Bonn Galactic Intergalactic B Field

Galactic and intergalactic magnetic fields
Course astro 848
U.Klein
Argelander-Institut für Astronomie

Bonn
March 2013
Preface
This lecture has been created in spring 2010 and was first held in the summer term of that
year. The motivation for this course was threefold: first, magnetic fields in the interstellar
and intergalactic medium had been a central research topic at Bonn (Max-Planck-Institut für
Radioastronomie and Universität) for more than three decades, while the students involved
in such projects had little or no education on this topic. Second, this topic was an obvious
gap to be filled in an otherwise rather complete astrophysical curriculum. And last but not
least, a so-called Research Unit entitled Magnetisation of Interstellar and Intergalactic Media:
The Prospects of Low-Frequency Radio Observations was granted by the DFG, for research
projects of which interested students had to be recruited and educated.
The second chapter of the lecture presents a treatment of the main radiation process which
the bulk of our knowledge about magnetic fields in the diffuse interstellar and intergalactic
medium rests upon, namely synchrotron radiation. Since the energy loss mechanism via syn-
chrotron radiation has the same energy dependence as that due to inverse-Compton radiation,
also the latter has to be treated in the context of this lecture. Furthermore, the chapter also
deals with the thermal free-free radiation, as in case of galaxies it contaminates the total radio
emission that we utilise to explore the properties of magnetic fields, and therefore needs to
be treated thoroughly.
Since there is no dedicated and comprehensive textbook on this topic available1 , these
lecture notes have been furnished. The lecture has its emphasis on phenomena that have
been observed with various tools at different wavelengths, with theoretical concepts addressed
wherever necessary. It is meant to give an overview of our knowledge on various astrophysical
phenomena related to cosmic magnetism, with scales ranging from molecular clouds in star-
forming regions and supernova remnants in the Milky Way, to clusters of galaxies on the
largest scales. The lecture is not dealing with magnetic fields in condensed matter (planets,
stars, neutron stars).
The lecture also attempts to convey modern tools of studying polarisation at various
wavelengths, from which information about the strength and/or structure of magnetic fields
is extracted more or less directly, The emphasis is almost naturally on synchrotron radiation,
which manifests our most powerful tool to study cosmic magnetic fields on a wide range of
scales, including the largest ones.
While the lecture is embedded in the Master of Science in Astrophysics at the University
of Bonn, it may be also helpful for Ph.D. students interested in this subject should it be
relevant for their research.
I would like to thank Prof. Dr. Richard Wielebinski: it was him who ’whetted my ap-
petite’ on this field of research back in the late 1970’s.
March 4, 2011
1
The textbook High-Energy Astrophysics by M.S. Longair works out the basics of this subject.
Contents
1 Introduction 1
1.1 Magnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Magnetic force on a moving charge . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Magnetisation of matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Some history . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Significance of interstellar and intergalactical magnetic fields . . . . . . . . . 6
2 Continuum radiation processes 9

2.1 Radiation of an accelerated electron . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Free-free radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Situation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Radiation in the non-relativistic case . . . . . . . . . . . . . . . . . . . 11
2.2.3 Total radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Synchrotron radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.1 Radiation from a single electron . . . . . . . . . . . . . . . . . . . . . 21
2.3.2 Synchrotron radiation from relativistic electrons with an energy spectrum 29
2.3.3 Polarisation properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.3.4 Losses and particle lifetimes . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4 Inverse-Compton radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.5 Spinning dust grains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3 Diagnostics 47
3.1 Optical polarisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Synchrotron radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3 Faraday rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3.1 Rotation measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3.2 Depolarisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3.3 RM synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.4 Zeeman effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.5 Polarised emission from dust and molecules . . . . . . . . . . . . . . . . . . . 70
4 Milky Way 71
4.1 Diffuse ISM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.2 Molecular clouds and star-forming regions . . . . . . . . . . . . . . . . . . . . 75
4.2.1 Cloud stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2.2 Flux freezing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
i
ii CONTENTS
4.2.3 Ambipolar diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

4.3 Supernova remnants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.4 Acceleration of cosmic rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.5 UHECR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5 External galaxies 91
5.1 Galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.2 Spiral galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.3 Dwarf irregular galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.4 Elliptical galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.5 CR containment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.6 Galactic dynamo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6 Active galactic nuclei 109

6.1 The AGN Zoo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.2 Taxonomy of AGN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.3 Unified scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.4 Radio galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.4.1 FR II sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.4.2 Jet formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.4.3 FR I sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.5 Quasars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.6 Seyfert galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.7 Origin of magnetic fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
7 Intergalactic magnetic fields 127

7.1 Clusters of galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7.2 Radio haloes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
7.3 Rotation measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.4 Radio relics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7.5 Mini haloes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.6 Inverse-Compton emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.7 Magnetisation of the IGM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
8 Cosmological magnetic fields 141

Chapter 1
Introduction
1.1 Magnetism
Historically, the phenomenon of magnetism was recognised in the form lodestones, which
are natural magnetised pieces of iron ore. The term magnet in Greek meant ”stone from
Magnesia”, a part of ancient Greece where lodestones were found. When lodestones were
suspended such that they could turn they served as the first magnetic compasses. The earliest
records of this phenomenon are from Greece, India, and China around 2500 years ago. Apart
from iron, magnetism was exhibited in a natural state by cobalt and manganese, and by
many of their compounds. The phenomenon was coined ”magnetism” (derived from city of
Magnesia in Asia Minor, where they say the phenomenon was first recognised). The strength
of magnetic fields is given in units of Gauss, and a frequently used unit is 1 Tesla = 104 G. In
order to get a feeling for this magnetic force, a comparison with some weight may be helpful:
holding a weight of 1 kg in the earth’s gravitational field corresponds to force of 9.81 N.
Keeping apart two bar magnets (B = 50 G) that are separated by 0.4 cm requires this force.
Our earth can be considered as a huge mag-
net: a magnetised rod (i.e. a compass) suspended
to rotate freely finally aligns with the earth’s
magnetic field, with the north pole of the com-
pass pointing northwards, towards the geographic
north pole, which is actually the magnetic south
pole! The earth’s magnetic field (Fig. 1.1), which
protects us against high-energy (charged) parti-
cles, the cosmic rays (CRs, see Sect. 4) by trap-
ping them in the so-called van Allen radiation
belts. This magnetic field has an overall dipo-
lar morphology, with a field strength near to the
earth’s surface of BE ≈ 0.3 . . . 0.5 G. For com-
parison, a small iron magnet has 100 G, the mag-
net used NMRI has between 5000 and 70000 G.
Neutron stars represent the most extreme envi-
ronment in this respect: they have ≤ 1015 G on
their surface. Figure 1.1: Earth’s magnetic field.
Throughout this lecture, we use cgs units
1
2 CHAPTER 1. INTRODUCTION
(Gaussian system) as it is quite common in astrophysics. In the context of electro-magnetism,

this bears the advantage that we do not have to deal with the constants 0 and µ0 , so that
e.g. the energy density of magnetic fields is simply B 2 /8π (instead of B 2 /2µ0 in SI units).
The magnetic field of the sun is of cardinal significance
for many processes on its surface. For instance by heating
the corona to its temperature of ∼ 106 K via magnetic recon-
nection continuously accelerates the charged particles. The
sunspots mark locations of an enhanced (0.1 T) magnetic field
with specific polarity (N or S), which means that sunspots
are bound to occur in pairs, connected by magnetic loops
(Fig. 1.2). The polarity of the preceding and following bipo-
lar spots (in the sense of the sun’s rotation) is always the
same in one (northern) solar hemisphere, and is opposite in
the other (southern). This polarity is reversed every 22 years
(solar cycle). We shall briefly come back to the solar magnetic Figure 1.2: Magnetic loops on
field when discussing cosmic dynamos. the sun.
1.2 Magnetic force on a moving charge

A charged particle (e.g an electron) moving in a magnetic field experiences a centripetal force
that accelerates it perpendicular to its direction of motion (Fig. 1.3)
e
F~ = ~v × B,
~ (1.1)
c
where B ~ is the magnetic field strength, e the unit charge, c the speed of light, and v the
velocity of the particle.
This so-called Lorentz force forms a thread through this lec-
ture, since it implies that as soon as we deal with charged par-
ticles, it is this force that controls their motion and thereby
leads to a tight coupling between charged particles and mag-
netic fields. Since interstellar and intergalactic matter is always
ionised to some extent1 , the gas and the magnetic fields are al-
ways strongly coupled, hence can never be considered seperately.
This is a major challenge for any theoretical modelling of pro-
cesses connected with gas flows.
Equating (1.1) to the centrifugal force, one can calculate the
so-called radius of gyration rL , also called the Larmor radius. If
~ we have
~v ⊥B, Figure 1.3: Lorentz force.
m v2 B
= ev (1.2)
rL c
and hence
mvc
rL = . (1.3)
eB
1
Most of the gas in the universe is ionised, and hence is a plasma. In galaxies, the UV and X-ray photons
as well as the CRs keep the ionisation at some level, while in clusters of galaxies the gas is nearly completely
ionised due to the energy input by large-scale shock waves.
1.3. MAGNETISATION OF MATTER 3
With
v = ω rL , (1.4)
the resulting gyration or Larmor frequency is
e
ω= B, (1.5)
mc
1.3 Magnetisation of matter

Atoms and molecules may or may not possess a net magnetic dipole moment, depending on
their symmetry and on the relative orientation of their electronic orbits. Matter in bulk,
with the exception of ferromagnetic substances, does not exhibit any net magnetic moment,
owing to the random orientation of the atoms or molecules. However, when applying an
external magnetic field, the distortion of the electronic motion results in a net magnetisa-
tion. Depending on the constitution of the substance, we distinguish the following kinds of
magnetism occurring in in solid-state material (Fig. 1.4):
• Diamagnetism: This can be considered as a general property of any substance! When

applying an external magnetic field to matter, this produces a magnetic dipole moment
opposite to the one applied2 , which is hence repulsive. This magnetisation, called
”diamagnetism” is, however, masked by paramagnetism in most substances.
• Paramagnetism: Substances with unpaired electrons possess a permanent magnetic

dipole moment, associated with the angular momentum of the unpaired electrons. Ap-
plying an external magnetic field tends to align all magnetic moments along the external
field, which results in a magnetisation, called ”paramagnetism”. So here, the magneti-
sation is in the same direction as the external magnetic field. It is much stronger than
diamagnetism and is attractive rather than repelling.
• Ferromagnetism: This form of magnetism implies permanent magnetisation, due to

the tendency of magnetic moments to align under their mutual magnets. It results
in so-called ”Weiss domains”, their sizes being of order 10 µm. . .0.2 mm. This is the
characteristic property of iron (Fe).
• Antiferromagnetism: In some substances, the electronic spins are antiparallel, which

results in a vanishing magnetisation. Such substances are, for instance, MnO, FeO,
CoC, and NiO.
• Ferrimagnetism: This kind of magnetism is similar to that of antiferromagnetism, but

the atomic or ionic magnetic moments in one direction are different from those in the
other. These substances are called ”ferrites” and are compounds of the form M OF e2 O3 ,
where M stands for Mn, Co, NI, CU, Mg, Zn, Cd etc. If M = Fe, then the compound
F e3 O4 , or magnetite, results.
2 ~
In a strong laboratory B-field (≤ 15 Tesla), strongly diamagnetic substances can levitate!
Concluding, we find that magnetism in solid-state matter is due to atomic structure and
electronic orbits. In astrophysical plasmas, magnetism is due to large-scale electric currents,
producing in particular the dynamo action (see Sect. 5.6). It is at work in the earth, in the
sun, in planets, in the interstellar medium (ISM), and even in the intergalactic medium via
hydrodynamic flows.
Figure 1.4: Different kinds of magnetism in solid bodies.
In every day’s life, ferrites have a wide range of applications, as they have little or no
electric conductivity:
- iron cores of transformers (low eddy-current losses)
- high-frequency electronic components (e.g. directional couplers)
- noise suppression chokes (”ferrite head” in computer cables)
- erase heads in tape recorders, computer disks
- coating of magnetic tapes
- loudspeaker magnets
1.4 Some history

Before diving into the subject of cosmic magnetism, it may be worth to briefly review the
history of cosmic magnetism by summarizing the landmarks of the notion that magnetic fields
are ubiquitously present in the universe 3 . Prior to World-War II it was noted (e.g. Alfvén,
1937) that magnetic fields in sun and stars or in interstellar space could have a strength
and scale sufficient to accelerate charged particles up to energies as observed in cosmic-ray
experiments, revealing protons, electrons, and α particles.
1946 The first discrete radio source Cyg A was discovered by Hey, Parsons & Phillips. They
noticed a fluctuation of the signal, which turned out to be low-frequency scintilla-
tion. Cyg A is the proto-typical radio galaxy emitting intense synchrotron radiation
(see Sect. 6.4).
3
The very first proof of the existence of magnetic fields outside the earth was made by G.E. Hale, who in
1908 discovered magnetic fields in the sunspots via the Zeeman effect.
1.4. SOME HISTORY 5
1948 Bolton & Stanley, using their ”sea interferometer”, determined the angular size of Cyg A
to be 80 . Smith measured its precise position in 1951, facilitating an optical identification
with the Palomar 200 inch telescope by Baade. The redshift implied a distance of
230 Mpc, resulting in a stunning radio luminosity of Lradio ≈ 1045 erg s−1 ! Later on,
interferometric observations revealed a central source, two jets and lobes. Such objects
with Lradio ≈ (103 ...104 ) × LM W were termed ”radio galaxies”.
1948 The detection of the Crab Nebula in the radio regime was reported by Bolton. Called
Tau A, it was identified with the supernova explosion reported by Chinese astronomers in
1054. Such supernova remnants also produce relativistic particles emitting synchroton
radiation (Sect.4.3).
1949 Alfvén, Richtmyer and Teller proposed that cosmic rays have a solar origin and are
confined to solar surroundings by an interplanetary magnetic field. With the Larmor
radius rL given by Eqn. 1.3, the particles with an inferred energy Emax ≈ 1014 eV must
have Larmor radii of up to rL = 3 · 1017 cm for a magnetic field strength B ≈ 10−6 G,
inferred from other evidence. For particles with E ≈ 1020 eV, such a µG field would keep
them within the Galactic disk, because the Larmor radius has a size scale comparable
to that of the thickness of the Galactic disk. To date, we know that the highest-energy
particles (E ≥ 1021 eV) must come from outside the Milky Way, simply because there
aren’t any conceivable acceleration mechanisms within the Galaxy capable of producing
such energies.
1949 Fermi proposed that CRs fill the whole Galaxy, so they would not be confined to the
solar system. He suspected that a magnetic field exists on still larger scales. He was
aware of the high conductivity of the ISM and of its moving inhomogeneties that could
serve as efficient accelerators (Sect. 4.4).
1949 Hiltner delivered impressive evidence for the existence of interstellar magnetic fields,
when he discovered the polarisation of starlight. This linear polarisation is explained in
terms of the so-called Davis-Greenstein mechanism (Sect. 3.1).
1950 The galactic nonthermal radio emission was proposed to be ”magneto-bremsstrahlung”

(synchrotron radiation, see Sect. 2.3) by Alfvén & Herlofson, by Kiepenheuer, and later
on also by Ginzburg (1953) and Shklovsky (1956). The discovery of linearly polarised
radio emission in the Milky Way strongly supported this view (Westerhout et al., 1962;
Wielebinski et al. 1962). All of this evidence suggested the presence of a ubiquitous
magnetic field in the Galaxy.
1955 Imprints of magnetic fields were noticed in the morphologies of nebulae in the Milky
Way, which exhibit filamentary structures. Shajn suggested that the elongation of dif-
fuse nebulae (along galactic plane) is connected with a large-scale interstellar magnetic
field.
1963 Maarten Schmidt identified otically the radio source 3C 2734 with star-like object at
z = 0.158, DL = 750 Mpc (ΛCDM). The name ”QSO” (quasi-stellar opject, also:
”quasar”) was coined. Later on, a whole ”zoo” of active galactic nuclei (AGN) was
revealed (radio galaxies, Seyfert galaxies, QSO, BL Lac, ...; see Sect. 6.1).
4
The 3C catalogue, meaning the Third Cambridge Catalogue of Radio Sources was published in 1959
1964 The first measurement of Faraday rotation was reported by Morris & Berge and (later
on) by Berge & Seielstad (1967) against extra-galactic polarised radio sources. These
measurements were followed by measurements of pulsars (Manchester 1972, 1974). A
magnetic-field strength of BISM ≈ 2 . . . 3 µG was inferred for the ordered field component
(there is also a random one that does not contribute to RM, See Sect. 3.3).
1968 The first detection of the Zeeman splitting in clouds of neutral hydrogen (HI) was made
by Verschuur. This discovery was of cardinal importance to directly prove the existence
of magnetic fields in the ISM. The problem is, however, that the splitting of the lines
in frequency is minute (see Sect. 3.4). Hence this discovery came late.
1971 The first models of a galactic dynamo were worked out by Parker, and by Vainstein
& Ruzmaikin. The concept of a hydromagnetic dynamo to explain the origin and
maintenance of magnetic fields of the sun and earth actually goes back to Larmor
(1919).
1978 This year marks the first detection of polarised radio emission from an external spiral
galaxy, viz. M 31, by Beck et al. These activities have developed to a major research
field (not only) in Bonn.
1.5 Significance of interstellar and intergalactical magnetic

fields
Magnetic fields may have a strong influence on the various astrophysical processes going on
in the interstellar and intergalactic medium. Its relative importance is essentially determined
by the energy density
B2
um = (1.6)
8π
Below, we summarise the most important aspects in which magnetic fields may be signif-
icant and may even play an active role. For instance, Eqn. (1.6) tells us that magnetic fields
cannot have any influence on the dynamics of galaxies, while on small scales (cloud collapse
and subsequent star formation) this is rather different.
• Coming back to the very first inferences from the notion of the presence of magnetic
fields in the ISM, it is obvious that our understanding of the strength and morphology
of magnetic fields is of paramount importance for our understanding of CR transport.
In particular, we need to know whether they arrive isotropically as suggested by ob-
servations of CRs at energies up to ≈ 1014 eV). What about ultra-high energy CRs
(UHECR)? Is there any preferred direction of their arrival, possibly connected with any
nearby AGN?
• Some more general and fundamental questions to be addressed are: What is the origin
of magnetic fields in the ISM and intra-cluster or intergalactic medium (ICM/IGM)?
Were they initially injected by AGN or by starburst (dwarf?) galaxies, together with
the relativistic particles? What kind of mechanisms maintain and amplifies magnetic
fields?
1.5. SIGNIFICANCE OF INTERSTELLAR AND INTERGALACTICAL MAGNETIC FIELDS7
• One of the most important issues that will be a thread over a large part of this lecture
is the role of magnetic fields in galactic evolution. Surely, magnetic fields take an active
role in star formation, as they serve to transport angular momentum outwards during
the cloud collapse; otherwise, star formation would be strongly inhibited. When stars
cease their lifes, they exert a strong feed-back on the ISM via galactic outflows or winds.
As we meanwhile know, magnetic fields are being dragged along with such outflows.
• Some more specific questions related to large-scale phenomena in clusters of galaxies

that are hitherto not fully understood: what is the nature of radio relics (cosmological
shocks) in galaxy clusters? What is the nature of radio haloes in galaxy clusters? Are
they produced by secondary electrons, or rather by shocks produced in galactic wakes?
Are intergalactic filaments magnetised?
• Perhaps the most fundamental question pertains to the origin of magnetic fields: have
there been primordial seed fields? Did they come into existence soon after the Big Bang,
before the recombination era, or in the course of cosmological structure formation?
The hope is that this course conveys basic knowledge about this field of research to the
students, and raise their interest in this. As far as contemporary knowledge allows, the lecture
shall try to answer some of the questions addressed above.
Chapter 2
Continuum radiation processes
2.1 Radiation of an accelerated electron

After the very first inference that nature of the Galactic
low-frequency radiation must be synchrotron radiation
by relativistic electrons moving in a magnetic field for
which there is other independent evidence (Sect. 1.4),
this radiation turns out to be the main diagnostic tool
to trace that magnetic field in interstellar and inter-
galactic space, by virtue of its radiation spectrum and
polarisation characteristics. It is therefore mandatory
to work out these spectral and polarisation properties
from first principles of electro-magnetic theory. The
radiation is naturally measured at radio frequencies as
shall be worked out below. However, this is not the
only radiation process measured at centrimetric and de-
cametric wavelengths, since the radiation of the ionised
gas present in HII regions and in the diffuse ionised
medium of the Milky Way and external galaxies con-
taminates the overall radio emission. Hence, we also Figure 2.1: Geometry for a moving
need to understand the properties of this thermal emis- charged particle as seen from some
sion, so we have to explore it here as well. We shall point P .
therefore in what follows work out the continuum radi-
ation spectra of both, the thermal (free-free) and nonthermal (synchrotron) emission. The
former is called ’thermal’, as it results from an ensemble of particles that have a Maxwellian
energy distribution, while the energy distribution of the latter follows a power-law distribu-
tion.
Let us consider the electric field of an accelerated electron, with velocity ~v and acceleration
˙~v as seen by an observer from some point P (Fig. 2.1). According to electro-magnetic theory1
it is given by
~ ~˙
~ = e · ~n × [(~n − β) × β] ,
E (2.1)
c R (1 − cos θ · β)3
1
see, e.g. J.D. Jackson, Classical Electrodynamics, Chapt. 14
9
10 CHAPTER 2. CONTINUUM RADIATION PROCESSES
where ~n is the unit vector pointing from the particle towards the observer and
~v ˙ ~v˙
β~ = , β~ = (2.2)
c c
The flux of radiation is given by the Poynting vector
S~ = c ·E ~ ×B ~ = c · |E ~ 2 | · ~n (2.3)
4π 4π
The power radiated into a unit solid angle per unit frequency and unit time is
2 ~ ~˙ 2
dP (t) ~ · (1 − β cos θ) R2 = e · |~n × [(~n − β) × β]|
= |S| (2.4)
dΩ 4πc (1 − β · cos θ)5
This equation will be used for the case of thermal radiation, where β 1, and non-
thermal radiation, where β ≤ 1 or γ 1, by integrating the radiated power over the total
4 π solid angle of the sphere. The angle θ = ∠(~v , ~n), hence
cos θ = ~n · β~ (2.5)
R is the distance of point P between the electron and the observer. When measuring
flux densities of a radio source, we can calculate its radio power or luminosity once we can
determine this distance using standard astronomical techniques. Hence R is not relevant in
the derivations that we shall work out below, and is hence just a matter of distance calibration,
by converting for instance flux density Sν to power Pν via
Pν = 4 π R2 Sν . (2.6)
2.2 Free-free radiation

2.2.1 Situation
In an ionised (HII) region, the hydrogen (and he-
lium) atoms are ionised, with the free electrons mov-
ing on hyperbolic orbits (s.b.) about the protons (or
ionised helium atoms or α particles). That’s why
the resulting radiation is termed ’free-free radiation’.
The electrons possess thermal velocities, according
to the temperature of the electron gas. In order to
work out the radiated power of such a region as a
function of frequency, we first need to compute the
emitted power of a single accelerated electron.
This power is determined by the physical condi-
tions. Consider a plasma (HII region) with temper-
ature T = 104 K, which corresponds to ≈ 1 eV. Let
us assume that we are dealing mostly with protons
and electrons (there will be some He ions, for which Figure 2.2: Trajectory of a thermal
one can correct the results later on). The number electron in a plasma.
densities, which are roughly identical for the electrons and protons, i.e. ne = np may range
from 0.03 cm−3 in the general diffuse ISM, to 103 cm−3 in the centres of HII regions such as
in the Orion Nebula (the radio source also called ’Ori A’).
2.2. FREE-FREE RADIATION 11
Due to their hyperbolic orbits about the pro-

tons the electrons permanently experience an ac-
celeration, the main component of which is in
the x-direction (Fig. 2.2). The motion is deter-
mined by the Coulomb law (me : electron mass,
Z: charge number, e: unit charge)
Z e2 ~r
~v˙ (t) = − · , (2.7)
me r 3
which has components (Fig. 2.3)
Z e2
ẍ = · cos φ (2.8)
me r 2 Figure 2.3: Acceleration of an electron.
and
Z e2
ÿ = · sin φ (2.9)
me r2
2.2.2 Radiation in the non-relativistic

case
We consider the non-relativistic case, for which β ≈ 0, i.e.
(1 − β cos θ) = 1. Hence, we have
dP e2 |~n × [(~n − β) ~˙ 2
~ × β]|
= · (2.10)
dΩ 4πc (1 − β · cos θ)5
e2
= · v̇ 2 (t) sin2 φ0 , (2.11)
4 π c3
and the integration over the solid angle yields the power
radiated by a single electron as a function of time:
Z2π Zπ
dP 2 e2 2
P (t) = · dΩ = · v̇ (t) (2.12)
dΩ 3 c3 Figure 2.4: Definition of an-
0 0
gles.
since
Z2π Zπ
8π
sin3 φ0 dφ0 dθ = . (2.13)
3
0 0
~˙
At this point, the various angles used above need to be explained once more: φ0 = ∠(~n, β),
~ φ = ∠(~x, β)
θ = ∠(~n, β), ~˙ (see Figs. 2.4 and 2.3).
We now have P (t), but we need to know the power as a function of frequency, P (ν), or
P (ω), from which we can then derive the frequency spectrum of a single particle. The usual
procedure to facilitate this conversion is to perform a Fourier analysis, or to calculate the

Fourier transform of the time-dependent quantity. We can thus express the time-dependent
acceleration ẍ(t) in terms of its frequency-dependent Fourier transform
Z∞
ẍ(t) = C(ω) · cos(ω t) · dω, (2.14)
0
and vice versa
+∞
Z
1
C(ω) = ẍ(t) · cos ω t · dt. (2.15)
π
−∞
The function ẍ(t) looks like a ”spike” as we have seen

(Fig. 2.3). If it were a pulse (a δ-function), then it would
have an infinitely wide frequency spectrum with constant Figure 2.5: Relation between
power. Its finite width, however, gives rise to a finite fre- ẍ(t) and C(ω).
quency spectrum, dropping to zero at some critical frequency (Fig. 2.5). We can now evaluate
C(ω) by performing the integration (2.15). We can utilise the modified Coulomb law to re-
place the integration over time t by an integration over the angle φ, which turns out to be
trivial. We rewrite Coulomb’s law, neglecting the y-term of the acceleration:
Z e2
ẍ = · cos φ (2.16)
me r 2
The particles move a distance v · dt during an infinitesimally small time step dt
dφ
v · dt = d(p · tan φ) = p , (2.17)
cos2 φ
where we have introduced the impact parameter p. The above equation holds because p
doesn’t change significantly while ẍ is large, i.e. dp = 0. We can now write the time step dt
in terms of the angular increment dφ,
p dφ
dt = · (2.18)
v cos2 φ
The radiation will be significant only for the time interval
p p
− < t < (2.19)
v v
for which ωt 1, hence cos ωt ≈ 1. With p = r · cos φ, and noting that the integration
over φ can be reduced to the range between -90◦ and +90◦ , we have
π
Z+ 2
1 Z e2 · cos3 φ p dφ
C(ω) = 2
· · (2.20)
π me p v cos2 φ
− π2
π
Z+ 2
Z e2
= · cos φ dφ (2.21)
π me p v
− π2
2 Z e2
= (2.22)
π me p v
We have managed to invert the problem such

that the function C does not depend on ω any-
more. It has rather become a step function that
approximates the real behaviour as sketched in
Fig. 2.6. At this point, Parseval’s theorem comes
into play, as in the end we need to compute the
radiated power ! As P (t) ∼ ẍ2 (Eqn. 2.11),
+∞
Z Z∞
v̇ (t) dt = π C 2 (ω) dω
2
(2.23)
−∞ 0 Figure 2.6: Approximation of C(ω).
+∞
Z Z∞
P (t) dt = P (ω) dω (2.24)
−∞ 0
Using (2.11) this yields
2 e2
P (ω) dω = P (t) dt = · π C 2 (ω) dω (2.25)
3 c3
and, plugging in (2.22) from above, we obtain
16 e6 Z 2
P (ω) dω = dν. (2.26)
3 m2e p2 v 2 c3
Recalling that P(ω) ≈ 0 for ν > v/p we have
16 e6 Z 2 v
P (ν) dν = dν for ν≤ (2.27)
3 m2e p2 v 2 c3 p
v
P (ν) dν = 0 for ν≥ (2.28)
p
The latter of the above two equations tells us that only for infinitely small values of the
impact parameter p would the frequency spectrum be constant out to infinity. We can make
an estimate of the expected maximum frequency. Let us assume that we have volume densities
ne = ni = 103 cm−3 of the plasma, which corresponds to the conditions in the centre of an
HII region. Let the electron temperature (definition s.b.) Te = 104 K, which is typical. As
we shall see below, this temperature corresponds to a mean velocity of the radiating electrons
of v̄ = 600 km s−1 . The mean density implies a mean separation of the ions of d¯ = 0.1 cm,
hence a mean impact parameter p̄ ≤ 0.05 cm, resulting in a frequency of ν ≈ 1.2 GHz. This
is measured then in the radio regime!
2.2.3 Total radiation

In order to calculate the power emitted by a certain ionised volume, we now consider the
emission of an ensemble of thermal electrons. Let ni be the number density of ions, ne that
of the electrons.
Let an ensemble of electrons be char-

acterised by collision parameters between p
and p + dp (Fig. 2.7). Then the number of
Coloumb collisions with this range of im-
pact parameters (p , p + dp) per second of
time that an electron with velocity v will
experience is equal to the number of ions
within a cylindrical ring with length (per
unit time) v and, respectively, inner and Figure 2.7: Ensemble of protons, through which
outer radii p and p + dp. This produces electrons with the impact parameter in the
Nc = v · 2 πp dp · ni collisions each second. range (p , p + dp) are moving with speed v.
The number of collisions per cm3 involving
electrons with velocities between v and v + dv then is
dN (v, p) = (v · 2πp · dp) · ni ne · f (v) dv cm−3 s−1 (2.29)
Here, f (v) is the Maxwellian velocity distribution

3
me 2 me v 2
f (v) dv = 4π e− 2 k Te v 2 dv (2.30)
2 π k Te
where Te is the so-called electron temperature. The total energy emitted per second, per
Hz and per cm3 is given by the emission coefficient ν , integrated over 4 π steradian:
Zp2 Z∞
4 π ν = P (v, p) dN (v, p) (2.31)
p1 0
Zp2 Z∞
16 e6 Z 2
= · ni ne · f (v) · 2 π p dp v dv (2.32)
3 c3 m2e p2 v 2
p1 0
Z∞ Zp2
32 π e6 Z 2 ni ne f (v) dp
= · dv · (2.33)
3 c3 m2e v p
0 p1
so that
Z∞ Zp2
8 e 6 Z 2 · ni ne f (v) dp
ν = · dv · (2.34)
3 c3 m2e v p
0 p1
Note that the dimension of the emission coefficient is [ν ] = erg s−1 cm−3 Hz−1 sr−1 . The
first integral is nothing but the mean reciprocal speed of the electrons, which upon plugging
in the Maxwellian distribution (2.30) becomes
Z∞ r
f (v) 1 2 me
dv = = , (2.35)
v v π k Te
0
where Te is the electron temperature, with a typ-

ical value of Te = 104 K. The second integral is called
’Gaunt factor’ for the free-free radiation,
Zp2
dp p2
= ln = gf f . (2.36)
p p1
p1
This finally leads to an emission coefficient of the

form
8 e 6 Z 2 ni ne
r
2
ν = 3/2
· · gf f . (2.37)
3
c me π k Te
We now need to discuss the most likely limits p1
p2 for the impact parameter, which are difficult to
estimate. Figure 2.8: Geometry of the hyperbolic
orbit of the electron.
Lower bound
The lower bound can be estimated in the following way. The electrons move on hyperbolic
paths (Fig. 2.8), for which energy conservation holds, i.e.
1 Z e2 1
E = Ekin + Epot = me v 2 − 2
= me v∞ >0 (2.38)
2 r 2
! Z e2
= (2.39)
2a
We can now simply set the minimum impact parameter p1 , by considering the relation
between the deflection angle and the physical ingredients of the particle (see textbooks on
classical mechanics):
χ a Z e2
tan = = 2
(2.40)
2 b b me v∞
The maximum deflection angle relevant here is χmax = 90◦ . Setting the corresponding
minimum impact parameter equal to b, we obtain
χmax π Z e2
tan = tan = 2
(2.41)
2 4 p1 me v∞
From kinetic gas theory, we know that

1 3
· me v 2 = · k Te (2.42)
2 2
so that we can solve for p1 to finally arrive at
Z e2
p1 = . (2.43)
3 k Te
We could think of still smaller impact parameters, hence even lower bounds on p, such
as the de-Broglie wavelength, but we would then leave the regime of classical mechanics and
would have to apply relativistic corrections. For normal HII regions, this is not necessary, as
their electron temperature is too low. Such considerations would only have to be made in
case of very hot, X-ray-emitting gas.
Upper bound
In order to calculate the upper bound p2 , we can argue that there must be a minimum
deflection angle causing significant radiation (otherwise, there would not be sufficient acceler-
ation to produce any radiation). Since this will correspond to the largest collision parameter,
we can relate p2 to the emitted frequency, corresponding to the inverse collision time:
1 hvi
2πν = = (2.44)
τcoll p2
so that
hvi
p2 = (2.45)
2πν
This time, we need to evaluate the mean speed of the electrons possessing a Maxwellian
velocity distribution, which upon plugging in (2.30) becomes
Z∞ r
8 k Te
hvi = f (v) v dv = , (2.46)
π me
0
and hence
r
1 8 k Te
p2 = · , (2.47)
2πν π me
Thus, the Gaunt factor for free-free radiation becomes
r
3 k Te 1 8 k Te
gf f = ln · (2.48)
Z e2 2 π ν π me
" 1.5 #
−1 Te ν −1
= 12.5 + ln Z · · ≈ 10 . . . 17, (2.49)
104 K GHz
which weakly depends on frequency and on the electron temperature. We now utilise
Kirchhoff’s law to obtain absorption coefficent. This law relates the emission and absorption
coefficients ν and κν via the so-called source function
ν
Sν (Te ) = (2.50)
κν
In case of local thermo-dynamical equilibrium, the source function equals Planck’s law,
2 h · ν3 1
Bν (T ) = 2
· hν (2.51)
c e k Te − 1
2 h ν2 hν
≈ · k Te for 1. (2.52)
c2 k Te
In the radio domain, the Rayleigh-Jeans approximation (2.52) always holds, as the reader
may easily verify. As an example of the situation discussed here, let us plug in an observing
frequency ν = 700 MHz and an electron temperature Te = 104 K, in which case we have
h ν/k Te ≈ 3 · 10−6 . The absorption coefficient κν , defined per unit path length, can now be
computed from (2.48) and (2.52) to yield
r
2 4 e6 Z 2 ni ne
κν = · gf f (2.53)
π 3 c (me k Te ) 32 ν 2
n n ν −2 T − 23
−3 i e e
= 8.77 10 cm−1 (2.54)
cm−3 cm−3 Hz K
Next we compute the optical depth, which is a dimensionless quantity obtained by inte-
grating the absorption coefficient over the total line-of-sight:
Zs0
τν = κν ds (2.55)
0
so that
ν −2 T − 32 Zs0
−3 e ni ne
τν = 8.77 10 gf f · ds. (2.56)
Hz K cm−3 cm−3
0
In general the plasma is neutral, as there is zero net charge over the whole volume, i.e.
we have ni = ne . At this point it is convenient to define a quantity called emission measure,
usually designated as EM , in the following way
Zs0
n e 2 ds
EM = pc cm−6 . (2.57)
cm−3 pc
0
With this definition, which makes use of convenient units, we can now express the optical
depth in the form
−1.5
−2 EM ν −2 Te
τν = 3.01 · 10 · · · · gf f , (2.58)
106 pc cm−6 GHz 104 K
again with convenient units. As an example, the Orion Nebula has an emission measure
EM ≈ 106 pc cm−6 in its central region. The optical depth for free-free radiation becomes
unity at a frequency of ν ≈ 0.6 GHz (with a Gaunt factor gef f = 12.3 and an electron
temperature Te = 104 K).
In order to account for the optical thickness, we need to employ a calculation of the
radiative transfer. It allows us to calculate the brightness or intensity Iν of a source, given
its optical depth2 .
Iν = Bν (Te ) · (1 − e−τν ) (2.59)
In order to work out the spectral behaviour of the radiation (important to judge its
contribution to the total radio emission), we consider the two extreme cases, namely the
optically thick and the thin one. In the first case, we have
τν 1
so that
2 h ν2
Iν = Bν (Te ) = · k Te . (2.60)
c2
The resulting emission spectrum follows exactly Planck’s law in the Rayleigh-Jeans limit,
i.e. we have a power law as a function of frequency with slope +2. This spectrum is ”feature-
less” and only depends on the electron temperature, with no other properties present! In the
second case, we have
τν 1
−0.5
−23 EM Te
Iν = τν · Bν (Te ) = 8.29 · 10 · · · gf f . (2.61)
pc cm−6 104 K
This produces a radiation spectrum that varies only very slowly with frequency, dropping
with increasing frequency Iν ∼ ν −0.1 , owing to the fequency dependence of the Gaunt factor.
The above intensities or brightnesses have units of erg s−1 cm−2 Hz−1 sr−1 . In radio astron-
omy, the term ”brightness temperature” is commonly used. It is defined via the Rayleigh-Jeans
approximation, and the temperature is therfore often referred to as the brightness temperature
Tb .
2 k ν2 c2
Iν := · Tb ⇐⇒ Tb = · Iν (2.62)
c2 2 k · ν2
We can now simply convert brightness into brightness temperature for the thermal free-
free emission, using the above Rayleigh-Jeans approximation so that
τν 1 : Iν = Bν (Te ) =⇒ Tν = Te (2.63)
τν 1 : Iν = τν · Bν =⇒ Tν = τν · Te . (2.64)
2
Eqn. (2.59) only describes the brightness emitted by a volume with a certain emissivity, given by Kirchhoff’s
law. In general, one would also have to account for the attenuation by any material located in the foreground.
We finally arrive at an important result,

namely for the frequency dependence of the
brightness and the brightness temperature (see
Fig. 2.9). In the first case, the intensity exhibits
the slope of the Rayleigh-Jeans law, and the ob-
served brightness temperature equals the phys-
ical temperature, while in the second case the
brightness drops slowly as a function of frequency,
and the brightness temperature drops off slightly
steeper than ν −2 . The conversion of the so-called
antenna temperature into brightness temperature
can be found in textbooks on radio astronomy3 :
Iν ∼ ν 2

for τν 1 (2.65)
Tb = Te
Iν ∼ ν −0.1

for τν 1 (2.66)
Tb ∼ ν −2.1
Measuring Iν or Tb over a sufficient frequency

range allows to determine the emission measure,
which is the dominant factor in the optical depth,

1/2
which depends on the physical parameter n2e ,
i.e. the variance of the number density of the Figure 2.9: Continuum spectra of free-free
thermal electrons in an astrophysical plasma. radiation.
Apart from the characteristic continuum spec-
trum described above, the thermal free-free radi-
ation is unpolarised, since the trajectories of the
electrons are randomly oriented. This, together
with the frequency dependence of the radiation,
is an important feature by which we can discern the thermal (free-free) and nonthermal (syn-
chrotron) radiation. Note that the dependence of the emission measure on n2e is in the end
due to the collisions occurring in the plasma!
In Fig. 2.10 (top) the thermal free-free radiation from the Large Magellanic Cloud is
juxtaposed to its Hα emission. The HII regions clearly show up in both images and are traced
nearly identically, as the radio continuum shows the radiation at a rather high frequency,
at which the free-free radiation dominates over the synchrotron. In Fig. 2.10 (middle) the
thermal radiation emitted by the central region of the Orion Nebula is shown, with the Balmer-
line emission, the radio free-free continuum, and the radio continuum spectrum. Fig. 2.10
(bottom) shows the imprint of thermal absorption in the starburst galaxy M 82, seen as a
’cavity’ in the overall nonthermal synchrotron radiation at 408 MHz. This ionised bubble
marks the footpoint of the outflow of hot gas seen in the X-ray emission of this galaxy.
3
See also the lecture notes on Radio astronomy: tools, applications and impacts, course astro 841.
Figure 2.10: Upper row: The Large Magellanic Cloud in Hα (left), and at 8.6 GHz (right).
Middle row: Centre of the Orion Nebula in Hα (left), contours of thermal free-free radio emis-
sion, superimposed onto an optical image (right), with the radio continuum spectrum below
it, showing the low-frequency turnover due to free-free absorption. Bottom row: Imprint of
thermal absorption in M 82, seen as a ’cavity’ in the overall nonthermal synchrotron radiation
at 408 MHz.
2.3. SYNCHROTRON RADIATION 21
2.3 Synchrotron radiation

As stated at the beginning of this chapter, synchrotron radiation is the main diagnostic tool to
trace magnetic fields in the ISM and IGM. It is produced by the relativistic electrons that have
been energized in the turbulent ISM and IGM via shock waves, notably caused by supernova
explosions in the former case, and by active galactic nuclei (AGN), by galactic wakes and by
merging of galaxy clusters in the latter. While moving at relativistic speeds, the particles
are subject to the Lorentz force, which forces them into helical paths. The corrsponding
acceleration makes them radiate synchrotron radiation, which has a characteristic frequency
spectrum and is partially polarised. In what follows, we shall work out these properties on
the basis of electro-magnetic theory.
2.3.1 Radiation from a single electron

As in the preceding section, we start out from the radiation by a single particle, this time
accounting for its relativistic motion and energy. The power radiated into a unit solid angle
per unit frequency and unit time is, as we have seen from (2.4),
dP e2 |~n × [(~n − β) ~˙ 2
~ × β]|
= · (2.67)
dΩ 4πc ~ 5
(1 − ~n · β)
When dealing with relativistic particles, we have to

˙
distinguish the linear accelerator, in which case β~ k
β~ and the acceleration is caused by an electric field,
and the transverse accelerator, in which case we have
~˙ β~ and the acceleration is due to a magnetic field.
β⊥
The former is used in laboratory accelerators such as
Tevatron or the planned ILC , while the latter is found
in the form of cyclotrons and synchrotrons, e.g. the
LHC , and in the ISM and IGM. We briefly consider
the linear accelerator, which accelerates the particles in
an electric field. In this case, we have
dP e2 v˙2 sin θ
= · , (2.68)
dΩ 4 π c (1 − β cos θ)5
3
and the radiation pattern has a strong dependence

on the angle θ and on the particle speed. Fig. 2.11
shows its two-dimensional illustration for β = 0.8, cor- Figure 2.11: Radiation pattern
responding to γ = 1.67. This is mildly relativistic. In of the linearly accelerated electron
case of β = 0.999 (γ = 22.4), the radiation pattern (β = 0.8).
would stretch out a stunning 10,500 km on the scale of
Fig. 2.11, nearly the diameter of the earth! The reason is the very strong dependence of the
maximum power radiated into the forward direction,
dP
(θmax ) ∼ γ 8 , (2.69)
dΩ
where
1 p
cos θmax = · 1 + 15 β 2 − 1 (2.70)
3β
and
1
θmax = . (2.71)
2γ
The strong dependence on the Lorentz factor is called ’relativistic boosting’ or ‘beaming’,
meaning that a charged particle moving with relativistic speed emits essentially its whole
radiation in the forward direction. In order to obtain the total radiation of the relativistic
particle, we integrate over the solid angle again and obtain
Z2π Zπ
dP 2 e2 v̇ 2
P (t) = dΩ = · 3 · γ 6 (2.72)
dΩ 3 c
0 0
where
1
γ=p (2.73)
1 − β2
is the Lorentz factor. We realise the cardinal difference to the non-relativistic case (2.12).
In both cases, the total radiated power of a single particle depends on the square of the
acceleration, v̇ 2 . In addition, there is the very strong dependence on the Lorentz factor γ.
Figure 2.12: Illustration of the various angles used in describing the transverse acceleration
of a relativistic electron in a magnetic field.
We now turn to the transverse accelerator , which is the more important process in the
context of this lecture. In Fig. 2.12 the geometry of a relativistic particle moving in a magnetic
~˙ φ is the angle measured along a circle
~ φ0 = ∠(~n, β),
field is illustrated. Here, θ = ∠(~n, β),
perpendicular to ~v , and θ is the angle measured in the (~v , ~v˙ )-plane. We are considering
particles moving in the interstellar magnetic field, the trajectory inclined by angle χ w.r.t.
the magnetic field vector. We refer to this angle as the ’pitch angle’. The charged particles
~ β~˙ and (2.4) becomes
are subject to the Lorentz force. This implies that β⊥
2
sin θ cos φ2
dP e2 v̇ 2 1 − γ 2 (1−β cos θ)2
= · . (2.74)
dΩ 4 π c3 (1 − β cos θ)3
As in case of the linear accelerator, the relativistic

motion causes a relativistic aberration of the radiation
of the charged particle, i.e. a strong distortion of the
radiation pattern. In case of the tranverse accelerator,
this relativistic beaming produces a radiation pattern
as illustrated in Fig. 2.13. Its main lobe can be shown to
have a half-power width4 that is inversely proportional
to the Lorentz factor 1/γ at half-maximum:
1 m0 c2
θHP ≈ = . (2.75)
γ E2
For example, consider a particle with energy E =

1 GeV, which as we shall see below is the typical en-
ergy of relativistic electrons in the ISM producing syn-
chrotron radiation at GHz frequencies. Such a particle
has γ = 2000 (β = 0.99999975). Hence, the opening
angle of its radiation pattern has θHP = 1.0 7. As in case
of the transverse accelerator, the radiated power has a
strong dependence on the Lorentz factor:
dP
∼ v̇ 2 γ 6 , (2.76)
dΩ
and
Z2π Zπ
dP
P (t) = · dΩ ∼ v̇ 2 γ 4 . (2.77)
dΩ
0 0 Figure 2.13: Radiation pattern of
A particle with γ = 2000 has a radiation pattern the transversely accelerated electron
the maximum of which stretches out to some 28 light (β = 0.8).
years on the scale of Fig. 2.13!
Apart from relativistic aberration, we have to apply the laws of special relativity in the
transition from the reference frame of the particle to that of the observer, which turns out
to significantly change the time dependence of the radiation seen in the observer’s frame of
reference. Before we look at that, we first have to work out the so-called Larmor circle, i.e.
the circle on which the charged particle is moving about the magnetic field (see Fig. 2.14,
left). We start out from the equation of motion, given by the Lorentz force:
e
m ~v˙ = m · (~v × ω ~
~ L ) = − · (~v × B) (2.78)
c
Here, m is the mass of the particle (electron in our case), ωL is the Larmor frequency, ~v
~ the vector of its magnetic field strength. In case of χ = 90◦ i.e. a particle
its velocity and B
motion perpendicular to the magnetic field the above equation can be rewritten in the form
v2 e
m ωL2 rL = m · = ·vB (2.79)
rL c
4
The half-power width is the angular width of the radiation pattern at which the power has dropped to
half its maximum value.
v eB
m = (2.80)
rL c
eB
ωL = (2.81)
mc
Note that since the particle is relativistic we have m = m0 γ, since E = m c2 = γ m0 c2

(m0 is the rest mass of the electron). It is then obvious that the Larmor frequency decreases
with increasing Lorentz factor γ:
eB
ωL = (2.82)
γ m0 c
From the above, the Larmor radius is obtained via (v ≈ c)
v m0 v c m0 c2 E
rL = = ·γ ≈ ·γ = , (2.83)
ωL eB eB eB
or, in general (χ 6= 90◦ )
E
rL = · sin χ. (2.84)
eB
In Table 2.1 the Larmor radii and frequencies are listed for different energies for a
magnetic-field strength of B = 10 µG.
E [eV] rL νL γ
109 3.3 · 106 km 1.4 · 10−2 Hz 2000
4.510 1.5 · 108 km 3.1 · 10−4 Hz 9 · 104
1020 10 kpc 1.4 · 10−13 Hz 2 · 1014
Table 2.1: Larmor radii and frequencies (B = 10 µG).
We realise that the Larmor radius does not depend on the mass of the particle, but just
on its energy (and on the magnetic-field strength). The particles moving on their helical
orbits about the interstellar magnetic field produce radiation that depends on the temporal
behaviour given by Eqn. (2.77), with the radiation pattern shown in Fig. 2.13 sweeping
around with an angular speed corresponding to the Larmor frequency (2.84). This frequency
corresponds to a mere sweeping frequency of 28 Hz for an electron with γ = 1 in a magnetic
field with B = 10 µG. The reader might now ask the question how, then, we can detect
synchrotron radiation from such particles in the radio regime? In order to see this we need
to calculate the time dependence of the radiation as seen by the observer.
In order to calculate the radiation spectrum, we have to follow the same path as in
Sect. 2.2.2, i.e. we perform a Fourier analysis of the time-dependent radiation power of single
particles, then ’fold in’ their energy spectrum and then calculate the emissivity. The frequency
of the emitted pulses of the gyrating relativistic electrons (Fig. 2.14, right) corresponds to
the inverse of the time that the radiation pattern needs to sweep across the observer. Note
in Fig. 2.14 how the amplitude is boosted by a slight change β. The spike for β = 0.99 is
Figure 2.14: Left: Geometry of the Larmor circle. Right: Sketch of the emitted pulses of the
gyrating relativistic electrons as seen in the particles’ reference frame. Pulse sequences are
shown for three different values of β, and have been shifted for the sake of clearity.
already way outside the frame of the graph. The duration of the pulse in the particle’s frame
of reference is equal to
rL θHP rL θHP
∆t = ≈ . (2.85)
v c
Since rL ≈ E/e B and θHP ≈ 1/γ, we find
m0 c
∆t = . (2.86)
eB
For example, this is about 6 ms in a 10-µG magnetic field, corresponding to ν ≈ 180 Hz.
We now transform this time duration from the particle’s reference frame t to that of the
observer t0 (Fig. 2.15).
Figure 2.15: Illustration of the geometry of the transformation from the particle’s to the
observer’s reference frame.
In order to do that we have to account for the motion of the particle while we measure
the pulse width ∆t, which means that t0 is equal to t plus the time elapsed while the particle
moves from A to B as shown in (Fig. 2.15), i.e.
|~r − ~rL (t)|

t0 = t + . (2.87)
c
This enables us to establish the relation between the two frames5
dt0 1 ~r − ~rL (t) d~rL (t)

= 1− · (2.88)
dt c |~r − ~rL (t)| dt
~n · ~v
q
= 1− 2
= 1 − β · cos θHP ≈ 1 − β · 1 − θHP (2.89)
cr
1 1
= 1 − β · 1 − 2 = 1 − β2 = 2 , (2.90)
γ γ
and we finally arrive at the important result that
∆t
∆t0 = . (2.91)
γ2
We now see that the frequency spectrum is shifted towards a γ 2 -times higher range, so that
for example an electron with an energy of E = 1 GeV, corresponding to γ = 2000, radiates
at a frequency of ν = 700 MHz in the presence of a 10-µG magnetic field! A more precise
determination of the radiation spectrum (of a single electron) requires a Fourier analysis (such
as in the non-relativistic case).
We define the critical frequency such that the particles produce significant power at that
frequency ωc = 2 π νc . There are different definitions around in the literature:
Schwinger 1
ωc := 2 (2.92)
3 · ∆t0
Ginzburg 1
ωc := (2.93)
∆t0
Jackson 1
ωc := 1 (2.94)
3 · ∆0
In what follows, we will use Schwinger’s6 definition. With (2.91) we then derive a critical
frequency that depends on the strength of the magnetic field and on the Lorentz factor:
3 e B⊥ 2
νc = · ·γ . (2.95)
4 π m0 c
Here
B⊥ = B · sin χ (2.96)
is the component of the magnetic field perpendicular to the line-of-sight. In Table 2.2 a few
values for critical frequencies are compiled, assuming a magnetic-field strength of B = 10 µG,
E [GeV] γ νc
1 200 170 MHz
45 9 · 104 340 GHz
1011 2 · 1014 1.7 · 1030 Hz
Table 2.2: Particle energies and their critical frequencies.
which is a typical value for the ISM. In Table 2.2, critical frequencies are listed for a range of
particle energies.
So far, the calculations towards the radiation spectrum of relativistic particles in the ISM
have been performed assuming electrons only. What about protons (or other nuclei)? The
cosmic-ray (CR) energy spectrum observed near earth exhibits ∼100 times more protons
than electrons (at the same energy). Nevertheless, they do not contribute significantly to the
emission, since
3 e B⊥ 2 3 e B⊥ E 2
νc = · ·γ = · (2.97)
4 π m0 c 4 π m30 c5
Thus, νc depends on the mass of the radiating particle like m−3 :
mp −3

= 1.6 · 10−10 ,
me
hence
νc,p
= 1.6 · 10−10 .
νc,e−
Put differently, we can calculate how much more kinetic energy a proton must have in
order to radiate at the same frequency as the electron.
mp 3/2

Ep = · Ee = 8 · 104 · Ee . (2.98)
me
Hence, even the ratio of number densities measured in the CR energy spectrum of np /ne ≈
100 does not help. In fact, as we shall see later, relativistic protons are much more long-lived,
owing to their very low radiation losses. They may remain relativistic for more than a Hubble
time, while electrons become non-relativistic within less than 100 Myr.
The Fourier analysis of the time-dependent radiated power (2.4) is obtained by using its
approximation for small angles θ and large Lorentz factors γ, which is
2 e2 v̇ 2 6 4 γ 2 θ2 cos2 φ

dP 1
= γ · · 1− , (2.99)
dΩ π c3 (1 + γ 2 θ2 )3 (1 + γ 2 θ2 )2
which upon integration over solid angle becomes
Z2π Zπ
dP 2 e2 v̇ 2
P (t) = dΩ = · 3 · γ 4 . (2.100)
dΩ 3 c
0 0
1
5 d
[~r − ~rL (t)]2 2

Here we make use of the derivative dt
6
J. Schwinger: 1949, Phys. Rev. 75, 1912 - 1925
The tedious Fourier analysis of this expression has been done for us by J. Schwinger,
yielding
√ 3
3e ν
P (ν) = 2
· B⊥ · F (2.101)
m0 c νc
where
Z∞
ν ν
F = · K5/3 (x) dx . (2.102)
νc νc
ν/νc
Figure 2.16: Sketch of the time dependence of the synchrotron pulses and their radiation
spectra.
The function F ( ννc ) is the so-called Airy integral of the modified Bessel function K5/3 (x).
It is well approximated by the so-called Wallis approximation7 , obtained after some lengthy
algebra:
0.3
ν ν ν
F = 1.78 · e− νc . (2.103)
νc νc
In Fig. 2.16, some synchrotron pulses emitted by relativistic particles with increasing
Lorentz factor are sketched, with the corresponding radiation spectra juxtaposed. In case
of γ = 1, trivially, the frequency spectrum consists of a single component (δ-function) only,
since the ’pulses’ are simply sinusoidal. As γ increases, more and more frequency components
are added, owing to the increasing distortion of the pulses. For γ 1 we obtain a quasi-
continuum of frequencies, its envelope given by (2.103).
7
G. Wallis; 1959, in Paris Symposium on Radio Astronomy, ed. R.N. Bracewell, p. 595 - 597.
2.3.2 Synchrotron radiation from relativistic electrons with an energy spec-

trum
In order to calculate the radiation spectrum
of an ensemble of relativistic electrons, we
need to know their energy spectrum. This
has been measured in the earth’s vicinity
by studying the CRs raining down onto the
atmosphere. From such measurements, a
power-law of the following form has been
found:
N (E) dE = A · E −g dE. (2.104)
Here, A is a ”constant”8 . It contains the

local number density of relativistic parti-
cles per energy interval. The power-law in-
dex is generally found to be g = 2.4, which
will turn out to be very convenient in our
computation below. In Fig. 2.17 the mea-
sured energy spectrum near earth is dis-
played. The energies of particles produc-
ing the galactic synchrotron radiation are
in the range of ≤ 1 GeV to several tens
Figure 2.17: CR energy spectrum.
of GeV. Unfortunately, the measured spec-
trum is strongly modulated by the solar wind below a few GeV, which explains the deviation
from the power-law there. Hence, nothing is known about the shape of the spectrum at the
lowest CR energies. At the highest energies, there are changes in the spectrum called ’knee’
∼ 10 eV) and ’ankle’ (at >
(at > 15 18
∼ 10 eV). The particles with the highest recorded energies
(> 20
∼ 10 eV, so-called ultra-high enrgy cosmic rays, or UHECR) are a real enigma, their origin
being totally unknown.
Next, we calculate the intensity via the emissivity as follows:
ZE2
4 π ν = P (ν) · N (E) dE. (2.105)
E1
Assuming that there is no background radiation, the radiation transport equation yields
the intensity from the brightness and the source function, the latter given by the emission
and absorption coefficient.
Iν = Sν (T ) · (1 − e−τν ) ≈ Sν (T ) · τν . (2.106)
With Kirchhoff’s law

ν
Sν (T ) = (2.107)
κν
8
‘constant’ refers to the local conditions near earth.
we have
Zs0
Iν = ν ds, (2.108)
0
and hence
Zs0 Z∞
1
Iν = P (ν) N (E) dE ds (2.109)
4π
0 0
It is readily seen that this brightness or intensity has dimension erg s−1 m−2 Hz−1 sr−1 .
Let us assume for simplicity that neither the power P (ν) nor the energy spectrum N (E)
depend on location, i.e. they are independent of s, or dP/ds = dN/ds = 0. Then the above
integral reduces to
√ 3 Z∞
s0 3e ν
Iν = · 2
· B⊥ A · F E −g dE, (2.110)
4 π m0 c νc
0
where s0 is the total path length. Plugging in Wallis’ approximation delivers

√ 3 Z∞ 0.3
s0 3e ν − νν −g
Iν = · · B ⊥ A · 1.78 · · e c · E dE (2.111)
4 π m0 c2 νc
0
We next define some quantities that render the for-

mulae easier to handle by lumping numerical factors and
constants into them.
√
3·e3
C := 1.78 · 4 π m0 ·c2
= 3.32 · 10−23 esu3 erg−1
3 e B⊥
νc = 4π · m30 c5
· E 2 := η B⊥ E 2
η = 6.26 · 1018 s4 g−5/2 cm−7/2
Using the substitution

r 1
νc η·B 2
=: x = · E, (2.112)
ν ν
i.e.
1
ν 2
dE = ηB dx,
Figure 2.18: Illustration of the
we obtain the following expression for the intensity: radiation spectrum.
g+1
Z∞
g−1 1
− g−1
Iν = s0 C A η 2 B⊥2
ν 2 · x−(g+0.6) e− x2 dx. (2.113)
0
It conveniently turns out that in the Milky Way, but also in external galaxies g = 2.4 so
that the integral has a trivial solution! With
1 2
x2
= u, i.e. − x3
· dx = du
it follows that
Z∞ Z∞
−3 − 1 1 1
x ·e x2 dx = e−u du = (2.114)
2 2
0 0
Inserting g = 2.4 we arrive at

s 1.7
−10 0 A B⊥ ν −0.7
Iν = 2.4 · 10 , (2.115)
cm erg cm−3
1.4 G Hz
which in this form has dimension erg s−1 cm−2 Hz−1 sr−1 . Close to the earth, the constant
in (2.104) takes the value A = 8.2 · 10−17 erg1.4 cm−3 . If this constant would hold over a
line-of-sight of 10 kpc, then with a magnetic-field strength of B = 10 µG we would expect a
synchrotron intensity of
Iν ≈ 10−18 erg s−1 cm−2 Hz−1 sr−1 (2.116)
at an observing frequency of ν = 1 GHz, which is close to what we measure towards the

Galactic plane. In general, with
N (E) dE ∼ E −g dE (2.117)
we arrive at
1+α
Iν ∼ B⊥ · ν −α , (2.118)
revealing that - apart from any spatial variations that we

have neglected in the above derivation - the synchrotron in-
tensity depends on the strength of the magnetic-field compo-
nent perpendicular to the line-of-sight, and on the frequency,
with a spectral index α that relates to the power-law index g
of the energy spectrum via
g−1
α= . (2.119)
2
Plugging g = 2.4 into the above equation, we obtain α =

0.7. This value is actually almost invariably found in the
ISM! The synchrotron spectrum steepens in regions of lacking
energy supply (see Sect. 2.3.4), while in the vicinity of star- Figure 2.19: Radiation spec-
forming regions in which stellar winds and supernovae cause trum with energy cutoff.
turbulence and (re-)accelerate the particles the canonical power-law radiation spectrum with
α = 0.7 is found. Note that the factor of 1/2 arises from the fact that E ∼ ν 1/2 .
What is happening in computing Iν is that for each electron the radiation spectrum
P (ν) of the single particle is successively multiplied by the particles’ number density for each
energy (see Fig. 2.18). The intergration over the whole energy range then yields the frequency
spectrum. In the log-log plot this means that we have to add (logarithmically) the ‘weighting
functions’, given by N (E). If the energy spectrum has a cut-off at some energy Emax , the
spectrum will fall off exponentially beyond the corresponding critical frequency (Fig. 2.19),
3 e · B⊥ 2
νc = · · γmax . (2.120)
4 π m0 c
Examples of synchrotron radiation are shown in Figs. 2.20 and 2.21. The Aitoff projection
of the radio emission from the Milky Way at 408 MHz is dominated by synchrotron radiation
almost everywhere, except for the most intense ridge along the Galactic plane, where there
is a significant contribution by thermal free-free radiation (see Chapt. 4). A large number of
discrete sources is seen superimposed onto the diffuse emission. Some of these are located in
the Galaxy, while the bulk of them is extragalactic. Fornax A is a classical radio galaxy with
two extended ‘radio lobes’, which are powered by a central AGN that must be connected with
a supermassive black hole (see Chapt. 6).
Figure 2.20: Radio emission from the Milky Way at 408 MHz.
Figure 2.21: Radio emission of the radio galaxy Fornax A (colour), superimposed onto an
optical image.
~ produced by the particle gyrating in a magnetic

Figure 2.23: Orientation of the electric field E
~
field B.
2.3.3 Polarisation properties

As we have seen in the previous section, relativistic perticles
radiate within a narrow beam of width θHP = γ −1 . Con-
sidering their helical motion around the magnetic field, the
radiation is emitted into a velocity cone (Fig. 2.22), which is
the cone described by the velocity vector ~v . Its axis is along
the direction of the magnetic-field vector. Hence, the open-
ing angle of the cone is twice the pitch angle χ. The cone
axis is parallel to the magnetic field, and ~v precesses about
this direction with the gyro-frequency. In the non-relativistic
case, looking along the magnetic field lines we would mea-
sure circular polarisation, or, if viewing at some angle w.r.t
the magnetic field B,~ elliptical polarisation. The fundamen-
tal difference in case of relativistic particles is that significant
radiation is only measured if the trajectory of the electrons
lies within the (very narrow) angle 1/γ of the line of sight.
The geometric situation is illustrated in Fig. 2.23.
The details are rather complicated. The full algebra
can be found in Longair (1994) 9 . The best way to under-
stand how a single particle produces elliptical polarisation
and how an ensemble of particles with a pitch-angle distribu-
Figure 2.22: Velocity cone of a
tion produces net linear polarisation is by looking at three-
particle moving in a magnetic
dimensional velocity cones (paper models). In case of non-
field.
or mildly relativistic particles the radiation pattern would be
~
broad and we would observe the (rotating) E-vector over the full gyro-circle around the mag-
netic field. If the particles are highly relativistic, we see the light pulse of width 1/γ only for
a very short time, hence we see it only from particles with one specific pitch angle. and thus
record the pulse only over a very small fraction of the velocity cone.
In order to see how the particles produce linear polarisation, let us first consider the case
9
M.S. Longair: High-Energy Astrophysics, Chapt. 8; Cambr. Univ. Press, 2011.
of those electrons whose velocity cones lie precisely along the line-of-sight to the observer
(Fig. 2.23). When the electron points directly to the observer, its acceleration ~v˙ is in the
direction ~v × B,~ so that the observed radiation is linearly polarised, the electric field is
oriented along the direction ~v × B,~ while it lies in the plane perpendicular to the wave vector
~k. Hence, in this case, the E-vector
~ is perpendicular to the projection of B~ onto the plane of
the sky.
If the electron does not precisely point to the
observer within the velocity cone, then there is
also a component of the electric field parallel to
the magnetic-field direction. The radiation from
a single electron is elliptically polarised because
the component parallel to the field has a differ-
ent time dependence within each pulse compared
with that of the perpendicular component. This
is reflected in the fact that the frequency spectra
of the two polarisations are different (see below).
When there is a distribution of pitch angles, how-
ever, all the electrons with velocity cones within
an angle 1/γ around the line-of-sight contribute
to the intensity measured by the observer. The
total net polarisation is found by integrating over
all particles which contribute to the intensity, If
these particles are relativistic, we observe their
radiation only for a very short time ∝ γ −2 , during Figure 2.24: Intensities radiated parallel
which their trajectory as seen by us is a straight and perpendicular to the magnetic field.
line. Hence the resultant polarisation is linear.
The precise calculation of the degree of polarisation is again obtained via a Fourier analysis
of P⊥ (t) and Pk (t), yielding
√ 3
3 e B⊥
P⊥ (ν) = · [F (x) + G(x)] (2.121)
m0 c2
√ 3
3 e B⊥
Pk (ν) = · [F (x) − G(x)] (2.122)
m0 c2
where
Z∞
F (x) = x · K5/3 (z) dz (2.123)
x
as before, and
G(x) = x · K2/3 (x), (2.124)
with x = ν/νc 10 . In Fig. 2.24 a sketch of the functions F (x) and G(x) is shown. Their
behaviour can be qualitatively understood as follows: since the amplitude of the electric field
10
Note that when doing the integration to obtain the total (unpolarised) intensity (2.113), the definition was
x = (ν/νc )1/2 .
compondent Ek is always smaller than that of the component E⊥ , the power Pk must be
lower than that of P⊥ . Furthermore, since Ek varies more slowly as a function of time, the
power Pk peaks at a lower frequency. The degree of linear polarisation at a single energy now
results in
p⊥ (ν) − pk (ν)
P (ν) = . (2.125)
p⊥ (ν) + pk (ν)
Integration over the power-law energy

spectrum, i.e. N (E) dE ∼ E −g dE, yields
g+1 α+1
p= 7 = . (2.126)
g+3 α + 35
With g = 2.4, i.e. α = 0.7, we expect

a maximum degree of linear polarisation of
a synchrotron source of p = 72%. This as-
sumes that the magnetic field is absolutely
uniform and that there are no depolarising
effects (Sect. 3.3.2) such as beam or Fara-
day depolarisation. Figure 2.25: Polarisation ellipse.
The polarisation properties of an electromagnetic wave are described by an ellipse swept
out by the Cartesian components of the electric-field vector. The geometry is shown in
Fig. 2.25. The orthogonal components of the electric field can be represented by
Ex = Ex0 · cos(k z − ωt − δ1 ) (2.127)
Ey = Ey0 · cos(k z − ωt − δ2 ), (2.128)
where k is the wave number, δ1 and δ2 the phases and z the direction of propagation of
the wave. Writing τ = kz − ωt, we have
Ex = Ex0 · cos(τ − δ1 ) (2.129)
Ey = Ey0 · cos(τ − δ2 ), (2.130)
which translates into
Ex
= cos τ · cos δ1 − sin τ · sin δ1 (2.131)
Ex0
Ey
= cos τ · cos δ2 − sin τ · sin δ2 . (2.132)
Ey0
We now multiply the first of the above equations by sin δ2 and the second one by sin δ1
and subtract these from one another; we then multiply the first of the above equations by
cos δ2 and the second one by cos δ1 and also subtract these from one another. This leads to
the following expressions:
Ex Ey
· sin δ2 − · sin δ1 = cos τ · sin(δ2 − δ1 ) (2.133)
Ex0 Ey0
Ex Ey
· cos δ2 − · cos δ1 = sin τ · sin(δ2 − δ1 ). (2.134)
Ex0 Ey0
Squaring and adding the above two equations, we obtain an ellipse equation of the form
Ex 2 Ey 2

Ex Ey
+ −2 cos δ = sin2 δ, (2.135)
Ex0 Ey0 Ex0 Ey0
where δ = δ2 − δ1 . Its orientation and
shape are defined by the angles χ and ψ,
respectively, where the latter is the polar-
isation angle and the former tells us how
much circular polarisation is contained in
the electro-magnetic wave. The electric
field can be also expressed in terms of cir-
cularly polarised waves:
El = El0 · cos(k z − ω t + δ 0 ) (2.136)
Er = Er0 · cos(k z − ω t) (2.137)
Poynting flux is given by

~ × H|
S0 = |E ~ (2.138)
= Ex20 + Ey20 (2.139)
= Er20 + El20 (2.140) Figure 2.26: Poincaré sphere.
= Ea2 + Eb2 (2.141)
At this point it is convenient to introduce the Stokes parameters and the Poincaré sphere
(Fig. 2.26). These definitions allow to measure the polarisation properties in a very simple
way, by adding, subtracting, or multiplying the orthogonal or circular components of the
electric field (I = S0 ):
I = Ea2 + Eb2 = Ex20 + Ey20 = Er20 + El20

(2.142)
Q = I cos 2 χ cos 2 ψ = Ex20 − Ey20 = 2 hEl0 Er0 i cos δ 0

(2.143)
U = I cos 2 χ sin 2 ψ = 2 hEx0 Ey0 i cos δ = 2 hEl0 Er0 i sin δ 0 (2.144)
= 2 hEx0 Ey0 i sin δ = El20 − Er20

V = I sin 2 χ (2.145)
In the above equations, the brackets <> represent time averages. It is evident that one
can retrieve the relevant polarisation properties by either evaluating the Ex − Ey or El − Er
components or by (cross-)correlating the electro-magnetic waves. The most reliable quantity
is obtained from correlation, since in such an operation
+∞
Z
1
< P >= lim E1 (t) E2 (t) dt
T →∞ T
−∞
which is the quantity delivered by the outputs of the so-called IF polarimeters that are
attached to the receivers of radio telescopes (see Sect. 3.2)11 .
11
See lecture notes on Radio astronomy: tools, applications and impacts. Course astro 841, U. Klein, Univ.
Bonn
Figure 2.27: Radio emission of the SNR 3C 10, with the magnetic-field orientation indicated
by the bars (see text for details).
In Figs. 2.27 and 2.28 ex-

amples of measured linear po-
larisation of synchrotron radia-
tion are shown. In the first ex-
ample we see the synchrotron
radiation from the supernova
remnant (SNR) 3 C10, which
is a relic of the supernova ob-
served by Tycho Brahe in 1572.
The radio contours indicate in-
creasing intensity when their
arrows go anti-clockwise, and
vice versa. Inspection of the
left panel of the image shows
that the SNR has a shell-type Figure 2.28: Radio emission of the spiral galaxy NGC 4631
radio structure. The superim- (contours), superimposed onto an optical image. The white
posed ‘vectors’ indicate the ori- bars indicate the orientation of the magnetic field.
entation of the magnetic field.
Their length is proportional to the polarised intensity. What we see here is the synchrotron
radiation from the relativistic electrons in this SNR. The right panel just shows the ‘B-
vectors’, this time with their length proportional to the degree of polarisation, as defined by
Eqn. (2.126) (for more details on supernova remnants, see Sect. 4.3). Fractions of polarisation
exceeding 10% are indicated by orthogonal dashes crossing the ‘vectors’. The second example
shows the radio emission of the spiral galaxy NGC 4631, which we happen to view ‘edge on’.
The radio emission (contours) is a mixture of free-free and synchrotron. Again, the magnetic
field as traced by the linear polarisation is represented by ‘B-vectors’ (for more details on
galaxies, see Chapt. 5).
2.3.4 Losses and particle lifetimes

Synchrotron radiation causes energy losses of the radiating particles (electrons or positrons),
which gives rise to a cutoff in their energy spectra above a certain energy, which in turn causes
a steepening of the radiation spectra above some frequency. As we shall see, this signature
in the spectra can be used as a tool to estimate the ‘ages’ of the particles, meaning the time
elapsed since their energy replenishment ceased. In what follows we derive a relation between
the energy and particle half-lifetime as a function of the strength of the magnetic field.
The energy losses due to synchrotron radiation correspond to the radiated power of the
particles, which is
4
2 e2 ˙2

E
P = · 3 ·v · (2.146)
3 c m0 c2
as we have seen. The acceleration can be written as
v2
v̇ = = ωL v, (2.147)
rL
and hence with
eB 1
ωL = (2.148)
m0 c γ
v e B m0 c2
v̇ = · . (2.149)
m0 c E
Since v ≈ c, we obtain
2 e4
P = · 4 7 · B2 E2. (2.150)
3 m0 c
Now, since
! dE
P =− (2.151)
dt
it follows that
2 2
dE B E
= −2.37 · 10−3 · · erg s−1 (2.152)
dt G erg
2 2
−3 B E
= −1.48 · 10 · · eV s−1 . (2.153)
µG eV
We rewrite this in the form
dE
= −a · B 2 · dt, (2.154)
E2
where a = 2.37 · 10−3 if E has units of erg, which upon integration yields
1 1
− = a · B 2 · (t − t0 ). (2.155)
E E0
E0 corresponds to the initial energy at time t0 that the particles have before the losses
set in. The half-lifetime t1/2 is defined as the time after which the particle has lost half its
energy, i.e.
E0
E(t1/2 ) = . (2.156)
2
Taking t0 = 0, we have
1 E0
E(t) = 1 = (2.157)
E0 + a · B2 · t 1 + a · B 2 · E0 · t
so that
E0 E0
= , (2.158)
2 1 + a · B 2 · E0 · t 1
2
or
1
t1 = · B −2 · E0−1 . (2.159)
2 a
Inserting a and expressing the magnetic-field strength B in µG and the energy E in GeV,
we arrive at
B −2 E0 −1

t1/2 = 8.34 · 109 · · yr. (2.160)
µG GeV
Assuming a magnetic-field strength B = 10 µG, as is typical for the ISM in the Milky-Way
and in external galaxies), we find half-lifetimes of order 107 . . . 108 yr for particles with GeV
energies (Table 2.3).
E t1/2
1 GeV 8·107 yr
10 GeV 8·106 yr
100 GeV 8·105 yr
Table 2.3: Particle half-lifetimes (B = 10 µG).
The evolution of a radio source is hence such that once the energy supply has been
switched off, the energy spectrum will have a cutoff beyond some critical energy Ec that
gradually migrates towards lower energies - initially rather rapidly at the highest energies,
and progressively slower as the cutoff moves towards lower ones. This is sketched in the left
panel of Fig. 2.29. As a result, the synchrotron radiation spectrum exhibits a corresponding
exponential decline beyond the cutoff frequency νc , which we obtain from
3 e
νc = · · B Ec2 . (2.161)
4 π m30 c5
This cutoff frequency νc beyond which the source is rendered undetectable tells us some-
thing about the age of the source. Strictly speaking, this is rather the duration of the ‘remnant
phase’, or the time elapsed since the energy source had been switched off, but it is commonly
used as the source age synonymously. Using convenient units, it can be written as
2.4. INVERSE-COMPTON RADIATION 41
Figure 2.29: Left: Migration of the cutoff frequency resulting from particle ‘ageing’. Right:
Radio continuum spectrum of the starburst dwarf galaxy NGC 1569, exhibiting a break in
the synchrotron spectrum above ∼ 12 GHz.
− 3
ν − 12

9 B 2
τsource = 1.06 · 10 · · yr (2.162)
µG GHz
The right panel of Fig. 2.29 shows an example: the radio continuum spectrum of the
starburst dwarf galaxy (see Sect. 5.3) NGC 1569 is shown here, with least-squares fits to the
synchrotron (blue) and total (black) radio continuum spectrum. The measured flux densities
(black squares) have been used, and the thermal free-free component (red) subtracted at each
frequency, such as to yield the pure nonthermal synchrotron component (blue stars). The
only fixed parameter in this fitting procedure is the spectral-index of the thermal radiation,
viz. αth = −0.1, which is well defined by theory (see Sect. 2.2). Obviously in this case, one
can ‘date’ the termination of the starburst producing the relativistic particles to about 5 Myr
using Eqn. 2.3.4, with a magnetic-field strength of about 20 µG inferred from the measured
synchtron intensity (see Sect. 3.2).
2.4 Inverse-Compton radiation

Another process causing energy losses of relativistic particles is the so-called inverse-Compton
process. In Compton scattering, a photon transfers energy and momentum to a free electron.
If the free electrons have a sufficiently high kinetic energy, m c2 h ν, the reverse may happen,
i.e. net energy will be transferred from the electron to the photon. Relativistic electrons
moving in a photon field, e.g. the far-infrared radiation from galaxies, or the ubiquitous
micro-wave background (CMB) radiation, may ‘boost’ such photons into the X-ray (or even
γ-ray) regime. The derivation of the radiated power is very similar to that for synchrotron
radiation, hence we only present the result here:
Z
dPIC 4
dΩ = PIC = · σT c β 2 γ 2 urad , (2.163)
dΩ 3
where σT is the Thomson cross section and

4σ 4
urad = T (2.164)
c
is the energy density of the radiation field, with σ the Stefan-Boltzman constant. We can
express the radiated synchrotron power in an analogous form,
2 e2 2 4
Psyn = · v̇ γ , (2.165)
3 c3 ⊥
where
e v⊥
v̇⊥ = . (2.166)
γ m0 c
Hence
2 e4
Psyn = · β2 γ2 B2, (2.167)
3 c3 m20 ⊥
2 . With β = β sin χ, we obtain
where we need to average over β⊥ ⊥
Z2 πZπ Z2π Zπ

2 β2 2 β2 2 2
β⊥ = sin χ dΩ = sin3 χ dχ dψ = ·β . (2.168)
4π 4π 3
0 0 0 0
Thus, we obtain
4 e4 2 2 2 32 π e4 2 2 B
2
Psyn = · β γ B = · β γ · (2.169)
9 m20 c3 9 m20 c3 8π
4
= · σT c β 2 γ 2 umag , (2.170)
3
where we have used the Thomson cross section and the classical electron radius,
8π 2
σT = r = 6.65 · 10−25 cm2
3 e
e2
re = = 2.82 · 10−13 cm.
m0 c2
We realise that the radiated power due to both, the synchrotron and the inverse-Compton
process have identical dependences, viz. on the square of the particle energy and on the
square of a photon or magnetic field.
As an example, let us assume that a synchrotron source (e.g. radio galaxy) immersed in the
CMB, which has T0 = 2.728 K. We now calculate ratio of both radiation powers by dividing
(2.163) by (2.170). which is nothing but the ratio of the energy densities of these fields, i.e.
that of the magnetic energy density and that of the energy density of the background-photon
field (which is a perfect black body):
Psyn ! umag B2 c
= = · · T −4 (2.171)
PIC urad 8π 4σ
2.4. INVERSE-COMPTON RADIATION 43
where the CMB temperature depends on red-

shift:
T = T0 · (1 + z). (2.172)
We can now ascribe this temperature an equiva-

lent magnetic field Beq of the CMB, i.e.:
2
Beq 4σ
= · T04 (1 + z)4 , (2.173)
8π c
so that for T0 = 2.728 K we finally obtain an
equivalent magnetic field strength of the CMB of
r
32 π σ
Beq = · T02 (1 + z)2 (2.174)
c
= 3.25 (1 + z)2 µG. (2.175) Figure 2.30: Synchrotron (contours)
and inverse-Compton (colour) radia-
A relativistic particle moving in such an environ- tion from the radio galaxy 3C 294.
ment ‘does not care’ whether it loses energy via syn-
chrotron or inverse-Compton, because of the identi-
cal dependence on their energy and on the energy density of the magnetic and the photon
fields. Accounting for both loss mechanisms reduces, of course, the particle lifetime is
1 i− 1
9 B2 h ν
2
t1/2 = 1.59 · 10 · 2 2
· · (1 + z) yr, (2.176)
B + Beq GHz
where B and Beq are in µG. One of the first examples in which inverse-Compton radiation
caused by the CMB was detected is the radio galaxy 3 C294. This source exhibits extended
Inverse-Compton emission (Fig. 2.30). Its redshift is z = 1.779, yielding an equivalent mag-
netic field of Beq ≈ 27 µG! The interpretation of the observed structures is quite intriguing:
while along the current jet axes (see Sect. 6.4 for the taxonomy and features of AGN) fresh
relativistic particles are transported into the radio lobes, producing synchrotron radiation
there (white contours), inverse-Compton radiation is seen in the X-rays (colour), produced
by particles that have cooled significantly over the past, but still able to boost the CMB
photons to X-ray energies. The implication is that the radio jets are precessing, which is a
not too rare phenomenon, indicating the presence of binary supermassive black holes. The
aged particles in the precessed jets and lobes should be detectable at low radio frequencies,
e.g. with LOFAR.
We will come back to the inverse-Compton process when dealing with radio galaxies
(Sect. 6.4) and clusters of galaxies (Sect. 7.6).
2.5 Spinning dust grains

In the course of studies of the Galac-
tic foreground emission as a potential
contaminent of the CMB unisotropy
studies (Kogut et al. 1996, Leitch et
al. 1997), ‘anomalous’ radio emission
was discovered at frequencies above
∼ 15 GHz that could not be at-
tributed to synchrotron, free-free, or
thermal (vibrational) emission from
dust. Clarification came witch a
seminal paper by Draine & Lazar-
ian (1998)12 on radio emission from
spinning dust grains. This electric
dipole radiation is produced by ro-
tationally excited, small interstellar
dust grains producting a bump in
the radio continuum spectrum in the
range 14 . . . 50 GHz. In Fig. 2.31 Figure 2.31: Superposition of various continuum radi-
the radio continuum spectrum of the ation components in the radio spectrum of M 82.
starburst galaxy M 82 is shown, with
the thermal free-free and synchrotron components as discussed in Sects. 2.2, 2.3, and 2.3.4.
At frequencies above about 100 GHz the thermal radiation from dust (not discussed in this
lecture) dominates the overall luminosity of galaxies. It is obvious from theis spectrum that
one can measure the thermal free-free radiation alone only within a small frequency range
between about 40 and 70 GHz (a frequency range that is not easily accesible from the ground,
owing to the atmospehric attenuation). It is this very range where one would expect the ra-
diation emitted by spinning dust grains to be seen. The expected(!)13 bump produced by this
kind of radiation is indicated in Fig. 2.31 by the dashed magenta line, with the contribution
to the total (solid black line) emission represented by the dotted black line, which is barely
visible in the frequency range of lowest overall emissivity of this galaxy.
The details of calculating the frequency spectrum of spinning dust grains are complex,
the present lecture not providing enough space for this. We will hence just have a look at the
general route to get there. A spinning grain with electric dipole moment µ ~ radiates a power
2 ω 4 µ2 sin2 Θ
P = · , (2.177)
3 c3
which if averaged over Θ average of
4 µ2 ω 4
hP iΘ = · 3 . (2.178)
9 c
Here, θ is the angle between the angular velocity ω
~ and the moment µ
~ . In order to calculate
the emissivity, one has to consider various damping and excitation mechanisms controlling the
12
Astroph. J. 508, 157, 1998
13
Note that the detection of emission from spinning dust grains has remained controversial up to now. The
bump shown here has not been measured yet, as this requires utter precision of the measurements!
2.5. SPINNING DUST GRAINS 45
spins of the grains - a tedious task, and in spite of the seminal paper quoted above probably
not yet develering a matured theoretical concept. In fact, all of the detections claimed to date
have not been all that convincing, also because of the difficulty to separate the radio emission
from spinning dust from the other emission components (synchrotron and free-free).
For damping, we need to consider:
• Collisional drag: species like H, H2 , and He may stick on the surface when hitting a
grain, and may be desorbed with a thermal velocity distribution.
• Plasma drag: a grain’s electric dipole moment µ~ interacts with passing ions, hence
it couples the grain rotation to the plasma. This process is the dominant damping
mechanism in molecular clouds.
• Infrared emission: a thermally excited photon of energy h ν released from a grain with
ω ν removes, on average, ~ ω/2 π ν of angular momentum from the grain. This is the
dominant process in the warm neutral medium (WNM).
• Electric dipole damping: this accounts for the radiated power P .
For excitation, we need to consider:
• Recoil from thermal collisions and evaporation: neutrals and ions deposit their angular
momentum when they impact the grain, and give it an additional kick when they
subsequently evaporate. This bombardement by neutrals and ions is the dominant
process for small grains in molecular clouds, in the cold neutral medium (CNM), in the
WNM, and in the WIM.
• Excitation by the plasma: this process implies the excitation of the grain rotation by
the fluctuating electric field of a passing plasma, and is at work in molecular clouds and
for large grains.
• Infrared emission: each IR photon carries away an angular momentum ~, corresponding

to a change in angular momentum of the grain.
• Photo-electric emission: photo-electrons are emitted randomly from the grains. Since
me mH , this mechanism is negligible.
In order to derive the emissivity, one has to account for the excitation and damping rates.
These depend on the temperature T and on the size distribution dn/da of the dust particles.
The emissivity per H atom finally reads:
1 2
dn µ2 ω 6 − 23hωω2 i
Z
jν 8 2 1
= · da · ·e (2.179)
nH 3π nH c3 da hω 2 i 23
This emissivity gives rise to an emission bump noticeable in the frequency range 15 ≤ ν ≤
50 GHz. Because of the superposition of the various continuum emissions in this frequency
domain, its seperation from other emission components is anything but easy (see Fig. 2.31).
Chapter 3
Diagnostics
In this chapter, we shall treat the main diagnostic tools that we have at hand to investigate
interstellar and intergalactic magnetic fields. There are essentially four different tracers of
magnetic fields in a diffuse astrophysical medium:
• Optical polarisation, caused by dust grains whose magnetic moments align along the
magnetic field.
• synchrotron radiation, bearing information about the total strength of the magnetic
field. Its linearly polarised component measures the strength and orientation of the
ordered or uniform component of the magnetic field perpendicular to the line of sight.
• Faraday rotation, yielding information about the line-of-sight component of the mag-
netic field and its direction.
• The Zeeman effect, yielding information about strong magnetic fields in dense gas clouds
and about their direction.
In what follows, the physics of these diagnostic tools is discussed, the observables pointed
out, and examples of measurements are given.
3.1 Optical polarisation

Optical (stellar) light is polarised by selective extinc-
tion, due to elongated dust particles that are aligned
with the interstellar magnetic field, owing to their mag-
netic moments (Fig. 3.1). The extinction is strongest
along the dust particles’ extended bodies, hence the re-
sulting net (linear) polarisation has E ~ k B,
~ which is
different from synchrotron radiation where we saw that
~ B.
E⊥ ~ The extinction cross-section is
Cext = Cabs + Csca = π a2 Qext (3.1)

Figure 3.1: Elongated grains lining
where a is the typical radius of dust grains, and Qext up in a magnetic field.
is the so-called extinction efficiency,
Qext = Qabs + Qsca . (3.2)
47
48 CHAPTER 3. DIAGNOSTICS
It is related to the albedo

Qsca Qsca
γ= = , (3.3)
Qext Qabs + Qsca
where γ = 1 implies full reflection, and γ = 0 implies no scattering at all. More realisti-
cally, the dust particles have a size distribution f (a) so that
Z∞
Cext (λ) = Qext (a, λ) · π a2 f (a) da. (3.4)
0
For silicates and graphites, f (a) ∝ a−3.5 . The resulting absorption coefficient is
κλ = π a2 · Qext · nd , (3.5)
where nd is the number density of dust particles. We define the extinction cross sections
Ck and C⊥ for the electric field parallel and perpendicular to dust particles’ longitudinal axes,
respectively. Then, the corresponding intensities can be written as
1 1
Ik = I0 · e−τk , I⊥ = I0 · e−τ⊥ (3.6)
2 2
with
I = Ik + I⊥ . (3.7)
The magnitude of the polarisation then is
Ik
∆mp = −2.5 · log = 1.086 · (τk − τ⊥ ), (3.8)
I⊥
where
τk = Ck · nd · r, τ⊥ = C⊥ · nd · r. (3.9)
In general, the alignment of the dust particles is not perfect. Hence, one introduces the
degree of alignment f ∈ [0, 1] via
∆mp = 1.086 f · (Ck − C⊥ ) nd r. (3.10)
The geometry of rotating dust particles exposed to an interstellar magnetic field is il-
lustrated in Fig. 3.2. Dust particles are likely to be non-spherical. Otherwise, anisotropic
extinction would not be possible. Let us consider them as prolate (rods or ellipsoids). Because
of their continuous collisions with gas particles (mainly hydrogen atoms), the dust particles
are permanently forced to rotate. In thermal equilibrium, the rotational energy around lon-
gitudinal and transverse axis is equal, i.e.
1 1
Erot,i = · Ii ωi2 = · k Tg . (3.11)
2 2
3.1. OPTICAL POLARISATION 49
Here, Tg is the gas temperature, Ii is the moment of inertia

of the dust particle, and the index i refers to the principle axes
of the elongated particle. √ Since Ii ωi2 = K is approximately
constant, we have ωi ∼ 1/ Ii , hence its angular momentum
can be expressed as
p
Ji = Ii · ωi = K Ii . (3.12)
Since usually I1 I2 , the total angular momentum J~ =

J1 + J~2 will have its vector close to J~1 , i.e. along the transverse
~
axis of the dust particle. Hence the rotation will primarily be
perpendicular to the symmetry axis. A rough estimate of
ωi can be made in the following way: consider a spherical
particle with radius a and density ρd , hence mass
4
md = π a3 ρd . (3.13)
3
Then
2 8
I= · md a2 = π a5 ρd , (3.14)
5 15
where
1 2 1 Figure 3.2: Geometry of ro-
Iω = k Tg . (3.15)
2 2 tating grain and magnetic
field.
Plugging in a = 5 · 10−5 cm, ρd = 1 g cm−3 , Tg = 100 K,
this yields ω = 2 · 104 rad s−1 ! The colliding particles are hydrogen atoms in the first place.
Many of them stick to the dust grain surface, combining to H2 with another hydrogen atom.
The particles thereby get rid of the excess energy in this exothermic process (dust is a catalyst
in H2 formation!), thereby transferring extra energy to the dust grain. Since the surface of
the grain is not uniform, this will cause a net increase of its angular momentum on average.
An estimate leads to an increase by a factor of about 104 so that ω ≈ 108 . . . 109 rad s−1 !
The alignment of the particles was worked out by Davis and Greenstein (1951) and ex-
plained in terms of paramagnetic relaxation. If the rotation axis of the particle does not
~
coincide with the direction of the B-field, each of its volume elements is exposed to a peri-
odically varying magnetic field and, correspondingly, to a varying magnetisation M ~ per unit
~
volume (Fig. 3.2). The direction of M permanently deviates from the instantaneous direction
~ lagging behind the latter by some amount. This behaviour is described in terms of
of B,
a complex magnetic susceptibility χ = χ0 + i χ00 , where the imaginary part determines the
portion of rotational energy that is continuously converted into heat by re-magnetisation. In
other words: it is associated with an additional magnetisation M ~ 00 , which is perpendicular to
~ ~
B and J. One finds
00
~ 00 = χ · ω
M ~
~ × B. (3.16)
ω
This gives rise to a torque
D ~ 00 × B
~ =V ·M ~ (3.17)
with modulus
~ 00 | · |B|
D = V · |M ~ sin Θ = V χ00 B 2 sin Θ, (3.18)
where V is the grain’s volume and the angle Θ = ∠(~ ~

ω , B). The torque D ~ is directed
~
perpendicular to B and has an opposite direction to ω⊥ , the component of ω ~ perpendicular to
~ The angular momentum I ·ω⊥ around the axis perpendicular to B
B. ~ is gradually diminished,
finally vanishing after a retardation time given by
D · tret = I · ω⊥ . (3.19)
~ while the magnetisation does
What remains is the rotation about an axis parallel to B,
not change anymore. For many paramagnetic substances the relation
ω⊥
χ00 ≈ 2.5 · 10−12 · (3.20)
Td
holds, where Td is the dust temperature.
An estimate of the retardation time tret =
I ω⊥ /D yields ∼ 4 · 107 yr, assuming a =
5 · 10−5 cm, ρd = 1 g cm−3 , Td = 15 K,
B = 3 µG, and sin Θ = 0.5. Now one has
to compare this with the time tc required to
significantly change the rotation of a dust
grain by collisions, which can be estimated
by assuming that this is the time required
to have collisions with particles making up
for the total mass of the dust grain. For
this we need N collisions
md
Ncoll = = tcoll νcoll . (3.21)
mH
Here, the collision frequency is
νcoll = π a2 vg nH , (3.22)
where Figure 3.3: Optical polarisation of the spiral

s galaxy M 51.
8 k Tg
vg = . (3.23)
π mH
Using the previous parameters (a = 5 · 10−5 cm, Tg = 100 K, nH = 1 cm−3 ), one arrives at
tcoll < 6
∼ 10 yr, which is much shorter than tret . This means that but few dust grains reach
perfect alignment. A ‘precise’ derivation of the degree of alignment delivers
00
tcoll χ 1
f ≈ 0.3 · ∼ B2 · · p (3.24)
tret ω a nH Tg
Hence, in order to obtain sizeable degrees of alignment, strong magnetic fields are required.
Yet, the polarisation of starlight has been readily observed in the general ISM, in the presence
of magnetic fields of several µG strength (see Fig. 3.3). It therefore looks as if the theory has
not been completely understood to date.
3.2 Synchrotron radiation

Synchrotron radiation is (partly) polarised, and hence provides a powerful tool to study cosmic
magnetic fields when- and whereever charged relativistic particles (electrons, positrons) are
present. As we have seen in Sect. 2.3.3, the polarisation state of an electromagnetic wave
is elegantly described by the so-called Stokes parameters I, U , Q and V , which define the
Poincaré sphere. The total power carried by the radiation (or, depending on the measured
quantity, the intensity, flux density, ...) is represented by Stokes I, such that
I 2 = Q2 + U 2 + V 2 . (3.25)
As we have also seen, the Stokes parameters can either be expressed in terms of parameters
defining the polarisation ellipse, or as combinations of the Cartesian or circular components
(Eqns. 2.142 through 2.145). By means of so-called IF polarimeters1 , radio astronomers
can relatively easily and reliably measure all Stokes parameters, from which one can then
deduce the total polarisation state of the observed electromagnetic wave. As we shall see
in Sect. 3.3.2, the actually measured degrees of polarisation are always much lower than the
theoretically predicted ones, owing to all kinds of depolarisation mechanisms. One therefore
makes a distinction between the measured total intensity, given by Stokes I,
Itot = I, (3.26)
and the measured polarised intensity as retrieved from

p
Ipol = U 2 + Q2 + V 2 . (3.27)
In many astrophysical applications we are dealing with linear polarisation, hence V = 0,

so that in this case
p
Ipol = U 2 + Q2 . (3.28)
The degree of polarisation is in general given by

p
U 2 + Q2 + V 2
ptot = , (3.29)
I
and the degree of linear polarisation by
p
U 2 + Q2
plin = . (3.30)
I
In this case, we also need to measure the polarisation angle ψ, which we obviously retrieve
from (Eqns. 2.143 and 2.144), i.e.
1 U
ψ= arctan . (3.31)
2 Q
1
The IF (intermediate frequency) polarimeter is a so-called quadrature network that allows to measure the
Stokes parameters by correlating the orthogonal components Ex , Ey or El , Er such as to yield the desired
quantities of Eqns. (2.142) through (2.145).
For sources with circular polarisation, the degree is obviously

V
pcirc = . (3.32)
I
Normal synchrotron sources (SNR, diffuse emission from the MW and from galaxies, from
radio galaxies, or from clusters of galaxies) emit linearly polarised radio emission, while circu-
larly polarised radiation is seen in sources emitting gyro-cyclotron or coherent emission (the
Sun, Jupiter, pulsars, cosmic masers). In what follows, we shall briefly discuss the physical
quantities that can be deduced from measurements of synchrotron radiation. Intensity of
synchrotron radiation is (Sect. 2.3.2)
1+α
Iν ∼ B⊥ · ν −α , (3.33)
where
g−1
α= , (3.34)
2
α being the slope of the power-law radiation spectrum, and g that of the energy spectrum
of relativistic particles,
N (E) dE = A E −g dE. (3.35)
Since radio waves traverse the universe essentially unhindered by matter, we can derive
the shape of the particle spectrum of synchrotron-emitting sources out to large distances.
The spectral break seen at high frequencies helps to estimate the particle age, i.e. the time
elapsed since these particles have been injected or accelerated (Sect. 2.3.4).
The brightness of the synchrotron radiation furthermore allows to estimate the total
strength of the magnetic field, or more precisely, that of B(⊥). In order to do so, we have to
assume energy equipartion between the relativistic particles and the magnetic field in which
they are gyrating. The radio luminosity is
−α
0 1+α −α ν
Pν = c A V B ν = P ν0 · , (3.36)
ν0
where V is the volume of the source and c0 a constant. The total energy of the source
responsible for the radio emission is the sum of the energies of the particles and the magnetic
field:
Z∞
B2
Etot = Epart + Emag = V · E N (E) dE + ·V (3.37)
8π
0
At this point, we have to take into account the energy of the relativistic protons in
this balance. Even though they do not radiate significantly, they make up for the major
contribution to the particle energy. We account for this by writing
Ep = β Ee (3.38)
and hence for the total particle energy
Epart = (1 + β) · Ee = η Ee (3.39)
we therefore write
Z∞
B2
Etot = η V · N (E) E dE + V · (3.40)
8π
0
Recalling the relation between the critical fre-

quency of the radio emission and the strength of the
magnetic field,
2
3 eB E
ν= · · = C E2 B (3.41)
4 π m0 c m0 c2
we can write the boundaries of the above inte-

gral, i.e. the minimum and maximum energies of
particles in terms of the corresponding frequencies
of the emission: Figure 3.4: Minimum-energy and
equipartition magnetic field strengths.
ν 1
min,max 2
Emin,max = , (3.42)
CB
for which the radio spectrum is known. It is generally assumed that νmin ≈ 10 MHz, and
νmax = 100 GHz, which are the extremes of the (classical) radio regime. Even though both
of these limits are somewhat arbitrary, one can argue: the lowest-energy particles2 contribute
little to the total energy budget. In normal radio sources with declining power-law spectra,
particles producing radio emission above the upper frequency bound are unlikely. We can
then calculate the total energy contained in the relativistic particles via
EZmax 2−g

η V A νmin 2 ν 2−g
−g max 2
Epart = η V · E AE dE = · − . (3.43)
g−2 CB CB
Emin
Solving (3.36) for A and substituting this into the above equation, we have

ηV Pν g−2
h
(2−g)/2 (2−g)/2
i
Epart = · · (C B) 2 · νmin − νmax . (3.44)
g−2 c V B 1+α ν −α
0
Preserving only the essential observables, we arrive at

3
Epart = G · η · Pν · B − 2 , (3.45)
where G is a constant that depends weekly on α, νmin , and νmax . Therefore, the total
energy is
3 B2
Etot = G η Pν B − 2 + V · . (3.46)
8π
2
Even though the low-energy CR spectrum is strongly modulated by the solar wind as mentioned before.
In Fig. 3.4, a plot of the magnetic-field, particle, and total energy is shown, with a mini-
mum total energy for a magnetic field Bmin = B(Emin ), which is near to the magnetic field
strength producing precise equipartition, Beq = B(Eeq ). Minimizing this w.r.t. B, we obtain
2
G η Pν 7
Bmin = 6π . (3.47)
V
Substituting Bmin into (3.45), we obtain
B2 3
Emag = V · = · Epart . (3.48)
8π 4
We realise that the condition for the minimum-energy requirement to produce the ob-
served radio emission from a given radio source corresponds very closely to a condition of
equipartition of energy between particles and magnetic fields. The minimum total energy is
7 3 4
Etot,min = V 7 · (6 π G η Pν ) 7 (3.49)
24 π
A useful approximation for this if we plug α = 0.75 (a typical value) into the quantity G
and assuming a spherical volume of radius R is
2 4 9
41
ν 7 Pν 7 R 7 4
Emin = 6 · 10 · · · · η 7 erg. (3.50)
100 MHz W Hz−1 kpc
The ratio of energy carried by protons to that carried by electrons is of order η = 102 .
Move recent estimates, based on the theory of ‘diffusive shock acceleration’ (Sect. 4.4) yield
η = 20 . . . 40. Assuming energy equipartition (which is not very different from minimum
energy, which a system usually maintains), we can utilise the synchrotron intensity to infer
the total magnetic-field strength. Even though this can only give rough estimates, it is a
powerful method in the absence of any other diagnostic tools, such as Faraday rotation, or
Zeeman measurements. We can then estimate the total magnetic-field strength locally by
using the brightness temperature given by
2 ν2 k
Sν = · Tb ΩS , (3.51)
c2
where Sν is the flux density and ΩS the solid angle of the observed source, Sν being related
to the monochromatic radio luminosity (the radio power) via
Pν
Sν = , (3.52)
4 π D2
where D is the distance to the source. For h ν k Tb , the Rayleigh-Jeans approximation
has been utilised as can be generally assumed in the radio domain (ν < 50 GHz). Noting
that the volume in (3.47) is V ≈ L3 (L = size of the source) and
2
L
ΩS = , (3.53)
D
3.3. FARADAY ROTATION 55
one can derive the minimum-energy magnetic-field strength of a galaxy disk with radius
L and effective ‘radio thickness’ H as
2 2
ν 74 H −7

2 Tb 7
Bmin = 1.7 · η ·
7 · · µG (3.54)
K GHz kpc
so that, inserting typical values (η = 40, Tb = 1 K, ν = 1 GHz, H = 1 kpc) we ob-

tain a value of Bmin ≈ 5 µG. This is in line with independent determinations based upon
measurements of Faraday rotation.
3.3 Faraday rotation

3.3.1 Rotation measure
In this section, we discuss the rotation of the electric-field
vector that a linearly polarised wave experiences when prop-
agating through a magnetised plasma, an effect called Fara-
day rotation. To this end, let us consider the propagation
of electro-magnetic waves through a magnetised plasma. In
order to treat this mathematically, we have to add the forces
acting on the free electrons. If some force is exerted on a free
charge in a magnetic field, the charged particles are forced
into helical paths, owing to the Lorentz force. Hence we have
• centrifugal force, −m0 · ω 2 · ~r
• the Lorentz force, ± ec · ω · Bk · ~r
• electric force of propagating wave, e · E(~

~ r, t)
A linearly polarised wave is a superposition of two circularly

polarised waves with opposite rotation (LHC, RHC). Bk is
the component of the magnetic field along (parallel to) the
direction of propagation of the electro-magnetic wave. The
geometry is illustrated in Fig. 3.5. The equation of motion
then reads: Figure 3.5: Geometry of cir-
cularly and linearly polarised
2 e ~
−m0 ω ~r = ± · ω Bk ~r + e E(~r, t) . (3.55) waves.
c
In the above, the ‘+’ sign stands for a right-hand circularly polarised wave (RHC), and the ‘−’
sign stands for a left-hand circularly polarised wave (LHC). Since the RHC polarised wave has
the same helicity as the electron, and the LHC polarised wave has opposite helicity to it, this
yields a difference in the interaction forces. This implies that if a circularly polarised wave has
same sense as the gyrating electron, the interaction is larger. The different interactions will
have different effects on the phases of the RHC and LHC components of the wave. Solving
the above equation for ~r, we obtain

e 1 ~ ,
~r = − · ·E (3.56)
m0 ω 2 ± ω ωc
where
e Bk
ωc = (3.57)
m0 c
is the cyclotron frequency. The force acting on the electron implies an electric displacement
~ described by the electric polarisation
D,
1 −1 ~
P~ = ~ − E)
· (D ~ = · E = ne e ~r (3.58)
4π 4π
e2 1 1
=⇒ −ne · · 2 = · ( − 1) , (3.59)
m0 ω ± ω · ωc 4π
or
s
ωp2 4 π ne e2
=1− ωp = . (3.60)
ω (ω ± ωc ) m0
The quantity ωp is the plasma frequency, which we encountered in the previous section.
√
Recalling that the refractive index is n = ( is the dielectric constant) we arrive at a
general expression for the refractive index, which accounts for the presence of a magnetic
field (contained in the cyclotron frequency ωc ):
s
ωp2
n= 1− . (3.61)
ω (ω ± ωc )
The difference to the refractive index for |B| ~ = 0 is that it implies different phase velocities
of the RHC component Er and the LHC component El of the wave, resulting in a phase shift
of Er with respect to El after traversing the magnetised plasma. This phase shift between
Er and El implies a rotation of the linearly polarised electric field of the electromagnetic
wave. Referring to the illustration on the previous page, the reader should be clear about the
~ to rotate by a certain amount, its individual
following: in order for the electric field vector E
RHC and LHC components have to rotate by twice this angle. Considering the rotation of
the electric-field vector along some path dr through the magnetised plasma and denoting the
corresponding rotation of the RHC component by dθr and that of the LHC component by
dθl , we calculate the net rotation d(∆ψ) along the path dr as we have
dθl = kl · dr (3.62)
dθr = kr · dr (3.63)
kl − kr
d(∆ψ) = · dr . (3.64)
2
The phase velocity is
ω c ω
vp = = =⇒ k = · n (3.65)
k n c
so that
∆k ω
d(∆ψ) = · dr = · ∆n dr (3.66)
2 2c
√
With n = , we can calculate the corresponding difference of the refractive index,
n2l − n2r = l − r = (nl + nr ) · (nl − nr ) = 2 n ∆n (nl ≈ nr ) (3.67)
l − r
=⇒ ∆n = (3.68)
2·n
Inserting here l , r yields
ωp2 ωc 1
∆n = · . (3.69)
n ω ω 2 − ωc2
The refractive index n was given by
" #1
ωp2 2
n= 1− , (3.70)
ω (ω ± ωc )
which for ω ωc and ω ωp is ap-
proximated by
1 ωp 2
n≈1− · . (3.71)
2 ω
For the difference of the refractive index
of the RHC and the LHC components this
yields
ωp2 ωc
∆n = . (3.72)
ω3
This change of phase ∆n is translated into a
change in polarisation angle ∆ψ of the lin-
early polarised wave penetrating the mag-
netised plasma via
ωp2 ωc
d(∆ψ) = dr . (3.73)
2 c ω2
Finally, the total Faraday rotation of the
electric vector is obtained by integrating
this equation over the total path length Figure 3.6: Observed RM distribution across the
Z r0 spiral galaxy NGC 6946.
2 π e3
∆ψ = 2 2 2 · ne Bk dr . (3.74)
m0 c ω 0
Conventionally, this is written in the form
∆ψ = RM · λ2 , (3.75)
where RM is the so-called rotation measure. This important quantity is measured in

units of rad m−2 , the wavelength λ being measured in m. In astrophysical applications, the
rotation measure is written such that it contains convenient units, i.e.
Z r0 n B dr
e k
RM = 0.81 · · · rad m−2 . (3.76)
0 cm−3 µG pc
Its dependence on λ2 tells us that Fara-

day rotation decreases rather quickly as we
go to higher radio frequencies. At the same
time, this makes it a very powerful tool
as a probe of the interstellar (and inter-
galactic) magnetic field. If RM can be de-
termined by measuring the Faraday rota-
tion angle ∆ψ over a sufficient frequency
range, then we have information about the
interstellar magnetic-field strength if ne is
known. The determination of RM requires
measurements of ψ at least three frequen-
cies, owing to an ambiguity to measure ψ.
With just two measurements, the rotation
could be ∆ψ + n · π, where n is an integer. Figure 3.7: RM ambiguity: Measured values
Below, we list some examples of rota- (blue) can be shifted by multiples of π (red) such
tion measures in different environments rel- as to produce a different RM.
evant for astronomy. An example of the
distribution of the rotation measure across a spiral galaxy is shown in Fig. 3.6 (see Chapt. 5
for more details on magnetic fields in galaxies). The absolute values of the rotation measure
in external galaxies are typically |RM | < −2
∼ 100 rad m , while in some radio galaxies they
can be very high, |RM | < −2
∼ 5000 rad m . The reason for the large values in such sources
(like e.g. in Hydra A) is that this object resides in the centre of a cooling-flow galaxy cluster,
so that we are dealing with a large line-of-sight through a magnetised and fully ionised region.
ne ≈ 106 cm−3 

Ionosphere: r0 ≈ 100 km RM ≈ 0.8 rad m−2

Bk ≈ 0.3 G

ne ≈ 0.03 cm−3


ISM: r0 ≈ 500 pc RM ≈ 60 rad m−2
Bk ≈ 5 µG

ne ≈ 0.005 cm−3 

−2
Cooling flows: r0 <
∼ 200 kpc RM <
∼ 3000 rad m
Bk ≈ 3 µG

It is difficult to derive the rotation measure unambiguously, since one cannot distinguish
between ψ and ψ + n π.3 The sketch in Fig. 3.7 shows a possible least-squares fit to measured
polarisation angles ψ(λ2 ) that allow for both, a positive and negative RM (or even larger ab-
solute values) if too few wavelengths are involved in the measurements! The newly developed
rotation-measure synthesis technique used at lower radio frequencies (Sect. 3.3.3) is a very
promising tool to resolve this problem. Otherwise, the recipe is to observe the target at as
many wavelengths as possible.
3
The reason for this is that the response of a dipole to a linearly polarised electromagnetic wave is identical
when the dipole is rotated by 180◦ , and any detector of linear polarisation is subject to this principle!
3.3.2 Depolarisation
Astrophysical objects with a particle energy spectrum of the form N (E) dE ∼ E −g dE produce
radio emission with a flux density Sν ∼ ν −α , where α = (g − 1)/2. In this case, we expect
(Sect. 2.3.3) a ‘theoretical’ degree of linear polarisation
g+1 α+1
p= 7 = , (3.77)
g+3 α + 35
which of course may vary with location, i.e. p = p(~r). This degree of polarisation depends
only weakly on frequency. Its maximum is, however, never reached, owing to a number of
effects that reduce this degree of polarisation, commonly referred to as depolarisation . We
shall see, however, that both, Faraday rotation as well as depolarisation deliver information
about the magnetised plasmas. If the instrumental effects are known, the ratio of the observed
to the theoretical degree of polarizaton bears astrophysical information that can then be
utilised.
An instrumental effect that reduces the degree
of depolarisation is the so-called bandwidth depo-
larisation. The reason for this is that the polari-
sation angle rotates over the bandwidth (Fig. 3.8)
in the presence of Faraday rotation. If the band-
pass has limiting wavelengths λ1 and λ2 , where
λ1 ≈ λ2 , then the electric-field vector will be ro-
tated across that band according to
∆ψ = RM · (λ21 − λ22 )
≈ (λ1 + λ2 ) · (λ1 − λ2 ) · RM
≈ 2 λ0 ∆λ · RM
where we have made use of
c dλ c ∆ν Figure 3.8: Bandpass.
λ= ⇒ = − 2 ⇒ ∆λ = −c 2
ν dν ν ν0
and hence
∆ν
∆ψ = −2 λ20 · RM · . (3.78)
ν0
We shall see below how this rotation across the bandpass leads to a reduction of the degree
of linear polarisation, given a particular rotation measure.
The degree of polarisation is generally obtained by integrating the degree of polarisation
P (~r) over some path r, weighted by the emissivity (λ, ~r), and neglecting any frequency
dependence of P (~r):
s 2
p(~r) · (λ, ~r) · ei 2 [ψ0 (~r)+λ Φ(~r)] dr dΩ
Pobs (λ) = source s . (3.79)
(λ, ~r) dr dΩ
source
Here,
P (~r) = p(~r) · ei 2 ψ0 (~r) (3.80)
is the intrinsic degree of polarisation of the

source. Note that the degree of polarisation defined
in this way is a complex quantity!4 If the polarised
emission comes from a region within which the Fara-
day rotation varies, it is convenient to define the
rotation or Faraday depth Φ(~r) (Fig. 3.9) for the ro-
tation measure appropriate to each volume element
at position ~r:
Z~r
Φ(~r) = K · ~ · d~r.
ne B (3.81)
0 Figure 3.9: Illustration of Faraday
depth.
Here K is the constant preceding the integral in
Eqn. (3.76), its numerical value depending on the units used for the thermal electron density,
the magnetic-field strength, and the path length. Eqn. (3.79) can now be simplifed by su-
perposing all radiation with the same Faraday depth or rotation depth. This simultaneously
accounts for lateral variations. If E(Φ) dΦ is the fraction of total emission (assumed inde-
pendent of λ) with a rotation depth between Φ and Φ + dΦ, and if the intrinsic polarisation
is
P (Φ) = p(Φ) · ei 2 ψ(Φ) , (3.82)
then the polarisation at wavelength λ is
+∞
Z +∞
Z
2 i 2 Φ λ2 2
P (λ ) = E(Φ) P (Φ) e dΦ = F (Φ) ei 2 Φ λ dΦ, (3.83)
−∞ −∞
where the source function F (Φ) = E(Φ) · P (Φ), also called Faraday dispersion function, is
the polarised intensity, expressed as a fraction of the total intensity of the source. Obviously,
P (λ2 ) is the Fourier transform of F (Φ). Unfortunately, we cannot simply invert this Fourier
integral, such as to obtain
+∞
Z
1 2
F (Φ) = · P (λ2 ) · e−i 2 Φ λ d(λ2 ), (3.84)
π
−∞
as P (λ2 ) is not defined for λ2 < 0. We shall, however, see how we can retrieve P (λ2 ) by
assuming simple model distributions for F (Φ) and perform the Fourier inversion (3.83). The
resulting degree of polarisation may be compared with the observations, and parameters of
the dispersion function F (Φ) adjusted until best agreement with the observations is achieved
(s.b.). Recently, however, a technique has been developed that allows to do the Fourier
inversion inspite of the restrictions inherent to it. This so-called ‘RM synthesis technique’
will be treated in (Sect. 3.3.3).
4
Note that we use upper-case P for the complex polarisation and lower-case p for its modulus!
Let us have a look at a few different situations, for

which plausible forms of the dispersion function can be
assumed. For this method, which was first proposed
in a seminal paper by Burn (1966)5 it is convenient to
work with the complex polarisation defined by
P = Q + i U = p · ei 2 ψ , (3.85)
which describes the pseudo-vector P in the complex

plane defined by the Stokes parameters Q and U . Its
magnitude is p, and its orientation is given by twice
the polarisation angle, in accord with Fig. 2.26 and
Eqns. (2.142) through (2.145). The Faraday dispersion
function F (Φ) measures the polarised flux coming from
a Faraday depth Φ, Figure 3.10: The complex (Q,U )-
+∞ plane.
Z
2
P (λ2 ) = F (Φ) · ei 2 Φ λ dΦ (3.86)
−∞
Resolved foreground screen

This is the trivial case, in which the Faraday-rotating screen has a uniform rotation measure
across the resolution element, i.e. the telescope’s half-power beam width (HPBW). In this
case, the Faraday depth is constant across the beam, hence there is no depolarisation, and
just Faraday rotation:
ψ(~r) = ψ0 (~r) + Φ(~r) · λ2 , (3.87)
where
ψ0 = ψ(λ = 0) (3.88)
is the intrinsic polarisation angle.
Unresolved foreground screen

In case of an unresolved foreground screen, the polarisation angle ψ does not rotate as a
function of λ2 anymore, since in this case there is a superposition of cells within the HPBW
having different rotation depths (Fig. 3.11). Suppose we have random fluctuations of the
magnetic field and/or electron density in a region extending a distance R from the observer.
If the size scale d of the fluctuations is sufficiently small, i.e. d θS · R, where θS is the
angular diameter of the radio source, the Faraday dispersion function is well represented by
a Gaussian with variance K 2 hne Bk i2c d R. Here hne Bk i2c is the variance of the product of the
electron density and the magnetic-field strength along the line-of-sight of a cell. The degree
of polarisation then is
2 hne Bk i2c d R λ4
P (λ2 ) = p0 · e−2 K . (3.89)
5
B.J. Burn, MNRAS 133, 67, 1966
Fig. 3.11 shows how the electric-field vectors are rotated by different amounts, depending
on their line-of-sight to the observer. They correspondingly add up with a zig-zag pattern in
the complex (Q,U )-plane.
The above equation rests upon the assumption that the dispersing cells fill the entire
region. If we now assume a filling factor η < 1 for cells of dimension d with an average
seperation D, then the probability for having m clouds along line-of-sight is η m e−η (m!)−1 .
In this case, the above equation takes the form
Figure 3.11: Illustration of an unresolved foreground Faraday screen.
h 2 2 2 4
i
P (λ2 ) = p0 · exp −η 1 − e−2 K hne Bk ic d λ (3.90)
where we have again assumed that d θS · R. The quantity
d2 · R
η=
D3
is the average number of cells along the line-of-sight. What this equation tells us is that
the fraction e−η not covered by cells is not affected while the rest is depolarised rather quickly.
Internal Faraday depolarisation

In case of Faraday rotation happening within the source, things become more complicated,
the resulting depolarisation being sketched in Fig. 3.12 for different geometries. The simplest
geometry one can think of is a uniform slab, in which the linear path along the line-of-sight
through the source is constant, assumed to be L here. In this case, Burn (1966) has shown
that
1 − e−S
P (λ2 ) = p0 · , (3.91)
S
where
S = (K ne Br )2 d L λ4 − i 2 K ne Bu,k L λ2 . (3.92)
Here, Br stands for the isotropic random field, Bu,k denotes the line-of-sight component
of the uniform field, ne the electron density, d the size scale of the fluctuation, and L is the
thickness of slab. In case of a uniform sphere, the above takes the form
3 [(S + 1) e−S + 21 S 2 − 1]
P (λ2 ) = p0 (3.93)
S3
where L is the diameter of sphere in this case. The properties of the above models depend
on the ratio of the real to imaginary part of the quantity S in the wavelength range at which
most of the depolarisation occurs, i.e. where |S| ≈ 1. This is given by the number of cells
along the line-of-sight N = L/d through the slab and by the ratio of the energy densities in
the uniform to total field strength (Br /2 Bu,k )2 . If the imaginary part of S dominates, i.e.
2
Br
N ,
2 · Bu,k
then we are dealing with a slab having a largely uniform magnetic field, which gives rise
to a depolarisation law of the form
sin Φ λ2
P (λ2 ) = p0 · (3.94)
Φ λ2
where
Φ = K ne Bu,k L (3.95)
Figure 3.12: Dependence of the degree of polarisation and polarisation angle on wavelength
in case of internal depolarisation.
Obviously, Eqn. (3.94) is a sinc function6 . This is not a surprise, since the sinc function
is the Fourier transform of a box function, which we have implicitly used to describe the
simple uniform slab. One can now think of more complicated models, representing regions
with different emission coefficients 1 , 2 , ... and Faraday depths Φ1 , Φ2 , .... Each of these will
have depolarisations
sin λ2 Φ1 i λ2 Φ1 sin λ2 Φ2 i λ2 Φ2
·e , ·e , ...
λ 2 Φ1 λ2 Φ2
If the intrinsic polarisation angle is constant (and set to zero for simplicity), the Stokes
parameters defining the linear polarisation can be written by assigning the cosine terms of
the exponential to Stokes Q and the sine terms to Stokes U :
6 sin π x
sinc(x) := πx
Figure 3.13: Internal depolarisation for three idealised source models.
X j sin λ2 Φj cos λ2 Φj X j sin 2 λ2 Φj

Q= = (3.96)
λ2 Φj 2 λ2 Φj
j j
X j sin λ2 Φj sin λ2 Φj X j sin2 λ2 Φj

U= = . (3.97)
λ 2 Φj 2λ2 Φj
j j
The individual values of Q will be positive and negative, hence tend to cancel, while all
values of U are positive. As a result, the zeros of the function P (λ2 ) disappear, and the
resulting polarisation angle becomes hψi ≈ 45◦ , with damped oscillations about this value.
Finally, we can state that if
2
Br
N , (3.98)
2Bu,k
the the imaginary part of S dominates and Faraday rotation is significant. On the other
hand, if
2
Br
N , (3.99)
2Bu,k
then the real part of S dominates, and the depolarisation goes along without any significant
Faraday rotation as sketeched in Fig. 3.12. In nature, there could be more sophisticated
geometries involving different distributions of the thermal electrons (with number density
ne ) and the synchrotron-emitting relativistic electrons (with number density Ne ). This is
sketched in Fig. 3.13. In Fig. 3.14 an example of depolarisation is shown. The synchrotron
radiation from the galaxy cluster A 2256 (see Chapt. 7 for magnetic fields in galaxy clusters)
at 20 cm wavelength is shown. The central region is characterised by a diffuse ‘halo’, which is
completely depolarised, while the peripheral ‘relic’ is strongly polarised, most likely because
of its location away from, and in front of, the central cluster gas.
Beam depolarisation
Assume that we have uniform field component Bu , with a random component Br superim-
posed. We are then faced with the phenomenon of ‘beam depolarisation’ such that (Burn,
1966)
Bu2
pobs = p(g) · (3.100)
Bu2 + Br2
is observed, where (see Sect. 2.3.3)
g+1
p(g) = . (3.101)
g + 37
The simple reason for beam depolarisation is that different field orientations within the
HPBW give rise to different polarisation angles which then partly cancel. Note that linearly
polarised emission from two regions with magnetic-field orientations perpendicular to each
other cancels out. The reason is that a difference of 90◦ of the polarisation angle in the sky
corresponds to a rotation by 180◦ in the (Q,U )-plane.
Figure 3.14: The galaxy cluster A 2256 at 20 cm wavelength, with the total (synchrotron)
radiation shown in the left panel, and the degree of polarisation in the right one. The
peripheral relic structure is strongly polarised, while the central halo is not.
Bandwidth depolarisation
This effect can be calculated using a slab with a constant dispersion function F (Φ) between
0 and Φ0 :
Φ
R0 2
ei 2 λ Φ dΦ
p(λ2 ) 0
= Φ
(3.102)
p(0) R0
dΦ
0
sin 2 λ2 Φ0 − i cos 2 λ2 Φ0 + i
= (3.103)
2 λ2 Φ0
2
sin λ Φ0 i2ψ
= ·e . (3.104)
λ 2 Φ0
The bandwidth depolarisation is hence given by
sin ∆ψ
p(λ) = p(0) · (3.105)
∆ψ
where (Sec. 3.3.2)
∆ν
∆ψ = −2 λ20 RM · . (3.106)
ν0
At low radio frequencies, bandwidth depolarisation can be very strong, The only way to
avoid it then is by splitting the band into many sub-bands and applying the RM synthesis
technique.
3.3.3 RM synthesis
Doing the Fourier transform of the observed P (λ2 ) has been impossible until recently. The
main reason was the scare coverage of λ2 -space, consisting of a rather limited number of
relatively narrow-band observations at discrete frequencies. This has changed, however, since
in particular at low frequencies radio-astronomical receivers and backends have been developed
with large relative (∆ν/ν) bandwidths and quasi-spectral modes, facilitating a reasonable
coverage and sampling of λ2 -space. Given that one can sample the λ2 space with a decent
coverage by observing at various wavelengths or by splitting the wavebands into individual
channels one can conveniently exploit the relation
+∞
Z
2 2
P (λ ) = F (Φ) · e2 i Φ λ dΦ. (3.107)
−∞
This equation can be inverted if one multiplies its right-hand side by a weight function
W (λ2 ) that vanishes for λ2 < 0 and at all wavelengths at which data are missing. Its inversion
then is
Fe(Φ) = F (Φ) ∗ R(Φ) (3.108)

+∞
Z
2
= K· Pe(λ2 ) · e−2 i Φ λ dλ2 , (3.109)
−∞
where
+∞
Z
2
R(Φ) = K · W (λ2 ) · e−2 i Φ λ dλ2 , (3.110)
−∞
and
1
K= +∞
. (3.111)
R
W (λ2 ) dλ2
−∞
In the above, use has been made

of the convolution theorem. The
weight function accounts for the lim-
ited wavelength coverage. Hence, the
original Faraday dispersion function
F (Φ) that we are after is convolved
with the Fourier transform R(φ) of
the weight function, thus yielding the
modified dispersion function Fe(Φ).
The tilde denotes either the observed
or the reconstructed quantities, and
therefore Fe(Φ) is an approximate re-
construction of F (Φ).
The convolution in Faraday depth
space corresponds to a Fourier filter
in λ2 -space. The quality of the recon-
Figure 3.15: Rotation-measure transfer function.
struction depends primarily on the
behaviour of W (λ2 ). Gaps in the λ2 -space increase the ‘diffraction effects’ in Φ-space. We call
R(Φ) the rotation-measure transfer function (RMTF). The qualitative behaviour of R(Φ) is
shown in Fig. 3.15. A thorough mathematical treatment has been presented by Brentjens & de
Bruyn (2005)7 . The technique of RM synthesis delivers images of Fe(Φ), i.e. one can evaluate
the Faraday dispersion function as a function of Faraday depth Φ, and thus ‘penetrate’ the
magneto-ionised medium. The reconstruction can be greatly improved by deconvolution, us-
ing the measured RMTF. This is similar to the clean algorithms used in radio-interferometry,
leading to a strong reduction or removal of the sidelobes (here: in Faraday-depth space). At
the time of establishing this lecture, RM synthesis is in its infancy and provides the most
powerful tool for exploiting polarisation measurements with the Westerbork Synthesis Radio
Telescope (WSRT) and the newly deployed Low-Frequency Array (LOFAR).
7
M.A. Brentjens & A.G. de Bruyn, Astron. Astroph. 441, 1217, 2005
3.4 Zeeman effect

One of the few tools in astrophysics to directly measure the strength and orientation of
magnetic fields is the measurement of the Zeeman effect. Because of the tiny frequency
splitting, it ‘probes’ the densest regions of the ISM, viz. the dense clouds, in particular
molecular clouds. Flux freezing (Sect. 4.2.2) implies that the magnetic-field strength is high
enough in these regions for the Zeemann effect to become observable.
Its measurement requires the precise determination of the left- and right-circularly po-
larised components of the radio spectral line. Once measured, it permits detailed studies of
the magnetic field in molecular clouds. This kind of information is important for our un-
derstanding of star formation out of molecular clouds, in which magnetic fields may play an
important role in terms of transporting the angular momentum outwards, which is crucial
for any cloud collapse and subsequent star formation to occur. As an example, the Zeeman
splitting of the HI line at a frequency of ν0 = 1420.406 MHz can be measured, delivering three
components called δ1 , representing the left-handed elliptical polarisation, π, representing the
linear polarisation, and δ2 , which represents the right-handed elliptical polarisation. If we
denote the total (unpolarised) intensity of the radiation as I, then one obtains the following
set of measurements:
Frequency Ir Il Component
(1−cos θ)2 (1+cos θ)2

ν0 − ∆ν0 I· 8 I· 8 δ1 (LH ellipt.)
sin2 θ sin2 θ
ν0 I· 4 I· 4 π (linear)
(1+cos θ)2 (1−cos θ)2

ν0 + ∆ν0 I· 8 I· 8 δ2 (RH ellipt.)
Here, θ is the angle between the line-of-

~ and
sight and the magnetic field B,
eB
∆ν0 = , (3.112)
4 π m0 c
where m0 is the electron mass, e the
unit charge, and c the speed of light.
Plugging in numbers, one obtains (for the
neutral-hydrogen line)
∆ν Hz
= 2.8 · cos θ . (3.113)
B µG
This is the Larmor frequency with

Figure 3.16: Illustration of the measurement of
which the magnetic moment of the electron
Zeeman splitting.
precesses in the external magnetic field. It
corresponds to the total splitting between the δ1 and the δ2 component. The above equation
tells us that the Zeeman splitting is extremely small! The typical strength of magnetic fields
in the ISM or in molecular clouds is
B ≈ 10 µG . . . 1 mG.
3.4. ZEEMAN EFFECT 69
which gives rise to a frequency splitting of

∆ν ≈ 28 Hz . . . 2.8 kHz .
This splitting is much smaller than the observed line widths, which are broadened by
pressure and turbulence. The only way of obtaining a useful Zeeman measurement is the
differential measurement of Ir and Il (see Fig. 3.16):
∆I(ν) = Ir (ν) − Il (ν) .

Fig. 3.16 shows that
dIν
∆I = · ∆ν ,
dν
where ∆ν is given by (3.113). Inserting (3.113) we obtain
dI(ν)
∆I(ν) = 2.8 · B cos θ · . (3.114)
dν
The magnetic-field strength B is

taken in µG, and dν in kHz. It should
be pointed out that the shape of the
differential Zeeman curve determines
~ i.e. we can figure
the direction of B,
out whether it points towards or away
from us, since ∆I changes its sign.
It is obvious that, owing to the
very small splitting of the δ1 and δ2
components, radio-astronomical Zee-
man measurements are extremely dif-
ficult and hence require utterly stable
equipment. In Fig. 3.17 two examples
of such measurements of molecular Figure 3.17: Examples of measurements of the Zeeman
clouds are displayed, which demon- effect in interstellar clouds.
strate how feable the circularly polarised components are compared to the total intensity of a
spectral line. The main difficulty is to measure the small splitting in the presence of Doppler
broadening of the emission lines, which is ubiquitous. Expressed in terms of velocity, the
Doppler broadening is
s
k Tg
∆vD = 8 · ln 2 (3.115)
mH
which via
ν0
∆νD = ∆vD · (3.116)
c
translates into a frequency broadening of (e.g.) the HI line of ∆νD ≈ 10 kHz for a gas
temperature of Tg = 100 K. This obviously exceeds the Zeeman splitting by a large amount.
3.5 Polarised emission from dust and molecules

The alignment of dust grains in the galactic magnetic field naturally also leads to polarisa-
tion of their continuum radiation, observed in the mm, submm, or far-infrared regime. The
process that aligns the dust particles has been discussed in Sect. 3.1. Contrary to the optical
polarisation, the polarised continuum from dust should have E⊥ ~ B,~ since the electric field
~
emitted as dipole radiation from elongated dust grains has E primarily along the long axis
of the grains. The expected degrees of polarisation are of order ≤ 10%. For a seminal article
on this topic see Hildebrand (1988)8 .
Weak linear polarisation in rotational emission lines of molecules was first predicted by
Goldreich and Kylafis, with the mathematical details being rather complicated, as the strength
of the effect depends on the ratio of the cloud’s density compared to the critical density of the
transition, on the optical depth of the line, and on the angle between the line of sight and,
the magnetic field9 . A weak magnetic field causes the rotational levels to split into magnetic
sublevels, and unequal populations of the different sublevels lead to partial linear polarisation.
This happens if there is a gradient in the optical depth of a line transition, which may be
caused by a velocity gradient or by an anisotropic radiation field.
However, this so-called Goldreich-
Kylafis effect is hard to detect, with the de-
gree of linear polarisation just above the ar-
tificial instrumental polarisation. Fig. 3.18
demonstrates how difficult it is to achieve
the required accuracy. Shown is the ob-
served linear polarisation observed using
the CS molecule in the envelope of the
evolved star IRC+10216. The derived de-
gree of linear polarisation is 5.1 ± 1.5%.
Still today, the detection of the Goldreich-
Kylafis effect is a challenge, with very few
successful observations reported thus far.
Figure 3.18: Spectropolarimetry of the CSJ =

2 → 1 line towards IRC+10216, with the an-
tenna temperature, the degree of linear polari-
sation, and the position angle plotted vs. the
local standard of rest velocity.
8
R.H. Hildebrand, Q. Jl. Astron. Soc., 29, 327, 1988
9
Goldreich, P., Kylafis, N.D., Ap.J. 253, 606, 1982
Chapter 4
Milky Way
4.1 Diffuse ISM

Polarised radio emission from the
Milky Way was first detected by
Westerhout et al. (1962) and by
Wielebinski et al. (1962). This was
the final proof that the nonthermal
component is synchrotron radiation
(free-free emission preponderates in
the galatic plane and in HII regions).
It has been noted early-on that there
is little correspondence between the
polarised and the total intensity of
the synchroton radiation, the reason
being - as we now know - depolarisa-
tion effects, mainly Faraday disper-
sion (Sect. 3.3.2). The magnetic field
is highly tangled: its ordered com-
ponent produces polarised emission
within the radio beam, while its ran-
dom component, together with the
ordered one, gives rise to the total
synchroton intensity. This leads to
beam depolarisation, since due to this
process the polarised intensity is al-
ways lower than the maximum that
Figure 4.1: Radio emission from the Milky Way at
would be possible. The total syn-
1420 MHz, with total intensity in the top panel, and
chrotron emission is given by the to-
polarised intensity in the lower one. Note the different
tal magnetic-field strength
appearances!
p
Bt = Bu2 + Br2 , (4.1)
and therefore has a rather smooth appearance, while the polarised intensity exhibits a
rather structured picture. Faraday dispersion and beam depolarisation therefore also limit
71
72 CHAPTER 4. MILKY WAY
the ‘polarisation horizon’ very strongly: while in total intensity one measures the synchrotron
radiation all along the line-of-sight through the Milky Way, the distance out to which we see
ordered magnetic fields is rather limited, especially within the Galatic plane. Polarisation
maps of the Milky Way exhibit conspicuous ‘canals’, i.e. thread-like structures that are
strongly depolarised (Fig. 4.1). These are regions of magnetic-field reversals that lead to
counter-Faraday rotation within the beam, and/or regions in which we pick up polarised
radiation from orthogonal orientations of the magnetic field within the beam: both effects
lead to strong depolarisation.
The diffuse synchrotron radiation is dif-
ficult to use as a tool to derive the strength
of the magnetic field. While its determina-
tion from the total intensity requires the
knowledge of the thickness of the source
(Sect. 3.2), which is not known for the dif-
fuse medium, the polarised intensity has a
very ‘limited horizon’, as stated above.
A better and more reliable method to
determine the magnetic-field strength is the
Zeemann effect. From measurements of dif-
fuse HI clouds, a value of hBi ≈ 2 . . . 10 µG
was obtained. Rotation-measure data from
pulsar measurements yield hBi = 2.2 ±
0.4 µG. Here, one utilises both, the rota-
tion measure RM as well as the disper-
sion measure DM , such as to extract the
mean magnetic-field strength, weighted by
the number density of the thermal electrons Figure 4.2: Rotation measures obtained from
pulsars with known distances, superimposed
Rr0 onto a sketch of the Milky-Way spiral arms.
ne Bk dr
RM 0
= K · r0 , (4.2)
DM R
ne dr
0
so that
RM
hBk i = 1.232 · µG . (4.3)
DM
The dispersion measure is retrieved from the dispersion of the pulses received from the
pulsars, caused by the lag in the refractive ionised medium between the pulsar and the ob-
server. In Fig. 4.2 the distribution of rotation measures from pulsars is shown. It should be
mentioned here that minimum-energy or equipartion values derived for external galaxies are
in the range B = 10 ± 4 µG (see Sect. 5.2), so absolutely in line with the above values.
Studying the rotation measure of extragalactic sources probes the whole line-of-sight
through the Milky Way, thus facilitating an assessment of the overall morphology of the
magnetic field. This can then be compared to models, viz. of the Galactic dynamo (see
Sect. 5.6). There have been two studies towards this goal:
4.1. DIFFUSE ISM 73
• A collection of RMs from all kinds of measurements of extragalactic radio sources in

dedicated studies, going back to Kronberg (1980). This has resulted in a couple of
thousand of RMs collected to date.
• An exploitation of the NVSS (NRAO VLA Sky Survey) at 20 cm wavelength. This

has resulted in 37543 RMs (Fig. 4.3), measured at 1435 and 1365 MHz, and has been
published by Taylor et al. (2009).
Figure 4.3: All-sky distribution of rotation measures.
From the Galactic distribution of rotation measures, two important pieces of knowledge
about the Galactic magnetic field have emerged. First, there is a local field, oriented perpen-
dicular to the Galactic plane, with a field strength of B = 0.3 µG below the Galactic plane
(bII < 0) and B = −0.14 µG above the plane (bII > 0). Second, there is a reversal of the
sign of the field strength across the Galactic plane, which is consistent with a quadrupolar
field geometry predicted for the halo field by dynamo models.
The rotation measures of extragalactic sources suggest an overall axi-symmetric configu-
ration of the magnetic field in Galactic disk (see Sect. 5.2 for a discussion of the magnetic-field
configuration in galaxies). We shall come back to these modes when discussing the observed
rotation measures in external galaxies, which provide the more convenient external view. The
current situation has been reviewed by Wielebinski (2005), who points out that constructing
a complete model is not possible at this stage. The current knowledge can be summarised as
follows:
• Magnetic fields are present everywhere in the Milky Way.
• A large-scale magnetic field, directed clockwise, exists in the Perseus spiral arm.
• A field reversal is evident in Sagittarius spiral arm.
• The field strengths are Bu ≈ 5 µG for the uniform field component and Br >
∼ 5 µG for
thw random one.
• The field strength drops as a function of galacto-centric distance, from Bt ≈ 10 µG at

R = 4 kpc, to Bt ≈ 4 µG at R ≥ 15 kpc.
Finally, it should be pointed out that the turbulent magneto-ionic medium can best be
illustrated by juxtaposing the radio continuum maps of the total and polarised radiation us-
ing interferometric data. An interferometer acts as a spatial filter that suppresses large-scale
structures1 . Hence, no large-scale structure is seen in total-intensity images of the Galactic
synchrotron radiation, whereas images of the polarised intensity exhibit copius small-scale
structure. Fig. 4.4 demonstrates this impressively: except for some structure in the Galac-
tic plane, the distribution of total intensity is essentially featureless (the image only shows
thousands of discrete mostly extragalactc - sources), while the polarised intensity discloses
copious plume-like structures emerging from the plane.
Figure 4.4: All-sky maps at 1.4 GHz from the NRAO VLA Sky Survey, with total (top) and
polarised intensity (bottom).
1
see U. Klein: Radio astronomy: tools, applications and impacts. Course astro 841
4.2. MOLECULAR CLOUDS AND STAR-FORMING REGIONS 75
4.2 Molecular clouds and star-forming regions

In order for stars to form, molecular clouds have to contract, which they can do under their
own gravity if there is a sufficient critical mass within a certain volume (Jeans mass). Once
they collapse, they must rid significant angular momentum to contract to a condensed pre-
stellar object2 . These aspects raise two interesting questions:
• Which process impedes the collapse of all molecular clouds on the scale of a Hubble
time?
• Why can stars form at all, given the conservation of angular momentum?
4.2.1 Cloud stability

It is readily shown that the internal thermal pressure and the rotation of molecular clouds
cannot impede their collapse. If we compute the ratio of the internal thermal to the potential
energy, we obtain
−1 −1
G M2

Eth M kT −3 M R T
≈ · = 3 · 10 · · · . (4.4)
|Epot | µ mH R 105 M 25 pc 15 K
Here, k is the Boltzmann constant and µ is the molecular weight. In case of primordial
gas, we have neutral hydrogen plus helium and traces of higher elements, and the molecular
weight is µ = 1.3. In molecular clouds having a solar composition, the molecular weight is
µ = 2.4. Molecular clouds hardly exhibit any signs of systematic rotation. Typical values
deduced for dense cores suggest that clouds are not stabilised against collapse by rotation:
2 3 −1
Ω 2 L3

Erot −3 Ω L M
≈ = 1 · 10 · · · . (4.5)
|Epot | 24 G M km s−1 pc−1 0.1 pc 10 M
Here L is the diameter of the clouds, and Ω their angular velocity. Comparing the potential
energy with the magnetic energy, the result is drastically different:
−1 2 4 −2
B 2 R3 G M2

Emag B R M
≈ · = 0.3 · · · (4.6)
|Epot | 6π R 20 µG 25 pc 105 M
The fact that magnetic fields can support a cloud against gravitational collapse is the result
of ‘flux freezing’ (Sect. 4.2.2). Typical field strengths derived from Zeeman measurements
support this picture: denser clouds or their cores possess stronger magnetic fields, reaching
values of mG there (Table 4.1 and Fig. 4.5). Of course, flux freezing does not imply that the
magnetic field is rigidly tied to the molecular clouds, otherwise they would not be able to
contract at all. Through a process called ‘ambipolar diffusion’ (Sect. 4.2.3) the magnetic field
manages to ‘slip through the gas’.
It is important to note that the random component of the interstellar magnetic field also
plays a role here: if the magnetic field were entirely uniform, then the gas would simply slip
along the field lines, leading to a flat (‘pancake’-like) morphology upon contraction. However,
this is never observed. The reason must be the random component, preventing such an
anisotropic contraction.
2
For a full treatment of this topic see the textbook by Stahler & Palla: ”The Formation of stars”, Whiley,
2004.
Object Type Diagn. Bk [µG]

Ursa Major diffuse cloud HI +10
L204 dark cloud HI +4
NGC 2024 GMC clump OH +87
Barnard 1 dense core OH -27
S 106 HII region OH +200
Sgr A/West molecular disk HI -3000
W75 N maser OH +3000
DR 21 HII region CN -360
Table 4.1: Magnetic-field strengths inferred from Zeeman absorption measurements in Galac-
tic clouds (see also Fig. 4.5).
Figure 4.5: Zeeman measurements in DR 21 (CN, left) and in S 106 (OH line, right.
4.2.2 Flux freezing

A magnetic field threading a molecular cloud must be generated by a current density ~j, which
~ by Ampere’s law:
is related to the magnetic field B
∇ ~ = 4 π · ~j
~ ×B (4.7)
c
Here, the displacement current 1c · ∂E

∂t has been neglected in view of the low frequencies of
interest. The current is carried by free electrons and ions (charged dust grains have too low
number densities to be significant). Therefore, we can write it in the form
~j = ni e v~i − ne e v~e ≈ ne e (~
vi − v~e ) (4.8)
Here, it was assumed that singly charged species (F e+ , M g + , . . .) preponderate, so that

charge neutrality requires ni = ne . In order to see how the current is generated by the
local electric field, we generalise Ohm’s law (which holds for a medium at rest) by shifting
temporarily to a reference frame that is moving at the local velocity ~v of the neutral matter.
Indicating this reference frame by primes, we have
~j 0 = σ · E
~ 0 = ~j, (4.9)
where σ is the electric conductivity. The second equality results from (4.8), noting that
neither ne nor v~i − v~e can change in the new reference frame. It should be pointed out that we
have neglected relativistic corrections of order (v/c)2 . To the same accuracy, the new electric
field is then given by
E ~ + ~v × B.
~0 = E ~ (4.10)
c
Eqns. (4.9) and (4.10) imply that the generalised Ohm’s law now reads

~j = σ · E~ + ~v × B
~ . (4.11)
c
We can now insert this current density into Ampere’s law to obtain

~ ~ 4πσ ~ ~v ~
∇×B = · E+ ×B . (4.12)
c c
~ by using Faraday’s law,

We can finally eliminate E
~
∇ ~ = − 1 · ∂B ,
~ ×E (4.13)
c ∂t
and obtain
~
∂B

c2 ~

~ × (~v × B)
=∇ ~ −∇
~ × ~
∇×B . (4.14)
∂t 4πσ
Eqn. (4.14) is the fundamental magneto-hydrodynamic (MHD) equation for the magnetic
field. The second term on the r.h.s. describes the Ohmic dissipation, which vanishes if the
conductivity becomes very large. The electric conductivity σ is a measure of the friction of the
charged particles in the conductor, i.e. of its resistance. The particles interact via Coulomb
collisions, the frequency of which depends on the particles’ density and temperature. The
conductivity hence also depends on that frequency:3
e2 ne
σ= (4.15)
me νcoll
The MHD equation is also frequently referred to as the induction equation. Recalling that
~ × (∇
∇ ~ × B)
~ =∇
~ (∇
~ · B)
~ − ∆B
~ = −∆B
~ (4.16)
3
L. Spitzer, 1956: ”The Physics of Fully Ionized Gases”, 2nd edition, Dover Pubn. Inc., 2006.
and defining the magnetic diffusivity
c2
η= , (4.17)
4πσ
we have, for a homogeneous diffusivity,
~
∂B ~ × (~v × B)
=∇ ~ + η ∆B
~ (4.18)
∂t
or, for an inhomogeneous one
~
∂B ~ × (~v × B)
=∇ ~ −∇
~ × η (∇
~ × B).
~ (4.19)
∂t
Assuming that ions and electrons have the same temperature, the resulting conductivity
becomes
3
T 2
σ ≈ 10 18
· s−1 . (4.20)
104 K
Let us assume static case, i.e. ~v = 0, and a constant diffusivity. Then
~
∂B ~
= η ∆B (4.21)
∂t
For an order-of-magnitude calculation, we simply rewrite the differential equation (4.21)
in the simple form
B B
=η· ,
τdif f L2
where τdif f is the diffusion time scale and L is a length scale over which the diffusion
happens. This leads to an estimate of the diffusion time, which is the time for the magnetic
field to decay, i.e.
L2
τdif f ≈ . (4.22)
η
For galaxies, this implies

2 3
27 L T 2
τdif f ≈ 10 · yr! (4.23)
1 kpc 104 K
What a large time scale! If this were true, then we would not expect magnetic fields to
change in galaxies over many orders of magnitude of a Hubble time. The situation changes,
however, dramatically if turbulent motions are added to the system. If ~v 6= 0 the magnetic field
does not at all survive that long. Let us apply this to the Sun, where we have L ∼ 1.4·1011 cm
and T ∼ 107 K in the corona, we obtain τdif f ≈ 108 yrs, while the solar magnetic cycle is a
mere 22 yr - an obvious contradiction unless there are (turbulent) motions at play.
The evolution of magnetic fields can be calculated

now, given any initial configuration once we know η and
~v . Neglecting the right-hand term of the hydrodynamic
equation (4.14) reduces it to the ideal MHD equation
~
∂B ~ × (~v × B),
=∇ ~ (4.24)
∂t
which is the mathematical representation of flux
freezing. To see why, we imagine a closed loop C of
comoving gas encircling a surface S (see Fig. 4.6). We
use the Lagrangian derivative
D ∂ ~
= + ~v · ∇, (4.25)
Dt ∂t
which is the comoving rate of change of a quantity
in a particular fluid element momentarily located at Figure 4.6: Illustration of flux freez-
~r. This operator describes the rate of change at that ing.
location ~r, plus the spatial derivative in the direction of the fluid velocity multiplied by the
magnitude of the fluid velocity. The flux is
Z
Φ= ~ · dS,
B ~ (4.26)
S
and its rate of change then is
DΦ
Z
∂B~ I
= · dS + (~v × d~l) · B.
~ ~ (4.27)
Dt S ∂t C
~ as a function of time:
There are two contributions to the change of B
• There is a change in the magnetic flux density, due to external causes (e.g. an external
force moving the loop), i.e. the change is simply

Z
∂B~
~
· dS
S ∂t
ext
• The motion of the loop induces an electric field
~ = ~v × B.
E ~ (4.28)
c
Then, because
~
∇ ~ = − 1 · ∂B ,
~ ×E (4.29)
c ∂t
there will be a contribution

Z
∂B~ Z
~
· dS =− ∇ ~ × ~v × B
~ · dS.
~
S ∂t S
move
Adding these two contributions then yields

D
Z Z ~
∂B
Z
~ · dS
B ~ = ~−
· dS ~ × ~v × B
∇ ~ · dS~ (4.30)
Dt S S ∂t S
Z " ~ #
∂B ~ ~

~ =! 0.
= − ∇ × ~v × B · dS (4.31)
S ∂t
So, having used Stokes’ theorem (also referred to as Gauss’ theorem), we see that the flux
through the comoving surface is constant in time.
We can now conceive how magnetic fields help to support molecular clouds to not collapse
instantly, i.e. within much less than a dynamical time scale. The question now arises why
then stars may form at all from gravitational collapse. The way out is ambipolar diffusion,
which describes how the neutral gas can ’slip through’ the plasma - which is tightly coupled
to the magnetic field - and collapse in a slowed-down mode.
4.2.3 Ambipolar diffusion

In the last phase of the contraction of a cloud towards a forming star the magnetic field
that has helped to get rid of angular momentum must be able to ’slip through’ the neutral
gas. Otherwise, flux conservation would lead to excessively large field strengths, which would
impede the final collapse and which are not observed anyway. This process is referred to
as ambipolar diffusion. Resisting the relative drift of the neutrals w.r.t. ions, there will be
a frictional (drag) force (per unit volume), due to collisions between the ions and neutrals
(electrons can be neglected here as they carry much less momentum):
f~d = γ ρn ρi · (~
vi − v~n ), (4.32)
where ρn is the mass density of neutrals, ρi that of the ions, and γ the drag coefficient.
To see how this equation comes about, let us first calculate the collision rate of the ions with
any neutral:
νcoll,i = ni · hω σin i . (4.33)
Here, ni is the number density of ions4 , ω is relative velocity of the ions as seen from rest
frame of the neutrals, σin is the cross section for inelastic scattering between the ions and the
neutrals, and hi stands for averaging over the distribution function of the ions. Denoting with
mi the mass of ions and with mn the mass of the neutrals, each collision transfers momentum
from the ion to neutral that is equal to ∆~ p = mi · (~vi − v~n )× the fraction of mass of the
collision pair contained by the neutral, i.e. mn /(mi + mn ). In actual molecular clouds, we
have mn ≈ 2.3 · mH and mi ≈ 29 · mH . We can thus write the momentum ∆~ p transferred per
unit volume and unit time ∆V from the ions to the neutrals as
∆~
p mn mi !
νcoll,i · = nn · vi − v~n ) · ni · hω σin i = f~d
· (~ (4.34)
∆V mn + mi
with
ρn = mn · nn , ρi = mi · ni .
4
In dense molecular clouds, HCO+ is the dominant ion.
We finally recover Eqn. (4.32) and realise that
hω σin i
γ= . (4.35)
mn + mi
Draine et al. (1983) found
γ = 3.5 · 1013 cm3 g−1 s−1 . (4.36)
Ionisation in molecular clouds is caused by cosmic rays. Theoretical calculations show

that for densities nn ≈ 104 cm−3 the fractional ionisation is ni /nn ≈ 10−7 , in good agreement
with what is derived observationally for molecular clouds (i.e. deriving abundances of neutrals
and ions). Now we calculate the drift speed. The force on a charged particle is
~ + q · ~v × B.
F~ = q · E ~ (4.37)
c
In order to calculate the force on a fluid of charged particles it is convenient to replace
charge by charge density and q ~v by current density:
q → σ, q~v → ~j.
Then the force per unit volume is the Lorentz force5
~ + 1 · ~j × B
f~L = σ · E ~ (4.38)
c
and, noting that in general E 2 /B 2 1, we have
1
f~L = ~j × B.
~ (4.39)
c
Using Maxwell’s equation
∇ ~ = 4 π · ~j,
~ ×B (4.40)
c
we finally derive for the Lorentz force density
1 ~
f~L = ~ × B.
(∇ × B) ~ (4.41)
4π
The dominating forces on the charged particles are now the drag and the Lorentz force,
all others can be neglected. These two forces must add to zero, i.e. f~d = f~L , and hence
1 ~ ~ × B.
~
γ ρn ρi · (~
vi − v~n ) = (∇ × B) (4.42)
4π
Then, to order of magnitude, the drift velocity vd = vi − vn is
B2 v2
vd ≈ ≈ A (4.43)
4 π γ ρn ρi L V
5
It is a bit inconvenient to the reader that at this point we are using the symbol σ for the electric conductivity,
the charge density, and for cross sections.
where we have introduced the Alfvén speed
B
vA = √ (4.44)
4πρ
by assuming ρ = ρn + ρi ≈ ρn for the low ionisation degrees (s.a.). The speed V contains
the product
V = γ ρi L.
Inserting typical values, i.e. L = 0.1 pc and nn = ρ/mn = 104 cm−3 , we obtain V =
6 km s−1 . With B ≈ 30 µG we have vA ≈ 0.4 km s−1 V , and the drift speed is a mere
vd = 27 m s−1 . The time scale for ambipolar diffusion then is
L
tAD ≈ ≈ 3.6 · 106 yr, (4.45)
vd
using the above quantities. This is an important process, as the time scale tAD is of the
same order of magnitude as the free-fall time of the cloud collapse,
1 n − 1
3π 2
n 2
τf f = = 3.1 · 107 · yr. (4.46)
32 G ρn cm−3
Note that the tight coupling of the ions to the magnetic field is hardly disturbed by the
collisions, since the collision frequency (or rate) is much lower than the Larmor frequency.
Hence, the ions may take many revolutions about the magnetic field before being knocked off
by a neutral. In Fig. 4.7 examples of magnetic-field structures observed in molecular clouds
are shown.
Figure 4.7: Examples of magnetic-field structures observed in molecular clouds. Left:

magnetic-field strength in S 106, obtained from OH observations at λ = 18 cm conducted
with the VLA. The contours depict the continuum emission. Right: structure of the magnetic
field in DR 21, inferred from polarisation measurements of the dust emission at λ = 1.3 mm
and in the 12 CO(1-0) line with BIMA.
4.3. SUPERNOVA REMNANTS 83
4.3 Supernova remnants

Among the brightest radio sources in the sky
are supernova remnants (SNR). These were
also among the first ‘discrete sources’ discov-
ered at radio frequencies, such as Cas A (or
3C 461), a nice radio image of which is pre-
sented in Fig. 4.8, showing the expanding relic
of a supernova explosion that occured6 in 1667,
and Tau A (3C 75, M 1), seen in 1054 by Chi-
nese astronomers. There are basically two
types of SNR:
• plerions (‘filled’ SNR)
• shell-type SNR.
From optical spectroscopy and VLBI ob-
servations one deduces the initial expansion
speeds, reaching vexp ≈ 20 000 km s−1 . Their Figure 4.8: Radio continuum image of the
shells initially expand freely, owing to their SNR Cas A obtained with the VLA at 5 GHz.
much larger density as compared to their sur-
roundings. A collisional shock is expected when a shell has swept over distance that is about
equal to the mean free path of the protons. A proton moving with a speed of v = 20 000 km s−1
has a kinetic energy of Ekin ≈ 2 MeV. Its mean free path can be calculated as follows. The
mean free path is given by the path it takes a cosmic-ray proton to lose its kinetic energy via
ionisation of the hydrogen atoms or H2 molecules. The cross section for this is (Bethe, 1933):
 
4
2π e Z 2  2 me v 2 v 
2
σCR = · 0.285 · ln + 3.04 − , (4.47)
m e χ0 v 2
h 2 i
 1− v ·χ c 
c 0
where χ0 the ionisation energy (13.6 eV in our case) and v the speed of the ionizing
protons. With v = 2 · 109 cm s−1 and Z = 1 we find σCR ≈ 10−17 cm2 , and with a mean gas
density of n̄HI ≈ 1 cm−3 we obtain a path between two subsequent ionisations of li ≈ 0.03 pc.
Since the proton loses only about 10−4 of its kinetic energy in each ionisation, the total mean-
free path becomes l = 104 · li = 300 pc. Obviously, no classical shock would develop under
such conditions. However, in the presence of a magnetic field of strength B ≈ 5 µG, these
protons have gyration radii of (Sect. 2.3.1)
mvc
rL = ≈ 4 · 1010 cm = 10−8 pc. (4.48)
eB
They are therefore tightly coupled to the magnetic field, which thus gives rise to a massless
barrier! The result is a hydromagnetic or MHD shock. The details of scattering in this process
are as yet not fully understood. The supernova blast wave will thus plow through the ISM and
compress the magnetic field, thus enhancing its strength within the shell. We schematically
distinguish four phases in the evolution of an SNR7 .
6
This means when the light of the event reached the earth.
7
Not all SNR fit into this scheme. For instance the Crab Nebula, and all plerions possess a morphology
that is dominated by a pulsar wind, which transfers rotational energy into the SNR.
Free expansion
At this early stage, the shell’s radius behaves like
R ∝ t,
and the mass of the swept-up gas is still less than

the initially injected mass MS of the shell
4
π RS3 ρext ≤ MS
3
Assuming for the surrounding gas density a value
of ρext ≈ 2 · 10−24 g cm−3 and an initial shell mass of
MS = 0.25 M , we obtain RS ≈ 1.3 pc, so this phase
will last for about
RS
t1 ≈ ≈ 60 yr
vexp
This phase can be observed in extragalactic envi-

ronments using VLBI-monitoring observations. Their
monochromatic radio luminosities decrease with a
power-law as a function of time, with emission ap-
pearing at a high frequency first, and showing up at
lower frequencies as time elapses (Fig. 4.9), because the
medium becomes less opaque.
Adiabatic phase
When the swept-up mass becomes greater than the Figure 4.9: VLBI observations of
ejected mass, the dynamics are decribed by the adi- the supernova SN1993j, showing its
abatic similarity solution that goes back to Sedov expansion (upper panel). As time
(1959)8 . At this stage, the temperature of the shocked elapses, the supernova emits syn-
gas is so high that its radiation is relatively weak. Hence chrotron radiation at higher fre-
the only important energy loss is through the adiabatic quencies (lower panel).
expansion of the gas. This phase terminates when the
temperature drops below T ∼ 106 K. During this phase, the radius of the shell behaves as
1 2
R ∝ n− 5 · t 5 . (4.49)
For the case of spherical symmetry, the one-dimensional numerical solution of the hy-
drodynamic equations has been presented by Chevalier (1974)9 , yielding a relation between
the supernova energy E0 , the surrounding gas density n0 , and the expansion speed and the
radius, such that
n 1.12 v 1.4 R 3.12
43 0
E0 = 5.3 · 10 · · · erg (4.50)
cm−3 km s−1 pc
8
L.I. Sedov, ”Similarity and Dimensional Methods in Mechanics”, Acedemic Press, New York, 1959.
9
R.A. Chevalier, Astroph. J. 188, 501, 1974.
4.3. SUPERNOVA REMNANTS 85
During this phase the total energy E0 , consisting of thermal energy and mass motion, is
constant. This phase is also called Sedov phase, as it goes back to a calculation describing
similarities in blast waves (nulcear bombs, supernovae, ...). Since this phase terminates when
the temperature drops below T ∼ 106 K (at which point radiative losses become important),
its duration can be estimated from
n − 2 t − 65 E
1
5
11 H 5
T = 1.5 · 10 · · K. (4.51)
cm−3 yr 4 · 1050 erg
Inserting nH = 1 cm−3 and E = 4 · 1050 erg, this yields t = 2 · 104 yr.
Isothermal or radiative phase

The main energy loss now is through radiation:
at a temperature of T ≤ 106 K, the ions of C, N
and O begin to recombine with the free electrons.
The collisional excitation of ions so formed and
the line radiation they emit increase cooling rate
of the gas by orders of magnitude. Hence, the
shock becomes radiative. The radius of the SNR
is now RS ≈ 15 pc for a density of nH = 1 cm−3 ,
the age is ∼ 4 · 104 yr, and the expansion speed
has dropped to vexp ≈ 85 km s−1 . Conservation
of momentum implies that M · vexp = constant,
i.e.
4 Figure 4.10: Illustration of the different

π RS3 ρext vs = const,
3 evolutionary phases of SNR.
which by integration yields
1
8 t 3 4
RS = Rrad · · − . (4.52)
5 trad 5
Here, Rrad and trad refer to the commencement of this radiatve phase. This last phase
ends with the dispersion of the shell when the expansion speed drops to ≤ 9 km s−1 , which
is the velocity dispersion of the ISM. Fig. 4.10 summarises the four evolutionary phases of
SNR.
The morphology of magnetic fields in SNR is kind of a mixed bag, even in shell-type
remnants. Some of them exhibit a radial, others a tangential field orientation. However, if
one subdivides them into young and old, the former show preferentially radial fields, while
the latter have tangential fields. The radial fields are probably caused by Rayleigh-Taylor
instabilities in the young, rapidly expanding remnants, as is indicated by MHD-simulations,
while the tangential field pattern in old remnants is likely to be the result of field compression
in the shock wave.
4.4 Acceleration of cosmic rays

Cosmic rays (CR) play an important role in plasma astrophysics. The enrichment of the intra-
cluster medium (ICM) and the intergalactic medium (IGM) with heavy elements, and along
with this also the magnetisation of the ICM or IGM, must have been furnished very early in
the universe via galactic winds (Sect. 5.5) produced by starburst galaxies, so dwarf galaxies in
the first place. The origin of cosmic rays is to some extent still a matter of speculation. While
particles with energies up to 1014 eV may well have been produced by SN shock fronts, those
with higher energies (> 1014 eV) still pose a problem in this respect. Active galactic nuclei
(AGN, Chapt. 6) are candidates for being factories, while the highest energies are likely to be
provided by γ-ray bursters (GRBs). These so-called ultra-high energy cosmic rays (UHECR)
have energies of up to E ∼ 1021 eV, and their sites of origin must be searched for within the
so-called GKZ horizon10 (Greisen, Zatsepin & Kuzmin 1966).
Cosmic-ray electrons are important for the astrophysicist because they reveal much of
the relativistic plasmas on all astrophysical scales by virtue of synchrotron radiation, and
we always must keep in mind that each relativistic electron is accompanied by a relativistic
proton, with its ∼ 2000 larger rest mass. It is hence justified to at least briefly consider
some basics of CR acceleration, while a full treatment would be beyond the scope of this
lecture. How much energy can a supernova and its remnant provide to the CR particles?
As an example, take Cas A, the brightest radio source in the nothern sky. Its flux density
is Sν = 2720 Jy at 1 GHz. With a distance D = 2.8 kpc, its monochromatic luminosity is
P1GHz = 2.6 · 1018 W Hz−1 . Using the minimum energy formula (3.50), we obtain Emin =
4
2.6 · 1048 η 7 erg. If the average time span between subsequent supernova explosions is tSN
and the escape time for particles, i.e. the time they need to leave the confinement volume, is
tc , then the mean energy transferred to the cosmic rays in a volume V corresponds to their
local energy density uCR , i.e.
tC E0
uCR = · erg cm−3 . (4.53)
tSN V
Here, E0 is the average energy release in
high-ernergy particles per supernova. Sup-
pose that the lifetime of such particles in
the Galaxy disk is 107 yr (Eqn. 2.160).
With a thickness of 700 pc and a radius of
10 kpc the volume becomes 6 · 1066 cm3 .
The local energy density of high-energy
particles is ∼ 1 eV cm−3 . Adopting a SN
rate of about 1 every 30 yr, then the average
energy release becomes ∼ 3·1049 erg, which
should be compared with the energy Emin
as derived above, using minimum-energy Figure 4.11: Illustration of particle trapping in
arguments for the relativistic particles and a ‘magnetic bottle’.
the magnetic field. With η ≈ 100, these
energies become comparable!
10
Particles (protons) with energies as high as this interact with the CMB photons, which they ‘see’ as γ-rays
in their rest frame. The CMB hence becomes opaque to them on a scale of about 50 Mpc. The interaction
implies photo-pion production.
4.4. ACCELERATION OF COSMIC RAYS 87
We now briefly discuss the process of particle acceleration by multiple shocks in the ISM.
This mechanism was first treated by E. Fermi (1949) and is referred to as 2nd-order Fermi
acceleration, meaning that it deals with energy gains that are ∝ (v/c)2 . In Fermi’s original
picture, charged particles are reflected from ‘magnetic mirrors’ associated with irregularities
in the Galactic magnetic field11 . The particles may get trapped in a ‘magnetic bottle’ and are
reflected forward and backward between moving clouds, which have a frozen-in magnetic field
(Fig. 4.11). It will be shown here that this process establishes a power-law energy distribution
of the accelerated particles, as is observed.
Let us calculate the energy gain of a
particle for an angle θ between the initial
motion of the particle and the normal to
the surface of the mirror surface (the cloud
surface, as sketched in Fig. 4.12). It is im-
portant here to carry out a proper relativis-
tic analysis. We assume the mass M of the
cloud to be infinitely large compared to the
particle’s mass m, M m, so that the
cloud’s velocity V is unchanged in the col-
lision. The centre-of-momentum frame is
therefore that of the cloud. The energy of Figure 4.12: Sketch of the geometry in Fermi-II
the particle in this frame is acceleration, with head-on (a) and following (b)
collision.
0
E = γV · (E + V p cos θ), (4.54)
where
− 21
V2

γV = 1 − 2 . (4.55)
c
is the Lorentz factor referring to the cloud’s motion. Here, the primed quantities refer
to the cloud’s frame, which is also the centre-of-momentum frame. The x-component of the
relativistic three-momentum in the centre-of-momentum frame is

0 0 VE
p x = p cos θ = γV p cos θ + 2 . (4.56)
c
0
In the collision, the particle’s energy is conserved, i.e. Ebef 0
ore = Eaf ter , and its momentum
0 0
in the x-direction is reversed, i.e. px → −px . Transforming back to the observer’s frame, we
hence find
E 00 = γV (E 0 + V p0x ). (4.57)
Inserting E 0 and p0x from above and recalling that
px cos θ
=v· 2 (4.58)
E c
11
1st-order Fermi acceleration happens when only head-on collisions occur.
we find
" 2 #
2 V v cos θ V
E 00 = γV2 E · 1 + + . (4.59)
c2 c
Expanding this to second order in Vc , we arrive at

" 2 #
00 2 V v cos θ V
E − E = ∆E = +2 ·E (4.60)
c2 c
We now need to integrate over θ. The probabilities of head-on and following collisions are
proportional to the relative velocities of approach of the particle and the cloud:
P ∝ v + V cos θ head on
P ∝ v − V cos θ following
Since v ≈ c, the propabilities are proportional to 1+V /c cos θ, where 0 < θ < π. Recalling
that the probability for the angle to lie in the range θ and θ + dθ is proportional to sin θ dθ,
we find, setting x = cos θ and averaging over all angles in the range 0 to π for the first term
of ∆E in (4.60):
+1
V
R
x· 1+ c x dx 2
2 V cos θ 2V −1 2 V
= +1
= . (4.61)
c c R V
3 c
1+ c x dx
−1
Thus, in the relativistic limit the average gain per collision is

2
∆E 8 V
= · . (4.62)
E 3 c
This illustrates the famous result derived by Fermi, namely that the average increase in
energy is only 2nd order in V /c. It is immediately apparent that this leads to an exponential
increase in energy since the same fractional increase occurs in each collision. If the mean free
path between the ‘mirrors’ along a magnetic-field line is L, then the time between collisions
is L/(c cos φ), where φ is the pitch angle of the particle w.r.t the the magnetic-field direction.
The average over cos φ yields the average time between the collisions, which is just 2 L/c.12
The typical rate of energy increase is therefore
2
dE 4 V
= · ·E (4.63)
dt 3 cL
The so-called diffusion-loss equation (Sect. 5.5) reads
dN ~ 2 N (E) + ∂ [b(E) · N (E)] − N (E) + Q(E).

= D∇ (4.64)
dt ∂E τesc
D E R
12 L L d(cos φ) 2L
c cos φ
= c
· R
cos φ d(cos φ)
= c
.
4.5. UHECR 89
This equation describes the rate of change of the particles in a given volume, where the
first term on the left-hand side accounts for diffusion, the second one for the energy losses or
gains, the third one for the escape of particles from the volume, while the last term is the
source term. We are interested here in the steady-state solution, i.e. we set dN/dt = 0. We
are not interested in diffusion, hence we set D∇ ~ 2 N = 0. In our case, the energy loss term
becomes a gain term and is given by
dE
b(E) = − = −a E. (4.65)
dt
Hence, the diffusion loss equation becomes
d N (E)
− [a E · N (E)] − = 0. (4.66)
dE τesc
Differentiating and integrating, we find

dN (E) 1 N (E)
=− 1+ · , (4.67)
dE a · τesc E
from which we hence obtain
N (E) dE = A · E −g dE, (4.68)
where
1 τacc
g =1+ =1+ . (4.69)
a · τesc τesc
This is the power-law that is found for the CR electron energy spectrum and that we
made use of in Sect. 2.3.2. The quantity a in the above equation can be considered as the
inverse acceleration time scale so that the slope g of the particle energy spectrum depends
on the relative balance between acceleration and loss time scales. The full and more realistic
calculations are a lot more complicated and have to account for the - here neglected - energy
losses of the accelerated particles, which would lose much of their acquired energy via ionisa-
tion losses in the first place. Furthermore, in the above treatment we have omitted the fact
that cloud motions in the ISM have a mere V /c ≈ 10−4 . In a full treatment, one needs to take
the energy losses into account and assume supernova shock fronts as the main accelerators
(O’Drury, 1983). Nevertheless, the main result, viz. a power-law for the energy spectrum
of relativistic electrons results in what is called the theory of diffusive shock acceleration. It
provides particle spectra up to energies relevant for the Galatic synchrotron radiation. Note
that even though in Fermi-II acceleration the particles lose energy in the following collisions,
there is always a net energy gain since the head-on collisions are somewhat more frequent.
4.5 UHECR
The measured CR energy spectrum (Fig. 2.17) exhibits particles with up to energies of ∼
1021 eV! Such particles have been coined ultra-high energy cosmic rays (UHECR). Their
origin is still unknown. The Larmor radius
E
rL =
eB
of a proton with this energy in a 1-µG magnetic field is about 1 Mpc! Hence, if these
particles originated within the Milky Way, their site of origin should be found within an angle
L
∆θ = ,
rL
where L is the traversed (curved) path on the way to us. Again, with B ≈ 1 µG, we
should find the responsible particle accelerator within ∼ 10◦ of the direction in which the CR
has been detected (with the Auger experiment, see Fig. 4.13). However, the recorded events
are distributed more or less isotropically across the sky, with no preference of locations in the
the Galactic plane. They must therefore be of extragalactic origin, also because there is not
likely to be any process within the Milky Way that could provide the required energies.
On the other hand, the particles must stem from a volume within the so-called GKZ
limit (Greisen, Kuzmin, Zatsepin). The reason is that protons with an energy E > 5 · 1019 eV
interact with the CMB photons, since in the proton restframe these photons appear as γ-rays,
with which they interact via the so-called ∆ resonance:
γ + p → ∆+ → p + π 0 (4.70)
In the ∆ resonance, a baryon13 with mass m∆ = 1.232 MeV/c2 decays into a proton and a
π0 (or a neutron and a π + ), the resulting proton having about 20% less energy. This process
continues until the proton energy falls below the GKZ threshold. Hence, the CR spectrum
should exhibit a so-called GKZ cutoff at the highest energies, which appears to have been
confirmed. As a result of this (strong) interaction, the highest-energy particles cannot reach
us from distances beyond about 50 Mpc, the so-called GKZ horizon.
Figure 4.13: UHECRs recorded with the Auger array.
13
It consists of ūūd quarks and has spin +3/2.
Chapter 5
External galaxies
5.1 Galaxies
The hadronic constituent of galaxies consists of gas and stars, with only a minor contribution
of dust. These components rotate in a gravitational potential made up by the dark matter.
In spiral galaxies, the gas has a neutral and an ionised component. The neutral component
is made up by (in order of increasing density) the molecular, the cold, and the warm phase.
The ionised component consists of the warm and the hot phase (Table 5.1).
Gas phase n [cm−3 ] T [K] fV [%] fM [%] h [pc]

Molec. 102 ...105 10 . . . 50 <1 ∼ 20 ∼70
CNM 40 . . . 80 50 . . . 200 2...4 ∼ 40 ∼140
WNM 0.1 . . . 0.6 5500 . . . 8500 ∼ 30 ∼ 30 ∼400
WIM ∼ 0.2 ∼ 8000 ∼ 20 ∼ 10 ∼900
HIM 10−3 ...10−2 105 . . . 107 ∼50 ∼1 ≥1000
Table 5.1: Gas phases in galaxies, with number densities, temperature, volume and mass
filling factor, and scale height of the disk.
In elliptical galaxies, the bulk of the gas is ionised (HIM). An exception to this is neutral
gas that has been captured from the surroundings or from neighbouring galaxies. The ionised
gas in galaxies immediately implies a high conductivity, which in the absence of turbulence
would result in a long-lived sustainment of magnetic fields.
Even though magnetic fields cannot have any dynamical significance on large scales1 , they
may play their role on smaller, local scales. For instance, as we have seen in Sect.4.2.1, they
must have a strong influence on the star-formation process. Another area in which magnetic
fields must play a cardinal role is cosmic-ray propagation and containment within a galaxy
(Sect. 5.5). Cosmic rays are produced in supernovae and are accelerated by their shock
fronts. Since they consist of charged particles (p, e− , ions), they are tightly coupled to the
magnetic field, which in turn governs their propagation. If the CR pressure is sufficiently
high, exceeding that of other pressure terms (also magnetic), they may escape from a galaxy
in a galactic wind (Sect. 5.5).
1 ~
There have been discussions in the literature as to whether B-fields could be responsible for flat rotation
curves, but this is flatly impossible in view of their energy density!
91
92 CHAPTER 5. EXTERNAL GALAXIES
In what follows, we shall discuss the properties of magnetic fields seperately for spiral,
dwarf irregular, and elliptical galaxies, as these have distinctly different properties concerning
their kinematics or dynamics and their gas content and distribution. Spiral galaxies possess
stellar and gaseous disks in which density waves can be excited, depending on the depth of
their gravitational potentials. Dwarf irregulars have shallow gravitational potentials, their
kinematics is therefore possibly influenced by local disturbances caused by star formation
and subsequent supernovae. Large elliptical galaxies have their baryonic mass dominated by
the stellar constituent, the hot gas forming a hot corona. They do not exhibit any systematic
rotation as seen in disk galaxies. Radio-quiet ellipticals are in the focus of this chapter, i.e.
those ellipticals that lack a central AGN. The radio-loud ones will be treated seperately in
Chapt. 6, since the physics of magnetic-field creation in these is fundamentally different from
that in radio-quiet ones.
5.2 Spiral galaxies

Magnetic fields in disk galaxies have been stud-
ied since late 70’s/early 80’s via the (polarised)
synchrotron radiation and, to a lesser extent,
via optical polarimetry. It soon turned out that
spiral galaxies are pervaded by large-scale, or-
dered magnetic fields, with field strengths of or-
der ≤ 10 µG. Similar findings were made a few
years later for low-mass, or dwarf irregular, galax-
ies. In the case of spirals the most conspicuous
phenomenon is that magnetic field closely follows Figure 5.1: Illustration of axi- and bisym-
the spiral patterns. There are two fundamental metric magnetic-field configurations.
magnetic-field configurations that are preponder-
ant in spiral galaxies:
• axisymmetric fields
• bisymmetric fields
They have been looked for early-on, making use of their different signatures in the observed
rotation measures of spirals. The motivation to search for these came from predictions of
calculations of the galactic dynamo (see Sect. 5.6). For an axisymmetric, differentially rotating
disk galaxy, the rotational invariance of the dynamo equations (s.b.) suggests a Fourier
decomposition of the from:
X
~ r, t) =
B(~ Bm (r, z, t) · eimΦ (5.1)
m
where m is an integer. It defines different modes that a galactic dynamo can actually
produce. The most fundamental modes are
m = 0 : axisymmetric mode
m = 1 : bisymmetric mode
5.2. SPIRAL GALAXIES 93
Higher modes may be present, too, but are expected to have

small amplitudes. With respect to the galactic plane, dynamo
models allow for even or odd configurations of the toroidal field
structure w.r.t. the galaxy plane (Fig. 5.2), with the following
nomenclature:
Configuration ASS BSS

even S0 S1
odd A0 A1
In the jargon of galactic magnetism, the horizontal disk field

is referred to as the toroidal field, while the vertical field is called
the poloidal.2
Let us assume that the galactic magnetic field and the dis-
tribution of thermal electrons give rise to a (constant) rotation
measure RM0 if we were looking exactly along the field. Now,
measuring from outside across a galaxy inclined by an angle i
(where i = 0 corresponds to the face-on view), the measured
rotation measure as a function of azimuthal angle θ is expected Figure 5.2: Possible modes
to vary as of halo fields.
RM (θ) = RM0 tan i · cos(θ − p). (5.2)
in case of an axisymmetric field. Here, p is the pitch angle of

the magnetic field, i.e. the angle that the field direction makes
with the tangent to the circle around the galactic centre at this
point (see Fig. 5.3). The above formula implies a single peri-
odicity of the observed rotation measure over one full circle in
the galaxy. This is different in case of a bisymmetric structure,
which yields a double periodicity:
1
RM = · RM0 tan i · [cos(2 θ − p − µ) + cos(p − µ)]. (5.3)
2
Figure 5.3: Pitch angle of
Here, µ is the position angle of the bisymmetric spiral field. the spiral structure.
Meanwhile, a small number of bona-fide cases are known with
magnetic-field structures complying with the above scheme.
M 31 : axisymmetric, m = 0
M 33 : m = 0 (1 kpc ≤ R ≤ 3 kpc), but RM phase inconsistent
M 51 : disk has m = 0, halo has m = 1
M 81 : bisymmetric
M 83 : RM double periodic, but phase in consistent with bisymmetric mode
NGC 6946 : m = 0 and m = +2
IC 342 : axisymmetric, m = 0
2
This somewhat sloppy nomenclature has been ‘copied’ from its more suitable relevance in the description
of stellar dynamos.
In contrast to the disk fields, the structures of magnetic fields in galaxy haloes are still
poorly explored. In galaxies viewed face-on their study requires Faraday tomography, such
as performed in case of M 51 to some extent (see below). When viewed edge-on, disk galaxies
exhibit the linear polarisation from entire lines of sight above and below the disk. Mildly
inclined galaxies such as M 31 provide the advantages of face- and edge-on views in one
observation.
Most edge-on galaxies exhibit a
nonthermal disk consisting of a thin
(∼ 300 pc) and thick (∼ 1.5 kpc)
component. An exception to this
is NGC 4631, which possesses the
brightest and most extended (in the
z-direction) radio halo, with a scale
height of z0 ≈ 2.5 kpc, where z0 is
defined by the intensity I of the syn-
chrotron radiation, i.e.
− zz
I(z) = I(0) · e 0 . (5.4)
Edge-on galaxies show the mag-

netic field to be plane-parallel to
the disk within the disk, while
it is directed away from the disk
at larger distances from the galac-
tic plane. Typical examples are
NGC 4631 (Fig. 2.28) and NGC 5775
(Fig. 5.5), with radio continuum
spurs associated with star-forming re-
gions in the underlying disk. The
magnetic field is probably dragged
outward by a galactic wind. Galax-
ies with massive disks and nor-
mal star formation rates exhibit X-
shaped magnetic-field morphologies Figure 5.4: Magnetic field in M 51. Contours depict
in their haloes (NGC 253, NGC 891, the radio continuum emission at λ = 6 cm. The yellow
NGC 3667, NGC 4565, NGC 5775). bars indicate the B-field. ~
Whether this is connected with field
topologies resulting from a galactic dynamo or just mimics the (known) flaring of the outer
gaseous disks is, at present, unknown.
The magnetic-field strengths in spiral galaxies have been inferred both, from the syn-
chrotron intensities using the minimum or equipartition argument, or from rotation-measure
analyses. The total field strengths are Bt <
∼ 10 µG, while for the regular (or uniform) field
the values are Breg < Bt (naturally). A rather detailed and thorough study exists for the
grand-design spiral M 51 (Fletcher et al. 2010, Fig. 5.4), which has resulted in the following,
rather detailed knowledge at this stage:
M 51 seems to be somewhat exceptional: higher magnetic-field strengths are indicated
than in other spiral galaxies. The possible reason for this could be its high star-formation
5.2. SPIRAL GALAXIES 95
total field Bt ordered field Breg

centre: 30 µG inner spiral arms: 11 − 13 µG
main spiral arms: 11 − 13 µG outer spiral arms: 8 − 10 µG
interarm regions: 15 − 20 µG interarm regions: 10 − 12 µG
rate, which causes enhanced turbulence, hence stronger dynamo action (see Sect. 5.6). The
cause for the enhanced star formation, which is also manifest in the high radio brightness
of this galaxy, could be its pronounced density waves, which give rise to shock compression
of the gas. The halo field (m = 1) can be separated from the disk field (m = 0, m = 2)
by observing at a high and a low frequency. At λ = 20 cm the Faraday rotation mainly
takes place in the halo of the galaxy (here: M 51), while at λ = 3.6 cm one observes Faraday
rotation from both, the disk and the halo! While at the longer wavelengths the disk emission
is almost entirely depolarised, much of the polarised emission at the short wavelengths is due
to anisotropic small-scale magnetic fields.
Figure 5.5: Magnetic field in the edge-on galaxy NGC 5775 (Soida et al., 2011) at λ = 20 cm
(left) and λ = 6.2 cm (right). Contours outline the total intensity, the bars denote the
orientation of the magnetic field.
An exceptional case is the starburst galaxy M 82. Its synchrotron brightness implies
Bt ≈ 50 µG in its central 700-pc area! This is the result of its intense star formation and
supernova rate. Not unexpectedly, its magnetic field exhibits a vertical or radial orientation.
This brings us to a particular class of disk galaxies, which host forming or active supermassive
black holes in their centres (Fig. 5.6). Superimposed onto their ‘normal’ radio disks, they
exhibit rather well-focused winds from their centres in the first case, or even a central source,
twin jets, and radio lobes in the latter - just like radio galaxies (Sect. 6.4). The most prominent
examples are complied below (see Sect. 6.1 for the AGN taxonomy).
As we shall see in Sect. 6.4, there are two classes of radio galaxies, namely so-called FR I
and FR II objects (Sects. 6.4.3 and 6.4.1), which are usually hosted by elliptical galaxies, i.e.
galaxies lacking any cold neutral gas, but having only hot, X-ray-emitting gas. It is most
surprising to find these phenomena also in some spiral galaxies. The magnetic-field strengths
in the anomalous radio features of spiral galaxies are ≥ 10 times higher than those responsible
for the normal disk emission. Seyfert-type galaxies are likely to possess a central black hole
with an accretion disk, from which twin jets are launched, probably at a large angle w.r.t to
the disk. An exception to this is probably NGC 4258.
Galaxy AGN type Radio features

0313-192 Sy 2 FRI
0421+040 Sy 2 FRII (?)
NGC 2992 Sy 2 ‘Figure 8’
NGC 3079 LINER ‘Figure 8’
NGC 3367 LINER FRII
NGC 3516 Sy 1 one-sided, curved, blobs
NGC 4258 Sy 2 ‘anomalous radio arms’; ejection in disk
So-called ‘Figure-8’ structures are the result of focussed winds from their active central
regions. In these galaxies, a wind with a speed of v ≈ 5000 km s−1 is focussed by the disk while
ploughing through the ISM ‘above’ and ‘below’ the central active region. The magnetic-field
strength in these somewhat collimated structures is between several tens and a few hundred
µG. There are many more spirals with Seyfert or LINER activity (see Sect. 6.6). In most
Seyferts, the central AGN is not powerful enough to produce any FRI/II morphology, as the
jets are quenched by the dense central ISM, so that such structures are inhibited, and the
synchrotron luminosities not correspondingly enhanced.
Figure 5.6: Potpourri of spiral galaxies with central AGN and radio-galaxy features. Shown
are: 0421+040 (top left), NGC 4258 (top right), 0313-192 (middle left), NGC 1068 (middle
right), NGC 3367 (bottom left), 0400-181 (bottom right).
5.3. DWARF IRREGULAR GALAXIES 97
5.3 Dwarf irregular galaxies

Low-mass (or dwarf) galaxies are mostly
weaker radio emitters, even though some of
them, being in a phase of intense star for-
mation, exhibit high radio surface bright-
nesses. With their shallow gravitational
potentials, their escape speeds are much
lower (∼ 1/4 . . . 1/3) than in massive spi-
rals, hence they are prone to mass loss
while experiencing strong bursts of star for-
mation. Such galactic winds not only en-
rich the intergalactic space (ICM, IGM)
with heavy elements (‘metals’), but at the
same time they transport relativistic parti-
cles and magnetic fields out of these galax-
ies into the surroundings. The majority of
dwarf galaxies is seen in quiescent phases
though, hence they are weak radio emit-
ters. Our knowledge of the magnetic-field
properties in low-mass galaxies is accord-
ingly poor, resting upon but a handful of
them. The compilation given further below
contains in brief our current state of knowl-
edge.
The magnetic-field structures in dwarf
galaxies are much less ordered than in mas-
sive spirals, which is probably due to their
slow rotation. In spite of their mostly rigid
rotation, however, they are able to produce
large-scale, coherent magnetic-field pat-
terns. Non-standard dynamos (Sect. 5.6)
may be at work in them. Some information
is available for our nearest galactic neigh- Figure 5.7: Magnetic fields in the dwarf galaxies
bours. For the Large Magellanic Cloud NGC 4449 (top) and NGC 1569 (bottom).
(LMC), a fit to the observed rotation measures has disclosed a spiral magnetic field with
m = 0. The favoured mechanism to produce this field in The LMC is the so-called CR-driven
dynamo (Sect. 5.6). In the Small Magellanic Cloud (SMC), the line-of-sight component of
the magnetic field appears to point uniformly away from us, while the projected field points
roughly in an east-western direction. There have been speculations whether this hints at a
so-called ‘pan-magellanic’ field connecting the LMC and the SMC. Also in case of the SMC,
a CR-driven dynamo seems to be favoured by the observations. For the prototypical star-
burst dwarf galaxies NGC 4449 and NGC 1569 (Fig. 5.7), the discussion is whether a dynamo,
perhaps supernova-driven, or a galactic wind is shaping the magnetic fields.
Dwarf galaxies may have been, or could still be, important in ‘magnetizing’ the ICM/IGM,
i.e. transporting relativistic particles and magnetic fields into their surroundings. Spectro-
scopic observations of emission lines (Hα, [NII]) indicate wind speeds reaching, or exceeding,
Galaxy Bt Bu Remarks
LMC 6 µG (equip.) 2 µG (filaments) diffuse pol. emission RM
grid (backgroud sources)
of 192 accurate RMs
SMC 4 µG (equip.) 1.7 µG RM grid; optical poln.;
70 accurate RMs
IC 10 14 µG 2 − 3 µG Bu associated with
giant bubble
NGC 1569 39 µG (centre) 3 − 9 µG radial field,
10 − 15 µG (halo) (strongest in halo) synchrotron halo
NGC 4449 14 µG 8 µG radial, with spiral structure;
counter-rotation; synchr. halo
NGC 6822 ≤ 5 µG 2 − 3 µG Bu just indicated (polarised
emission)
the escape speed

r
2GM
vesc = . (5.5)
R
By the same token, the measured temperatures of the hot (X-ray-emitting) gas reaches,
or exceeds, the virial temperature
vesc
Tvir = 1.5 · 105 · K (5.6)
100 km s−1
It is therefore quite likely that magnetic fields and relativistic particles have been ejected
into the ICM/IGM by low-mass galaxies, in particular soon after their formation in the young
universe. They were numerous in a ΛCDM cosmology, they had low escape speeds, hence a low
containment of CR (Sect. 5.5). This process must have been competing with magnetisation
by AGN (Chapt. 6).
5.4 Elliptical galaxies

Elliptical galaxies, and likewise dwarf ellipticals and spheroidals, are fundamentally different
from disk galaxies, also as far as their gaseous content is concerned. They lack the cold gas,
but rather have an ISM of hot ≥ 106 K gas. Dwarf ellipticals and spherodials may not have
any gas at all, as their gravitational potentials may not be deep enough in order to retain it.
The obvious reason for this is that their evolution is ‘completed’, in the sense that their stars
are evolved and their cold gas has been consumed in star formation, or expelled. The hot gas
needs a long time to cool,
n −1 T 21
6 p
τcool = 1.2 · 10 · · yr (5.7)
cm−3 106 K
so that star fomation is inhibited. Without star formation, there will not be any related
radio emission, due to the lack of production of relativistic particles. By the same token, there
are no sources for magnetic fields to be generated and/or sustained. The radio continuum
5.5. CR CONTAINMENT 99
emission from elliptical galaxies is hence confined to the central regions of those ellipticals that
are sufficiently massive, so that sufficient amounts of cooled gas may feed a central machine,
viz. an AGN (Chapt. 6). Diffuse, extended emission is therefore not seen in ellipticals.
Hence, elliptical galaxies only possess central synchrotron-emitting sources, if any. These
may be AGN producting jets and lobes (such as in radio galaxies), or are just confined compact
sources, probably because the surrounding ISM is too dense, and the AGN too weak, for the
jets to make it out of the galaxy. Some properties may be worth mentioning here:
• A correlation exists between the isophotal shape of ellipticals and their radio luminosity.
Ellipticals with boxy or irregular optical morphologies are radio-loud and tend to possess
X-ray haloes, whereas galaxies with pointed isophotes3 are radio-quiet and do not show
any X-rays.
• A radio-X-ray correlation exists, but has a large scatter. This hints at an accretion flow
to fuel central radio sources.
• Ellipticals possess, if any, mostly unresolved central radio sources. Some show extended
(linear) features, similar to the central sources in star-forming spiral galaxies.
• Most ellipticals depart from the radio-FIR correlation established for normal disk galax-
ies. The excess radio emission is obviously produced by the AGN, hence a different
mechanism for the particle production and acceleration has to be invisaged here.
• The extended radio lobes of elliptical galaxies hosting an AGN come along with low
X-ray to optical ratios. This hints at a more tenuous ISM, which is less impeding to
the formation of radio jets.
5.5 CR containment
Relativistic particles (e− , p, ...) are charged, hence tightly coupled to the magnetic field,
The latter in turn is coupled to the gas via the ionised (thermal) component. Hence, all of
these constituents form a disk (with different scale heights) which is subject to hydrostatic
equilibrium unless it is overpressured by strongly enhanced star formation and subsequent
supernova activity. This process may lead to a galactic wind in which the relativistic particles
and magnetic fields are transported out of a galaxy into intergalactic space. Without such
convective transport, the propagation of relativistic particles is subject to diffusion, in which
they experience multiple scattering from Alfvén waves, thereby moving at the Alfvén velocity
B
vA = √ , (5.8)
4 π ρion
where ρion is the density of the ionised gas. Supposing that the galaxy disk is pervaded
by a uniform, plane-parallel magnetic field, hydrostatic equilibrium in the z-direction (per-
pendicular to the disk) is given by
dP dΦ
= −ρ · . (5.9)
dz dz
3
The term ‘pointed’ means weak edge-on disks.
The density ρ is related to the gravitational

potential by the Poisson equation
d2 Φ
= 4 π G ρ. (5.10)
dz 2
For |z| ≤ 250 pc, the gravitational accelera-
tion az has a roughly constant slope (Fig. 5.8) so
that we can write
dΦ daz
= −z · . (5.11)
dz dz
Figure 5.8: Sketch of the gravitational ac-
Hence, from (5.9) we see that celeration as a function of height above the
galactic disk.
dP daz
= ρz · (5.12)
dz dz
The pressure is the sum of pressure contributions from the gas, cosmic rays and the
magnetic-field. The gas pressure is dominated by macrosopic turbulence rather than the
thermal one, so that
1
· ρ v 2 = 1.0 · 10−12 dyn cm−2

Pg = (5.13)
3
The CR pressure is one third of the energy density of the relativistic particles, uCR ≈
1.3 · 10−12 erg cm−3 , i.e. PCR = 0.4 · 10−12 dyn cm−2 . The magnetic pressure is
B2
Pmag = = 1.0 · 10−12 dyn cm−2 (5.14)
8π
for a magnetic-field strength B = 5 µG. Now, assuming that the three pressures vary with
height z above (and below) the galactic plane in the same way, i.e.
Pmag = α · Pg
PCR = β · Pg
the equation for hydrostatic equilibrium reads
1 d daz
ρ v2 = z ·

(1 + α + β) · · . (5.15)
3 ρ dz dz
Assuming that v 2 is constant, the solution is

z 2
ρ(z) = ρ(0) · e−( h ) , (5.16)
where
"
#1
2 · (1 + α + β) v 2 2 1
h= daz
= 100 · (1 + α + β) 2 pc. (5.17)
−3 · dz
is the scale height of the disk. One usually quotes the effective thickness, defined as the
total mass per unit area, divided by ρ(0):
R∞
ρ(z) z dz
0 1 1
2H = = π 2 h = 180 · (1 + α + β) 2 pc. (5.18)
R∞
ρ(z) dz
0
Inserting α = 1 and β = 0.4, we find 2 H = 280 pc. Without any enhanced star formation,
hence supernova activity, the disks of galaxies would attain a more or less constant thickness of
the above order. The CR particles propagate through the disk via diffusion, owing to multiple
scattering by the Alfvén waves (s.b.). The discussion of CR propagation is complicated, and
a thorough treatment would be beyond the scope of this lecture. We therefore indicate the
issues of interest with a few simplified arguments. The morphology of magnetic fields in
interstellar or intergalactic space is never such that they are homogeneous. Even if they are,
the streaming motion of the relativistic particles imposes fluctuations onto the fields. If the
size scale of the fluctuations is much smaller than the Larmor radii of the particles, then these
will just move on their helical orbits in the mean field. However, if the Larmor radii are smaller
than the fluctuation scales, then the relativistic particles will experience frequent pitch-angle
changes, i.e. they will be permanently scattered by the irregularities of the magnetic field.
This readily explains their isotropic arrival directions. Without derivation, the mean free
path of this scattering process is
2
B0
λsc ≈ rg · , (5.19)
B1
where rg is the Larmor radius of the particles, B0 is the mean strength of the magnetic
field, and B1 is the strength of the perturbations inherent to this field. The diffusion coefficient
then is
D ≈ v · λsc . (5.20)
The irregularities can be considered as transverse waves propagating along the field lines.
The bulk streaming motion of the charged particles is limited by the Alfvén speed
ne − 12

B B
vA = √ = 2.2 · · km s−1 , (5.21)
4 π ρion µG cm−3
owing to the multiple scattering off the field irregularities. This scattering also implies
an efficient ‘storage’ of the CR particles in a galactic disk that they are confined in. Their
containment is desribed by the diffusion loss equation:
dN (E) d ~ 2 N (E),
= [b(E) N (E)] + Q(E, t) + D · ∇ (5.22)
dt dE
where
dE
b(E) = − . (5.23)
dt
describes the energy losses (synchrotron, inverse-Compton, adiabatic,...). Q(E, t) is a

source term, since the particles may be continuously injected (per unit volume), and the
last term on the right-hand-side stands for the diffusion losses, i.e. the term quantifying the
number of particles entering and leaving a volume dV by diffusion4 .
What happens in a real disk? In a
real disk, there is star formation, produc-
ing energetic (long-lasting) stellar winds,
due to (massive) star formation, followed
by supernova explosions, which are even
more energetic. They inject both, ener-
getic particles as well as mechanical en-
ergy, i.e. they produce turbulence. As a
result, the pressure increases so that the
disk thickness in the vertical (z-) direction
is strongly increased (Fig. 5.9). The forma-
tion of dense clouds, requiring cloud con-
traction and collapse, implies that under
magnetic-flux conservation the magnetic-
field strength is enhanced within the denser
regions, while in between the field it is di-
luted. This is the famous Parker insta-
bility (Parker, 1966), producing so-called
‘Parker Loops’ (Fig. 5.9). Eventually, when
there is star-formation activity between the Figure 5.9: Sketch of a stable and an unstable
clouds (with some time delay), the particles galactic disk.
and fields may be pushed out of the disk,
whereby magnetic-field lines may reconnect to form buyoant loops, transporting magnetic
fields and particles away from the disk into the galaxy halo and, if sufficient energy is injected,
out of the gravitational potential into the IGM. The condition for this is that the speed of
the convective wind exceeds the escape speed (Eqn. 5.5) or, equivalently, the temperature of
the heated gas exceeds the virial temperature (Eqn. 5.6).
There are quite a few starburst galaxies fulfilling these criteria. In particular, galaxies
with starburst centres such as M 82 (Fig. 5.10) and NGC 253 exhibit clear signs of convective
winds transporting thermal and relativistic plasma out into intergalactic space. Low-mass, or
dwarf, irregular galaxies possessing large amounts of gas and experiencing episodes of intense
star formation are clear candidates for producing galactic winds. In fact, spectroscopic and
X-ray observations corroborate the above inferences.
We distinguish ‘blow-out’ and ‘blow-away’ scenarios: depending on the energy input and
galaxy mass, the (thermal and relativistic) gas is either just blown out of the disk and falls
back onto it, which is referred to as a ‘galactic fountain’, or the material and the magnetic
fields escape irreversably and ‘pollute’ the ICM/IGM (Mac Low & Ferrara, 1999). This has
cosmological significance. In a ΛCDM cosmology, small dark-matter haloes are formed first,
with baryonic material subsequently falling into them. Hence, because of the preponderate
number of dwarf galaxies, along with their shallow gravitational potentials and their intense
star-formation rates (Pop III stars), low-mass galaxies may have (had) a significant role in
4
M.S. Longair, ”High-Energy Astrophysics”, 3rd edition, 2011, Cambridge University Press.
magnetizing their surroundings by injecting relativistic particles and magnetic fields into the
ICM/IGM. In fact, (mildy) relativistic particles and magnetic fields are ubiquitously detected
in the intra-cluster environment (Sect. 7). In this process, dwarf galaxies must have been
competing with AGN. A discussion of this competition must account for the relative powers
and duty cycles of activity of dwarf galaxies and AGN (see Sect. 7.7).
Figure 5.10: The M 81/82 galaxy group mapped at λ = 90 cm with the WSRT. The extended
synchrotron halo of the starburst galaxy M 82 in the north-west is obvious.
Finally, it is interesting to make a quick estimate of the total mass of relativistic particles
in a galaxy. This is readily done by assuming equipartition between particles and fields so that
the energy densities of the particles (electrons and protons) and magnetic fields are equal:
B2
= nrel γ η me c2 , (5.24)
8π
where η accounts for the energy contained in the protons (see Chapt. 3). This yields the
number density of relativistic particles nrel . Multiplying by the likely volume of the relativistic
plasma and by the mass of the relativistic protons (which do count in this balance), we obtain
B 2 mp
Mrel = nrel V mp = · π R2 H, (5.25)
8 π γ η me c2
where we have simply assumed a volume V made up by a disk of radius R and thickness
H. Assuming B = 10 µG, γ = 3000, R = 15 kpc, and H = 500 pc, we obtain the (perhaps
stunningly) tiny amount of Mrel = 0.4 M ! Note that we have chosen a relatively large
magnetic-field strength and a small Lorentz factor. This, of course, is the rest mass of the
particles, but even accounting for their relativistic energies, meaning that we would have
to multiply the above mass by γ 2 , would not make this constituent of the ISM of a galaxy
dynamically significant.
5.6 Galactic dynamo

We now briefly discuss the likely mechanism that produces and sustains the large-scale mag-
netic fields in the ISM of galaxies. A very likely process that was first worked out by Parker
(1966) is the so-called α − ω dynamo, originally applied to the Sun and Earth (Parker, 1955).
A dynamo is a machine in which the motion of a conductor amplifies a given seed magnetic
field against the diffusive losses. Looking at the induction equation
~
∂B ~ × (~v × B)
=∇ ~ −∇
~ × η (∇
~ × B),
~ (5.26)
∂t
we note that there is no source term in it, hence an initial seed field is needed. One
possibility is the Biermann battery (1950). Having this, the dynamo can work against the
diffusive term on the right-hand-side of the above equation. In order to produce a magnetic
field, one needs currents, according to Ampere’s law. Of course, the ISM is neutral on large
scales, hence currents cannot be produced. However, the rather different masses of protons
(or ions) and electrons lead to a charge seperation, owing to the different drag forces on the
protons and the electrons.5
The mathematical details of the galactic dynamo are rather intricate so that we cannot
treat them here in any depth. We shall rather have a look at the basic ingredients of the
induction equation, try to infer observational parameters, and finally compare theoretical
predictions with observations.
Cowling (1976) has demonstrated that an axisymmetric magnetic field, such as the solar
dipole field, cannot be sustained. This finding is referred to as Cowling’s Theorem. Hence,
deviations from axisymmetric motions are indispensable for the dynamo action. These devia-
tions are caused by turbulence. Turbulence imposes fluctuations onto both, the velocity and
the magnetic field, in the sense of writing them as
~ = B~0 + δ B
B ~ (5.27)
and
~v = v~0 + δ~v , (5.28)

5
This is also the basis of Biermann’s battery.
5.6. GALACTIC DYNAMO 105
D E
where B~0 = B ~ is the mean field, v~0 = h~v i is the mean velocity, and δ B
~ and δ~v are
the perturbations. Then the ensemble average of the induction equation in case of infinite
conductivity reads
~
∂hBi
=∇ ~ × h~v i × hBi)
~ +∇ ~ × hδ~v × δ Bi,
~ (5.29)
∂t
with the residual
~
∂δ B h D Ei
=∇ ~ × δ~v × hBi ~ + h~v i × δ B
~ + δ~v × δ B
~ − δ~v × δ B
~ . (5.30)
∂t
In order to understand how Cowl-
ing’s theorem can be ‘beaten’, we
have to have a closer look at the gas
kinematics in a galaxy, which is well
known from stellar spectra, HI and
CO measurements, making efficient
use of the Doppler effect. The re-
sulting rotation curves (Fig. 5.11) ex-
hibit rigid rotation in the inner part
of galaxies, and differential rotation
(‘flat rotation curves’) out to any de-
tectable tracer (essentially HI). This
induces shear. On top of this, there is
star formation and hence supernova
activity, which causes turbulence. In
particular, it gives rise to vertical
expansion of (heated, hence ionised)
gas bubbles expanding into the galax-
ies’ haloes. While discussing this, do
not forget that the gas is partially
ionised, hence the magnetic field is
frozen into the gas. Now suppose that
the magnetic field is initially plane-
parallel to the disk. A region with Figure 5.11: Collection of galaxy rotation curves.
star formation and supernova activ-
ity produces a plasma bubble that expands out of the disk, the particles consequently ex-
periencing a Coriolis force in the course of the expansion. The result is that the magnetic
field is twisted into a loop, with its angular-momentum vector opposed to that of the galactic
rotation. As illustrated in Fig. 5.12, this gives rise to a current ~j directed anti-parallel to the
~ t . Having a multitude of such currents, the result is a poloidal magnetic
initial torodial field B
~
field Bp , which in turn gives rise to a current in the galaxies’ planes. This process can grow
in such a way that the decay of the magnetic field by diffusion is overcompensated, giving
rise to an overall galactic magnetic field.
The dynamo theory leads to the so-called dynamo equation
~
∂B ~ × [(∇
=∇ ~ × B)
~ + αB
~ − ηt ∇
~ × B],
~ (5.31)
∂t
where B~ now stands for hBi,

~ and ~v is the mean
velocity. The parameter ηt is the turbulent diffusivity,
with
c2
ηt = + β. (5.32)
4πσ
The quantities α and β are given by
τ ~ × ~vt )i,
α = − · h~vt · (∇ (5.33)
3
τ
2
β= v , (5.34) Figure 5.12: Illustration of the α-ω
3 t dynamo.
where ~vt is the turbulent velocity. The first term
describes the α-effect, which arises inevitably, owing to the fact that turbulence appears
automatically in a rotating system. The ascending clouds (pushed upward by star formation)
are subject to the Coriolis force, which is proportional to Ω · vr , where Ω is the rotational
frequency of the galaxy and vr is the lateral speed of the expanding clouds. The quantity β
reduces the electric conductivity, i.e. it increases the diffusivity so that the mean magnetic
field diffuses faster in the presence of turbulent motions since ηt η. The fluctuating plasma
motions shear, twist, stretch, fold and unfold the field lines. In Fig. 5.13, a sketch of this
dynamo action involving these processes is shown. Locally, magnetic reconnection transfers
the magnetic overload to heat and particle acceleration.
The radial velocity component vr
is created by the expansion of the
ascending bubbles. The correlation
time τ of the turbulent velocity field
is essentially the lifetime of the tur-
bulent cells producing the helicity.
Roughly speaking,
lmf p
τ= , (5.35)
vt
Figure 5.13: Illustration of the effects of dynamo ac-
where lmf p is the typical mean tion producing a bisymmetric magnetic field.
free path of the turbulent cells and vt
is the typical turbulent velocity. The
turbulence and bubbles can be driven
by supernovae and by CRs (‘SN-driven’ and ‘CR-driven’ dynamos). The helicity parameter
α becomes
2
lmf p·Ω
|α| = , (5.36)
h
where h is the scale height of the disk. So α represents the effect of cyclonic motions on
the mean field, in such a way that they amplify it. One can think of the effect as the result
of random steps taken by the field lines:
5.6. GALACTIC DYNAMO 107
• In one step a given field line of the mean field is rotated by τ · (∇

~ × ~v ) about the step
~ × ~v )
direction ~v · τ so that the mean field is turned and displaced by an amount τ ~v · (∇
per unit time. The β term represents the mixing of the field, and can be thought of as
resulting from a random walk of field lines:
• There is a random walk of various field lines, with a mean step size of v · τ , and a rate
of steps of τ −1 . During the steps, the field strength of the lines does not change, and
~ at any point is the mean of all field strengths of all the flux
the resulting mean field hBi
lines brought to that point by the ensemble.
Figure 5.14: Sketch of the α-ω dynamo working in a galaxy disk. The field lines of the
poloidal component are depicted in the left half of the disk, while for the toroidal field the
field strength is indicated by the contours and grey-scale in the right half.
The amplification of the magnetic field has to happen fast enough in order to explain the
existence of equipartion fields in distant galaxies (i.e. in a relatively young universe), starting
out from weak seed fields. Hence, the galactic dynamo has to be able to amplify very weak
(B ∼ 10−15 . . . 10−10 G) seed fields to equipartion fields (B ∼ 10−6 . . . 10−5 G) within a few
Gyr. While this is difficult to achieve even in massive spiral galaxies with sufficient differential
rotation (an important ingredient to make the α − ω dynamo work), this is deemed even more
difficult in low-mass galaxies, which largely lack differential rotation. In Fig. 5.14 a schematic
representation of the magnetic field produced by an α-ω dynamo is shown.
A solution to this problem is the so-called CR-driven dynamo, which was first proposed by
Parker (1992). Its main difference to previous dynamo models is a strongly enhanced growth
rate for the amplification, due to the action of SN-driven CRs that quickly stream out of
the star-forming galaxy disk, thus producing large magnetic loops or ‘lobes’, which strongly
enhance the α-effect. Meanwhile, successful numerical MHD models have been published by
Hanasz et al. (2009, see Fig. 5.15). A three-dimensional view of a simulated galactic magnetic
field is presented in Fig. 5.16. Note the field lines pervading the halo, i.e. stretching out of
the disk and coming back into it.
Figure 5.15: Numerical MHD simulations of the CR-driven dynamo. The colour coding
represents the magnetic-field strength in units of µG, with an edge-on (top) and a face-on
(bottom) view.
Figure 5.16: ’Spaghetti’ diagramme of the three-dimensional configuration of the magnetic

field in a spiral galaxy from numerical simulations of a dynamo.
Chapter 6
Active galactic nuclei
6.1 The AGN Zoo

So far, we have discussed the radio con-
tinuum (and specifically: its synchrotron)
emission of normal, i.e. star-forming galax-
ies, along with their magnetic field. The
production of the relativistic plasma, i.e.
particles and magnetic fields, in them is
closely linked to star formation activity in
their disk, with massive stars ionising the
gas surrounding them and supernova ex-
plosions injecting and accelerating parti-
cles, mainly protons and electrons. The
UV radiation of the stars is very efficiently
absorbed by the dust and reradiated in
the FIR regime, this naturally explaining
the well-established radio-FIR correlation
(Fig. 6.1).
Note that this correlation is one be-
tween the (nonthermal) synchrotron radio
emission and the (thermal) FIR, while a Figure 6.1: Radio-FIR correlation of star-
correlation between the thermal (free-free) forming galaxies and galaxies hosting AGN.
radio and the FIR emission would be trivial, since it would reflect nothing but the luminosity
of Lyman continuum photons in each galaxy: these photons ionise the gas, giving rise to the
thermal radio continuum, until they are ultimately absorbed, their energy being reradiated
in the FIR. Both luminosities are therefore directly proportional to the underlying UV lumi-
nosity, hence the star formation rate. The fact that the radio-FIR correlation also holds for
the non-thermal radio continuum is still due to the star-formation rate, more specifically, to
that integrated over the past ∼ 108 yr. The reason is that the relativistic particles produced
by the supernove are stored within the galaxies for some time and are kept relativistic, owing
to the ongoing supernova activity and Fermi acceleration processes. This is reflected by the
tight correlation for normal star-forming galaxies evident in the lower distribution of data
points in Fig. 6.1.
109
110 CHAPTER 6. ACTIVE GALACTIC NUCLEI
This tight correlation does not hold, however, for galaxies hosting active galactic nuclei
(AGN). They still obey such a law, however, one with a lot more scatter and with the radio
luminosity being about two orders of magnitude higher (at the same FIR luminosity), as is
obvious in Fig. 6.1. The production of relativistic particles emitting their radio (synchrotron)
emission is fundamentally different from that of star-forming galaxies. AGN show up in a
variety of forms and have been identified as a new class of galactic systems in the second half
of the last century.
The first recognition that there are extreme phenomena in the centres of certain galaxies
was made for the so-called Seyfert galaxies (Seyfert, 1943). With their two categories, viz.
Sy 1 and Sy 2, they are templates for the standard AGN paradigm. Sy-1 galaxies exhibit
broad emission lines, with F W HM ∼ 5000 . . . 10 000 km s−1 , and a (partially) nonthermal,
nonstellar continuum, recognised as synchrotron radiation. The central region emitting these
braod spectral lines is therefore called ‘broad-line region’ (BLR). In contrast, Sy-2 galaxies
show ’narrow’ lines, F W HM ∼ 200 . . . 400 km s−1 , which is still broader than the correspond-
ing linewidths of typical normal galaxies. The central regions of this type of Seyfert galaxies
is called ‘narrow-line region’ (NLR). From the observed variability of the emission from Sy-1
galaxies one infers a linear size of the BLR < 1016 cm (1/100th of a lightyear), while for NLR
this is ∼ 102 . . . 103 times larger. The difference between the BLR in Sy 1 and the NLR Sy 2
is the viewing angle (see the so-called ’unified scheme’ explained further below).
Figure 6.2: Radio image of Cyg A.
After World War II, radio astronomy saw a fast development. Bolton & Stanley (1948)
detected the radio source Cyg A (3C 405) with their ‘Sea Interferometer’. Smith (1951) ob-
tained a more accurate position at Cambridge, and Baade identified the radio source with a
distant (D ≈ 700 million ly) optical galaxy. The inference therefore was that Cyg A (Fig. 6.2)
is an extremely radio-luminous galaxy, having ∼ 104 times the radio luminosity of M 31. M.
Schmidt (1963) identified the radio source 3C 273 with a 13th magnitude star-like object, hav-
ing a redshift of z = 0.158 (obtained from the spectroscopy of the Balmer series). Matthews
& Sandage had discovered a 16th magnitude star-like object at the position of the radio source
3C 48. However, he did not yet get a spectrum, hence there was no known redshift for this
source at that time (1960). The broad emission lines could not be identified, and the spectrum
exhibited much more UV light than that of an ordinary main-sequence star.
6.2. TAXONOMY OF AGN 111
6.2 Taxonomy of AGN

Quasars
Quasars are very luminous and compact centres of galaxies, outshining their host galaxies.
They exhibit nearly featureless spectra, and are visible from radio wavelengths through X-ray
energies. Their luminosities are 1045 . . . 1049 erg s−1 . To date, ∼ 104 QSOs are known, of
which about 10% are strong radio emitters, so-called radio-loud QSOs.
Radio galaxies
This kind of AGN phenomenon was first recognised in the form of its prototype Cyg A. The
host galaxies are ordinary ellipticals in the optical regime. They are very strong radio emitters,
with radio luminosities of Lr ≥ 3 · 1041 erg s−1 . They mostly consist of a compact core, they
show a jet or twin jets, which feed double lobes with a relativistic plasma. Sometimes they
exhibit so-called ‘hotspots’ in their outer lobes. These radio galaxies show two morphologically
different categories, which can also be subdivided by their radio luminosities:
• FR I radio sources
FR I radio sources have monochromatic radio luminosities of L1.4GHz ≤ 1032 erg s−1 Hz−1 .
They possess less collimated jets, which exhibit a loss of collimation beyond the optical
peripheries of their host galaxies, with a strong flaring further out. They then become
fainter and have steeper synchrotron spectra towards the outer edges of the lobes.
• FR II radio sources
FR II radio sources have monochromatic radio luminosities of L1.4 GHz ≥ 1032 erg s−1 Hz−1 .
Their highly collimated jets are often one-sided), with pronounced hotspots at their
termination points in the lobes. These lobes extend out of the host galaxies for many
kiloparsecs, the biggest such object known to date being the radio galaxy 3C 236, with
a total extent of ∼ 4.5 Mpc (H0 = 71 km s−1 Mpc−1 )!
This difference is likely due to the jets in FR Is being subsonic, and supersonic in FR IIs.
The difference in jet speed may be also connected with the different properties of the ISM in
the host galaxies, i.e. its density in the first place.
Seyfert galaxies
Seyfert galaxies are spiral galaxies hosting an AGN in their centre. Their luminosity is in
the range ∼ 1043 . . . 1045 erg s−1 , with the brightest Seyfert galaxies being as luminous as
the faintest QSOs. As already pointed out in Sect. 6.1, the main difference between the two
types, Sy 1 and Sy 2, is that they show rather different line widths when looked at with optical
spectroscopy:
• In a Sy 1 galaxy, the observer measures a large velocity dispersion, as one is looking
directly at the immediate surroundings of the central supermassive black hole.
• In a Sy 2 galaxy, our view onto it has a different aspect angle, which causes the central
region to be obscured by a dust torus.
Nearly 15% of Seyferts have close companions, which is indicative of merging or tidal
forces at work.
Blazars
Blazars emit polarised light with a fea-
tureless, nonthermal spectrum that can
be traced over a large spectral range.
They also exhibit very strong variability
(Fig. 6.3), therefore one subdivides them
into two types, viz. BL Lacs, named after
its prototype BL Lacertae. These do not
exhibit any emission lines at all. Optically
violently variable sources (OVV) do show
emission and absorption lines when scru-
tinised. Blazars are most likely AGN in
which we look almost directly into the jet
coming towards us, which results in strong
relativistic boosting, the so-called ‘Doppler
boosting’. This process enhances the un-
boosted flux density S0 towards the ob-
Figure 6.3: An example of the variability of the
server to a value of
emission from a BL Lac-type object.
3+α
Sobs = S0 · D+ , (6.1)
where
1
D+ = , (6.2)
γ · (1 − β cos θ)
is the Doppler factor,
1
γ = (1 − β 2 ) 2
is the Lorentz factor and
v
β= .
c
The source emits synchrotron radiation with the usual power-law (in the rest frame):
Sν0 ∝ ν0−α . (6.3)
Correspondingly, the emission of the receeding jet is diminished to
3+α
Sobs = S0 · D− , (6.4)
where
1
D− = . (6.5)
γ · (1 + β cos θ)
This is the reason why radio galaxies generally show a weaker counter-jet, and why the
counter-jet is mostly invisible in quasars. The extreme Doppler boosting in blazars is the
reason why one does not see any line emission in their spectra. the line emission is simply
swamped out by the vastly dominating and outshining, Doppler-boosted synchrotron radia-
tion that we measure when looking into the jet. The variability is readily explained in terms
of wiggeling jets, implying that the direction θ of the boosting w.r.t. the observer varies,
hence a variable Doppler factor.
6.3. UNIFIED SCHEME 113
6.3 Unified scheme

In this section, we summarise the properties of
the AGN phenomenon, which is largely connected
with the production and action of magnetic fields.
A summary of the signs of nuclear activity is given
below. It has to be pointed out that not all of
these features are always present.
• There is a compact (≤ 3 pc), luminous cen-

tre.
• It shows a spectrum with strong emis-

sion lines, and these are strongly Doppler-
broadened.
• The central region emits strong nonthermal

emission.
• Strong UV radiation from their compact

central region is observed.
• Frequently, jets are formed that feed double Figure 6.4: Sketch of the central region of
radio sources. a galaxy illustrating the unified scheme.
• There may be variability of the emis-

sion from the central region over short
timescales, which encompass the whole
spectrum.
These properties can be explained in terms of the so-called ‘unified scheme’ (Fig. 6.4) of
AGN, with the following ingredients:
• In the centre of these galaxies there is a suppermassive black hole.
• The central black hole is surrounded by an accretion disk.
• Twin jets are ejected into two opposite directions.
• The BLR reflects the immediate surroundings of the black hole and the accretion disk.
• The NLR marks the gas motions further away from the compact nuclear source.
• There is a molecular and/or dust torus surrounding this central BLR which, depending
on the viewing angle blocks the view onto the central region, hence showing only the
NLR, while the view onto the BLR is unblocked in the other case.
The linear dimensions of the main AGN components are given in the tabular compilation
below:
Component Typical size

BH (RS for 109 M ) 10−4 pc
Accretion disk 0.01 pc
BLR 1 pc
dust torus 10 pc
NLR 1 kpc
host galaxy 10 kpc
The brightness temperature
λ 2 Sν
Tb = · (6.6)
2 k ΩS
is an important tool to diagnose the radiation process of a source. As we have seen

in Sect. 2.2.2, this quantity is defined via the Rayleigh-Jeans approximation of a radiation
process. Nonthermal radiation processes are generally characterised by high brightness tem-
peratures, which of course do not have any physical meaning, except that they tell something
about the intensity of the source. For AGN the intensities can be extreme. However, the
brightness temperature that a radio source can possess is limited by the fact that for very high
number densities and energies of relativistic electrons, inverse-Compton losses become dra-
matic. The maximum brightness temperature is constrained by the ratio of inverse-Compton
to synchrotron powers:
" #
Tmax 5 νc Tmax 5 νc

PIC 1 1
= · · · 1+ · · . (6.7)
Psyn 2 1012 K MHz 2 1012 K MHz
For temperatures T > 1012 K, the inverse-Compton losses become catastrophic, as in this
10 . This situation is referred to as the inverse-Compton catastrophe.
case PIC /Psyn ∼ Tmax
6.4 Radio galaxies

Radio galaxies are probably the
most well-defined type of AGN
in terms of their radio struc-
tures, in spite of their bimodal
distribution of morphologies
and luminosities (FR I, FR II).
They possess (almost) all of
the ingredients mentioned so
far, except that, or because of, Figure 6.5: Sketch of the main features of a classical FR II
their host galaxies are elliptical radio galaxy.
galaxies the ISM of which is a
tenuous and hot one. We will first discuss FR II sources, since they represent the radio-galaxy
phenomenon as it had first been noted in the form of its prototype, Cyg A (3C 405), which
shows the two jets ploughing thorough the ISM and IGM, finally piling up at the hotspots
and, from there, depositing the relativistic plasma in the outer lobes.
6.4. RADIO GALAXIES 115
6.4.1 FR II sources
In the radio regime, the main features of FR II
sources are as follows (see Fig. 6.2 and the sketch
in Fig. 6.5):
• A central compact (pc-scale) radio source.
• Twin jets ejected in opposed directions

(with different synchrotron brightness).
• The jet opening angle ξ = drj /d` < 0.1 rad,

with typical values around 0.05 (rj being
the jet radius and ` the length).
• Hotspots marking the terminations of the

jets.
• Double lobes, ‘floating’ back onto the host

galaxies from the hotspots (Fig. 6.6).
• They are edge-brightened.
• They show radio luminosities at least as Figure 6.6: Cyg A at 327 MHz (top) and
high as L1.4GHz ≥ 1032 erg s−1 Hz−1 . at 15 GHz (bottom).
As said above, the most prominent representative of this species is the radio galaxy first
discovered and identfied as such, Cyg A. Since it is one of the radio galaxies that has been
studied in enormous detail1 , it is useful to discuss the properties of radio galaxies using
Cyg A as a template. Since we are discussing magnetic fields and hence, by virtue of these,
the synchrotron radiation used to study radio galaxies, the appearance of Cyg A in the radio
domain is first briefly outlined, as it has been studied in the radio continuum with resolved
linear polarisation on all scales encompassing the jets and lobes:
• Looking at Fig. 6.6, we see the basic morphology at this somewhat lower angular res-
olution, with a core and the outer lobes. Note, however, the different appearance at
different frequencies: at the lower frequency, we see thick lobes, with aged particles
streaming back onto the host galaxy, whereas at the higher frequency, we only see the
high-energy particles, mainly from the jets and lobes.
• The magnetic field is mostly circumferential in the lobes, following their boundaries
(Fig. 6.7). This is a general phenomenon seen in radio galaxies.
• There are indications of the magnetic field to be oriented parallel to the jet (like in most
FR IIs). Within the lobes, the magnetic field is more tangled, with field-ordering on
scales of ∼ 20 . . . 30 kpc.
• One frequently finds large rotation measures, ∼ 4000 rad m−2 < RM < +3000 rad m−2 .
These must be due to a foreground Faraday screen, i.e. the hot ICM gas and the
intergalactic magnetic field.
Figure 6.7: The magnetic field in Cyg A as deduced from measurements of the linear polari-
sation.
The inferred intergalactic magnetic field has a strength of BICM ∼ 2 . . . 10 µG (see

Chapt. 7 for more details on this topic). Below we list the strength of the magnetic field
within Cyg A from the observed synchrotron intensity, assuming equipartition:
Bt [µG] Location
100 core
80 jets
200 hot-spots
70 lobe’s head
40 lobe’s tail
The fact that in FR-II radio galaxies the magnetic

field is parallel to the jet over most of its length can
be explained in terms of the large bulk velocities of
the jet material. This is perhaps best illustrated by
the ’flux-rope’ model (Fig. 6.8), which was ‘invented’
to explain the magnetic flux rope on the planet Venus.
Flux conservation requires that
Bk ∝ r−2 , B⊥ ∝ r−1 .
Figure 6.8: Illustration explain-
This is a strongly simplified picture, as it cannot ac- ing flux conservation in a flux
count for the intricate physics, such as the entrainment rope made up by longitudinal and
of surrounding material from the ISM and/or ICM, or poloidal magnetic fields.
(oblique) shocks. In fact, for instance, the very promi-
nent jet in the giant radio galaxy NGC 6251 exhibits a preferentially longitudinal magnetic-
field orientation over the first (projected!) ∼40 kpc, while further out it is dominated by
the perpendicular, or poloidal, component. A helical magnetic-field structure has meanwhile
been inferred for several jets (3C 120, 3C 273, ...) from rotation measure analyses. A different
morphology is seen in some pc-scale jets (e.g. the blazar 1055+018), which show two com-
ponents of polarisation. An inner ‘spine’ with a transverse field is bracketed by a boundary
1
see ”Cygnus A - Study of a Radio Galaxy”, eds. C.L. Carilli & D.E. Harris, 1996
layer having a longitudinal field. The interpretation here is that the inner spine is dominated
by oblique shocks, while the jet’s mantle interacts with the surrounding medium. This inter-
action can be twofold: (i) shear of the jet flow stretching the magnetic field along the jet, and
(ii) the jet fluid can be compressed against its boundaries. In Fig. 6.9 the prominent jet in
NGC 6251 is shown.
Figure 6.9: Sketch of the magnetic field in a flux rope, with a real magnetised jet (NGC 6251)
juxtaposed.
It is hitherto not clear what the jets consist of. We could be dealing with
(i) a mildly relativistic p-e− jet, with a jet speed of vj ≤ 0.5 · c. It would carry most of the
mass and kinetic power ejected by the AGN, and would be responsible for the formation
of the kpc-scale jets, hot-spots, and lobes;
(ii) a relativistic e± beam (”pair plasma”), with a jet speed of vb ≈ c and 3 ≤ γb ≤ 10. This
jet material would move in a channel through the jet and would be responsible for the
superluminal motion of, and γ-ray emission seen from, the jets.
In order to test this model, sensitive measurements of the circular polarisation of jets seen
at very small viewing angles have to be made2 . Circular polarisation is to be expected in a
very homogeneous magnetic field from a p-e− plasma. The pair plasma should be inferred
from detections of the 511-keV emission line from jets. This is difficult, however, since the
line will be strongly ‘washed out’ by relativistic effects.
Quite naturally, the magnetic fields lead to synchrotron losses of the radiating particles,
which is reflected in a spectral steepening of the synchrotron spectra. In FR-II sources, this
steepening starts away from the hotspots, and becomes progressively stronger further away
2
These have been started in the recent past!
from them, towards the host galaxy, as the relativistic plasma is slowly ‘falling back’ from
the hotspots onto the peripheral ISM of that galaxy. These radiation losses in particular give
rise to a break in the synchrotron spectra, given by
2
B E
νc = 16.1 · · MHz. (6.8)
µG GeV
The analysis of the synchrotron spectra in radio
galaxies is a lot easier than in normal spirals: in the
latter, there is contamination by thermal free-free
radiation, which is totally absent in the former. It
should be noted here that there are two other loss
mechanisms that may play a role here: (i) Inverse-
Compton losses take over as soon as the magnetic-
field strength drops below BCM B = 3.25 · (1 + z) µG.
(ii) Adiabatic-expansion losses are dominant in the
overpressured hot-spots.
In quite a few cases, the central activity of QSOs
or radio galaxies has ceased, the jets have been
’switched off’, and the lobes are expanding, with
their synchrotron spectra steepening (i.e. the break
frequency is migrating towards lower and lower val-
ues). The radio morphology of such sources is re-
ferred to as ‘fat doubles’, or radio relics in extreme Figure 6.10: Particle ageing as derived
cases. A nice example of such an aged, ‘fat double’, for Cyg A, with the break frequencies
was shown in Fig. 2.21. One of the first ‘died’ radio (top) and the ages (bottom) shown.
galaxies found is 0924+30, which has an ’age’ of ∼ 50 Myr (Fig. 6.11).
Figure 6.11: The died radio galaxy 0924+30 at 151, 327, 610, 1400, and 4750 MHz. The total
continuum spectrum in the lower right clearly shows the spectral break.
6.4.2 Jet formation

An inevitable question is how the jets
come into existence. It was spec-
ulated early-on that it is the mag-
netic field near to the accretion disk
that plays a cardinal role in jet for-
mation. Meanwhile, a large number
of MHD simulations has been carried
out (an example shown in Fig. 6.12),
which in essence reflect the response
of the magnetic field to the rotation
of the accretion disk. Since the field is
tightly coupled to that disk, its orini-
gally poloidal structure is quickly de-
formed, with more and more toro-
dial components being generated, un-
til the the field is strongly wound up.
The plasma is now accelerated in this
torodial field by what can be inter-
preted as a centrifugal force in a co- Figure 6.12: MHD simulations of a jet.
rotating frame (~j × B ~ forces). In the
end, this force is due to the magnetic pressure gradient.
Hydrodynamic numerical simulations have been carried out quite successfully, these al-
lowing to deduce the physical conditions in the ICM or IGM into which the jets and lobes of
radio galaxies are expanding, i.e. the pressure, temperature, and magnetic-field strength and
structure. Apart from getting a clue how the jets are produced via twisting of the magnetic
fields, these were found to have two important effects on the shape and evolution of radio
galaxies:
(i) Magnetic fields damp Kelvin-Helmholtz instabilities at the interface between the cocoon
and the ambient gas (that of the IGM), this stabilizing the contact surface. This way,
pronounced jet heads and lobes such as seen in Cyg A and other FR IIs are produced.
(ii) When including magnetic fields in the simulations, the amount of entrainment of the
ambient gas is much lower than without magnetic fields.
6.4.3 FR I sources
FR I sources are quite different from FR IIs both, in terms of their luminosity and their
morphology. The main features and properties of FR I radio galaxies are:
• They are edge-darkened, i.e. their brightness decreases steadily from their central jets
towards their outer lobes.
• They have radio luminosities of L1.4GHz ≤ 1032 erg s−1 Hz−1 .
• They show one-sided jets close to the core.

• Their jet opening angles are ξ = drj /d` ≈ 0.1 within the first kpc. Between 1 and
10 kpc flaring occurs, with an opening angle drj /d` ≈ 0.25 . . . 0.6 beyond that point.
• At this transition point, the flaring is accompanied by a brightening and broadening of

the jet.
• Along the inner part, the magnetic field is prevailingly parallel to the jet, while a
perpendicular orientation of the magnetic field w.r.t. the jet dominates further out.
• The synchrotron spectra steepen already in the diffuse lobes, with the steepening in-
creasing further outward.
Essentially all of these properties are different from, or

even opposed to, those of FR II sources. This dichotomy
between FR I and FR II sources must be connected to a
combination of mass infall rates and the density profile of
the ISM in the host (elliptical) galaxies. A prototypical
example of an FR-I radio galaxy is 3C 31, with a radio
image shown in Fig. 6.13.
A variant of FRI’s is seen in the form of so-called ‘head-
tail radio sources’, where one distinguishes further between
narrow-angle tails (NAT) and wide-angle tails (WAT). The
nature of these sources is the same as for ‘untailed’ ones,
except that they occur in denser cluster environments.
What happens here is that the twin jets ‘drill’ their chan-
nels through the ISM of the host galaxy and suddenly are
exposed to the ram pressure of the surrounding intergalac-
tic gas. From Euler’s equation, we obtain
R
ρj · vj2 ≈ Pram · , (6.9)
rj
where R is the curvature radius of the jet and rj , vj , Figure 6.13: Radio image of the
and ρj are the jet’s radius, speed and density, respectively. FR-I radio galaxy 3C 31.
With
1 2
Pram = · ρICM · vICM (6.10)
2
one has a handle on the (relative) wind speed of the ICM and may thus explore the ‘cluster
weather’, since all other quantities can (at least roughly) be inferred from observations. It is
meanwhile clear that such sources preferentially exist in unrelaxed galaxy clusters, i.e. those
which just have undergone or are experiencing a merger. Such mergers give rise to large-scale
turbulence, hence to strong winds inside the clusters. The particular case of the Perseus
Cluster, in which the radio ‘trails’ of the NAT source NGC 1265 has been traced over a huge
projected distance, provides very strong evidence for this. The radio tail obvioulsy does not
reflect any ‘ballistic’ motion of this radio galaxy in the gravitational potential of the cluster,
as it is characterised by several kinks. These must be the result of intracluster turbulence
deflecting the trailing radio tails of this source (see Fig. 6.14).
This scenario is nicely discussed in an article by Burns (1998)3 , in which he describes the
effect of continuous infall of dark matter and baryons in cluster mergers from their peripheries
inward, this producing shocks, turbulence and winds exceeding speeds of 1000 km s−1 in hy-
drodynamic simulations, corroborated by the observed X-ray morphologies and temperatures,
and their gradients.
Figure 6.14: The head-tail radio galaxy NGC 1265 at different frequencies and angular reso-
lutions. On the right a radio image of the central Perseus Cluster is shown, with NGC 1265
in the north.
Another variant are the so-called X-shaped radio galaxies (Fig. 6.15). Their morphology
is best explained in terms of supermassive binary black holes, which cause precession of the
accretion disk of the AGN showing the jets. This leads to the X-shape, consisting of young
double lobes, which are still being energised by the jets, while the aged lobes exhibit radio
continuum spectra with low break frequencies.
In Sect. 7.7 we will discuss the interaction of radio galaxies with the ICM/IGM, and
discuss the feedback of AGN onto these media. In particular, it turns out that the observed
distribution of rotation measures across the radio galaxies provide valuable information about
the surroundings of radio galaxies, rather than being intrisic to them. Hence, radio galaxies
constitute invaluable tools to study the physical parameters of the magneto-ionic medium in
galaxy clusters.
3
”Stormy Weather in Galaxy Clusters”, J.A. Burns, Science 280, 345, 1998.
Figure 6.15: Potpourri of so-called X-shaped radio galaxies.

6.5. QUASARS 123
6.5 Quasars
Quasars are the most luminous objects in the universe. Their
formation commenced early-on, the most distant QSO known to
date having a redshift of z = 7.1 (770 Mio. years after the Big
Bang). The redshift distribution (Fig. 6.16) shows a maximum
just below z ≈ 2, with a pronounced truncation of QSO forma-
tion below z ≈ 0.4. QSOs reflects the history of major mergers
in the universe, leading to supermassive black holes (SMBH),
with the gravitational energy representing the biggest reservoir
for energy production, it is these objects that are able to produce
the highest luminosities via infall of matter onto accretion disks
surrounding the SMBHs. Local templates of QSOs in the pro- Figure 6.16: Redshift dis-
cess of formation are so-called ultra-luminous infrared galaxies tribution of QSOs.
(ULIRGs), such as Arp 220 and NGC 6240. The enormous lumi-
nosity of QSOs is due to friction in their accretion disks, which heats these up to Tad ≤ 106 K.
This friction arises from the fact that conservation of angular momentum does not allow mat-
ter to fall onto or into the black hole straight, but rather ‘spirals’ inward in an accretion disk,
which is subject to very strong differential rotation.
At a redshift of z ≈ 3, the feeding rate
onto QSO SMBHs was about 300 times that
of what is found in the local universe. Most
giant elliptical galaxies can be considered as
‘dead quasars’ - the relics of a period with
stronger merger and hence QSO activity. Not
all QSOs are also strong radio (i.e. syn-
chrotron) sources. In fact, there are many
more ‘radio-quiet’ than ‘radio-loud’ quasars.
This is quantified by the R-parameter, which
is the ratio of optical-to-radio flux density:
S(4400 Å)
R= , (6.11)
S(5 GHz)
and R = 10 being defined as the dividing

line between radio-quiet and radio-loud. Re-
cent MHD simulations of rotating black holes
with accretion disks seem to indicate that the
spin of the central SMBH may be the under-
lying reason for this dichotomy.
Radio-loud quasars show one-sided jets,
which sometimes produce large-scale lobes Figure 6.17: The quasar 3C 120 at different
(Fig. 6.17). Their jets can be traced down to wavelengths and angular resolutions.
pc-scales by means of (space-) VLBI. Monitoring the jet components4 discloses changes in
the jet morphologies, with mulitple components being traced over decades. It is thus possible
to construct kinematic models of the ejection and propagation of the relativistic plasma.
4
MOJAVE: Monitoring of Jets in AGN with VLBA Experiments.
Quite a bit of information has recently been gained by measurements of the linear polari-
sation and rotation-measure studies of the central regions of QSOs with the VLBA. However,
the interpretation of observed rotation measures in the central regions of AGN may not be
straight forward. There are several possible effects that could mimic Faraday rotation where
there is none. High-resolution VLBI observations reveal an unresolved core in quasars, plus
jet components that have recently emerged from that core regon. These components might
exhibit different polarisation angles, owing to
(i) different intrinsic polarisation angles,
(ii) relativistic aberration of subcomponents having identical polarisation angles but differ-
ent relativistic speeds,
(iii) a rotation-measure gradient across the core.
If the synchrotron opacity of the components is different or if they have different spec-
tral indices, this could change their relative preponderance, which would result in dramatic
changes of the polarisation angle as a function of frequency.
Figure 6.18: Rotation measures in the central regions of quasars.
Recent VLBA observations of the quasars 3C 273, 3C 279, and 3C 380 have revealed the
following interesting properties (see also Fig. 6.18): the central cores have rotation measures
of RM > 1000 rad m−2 . Beyond about 20 pc h−1 75 (h75 = H0 /75 km s
−1 Mpc−1 ). They show
−2
rotation measures of RM < 100 rad m . Such sharp rotation-measure gradients cannot be
produced by magnetic fields in the forground, i.e. either in galaxy clusters or in the Milky
Way. They must be produced by ordered magnetic fields in the central regions of these QSOs
on scales of 1 . . . 100 pc. If associated with the NLR5 , then Bu ≈ 0.05 G on a 10-pc scale is
inferred from the observed rotation measures. The high rotation measures found in these QSO
cores account for the depolarisation found at cm wavelengths. They result from irregularities
in the Faraday screens on scales smaller than the telescope beam.
5
The BLR can be ruled out: it is too dense, hence it must be depolarised.
6.6. SEYFERT GALAXIES 125
6.6 Seyfert galaxies
As mentioned before, Seyfert galaxies are mostly spiral galaxies hosting an AGN in their
centre. Depending on the viewing angle, the observer sees a Seyfert-1 galaxy with broad
emission lines, or a Seyfert-2 type with narrow emission lines. What is naturally common to
all Seyfert galaxies is that they stick out of the radio-FIR correlation (Fig. 6.1), in the sense
that they are radio-overluminous compared to normal star-forming galaxies. The radio excess
is not as large as in case of radio galaxies or QSOs, but since this correlation is so tight for
normal galaxies, it can be used to pick out most active spirals right away.
Most Seyfert galaxies host compact sources of synchrotron radiation produced by the
central AGN. However, there are quite a few examples in which the spiral disk galaxies have
produced extended radio lobes - in some cases having very much the same morphology as
classical radio lobes produced by radio galaxies (see Fig. 5.6). Apparently, in these cases the
orientation of the jets is such that they avoid the densest ISM and make it out of the gaseous
disk. Here, we briefly feature these objects by describing their radio continuum properties in
tabular form. Note that all sizes are projected sizes, so the observed structures could still be
larger.
Galaxy/object Morphology jet orientation ~

B-field
0313-192, 350 kpc double source 30◦ off the poles not known
QSO / Blazar? with jets
0400-181 jets & double lobes; not known not known
size unknown
(no distance known)
0412+040 90 kpc; not known not known
Seyfert 2 core & double lobes
IC 2497 extended radio continuum ∼ ⊥ disk, not known
⊥ to disk; likely to be
LINER† an exhausted AGN
NGC 1068 1-kpc lobes; nucleus, slightly inclined similar to that in
Seyfert 2 inner ’hot-spots’ out of the disk classical lobes;
at lobes’ head:
Bt ∼ 400 − 600 µG
NGC 3367 12 kpc; double lobes, slightly inclined not known
Seyfert-like S-shaped ⇒ interaction out of the disk (Faraday rotation)
with ambient gas? (poln. asymmetry)
NGC 4258 14 kpc; ”anomalous radio most likely in plane k to jet
Seyfert/LINER arms, S-shaped; bifurcation (from RM)
†A low-ionisation nuclear emission-line region (LINER) is a galactic nucleus showing spec-

tra that typically include line emission from weakly ionised or neutral atoms, such as O, O+ ,
N++ , and S+ , with emission from strongly ionised atoms, such as O++ , Ne++ , and He+ being
relatively weak.
6.7 Origin of magnetic fields

The origin of magnetic fields in AGN is still a matter of speculation. The easiest path of
arguing is probably connected with the rotational energy of the central black hole. M. Rees
(2006) takes an optimistic view in saying that the time scales for any relevant processes are
extremely short close to a black hole - maybe as short as hours, so that there has always been
enough time for a battery process to operate, or for a dynamo to be seeded by an infitesimal
field.
Following Hoyle (1969), one can give an order of magnitude for the possible field strengths
in AGN. The rotational energy of a compact central object from which the magnetic energy
is tapped can be parametrised as
Erot = f M• c2 , (6.12)
where M• is the mass of the compact object and the efficiency factor f < 1. If there is
equipartition between the magnetic field and the rotational energy in the fluid (achieved by
differential rotation and/or by the dynamo action) this implies a field strength of
12
8 π f M• c2

Bc ≈ , (6.13)
Vc
where Vc is the volume of the central region. The accretion disks offer optimum conditions
to drive a dynamo. For a region of size ∼ 1 pc, we then have
1
M• 2 1
Bc = 0.6 · · f 2 G. (6.14)
M
Chapter 7
Intergalactic magnetic fields
7.1 Clusters of galaxies

Clusters of galaxies are the largest gravitationally bound entities in the universe. They
comprise up to several thousand galaxies, the majority of which are dwarf galaxies though.
For example, the Virgo Cluster contains some 250 large galaxies and more than 2000 smaller
ones. Small assemblies of galaxies are referred to as groups. For instance, the local group
comprises two really massive galaxies, the Milky Way and M 31, and a third one, M 33, being
somewhat less massive, followed by the Large and Small Magellanic Cloud and some 30 other
dwarf galaxies with masses down to what is comparable to the largest globular clusters.
An important ingredient of galaxy clusters is the ICM (or IGM), which is hot (T ∼ = 108 K),
tenuous (ne ≈ 10−3 cm−3 ) gas, and which is well ‘visible’ through its bright X-ray emission
(LX ≈ 1043 . . . 1045 erg s−1 ), produced by thermal bremsstrahlung. Another constituent of the
ICM is its relativistic plasma, consisting of relativistic particles and magnetic fields - the topic
of this lecture, which we shall come back to. The cluster mass is, as is well-known, dominated
by dark matter, as inferred from the galaxies’ velocity dispersion, from the temperature of
the hot gas, and from gravitational lensing. In the framework of the hierarchical scenario of
galaxy cluster formation in a ΛCDM cosmology, large clusters have formed by gravitational
mergers from smaller entities, i.e. from groups or subclusters. In these mergers, high kinetic
energies are involved, viz. up to 1064 erg. A large fraction of this energy is dissipated in the
ICM in the form of shocks, turbulence and bulk motion, which heat up the gas.
Many clusters exhibit strongly peaked X-ray surface brightnesses in their centres, implying
high densities, and cooling times within the inner 100 kpc of tHubble . Hence, in order to
maintain hydrostatic equilibrium, inward flows may be required. X-ray observations with
XMM-Newton, which has a high spectral resolution, did not indicate any evidence for large
amounts of cooling and condensing gas in the centres of galaxy clusters. It thus seems that
such cooling flows’ are hindered by some as yet unknown mechanism. There is consensus that
‘cooling core’ clusters are dynamically more relaxed than non-cooling core ones.
A full understanding of the ICM necessitates adequate knowledge of the role of the non-
thermal component, the best information about which is obviously provided by radio contin-
uum observations. A very strong argument comes from the fact that quite a number of clusters
exhibits diffuse radio sources that cannot be associated with any discrete sources powering
them, but are rather connected with the general ICM. The observed synchrotron emission
requires ∼ GeV electrons, hence highly relativistic particles, and magnetic fields at the level
127
128 CHAPTER 7. INTERGALACTIC MAGNETIC FIELDS
of µG strengths, which must pervade - at least part of - the cluster volumes. Independent ev-
idence for such magnetic fields comes from rotation measures of polarised background sources
or radio galaxies located within these clusters. Finally, independent evidence for such a popu-
lation of relativistic particles also comes from nonthermal emission of inverse-Compton origin
seen in the regime of hard X-rays and possibly in the EUV. The combination of the diffuse
radio continuum emission with the hard X-ray radiation can be used to estimate intra-cluster
magnetic-field strengths.
7.2 Radio haloes

Cluster radio haloes are the most spectacular manifestation of diffuse nonthermal emission
and widespread magnetic fields seen over large portions of galaxy clusters. They permeate
them on scales of more than a Mpc, with the following properties:
- They have intensities of I1.4GHz ≈ 10−6 Jy/u

t00 .
- Their radio spectral index is α ≥ 1 (Iν ∝ ν −α ).
- They are unpolarised down to the current detection levels for polarised radio synchrotron
emission.
For comparison, the mean radio brightness of the radio lobes of the classical radio galaxy
Cyg A is of order 0.5 Jy/ut00 , hence 500 000 times more intense1 . The first such diffuse cluster
radio source was reported by Willson (1970), and this structure was later on mapped with
ever increasing resolution and sensitivity. This structure, discovered in the Coma Cluster of
galaxies (A 1656), soon showed a spectral steepening from the centre towards the periphery,
with α = 0.8 in the centre to α = 1.8, and hαi = 1.3. The break frequency, reflecting the
synchrotron age of the radiating particles, was determined as νc ≥ 1 GHz, and the average
total magnetic-field strength as Bt ≈ 0.6 µG or, more precisely (the equipartition strength)
Beq = 0.57 · (1 + k)0.26 µG, (7.1)
where k is the ratio of the energy contained by the protons to that contained by the
electrons. It is a common feature of all clusters with radio haloes that they are X-ray
luminous. Such radio haloes have been found in about one quarter of all clusters with
LX > 5 · 1044 erg s−1 . Studies of several other cluster haloes have been performed, such
as A 2163, A 2256, 1E0657-57, and distant clusters like A 2744 (z = 0.308) and CL 0016716
(z = 0.5545), the latter being the most distant cluster with a radio halo known to date.
Smaller-sized haloes (500 . . . 600 kpc) have been detected in some cases (A 2218, A 3562).
A common property of cluster radio haloes is their lack of any strong (linear) polarisation.
The degree of polarisation is < 10% (Coma), or even < 5% in others, obtained at 1.4 GHz.
Obviously, any polarised radiation emitted by these central radio haloes traverse large lines
of sight through the cluster medium. This, along with the arguments given below, delivers
the likely reasons for the low degrees of polarisation:
1
Recall that brightness is distance-independent!
7.2. RADIO HALOES 129
- Internal depolarisation, owing to mixing of the magnetic fields with the thermal gas.
- Turbulence causing highly disordered fields, which could come along with a likely mech-
anism that accelerates the particles, viz. galactic wakes.
The low surface brightness of the radio haloes has so far restricted any studies to mea-
surements with low angular resolution. This is changing now since LOFAR starts producing
high-quality images of galaxy clusters at the lowest radio frequencies.
What is the origin of cluster haloes? They are not associated with any discrete sources
such as radio galaxies. The various possible models can basically be divided into two classes:
• Primary electron models, in which the relativistic electrons are injected into the IGM
by AGN (QSOs, radio galaxies,...) and/or by starburst galaxies (SNe, galactic winds).
However, the radiative lifetime of such particles is rather short,
B −2
−1
9 E
t 1 = 8.35 · 10 · · yr; (7.2)
2 µG GeV
hence such particles would be rendered invisible after 107 . . . 108 yrs, and would thus
require continuous injection by sources (which is in conflict with the observations) or by
some reacceleration mechanism. One possible process is turbulence caused by galactic
wakes, although its efficiency is a matter of debate.
• Secondary electron models, in which the relativistic electrons result as secondary prod-
ucts in hadronic collisions. Relativistic protons have lifetimes exceeding a Hubble time.
They can thus propagate over large distances from their site of origin. In the cen-
tral regions of clusters, they frequently collide with the thermal protons of the ICM,
thereby producing relativistic positrons and γ-rays from the π 0 decay. Future γ-ray
observatories can therefore test such models independently.
The production of secondaries proceeds ac-
cordingly to the following chain (see also Fig. 7.1):
p + p → π +/0 + 2 N,
where 2 N is any combination of particles.
This is followed by a mesonic decay
π + → µ+ + νµ ,
π 0 → 2 γ,
and by the leptonic decay
µ+ → e+ + νe + ν̄µ .
So this finally results in relativistic electrons Figure 7.1: Illustration of a hadronic col-
or positrons that we could see in the form of syn- lisions in the ICM.
chrotron radiation. A correlation between the ra-
dio luminosity of cluster haloes and the X-ray luminosity of their hot gas has been established,
which is
P1.4GHz ∝ L1.97±0.25
X , (7.3)
Figure 7.2: The Coma Cluster at X-ray energies (top) and at 90 cm wavelength.
7.3. ROTATION MEASURES 131
where the X-rays have been measured in the 0.1 - 2.4-keV energy band. This correlation
obviously supports models of secondary-electron production in hadronic collisions. On the
other hand, however, the fact that central radio haloes appear to be connected with clusters
having substructure, hence are undergoing or have undergone a merger, speaks in favour of
primary models, with turbulence and shocks as energy sources. Hence, this issue still awaits
clarification. This must come with future γ-ray observatories, with the LOFAR harvest, and
aided by numerical (MHD) simulations that are currently also underway.
As another nice example, the radio halo of A 2256 is shown in Fig. 7.3, which has been
thoroughly studied. It constitutes a perfect template for cluster radio haloes, but also exhibits
a so-called radio relic, which is a diffuse, peripheral structure (see Sect. 7.4).
Figure 7.3: Radio continuum images of the galaxy cluster A 2256.
7.3 Rotation measures

Faraday rotation of both, background as well as cluster-source emission is another, indepen-
dent hint at the ICM being magnetised. Clarke et al. (2001) analyzed the rotation measures
of background sources probing 16 low-redshift (z ≤ 0.1) galaxy clusters (Fig. 7.4), selected
to not possess any radio haloes or being indicative of any cooling flows. Combined with an
analysis of cluster-embedded sources, they found a cluster-generated rotation-measure excess
out to ∼ 0.5 Mpc h−1 75 from the cluster centres. The results imply magnetic fields with a high
filling factor and
− 1
` 2
1/2
hBi = (5 . . . 10) · · h75 µG, (7.4)
10 kpc
where ` is the correlation length of the magnetic field. Rotation-measure mapping in

clusters of galaxies has been performed in quite a number of cases using extended radio
galaxies. The first such high-resolution rotation-measure study was performed using Cyg A.
These results imply that a Faraday screen with −4000 rad m−2 < RM < +3000 rad m−2 is
required! The screen is implied because ∆ψ ∼ λ2 . Recall that strong deviations from this law
imply mixing of Faraday-rotation with the emitting medium (Sect. 3.3.2). In line with this
scenario of a foreground Faraday screen, the Faraday rotation and depolarisation is generally
lower for the lobe fed by the brighter jet, which is nearer side as seen from the observer. This
phenomenon has been coined ‘Laing-Garrington effect’. The observed variations or gradients
of the rotation measures, which are of the order of |RM | ≈ 300 rad m−2 per arcsec, have size
scales that are too small to be attributable to any variations in the Galactic foreground. From
these observations, magnetic fields of strength 2 . . . 10 µG, ordered over scales of 20 . . . 30 kpc,
are inferred.
Further detailed rotation-measure studies have been per-
formed on what is deemed cooling-core clusters, characterised
by powerful radio galaxies in their centres. Like in case of
Cyg A, these sources exhibit large rotation measures, with
strong spatial variations. Prominent examples are Hyd A,
3C 295, or Vir A. The inferred magnetic-field strengths range
from ∼ 5 µG to ∼ 30 µG. The ratio of thermal to magnetic
energy density, the so-called ‘plasma-β’,
nkT
β= , (7.5)
B 2 / 8π
is found to exceed unity significantly. For instance in Figure 7.4: Radial distribu-
3 C31 one finds β ≈ 10 at the radius of the galaxy group, tion of RMs in galaxy clusters.
and in 3C 449 one finds β ≈ 30 at the group radius, and β ≈ 400 in the centre, showing that
the magnetic field is likely to be dynamically insignificant in the central regions of groups and
clusters.
Rotation measures have also been mapped to some extent outside of the central regions
(Coma Cluster, A 119, A 514, A 400, A 2634, 3C 129). These data lead to estimates of the
magnetic-field strength of ∼ 2 . . . 8 µG, with ordered magnetic fields on scales of ∼ 5 . . . 15 kpc.
The variance σRM of the rotation measure as a function of distance from the cluster centre
can be inferred from the distribution of the thermal gas density, assuming that the strength
of the magnetic field obeys the same radial law. X-ray observations have revealed that the
thermal gas density can be best described by
− 32β
r2

ne (r) = n0 · 1 + 2 , (7.6)
rc
where n0 is the central density, rc is the core radius, and β is a free parameter (not to be
confused with the plasma-β!). Typical values are rc = 200 kpc, n0 = 10−2 cm−3 , β ≈ 32 . For
a Gaussian distribution of rotation measures with hRM i = 0, the variance is given by
Z
σRM = RM = (812) Lc (ne Bk )2 dr,
2 2 2

(7.7)
where Lc is the cell size, i.e. the size scale on which the magnetic field is tangled. Inserting
ne (r) from (7.6), the following expression results:
1 1 s
K B n0 rc2 Lc2 Γ(3 β − 0.5)
σRM (r) = 6 β−1 · , (7.8)
2 4 Γ(3 β)
1 + rrc
7.4. RADIO RELICS 133
where K = 624 if the source lies completely beyond the cluster and K = 441 if the source
is located half-way through the cluster.
Figure 7.5: Rotation measure and magnetic field in the cooling-core radio galaxy Hyd A.
7.4 Radio relics

Relic sources are a class of diffuse radio sources
located near the peripheries of galaxy clusters.
What they have in common with radio haloes is
the lack of any obvious radio galaxies that could
be powering them. However, unlike haloes, they
have an elongated or irregular shape, and they
are strongly polarised. The elongation is tangen-
tial to the cluster circumference. The prototype
of these is located at the south-western periphery
of the Coma Cluster (see Fig. 7.2), and is called
1253+275. It marks the contact surface between
the main cluster and a subcluster that is falling
into the main one from the south-west (projected
direction). The observed magnetic-field orienta-
tion is parallel to the (tangential) elongation of
the diffuse source.
Another prominent example is located in
A 2256. This is the cluster that also possesses a
prominent radio halo. Meanwhile, about a dozen
of clusters is known to possess radio haloes and/or Figure 7.6: Radio relic in the galaxy clus-
relics. One of the most spectacular structures ter CIZA J2242.8+5301.
of this kind has been discovered recently in the northern outskirts of the galaxy cluster
CIZA J2242.8+5301 at a redshift of z = 0.1921. Fig. 7.6 shows the pronounced gradient
in spectral index across the ‘sausage’-like structure (top panel) as well as the strong linear
polarisation, with degrees of p ≈ 50 . . . 60%. Its total length is 2 Mpc. A double radio relic
in the merging galaxy cluster ZwCl 0008.8+5215 is shown in Fig. 7.7, which illustrates the
opposite location of the relics, with the X-ray emission shown in colour, and the galaxy dis-
tribution as dashed contours. Note how both, the gas and the galaxy distribution have been
been stretched in the direction of the merging, with the relic structures marking the opposite
locations of the shocks produced in this process.
Figure 7.7: Double relic in the galaxy cluster ZwCl 0008.8+5215.
In brief, the properties of relics are:
- They have sizes of up to several Mpc.
- They are highly polarised, p1.4GHz ∼ 10 . . . 50%.
- They possess steep radio spectra, α ≥ 1.0.
- Their locations are mostly peripheral.
- They are commonly found in merging clusters.
The most straight-forward explanation for this kind of diffuse synchrotron source in galaxy
clusters is considered in the framework of the cluster-merging process:
(i) Fermi-I type, diffusive shock acceleration of ICM electrons (the somewhat stupid term
‘radio gischt’ has been coined for this phenomenon).
(ii) Adiabatic compression of a fossile radio plasma in cluster shocks. The plasma is the
relic of mildly relativistic particles injected by AGN (or starburst galaxies) in the past
(this phenomenon is referred to as ‘phoenix’).
7.4. RADIO RELICS 135
The term ‘radio gischt’ stands for large sources. These cannot have been produced by
compression, because the time scale for this compression would be too long to still see the
particles radiating at GHz frequencies. The assumption here is that elongated the structures
were equally extended in all directions prior to compression. The term ‘phoenix’ is used for
smaller sources, for which the compression time scale is short enough to still see the relativistic
particles radiating.
First analytical models of shock acceleration were able to reproduce the observed polarisa-
tion properties and spectral indices. In particular, the measured degree of linear polarisation
is a good test, since the shock geometry predicts a dependence of the viewing angle of the relic
structure. This relation has in fact been confirmed by observations. The predicted degree of
linear polarisation as a function of viewing angle δ is
g+1 sin2 δ
hP (δ)i = · , (7.9)
g + 73 F (R) − sin2 δ
where g is the power-law spectral index of the relativistic particles. The function F(R)
depends differently on the compression ratio R:
(
2 13 R−7
15 · R−1 strong magnetic fields
F (R) = 2R 2 (7.10)
R2 −1
weak magnetic fields
‘Weak’ and ‘strong’ here refer to the relative importance of the magnetic vs. the thermal
pressure (plasma-β),
> B2
nkT < .
8π
Figure 7.8: Three-dimensional MHD simulation of cluster shock waves.
Three-dimensional MHD simulations of such passages of shock waves through a relic radio
plasma have also been performed, with synthetic radio maps resulting from these (Enßlin &
Brüggen, 2002; see also Fig. 7.8). Any such MHD simulation reproducing the phenomenon
of cluster radio haloes is obvioulsy a much more complex task and would require the imple-
mentation of MHD in large-scale cosmological simulations of hierarchial structure formation.
7.5 Mini haloes

The so-called mini-haloes in some cluster centres have a nature that is entirely different from
the diffuse radio sources discussed in the preceding sections. As the name says, they are a lot
smaller (≤ 100 kpc) and are found around powerful radio galaxies residing at the centres of
what are deemed to be cooling-flow clusters. The prototype is the ’radio halo’ around Per A
(NGC 1275, 3C 84).
Figure 7.9: The radio galaxy Vir A and its halo.
Other examples are Vir A (Fig. 7.9), PKS 0745-191, and possibly A 2390. Like other
extended cluster sources, they also show steep radio spectra. The magnetic-field strength
inferred from their (high) radio brightness is of order Bt ≈ 10 . . . 20 µG. The radiative lifetimes
of relativistic electrons in magnetic fields as strong as this are short: such particles would be
observable for at most a few times 107 yr - much too short to make it to the distances marked
by the mini-haloes, away from the central radio galaxies. It has therefore been proposed that
mini-haloes are powered by MHD turbulence in these cooling-flow centres. Gitti et al. (2004)
have argued that the power necessary to accelerate the relic electron population is a mere
0.7% of the maximum power that can be tapped from the cooling flow.
But, also here, there is an alternative explanation: the relativistic electrons could again
be of secondary origin, being produced by hadronic collisions of the relativitic protons with
the thermal ones.
7.6 Inverse-Compton emission

A completely independent way of determining magnetic-field strengths
would rest upon inverse-Compton X-ray emission, together with synchrotron radiation.
Inverse-Compton emission is the relativistic extrapolation of the Sunyaev-Zel’dovich effect2 ,
2
The SZ effect implies scattering of CMB photons off the thermal free electrons off the hot plasma of galaxy
clusters, leading to a decrement of the CMB brightness towards such clusters.
7.6. INVERSE-COMPTON EMISSION 137
in which the ambient photon field is up-scattered into the hard X-ray regime by the same
relativistic particles that produce the synchrotron radiation. This inverse-Compton process
involves two Lorentz transformations (to and from the rest frame of the electron), plus Thomp-
son scattering in the rest frame of the electrons, resulting in the emergent photon frequency
4 2
νIC = · γ νCM B , (7.11)
3
where νCM B is the frequency of the incident CMB photons, and γ the Lorentz factor of the
relativistic electrons. Detecting the nonthermal hard X-ray and synchrotron radiation that is
produced by the same population of relativistic electrons, allows us to unambigously estimate
the volume-averaged strength of the intra-cluster magnetic-field. The exact derivations of the
respective flux densities by Blumenthal & Gould (1970) yield the following expressions:
g−1
V0 4 π e3

g+1 3e 2 g−1
Ssyn (νr ) = 2 · 2 g
· N0 B 2 · · A(g) ν − 2 (7.12)
4 πDL (me c ) 4 z π me c
V0 8 π 2 a20 − g+3 g+5 g−1

SIC (νx ) = 2 · 2
· h 2 N0 (me c2 )1−g · (k TCM B ) 2 · F (g) ν − 2 (7.13)
4 π DL c
Here:
a0 = classical electron radius, a0 = e2 /me c2 = 2.82 · 10−13 cm
h = Planck’s constant, h = 6.6262 · 10−27 erg s
g = power-law index of the relativistic electron spectrum
V0 = effective source volume, V 0 = f · V , where f = filling factor
B = magnetic-field strength
TCM B = radiation temperature of CMB, T = T0 · (1 + z)
N0 = amplitude of the electron spectrum
A(g), F (g) = tabulated functions given by Blumenthal & Gold (1970)
DL = luminosity distance of the cluster
These two equations can be solved for the magnetic-field

strength to yield
2 g−1
Ssyn (νr ) g+1 νr g+1
B∝ · . (7.14)
SIC (νx ) νx
The CMB Planck function has its maximum at νCM B ≈
1.6 · 1011 Hz, hence hard X-rays observed at 20 keV (νIC =
4.8 · 1018 Hz are produced by electrons with γ = 5000 (in-
dependent of redshift, since νCM B = νCM B,0 · (1 + z)). The Figure 7.10: Possible inverse-
corresponding synchrotron emission of such relativistic elec- Compton excess emission in
trons peaks at a (rest frame) frequency of the Coma Cluster.

B B
νsyn ≈ 4 · · γ 2 Hz = 100 · MHz. (7.15)
µG µG
The principal difficulty of measuring the inverse-Compton emission from galaxy clusters
in the hard X-ray regime is the contamination or confusion by the thermal emission from the
cluster atmosphere. One hence has to seperate the thermal from the (power-law) nonthermal
radiation through very sensitive spectroscopic X-ray measurements. These were rendered
feasible with the Beppo/Sax and RXTE missions, allowing sensitive measurements well above
10 keV. This way, field strengths in the range B ≈ 0.2 . . . 1 µG have been derived (Fig. 7.10).
These appear to be lower than thosed inferred from radio data alone. An interesting solution
of this problem is to account for small pitch angles of the electrons, which would produce
weaker synchrotron but still the same inverse-Compton emission!
7.7 Magnetisation of the IGM

The question of what injects particles and fields in galaxy clusters finally needs to be discussed.
In view of the observed nonthermal phenomena, one might speculate whether or not galaxy
clusters are filled with a (mildly) relativistic plasma throughout, which at least in view of
the statistics of rotation measures (Fig. 7.4) seems to be the case. A primoridal origin of
magnetic fields (Biermann battery) is conceivable; however, the time scale to amplify them to
the observed strengths are just too long, since there must be a mechanism that amplifies the
magnetic fields from > −18 G to > 1 µG. It is therefore near at hand to consider injection
∼ 10 ∼
of particles and fields by active galaxies. These could have been
• AGN
• starburst (dwarf) galaxies
Figure 7.11: Cavities produced by the radio plasma of 3C 84 (left) and MS 0735.6+7421
(right).
The injection of radio plasma into the ICM by AGN can still be seen, even in the local
universe. Radio galaxies fill large volumes with the plasma, sometimes leaving behind pro-
nounced cavities in the thermal plasma as evident in many X-ray observations. Prominent
examples are Cyg A and Per A (Fig. 7.11, left). The central galaxies in the corresponding
clusters have luminosities between ∼ 2 · 1038 and 7 · 1044 erg s−1 . The X-ray cavities have
average radii of ∼ 10 kpc and average projected distances of ∼ 20 kpc from the central
galaxy. The minimum energies associated with these activities are between P · V = 1055 erg
(in galaxy groups) and P · V = 1060 erg (in rich clusters). The largest such structure was,
however, found in the z = 0.22 cluster MS 0735.6+7421 (Fig. 7.11, right). The cavities there
7.7. MAGNETISATION OF THE IGM 139
measure ∼ 200 kpc in diameter, the inflation work is P · V = 1061 erg for each cavity. The
enthalpy
H = U + P V, (7.16)
where U is the internal energy, is much larger than that of other sources with prominent
cavities:
H0735.6 ≈ 10 · HHydA ,
≈ 15 · HCygA ,
≈ 250 · HP erA ,
≈ 104 · HV irA
Another source of relativistic plasma can be seen in starburst galaxies, in particular low-
mass ones. Kronberg et al. (1999) were the first to discuss possible seeding of magnetic
fields (along with the ejection of relativistic particles) in the ICM by dwarf galaxies. This
was treated more quantitatively, with predictions of magnetic-field strengths, by Bertone et
al. (2006). Energetic winds from galaxies should (have) be(en) able to provide seed fields
with a strength of Bt ≈ 10−12 . . . 10−8 G in the ICM, strong enough to be amplified to the
observed values by the turbulent dynamo. Dwarf galaxies could have been responsible for (at
least part of) the magnetisation of the ICM because of two advantages they bring:
1. They were/are numerous in the framework of the ΛCDM bottom-up scenario of struc-
ture formation.
2. Their shallow gravitational potentials imply low escape speeds,

r
2GM
vesc = .
R
In judging their relative importance in magnetizing the ICM one has to compare radio
powers:
(i) typical starburst dwarf galaxies produce P1.4GHz ≈ 1020.5 : W Hz−1 .
(ii) radio galaxies in the transition regime between FR I and FR II produce P1.4GHz ≈
1024.7 W Hz−1 ,
i.e. PF RI/II ≈ 15 000 · PSB . However, cosmology helps a lot here, since dwarf galaxies are
much more numerous than massive spiral or ellipticals. This was all the more true in the early
universe, which was a lot smaller then, helping to ‘pollute’ it with heavy elements, relativistic
particles, and magnetic fields. Furthermore, the duty cycle of radio galaxies is probably low.
From source statistics, Bird et al. (2008) estimate the lifetime of a radio galaxy to be
τRG ≈ 1.5 · 107 yr (7.17)
and a duty cycle of
τDC ≈ 8 · 108 yr, (7.18)

which yields an effective activity period of
τAP ≈ 2 · 108 yr ≈ 0.018 · tHubble . (7.19)
This time scale can be estimated via

τuniv
τAP = · τRG , (7.20)
τRG + τDC
where the ratio on the right-hand-side of this equation is the number of activity periods.
There are local template dwarf galaxies that are undergoing a starburst (e.g. NGC 4449),
or have been undergoing one recently (e.g. NGC 1569, see Sect. 5.3). During the lifetime
of the ejected relativistic particles these could propagate a considerable distance away from
the galaxy, once they escaped from it. Outside the galaxy the energy losses are governed by
inverse-Compton losses against the CMB photons.
Even though such particles age fairly rapidly so that they become invisible at GHz fre-
quencies after ∼ 108 yr, the new low-frequency developments in radio astronomy provide ideal
tools to trace them. With LOFAR working in two bands, the ‘low band’ (10 - 80 MHz) and
the ‘high band’ (120 - 250 MHz), it will have the best low-frequency coverage to disclose the
secrets of the nonthermal universe, including the linear polarisation. It will thus be possible to
trace relic haloes around all kinds of formerly active galaxies, also dwarf galaxies. Previously
starbursting dwarf galaxies should be ‘wrapped’ in nonthermal radio haloes, detectable only
at the lowest frequency.
If the relativistic electrons would propagate without ‘intergalactic weather’, i.e. just by
diffusion, then they could maximally move with the Alfvén speed:
ne − 21

B
vA = 2.2 · · km s−1 (7.21)
µG cm−3
where ne is thermal electron density. Taking B = 1 µG and ne = 0.001 cm−3 , the parti-
cales radiating at 120 MHz could move out to 30 . . . 40 kpc within 500 Myr, or correspondingly
further if caught at still lower frequencies. Hence, such haloes in which we should see star-
burst dwarf galaxies wrapped could have sizes of 60 . . . 80 kpc. LOFAR should be sensitive
enough to detect such nonthermal haloes.
In case of ‘stormy cluster weather’, however, the situation will change dramatically. If
located in a massive cluster undergoing merging, the ejected radio plasma would ‘fly away’
at high speeds, of order 1000 km s−1 . This is what one actually observes in the tails of the
much more powerful WAT/NAT sources. Both, particles and fields would in this way spread
out very quickly over the cluster volume, also taking into account the large number of such
dwarf galaxies.
In any case, and doubtlessly, LOFAR will make a significant contribution to the scenarios
discussed above. Even though we cannot trace the mechanisms back to the era of initial
galaxy formation, we can test the hypotheses using observations in the local universe.
Chapter 8
Cosmological magnetic fields
Finally, a few words are mandatory about a cardinal question, the discussion of which can
become very complex and is still a matter of much speculation. Whatever mechanism (in
clusters, galaxies or smaller scales) produces magnetic fields, there is the requirement of seed
fields that have to be amplified. This holds especially for systems with large dynamical time
scales (formation, rotation, turbulence). For instance, the sun has existed for some 1011
dynamical time scales, while the Milky Way is likely to have seen a mere 50 or so, which
means that there were only 50 e-folding times to build up the magnetic field. Even worse:
strong magnetic fields are already present at large redshifts.1
Various authors have thought about magnetic-field creation in exotic, ultra-dense stages
of the Big Bang, by considering what might have happened during phase transitions:
- electro-weak: 10−12 s.
- quark-hadron: 10−5 s; the event horizon then encompassed 106 M of baryons, hence
the resulting field on galactic scales would be only 10−30 G.
- GUT era: 10−36 . . . 10−34 s, 1015 GeV; the horizon then contained a mere 104 baryons;
even with a high local energy density, there was no chance to get to 10−20 G on galactic
scales.
Other scenarios deal with proto-galactic batteries, such as the Compton drag: since the
cross section for Inverse-Compton scattering is much larger for electrons than for protons, the
former lose energy against the photon background, this changing their bulk motion w.r.t the
latter. This produces a current, which in turn produces a magnetic field of 10−21 G at z = 5.
Another ingenious idea (Harrison 1970) rests upon the speculation whether there has been
vorticity in the primordial fluctuations. If so - and nobody really knows - there would have
been an interesting mechanism leading to a proton-electron seperation. Consider a spherical
region of radius r in the pre-recombination era, which consisted largely of a soup of photons,
protons, and electrons (plus dark matter, neutrinos,...). The electrons were tightly coupled
to the photons by Thomson scattering. The sphere has photons with uniform density ργ and
matter with uniform density ρm . As the eddy expands, we have
1
Cosmological magnetic fields have been discussed to some extent in the excellent review by L.M. Widrow,
Rev. Mod. Ph. 74, 775, 2002.
141
142 CHAPTER 8. COSMOLOGICAL MAGNETIC FIELDS
ρm r3 = const
and
ργ r4 = const.
If ωγ and ωm denote the angular velocities of the photon and matter spheres, respectively,
then the angular momenta
Lγ = ργ ωγ r5 (8.1)
and
Lm = ρm ωm r5 , (8.2)
are are seperately conserved. Hence we find
ωγ ∝ r−1 (8.3)
and
ωm ∝ r−2 , (8.4)
because the eddy expands with the cosmic scale factor. We thus see that radiation spins
down more slowly than matter. Because of the coupling of the electrons and photons by
Thomson scattering, the electron soup will rotate faster than the ion soup. The difference in
angular velocities leads to a magnetic field
~ = −2 · mH c · ω
B ~ = −2.1 · 10−4 · ω
~ G (8.5)
e
There may be other scenarios, but the intrigued student is referred here to specific liter-
ature (see the review by Widrow).
List of Figures
1.1 Earth’s magnetic field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Magnetic loops on the sun. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Lorentz force. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Different kinds of magnetism in solid bodies. . . . . . . . . . . . . . . . . . . 4
2.1 Geometry for a moving charged particle as seen from some point P . . . . . . 9
2.2 Trajectory of a thermal electron in a plasma. . . . . . . . . . . . . . . . . . . 10
2.3 Acceleration of an electron. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Definition of angles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 Relation between ẍ(t) and C(ω). . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.6 Approximation of C(ω). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.7 Ensemble of protons, through which electrons with the impact parameter in
the range (p , p + dp) are moving with speed v. . . . . . . . . . . . . . . . . . 14
2.8 Geometry of the hyperbolic orbit of the electron. . . . . . . . . . . . . . . . . 15
2.9 Continuum spectra of free-free radiation. . . . . . . . . . . . . . . . . . . . . . 19
2.10 Upper row: The Large Magellanic Cloud in Hα (left), and at 8.6 GHz (right).
Middle row: Centre of the Orion Nebula in Hα (left), contours of thermal
free-free radio emission, superimposed onto an optical image (right), with the
radio continuum spectrum below it, showing the low-frequency turnover due
to free-free absorption. Bottom row: Imprint of thermal absorption in M 82,
seen as a ’cavity’ in the overall nonthermal synchrotron radiation at 408 MHz. 20
2.11 Radiation pattern of the linearly accelerated electron (β = 0.8). . . . . . . . . 21
2.12 Illustration of the various angles used in describing the transverse acceleration
of a relativistic electron in a magnetic field. . . . . . . . . . . . . . . . . . . . 22
2.13 Radiation pattern of the transversely accelerated electron (β = 0.8). . . . . . 23
2.14 Left: Geometry of the Larmor circle. Right: Sketch of the emitted pulses of the
gyrating relativistic electrons as seen in the particles’ reference frame. Pulse
sequences are shown for three different values of β, and have been shifted for
the sake of clearity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.15 Illustration of the geometry of the transformation from the particle’s to the
observer’s reference frame. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.16 Sketch of the time dependence of the synchrotron pulses and their radiation
spectra. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.17 CR energy spectrum. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.18 Illustration of the radiation spectrum. . . . . . . . . . . . . . . . . . . . . . . 30
2.19 Radiation spectrum with energy cutoff. . . . . . . . . . . . . . . . . . . . . . 31
143
144 LIST OF FIGURES
2.20 Radio emission from the Milky Way at 408 MHz. . . . . . . . . . . . . . . . . 32

2.21 Radio emission of the radio galaxy Fornax A (colour), superimposed onto an
optical image. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.23 Orientation of the electric field E~ produced by the particle gyrating in a mag-
~
netic field B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.22 Velocity cone of a particle moving in a magnetic field. . . . . . . . . . . . . . 34
2.24 Intensities radiated parallel and perpendicular to the magnetic field. . . . . . 35
2.25 Polarisation ellipse. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.26 Poincaré sphere. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.27 Radio emission of the SNR 3C 10, with the magnetic-field orientation indicated
by the bars (see text for details). . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.28 Radio emission of the spiral galaxy NGC 4631 (contours), superimposed onto
an optical image. The white bars indicate the orientation of the magnetic field. 38
2.29 Left: Migration of the cutoff frequency resulting from particle ‘ageing’. Right:
Radio continuum spectrum of the starburst dwarf galaxy NGC 1569, exhibiting
a break in the synchrotron spectrum above ∼ 12 GHz. . . . . . . . . . . . . . 41
2.30 Synchrotron (contours) and inverse-Compton (colour) radiation from the radio
galaxy 3C 294. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.31 Superposition of various continuum radiation components in the radio spec-
trum of M 82. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.1 Elongated grains lining up in a magnetic field. . . . . . . . . . . . . . . . . . . 47

3.2 Geometry of rotating grain and magnetic field. . . . . . . . . . . . . . . . . . 49
3.3 Optical polarisation of the spiral galaxy M 51. . . . . . . . . . . . . . . . . . . 50
3.4 Minimum-energy and equipartition magnetic field strengths. . . . . . . . . . . 53
3.5 Geometry of circularly and linearly polarised waves. . . . . . . . . . . . . . . 55
3.6 Observed RM distribution across the spiral galaxy NGC 6946. . . . . . . . . 57
3.7 RM ambiguity: Measured values (blue) can be shifted by multiples of π (red)
such as to produce a different RM. . . . . . . . . . . . . . . . . . . . . . . . . 58
3.8 Bandpass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.9 Illustration of Faraday depth. . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.10 The complex (Q,U )-plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.11 Illustration of an unresolved foreground Faraday screen. . . . . . . . . . . . . 62
3.12 Dependence of the degree of polarisation and polarisation angle on wavelength
in case of internal depolarisation. . . . . . . . . . . . . . . . . . . . . . . . . 63
3.13 Internal depolarisation for three idealised source models. . . . . . . . . . . . . 64
3.14 The galaxy cluster A 2256 at 20 cm wavelength, with the total (synchrotron)
radiation shown in the left panel, and the degree of polarisation in the right
one. The peripheral relic structure is strongly polarised, while the central halo
is not. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.15 Rotation-measure transfer function. . . . . . . . . . . . . . . . . . . . . . . . 67
3.16 Illustration of the measurement of Zeeman splitting. . . . . . . . . . . . . . . 68
3.17 Examples of measurements of the Zeeman effect in interstellar clouds. . . . . 69
3.18 Spectropolarimetry of the CSJ = 2 → 1 line towards IRC+10216, with the
antenna temperature, the degree of linear polarisation, and the position angle
plotted vs. the local standard of rest velocity. . . . . . . . . . . . . . . . . . . 70
LIST OF FIGURES 145
4.1 Radio emission from the Milky Way at 1420 MHz, with total intensity in the top
panel, and polarised intensity in the lower one. Note the different appearances! 71
4.2 Rotation measures obtained from pulsars with known distances, superimposed
onto a sketch of the Milky-Way spiral arms. . . . . . . . . . . . . . . . . . . . 72
4.3 All-sky distribution of rotation measures. . . . . . . . . . . . . . . . . . . . . 73
4.4 All-sky maps at 1.4 GHz from the NRAO VLA Sky Survey, with total (top)
and polarised intensity (bottom). . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.5 Zeeman measurements in DR 21 (CN, left) and in S 106 (OH line, right. . . . 76
4.6 Illustration of flux freezing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.7 Examples of magnetic-field structures observed in molecular clouds. Left:
magnetic-field strength in S 106, obtained from OH observations at λ = 18 cm
conducted with the VLA. The contours depict the continuum emission. Right:
structure of the magnetic field in DR 21, inferred from polarisation measure-
ments of the dust emission at λ = 1.3 mm and in the 12 CO(1-0) line with
BIMA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.8 Radio continuum image of the SNR Cas A obtained with the VLA at 5 GHz. 83
4.9 VLBI observations of the supernova SN1993j, showing its expansion (upper
panel). As time elapses, the supernova emits synchrotron radiation at higher
frequencies (lower panel). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.10 Illustration of the different evolutionary phases of SNR. . . . . . . . . . . . . 85
4.11 Illustration of particle trapping in a ‘magnetic bottle’. . . . . . . . . . . . . . 86
4.12 Sketch of the geometry in Fermi-II acceleration, with head-on (a) and following
(b) collision. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.13 UHECRs recorded with the Auger array. . . . . . . . . . . . . . . . . . . . . . 90
5.1 Illustration of axi- and bisymmetric magnetic-field configurations. . . . . . . . 92

5.2 Possible modes of halo fields. . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.3 Pitch angle of the spiral structure. . . . . . . . . . . . . . . . . . . . . . . . . 93
5.4 Magnetic field in M 51. Contours depict the radio continuum emission at λ =
~
6 cm. The yellow bars indicate the B-field. . . . . . . . . . . . . . . . . . . . 94
5.5 Magnetic field in the edge-on galaxy NGC 5775 (Soida et al., 2011) at λ = 20 cm
(left) and λ = 6.2 cm (right). Contours outline the total intensity, the bars
denote the orientation of the magnetic field. . . . . . . . . . . . . . . . . . . 95
5.6 Potpourri of spiral galaxies with central AGN and radio-galaxy features. Shown
are: 0421+040 (top left), NGC 4258 (top right), 0313-192 (middle left), NGC 1068
(middle right), NGC 3367 (bottom left), 0400-181 (bottom right). . . . . . . . 96
5.7 Magnetic fields in the dwarf galaxies NGC 4449 (top) and NGC 1569 (bottom). 97
5.8 Sketch of the gravitational acceleration as a function of height above the galac-
tic disk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.9 Sketch of a stable and an unstable galactic disk. . . . . . . . . . . . . . . . . 102
5.10 The M 81/82 galaxy group mapped at λ = 90 cm with the WSRT. The extended
synchrotron halo of the starburst galaxy M 82 in the north-west is obvious. . 103
5.11 Collection of galaxy rotation curves. . . . . . . . . . . . . . . . . . . . . . . . 105
5.12 Illustration of the α-ω dynamo. . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.13 Illustration of the effects of dynamo action producing a bisymmetric magnetic
field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
146 LIST OF FIGURES
5.14 Sketch of the α-ω dynamo working in a galaxy disk. The field lines of the
poloidal component are depicted in the left half of the disk, while for the
toroidal field the field strength is indicated by the contours and grey-scale in
the right half. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.15 Numerical MHD simulations of the CR-driven dynamo. The colour coding
represents the magnetic-field strength in units of µG, with an edge-on (top)
and a face-on (bottom) view. . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.16 ’Spaghetti’ diagramme of the three-dimensional configuration of the magnetic
field in a spiral galaxy from numerical simulations of a dynamo. . . . . . . . . 108
6.1 Radio-FIR correlation of star-forming galaxies and galaxies hosting AGN. . . 109
6.2 Radio image of Cyg A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.3 An example of the variability of the emission from a BL Lac-type object. . . . 112
6.4 Sketch of the central region of a galaxy illustrating the unified scheme. . . . . 113
6.5 Sketch of the main features of a classical FR II radio galaxy. . . . . . . . . . . 114
6.6 Cyg A at 327 MHz (top) and at 15 GHz (bottom). . . . . . . . . . . . . . . . 115
6.7 The magnetic field in Cyg A as deduced from measurements of the linear po-
larisation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.8 Illustration explaining flux conservation in a flux rope made up by longitudinal
and poloidal magnetic fields. . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.9 Sketch of the magnetic field in a flux rope, with a real magnetised jet (NGC 6251)
juxtaposed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.10 Particle ageing as derived for Cyg A, with the break frequencies (top) and the
ages (bottom) shown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.11 The died radio galaxy 0924+30 at 151, 327, 610, 1400, and 4750 MHz. The
total continuum spectrum in the lower right clearly shows the spectral break. 118
6.12 MHD simulations of a jet. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.13 Radio image of the FR-I radio galaxy 3C 31. . . . . . . . . . . . . . . . . . . . 120
6.14 The head-tail radio galaxy NGC 1265 at different frequencies and angular res-
olutions. On the right a radio image of the central Perseus Cluster is shown,
with NGC 1265 in the north. . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.15 Potpourri of so-called X-shaped radio galaxies. . . . . . . . . . . . . . . . . . 122
6.16 Redshift distribution of QSOs. . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.17 The quasar 3C 120 at different wavelengths and angular resolutions. . . . . . 123
6.18 Rotation measures in the central regions of quasars. . . . . . . . . . . . . . . 124
7.1 Illustration of a hadronic collisions in the ICM. . . . . . . . . . . . . . . . . . 129

7.2 The Coma Cluster at X-ray energies (top) and at 90 cm wavelength. . . . . . 130
7.3 Radio continuum images of the galaxy cluster A 2256. . . . . . . . . . . . . . 131
7.4 Radial distribution of RMs in galaxy clusters. . . . . . . . . . . . . . . . . . . 132
7.5 Rotation measure and magnetic field in the cooling-core radio galaxy Hyd A. 133
7.6 Radio relic in the galaxy cluster CIZA J2242.8+5301. . . . . . . . . . . . . . . 133
7.7 Double relic in the galaxy cluster ZwCl 0008.8+5215. . . . . . . . . . . . . . 134
7.8 Three-dimensional MHD simulation of cluster shock waves. . . . . . . . . . . 135
7.9 The radio galaxy Vir A and its halo. . . . . . . . . . . . . . . . . . . . . . . . 136
7.10 Possible inverse-Compton excess emission in the Coma Cluster. . . . . . . . . 137
LIST OF FIGURES 147
7.11 Cavities produced by the radio plasma of 3C 84 (left) and MS 0735.6+7421

(right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
148 LIST OF FIGURES
List of Tables
2.1 Larmor radii and frequencies (B = 10 µG). . . . . . . . . . . . . . . . . . . . 24

2.2 Particle energies and their critical frequencies. . . . . . . . . . . . . . . . . . . 27
2.3 Particle half-lifetimes (B = 10 µG). . . . . . . . . . . . . . . . . . . . . . . . . 40
4.1 Magnetic-field strengths inferred from Zeeman absorption measurements in

Galactic clouds (see also Fig. 4.5). . . . . . . . . . . . . . . . . . . . . . . . . 76
5.1 Gas phases in galaxies, with number densities, temperature, volume and mass
filling factor, and scale height of the disk. . . . . . . . . . . . . . . . . . . . . 91
149
Index
α-effect, 106 Alfvén speed, 82, 99, 101, 140

γ-ray bursts, 86 ambipolar diffusion, 75, 80
γ-rays, 129 Ampere’s law, 76
0313-192, 96 antenna temperature, 19
0421+040, 96 antiferromagnetism, 3
1055+018, 116 Arp 220, 123
1253+275, 133 Auger experiment, 90
1E0657-57, 128
3C 10, 38 bandwidth depolarisation, 59, 66
3C 120, 116 beam depolarisation, 65
3C 236, 111 Bessel function, 28
3C 273, 5, 110, 116, 124 Biermann battery, 104, 138
3C 279, 124 Big Bang, 141
3C 294, 43 BL Lac, 112
3C 295, 132 black hole, supermassive, 113
3C 31, 120, 132 black holes, 126
3C 380, 124 black holes, supermassive, 32, 95, 111, 123
3C 405, 110, 114 black holes, supermassive binary, 43, 121
3C 449, 132 blazars, 112
3C 461, 83 brightness, 18
3C 48, 110 brightness temperature, 18
3C 75, 83 broad-line region, 110, 113
3C 84, 136
Cassiopeia A, 83, 86
A 1656, 128 circular polarisation, 37, 55
A 2163, 128 CL 0016716, 128
A 2218, 128 classical electron radius, 42
A 2256, 128, 131 clusters of galaxies, 127
A 2744, 128 CMB, 41, 42
A 3562, 128 cold neutral medium, 45
absorption coefficient, 16, 29 Coma Cluster of galaxies, 128, 133
accretion disk, 113, 123 cooling flows, 127, 136
active galactic nuclei, 21, 110 Coriolis force, 106
adiabatic compression, 134 cosmic rays, 5, 27, 29, 86
adiabatic expansion, 118 Coulomb force, 11
age of radio source, 40 Cowling’s Theorem, 104
AGN taxonomy, 111 Cowling’s theorem, 105
Airy integral, 28 Crab Nebula, 83
albedo, 48 critical frequency, 26
150
INDEX 151
cutoff frequency, 40 FR II radio sources, 95, 111, 115

cyclotron frequency, 56 free-free radiation, 10
Cygnus A, 4, 110, 114, 131, 132, 138
galactic fountain, 102
dark matter, 102, 121, 127 galactic wind, 94
depolarisation, 59 Gaunt factor, 15
diamagnetism, 3 Gauss’ theorem, 80
dielectric constant, 56 GKZ horizon, 90
diffusion loss equation, 88 GKZ limit, 90
diffusion time scale, 78 Goldreich-Kylafis effect, 70
diffusive shock acceleration, 89, 134
disk scale height, 101, 106 half-power beam width, 61
disk thickness, 101 head-tail radio sources, 120
dispersion measure, 72 HII regions, 10
Doppler boosting, 112 hot gas, 127
Doppler broadening, 69 hotspots, 111, 115
Doppler factor, 112 Hydra A, 58, 132
dust, 47, 70 hydrostatic equilibrium, 100
dwarf galaxies, 97, 139
dynamo, 6 IC 10, 98
dynamo, galactic, 104 IC 342, 93
ICM, 127
Earth, 1, 104 IF polarimeter, 51
electric conductivity, 77 IGM, 127
electric displacement, 56 ILC, 21
electric polarisation, 56 induction equation, 77, 104
electron temperature, 14, 15 intensity, 18
electrons, prelativistic, 9 internal depolarisation, 62
elliptical galaxies, 98 inverse-Compton catastrophe, 114
emission coefficient, 14, 16, 29 inverse-Compton emission, 136
emission measure, 17 inverse-Compton losses, 41, 114
energy density, 6
energy equipartition, 54 jet opening angle, 115
escape speed, 98, 102 jets, 115
extinction, optical, 47
Kelvin-Helmholtz instability, 119
Faraday depth, 60 Kirchhoff’s law, 16
Faraday dispersion function, 60
Faraday rotation, 6, 55 Lagrangian derivative, 79
Faraday screen, 132 Laing-Garrington effect, 132
Faraday’s law, 77 Large Magellanic Cloud, 19
Fermi acceleration, 2nd-order, 87 Larmor circle, 23
ferrimagnetism, 3 Larmor frequency, 3, 23, 68
ferromagnetism, 3 Larmor radius, 2, 24, 89
flux freezing, 68, 75, 79 LHC, 21
flux rope, 116 linear accelerator, 21
Fornax A, 32 linear polarisation, 36
FR I radio sources, 95, 111, 119 LINER, 96, 125
152 INDEX
LMC, 98 NGC 3367, 96

LOFAR, 43, 67, 129, 140 NGC 3516, 96
Lorentz factor, 22 NGC 3667, 94
Lorentz force, 2, 22, 23, 55 NGC 4258, 96
NGC 4449, 98, 140
M 1, 83 NGC 4565, 94
M 31, 93 NGC 4631, 38, 94
M 33, 93 NGC 5775, 94
M 51, 93 NGC 6240, 123
M 81, 93 NGC 6251, 116
M 82, 19, 95, 102 NGC 6822, 98
M 83, 93 NGC 6946, 93
M 51, 50 NGC 891, 94
M 82, 20 NGC 6251, 117
magnet, 1 NGC 6946, 57
magnetic bottle, 87 NVSS, 73
magnetic diffusivity, 78
magnetic field, 58 Ohm’s law, 76
magnetic field, disk, 93 optical depth, 17
magnetic field, dwarf galaxies, 97 Orion Nebula, 10, 19
magnetic field, elliptical galaxies, 98 Orion nebula, 17
magnetic field, Galactic, 73 Orion A, 10
magnetic field, halo, 93 OVV, 112
magnetic field, intergalactic, 116
magnetic field, spiral galaxies, 92 paramagnetic relaxation, 49
magnetic moment, 47 paramagnetism, 3
magnetic reconnection, 106 Parker instability, 102
magnetic susceptibility, 49 Parseval’s theorem, 13
magnetic-field strength, equipartition, 54 particle acceleration, 5
magnetisation, 49 particle ageing, 39
magnetisation of ICM, 139 Perseus Cluster, 120
magnetism, 1 Perseus A, 136, 138
magneto-hydrodynamic equation, 77 phase velocity, 56
MHD shocks, 83 pitch angle, 22, 34
Milky Way, 71 Planck’s law, 16
minimum energy, 54 plasma, 57
molecular clouds, 68, 75 plasma-β, 132, 135
MS 0735.6+7421, 138 Poincaré sphere, 37
Poisson equation, 100
narrow-line region, 110, 113 polarisation, 36
neutron stars, 1 polarisation angle, 37, 57
NGC 1265, 120 polarisation ellipse, 36
NGC 1275, 136 polarisation of starlight, 5
NGC 1569, 98, 140 polarisation, circular, 34, 68
NGC 253, 94, 102 polarisation, complex, 61
NGC 2992, 96 polarisation, elliptical, 34
NGC 3079, 96 polarisation, linear, 34
INDEX 153
polarisation, optical, 47 Sunyaev-Zel’dovich effect, 136

poloidal field, 93 supernova remnants, 38, 83
Pop III stars, 102 supernovae, 83
Poynting flux, 37 Sy 1 galaxies, 110
Poynting vector, 10 Sy 2 galaxies, 110
pressure, CR, 100 synchrotron losses, 39
pressure, gas, 100 synchrotron radiation, 5, 9, 21
pressure, magnetic, 100 synchrotron radiation, Galactic, 32
primary electrons, 129
protons, relativistic, 27 Taurus A, 5, 83
Tevatron, 21
QSO, 111, 123 Thomson cross section, 42
quasars, 111, 123 Thomson scattering, 141
quasars, radio-loud, 123 toroidal field, 93
quasars, radio-quiet, 123 transverse accelerator, 21, 22
Tycho Brahe, 38
radiative transfer, 18
radio core, 111 UHECR, 6, 29, 86, 89
radio galaxies, 111 ultra-luminous infrared galaxies, 123
radio gischt, 134 unified scheme, 113
radio haloes, 128
radio jets, 111 velocity cone, 34
radio lobes, 111, 115 Venus, 116
radio phoenix, 134 Virgo A, 132
radio relics, 133 vorticity, 141
radio-FIR correlation, 109, 125
ram pressure, 120 warm ionised medium, 45
refractive index, 56 warm neutral medium, 45
relativistic beaming, 22, 23 WSRT, 67
relativistic boosting, 22 X-shaped radio galaxies, 121
rotation curves, 105
rotation measure, 57, 131 Zeeman effect, 6, 68
rotation-measure transfer function, 67
secondary electrons, 129

Sedov phase, 85
Seyfert galaxies, 95, 110, 111, 125
shocks, 83
sinc function, 63
SMC, 98
SN-driven dynamo, 106
source function, 16
spectral index, 31
spinning dust grains, 44
spiral galaxies, 92
Stokes parameters, 37, 51
Stokes’ theorem, 80
Sun, 2, 78, 104

Bonn Galactic Intergalactic B Field

Uploaded by

Copyright:

Available Formats

Bonn Galactic Intergalactic B Field

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bonn Galactic Intergalactic B Field

Uploaded by

Copyright:

Available Formats

Galactic and intergalactic magnetic fields

Course astro 848

Argelander-Institut für Astronomie

2 Continuum radiation processes 9

4.2.3 Ambipolar diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

6 Active galactic nuclei 109

7 Intergalactic magnetic fields 127

8 Cosmological magnetic fields 141

(Gaussian system) as it is quite common in astrophysics. In the context of electro-magnetism,

1.2 Magnetic force on a moving charge

the resulting gyration or Larmor frequency is

1.3 Magnetisation of matter

• Diamagnetism: This can be considered as a general property of any substance! When

• Paramagnetism: Substances with unpaired electrons possess a permanent magnetic

• Ferromagnetism: This form of magnetism implies permanent magnetisation, due to

• Antiferromagnetism: In some substances, the electronic spins are antiparallel, which

• Ferrimagnetism: This kind of magnetism is similar to that of antiferromagnetism, but

Figure 1.4: Different kinds of magnetism in solid bodies.

1.4 Some history

1950 The galactic nonthermal radio emission was proposed to be ”magneto-bremsstrahlung”

1.5 Significance of interstellar and intergalactical magnetic

• Some more specific questions related to large-scale phenomena in clusters of galaxies

Continuum radiation processes

2.1 Radiation of an accelerated electron

2.2 Free-free radiation

Due to their hyperbolic orbits about the pro-

2.2.2 Radiation in the non-relativistic

procedure to facilitate this conversion is to perform a Fourier analysis, or to calculate the

The function ẍ(t) looks like a ”spike” as we have seen

We have managed to invert the problem such

Using (2.11) this yields

Recalling that P(ω) ≈ 0 for ν > v/p we have

2.2.3 Total radiation

Let an ensemble of electrons be char-

dN (v, p) = (v · 2πp · dp) · ni ne · f (v) dv cm−3 s−1 (2.29)

Here, f (v) is the Maxwellian velocity distribution

where Te is the electron temperature, with a typ-

This finally leads to an emission coefficient of the

From kinetic gas theory, we know that

Thus, the Gaunt factor for free-free radiation becomes

Iν = Bν (Te ) · (1 − e−τν ) (2.59)

We finally arrive at an important result,

Measuring Iν or Tb over a sufficient frequency

2.3 Synchrotron radiation

2.3.1 Radiation from a single electron

When dealing with relativistic particles, we have to

and the radiation pattern has a strong dependence

As in case of the linear accelerator, the relativistic

For example, consider a particle with energy E =

Note that since the particle is relativistic we have m = m0 γ, since E = m c2 = γ m0 c2

From the above, the Larmor radius is obtained via (v ≈ c)

or, in general (χ 6= 90◦ )

Table 2.1: Larmor radii and frequencies (B = 10 µG).

moves from A to B as shown in (Fig. 2.15), i.e.

|~r − ~rL (t)|

dt0 1 ~r − ~rL (t) d~rL (t)

Table 2.2: Particle energies and their critical frequencies.

Thus, νc depends on the mass of the radiating particle like m−3 :

2.3.2 Synchrotron radiation from relativistic electrons with an energy spec-

Since usually I1 I2 , the total angular momentum J~ =