Metal Content and Production in Clusters

of Galaxies
A. Renzini
1,2
1
European Southern Observatory, Garching bei Munchen, Germany
[email protected]
2
INAF - Osservatorio Astronomico di Padova, Padova, Italy
[email protected]
1 Introduction
X-ray observations allow the abundance of several elements in the intracluster
medium (ICM) to be measured and optical observations allow similar stud-
ies for galaxies. Taken together, we have at hand the possibility of studying
chemical evolution on the large scale of clusters of galaxies, and in doing so
we have the opportunity to learn several aspects of how both galaxies and
clusters work and evolve.
The topics to be covered by these lectures include:
Metals in the ICM and galaxies: how much of them and how they got there
Metal Production: Supernovae of Type Ia, their progenitors and rate
Metal Production: Supernovae of Type II and the IMF
The main epoch of metal production in clusters
Clusters as anempirical nucleosynthesis template
The chemical evolution of the universe
The global chemical evolution of the Milky Way galaxy.
2 Metals in the ICM
In one of the rare cases in which theory anticipates observations, the exis-
tence of large amounts of heavy elements in the intracluster medium (ICM)
was predicted shortly before it was actually observed [64]. This came from
(now old-fashioned) so-called monolithic models of elliptical galaxy forma-
tion, in which the observed color-magnitude relation is reproduced in terms
of a metallicity trend. In turn, this trend is established by supernova-driven
galactic winds being more eective in less massive galaxies with shallower po-
tential wells, compared to more massive galaxies, harbored in deep potential
A. Renzini: Metal Content and Production in Clusters of Galaxies, Lect. Notes Phys. 740,
177211 (2008)
DOI 10.1007/978-1-4020-6941-3 6 c Springer Science+Business Media B.V. 2008
178 A. Renzini
wells. While these models are now inadequate in many respects, their predic-
tion was conrmed the following year by the discovery of the strong Iron-K
line in the X-ray spectrum of galaxy clusters [73].
The iron and other heavy element production, and circulation on the
galaxy cluster scale, has been widely discussed since their early discovery
[1, 6, 21, 23, 57, 68, 71, 84, 91, 93, 99, 119].
Figure 1 shows the ASCA X-ray spectrum of the inner and outer regions
of the cluster A496 [34], which best-t temperatures are respectively 3.4
and 4.2 keV. Emission lines of several elements are clearly visible, with the
most prominent feature being the Iron-K line at 7 keV. Figure 2 shows
the XMM-Newton X-ray spectrum of the core of the Virgo cluster, with the
individual contribution of the various elements from a synthetic spectrum
[46]. As fully detailed in the lectures by Craig Sarazin, the continuum X-ray
emission is due to electron bremsstrahlung, while the lines come from decays
from higher energy levels which are populated by collisional excitations. Thus,
for example, the Fe K line comes from the transitions down to the K shell of
H-like and He-like iron ions, while the Fe L complex comes from transitions
down to the L shell of iron ions with 3 or more bound electrons. Obtaining
abundances from such spectra is universally achieved using packages [90],
which are based on theoretically calculated collisional excitation probabilities.
Therefore, the derived abundances of simple ions (e.g., H-like and He-like)
should be regarded as more reliable than those of ions still with many bound
electrons.
0
.
1
0
.
0
1
0
.
0
1
1
0

3
1
0

3
K
e
v
/
c
m
2

s

K
e
v
1
0

4
0.5 1
Outer 3-12 arcmin
Inner 0-2 arcmin
---FeL---
O
Ne
Mg Si
S
Ar
FeK
Ni
2
Channel Energy (keV)
5 10
Fig. 1. The X-ray spectrum of the cluster A496 with the main emission lines iden-
tied (from [34])
Metals in Clusters of Galaxies 179
5
2
1
c
o
u
n
t
s
/
s
e
c
/
k
e
V
c
o
u
n
t
s
/
s
e
c
/
k
e
V
0.5
0
.
5
0
.
2
s
Ar
Ca
Si
Fe
O
Me
Ne
Si
Ni
0
.
1
3
Energy (keV)
Energy (keV)
4
0.5
1 2
Fig. 2. The X-ray spectrum of the core of the Virgo cluster A496 and the superpo-
sition of the contribution of individual ions from a synthetic spectrum [46]
2.1 Iron
Iron is the best studied element in clusters of galaxies since ICM iron emis-
sion lines are present in all clusters and groups, either warm or hot. Figure 3
shows the iron abundance in the ICM of clusters and groups as a function of
ICM temperature from an earlier compilation [97]. For kT

>
3 keV the ICM
iron abundance is constant at Z
Fe
0.3Z

Fe
, independent of cluster tempera-
ture [29]. Abundances for clusters in this horizontal sequence come from the
Iron-K complex at 7 keV. At lower temperatures the situation is much less
straightforward. Figure 4 shows data from [17], with the iron abundance hav-
ing been derived with both one-temperature and two-temperature ts. The
one-temperature ts give iron abundances for these cool groups which are
more or less in line with those of the hotter clusters. The two-temperature-t
abundances, instead, form an almost vertical sequence, with a great deal of
dispersion around a mean 0.75 solar. Earlier estimates gave extremely low
values for cooler groups, kT

<
1 keV [75]. Compiling values from the litera-
ture a strong dependence of the abundance on ICM temperature is apparent,
being very low at low temperatures, steeply increasing to a maximum around
kT 1.52 keV, then decreasing to reach 0.3 solar by kT 3 keV [78, 93].
180 A. Renzini
Z
F
e
/
Z
F
e
1.2
1
0.8
0.6
0.4
0.2
0
0 2 4 6
kT (keV)
8 10
Fig. 3. A compilation of the iron abundances in the ICM as a function of the ICM
temperature for a sample of clusters and groups [97]. Clusters at moderately high
redshift (z 0.35) are represented by small lled circles. For kT

<
2 keV 11 groups
are shown [17], whose temperatures and abundances are determined from one- and
two-temperature ts (lled squares and open triangles, respectively)
Is this strong temperature dependence real? Perhaps some caution is in
order? Besides the ambiguity as to whether one- or two-temperature ts are
preferable, additional uncertainties on the iron abundances at kT

<
2 keV are
due to them being derived from the Iron-L complex at 1 keV, where the
emission lines are due to transitions to the L level of iron ions with three or
more electrons. In these cooler groups/clusters most of the iron is indeed in
such lower ionization stages, and the iron-K emission disappears. The atomic
congurations of these more complex ions are not as simple as those giving
rise to the iron-K emission, and their (calculated) collisional excitation prob-
abilities may be more uncertain. In summary, iron abundances derived from
the iron-L emission should be regarded with a little more caution, compared
to those from the iron-K emission.
0.6
0.4
0.2
M
e
t
a
l
l
i
c
i
t
y

(
s
o
l
a
r
)
M
e
t
a
l
l
i
c
i
t
y

(
s
o
l
a
r
)
0
0.6
0.4
0.2
0
0 0.25
r/r
180
0 0.25
r/r
180
Fig. 4. Projected metallicity distributions (mostly iron) for non cool-core clusters
(left panel) and cold-core clusters (right panel), from Beppo-SAX data [30]. The
radial coordinate is normalized to the radius with overdensity factor 180
Metals in Clusters of Galaxies 181
Abundances shown in Fig. 3 refer to the cluster central regions. However,
radial gradients in the iron abundance have been reported for several clus-
ters, starting with ASCA and then ROSAT data [34, 38, 40, 43, 122]. From
Beppo-SAX data a systematic study of the radial distribution of iron in many
clusters has been conducted [30], and Fig. 4 shows that clusters break up in
two distinct groups. The so-called cool core clusters (CC, formerly known as
cooling ow (CF) clusters before the failure of the CF model was generally
acknowledged) are characterized by a steep iron gradient in the core, reaching
0.6 solar near the center. Instead, in non-CC clusters (where no tempera-
ture gradient is found) no metallicity gradient appears either. The origin of
the dichotomy remains to be fully understood and will be discussed later. It
appears that silicon follows iron in the CC clusters, while instead there ap-
pears to be no abundance gradient in the lighter elements, such as oxygen,
sulfur, etc. [70]. The fact that metallicity gradients are found in association
with large temperature gradients in the central regions may look suspicious, as
noted for the strong dependence of Z
Fe
on ICM temperature, but the existence
of such gradients appears to be well established.
Figure 5 shows the the iron abundance in clusters at high redshifts, in-
cluding some at z > 1 [111]. Clearly, there is no appreciable departure from
the 0.3 solar abundance typical of the nearby clusters, indicating that the
bulk of iron was produced and distributed in the ICM well before z = 1.
10
8
6
k
T

(
k
e
V
)
4
2
0
0.8
0.6
0.4
0.2
0
0.5 1
0 0.5 1
z
Z
/
Z
Fig. 5. The iron abundance in the ICM (lower panel) and the ICM temperature
(upper panel) of a sample of high redshift clusters as a function of redshift [111].
The dashed line shows the typical abundance of z 0 clusters
182 A. Renzini
2.2 Elemental Ratios
X-ray observatories (especially ASCA, Beppo-SAX and XMM-Newton) have
such high spectral resolution that, besides those of iron, the emission lines of
many other elements can be detected and measured. These include oxygen,
neon, magnesium, calcium, silicon, sulfur, argon, and nickel. Most of these are
elements (i.e., made by an integer number of particles), which are pre-
dominantly synthesized in massive stars exploding as Type II supernovae. As
it is well known, iron-peak elements are also produced by Type Ia supernovae,
and 5075% of iron in the sun may come from them.
Early estimates from ASCA suggested a sizeable -element enhancement,
with [/Fe] +0.4 [77], later reduced to +0.2 [79] and eventually found
to be consistent with solar proportions [/Fe] 0.0 [57]. More recently,
Finoguenov et al. (2000) report near solar Ne/Fe, slightly enhanced Si/Fe,
and slightly depleted S/Fe, although with rather large error bars. From a sys-
tematic re-analysis of the ASCA archival data, a systematic increase of silicon
and nickel and a decrease of sulfur with ICM temperature was reported [7] (see
Fig. 6). Note that both silicon and sulfur are elements, and apparently they
do not follow the same trend. Note that the iron abundance has a pronounced
1.4
1.2
1.0
0.8
A
b
u
n
d
a
n
c
e

[
w
r
t

s
o
l
a
r
]
A
b
u
n
d
a
n
c
e

[
w
r
t

s
o
l
a
r
]
0.6
0.4
0.2
0.0 0.0
0.5
1.0
1.5
2.0
0 2 4 6 8 10 12
Temperature [keV]
1.4
1.2
1.0
0.8
A
b
u
n
d
a
n
c
e

[
w
r
t

s
o
l
a
r
]
0.6
0.4
0.2
0.0
0 2 4 6 8 10 12
Temperature [keV]
1.4
1.2
1.0
0.8
A
b
u
n
d
a
n
c
e

[
w
r
t

s
o
l
a
r
]
0.6
0.4
0.2
0.0
0 2 4 6 8 10 12
Temperature [keV]
Silicon Sulfur
Iron Nickel
0 2 4 6 8 10 12
Temperature [keV]
Fig. 6. The iron. nickel, silicon and sulfur abundances in the ICM (relative to solar
abundances), as a function of ICM temperature, from archival ASCA data [7]
Metals in Clusters of Galaxies 183
peak at T 2 keV, which in conjunction with the increase of silicon would
imply a very strong increase of the Si/Fe ratio with cluster temperature. If
taken at face value, such a trend implies a progressive dominance of SN IIs
over SN Ias in hotter and hotter clusters. No simple interpretation has so far
emerged of why the SN mix should be sensitive to the cluster environment,
and some caution is in order before accepting the reality of the trend. Indeed,
one can note that Fe is most accurately determined from Fe K, i.e., in clusters
hotter than 2.5 keV, while Si is best determined in cooler clusters, because
this element becomes virtually fully ionized in hot clusters. As a consequence
the Si lines become very weak, hence with low S/N (note the increasing size
of the Si error bars in Fig. 6). As a result, the Si/Fe ratio is poorly determined
in cool as well as hot clusters.
Figure 7 shows more recent data from XMM-Newton, in which no apparent
trend with cluster temperature is present in either iron or -elements [108].
Therefore, I would conclude that no compelling evidence has so far emerged
for other than near solar [/Fe] ratios globally in the ICM, especially when
all elements are lumped together. This leaves the possibility open for stellar
nucleosynthesis having proceeded in much the same way in the solar neighbor-
hood as well as at the galaxy cluster scale. In turn, this demands a similar ratio
of the number of Type Ia to Type II SNs, as well as a similar IMF, suggesting
that the star formation process (IMF, binary fraction, etc.) is universal, with
little or no dependence on the global characteristics of the parent galaxies
1
0
1
1 2
Fe/H
O/Fe
Si/Fe
S/Fe
T
ICM
(keV)
5 10
0
.
1
A
b
u
n
d
a
n
c
e
s

(
s
o
l
a
r
)
Fig. 7. The abundance of iron and the -elements oxygen, sulfur, and silicon as a
function of cluster temperature, as from XMM-Newton data [108]
184 A. Renzini
(and their large scale structure environment) in which molecular clouds are
turned into stars.
Alternatively, one can take at face value the variations of the abundance
ratios with cluster temperature, as well as the overabundance of some ele-
ments and the underabundance of others. One can then be forced to accept
rather contrived conclusions, such as the mix of the two SN types, and perhaps
even the nucleosynthesis of massive stars, depends on what the temperature
of the ICM will become billions of years after star formation has ceased. On
the other hand, one may argue that rich galaxy clusters are special places
in many senses, and that ICM abundances reect not only supernova nucle-
osynthesis yields, but also how eciently these are ejected, mixed into, and
retained by the ICM. However, no simple understanding of the apparent em-
pirical trends has yet emerged [41, 47, 67]. Still, the story may be dierent
for the cool-core clusters, which are characterized by a strong gradient in the
iron abundance within the cool core. I shall return to cool-core clusters later.
In summary, in the following I will assume that clusters, on a global scale,
have solar elemental ratios and the total heavy element abundance is 0.3 solar,
or 0.006 by mass.
2.3 The Iron Mass-to-Light Ratio
One useful quantity is the iron-mass-to-light ratio (FeM/L) of the ICM, i.e.
the ratio M
ICM
Fe
/L
B
of the total iron mass in the ICM over the total B-band
luminosity of the galaxies in the cluster. In turn, the total iron mass in the
ICM is given by the product of the iron abundance times the mass of the
ICM, i.e., M
ICM
Fe
= M
ICM
Z
Fe
ICM
. Figure 8 shows the resulting FeM/L from an
earlier compilation [97]. The drop of the FeM/L in poor clusters and groups
(i.e. for kT

<
2 keV) can be traced back to a drop in both the iron abundance
(which however may not be real, see above) and in the ICM mass. Such groups
appear to be gas poor compared to clusters, which suggests that they may
have been subject to baryon and metal losses due to strong galactic winds
driving much of the ICM out of them [28, 91, 93], or such winds having pre-
heated and inated the gas distribution around galaxies, thus prevents it to
fall inside groups. In one way or another, the break seen in Fig. 8 is likely to
be related to the break of self-similarity in the X-ray luminosity-temperature
relation (see later).
In these lectures I will mainly deal with clusters with kT

>
23 keV,
for which the interpretation of the data appears more secure, but several
cautionary remarks are in order, even concerning these hotter clusters. The
rst is that the iron abundances used to construct Fig. 8 did not take into
account that some clusters have sizable iron gradients. In principle, X-ray
observations can provide both the run of gas density and abundance with
radius, making it possible to integrate their product over the cluster volume
and get M
ICM
Fe
. To my knowledge, so far this has been completed only for one
cluster [89]. However, the extra-iron contained within the iron gradient core
Metals in Clusters of Galaxies 185
4 6 2
2
4
5
L
o
g
(
M
F
e
/
L
B
)

(
M

/
L

)
3
1
kT (keV)
The ICM Iron Mass to Light Ratio
8 10
Fig. 8. The iron mass to light ratio of the ICM of clusters and groups as a function
of the ICM temperature from an earlier compilation [93]. Data are taken from the
following sources: lled circles [6]; lled triangles [113]; open triangle [27]; open
square [75]; lled square [85]; open circles [76]
seems to be the product of the central cD galaxy, and may represent only a
small fraction of the whole M
ICM
Fe
[31].
Another concern is that two of the three ingredients entering in the calcu-
lation of the FeM/L values shown in Fig. 8 (namely M
ICM
and L
B
) may not
be measured precisely in the same way in the various sources used for the com-
pilation. Both quantities come from a radial integration up to an ill-dened
cluster boundary, e.g., the Abell radius, the virial radius, or to a radius of
some xed overdensity. Sometimes it is quite dicult to ascertain what def-
inition has been used by one author or another, with the complication that
X-ray and optical data have generally been collected by dierent groups using
dierent assumptions. There is certainly room for improvement here, and a
new compilation paying attention to analyse all clusters in a homogeneous
way would be highly desirable. Finally, estimated total luminosities L
B
refer
to the sum over all cluster galaxies, and do not include the population of stars
which are diused through the cluster, and which may account for at least
10% of the total cluster light, and perhaps more [4, 37]. In any event, Fig. 9
shows a more recent compilation of FeM/L values [29], and compare them to
the old ones nding fair agreement.
While keeping these cautions in mind, we see from Figs. 8 and 9 that the
FeM/L runs remarkably at with increasing cluster temperature, for kT

>
2
3 keV. This constancy of the FeM/L comes from both Z
Fe
ICM
and M
ICM
/L
B
showing very little trend with cluster temperature, see Fig. 3 in this paper, and
Fig. 4 in [93], where M
ICM
/L
B
25h
1/2
70
(M

/L

). The resulting FeM/L is

therefore
(FeM/L)
ICM
= Z
Fe
ICM
M
ICM
L
B
0.3Z

Fe
25h
1/2
70
0.01 h
1/2
70
(M

/L

) , (1)
186 A. Renzini
4 6
2
L
o
g

M
F
e
/
L
B

(
s
o
l
a
r

u
n
i
t
s
)
3
1
T
ew
(keV)
8 10
= 2500
= 1000
= 500
Fig. 9. The iron mass to light ratio of the ICM of clusters and groups as a function
of the ICM temperature from a more recent compilation [29]. The values refer to
the overdensity within which the ratio is calculated. Note the good agreement with
an old compilation [93], shown here as lled circles, when virtually all the cluster
ICM mass is included (i.e., for the lowest value of )
i.e., in the ICM there are about 0.01 solar masses of iron for each solar lu-
minosity of the cluster galaxies. This value is 30% lower than adopted
in [93] and shown in Fig. 8, having consistently adopted here for Z

Fe
the rec-
ommended meteoritic iron abundance [2], i.e., Z

Fe
= 0.0013. Assuming solar
elemental proportions for the ICM, the ICM metal mass to light ratio is there-
fore 0.3 0.02 25h
1/2
70
= 0.15 (M

/L

), having adopted Z

= 0.02 and
for h
70
= 1.
A very accurate analysis was performed recently for the A1983 cluster [89],
paying attention to measure M
ICM
and L
B
within the same radius. The result
is FeM/L= (7.5 1.5) 10
3
h
1/2
70
(M

/L

), in fair agreement with the

estimate above.
The most straightforward interpretation of the constant FeM/L is that
clusters did not lose iron (hence baryons), nor dierentially acquired pristine
baryonic material, and that the conversion of baryonic gas into stars and
galaxies has proceeded with the same eciency and the same stellar IMF in
all clusters [93]. Otherwise, there should be cluster to cluster variations of
Z
Fe
ICM
and FeM/L. All this is true in so far as the baryon to dark matter ratio
is the same in all kT

>
2 keV clusters [123], and the ICM mass-to-light ratio
and the gas fraction are constant. Nevertheless, there may be hints for some
of these quantities showing (small) cluster to cluster variations [5, 74, 89], but
no rm conclusion has yet been reached.
Metals in Clusters of Galaxies 187
2.4 The Iron Share Between ICM and Cluster Galaxies
The metal abundance of the stellar component of cluster galaxies is derived
from integrated spectra coupled to synthetic stellar populations. Much of the
stellar mass in clusters is conned to passively evolving spheroids (ellipticals
and bulges) for which the iron abundance Z
Fe

may range from 1/3 solar

to a few times solar. For example, among ellipticals metal-sensitive spectral
features such as the magnesium index Mg
2
range from values slightly lower
than those in the most metal rich globular clusters of the Milky Way bulge
(which are nearly solar), to values for which models indicate a metallicity a
few times solar [69]. The FeM/L of cluster galaxies is then given by:
(FeM/L)
gal
= Z
Fe

L
B
0.0046 h
70
(M

/L

) , (2)
where Z
Fe

= Z

Fe
and one has adopted M

/L
B
= 3.5 h
70
. This estimate comes
from the M/L
B
determinations for a sample of ellipticals [116], which have
been used to derive the an average M/L
B
= 4.2 h
70
when adopting the Coma
cluster luminosity function [123]. This value is nally reduced to 3.5 h
70
taking
into account a likely 30% dark matter contribution to the total mass within
the galaxy eective radius [18]. The total cluster FeM/L (ICM+galaxies) is
therefore 0.015 (M

/L

), for h
70
= 1. The ratio of the iron mass in the
ICM to the iron mass locked into stars and galaxies is then
Z
Fe
ICM
M
ICM
Z
Fe

2.2h
3/2
70
, (3)
having adopted Z
Fe
ICM
= 0.3 solar, Z
Fe

= 1 solar, and M
ICM
/M

= 9.3h
3/2
70
as for the Coma cluster [123]. So, it appears that there is 2 times more iron
mass in the ICM than locked into cluster stars (galaxies), perhaps even more
if Z
Fe

is subsolar due to an abundance gradient within individual galaxies [3].

In turn, this empirical iron share (ICM vs. galaxies) sets a strong constraint
to models of the chemical evolution of galaxies. Under the same assumptions
as above, the total metal mass to light ratio (ICM + galaxies) is therefore
0.15h
1/2
70
+ 0.07h
70
0.2 (M

/L

). This can be regarded as a fully

empirical determination of the metal yield of (now) old stellar populations.
I would like to emphasize that the values of the total, cluster FeM/L and
of the iron share derived in this section strictly depend on the adopted values
of M

/L
B
and M
ICM
/L
B
, which may be subject to change as better estimates
become available.
3 Metal Production: the Parent Stellar Population
The constant FeM/L of clusters means that the total mass of iron in the ICM
is proportional to the total optical luminosity of the cluster galaxies [6, 23,
91, 104]. The simplest interpretation is that the iron and all the metals now
188 A. Renzini
in the ICM have been produced by supernovae belonging to the same stellar
generation whose surviving low-mass stars now radiate the bulk of the cluster
optical light. As much of the cluster light comes from old spheroids (ellipticals
and bulges), one can conclude that the bulk of cluster metals were produced by
the same stellar generations that make up the old spheroids that we see today
in clusters.
It is also interesting to ask which galaxies have produced the bulk of the
iron and the other heavy elements, i.e. the relative contribution as a function of
the present-day luminosity of cluster galaxies. From their luminosity function
it is easy to realize that the bright galaxies (those with L

>
L

) produce the
bulk of the cluster light, while the dwarfs contribute a negligible amount of
light in spite of dominating the galaxy counts by a large margin [109]. In
practice, most galaxies dont do much, while only the brightest 3% of all
galaxies contribute 97% of the whole cluster light. Giants dominate the
scene while dwarfs dont count much. Following the simplest interpretation,
according to which the metals were produced by the same stellar population
that now shines, one can conclude that also the bulk of the cluster metals have
been produced by the giant galaxies that contain most of the stellar mass. The
relative contribution of dwarfs to ICM metals may have been somewhat larger
than their small relative contribution to the cluster light, since metals can
more easily escape from their shallower potential wells [109]. This is, however,
unlikely to alter the conclusion that the giants dominate metal production by
a very large margin.
Up to about 3/4 of the whole mass in stars in the local universe is now in
spheroids, 1/4 in disks, and less than 1% in irregular galaxies [10, 33, 44].
In clusters the dominance of spheroids is likely to be even stronger than in the
general eld. The prevalence of spheroids oers an opportunity to estimate the
epoch (redshift) at which (most) metals were produces and disseminated, since
we now know quite well when most stars in cluster spheroids were formed.
Following the rst step in this direction [16], I believe that the most precise
estimates of the age (redshift of formation) of stellar populations in cluster
elliptical galaxies come from the tightness of several correlations, such as the
color-magnitude, fundamental plane, and the Mg
2
relations, and especially
by such relations remaining tight all the way to z 1 [96, 105, 117]. This has
taught us that the best way of breaking the age-metallicity degeneracy in
the global proprties of stellar populations is to look back at high redshift
galaxies. The collective evidence indicates that most stars in cluster ellipticals
formed at z

>
3, while only minor episodes of star formation may have occurred
later.
With most of the star formation having taken place at such high red-
shift, the major fraction of cluster metals should also have been produced and
disseminated at z

>
3. Little evolution of the ICM composition is then ex-
pected all the way to high redshifts, with the possible exception of iron from
SNIas, for which the rate of release does not closely follow the star formation
rate (SFR), as does the SNII rate, but for which the rate of release of iron is
Metals in Clusters of Galaxies 189
modulated by the distribution of the delay times between formation of the pre-
cursor and explosion time. As illustrated in the next section, one expects that
the SNIa rate peaks shortly after a burst of star formation and then rapidly
declines, with most events taking place within 12 Gyr after formation. If this
is the case, no appreciable evolution of the iron abundance in clusters should
be detectable from z = 0 to z 1 an argument supported by observational
evidence [111]. Note, however, that late winds will keep enriching the ICM at
a decreasing rate [23].
4 Metal Production: Type Ia vs. Type II Supernovae
Most heavy elements (metals) are produced by supernovae (SN), of which
there are two main types: supernovae of Type II (SNII) result from the core
collapse of massive stars (M

>
8 M

), while supernovae of Type Ia (SNIa)

result from the thermonuclear explosion of a degenerate star, i.e., a white
dwarf. Their relevance to the metal enrichment in clusters is discussed next.
4.1 Iron from SNIas, SNIa Progenitors and SNIa Rate
As it is well known, clusters are now dominated by E/S0 galaxies, which
produce only Type Ia SNs at a rate of (0.16 0.06)h
2
70
SNU [19], with
1 SNU corresponding to 10
12
SNs yr
1
L
1
B
. Assuming such rate to have
been constant through cosmological times (13 Gyr), the number of SNIas
exploded in a cluster of present-day luminosity L
B
is therefore 1.610
13

1.3 10
10
L
B
h
2
70
2 10
3
L
B
. With each SNIa producing 0.7 M

of iron,
the resulting FeM/L of clusters would be:

M
Fe
L
B

SNIa
(1.4 0.5) 10
3
h
2
70
, (4)
which falls short by a factor 10 compared to the observed cluster FeM/L
(0.015 for h
70
= 1). The straightforward conclusion is that either SNIas did
not play any signicant role in manufacturing iron in clusters, or their rate in
what are now E/S0 galaxies had to be much higher in the past. This argues
for a strong evolution of the SNIa rate in E/S0 galaxies and bulges, with the
past average being 5 10 times higher than the present rate [23].
In the case of SNIas we believe we know fairly precisely the amount of
iron released by each event, while we still dont know for sure what are the
progenitors producing the events. There is universal agreement that SNIas
originate from the thermonuclear explosion of white dwarfs (WD) made of
carbon and oxygen (CO), once they reach the Chandrasekhar limit ( 1.4 M

)
by having accreted mass from a donor binary companion. Carbon is then
ignited explosively and the star disrupts completely. The event produces
0.7 M

of iron-peak elements, mostly

56
Ni, which decays into
56
Co and nally
190 A. Renzini
into
56
Fe, powering the SN light curve. Along with iron, 0.150.28 M

of
silicon are also synthetized [58].
Several dierent evolutionary paths, however, may lead to the SN explo-
sion, and each of them would be characterized by a dierent evolution of the
past SNIa rate. Two main channels are currently considered as viable, the so-
called single-degenerate (SD) and the double-degenerate (DD) channel. In the
SD case, a CO-WD accretes hydrogen-rich material from a companion-star,
and processes it through H and He burning, increasing the CO mass until the
WD exceeds the Chandrasekhar limit [121]. The case of helium being ignited
explosively, even before the Chandrasekhar limit is reached, has also been
considered (SD/Sub-Chandra exploders), but the resulting synthetic spectra
dier markedly compared to those of observed SNIas [66]. In the DD option,
the secondary star in the binary is also a WD, and the two stars spiral in
towards eachother, due to angular momentum loss via gravitational wave ra-
diation (GWR), until the less massive star lls its Roche lobe and the two
WDs merge together [56]. Once again, a SNIa may result if the combined mass
exceeds the Chandrasekhar limit.
In the SD channel, the time of explosion, i.e. the delay between the star
formation event and the supernova, is set by the time it takes the secondary
(less massive) star to evolve o the main sequence and ll its Roche lobe.
Hence the delay strongly increases with decreasing mass and, in principle,
delay times of order the Huble time are possible, provided 1M

donors are
able to transfer enough mass to grow the WD beyond the threshold mass for
ignition. In the DD channel, the delay time is further augmented by the time
it takes the secondary WD to spiral in, due thanks to the GWR, which is
given by:

GWR
=
0.15A
4
(M
WD
1
+M
WD
2
)M
WD
1
M
WD
2
Gyr , (5)
where A is the intial separation of the DD system in units of the solar radius,
and the two WD masses are also in solar units. Also in this case delay times
can easily exceed the Hubble time. In both cases, the minimum delay time is
set by the lifetime of the maximum initial mass that produces a WD remnant,
i.e. 35 Myr, the lifetime of 8 M

stars [55].
The run of the past SNIa rate is therefore proportional to the distribution
function of the delay times, and hence depends on the distributions of the
initial binary parameters (i.e., masses M
1
and M
2
, and separations), since a
wide range of them can lead to a successful explosion, as well as additional
parameters describing the mass transfer phases that take place in the course
of the binary evolution and their outcome.
The distributions of the delay times f
Ia
() for both the SD and DD chan-
nels have been recently calculated in a fashion that allows one for an eec-
tive exploration of the parameter space [49], and the results are shown in
Figs. 10 and 11, respectively. The distributions can dier widely, but have
several characteristics in common: (a) f
Ia
() = 0 for < 35 Myr, (b) from
Metals in Clusters of Galaxies 191
1
0
1
I
a
k

+

l
o
g

f
S
D
2
3
8 8.5
Color Coding:
= 2.35 = 1
= 2.35 = 0.5
= 2.35 = 0.005
= 2.6 = 2
9
log
9.5 10
Fig. 10. The distribution function of the delay times for the SD model for the
production of Type Ia supernovae, and for various values of the IMF slope (IMF
M

) and and the slope of the distribution function of the binary initial mass
ratios q = M
2
/M
1
(with f(q) q

). The distributions present two cusps, one at

1 Gyr, which is due to the requirement that the primary produces a CO-WD,
and the second at 8 Gyr is due to the requirement that the mass of the WD
plus the mass of the envelope of the secondary star exceeds the Chandrasekhar limit
(from [49])
zero, f
Ia
() steeply increases, reaching is maximum by

<
10
8
yr, (c) the
maximum is followed by a plateau phase with a duration

<
1 Gyr and is not
model-dependent, (d) the plateau is followed by a decline with a rate which
is extremely model-dependent, especially in the SD scenario, and (e) in the
SD case the late decline is much steeper than in the DD case. This is caused
by the fact that at late times (e.g., 10 Gyr after the burst of star formation)
the secondary components have quite low mass 1 M

, their envelope mass

available for transfer to the WD is just a fraction of this, and therefore suc-
cessful exploders are restricted only to those few systems with very massive
WDs. In general, the late decline is primarily controlled by the distributions
of the binary masses that lead to a SNIa event in the SD case, and by the
distribution of the initial separations of the WD+WD systems in the DD case,
which are both dicult to predict. Nevertheless, properties (a), (b), (c) and
(e) are generic, i.e., common to all combinations of model assumptions and
parameters.
At rst sight the various curves in Figs. 10 and 11 look very similar to
each other, but it is only the use of a log-log plot that gives this impression.
192 A. Renzini
1
0
1
k

+

l
o
g

f
I
a

+

c
o
s
t
v
v
2
3
8 8.5

n
= 0

n, x
= 1Gyr

g
= 0.75

gw, i
= 1Myr
SDChandra
CLOSE DD
WIDE DD
9
log
9.5 10
Fig. 11. The distribution function of the delay times for the DD case in the pro-
ductiob of SNIa. Two dierent assumptions are considered which concern the rst
common envelope event (labelled CLOSE DD and WIDE DD respectively), com-
pared to one rendition of the delay time for the SD case. The relative normalizations
are arbitrary. (from [49])
Figure 12 shows as function of time (not log-time) the distributions of delay
times for various model parameters, both for the SD and the DD cases. The
curves are all normalized to give the same SNIa rate at t = 10 Gyr, equivalent
to force all models to account for the present SNIa rate in ellipticals. Clearly,
the time integrals of the delay times (which are proportional to the total
number of SNIa events) dier dramatically from one case to another, and
therefore so does the amount of iron from SNIas predicted by the various
models. This is shown in Fig. 13, where the FeM/L predicted by various
models is plotted as a function of the SNIa rate at t = 12 Gyr divided by the
average rate in the past [48]. Clearly some models dramatically overpredict the
FeM/L compared to the cluster value, while others dramatically underpredict
it. By and large, SD models appear to be excluded, because the late decline
of their rate is too fast: if forced to account for the present rate in ellipticals
their past rate would have been too high. Therefore, this comparison favors
DD models, actually some particular version of them [48].
This conclusion, however, rests on the assumption that nature has chosen
only one path to make SNIas (either DD or SD), but by no means can we
exclude that nature is able to make SNIas from both SD and DD precursors.
If so, the theoretical delay times shown in these gures suggest that SDs would
Metals in Clusters of Galaxies 193
0 2 4 6
(Gyr)
8 10 12
1
0
1
SD Chandra
SD SubChandra
CLOSE DD,
n,x
= 0.4 Gyr
WIDE DD,
n,x
= 0.4 Gyr
CLOSE DD,
n,x
= 1 Gyr
WIDE DD,
n,x
= 1 Gyr
l
o
g

[
f
I
a
/
f
I
a

(
1
0
)
]
2
3
Fig. 12. The distribution function of the delay times for the DD (from [49])
0.5
SD Ch, e=1
SD SCh
DDC (0.4; 0.975)
DDC (0.4; 75)
DDC (0.4; 0.525)
DDW (0.4; 0)
DDW
(0.4; 0.9)
DDC (1; 0.975)
DDC (1; 0.75)
DDC
(1; 0.525)
DDW (0.4; 0.9)
1
1.5
L
o
g

(
M
F
E
/
L
B
)
Log (f
Ia
(12)/<f
Ia
>
12
)
I
a
2
2 1.5 1 0.5 0
2.5
Fig. 13. The distribution function of the delay times for the DD (from [49])
194 A. Renzini
dominate at early times, and DDs would dominate at late times, as indicated
in Fig. 11. In any event, the bottom line is that these theoretical SNIa rates
predict that the bulk of iron should be produced within 12 Gyr from the
major phase of star formation phase in ellipticals.
The distributions of delay times derived for the SD and DD channels are
sharply at variance with those derived from the SNIa statistics at high redshift
by the GOODS Piggyback transient survey [107]. From the counts of SNIas
up to z = 1.55 in the GOODS elds, and from the cosmic history of star
formation, it was claimed that such data would require the minimum delay
time to be of the order of 2 Gyr, with an average delay time of 4 Gyr.
Indeed, their best-t solution is a Gaussian distribution uctuation centered
at = 4 Gyr, with = 0.8 Gyr [107]. With only one possible exception,
a distribution such as this bears no resemblance whatsoever to any of the
theoretical distributions shown here or in [49], or ever considered for SNIas
existing scenarios (e.g., [126]). The exception would be a DD model in which
the donor is a helium WD, which typically requires 1 Gyr to appear after
the formation of a stellar population. The explosion, however, would need to
be a helium detonation that could ignite the underlying CO core, but which
would give a spectrum totally at variance with the observed SNIa spectra.
I conclude that either we have so far completely missed the identication of
the nature of the SNIa progenitors, or the mentioned estimate of the rate
at z 1.5 is severely biased. Being based on only 2 events at z > 1.4. Im
inclined to favor this latter option.
4.2 Iron and Metals from SNIIs and the IMF
In the case of SNIas, we currently believe to have a fairly precise knowledge of
the amount of iron produced by each event, while the nature of the progenitors
and the evolution of the SN rate still remain as open issues. The case of Type
II SNs is quite the opposite: we believe to have unambiguously identied
the progenitors (stars more massive than 8M

), while a great uncertainty

aects the amount of iron M
II
Fe
(M) produced by each SNII event as a function
of the progenitors mass. This is due to the fundamental diculty for core-
collapse SN models to precisely locate the mass cut between the collapsing
core that forms the neutron star remnant, and the ejected envelope. This cut
is often within the iron-peak layer. On the other hand, the SN luminosity at
late times can be used to infer the amount of radioactive Ni-Co (and hence
eventually iron) that was ejected. An early study indicated small variations
from one event to another (0.040.10 M

) [81]. This led to the assumption

that M
II
Fe
is a weak function of initial mass, with an average yield of 0.07 M

of iron per SN event (as in SN 1987A) [93]. More recent studies, based on a
larger sample of SNII events, have actually detected very large dierences from
one event to another (ranging from 0.002 M

to 0.3 M

[114]). Figure 14
shows the mass of ejected
56
Ni as a function of the velocity of the ejecta [51],
Metals in Clusters of Galaxies 195
0.1
0.01
0.001
1000 2000
99br
92am
4000 6000 8000
M
N
I

(
M

)
v
50
(km s
1
)
Fig. 14. The mass of
56
Ni ejected by a well studied sample of Type II supernovae,
as a function of the velocity of the ejecta; from [51]
and averaging over the 22 SNIIs in the sample one gets M
Ni
= 0.073 M

,
so close to the adopted value.
The total number of SNIIs N
SNII
is obtained by integrating the stellar
IMF from 8 to 40 M

for example, with the IMF being expressed as [95]:

(M) = a(t, Z)L
B
M
s
, (6)
where a(t, Z) is a (slow) function of the SSP age and metallicity. For example,
for t = 12 Gyr, a(Z) = 2.22 and 3.12, respectively for Z = Z

and Z = 2Z

[69], with L
B
and M expressed in solar units.
Clearly, the atter the IMF slope, the larger the number of massive stars
per unit present luminosity, the larger the number of SNIIs, and the larger the
implied FeM/L. Thus, adopting M
II
Fe
= 0.07M

and a = 3 and integrating

over the IMF one gets:

M
Fe
L
B

SNII
=
M
II
Fe
N
SNII
L
B

0.003 for s = 2.7

0.009 for s = 2.35
0.035 for s = 1.90
(7)
Hence, if the Salpeter IMF (s = 2.35) applies also to clusters ellipticals, then
SNIIs underproduce iron by less than factor of 2.
Constraints on the IMF slope in cluster ellipticals can be derived from the
evolution of their M/L ratio with redshift, as inferred from the shift of the
fundamental plane in clusters at increasingly high redshifts. This is illustrated
in Fig. 15, showing the evolution of the M/L
B
ratio all the way to z = 1.27,
which indeed favors a Salpeter IMF [100].
196 A. Renzini
0.5 0 1.5 1
0.6
0.4

L
o
g

M
*
/
L
B
0.2
0
redshift
z
p
= 5 s = 2.35
z
F
= 3 s = 1.5
z
F
= 3 s = 3.5
M
TO
= ~1 M
o
M
TO
= ~1.4 M
o
z
p
= 3 s = 2.35
z
p
= 2 s = 2.35
Fig. 15. The evolution with redshift of the M

/L
B
ratio of simple stellar populations
(SSP) of solar metallicity and various IMF slopes (dN M
s
dM), normalized
to its value at z = 0 [100]. Convergence cosmology (
m
= 0.3,

= 0.7, H

=
70 kms
1
Mpc
1
) and SSP formation redshifts as indicated. The data points [117]
refer to the shifts in the fundamental plane locations for clusters of galaxies at
various redshifts. Note that for such high formation redshifts the stellar mass at the
main sequence turno is 1.4 M

at z = 1.5 and M

at z = 0, as indicated by
the arrows
At z = 0, however, elliptical galaxies harbour 12 Gyr old stellar popula-
tions, with stars of 1 M

at the main sequence turno (MSTO). By z = 1.5

the precursors of such populations have an age of only 3 Gyr, and corre-
spondingly a higher mass at the MSTO, but not by much so. Specically, the
MSTO mass at an age of 3 Gyr is 1.41.5 M

, and therefore by following

the evolution of the FP with redshift up to z 1.5 (or equivalently of the
mass-to-light ratio) we explore the IMF slope within the rather narrow mass
interval 1

<
M

<
1.4 M

. In practice, we measure the slope of the IMF only

near M = M

.
Given that iron is produced by both types of supernovae, iron is not the
best element to constrain the IMF slope in the high mass range. Instead,
elements are produced almost exclusively by SNIIs, and therefore the IMF
slope can better be constrained by them, and in particular by oxygen and
silicon, which abundance in the ICM is aected by relatively small errors.
Therefore, in a similar way to the case for iron, the metal-mass-to-light ratio
for the X element can be calculated in a straightforward manner from:
M
X
L
B
= a(t, Z)

40
8
m
X
(M)M
s
dM , (8)
where m
X
(M) is the mass of the element X which is produced by a star of
mass M. Adopting a(t, Z) = 3, m
X
(M) for oxygen and silicon from [124], and
integrating (8), one obtains the oxygen- and the silicon-mass-to-light ratios
Metals in Clusters of Galaxies 197
1.5
1
0.5
0
Salpeter
Silicon
Oxygen
1.5
2
2 2.5 3
L
o
g

1
0
*
M
s
i
/
L
B
L
o
g

M
o
/
L
B
IMF Slope in the range 1<M<40 M.
Fig. 16. The Oxygen- and the Silicon-Mass-to-Light Ratios as a function of the
IMF slope calculated from (8) with a(t, Z) = 3 and using standard nucleosynthesis
prescriptions [124]. The horizontal lines show the observed average values of these
ratios in clusters of galaxies, and their range of uncertainty
which are shown in Fig. 16 as a function of the IMF slope [100]. As expected,
the M
O
/L
B
and M
Si
/L
B
ratios are extremely sensitve to the IMF slope. The
values observed in clusters of galaxies (ICM plus galaxies)
1
are 0.1 and
0.01 M

/L

respectively for oxygen and silicon [39, 88]. These empiri-

cal values are reported in Fig. 16 showing that with the Salpeter IMF slope
(s = 2.35) the standard explosive nucleosynthesis from Type II supernovae
produces just the right amount of oxygen and silicon to account for the ob-
served M
O
/L
B
and M
Si
/L
B
ratios in cluster of galaxies, having assumed that
most of the cluster B-band light comes from

>
12 Gyr old stellar populations.
Figure 16 also shows that with s = 1.35 such a top heavy IMF (in various
circumstances invoked to ease discrepancies between theories and observa-
tions) would overproduce metals by more than a factor of 20. This is indeed
the change one expects in M
O
/L
B
, M
Si
/L
B
, etc. for a change in the IMF slope
s = 1 when considering that the light L
B
is provided by M

stars and
the metals by 25 M

stars.
In summary, it appears that, with a Salpeter IMF and standard nucleosyn-
thesis prescriptions, massive stars can produce the observed amonts of oxygen
and silicon which are present in clusters of galaxies, while perhaps falling short
by a factor 2 to produce the observed iron. Yet, with an IMF just slightly
shallower than Salpeter, SNIIs could make also all the iron. There should
then however be clearer evidence for an -element overabundance than cur-
rently indicated by the observations(cf. Fig. 7). Hence, nucleosynthesis may
1
These values result from averaging over the reported values for individual clusters
with dierent ICM temperature, and take into account that 1030% of the
stellar mass in clusters is not bound to individual galaxies [4, 45].
198 A. Renzini
well have proceeded in clusters not unlike in the Milky Way, where we cur-
rently believe that about 1/2 of iron has been produced by core-collapse
Type II supernovae, and the other half by thermonuclear Type Ia supernovae.
I shall discuss later the chemical evolution of clusters and the Milky Way.
With 2.35

<
s

<
2.7 and a past average rate of SNIas in ellipticals

>
5 times
the present rate, the iron content of clusters and the global ICM [/Fe] ratio
are grossly accounted for, with SNIas then having produced 1/2 of the
total cluster iron, not unlike in standard chemical models for the Milky Way.
This is not to say that this has been rmly proved, but it seems to me to be
premature to abandon the attractive simplicity of a universal nucleosynthe-
sis process (i.e., IMF and SNIa/SNII ratio) before embarking towards more
complex, multi-parametric scenarios.
5 Metals from Galaxies to the ICM:
Ejection vs. Extraction
Having established that most metals in clusters are in the ICM and not in
the ISM of their parent galaxies, it remains to be understood how they were
transfered from galaxies to the ICM. There are two main possibilities: extrac-
tion by ram pressure stripping as galaxies plow through the ICM, and ejection
by galactic winds powered from within the galaxies themselves. In the latter
case the power can be supplied by supernovae (the so-called star formation
feedback) and/or by AGN activity.
Ram pressure stripping certainly exists in clusters [102], as clusters are
assembled by growing group and isolated galaxies which, by entering a dense
ICM, are stripped of their gas and then become quiescent. Several arguments,
however, favor ejection over extraction [99]:
There appears to be no trend of either Z
Fe
ICM
or the FeM/L with cluster
temperature or cluster velocity dispersion (
v
), while the eciency of ram
pressure stripping increases steeply with increasing
v
.
Field ellipticals appear to be virtually identical to cluster ellipticals. They
follow basically the same Mg
2
and fundamental plane relations [11, 12],
which does not show any appreciable trend with the local density of galaxies,
or at most a very weak one. If stripping was responsible for extracting metals
from galaxies one would expect galaxies in low density environments to have
retained more metals, hence showing higher metal indices for given , which
is not seen.
Non-gravitational energy injection of the ICM seems to be required to ac-
count for the break of the self-similar X-ray luminosity-temperature relation
for groups and clusters [86]. While galactic winds are an obvious vehicle
for such pre-heating, no pre-heating is associated to metal transfer by ram
pressure.
Metals in Clusters of Galaxies 199
Strong galactic (super)winds are actually observed in starburst galaxies at
low and high redshift, which are thought to be the progenitors of local ellip-
ticals [52, 83, 106, 118].
Most metals have been produced at very early times (z

>
3), probably well
before the clusters were assembled, hence before the dense ICM was in place,
hence when there was not much ram pressure exerted on the galaxies.
One can quite safely conclude that metals in the ICM have been ejected
from galaxies by supernova (or AGN) driven winds, rather than stripped by
ram pressure [34, 91]. Two kinds of galactic winds are likely to operate: early
winds driven by the starburst forming much of the galaxys stellar mass itself,
and late winds or outows where the gas comes from the cumulative stellar
mass loss as the stellar populations passively age. Late winds are also likely
to operate, as the stellar mass loss from the aging population ows out of
spheroids, being either continuously driven by a declining SNIa rate [23], or
intermittently by recurrent AGN activity [24].
5.1 The Metallicity Gradients in Cool-Core Clusters
Besides showing a radial gradient in the iron abundance (see Fig. 4), virtu-
ally all CC clusters host a cD galaxy at their center [30]. Hence, it is quite
natural to associate the two phenomena, and attribute to the cD galaxy the
responsibility to have further enriched the central regions of the clusters. This
possibility has been recently explored in some detail [15], and Fig. 17 shows
0.5
0.6
0.4
0.2
0.0
0.2
0.4
0.0
10%
SN II
40% 60%
fraclion of Fe contributed by SN la
80%
Cent.
Sun
N5044
RGH80
Cent.
WDD1
WDD2
WDD3
N1550
M87
90% 95% 97.5%
0.5
[Fe/O]
[
F
e
/
S
i
]
1.0
Fig. 17. The [Fe/Si] and [Fe/O] ratios for various proportions of SNIa contribution
to the total iron abundance, an indicated in the inner scale [15]. The lines refer to
dierent SNIa models, while the data refer to the innermost (r

<
100 kpc) of some
cool-core clusters. An asterisk shows the pure SNII elemental retios, and another
asterisk refer to the solar proportions
200 A. Renzini
the main results of this analysis. Both ratios [Fe/Si] and [Fe/O] depend on the
relative proportion of the two supernova types that contribute to the nucle-
osynthesis, with the asterisk showing the ratio from SNIIs only. By increasing
the proportion of iron contributed by SNIas, the ratios move along the tra-
jectories shown by the various lines in Fig. 17, which refer to dierent SNIa
models. The data points refer to the central regions (typically within 100 kpc)
of the indicated (CC) clusters. It appears that these regions are exceptionally
rich in SNIa products, which can be interpreted as due to the late winds for the
cDs, powered by SNIas themselves [23], and/or recurrent AGN activity [24].
By the way, sitting at the bottom of the cluster potential well, cD galaxies
experience no ram pressure, yet, they appear to have further enriched the
Cool Core of CC clusters. This is a further argument for a dominant role
of winds in transfering metals from galaxies to the ICM.
6 Metals as Tracers of ICM Pre-heating
The total amount of iron (and metals) in clusters represents a record of the
overall past supernova activity as well as the history of the mass and energy
ejected from cluster galaxies. The empirical values FeM/L can be used to set
a constraint on the energy injection into the ICM by supernova-driven galactic
winds [92]. The total SN heating is given by the kinetic energy released by one
SN ( 10
51
erg) times the number of SNs that have exploded. It is convenient
to express this energy per unit present optical light L
B
, i.e.:
E
SN
L
B
= 10
51
N
SN
L
B
= 10
51

M
Fe
L
B

TOT
1
M
Fe

10
50
(erg/L

) , (9)
where the total (ICM+galaxies) FeM/L=0.015 M

/L

is adopted, and the

average iron release per SN event is assumed to be 0.15 M

(appropriate if
SNIas and SNIIs contribute equally to the iron production). This estimate
should be accurate to within a factor 2.
The kinetic energy injected into the ICM by galactic winds, again per unit
cluster light, is given by 1/2 the ejected mass (M
ICM
Fe
/Z
Fe
w
) times the typical
wind velocity squared, i.e.:
E
w
L
B
=
1
2
M
ICM
Fe
L
B

v
2
w
Z
Fe
w

1.5 10
49
Z
Fe

Z
Fe
w

v
w
500 kms
1

2
10
49
(erg/L

) ,
(10)
where the empirical FeM/L for the ICM has been used, and the average
metallicity of the winds Z
Fe
w
is assumed to be two times solar. As usual in
the case of thermal winds, the wind velocity v
w
is of the order of the escape
velocity from individual galaxies. Again, this estimate may be regarded as
accurate to within a factor of 2, or so.
A rst inference is that of order of 5%20% of the kinetic energy released
by SNs is likely to survive as kinetic energy in galactic winds, thus contributing
Metals in Clusters of Galaxies 201
to the heating of the ICM. A roughly similar amount goes into work to extract
the gas from the potential well of individual galaxies, while the rest of the SN
energy has to be radiated away locally and does not contribute to the feedback.
This estimated energy injection represents a small fraction of the thermal
energy of the ICM of rich (hot) clusters and so it has only a minor impact
on the history of the ICM. However, in groups it represents a non-negligible
fraction of the thermal energy of the ICM, thus aecting its evolution and
present structure. The necessity of some non-gravitational heating (or pre-
heating) was recognized from the break of the self-similarity demanded by the
observed X-ray luminosity-temperature relation, especially when groups are
included [86].
The estimated 10
49
erg/L

correspond to a pre-heating of 0.1 keV

per particle, for a typical cluster M
ICM
/L
B
25 M

/L

. This is

>
10 times
lower than the 1 keV/particle pre-heating that some models require to
t the cluster L
X
T relation [13, 39, 84, 110, 125]. This estimate depends
somewhat on the gas density (hence environment and redshift) where/when
the energy is injected, because what matters is the entropy change induced
by the pre-heating, S = kT/n
2/3
e
[13, 20, 60]. Hence the required energy
decreases if it is injected at a lower gas density. Nevertheless, this extreme (1
keV/particle) requirement would be met only if virtually all the SN energy
were to go into increasing the thermal energy of the ICM. Such extreme pre-
heating requirement points toward an additional energy (entropy) source, such
as AGN energy injection [115, 125]. Note however that in powerful starbursts
most SNs explode inside hot bubbles made by previous SNs, thus reducing
radiative losses, and the feedback eciency may approach unity [53]. More
recently it has been suggested that pre-heating requirements may be relaxed
somewhat if the energy injection takes place at relatively low density, so as
to boost the entropy increase with less energy deposition [87]. For example,
pre-heating could take place within the laments, prior to the time when they
coalesce to form clusters. Indeed, if much of the star formation in cluster
ellipticals took place at z

>
3, it likely predates by a long time the assembly
of clusters. Further exploration of the metal enrichment connections to pre-
heating are found in [14, 38].
7 Clusters vs. Field at z = 0 and the Overall Metallicity
of the Universe
To what extent are clusters fair samples of the z 0 universe as a whole? In
many respects clusters look much dierent from the eld, e.g., in the morpho-
logical mix of galaxies, or in the star formation activity, which in clusters has
almost completely ceased while it is still going on in the eld. Yet, when we re-
strict ourselves to some global properties, clusters and eld are not so dierent.
For example, the baryon fraction of the universe is
b
/
m
0.16 0.02 [9],
which compares 0.15 as estimated for clusters [123] adopting h
70
= 1. This
202 A. Renzini
tells us that no appreciable baryon vs. dark matter segregation has taken place
at a cluster scale [123], a prediction that X-ray observations should be able
to check.
Even more interesting may be the case of the stellar mass over baryonic
mass in clusters and in the eld. For the eld, i.e., the local universe, the contri-
bution of stars to is estimated as

= 0.0035h
1
70
[44], or

= 0.0041h
1
70
from the 2dF K-band luminosity function [25], with a 15% uncertainty
(adopting a Salpeter IMF). The total baryon density is
b
= 0.039h
2
70
, as de-
rived from the Standard Big Bang nucleosynthesis (and conrmed by WMAP
[9]). This gives a global baryon to star conversion eciency

/
b
0.10h
70
,
i.e., over the whole cosmic time 10% of the baryons have been converted
and locked into stars. At the galaxy cluster level, the same eciency can be
measured directly, and following [123] one gets:
M

M
ICM
+M

1
9.3h
3/2
70
+ 1
0.1. (11)
One can safely conclude that the eciency of baryon to galax-
ies/stars conversion has been 10%, quite the same in the eld
as well as within rich clusters of galaxies. At this very basic level, the
environment seems to be irrelevant! Note that we may be living in a rather
special time, as the friction of baryons locked in stars must have evolved at
a dierent rate in clusters and in the eld, with clusters freezing at the 10%
level at a much higher redshift (z 3) compared to the eld (z

<
1).
Two interesting inferences can be drawn from this intriguing cluster-eld
similarity:
The metallicity of the present universe is 1/3 solar. The metal-
licity of the local universe has to be virtually identical to that measured in
clusters ( 1/3 solar), since star formation, hence the ensuing metal enrich-
ment, have proceeded to the same level of baryon consumption ( 10%). In an
analogy to clusters, a majority share of the metals now reside out of galaxies
in a warm/hot intergalactic medium (WHIM) containing the majority of the
baryons. Most baryons as well as most metals in the local universe remain
unaccounted for, but observational eorts are currently being made to detect
them [26, 80].
The thermal energy (temperature) of the local universe is about
the same as the pre-heating energy of clusters. Similar overall star for-
mation activities most likely result not only in similar metal productions but
also in similar energy depositions by galactic winds. Hence, the temperature
of the local IGM is likely to be kT 0.11 keV, whatever the physical na-
ture of the cluster pre-heating turns out to be. Again, attempts are currently
going on to detect this metal rich WHIM. The detection of OVI absorption
clouds, physically located within the Local Group [80] as well as at moderate
redshift [101], are important steps in this direction.
Metals in Clusters of Galaxies 203
At z 0 eld early-type galaxies (ETG) show very little, yet detectable
dierences with respect to their cluster analogues [12, 35], implying typical
ages only 1 Gyr younger than ETGs in clusters. Moreover, bulges appear
very similar to ellipticals in their integrated properties, such as the Mg
2

and fundamental plane relations [36, 59]. In the well-studied case of the Milky
Way bulge no trace of stars younger than halo-bulge globular clusters could
be found [127]. At z 1 old ETGs are also found in sizable numbers in the
general eld, while it appears that star formation may have been a little more
extended than in clusters [8, 22, 112].
Therefore, spheroids in the general eld appear almost as old as cluster
ellipticals, i.e., with the bulk of their stellar populations having formed at
z

>
23. Given this, it is estimated that at least 65% of the stellar mass is at
least 8 Gyr old, or formed at z > 1 [54]. With 50% of the stellar mass in
spheroids that formed

>
80% of their mass at z

>
23, one can conclude that

>
30% of the stellar mass we see today was already in place by z 3 [94].
This indirect estimate is 3 times higher than directly measured in the HDF-
N [32]. However, this latter result may be subject to cosmic variance given
the small size of the explored eld, and a value as high as 30% cannot be
excluded by current observations [42].
7.1 The Metallicity of the Universe at z = 3
With 30% of all stars having formed by z = 3, also 30% of the met-
als should have been formed before such an early epoch. I have argued that
the global metallicity of the present-day universe is 1/3 solar, hence, the
metallicity of the z = 3 universe should be 1/10 solar [94], because by that
redshift the universe has experienced only 1/3 of the cumulative star for-
mation all the way to the present. This simple argument supports the notion
of a prompt initial enrichment of the early universe. While the 10% solar
metallicity at z = 3 is a very straightforward estimate however, its direct
observational test is not so easy.
Figure 18 [82] shows that at z = 3 the universe had already developed to
become extremely inhomogeneous in chemical composition, with the metallic-
ity ranging from supersolar in the central regions of young/forming spheroids
and in QSOs likely hosted by them, down to 10
3
solar in the Ly forest.
Making the proper (mass-) average abundance of the heavy elements requires
to know the fractional mass of each baryonic component at z = 3 not an
easy task. Sometimes the Ly forest is considered as representative of the
global metallicity of the high-z universe, as it may ll most of the volume and
perhaps contain most of the baryons. But at best it may provide an estimate
of the volume-averaged metallicity, which is irrelevant. What matters is in
fact the mass-averaged metallicity, which I argue can be 100 times higher
at z 3 than the volume-averaged one. At this early time most metals are
likely to be locked into stars, in metal rich winds, and in shocked IGM which
has already diluted wind materials, and none of these components qualify as
204 A. Renzini
Log Physical Scale (pc)
Abundances at z = 3
2 4
4
M
e
t
a
l
l
i
c
i
t
y
3
2
1
0
QSOs
LBGs
DLAs
Lyforest
OVI Preheated
gas?
Young
Spheroids
1
6
Fig. 18. Summary of current knowledge of metal abundances at z 3. On the
vertical axis the logarithmic abundance relative to solar is reported. The horizontal
axis gives the typical linear dimensions of the structures for which direct abundance
measurements are available. This gure has been adapted from [82] by the inclusion
of the box for young spheroidals for which the estimate is indirect, as based on the
present day observed metallicity range and on the estimated redshift of formation.
The gure also includes the approximate location of the OVI absorbers [103], and
the hypothetical location of the intergalactic medium enriched and pre-heated by
early galactic winds
Ly absorbers. Instead, a metal rich WHIM may have been detected thanks
to its OVI absorption [103]. By and large, Ly absorbers are very poor tracers
of cosmic chemical evolution.
8 Clusters vs. the Chemical Evolution of the Milky Way
The Galactic bulge luminosity is 5.5 10
10
L
K,
[61] and L
BULGE
B

6 10
9
L
B,
respectively in the K and in the B band. If we take the cluster
empirical yield of metals ( 0.2 L
B
M

) as universal, it follows that the

Galactic bulge has produced M
Z
0.2L
BULGE
B
= 0.2 6 10
9
10
9
M

of
metals. Where are all these metals? One billion solar masses of metals should
not be easy to hide: part must be in the stars of the bulge itself, part must
have been ejected by winds. The stellar mass of the bulge follows from its K-
band mass to light ratio, M
BULGE

/L
K
= 1 [62, 127], and its luminosity, and
hence M
BULGE

10
10
M

. Its average metallicity is about solar or slightly

lower [72, 127], i.e. Z = 0.02, and therefore the bulge stars all together contain
2 10
8
M

of metals. Only 1/5 of the metals produced when the bulge

was actively star forming some 1113 Gyr ago are still in the bulge! Hence,
80%, or 10
9
M

were ejected into the surrounding space by an early wind.

Metals in Clusters of Galaxies 205
At the time of bulge formation, such 10
9
M

of metals ran into largely

pristine (Z = 0) material, experienced RT instabilities leading to chaotic
mixing, and establishing a distribution of metallicities in a largely inhomo-
geneous IGM surrounding the young Milky Way bulge. For example, this
enormous amount of metals was able to bring to a metallicity 1/10 solar (i.e.
Z = 0.002) about 5 10
11
M

of pristine material, several times the mass of

the yet to be formed Galactic disk. Therefore, it is likely that the Galactic
disk formed and grew out of such pre-enriched material, which provides a
quite natural solution to the classical G-Dwarf problem [98].
But there is another very intriguing aspect of chemical evolution that is
revealed by the comparison of the Milky Way to clusters of galaxies [88].
Both for the Galactic disk and for clusters one can estimate the empirical
metal yield, i.e., the ratio of the mass of metals to the mass of stars. Thus, in
the MW disk we have:
y
disk

(M

+ 0.2M

)
M

= 1.2Z

(12)
assuming that both stars and the ISM are solar metallicity on average and the
mass of the disk ISM is 20% the mass of the stars in the disk. In clusters
we have instead:
y
clusters

(M

+ 0.3 5 M

)
M

= 2.5Z

, (13)
where stars in clusters are again assumed to be solar metallicity, the ICM is
assumed 0.3 solar, and 5 times more massive than stars in galaxies. Thus, the
apparent yield of clusters is about twice that of the Galactic disk. Actually,
the dierence could be even larger, if one adopts a M
ICM
/M

ratio as high as
in [88].
Two opposite solutions of the discrepant yields are discussed in [88]:
Option A: The IMF in galaxy clusters is atter than in the Galactic disk,
hence with a top-heavy IMF more massive stars are produced, hence more
metals.
Option B: The IMF is the same in the disk as in clusters, but the discrepancy
arises from not having counted metals produced by the MW disk stars which
have been ejected by disk winds, i.e., the disk has lost metals (just like the
bulge).
Option A is favored by [88] based on two arguments: (1) if B were true then
most chemical evolution models of the MW galaxy would be wrong, and (2)
there is no evidence for star formation in the disk causing mass loss, but
material ejected in galactic fountains sooner or later would fall back to the
disk. I think that this choice is premature, but cannot be ruled out either.
It is certainly true that Option B would cause some problems to chemical
evolution models, as they usually rest on three assumptions that may not be
valid. Namely, (1) that disks started forming out of pristine (Z = 0) material,
206 A. Renzini
(2) that they grow by accumulating pristine (Z = 0) material, and (3) that
disks dont lose any mass.
In favor of Option A, the argument can be put forward that, even if indi-
vidual star-formation events follow a universal IMF, the resulting global IMF
depends on the distribution function of the mass of individual (star cluster
size) formation events [63, 120]. Hence galaxies, where most of star forma-
tion originated in powerful starbursts, will closely follow the universal IMF,
while quiescent star formation resulting from many small individual events
may have a somewhat steeper IMF. Therefore, it is not so inconceivable that
elliptical galaxies (which formed in powerful starbursts and which dominate
clusters) may have a atter IMF than the MW disk.
On the other hand, while the present SFR in the MW disk is very low
( 1 2M

yr
1
), it may have been much higher in the past. Actually, it has
been argued that most of the factor of 10 increase in the global SFR between
z = 0 and z 1 is due to an increase within disks [50, 65]. If so, just a few
Gyr ago the MW disk was forming stars much more violently than is currently
observed, and substantial ejection of metals from the disk is therefore not at
all inconceivable. By the same token, if most of the disk build-up was through
stronger bursts than observed today, then the global disk IMF may also be
close to that in ellipticals. Thus the choice between Option A and B is still
an open debate.
9 Summary
With these lectures I hope to have conveyed the feeling that the chemistry
of galaxy clusters is at the crossroads of many interesting astrophysical and
cosmological issues, and that we can learn a lot from their study. A number of
unexpected inferences are derived, starting from a few empirical facts, namely,
the iron and metal content of the ICM and cluster galaxies, the fraction of
the baryons locked into stars in clusters and in the eld, and the age and
baryon-fraction of stellar populations of galactic spheroids. Such inferences
include:
In clusters and in the general eld alike there are more metals in the gas that
has diused out of galaxies (ICM and IGM) than there are locked into stars
inside galaxies (only 10%). The loss of metals to the surrounding media is
therefore a major factor in the chemical evolution of galaxies. Furthermore, at
this global level, the outcome of star formation through cosmic time is largely
independent of environment, most likely just because a major fraction of all
stars formed before cluster formation.
Various arguments support the notion that the metals now in the ICM/IGM
were ejected by galactic winds, rather then being extracted from galaxies by
ram pressure.
Metals in Clusters of Galaxies 207
Having processed the same fraction of baryons into stars, the global metal-
licity of the local universe has to be nearly the same that one can measure in
clusters, i.e., 1/3 solar.
For the same reason, one expects the IGM to have experienced nearly the
same amount of pre-heating as the ICM, and therefore to be at a temperature
of 0.11 keV, regardless of the amount of pre-heating that is required for
clusters.
Given the predominance and formation redshift of galactic spheroids, both
in clusters as well as globally in the universe, it is likely that the universe
experienced a prompt metal enrichment, with the global metallicity possibly
reaching 1/10 solar already by z 3. Most metals remain unaccounted for
at low-redshift as well as at high-redshift, however, and are likely to reside
in a warm/hot IGM (WHIM) whose existence may have been revealed by
observational data.
This same scenario may be valid down to the scale of our own Milky Way
galaxy, with early winds from the forming Galactic bulge having pre-enriched
to 1/10 solar a much greater mass of gas, out of which the Galactic disk
started to form and evolve.
The empirical metal yield of clusters is at least twice that of the MW disk.
This signals that either the stellar IMF of the disk is a little steeper than that
of ellipticals, or that the disk has lost at least as many metals than it has
produced.
Acknowledgements
I would like to thank Manolis Plionis and Omar L opez-Cruz for having invited
me to Tonantzintla to give these lectures, and for their friendly hospitality.
I would also like to thank Laura Greggio for having modied her original
gures, specically for these proceedings.
References
1. Aguirre, A., Hernquist, L., Schaye, J., Katz, N., Weinberg, D.H., Gardner, J.:
ApJ 561, 521 (2001)
2. Anders, E., Grevesse, N.: Geochimica et Cosmochimica Acta 53, 197 (1989)
3. Arimoto, N., Matsushita, K., Ishimaru, Y., Ohashi, T., Renzini, A.: ApJ 477,
128 (1997)
4. Arnaboldi, M., et al.: AJ 125, 514 (2003)
5. Arnaud, M., Evrard, A.E.: MNRAS 305, 631 (1999)
6. Arnaud, M., Rothenug, R., Boulade,O., Vigroux, R., Vangioni-Flam, E.:
A&A 254, 49 (1992)
7. Baumgartner, W.H., Loewenstein M., Horner D.J., Mushotzsky, R.F.: ApJ
620, 680 (2005)
208 A. Renzini
8. Bell, E.F., Wolf, C., Meisenheimer, K., Rix, H.-W., Borch, A., Dye, S.,
Kleinheinrich, M., McIntosh, D.H.: ApJ 608, 752 (2004)
9. Bennett, C.L., et al.: ApJS 148, 1 (2003)
10. Benson, A.J., Frenk, C.S., Sharples, R.M.: ApJ 574, 104 (2002)
11. Bernardi, M., Renzini, A., da Costa, L.N., Wegener, G., Alonso, M.V.,
Pellegrini, P.S., Rite, C., Willmer, C.N.A.: ApJ 508, L43 (1998)
12. Bernardi, M., et al.: AJ 125, 1882 (2003)
13. Borgani, S., Governato, F., Wadsley, J., Menci, N., Tozzi, P., Lake, G.,
Quinn, T., Stadel, J.: ApJ 559, L71 (2001)
14. Bohringer, H., Matsushita, K., Churazov, E., Finoguenov, A., Ikebe, Y.: A&A
416, L21 (2004)
15. Bohringer, H., Matsushita, K., Finoguenov, A., Xue, Y., Churazov, E.: AdSpR
36, 677 (2005)
16. Bower, R.G., Lucey, J.R., Ellis, R.S.: MNRAS 254, 613 (1992)
17. Buote, D.A.: MNRAS 311, 176 (2000)
18. Cappellari, M., Bacon R., Bureau, M., Damen, .C., et al.: MNRAS 366, 1126
(2006)
19. Cappellaro, E., Evans, R., Turatto, M.: A&A 351, 459 (1999)
20. Cavaliere, A., Colafrancesco, S., Menci, N. ApJ 415, 50 (1993)
21. Chiosi, C.: A&A 364, 423 (2000)
22. Cimatti, A., et al.: A&A 381, L68 (2002)
23. Ciotti, L., DErcole, A., Pellegrini, S., Renzini, A.: ApJ 376, 380 (1991)
24. Ciotti, L., Ostriker, J.P.: ApJ 551, 131 (2001)
25. Cole, S., et al.: MNRAS 326, 255 (2001)
26. Danforth, C.W., Shull, J.M.: ApJ 62, 555 (2005)
27. David, L.P., Jones, C., Forman, W, Daines S.: ApJ 428, 544 (1994)
28. Davis, D.S., Mulchaey, J.S., Mushotzky, R.F.: ApJ 511, 34 (1999)
29. De Grandi, S., Ettori, S., Longhetti, M., Molendi, S.: A&A 419, 7 (2004)
30. De Grandi, S., Molendi, S.: ApJ 551, 153 (2001)
31. De Grandi, S., Molendi, S.: ASP Conf. Ser. 253, 3 (2002)
32. Dickinson, M., Papovich, C., Ferguson, H., Budavari, T.: ApJ 587, 25 (2003)
33. Dressler, A., Gunn, J. E.: ASP. Conf. Ser. 10, 204 (1990)
34. Dupke, R.A., White, R.E. III: ApJ 537, 123 (2000)
35. Eisenstein, D.J., Hogg, D.W., Fukugita, M., Nakamura, O., Bernardi, M.,
et al.: ApJ 585, 694 (2003)
36. Falcon-Barroso, J., Peletier, R.F., Balcells, M.: MNRAS 335, 741 (2002)
37. Ferguson, H.C., Tanvir, N.R., von Hippel, T.: Nature 391, 461 (1998)
38. Finoguenov, A., Arnaud, M., David, L.P.: ApJ 555, 191 (2001)
39. Finoguenov, A., Borgani, S., Tornatore, L., Bohringer, H.: A&A 398, L35
(2003)
40. Finoguenov, A., David, L.P., Ponman, T.J.: ApJ 544, 188 (2000)
41. Finoguenov, A., Jones, C., Bohringer, H., Ponman, T.J.: SpJ 578, 74 (2002)
42. Fontana, A., Pozzetti, L., Donnarumma, I., Renzini, A., Cimatti, A., et al.:
A&A 424, 23 (2004)
43. Fukazawa, Y, Ohashi, T., Fabian, A.C., Canizares, C.R., Ikebe, Y.,
Makishima, K., Mushotzky, R.F., Yamashita, K.: PASJ 46, L55 (1994)
44. Fukugita, M., Hogan, C.J., Peebles, P.J.E.: ApJ 503, 518 (1998)
45. Gal-Yam, A., et al.: AJ 125, 1087 (2003)
46. Gastaldello, F., Molendi, F.: ApJ 572, 160 (2002)
Metals in Clusters of Galaxies 209
47. Gibson, B., Matteucci, F.: MNRAS 291, L8 (1997)
48. Greggio. L.: ASP Conf. Ser. 342, 459 (astro-ph/0410187) (2005a)
49. Greggio. L.: A&A 441, 1055 (2005b)
50. Hammer, F., Flores, H., Elbaz, D., Zheng, X.Z., Liang, Y.C., et al.: A&A 430,
115 (2005)
51. Hamuy, M.: ApJ 582, 905 (2003)
52. Heckman, T.M.: Asp. Conf. Ser. 254, 292 (2002)
53. Heckman, T.M., Lehnert, M.D., Strickland, D.K., Armus, L.: ApJS 129, 493
(2000)
54. Hogg, D.W., et al.: AJ 124, 646 (2002)
55. Iben, I. Jr., Renzini, A.: ARA&A 21, 271 (1983)
56. Iben, I. Jr., Tutukov A.V.: ApJS 54, 335 (1984)
57. Ishimaru, Y., Arimoto, N.: PASJ 49, 1 (1997)
58. Iwamoto, K., Brachwitz, F., Nomoto, K., Kishimoto, N. et al.: ApJS 125, 439
(1999)
59. Jablonka, P., Martin, P., Arimoto, N.: AJ 112, 1415 (1996)
60. Kaiser, N.: ApJ 383, 104 (1991)
61. Kent, S.M.: ApJ 387, 181 (1994)
62. Kent, S.M., Dame, T.M., Fazio, G.: ApJ 378, 131 (1991)
63. Kroupa, P., Weidner, C.: ApJ 598, 1076 (2003)
64. Larson, R.B., Dinerstein, H.L.: PASP 87, 511 (1975)
65. Lilly, S.J., Tresse, L., Hammer, F., Crampton, D., Le Fevre, O.: ApJ 455,
108 (1995)
66. Livio, M.: In: Niemeyer, J.C., Truran, J.W. (eds.) The Progenitors of Type Ia
Supernovae, p. 33. CUP, Cambridge (2000)
67. Loewenstein, M.: ApJ 557, 573 (2001)
68. Loewenstein, M., Mushotzky, R.F.: ApJ 466, 695 (1996)
69. Maraston, C., Greggio, L., Renzini, A., Ortolani, S., Saglia, R. P., Puzia, T.H.,
Kissler-Patig, M.: A&A 400, 823 (2003)
70. Matsushita, K., Finoguenov, A., Bohringer, H.: A&A 401, 443 (2003)
71. Matteucci, F., Vettolani, G.: A&A 202, 21 (1988)
72. McWilliam, A., Rich, R.M.: ApJS 91, 794 (1994)
73. Mitchell, R.J., Culhane, J.L., Davison, P.J., Ives, J.C.: MNRAS 175, 29 (1976)
74. Mohr, J.J., Mathiesen, B., Evrard, A.E.: ApJ 627, 649 (1999)
75. Mulchaey, J.S., Davis, D.S., Mushotsky, R.F., Burnstein, D.: ApJ 404, L9
(1993)
76. Mulchaey, J.S., Davis, D.S., Mushotsky, R.F., Burnstein, D.: ApJ 456, 80
(1996)
77. Mushotzky, R.F.: In: Durret, F., Mazure, A., Tran Thanh Van, J. (eds.) Clus-
ters of Galaxies. p. 167. Editions Frontieres, Gyf-sur-Yvette (1994)
78. Mushotsky, R.: Phil. Trans. R. Soc. Lond. A 360, 2019 (2002)
79. Mushotzky, R.F., et al.: ApJ 466, 686 (1996)
80. Nicastro, F., Zezas, A., Elvis, M., Mathur, S., Fiore, F., Cecchi-Pestellini, C.,
Burke, D., Drake, J., Casella, P.: Nature 421, 719 (2003)
81. Patat, F., Barbon, R., Cappellaro, E., Turatto, M.: A&A 282, 731 (1994)
82. Pettini, M., In Cosmochemistry: The melting pot of the elements, eds.
Esteban, C. et al., Cambridge Contemporary Astrophysics, Cambridge. Univ.
Press (2004)
83. Pettini, M., Shapley, A.E. Steidel, C.C., Cuby, J.-G., Dickinson, M.,
Moorwood, A.F.M., Adelberger, K.L., Giavalisco, M.: ApJ 554, 981 (2001)
210 A. Renzini
84. Pipino, A., Matteucci, F., Borgani, S., Biviano, A.: New Astr. 7, 227 (2002)
85. Ponman, T.J., et al.: Nature 369, 462 (1994)
86. Ponman, T.J., Cannon, D.G., Navarro, J.F.: Nature 397, 135 (1999)
87. Ponman, T.J., Sanderson, A.J.R., Finoguenov, A.: MNRAS 343, 331 (2003)
88. Portinari, L., et al.: ApJ 604, 579 (2004)
89. Pratt, G.W., Arnaud, M.: A&A 408, 1 (2003)
90. Raymond, J.C., Smith, B.W.: ApJS 35, 419 (1977)
91. Renzini, A., Ciotti, L., DErcole, A., Pellegrini, S.: ApJ 419, 52 (1993)
92. Renzini, A.: In: Durret, F. et al. (eds.) Clusters of Galaxies, p. 221. Edition
Frontieres, Gyf-sur-Yvette (1994)
93. Renzini, A.: ApJ 488, 35 (1997)
94. Renzini, A.: ASP Conf. Ser. 146, 298 (1998a)
95. Renzini, A.: AJ 115, 2459 (1998b)
96. Renzini, A.: In: Carollo, C.M., et al. (eds.) The Formation of Galactic Bulges,
p. 9. CUP, Cambridge (1999)
97. Renzini, A.: In: Plionis, M., Georgantopoulos, I. (eds.) Large Scale Structure
in the X-ray Universe, p. 103. Atlantisciences, Paris (2000)
98. Renzini, A.: ASP Conf. Ser. 253, 331 (2002)
99. Renzini, A.: In: Mulchaey, J.S., Dressler, A., Oemler, A. (eds.) Clusters of
Galaxies: Probes of Cosmological Structures and Galaxy Evolution, p. 261.
CUP, Cambridge (2004)
100. Renzini, A.: In: Corbelli, E., Palla, F., Zinnecker, H. (eds.) The Initial Mass
Function 50 Years Later, p. 221. Springer, Berlin (2005)
101. Richter, P., Savage, B.D., Tripp, T.M., Sembach, K.R.: ApJS 153, 165 (2004)
102. Schindler, S., et al.: A&A 435, L25 (2005)
103. Simcoe, R.A., Sargent, W.L.W., Rauch M.: ApJ 578, 737 (2002)
104. Songaila, A., Cowe, L.L., Lilly, S.J.: ApJ 348, 371 (1990)
105. Stanford, S.A., Eisenhardt, P.R., Dickinson, M.: ApJ 492, 461 (1998)
106. Steidel, C.C., et al.: ApJ 604, 534 (2004)
107. Strolger, L.-G., et al.: ApJ 613, 200 (2004)
108. Tamura, T, Kaastra, J.S., Makishima, K, Takahashi, I.: A&A 399, 497 (2004)
109. Thomas, D.: In: Redshift, ed. Walsh, J.R., Rosa, M.R. (eds.) Chemical Evo-
lution from Zero to High, p.197. Springer, Berlin (1999)
110. Tozzi, P., Norman, C.: ApJ 546, 63 (2001)
111. Tozzi, P., Rosati, P., Ettori, S., Borgani, S., Mainieri, V., Norman, C.: ApJ
503, 705 (2003)
112. Treu, T., Ellis, R.S., Liao, T.X., van Dokkum, P.G., Tozzi, P., et al.: ApJ 633,
174 (2005)
113. Tsuru, T.: PhD Thesis, University of Tokyo (1992)
114. Turatto, M., LNP, 598, p. 21 (Springer)
115. Valageas, P., Silk, J.: A&A 350, 725 (1999)
116. van der Marel, R.: MNRAS 253, 710 (1991)
117. van Dokkum, P.G., Stanford, S.S.: ApJ 585, 78 (2003)
118. Veilleux, S., Cecil, G., Bland-Hawthorn, J.: ARA&A 43, 769 (2005)
119. Vigroux, L.: A&A 56, 473 (1977)
120. Weidner, C., Kroupa, P.: MNRAS 365, 1333 (2006)
121. Whelan, J., Iben, I. Jr.: ApJ 186, 1007 (1976)
122. White, D.A.: MNRAS 312, 663 (2000)
123. White, S.D.M., Navarro, J.F., Evrard, A.E., Frenk, C.S.: Nature 366, 429
(1993)
Metals in Clusters of Galaxies 211
124. Woosley, S.E., Weaver, T.A.: ApJS 101, 181 (1995)
125. Wu, K.K.S., Fabian, A.C., Nulsen, P.E.J.: MNRAS 318, 889 (2000)
126. Yungelson, L.R., Livio, M.: ApJ 528, 108 (2000)
127. Zoccali, M., Renzini, A., Ortolani, S., Greggio, L., Saviane, I., Cassisi, S.,
Rejkuba, M., Barbuy, B., Rich, R. M., Bica, E.: A&A 399, 931 (2003)