Gas Dynamics in Clusters of Galaxies

C. L. Sarazin
Department of Astronomy, University of Virginia, P. O. Box 3818, Charlottesville,
VA 22903-0818, USA
[email protected]
1 Introduction
One of the more surprising results from X-ray astronomy is that the great
volumes of space between galaxies in clusters of galaxies are not empty, as
they appear in optical images. Instead, they are lled with a diuse, hot
plasma, with typical temperatures of T 10
7
10
8
K. At this temperature,
the sound speed in the gas is comparable to the orbit velocities of the galaxies
in the cluster, which is consistent with the gas being in hydrostatic equilib-
rium with the same gravitational potential as binds the galaxies. This intr-
acluster medium (ICM) is highly rareed, with electron number densities of
n
e
10
4
10
2
cm
3
. At least on large scales, the gas is stably stratied,
with the density decreasing with increasing radius r. The gas extends out to
distances of r

> Mpc from the cluster center. The total mass of hot gas is
typically M
gas
10
14
M

; this mass exceeds the total mass of all the galaxies

in a typical rich clusters, although even more of the mass is in the form of
unseen dark matter.
At temperatures of 10
6
10
8
K, the dominant radiation mechanism of a
plasma is X-ray emission. As a result, clusters of galaxies are generally very
luminous X-ray emitters, with luminosities of L
X
10
43
10
45
ergs s
1
.
Clusters are second only to quasars as the most luminous X-ray sources in the
Universe, and are the most luminous extended sources. While X-ray emission
is the primary observational diagnostic for the intracluster medium, the ICM
has a number of other important physical eects. It connes and distorts radio
galaxies within the cluster. The cosmic ray and magnetic eld components of
the intracluster medium can also produce diuse radio emission (see Feretti &
Giovannini this volume). The ICM can strip interstellar gas from galaxies as
they move through the cluster. Intracluster gas cools at the centers of many
clusters, producing lower temperature gas. If the ICM contains dust, the dust
will be strongly heated by the plasma, and may emit strongly in the infrared.
The ICM also has a number of opacity eects; for example, it scatters and
C. L. Sarazin: Gas Dynamics in Clusters of Galaxies, Lect. Notes Phys. 740, 130 (2008)
DOI 10.1007/978-1-4020-6941-3 1 c Springer Science+Business Media B.V. 2008
2 C. L. Sarazin
heats the cosmic background radiation which passes through it. The magnetic
eld in the ICM leads to Faraday rotation and depolarization (Sect. 4.2).
In this chapter, I will review the physical state and X-ray emission pro-
cesses of the ICM (Sect. 2 and 3). The ICM is shown to act as a uid in
Sect. 4. In Sect. 5, the transport properties of the gas, particularly thermal
conduction, are discussed. The hydrodynamical equations for the ICM are
given in Sect. 6. Models for the distribution of the gas, and the use of the gas
to determine the total mass distributions of clusters are described in Sect. 7
and Sect. 8. The heating and cooling processes in the ICM are discussed in
Sect. 9. Much of the heating is due to the hierarchical formation of clusters,
and cluster mergers are introduced in Sect. 10. The thermal eects of merger
shocks are discussed in Sect. 11. In Sect. 12, the eects of mergers on cluster
cooling cores and the phenomena of cold fronts are described.
As much as possible, comparisons to observations in this chapter assume
the standard WMAP cosmology [2], with a Hubble constant of H
0
= 71 km
s
1
Mpc
1
, a ratio of the mass density to the critical density of
m
= 0.27,
and the ratio of the dark energy density to the critical density of

= 0.73.
2 Physical State of the Intracluster Gas
2.1 Local Thermal State
At the very high temperatures of the intracluster gas, the gas is very highly
ionized, but not completely so for the heavy elements. Thus, to describe the
local thermal state of the gas, we need to specify three things. First, there are
the motions of free particles (electrons and ions), or the kinetic state of the
gas. Then, we need to give the ratios of electrons which are free to those which
are bound to ions, or the ionization state of the gas. Finally, for the bound
electrons, we need to which energy levels they occupy; this is the excitation
state of the gas.
Kinetic Equilibrium
If Coulomb collisions are suciently rapid, the free particles in the gas (free
electron, free proton, and ions) will be brought into kinetic equilibrium and
develop a Maxwellian distribution. The time scale for a particle of mass m
1
and charge Z
1
e to collide with eld particles of mass m
2
and charge Z
2
e with a
number density of n
2
in a Maxwellian distribution at a temperature T is [43]:
t
eq
(1, 2) =
3m
1

2 (kT)
3/2
8

m
2
n
2
Z
2
1
Z
2
2
e
4
ln
. (1)
Here, ln ln(b
max
/b
min
) 40 is the Coulomb logarithm, and b
min
and b
max
are the minimum and maximum impact parameters for Coulomb collisions
Gas Dynamics in Clusters of Galaxies 3
in the gas. The Coulomb logarithm has a weak (logarithmic) dependence
on density and temperature, but is nearly constant under ICM conditions.
Coulomb collisions between electrons will bring the electrons into equilibration
(an isotropic Maxwellian velocity distribution) on a time scale of roughly
t
eq
(e, e) 3 10
5
yr
_
T
10
8
K
_
3/2
_
n
e
10
3
cm
3
_
1
yr . (2)
The time scale for protons to equilibrate among themselves is t
eq
(p, p)
(m
p
/m
e
)
1/2
t
eq
(e, e), or roughly 43 times longer than the value in (2). Follow-
ing this time, the protons and ions would each have Maxwellian distributions,
but generally at dierent temperatures. The time scale for protons to col-
lide with electrons and exchange energy is t
eq
(p, e) (m
p
/m
e
)t
eq
(e, e), or
roughly 1870 times the value in (2). The time scale for the electrons and pro-
tons to come into equipartition (equal temperatures) is roughly one half of
t
eq
(p, e) [43].
Under typical conditions in the intracluster gas, these time scales are
t
eq
(e, e) 10
5
yr, t
eq
(p, p) 4 10
6
yr, and t
eq
(p, e) 2 10
8
yr. Most
clusters have existed for

> 10
9
yr, so one would expect the gas to generally be
in kinetic equilibrium, with the distributions of free particles being isotropic
Maxwellians. Moreover, the electrons and ions should generally be in equipar-
tition, with a common kinetic temperature T = T
e
= T
p
. Possible exceptions
might be the outermost regions of clusters (where the gas density is low), or
regions where the gas properties have changed rapidly, such as shocks [26].
Collisional Ionization Equilibrium
The main ionization process in the intracluster gas is collisional ionization,
e

+X
+i
e

+e

+X
+i+1
. (3)
The main recombination processes are radiative and dielectronic recombina-
tion,
e

+X
+i+1
X
+i
+ photon(s) . (4)
Here, X
+i
represents some element X which has been ionized i times. Note
that neither radiative nor dielectronic recombination (4) are the inverse of
collisional ionization (3), which implies that the ionization state in the intra-
cluster gas is not that in thermodynamic equilibrium (the Saha equation).
Let C(X
i
, T) be the rate coecient for collisional ionization out of the ion
X
i
(3), while (X
i
, T) is the rate coecient for recombination to the ion X
i
(4). If the gas starts in a lower ionization state than in equilibrium, it will be
ionized up towards equilibrium on a time scale of roughly:
t
ion

_
C(X
i
, T)n
e

1
3 10
8
_
C(X
i
, T)
10
13
cm
3
s
1
_
1
_
n
e
10
3
cm
3
_
1
yr .
(5)
4 C. L. Sarazin
In general, the collisional ionization rates are high enough that one would
expect he ICM to generally be in collisional ionization equilibrium. Again,
possible exceptions might be the outermost regions of clusters (where the gas
density is low), or regions where the gas properties have changed rapidly, such
as shocks.
In collisional ionization equilibrium, the rates of collisional ionization and
radiative and dielectronic recombination balance, which implies that
_
C(X
i
, T
g
) +(X
i1
, T
g
)

n(X
i
) = C(X
i1
, T
g
)n(X
i1
)+(X
i
, T
g
)n(X
i+1
) .
(6)
Here, n(X
i
) is the number density of the X
i
ion. Note that, unlike thermo-
dynamic equilibrium (Saha equilibrium), the state of ionization in collisional
ionization equilibrium is independent of density, and only depends on the
electron kinetic temperature T. Generally, each ionization fraction reaches a
maximum at a temperature that is some fraction of its ionization potential.
At the temperatures which predominate in clusters, iron is mainly in the fully
stripped, hydrogenic, or heliumlike stages.
Excitation Equilibrium
For ions with bound electrons, the population of excited states are determined
mainly by a balance between collision excitation by free electrons and radia-
tive de-excitation. In general, the spontaneous radiative de-excitation rates
are much higher than the excitation rates, and the electrons are almost al-
ways found in the ground level. The population of excited states are much
lower than would be expected in thermodynamic equilibrium (Boltzmann dis-
tribution). Collisional de-excitation rates are much lower than radiative de-
excitation rates; this means that there are no X-ray spectral diagnostics which
determine the local density in the gas.
3 X-ray Emission
The X-ray emission of the intracluster gas is mainly due to thermal
bremsstrahlung and line emission. There are smaller contributions of contin-
uum from bound-free (recombination) emission and from two-photon decays
of 2s levels in hydrogenic and helium-like ions.
The emissivity due to thermal bremsstrahlung (freefree emission) is given
by

=
2
5
e
6
3m
e
c
3
_
2
3m
e
k
_
1/2
n
e
T
1/2
exp(h/kT)

i
Z
2
i
n
i
g

(Z
i
, T, ) , (7)
where the emissivity

is dened as the emitted energy per unit time, fre-

quency, and volume. The sum is over the various ions in the plasma, but is
Gas Dynamics in Clusters of Galaxies 5
dominated by hydrogen and helium for Solar abundances. The Gaunt factor
g

(Z
i
, T, ) corrects for quantum mechanical eects and for the eect of dis-
tant collisions, and is a slowly varying function of frequency and temperature.
As a result, the dominant dependence of the freefree emissivity on frequency
is the Boltzmann exponential factor, and the main dependences on tempera-
ture are this factor and the squareroot factor T
1/2
. Thermal bremsstrahlung
produces a roughly exponential continuum component in the X-ray spectrum.
At high temperatures T

> 310
7
K, thermal bremsstrahlung is the dominant
emission mechanism.
At lower temperatures, the main X-ray radiation is from lines. The
strongest line feature observed from most clusters of galaxies is the complex of
iron Fe K lines at about 6.7 keV. This line feature is actually a blend of lines
from iron ions (mainly Fe
+24
and Fe
+25
) and weaker lines from nickel ions.
The notation K gives the principal quantum number n of the lower level of
the transition and the change in the principal quantum number n n

n,
where n

is the principal quantum number of the upper level of the transition.

K indicates that the lower level is in the K-shell (n = 1), L indicates the lower
level is in the L-shell (n = 2), and so on, while indicates that n = 1,
indicates that n = 2, etc. In addition to the Fe K line complex, the X-ray
spectra of clusters of galaxies contain a large number of lower energy lines.
These include the K lines of the common elements lighter than iron, such as
C, N, O, Ne, Mg, Si, S, Ar, and Ca, as well as the L lines of Fe and Ni. These
lines become very strong at lower temperatures (T

< 10
7
K).
As an illustration, Fig. 1 shows the predicted X-ray spectrum of an X-
ray cluster [53]. The model cluster is isothermal in its outer regions (with a
temperature of 8 10
7
K), and has a cooling core at its center. The gure
shows the overall exponential continuum from thermal bremsstrahlung, the
Fe K lines at about 7 keV (which come mainly from the region of the cluster
outside of the cooling ow), and the lower energy lines from the cooling core.
Most X-ray lines are excited by collisional excitation by electrons, although
radiative and dielectronic recombination and inner shell collisional ionization
also play a role. The emissivity due to a collisionally excited line is usually
written [36]:
_

line

d = n(X
i
)n
e
h
3
(T)B
4
gs
(X
i
)
_
2

3
m
3
e
kT
_
1/2
e
E/kT
, (8)
where h is the energy of the transition, E is the excitation energy above
the ground state of the excited level, B is the branching ratio for the line (the
probability that the upper state decays through this transition), and is the
collision strength, which is often a slowly varying function of temperature.
The intracluster gas is almost certainly in collisional ionization equilibrium
(Sect. 2.1); under these circumstances, the ionization fractions depend only
on the electron temperature T, and are independent of the density of the gas.
Then, the density of any ion is just proportional to the proton density in the
6 C. L. Sarazin
.1 .2
10
43
10
44
10
45
10
46
.5 1 2
E (keV)
d
L
/
d
E

(
e
r
g
s

s

1

k
e
V

1
)
Fe L
Fe K
0 K
5 10
Fig. 1. Model X-ray spectrum of a cluster of galaxies. The cluster was assumed to
be isothermal at T = 8 10
7
K in its outer regions, and to have a large amount of
cool gas (a cooling core) in its inner regions
gas times the abundance of the relevant element relative to hydrogen. Thus,
all of the X-ray emission processes in the gas scale with the product n
p
n
e
of
the proton and electron densities, respectively. If L

is the X-ray luminosity

per unit frequency emitted by a cluster, then this can be written as
L

(T, Abundances)
_
n
e
n
p
dV , (9)
where the integral is over the volume V of the cluster. The total emissivity

n
e
n
p
, where

depends only on the temperature and the abundances of the

heavier elements relative to hydrogen. Similarly, the X-ray surface brightness
is given by
I

(T, Abundances)
_
n
e
n
p
dl . (10)
Here, the integral is along the line of sight distance l through the cluster.
The emissivity of a line is then proportional to the square of the density
and to the abundance of the relevant element, and depends signicantly on
the electron temperature. Because the thermal bremsstrahlung emissivity also
is proportional to the square of the density (7), the ratio of line emission to
thermal bremsstrahlung continuum emission is independent of density. Line
ratios or the shape of the X-ray continuum spectrum can be used to derive a
temperature for the gas in a cluster. Then, the ratio of line emission to thermal
bremsstrahlung continuum emission can be used to determine the abundance
of the heavy element responsible for the line.
Gas Dynamics in Clusters of Galaxies 7
4 The Intracluster Medium as a Fluid
4.1 Mean Free Paths
The mean free paths of electrons and ions in a plasma without a magnetic
eld are determined by Coulomb collisions [43]. The electrons in a Maxwellian
plasma undergo Coulomb collisions in a time which is a factor of
_
m
e
/m
p
shorter than the protons (Sect. 2.1). On the other hand, the electrons move
faster by the inverse of this factor. Thus, the mean free paths of electrons and
protons are essentially equal, with

p
=
e
=
3
3/2
(kT)
2
4
1/2
n
e
e
4
ln
23
_
T
10
8
K
_
2
_
n
e
10
3
cm
3
_
1
kpc . (11)
These mean free paths are smaller than most scales of interest in clusters; they
are only about 1% of the radius of a cluster (2 Mpc). Thus, it is reasonable to
treat the ICM as a uid under most circumstances. The uid approximation
might breakdown in the outer parts of a cluster (where the lower density
increases
e
), in interactions with galaxies (whose sizes are comparable to

e
), if the ICM is very inhomogeneous, or in sharp transitions in the ICM
properties at shocks or cold fronts (Sects. 11 and 12.2).
4.2 Magnetic Fields and Gyroradii
In any case, the ICM apparently contains a signicant magnetic eld, with
typical values of B 1 G. (See the chapter by Feretti & Giovannini for
more details concerning the magnetic eld in clusters.) Stronger elds occur
in some smaller volumes of clusters. These elds are probably too weak to be
very important dynamically, as the magnetic pressure, P
B
= B
2
/(8), is much
smaller than the typical gas pressures. However, the magnetic eld does very
strongly eect the microscopic motions of electrons and ions. In the presence
of a magnetic eld, electron and ions follow helical orbits, gyrating about
magnetic eld lines. The gyroradii of electrons and ions in cluster magnetic
elds are very small. For example, the gyroradius of a typical electron is
r
g
3 10
8
_
T
10
8
K
_
1/2
_
B
1 G
_
1
cm . (12)
These very small gyroradii probably insure that the ICM acts as a uid even
when the Coulomb mean free paths are long.
5 Transport Processes
The fact that the mean free paths are small but nite implies that the local
properties of the gas will be inuenced by the properties of the surrounding
gas through diusive processes, also called transport processes. These include
8 C. L. Sarazin
the thermal conduction of heat energy in non-isothermal gases, the viscous
transport of momentum, and the diusion and settling of heavy elements
within the intracluster gas. I will concentrate on thermal conduction here;
viscosity and diusion are discussed in [40].
5.1 Thermal Conduction
In a plasma with a gradient in the electron temperature, heat is conducted
down the temperature gradient. If the scale length of the temperature gradient
l
T
T/[T[ is much longer than the mean free path of electrons, l
T

e
,
then the heat ux is given by
Q = T , (13)
where the thermal conductivity for a hydrogen plasma is [43]:
= 1.31n
e

e
k
_
kT
e
m
e
_
1/2
4.6 10
13
_
T
e
10
8
K
_
5/2
_
ln
40
_
1
erg cm
1
s
1
K
1
. (14)
Because of the inverse dependence on the particle mass, thermal conduction
is primarily due to electrons. If the very weak dependence of ln on density
is ignored, then is independent of density but depends very strongly on
temperature.
If heat conduction operates at this Spitzer rate, then the gas in the cen-
tral regions of clusters is likely to be isothermal. In addition, heat conduction
would be very important at and would tend to eliminate any large local tem-
perature gradients, such as appear to occur in the cooling core of clusters or
near cold fronts (Sect. 12). On the other hand, the rate of thermal conduction
along a thermal gradient perpendicular to the magnetic eld is very low, as
a result of the small gyroradii of electrons (12). Thus, transverse or tangled
magnetic elds may be able to suppress thermal conduction in clusters, at
least in some regions. The existence of very steep temperature gradients in
cold fronts has been used to argue that heat conduction is suppressed by a
factor of

>10
2
in these regions (Sect. 12.2).
6 Hydrodynamics
In the uid limit, the ICM can be characterized by the local values of the gas
density , the gas pressure P, the gas temperature T or internal energy, and
the gas velocity v. The gas pressure is determined by the ideal gas law:
P =
kT
m
p
, (15)
Gas Dynamics in Clusters of Galaxies 9
where is the mean mass per particle in terms of the mass of a proton m
p
.
The dynamical equation for a single component uid is [20]:

Dv
Dt
+P + = 0 , (16)
where is the gravitational potential, and D/Dt is the Lagrangian derivative
with respect to time. Equation (16) ignores non-gravitational forces, such as
magnetic stresses or viscosity. The continuity equation (mass conservation)
is [20]:

t
+ (v) = 0 . (17)
There is also an equation giving the variation in the energy in the uid.
However, it is simpler to give this equation in terms of the entropy in the
gas, S. A useful quantity to consider is the specic entropy per particle in the
gas, s S/N, where N is the total number of particles. To within additive
constants, the specic entropy of an ideal gas is
s =
3
2
k ln
_
P

5/3
_
=
3
2
k ln
_
T

2/3
_
. (18)
To avoid the logarithmic character of the entropy, it is conventional to dene
an entropy parameter K as
K
kT
(n
e
)
2/3
(19)
with units of keV cm
2
. Thus, s ln K. Then, the equation for the change in
the gas entropy can be written [20]:

m
p
k
Ds
Dt
= HL , (20)
where H and L are the rate of heating and cooling per unit volume in the gas.
In the absence of irreversible processes like heating or cooling or shocks, the
specic entropy of a parcel of gas is constant.
7 Hydrostatic Equilibrium
Unless it is disturbed in some way, one would expect the gas in a cluster to
relax into hydrostatic equilibrium on roughly the sound crossing time of the
cluster,
t
s

D
c
s
6.6 10
8
yr
_
T
10
8
K
_
1/2
_
D
1 Mpc
_
. (21)
Here, D is the diameter of the cluster, and c
s
is the sound speed. Since this
time scale is shorter than the age of a typical cluster, which is a fraction
10 C. L. Sarazin
of the Hubble time, the gas in many clusters should be close to hydrostatic
equilibrium. Exceptions would include clusters which are undergoing or have
recently undergone a major merger, and regions of a cluster where an AGN
has injected energy recently.
In hydrostatic equilibrium, the pressure forces balance gravity:
P = ,
1

dP
dr
=
GM(r)
r
2
, (22)
where M(r) is the total cluster mass within r, and the second form assumes
spherical symmetry. Because (22) gives a single relation for two gas properties
(density and pressure), one must also specify the entropy distribution of the
gas to determine its distribution.
7.1 Isothermal Models
A very simple model follows if the gas is assumed to be isothermal (T =
constant); isothermality might result if thermal conduction were ecient in
the cluster (Sect. 5.1). Then, the solution of the hydrostatic equation is
ln
_
(r)

0
_
=
m
p
kT
[
0
(r)] , (23)
where
0
and
0
are the central values of the the gas density and gravitational
potential, respectively. Note that the gas density will generally go to a nite
value as r .
Numerical simulations suggest that the dark matter distribution in clusters
should have a power-lawdrop o at large radii, and a atter power-lawat small
radii [35]. Thus, the dark matter distribution should have a cusp at the center
of the cluster. The NFW dark matter prole [35] has:

DM
(r) =
s
_
_
r
r
s
__
1 +
r
r
s
_
2
_
1
, (24)
where r
s
and
s
are the characteristic scaling radius and density, respectively.
If this distribution applies to the sum of all the matter in a cluster, then the
potential is
(r) =
0
ln (1 +r/r
s
)
r/r
s
, (25)
and the central potential is
0
= 4G
s
r
2
s
.
However, in the past the dark matter and/or galaxy distributions in clus-
ters were modeled using a function with a constant density core,

DM
(r) =
DM,0
_
1 +
_
r
r
c
_
2
_
3/2
, (26)
Gas Dynamics in Clusters of Galaxies 11
where
DM,0
is the central density and r
c
is the core radius. If this form is
assumed for the total matter density in a cluster, or if it applies to the galaxy
distribution, and the galaxies have an isotropic gaussian velocity distribution,
then the resulting gas density distribution is the beta model [4]:
(r) =
o
_
1 +
_
r
r
c
_
2
_
3/2
. (27)
If the gas is isothermal, then this density distribution gives an X-ray surface
brightness distribution of the form
I
X
(r) = I
o
X
_
1 +
_
r
r
c
_
2
_
3+1/2
. (28)
This beta-model provides a reasonable t to the X-ray surface brightness in
the outer regions of many cluster, with a typical value of 2/3. However,
it does not t the inner parts of cooling core clusters.
7.2 Adiabatic or Polytropic Models
The temperature proles in clusters of galaxies are generally more consistent
with a gradual decline with radius at large radii, rather than isothermal gas
[48]. A simple alternative would be if the gas in clusters was adiabatic (had a
constant specic entropy); then the pressure and density would vary together
as P

with = 5/3. Often, one also considers distributions with the same
pressure-density relationship, but for values of in the range 1 5/3.
We will refer to these distributions as polytropic. Then, the hydrostatic
equation can be solved to give
T(r)
T
0
= 1 + ( 1)
_
1
(r)

0
_
, (29)
(r)

0
=
_
T(r)
T
0
_
1/(1)
. (30)
Here, T
0
is the central temperature, and T()/T
0
. The temperature
proles in the outer parts of clusters can generally be t with intermediate
values of 1.21.3 [24].
7.3 Surface Brightness Deprojection
The gas distributions in clusters can be derived directly from observations of
the X-ray surface brightness of the cluster, if the shape of the cluster is known
and if the X-ray observations are suciently detailed and accurate. The X-ray
12 C. L. Sarazin
surface brightness at a photon frequency and at a projected distance b from
the center of a spherical cluster is
I

(b) =
_

b
2

(r)dr
2

r
2
b
2
, (31)
where

is the X-ray emissivity. This Abel integral can be inverted to give

the emissivity as a function of radius,

=
1
2r
d
dr
_

r
2
I

(b)db
2

b
2
r
2
. (32)
The emissivity

is proportional to the square of the density, and its

spectral dependence is determined by the gas temperature and abundances
(7), (8) and (9). Thus, the radial dependence of the spectrum and intensity
of X-rays can be de-projected to the the gas density (r), gas temperature
T(r), and abundances as a function of physical radius. The gas pressure is
then given by the ideal gas law (15).
8 Cluster Masses
Once the gas density has been determined by either model tting or de-
projection, the gas mass can be derived simply as
M
gas
(r) = 4
_
r
0
(r

)(r

)
2
dr

. (33)
Here, M
gas
(r) is the gas mass interior to the radius r.
The total gravitational mass can be derived from the condition of hydro-
static equilibrium (22), which can be written as
M(r) =
r
2
G(r)
dP
dr
, (34)
where M(r) is the total mass interior to r. This equation can also be written
as
M(r) =
kT(r)r
m
p
G
_
d ln (r)
d ln r
+
d lnT(r)
d ln r
_
. (35)
Optical observations of the galaxies can be used to estimate the total mass
of galaxies interior to r, M
gal
(r). Any diuse stellar light can be included
in this; although these values can be dicult to determine, the stars and
galaxies constitute only a small fraction of the mass, so this correction is not
so important. Then, the mass of dark matter in the cluster (interior to r) is
given by
Gas Dynamics in Clusters of Galaxies 13
M
DM
(r) = M(r) M
gas
(r) M
gal
(r) . (36)
In typical clusters, the masses of stars and galaxies are much smaller than
those of the hot gas, with M
gal
0.15M
b
at large radii [49]. Thus, hot plasma
is the dominant form of baryonic matter in clusters of galaxies, with M
gas

6M
gal
at large radii. It appears that the same may be true on large scales
throughout the present day Universe; it seems that most of the baryons in the
Universe today are in hot, diuse intergalactic gas (often called WHIM, or
Warm Hot Intergalactic Medium), rather than stars and galaxies (e.g., [8]).
In this sense, cluster represent the tip of the iceberg. With their very high
densities, they are the one place it has been easy to detect the bulk of the
baryons, which are in intergalactic gas.
The gas mass fraction f
gas
(r) and baryon fraction f
b
(r) are then
f
gas
(r) =
M
gas
(r)
M(r)
, f
bary
(r) =
M
gas
(r) +M
gal
(r)
M(r)
. (37)
The observations of most clusters show an increase in f
gas
(r) with radius r
in the inner parts of clusters [1]. Thus, the gas is more broadly distributed
than the dark matter in clusters. The gas fractions level out at radii which are
r

> 0.2r
vir
. Rich clusters have gas fractions which average f
gas
(r
2500
)) = 12%
at a radius where the mean interior density is 2500 times the critical density
[1]. For typical clusters, r
2500
0.25r
vir
. The total gas fraction within r
vir
will be a bit larger than this. Thus, clusters appear to consist of about 23%
stars and galaxies, 14% hot gas, and 84% dark matter. Although these are
recent values, clusters of galaxies provided some of the earliest evidence that
the mass in the Universe was predominantly dark matter.
Clusters of galaxies are very useful cosmological probes. Arguably, they
are the largest objects in the Universe which are dynamically relaxed. On
the other hand, they are probably the smallest objects which formed from a
suciently large volume that they contain a fair sample of the material in
the Universe. Thus, the ratio of baryons to dark matter in clusters should
be close to the universal value. Numerical simulations do indicate that the
baryon fraction in clusters is nearly the general value in the Universe; even at
r
2500
, f
bary
is about 82% of the universal value [10].
When combined with the density of baryons inferred from Big Bang nu-
cleosynthesis, the observed baryon fraction in clusters indicates that the total
mass density in the Universe is
m
0.3 [52]. Thus, cluster have provided
some of the earliest and strongest evidence that we live in a low density Uni-
verse, with too little matter to close the Universe and reverse the expansion
of the Big Bang.
The measured values of f
gas
and f
bary
depend on the distance d to a
cluster as d
3/2
. On the other hand, if clusters are fair samples of the materials
in the Universe, then f
bary
should be independent of redshift or distance.
Thus, a comparison of f
bary
in low redshift and high redshift clusters provides
a measure of the distance to the clusters which is independent of the redshift.
14 C. L. Sarazin
Such measurements provide evidence that we live in an accelerating Universe,
with an eective cosmological constant of

0.7 [1]. This is in concordance

with the results from WMAP [2] and supernova Type Ia observations at high
redshifts.
9 Heating and Cooling of Intracluster Gas
9.1 Why Is the ICM So Hot?
When it was rst observed in X-rays, one of the most surprising features about
the intracluster gas was its very high temperature. Why is this gas so hot? In
fact, this is one of the easiest aspects of the ICM to understand. At least in
rich clusters, most of the heating is gravitational in origin. The basic idea is
that clusters have huge masses, and very deep gravitational potential wells.
Essentially, any means of introducing the gas into a cluster will cause it to
move very rapidly, and collide with other gas, and be shocked. For example,
if the gas fell into the cluster (either at the same time as the dark matter,
or subsequently), cluster gravitational potentials imply that the gas would
fall in at a speed

> 1000 km s
1
. Unless the gas motions were very carefully
controlled, the gas would encounter other gas moving at similar velocities, and
the intersecting gas streams would collide and shock. Since the ICM has heavy
elements, a portion of it came out of galaxies. If it did so after the clusters
formed, then the galaxies would be moving at orbital speeds of

> 1000 km s
1
in the cluster, and gas ejected from dierent galaxies would collide and shock
at these sorts of speeds. (If the gas came out of galaxies before clusters formed,
then it had to fall into a cluster, and was shocked as described previously.)
Thus, it is likely that essentially all of the gas in the ICM medium shocked at
speeds of

> 1000 km s
1
, and was heated in this way.
In actually, we believe that clusters form hierarchically from the merger
of smaller groups and clusters. Such mergers are discussed extensively below
(Sect. 10). Thus, the specic mechanism for much of the heating of the ICM
is likely to be merger shocks.
9.2 Simple Scaling Laws for Gravitational Heating
If one assumes that gravitational heating dominates in clusters, and makes
a few other simple assumptions, it is possible to derive a number of simple
scaling laws for the X-ray properties of clusters [18]. If the gas in clusters
is in hydrostatic equilibrium and is distributed similarly to the dark matter,
then the typical gas temperature should be kT m
p
GM/R, where M is
the total cluster mass, and R is the cluster radius. If one can treat the for-
mation of a cluster as equivalent to the collapse of a isolated, spherical region
of overdensity in the Universe, then the post-collapse average density in the
Gas Dynamics in Clusters of Galaxies 15
cluster should be
tot
) 180
crit
(z
form
), where
tot
is the total mass den-
sity (dark matter and baryons), and
crit
(z
form
) is the critical density for the
Universe to collapse at the epoch of formation of the cluster. If most clusters
have formed recently, then one could approximate z
form
0. Finally, clusters
are large enough to contain a fair sample of the material in the Universe, and
thus it is reasonable to assume that the baryon fraction in clusters (which is
predominantly in the hot gas) is the universal value (Sect. 8). Then, the radii
of clusters should scale with mass as
R M
1/3
. (38)
The gas temperature would scale as
T M
2/3
, (39)
and the X-ray luminosity vs. temperature relationship would be
L
X
T
2
. (40)
The latter scaling assumes that the X-ray emission is mainly due to thermal
bremsstrahlung, which is true for hot clusters.
9.3 Non-Gravitational Heating
There are a number of indications that non-gravitational heating or cooling
processes may aect the ICM, particularly in smaller clusters and groups.
First, the observed cluster X-ray properties do not agree very well with the
scaling relations for purely gravitational heating (38), (39) and (40). Prob-
ably, the most signicant deviation is that the measured X-ray luminosity
temperature relation is much steeper than (40) [23]. The departures for the
scaling relations are particularly strong for cooler clusters and groups. Sec-
ond, the observed gas distributions in clusters are more extended than would
be expected from purely gravitational heating. The gas distributions often
have central cores. This suggests that some non-gravitational heating pro-
cesses have occurred and have pued up the gas distributions, particularly in
the poorer clusters. This would lower the average density in the gas, and thus
reduce the X-ray luminosity. An alternative possibility is that inhomogeneous
cooling has removed the cooler ICM, increasing the average temperature of
the gas which remains. Presumably, this cooling would also lead to star and
galaxy formation. These topics have been reviewed extensively in [50].
If the non-gravitational heating occurred just prior to the collapse of a
cluster, then the amount of heat needed is 2 keV per particle [21]. How-
ever, a more useful quantity to describe the preheating is probably the extra
entropy per particle s (18). As noted in Sect. 6, the specic entropy is a
Lagrangian quantity which moves with the gas, and which remains constant
for reversible changes. As discussed in Sect. 6, it is conventional to use the
16 C. L. Sarazin
entropy parameter K (19) rather than s. For purely gravitational heating, the
scaling laws described above (Sect. 9.2) imply that the entropy parameter is
expected to scale as
K T M
2/3
. (41)
Observations of clusters and groups initially suggested that preheating pro-
duced an extra entropy of K 135 keV cm
2
[22, 37]. It now appears that
such an entropy oor may be to simplied to explain the detailed varia-
tions in entropy between clusters and the radial variations within clusters.
Also, the existence of the Lyman alpha forest and other quasar absorption
lines indicates that not all of the intergalactic gas underwent the same level of
preheating. Nonetheless, this value provides a useful value in assessing models
for the thermal history of the ICM.
The radial variation of the entropy in the ICM also appears to be inconsis-
tent with purely gravitational heating. Gravitational heating models predict
that the entropy vary roughly as K r
1.1
. The observed entropy proles in
clusters are much atter in the center [38].
Supernovae could provide a signicant source of heating of the ICM. These
would include core collapse supernova associated with the deaths of massive
stars. Since the galaxies in clusters today contain very few such stars, this
would have occurred during the epoch of star formation and galaxy forma-
tion. The supernovae might have driven galactic winds. The second type of
supernovae are Type Ias, which are produced by older binary star systems.
They would provide a more continuous source of heating.
Supernovae also eject heavy elements. Thus, the abundances in clusters
can be used to limit the total number of supernova which have occurred. The
observed abundances suggest that the extra energy added is probably 0.3
keV per particle [21]. This is a bit low to explain the required preheating, but
might be possible. However, this mechanism would also require that a large
fraction of the supernova explosion energy be converted into heat in the ICM,
which may also be a diculty.
Active galactic nuclei (AGNs) within clusters might also provide a signif-
icant amount of heating. As with the supernovae, it is dicult to determine
what fraction of the energy produced by AGNs goes into heating the surround-
ing medium. In this regard, it is only the AGN output in kinetic energy in
jets or in relativistic particles which is likely to be useful. It may be important
that the early-type galaxies found in clusters generally host radio galaxies and
radio quasars, which are more likely to deposit energy into the ICM.
One way to limit the total energy input by AGNs is to look at the total
masses of supermassive black holes contained in clusters today. In general, all
large bulges appear to contain supermassive black holes, and there is a strong
correlation of black hole mass with bulge mass or velocity dispersion. If the
growth of black holes occurred largely by accretion (rather than merging of
existing massive black holes), then the total accretion energy from black holes
Gas Dynamics in Clusters of Galaxies 17
can be derived from their total mass. This could provide a signicant level of
heating for the ICM if the fraction of accreted energy which goes into heating
is

>10% [5].
9.4 Cooling in the Intracluster Medium
The primary cooling process for the ICM is the emission of X-ray radiation.
The emission is proportional to the square of the density and varies with
temperature (7), (8), and (9). Thus, the total cooling rate per unit volume L
in the gas can
L = (T, Abundances)n
e
n
p
, (42)
where depends only on the temperature and the abundances of the heav-
ier elements relative to hydrogen. At high temperature (kT

> 2 keV), the

dominant radiation is thermal bremsstrahlung, and T
1/2
. At lower tem-
perature, line emission becomes dominant, and decreases with increasing
temperature.
At high temperatures where thermal bremsstrahlung dominates, the time
required for gas to cool to low temperatures at constant pressure is
t
cool
= 69
_
n
e
10
3
cm
3
_
1
_
T
10
8
K
_
1/2
Gyr . (43)
Note that cooling accelerates as the gas cools; this tendency is even stronger
below kT

< 2 keV due to line emission. The cooling time is much longer than
the Hubble time in the outer parts of clusters. However, it can be quite short
(t
cool
300 Myr) in the inner regions of cooling core clusters.
It is interesting to write the cooling time as a function of the entropy and
temperature rather than the density and temperature:
t
cool
= 17
_
K
130 keVcm
2
_
3/2
_
kT
2 keV
_
1
Gyr . (44)
Note that the cooling time is less than the Hubble time for K

< 130 keV cm

2
for kT 2.5 keV. If clusters start with gas with a wide range of entropies,
the lower entropy gas will cool rapidly and be removed from the ICM. Thus,
cooling can increase the average entropy of the gas and provide an eective
oor to the ICM entropy [51]. The cooled gas presumably goes into forming
galaxies and stars. Feedback heating from supernovae, galactic winds, and
AGNs (Sect. 9.3) might result in some of the cooled gas actually becoming
hotter ICM. However, the result is to remove the cooler gas and raise the
average entropy of the ICM.
Thus, while the bulk of the heating of the ICM in large clusters is due
to gravitational heating, mainly by merger shocks, smaller clusters and the
centers of clusters show evidence for the eects of non-gravitational heating
and cooling. The cooling leads to star and galaxy formation, and which leads
18 C. L. Sarazin
to possible heating by supernovae, galactic winds, and AGNs. Thus, the ICM
(and intergalactic medium more generally) preserve a unique record of the
thermal history of the Universe.
10 Cluster Mergers
Major cluster mergers are the most energetic events in the Universe since the
Big Bang. Cluster mergers are the mechanism by which clusters are assem-
bled. In these mergers, the subclusters collide at velocities of 2000 km/s,
releasing gravitational binding energies of as much as

>10
64
ergs. During
mergers, shocks are driven into the intracluster medium. In major mergers,
these hydrodynamical shocks dissipate energies of 3 10
63
ergs; such shocks
are the major heating source for the X-ray emitting intracluster medium. The
shock velocities in merger shocks are similar to those in supernova remnants
in our Galaxy, and we expect them to produce similar eects. Mergers shocks
should heat and compress the X-ray emitting intracluster gas, and increase
its entropy. We also expect that particle acceleration by these shocks will pro-
duce relativistic electrons and ions, and these can produce synchrotron radio,
inverse Compton (IC) EUV and hard X-ray, and gamma-ray emission. (See
the chapter by Feretti & Giovannini for more details relativistic particles and
non-thermal emission in clusters.)
11 Thermal Physics of Merger Shocks
The intracluster medium (ICM) is generally close to hydrostatic equilibrium
in clusters which are not undergoing strong mergers. The virial theorem then
implies that the square of the thermal velocity (sound speed) of the ICM is
comparable to the gravitational potential. During a merger, the infall veloc-
ities of the subclusters are comparable to the escape velocity, which implies
that the square of the infall velocity is larger (by roughly a factor of two) than
the gravitational potential. Thus, the motions in cluster mergers are expected
to be supersonic, but only moderately so. As a result, one expects that cluster
mergers will drive shock waves into the intracluster gas of the two subclusters.
Let v
s
be the velocity of such a shock wave relative to the preshock intraclus-
ter gas. The sound speed in the preshock gas is c
s
=
_
(5/3)P/, where P is
the gas pressure and is the density. Then, the Mach number of the shock is
/ v
s
/c
s
. Based on the simple argument given above, one expects shocks
with Mach numbers of /

< 2. Stronger shocks may occur under some cir-
cumstances, such as in the outer parts of clusters, or in low mass subclusters
merging with more massive clusters.
Shocks are irreversible changes to the gas in clusters, and thus increase
the entropy S in the gas. A useful quantity to consider is the specic entropy
per particle in the gas, s (18). Observations of X-ray spectra can be used to
Gas Dynamics in Clusters of Galaxies 19
determine T, while the X-ray surface brightness depends on
2
. Thus, one can
use X-ray observations to determine the specic entropy in the gas just before
and just after apparent merger shocks seen in the X-ray images. Since merger
shocks should produce compression, heating, pressure increases, and entropy
increases, the corresponding increase in all of these quantities (particularly
the entropy) can be used to check that discontinuities are really shocks (e.g.,
not cold fronts or other contact discontinuities, Sect. 12.2).
In [26], this test was applied to ASCA temperature maps and ROSAT
images of Cygnus-A and Abell 3667, two clusters which appeared to show
strong merger shocks. Recent Chandra images have shown that the feature in
Abell 3667 is a cold front [47]. In Cygnus-A, the increase in specic entropy
in the shocked regions is roughly s (3/2)k. The specic heat per particle
q which must be dissipated to produce this change in entropy is q Ts
(3/2)kT, or about the present specic heat content in the shocked gas. Thus,
these observations provide a direct conrmation that merger shocks contribute
signicantly to the heating of the intracluster gas.
The most dramatic merger shock which has been seen with Chandra is in
the Bullet Cluster 1E0657-56 [7, 29, 30]. This is a very high velocity (4500
km s
1
) merger occurring nearly in the plane of the sky, with a merger bow
shock located ahead of a cold front (Sect. 12.2). Another prominent merger
shocks with a Mach number of / 2.1 is seen in Abell 520 [25]. In both
cases, the merger shocks appear to have associated diuse radio emission (See
the chapter by Feretti & Giovannini for more details.)
11.1 Shock Kinematics
The variation in the hydrodynamical variables in the intracluster medium
across a merger shock are determined by the standard RankineHugoniot
jump conditions [20], if one assumes that all of the dissipated shock energy
is thermalized. Consider a small element of the surface of a shock (much
smaller than the radius of curvature of the shock, for example). The tangential
component of the velocity is continuous at the shock, so it is useful to go to
a frame which is moving with that element of the shock surface, and which
has a tangential velocity which is equal to that of the gas on either side of the
shock. In this frame, the element of the shock surface is stationary, and the
gas has no tangential motion. Let the subscripts 1 and 2 denote the preshock
and post-shock gas; thus, v
1
= v
s
is the longitudinal velocity of material into
the shock (or alternative, the speed with which the shock is advancing into
the preshock gas). Conservation of mass, momentum, and energy then implies
the following jump conditions

1
v
1
=
2
v
2
,
P
1
+
1
v
2
1
= P
2
+
2
v
2
2
,
w
1
+
1
2
v
2
1
= w
2
+
1
2
v
2
2
. (45)
20 C. L. Sarazin
Here, w = P/+ is the enthalpy per unit mass in the gas, and is the internal
energy per unit mass. If the gas behaves as a perfect uid on each side of the
shock, the internal energy per unit mass is given by: = /( 1) P/, where
is the ratio of specic heats (the adiabatic index) and is = 5/3 for fully
ionized plasma. The jump conditions can be rewritten as:
P
2
P
1
=
2
+ 1
/
2

1
+ 1
v
2
v
1
=

1

1
C
=
2
+ 1
1
/
2
+
1
+ 1
, (46)
where C
2
/
1
is the shock compression.
If one knew the velocity structure of the gas in a merging cluster, one could
use these jump condition to derive the temperature, pressure, and density
jumps in the gas. At present, the best X-ray spectra for extended regions in
clusters of galaxies have come from CCD detectors on ASCA, Chandra, and
XMM/Newton. CCDs have a spectral resolution of >100 eV at the Fe K line
at 7 keV, which translates into a velocity resolution of >4000 km/s. Thus, this
resolution is (at best) marginally insucient to measure merger gas velocities
in clusters. In a few cases with very bright regions and simple geometries,
the grating spectrometers on Chandra and especially XMM/Newton may be
useful.
At present, X-ray observations can be used to directly measure the tem-
perature and density jumps in merger shocks. Thus, one needs to invert the
jump relations to give the merger shock velocities for a given shock tempera-
ture, pressure, and/or density increase. If the temperatures on either side of
the merger shock can be measured from X-ray spectra, the shock velocity can
be inferred from [26]:
v
s
=
_
kT
1
m
p
(C 1)
_
T
2
T
1

1
C
__
1/2
, (47)
where v
s
= v
1
v
2
= [(C 1)/C]v
s
is the velocity change across the shock,
and is the mean mass per particle in units of the proton mass m
p
. The shock
compression C can be derived from the temperatures as
1
C
=
_
1
4
_
+ 1
1
_
2
_
T
2
T
1
1
_
2
+
T
2
T
1
_
1/2

1
2
+ 1
1
_
T
2
T
1
1
_
. (48)
Alternatively, the shock compression can be measured directly from the X-ray
image. However, it is dicult to use measurements of the shock compression
alone to determine the shock velocity, for two reasons. First, a temperature
is needed to set the overall scale of the velocities; as is obvious from (46),
the shock compression allows one to determine the Mach number / but
not the shock velocity. The second problem is that temperature or pressure
Gas Dynamics in Clusters of Galaxies 21
information is needed to know that a discontinuity in the gas density is a shock,
and not a contact interface (e.g., the cold fronts discussed in Sect. 12.2
below).
X-ray temperature maps of clusters have been used to derive the merger
velocities using these relations. Reference [26] used ASCA observations to de-
termine the kinematics of mergers in three clusters (Cygnus-A, Abell 2065,
and Abell 3667). Because of the poor angular resolution of ASCA, these
analyses were quite uncertain. More recently, possible shocks have been de-
tected in Chandra images of a number of merging clusters (e.g., Abell 85 [19],
Abell 665 [27], Abell 3667 [47]), and the shock jump conditions have been
applied to determine the kinematics in these clusters.
The simplest case is a head-on symmetric merger (b = 0 and M
1
= M
2
) at
an early stage when the shocked region lies between the two cluster centers.
Reference [26] suggests that the Cygnus-A cluster is an example. If the gas
within the shocked region is nearly stationary, then the merger velocity of the
two subclusters is just v = 2v
s
. Applying these techniques to the ASCA
temperature map for the Cygnus-A cluster, Reference [26] found a merger
velocity of v 2200 km/s. This simple argument is in reasonable agreement
with the results of numerical simulations of this merger [39].
One can compare the merger velocities derived from the temperature
jumps in the merger shocks with the values predicted by free-fall from the
turn-around radius. In the case of Cygnus-A, [26] found good agreement with
the the free-fall velocity of 2200 km/s. This consistency suggests that the
shock energy is eectively thermalized, and that a major fraction does not go
into turbulence, magnetic elds, or cosmic rays. Thus, the temperature jumps
in merger shocks can provide an important test of the relative roles of thermal
and non-thermal processes in clusters of galaxies.
11.2 Nonequilibrium Eects
Cluster mergers are expected to produce collisionless shocks, as occurs in
supernova remnants. As such, nonequilibrium eects are expected, including
non-equipartition of electrons and ions and nonequilibrium ionization [26, 44,
45]. Collisionless shocks are generally not as eective in heating electrons as
ions. Assuming that the post-shock electrons are somewhat cooler than the
ions, the time scale for electron and protons to approach equipartition as a
result of Coulomb collisions in a hot ionized gas is (2) [43]:
t
eq
=
3m
p
m
e
8

2n
e
e
4
ln
_
kT
e
m
e
_
3/2
2.1 10
8
_
T
e
10
8
K
_
3/2
_
n
e
0.001 cm
3
_
1
yr , (49)
where n
e
and T
e
are the electron number density and temperature, re-
spectively, and is the Coulomb factor. The relative velocity between the
22 C. L. Sarazin
post-shock gas and the shock front is (1/4)v
s
; thus, one would expect the
electron temperature to reach equipartition a distance of
d
eq
160
_
v
s
3000 km/s
_ _
T
e
10
8
K
_
3/2
_
n
e
0.001 cm
3
_
1
kpc (50)
behind the shock front. Of course, it is the electron temperature (rather than
the ion or average temperature) which determines the shape of the X-ray spec-
trum. This distance is large enough to insure that the lag could be spatially
resolved in X-ray observations of low redshift clusters. Similar eects might
be expected through non-equilibrium ionization.
On the other hand, it is likely that the nonequilibrium eects in cluster
merger shocks are much smaller than those in supernova blast wave shocks
because of the low Mach numbers of merger shocks. That is, the preshock gas
is already quite hot (both electrons and ions) and highly ionized. Moreover, a
signicant part of the heating in low Mach number shocks is due to adiabatic
compression, and this would still act on the electrons in the post-shock gas
in merger shocks, even if there were no collisionless heating of electrons. For
example, in a /= 2, = 5/3 shock, the total shock increase in temperature
is a factor of 2.08 (46). The shock compression is C = 2.29, so adiabatic
compression increases the electron temperature by a factor of C
2/3
= 1.74,
which is about 83% of the shock heating.
11.3 Mergers and Basic Gravitational Physics Eects
Merging clusters also provide several very direct tests of basic gravitational
physics. These tests are possible because of the dynamical nature of mergers,
and the dierence in the behavior of collisional and non-collisional components
of clusters. The gas is clusters is a collisional uid (Sect. 4) with a mean-
free-path which is small compared to the scale of clusters (11). Thus, when
clusters collide, the motion of the gas will be retarded by ram pressure and
shocks. On the other hand, the galaxies in clusters are essentially collisionless.
When clusters collide, the galaxies will y by one another. Thus, the galaxies
in a merging subcluster will often be found ahead of the gas from the same
subcluster. This is particularly obvious in late stage mergers with cold fronts
(Sect. 12.2), where the gas which was initially at the center of a subcluster
will be found lagging behind the central dominant and other galaxies from the
subcluster. Perhaps the most prominent example of this is the in the Bullet
Cluster 1E0657-56 [7, 29, 30], where the cold front and dense gas from the
subcluster are clearly separated for the galaxies from the same subcluster.
In the most widely accepted model for dark matter (Sect. 8), the dom-
inant component of the mass of the Universe is made up of collisionless
elementary particles. For example, this would be true of Cold Dark Matter
(e.g., [2]). If this is the case, one would expect that the dark matter would be
located in the same regions of merging clusters as the galaxies. An alternative
Gas Dynamics in Clusters of Galaxies 23
idea is that the law of gravity or laws of motions dier from the Newtonian
form at large distances or small accelerations [33]. In these MOdied Newto-
nian Dynamics (MOND) theories, there is no dark matter component of the
Universe, and gravity is just due to ordinary baryonic matter. In clusters of
galaxies, the vast majority of the ordinary baryonic matter is in the hot X-ray
gas (Sect. 8). Thus, MOND theories predict that in a merger, the gravity (or
apparent dark matter) should mainly be located where the gas is located. In
an advanced merger, the gas is located behind the galaxies.
The location of the gravity (or apparent dark matter) can be determined
from weak gravitational lensing observations of the cluster [7, 30]. This test
has been performed on the Bullet Cluster 1E0657-56 [7, 30], where the
weak lensing measurement show that the gravity of the merging subcluster is
centered on the galaxies, and is clearly displaced from the located of the sub-
cluster gas. Thus, these measurements provide what is arguably the strongest
proof of the existence of dark matter, rather than a change in the laws of
gravity.
Another alternative to the conventional Cold Dark Matter hypothesis is
that the dark matter consists of weakly interacting elementary particles, but
the particles do have a small but signicant cross-section for self-interaction
[42]. If this were the case, the dark matter would act as a collisional uid,
and would be displaced from the position of the galaxies in a merging cluster
towards the center of the subcluster gas. In the Bullet Cluster 1E0657-56 no
such displacement is evident [30], and the lack of such a displacement can be
used to set an upper limit on the self-interaction cross-section per unit mass
of dark matter of < 1 cm
2
g
1
. This is a very serious constraint on models of
self-interacting dark matter.
12 Mergers and Cool Cluster Cores
12.1 Cooling Flows vs. Mergers
The centers of a signicant fraction of clusters of galaxies have luminous cusps
in their X-ray surface brightness known as cooling ows (see [12] for an
extensive review). In every case, there is a bright (cD) galaxy at the center
of the cooling ow region. The intracluster gas densities in these regions are
much higher than the average values in the outer portions of clusters. X-
ray spectra indicate that there are large amounts of gas at low temperatures
(down to 10
7
K), which are much cooler than those in the outer parts of
clusters. The high densities imply rather short cooling times t
cool
(the time
scale for the gas to cool to low temperature due to its own radiation). The
hypothesis is that the gas in these regions is cooling from higher intracluster
temperature (10
8
K) down to these lower temperatures as a result of the
energy loss due to the X-ray emission we observe. Typical cooling rates are
100 M

yr
1
. The cooling times, although much shorter than the Hubble
24 C. L. Sarazin
time, are generally much longer than the dynamical (i.e., sound crossing time)
of the gas in these regions. As a result, the gas is believed to remain nearly in
hydrostatic equilibrium. Thus, the gas must compress as it cools to maintain
a pressure which can support the weight of the overlying intracluster medium.
The primary observational characteristics of cooling ows are very bright
X-ray surface brightnesses which increase rapidly toward the center of the clus-
ter. The high surface brightnesses imply high gas densities which also increase
rapidly towards the cluster center. These regions contain cooler cluster gas.
Empirically, there is signicant indirect evidence that mergers disrupt cool-
ing ows. There is a strong statistical anticorrelation between cooling ows
and/or cooling rates, and irregular structures in clusters as derived by statisti-
cal analysis of their X-ray images [3]. Looked at individually, very strong cool-
ing cores are almost never associated with very irregular or bimodal clusters,
which are likely merger candidates [9, 17]. There are some cases of moderate
cooling ows in merging clusters; in most cases, these appear to be early-
stage mergers where the merger shocks havent yet reached the cooling core
of the cluster. Examples may include Cygnus-A [26] and Abell 85 [19]. There
also are a large number of merging clusters at a more advanced stage with
relatively small cooling cores; Abell 2065 [26] may be an example. Recently,
Chandra Observatory X-ray images have shown a number of merging clusters
with rapidly moving cores of cool gas (the cold fronts discussed below in
Sect. 12.2). In these systems, the cooling ows appear to have survived, at
least to the present epoch in the merger.
It is unclear exactly how and under what circumstances mergers disrupt
cooling ows. The cooling ows might be disrupted by tidal eects, by shock
heating the cooler gas, by removing it dynamically from the center of the
cluster due to ram pressure, by mixing it with hotter intracluster gas, or by
some other mechanism. Numerical hydrodynamical simulations are needed to
study the mechanisms by which cooling ows are disrupted. This is a relatively
unexplored area, largely because the small spatial scales and rapid cooling time
scales in the inner regions of cooling ows are still a signicant challenge to
the numerical resolution of hydrodynamical codes. McGlynn and Fabian [32]
argued that mergers disrupted cooling ows, but this was based on purely N-
body simulations. Hydrodynamical simulations have been made of the eects
of head-on mergers with relatively small subclusters (1/4 or 1/16 of the mass
of the main cluster) on a cooling ow in the main cluster [16]. They nd
that the mergers disrupt the cooling ow in some cases, but not in others.
Their simulations suggest that the disruption is not due to tidal or other
gravitational eects.
Another possibility is that the merger shocks heat up the cooling ow gas
and stop the cooling ow. In the simulations, this does not appear to be the
main mechanism of cooling ow disruption. There are a number of simple
arguments which suggest that merger shocks should be relatively inecient
at disrupting cooling ows. First, it is dicult for these shocks to penetrate
the high densities and steep density gradients associated with cooling ows,
Gas Dynamics in Clusters of Galaxies 25
and the merger shocks would be expected to weaken as they climb these
steep density gradients. Even without this weakening, merger shocks have
low Mach numbers, and only produce rather modest increases in temperature
(

< a factor of 2). These small temperature increases are accompanied by

signicant compressions. As a result, shock heating actually decreases the
cooling time due to thermal bremsstrahlung emission for shocks with Mach
numbers / (21 + 12

3)
1/2
6.5. It is likely that the shocked gas will
eventually expand, and adiabatic expansion will lengthen the cooling time.
However, even if the gas expands to its preshock pressure, the increase in the
cooling time is not very large. For a /= 2 shock, the nal cooling time after
adiabatic expansion to the original pressure is only about 18% longer than
the initial cooling time.
The simulations by [16] suggest that the main mechanism for disrupting
cooling ows is associated with the ram pressure of gas from the merging
subcluster. The gas in the cooling ow is displaced, and may eventually mix
with the hotter gas [39]. Earlier, [13] had argued that ram pressure, rather
than shock heating, was the main mechanism for disrupting cooling ows.
Assuming this is the case, one expects that the merger will remove the cooling
ow gas at radii which satisfy

sc
v
2
rel

> P
CF
(r) , (51)
where P
CF
(r) is the pressure prole in the cooling ow,
sc
is the density
of the merging subcluster gas at the location of the cooling ow, and v
rel
is the relative velocity of the merging subcluster gas and the cooling ow.
Reference [16] nds that this relation provides a reasonable approximation to
the disruption in their simulations.
The pressure prole in the cooling ow gas prior to the merger is deter-
mined by the condition of hydrostatic equilibrium. If the cluster gravitational
potential has a wide core within which the potential is nearly constant (e.g.,
as in a King model), then the cooling ow pressure will not increase rapidly
into the center. In this case, once the merger reaches the central regions of
the cluster, if the ram pressure is sucient to remove the outer parts of the
cooling ow, it should be sucient to remove nearly all of the cooling ow.
On the other hand, if the cluster potential is sharply peaked as is predicted
by numerical simulations [35], the merger may remove the outer parts of the
cooling ow but not the innermost regions. Thus, the survival and size of
cool cores in merging clusters can provide evidence on whether clusters have
sharply peaked potentials [26].
12.2 Cold Fronts
One of the more dramatic early discoveries with the Chandra X-ray Obser-
vatory was the presence of very sharp surface brightness discontinuities in
merging clusters of galaxies. A pair of such discontinuities were rst seen in
26 C. L. Sarazin
the public science verication data on the Abell 2142 cluster [28]. Initially, it
seemed likely that these were merger shocks. However, temperature measure-
ments showed that this was not the case. The high X-ray surface brightness
regions were both dense and cool; thus, the gas in these regions had a lower
specic entropy than the gas in the less dense regions. The lack of a pressure
jump and the incorrect sign of the temperature and entropy variations showed
that these features could not be shocks [28]. Instead, they appear to be contact
discontinuities between hot, diuse gas and a cloud of colder, denser gas [28].
In [47], these contact discontinuities were named cold fronts. Reference [28]
argues that the source of the cold clouds are the cooling cores of one or both
of merging subclusters. As noted above, cooling ows do appear to be able to
partially survive in mergers, at least for some period. Subsequently, cold fronts
have been observed in a number of other clusters; for an extensive review of
the observations of these cold fronts, see [14].
Kinematics of Cold Fronts
As discussed extensively in [47], the variation in the density, pressure, and
temperature of the gas in a cold front can be used to determine the relative
velocity of cold core. This technique is analogous to that for merger shocks
discussed above (47) and (48). The geometry is illustrated in Fig. 2, which is
drawn in the rest frame of the cold core. We assume that the cold core has
a smoothly curved, blunt front edge. The normal component of the ow of
hot gas past the surface of the cold core will be zero. There will be at least
one point where the ow is perpendicular to the surface of the cold core, and
the ow velocity of the hot gas will be zero at this stagnation point (st
in Fig. 2). Far upstream, the ow of the hot gas will be undisturbed at the
velocity of the cold core relative to the hotter gas, v
1
. Let c
s1
be the sound
speed in this upstream gas, and /
1
v
1
/c
s1
be the Mach number of the
motion of the cold core into the upstream gas. If /
1
> 1, a bow shock will
be located ahead of the cold front.
The ratio of the pressure at the stagnation point to that far upstream is
given by [20]
P
st
P
1
=
_
_
_
_
1 +
1
2
/
2
1
_

1
, /
1
1
/
2
1
_
+1
2
_
+1
1
_

1
2M
2
1
_

1
1
, /
1
> 1 .
(52)
The ratio (P
st
/P
1
) increases continuously and monotonically with /
1
. Thus,
in principle, measurements of P
1
and P
st
in the hot gas could be used to
determine /
1
. The pressures would be determined from X-ray spectra and
images. In practice, the emissivity of the hot gas near the stagnation point is
likely to be small. However, the pressure is continuous across the cold front,
so the stagnation pressure can be determined just inside of the cold core,
where the X-ray emissivity is likely to be much higher. Once /
1
has been
determined, the velocity of the encounter is given by v
1
= /
1
c
s1
.
Gas Dynamics in Clusters of Galaxies 27
1 2
cool, dense gas
bowshock?
hot, diffuse gas
st
Cold Front
Fig. 2. A schematic diagram of ow around a cold front in a cluster merger. The
heavy solid arc at the right represents the contact discontinuity between the cold,
dense cold core gas, and the hotter, more diuse gas from the outer regions of the
other cluster. The cold core is moving toward the left relative to the hotter gas. The
narrow solid lines are streamlines of the ow of the hotter gas around the cold core.
The region labelled 1 represent the upstream, undisturbed hot gas. If the cold
front is moving transonically (M
1
> 1), then the cold front will be preceded by a
bow shock, which is shown as a dashed arc. The stagnation point, where the relative
velocity of the cooler dense gas and hotter diuse gas is zero, is marked st
If the motion of the cold core is transonic (/
1
> 1), one can also determine
the velocity from the temperature and/or density jump at the bow shock (47)
and (48). If the bow shock can be traced to a large transverse distance and
forms a cone, the opening angle of this Mach cone corresponds to the Mach
angle,
M
csc
1
(/
1
). However, variations in the cluster gas temperature
may lead to distortions in this shape.
The distance between the stagnation point and the closest point on the
bow shock (the shock stand-o distance d
s
) can also be used to estimate the
Mach number of the motion of the cold front [47]. The ratio of d
s
to the radius
of curvature of the cold front R
cf
depends on the Mach number /
1
and on
the shape of the cold front. Figure 3 shows the values of d
s
/R
cf
as a function
of (/
2
1
1)
1
for a spherical cold front [41]. Although there is no simple
analytic expression for the stand-o distance which applies to all shapes of
objects, a fairly general approximate method to calculate d
s
has been given
by [34]. and some simple approximate expressions exist for a number of simple
geometries. The stand-o distance increases as the Mach number approaches
unity; thus, this method is, in some ways, a very sensitive diagnostic for the
Mach number for the low values expected in cluster mergers. On the other
hand, the stand-o distance also depends strongly on the shape of the cold
front as the Mach number decreases. The application of this diagnostic to
observed clusters is strongly aected by projection eects. Because the radius
28 C. L. Sarazin
.01
.1
.3
1
3
10
.1 1 10
1/(
2
1)
d
s
/

R
c
f
Fig. 3. The ratio of the stand-o distance of the bow shock d
s
to the radius of
curvature R
cf
of the stagnation region of the cold front, as a function of 1/(M
2
1
1),
where M
1
is the Mach number. This is for a spherical cold front and = 5/3
of curvature of the bow shock is usually greater than that of the cold front,
projection eects will generally cause d
s
to be overestimated and /
1
to be
underestimated. Projection eects also make the true shape of the cold front
uncertain.
These techniques have been used to determine the merger velocities from
cold fronts in Abell 3667 [47], RXJ1720.1+2638 [31], and Abell 85 [19]. The
most spectacular application is the Bullet Cluster 1E0657-56 [7, 29, 30],
which contains a very high Mach number merger.
Width of Cold Fronts
One remarkable aspect of the cold fronts observed with the Chandra Obser-
vatory in several clusters is their sharpness. In Abell 3667, the temperature
changes by about a factor of two across the cold front [47], and the accom-
panying change in the X-ray surface brightness occurs in a region which is
narrower than 2 kpc [47]. This is less than the mean-free-path of electrons in
this region. The existence of this very steep temperature gradient and similar
results in other merging clusters with cold fronts requires that thermal con-
duction be suppressed by a large factor [11, 46, 47] relative to the classical
value in an unmagnetized plasma (14) [43]. It is likely that this suppression is
due to the eects of the intracluster magnetic eld. It is uncertain at this point
whether this is due to a generally tangled magnetic eld (in which case, heat
conduction might be suppressed throughout clusters), or due to a tangential
magnetic eld specic to the tangential ow at the cold front [46].
Gas Dynamics in Clusters of Galaxies 29
Because of the tangential shear ow at the cold front (Fig. 2), the front
should be disturbed and broadened by the KelvinHelmholtz (KH) instabil-
ity. Reference [46] argues that the instability is suppressed by a tangential
magnetic eld, which is itself generated by the tangential ow. This suppres-
sion requires that the magnetic pressure P
B
be a non-trivial fraction of the gas
pressure P in this regions, P
B

> 0.1P. The required magnetic eld strength

in Abell 3667 is B 10 G. Alternatively, cold fronts might be stabilized by
gravity [15] or acceleration along the front [6].
References
1. Allen, S.W., Schmidt, R.W., Ebeling, H., Fabian, A.C., van Speybroeck, L.:
MNRAS 353, 457 (2004)
2. Bennett, C.L., et al.: ApJS 148, 1 (2003)
3. Buote, D.A., Tsai, J.C.: ApJ 458, 27 (1996)
4. Cavaliere, A., Fusco-Femiano, R.: A&A 49, 137 (1976)
5. Cavaliere, A., Lapi, A., Menci, N.: ApJ 581, L1 (2002)
6. Churazov, E.: Physics of cluster cores. In: The X-ray Universe 2005, (in press)
7. Clowe, D., Gonzalez, A., Markevitch, M., et al.: ApJ 604, 596 (2004)
8. Dave, R.: ApJ 552, 473 (2001)
9. Edge, A.C., Stewart, G.C., Fabian, A.C.: MNRAS 258, 177 (1992)
10. Eke, V.R., Navarro, J.F., Frenk, C.S.: ApJ 503, 569 (1998)
11. Ettori, S., Fabian, A.C.: MNRAS 317, L57 (2000)
12. Fabian, A.C.: ARA&A 32, 277 (1994)
13. Fabian, A.C., Daines, S.J.: MNRAS 252, 17p (1991)
14. Forman, W., Jones, C., Markevitch, M., Vikhlinin, A., Churazov, E.: High an-
gular resolution cluster observations with Chandra: a new view. In: Merging
Processes in Galaxy Clusters, pp. 109132. Kluwer, Dordrecht (2002)
15. Fujita, Y., et al.: ApJ 575, 764 (2002)
16. Gomez, P.L., Loken, C., Roettiger, K., Burns, J.O.: ApJ 569, 122 (2002)
17. Henriksen, M.J.: ApJ 407, L13 (1988)
18. Kaiser, N.: MNRAS 222, 323 (1986)
19. Kempner, J., Sarazin, C.L., Ricker, P.R.: ApJ 579, 236 (2002)
20. Landau, L.D., Lifshitz, E.M.: Fluid Mechanics, Pergamon, Oxford (1959)
21. Loewenstein, M.: ApJ 532, 17 (2000)
22. Lloyd-Davies, E.J., Ponman, T.J., Cannon, D.B.: MNRAS 315, 689 (2000)
23. Markevitch, M.: ApJ 504, 27 (1998)
24. Markevitch, M., Forman, W.R., Sarazin, C.L., Vikhlinin, A.: ApJ 503, 77
(1998)
25. Markevitch, M., Govoni, F., Brunetti, G., Jerius, D.: ApJ 627, 733 (2005)
26. Markevitch, M., Sarazin, C.L., Vikhlinin, A.: ApJ 521, 526 (1996)
27. Markevitch, M., Vikhlinin, A.: ApJ 563, 95 (2001)
28. Markevitch, M., et al.: ApJ 541, 542 (2000)
29. Markevitch, M., et al.: ApJ 567, L27 (2002)
30. Markevitch, M., et al.: ApJ 606, 542 (2004)
31. Mazzotta, P., Markevitch, M., Vikhlinin, A., Forman, W.R., David, L.P., Van
Speybroeck, L.: ApJ 555, 205 (2001)
30 C. L. Sarazin
32. McGlynn, T.A., Fabian, A.C.: MNRAS 208, 709 (1984)
33. Milgrom, M.: ApJ 270, 365 (1983)
34. Moekel, W.E.: Approximate Method for Predicting Forms and Location of De-
tached Shock Waves Ahead of Plane or Axially Symmetric Bodies, NACATe-
chnical Note 1921 (1949)
35. Navarro, J.F., Frenk, C.S., White, S.D.M.: ApJ 490, 493 (1997)
36. Osterbrock, D.E.: Astrophysics of Gaseous Nebulae and Active Galactic Nuclei,
pp. 5365. University Science, Mill Valley (1989)
37. Ponman, T.J., Cannon, D.B., Navarro, J.F.: Nature 397, 135 (1999)
38. Ponman, T.J., Sanderson, A.J., Finoguenov, A.: MNRAS 343, 331 (2003)
39. Ricker, P.M., Sarazin, C.L.: ApJ 561, 621 (2001)
40. Sarazin, C.L.: X-ray Emission from Clusters of Galaxies, Cambridge University
Press, Cambridge (1988)
41. Schreier, S.: Compressible Flow, pp. 182189. Wiley, New York (1982)
42. Spergel, D.N., Steinhardt, P.J.: Phys. Rev. Lett. 84, 3760 (2000)
43. Spitzer Jr., L.: Physics of Fully Ionized Gases, Interscience, New York (1956)
44. Takizawa, M.: ApJ 520, 514 (1999)
45. Takizawa, M.: ApJ 532, 182 (2000)
46. Vikhlinin, A., Markevitch, M., Murray, S.M.: ApJ 549, L47 (2001a)
47. Vikhlinin, A., Markevitch, M., Murray, S.M.: ApJ 551, 160 (2001b)
48. Vikhlinin, A., Markevitch, M., Murray, S.S., Jones, C., Forman, W., Van
Speybroeck, L.: ApJ 628, 655 (2005)
49. Voevodkin, A., Vikhlinin, A.: ApJ 601, 610 (2004)
50. Voit, G.M.: RMP 77, 207 (2005)
51. Voit, G.M., Bryan, G.L.: Nature 414, 425 (2001)
52. White, S.D.M., Frenk, C.S.: ApJ 379, 52 (1991)
53. Wise, M.W., Sarazin, C.L.: ApJ, 415, 58 (1993)