto0.1solar.Theblackpointsindicategalaxies
Ifgalaxies haveexperienced burstsof starformation in thathaveexperiencedpastbursts(seetext).Theeffectofdustis
their recent past, the situation becomes significantly more illustratedbytheredvector.
complicated.ThefourdifferentpanelsinFigure6showfour
different possibilities. In the top-left panel we show what
happens if a metal-rich burst is superposed on a smooth mag and R(V)=3.1. 2. Extinction will move galaxies away
metal-poor τ model, while the bottom left panel shows the from theloci definedbythedust-freeτ models in thesense
opposite case of ametal-poor burst superposed on a metal- ofcausingthemtohaveredder-than-predictedr−z colours
rich τ model. The top and bottom-right panels show the at a given metallicity.
casesofmetal-poorburstwithmetal-poorτ,andmetal-rich
burst with metal-rich τ, respectively. In all cases, “metal-
rich” corresponds to metallicities between solar and 1.5 so- 2.4 Application to outer stellar populations
lar,while“metal-poor”correspondstometallicitiesbetween
Given the complications discussed in the previous section,
0.1 and 0.25 solar. The burstsare allowed to begin at look-
we do not attempt a detailed quantitative analysis of age
back times ranging from 2 Gyr to 0.1 Gyr past and last
or metallicity trends as a function of radius in galaxies in
for 0.1 Gyr. Between 1% and 90% of the total mass of the
this paper. However, the strong trend towards bluer outer
galaxy is allowed to form in the burst.
g−r andr−z coloursinthegalaxy populationasawhole,
As can be seen, the power of the r−z colour to dif- andin in HI-richgalaxies in particular, isworthy offurther
ferentiatemetallicitycanbecompromisedbybursts.Super- analysis. The outer regions of galaxies viewed face-on are
posing a metal-rich starburst on a metal-rich τ model can sites where one might reasonably conjecture that neither
lead to g−r and r−z colours that overlap the locus of τ recent bursts of star formation nor dust play a significant
models with 0.1 solar metallicity. Ourexperiment of super- role.
posing different model star formation histories leads to the Onequestionthatarisesistheextenttowhichthe“blue
following general conclusion: if, in the absence of dust, the tails” seen in the outer radii in Figures 1-4 are simply due
r−z colourofagalaxyisred,thenthegalaxyismetal-rich. to increasing errors in the measured colours at large radii.
However,theoppositedoesnotnecessarilyhold.Ifther−z InparticulartheSDSSz-bandphotometryhaslowersignal-
colourisblue,thenthegalaxymaybemetalpoor,oritmay to-noise than the g− and r-band photometry, so the r−z
haveundergonea recent burst of star formation. colourswillexhibitgreaterspreadsimplyduetophotometric
errors. To answer this, we construct a higher S/N estimate
The effect of dust on the r −z versus g −r colours ofoutercolourforeachgalaxybybinningtogetherthelight
is illustrated by the red vector plotted in the bottom-right
corner of the top-right panel in Figure 6. We adopt a stan-
dardMilkyWayextinctionlaw(Cardelli,Clayton&Mathis 2 Thechoiceofextinctionlawmakesalmostnodifferencetothe
1989) andthelengthofthevectorcorrespondstoτ(V)=0.5 orientationofthevector inther−z versusg−r plane
(cid:13)c 0000RAS,MNRAS000,000–000
Unusually HI-rich Galaxies 7
fromradiilargerthan1.5R50.Weselectthesubsetofgalax-
ies with errors on both their g−r and r−z outer colours
of less than 0.04 and plot these in Figure 7 on top of our
gridofτ models.Theleft panelsshowresultsforearly-type
galaxies, while the right panels show results for late-type
galaxies. Wehaveaddedayellowtrack,indicatingtheloca-
tion of single stellar population (SSP) models of 0.01 solar
metallicity.Theyellowtrackmayberegardedasatheoreti-
callowerlimittotheallowedrangeofr−zcoloursatagiven
valueof g−r.3 Ascan beseen, eventhough the0.01 solar
SSPmodelsareveryblueing−r,thereisalmost nodiffer-
enceinr−z atfixedg−r colourcomparedtothe0.1solar
τ models. This means that stellar populations with metal-
licitiesbelow0.1solarcannolongerbeclearlydistinguished
in thisdiagram.
ThereddashedlinesinthetoppanelsofFigure7show
ourchosendemarcationlineforgalaxiesthatare“unusually
blue”inouterr−zcolour.Ourdemarcationlineischosento
passthroughthe0.1solarτ modeltrack.Galaxieslyingbe-
lowthislineareplottedascyancolouredpoints,whilethose
lyingabovethelineareplottedasblackpoints.Around10%
ofbothearly-typeandlate-typegalaxiesareincludedinthe
“unusuallyblue”sub-sample.Inthebottompanels,weplot
the same galaxies in the plane of 4000 ˚A break strength
(D (4000)) versus stellar surface mass density. The 4000 ˚A Figure 7. Top: The outer (R > 1.5Re) colours of galaxies are
n plotted in the r−z versus g−r colour-colour plane. We have
breakstrengthisarelativelycleanmeasureoftheageofthe selectedgalaxieswitherrorsinbothg−randr−zoutercolours
central stellar population of thegalaxy within the3 arcsec- ofless than 0.04inthe stellar massrange 10.3<logM∗ <10.8.
ond diameter fibre aperture. Because the index is defined Results are shown separately for early-type (C > 2.6) and late-
over a very narrow range in wavelength, it is insensitive to type(C<2.6)galaxies.Thecolouredtracksindicatethelocation
dust extinction. The stellar surface mass density is defined ofgalaxiesthathavehadsmoothstarformationhistories.Differ-
as 0.5M∗/πR520, where R50 is the half-light radius of the ent colours indicate different metallicities: magenta corresponds
galaxy. As can be seen, early-type galaxies with unusually to 1.5 solar, green to 0.5 solar and blue to 0.1 solar. The yellow
blueouterr−z colourshavecentralstellar populationages trackshowsthelocusof0.01solarmetallicitysimplestellarpop-
ulations (SSPs). Galaxies above the red-dashed line are plotted
and densities that are indistinguishable from the underly-
in black; galaxies below the red-dashed line are plotted in cyan.
ingpopulation.Late-typegalaxieswithunusuallyblueouter
Bottom: We show the same galaxies in the top panels in the
colours have similar central ages to the underlying popula- planeof4000˚Abreakstrength (Dn(4000)) versusthelogarithm
tion, but are shifted to lower mass densities by a factor of of the stellar surface mass density. Colour-coding of points is as
3 (implying that at fixed stellar mass, their half-light radii inthetoppanels.
are a factor of 1.7 larger). These differences are quantified
in more detail in Figure 8, in which we show histograms of
Dn(4000) and log µ∗ for the two samples. of HI gas mass fraction and with colour errors less than
Ifweinterpretgalaxieslyingbelowthereddashedlines 0.08.Thebluepointsindicatethelocusofcontinuousmodels
in Figure 7 as having metal-poor outer-stellar populations, with0.1solarmetallicityforreference.Resultsareshownfor
then the results in Figure 8 show that early-type galax- two stellar mass ranges: 10.0 < logM∗ < 10.5 and 10.5 <
ies with metal-poor outer stellar populations have central logM∗ <11.0.
stellar mass densities and ages that are indistinguishable Ascanbeseen,theouterstellar populations ofHI-rich
fromtheearly-typegalaxypopulationasawhole.Late-type early-typegalaxiesinthelowermassbinliealongatrackin
galaxies with metal-poor outer stellar population also have ther−z versusg−rcolour/colourdiagram consistentwith
central stellar population ages that are indistinguishable youngeragesbutmetallicitiesthatarethesameastheouter
from other late-type galaxies, but they have significantly stellar populations of the parent population. We infer that
lower stellar densities than theparent population. if the extra HI gas in these systems is of external origin, it
Catinellaetal(2010)showthattheHImassfractionin must have been accreted at high metallicity. We also note
galaxiesisstronglycorrelatedwithstellardensityinaddition that there are only a few HI-rich early-type systems that
to UV/optical colour. In Figure 9, we plot the outer (R > overlap theunusuallybluer−z population plotted in cyan
1.5R50) r −z versus g −r colours of HI-rich galaxies on intheleftpanelsofFigures7and8.Theearly-typegalaxies
topoftheparentpopulation.ToincreaseourHI-richsample with the bluest outer r−z colours in Figures 1 and 3 have
statistics,weincludeallgalaxiesintheupper25thpercentile have errors on their R > 1.5R50 colour measurements that
aretoolargetoallowthemtobeincludedinthissample.In
contrast, the outer stellar populations of low-mass HI-rich
3 We have compared the low metallicitySSPs fromthe Bruzual late-typegalaxiesareshiftedtobothyoungeragesandlower
& Charlot (2003), Maraston (2005) and Conroy, Gunn & White metallicities.TheshiftinmetallicityislargestfortheHI-rich
(2009)modelsandfindthattheyallyieldcomparableresults. galaxies with the youngest outer disks. The most extreme
(cid:13)c 0000RAS,MNRAS000,000–000
8 G.Kauffmann
& Haynes 1985; Kenney & Young 1989; Solanes et al 2001;
Chunget al 2009), showing that HIhasbeen removedfrom
thedisksof spirals in thecentral regions of clusters by pro-
cesses such as ram-pressure stripping or tidal interactions.
Inrecentwork,Catinellaetal(2013)cross-correlatedGASS
galaxies with the updated SDSS galaxy group catalogue of
Yangetal(2007)andshowedthatgalaxiesinhaloswithes-
timatedvirialmassesgreaterthan1013M⊙ aregas-deficient
compared to galaxies in lower mass halos.
Wenotethatthesepaststudieshavealmostexclusively
probed physical processes leading to the removal of atomic
gasfromgalaxiesindenseenvironmentssuchasclustersand
groups. This is true also for the correlation function stud-
ies,because theamplitudeon small scales scales (<1Mpc)
is simply proportional to the square of the galaxy density
within individual dark matter halos. In this paper, we take
adifferentapproachandstudyenvironmentinawaythatis
designedtoprobepossiblegasaccretion processesinHI-rich
galaxies. Our analysis focuses on whether HI-rich galaxies
ofdifferentstellarmassesandmorphologicaltypesaremore
likely to be central or satellite systems within their dark
matter halo. We also analyze not only the numberof satel-
lites around HI-rich galaxies, but also properties such as
theirstellarmassesandtheagesoftheirstellarpopulations.
Figure8.Histogramsof4000˚Abreakstrengthandstellarsurface
Finally,welook atwhetherthesatellites ofHI-richgalaxies
mass density for the same galaxies plotted in Figure 7. Black
display any preferred orientation with respect to the major
histogramsareforthegalaxieslyingabovethedashedredlinein
Figure 7, while cyan histograms are for the galaxies lying below axis of the primary (theso-called Holmberg effect).
this line. Errorbars are computed by bootstrap resampling the
samples.
3.1 Central and satellite properties
In a redshift survey,there is no completely accurate way of
amongst these systems have estimated stellar metallicities
atR>1.5R50 ofatenthsolar orless.4 Theshiftsincolour separating galaxies into those that reside at the centers of
theirdarkmatterhalos(so-calledcentralgalaxies)andthose
for the high mass HI-rich galaxies are very much weaker.
that are satellite systems gravitationally bound to a more
In particular, the high-mass HI-rich early-types display no
massive central object. We adopt a simple definition used
colour offsets at all.
in previous analyses (e.g Kauffmann et al 2012) whereby a
galaxy is classified as central if it has stellar mass larger
thanall itsneighbourswithin aprojected radiusof500 kpc
3 DARK MATTER HALO-SCALE withvelocitydifference∆V <500km/s.Itisalso classified
ENVIRONMENTS OF UNUSUALLY as a central if no neighbours are found within this volume.
HI-RICH GALAXIES A galaxy is classified as a satellite if it has a more massive
neighbour within ∆R < 500 kpc and ∆V < 500 km/s. In
In this section we compare the environments of HI-rich
this paper, we add an addition class of “binary” galaxies,
galaxies to the underlying population. We focus on scales
defined that those with a neighbour within a factor of two
less than 500 kpc, roughly corresponding to the predicted
of their own stellar mass.
virial radii of the dark matter halos that host L∗ galaxies.
Our HI-rich sample is constructed as described in the
The influence of environment on atomic gas in galax-
previous section, except that for the environmental anal-
ies has been studied in the past using two main meth-
ysis we include all galaxies regardless of their inclination.
ods: 1) In wide-field surveys such as HIPASS (Barnes et
ForeveryHI-richgalaxy,weselect asampleof10 “control”
al 2001) or ALFALFA (Giovanelli et al 2005), it is possi-
galaxies,randomlydrawnfromtheDR7spectroscopiccata-
ble to compute the auto-correlation function of HI-selected
logue. The control galaxies arechosen tomatch theHI-rich
galaxies and compare this to the auto-correlation function
of optically-selected galaxies (Passmoor et al 2011; Martin galaxytowithin0.2dexinlogM∗,0.2dexinlogµ∗,0.01in
redshift, and 0.2 in axial ratio. The environmental proper-
et al 2012; Papastergis et al 2013). These studies find that
ties of the HI-rich sample are then analyzed relative to the
HI-selected galaxies cluster very weakly, with a clustering
length r0= 3.3 h−1 Mpc, and are more strongly anti-biased control sample, which is selected without regard to atomic
gas fraction.
withrespecttothedarkmatterdistributiononsmallscales
InFigures10and11,weshowresultsforearlyandlate-
comparedtolargescales.2)Therehavealsobeenstudiesfo-
typeHI-rich and control galaxies with stellar masses in the
cusingontheHIpropertiesofgalaxiesinclusters(Giovanelli
range 10.0 < logM∗ < 10.3. In the top panels, we plot the
fractionofgalaxiesclassifiedascentral/satellite/binary and
4 Wenotethat thescatter belowthetenth solartrackisconsis- thedistributionofthenumberofneighbourswith∆R<500
tentwiththetypical erroronther−z outercolour kpc and ∆V < 500 km/s. In the bottom panels we plot
(cid:13)c 0000RAS,MNRAS000,000–000
Unusually HI-rich Galaxies 9
Figure9.Theouter(R>1.5Re)coloursofgalaxiesareplottedinther−z versusg−rcolour-colourplane.Wehaveselectedgalaxies
witherrorsinbothg−randr−zoutercoloursoflessthan0.04inthestellarmassranges10.0<logM∗<10.5and10.5<logM∗<11.0.
Resultsareshownseparatelyforearly-type(C>2.6)andlate-type(C<2.6)galaxies.Thebluetrackindicatesthelocationofgalaxies
of 0.1 solar metallicity that have had smooth star formation histories. Cyan points indicate galaxies with HI gas mass fractions in the
upper25thpercentilefortheirstellarmassdetectedintheCOLDGASSorALFALFAsurveys.Weonlyplotgalaxieswithcolourerrors
lessthan0.08.
the stellar mass and 4000 ˚A break strength distributions of fractions and the number and stellar mass distributions of
thesatellites. Black histograms show results for the HI-rich neighbours.
galaxies,whileredpointsshowresultsforthecontrolsample,
whichis10timeslargerinsize.Errorbarsarecomputedby
3.2 Holmberg effect
boot-strap resampling both samples.
Interestingly,environmentaldifferencesbetweenHI-rich In 1969, Holmberg wrote a paper claiming that satellite
galaxiesandthecontrolsamplearemuchmoreapparentfor galaxies were preferentially located along the minor axis of
early-typegalaxies than for late-typegalaxies. Thefraction diskgalaxies.Zaritskyetal(1997)foundasimilarresultfor
ofHI-richearly-typegalaxies thatarecentralsis0.75, com- asampleofisolatedspirals,butmorerecentstudieshaveled
pared to 0.55 for the control sample. Their satellite popu- todifferentconclusions.Themostrecentstudiesusinglarge
lations consists of lower mass galaxies with younger stellar samplesdrawnfromredshiftsurveyssuchas2dFandSDSS
populations.SimilardifferencesarefoundforHI-richspirals, find that satellite alignments occur along the major axis of
but theeffect is somewhat weaker. the primary (Brainerd 2005; Yang et al 2006). In addition,
Figure 12 and 13 are the same as figures 10 and 11, alignments are are strongest for massive red galaxies (Yang
except for HI-rich and control galaxies with stellar masses etal2006; Faltenbacheretal2007,2009). Thestandardex-
in the range 10.6 < logM∗ < 10.9. Here we see no differ- planationforsatellitealignmenteffectsisthattheyoccuras
ence between the environments of HI-rich early-type galax- aresultofthelarge-scaletidalfieldandthepreferredaccre-
iesandthecontrolsample.Thereisaweaktendencyforthe tionofmatteralongfilamentsinthelarge-scaledistribution
satellites of massive HI-rich spirals to have younger stellar ofgalaxies.Theoreticalsupportforthisscenariocomesfrom
populations compared to the control sample, but otherwise studying the origin of alignment effects between dark mat-
the two samples are identical in terms of central/satellite ter halos in N-body simulations (Li et al 2013). The most
(cid:13)c 0000RAS,MNRAS000,000–000
10 G.Kauffmann
Figure 10. Distribution of central/satellite/binary galaxy frac- Figure 12. As in Figure 10, except for early-type (C > 2.6)
tions(upperleft),satellitenumbers(upperright),satellitestellar galaxiesinthestellarmassrange10.6<logM∗<10.9.
masses (lower left)andsatellite4000 ˚A breakstrengths. Results
are shown for early-type (C > 2.6) galaxies in the stellar mass
range10<logM∗<10.3.BlackhistogramsshowresultsforHI-
richgalaxieswithHImassfractionsintheupper10thpercentile,
whileredsymbolsshowresultsforcontrolsamples(seetext).Er-
rorbarshavebeencomputedbybootstrapresampling.Notethat
in the top left panel, Type=1,2,3 refers to central, satellite, and
binary,respectively.
Figure13.AsinFigure10,exceptforlate-type(C<2.6)galax-
iesinthestellarmassrange10.6<logM∗<10.9.
massive red galaxies are predictedto belocated at thecen-
tersofthemostmassivedarkmatterhalos,whereaccretion
ofstellarmaterial frominfalling satellites occursfrequently,
even at the present day (e.g. Oser et al 2010). Alignment
Figure11.AsinFigure10,exceptforlate-type(C<2.6)galax-
effects are thusexpected to bestrong for these systems.
iesinthestellarmassrange10<logM∗<10.3.
It is thus interesting to consider whether alignments
canbeusedasaprobeofgasaccretionprocesses.Accretion
of new gas onto galaxies is thought to occur in two main
(cid:13)c 0000RAS,MNRAS000,000–000