Mon. Not. R. Astron. Soc. 425, 273–286 (2012) doi:10.1111/j.1365-2966.2012.21454.

The gas-phase metallicity of central and satellite galaxies in the Sloan

Digital Sky Survey

Anna Pasquali,1 Anna Gallazzi2 and Frank C. van den Bosch3

1 Astronomisches Rechen-Institut, Zentrum für Astronomie der Universität Heidelberg, Mönchhofstrasse 12–14, 69120 Heidelberg, Germany
2 DarkCosmology Centre, Niels Bohr Institute, University of Copenhagen, Juliane Maries Vej 30, 2100 Copenhagen, Denmark
3 Department of Astronomy, Yale University, PO Box 208101, New Haven, CT 06520-8101, USA

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

Accepted 2012 June 7. Received 2012 May 22; in original form 2012 March 26

ABSTRACT
We exploit the galaxy groups catalogue of Yang et al. and the galaxy properties measured in
the Sloan Digital Sky Survey Data Releases 4 and 7 to study how the gas-phase metallicities
of star-forming galaxies depend on environment. We find that satellite and central galaxies
follow a qualitatively similar stellar mass (M  )–gas-phase metallicity relation, whereby their
gas-phase metallicity increases with M  . Satellites, though, have higher gas-phase metallicities
than equally massive centrals, and this difference increases with decreasing stellar mass. We
find a maximum offset of 0.06 dex at log(M  /h−2 M )  8.25. At fixed halo mass, centrals are
more metal rich than satellites by ∼0.5 dex on average. This is simply due to the fact that, by
definition, centrals are the most massive galaxies in their groups, and the fact that gas-phase
metallicity increases with stellar mass. More interestingly, we also find that the gas-phase
metallicity of satellites increases with halo mass (M h ) at fixed stellar mass. This increment
is more pronounced for less massive galaxies, and, at M   109 h−2 M , corresponds to
∼0.15 dex across the range 11 < log (M h /h−1 M ) < 14. We also show that low-mass satellite
galaxies have higher gas-phase metallicities than central galaxies of the same stellar metallicity.
This difference becomes negligible for more massive galaxies of roughly solar metallicity.
We demonstrate that the observed differences in gas-phase metallicity between centrals and
satellites at fixed M  are not a consequence of stellar mass stripping (advocated by Pasquali
et al. in order to explain similar differences but in stellar metallicity), nor to the past star
formation history of these galaxies as quantified by their surface mass density or gas mass
fraction. Rather, we argue that these trends probably originate from a combination of three
environmental effects: (i) strangulation, which prevents satellite galaxies from accreting new,
low-metallicity gas which would otherwise dilute their interstellar medium; (ii) ram pressure
stripping of the outer gas disc, thereby inhibiting radial inflows of low-metallicity gas and
(iii) external pressure provided by the hot gas of the host halo which prevents metal-enriched
outflows from escaping the galaxies. Each of these three mechanisms naturally explains why
the difference in gas-phase metallicity between centrals and satellites increases with decreasing
stellar mass and with increasing host halo mass, at least qualitatively. However, more detailed
simulations and observations are required in order to discriminate between these mechanisms,
and to test, in detail, whether they are consistent with the data.
Key words: galaxies: abundances – galaxies: evolution – galaxies: fundamental parameters
– galaxies: groups: general – galaxies: star formation.

 E-mail: [email protected]


C 2012 The Authors

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
274 A. Pasquali, A. Gallazzi and F. C. van den Bosch
in higher density environments (Pustilnik, Tepliakova & Kniazev
1 I N T RO D U C T I O N
2011).
Star-forming galaxies in the local Universe define a Fundamental Nearby clusters (such as Coma, Hercules and Virgo) have been
Plane, where their star formation rate (SFR) and gas-phase metallic- extensively observed in search of environmental effects on the prop-
ity [typically 12 + log (O/H)] correlate with their stellar mass (M  ; erties of star-forming galaxies. For example, Shields, Skillman &
cf. Lara-López et al. 2010; Mannucci et al. 2010). The individual Kennicutt (1991) and Skillman et al. (1996) found that the ISM
relations in this plane have been long known; for example, Lequeux of spiral galaxies in Virgo is on average metal richer than that of
et al. (1979) were the first to recognize the existence of a correlation comparable field galaxies. When a distinction is made on the ba-
between galaxy magnitude and gas-phase metallicity, whereby more sis of their H I content, the Virgo spirals that are H I deficient and
luminous galaxies exhibit higher metallicities. More observational also located closer to the cluster core are seen to have a higher
evidence in support of such a dependence was later provided by gas-phase metallicity and a significantly lower birth rate (i.e. the
Garnett & Shields (1987), Skillman, Kennicutt & Hodge (1989) and ratio of newly born stars to the stars formed in the past; Gavazzi
Zaritsky, Kennicutt & Huchra (1994) among others. The advent of et al. 2002) than those with normal H I content, which are usually

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

the Sloan Digital Sky Survey (SDSS) made it possible for Tremonti located in the cluster outskirts and have similar metallicity as field
et al. (2004) to transform the galaxy luminosity–metallicity relation galaxies. More recently, Petropoulou et al. (2011) have shown that
into the more fundamental dependence of gas-phase metallicity on spiral galaxies in the Hercules cluster are chemically evolved with
galaxy stellar mass, where the interstellar medium (ISM) of more an oxygen abundance close to solar, and their gas-phase metallicity
massive galaxies is metal richer. Using the SDSS, Brinchmann et al. does not depend on the local density within the cluster. Interestingly,
(2004) and Salim et al. (2005) established a correlation between the their Hα emission is less spatially extended than their optical discs
SFR of galaxies and their stellar mass, whereby more massive, star- and their star formation activity takes place preferentially in their
forming galaxies sustain larger SFRs (see also Ellison et al. 2008). inner regions. This result supports the picture where ram pressure
The SFR has been seen to correlate also with the gas-phase metal- stripping due to the intergalactic medium (IGM) has truncated the
licity of galaxies, although in a way that differs according to the H I disc of these spirals, with the result of preventing inflows of
galaxy stellar mass. As pointed out by Mannucci et al. (2010), the metal-poor gas from the outer disc and halo which would otherwise
gas-phase metallicity of low-mass galaxies decreases as their SFR dilute the gas-phase metallicity in the inner regions of these galax-
increases, while high-mass galaxies do not show any significant ies. Suppression of infall of metal-poor gas due to environment
dependence of their gas-phase metallicity on SFR (see also Yates, would also be an explanation for the higher gas-phase metallicity
Kauffmann & Guo 2012). of cluster spirals.
Such an interplay among M  , SFR and gas-phase metallicity of As for dwarf galaxies, the studies of Vı́lchez (1995) and Lee,
galaxies is usually explained in terms of stellar feedback triggering McCall & Richer (2003) did not reveal a significant dependence of
gas and metals outflows (see Edmunds 1990; Lehnert & Heckman gas-phase metallicity on environment. On average, equally bright
1996; Frye, Broadhurst & Benı́tez 2002; Garnett 2002; Tremonti dwarfs residing in the field and in Virgo do not differ in their O/H
et al. 2004; Kobayashi, Springel & White 2007; Scannapieco et al. abundance, but some Virgo dwarfs can show a gas deficiency when
2008; Weiner et al. 2009; Spitoni et al. 2010; McCarthy et al. compared with equally metal-rich dwarfs in the field. Such a gas
2011; Peeples & Shankar 2011). Mannucci et al. (2010) envisaged a deficiency seems to correlate with the X-ray surface brightness of
condition of steady state in the local Universe, where infall of metal- the IGM, thus suggesting that ram pressure stripping due to the
poor gas (diluting the galaxy gas-phase metallicity and sustaining its IGM deprived these Virgo dwarfs of some of their gas. Ram pres-
star formation activity) occurs together with outflow of metal-rich sure stripping has been observed ‘in action’ for already a number
gas whose efficiency should depend on the galaxy mass and SFR. of galaxies in Virgo (e.g. Crowl et al. 2010), whose H I discs ap-
Alternative explanations of the Fundamental Plane of star-forming pear distorted and truncated with respect to their stellar discs. A
galaxies invoke downsizing (where the star formation efficiency somewhat stronger dependence on environment was established by
depends on galaxy mass; e.g. Brooks et al. 2007; Mouhcine et al. Petropoulou et al. (2011) for the dwarf galaxies in the Hercules
2008; Calura et al. 2009), a dependence of the initial mass function cluster, whose gas-phase metallicity is seen increasing with local
on galaxy mass (Köppen, Weidner & Kroupa 2007) or a model of density. Metal-poorer dwarfs are associated with lower densities
pure infall of metal-poor gas (Finlator & Davé 2008; Davé et al. (typically the infall regions of the cluster) and with a stronger Hα
2010). emission with respect to the global dwarf population in the cluster
One can easily expect that environment may intervene in shap- (Mahajan, Haines & Raychaudhury 2010), a sign that they undergo
ing the Fundamental Plane of star-forming galaxies. Environmental a starburst phase upon accretion on to the cluster. Such a star forma-
processes, such as strangulation (i.e. the removal of the gas reservoir tion activity has to get quenched at later times in order to reproduce
of satellites after they are accreted by their host halo), ram pressure the observed fractions of red-and-dead satellites (e.g. Kang & van
stripping (i.e. the removal of gas from satellites moving through a den Bosch 2008; van den Bosch et al. 2008a,b; Kimm et al. 2009;
dense intracluster medium), galaxy harassment (i.e. high-speed im- Pasquali et al. 2009; Weinmann et al. 2009).
pulsive galaxy encounters) and galactic wind confinement (i.e. due The large statistical power of the SDSS survey has not yet pro-
to the pressure exerted by the hot gaseous atmosphere of the host vided a definitive assessment of the environmental dependence of
halo, which can inhibit galactic winds), are believed to alter the galaxy gas-phase metallicity. In Mouhcine, Baldry & Bamford
inflows/outflows of gas experienced by galaxies, modifying their (2007) the gas-phase metallicity–mass relation turns out to be
star formation activity and their gas-phase metallicities with respect weakly dependent on environment, defined as the number den-
to that predicted by closed-box evolution. sity of galaxies within the projected distance to the closest 4th and
Observations of the ISM in galaxies residing in the Lynx-Cancer 5th neighbours. Star-forming galaxies show a marginal increase of
void show that the gas-phase metallicity of these galaxies is lower their O/H abundance at fixed stellar mass in denser environments,
by 30 per cent on average than that of equally massive galaxies equivalent to a few per cent for the more massive galaxies up to


C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
 Gas metallicity of centrals and satellites 275
∼20 per cent at log (M  /M ) < 9.5. The authors concluded that the most massive group members on the basis of their M  ) and ‘satel-
evolution of star-forming galaxies is mainly driven by their intrinsic lites’ (all group members that are not centrals). Galaxy magnitudes
properties, and does not depend on environment at large. Cooper and colours are based on the standard SDSS Petrosian technique
et al. (2008) found a stronger dependence of gas-phase metallicity (Petrosian 1976; Strauss et al. 2002), are corrected for Galactic
on galaxy environment (defined as the number density of galaxies foreground extinction (Schlegel, Finkbeiner & Davis 1998) and
within the projected distance to the closest 3rd neighbour), in that K-corrected and evolution corrected to z = 0.1 according to the
galaxies residing in higher density environments are metal richer procedure of Blanton et al. (2003). We use the notation 0.1 M X to in-
(see also Ellison et al. 2009). The question arises whether these dif- dicate the resulting absolute magnitude in the photometric X band.
ferences between Mouhcine et al. (2007) and Cooper et al. (2008) Galaxy stellar masses are computed using the relation between stel-
may be due to different definitions of environment. As already dis- lar mass-to-light ratio and colour of Bell et al. (2003), and galaxy
cussed by e.g. Weinmann et al. (2006) and recently Muldrew et al. morphology is defined in terms of the concentration parameter (C =
(2012), the projected number density of galaxies is of ‘tricky’ in- r90 /r50 , where r90 and r50 are the radii encompassing 90 and 50 per
terpretation as it depends on environment itself. In massive haloes cent of the Petrosian r-band flux, respectively). Values of C larger

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

such as clusters, it probes environment on a scale smaller than the than 2.6 typically refer to bulge-dominated galaxies (cf. Strateva
halo virial radius, while in low-mass haloes the environment probed et al. 2001).
typically encompasses multiple dark matter haloes. For each group in the catalogue Yang et al. (2007) provided two
In this paper, we revisit the issue of gas-phase metallicity ver- estimates of its dark matter halo mass, M h , one based on the ranking
sus galaxy environment by making use of the catalogue of galaxy of its total luminosity, the other on the ranking of its total stellar
groups extracted from the SDSS Data Release 4 (DR4) by Yang et al. mass. These halo masses deviate from each other by about 0.1 dex
(2007). In this catalogue, environment is defined as the total amount at low M h and 0.05 dex at the massive end. The method of Yang
of dark matter associated with each galaxy group, and also allows us et al. (2007) is able to assign M h only to groups more massive than
to distinguish (and treat separately) central from satellite galaxies, ∼1012 h−1 M and with one or more members brighter than 0.1 M r −
for which semi-analytic models of galaxy formation and evolution 5 log h = −19.5 mag. For all the other groups, Yang, Mo & van den
predict rather different evolutionary paths. We aim at achieving a de- Bosch (2008) used the relations between the luminosity (or M  ) of
tailed assessment of how gas-phase metallicities of galaxies depend central galaxies and the halo mass of their groups to estimate the halo
on galaxy stellar mass (over the interval 8 < log (M  /h−2 M ) < mass of single central galaxies down to M h ∼ 1011 h−1 M . In what
11), hierarchy (centrals versus satellites) and environment across a follows we will use the halo masses derived from the group’s total
large range of halo masses (11 < log (M h /h−1 M ) < 14), in order stellar mass, because, as shown by More et al. (2011), stellar mass
to gain important insights into how environment regulates gas in- is a better indicator of halo mass than luminosity. Since the groups
flows feeding the star formation activity of galaxies, and affects the catalogue of Yang et al. (2007) is not volume limited, it suffers from
efficiency of galactic outflows through the pressure applied by the the Malmquist bias, which causes an artificial increase of the average
intragroup hot gas. luminosity (and M  ) of galaxies with increasing redshift. This effect
This paper is organized as follows. In Section 2 we describe is especially important for satellites, since they span a large range
the catalogue of galaxy groups obtained by Yang et al. (2007) for of stellar masses within each halo. We take the Malmquist bias into
the SDSS DR4. The stellar mass–gas-phase metallicity relation ob- account by weighting each galaxy by 1/V max , where V max is the
tained for satellites and centrals separately is discussed in Section 3, comoving volume of the Universe out to a comoving distance at
while the halo mass–gas-phase metallicity relation is presented in which a galaxy would still have made the selection criteria of the
Section 4. The results are discussed in Section 5 and the conclu- groups sample. Therefore, all distributions in the following sections
sions follow in Section 6. Throughout this paper we adopt a flat  are weighted by 1/V max unless specified otherwise.
cold dark matter (CDM) cosmology with m = 0.238 and  =
0.762 (Spergel et al. 2007) and we express units that depend on the
2.1 Stellar population parameters: sample S
Hubble constant in terms of h = H 0 /(100 km s−1 Mpc−1 ).
We matched sample II with the catalogue of stellar ages and metal-
licities obtained by Gallazzi et al. (2005) for the galaxies in the
2 DATA
SDSS DR4. Briefly, Gallazzi et al. (2005) compared the strength
Our analysis is based on the SDSS DR4 catalogue of galaxy groups of stellar absorptions in the observed spectrum of each galaxy with
constructed by Yang et al. (2007) by applying the halo-based group the predictions of a Monte Carlo library of 150 000 star formation
finder algorithm of Yang et al. (2005) to the New York University histories (SFHs), based on the Bruzual & Charlot (2003) population
Value-Added Galaxy Catalogue (NYU-VAGC; Blanton et al. 2005) synthesis code and the Chabrier (2003) initial mass function. The
extracted from SDSS DR4 (Adelman-McCarthy et al. 2006). From SFHs in the library have a declining SFR over varying time-scales,
the NYU-VAGC Main Galaxy Sample Yang et al. (2007) selected and are superposed with random bursts of varying intensity and
all galaxies with an apparent magnitude (corrected for Galactic fore- duration. These bursts are added in such a way that only 10 per cent
ground extinction) brighter than r = 18 mag, in the redshift range of the models can experience a burst of star formation in the last
0.01 ≤ z ≤ 0.20 and with a redshift completeness Cz > 0.7. These 2 Gyr. For each galaxy a probability density distribution of stellar
galaxies were used to build three galaxy groups samples: sample I, r-band flux-weighted stellar ages and metallicities can be derived,
which only uses the 362 356 galaxies with measured redshifts from whose median values we will use for our analysis below. It has
the SDSS; sample II, which includes an additional 7091 galaxies to be noted that the stellar ages and metallicities derived as above
with SDSS photometry but redshifts taken from alternative surveys; refer to the redshift at which galaxies are observed. No correction
sample III which lists an additional 38 672 galaxies lacking a red- to z = 0 was attempted because this would require an accurate
shift due to fibre collisions, but being assigned the redshift of their knowledge of the SFH from the redshift of the observations to
nearest neighbour (cf. Zehavi et al. 2002). This paper focuses solely the present. The uncertainty on the stellar ages and metallicities de-
on sample II, where galaxies are split into ‘centrals’ (defined as the pends on the spectral signal-to-noise ratio (S/N); a S/N ≥ 20 ensures

C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
276 A. Pasquali, A. Gallazzi and F. C. van den Bosch
Table 1. Samples statistics.

Description Centrals Satellites Satellite fraction

(per cent)

Sample S Galaxies with spectral S/N ≥ 20 and measured age and Z 70 067 13 626 19.4
Sample G Galaxies with EW (Hβ) ≤ −3 Å and measured 12 + log (O/H) 69 688 14 182 20.4
Sample C Galaxies that are both in samples S and G 8134 1595 19.6

log (Z) < 0.3 dex, log (Age) < 0.2 dex and an average uncer- emission lines with the models of Charlot & Longhetti (2001). The
tainty on both parameters of ∼0.12 dex. global SFR of the galaxies in DR7 was derived by fitting the pho-
All galaxies in sample II which have a spectral S/N ≥ 20 and a tometry of the outer regions of galaxies with the synthetic stellar
stellar age and metallicity estimates computed as in Gallazzi et al. population models of Bruzual & Charlot (2003) convolved with the

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

(2005) hereafter constitute our sample S. This sample was exten- extinction law of Charlot & Fall (2000) (cf. Salim et al. 2007). Using
sively studied by Pasquali et al. (2010), who found that satellites this same procedure, Brinchmann et al. (2004) derived the global
are older and metal richer than centrals of the same M  , and this dif- and fibre stellar masses, M  and mfib , from the galaxy integrated
ference increases with decreasing M  . In addition, the average age magnitudes and the magnitudes measured within the fibre.
and stellar metallicity of low-mass satellites (M  ≤ 1010 h−2 M )
increase with the mass of the halo in which they reside. The authors
2.3 Combined sample
interpreted these trends as due to the quenching of star formation in
satellites, which leaves their stellar populations to evolve passively We extracted from sample G all galaxies for which the spectral
(thus explaining their older ages), and to tidal stripping, which re- S/N is larger than 20, and the stellar age and metallicity are avail-
moves a non-negligible fraction of their stellar mass. In this picture, able together with their gas-phase metallicity and SFR. We refer
the stellar metallicity of present-day satellites reflects the maximum to this subset of objects as the composite sample, i.e. sample C.
stellar mass that they reached during their lifetime and which can Table 1 summarizes the number of central and satellite galaxies in
be significantly higher than their present-day M  . the samples S, G and C defined above. In Fig. 1 we compare the
normalized distributions of the galaxies in sample II (black dashed
line), sample S (grey solid line) and G (black solid line) in colour,
2.2 Gas metallicities and star formation rates: sample G
concentration, stellar and halo masses. We note that the sample
Sample II was also cross-matched with the catalogue of gas-phase definition introduces a clear separation in colour and concentration
metallicities [i.e. 12 + log (O/H)] published by Tremonti et al. between sample G and sample S. Being star forming, the galaxies
(2004) for the galaxies in SDSS DR4. The analysis performed by in sample G turn out to be bluer and more disc like (with lower con-
Tremonti et al. (2004) can briefly be summarized in the following centrations) than those in sample S. As already shown by Pasquali
steps. (i) The continuum emission in each galaxy spectrum was et al. (2009), star-forming galaxies have on average lower M  and
modelled with a combination of single stellar populations from preferentially reside in low-mass haloes. This trend is mainly due to
Bruzual & Charlot (2003) of the same stellar metallicity, also con-
volved with the extinction law of Charlot & Fall (2000), and sub-
sequently subtracted from the observed spectrum. (ii) The lines in
the resulting pure emission-lines spectrum were simultaneously fit-
ted with Gaussians by imposing that all the Balmer lines have the
same width and velocity offset, and likewise for the forbidden lines.
(iii) The gas-phase metallicity was statistically estimated for all star-
forming galaxies by fitting the more prominent emission lines with
the models of Charlot & Longhetti (2001). These models combine
the synthetic stellar populations of Bruzual & Charlot (2003) with
the photoionization code CLOUDY (Ferland 1996) and the Charlot &
Fall (2000) extinction law, so that the galaxy ISM conditions (e.g.
gas ionization, metallicity and dust attenuation) are intertwined
with the radiation field and nuclear synthesis of the underlaying
stellar populations. For each galaxy Tremonti et al. (2004) derived
a likelihood distribution of the gas-phase metallicity from which
we extracted the median, the 16th and 84th percentile values. All
galaxies in sample II whose Hβ equivalent width is EW(Hβ ≤ −3 Å
(corresponding to a S/N ≥ 10) and whose gas-phase metallicity is
available from Tremonti et al. (2004) make up our sample G.
For the galaxies in sample G we retrieved the median values
of stellar mass, mfib , SFR, SFRfib , and specific SFRfib (sSFRfib =
SFRfib /mfib ) in the fibre, the median values of global SFR, SFRglo ,
and specific SFRglo (sSFRglo = SFRglo /M  ) from the SDSS Data Re- Figure 1. The normalized distributions of sample II (black dot–dashed
lease 7. For star-forming galaxies (like those in sample G), the SFRfib line) and samples S and G (solid grey and black, respectively) in colour,
was directly computed from their emission lines following the pro- concentration, stellar mass and halo mass. The distributions are not weighted
cedure of Brinchmann et al. (2004), who modelled the observed by 1/V max .


C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
 Gas metallicity of centrals and satellites 277

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

Figure 2. The normalized distributions of centrals (grey) and satellites (black) in sample G in stellar mass, gas-phase metallicity and halo mass. The distributions
are not weighted by 1/V max .

Figure 3. The normalized distributions of galaxies in sample S (grey) and

sample C (black) in stellar age and metallicity. The distributions are not
weighted by 1/V max .

the central galaxies of sample G as it can be inferred from the right-

hand panel of Fig. 2, while the satellites in the same sample span
a wide range of environments from M h  1011 to 1015 h−1 M . In
addition, centrals seem to be slightly more massive and metal poorer
than satellites in sample G (left-hand and middle panels of Fig. 2).
Figure 4. The M  –gas-phase metallicity relation for central (grey) and satel-
It is also interesting to compare sample S and sample C in stellar
lite (black) galaxies. The solid lines represent the median metallicity, while
age and metallicity; we find in Fig. 3 that the star-forming galax- the dashed lines the 16th and 84th percentiles of the metallicity distribution
ies in sample C have on average younger and metal-poorer stellar of centrals and satellites in each bin of stellar mass. Centrals and satellites
populations than those in sample S. We are thus probing different belong to sample G.
ranges of galaxy properties with respect to the analysis in Pasquali
et al. (2010). For each bin of stellar mass, we constructed the cumulative dis-
tributions in 12 + log (O/H) of centrals and satellites, and used the
Kolmogorov–Smirnov (KS), two sample test to assess the statistical
3 THE STELLAR MASS–GAS-PHASE
significance of the metallicity offset seen in Fig. 4. The probability
M E TA L L I C I T Y R E L AT I O N
at which the null hypothesis (i.e. centrals and satellites are drawn
The dependence of gas-phase metallicity on stellar mass is plotted from the same population) is rejected is larger than 95 per cent in
in Fig. 4 for centrals (grey) and satellites (black). We computed the the range 8 ≤ log (M  /h−2 M ) ≤ 10.5, and drops to ∼19 per cent
median (solid lines), the 16th and 84th percentile values (dashed for galaxies more massive than M  = 1010.5 h−2 M . Thus, over
lines) of the gas-phase metallicity distribution of centrals and satel- the stellar mass range spanned by our sample, the cumulative distri-
lites in sample G per bin of M  . It turns out that centrals and satellites butions indicate that satellites have a higher gas-phase metallicity
follow a qualitatively similar relation between gas-phase metallic- than centrals.
ity and M  , whereby the metallicity of their gas increases with their In order to check whether/how the M  –gas-phase metallicity rela-
stellar mass. The gas in satellites is, though, metal richer than in tion of satellites depends on environment, we derived it for satellites
centrals at nearly all stellar masses, and this is true for the median split among different bins of halo mass. The results are shown in
values as well as for the 16th and 84th percentiles. Fig. 5, where satellites residing in different groups are colour coded;
The difference in the median 12 + log (O/H) between satellites the black dashed line traces the median 12 + log (O/H) of centrals
and centrals decreases from 0.06 dex at log (M  /h−2 M )  8.25 from Fig. 4 and the grey band represents the 16th–84th percentile
to 0.004 dex at log (M  /h−2 M )  10.75. This trend holds also for range of the distribution of centrals also from Fig. 4. The 1σ error
volume-limited subsamples extracted from sample G. bars represent the error on the mean gas-phase metallicity weighted

C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
278 A. Pasquali, A. Gallazzi and F. C. van den Bosch

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

Figure 5. The M  –gas-phase metallicity relation for centrals (grey) and Figure 6. The M h –gas-phase metallicity relation for central (grey) and
satellites (colour) split among different bins of halo mass. The black dashed satellite (black) galaxies. The solid lines represent the median metallicity,
line represents the median 12 + log (O/H) of centrals, while the grey band while the dashed lines the 16th and 84th percentiles of the metallicity dis-
marks the 16th–84th percentile range of the distribution of centrals. tribution of centrals and satellites in each bin of stellar mass. Centrals and
satellites belong to sample G.
by 1/V max , and were computed by propagating the uncertainties
associated with the individual measurements of 12 + log (O/H). groups. Overall, it is seen increasing by ∼0.5 dex across the full
Satellites less massive than log (M ∗ /h−2 M )  9.3 and in groups range of environments probed by sample G, in agreement with the
more massive than log (M h /h−1 M )  12 appear to follow the findings of Cooper et al. (2008).
same M  –gas-phase metallicity relation. The overall shift to higher We split the satellites in sample G among different bins of stellar
gas-phase metallicities of satellites with respect to centrals is due mass and computed their M h –gas-phase metallicity relation as a
only to satellites in haloes more massive than 1012 h−1 M , with function of M  . The results are illustrated in the left-hand panel of
their stellar mass–gas-phase metallicity relation slightly steepening Fig. 7, where satellites are colour coded on the basis of their stellar
as M h increases. mass. The black dashed line traces the median 12 + log (O/H) of
Satellites in the least massive groups (11 < log (M h /h−1 M ) < centrals as a function of M h (from Fig. 6), while the grey band
12) and more massive than log (M  /h−2 M ) = 9.3 are very similar indicates their 16th–84th percentile distribution in gas-phase metal-
to centrals in terms of their gas-phase metallicity, while less massive licity as shown in Fig. 6. The 1σ error bars for satellites represent the
satellites exhibit higher 12 + log (O/H) values than centrals as error on the mean gas-phase metallicity weighted by 1/V max , and
their stellar mass decreases [the maximum difference, +0.11 dex, were computed by propagating the uncertainties associated with
is registered at log (M  /h−2 M )  8.2]. the individual measurements of 12 + log (O/H). The offset in 12 +
log (O/H) among the different M h –gas-phase metallicity relations is
simply due to stellar mass. We see that the gas-phase metallicity of
4 THE HALO MASS–GAS-PHASE satellites increases with their halo mass in all bins of M  . The am-
M E TA L L I C I T Y R E L AT I O N plitude of this trend across the full M h range amounts to ∼0.15 dex
We show in Fig. 6 the dependence of the gas-phase metallicity for satellites with log (M  /h−2 M ) < 9, ∼0.1 dex for those with
of central and satellite galaxies on the halo mass of the group in 9 < log (M  /h−2 M ) < 10.5 and only ∼0.02 dex for the most
which they reside. The grey and black lines refer to centrals and massive satellites. These findings are consistent with the results
satellites, respectively; the solid and dashed lines represent the 50th of Mouhcine et al. (2007). To better illustrate the dependence
and 16th/84th percentiles of the gas-phase metallicity distribution of gas-phase metallicity on M h , we computed the slope d[12 +
within each bin of M h . In haloes with M h  1012 h−1 M the gas- log (O/H)]/d[log (M h )] in each stellar mass bin, and plotted it as a
phase metallicity of central galaxies increases sharply with halo function of M  in the right-hand panel of Fig. 7. A weak trend can be
mass. Centrals in more massive haloes, though, have roughly con- recognized whereby the strength of the M h –gas-phase metallicity
stant gas-phase metallicities. Note that the small number statistics relation decreases with increasing stellar mass.
for massive groups in the Yang et al. (2007) catalogue prevent us
from probing this halo mass dependence beyond M h ∼ 1014 h−1 M .
5 DISCUSSION
Regardless of halo mass, central galaxies are metal richer than satel-
lites in haloes of the same mass by ∼0.5 dex on average. This is In the previous sections we have shown that the gas-phase metallic-
simply a reflection of the fact that, by definition, central galaxies are ity of central and satellite galaxies depends primarily on their stellar
the more massive galaxies in their groups, combined with the fact mass. In agreement with the results of Tremonti et al. (2004), their
that gas-phase metallicity is an increasing function of stellar mass. gas becomes metal richer as their stellar mass increases. However,
The oxygen abundance of satellites smoothly increases with halo at fixed M  the gas-phase metallicity is higher in satellites than in
mass up to M h  1013 h−1 M , and is constant in more massive centrals, and this observed offset increases with decreasing stellar

C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
 Gas metallicity of centrals and satellites 279

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

Figure 7. Left: the M h –gas-phase metallicity relation for satellite galaxies (colour) split among different bins of stellar mass. The black dashed line represents
the median metallicity of centrals, while the grey band shows the 16th–84th percentiles distribution of centrals as a function of halo mass. Centrals and satellites
belong to sample G. Right: the slope of the M h –gas-phase metallicity relations shown in the left-hand panel is plotted as a function of stellar mass. The error
bars indicate the 1σ uncertainty.

mass. In addition, at fixed M  the oxygen abundance of satellites

is found to be higher in more massive haloes. This trend is ob-
served for satellites less massive than M   1010.5 h−2 M with
an amplitude of ∼0.1 dex over the range 11 < log (M h /h−1 M ) <
14, and becomes considerably weaker for the most massive satel-
lites. Hence, there is a clear and ‘pure’ dependence of the gas-phase
metallicity of satellites on environment, but this dependence is much
weaker than the dependence on stellar mass. In what follows we in-
vestigate the physical mechanisms that may be responsible for the
(M  , M h )–gas-phase metallicity relations of satellite galaxies.

5.1 Gas-phase and stellar metallicities

We start by comparing the properties in common between the galax-
ies in sample C and those in sample S. We recall that sample C is built
from sample S and includes only galaxies with gas-phase metallic-
ity, stellar age and metallicity, while sample S contains galaxies with
measured stellar age and metallicity, regardless of whether their gas-
phase metallicity is available (see Section 2). Similarly to sample G
(cf. Fig. 4), also the satellites in sample C display a higher median Figure 8. The M  –gas-phase metallicity relation for central (grey) and
oxygen abundance when compared with equally massive centrals. satellite (black) galaxies in sample C, i.e. with measures of both gas and
Fig. 8 shows that their difference in median 12 + log (O/H) at fixed stellar metallicity. The solid lines represent the median metallicity, while
M  decreases from ∼0.14 dex at M  < 109 h−2 M to ∼0.02 dex the dashed lines the 16th and 84th percentiles of the metallicity distribution
at M  < 1010.3 h−2 M . Note that this difference is larger than that of centrals and satellites in each bin of stellar mass.
measured between centrals and satellites in sample G as shown in
Fig. 4. This is likely to be a consequence of the fact that sample C is Using sample S, Pasquali et al. (2010) found that satellites have
biased towards galaxies with weaker emission lines and lower star older, luminosity-weighted ages and higher stellar metallicities than
formation activity than those in sample G. In this respect, satellites equally massive centrals. Both these differences become smaller for
in sample C may have been more significantly quenched by their more massive galaxies. In addition, they showed that satellites with
environment than satellites in sample G. We caution, though, that M   1010 h−2 M have ages and metallicities that increase with
sample C suffers from small number statistics when compared to the mass of the host halo in which they reside. Pasquali et al. (2010)
sample G. The KS two sample test performed on the cumulative explained these trends by invoking (i) strangulation, i.e. the removal
distributions in 12 + log (O/H) of centrals and satellites in sample of the gas reservoir of satellites after they are accreted by their
C in different bins of stellar mass establishes that satellites are metal host halo. The consequent quenching of star formation leaves the
richer than centrals at a confidence level between 78 and 100 per cent stellar populations of satellites to evolve passively, while the more
in the range 8 < log (M  /h−2 M ) < 10.5, while the most massive prolonged star formation activity of centrals keeps their luminosity-
satellites cannot be distinguished from equally massive centrals. weighted age younger. (ii) Tidal stripping, which causes satellites to

C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
280 A. Pasquali, A. Gallazzi and F. C. van den Bosch

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

Figure 9. The distributions of stellar age and metallicity as a function of stellar mass for centrals (grey) and satellites (black) belonging to sample S (left-hand
panels) and to sample C (right-hand panels). The solid lines represent the median age/metallicity, while the dashed lines the 16th and 84th percentiles of the
age/metallicity distribution of centrals and satellites in each bin of stellar mass.

lose stellar mass once they have been accreted on to their host group. ple C indicates that centrals and satellites with log(M /h−2 M ) >
The fact that present-day satellites descend from more massive 9.5 are drawn from two different populations at a >92 per cent
progenitors is left imprinted in their stellar metallicity, which is not confidence level, while less massive centrals and satellites belong
expected to change significantly after a satellite is accreted on to a to the same parent population. In sample S the age difference be-
bigger halo and its star formation activity is quenched. The effects tween satellites and centrals is larger, and decreases from 0.19 dex at
of strangulation and tidal stripping likely depend on the time of M  < 109 h−2 M to 0.02 dex at M  < 1010.5 h−2 M . The KS two
infall, i.e. when a galaxy became a satellite for the first time. The sample test applied to the age cumulative distributions of centrals
earlier the infall time is, the older and less massive a satellite is, and satellites in different M  bins shows that the hypothesis of cen-
and the more massive the group in which the satellite resides (given trals and satellites belonging to the same population is rejected at a
that more massive haloes assemble earlier). This dependence on >99 per cent confidence level. We may thus conclude that the
infall time naturally explains why the ages and stellar metallicities satellites in sample C were accreted later and their star formation
of satellites increase with M h . activity has been quenched less than the satellites in sample S. Con-
Since both stellar and gas-phase metallicities correlate with M  , sequently, they should have also experienced less tidal stripping, so
one might also expect the higher gas-phase metallicities of satel- that their stellar metallicities should be closer to those of equally
lites (with respect to centrals of the same M  ) to be a manifestation massive centrals. This is indeed the case; the bottom panels of Fig. 9
of tidal stripping. In this scenario, satellites have higher gas-phase show that satellites in sample C have stellar metallicities that are
metallicities than centrals of the same present-day mass, simply basically indistinguishable from those of centrals of the same stellar
because they had a higher stellar mass at infall: the gas-phase mass. The differences are less than 0.02 dex, implying that satel-
metallicity serves as an indicator of the maximum stellar mass lites in sample C have lost <10 per cent of their pre-infall stellar
the galaxy reached during its lifetime. In order to test this hypoth- mass. For comparison, the difference in the median log(Z/Z ) be-
esis, we compare in Fig. 9 the M  –stellar age and the M  –stellar tween satellites and centrals in sample S decreases from ∼0.3 dex at
metallicity relations of centrals and satellites in sample S with those M  < 109 h−2 M to <0.02 dex at M  > 1010 h−2 M , indicating
of centrals and satellites in sample C. As already pointed out by that satellites in sample S have lost between ∼75 (for a present
Fig. 3, the galaxies in sample C are characterized by younger stel- day M  = 109 h−2 M ) and ∼20 per cent (for a present day M  =
lar ages and lower stellar metallicities than those in sample S. At 1010.5 h−2 M ) of their stellar mass at infall. To summarize, Fig. 9
fixed stellar mass satellites are systematically older than centrals indicates that satellites in sample C have not suffered significant
in both samples. This age difference is rather small for sample C tidal stripping (yet), but have undergone quenching of their star
(about 0.04 dex at M  < 1010 h−2 M ), and statistically significant formation activity at some, small degree. Therefore, we conclude
only for galaxies more massive than 109.5 h−2 M . In fact, the KS that stellar mass stripping cannot explain why satellites have higher
two sample test applied to the age cumulative distributions of sam- gas-phase metallicities than centrals of the same stellar mass.


C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
 Gas metallicity of centrals and satellites 281
lar surface mass density, SFR and gas mass fraction (Gallazzi et al.
2005). In the next sections we test whether any of these param-
eters may indeed be associated with the difference in gas-phase
metallicity between centrals and satellites.

5.2 Surface mass densities

Kauffmann et al. (2003) have shown that the star formation activity
of low-mass galaxies depends more on their surface mass density
than on their stellar mass. In a naive closed-box model for chem-
ical evolution, this would imply that, at fixed M  , galaxies with a
higher surface mass density have transformed a larger amount of
gas into stars and hence increased their gas-phase metallicity. For

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

the galaxies in sample G we define the surface mass density in the
fibre as fib = mfib /areafib , where areafib is the physical area covered
by the SDSS fibre (3 arcsec in diameter) at the galaxy distance. We
use fib as a proxy for the past SFH that took place over the same
(central) regions from which the gas-phase metallicities have been
measured. The solid lines in the left-hand panel of Fig. 11 plots the
median fib of centrals (grey) and satellites (black) as a function of
Figure 10. The dependence of the gas-phase metallicity of centrals (grey) their stellar mass; the dashed lines indicate the corresponding 16th
and satellites (black) in sample C on their stellar metallicity. The solid lines and 84th percentiles. The difference in fib between satellites and
trace the median 12 + log (O/H), while the dashed lines the 16th and 84th centrals is ∼0.02 dex at most, and changes from being negative at
percentiles of the gas-phase metallicity in each bin of stellar metallicity.
log (M  /h−2 M ) < 9.5 ( fib is lower in satellites) to positive at
higher M  . In the right-hand panel of Fig. 11 the median oxygen
In further support of this conclusion, we directly compare the gas- abundance (solid line) of centrals (grey) and satellites (black) is
phase metallicities of centrals and satellites at fixed stellar metallic- shown as a function of their surface mass density; as expected from
ity in Fig. 10. If the same stellar stripping responsible for creating the findings of Kauffmann et al. (2003), the gas-phase metallicity
the (small) offset in stellar metallicity between star-forming cen- increases with fib for both centrals and satellites. When matched
trals and satellites were the only responsible for the observed offset in fib , satellites with log ( fib ) < 8.5 display a higher gas-phase
in gas-phase metallicities, we would expect the relation between metallicity than centrals in the median and 16th and 84th percentiles
gas-phase and stellar metallicity to be the same for central and values. We therefore conclude that the offsets in 12 + log (O/H) are
satellite galaxies. Fig. 10 shows that clearly this is not the case: not associated with any offsets in surface mass density, as expected
at log (Z/Z ) < −0.4 (corresponding to M  < 109.5 h−2 M , see in a simple closed box model if the different gas-phase metallicities
Fig. 9) satellites have higher gas-phase metallicities than centrals of centrals and satellites owe to differences in their SFHs.
of the same stellar metallicity. Rather, Fig. 10 suggests that the
differences in gas-phase metallicity between centrals and satellites
5.3 Gas mass fractions
have their origin in something that is specific to the gas and that
has a larger impact on less chemically evolved (hence preferen- Similarly to fib , the gas mass fraction can be considered an indi-
tially low-mass) galaxies. The scatter between gas-phase and stellar rect measurement of the past star formation activity undergone by
metallicity has been shown to be associated with variations in stel- galaxies. It has been shown by Zhang et al. (2009) that the H I gas

Figure 11. Left: the surface mass density in the fibre as a function of stellar mass for centrals (grey) and satellites (black). The solid lines show the median
fib while the dashed lines the 16th and 84th percentiles in different bins of stellar mass. Right: the dependence of gas-phase metallicity on fib for central
(grey) and satellite (black) galaxies.


C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
282 A. Pasquali, A. Gallazzi and F. C. van den Bosch

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

Figure 12. Left: the gas mass fraction in the fibre as a function of stellar mass for centrals (grey) and satellites (black). The solid lines show the median μgas
while the dashed lines the 16th and 84th percentiles in different bins of stellar mass. Right: the dependence of gas-phase metallicity on μgas for central (grey)
and satellite (black) galaxies.

mass fraction of galaxies decreases along the gas-phase metallicity– of Fig. 12. For both centrals and satellites, the oxygen abundance
stellar mass relation as both M  and the oxygen abundance increase. decreases with increasing gas mass fraction, in agreement with the
Moreover, at fixed stellar mass, gas-poor galaxies exhibit higher gas- findings of Zhang et al. (2009), Erb et al. (2006) and Erb (2008 ,
phase metallicities. This poses the question whether the difference for galaxies at z ∼ 2). However, at fixed μgas satellites are typically
in gas-phase metallicity between central and satellite galaxies re- metal richer than centrals, by 0.06 dex at μgas  0.1 to 0.12 dex at
flects a difference in their gas mass fractions. In order to test this, we μgas  0.6. This trend likely arises from the anticorrelation between
computed the gas mass fraction in the fibre for central and satellite μgas and M  , combined with the fact that, at fixed M  , satellites are
galaxies using equation (5) of Tremonti et al. (2004), i.e. we derive metal richer than centrals. We conclude that the difference in gas-
the surface gas mass density, gas , from the star formation surface phase metallicity between centrals and satellites is not associated
density in the fibre by inverting the Schmidt law given by Kennicutt with a difference in (indirectly inferred) gas mass fraction, as would
(1998). The gas mass fraction in the fibre is then defined as μgas = have been expected in a simple closed box model.
gas /( gas + fib ), where fib is the surface mass density calcu-
lated in Section 5.2. The gas mass fraction is plotted as a function
of M  in the left-hand panel of Fig. 12, where centrals are in grey 5.4 Star formation rates
and satellites in black. Solid lines represent the median μgas in bins The results in the previous two subsections show that the enhanced
of stellar mass, while the dashed lines indicate the 16th and 84th gas-phase metallicities of satellite galaxies are not associated with
percentiles. The gas mass fraction decreases as galaxies become any enhancement or suppression of the stellar surface mass den-
progressively more massive, and no significant difference is seen sity or the gas mass fractions. We now test whether there is any
between centrals and satellites. We also determined the 16th, 50th correlation with the SFRs.
and 84th percentiles of the gas-phase metallicity of centrals and Fig. 13 shows the specific SFR in the fibre [sSFRfib = log
satellites in different bins of μgas , as shown in the right-hand panel (SFRfib /mfib )] and the specific global SFR [sSFRglo = log

Figure 13. The dependence of the specific SFR in the fibre (sSFRfib ) and the specific global SFR (sSFRglo ) as a function of stellar mass for centrals (grey) and
satellites (black) in sample G. The solid lines indicate the median specific SFR and the dashed lines the 16th and 84th percentiles of the distribution in specific
SFR in different bins of stellar mass.


C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
 Gas metallicity of centrals and satellites 283
(SFRglo /M  )] as a function of stellar mass for centrals (grey) and might explain why the gas-phase metallicity of satellites increases
satellites (black) in sample G. The sSFRfib of satellites is higher with host halo mass (see Fig. 7), at least qualitatively.
by only 0.02 dex than that of equally massive centrals in the range However, strangulation is unlikely to be the entire picture. First of
9 < log (M  /h−2 M ) < 10.5. The KS two sample test applied to all, if the differences in gas-phase metallicity (within the fibre) are
the cumulative distributions in sSFRfib of centrals and satellites in due to differences in the amount of dilution, this should probably
different bins of stellar mass indicates that such a difference is not show up as differences in the gas mass fraction (again within the fi-
statistically significant.1 We checked that the differences in gas- bre). As is evident from Fig. 12, such difference is absent, although
phase metallicity between satellites and centrals do not disappear we caution that these ‘gas mass fractions’ are indirectly inferred
when galaxies are matched in sSFRfib and stellar mass. from the SFRs (within the fibre). We conclude that strangulation
Interestingly, the median sSFRglo of satellites is systematically is likely to play a role in regulating the gas-phase metallicities of
lower than that of centrals by 0.03 dex on average at any stellar satellite galaxies, but that direct data on the gas mass fractions of
mass. The KS two sample test run on the cumulative distributions in satellite galaxies, combined with detailed hydrodynamical simula-
sSFRglo of centrals and satellites in different bins in M  indicates that tions of strangulation, are needed to test whether strangulation in
this difference is statistically robust at >95 per cent confidence level,

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

itself can explain the various trends presented in this paper.
at any stellar mass. The fact that satellites have lower global SFRs
than centrals of the same stellar mass is consistent with numerous
other studies (e.g. Weinmann et al. 2006, 2009; Cooper et al. 2007;
5.5.2 Ram pressure stripping and galaxy harassment
Kimm et al. 2009; van den Bosch et al. 2008a,b; Pasquali et al. 2010;
Peng et al. 2010; Wetzel, Tinker & Conroy 2011), and indicates that In addition to strangulation, there are two other mechanisms that
certain processes operating on satellite galaxies causes them to may impact gas-phase metallicities and which operate on satellites
quench their star formation. The fact that our results indicate that but not on centrals: ram pressure stripping (e.g. Gunn & Gott 1972)
satellites have lower sSFRglo than centrals, but are equally active in and galaxy harassment (e.g. Farouki & Shapiro 1981; Moore et al.
their central regions (within the fibre), suggests that this quenching 1996). We now discuss how these two processes may cause satellite
is an outside-in process. galaxies to have higher gas-phase metallicities than centrals of the
same stellar mass.
An important mechanism for regulating the gas-phase metal-
5.5 Beyond a closed box licities of (the inner regions of) disc galaxies is radial gas flows.
The results in the previous subsections have shown that the dif- Angular momentum redistribution within the disc, due to torques
ferences in gas-phase metallicities between centrals and satellites from bars, spirals and/or interacting galaxies can cause the cold gas
cannot be understood within a simple closed-box model for their at large galactocentric radii to move inwards (cf. Roškar et al. 2008;
chemical evolution. This should not come as a surprise, since galax- Minchev et al. 2012). Since galaxies typically have metallicity gra-
ies are not expected to be closed boxes. Rather they experience in- dients that decline outwards, such radial gas flows will dilute the gas
flows, outflows and even mass stripping. Interestingly, the efficiency in the inner regions. Since satellite galaxies are exposed to stronger
of these processes is expected to depend strongly on environment, tidal forces, and to high-speed impulsive encounters (i.e. galaxy
which, as we argue below, has the potential of explaining the trends harassment), they are more likely to experience such diluting radial
identified in this paper. flows than central galaxies. If this were an important mechanism,
satellite galaxies would have central gas-phase metallicities that are
lower than central galaxies, opposite to what is observed. This sug-
5.5.1 Strangulation gests that, contrary to what we postulated above, satellite galaxies
do not experience (an enhancement in) radial gas flows. This could
Central galaxies that reside in host haloes with M h  1012 h−1 M come about if ram pressure and/or tidal forces strip the outer, metal-
are expected to accrete cold gas from their surroundings (‘cold poor gas before it has the opportunity to migrate inwards. In that
flows’), whereas centrals in more massive haloes are expected to ac- case, satellite galaxies are less likely to experience diluting, radial
crete most of their gas through cooling flows from their hot gaseous gas flows than centrals, which again might explain why the latter
haloes (e.g. Birnboim & Dekel 2003; Kereš et al. 2005). Satellite have lower gas-phase metallicities than the former (see also Shields
galaxies, however, are likely to be ‘deprived’ of these inflows, sim- et al. 1991; Petropoulou et al. 2011).
ply because their gas reservoir, be it hot or cold, is likely to be There is ample observational evidence to support the notion
stripped by tidal forces and/or ram pressure from the hot gaseous that ram pressure stripping is an important mechanism, at least
atmosphere associated with the host halo. Since inflows are likely in cluster environments (e.g. Gavazzi et al. 2001; Solanes et al.
to be metal poor, this ‘strangulation’ (Larson, Tinsley & Caldwell 2001; Koopmann & Kenney 2004; Abramson et al. 2011). This is
1980) of satellite galaxies is expected to suppress the dilution of also supported by detailed hydrodynamical simulations (e.g. Abadi,
their ISM, resulting in gas-phase metallicities that are higher than Moore & Bower 1999; Roediger & Hensler 2005; Kronberger et al.
those of central galaxies, in qualitative agreement with the data. 2008; Bekki 2009). However, the overall efficiency of ram pres-
Furthermore, as already discussed in Pasquali et al. (2010), present- sure stripping (i.e. what fraction of the gas is stripped and on what
day satellites in more massive haloes were, on average, accreted time-scale) is still a matter of debate, mainly because the simu-
earlier (see also De Lucia et al. 2012). Hence, satellites in more lations have shown that it is a strong function of the porosity of
massive haloes have been ‘deprived of dilution’ for a longer time the galaxy’s ISM (e.g. Quilis, Moore & Bower 2000; Tonnesen &
than satellites of the same stellar mass in less massive haloes, which Bryan 2009). What is well established, though, is that ram pressure
stripping is more efficient for less massive galaxies in more massive
environments (e.g. Gunn & Gott 1972; Bekki 2009). This makes
1The same result is obtained when central and satellite galaxies are com- the simple prediction that the suppression of metallicity dilution in
pared in their sSFRfib defined as log(SFRfib /M  ). satellite galaxies increases with decreasing stellar mass and with

C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
284 A. Pasquali, A. Gallazzi and F. C. van den Bosch
increasing host halo mass; this is consistent with the observed than the confinement pressure independently of halo mass. Only a
trends, at least qualitatively. Another effect that may contribute here small fraction of the simulated galaxies (∼16 per cent) appears to
is again the distribution of infall times; since low-mass satellites are be confinement dominated, but this occurs when they have already
typically accreted earlier than their massive counterparts (e.g. De gone through the first pericentre passage, where ram pressure is
Lucia et al. 2012), they have been exposed to the ‘corrosion’ exerted highest, and have hence lost most of their hot gas.
by their environments for a longer period. If ram pressure stripping
indeed causes satellite galaxies to have gas-phase metallicities that
are higher (i.e. less diluted) than those of centrals of the same stellar 6 CONCLUSIONS
mass, then the statistics of infall times would contribute to explain- We used the groups catalogue of Yang et al. (2007) together with
ing the observed trends of gas-phase metallicities of satellites as a the DR4 catalogue of stellar ages and metallicities of Gallazzi et al.
function of M  and M h . (2005), the DR4 compilation of gas-phase metallicities of Tremonti
An interesting consequence of ram pressure stripping, which may et al. (2004) and the DR7 catalogue of SFRs of Brinchmann et al.
have important consequences for our discussion, is that the gas that (2004) to study the gas-phase metallicity of central and satellite

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

is not stripped away, which is mainly the gas at small galactocentric galaxies and its dependence on their stellar mass and the dark matter
radii, may actually be compressed by the ram pressure, giving rise mass of their host environments. We find the following.
to (significantly) enhanced star formation in the disc (e.g. Dressler
& Gunn 1983; Gavazzi et al. 1995; Kronberger et al. 2008; Kapferer (i) Satellites have on average higher gas-phase metallicities
et al. 2009). Our results that satellite galaxies have the same SFRs in than central galaxies of the same M  . The magnitude of this dif-
their central regions (within the fibre) as centrals of the same stellar ferences depends on stellar mass, increasing from 0.004 dex at
mass (see Fig. 13), therefore, seems to argue that ram pressure log (M  /h−2 M )  10.75 to 0.06 dex at log (M  /h−2 M )  8.25.
stripping is not very efficient. On the other hand, there are clear (ii) Within the same host halo, satellites have lower gas-phase
indications that the global SFRs of satellites are suppressed with metallicities than their central galaxies by ∼0.5 dex on average.
respect to those of centrals, which might be an indication that ram This simply reflects that, by definition, central galaxies are the most
pressure stripping has started to remove gas from the outer discs. A massive galaxies in their haloes, combined with the fact that gas-
ram pressure stripping induced boost in the SFR will also lead to phase metallicity increases with stellar mass.
an increase in the amount of metals returned to the ISM. However, (iii) At fixed M  , the gas-phase metallicity of satellites increases
whether these metals can be retained or whether they are ejected with halo mass; this trend is more pronounced for less massive
and/or stripped is likely to depend on many factors, to the extent galaxies. Low-mass satellites with 108.5 ≤ M  ≤ 109 h−2 M have
that it is unclear whether this boost in SFR is expected to increase average oxygen abundances that increase by ∼0.15 dex between
or decrease the gas-phase metallicities. More detailed studies, in M h = 1011 and 1014 h−1 M .
particular of the H I content of satellites versus centrals, will be (iv) Satellite galaxies have higher gas-phase metallicities than
needed to investigate the potential impact of ram pressure stripping central galaxies with the same stellar metallicities. The magnitude
for explaining the gas-phase metallicities of satellite galaxies. of this difference increases with decreasing metallicity.
We have contrasted the gas-phase metallicities of centrals and
satellites with other galaxy properties, such as stellar age, surface
5.5.3 Galactic wind confinement mass density, gas mass fraction and specific SFR, in order to iden-
tify the mechanism(s) responsible for the observational results listed
Another environment-dependent process that may potentially play
above. Based on these comparisons we draw the following conclu-
a role in ‘regulating’ the gas-phase metallicities of satellite galaxies
sions.
is galactic wind confinement. When a galaxy, central or satellite, is
located within a dark matter halo that has a hot gaseous atmosphere, (i) Star-forming centrals and satellites have very similar stellar
the associated pressure may be able to inhibit a potential supernova- metallicities at fixed M  . This basically rules out that the differ-
driven galactic wind from escaping the galaxy. Since galactic ence in gas-phase metallicities between centrals and satellites of
winds are typically metal enhanced (e.g. Mac-Low & Ferrara 1999; the same stellar mass is a consequence of stellar mass stripping.
Dalcanton 2007), systems in which the winds can escape are ex- Interestingly, this mechanism was used by Pasquali et al (2010) in
pected to have lower gas-phase metallicities than systems in which order to explain a similar difference between centrals and satellites
the wind material is confined to be recycled within the ISM of the but in stellar metallicity for mostly quiescent galaxies. The fact that
galaxy itself. Since satellite galaxies reside in more massive host this same mechanism cannot explain the gas-phase metallicities is
haloes than central galaxies of the same stellar mass, galactic wind also immediately evident from the fact that satellites have higher
confinement might explain why the former have higher oxygen gas-phase metallicities than centrals of the same stellar metallicity
abundances than the latter. (see Fig. 10). Rather, these results suggest that the differences in
This mechanism also nicely explains the observed trends with 12 + log (O/H) have their origin in something that is specific to the
both stellar mass and halo mass, at least qualitatively. After all, the gas.
hot gaseous atmosphere in more massive haloes has higher pressure, (ii) Star-forming satellites and centrals show very similar surface
and is therefore better able to confine winds. This might explain why mass densities and gas mass fractions (both measured within the
the difference in gas-phase metallicities between centrals and satel- fibre) at fixed M  . Consequently, their observed offset in gas-phase
lites is larger for satellites in more massive host haloes. In addition, metallicity remains the same when they are matched in surface
since winds are expected to be more efficient in less massive haloes, mass density or gas mass fraction, rather than stellar mass. Since
which host less massive centrals, the potential impact of winds is these quantities are an indicator of the galaxy’s past, integrated SFH
more important for less massive galaxies, providing a qualitative ex- within the fibre we conclude that the differences in 12 + log(O/H)
planation for the scalings with stellar mass. The recent simulations between centrals and satellites (which are also measured within
developed by Bahé et al. (2012) indicate that ram pressure is larger the fibre) are not a consequence of centrals and satellites having

C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
 Gas metallicity of centrals and satellites 285
experienced different closed-box SFHs. Rather, we argue that the University of Pittsburgh, University of Portsmouth, Princeton Uni-
mechanism(s) responsible for this offset must be associated with versity, the United States Naval University and the University of
differences in their inflow and/or outflow histories. Washington.
(iii) Despite their difference in gas-phase metallicities, centrals
and satellites of the same stellar mass have present-day SFRs, mea-
sured within the fibre, that are indistinguishable. However, when REFERENCES
corrections are made to include the star formation occurring on
Abadi M. G., Moore B., Bower R. G., 1999, MNRAS, 308, 947
larger scales, outside the fibre, we find that satellites have slightly Abramson A., Kenney J. D. P., Crowl H. H., Chung A., van Gorkom J. H.,
reduced global SFRs compared to centrals of the same M  . In ad- Vollmer B., Schiminovich D., 2011, AJ, 141, 164
dition, we find that star-forming satellite galaxies are characterized Adelman-McCarthy J. K. et al., 2006, ApJS, 162, 38
by somewhat older stellar ages than equally massive star-forming Bahé Y. M., McCarthy I. G., Crain R. A., Theuns T., 2012, MNRAS, in
centrals. These results suggest that the satellite galaxies have started press (arXiv:1205.2549)
an outside-in process of star formation quenching. We emphasize Bekki K., 2009, MNRAS, 399, 2221
Bell E. F., McIntosh D. H., Katz N., Weinberg M. D., 2003, ApJS, 149, 289

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

that the overall impact of quenching is small in our sample, which
is a consequence of the fact that our sample is biased towards star- Birnboim Y., Dekel A., 2003, MNRAS, 345, 349
forming galaxies, for which gas-phase metallicities can be measured Blanton M. R. et al., 2003, ApJ, 594, 186
Blanton M. R. et al., 2005, AJ, 129, 2562
from their emission lines.
Brinchmann J., Charlot S., White S. D. M., Tremonti C., Kauffmann G.,
We argue that there are three mechanisms that operate on satellite Heckman T., Brinkmann J., 2004, MNRAS, 351, 1151
galaxies and that can potentially explain why they have gas-phase Brooks A. M., Governato F., Booth C. M., Willman B., Gardner J., Wadsley
J., Stinson G., Quinn T., 2007, ApJ, 655, L17
metallicities that are higher than those of central galaxies of the
Bruzual G., Charlot S., 2003, MNRAS, 344, 1000
same stellar mass: (i) strangulation, which inhibits the inflow of Calura F., Pipino A., Chiappini C., Matteucci F., Maiolino R., 2009, A&A,
metal-poor gas; (ii) ram pressure stripping of the outer, metal-poor 504, 373
gas disc, thereby preventing dilution of the central gas due to radial Chabrier G., 2003, PASP, 115, 763
gas flows promoted by torques and (iii) pressure confinement of Charlot S., Fall S. M., 2000, ApJ, 539, 718
metal-enriched outflows. As we have discussed in the text, each of Charlot S., Longhetti M., 2001, MNRAS, 323, 887
these three mechanisms naturally explains why the difference in Cooper M. C. et al., 2007, MNRAS, 376, 1445
gas-phase metallicity between centrals and satellites increases with Cooper M. C., Tremonti C. A., Newman J. A., Zabludoff A. I., 2008,
decreasing stellar mass and with increasing host halo mass, at least MNRAS, 390, 245
qualitatively. More work needs to be done in order to quantitatively Crowl H. H., Kenney J. D. P., Chung A., Blanton M., van Gorkom J. H., 2010,
in Smith B., Bastian N., Higdon S. J. U., Higdon J. L., eds, ASP Conf.
discriminate between these environmental processes. For example,
Ser. Vol. 423, How Gas Stripping Transforms the Stellar Populations of
detailed simulations of ram pressure stripping and galactic wind Cluster Spiral Galaxies. Astron. Soc. Pac., San Francisco, p. 97
confinement, which include also star formation and stellar/active Dalcanton J. J., 2007, ApJ, 658, 941
galactic nucleus (AGN) feedback, are required in order to predict Davé R., Finlator K., Oppenheimer B. D., Fardal M., Katz N., Kereš D.,
the effect of these mechanisms on the observed stellar and gas- Weinberg D. H., 2010, MNRAS, 404, 1355
phase metallicities of satellites. From an observational point of De Lucia G., Weinmann S., Poggianti B. M., Aragón-Salamanca A., Zartsky
view, determining the SFH of satellites of different stellar mass and D., 2012, MNRAS, 423, 1227
in different environments will constrain the time-scale over which Dressler A., Gunn J. E., 1983, ApJ, 270, 7
strangulation/ram pressure stripping quench star formation, while a Edmunds M. G., 1990, MNRAS, 246, 678
direct measurement of the (central and total) gas mass of satellites Ellison S. L., Patton D. R., Simard L., McConnachie A. W., 2008, AJ, 135,
1877
will test the efficiency of strangulation and ram pressure stripping.
Ellison S. L., Simard L., Cowan N. B., Baldry I. K., Patton D. R.,
McConnachie A. W., 2009, MNRAS, 396, 1257
Erb D. K., 2008, ApJ, 674, 151
AC K N OW L E D G M E N T Erb D. K., Shapley A. E., Pettini M., Steidel C. C., Reddy N. A., Adelberger
We would like to thank Christy Tremonti for helpful discussions. K. L., 2006, ApJ, 644, 813
Farouki R. T., Shapiro S. L., 1981, ApJ, 243, 32
Funding for the SDSS and SDSS-II has been provided by the Alfred
Ferland G., 1996, Hazy: A Brief Introduction to CLOUDY. Internal Report,
P. Sloan Foundation, the Participating Institutions, the National Sci-
Univ. Kentucky
ence Foundation, the U.S Department of Energy, the National Aero- Finlator K., Davé R., 2008, MNRAS, 385, 2181
nautics and Space Administration, the Japanese Monbukagakusho, Frye B., Broadhurst T., Benı́tez N., 2002, ApJ, 568, 558
the Max Planck Society and the High Education Funding Council Gallazzi A., Charlot S., Brinchmann J., White S. D. M., Tremonti C. A.,
for England. The SDSS website is http://www.sdss.org/ he SDSS is 2005, MNRAS, 362, 41
managed by the Astrophysical Research Consortium for the Partic- Garnett D. R., 2002, ApJ, 581, 1019
ipating Institutions. The Participating Institutions are the American Garnett D. R., Shields G. A., 1987, ApJ, 317, 82
Museum of Natural History, Astrophysical lnstitute Postdam, Uni- Gavazzi G., Contursi A., Carrasco L., Boselli A., Kennicutt R., Scodeggio
versity of Basel, University of Cambridge, Case Western Reserve M., Jaffe W., 1995, A&A, 304, 325
Gavazzi G., Boselli A., Mayer L., Iglesias-Paramo J., Vı́lchez J. M., Carrasco
University, University of Chicago, Drexel University, Fermilab, the
L., 2001, ApJ, 563, L23
lnstitute for Advanced Study, the Japan Participation Group, John
Gavazzi G., Boselli A., Pedotti P., Gallazzi A., Carrasco L., 2002, A&A,
Hopkins University, the Joint lnstitute for Nuclear Astrophysics, the 396, 449
Kavli lnstitute for Particle Astrophysics and Cosmology, the Korean Gunn J. E., Gott J. R., III, 1972, ApJ, 176, 1
Scientist Group, the Chinese Academy of Sciences (LAMOST), Los Kang X., van den Bosch F. C., 2008, ApJ, 676, L101
Alamos National Laboratory, the Max-Planck-Institute for Astron- Kapferer W., Sluka C., Schindler S., Ferrari C., Ziegler B., 2009, A&A, 499,
omy (MIPA), New Mexico State University, Ohio State University, 87


C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS
286 A. Pasquali, A. Gallazzi and F. C. van den Bosch
Kauffmann G. et al., 2003, MNRAS, 341, 54 Roškar R., Debattista V. P., Stinson G. S., Quinn T. R., Kaufmann T., Wadsley
Kennicutt R. C., Jr, 1998, ApJ, 498, 541 J., 2008, ApJ, 675, 65
Kereš D., Katz N., Weinberg D. H., Davé R., 2005, MNRAS, 363, 2 Salim S. et al., 2005, ApJ, 619, L39
Kimm T. et al., 2009, MNRAS, 394, 1131 Salim S. et al., 2007, ApJS, 173, 267
Kobayashi C., Springel V., White S. D. M., 2007, MNRAS, 376, 1465 Scannapieco C., Tissera P. B., White S. D. M., Sringel V., 2008, MNRAS,
Koopmann R. A., Kenney J. D. P., 2004, ApJ, 613, 866 389, 1137
Köppen J., Weidner C., Kroupa P., 2007, MNRAS, 375, 673 Schlegel D. J., Finkbeiner D. P., Davis M., 1998, ApJ, 500, 525
Kronberger T., Kapferer W., Ferrari C., Unterguggenberger S., Schindler S., Shields G. A., Skillman E. D., Kennicutt R. C., 1991, ApJ, 371, 82
2008, A&A, 481, 337 Skillman E. D., Kennicutt R. C., Hodge P. W., 1989, ApJ, 347, 875
Lara-López M. A. et al., 2010, A&A, 521, L53 Skillman E. D., Kennicutt R. C., Shields G. A., Zaritsky D., 1996, ApJ, 462,
Larson R. B., Tinsley B. M., Caldwell C. N., 1980, ApJ, 237, 692 147
Lee H., McCall M. L., Richer M. G., 2003, AJ, 125, 2975 Solanes J. M., Manrique A., Garcı́a-Gómez C., González-Casado G.,
Lehnert M. D., Heckman T. M., 1996, ApJ, 462, 651 Giovanelli R., Haynes M. P., 2001, ApJ, 548, 97
Lequeux J., Peimbert M., Rayo J. F., Serrano A., Torres-Peimbert S., 1979, Spergel D. N. et al., 2007, ApJS, 170, 377
A&A, 80, 155 Spitoni E., Calura F., Matteucci F., Recchi S., 2010, A&A, 514, A73

Downloaded from https://academic.oup.com/mnras/article/425/1/273/998675 by guest on 27 December 2024

MacLow M.-M., Ferrara A., 1999, ApJ, 513, 142 Strateva I. et al., 2001, AJ, 122, 1861
McCarthy I. G., Schaye J., Bower R. G., Ponman T. J., Booth C. M., Dalla Strauss M. A. et al., 2002, AJ, 124, 1810
Vecchia C., Springel V., 2011, MNRAS, 412, 1965 Tonnesen S., Bryan G. L., 2009, ApJ, 694, 789
Mahajan S., Haines C. P., Raychaudhury S., 2010, MNRAS, 404, 1745 Tremonti C. et al., 2004, ApJ, 613, 898
Mannucci F., Cresci G., Maiolino R., Marconi A., Gnerucci A., 2010, van den Bosch F. C., Aquino D., Yang X., Mo H. J., Pasquali A., McIntosh
MNRAS, 408, 2115 D. H., Weinmann S. M., Kang X., 2008a, MNRAS, 387, 79
Minchev I., Famaey B., Quillen A. C., Di Matteo P., Combes F., Vlajic M., van den Bosch F. C., Pasquali A., Yang X., Mo H. J., Weinmann S., McIntosh
Erwin P., Bland-Harwthorn J., 2012, A&A, submitted (arXiv:1203.2621) D. H., Aquino D., 2008b, preprint (arXiv:0805.0002)
Moore B., Katz N., Lake G., Dressler A., Oemler A., 1996, Nat, 379, Vı́lchez J. M., 1995, AJ, 110, 1090
613 Weiner B. J. et al., 2009, ApJ, 692, 187
More S., van den Bosch F. C., Cacciato M., Skibba R., Mo H. J., Yang X., Weinmann S. M., van den Bosch F. C., Yang X., Mo H. J., 2006, MNRAS,
2011, MNRAS, 410, 210 366, 2
Mouhcine M., Baldry I. K., Bamford S. P., 2007, MNRAS, 382, 801 Weinmann S. M., Kauffmann G., van den Bosch F. C., Pasquali A., McIntosh
Mouhcine M., Gibson B. K., Renda A., Kawata D., 2008, preprint D. H., Mo H., Yang X., Guo Y., 2009, MNRAS, 394, 1213
(arXiv:0801.2476) Wetzel A. R., Tinker J. L., Conroy C., 2012, MNRAS, 424, 232
Muldrew S. I. et al., 2012, MNRAS, 419, 2670 Yang X., Mo H. J., van den Bosch F. C., Jing Y. P., 2005, MNRAS, 356,
Pasquali A., van den Bosch F. C., Mo H. J., Yang X., Somerville R., 2009, 1293
MNRAS, 394, 38 Yang X., Mo H. J., van den Bosch F. C., Pasquali A., Li C., Barden M.,
Pasquali A., Gallazzi A., Fontanot F., van den Bosch F. C., De Lucia G., Mo 2007, ApJ, 671, 153
H. J., Yang X., 2010, MNRAS, 407, 937 Yang X., Mo H. J., van den Bosch F. C., 2008, ApJ, 676, 248
Peeples M. S., Shankar F., 2011, MNRAS, 417, 2962 Yates R. M., Kauffmann G., Guo Q., 2012, MNRAS, 422, 215
Peng Y. et al., 2010, ApJ, 721, 193 Zaritsky D., Kennicutt R. C., Huchra J. P., 1994, ApJ, 420, 87
Petropoulou V., Vı́lchez J., Iglesias-Páramo J., Papaderos P., Magrini L., Zehavi I. et al., 2002, ApJ, 571, 172
Cedrés B., Reverte D., 2011, ApJ, 734, 32 Zhang W., Li C., Kauffmann G., Zou H., Catinella B., Shen S., Guo Q.,
Petrosian V., 1976, ApJ, 209, L1 Chang R., 2009, MNRAS, 397, 1243
Pustilnik S. A., Tepliakova A. L., Kniazev A. Y., 2011, MNRAS, 417, 1335
Quilis V., Moore B., Bower R. G., 2000, Sci, 288, 1617
Roediger E., Hensler G., 2005, A&A, 433, 875 This paper has been typeset from a TEX/LATEX file prepared by the author.


C 2012 The Authors, MNRAS 425, 273–286

Monthly Notices of the Royal Astronomical Society 

C 2012 RAS