Astrophysics at Very High Energies - Saas-Fee Advanced Course 40. Swiss Society For Astrophysics and Astronomy (PDFDrive)

Saas-Fee Advanced Course 40
Swiss Society for Astrophysics and Astronomy
Felix Aharonian
Lars Bergström
Charles Dermer
Astrophysics at
Very High Energies
1 5 10 1155 20 25 30 5
2010 September October
For further volumes:

http://www.springer.com/series/4284
Felix Aharonian Lars Bergström
•
Charles Dermer
Astrophysics at Very High

Energies
Swiss Society for Astrophysics and Astronomy

Edited by R. Walter and M. Türler
123
Felix Aharonian Charles Dermer
DIAS Dublin and MPIK Heidelberg Space Science Division
Dublin U.S. Naval Research Laboratory
Ireland Washington
USA
and
Max Planck Institute für Kernphysik Volume Editors
Heidelberg Roland Walter
Germany Marc Türler
ISDC, Geneva Observatory
Lars Bergström University of Geneva, Versoix
Department of Physics Switzerland
Stockholm University
Stockholm
Sweden
This Series is edited on behalf of the Swiss Society for Astrophysics and Astronomy:
Société Suisse d’Astrophysique et d’Astronomie Observatoire de Genève, ch. des Maillettes 51,
CH-1290 Sauverny, Switzerland
Cover Illustration: A composite image of the Crab Nebula showing the X-ray (blue), and optical (red) images
superimposed. The X-ray emitting region is smaller because the higher energy X-ray emitting electrons radiate
away their energy more quickly than the lower energy optically emitting electrons. The inner ring is about one
light year across (Credit: X-ray: NASA/CXC/ASU/J. Hester et al.; Optical: NASA/HST/ASU/J. Hester et al.).
The inset features the gamma-ray variability of the central part of the pulsar wind nebula, as observed by the
Fermi/LAT, indicating that electrons are accelerated up to PeV energies, the highest energy electrons ever
measured (Credit: ISDC/M. Balbo).
ISSN 1861-7980 ISSN 1861-8227 (electronic)

ISBN 978-3-642-36133-3 ISBN 978-3-642-36134-0 (eBook)
DOI 10.1007/978-3-642-36134-0
Springer Heidelberg New York Dordrecht London
Library of Congress Control Number: 2013932553
Ó Springer-Verlag Berlin Heidelberg 2013

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection with
reviews or scholarly analysis or material supplied specifically for the purpose of being entered and executed on
a computer system, for exclusive use by the purchaser of the work. Duplication of this publication or parts
thereof is permitted only under the provisions of the Copyright Law of the Publisher’s location, in its current
version, and permission for use must always be obtained from Springer. Permissions for use may be obtained
through RightsLink at the Copyright Clearance Center. Violations are liable to prosecution under the
respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Preface
The 40th ‘‘Saas-Fee Advanced Course’’ of the Swiss Society for Astrophysics and
Astronomy devoted to Astrophysics at Very-High Energies was held from March
14 to 20, 2010 in Les Diablerets, in the Swiss Alps. It gathered 105 participants
and included a Fermi hands-on tutorial and an INTEGRAL data analysis tutorial.
The course was organized as 28 lectures reviewing the state of knowledge, open
questions, and forecasts in the field of high and very-high energy gamma-ray
astrophysics, a field that has encountered a revolution in the last years with the
success of Cherenkov astronomy and the launch of the Fermi mission. Impact of
gamma-ray observations on our knowledge of particle acceleration in galactic and
extragalactic sources were reviewed as well as the prospects for dark matter
detection and advances from the multi-messenger approaches.
The lectures were given by three world experts in the field:
Prof. Felix A. Aharonian is Professor of Astrophysics at the Dublin Institute for
Advanced Studies (Ireland) and the Head of the High Energy Astrophysics Theory
Group at the Max-Planck-Institut für Kernphysik in Heidelberg (Germany). His main
scientific interests are the processes in thermal and non-thermal relativistic plasmas,
physics and astrophysics of relativistic winds and jets, non-thermal processes in
large-scale AGN Jets and in clusters of galaxies, the origin of galactic and extra-
galactic cosmic rays, the diffuse extragalactic background radiation, observational
gamma-ray cosmology, and the imaging atmospheric Cherenkov array technique.
Prof. Aharonian is involved in a number of major high energy experiments, in
particular as a member of the Science Working Group of ASTRO-H, of the
Collaboration Board of H.E.S.S. and of the Consortium Board of KM3NeT.
Prof. Lars Bergström is Professor at the University of Stockolm (Sweden) and
the Head of the Cosmology, Particle Astrophysics and String Theory Group. He is
also Director of the Oskar Klein Centre for Cosmoparticle Physics. One of his
main lines of research is the investigation of the nature of dark matter, in particular
supersymmetric and Kaluza-Klein particles and the prediction of indirect detection
rates of various dark matter particle candidates. His group is active in many
aspects of observational and theoretical supernova cosmology, gravitational
lensing, determination of cosmological parameters, models for dark matter, and
v
vi Preface
string cosmology. He is collaborating in various experiments to search for evi-

dences for dark matter, in particular Fermi, IceCube and PAMELA.
Dr. Charles D. Dermer is the Head of the Space Radiations Section in the Space
Science Division of the Naval Research Laboratory in Washington, DC. His
interests cover many areas of astrophysics, including cosmic rays, the multi-
wavelength astronomy of blazars, the physics of neutron stars and black holes,
gamma-ray bursts, merging clusters of galaxies, and solar flares. He uses theo-
retical modeling, supported by numerical simulations of the basic physical pro-
cesses involving high-energy interactions between particles and photons in
magnetized plasma, to identify the nature of high-energy astronomical sources and
the physical mechanisms responsible for the observed gamma ray and particle
emissions. Dr. Dermer was one of four GLAST Interdisciplinary scientists, and is
currently a full Fermi Collaboration Member. He has served on numerous review
and study panels, including the Advanced Compton Telescope Working Group and
the VERITAS External Oversight Committee.
This volume of the Saas-Fee lecture notes provides a broad overview of
astrophysics at high and very-energy energies, as well as an introduction to multi-
messenger astronomy and the possible nature of dark matter. Prof. Felix Aharonian
presents the breakthrough in very-high energy gamma-rays achieved by the current
generation of Cherenkov telescopes. He describes the main results and their
implications for theoretical models of the TeV gamma-ray emission with a focus
on Galactic sources. Dr. Charles Dermer follows a similar approach for the other
breakthrough in high-energy observations achieved by the Fermi gamma-ray space
telescope in the spectral window of GeV gamma-rays. In this second part,
emphasis is given on the physics at play in blazars—the most extremely luminous
and variable active galactic nuclei—as constrained by the unprecedented gamma-
ray observations by Fermi. The third contribution to this book is of a quite different
nature. Prof. Lars Bergström gives us a broad overview of multi-messenger
astronomy and the quest of identifying the nature of dark matter both theoretically
and experimentally. He perfectly succeeded in making the challenges of current
astroparticle physics and theoretical cosmology understandable to astronomers.
We are very grateful to the lecturers for their enthusiasm in communicating
their deep knowledge, their brilliant lectures, as well as for writing the rich
manuscripts composing this book. We extend our warmest thanks to the course
secretary, Martine Logossou, for her effective administration of registrations, of
the budget, and her organizational help during the course. We acknowledge the
design of the course poster by Jean-Christophe Leyder. We also would like to
thank all speakers of the INTEGRAL tutorial session held on Thursday afternoon.
In particular, Peter Kretschmar from ESA and our colleagues from the ISDC Data
Centre for Astrophysics: Enrico Bozzo, Carlo Ferrigno, Lucia Pavan, Nicolas
Produit, Claudio Ricci, and Reiner Rohlfs. Last but not least, we thank Elizabeth
Hays and Elizabeth Ferrara from the NASA Fermi Science Support Center and
Preface vii
Andrea Tramacere from the ISDC to offer the course participants the opportunity
of a hands-on session on Fermi data analysis.
One of the highlights of the course was the concert ‘‘Il Viaggio d’Amore’’,
a love journey from the Renaissance to nowadays, offered by Arianna Savall
Figueras and Petter Udland Johansen. Many participants made a memorable walk
with torches from Les Diablerets to the little church of Vers l’Eglise where the
concert took place. It was a magical evening and we would like to thank again the
two outstanding performers for their delighting music.
The Eurotel-Victoria provided—as so often in the past—a pleasant environment
for the Saas-Fee Course and a generous banquet dinner. The organization of this
course would not have been possible without the financial support of the Swiss
Society for Astrophysics and Astronomy (SSAA), the Swiss Institute of Particle
Physics (CHIPP), and the Swiss Academy of Sciences (SCNAT). We are very
grateful to these organizations for their contribution, which allowed the partici-
pants to attend a very diverse, interesting, and successful 40th Saas-Fee Course.
Versoix, October 2012 Roland Walter

Marc Türler
Contents
Gamma Rays at Very High Energies . . . . . . . . . . . . . . . .......... 3

Felix Aharonian
1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1 Status of Observational Gamma Ray Astronomy . . . . . . . . . . 4
1.2 Links to Other Disciplines . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Astrophysical Potential of Ground-Based Detectors . . . . . . . . . . . . . 10
2.1 IACT Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Potential of EAS Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Future IACT Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Radiation Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.1 General Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Brief Overview of Important Processes . . . . . . . . . . . . . . . . . 28
4 SNRs and Origin of Galactic Cosmic Rays . . . . . . . . . . . . . . . . . . . 50
4.1 Gamma-Ray Signatures of SNRs . . . . . . . . . . . . . . . . . . . . . 51
5 TeV Emission of Young SNRs . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.1 RX J1713.7-3946: An Atypical SNR . . . . . . . . . . . . . . . . . . 54
5.2 SN1006, Tycho and Cas A . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3 Radiation Signatures of Proton PeVatrons . . . . . . . . . . . . . . . 60
5.4 Expectations from Future Studies . . . . . . . . . . . . . . . . . . . . . 63
6 Galactic Center . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.1 Sgr A* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.2 Diffuse Gamma-Ray Emission from the Central
10 pc Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7 Pulsars, Pulsar Winds, Pulsar Wind Nebulae . . . . . . . . . . . . . . . . . . 77
7.1 Radiation of Pulsar Magnetospheres . . . . . . . . . . . . . . . . . . . 77
7.2 Radiation of Pulsar Winds . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.3 Pulsar Wind Nebulae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
8 Gamma-Ray Loud Binaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.1 Microquasars: Not yet Proved TeV Emitters . . . . . . . . . . . . . 97
8.2 Binary Pulsars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
ix
x Contents
8.3 Enigmatic TeV Binaries . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Multi-Messenger Astronomy and Dark Matter . . . . . . . . . . . . ...... 125

Lars Bergström
1 Preamble. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
2 The Particle Universe: Introduction . . . . . . . . . . . . . . . . . . . . . . . . 127
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
2.2 Basic Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
2.3 Energy and Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
2.4 Contributions to Vacuum Energy . . . . . . . . . . . . . . . . . . . . . 131
2.5 Summary of Observations . . . . . . . . . . . . . . . . . . . . . . . . . . 134
3 Relic Density of Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3.1 Coannihilations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
3.2 Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
4 Basic Cross Sections for Neutrinos and c-Rays . . . . . . . . . . . . . . . . 148
4.1 Estimates of Cross Sections . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.2 Examples of Cross Section Calculations . . . . . . . . . . . . . . . . 152
4.3 Definition of the Cross Section . . . . . . . . . . . . . . . . . . . . . . 152
4.4 The ccee System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4.5 Processes Involving Hadrons . . . . . . . . . . . . . . . . . . . . . . . . 158
4.6 Neutrinos. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
4.7 Neutrino Interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
4.8 Atmospheric Neutrinos . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
4.9 Neutrinos as Tracers of Particle Acceleration. . . . . . . . . . . . . 166
4.10 AMANDA, IceCube and Direct Detection of WIMPs . . . . . . . 167
4.11 Water and Ice Cherenkov Telescopes . . . . . . . . . . . . . . . . . . 168
5 Supersymmetric Dark Matter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.1 Supersymmetric Dark Matter Particles . . . . . . . . . . . . . . . . . 170
5.2 Higgs and Supersymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.3 The Neutralino Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.4 Experimental Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.5 Supersymmetry Breaking. . . . . . . . . . . . . . . . . . . . . . . . . . . 175
5.6 Other Supersymmetric Candidates . . . . . . . . . . . . . . . . . . . . 176
6 Detection Methods for Neutralino Dark Matter . . . . . . . . . . . . . . . . 176
6.1 Indirect Searches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
6.2 Indirect Detection by c-Rays from the Halo. . . . . . . . . . . . . . 181
6.3 Antimatter Detection of Dark Matter . . . . . . . . . . . . . . . . . . 189
7 Particular Dark Matter Candidates . . . . . . . . . . . . . . . . . . . . . . . . . 193
7.1 WIMP Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
7.2 Dark Stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
7.3 Inelastic Dark Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
7.4 Dynamical Dark Matter. . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Contents xi
7.5 Leptophilic Dark Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

7.6 Supersymmetric Models Beyond the MSSM . . . . . . . . . . . . . 194
7.7 Asymmetric Dark Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
7.8 Kaluza-Klein Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
7.9 Inert Higgs Doublet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
7.10 Non-WIMP Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
7.11 The Axion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
8 Dark Matter Detection: Status . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
9 A Detailed Calculation: The Saas-Fee WIMP . . . . . . . . . . . . . . . . . 200
9.1 The Flux in a Smooth Universe . . . . . . . . . . . . . . . . . . . . . . 202
9.2 Including Effects of Cosmic Structure. . . . . . . . . . . . . . . . . . 204
9.3 The Saas-Fee WIMP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
10 Primordial Black Holes as Dark Matter?. . . . . . . . . . . . . . . . . . . . . 207
10.1 Primordial Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
10.2 Hawking Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
10.3 Thermodynamics of Black Holes . . . . . . . . . . . . . . . . . . . . . 209
10.4 Formation of Primordial Back Holes. . . . . . . . . . . . . . . . . . . 210
11 Gravitational Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
11.1 The Gauge Choice for Electromagnetism . . . . . . . . . . . . . . . 212
11.2 Gauge Choice for the Metric Perturbation . . . . . . . . . . . . . . . 213
11.3 Solutions to the Wave Equation . . . . . . . . . . . . . . . . . . . . . . 213
12 Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Sources of GeV Photons and the Fermi Results . . . . . . . . . . . . . .... 227

Charles D. Dermer
1 GeV Instrumentation and the GeV Sky with the Fermi
Gamma-Ray Space Telescope . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
1.1 Historical Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
1.2 Fermi Gamma-Ray Space Telescope. . . . . . . . . . . . . . . . . . . 234
1.3 Energy, Flux, and Luminosity . . . . . . . . . . . . . . . . . . . . . . . 238
1.4 Limits to the Extreme Universe . . . . . . . . . . . . . . . . . . . . . . 242
2 Fermi Gamma-Ray Source Catalogs and Fermi Pulsars . . . . . . . . . . 242
2.1 First Fermi Catalog of Gamma-Ray Sources: 1FGL . . . . . . . . 243
2.2 Second Fermi Catalog of Gamma-Ray Sources: 2FGL . . . . . . 245
2.3 Fermi Pulsars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
3 Fermi AGN Catalogs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
3.1 LAT Bright AGN Sample (LBAS) and First LAT AGN
Catalog (1LAC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
3.2 Classification of Radio-Emitting AGNs and Unification . . . . . 256
3.3 Properties of Fermi AGNs . . . . . . . . . . . . . . . . . . . . . . . . . . 258
3.4 Second LAT AGN Catalog (2LAC) . . . . . . . . . . . . . . . . . . . 262
4 Relativistic Jet Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
4.1 GeV Spectral Break in LSP Blazars . . . . . . . . . . . . . . . . . . . 264
xii Contents
4.2 Leptonic Jet Models . . . . . . . . . . . . . . . . . . . . . . . . . ..... 268

4.3 Hadronic Jet Models . . . . . . . . . . . . . . . . . . . . . . . . . ..... 272
4.4 Cascade Halos and the Intergalactic Magnetic
Field (IGMF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
5 c Rays from Cosmic Rays in the Galaxy . . . . . . . . . . . . . . . . . . . . 290
5.1 c Rays from Solar System Objects . . . . . . . . . . . . . . . . . . . . 291
5.2 GeV Photons from Cosmic Rays . . . . . . . . . . . . . . . . . . . . . 294
5.3 Fermi Bubbles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
5.4 c-Ray Supernova Remnants . . . . . . . . . . . . . . . . . . . . . . . . . 299
5.5 Nonrelativistic Shock Acceleration of Electrons . . . . . . . . . . . 304
6 c Rays from Star-Forming Galaxies and Clusters of Galaxies,
and the Diffuse Extragalactic c-Ray Background . . . . . . . . . . . . . . . 308
6.1 c Rays from Star-Forming Galaxies . . . . . . . . . . . . . . . . . . . 308
6.2 c Rays from Clusters of Galaxies . . . . . . . . . . . . . . . . . . . . . 310
6.3 Extragalactic c-Ray Background and Populations . . . . . . . . . . 311
7 Microquasars, Radio Galaxies, and the EBL . . . . . . . . . . . . . . . . . . 313
7.1 c-Ray Binaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
7.2 Misaligned Blazars and Radio Galaxies. . . . . . . . . . . . . . . . . 317
7.3 The EBL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
8 Fermi Observations of Gamma Ray Bursts . . . . . . . . . . . . . . . . . . . 322
8.1 Fermi LAT Observations of GRBs . . . . . . . . . . . . . . . . . . . . 322
8.2 GRB Luminosity Function. . . . . . . . . . . . . . . . . . . . . . . . . . 325
8.3 Closure Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
9 Fermi Acceleration, Ultra-High Energy Cosmic Rays,
and Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
9.1 Maximum Particle and Synchrotron Photon Energy . . . . . . . . 337
9.2 L-C Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
9.3 Luminosity Density of Extragalactic c-Ray Jet Sources . . . . . . 339
9.4 Origin of UHECRs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
9.5 Black Holes, Jets, and the Extreme Universe . . . . . . . . . . . . . 341
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
Gamma Rays at Very High Energies
Felix Aharonian
1 Introduction
Cosmic gamma rays carry key information about high energy phenomena in a large
variety of astrophysical environments. Being a part of modern astrophysics and
astroparticle physics, gamma-ray astronomy is a discipline in its own right [52]. It
addresses an impressively broad range of topics related to the non thermal processes
in the Universe, including acceleration, propagation, and radiation of relativistic
particles on all astronomical scales: from compact objects like pulsars (neutron-
stars) and microquasars (accreting stellar mass black holes) to giant jets and lobes
of radio-galaxies and galaxy clusters.
The gamma-ray phenomena generally proceed under extreme physical conditions
in environments characterized with huge gravitational, magnetic and electric fields,
relativistic bulk motions and shock waves, highly excited (turbulent) plasma, etc.
Consequently, any coherent description and interpretation of phenomena related to
gamma-rays requires deep knowledge of many disciplines of physics like nuclear
and particle physics, quantum and classical electrodynamics, special and general
relativity, plasma physics, magnetohydrodynamics, etc.
The energy range covered by gamma ray astronomy spans from 0.1 MeV to
100 EeV (throughout these lectures I will use the energy units which are common in
high energy physics and astrophysics: 1 keV = 103 eV, 1 MeV = 106 eV, 1 GeV =
109 eV, 1 TeV = 1012 eV, 1 PeV = 1015 eV, 1 EeV = 1018 eV). While the lower
bound associates with the region of nuclear gamma-ray lines, the upper bound is
determined by the highest energy particles observed in cosmic rays. Because of the
F. Aharonian (B)
Dublin Institute for Advanced Studies, 31 Fitzwillam Place, Dublin 2, Ireland
e-mail: [email protected]
F. Aharonian
Max Planck Institute für Kernphysik, Saupfercheckweg 31, 69117 Heidelberg, Germany
F. Aharonian et al., Astrophysics at Very High Energies, 3

Saas-Fee Advanced Course 40, DOI: 10.1007/978-3-642-36134-0_1,
© Springer-Verlag Berlin Heidelberg 2013
4 F. Aharonian
essentially different detection methods and approaches applicable to different energy

bands, currently this enormous energy domain of cosmic electromagnetic radiation
is covered inhomogeneously. In particular, so far cosmic gamma-rays are detected in
‘low’ (LE or MeV), ‘high’ (HE or GeV) and very high (VHE or TeV) energy bands.
1.1 Status of Observational Gamma Ray Astronomy

(i) Low Energy Band: 0.1–100 MeV
This energy interval is uniquely linked to several astrophysical phenomena, in
particular to the nucleosynthesis of heavy elements related to the type Ia supernovae
(SNIa), Gamma Ray Bursts, Solar flares, interactions of sub-relativistic cosmic rays
with the interstellar gas and dust, production and annihilation of positrons, etc. While
many aspects of these phenomena can be best probed with low energy gamma-rays,
the MeV gamma-ray sky remains an almost unexplored frontier. The main challenge
of low energy gamma-ray astronomy is the design and construction of detectors
with a sensitivity compatible to the conservative flux predictions. Unfortunately, the
combination of several principal factors—the low detection efficiency, the modest
angular resolution and the high level of backgrounds of different origin—severely
limit the potential of detectors operating in this energy region.
The minimum detectable energy fluxes, even after significant improvements
as foreseen for the next generation of low-energy gamma-ray detectors, will still
remain modest, hardly better than 10−12 erg/cm2 s. Even so, low-energy gamma-
rays are messengers of crucial astronomical information that cannot be obtained by
other means. This concerns, for example the probes of the flux of sub-relativistic
(E ≤ 100 MeV) cosmic rays in the Interstellar Medium (ISM) through the prompt
de-excitation gamma-ray lines (see e.g. Ref. [178]). On the other hand, gamma-ray
continuum at MeV energies produced via bremsstrahlung of electrons (positrons), as
well as at the positron annihilation in flight (in sources with high positrons-to-electron
ratio) contains unique information about relativistic electrons below 100 MeV [58]. In
the environments with low magnetic field, in particular in the ISM (B ≤ 10 µG), such
an information is not accessible via synchrotron radio emission because it appears
at non-visible frequencies below 1 MHz. The information about low-energy elec-
trons and protons (nuclei) is important for the understanding of the energy balance
between different forms of matter, magnetic fields and cosmic rays. In the galactic
disk, such measurements have some other astrophysical implications, for example
they provide direct estimates of the ionization and heating rates of the interstellar
gas by low-energy cosmic-ray protons and electrons.
Another important implication of MeV gamma-ray emission is related to stud-
ies of mildly relativistic thermal plasmas formed in the vicinity of compact rela-
tivistic objects like neutron stars and black holes. The detection of characteristic
MeV radiation dominated by the Comptonized free-free (bremsstrahlung) emission
and electron–positron annihilation, gives direct information on electrons in such
extreme thermal plasmas (see e.g. [223]). However, in most cases the fast radiative
Gamma Rays at Very High Energies 5
cooling of electrons prevents their heating to temperatures beyond 109 K, thus the
electron cooling proceeds through radiation in hard X-rays. Consequently,the pro-
ton temperature can significantly exceed the electron temperature. The formation
of hot two-temperature plasmas, Ti Te , in strong shock waves or in accretion
flows close to black holes, can be studied by detection of characteristic gamma-ray
line emission produced through the chain of spallation and excitation reactions. A
clear signature of radiation at the final stage of such hot plasmas when all nuclei are
destroyed and the nucleonic component of plasma basically consists of protons and
neutrons (with a small fraction of deuterium in equilibrium), is the continuum due to
the proton-neutron bremsstrahlung and the broadened and “blue-shifted” deuterium
line [69]. In accreting solar-mass black holes, this radiation is released, depending
on Ti , typically between 1 and 30 MeV.
Finally, one of the major objectives of MeV gamma-ray astronomy remains the
exploration of 0.511 MeV line emission due to annihilation of the positrons which
are copiously produced in various astrophysical environments (see e.g. the recent
review [210] on the annihilation line from the Galactic Center).
(ii) High Energy Band: 0.1–100 GeV
Before the launch of the Fermi Gamma-ray Space Telescope (formerly GLAST) in
May 2008, the high energy space-based gamma-ray astronomy has been dominated
by the results obtained with the EGRET telescope aboard Compton Gamma Ray
Observatory. Because of the rather modest angular resolution of EGRET (of order of
a few degrees), only two source populations—the active galactic nuclei and pulsars—
have been clearly identified as high energy gamma-ray emitters. With the Fermi LAT
(Large Area Telescope) the HE gamma-ray astronomy entered a new era. This instru-
ment with significantly improved (compared to EGRET) angular resolution (0.6◦ at
1 GeV and better than 0.15◦ at energies above 10 GeV) and flux sensitivity (better
than 10−12 erg/cm2 s) [86], is a perfectly designed tool for deep gamma-ray surveys
with an effective field of view of order of 2 steradian. Over the last three years, Fermi
LAT has been releasing vast amount of important astronomical information. The high
energy gamma-ray sky revealed by Fermi (see Fig. 1) is really very impressive! These
results confirm, to a large extent, the optimistic pre-launch expectations concerning,
in particular, the dramatic increase of the number of gamma-ray emitting pulsars
and AGN, discovery of new classes of compact/variable and extended galactic and
extragalactic gamma-ray sources, the detection of multi-GeV components of GRBs,
etc. The second Fermi LAT gamma-ray source catalogue [200], based on the first two
years of observations, consists of almost 2000 galactic and extragalactic gamma-ray
emitters. While more than half of these objects are associated with counterparts rep-
resenting known source populations (more than one hundred sources being firmly
identified), the origin of approximately 1/3 of these objects remains an open issue.
This concerns, first of all, the extended sources located in the galactic plane, e.g.
SNRs and PWNe, for which the chance of confusion with the diffuse emission of
the galactic disk is especially high. Because of the limited angular resolution, the
most reliable approach for identification of GeV gamma-ray sources is the analysis
based on temporal studies. In this regard it is quite natural that the best “astronomical
6 F. Aharonian
Fig. 1 The overall MeV/GeV gamma-ray sky (the blue to red color background) as seen by Fermi
[200] and the positions of discrete TeV gamma-ray sources detected with ground-based instruments;
the regularly updated version of this figure can be found on the “TeVCat” webpage: http://tevcat.
uchicago.edu. The SNRs (in association with GMCs or without) are shown with a green symbol,
the pulsar wind nebulae are shown in violet, the binary systems—in yellow, starburst galaxies—in
brown, AGN (all types)—in red. The so-called “dark” sources without any reliable association with
the well known objects are shown with a grey symbol
clocks”—the Pulsars—constitute the largest population of identified galactic GeV

gamma-ray sources. The periodic character of gamma-ray emission of the galactic
binary systems or the sporadic flares of AGN provide another tool for identification of
variable gamma-ray sources based on simultaneous observations in different energy
bands. In general, the multi-wavelength observations is a key component for identi-
fication of gamma-ray emitters, as well as for deeper understanding of the nature of
these objects.
The Fermi observations significantly enhance our knowledge about the diffuse
gamma-ray backgrounds of different origin. In particular, Fermi has extended the
range of observations of the diffuse emission of the Galactic Disk and the isotropic
(extragalactic) gamma-ray background to several hundreds of GeV and helped to
clarify some controversial issues related to the contributions of different source pop-
ulations and the relevant radiation mechanism.
A number of important results, especially at energies below 1 GeV, have been
recently reported also by the Italian gamma-ray satellite AGILE [207]. In general,
the observations by Fermi and AGILE support many phenomenological concepts
and theoretical models in different areas of astrophysics. At the same time, these
observations resulted in a number of “unplanned” discoveries and revealed some
puzzling phenomena like flares of the Crab Nebula or existence of multi-kpc scale
non-thermal structures—giant reservoirs of relativistic particles centered on the core
of the Galaxy (“Fermi bubbles”).
Concerning the next generation space-based gamma-ray detectors, it is likely that
in the foreseeable future one cannot expect significant developments beyond the level
achieved by Fermi LAT, except perhaps for the energy band below 1 GeV down to
several tens of MeV. The design and construction of a space-based instrument in
this energy interval with an affordable effective detection area of order of 1 m2 and
angular resolution of about 1◦ , would increase significantly the detection rate (photon
statistics) and improve the flux sensitivity of Fermi LAT below 1 GeV potentially by
an order of magnitude. The optimization of the pair-conversion tracking detection
technique with a focus on energies around 100 MeV would be an attractive and
promising strategy given the number of outstanding astrophysical questions (relevant
to nearly all source populations) not fully addressed by Fermi LAT.
(iii) Very High Energy Band: 0.1–100 TeV
One of the most remarkable achievements of recent years in astrophysics was the
sudden emergence of very-high-energy gamma-ray astronomy as a truly astronom-
ical discipline. The observations conducted by HESS, MAGIC, VERITAS (see
Figs. 4, 5, 6) and MILAGRO groups resulted in the discovery of many sources with
a number in excess of 130 (see Fig. 1). These sources represent almost all major
non-thermal astrophysical source populations, including shell type Supernova Rem-
nants, Pulsar Wind Nebulae, Star Forming Regions, Giant Molecular Clouds, X-ray
Binary Systems, Blazars, Radio-galaxies, Starburst Galaxies (for a review see e.g.
[27, 154]). In general, this success was a big surprise, especially given the rather
difficult past and the controversial history of the field over the last four decades (see
e.g. [52, 248]). In this regard, a question naturally arises concerning the reasons
which made possible this success. A likely answer to this question perhaps can be
formulated as a fortunate combination of two independent factors:
(a) the practical realization of the great potential of stereoscopic arrays of Imaging
Atmospheric Cherenkov Telescopes (IACTs) as effective multifunctional tools
for spectral, temporal, and morphological studies of VHE gamma-ray sources;
(b) the existence of a large variety of perfectly designed machines—TeVatrons, PeVa-
trons, and EeVatrons—“factories” of relativistic matter where the effective par-
ticle acceleration is accompanied by effective radiation processes.
The discussion of the nature of VHE gamma-ray source populations constitutes
a major purpose of this chapter.
1.2 Links to Other Disciplines
Gamma-ray astronomy has deep intrinsic links to other astronomical disciplines and
cosmology. For example, relativistic electrons responsible for radio emission of the
Galactic Disk, produce, at interactions with ambient gas and radiation fields, diffuse
gamma-ray emission in the MeV and GeV energy bands. At higher (TeV) energies
of electrons, the synchrotron radiation appears in the X-ray energy band, while the
inverse Compton (IC) scattering of same electrons results in high energy gamma-
rays in the TeV energy domain. The link between synchrotron X-rays and very high
energy IC gamma-rays is common for source population of quite different origin, in
particular for Supernova Remnants, Pulsar Wind Nebulae, X-ray binaries, Blazars,
8 F. Aharonian
etc. Another example is the interactions of cosmic ray (CR) protons with matter and
radiation that produce π 0 mesons, the decay of which results in the so-called hadronic
component of gamma-rays. Neutrinos from decays of π ± mesons produced in same
interactions, constitute the basis of high energy neutrino astronomy. The common
origin of high energy neutrinos and hadronic gamma-rays as secondary products
of interactions of cosmic rays is an indication of the important role of gamma-ray
astronomy for realization of the so-called multi-messenger approach in the solution
of the problem of origin of galactic and extragalactic cosmic rays—one of the key
objectives of the new interdisciplinary research area called Astroparticle Physics.
The topics related to gamma-ray astronomy cannot be reduced to merely non-
thermal phenomena. For example diffuse galactic and extragalactic infrared/optical
radiation components have thermal origin and formally do not associate with high
energy processes. On the other hand, they play an important role in production
and absorption of high energy gamma-rays through the inverse Compton scattering
and photon–photon pair production processes. This determines the deep links of
gamma-ray astronomy to infrared and optical astronomies, as well as to cosmology.
For example, the characteristic absorption features in the spectra of high energy
gamma-rays arriving from distant extragalactic objects caused by interactions with
the Extragalactic Background Light (EBL) contain unique cosmological information
about the epochs of formation of galaxies and their evolution in the past. The indirect
search for non-baryonic Dark Matter through high energy gamma-ray, as well as
detection of gamma-radiation related to large scale cosmological structures (clusters
of galaxies) are two other important (not yet realized) issues at the interface of
gamma-ray astronomy and cosmology.
The above examples demonstrate the distinct feature of high energy gamma-ray
astronomy as a multi-disciplinary research area and a key component of astroparticle
physics in the context of multi-wavelength and multi-messenger approach in the
studies of the most energetic processes in the Universe.
The links of the TeV domain to its closest neighbour—the GeV domain—are
expected to be especially tight. It is generally believed that the TeV and GeV gamma-
ray fluxes should strongly correlate, therefore the TeV gamma-ray sources should be
also prominent GeV sources. However the GeV–TeV links are not so straightforward
as it may look at first glance.
On the GeV–TeV Links
Over the last several years the number of VHE sources have been dramatically
increased; presently (first half of 2012) it exceeds 130. At first glance, it looks a
modest achievement compared to the almost 2000 sources detected by Fermi LAT.
However, when one takes into account the limited observation time (the duty factor
of observations with Cherenkov telescopes does not exceed 10 %) and the small
field of view (less than 0.01 steradian) of IACT arrays, one may conclude that there
should be a plenty of VHE gamma-ray emitters to be discovered by next-generation
instruments.
It is expected that the next generation major ground based detector, the Cherenkov
Telescope Array(CTA), with an order-of-magnitude improved sensitivity (see Sect. 2),
will increase dramatically, by one or perhaps even two orders of magnitude, the num-
ber of VHE gamma-ray sources. However, the predictions on the number of VHE
sources based on extrapolations from observations at other wavelengths, should be
taken with a caution. Indeed, as long as we deal with poorly understood phenom-
ena in a new energy band, often seemingly reasonable extrapolations could appear
wrong. Moreover, since in sources of different power the acceleration, radiation and
absorption processes can proceed with different efficiencies, the predictions based
on the so-called ‘LogN–LogS’ relations, could lead to misleading conclusions even
in the case of well established source populations. This rather general statement
concerns all wavelengths of non thermal electromagnetic radiation, including the
two closest neighbours—the GeV and TeV bands. For example, the increase of the
overall luminosity of a compact object could reduce the maximum achievable energy
of electrons (because of the enhanced inverse Compton cooling), and, at the same
time, increase the probability for absorption of VHE gamma-rays inside the source.
This kind of non-linear effects should have a strong impact on gamma-ray fluxes,
but act in different energy bands quite differently. Consequently one should expect
essentially different ‘LogN–LogS’ relations applied to GeV and TeV energies.
For copious gamma-ray production two conditions are required—an effective
particle accelerator and a surrounding dense target in the form of gas, radiation
or magnetic field. In the case of energy-dependent escape of particles (protons or
electrons) from the accelerator, the resulting spectrum of particles inside the ‘old’
accelerator can be significantly softer than the spectrum of escaped particles outside
the accelerator. Correspondingly, the gamma-ray spectrum from the accelerator will
be softer compared to the energy spectrum of gamma-ray produced outside the accel-
erator. Since the diffusion of low energy protons is significantly slower, the impact
of the escape is less critical for GeV gamma-rays. Thus an observer detecting GeV,
but not TeV gamma-ray, from an accelerator at late epochs of its evolution, might
conclude that we see an active source which however does not accelerate particles
beyond TeV energies. But in reality, the accelerator could be a dead TeVatron or
PeVatron. The ambiguity can be resolved through comparison of gamma-ray spectra
detected from inside and outside the accelerator.
Because of propagation effects, the hadronic gamma-ray sources are expected to
be more extended at TeV than at GeV energies. For example, the gamma-ray sources
can be due to protons which left the accelerator and interact with nearby dense
molecular clouds. In such cases gamma-rays can be produced predominantly from
the clouds but not from the accelerator itself; the latter could be not active anymore,
or the gas density inside the source cannot be sufficient for production of detectable
gamma-ray fluxes. This might be a quite natural explanation of the so-called ‘dark
accelerators’—VHE gamma-ray sources from regions without counterparts observed
at other wavelengths. For electrons, which suffer significant radiative (synchrotron
and inverse Compton) losses on time-scales shorter than the escape time, just an
opposite picture is expected—strong VHE gamma-ray emission of IC origin from a
10 F. Aharonian
compact region which coincides with the accelerator, and more extended emission
at lower energies from run-away electrons.
Finally, the energy dependent absorption of gamma-rays due to photon–photon
collisions inside the compact objects can significantly change the original (produc-
tion) spectrum of gamma-rays. The impact of absorption can be different for different
energy intervals of gamma-rays depending on the spectral energy distribution (SED)
of the surrounding target photon gas. For example, in compact X-ray regions, e.g. in
accretion flows close to black holes, the absorption effect is strongest at GeV ener-
gies, while in binary systems containing luminous optical stars, TeV gamma-rays
suffer most severe absorption.
In summary, depending on specifics of acceleration and propagation of charged
particles (protons and electrons), as well as depending on the combination of gamma-
ray production/absorption mechanisms, one can expect quite different relations
between the GeV and TeV fluxes. They can correlate, anti-correlate, or behave
in a random fashion. Thus, the energy spectra of GeV gamma-ray sources should
not necessarily extend to TeV energies, while the TeV gamma-ray sources could
appear without a GeV counterpart. Therefore the GeV and TeV gamma-ray sources
can be represented quite differently in the given source population. For example,
while the Fermi observations have proved that a large fraction of pulsars are prolific
MeV/GeV gamma-ray emitters, the ground-based observations show that Pulsar
Wind Nebulae radiate most effectively in the TeV gamma-ray band. On the other
hand it is remarkable that both GeV and TeV gamma-rays are detected from sources
representing almost all ‘suspected’ non thermal source populations of galactic and
extragalactic origin, in particular from Supernova Remnants (SNRs), Pulsar Wind
Nebulae (PWNe), Giant MolecularClouds (GMCs), Compact Binaries (CBs), Star-
burst Galaxies, Radio Galaxies and Blazars. High energy gamma-rays are detected
also from the Sun and Moon, as well as from Gamma Ray Bursts.
Figure 2 shows the populations of sources established as GeV and TeV gamma-ray
emitters, and demonstrates the complex links related to several major research areas
of astrophysics and astroparticle physics: Origin of Galactic and Extragalactic Cos-
mic Rays, Physics and Astrophysics of Compact Objects (Black Holes and Neutron
stars), Relativistic Outflows (AGN jets and Pulsar Winds), Cosmology (Dark Matter,
Extragalactic radiation and magnetic fields), etc.
2 Astrophysical Potential of Ground-Based Detectors
Earth’s atmosphere is not transparent to gamma-rays of any energy. Therefore, their

registration requires detectors installed on space platforms. However, the satellites
cannot offer, at least in the foreseeable future, detection areas significantly exceeding
1 m2 ; this constrains the effective studies of tiny fluxes of cosmic gamma-rays to
energies ≤100 GeV. Fortunately, at higher energies an alternative method can be used
for detection of gamma-rays. The method is based on the registration of atmospheric
Fig. 2 GeV and TeV source populations and the links between the major scientific topics
showers (initiated by interactions of gamma-ray) either directly or through their

Cherenkov radiation.
The faint and brief Cherenkov signal which lasts only several nanoseconds can
be detected by large optical reflectors equipped with fast multi pixel cameras. With a
telescope consisting of an optical reflector of diameter D ≈ 10 m, and a multichannel
camera with pixel size 0.1◦ –0.2◦ and field-of-view Θ ≥ 3◦ , primary gamma-rays of
energy ≥100 GeV can be collected from distances as large as 100 m. This provides
huge detection areas, A ≥ 3 × 104 m2 , which largely compensate the weak gamma-
ray fluxes at these energies. The total number of photons in the registered Cherenkov
light image is a measure of energy, the orientation of the image correlates with the
arrival direction of the gamma-ray, and the shape of the image contains information
about the origin of the primary particle (a proton or photon). The basic principles of
operation of the IACT technique is illustrated in Fig. 3.
The stereoscopic observations of air showers with two or more 10 m diameter
telescopes located at distances of about 100 m from each other provide a quite
low energy threshold around 100 GeV, effective (by a factor of 100) rejection of
hadronic showers, and good angular (≈ 0.1◦ ) and energy (≈15 %) resolutions (see
e.g. [53]). At energies around 1 TeV, this results in a minimum detectable energy flux
of 10−13 erg/cm2 (see Fig. 7), a quite impressive sensitivity even in the standards
of advanced branches of astrophysics. In particular, it is much better than in any
other gamma-ray domain, including the GeV energy band, where the sensitivity of
Fermi LAT, even after dramatic improvement compared to the performance of the
previous gamma-ray space-borne instruments, still cannot compete with the perfor-
mance already achieved in the TeV energy band. Thanks to very large collection
area, the IACT technique provides large gamma-ray photon statistics even from
12 F. Aharonian
Fig. 3 The operation of the imaging atmospheric Cherenkov telescope technique (from Ref. [154])
Fig. 4 The HESS system of four 13 m diameter Cherenkov telescopes. While this system has been
operating in Namibia since 2003, the new 28 m diameter telescope is under construction (the picture
in the center of the original HESS array is a photo montage)
relatively modest TeV gamma-ray emitters. Coupled with good energy and angu-
lar resolutions, the rich photon statistics allows deep morphological, spectral, and
temporal studies. This makes the IACT arrays perfect multifunctional and multi-
purpose astronomical tools for exploration of a broad range of non-thermal objects
and phenomena, both of galactic and extragalactic origin. Currently, three major
IACT arrays (see Figs. 4, 5, 6) HESS (High Energy Stereoscopic System), MAGIC
(Major Atmospheric Imaging Cherenkov) and VERITAS (Very Energetic Radiation
Imaging Telescope Array System)—located both in the northern (MAGIC, VERI-
TAS) and southern (HESS) hemispheres, determine the status of VHE gamma-ray
Fig. 5 The system of two 17 m diameter MAGIC gamma-ray telescopes at the Roque de los
Muchachos observatory (La Palma, Canary Islands)
Fig. 6 The VERITAS system of four 12 m diameter Cherenkov telescopes located in southern
Arizona
astronomy. What concerns the previous generation instruments, one should mention,
amongst others, the 10 m diameter single dish of the Whipple Observatory (south
Arizona) and the HEGRA array of five relatively modest (4 m diameter) Cherenkov
telescopes (La Palma, Canary Islands). These instruments, which can be considered
as prototypes of the current IACT arrays, played a crucial role in the development
of ground-based gamma-ray astronomy. While the Whipple collaboration pioneered
the implementation and successful realization of the imaging atmospheric Cherenkov
technique, the HEGRA collaboration convincingly demonstrated the power of the
stereoscope approach. In this regard, it is not a big surprise that the performance of
the current (and the next) generation IACT arrays are not far from the early predic-
tions based on the extrapolation of the performance of the single Whipple dish and
the HEGRA telescope array (see e.g. [52]).
14 F. Aharonian
Fig. 7 The energy-flux

sensitivities of the current
and future ground-based
detectors—the IACT and EAS
arrays in the energy range
1010 to 1016 eV (courtesy of
Gus Sinnis)
2.1 IACT Arrays
Figures 7, 8, 9, and 10 represent three characteristic examples demonstrating the

great performance of the stereoscopic IACT technique for morphological, spectral
and temporal studies, respectively.
(i) Figure 8 shows the VHE gamma-ray map of an extraordinary site—the central
several hundred parsec region of our Galaxy which harbours a variety of poten-
tial gamma-ray emitters. This region has been predicted as a possible gamma-ray
source also because of predicted sharp concentration of Dark Matter. Deep obser-
vations of HESS did reveal that this compact region (angular size less than 2◦ ) is
packed with several gamma-ray sources, including a point like source in the very
center of the Galactic Center, diffuse emission contributed by giant molecular
clouds, a composite supernova remnant, as well as an interesting but not yet
identified extended source.
(ii) Figure 9 demonstrates the power of the IACT technique for spectroscopic stud-
ies. The energy spectra of two active galactic nuclei, Mkn 421 and Mkn 501,
have been measured by the HEGRA IACT array. The spectra based on very
large gamma-ray statistics (60,000 and 40,000 from Mkn 421 and Mkn 501,
respectively), detected during high states of these objects, can be fitted with the
canonical “power-law with exponential cut-off” function. Note that the mea-
sured spectral points extend beyond 3E 0 ; this is a remarkable result even for the
standards of laboratory experiments.
(iii) High detection rates of gamma-rays by IACT arrays are possible even for
relatively modest energy gamma-ray flux of about 10−11 erg/cm2 s. This makes
these instruments powerful tools for temporal studies of highly variable VHE
sources. This is demonstrated in Fig. 10 for a major flare of the blazer
Fig. 8 The image of the several-hundred parsec region of the Galactic Center in TeV gamma-
rays. It contains a point like source (angular radius less than a few arcminutes) the gravity center
of which coincides with an accuracy of 13 arcseconds with the compact radio source Sgr A*—a
super massive black hole at the dynamical center of the Milky Way [21, 32]. The second point
like source located about one degree away positionally coincides with the composite supernova
remnants G09+0.1 [37]. A prominent feature of this region is the ridge of diffuse emission tracing
several well identified giant molecular clouds [39]. This complex region contains some other, not
yet firmly identified, “hot spots”
PKS 2155-304 on the night of July 29–30, 2006. The outburst was so powerful
that the detection rate of VHE gamma-rays by the HESS telescopes “jumped” a
level of several (background-free) events per second leading to the discovery of
variability of the source on an exceptionally short time-scale of about 2–3 arcmin.
2.2 Potential of EAS Arrays
The IACT arrays are designed for observations of point-like or moderately extended
(with angular size 1◦ or less) objects with known celestial coordinates. However, the
high sensitivity and relatively large (≥4◦ ) field of view of IACT arrays allow effective
all-sky surveys as demonstrated by the HESS collaboration. On the other hand, the
potential of IACT arrays is limited for the search of very extended structures (like
16 F. Aharonian
HEGRA CT System
-10
10
E dN/dE TeV/(cm2 s)
-11
10
-12
10 Mkn 501 (1997)
2
Mkn 421 (2000/2001)

-13
10
1
Ratio
-1
10
Ratio Mkn 421/Mkn 501
1 10
E/TeV
Fig. 9 The energy spectra of active galactic nuclei Mkn 501 and Mkn 421 measured by the HEGRA
stereoscopic system of Cherenkov telescopes in the high states (from Ref. [30]). The spectra are
well described by “power-law with exponential cut-off”, E −Γ exp (−E/E 0 ), with Γ = 1.92 and
E 0 = 6.2 TeV for Mkn 501, and Γ = 2.19 and E 0 = 3.6 TeV for Mkn 421. To demonstrate the
difference in two energy spectra and to reduce the impact of possible systematic effects, the ratio
of the Mkn 421 and Mkn 501 spectra is shown in the lower panel
Fig. 10 The light curve of the exceptional outburst of the blazer PKS 2155-304 on the night of July
29–30 2006. More than 10,000 gamma-rays have been detected during 90 min leading to extraction
of sharp flares on minute time-scales
diffuse emission of the galactic disk), as well as for the solitary events like GRBs.
In this regard, the detection technique based on direct registration of particles that
comprise the extensive air showers (EAS), is a complementary approach to the IACT
technique.
The traditional EAS technique, based on scintillators or water Cherenkov detectors
spread over large areas, works quite effectively for detection of cosmic rays at
ultra-high energies, E ≥ 100 TeV. In order to make this technique more adequate to
purposes of gamma-ray astronomy, the detection energy threshold should be reduced
by two orders of magnitude. This can be achieved using dense particle arrays located
on very high altitudes. The feasibility of both approaches recently have been success-
fully demonstrated by the ARGO and Milagro groups. The significance map of the
galactic plane region l ∈ [30◦ , 220◦ ] and b ∈ [−10◦ , 10◦ ] obtained with the Milagro
detector [13] is shown in Fig. 11. Eight candidate sources at a median energy of ∼20
TeV have been found with pre-trial significance ≥4.5σ. After accounting for the
trials over ≈ 400 square degree region, four of these candidates survived as reliable
detections with an after-trial statistical significance exceeding 4σ.
These results, as well as the prospects of continuous monitoring of a significant
part of the sky, which might lead to exciting discoveries of yet unknown VHE tran-
sient phenomena in the Universe, justifies the new proposals of high altitude EAS
detectors (see for a review [27]) like HAWK, a High Altitude Water Cherenkov
Experiment under construction on a site close to Sierra Negra, Mexico. The 5 year
survey sensitivity of HAWK at energies between 1 and 10 TeV is expected to be
comparable to the sensitivity of Fermi around 1 GeV. In this regard HAWK will be
complementary to Fermi for continuous monitoring of more than 1 steradian fraction
of the sky at TeV energies. At higher energies, one should mention the ambitious
LHAASO (Large High Altitude Air Shower Observatory) detector facility at Yang-
bajing, Tibet. This array consisting of several types of detectors of electromagnetic
and muon components of air showers will cover huge area and achieve an impressive
sensitivity at energies of several tens of TeV.
2.3 Future IACT Arrays
Planning of the next generation of Imaging Atmospheric Cherenkov Telescope

(IACT) arrays has two major objectives: (i) an order of magnitude improvement
of the flux sensitivity in the standard 0.1–10 TeV energy interval (TeV regime), and
(ii) an aggressive expansion of the energy domain of IACT arrays in both directions—
down to 10 GeV (multi-GeV regime) and well beyond 10 TeV ( sub-PeV regime).
TeV Regime
The best performance the IACT technique is achieved in this energy regime, and
still the potential is not saturated. The combination of three basic factors, (i) high
18 F. Aharonian
Fig. 11 Significance map of the Galactic plane at energies above 20 TeV produced on the basis of
MILAGRO data [13]. The color code shows the pre-trials significance in this PSF-smoothed map.
The maximum positive value of the color code saturates at 7σ, although three of the gamma-ray
sources are detected with higher statistical significance
efficiency of detection/identification of electromagnetic showers, (ii) good accuracy

of reconstruction of the direction and energy of primary gamma-ray, and (iii) large
gamma-ray photon statistics, allows reduction of the minimum detectable energy
flux to the level of 10−14 erg/cm2 s, and improvement of the angular resolution to
δθ ≈ 2–3 arcmin.
Such an impressive performance can be achieved by stereoscopic arrays con-
sisting of tens of 10 m diameter class (HESS-type) telescopes. The flux sensitivity
10−14 erg/cm2 s at TeV energies would be a great achievement even in the stan-
dards of the most advanced branches of observational astronomy. This should allow
Fig. 12 Layout of the future Cherenkov telescope array
us to probe the gamma-ray luminosities of potential TeV emitters at the levels of

1032 (d/10 kpc)2 erg/s for galactic sources and 1040 (d/100 Mpc)2 erg/s for extra-
galactic objects. Although for moderately extended sources, e.g. of angular size
Ψ ∼ 1◦ , the minimum detectable energy flux will be by a factor of Ψ/δθ ∼10–30
higher, yet it would be better than the energy flux sensitivities of the best current
X-ray satellites, Chandra, XMM-Newton, INTEGRAL and Suzaku, i.e. should allow
the deepest probes of non thermal high energy phenomena in extended sources,
in particular in shell type Supernova Remnants (SNRs), Giant Molecular Clouds
(GMCs), Pulsar Driven Nebulae (Plerions), Clusters of Galaxies, hypothetical Giant
Pair Halos around AGN, etc. Such a system of 10–12 m diameter class IACTs with
a field of view (FoV) of 6−8◦ , most likely will constitute the core of the Cherenkov
Telescope Array (CTA)—an initiative towards the major ground-based gamma-ray
detector (see Fig. 12). The huge area covered by tens of telescope should provide
dramatic increase of the gamma-ray photon statistics. On the other hand, the detec-
tion of the cascades in many projections will improve the angular resolution, and
the efficiency of suppression of hadronic showers. All these factors should allow an
improvement of the sensitivity of CTA in the TeV region, compared to the current
instruments, by a factor of five (see Fig. 7). Correspondingly, the required observation
time of the objects would be reduced by a factor of 25.
Sub-PeV Regime
The general tendency of decreasing gamma-ray fluxes with energy becomes

especially dramatic above 30 TeV. The reasons could be different, e.g. external
and internal absorption of gamma-ray, limited efficiency of particle acceleration
processes, escape of highest energy particles from the production region, etc. Any
meaningful study of cosmic gamma-rays beyond 30 TeV requires detection areas
20 F. Aharonian
well exceeding 1 km2 . An effective and straightforward approach would be the use
of IACT arrays optimized for detection of gamma-rays in the region up to 100 TeV
and beyond. This can be realized by a modest, approximately 10–30 m2 area reflectors
separated from each other, depending on the scientific objectives and the configura-
tion of the imagers, between 300 and 500 m. The requirement to the pixel size of
imagers is also rather modest, 0.25◦ or so, however they should have large, up to 10◦
FoV for simultaneous detection of showers from distances of several 100 m [209].
A sub-array consisting of several tens of such telescopes is included in the concept
of CTA with a primary goal of study of energy spectra of gamma-ray sources well
beyond 10 TeV. It will serve as a powerful too for searches of galactic cosmic ray
‘PeVatrons’, albeit nearby (R 10 Mpc radio galaxies and starburst galaxies can
be considered high priority targets as well.
Sub-100 GeV Regime
The energy threshold εth of IACTs is generally defined as a characteristic energy

at which the gamma-ray detection rate for a primary power-law spectrum with a
photon index Γ = 2−3 achieves its maximum. It is well known, from Monte Carlo
simulations and from the operation of previous generation IACTs, that in practice the
best performance, in particular the minimum detectable energy flux, is achieved at
energies exceeding several times εth . In this regard, for optimization of gamma-ray
detection around 100 GeV, one should reduce the energy threshold of telescopes to
εth ≤ 30 GeV. This can be done by using very large, 20 m-diameter class reflectors.
On the other hand, the reduction of the detection threshold to 30 GeV is an important
scientific issue in its own right; the intermediate interval between 30 and 300 GeV is a
crucial energy regime for certain class of galactic and extragalactic gamma-ray source
populations. A sub-array consisting of several very large telescopes foreseen in CTA
(see Fig. 12) will indeed significantly broaden the topics and scientific objectives of
CTA.
The above three sub-arrays of Cherenkov telescopes are characterized by specific
energy intervals in which the best energy flux sensitivity is achieved. On the other
hand, each of these arrays covers at least 2 decades in energy. thus the energy domains
of the sub-arrays largely overlap. Since these arrays contain the same basic elements,
and generally have overlaps in scientific motivations, an ideal arrangement would
be the combination of these sub-arrays in a single facility which would have a sen-
sitive and homogeneous coverage throughout the energy region from approximately
30 GeV–300 TeV. The conceptual design of CTA is motivated, to a large extent, by
this objective [23]. The high detection rates, coupled with good angular and energy
resolutions over four energy decades will make CTA a powerful multi-function and
multi-purpose gamma-ray observatory with a great capability for spectrometric, mor-
phological and temporal studies of a diverse range of persistent and transient high
energy phenomena in the Universe.
Multi-GeV Regime: Gamma-Ray Timing Explorers
Despite the recent great achievements of high energy gamma-ray astronomy, there are
obvious shortcomings in the performance of the current so-called “pair-conversion”
tracking detection technique—the most effective approach used in the satellite-borne
instruments for detection of gamma-rays at energies above several tens of MeV. First
of all this concerns the low angular resolution, especially at energies below 1 GeV
and poor photon statistics at energies above 10 GeV. While there is a room for
improvement of the angular resolution at low energies, e.g. reducing it to 1◦ around
100 MeV (by using very thin e± pair-converters above the tracking detectors), the
increase of the gamma-ray photon statistics at high energies is a more difficult task.
One should note in this regard that the flux sensitivity of Fermi LAT at 1 GeV of
about 10−12 erg/cm2 s can be achieved only after one year all-sky survey. While for
persistent gamma-ray sources this seems to be an adequate sensitivity (given that a
huge number of sources are simultaneously monitored within the large, almost ∼2π
steradian homogeneous FoV), the small, ≈1 m2 detection area limits significantly
the potential of this instrument for detailed studies of the temporal and spectral
characteristics of highly variable sources like blazars or solitary events like gamma-
ray bursts (GRBs). The sensitivity achieved by Fermi at high energies will be not easy
to improve significantly by any future space-based mission, unless the Moon would
be used in (far) future as a possible platform for installation of very large (10 m2 )
area pair-conversion tracking detectors. It is clear that the space-based resources of
GeV gamma-ray astronomy have achieved a point where any further progress would
appear extremely difficult and very expensive. In any case, for the next decades
to come there is no space-based mission planned for the exploration of the high-
energy gamma-ray sky. On the other hand, the principal possibility of extension of
the Imaging Atmospheric Cherenkov Telescope (IACT) technique towards 10 GeV
promises a new breakthrough in gamma-ray astronomy [27]. The (relatively) large
gamma-ray fluxes in this energy interval, together with the huge detection areas
offered by the IACT technique, can provide the highest gamma-ray photon statistics
compared to any other energy band of cosmic gamma-radiation. Thus, in the case of
realization of 10 GeV threshold IACT arrays, the presently poorly explored interval
between 10 and 100 GeV could become one of the most advanced domains of gamma-
ray astronomy with a great potential for studies of highly variable phenomena.
The reduction of the energy threshold down to 10 GeV or even less is principally
possible within the basic concept of the IACT technique, but it requires an extreme
approach—25 to 30 m diameter class telescopes equipped with very high (≥40 %)
quantum efficiency focal plane imagers, operating in a robotic regime at very high
(5 km or) mountain altitudes [66].
The energy range from several GeV to 30 GeV has very specific astrophysical and
cosmological objectives: exploration of highly variable non thermal phenomena, in
particular in the remote universe at redshifts of z = 5, as well as in compact galactic
objects like pulsars and microquasars. The successful realization of such a gamma-
ray timing explorer, hopefully during the lifetime of the Fermi observatory would be
a great achievement for gamma-ray astronomy.
22 F. Aharonian
The remarkable success of observational gamma-ray astronomy, together with

recent intensive theoretical and phenomenological studies of acceleration and radi-
ation processes in astrophysical objects, supply a strong rationale for the further
exploration of the sky at high and very-high energies. Although generally the main
motivations of gamma-ray astronomy remain unchanged, the recent observational
results have introduced important corrections to our understanding of many relevant
phenomena. They revealed new features which in many cases require revisions of
current theoretical models and formulations of new concepts. It is expected that over
the next decade the ongoing operation of Fermi will be accompanied by observations
with the next generation ground-based detectors. The data obtained in the enormous
energy range from 100 MeV to 1 PeV will provide very deep insight into a number
of problems of high energy astrophysics and fundamental physics.
3 Radiation Mechanisms
3.1 General Comments
Basic Processes
Any interpretation of an astrophysical observation starts with identification of

relevant radiation mechanism(s). Therefore the comprehensive knowledge on prin-
cipal radiation and absorption processes is a key issue in astrophysics. With some
exceptions, all basic processes of gamma-ray production and absorption have been
studied in great detail using the methods and tools of modern experimental and the-
oretical physics. Each gamma-ray domain has its specifics and is characterized by
several radiation processes. Generally, the most important mechanisms in the high
and very-high energy domains are
• electron bremsstrahlung
• decay of neutral π 0 -mesons produced at p–p and p–A interactions
• inverse Compton scattering
The first two mechanisms operate effectively in (relatively) dense and extended
gaseous environments. While the second mechanism is important in all energy bands
above 100 MeV, the electron bremsstrahlung contributes mostly to sub-TeV gamma-
rays. The reason is the synchrotron and inverse Compton (IC) cooling of electrons
which typically at very high energies starts to dominate over the bremsstrahlung
losses. For the same reason the inverse Compton scattering is a very important process
of gamma-ray production at TeV energies, but it could be a dominant radiation
mechanism also at lower energies. It works effectively almost everywhere, from
compact objects like pulsars and AGN to extended sources like supernova remnants
and clusters of galaxies. Since the Compton cooling time decreases linearly with
energy, the process becomes especially effective at very high energies.
Because of universal presence of dense radiation fields, ultrarelativistic electrons

always release a significant fraction of their energy through IC gamma-ray, except
for regions where the energy density of the magnetic field significantly exceeds
the energy density of the surrounding radiation fields. In such cases, the energy
of electrons is released in synchrotron radiation which typically does not extend
beyond X-ray energies. Therefore, from the point of view of production of high
energy gamma rays, the synchrotron cooling generally plays a negative (dissipative)
role. However, in the so-called extreme accelerators, where the particle acceleration
proceeds at the maximum possible rate allowed by classical electrodynamics [61], the
synchrotron radiation of both electrons and protons can become the dominant channel
of gamma-radiation, especially in highly magnetized environments moving with
relativistic speeds towards the observer [51]. In radiation dominated environments
the synchrotron radiation of secondary charged μ-mesons and π-mesons also may
contribute significantly to very high energy gamma-radiation [198].
In strong magnetic fields, when particles move along the field lines, they radiate
through the so-called curvature radiation (see e.g. [201]). This mechanism is gener-
ally believed to be the most important process responsible for gamma-radiation of
pulsars, but it can effectively work also in other astronomical objects, in particular
in the vicinity of rotating supermassive black holes [61, 181]. Thus, in compact and
highly magnetized objects, high energy gamma-rays can be effectively produced
through
• magnetobremsstrahlung of electrons and protons
under which one should understand the radiation of electrons moving in magnetic
fields in a general sense, i.e. both in the synchrotron and curvature radiation regimes.
In the case of combination of certain favourable conditions regarding, in particular,
the existence of dense surrounding radiation fields and an effective confinement of
relativistic particles, gamma-rays can be effectively produced at interactions of pro-
tons with low-energy photons, directly through decays of π 0 -mesons, the products of
• photomeson reactions: p + γ → π 0 → γ
or indirectly, through synchrotron radiation of secondary electrons, the products of
photomeson reactions and
• Bethe-Heitler pair production: p + γ → e+ e−
Another mechanism of gamma-ray production related to interactions of
accelerated particles with low-frequency photons is (e.g. Ref. [219]) the de-excitation
of nuclei following
• photo–disintegration interactions: A + γ → A ∗ + X ; A∗ → A + γ
However, despite some recent optimistic claims (see e.g. Refs. [77, 199]), the
efficiency of conversion of energy of accelerated nuclei to gamma-rays in optically
thin objects appear very small, η 1. This is a rather generic and robust limit [50]
which makes negligible the role of this mechanism in almost all realistic models of
24 F. Aharonian
cosmic gamma-ray production. In any case, the photo-disintegration reaction rates

are significantly below the rates of photomeson and Bethe-Heitler pair-production
(BHPP) processes.
Generally, because of small cross-sections, the interactions of ultrarelativistic
protons and nuclei with photon fields are rather slow processes. Of course, the rates
can be higher in compact objects at the presence of dense radiation fields. However,
the rates are always limited because of the requirement to the source to be optically
thin for gamma-rays regarding the
• photon–photon pair production: γ + γ → e+ e−
The cross-section of this process exceeds by orders of magnitude the cross-sections
of the photomeson and pair-production cross-sections, therefore the condition
τγγ ≤ 1 implies very low efficiency of photomeson and BHPP processes unless
the protons are very effectively confined in the source. And, vice versa, the high
efficiency of the photomeson and BHPP processes implies large optical depth of
the source. In this case, the products of these reactions, electrons, positrons and
gamma-rays, initiate electromagnetic cascades in the low-energy photon gas, which
wash out the signatures of primary particles, thus the energy spectrum of gamma-
rays that emerge the source is basically determined by the specifics of the cascade
development, rather than the initial spectra of the first generation particles.
Gamma-rays are also absorbed due to pair production at interactions with the gas
and magnetic fields. While the first process generally has a negligible effect in most
of the astrophysical environments, the pair production of very high energy gamma-
rays is a very important process in objects with very strong magnetic field—first of
all in pulsars and perhaps also in the magnetospheres of rotating black holes.
Relativistic Electron–Photon Cascades
The interactions of relativistic electrons, either accelerated directly or being secondary

products of various hadronic processes, with ambient targets in forms of matter,
radiation and magnetic fields result in gamma-rays. Generally, the gamma-ray pro-
duction is effective when the radiative cooling time does not significantly exceed
(i) the source age, (ii) the time of non-radiative losses caused by adiabatic expansion
or by particle escape, and (iii) the cooling time of competing radiation mechanisms
resulting in low-energy photons outside the gamma-ray domain. Note that although
an effective gamma-ray production requires dense targets, if the charged particles
are effectively confined to the gamma-ray production region, these conditions could
be fulfilled even in environments with relatively low gas and photon densities or a
weak magnetic field. For example, the gamma-ray production efficiency could be
close to 1 even in very extended objects, when trad R/c; R is the characteristic
linear size of the production region, c is the speed of light.
The major gamma-ray production mechanisms associated with relativistic elec-
trons have their ‘counterparts’—gamma-ray absorption mechanisms of the same
electromagnetic origin, which result in electron–positron pair production in matter
(the counterpart of bremsstrahlung), in photon gas (the counterpart of inverse Comp-

ton scattering), and in magnetic field (the counterpart of synchrotron radiation). The
gamma-ray production mechanisms and their absorption counterparts have similar
cross-sections, therefore the condition for radiation trad ≥ R/c generally implies
small optical depths for the corresponding gamma-ray absorption process, τabs ≤ 1.
But in many astrophysical scenarios, in particular in compact galactic and extra-
galactic objects with favourable conditions for particle acceleration, the radiation
processes can be very fast, so that trad ≤ R/c. Since the energy threshold of pair pro-
duction implies that the corresponding ‘counterpart’ (production) processes proceed
in relativistic regimes, namely (i) E γ,e 2m e c2 in matter, (ii) E γ,e ω0 m 2e c4 in
photon gas (Klein-Nishina regime), or (iii) (E ,e /m e c2 )(B/Bcrit ) 1 in the magnetic
field (quantum regime), when a large fraction of the electron energy is transferred to
the secondary photon, the problem cannot be reduced to the absorption effect. In this
regime, the secondary electrons produce a new generation of high energy gamma-
ray, these photons again produce electron–positron pairs, etc., so an electromagnetic
cascade develops.
The characteristics of electromagnetic cascades in matter have been
comprehensively studied in literature, in particular the context of interactions of
cosmic rays with the Earth’s atmosphere. These results can be applied to a broad
class of the so-called “hidden sources”, objects of high energy radiation with dense
surrounding gas: super-massive black holes, compact X-ray binaries, young pul-
sars inside dense shells of recent supernovae explosions, etc. (see e.g. Ref. [96]).
If the thickness of the surrounding gas significantly exceeds 100 g/cm2 , the pro-
tons accelerated in the central source would initiate electron–photon cascades While
“hidden sources” are considered as potential neutrino sources, they are less attractive
targets for high energy gamma-ray astronomy. However, the gamma-ray emission
in these objects is not fully suppressed. The recycled radiation with spectral fea-
tures determined by the thickness (‘grammage’) of the gas shell, should be seen in
gamma-rays, unless the synchrotron radiation of secondary electrons dominates over
the bremsstrahlung, and thus channels the main fraction of the non thermal energy
into the sub-gamma-ray domain.
The development of electromagnetic cascades in photon gas and magnetic fields
is a more common phenomenon in astrophysics. In photon fields, such cascades are
created on broad astronomical scales, from compact objects like gamma-ray bursts
to the intergalactic medium as a whole. Very high energy gamma-rays emitted by
astronomical objects and interacting with diffuse extragalactic photon fields initiate
electromagnetic cascades in the entire Universe. The superposition of contributions of
gamma-rays from these cascades can constitute a significant fraction of the observed
diffuse extragalactic background.
The importance of electron–photon cascades in astrophysical objects supported
by γ−γ pair-production and inverse Compton scattering has been first realized in
Ref. [106]. The development of these cascades is well described by the so-called
compactness parameter [147] l = LσT /Rm e c3 , where L is the luminosity and R is
the radius of the source. When l is less than 10, the cascade develops in the linear
26 F. Aharonian
regime, i.e. when the soft radiation produced by cascade electrons does not have a
significant feedback effect on the cascade development [67].
The development of cascades in magnetic fields are best studied for pulsar mag-
netospheres (see e.g. Refs. [89, 148, 236]. Such cascades could be triggered in some
other sites as well, e.g. in the Earth’s geomagnetic field [208] and in blazars [91],
etc. The pair cascades in magnetic fields are effective when the product of the par-
ticle (photon or electron) energy and the strength of the B-field becomes close to
the “quantum threshold” of about E B ≥ Bcrit m e c2 2 × 107 TeV Gauss. A simple
approach, similar to the so-called approximation A used for cascades in matter, in
general can satisfactorily describe the basic features of cascades in magnetic field
[70], but it does not provide adequate accuracy for a quantitative description of the
cascade characteristics [78].
As long as we are interested in the one-dimensional cascade development (which
seems to be quite sufficient for many astrophysical purposes), all 3 types of cascades
can be described by the same integro-differential equations as the ones derived by
Landau and Rumer [180], but in each case specifying the cross-sections of the relevant
interaction processes. These equations over a broad range of energies has been studied
using numerical solutions of the so-called adjoint cascade equations has been recently
conducted by [68].
Although for certain astrophysical scenarios the development of cascades in ‘pure’
environments can be considered as an appropriate and fair approximation, in some
conditions the interference of the processes associated with interactions of cascade
electrons and gamma-rays with both the ambient photon gas and magnetic field (or
matter) can significantly change the character of cascade development, and conse-
quently the spectra of observed gamma-rays. The impact of such interference is very
complex and quite sensitive to the choice of the principal parameters. Therefore each
practical case is a subject to independent studies.
Basic Definitions
Calculations of emissivity of gamma rays resulting from interactions of accelerated

dσ
particles require two functions—the differential cross section of the process, d EdΩ ,
and the energy distribution of parent particles, f (E). While for given energy of
primary particles E, the cross-sections are generally well know functions of the
energy and emission angle of secondary products, the distribution functions of parent
particles depends on the acceleration mechanism (e.g. diffuse shock acceleration or
stochastic acceleration) and the local conditions of the ambient medium—the strength
of magnetic field, the shock speeds, the level of turbulence, etc.
The general kinetic equation that describes the evolution of the particle energy
distribution f (E, r, t) can be written in the form
∂f ∂ ∂ ∂2
= ∇ · (Dr ∇ f ) − ∇ · (ur f ) + (Pr f ) − (br f ) + (dr f ) , (1)
∂t ∂E ∂E ∂ E2
which includes the terms responsible for the diffusion, acceleration and energy losses
of particles. For the derivation of this equations, its approximate solutions under
specific assumptions, and for implications of these solutions the reader is referred to
the original source [140].
Strictly speaking, this equation should be invoked in any detailed treatment of
gamma-ray production. However, this implies that one has to make an assumption
about the acceleration mechanism which in many cases is highly unknown. Thus
the inclusion of acceleration aspect in the overall treatment of the problem not only
would introduce additional uncertainties, but in some cases could be redundant.
In such cases, it would be reasonable to assume an injection of particles, protons
and/or electrons, into the gamma-ray production rate with a given history of rate
and energy spectrum, Q(E, t). Such an approach could be quite appropriate, for
example when the accelerator of particles and the target (the gamma-ray production
region) are separated, or if the acceleration zone is much smaller than the gamma-ray
production zone. In this case Eq. (1) integrated over the volume, can be reduced to
the following simple form
∂N ∂ N
= (P N ) − + Q(E, t) (2)
∂t ∂E τesc
where
N (E, t) ≡ f d 3 r. (3)
Here P(E, t) = −∂ E/∂t is the energy loss rate for a particle with energy E,
and the parameter τesc (E, t) is the characteristic escape time of particles from the
source due to diffusion and convection, τesc = (1/τdiff + 1/τconv )−1 . The injec-
tion function generally is assumed to be a power-law with a high-energy cut-off,
Q(E, t) ∝ E −α exp −(E/E 0 )β .
Here β characterizes the sharpness of the spectrum in the region of the cut-off.
The cut-off energy can be estimated equating the acceleration rate to the cooling
rate, tcool = tacc . Due to energy losses, the particle spectrum in the emission region
is modified, and can significantly deviate from the original injection spectrum. The
steady-state particle distribution can be obtained by setting ∂ N /∂t = 0 in Eq. (2). If
we further assume that the term proportional to τesc −1 can be neglected compared to
the term containing the energy loss rate, the solution for N (E) is very simple,

d E −1 ∞

N (E) = Q(E) d E. (4)
dt E
Several specific realizations for energy losses are of particular interest. For example,
if the energy losses are proportional to the particle energy, d E/dt ∝ E, which is the
case of relativistic Bremsstrahlung, proton-proton inelastic collisions and adiabatic
losses, the initial (injection) power-law spectrum of particles remains unchanged,
N (E) ∝ E −α . The steady-state distribution is steeper if electrons are cooled due to
28 F. Aharonian
synchrotron radiation or inverse Compton scattering (IC) in the Thomson regime,

d E/dt ∝ E 2 , namely N (E) ∝ E −(α+1) . On the other hand, the spectrum becomes
harder if d E/dt is constant (as for ionization losses) or d E/dt ∝ E −1 (the case of
IC scattering in the Klein-Nishina limit): N (E) ∝ E −(α−1) and N (E) ∝ E −(α−2) ,
respectively. These features are reflected in the steady-state gamma-ray spectra.
An order-of-magnitude estimate of gamma-ray luminosity can be obtained using
the cooling time (tcool = E/ |d E/dt|) of the corresponding radiation mechanism,
−1
L γ ≈ We( p) tcool , (5)
where We( p) is the total energy budget of relativistic particles. If V is the volume
of the region, and E min and E max are the minimum and maximum energy of the
accelerated particles, respectively, We( p) can be calculated as
E max
We( p) = E Ne( p) (E) d E d 3 x. (6)
V E min
The shape of the gamma-ray spectral energy distribution can be obtained, in some
cases with a good accuracy, by applying the so-called δ-function approximation
which assumes that the same fixed fraction of the energy of parent particle goes to
secondary gamma-rays. However, the δ-function formalism should be applied with
care, especially near the cut-off of distribution. Otherwise one may arrive at wrong
conclusions (see below) about the distribution of parent particles derived from the
energy distribution of the radiation using the δ-function approximation.
3.2 Brief Overview of Important Processes
Interactions with Matter
The electron bremsstrahlung and the decay of neutral π-mesons produced at inelastic
pp collisions are the most effective high energy gamma-ray production processes in
astrophysical environments related to interactions with matter.
Electron Bremsstrahlung
Often the electron bremsstrahlung is treated together with its counterpart process—
the Bethe-Heitler pair production (BHPP) by gamma-rays at interactions with elec-
trons and nuclei. Comprehensive description of cross-sections of these two processes
can be found in many monographs, in particular in Ref. [149]. Although these two
processes have many similarities, the electron bremsstrahlung has much broader
10
0
/
1
p
/ 0,
br -1
10
-2
10
2 3 4 5
10 10 10 10 10
0
Fig. 13 Total cross-sections of the bremsstrahlung (σbr ) and pair production (σp ) processes in
hydrogen normalised to the asymptotic value (σ0 ) of the pair production cross-section at ε0 → ∞.
The bremsstrahlung cross-sections are calculated for secondary gamma-ray produced with energies
exceeding (1) the pair-production threshold, εth = 2; (2) the critical energy, εth 700; (3) half of
the energy of the primary electron, εth = ε0 /2
astrophysical applications than BHPP. The latter becomes effective only in rare astro-
physical environments, e.g. in very dense shell surrounding particle accelerators.
The integral cross-sections of the bremsstrahlung and pair production processes in
hydrogen gas are shown in Fig. 13. The energies of electrons and γ-rays are expressed
in units of m e c2 . The cross-sections are normalised to the asymptotic value of the
pair production cross-section at ε0 → ∞:
ln(183Z −1/3 )
σ0 = 7/9 × 4αf re2 Z (Z + 1) (7)
1 + 0.12(Z /82)2
where Z is the charge of the target nucleus, and re is the classical electron radius.
Based on σ0 , a new parameter called radiation length is introduced
(m)
X0 = 7/9(nσ0 )−1 , (8)
the physical meaning of which is the average distance over which the ultrarelativistic
electron loses the main fraction of its energy due to bremsstrahlung. The same para-
meter also implies the mean free path of γ-rays. This parameter is widely used to
describe the cascade development in optically thick sources. The cascade effectively
develops at depths exceeding the radiation length. Usually the radiation length is
(m)
expressed in units of g/cm2 . For hydrogen gas X 0 60 g/cm2 . The second impor-
tant parameter that characterises the cascade development is the critical energy below
which ionization energy losses dominate over bremsstrahlung losses. In hydrogen
gas, εcr 700. Effective multiplication of particles in a cascade is possible only at
30 F. Aharonian
102
), arb. units
10
0wbr( 0,
1
-1
10
0 0.2 0.4 0.6 0.8 1
x= / 0
arb. units
1
0wp( 0,, e),
-1
10
0 0.2 0.4 0.6 0.8 1

x= e/ 0
Fig. 14 Differential cross-sections of the bremsstrahlung (upper panel) and pair production (bot-
tom panel) processes in hydrogen. The cross-sections are normalised to one radiation length. The
energies of primary electrons and gamma-ray ε0 (in units of m e c2 ) are indicated at the curves
energies ε ≥ εcr . At lower energies electrons dissipate their energy by ionization

rather than producing more high energy gamma-ray which would support further
development of the electron–photon shower.
In Fig. 13 the bremsstrahlung total cross-sections are shown calculated for 3 dif-
ferent values of minimum energy of emitted gamma-ray: εth = 2, εcr and εe /2. It is
seen from Fig. 13 that while for εth = 2 the pair-production cross-section is an order
of magnitude smaller compared to the bremsstrahlung cross-section, for εth = εe /2
the cross-sections of two processes become almost identical at ε ≥ 100.
The differential cross-sections of bremsstrahlung and pair production are pre-
sented in Fig. 14. The pair-production cross-section obviously is a symmetric function
around the point x = εe /ε0 = 0.5. The bremsstrahlung differential cross-section has
a 1/εγ type singularity at εγ → 0, but because of the hard spectrum of bremsstrahlung
photons the energy losses of electrons contribute mainly to high energy γ-rays. Thus
bremsstrahlung should be treated as an essentially catastrophic process. Neverthe-
less, it is convenient to introduce the so-called average energy loss-rate,

dE e cm p n
− = Ee . (9)
dt br X0
Correspondingly, the lifetime of electrons is
Ee
tbr = 4 × 107 (n/1 cm−3 )−1 year, (10)
−dE e /dt
where n is the number density of the ambient gas.

Note that the electron energy loss rate given by Eq. (9) is proportional to the
electron energy, and, correspondingly, the lifetime given by Eq. (10) is energy inde-
pendent. This implies that for a initial (acceleration) power-law spectrum Q(E e ),
bremsstrahlung losses do not change the original electron spectrum, taking into
account that the cooled steady-state spectrum N (E e ) is given by Eq. (4).
Interestingly, in the case of power-law spectrum of electrons N (E e ), the spec-
trum of bremsstrahlung gamma-rays is also power low with the same power-law
index (this is a result of 1/E γ dependence of the differential cross-section). Thus,
the bremsstrahlung gamma-ray spectrum simply repeats the shape of the electron
acceleration spectrum Q(E e ) ∝ E e−Γ .
This is true, however, only when the energy losses are dominated by brems-
strahlung. In hydrogen gas, at energies below ∼700m e c2 350 MeV ionization
dominates over the bremsstrahlung. Because both the ionization and bremsstrahlung
loss rates are proportional to n, this condition does not depend on the ambient gas
density. On the other hand, in the relativistic regime, the ionization loss rate does
not depend on the electron energy. Thus, in accordance with Eq. (4), the steady-
state electron spectrum becomes flatter, N (E e ) ∝ E e−Γ +1 , and correspondingly at
energies below several hundred MeV we should expect a very hard bremsstrahlung
gamma-ray spectrum with power-law photon index Γ − 1.
In many astrophysical scenarios the inverse Compton and synchrotron losses may
well dominate, especially at very high energies, over bremsstrahlung, depending on
the ratio of the energy density of the radiation and magnetic fields to the number
density of the ambient gas. The synchrotron and inverse Compton energy loss rates
are proportional to the electron energy. This makes the steady-state electron spectrum
steeper, N (E e ) ∝ E e−Γ −1 , and correspondingly the bremsstrahlung gamma-rays
emerge with photon index Γ + 1.
The positron annihilation is another channel of gamma-ray production related
to interactions with matter. The astrophysical significance of this process generally
is linked to the annihilation line at energy m e c2 = 0.511 MeV, as well as to the
3-photon positronium continuum, produced by annihilation of thermalized positrons
with background cold electrons of the ambient gas/plasma. However, if positrons are
injected into the production region with relativistic energies, a significant fraction of
positrons (from 10 to 20 %, depending on the ionization state of the ambient plasma)
annihilate in flight before they cool down to the temperature of the thermal gas [55].
The energy spectrum of gamma-rays has a symmetric form with two sharp peaks
at E γ ∼ m e c2 and E γ ∼ E + . This implies that the energy of relativistic positron
32 F. Aharonian
is transferred mainly to the leading photon. For a power-law spectrum of positrons,

−α
N+ ∝ E + the spectrum of annihilation radiation at E γ m 2c has a power-law
form
Jann (E γ ) ∝ E γ−(α+1) [ln(2E γ /m 2c ) − 1] . (11)
The total cross-section of annihilation of a relativistic positron of energy ε+ is

given by σann = 38 σT2 (E + /m e c2 )−1 [ln(2E + /m e c2 ) − 1]. Correspondingly the anni-
hilation time is
(E + /m 2c )
tann 4 × 106 (n/1 cm−3 )−1 year. (12)
ln(2E + /m 2c ) − 1
Comparing Eq. (12) with Eq. (10) one finds that for a positron of energy E + ≤
15 MeV, the annihilation times becomes shorter than the bremsstrahlung cooling
time, i.e. at these energies the annihilation continuum starts to dominate over the
bremsstrahlung spectrum, taking into account that in both processes the leading
photon receives a substantial part of the positron kinetic energy.
Gamma Rays and Neutrinos from pp Interactions
Relativistic protons and nuclei produce high energy gamma-rays and neutrinos
in inelastic collisions with ambient gas due to the production and decay of sec-
ondary pions, kaons and hyperons. The neutral π 0 -mesons provide the main chan-
nel of conversion of the kinetic energy of protons to high energy gamma-rays.
For the production of π 0 -mesons the kinetic energy of protons should exceed
E th = 2m π c2 (1 + m π /4m p ) ≈ 280 MeV, where m π = 134.97 MeV is the mass
of the π 0 -meson. This particle immediately decays to two gamma-rays. The mean
lifetime of π 0 -decay, tπ0 = 8.4 × 10−17 s, is significantly shorter than the lifetime
of charged π-mesons (≈2.6 × 10−8 s). At high energies, all three types of pions are
produced with comparable probabilities. The spectral form of π-mesons is generally
determined by a few (one or two) leading particles (that carry a significant fraction
of the nucleon energy) rather than by the large number of low-energy secondaries.
The decays of charged pions lead to neutrinos with spectra quite similar to the
spectrum of the accompanying π 0 -decay gamma-rays. However, this symmetry can
be violated in environments with high gas or radiation densities. In certain conditions,
the characteristic time for inelastic interactions of charged pions with nucleons or
photons could be shorter than the decay time, so the energy of pions degrades before
they decay. At very high energies this would result in significantly smaller fluxes of
neutrinos compared to gamma-rays.
The distinct feature of the spectrum of π 0 -decay γ-rays is the maximum at E γ =
m π c2 /2 67.5 MeV, independent of the energy distribution of π 0 mesons, and
consequently of the parent protons. The appearance of such a bump in the gamma-
ray spectrum is a result of the π 0 → 2γ decay kinematics. It is easy to show (see e.g.
Ref. [220] that the spectrum of gamma-rays from decays of mono energetic pions
of energy E π and velocity vπ is constant f (E π ) = c/(vπ E π ) within the interval

between E 1 = 0.5E π (1 − vπ /c) and E 2 = 0.5E π (1 + vπ /c). The spectrum of
gamma-rays for an arbitrary distribution of π 0 -mesons Π (E π ) can be presented as
superposition of rectangles for which only one point at m π c2 /2 is always presented.
Obviously this should result in a spectral maximum independent of the distribution
of parent pions.
Although precise calculations of gamma-ray spectra require quite heavy integra-
tions over differential cross-sections obtained experimentally at particle accelerators,
the emissivity of γ-rays for an arbitrary broad energy distribution of protons without
sharp features can be derived within a simple formalism of δ-function approxima-
tion which nevertheless provides surprisingly good over a broad gamma-ray energy
range [58].
The gamma-ray emissivity qγ (E γ ) is directly defined by qπ (E π ) as
∞ qπ (E π )
qγ (E γ ) = 2 dE π , (13)
E min E π2 − m 2π c4
where E min = E γ + m 2π c4 /4E γ . The emissivity of secondary pions qπ from inelas-

tic proton–proton interactions can be calculated with high accuracy using accel-
erator measurements of the inclusive cross-sections σ(E i , E p ). The emissivity of
π 0 -mesons calculated in the δ-function approximation for the cross-section σ(E π , E p )
then becomes

qπ (E π ) = c n H δ(E π − κπ E kin )σpp (E p )n p (E p )dE p

c nH Eπ Eπ
= σpp m p c2 + n p m p c2 + (14)
κπ κπ κπ
where σpp (E p ) is the total cross section of inelastic pp collisions, and κπ is the
mean fraction of the kinetic energy E kin = E p − m p c2 of the proton transferred
to the secondary π 0 -meson per collision; n p (E p ) is the energy distribution of the
protons. In a broad region from GeV to TeV energies κπ ≈ 0.17 which includes a
∼6 % contribution from η-meson production [139]. From the threshold at E kin
0.3 GeV, σpp rises rapidly to about 30 mb. But after E kin ∼ 2 GeV, σpp increases
only logarithmically. The measurements of the total inelastic cross-section are shown
in Fig. 15 together with a simple analytical approximation [165]:
σinel (E p ) = (34.3 + 1.88 L + 0.25 L 2 ) × [1 − (E th /E p )4 ]2 mb, (15)
where E th = m p +2m π +m 2π /2m p = 1.22 GeV is the threshold energy of the proton
for π 0 production. This expression correctly describes σinel E p near the threshold,
and fits the experimental data and SYBILL simulations up to at least E p ∼ 104 TeV.
34 F. Aharonian
Fig. 15 Inelastic cross-section of pp interactions approximated by Eq. (15). The experimental data
are from http://wwwppds.ihep.su:8001/c5-5A.HTML, the open points correspond to the cross-
sections which are used in the SIBYLL code
The δ-function approach yields the gamma-ray emissivity with good accuracy
for a broad energy distribution of protons without sharp features. Indeed, in the
δ-function formalism, the shape of the photon spectrum is similar to the shape of the
parent proton spectrum, shifted in energy by a factor κπ . However, this approximation
does not give correct results if the proton energy distribution contains sharp features
like a pile-up or a cut-off [165].
The energy spectra of gamma-rays from pp interactions have been comprehen-
sively studied recently by Kelner et al. [165] and Kamae et al. [162] who also
developed convenient procedures for calculations of gamma-ray spectra for arbi-
trary energy distributions of protons. In particular, in Ref. [165] simple analytical
parametrizations of energy spectra of secondary particles—gamma rays, electrons,
and neutrinos have been obtained based on simulations of proton-proton interactions
using the public available SIBYLL code.
For a proton of energy E p , the number of gamma-rays per a pp collision from the
decays of secondary π 0 and η-mesons, in the interval (x, x + d x), where x = E γ /E p ,
can be presented as
4
ln x 1 − x βγ
Fγ x, E p = Bγ
x 1 + kγ x βγ 1 − x βγ
1 4βγ x βγ 4kγ βγ x βγ 1 − 2x βγ
× − − (16)
ln x 1 − x βγ 1 + kγ x βγ 1 − x βγ
Fig. 16 Spectra of gamma-ray, electrons and neutrinos from inelastic pp collisions, for power-law
proton distributions given by Eq. (18) with α = 2, β = 1, and E 0 = 1000 TeV. The dashed curve
is the photon spectrum calculated in the δ-function approximation, with κπ = 0.17. Two curve for
all muonic neutrinos (νμ ) corresponds to the contributions from two channels—νμ(1) : π → μνμ ,
and νμ(2) : μ → eνe νμ . The spectrum of electronic neutrinos coincides, with an accuracy better than
5 %, with the electron spectrum
with Bγ = 1.30 + 0.14L + 0.011L 2 , βγ = (1.79 + 0.11L + 0.008L 2 )−1 , and

kγ = (0.801 + 0.049L + 0.014L 2 )−1 , where L = ln(E p /1 TeV).
Equation (16) allows one to derive the gamma-ray emissivity qγ (E γ ) for an arbi-
trary distribution of protons N p E p :
∞
Eγ dEp
qγ E γ = c n H σinel E p N p E p Fγ
, Ep
Eγ Ep Ep
1
Eγ Eγ Eγ d x
= c nH σinel Np Fγ x, , (17)
0 x x x x
where the inelastic pp cross section σinel E p is given by Eq. (15).

Figure 16 shows the spectra of secondary gamma-ray, electrons and neutrinos
obtained for the proton distribution given in the form
β
−α E
N (E) = A E exp − , (18)
E0
with α = 2, β = 1, and E 0 = 1000 TeV. The constant A is set from the condition
that the energy density of protons w = E N (E)d E = 1 erg/cm3 .
It should be noted that, the spectrum of gamma-rays deviates, in contrast to
commonly used assumptions, from the power-law form of the parent proton spec-
trum even at energies well below the cut-off energy and well above the thresh-
old of π 0 production. This is explained by the increase of the total cross-section
36 F. Aharonian
with energy; the effect becomes significant especially at energies of protons above
1 TeV (see Fig. 15). Another interesting feature is the shape of the gamma-ray spec-
trum in the cut-off region. The δ-function approximation which agrees quite well
with the results of accurate calculations at low energies, fails to describe correctly
the spectrum at highest energies around the cut-off (see Fig. 16). In particular, it
predicts exp[−(E γ /E γ,0 )] type cut-off term with E γ,0 = kπ E 0 0.17E 0 , i.e.
the same shape as the proton spectrum but shifted by a factor of 6 towards low
energies. The accurate calculations result in a significantly different shape, namely
exp[−(E γ /E 0,γ )1/2 ] with E 0,γ 0.06E 0 . This implies that the gamma-ray spec-
trum displays a smoother cut-off, and the impact of the proton cut-off starts earlier,
already around E γ ∼ 0.01E 0 .
Finally we note that for α = 2, the amplitude of the gamma-ray spectrum at
energies E E 0 exceeds the level of the flux of muonic neutrinos. The difference
is due to the contribution of η-mesons in the production of gamma-rays; If one takes
into account only the decays of π 0 -mesons, the spectra of gamma-rays and muonic
neutrinos appear almost identical at energies well below the cut-off region. This
conclusion does not contradict the fact that in each pp interaction the number of
muonic neutrinos is a factor of 2 larger, on average, than the number of gamma-rays.
The imbalance is compensated, in fact, at low energies (see Fig. 16).
Processes in Magnetic Fields
Synchrotron Radiation
Many important results of the theory of synchrotron radiation can be obtained within
the framework of classical electrodynamics (see e.g. Ref. [140]) which is limited by
the condition
E B
χ= 1, (19)
m e c2 Bcr
where Bcr = m 2e c3 /e ≈ 4.4 × 1013 G is the critical value of the magnetic field.
The synchrotron cooling time for a particle of mass m and energy E in a magnetic
field of strength B is
3
m m e c2 mc2
tsynchr = 6π . (20)
me cσT B 2 E
4
The cooling time of a protons is much longer, by a factor of m p /m e ≈ 1013 ,
compared to the synchrotron cooling time of an electron of same energy. Therefore
in astrophysics the electron synchrotron radiation is far more important process
compared to the proton synchrotron radiation, although the latter in some specific
conditions can also be an effective mechanism of gamma-radiation.
The energy distribution of photons radiated by an electron of energy E is given
by equation (see e.g. Ref. [140])
√
d Nγ 3 e3 B Eγ
= F , (21)
d E γ dt 2π m e c E γ
2 Ec
where
∞
F(x) = x K 5/3 (τ )dτ . (22)
x
Here K 5/3 is the modified Bessel function of order 5/3, and

2
3eB E
Ec = (23)
2m e c m e c2
is the so-called characteristic energy of synchrotron radiation.

In a regular magnetic field, the strength B in these equations should be considered
as the perpendicular component of the field, B⊥ = Bsinθ, where θ is the pitch-angle.
In chaotic magnetic fields, Eq. (21) should be averaged over directions of magnetic
field, and the function F(x) should be replaced by the function G(x) which can be
presented in the form [65]
x
G(x) = [(8 + 3x 2 )(κ1/3 )2 + xκ2/3 (2κ1/3 − 3xκ2/3 )], (24)
20
where κ1/3 = K 1/3 (x/2), κ2/3 = K 2/3 (x/2) (for a different presentation of this
function in terms of Whittaker’s function see Ref. [120]). Note that while the function
F(x) has a maximum at x = 0.2858 (max F(x) = 0.9180), the maximum of the
function G(x) is shifted towards smaller values: x = 0.2292 (max G(x) = 0.7126).
The functions F(x) and G(x), as well as the ratio G(x)/F(x) are shown in Fig. 17.
Although the function G(x) in Eq. (24) has a quite compact form, for practical
purposes often it is desirable to have an approximation which does not contain special
functions. In particular, the following convenient approximations for functions F(x)
and G(x) obtained in Ref. [65]
1 + 0.884x 2/3 + 0.471x 4/3 −x

F(x) ≈ 2.15x 1/3 (1 + 3.06x)1/6 × e , (25)
1 + 1.64x 2/3 + 0.974x 4/3
1.808x 1/3 1 + 2.21x 2/3 + 0.347x 4/3 −x
G(x) ≈ √ e . (26)
1 + 3.4x 2/3 1 + 1.353x 2/3 + 0.217x 4/3
provide an accuracy better than 0.2 % over the entire range of variable x.
The spectral energy distributions (SED) of synchrotron photons corresponding to
distributions of electrons presented in the form of Eq. (18), can be described by a
simple analytical expression
− α−1

λ
L γ ∝ Eγ 2
exp − E γ /E γ,0 , (27)
38 F. Aharonian
Fig. 17 The emissivity

functions for synchrotron
radiation F(x) and G(x). The
dashed line shows the ratio
G(x)/F(x)
In Fig. 18 the results of accurate calculations are compared with the ones obtained
within the δ-function approximation. It is seen that while the latter describes correctly
the power-law part of the spectrum of synchrotron radiation, it fails to reproduce
correctly the radiation spectrum in the cut-off region. Indeed, this approximation
for the index in the exponential cut-off in Eq. (27) predicts λ = β/2 which is quite
different from the results of exact analytical calculations, λ = β/(β + 2) [135, 253].
Fig. 18 Spectral energy distributions of synchrotron radiation from a distribution of electrons as

in Eq. (18), with indices α = 2, 3 and β = 1 (left panel) and β = 2 (right panel). In all cases
E 0 = 1015 eV and B = 1 mG. The solid and dashed curves are the result of exact calculations,
whereas the dotted and dash-dotted curves were obtained applying the δ-function approximation
The spectra of synchrotron radiation in the cut-off region appear much smoother
compared to the ones calculated in the δ-function approximation.
The shape of the particle spectrum in the cut-off region depends, among other
factors, on the specific mechanisms of acceleration and energy losses. The cut-off
energy itself can be estimated from the balance between the rates of acceleration and
energy losses. The particle acceleration time can be written as
E
tacc = η (E) . (28)
ecB
The parameter η (E) ≥ 1 characterizes the acceleration rate, and generally (in most
models) is treated as a free parameter. Note that E eff = η −1 B can be treated as the
projection of an electric field E on the particle trajectory, effectively averaged as the
particle moves along the trajectory. This representation is useful, in particular, in a
formal treatment of the diffusive shock acceleration [61]. In the majority of cosmic
accelerators B E, and correspondingly η 1. However in certain situations
η can be close or even less than 1, namely when E ≥ B. When such a condition is
satisfied, the particles are unbound in the acceleration region, and the energy gain is
determined by the difference of the electric potential across the acceleration region.
If energy losses are dominated by synchrotron radiation, the condition tacc =
tsynchr yields a maximum energy
3/4
3 m 2 c4
E0 = . (29)
2 η e3 B
The corresponding cut-off energy of the synchrotron spectrum appears independent

of the magnetic field B (see e.g. Ref. [51]).
9 mc2
E γ,0 = . (30)
4 αfη
Here α f = e2 /c ≈ 1/137 is the fine-structure constant. For η = 1, the cut-offs

appear at energies ≈160 MeV and ≈300 GeV for synchrotron radiation by electrons
and protons, respectively. Any violation of these limits would imply acceleration of
particles at conditions when E > B. Realization of this condition is difficult, but
a priori cannot be excluded in compact relativistic objects.
In more standard scenarios, B E, i.e. η 1. For example in the case of
non-relativistic diffusive shock acceleration,
2
D(E) c
η ≈ 10 , (31)
rg c vs
where rg = E/eB is the particle gyroradius, D(E) is the diffusion coefficient, and
vs is the shock speed. If diffusion proceeds in the Bohm regime, D(E) = rg c/3, thus
40 F. Aharonian
27 mc2 vs 2
E γ,0 ≈ . (32)
40 α f c
For young supernova remnant in the Sedov phase, the shock speeds are typically of
order of several thousand km/s, therefore the synchrotron peaks appear in the range
of a few keV.
Magnetic Pair Production
In general, the energy of the synchrotron photons is much less than the energy of
the parent particle. But in certain astrophysical environments, synchrotron radiation
can take place near the quantum threshold. In this limit, the production of electron–
positron pairs in the magnetic field by high-energy photons,
γ + B −→ e+ + e− . (33)
becomes equally important. In the quantum regime, synchrotron radiation and mag-
netic pair production are tightly coupled. The operation of these two processes
together may initiate and support an effective electron–photon cascade development
in environments with very large magnetic field, provided that the mean free path of
gamma-rays or electrons are smaller than the size of the source.
The probabilities of interaction per unit length (also called differential cross sec-
tions) of an electron or a photon with the magnetic field are [78]

α f m e c3 1 Ee − Eγ Ee 2u
ωsynchr E e , E γ = √ + K2
π 3 E 2 E Ee − Eγ 3 3χ
∞
− K 1 (y) dy , (34)
2u 3
3χ
and

α f m e c3 1 Eγ − Ee Ee 2ũ
ω p Eγ , Ee = √ + K2
π 3 E γ2 Ee Eγ − Ee 3 3χ
∞
+ K 1 (y) dy , (35)
2ũ 3
3χ
respectively. Here u = E γ /(E e − E γ ) and ũ = E γ2 /E e (E γ − E e ). The parameter χ

is given by Eq. (19), with E = E e or E = E γ depending on the process.
Figure 19 shows the total cross sections σsynchr and σ p as a function of χ. These
are obtained integrating Eqs. (34) and (35) over E γ and E e , respectively. For χ 1,
the probability of synchrotron radiation is approximately constant, but the probability
Fig. 19 Interaction probabilities of synchrotron radiation and magnetic pair production as a function
of χ
Fig. 20 Differential interaction probabilities (differential cross sections) of synchrotron radiation

(left panel) and pair production in a magnetic field (right panel). Both quantities are normalised to
the corresponding total cross section. The values of χ are indicated at each curve
of pair production drops abruptly as exp(−8/3χ). For χ 10, both cross sections
decrease with χ as χ−1/3 , but the probability of synchrotron radiation is several times
larger.
Figure 20 shows the differential cross sections for both processes for different
values of χ. The distribution of pairs created through interaction of gamma rays
with the magnetic field is symmetric around E e = 0.5E γ . The pair spectrum is
flat, and the photon transfers most of its energy to one of the created particles. The
energy spectrum of synchrotron photons is very steep for χ 1, but becomes flat
at χ ≥ 1.
42 F. Aharonian
Interactions with Photon Fields
The interaction of relativistic electrons with radiation fields through inverse Compton
scattering provides one of the principal gamma-ray production processes in astro-
physics. It works effectively almost everywhere, from compact objects like pulsars
and AGN to extended sources like supernova remnants and clusters of galaxies. Since
the Compton cooling time decreases linearly with energy, the process becomes espe-
cially effective at very high energies.
The electron–positron pair production in photon–photon collisions is tightly cou-
pled with the inverse Compton scattering. First of all, it is an absorption process that
prevents the escape of energetic gamma-rays from compact objects, and determines
the “gamma-ray horizon” of the Universe. At the same time, in an environment where
the radiation pressure dominates over the magnetic field pressure, the photon–photon
pair production and the inverse Compton scattering “work” together supporting an
effective transport of high energy radiation via electromagnetic “Klein-Nishina” cas-
cades.
Although the IC scattering of protons is suppressed, compared to the scattering of
electrons, by a factor of (m e /m p )4 , very high energy protons effectively interact with
the ambient photon fields through electron–positron pair production and photomeson
processes. While in the first process gamma-rays are produced indirectly, via inverse
Compton scattering of the secondary electrons, photo-meson reactions result in the
direct production of π 0 -mesons and their subsequent decay to γ-rays. Typically,
at extremely high energies these interactions proceed effectively both in compact
objects and large scale structures.
Inverse Compton Scattering
The total cross-section of inverse Compton scattering depends only on the product
of energies of the interacting electron ε and photon ω0 , κ0 = ω0 εe (all energies are
in units of m e c2 ). In the non relativistic regime (κ0 1) it approaches the classical
(Thomson) cross-section σIC ≈ σT (1 − 2κ0 ), while in the ultrarelativistic regime
(κ0 1) it decreases with κ0 as σIC ≈ (3/8)σT κ−1 0 ln(4κ0 ). With an accuracy of
better than 10 % in a very broad range of κ0 , the cross-section can be represented in
the following simple form [118].

3σT 2 2 1 4 1
σIC = 1− − 2 ln(1 + 2κ0 ) + + − (36)
8κ0 κ0 κ0 2 κ0 2(1 + 2κ0 )2
The total cross-section of Compton scattering as a function of κ0 is shown in Fig. 21.

The energy distribution of up-scattered gamma-rays is determined by the differen-
tial cross-section of the process. Assuming that a mono energetic beam of low energy
photons ω0 penetrates an isotropic and homogeneous region filled with relativistic
electrons of energy εe , the spectrum of radiation scattered at the angle θ relative to
Fig. 21 Total cross-sections 1

of inverse Compton scat-
tering and photon–photon
pair production in isotropic
radiation fields. Two spectral 10 -1
distributions for the ambient
photon gas are assumed:
(i) mono energetic with 10 -2
energy ω0 (curves 1 and 3),
T
and (ii) Planckian with the
/
same mean photon energy
ω0 3kT /m e c2 (curves 2 10 -3
and 4)
10 -4
-2 -1 2 3 4 5
10 10 1 10 10 10 10 10
0
= 0 0
the initial photon beam is written as [54].
d2 N (θ, εγ ) 3σT z2 2z 2z 2
= 1+ − + 2 , (37)
dεγ dΩ 16πω0 εe
2 2(1 − z) bθ (1 − z) bθ (1 − z)2
where bθ = 2(1−cos θ)ω0 εe , z = εγ /εe . The energy of the high energy γ-ray photon
εγ varies in the limits ω0 εγ εγ,max , where εγ,max = εe bθ /(1 + bθ ), b = 4κ0 .
In the case of isotropically distributed electrons and photons, the integration of
Eq. (37) over the angle θ gives [54, 98, 161].

dN (εγ ) 3σT z2 z 2z 2
= 1 + + −
dεγ 4ω0 ε2e 2(1 − z) b(1 − z) b2 (1 − z)2

z3 2z b(z − 1)
+ − ln . (38)
2b(1 − z)2 b(1 − z) z
The differential energy spectra of gamma-rays for several fixed values of κ0 are
shown in Fig. 22. In the deep Klein-Nishina regime (κ0 1) the spectrum grows
sharply towards the maximum at εγ,max . This implies that in this regime just one
interaction is sufficient to transfer a substantial fraction of the electron energy to
the upscattered photon. In the Thomson regime (κ0 1) the average energy of
the upscattered photon is εγ ≈ ω0 ε2e , thus only a fraction εγ /εe ∼ κ0 1 of the
primary electron energy is released in the upscattered photon.
For a power-law distribution of electrons, dNe /dεe ∝ ε−Γ e , the resulting gamma-
ray spectrum in the non relativistic regime (a = 4ω0 εγ 1) has a power-law form
44 F. Aharonian
wics(ε0,εγ), arbitrary units

10
-1
10
-2
10
0 0.2 0.4 0.6 0.8 1
x=εγ/εγmax
wp(ε0,εe), arbitrary units
10
-1
10
0 0.2 0.4 0.6 0.8 1

x=(εe-εemin)/(εemax-εemin)
Fig. 22 Differential spectra of gamma-rays from inverse Compton scattering (upper panel) and
electrons from photon–photon pair production (bottom panel) in an isotropic and mono-energetic
√ εγ,max , εe,min and εe,max are defined as εγ max = 4ε0 (κ0 /1 + 4κ0 ) and
photon field. The parameters
εe min,e max = 0.5ε0 (1 ∓ 1 − 1/κ0 ). The same values of the parameters κ0 = εe ω0 and s0 = εγ ω0
are indicated by the curves
with photon index α = (Γ + 1)/2 [140]. In the ultrarelativistic (a 1) regime the

gamma-ray spectrum is noticeably steeper, ∝ ε−α γ (ln a + const) with α = (Γ + 1)
[98]. Useful analytical approximations for gamma-ray spectra over a broad energy
interval have been obtained in Refs. [54, 118].
The energy-loss rate of relativistic electrons in a mono-energetic field of photons
with energy ω0 and number density n ph is given by the following equation [54]

dεe 3σT cn ph b 6
= 6+ + ln(1 + b) − ln2 (1 + b)
dt 4ω0 b 2 b

1 (11/12)b3 + 8b2 + 13b + 6
− 2Li − . (39)
1+b (1 + b)2
y)−1 ln(y) dy.

1
where Li(x) = x (1 −
In the Thomson and Klein-Nishina regimes Eq. (39) reduces to the well known
expressions [98].
dεe 4
= σT cω0 n ph ε2e at b 1, (40)
dt 3
and
dεe 3 σT cn ph
= (ln b − 11/6) at b 1. (41)
dt 8 ω0
The energy losses in these two regimes have quite a different dependence on the
electron energy. While in the Thomson regime the loss rate is proportional to ε2e , in
the Klein-Nishina regime it is almost energy independent. This implies that in the
first case the energy losses make the electron spectrum steeper, whereas the in the
Klein-Nishina regime the electron spectrum becomes harder.
Photon–Photon Pair Production
Photon–photon pair production is the inverse process to pair annihilation. Therefore

the differential cross-section is identical to the pair annihilation cross-section, except
for a different phase-space volume. In the relativistic regime this process is quite
similar also to inverse Compton scattering. However, unlike the pair annihilation
and Compton scattering, the photon–photon pair production has a strict kinematic
threshold given by
εγ1 εγ2 (1 − cos θ) ≥ 2, (42)
where εγ1 and εγ2 are the energies of two photons in units of m e c2 colliding at an
angle θ (in the laboratory frame).
The large cross-section makes the photon–photon pair production one of the
most relevant elementary processes in high energy astrophysics. Several convenient
approximations for the total cross-section of this process in the isotropic radiation
field have been proposed in Refs. [60, 118, 145]. With an accuracy of better than 3 %,
the total cross-section in the monoenergetic isotropic photon field can be represented
in the following analytical form

3σT 1 1 1 √
σγγ = 2 s0 + ln s0 − + ln( s0 + s0 − 1)
2s0 2 6 2s0

4 1 1
− s0 + − 1− . (43)
9 9s0 s0
The total cross-sections of inverse Compton scattering and pair production in an

isotropic monoenergetic photon field of energy ω0 are shown in Fig. 21 (curves 1
and 3, respectively). Both cross sections depend only on the product of the primary
46 F. Aharonian
(εe or εγ ) and target photon (ω0 ) energies, κ0 = εe ω0 and s0 = εγ ω0 . While as

κ0 → 0, the inverse Compton cross-section approaches the Thomson cross-section,
σIC ≈ σT (1 − 2κ0 ), as s0 → 1 the pair production cross-section approaches zero,
σγγ ≈ (1/2)σT (s0 −1)3/2 . For κ0 , s0 1 the two cross-sections are quite similar and
decrease with κ0 and s0 : σIC ≈ (3/8)σT κ−1 −1
0 ln(κ0 ), σγγ ≈ (2/3)σT s0 ln(s0 ). The
pair-production cross-section has a maximum at the level of σγγ ≈ 0.2σT achieved
at s0 ≈ 3.5–4.
The parameter that characterises gamma-ray absorption at photon–photon inter-
actions in a source of size R is the so-called optical depth
R ω2
τ (εγ ) = σ(εγ , ω)n ph (ω, r )dωdr, (44)
0 ω1
where n ph (ω, r ) describes the spectral and spatial distribution of the target photon
field in the source.
Because of narrowness of the pair-production cross-section, for a large class of
broad band target photon energy distributions n ph (ω), the optical depth at given
gamma-ray energy εγ is essentially determined by a relatively narrow band of tar-
get photons with energy centred on ω∗ = 4/γ . Therefore, the optical depth can be
written in the form τ (εγ ) = η(σT /4)ω∗ n ph (ω∗ )R, where the normalization factor
η depends on the spectral shape of the background radiation. For a power-law tar-
get photon spectrum, n ph (ω) = n 0 ω −α , the parameter η is calculated analytically,
η = (7/6)4α α−5/3 (1 + α)−1 [224].
The energy spectrum of electrons produced at photon–photon pair production has
been studied in Refs. [60, 99, 118]. For a low-energy monoenergetic photon field
(ω0 1), and correspondingly εγ 1, the spectrum of electron–positron pairs
can be represented, with an accuracy of better than a few percent, in the following
analytical form [60]:
dN (εe ) 3σT 4ε2γ

4ω0 (εγ − εe )εe
= − 8ω0 εγ
ln
dεe 32ω02 ε3 (εγ − εe )εe εγ

2(2ω0 εγ − 1)ε2γ 1 ε4γ
+ − 1− . (45)
(εγ − εe )εe ω0 εγ (εγ − εe )2 ε2e
The kinematic range of variation of εe is

εγ 1 εγ 1
1− 1− ≤ εe ≤ 1+ 1− . (46)
2 ω 0 εγ 2 ω 0 εγ
The differential energy spectra of gamma-rays for several fixed values of the
parameters s0 = ω0 εγ are shown in Fig. 22. The spectra are symmetric around the
point x = εe /εγ . Although the average energy of the secondary electrons is εγ /2,
for very large s0 the interaction has a catastrophic character—the major fraction of
the energy of the primary photon is transferred to the leading electron.
These two processes—inverse Compton scattering and photon–photon pair pro-
duction—determine the basic features of interactions of electrons and gamma-rays in
the radiation dominated environments. At extremely high energies, the higher order
QED processes may compete with these basic channels. Namely, when the product
of the energies of colliding cascade particles (electrons or photons) E and the back-
ground photons ω significantly exceed 105 m 2e c4 , the processes γγ → e+ e− e+ e− and
eγ → eγe+ e− may dominate over the single (e+ , e− ) pair production and Compton
scattering, respectively. Because the γγ → 2e+ 2e− and eγ → eγe+ e− channels
result in production of 2 additional electrons, they substantially change the charac-
ter of interactions. Note, however, that an effective realization of these processes is
possible only under very specific conditions with an extremely low magnetic field
and narrow energy distribution of the background photons.
Interactions of Hadrons with Radiation Fields
In astrophysical environments the radiation density often exceeds the density of gas
component. In these conditions the interactions of high energy hadrons with radiation
can dominate over interactions with matter, albeit the relevant cross-sections are
relatively small. The main processes of hadron–photon interactions include (i) inverse
Compton scattering: p+γ → p+γ , (ii) electron–positron pair production: p+γ →
pe+ e− , (iii) photo-disintegration of nuclei: A + γ → A + kN, (iv) photomeson
production: N + γ → N + kπ. In extremely dense radiation fields the secondary
π ± -mesons may effectively interact with photons before they decay.
Except for the inverse Compton scattering, all other processes take place only
above certain kinematic thresholds: ∼1, 10, and 140 MeV (in the rest frame of pro-
jectile particles) for the pair production, photo-disintegration, and pion production,
respectively.
The process of inverse Compton scattering of protons is identical to the inverse
Compton scattering of electrons, but the energy loss rate of protons is suppressed,
for the fixed energy of both particles, by a factor of (m e /m p )4 ≈ 10−13 . Generally,
this process does not have noticeable astrophysical applications. At energies above
the pair production threshold, the inverse Compton energy loss rate is significantly
(by a factor of α(m p /m e )2 ∼ 104 ) slower compared to the losses caused by pair-
production.
In certain conditions the pair-production may result in significant spectral dis-
tortions of highest energy protons propagating through dense photon fields. The
cross-section of this process, which in fact is the same Bethe-Heitler pair produc-
tion, is quite large, but in each interaction only a small fraction of the proton energy
is transferred to the secondary electrons. Therefore the energy loss rate of protons
remains relatively slow. Moreover, the energy region where this process dominates
is quite narrow.When the proton energy exceeds the pion production threshold, the
photomeson interactions immediately start to dominate over the pair production.
48 F. Aharonian
The photo-disintegration of nuclei may have a strong impact on the formation of

the chemical composition of very high energy cosmic rays in compact astrophysical
objects as well as in the intergalactic medium. However, this process is an inefficient
mechanism for gamma-ray production [50].
Photomeson production is much more important channel for transformation of
the kinetic energy of protons into high energy gamma-rays, electrons and neutrinos.
Close to the energy threshold, th = m π c2 (1 + m π /2m p ) ≈ 145 MeV, the process
proceeds through single-pion production, p + γ → p + π 0 , and p + γ → n + π + .
At higher energies, multi-pion production channels begin to dominate. The cross-
sections of these processes are basically well known from particle accelerator exper-
iments. For astrophysical applications the data obtained with gamma-ray beams at
energies from 140 MeV to 10 GeV are quite sufficient, if one takes into account
the fact that for typical broad-band target photon spectra the hadron–photon interac-
tions are contributed mainly from the region not far from the energy threshold, i.e.
E γ ≤ 1 GeV.
The cross-sections of interactions of secondary electrons and gamma-rays with
the ambient photons exceed by three orders of magnitude the photomeson cross-
sections. Therefore the electrons and gamma-rays cannot leave the active region of
pion production, but rather initiate electromagnetic cascades in the surrounding pho-
ton and magnetic fields. The standard spectra of the low-energy cascade gamma-rays
that eventually escape the source are not sensitive to the initial spectral distributions,
and thus contain information only about the total hadronic power of the source. On
the other hand the secondary neutrinos freely escape the production region, and thus
carry direct information about the energy spectra of accelerated protons.
Another interesting feature of the mixed hadronic/electromagnetic cascades in
radiation dominated environments is the effective transport of primary nonthermal
energy released in accelerated protons further away from a central engine through
production and escape of secondary neutrons [129].
The presence of dense photon fields in the compact particle accelerators may have
an even more fundamental impact. In relativistic flows, the multiple conversions
of relativistic particle from charged to neutral state (proton→neutron→proton...,
e → γ → e...) may allow a strong (up to the bulk Lorenz factor squared) energy gain
in each cycle, whereas in the standard relativistic shock acceleration scenario the
energy gain ∼Γ 2 occurs only in the first circle. This acceleration mechanism [123]
is capable to boost protons in GRBs and AGN jets to maximum available energies,
thus could be a key to the solution of the problem of the highest energy, E ≥ 1020 eV,
particles observed in cosmic rays.
A very useful tool for detailed calculations of photomeson processes in radiation
fields is the SOPHIA code [197]. For the given distributions of relativistic protons
N p (E p ) and target photons n ph (), the gamma-ray emissivity can be written as

d Nγ dEp
= N p (E p ) n ph () Φ (κ, x) d. (47)
d Eγ Ep
Here κ = 4E p /m 2p c4 and x = E γ /E p .

Using numerical results obtained with the code SOPHIA, the function Φ (κ, x)
can be approximated with an accuracy better than 10 % by a simple analytical expres-
sions [164]. In the region x− < x < x+ , the function Φγ (κ, x) can be presented in
the form
δγ 2.5+0.4 ln(κ/κ0 )
x 2
Φγ (κ, x) = Bγ exp −sγ ln ln , (48)
x− 1 + y2
where
x − x−
y= . (49)
x+ − x−
and
1
x± = κ + r 2 ± κ − r 2 − 2r (κ − r 2 + 2r ) . (50)
2(1 + κ)
For x < x− ,
Φγ (κ, x) = Bγ [ln 2]2.5+0.4 ln(κ/κ0 ) , (51)
and, finally, for x > x+ Φγ (κ, x) = 0. The parameters Bγ , sγ and δγ as functions

of κ are tabulated in Ref. [164].
Similar approximations have been obtained in Ref. [164] also for other secondary
products. Figure 23 shows an example of energy spectra of the secondary gamma
rays, neutrinos, and electrons produced in photomeson collisions of protons of energy
E p = 1020 eV with 2.7 K CMBR. The chosen value of E p is of particular interest in
the study of the origin and propagation of cosmic rays. Due to interaction with the
CMBR, only the most energetic gamma rays created at distances less than 1 Mpc can
reach the observer. Pairs also cool rapidly in the intergalactic medium due to syn-
Fig. 23 Energy spectra of secondary products from photomeson interactions of protons of energy
E p = 1020 eV with the photons of 2.7 K CMBR. Left panel Gamma-rays and electrons; Right panel
Electronic and muonic neutrinos and anti-neutrinos
50 F. Aharonian
Fig. 24 Left panel Production spectra of electrons and positrons from direct Bethe-Heitler pair
production (curve 1) and from photomeson interactions—curve 2 (positrons) and curve 3 (electrons).
The parent proton distribution is a power-law with an exponential cutoff as described in the text.
Right panel Synchrotron (S) and IC radiation (IC) radiation spectra of cooled electrons and positrons
in a magnetic field of strength B = 1 µG. Curve 1 is the sum of the contributions of secondary
electrons, curve 2 is the direct gamma-ray spectrum from the decay of π 0 -mesons
chrotron and inverse Compton losses. And only neutrinos can propagate through the
intergalactic medium without distortion of their original energy spectra and direction.
At energies below the pion production threshold, protons continue to interact
with background photons through Bethe-Heitler pair production. Although these
interactions are not accompanied by emission of gamma rays, the secondary electrons
quickly lose their energy via inverse Compton scattering and synchrotron radiation,
and in this way result in gamma-rays. In this regard, the inverse Bethe-Heitler pair
production often operates as an indirect process of gamma-ray emission. The broad-
band gamma-ray spectra produced at interactions of ultrarelativistic protons with
2.7 K CMBR are shown in Fig. 24. It is assumed that protons are distributed as
N p (E p ) = AE −2
p exp(−E p /E 0 ) with the cutoff energy E 0 = E ∗ = m p c /4kT =
2 4
3 × 10 eV (E ∗ corresponds to the energy of protons when π-mesons are effectively

20
produced at interactions with 2.7 K CMBR). The production spectrum of pairs drops
rapidly below E e ∼ 1015 eV, however this spectral energy distribution is modified
by energy losses, to give a steady-state pair distribution Ne ∝ E e−2 . For the adopted
magnetic field B = 1 µG, the synchrotron radiation spectrum of cooled electrons has
two maximums around 1 MeV and 10 TeV corresponding to the Bethe-Heitler and
photomeson electrons, respectively. For the assumed magnetic field the contribution
of IC gamma-rays is important only above 100 TeV.
4 SNRs and Origin of Galactic Cosmic Rays
SNRs are believed to be the major contributors to the locally measured cosmic ray
(CR) flux up to the so-called “knee”, a distinct spectral feature around 1015 eV. The
main (phenomenological) argument in favour of this hypothesis is the CR production
rate in the Galaxy, ẆCR ≈ (0.3−1) × 1041 erg/s. It can be supported by SNRs if
approximately 10 % of the kinetic energy of galactic SN explosions is released in
CRs (see e.g. Refs. [96, 139, 140]). The second argument has more theoretical
background; it is linked to the potential of the so-called diffusive shock acceleration
(DSA) mechanism which may convert with a high efficiency the available kinetic
energy of bulk motion to relativistic particles.
Yet, despite the recent advances in the field, the SNR paradigm of the origin of
galactic CRs remains a hypothesis. Moreover, it is not free of ‘nasty problems’ (see
e.g. Ref. [152]). The direct measurements of cosmic rays are very important, but they
alone hardly can address all outstanding issues, in particular the ones related to the
localization and identification of the sites of particle accelerators. It is believed that
gamma-ray astronomy should play a key role in the solution of this long-standing
problem. Over the last decade, the space- and ground-based gamma-ray observations
have significantly contributed to the understanding of particle accelerators in our
Galaxy, and thus have proved the early predictions on the potential of the field in this
regard.
4.1 Gamma-Ray Signatures of SNRs
The effective acceleration of cosmic ray protons and nuclei in supernova remnants
makes them potentially detectable sources of GeV and TeV gamma-rays resulting
from hadronic interactions through the production and decay of π 0 -mesons. Thus,
a straightforward test of acceleration of cosmic rays in SNRs would be detection
of hadronic gamma-rays—directly from young remnants [124] and/or from dense
clouds overtaken by the expanding shells [63].
The theory of DSA allows definite observational predictions. In particular, the
distinct feature of shock acceleration in the non-linear regime is the concave shape
of the particle energy distribution. At low (GeV) energies it is relatively steep with
differential spectral index larger than 2, but at highest energies the spectrum becomes
very hard. In the case of strongly modified shocks, the proton spectrum, just before the
high energy cutoff, can be as hard as E −1.5 (see Ref. [190]). Generally, these features
are reflected in the spectrum of secondary gamma-rays [95, 131, 195, 254]. However,
the energy-dependent propagation effects may introduce significant modifications in
the proton spectra, in particular in the dense regions where the major fraction of
gamma-rays is produced. This concerns massive molecular clouds located outside
the mid-age SNRs [130, 137], as well as possible dense compact condensations
inside young SNR shells [254]. Therefore, the observed gamma-ray spectra could
significantly deviate from the acceleration spectrum of protons.
Presently seven shell type SNRs—Cas A [19, 28, 73], Tycho [16], SN 1006 [20],
RX J1713.7-3946 [64, 132], RX J0852-4622 [33], RCW 86 [45], and G353.6-0.7
[14] are identified as TeV gamma-ray emitters. Remarkably, while the first six sources
are well established young SNRs, the object G353.6-0.7 is the first SNR discovered
52 F. Aharonian
serendipitously in TeV gamma-ray, and only later confirmed by radio and X-ray
observations [14].
While TeV gamma-rays from young SNRs prove the effective acceleration of
CRs to energies up to 100 TeV, the relative contributions of accelerated protons
and electrons to the gamma-ray production remain unknown. The problem is that
the ratio of gamma-ray produced by accelerated protons interacting with the sur-
rounding gas and by ultra-relativistic electrons upscattering the 2.7 K MBR, is very
sensitive to generally unknown parameters, in particular to the gas density and the
magnetic field of the ambient medium. The efficiency of the inverse Compton (IC)
scattering is especially high at TeV energies (up to E e ∼ 100 TeV, it proceeds in
the Thomson regime, and therefore the cooling time tcool IC ∝ 1/E ∝ 1/E 1/2 ). For
e γ
example, the typical production times of an 1 TeV photon by an electron and a
proton of the same characteristic energy of about 20 TeV, are ≈5 × 104 year and
5 × 107 (n/1 cm−3 )−1 year, respectively (see e.g. Ref. [52]). Correspondingly, at
1 TeV the ratio of production rates of IC gamma-rays to π 0 -decay gamma-rays, is
approximately 103 (We /Wp )(n/1 cm−3 )−1 , where We and We are the total energies
in 20 TeV electrons and protons, respectively. Thus even for a very small electron-
to-proton ratio (at the stage of acceleration), e/ p = 10−3 , the contribution of the
IC component will dominate over the π 0 decay gamma-rays (in the shell with a
typical gas density n ≤ 1 cm−3 ), unless the magnetic field in the shell significantly
exceeds 10 µG. In this case, the accelerated electrons are cooled predominantly via
synchrotron radiation, thus only a small fraction, w M B R /wB ≈ 0.1(B/10 µG)−2 , is
released in IC gamma-rays. Alternatively, the proton-to-electron acceleration ratio
should exceed 103 which, in principle, cannot be excluded given the uncertainty
in one of the key aspects of DSA related to the so-called ‘injection problem’ (see
Ref. [190]).
Young SNRs are less prominent at low energies. Only Cas A [7], Tycho [141],
RX J1713.7-3946 [12], and RX J0852-4622 [228] have been detected by Fermi, and
all of them are characterized by modest GeV gamma-ray fluxes. On the other hand,
strong GeV gamma-ray emission has been reported from a number of mid-age SNRs
(see Fig. 25), in particular from W28 [6, 142], IC 433 [9, 234] and W44 [8, 143]
Remarkably, W28 [44] and IC 433 [15, 72] have been reported as VHE gamma-ray
emitters as well. In this regard one should note also that for any reasonable model
parameters, the multi-TeV particles cannot be confined in the shell of a mid-age
remnant. Therefore the gamma-ray emission is likely to be the result of cosmic rays
which already have left the remnants and interact with the nearby dense gaseous
complexes [56]. In the case of mid-age SNRs we should expect both GeV and TeV
gamma-rays from such regions [137, 237]. The detection of GeV gamma-ray emis-
sion by Fermi LAT from the molecular cloud complex that surrounds the supernova
remnant W44 is a strong evidence of realization such a scenario [239].
One should note also that while for any reasonable model parameters the TeV
particles cannot be confined in the shell of a mid-age remnant, the latter can still
contain low-energy particles. Therefore one we should expect MeV/GeV gamma-
rays not only from the regions outside the remnant, but also from the shell itself.
Generally, in the case of energy-dependent escape, the proton spectrum should be
(a) (b)
(c) (d)
Fig. 25 2–10 GeV gamma-ray images of four mid-age SNRs reported by Fermi LAT collaboration:
a W51C; b W44; c IC443; and d W28. Superposed are the contours from the VLA radio maps
(from Ref. [240])
modified, namely become steeper. This seems to be the case of W44. The gamma-
ray spectrum of 50 MeV–10 GeV gamma-rays from the shell of W44 reported by
the AGILE team hardly can be explained by bremsstrahlung or inverse Compton
scattering of electrons, but it agrees quite well with gamma-rays from pp interactions
for a steep, E −3 type proton spectrum [143].
The detection and identification of gamma-ray components produced in the
shell and in nearby clouds provide an important insight into the understanding
of acceleration and escape mechanisms of cosmic rays in SNRs. Although GeV
gamma-rays tell us only about low-energy particles, they in fact serve unique carri-
ers of information about the sites of “ancient” PeVatrons. It is important to note in
this regard that no TeV gamma-rays are expected from the shells of old and mid-age
SNRs. On the other hand, this belief should not prevent us from future searches for
TeV gamma-rays from the shells of old SNRs. Although so far the DSA mechanism
in general, and in the context of CR acceleration in SNRs in particular, seems to
work without a major problem, yet it is worth to keep in mind that DSA remains a
theoretical paradigm the predictions of which should not be overestimated as long
as it concerns the interpretation of gamma-rays from SNRs.
54 F. Aharonian
Finally, several galactic TeV gamma-ray sources spatially coincide with the
so-called composite SNRs, objects with characteristic features of both standard shell-
type SNRs and pulsar wind nebulae. At least in one case, the association of a TeV
source with the composite SNR G0.9+0.1 is clearly established [37]. The point-like
gamma-ray image with an angular size less than 1.3 indicates that TeV gamma-rays
originate in the plerionic core of the remnant, rather than in the 4 radius shell.
5 TeV Emission of Young SNRs
Both the particle acceleration and radiation processes are very sensitive to the initial
conditions of SN explosions, as well as to the parameters characterizing the sur-
rounding regions. This explains the considerable diversity in the multi wavelength
radiation properties of young SNRs reported as TeV gamma-ray sources. So far, the
most unusual representative of this class of objects is RX J1713.7-3946.
5.1 RX J1713.7-3946: An Atypical SNR
While the synchrotron radio emission and thermal X-rays are two distinct compo-
nents of shell type SNRs in general, RX J1713.7-3946 shows weak radio emission,
and no thermal X-radiation at all. On the other hand, this object is a powerful non
thermal X-ray and TeV gamma-ray emitter. The X- and VHE gamma-ray images of
this remnant are shown in Fig. 26a. The overall shell type structure and its correlation
with the non thermal X-ray image is visible, although the correlation is less evident
on smaller scales [217].
The broad-band gamma-ray spectrum of the entire remnant based on the Fermi
LAT [12] and HESS [43] measurements is shown in Fig. 26b. It extends over five
decades in energy, from 1 GeV to 100 TeV. The theoretical curves correspond to the
leptonic (IC) and hadronic (π 0 -decay) model predictions calculated within a simple
one-zone model, assuming that the GeV and TeV gamma-ray regions fully overlap.
It is seen that although both hadronic and leptonic models do satisfactorily explain
the spectral points above 1 TeV, the one-zone leptonic model fails to explain the
GeV fluxes reported by Fermi. The problem here is related to the cooling break in
the electron spectrum, and correspondingly to the position of the Compton peak
which in the spectral energy distribution (SED) appears above 1 TeV [229]. Thus,
the reduction of the break energy down to 200 GeV could in principle solve the
problem. Since the magnetic field in this model cannot significantly exceed 10 µG,
the only possibility to shift the Compton peak to sub-TeV energies is to assume that
the remnant is much older than 103 year, which however is not supported by multi
wavelength data. On the other hand, the constraints on the strength of the magnetic
field are less robust, if the IC and synchrotron components of radiation are formed in
different zones [57]. Such a scenario in young SNRs is not only possible, but, in fact,
(a) (b)
Fig. 26 Spatial and spectral characteristics of RX J1713.7-3946 at very high energies. a (left panel)
The X- and VHE gamma-ray images of RX J1713.7-3946 obtained with the ASCA and HESS
telescope array, respectively. b (right panel) The spectral energy distribution of RX J1713.7-3946
based on the Fermi [12] and HESS [43] data. The theoretical “hadronic” and “leptonic” gamma-ray
spectra calculated within a simple one-zone model are from Ref. [229]. The IC curve is obtained
for the electron spectrum derived from the synchrotron X-ray flux assuming magnetic field of
about 14 µG. The π 0 -decay gamma-ray spectrum corresponds to the spectrum of protons with the
power-law index Γ = 1.7 and exponential cutoff at 25 TeV (from Ref. [43])
can be naturally realized in the forward and inverse shocks in which the magnetic
fields are essentially different [254].
The agreement of the spectrum of hadronic gamma-rays with the measurements
over the entire GeV to TeV region can achieved assuming a very hard spectrum of
protons with power-law index 1.7 and an exponential cutoff at 25 TeV. Although this
spectrum is harder than the nominal E −2 type acceleration spectrum predicted by
the models applied to this source [95, 131, 195, 254], such a hard proton distribution
cannot be excluded. Moreover, in the case of inhomogeneous distribution of gas in
the shell, the proton spectrum in the densest regions, where the major fraction of
gamma-rays is produced, can significantly deviate, due to the propagation effects,
from the acceleration spectrum [189, 254] (see below).
The total energetics in accelerated electrons and protons in the relevant leptonic
and hadronic models of gamma-rays can be estimated by invoking minimum model
parameters. For the given distance to the source of about 1 kpc, the required budget in
electrons is determined only by the reported gamma-ray fluxes, We 3 × 1047 erg,
while the total energy budget of protons in hadronic models depends on the ambi-
ent gas density, Wp 1050 (n/1 cm−3 )−1 erg [229]. The lack of the thermal X-ray
emission from this source requires gas density as low as 0.1 cm−3 which makes the
realization of standard hadronic scenarios rather problematic [131, 163, 254]. Still,
even in the case of very low gas density of the shell, the contribution of hadronic
gamma-rays can be significant, if accelerated protons interact with the dense cores of
molecular clouds embedded in the shell [136]. The corresponding gamma-emission
may exceed the gamma-emission from the shell by a factor that is the ratio of the
56 F. Aharonian
total mass of clouds to the mass swept up by the forward shock. Thus, in the scenario
with very low gas density of the shell, in addition to the IC radiation by electrons, one
may expect significant contribution of hadronic gamma-rays produced in dense gas
condensations [254], provided that all particles freely enter the dense clouds. This
however could not be the case, especially for the low energy particles. Because of
slow diffusion, the penetration of low energy particles into the dense cores of these
condensations can take longer than the age of the SNR. Correspondingly, the low
energy gamma-ray emission can be suppressed. This effect offers a possible expla-
nation [136, 159, 254] for the hard gamma-ray spectrum below 100 GeV reported
by the Fermi collaboration from RX J1713.7-3946 [12].
The production of gamma-rays at pp interactions in dense gas condensations
embedded in the low density shell is an interesting scenario which keeps the hadronic
origin of radiation as a viable option with several attractive features. The increase of
photon statistic in future observations with Fermi and HESS should help, but hardly
could be sufficient to distinguish unambiguously the contributions of leptonic and
hadronic interactions to different bands of gamma-ray spectrum. In this regard, CTA
has a great role to play. This concerns, first of all, to the precise measurements of
the energy spectrum below 1 TeV down to tens of GeV and above 10 Tev up to
100 TeV. The morphological studies provide an independent and complementary
information about the radiation mechanism. The low magnetic field, which is a
key element of any IC model, allows the multi-TeV electrons to propagate to large
distances, and thus to fill a quite large volume. Because of homogeneous distribution
of the target photon fields, the spatial distribution of resulting IC gamma-rays appears
quite broad. The hadronic model predicts narrower and sharper spatial distribution,
mainly due to the enhanced emission in the compressed region of the shock, as it
is seen in Fig. 27. However, because of limited angular resolution of gamma-ray
telescopes, it is hard to distinguish between the radial distributions predicted by
two models. It is demonstrated in Fig. 27 where the radial profiles are smoothed
with a typical for the current Cherenkov telescope arrays point spread function of
δψ = 3 arcmin. Both smoothed profiles reasonably agree with the reported angular
distribution of TeV gamma-rays. For decisive conclusions about the sharpness of the
shell emission, and therefore on the nature of parent particles, the detector’s angular
resolution should be around 1–2 arcmin. This seems to be at the edge of capability of
the atmospheric Cherenkov imaging technique [156], but its practical realization by
CTA would greatly contribute to the understanding of the role of SNRs in the origin
of galactic CRs.
5.2 SN1006, Tycho and Cas A
SN 1006 was the first SNR from which the non thermal component of X-rays had been
unambiguously identified with synchrotron radiation of multi-TeV electrons. Like the
X-ray image, gamma-ray emission is concentrated in two extended regions in North-
East and South-West (see Fig. 28a). In this object we deal with a quite homogeneous
Fig. 27 Radial Profiles of 1 TeV gamma-rays calculated for the hadronic and electronic scenarios
[254] in the uniform medium (solid) and for the leptonic scenario with the unmodified forward
shock (dashed). The profiles smoothed with a Gaussian point spread function with σ = 0.05◦
are also shown (thin lines). The triangles correspond to the azimuthally averaged TeV gamma-ray
radial profile as observed by HESS
-6
10
(a) (b)
50
-41.6 40 10-7
F(E) (TeV-1 m-2 s-1)
30 -8 NE Region
10
20 SW Region
-41.8
Dec (deg)
10 10-9
0
-42 -10 10-10
-20
PSF 10-11
-42.2
10-12
-42.4 -13
15h04m 15h02m 10
1 10
RA (hours) E (TeV)
Fig. 28 The spatial and spectral distributions of VHE gamma-rays in SN 1006. a (right panel)
TeV gamma-ray image [20]. The white contours correspond to a constant X-ray intensity as derived
from the XMM-Newton flux map and smoothed to the HESS point spread function. b Differential
energy spectra extracted from the NE and SW regions. The shaded bands correspond to the range
of the power-law fit
environment, therefore such a morphology indicates the sites of concentration of

highest energy particles. It has been argued that a possible reason for the bipolar
picture could be the dependence of efficiency of injection of supra-thermal particles
on the angle between the ambient magnetic field and the shock normal [94].
The gamma-ray emission of SN 1006 is very weak; the integral flux above 1 TeV
is of the order of 1 % of the Crab flux. The energy spectra of the NE and SW regions
are similar and compatible with a power law distribution, F(E) ∝ E −Γ , Γ ≈ 2.3
58 F. Aharonian
(see Fig. 28b), i.e. significantly different from the energy spectrum of RX J1713.7-
3946 which is flatter around 1 TeV (Γ 2) and steeper above 10 TeV (Γ 3).
This might imply that in this case we have stronger evidence of hadronic origin
of gamma-rays than for RX J1713.7-3946. Indeed, the relatively hard power-law
spectrum of SN 1006 extending over two decades, from 0.2 to 20 TeV, significantly
deviates from general predictions of IC models. The so called one-zone model which
assumes a common region of production of the IC and synchrotron components of
radiation, satisfactorily explains the gamma-ray flux around 1 TeV (for a reasonable
magnetic field of order of 30 µG), but fails to explain the hard power-law spectrum
at multi-TeV energies [20]. However, the one zone-model is a simplification [57]
which, in fact, hardly can be realized in SNRs. The difference of the magnetic field
in the upstream and downstream regions results in the positional shift of production
regions of synchrotron X-rays and IC gamma-rays. While X-rays are predominantly
produced in the downstream region, IC gamma-rays are contributed by both the
downstream and upstream regions, with a strong dominance of the latter at multi-
TeV energies [253]. This leads to a shift of the overall gamma-ray spectrum towards
higher energies.
Alternatively, the existing VHE gamma-ray data can be explained by interactions
of accelerated protons with power-law distribution E −2.3 between 1 and 100 TeV.
Below 1 TeV, the proton spectrum should become flatter than E −2 . This is a rather
robust conclusion opposed by the available energy budget, given that for the ambient
gas density n ∼ 0.1 cm−3 , the required acceleration efficiency of protons in the
energy interval 1–100 TeV already exceeds 20 % of the total mechanical energy of
the SN explosion, E SN = 1.4 × 1051 erg [20]. The proton spectrum can be described,
for example, by a power low E −2.3 with a break below 1 TeV, and a high energy
cutoff at E 100 TeV. This is an interesting option because it requires effective
acceleration of protons to extremely high energies, formally to 1 PeV. However,
there is an alternative option when the proton spectrum is described by a flat E −2
type spectrum with an exponential cutoff at E 80 TeV [20]. The extension of
gamma-ray measurements by CTA to energies well beyond 10 TeV should allow
us to distinguish between these two realizations, and thus to answer to a principle
question as whether SN 1006 acts as a PeVatron.
Gamma-ray signals have been reported, both at high and very high energies
(see Fig. 29), from another prominent object, the Type Ia supernova remnant Tycho
[16, 141]. While the Fermi data between 0.4 and 100 GeV can be described by a
power-law with a photon index ΓGeV ≈ 2.3, the spectrum reported by the VERI-
TAS collaboration between 1 and 10 TeV is somewhat harder, ΓTeV ≈ 2. Within the
uncertainties of the ambient gas density and the distance to the source, the reported
fluxes at both GeV and TeV energies agree with the early phenomenological predic-
tions [124], as well as with the recent theoretical studies [196, 244] of production of
hadronic gamma-rays in Tycho. The flat gamma-ray spectrum up to 10 TeV implies
that the corresponding spectrum of parent protons continues without a significant
steepening or a cutoff to at least several hundred TeV [165]. The extension of the
flat gamma-ray spectrum just by a factor of two or three beyond 10 TeV, would be
sufficient to claim the source as a PeVatron. Such observations also would robustly
Fig. 29 The reported gamma-

ray emission of Tycho from
GeV [141] to TeV [16] ener-
gies. The expected contribu-
tions from the pp-interactions,
(dot-dashed line), relativistic
bremsstrahlung (dot-dot-
dashed) and Inverse Compton
scattering (on three differ-
ent radiation fields) are also
shown (from Ref. [196])
exclude the IC origin of radiation. Note that although the present data give a pref-
erence to hadronic models [196] (see Fig. 29), yet a specifically designed two-zone
leptonic model cannot be discarded as a possible scenario for explanation of the
current gamma-ray data from Tycho [80].
The shell type supernova remnant Cas A is one of the best studied non thermal
objects in our Galaxy. Although some of its general features are common for young
SNRs, Cas A is a rather unique representative of the remnants of recent supernovae
explosions. Its synchrotron emission spans from radio to hard X-rays indicating the
presence of relativistic electrons from sub-GeV to multi-TeV energies. For any rea-
sonable assumption on the nebular magnetic field, this object contains enormous total
energy in the form of relativistic electrons, We 3 × 1048 erg [85]. The average rate
of accumulation of this energy over a short time period, tacc ≤ tage ∼ 300 year, is even
more striking, Ẇe = We /tacc ≈ 3×1038 erg/s. It is larger, by at least an order of mag-
nitude, than the electron production rate in any other supernova remnant. On the other
hand, the content of protons in this object could be relatively modest. As discussed
below, the amount of relativistic protons and nuclei is constrained by gamma-ray
fluxes detected in the GeV and TeV energy bands, Wp ≤ 3 × 1049 (n/10 cm−3 )−1
erg [7]. This constitutes less than 2 % of the total explosion energy, if gamma-rays are
produced in the reverse shock where the plasma could be quite dense, n ≥ 10 cm−3 .
In this case, the ratio of relativistic protons to electrons in Cas A is less than 10, i.e. an
order of magnitude below the level observed in cosmic rays. This is in sharp contrast
to the hadronic models of γ-radiation of other SNRs, e.g. SN 1006 and RX J1713.7-
3946, which require p/e ≥ 103 . However, if the reported GeV and TeV gamma-ray
fluxes are produced at hadronic interactions in the forward shock, which propagates
through a low density circumstellar medium, the total energy in accelerated protons
can exceed 1050 erg.
Another unique feature of this object is the reported gamma-ray [160] and
X-ray [213] emission lines which associate with 44 Ti. This radiation component
provides direct information about the ejected mass of radioactive titanium-44,
M44Ti ≈ 2 × 10−4 M . It has been recently suggested [255] that the unusually
60 F. Aharonian
high content of relativistic electrons in Cas A could have a link to the ejection of
large amount of radioactive material, first of all 44 Ti and 56 Ni. The decay products
of these nuclei provide a vast pool of supra-thermal positrons and electrons which
can be further accelerated, by both the reverse and forward shocks, to multi-TeV
energies.
The flux measured by Fermi in the 0.3–30 GeV interval is flat with a slope
Γ = 2.0 − 2.2 [7]. The VHE gamma-ray signal from Cas A has been discovered in
the 1–10 TeV energy interval by the HEGRA collaboration [28], and confirmed by
the MAGIC [73] and VERITAS [19] collaborations. The gamma-ray spectrum is com-
patible with power low, d N /d E ∝ E −Γ with a photon index Γ = 2.6 and total flux
above 1 TeV Fγ (≥1 TeV) = 7.7 × 10−13 ph/cm2 s [19]. The GeV and TeV electrons
which are responsible for the broad-band synchrotron radiation of Cas A, inevitably
produce also gamma-rays—through bremsstrahlung and inverse Compton scattering.
The calculations of gamma-ray fluxes based on the radio data are straightforward,
but strongly depend on distributions of the gas and magnetic field in the nebula.
Since the average magnetic field in Cas A cannot be significantly less than 0.3 mG
(otherwise the contribution of bremsstrahlung would lead to overproduction of low-
energy gamma-rays [85]), the TeV gamma-ray emission of IC origin its expected to
be very low, unless one invokes regions with low magnetic field but yet with adequate
conditions for effective acceleration of electrons to multi-TeV energies. Although at
first glance this sounds a rather superficial assumption, the regions with very low
magnetic field in SNRs cannot be excluded. Moreover, in the case of Cas A this can
be realized in a quite natural way, through the operation of the reverse shock.
On the other hand, the overall GeV–TeV gamma-ray spectrum can be readily
explained by interactions of accelerated protons and nuclei with the ambient gas,
assuming a power-law spectrum of protons extending to 100 TeV with a power-law
index α = 2.3, or by a harder spectrum with α = 2.1 but suffering an exponential
cutoff at 10 TeV [7] (see Fig. 30). This ambiguity, which is a result of large statistical
uncertainties of TeV gamma-ray fluxes, leaves open the question whether Cas A
accelerate particles to PeV energies. As discussed before, this is true also for other
SNRs. Meanwhile, without a clear answer to this question, we cannot be sure that
SNRs are the major contributors to the galactic cosmic rays up to the “knee” around
1 PeV. The “hunt” for galactic PeVatrons continues.
5.3 Radiation Signatures of Proton PeVatrons
The most straightforward search for galactic PeVatrons can be conducted by gamma-
ray detectors designed for operation in the energy regime between 10 and 100 TeV.
In SNR shocks with relatively low acceleration rate, the synchrotron losses prevent
acceleration of electrons to energies beyond 100 TeV. Also, at such high energies the
contribution of the IC component is suppressed because of the Klein-Nishina effect.
Therefore, the contribution of the IC gamma-rays to the radiation above 10 TeV is
Fig. 30 The differential energy spectra of gamma-rays reported from Cas A by the Fermi, MAGIC
and VERITAS collaborations (from [7]. Two curves correspond to calculations of theoretical gamma-
ray spectra from pp interactions assuming for protons (i) a single power low distribution with an index
α = 2.3 (red line) (ii) a power-law distribution with α = 2.1 an exponential cutoff at E 0 = 10 TeV
expected to gradually fade out. Thus, in the case of detection of gamma-rays up to

100 TeV, the hadronic origin of radiation would be unambiguously established.
Figure 31 shows the X-ray and gamma-ray luminosities from a 1000 year old
proton PeVatron calculated for three different distributions of accelerated protons.
Both radiation components are initiated by interactions of accelerated protons for the
ambient gas density n = 1 cm−3 and magnetic field B = 300 µG. While gamma-rays
(a) (b)
Fig. 31 The broad-bad radiation of a PeVatron initiated by interactions of protons with the ambient
gas. a (left panel) Three different distributions of protons: 1,2—“power-law with an exponential
cutoff”, E −α exp(−E/E 0 )β with α = 2, E 0 = 3 PeV, and β = 1 (solid curve), β = 1/3 (dashed
curve), and 3—“broken power-law” when the spectral index is changed at E = 1 PeV from α = 2
to α = 3. b (right panel) Luminosities of brand-band emission produced at pp interactions. The
gas density n = 1 cm−3 , magnetic field B = 300 µG, and the age of the source t = 103 years
62 F. Aharonian
arise directly from decays of π 0 -mesons, X-rays are result of synchrotron radiation of
secondary electrons, the products of π ± -decays. The lifetime of electrons producing
−3/2
X-rays, tsynch 1.5BmG (E X /1 keV)−1/2 year, is very short (≤50 year) compared
to the age of the source. Therefore, X-rays could be treated as a “prompt” radiation
emitted simultaneously with gamma-rays.
The X- and gamma-ray fluxes depend on the present content of total energy
of accelerated protons accumulated in the source, and on the density of the ambient
matter. Approximately the same fraction of energy of the parent protons is transferred
to secondary electrons and gamma-rays. However, since the energy of sub-TeV elec-
trons is not radiated away effectively, the direct (π 0 -decay) gamma-ray luminosity
exceeds the synchrotron luminosity. The L X /L γ ratio depends on the proton spec-
trum as well as on the particle injection history; typically it does not exceed 0.2–0.3.
The spectrum of highest energy gamma-rays contains an important information
about the shape of the proton spectrum around the cutoff E 0 which is crucial for
identification of acceleration mechanisms in SNRs, as well as for understanding
of the role of different processes responsible for the formation of the knee in the
CR spectrum. The X- and gamma-ray luminosities in Fig. 31 are calculated for a
proton accelerator operating during 103 year with a constant rate L p = 1039 erg/s;
the total energy in protons is Wp = L p · T 3 × 1049 erg. To estimate the X-and
2
gamma-ray energy fluxes (in units of erg/cm s) from an arbitrary PeVatron, one
should multiply the luminosities in Fig. 31b by the factor κ ≈ 10−44 (nWp /3 ×
1049 erg/cm3 )(d/1 kpc)−2 . Thus, all galactic PeVatrons up to distances of 10 kpc
and with nWp ≥ 1049 erg/cm3 , can be probed by CTA, the sensitivity of which
above 10 TeV is expected to be better than 10−13 erg/cm2 s [23]. However, because
of tiny gamma-ray fluxes, detailed spectroscopic measurements in the cutoff region
could be possible only for powerful and/or nearby objects. Note that the extension
of the proton spectrum to 1 PeV is crucial for effective production of neutrinos in the
10–100 TeV range, the most optimal energy interval for the TeV neutrino detectors
like IceCube or KM3NeT [243]. However, the sensitivity of neutrino detectors is
quite limited, and even the brightest in gamma-ray SNRs can be only marginally
detected by these instruments.
In such circumstances, the search for the PeVatrons via synchrotron X-radiation
of secondary (π ± -decay) electrons has been suggested as an alternative tool [52].
In principle, Chandra and XMM-Newton have sufficient sensitivity to perform such
studies. However, the limited effective energy domain (≤10 keV) of these instruments
is not optimal for detection of synchrotron radiation of secondary electrons. Indeed
the major challenge of this method is the extraction of the “hadronic” component of
X-rays from the synchrotron radiation of directly accelerated electrons. These two
components can be separated if the magnetic field in the SNR exceeds 100 µG and the
proton spectrum extends to 1 PeV. These two conditions are, in fact, connected since
the acceleration of protons in SNRs to PeV energies is possible only at the presence of
large magnetic fields. The second key condition for operation of SNRs as PeVatrons is
the diffusion in the Bohm limit. In this case, the proton cutoff energy is proportional to
the strength of the magnetic field. Thus, the corresponding energy in the spectrum of
secondary synchrotron radiation hν ∝ B E 02 ∝ B 3 . On the other hand, the position of

the cutoff of synchrotron radiation of directly accelerated electrons does not depend
on the magnetic field and typically appears in the soft X-ray domain, hν ≤ 1 keV
(see e.g. [57]). Thus, if the spectrum of synchrotron radiation of secondary electrons
extends well beyond 10 keV, the background caused by the synchrotron radiation of
directly accelerated electrons is dramatically reduced.
The spectrum of the secondary synchrotron radiation in the cutoff region is
smoother and broader than the gamma-ray spectrum around the cutoff. For the proton
spectrum written in a general form (d N /d E)p ∝ exp[−(E/E 0 )β p ], the distributions
of secondary gamma-rays and electrons in the cutoff region is (d N /d E)γ/e ∝
exp[−(E/E 0,γ/e )βγ/e ], with βγ/e ≈ 0.5β p [165]. The spectrum of the synchrotron
radiation in the cutoff region is described by a function proportional to
exp[−(ε/ε0 )βs ], with βs = βe /(2 + βe ) [253]. For example, in the Bohm diffusion
regime, β p = 1, thus in the cutoff region the distribution of gamma-rays is pro-
portional to exp[−(E γ /E γ/e )1/2 ], while the spectrum of the secondary synchrotron
radiation behaves as exp[−(ε/ε0 )1/5 ]. This is an important feature (see Fig. 31),
which should allow detailed studies of the proton spectra around and beyond the
cutoff with the planned hard X-ray imagers of the NuSTAR and ASTRO-H X-ray
satellites.
5.4 Expectations from Future Studies
As discussed above, the gamma-ray emission from several famous representatives

of young SNRs can be explained by interactions of shock accelerated protons and
nuclei. However, this cannot be taken as a proof of the major contribution of SNRs
to the production of galactic cosmic rays because the data can be explained also
by the competing leptonic processes. Moreover, we do not have a single example
which would be considered as an evidence of accelerations of protons up to 1 PeV.
A natural reason for the deficit of such energetic particles could be their leakage from
the shell [114, 254]. Indeed, the acceleration and confinement of multi-TeV particles
in the remnants can last less than several hundred years after the explosion, so one
should be lucky to “catch” a SNR while it is still accelerating particles to multi-TeV
energies. This significantly constraints the number of SNRs emitting very high energy
gamma-rays. Moreover, the sensitivity of current IACT arrays limits the detection
of TeV gamma-rays by young SNRs located within 1–2 kpc distances. An increase
of the ‘detectability distance’ by an order magnitude would be a great achievement,
since it would make visible most of young SNRs in our Galaxy. The inner parts
of the Galaxy with enhanced star formation rate are of special interest. Because of
reduction of gamma-ray fluxes proportional to the distance squared, the increase of
the distance by an order of magnitude formally demands an improvement of the
sensitivity for point-like sources by two orders of magnitude. But, fortunately, in the
case of young SNRs an order of magnitude improvement of the sensitivity would be
sufficient. Indeed, the sensitivity of IACT arrays for extended sources is proportional
64 F. Aharonian
to the angular size of the source, Jmin ∝ Ψ , as long as Ψ is larger than the angular
resolution of the IACT arrays, δθ ≈ 0.1◦ . On the other hand, the angular sizes of
young SNRs scales with the distance approximately as Ψ ∼ 1◦ (d/1 kpc)−1 . Thus, the
expected improvement of the sensitivity of CTA by an order of magnitude compared
to the current IACT arrays, will dramatically increase the number of potentially
detectable SNRs. This should allow compelling population studies—a key issue for
the proof of the SNR origin of galactic cosmic rays.
The acceleration of protons in SNRs to multi-TeV and PeV energies can be studied
also in an indirect way using the gamma-ray “echos” of old, faded out accelerators.
In certain environments, TeV gamma-rays can be detected even ten thousand years
after the so-called Sedov phase when all particles already have left the remnant.
Indeed, the particles, after they escape the SNR shell, interact with the surrounding
atomic and molecular gas. Before being fully diffused away and integrated into the
“sea” of galactic CRs, these particles produce gamma-rays the spectrum of which can
significantly differ from both the radiation of the SNR shell and the diffuse galactic
gamma-ray emission [56]. The massive molecular clouds located in the vicinity of
the supernova remnant, provide dense targets for hadronic interactions, and thus
dramatically increase the chances of tracing the run-away protons via the secondary
gamma-rays. For parameters of a typical SNR at a distance of 1 kpc, a molecular
cloud of mass 104 M can emit multi-TeV emission at a detectable level, if the cloud
is located within 100 pc from the SNR [137]
The location of molecular clouds close to SNRs could be accidental, but in general
there is a deep link between SNRs and MCs, especially in the star-forming regions
[194]. Depending on the location of massive clouds, the time of particle injection
into the interstellar medium, as well as on the diffusion coefficient, we might expect a
broad variety of energy distributions of gamma-rays—from very hard spectra (much
harder than the spectrum of the SNR itself) to very steep ones [56]. Correspondingly,
the ratio of GeV to TeV gamma-ray fluxes can significantly vary from site to site.
One should note that the role of giant molecular clouds is not limited by SNR
studies. These massive objects are intimately connected with the star formation
regions that are strongly believed to be most probable sites of CR production (with
or without SNRs) in our Galaxy. They serve also as a unique “barometers” for mea-
surements of the energy density of cosmic rays in remote parts of the Galaxy [113].
It is generally believed that the local CR flux directly measured at the Earth, gives
a correct estimate for the level of the “sea” of galactic CRs. However, one can-
not exclude that the flux of local CRs could be dominated by a single or few local
sources, especially given that the Solar system is located in a rather extraordinary
region—inside active star formation complexes which constitute the so-called Gould
Belt. The recent anomalies discovered in cosmic rays, such as very high content of
positrons in the leptonic component of cosmic rays [24], or the significant differences
between energy spectra of protons and alpha particles [25], tell us that the generally
adopted picture of homogeneously distributed galactic CRs contributed by a single
class of accelerators (SNRs) could be an oversimplification. In fact, we may deal
with a diverse variety of cosmic ray accelerators. The fortunate location of giant
molecular clouds in the vicinity of particle accelerators provides us with a unique

tool to probe cosmic rays in these environments.
Despite the remarkable advances of recent years, we do not have yet a robust evi-
dence of acceleration of nucleonic component of cosmic rays in supernova remnants.
It is expected that explorations of these objects with the major future ground-based
gamma-ray detector, the CTA, with significantly improved performance regarding the
minimum detectable flux, the angular resolution, and broader (almost five decades!)
energy coverage, will remove, to a large extent, many current uncertainties and
ambiguities concerning the origin of gamma-rays from SNRs. A breakthrough in
this regard is especially expected in the unexplored so far energy interval above
30 TeV. Indeed, the contribution of the inverse Compton component of radiation
at these energies is largely suppressed, thus there will be little doubt that we deal
with gamma-rays of hadronic origin which would carry direct information about
protons accelerated up to 1 PeV. Also, the best angular resolution of CTA (as good as
1–2 arcmin; J. Hinton, private communication) is expected at energies above 10 TeV.
This is very important for the study of several key, but so far poorly understood issues
related, for example to the escape of highest energy particles from the acceleration
zone. In general, the discovery of the first cosmic PeVatrons, can be considered as
one of the highest priority objective of CTA. The galactic PeVatrons are guaranteed
sources of neutrinos in the 10–100 TeV range, the most optimal energy interval
for their registration by the km3 scale detectors like IceCube and KM3NeT. Unfor-
tunately, because of the limited sensitivity, even the most promising young SNRs
suspected to be the brightest neutrino sources in our Galaxy in the multi-TeV band,
can be robustly detected only after 5–10 years of continuous exposure. Together
with studies of ultra-high energy gamma-rays and neutrinos, the future hard X-ray
imaging detectors like NuSTAR and ASTRO-H, should be able to conduct an effec-
tive search for the currently active PeVatrons through the synchrotron radiation of
secondary electrons produced at hadronic interactions. Finally, the coverage of the
GeV energy domain by the Fermi space telescope is very important, especially for
the search of ‘smoking guns’ of former PeVatrons, as well as for exploration of the
environments which harbour these mysterious cosmic ray “factories”.
Of course, the search for PeVatrons should not necessarily be linked to SNRs.
Some other classes of sources, e.g. superbubbles [151, 205] or remnants of gamma-
ray bursts in our Galaxy [79] have been proposed as effective accelerators of particles
to PeV energies and potential contributors to galactic cosmic rays. Moreover, the
galactic plane survey by CTA might result in an exciting discovery of a new class
of E 10 TeV gamma-ray sources of unknown origin. The outcome could be a
dramatic revision of the concept of origin of galactic cosmic rays without a major
contribution from SNRs.
66 F. Aharonian
Fig. 32 Energy spectrum of the source in the GC based on the combined Fermi (green points) and
H.E.S.S. (blue points) data. The curves correspond to the calculated contributions from a 10-day
proton flare that occurred 300 years ago (dashed line) and from a constant proton source that switched
on 104 years ago (dotted line). The solid curve is the superposition of these two components. The
upper and lower dashed lines correspond to the 100 and 50 TeV cut-offs in the proton spectrum
while the solid curve assumes E cut = 75 TeV
6 Galactic Center
The Galactic Center (GC) is an extraordinary site that harbours many remarkable
objects, in particular the compact radio source Sgr A*, a suspected super-massive
black hole (SMBH) located at the dynamical center of the Galaxy. Over the last four
decades this regions has been repeatedly observed by different gamma-ray detectors.
GC contains a bright gamma-ray source with a broad-band spectrum that spans from
100 MeV [117] to 30 TeV [46]. Spatially it coincides with the position of Sgr A*
(see Fig. 8). However the upper limit on the angular size of a few arcminutes is still
large to exclude any other sources located within the central ≤10 pc region. Unlike
the radio and X-ray bands, no variability has been observed both at GeV and TeV
energies. This disfavours but still cannot discard Sgr A* as a possible gamma-ray
source. Further constraints on the gamma-ray emission models come from the energy
spectrum which has an interesting shape; at low energies it is hard with a photon
index Γ ≈ 2.2. Above 2 GeV the spectrum becomes stepper by ΔΓ ≈ 0.5 [117],
but at TeV energies it hardens again with a photon index Γ 2.1 and an apparent
break or a cutoff above 10 TeV (see Fig. 32).
6.1 Sgr A*
The temporal and spectral features of Sgr A* are unusual and, as a whole, very
different from other compact galactic and extragalactic black hole candidates. This
concerns, in particular, the extraordinary low luminosity of Sgr A*. In addition
to other important astrophysical implications, the low luminosity of Sgr A* has

a dramatic effect on the visibility of the source in gamma-rays [48]. Because of
internal photon–photon pair production, the high energy gamma-ray emission of
black holes is strongly suppressed. However this is not the case of Sgr A* which
makes it (together with the black hole in the radio galaxy M87) a unique laboratory
for study of possible particle acceleration processes in the accretion flows near the
event horizon of black holes. High energy gamma-rays from compact regions close
to the SMBH can be produced in various ways.
• Synchrotron and curvature radiation of protons
Gamma-ray production at interactions of protons with magnetic field are effective
only in the so-called extreme accelerators [61] where particles are accelerated at the
maximum possible rate, Ė = eB. Such an effective acceleration can be realized in
the vicinity of SMBHs [181]. Even for an “ideal” combination of parameters allowing
the most favourable acceleration/cooling regime, the characteristic energy of syn-
chrotron radiation is limited to max = (9/4)α−1 f m pc
2 0.3 TeV. This implies
that the proton-synchrotron radiation cannot explain the flux observed from the
direction of GC since it would require a source moving with unrealistically large
Doppler-factor, D ≥ 30. More promising seems to be the curvature radiation [181]
when the protons emit while moving along the regular magnetic field lines. For the
SMBH in GC, 3 × 106 M , max = 3E 3p /2m 3 R 0.2(B/104 )3/4 TeV. Formally,
this allows extension of the spectrum to 10 TeV, provided that the magnetic field
exceeds B 106 G. However, such a strong field would make the source opaque for
E ≥ 1 TeV gamma rays [48].
• Photomeson interactions of protons
Close to the event horizon of SMBH in GC protons can be accelerated to energies
E ∼ 1018 eV [61, 181] and interact effectively with the radiation of the compact
infrared source. The mean free path of protons through the photon field is estimated
−1
pγ ∼ (σ pγ f n ph ) 1015 (RIR /1013 cm)2 cm. This means that approximately
R/ pγ ∼ 0.01 fraction of the energy of protons is converted into secondary parti-
cles [48]. However, while neutrinos and neutrons, as well as gamma-rays of energies
below 1012 eV escape freely the emission region, TeV gamma-rays effectively inter-
act with the ambient photon and magnetic fields, and thus initiate IC and/or (depend-
ing on the strength of the B-field) synchrotron cascades. Gamma-rays produced in
this way can explain the observed absolute TeV flux, if the acceleration power of
1018 eV protons is about 1037 erg/s. However the energy spectrum of the cascade
gamma-rays does not agree with the observed overall GeV–TeV energy spectrum.
• Proton–proton interactions
While acceleration of protons to energies E ∼ 1018 eV requires extreme conditions
which not always are satisfied, the boost to TeV energies can be more routinely
realized. In these cases, interactions of protons with the ambient thermal gas become
the main source of gamma-ray production. Protons can be accelerated to TeV energies
also in the accretion disk, e.g. through strong shocks developed in the accretion flow.
However, the efficiency of gamma-ray production, which is determined by the ratio
68 F. Aharonian
of accretion time to the pp cooling time, appears less than 10−4 . Correspondingly,
the required current acceleration rate of high energy protons L p ≥ 1039 erg/s is
unacceptably high for GC. This makes unlikely the gamma-ray production via pp
interactions.
• Gamma rays of leptonic origin
The models of gamma-ray emission associated with accelerated protons provide
rather modest efficiencies of conversion of the energy of accelerated protons to
gamma-rays. The radiative energy loss rate of electrons is much higher, and there-
fore the models that link gamma-rays to accelerated electrons provide more eco-
nomic ways of gamma-ray production. Obviously, these electrons should be accel-
erated to at least E max 10 TeV. This immediately constrains the strength of the
chaotic component of the magnetic field; even under an extreme assumption that the
acceleration proceeds at the maximum possible rate, (d E/dt)acc eB, one gets
B ≤ 10(E max /10 TeV)−2 G.
The requirement of particle acceleration at the maximum rate imposes strong
restrictions on the geometry of magnetic field and possible acceleration mechanisms.
In this regard, acceleration in ordered electric and magnetic fields, may provide cer-
tain advantages. Moreover, in the ordered field, the energy dissipation of electrons
is reduced to the curvature radiation loses. This allows significant gain in the energy
of accelerated electrons, up to E e,max 100(B/10G)1/4 TeV. In this scenario, elec-
trons radiated through curvature and inverse Compton mechanisms. The curvature
radiation peaks at curv 200(E e /1014 eV)3 MeV. The Compton scattering of same
electrons leads to the second peak at much higher energies, E γ ∼ E e 100 TeV
(because the scattering proceeds in the Klein-Nishina limit). However, because of
interactions with infrared photons, gamma-rays of energy exceeding 10 TeV can
not freely escape the source. Synchrotron radiation and Compton scattering of the
secondary (pair-produced) electrons lead to re-distribution of the initial gamma-ray
spectrum. The results of calculations strongly depend on the maximum energy of
electrons, therefore the interpretation of the observed gamma-ray spectrum from
GeV to TeV energies requires fine-tuning of the model parameters. What concerns
the acceleration rate of electrons, it could be quite reasonable—as low as 1037 erg/s.
In this regard, the major problem of this interpretation is the compact production
region and very short acceleration and radiative cooling times. They naturally imply
fast variability which has not been yet detected in both GeV and TeV energy bands.
Although this cannot yet rule out the gamma-ray production in the vicinity of SMBH,
the lack of gamma-ray variability of the source is, of course, a serious argument
against Sgr A* as the source of gamma-radiation of GC. However, Sgr A* remains
a potential source to be (indirectly) responsible for the gamma-ray signal through
interactions of runaway particles accelerated in Sgr A*, but later injected into the
surrounding dense gas environment [49, 183, 184].
Fig. 33 Energy spectra of gamma-ray emission from GC. The Fermi and HESS data are shown
together with calculations of gamma-rays from pp interactions within radial cones of various size
up to 50 pc [183]. The flux falls off rapidly after 3 pc because the main contribution comes from
the 1.2–3 pc circum-nuclear ring
6.2 Diffuse Gamma-Ray Emission from the Central

10 pc Region
In the scenario of acceleration of protons in Sgr A* and their subsequent injection

into the interstellar medium, the total gamma-ray flux, as well as the spectral and
angular characteristics of π 0 -decay gamma-rays significantly depend on the proton
injection rate and the regime of their diffusion. In particular, it has been realized [49]
that for a certain combination of model parameters the energy spectrum of resulting
gamma-rays can contain several spectral features related to the effects of proton
propagation. The analysis of the combined Fermi LAT and HESS data show that
the complex shape of the GeV–TeV radiation can be indeed naturally explained by
the propagation effects of protons interacting with the dense gas within the central
10 pc region [117, 183]. The good agreement between the data and calculations
are shown in Fig. 33 where the radial profile of the gas density has been carefully
taken into account. The flat spectra in the segments of the proton spectrum around
1 GeV, and at TeV energies (below 10 TeV) have different explanations. While at
GeV energies the protons are diffusively trapped, so they lose a large fraction of
their energy before they leave the dense 3 pc region, at TeV energies they propagate
rectilinearly. At intermediate energies the protons start to effectively leave the inner
70 F. Aharonian
3 pc region, and the steepening of the energy spectrum can be naturally referred
to the energy dependent diffusion coefficient. What concerns the proton injection
spectrum, it should be hard power-law, close to E −2 , with an intrinsic cutoff around
100 TeV. The required total energy of protons currently trapped in the gamma-ray
production region, Wp L γ tpp→γ 1049 (n/10−3 cm−3 )−1 erg is quite modest,
given that the density in the circum-nuclear ring could be as large as 105 cm−3 [183].
The total injection power of protons depends on the gas density and the diffusion
coefficient; in the case of most economic realization of the scenario, the average rate
of proton injection over the 103 year could be close to 1037 erg/s.
It is remarkable that this simple scenario allows a successful interpretation of the
complex spectrum of gamma-rays extending over 5 decades from 0.3 GeV to 30 TeV,
based only on an assumption of a single component of protons and a small number of
reasonable model parameters. This, however, does not exclude other possible expla-
nations, e.g. by two different components of protons, e.g. by a short flare that took
place some 300 years ago on top of a long-term quasi-continuous proton injection
(see Fig. 32). Moreover, diffuse gamma-ray emission can be produced also by high
energy electrons.
In this regard, one should mention the leptonic models of extended (although
not resolvable) sources of gamma-ray emission in GC. In particular, an interesting
idea of formation of a hypothetical “black-hole plerion” powered by the advection-
dominated accretion flow (ADAF) has been suggested in Ref. [81]. It has been shown
that the electrons accelerated to Lorentz factor γ ≥ 108 at the termination shock at
0.01 pc from the SMBH in GC can in principle reproduce the HESS data. However,
the GeV emission predicted by this model appears well below the flux detected by
Fermi LAT.
The recently discovered synchrotron nebula G359.95-0.04 located at 8 arcsecond
from Sgr A* is another interesting site where TeV gamma-rays can be produced
[153, 246]. Because of the high radiation density, the radiative cooling of electrons in
this nebula is dominated by intense Compton scattering in the Klein-Nishina regime;
this results in formation of an electron energy distribution which differs dramatically
from those produced in PWNe located in the galactic plane [153]. Correspondingly,
the broad-band radiation of electrons consisting of synchrotron and IC components
also is characterized by a quite non-standard spectrum. It can reproduce the TeV data,
however fails to explain the flux at GeV energies, unless we postulate existence of an
additional, low energy component of electrons, or assume that the GeV gamma-ray
emission belongs to a different source. In this regard, one should mention that recently
a new leptonic model has been suggested [179] in which the electrons accelerated
during flares in Sgr A*, are accumulated in the central ≤1 pc region around the
super massive black hole. Since multi-TeV electrons cannot escape Sgr A*, the IC
radiation of this region cannot be extended beyond the GeV domain.
Note that although all these leptonic models imply gamma-ray production in
extended regions, they are still too compact to be resolved by current gamma-ray
detectors. On the other hand, in the hadronic models the bulk of gamma-ray emission
is produced in the circum-nuclear region of radius 3 pc, which for the distance to
the galactic center of 8 kpc implies an angular size slightly more than 1 arcmin, i.e.
quite close to the angular resolution of CTA expected at energies above 10 TeV.
The detection of a 1 arcmin TeV gamma-ray source in GC would prove the hadronic
origin of gamma-ray emission and its connection to protons accelerated in Sgr A*.
An independent argument in favour of Sgr A* as a source of multi-TeV protons is
provided by the HESS observations of the huge molecular clouds located within
100 pc around GC.
Gamma Rays from the Central Molecular Zone
The interpretation of the spatially unresolved gamma-ray emission towards Sgr A*

by interactions of runaway protons with the dense gas in the central several pc ring,
implies no-sharp edges of the gamma-ray production region but rather a smooth
transition to a new radiation component formed in more extended regions of GC.
The energy and spatial distributions of this radiation depend on the injection history
of protons and the character of their diffusion. For the given injection rate, slower
diffusion allows more effective accumulation of protons, and correspondingly higher
gamma-ray luminosity. And vice versa—fast diffusion may dramatically reduce the
gamma-ray luminosity down to the level below the detection threshold even for very
high injection rate of protons.
The diffusion time of protons increases with the distance as t ∝ r 2 /D (D is
the diffusion coefficient), thus the time-scales that characterize the processes of CR
accumulation in the so-called Central Molecular Zone (CMZ) of radius ≈200 pc
are much longer than in the inner 3 pc ring. Generally, if the accelerator is located
in Sgr A*, the proton density is expected to fall with the distance from the galactic
center. However, in the case of rectilinear propagation of particles (as expected at
highest energies), at certain epochs after strong flares in Sgr A* we may expect, in
principle, higher CR fluxes at larger distances compared to the inner parts of GC.
In any case, independent of the spatial distribution of CRs, the large-scale diffuse
gamma-ray emission from CMZ is expected to be highly inhomogeneous, because it
largely reflects the distribution of the gas concentrated primarily in giant molecular
clouds.
The HESS observations of the galactic center region indeed revealed an extended
TeV gamma-ray emission from CMZ [39] with a clear indication of correlation
with the most prominent giant molecular clouds (see Fig. 8). Using the maps of
TeV gamma-ray emission obtained with HESS, and the maps of CS (J = 1 − 0)
emission which contains information about the column density in dense cores of
molecular clouds, we can “measure” the density of CRs in CMZ. In particular,
assuming that the locally measured CR spectrum is valid in the GC, we can calculate
the expected π 0 -decay γ-ray flux from CMZ. It is seen from Fig. 34 that the gamma-
ray spectra measured by HESS from the overall 200 pc region, as well as from
individual clouds like the Sgr B2 complex, are significantly harder than the predicted
one. The absolute CR flux derived from gamma-ray observations, also appears to be
significantly enhanced (by an order of magnitude at multi-TeV energies) in this region
relative to the local cosmic ray flux in the solar neighbourhood.
72 F. Aharonian
Fig. 34 Gamma ray flux per unit solid angle from different parts of the GC region (data points),
compared with the contribution of gamma-rays expected from interactions of galactic cosmic rays
[39]. The grey shaded zone shows the flux expected from CMZ within 200 pc due to hadronic
interactions assuming for cosmic rays flux the local one as measured in the solar neighbourhood.
The overall gamma-ray flux from the region of CMZ (full circles) consists of contributions from
several clouds, one of which, the Sgr B2 complex, is shown separately (open circles). The flux of
the central source (HESS J1745-290) towards Sgr A* is also shown (using an integration radius of
0.148)
This indicates to a strong nonthermal activity accompanied with proton acceler-

ation which perhaps in the past was higher than at the present epoch. The HESS
observations show that the ratio of gamma-ray flux to the molecular gas column
density varies with galactic longitude, with a noticeable “deficit” of gamma rays
at l ≈ 1.3◦ . This interesting feature implies a non-uniform spatial distribution of
cosmic rays in the ≤200 pc region of GC. Amongst other possible considerations,
an interesting explanation of this effect could be the high activity of the particle
accelerator in the recent past. Indeed, in this case protons accelerated in Sgr A*
or in a nearby object, have not yet had time to diffuse out to the periphery of the
200 pc region. The epoch of the high activity of the accelerator depends on the pro-
ton diffusion coefficient. Assuming, for example, that the propagation of multi-TeV
protons in the galactic center proceeds with a speed similar to the one in the Galactic
Disk, the epoch of the high activity of the accelerator and the total energy release in
relativistic particles during the outburst are estimated 104 year and 1050 ergs, respec-
tively [39]. Two potential cosmic ray accelerators are located within the error box
of HESS J1745-290. These are the supernova remnant Sgr A East the age of which
is estimated between 1,000 and 10,000 years, and Sgr A* assuming that it was more
active in the past.
Fermi Bubbles
Recently, remarkable evidence has emerged for two enormous gamma-ray structures
centred on the core of the Galaxy and extending to approximately 10 kpc above and
below the Galactic plane. Since these spectacular radiation features have been found
in the Fermi LAT dataset [222], they received a nickname Fermi Bubbles. At lower
galactic latitudes they coincide with somewhat smaller structures found earlier in the
X-ray data of ROSAT [218] and in 20–60 GHz data of WMAP (the so-called WMAP
“haze”) [133]. Although currently the origin of these mysterious structures is under
intense debates, it is clear that they are connected, in one way or another, to the
activity of the galactic nucleus. While X-rays are of thermal origin, the microwave
“haze” most likely is a result of synchrotron radiation of relativistic electrons [133],
although a significant contribution of the dust to the microwave emission cannot be
excluded.
What concerns the Fermi Bubbles—they undoubtedly have nonthermal origin.
The parent relativistic particles (primarily protons) can be accelerated in the nucleus,
and then injected into Fermi Bubbles. Alternatively, protons and electrons can be pro-
duced in situ through the first and/or second order Fermi acceleration mechanisms
supported by hydrodynamical shocks or plasma waves in highly turbulent medium.
The processes that create and support these structures can originate either from an
AGN type activity related to the central black hole (Sgr A*) or from the ongoing star
Fig. 35 The location and geometry of the Fermi Bubbles shown schematically together with the
WPAP haze and the X-ray emission structure detected by ROSAT (from Ref. [222]). All three
structures may have a common origin (although produced in different ways) related to much higher
activity of the central black hole (Sgr A*) in the past or to the nuclear starburst in GC
74 F. Aharonian
formation in the galactic nucleus. The Dark Matter is another alternative for produc-
tion of the WMAP “haze’ and Fermi Bubbles, either directly or due to synchrotron
radiation and IC scattering of electrons (the products of Dark Matter annihilation or
decay), but the morphology of Fermi Bubbles disfavours the Dark Matter origin of
these structures [222].
Figure 35 illustrates the location, geometry and scales of Fermi Bubbles, and their
relation to the WPAP “haze” and the large-scale X-ray structures. The luminosity
of gamma-rays with hard, E −2 type spectrum in the energy interval 1–100 GeV
is quite impressive, L γ ≈ 4 × 1037 erg/s—an order of magnitude larger than the
microwave luminosity of the “haze”, and only an order of magnitude less than the
thermal X-ray luminosity [222]. Clearly, given the overall limited energy budget
of GC, the mechanisms of particle acceleration and gamma-ray emission in Fermi
Bubbles should proceed with very high efficiency.
In this regard, a natural explanation of Fermi Bubbles and the WMAP “haze”
would be that they are produced by the same population of highly relativistic
(10 GeV) electrons through IC scattering and synchrotron radiation, respectively
[222]. Because of severe radiative energy losses, the mean free path of ≥100 GeV
electrons is significantly shorter than the size of Fermi Bubbles. Therefore it is
difficult to explain how the electrons could fill these vast structures, unless we assume
very fast propagation of electrons with a speed exceeding 10,000 km/s, or postulate
in situ electron acceleration throughout the entire volume of the bubbles [119, 222].
The requirement for ultra fast propagation of electrons seems to be unrealistic, both
for the diffusion and the transport via galactic wind. On the other hand, in situ particle
acceleration can be realized through stochastic (2nd order Fermi) acceleration [192]
or due to series of shocks propagating through the bubbles and accelerating relativistic
electrons [115]. It has been argued [192] that stochastic acceleration of electrons may
reproduce rather well the energy dependent intensity profiles of gamma-radiation,
and predict significant limb brightening above a few hundred GeV. The reason is
the presumably higher turbulence due to presence of shocks at the bubble edges
which would imply acceleration of electrons to higher energies close to these edges.
Importantly, the suggested acceleration mechanism cannot boost the electron energy
beyond 1 TeV, thus in order to explain the extension of the observed gamma-ray
spectrum up to 100 GeV, one has to invoke FIR and optical/UV background emission
supplied by the galactic disk (see Fig. 36). This model provides robust predictions.
In particular, since the FIR and optical/UV contributions to the target field for IC
scattering decrease quickly with distance from the disk, the spectrum of gamma-rays
from high latitudes should contain a cutoff above tens of GeV. The limb brightening
at highest energies is another characteristic feature predicted by this model. These
spectral and spatial features can be explored in the near future, after the gamma-ray
photon statistics in the Fermi LAT dataset is achieved an adequate level.
Hadronic origin of gamma-rays is an alternative interpretation suggested for Fermi
Bubbles [119, 186]. Generally it is claimed to be inefficient channel for gamma-
radiation, especially because of the low plasma density, n ≤ 10−2 cm−3 . However,
these claims are based on a misleading interpretation of the efficiency of gamma-ray
production processes in general, and their specifics in Fermi Bubbles, in particular. In
5
10
Aharonian and Crocker
Simple disk IC template
1
Cheng et al.
sr
Fermi 0.5 1.0 GeV IC template
1
6 this work
10
2
s
E2 J GeVcm
7
10
8
10 1 2 3
10 1 10 10 10
Energy GeV
Fig. 36 The spectral energy distribution of gamma-rays from the Fermi bubbles compared to the
theoretical predictions. (i) IC model of Ref. [192] (solid line) assuming stochastic acceleration of
electrons in the bubbles (the contributions from the scattering on the CMB, FIR, and optical/UV
backgrounds are shown separately); (ii) IC model of Ref. [115] (dotted line) assuming diffusive
shock acceleration of electrons; (iii) hadronic model of Ref. [119] (dashed line). The figure is from
Ref. [192]
fact, the Fermi Bubbles can be considered as an exemplary case when pp interactions
proceed with very high efficiency. For example, the efficiency of this process can
dramatically exceed, by several orders of magnitude, the gamma-ray production effi-
ciency in young SNRs. Indeed, the plasma density alone does not tell us much about
the efficiency; obviously, this issue should be discussed in the context of time-scales
that characterize the gamma-ray production and the confinement of protons in the
bubbles. If protons have been continuously injected and trapped in the bubbles over
the time-scales of approximately 1010 year (the age of the Galaxy), the main energy
in accelerated protons would be lost in pp collisions given that the characteristic
time of the latter, tpp = 1/(k p nσpp c) ≈ 5 × 109 (n/10−2 )−1 cm−3 year, is shorter
than the confinement time. This implies that we deal with the so-called “thick target”
scenario, when the system is in saturation. The hadronic gamma-ray luminosity is
equal to L γ ≈ Wp /tpp→π0 , where Wp is the total energy of protons in the bubbles,
and tpp→π0 is the time-scale for neutral pion production in pp interactions. In the
saturation regime, Wp = Q̇ p tpp , assuming that the energy dissipation through pp
collisions is the dominant loss process. Since tpp = 1/3 tpp→π0 , we have L γ = Q̇ p /3,
thus about a third of the power injected into relativistic CRs emerges in gamma-rays
(of all energies) independent of the local density, the source volume and the injection
time. Note that since the time-scale of pp interactions is comparable to the supposed
age of the bubbles of 1010 year, the efficiency would be somewhat less. Also, one
should take into account that at low energies the ionization and adiabatic losses of
protons play a non-negligible role, thus the overall efficiency for a broad energy
spectrum of protons would be reduced to several percent. The fluxes of hadronic
gamma-rays shown in Fig. 36 confirm these simple estimates. The numerical cal-
culations, which have been performed assuming E −2.1 type spectrum of protons
76 F. Aharonian
up to 1015 eV injected into the plasma of density n = 5 × 10−3 cm−3 , require an

average rate of proton injection 1039 erg/s. Note that independent of the history of
injection of relativistic protons, the current total energy in protons should be as high
as Wp = L γ tpp→π0 1055 erg which is comparable to the magnetic field energy in
the bubbles.
It is remarkable that the current rate of star formation in GC can supply the required
nonthermal power in relativistic protons. The total infrared luminosity of the inner
few hundred parsec region of GC, L I R 1.6 × 1042 erg/s, gives an estimate for the
star formation rate of about 0.08 M /year and, correspondingly, the supernova (SN)
rate of 0.04/century. For the average mechanical energy released per SN, 1051 erg,
the total power injected by supernovae is estimated Ė S N 1.3 × 1040 erg/s. Thus,
for the “standard” 10 % efficiency of transformation of the SN explosion energy to
accelerated particles, GC seems to be able to provide the required 1039 erg/s power in
relativistic protons. This is a strong argument in favour of interpretation of gamma-
ray emission of Fermi Bubbles by a population of relic cosmic ray protons and
ions associated with star-formation in GC. This scenario implies a relatively slow
release of energy from supernovae explosions and requires effective confinement of
protons in Fermi Bubbles on multi-Gyr time-scales. Propagation of protons in the
diffusion regime allows an effective confinement on these scales provided that the
diffusion coefficient in Fermi Bubbles is smaller, by 1 or 2 orders of magnitude, than
in the Galactic plane. This sounds a quite reasonable assumption given the very high
turbulence of plasma in the Fermi Bubbles.
The gamma-ray production in a saturated regime minimizes the dependence of
the production rate on the ambient plasma density which implies a rather constant
volume emissivity despite possible variations of the plasma density. This is true,
of course, for homogeneous distribution of cosmic rays which can be realized, for
example, in the case of nearly constant diffusion coefficient throughout the bubbles.
However, the spatial variations of the diffusion coefficient could be rather significant.
For example, the higher plasma turbulence around the edges of the bubbles, would
slow down the particle diffusion speed compared to the central parts of the bubbles,
and thus would increase the concentration of protons closer to the edges. Corre-
spondingly, this would lead to more homogeneous projected gamma-ray intensity
(surface brightness) as it follows from observations [222]. In this regard, the criticism
[192] of hadronic model of gamma-ray production in Fermi Bubbles, in particular
in the context of expected and observed morphologies of gamma-ray emission, can
be readily overcome assuming a strong gradient of turbulence towards the edges.
Finally, it should be noted that the hadronic origin of gamma-ray emission of
the Fermi Bubbles does not exclude other “hadronic” scenarios with faster energy
release related, for example to the activity of the central black hole Sgr A*. Fast
energy release can be provided, for example, by the star capture processes by Sgr
A* over the last 10 Myr with an average capture rate of 3 × 10−5 year −1 and energy
release of 3 × 1052 erg per capture [116]. It has been argued in Ref. [116] that the
quasi-periodic injection of hot plasma can produce a series of shocks in the bubbles
which should be able to (re)accelerate protons to energies well beyond the “knee” up
to 1018 eV. If confirmed by independent detailed hydrodynamical calculations, this
Fig. 37 Three regions of gamma-ray production related to rotation-powered pulsars: Pulsar Mag-
netosphere, Pulsar Wind, and Pulsar Wind Nebula. The specific numbers shown in the sketch
characterize the pulsar in the Crab Nebula
could appear a quite natural solution for the origin of one of the most “problematic”
(poorly understood) energy intervals of cosmic rays.
7 Pulsars, Pulsar Winds, Pulsar Wind Nebulae
Pulsars, rapidly rotating neutron stars left over after supernova explosions, are promi-
nent sources of high energy gamma-rays produced in three physically distinct regions:
(i) Pulsar Magnetospheres (PM), (ii) Cold Ultrarelativistic Winds (CRW) which
carries almost the entire rotational energy of the pulsar in the form of Poynting
flux and/or kinetic energy of the bulk motion, and (iii) the Pulsar Wind Nebulae
(PWN)—the product of termination of the pulsar wind with subsequent acceleration
of relativistic particles and their nonthermal emission. These three zones schemati-
cally are shown in Fig. 37.
7.1 Radiation of Pulsar Magnetospheres
In the pre-Fermi era only a handful of pulsars have been established as high energy
gamma-ray sources. Observations with Fermi LAT dramatically increased the num-
ber of gamma-ray emitting pulsars to more than 100 from which almost one third
have been discovered through their pulsed gamma-ray emission alone. The high qual-
ity phased-resolved energy spectra obtained with Fermi LAT provide a new insight
into the pulsar demographics and physics. In particular, the measured light curves
78 F. Aharonian
and energy spectra indicate that gamma-ray emission from the brightest pulsars is
produced in the outer magnetosphere with fan-like beams scanning over a large por-
tion of the celestial sphere. The energy spectra of most of gamma-ray pulsars can be
described by a simple function of “power-law with exponential cutoff” written in a
generalized form
E −Γ exp [−(E/E 0 )b ], (52)
with b ≤ 1, and the cutoff energy between 1 and 10 GeV [1]. The detection of
gamma-rays beyond a few GeV with no indication of super-exponential attenuation
(i.e. b > 1) effectively excludes the so-called polar cap model and gives a preference
to models of gamma-ray production in the outer magnetosphere (to avoid severe pair
production in the strong magnetic field in low-altitude zones).
The energy spectra given in the form of Eq. (52) with b = 1 generally are well
explained by the mechanism of curvature radiation. Although most of the reported
spectra can be fitted with the fixed value of b = 1 [1], however the extension of
spectral measurements of the brightest gamma-ray pulsars towards both higher and
lower energies revealed that the spectra beyond the cutoff could be smoother with b
closer to 0.5. For example, the phase-averaged spectrum of the Crab pulsar is better
fitted with the combination of parameters b = 0.43, Γ = 1.59 and E 0 = 0.50 GeV
[110], rather than b = 1, Γ = 1.97 and E 0 = 5.8 GeV as reported earlier by the
Fermi collaboration based on smaller gamma-ray statistics [5].
The classical radio pulsars are young astronomical objects of age between 103
and 105 year. Because of rotational losses, pulsars gradually slow down with time,
and typically after 105 year become too faint to be detected. However, the pulsars
produced in binary systems may have the “second” life. The matter of the companion
star falling onto the neutron star spins it up with a rotation period as short as a
few milliseconds. As a result, the “recycled” pulsars after 108 −109 years become
visible again under a new name of millisecond pulsars (MSPs). In this regard, the
detection of gamma-rays from tens of MSPs by Fermi LAT [2] can be considered as a
natural outcome, although not without some surprise. This concerns, in particular, the
discovery of gamma-rays from the millisecond pulsar J1823-3021A [134]. The large
rate of change of the period of this pulsar derived from its gamma-ray luminosity
implies an unexpectedly small age of 2.5 · 106 year, two orders of magnitude shorter
than other MSPs, indicating the need for a revision of the current concept of formation
of millisecond pulsars. Another interesting (to some extent unexpected) feature of
MSPs is that despite significant differences in conditions (e.g. the magnetic field at
the neutron star surface in MSPs is smaller than in young pulsars by four orders of
magnitude), both the spectral shapes and time profiles of MSPs resemble those of
young pulsars [2]. This indicates to the similar conditions formed in the gamma-ray
production regions in both type of pulsars.
In general, no VHE gamma rays are expected from pulsar magnetospheres, except
for realization of some specific conditions in MSPs [111], and perhaps also in some
young (Vela-type) pulsars [214], when an additional inverse Compton component of
emission can contribute to the radiation of electrons. Such a component seems to be
needed if the recently reported by the VERITAS [76] and MAGIC [75] collaborations
pulsed VHE gamma-ray emission from the Crab pulsar has magnetospheric origin.
However, the pulsar magnetosphere is not the only region from which we should
expect pulsed emission. Periodic gamma-rays can be produced also in the pulsar
wind.
7.2 Radiation of Pulsar Winds
The generally accepted paradigm of pulsars and their synchrotron nebulae [166, 212]
postulates the existence of a relativistic electron–positron wind, which originates in
the pulsar’s magnetosphere and terminates in the interstellar medium. The large
electric fields in the magnetosphere accelerate electrons, the interactions of which
with the surrounding magnetic field initiate electromagnetic cascades. This leads
to the formation of a relativistic outflow of a dense electron–positron plasma—the
wind. Initially, the rotational energy losses are released in the form of Poynting
flux (electromagnetic energy), thus the wind’s Lorentz factor cannot be very large.
The conversion of the Poynting flux leads to the ‘acceleration’ of the wind with
Lorentz factor as large as Γw ∼ 104 . Since the observations of the synchrotron nebula
require an energy injection into the nebula close to the pulsar’s spin-down luminosity
[83, 121, 166, 245], the wind should transport the rotational energy of the pulsar to
the nebula in the most effective and economical way. In particular, the wind should be
‘cold’ (in the sense of the low energy of the electrons in the frame of moving plasma)
in order to avoid synchrotron losses. The presence of cold ultrarelativistic winds has
been argued also in the context of radiation of binary pulsars [170, 173, 232].
Despite the general consensus on the concept, the formation of cold ultrarela-
tivistic pulsar winds remains a mysterious phenomenon. To a certain extent, this is
explained by the lack of information about the site(s) and the rate of acceleration
of plasma (dissipation of the Poynting flux) in pulsar winds. And, of course, the
very existence of pulsar winds still needs to be observationally confirmed. This is
not an easy task. It is often argued that the cold ultrarelativistic wind is a ‘dark sub-
stance’ which cannot be directly observed. This statement is true for the synchrotron
radiation of the wind—relativistic electrons move together with magnetic field and
therefore do not emit synchrotron radiation. However, the cold ultrarelativistic winds
of pulsars can be directly detected through their characteristic inverse Compton radi-
ation [100]. Moreover, we already do have a strong evidence in this regard; it is quite
plausible that the VHE pulsed gamma-radiation recently reported from the Crab
pulsar [75, 76], in fact originates from the pulsar’s wind [62].
The phase-averaged spectral points of the Crab pulsar measured by Fermi LAT [5]
are shown together with VHE points reported by VERITAS [76] and MAGIC [75]
in Fig. 38. It is seen that the VHE points are located well above any extrapolation
of Fermi data fitted by the standard form of the pulsar spectra given by Eq. (52). On
the other hand, the extrapolation of GeV fluxes to the VHE-domain as a power-law
with photon index Γ ≈ 3.8, and its interpretation as an evidence for the same mag-
80 F. Aharonian
10-12
PULSED GAMMA-RAY EMISSION OF CRAB
10-13
KEY MODEL PARAMETERS
Pulsar Emission w=10
6
, Rw: (from 20 to 50)RL
Flux (Jl m-2s-1)

5
w=4x10 , Rw=32RL
5
w=6x10 , Rw=32RL
5
w=6x10 , Rw=30RL
10-14
Wind Emission
MAGIC mono
MAGIC stereo
10-15 VERITAS
Fermi
10-16
2 3
100 101 10 10
Energy (GeV)
Fig. 38 SED of gamma-radiation produced by the pulsar magnetosphere and by the pulsar wind.
The Fermi LAT points are best-fitted by Eq. (52) with b = 1, E 0 = 5.8 GeV and Γ = 1.97
(dashed grey line) or by slightly harder spectrum in the cutoff region with b = 0.85, E 0 = 7 GeV
and Γ = 1.97. These spectra can be explained by curvature radiation of electrons in the pulsar’s
magnetosphere. The IC emission of the ‘cold’ ultrarelativistic wind can explain the pulsed gamma-
ray fluxes above 100 GeV. The solid light blue, blue and green curves are calculated under the
assumption of an instant acceleration of the wind to the fixed value of the Lorentz factor Γw and
at the fixed radius Rw (these parameters are indicated in figure). The solid red curve corresponds
to the case when the Poynting flux transformation takes place within the zone 20–50RL with linear
dependence of the wind’s acceleration rate with the distance; the maximum Lorentz factor achieved
at 50R is set to 106 (from Ref. [62])
netospheric origin of GeV and TeV gamma-rays (see e.g. Ref. [75]) requires, in fact,
a rather dramatic revision of basic concepts currently employed in magnetospheric
models. Also, the assumption of the magnetospheric origin of radiation over the
entire gamma-ray domain contradicts the essentially different lightcurves reported
at GeV and TeV energies (unless the production sites of these two components are
well separated), as well as the apparent tendency of spectral flattening above 100 GeV.
A natural and more plausible site of production of pulsed VHE gamma-rays is the
ultrarelativistic wind being illuminated by pulsed X-rays arriving from the pulsar’s
magnetosphere [62]. It is demonstrated in Fig. 38 that the entire gamma-ray region
can be considered as a superposition of two separate components—the nominal
(magnetospheric) GeV gamma-rays and the VHE component of the Comptonized
wind.
Although the inverse Compton gamma-rays are produced by monoenergetic elec-
trons, the SED of gamma-rays in the range of tens to hundreds of GeV is quite
flat. This is caused by the combination of effects related to the broad power-law
distribution of seed photons and the transition of the Compton cross-section from the
Thomson to the Klein-Nishina regime. On the other hand, the spectrum is expected
to have a very sharp cutoff at E = Γw mc2 . This not only serves as a distinct feature
for identification of the wind origin of gamma rays, but also allows derivation of
wind’s Lorentz factor.
Generally, the lightcurve of target photons should be reflected in the time structure
of the IC gamma ray signal, however, they cannot be identical because of the effects
related to the geometry as well as the specifics of the anisotropic IC scattering. The
geometrical effects may lead to significant differences between the arrival times of
the target photon and the secondary gamma ray pulses as illustrated in Fig. 39. It is
seen that whilst for the wind located close to the light cylinder the gamma-ray signal
appears significantly shifted relative to the reported gamma-ray data, Δt ∼ 0.1T ,
for the wind acceleration at Rw = 30RL the widths and the positions of the predicted
and observed gamma ray peaks P1 and P2 are in very good agreement. On the other
hand, in the case of isotropic wind the predicted P1/P2 ratio of the gamma-ray signal
mimics the X-ray lightcurve (≈2), while the reported VHE gamma-ray data [75, 76]
show P1/P2 < 1. This can be explained by anisotropy of the wind which introduces
significant corrections to the shape of the gamma ray lightcurve in general, and in
the P1/P2 ratio, in particular (see Fig. 39).
In summary, the interpretation of Crab’s pulsed VHE gamma-ray emission in
terms of Comptonization of the pulsar wind implies the first direct evidence of cold
ultrarelativistic electron–positron wind in a pulsar. The conversion of the Poynting
flux to kinetic energy of bulk motion should take place abruptly in the narrow cylin-
drical zone between 20 and 50 light-cylinder radii and should accelerate the wind to
a Lorentz factor of (5 − 10) × 105 .
7.3 Pulsar Wind Nebulae
Before the discovery of the Crab pulsar, the compact stellar object in the center of
the Crab Nebula has been proposed as a potential source of the nebular magnetic
fields and relativistic particles [206]. Immediately after the discovery of pulsars
[150] and their identification with neutron stars [144, 150], the rotation energy of
these compact stars was recognized as the ultimate energy source powering PWNe
[203]. A bit later, the existence of relativistic electron–positron pulsar winds has
been postulated to explain the link between pulsars and PWNe [212]. The current
paradigm of formation of PWNe is based, to a large extent, on the MHD model of
interaction of a cold ultrarelativistic electron–positron wind with interstellar medium
[166] which, despite a number of introduced simplifications, satisfactorily describes
the main features of the Crab Nebula [83, 121]. The recent detailed two-dimensional
MHD simulations [101, 245] confirm the basic concept, at least for the Crab Nebula—
the most prominent representative of this source population. Moreover, even the fine
structures, like the toroidal and jet-like features in the inner part of the Crab Nebula
and some other PWNe discovered in X-rays by Chandra, can be explained within
the framework of the model of Kennel and Coroniti [166], after introducing initial
anisotropy of the energy flux in the wind [102, 122, 175]. For a comprehensive
review on observational and phenomenological studies of evolution and structure of
82 F. Aharonian
(a) Observer
X−ray photons
target
photo
θ ns
RL γ −Rays
Pulsar Wind
Rw
Δ t = − θ T+ w (1−cos θ )
R
2π c
(b) TIME PROFILES OF PULSED EMISSION

1.2
MAGIC stereo
1
VERITAS P1
X−ray photons
Normalized Flux
−ray R w=R L
0.8 P2 −ray R w=30R L
−ray anisot. wind
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2
Phase
Fig. 39 Formation of the pulsed VHE gamma-ray signal in the wind of the Crab pulsar. a Geometry
of the IC scattering of magnetospheric X-rays by the ultrarelativistic cold wind; b the reported
lightcurves of VHE gamma-rays together with theoretical lightcurves of the wind IC emission
calculated for different assumptions concerning the acceleration site of the wind and it anisotropy
(from Ref. [62])
PWNe the reader is referred to [138], and for the theoretical models and challenges
of the physics of pulsar winds and nebulae, to the review article [174].
Crab Nebula
The gamma-ray emission of the Crab Nebula has been studied by many groups
using different detection techniques. Presently the reported fluxes cover the energy
interval from 1 MeV to 100 TeV (see Fig. 40). Remarkably, even this impressive
coverage of 6 decades in energy is only a fraction of the broad band non thermal
spectrum that extends over 21 (!) decades of frequencies—from radio wavelengths
to very high energy gamma-rays. Within the MHD wind paradigm of the Crab,
the ultrarelativistic pulsar wind terminates by a standing reverse shock at which
the electrons and positrons of the wind are accelerated and randomized. Although
there is no doubt in the important role of the relativistic shock for acceleration
Fermi
CGRO COMPTEL
CGRO EGRET
HESS
10−9 MAGIC
CANGAROO
VERITAS
E 2.F [erg cm−2 s −1 ] HEGRA
CELESTE
10−10
10−11
10 −12
2 3 4 5 6 7 8
1 10 10 10 10 10 10 10 10
Energy [MeV]
Fig. 40 SED of gamma-rays of the Crab Nebula. The spectral points from low to VHE gamma-rays
are shown together with the fit of the synchrotron component (blue dashed line) and predictions of
IC gamma-rays calculated for three different values of the mean magnetic field: B = 100 µG (solid
red line), B = 200 µG, and the equipartition field of the nebula of 300 µG [83] (from Ref. [5])
of electrons, the acceleration mechanism itself remains unknown (see for a review
Ref. [174]). In principle, the electrons can be accelerated in the course of reconnection
of the alternating magnetic field at the termination shock, but it is not clear that this
mechanism alone would be sufficient to boost the electrons to highest energies. For
the hard power-law spectrum of electrons derived from observations of PWNe up to
100 TeV and beyond, the first-order Fermi mechanism seems to be a more feasible
scenario [22], although it is not free certain difficulties, in particular it requires
significant amplification of turbulence downstream of the shock [202].
The major fraction of the rotational energy of the pulsar transported through the
wind is eventually released in non thermal synchrotron radiation observed in radio,
optical, X-ray and gamma-rays extending to energies ≥100 MeV. The mean energy
of synchrotron photons εs radiated by an electron of energy E TeV = E/1 TeV in the
magnetic field of strength BμG = B/1 µG is
εs 2
0.02E TeV BμG eV. (53)
For the typical magnetic field of order of 100 µG at the shock [166], the energy
of electrons responsible for ≥100 MeV synchrotron radiation should be as large
as 1015 eV. Thus we deal with a cosmic PeVatron. It is remarkable that the Crab
Nebula is not simply the first (and so far the only) cosmic accelerator identified
as a PeVatron, but, more importantly, it operates as an electron PeVatron. Because
of severe radiative losses of electrons, their acceleration to such high energies is
possible only if the acceleration proceeds at a rate close to the theoretical limit
given by Eq. (28) with η ∼ 1. Note that the very fact of extension of the synchrotron
spectrum up to ≥100 MeV is a clear indication, independent of the model parameters,
84 F. Aharonian
in particular the strength of the magnetic field, that the acceleration in this object
proceeds at a rate close to the theoretical limit (see Sect. 3.2). The recent discovery of
gamma-ray flares [11, 110, 221, 235] confirms these early conclusions (see e.g. Ref.
[52]) but, at the same time, brings new puzzles and challenges in the understanding
of the physics behind the extreme acceleration of electrons and related to them the
spectacular gamma-ray flares.
The inverse Compton scattering of the same population of ultrarelativistic elec-
trons provides the second channel for gamma-ray production resulting in formation
of very high energy tail of gamma-radiation. Although the energy release in IC con-
stitutes only a small fraction of the synchrotron luminosity of the nebula, the high
energy gamma-radiation provides a unique channel of information about conditions
in the nebula. In particular, the comparison of X- and TeV gamma-ray fluxes observed
from the Crab Nebula leads to a robust estimate of the mean magnetic field between
100 and 200 µG (see Fig. 40), in a remarkably good agreement with predications of
the theory of termination of the MHD pulsar wind [166].
Figure 40 demonstrates the homogeneous coverage of the entire gamma-ray
domain with high quality data. While the COMPTEL and EGRET results carry
information about synchrotron radiation in the cutoff region, the Fermi LAT data
show the sharp transition from the synchrotron to the IC component of radiation
around 1 GeV. Another important feature seen in Fig. 40, is the clear indication
of the IC maximum around 100 GeV supported by both satellite (Fermi LAT [5]),
and ground-based (MAGIC [75] and VERITAS [76]) measurements which show
remarkable agreement with each other. The measurements of ground-based observa-
tions have almost approached 100 TeV [31, 42, 230], but current instruments are not
able to explore the region beyond 100 TeV. Meanwhile, the IC component should
extend to the energy region set by the maximum energy of accelerated electrons, i.e.
1 PeV. Although the production of gamma-rays at such energies takes place in the
Klein-Nishina regime, and therefore is suppressed, still it can be detected by next
generation gamma-ray detectors as it is demonstrated in Fig. 7.
As the brightest persistent point-like TeV gamma-ray source seen effectively from
both hemispheres, the Crab Nebula has been considered as a standard candle for
cross-calibration of detectors, and correspondingly this technical issue often has been
claimed as the major motivation for high energy gamma-ray observations, assuming
that the “astrophysical” objectives already have been achieved. However, the Crab
flares above 100 MeV [11, 235] indicate that many details remain unresolved, and
that we are far from proper understanding of many aspects related to this unique
source. The new, exceptional activity of the source in April 2011 [110, 221] provided
higher quality data which however only deepened the challenges. Two most striking
features of these flares are the short times scales of flares with a rise time as short
as 6 h and extension of the energy spectrum to GeV energies (one should note that
similar features have been found earlier, in an independent analysis of Fermi LAT
data on the 2010 flares [87]).
The exceptionally high fluxes during the active state in April 2011 allow detailed
spectroscopy for different flux levels (see Fig. 41). In order to study the spectral
evolution of the flaring component, a steady-state (constant) background has been
Energy [MeV]
3 3
102 10 10 2
10 10 2 10 3 102 10 3
-9
10
-10 0
10
1 2 3 4
F [ ergs cm-2 s-1 ]
10-9
10-10 0
5 6 7 8
10-9
10-10 0
9 10 11
Fig. 41 Time evolution of the Crab SED during the April 2011 flare. The flare duration of approxi-
mately 9 days has been divided in 11 time windows of approximately constant flux. The dot-dashed
line indicates the assumed constant background from the synchrotron nebula. The dotted lines show
the flaring component, and the dashed lines are the sums of the background and flaring components
(from Ref. [110])
assumed with a steep power-law spectrum described by a photon index Γb = 3.9.

The spectrum of the flaring component has been assumed in the form of power-law
with exponential cutoff, ν Fν = f 0 E 2−Γf exp[−(E/E 0 )κ ]. The spectra during all
selected windows can be well described by the same photon index Γf = 1.27 ± 0.12
and the exponential cutoff index κ = 1, but with variable total flux f 0 and the cutoff
energy E 0 . It has been found that while the cutoff energy varies only by a factor
of two, the total flux can be changed more than an order of magnitude. If the flares
are caused by synchrotron radiation of electrons, the strictly exponential cutoff with
κ = 1 and photon index Γf = 1.27 imply a narrow distribution of electrons with
a very sharp high-energy cutoff. Such a distribution hardly can be explained by the
diffusive shock acceleration, at least in the standard scheme of its operation. An
alternative mechanism of acceleration due to the magnetic reconnection (see e.g.
Ref. [252]) seems a viable option. In particular, in the recently suggested scenario
of reconnection-powered linear electric accelerator [241], the electrons are acceler-
ated by the electric field to highest (1 PeV or so) energies before suffering significant
radiative losses, therefore one should expect very narrow distribution of electrons.
Another attractive feature of this mechanism is that the synchrotron radiation of elec-
trons can well exceed the “standard” maximum energy of about 100 MeV, indicated
86 F. Aharonian
earlier in Ref. [172] for relativistic current sheets. The obvious reason of this effect
is the assumption that the electric field exceeds the strength of the magnetic field.
This follows from Eq. (30) in Sect. 3.2. The condition E > B implies that the η
parameter defined as η = B/E is less than 1. Correspondingly the maximum energy
of synchrotron radiation is shifted to higher energies by the factor of E/B > 1
[231]. The episodic nature of radiation in this model is explained by the motion of
PeV electrons along the current sheet the orientation of which has rather random
character. Therefore one can observe sudden increase of synchrotron emission with
shifted synchrotron peak only when the particle beam is directed by chance towards
the observer [241]. It should be noted that the condition E > B implies a breakdown
of MHD. Wile this “phase transition” of plasma can (or even should) take place at
the acceleration of the pulsar wind [97], the formation of reconnection current sheets
in the termination shock region seems a less obvious realization; this question needs
further theoretical studies.
There are other ways to explain the extension of the gamma-ray spectra of flares
beyond the “nominal” synchrotron cutoff. The simplest (although somewhat arti-
ficial) solution could be to assume that the acceleration and radiation regions are
effectively separated, namely, the electrons are accelerated in a region of low mag-
netic field, then enter and radiate in a region with stronger magnetic field. A more
realistic modification of this possibility is the acceleration and radiation of electrons
in the termination shock region at the presence of stochastic magnetic field [112].
The strong magnetic field fluctuations, which can be facilitated by multiple forced
magnetic field reconnections, is a key element in this scenario. While the acceleration
and energy loss rates depend on B 2 , the synchrotron radiation in the cutoff region
is contributed by higher statistical moments of the stochastic field distribution. This
might shift the position of the synchrotron cutoff to higher energies. For example,
in the case of Gaussian type probability distribution function of magnetic field, the
cutoff energy can be increased up to 300–400 MeV [112]. This will be accompanied
by a dramatic, by an order of magnitude increase of the gamma-ray flux around
1–2 GeV as has been observed during the Crab flares. This model does not address,
however, another important feature of flares, namely their extremely fast variability,
Δt ≤ 6 h, which implies that the small size of the emitting region should be as
small as l ≤ Δt × c ≈ 1015 cm. On the other hand, the gamma-ray luminosity at
the peak of the April 2011 flare was 4 × 1036 erg/s, close to 1 % of the spin-down
luminosity of the pulsar. It is difficult to propose a realistic scenario of channelling
of substation fraction of the available energy into a compact regions with a linear
size less than 10−3 of the termination shock radius! In phenomenological terms, the
simplest (and quite obvious) solution of this general problem is to propose that we
deal with an apparent but not intrinsic luminosity, i.e. assume that the emission is
strongly beamed towards the observer within a small angle [241], or the flares are a
result of strongly Doppler boosted radiation from compact structures (see e.g. Ref.
[231]). These could be fragments of the wind termination shock (see e.g. Refs. [93,
176, 187]) or some specific compact features in the inner nebula (see e.g. [185, 249]).
The hypothesis of Doppler boosted radiation is especially attractive because it
allows reduction of energy requirements by orders of magnitude, relaxes the con-
strains on the emission size, and offers different ways of sudden increase of the
Doppler factor of the magnetized non thermal plasma. Yet, there are key issues, e.g.
the origin of the relativistic motion of the synchrotron source and its large (mG scale)
magnetic field, which can be addressed only by detailed relativistic MHD simula-
tions. Such studies already have started, and it is likely that soon they will tell us much
more about the Crab flares. On the other hand, it is likely that the Crab flares are not a
rare events, so it is expected that the detection of new flares by Fermi and AGILE will
provide more information about the features of this unique phenomenon. The search
for IC gamma-rays at TeV energies during GeV synchrotron flares seems an obvious
motivation for ground-based observations, although the expectations concerning the
detectable counterpart TeV gamma-ray flares should not be overestimated.
TeV PWNe
The strong magnetic field in the Crab Nebula initiated by termination of the pow-
erful pulsar wind dramatically reduces the efficiency of production of gamma-rays
through IC scattering. The energy density of the magnetic field exceeds by more
than two orders of magnitude the radiation density. Thus less than 1 % of the energy
of accelerated electrons is released in IC gamma-rays, the rest being emitted trough
synchrotron radiation. In the case of the Crab Nebula, the low gamma-ray produc-
tion efficiency is compensated by the huge energy budget of the source provided by
rotation of the pulsar and converted very effectively to multi-TeV electrons.
In other PWNe with less powerful winds, the resulting nebular magnetic fields
can be weaker by a factor of 10–100, which makes these objects as potentially
(very) effective gamma-ray emitters. The main target photons for production of VHE
gamma-rays in these objects are contributed by the 2.7 K CMBR, thus the average
energy of IC gamma-rays produced in the Thomson limit is
εic 5 × 109 E TeV

2
eV. (54)
Equations (53) and (54) give a direct relation between the characteristic energies of
synchrotron and IC photons produced by the same electrons:
εs 7(εic /1TeV)BμG eV. (55)
One should note that at the highest energies the Compton scattering proceeds in the
Klein-Nishina regime, thus a non-negligible correction to the energy is needed. As
the maximum energy of the upscatered photon is εic,max = (1 + 1/b)−1 E e , where
b ≈ 0.005E TeV , the Klein-Nishina effect becomes significant at E e ≥ 100 TeV.
For a given strength of magnetic field, the energy losses of electrons are shared
between the synchrotron and IC radiation components as L γ /L X = wMBR /wB
1 (B/3 µG)−2 . This implies that in a PWN with a nebular magnetic field of about
10 µG or less, the IC gamma-ray production efficiency could be as large as 10 %.
Given that the rotational energy of pulsars is eventually released in relativistic elec-
88 F. Aharonian
trons accelerated at the termination shock, the detectors with sensitivities typical for
the current IACT arrays should be able to reveal TeV gamma-ray emission from tens
of PWNe of young pulsars with spin-down luminosities L 0 ≥ 1034 dkpc 2 erg/s [59].
The discovery of more than two dozens of extended TeV gamma-ray sources in the
close vicinity of pulsars in the indicated rage of luminosities, supports this phenom-
enological prediction. Presently PWNe constitute the largest galactic TeV source
population. Many previously dubbed “dark” TeV gamma-ray sources, including the
first unidentified TeV gamma-ray source discovered by the HEGRA collaboration,
TeV J2032+4130 [29], later have been identified with PWNe. Most of these iden-
tifications with PWNe are quite convincing, but still tentative, except for several
HESS sources which are firmly identified with famous PWNe. VHE gamma-ray
images of four PWNe, Vela X, MSH 15-52, the nebula associated with the pulsar
PSR J1826-1334, and the Kookaburra complex are shown in Fig. 42.
A common feature for all TeV sources identified, either firmly or tentatively, with
PWNe is that the gravity centres of the extended TeV images do not coincide with the
positions of parent pulsars. Asymmetric, one-sided images of these PWNe have been
found also in X-rays, but on significantly smaller scales. Although the mechanism
which causes PWN offsets from the pulsar positions, is not yet firmly established,
this effect could be linked to the propagation of the reverse shock created at the
termination of the pulsar wind in highly inhomogeneous medium (for a discussion
of this issue see [138]).
The second interesting feature of TeV PWNe is the significantly larger extension of
TeV images compared to the X-ray images. Basically, this is the result of combination
of several factors. (i) Generally, in PWNe with magnetic field of order of 10 µG or
less, which seems to be the case of most of TeV PWNe, the electrons responsible
for X-ray emission are more energetic than the electrons emitting TeV gamma-rays.
Therefore the synchrotron burning of highest energy electrons results in a smaller
size of the X-ray source. (ii) When electrons diffuse beyond the PWN boundary,
they emit less synchrotron radiation (because of the reduced magnetic field), but they
still effectively radiate gamma-rays via inverse Compton scattering on the universal
CMBR. (iii) Finally, because of the high X-ray background, the sensitivities of X-ray
detectors like Chandra and XMM-Newton are dramatically reduced beyond several
angular minutes. This significantly limits the potential of these instruments for weak
extended X-ray sources. In contrast, the sensitivity of IACT arrays remains almost
unchanged approximately within 1◦ radius of FoV. All these factors make the IACT
technique as the most powerful tool for studies of non thermal population of electrons
in PWNe.
The similar morphologies of PWNe in X-ray and TeV gamma-ray bands supports
the widely accepted view that both emission components are due to the radiation of
the same population of multi-TeV electrons via synchrotron and inverse Compton
channels, respectively. Generally, the inverse Compton models predict energy depen-
dent morphology of PWNe in TeV gamma-rays caused by the energy losses and the
propagation effects of multi-TeV electrons. Remarkably, a tendency of softening of
the energy spectrum of gamma-rays with distance from the position of the pulsar
has been found in the gamma-ray image of HESS J1825-137 [40]. This is indeed
Fig. 42 TeV gamma-ray images of four PWNe: top left Vela X. The white contours correspond
to the distribution of X-rays obtained with ROSAT. Top right The pulsar wind nebula MSH 15-52.
The smoothed excess map is shown together with X-ray distribution (white contours) measured
by ROSAT. The black point and the black star correspond to the pulsar position and the center
of gravity of gamma-ray distribution, respectively. Bottom left The PWN HESS J1825-137. The
position of the pulsar PSR J1826-1334 is marked by a white triangle (the bright point-source to the
south is the microquasar LS 5039). Bottom right The map of TeV gamma-rays within the 1◦ × 1◦
FoV in the Kookaburra complex region. The positions of two separated sources are marked with
error crosses. The position of the pulsar PSR J1420-6048 is marked with a star, the position of
the source called Rabbit (G313.3+0.1) is marked with black triangle. In all figures the PSF of the
instrument is shown in the lower left or right hand corners
a strong argument in favour of the inverse Compton origin of TeV emission of this
source. On the other hand, the TeV gamma-ray luminosity of this source is about
10 % of the pulsar spin-down luminosity. While formally this is still within the avail-
able energy budget of the system, the IC cooling time for electrons responsible for
≤1 TeV gamma-rays appears longer compared to the age of the source. This makes
quite difficult the interpretation of gamma-ray emission by the current spin-down
power of the pulsar. A natural solution to the problem of the energy budget crisis
could be the contribution of “relic” electrons produced at the early epochs when the
pulsar’s spin-down luminosity was significantly higher [40].
90 F. Aharonian
Cooling Time (Syn+IC)

5
10 yr
4
10 yr
time, yr
3
10 yr
B=0.2 G
B= 2 G
B=20 G
Acceleration Time
E, eV
Fig. 43 Acceleration and radiative (synchrotron and IC) cooling time-scales of electrons in a
PWN calculated for the strength of the magnetic field B = 2 × 10−7 G, B = 2 × 10−6 G and
B = 2 × 10−5 G
Note that the significant enhancement of production rate of electrons at early

epochs of pulsars has been independently argued in the context of observations
of radio data [82, 109, 227]. However, unlike the low energy (radio) electrons, the
“relic” TeV electrons from early epochs can survive only if the magnetic field does not
significantly exceed 1 µG. This is demonstrated in Fig. 43, where the characteristic
lifetimes of electrons are shown for 3 different values of the magnetic field—0.2,
2 and 20 µG. It is seen that while at the presence of magnetic field of 20 µG the
radiative cooling time of electrons of energy 10 TeV is approximately 103 year, for
the field of ≈1 µG their lifetime is increased almost by two orders of magnitude
allowing accumulation of electrons from epochs as early as 105 year. This not only
increases dramatically the number of potential TeV PWNe, but also can make very
bright the TeV gamma-ray emission of some individual representatives of this source
population. The key condition for realization of this scheme is the requirement of a
weak nebular magnetic field, B 10 µG.
The strength of the magnetic field varies over the nebula. In the cold ultra-
relativistic wind zone, the toroidal component of the field (dominant at large scales)
decreases linearly with the distance to the pulsar. In the shocked region, magnetic
field may follow different trends depending on the magnetization parameter σ. At
the termination shock with a characteristic radius
1/2 −1/2
L SD pext
rsh 5 · 1018 −12
cm, (56)
37
10 erg/s 10 erg/cm3
the magnetic field is estimated

Fig. 44 The efficiency of radiation of IC gamma-rays in a PWN as a function of time, calculated

for different strengths of magnetic field and for the history of the electron injection rate given by
Eq. (58)
1/2
σ 1/2 pext
B 6 µG, (57)
3 · 10−3 10 −12 erg/cm3
where pext is external to the nebula pressure contributed by the matter, magnetic field,
and cosmic rays outside the nebula. Interestingly, the B-field strength doesn’t depend
on the spin-down luminosity of the pulsar, but only on the external medium pres-
sure. The estimated magnetic field strength is significantly smaller than interstellar
magnetic field, Bism ≈ 3 µG. This implies a rather specific cooling regime when
electrons loose their energy mainly through IC scattering on the CMBR photons.
The higher spin-down luminosity of pulsars in the past has a dramatic effect on
the radiation efficiency which usually is determined as the IC gamma-ray luminosity
of the nebula to the current spin-down luminosity of the pulsar. In principle, the
efficiency determined in this way can be larger than 1. The “trick” here is related
to the relic electrons that have been accelerated in the past, but survived the energy
losses, and today provide the dominant contribution to production of gamma-rays. It
is seen from Fig. 44 where the radiation efficiency is shown for the time-dependent
injection of electrons (assumed to be proportional to the spin-down luminosity) in
the following form:

−2 t −2
Q∝γ 1+ , (58)
t0
with t0 = 5.3 × 103 year. This presentation corresponds to the case when the
current spin-down luminosity of the pulsar of age 1.7 × 104 year is smaller by
a factor of 10 compared to the spin-down luminosity at the birth of the pulsar
(the so-called braking index was assumed to be equal 3). Note that for a strong
magnetic field, the impact of the high initial luminosity is not significant. This is
explained by the fact that multi-TeV electrons produced in the past do not survive
92 F. Aharonian
Fig. 45 The spectral energy distributions of the synchrotron and inverse Compton radiation of a
PWN. The calculations are performed calculated for the steady injection (dotted lines) and time-
dependent injection (solid lines) of electrons given by Eq. (58). The curves correspond to different
strengths of the magnetic field. The shown fluxes correspond to the spin-down luminosity of the
pulsar L 0 = 1034 (d/1 kpc)2 erg/s, where d is the distance to the source
because of fast synchrotron losses. On the other hand, for a very weak magnetic field,
B ≤ 1 µG, the gamma-ray efficiency can significantly exceed 100 %. Obviously, this
does not imply a violation of conservation laws, but simply is a result of the cho-
sen definition of the efficiency when the gamma-ray luminosity is compared with
the current spin-down luminosity of the pulsar. In Fig. 45 the broad band spectral
energy distributions are shown for three different values of the magnetic field, and
assuming two different time histories of electron injection (following the evolution of
the pulsar): (i) constant injection rate (dotted lines) and (ii) time-dependent electron
injection given by Eq. (58) (solid lines).
For a very weak nebular magnetic field, B ≤ 1 µG, a pronounced IC peak is
formed around 10 TeV due to the radiative cooling break and the Klein-Nishina
effect. With an increase of the magnetic field, the IC peak moves towards low energies
and gradually disappears. However, in the case of higher injection rate of electrons
in the past, the IC peak survives even for very large nebular magnetic field. Quite
interestingly, a similar trend is observed also in the SED of synchrotron radiation;
with the increase of the magnetic field, both the synchrotron and IC peaks are shifted
towards low energies.
One of the nice features of VHE IC radiation on CMBR photons is its inde-
pendence of any model parameter, except for the spatial and energy distributions
of relativistic electrons. Thus, the high quality gamma-ray data can provide unam-
biguous information about the relativistic electrons at the present epoch. Coupled
with the synchrotron radiation, IC gamma-rays give us precise information about the
absolute strength and the spatial distribution of magnetic field over the nebula. They
also allow us to explore, under certain model assumptions, the history of electron
injection.
Fig. 46 The spectral energy

distributions of IC gamma-
rays arriving from different
regions of the nebula (see the
text for details)
Because of the effects related to the propagation and radiative cooling of electrons,
the IC radiation is characterized by strong variation of both the surface brightness
and the spectral shape of gamma-rays from different regions of the nebula. It is
demonstrated in Fig. 46 for SED of synchrotron and IC components of radiation
integrated over four different regions within 3, 10, 30, and 100 termination shock
radii. The pulsar’s spin-down luminosity is fixed at the level L 0 = 1.8 × 1037 erg/s,
and the magnetization parameter σ = 10−3 . The injection spectrum of electron has
been assumed power-law, dN /dE ∝ E −2.2 , with an exponential cutoff at 300 TeV.
Note that while in the inner region of the nebula (r ≤ 3rsh ) the IC component strongly
dominates over the synchrotron radiation, beyond r = 10rsh the contributions of
these components become comparable. This is explained by the hydrodynamical
amplification of the magnetic field which is implemented in calculations following
the formalism of Ref. [166]. While at the shock the field is expected quite week, less
than 1 µG (see Eq. (57), at larger distances it is increases up to a few µG.
Figure 46 shows that the SEDs of IC gamma-rays from the inner regions of the
nebula sharply peak at energy E 10 TeV. The reason is that we see the radiation of
uncooled electrons which quickly leave the compact region before suffering signifi-
cant radiative losses even in the case of relatively large magnetic field. This seems to
be the case of Vela X, a nearby PWN belongs to the powerful pulsar PSR J0835-4510
of age ≈11,000 year and spin-down luminosity L 0 = 7 × 1036 erg/s. Vela X has
been established [41] as one of the strongest TeV gamma-ray sources in the Galaxy.
The energy spectrum of the source is quite different from other galactic sources; it
is very hard at low energies with photon index Γ ≈ 1.5, and contains a high energy
exponential cutoff resulting in a distinct maximum in the SED at 10 TeV (see Fig. 47).
Because of the nearby location of the source (d ≈ 300 pc) we see, despite the
large angular size of the gamma-ray image of order of 1◦ (see Fig. 42), only the
central region with a linear size less than several pc. In this regard, Vela X is a
perfect object for exploration of processes in the inner parts of the nebula close to the
termination shock. The significantly improved sensitivity of the future Cherenkov
Telescope Array and its superior angular resolution (1–2 arcmin at 10 TeV) should
94 F. Aharonian
Fig. 47 The spectral energy

distribution of Vela X obtained
with the HESS telescope array.
The gamma-ray fluxes are
shown per solid angle from
two regions: (i) Θ < 0.8◦ (full
circles) and (ii) 0.8◦ ≤ Θ ≤
1.2◦ (open circles)
allow unique probes of relativistic electrons inside the region of the termination
shock, i.e. just at the heart of the accelerator.
On the other hand, for exploration of distributions of relativistic electrons and
magnetic fields beyond the termination shock, we need gamma-ray data up to at
least several degrees from the pulsar. In general, to image Vela X on such large
angular scales is a hard task for IACT arrays, especially given the expected decrease
of the surface brightness of the source, as well as its location in a very complex
region in the sky. Presently, the VHE gamma-ray studies of Vela X are limited by the
region of angular size 1.2◦ (see Fig. 47). Hopefully CTA will extend observations
beyond 2 or 3◦ , but even this will be not sufficient for calorimetric studies of the
entire source. Instead, the observations of other PWNe in environments of similar
properties and powered by pulsars of comparable age and spin-down luminosity,
but located at much larger distances, promise more information in this regard. In
particular, the PWN MSH 52-15 seems a good counterpart of Vela X, given the spin-
down luminosity L 0 = 1.7 × 1037 erg/s of the pulsar PSR B1509-58 that powers
MSH 52-15. The TeV gamma-ray image of this source is about 0.2◦ [34] which
implies, for the distance to the source of 5 kpc, a linear size of the gamma-ray nebula
of about 15 pc.
Another important representative of middle-age PWNe is the TeV gamma-ray
source HESS J1825-137, the nebula of the pulsar PSR J1826-1334. A remarkable
feature of this source is its unusually large angular size (of order of 1◦ ; see Fig. 42)
despite its far location (d ≈ 4 kpc) and relatively modest spin-down luminosity of
2.8 × 1036 erg/s. This implies huge linear size of the gamma-ray source extending
up to 80 pc which can be interpreted as the result of explosion of the supernova in an
extremely low-density environment to allow such a quick expansion of the nebula.
An alternative explanation could be that the external parts of the gamma-ray source
represent a IC halo produced by the cloud of ultrarelativistic electrons which have
left the nebula and propagate diffusively in the interstellar magnetic fields, a scenario
discussed in Ref. [52]. High quality data from this source have been recently obtained
Fig. 48 SED of broad-band IC gamma-ray emission of the pulsar wind nebula HESS J1825-137
calculated for 12 zones with a constant 6 arcmin width of the zones: 0 − 6 , 6 − 12 , . . . 66 − 72 .
The theoretical curves are shown together with observational points obtained with the Suzaku (the
inner 6 zones), Fermi LAT (the entire nebula), and HESS (all 12 zones) telescopes [242]
also in X-ray and GeV gamma-ray bands by the Suzaku [238] and Fermi LAT [146]
telescopes, respectively. Together with the reported spatial and spectral distributions
of TeV gamma-rays [40], these data allow detailed multi-zone and time-dependent
study of distributions of electrons and magnetic fields in the entire nebula. Such an
attempt has been made recently in Ref. [242]; a careful three-dimensional treatment
of the problem with inclusion of several important aspects like the magnetic field
and velocity profiles, diffusion of electrons, etc., allowed the authors to make a quite
strong conclusions [242]. While the energy-dependent profile of the gamma-ray mor-
phology, as well as the significantly smaller size of the X-ray image compared to
the gamma-ray image are naturally explained by the effects related to the radiative
cooling of electrons, the time-dependent 3D modelling of this source shows for that
for explanation of 10 TeV gamma-ray from outer parts of the nebula the diffusion
of electrons should be very fast, in particular the time of diffusive escape adopted in
the form τesc = 90(R/10 pc)2 (Ee /100 rmTeV)−1 , provides a very good fit of data
shown in Fig. 48. Also, the results of this study demand an initial period of the pulsar
close to 13 ms with the braking index n = 1.9. This implies an age of the source
of about 40,000 year, twice older that thought before. The best-fit model requires
96 F. Aharonian
for the interior velocity a profile close to v(r ) ∝ v −0.5 , while for the magnetic field
it demands B(r ) ∝ r −0.7 , with the field falling from a very large value of 400 µG
at the termination shock to ≈2 µG at the periphery of the nebula. Also, the model
requires quite
√ strong dependence of the magnetic field from the spin-down luminos-
ity, B(t) ∝ L(t), i.e. it should be much higher at the early epochs. Actually, this is a
quite significant deviation from the standard MHD picture [166] concerning both the
strength of the field at the termination shock and its radial dependence. Obviously,
these conclusions derived phenomenologically from the “bet-fit” requirement, need
also a thorough theoretical justification. Also, it is possible that one should treat the
TeV source as a combination of a more compact pulsar wind nebula and an extended
IC gamma-ray halo produced by electrons which already have left the nebula, and
diffusively expand in the interstellar medium. But in any case, the results of this work
nicely demonstrate the potential of gamma-ray observations, especially in the TeV
regime, for understanding the complex phenomena in PWNe.
In summary, the basic concept of formation of PWNe due to termination of
ultrarelativistic electron–positron pulsar winds is strongly supported by X-ray and
gamma-ray data both in GeV and TeV bands. In the case of most of TeV PWNe, the
inverse Compton origin of radiation is the only real option to explain the reported
fluxes, although in a few cases, e.g. for Vela X [158], the hadronic origin of gamma-
ray emission formally cannot be fully excluded (see for a review Ref. [92]). However,
the hadronic interpretation of TeV emission not only implies dramatic revision of the
generally accepted paradigm of PWNe, but it is a redundancy and, more importantly,
is not supported by observations.
8 Gamma-Ray Loud Binaries
In a binary system containing a relativistic object, a neutron start or a black hole,

the gravitational energy of the compact object is transformed into thermal emission
(typically in the X-ray band) radiated away by the hot accretion plasma. Therefore
these objects traditionally are treated as sources of thermal radiation. However, they
can operate also as particle accelerators. In fact, several different scenarios of effective
particle acceleration can be realized in these systems linked to either the termination
of the pulsar wind or to the internal shocks in the jet formed in the vicinity of the
black hole. Moreover, thanks to the presence of dense target material provided by
the companion star in the form of optical photons or gas, favourable conditions for
gamma-ray production can be established as well.
The history of developments related to gamma-radiation of binary systems is
rather controversial (see for a review Refs. [27, 247]). It is almost forgotten that
in the 1980s these objects constituted the highlight source population of ground-
based gamma-ray astronomy. The reports by several independent groups on detec-
tion of modulated TeV and PeV signals from Cyg X-3, Her X-1 and some other
similar objects [247] initiated a significant interest to VHE gamma-ray astronomy.
However, since the next generation of ground-based instruments with significantly
improved performance failed to confirm the early claims, these objects have not been
anymore treated as important targets for gamma-ray observations. This stance has
again changed after the discovery of galactic sources with relativistic jets, dubbed
microquasars [193]. A clear message of this discovery was the important role of
non-thermal processes, and hence possible effective gamma-ray production in these
accretion-driven objects.
8.1 Microquasars: Not yet Proved TeV Emitters
During strong flares, the non-thermal power of synchrotron jets of microquasars

(in the form of accelerated electrons and kinetic energy of the relativistic outflows)
could be comparable or even exceed the thermal X-ray luminosity of the central black
hole. The jets in microquasars are typically trans-relativistic, thus the acceleration of
electrons can proceed, in principle, at a very high rate. Correspondingly the spectrum
of synchrotron radiation of the jet can extend to hard X-rays [84, 191]. On the
other hand, the high density photon fields provided by the accretion disk around the
compact object, by the companion star, as well as produced by the jet itself, can
create favourable conditions for effective production of high energy IC gamma-rays
[84, 182, 215] provided that the ratio of energy densities of accelerated electrons
and the magnetic field does not significantly deviate from the equipartition condition.
Generally, this radiation is expected to have an episodic character associated with
strong radio flares in objects like the classical microquasar GRS 1915+105.
The shocks in microquasars can accelerate protons as well. But the gamma-ray
production through pp interactions may be effective only in high gas density regions
(see for a review Ref. [107]). In particular, the old “atmospheric target” (bombard-
ment of the normal star’s atmosphere by a relativistic beam of particles accelerated
at the compact object) or “target crosses beam” (interaction of a moving gas target
with a beam of relativistic particles) scenarios can provide non-negligible produc-
tion of high energy gamma-rays of hadronic origin. Although these models had been
inspired by the reports of detection of VHE signals from Cyg X-3, Her X-1, which
later were not confirmed (and eventually, discredited), the models themselves still
present a certain conceptual interest as possible scenarios of efficient gamma-ray
production.
Despite the several proposed viable scenarios of gamma-ray production in
microquasars, and intensive searches for VHE gamma-ray signals from some famous
representatives of this source population, we do not have yet a solid evidence in favour
of particle acceleration and their radiation in the VHE regime. A possible exception
is the marginal signal of VHE gamma-rays reported by the MAGIC collaboration
from Cyg X-1 [74]. However the astrophysical origin of the claimed episodic event
obviously needs further confirmation. Interestingly, a tentative (not yet confirmed)
detection of GeV gamma-ray flare of Cyg X-1 has been claimed by the AGILE col-
laboration [216]. It is difficult to overestimate the astrophysical significance of these
results in the case of their confirmation. AGILE and Fermi LAT telescopes are per-
98 F. Aharonian
fectly designed for such studies at GeV energies. In this regard, the detection by both
instruments of several flaring episodes from another famous microquasar, Cyg X-3
[4, 233], is a remarkable observational achievement with important astrophysical
implications (see e.g. Ref. [251]). On the other hand, the search for random/episodic
VHE gamma-ray events from variable sources with small FoV Cherenkov telescopes
is a quite hard task. Continuous monitoring of these source with future low-energy
threshold particle arrays like HAWK, seems to be a more promising approach.
8.2 Binary Pulsars
The second type of binary systems predicted as potential VHE gamma-ray emitters
are the systems which contain high spin-down luminosity pulsars [173, 232]. These
sources work as a compact pulsar wind nebula located in an environment where
the radiation and gas densities, as well as the magnetic field pressure are enhanced
by more than ten orders compared to the interstellar medium. Correspondingly, all
processes related to the termination of the shock, particle acceleration and radiation
proceed in much shorter time-scales. The best candidate representing this class of
sources for gamma-ray observations is PSR B1259-63/LS2883—a binary system
consisting of a 48 ms pulsar in highly eccentric orbit around a massive companion
star with a period 3.4 years. PSR B1259-63/LS 2883 was observed by the HESS array
of telescopes around three periastron passages in 2004, 2007 and 2010 with detection
at high statistical level on all three occasions [35, 47]. These observations revealed
a variable character of gamma-radiation. Three observations have been conducted
mainly either before or after the periastron passage. But fortunately some overlaps
of same orbital phases was possible indicating on the general repetitive behaviour
of the source. The cumulative lightcurves based on these observations periastron
show a hint of two peaks around the periastron with a tendency of flux decrease
towards the periastron. Unfortunately the source never has been observed exactly at
the periastron passage (it was possible in 2004, but unfortunately, due to the moon and
bad weather, such an opportunity was missed). The positions of the gamma-ray peaks
coincide with the eclipse of the radio pulsed emission. This indicates that the origin
of these peaks should have a relation to the position of the circumstellar disk. The
inverse Compton scattering seems to be the most plausible gamma-ray production
mechanism. However, while the measured spectrum is close to expectations [173],
neither the flux minimum at the periastron, nor the enhanced fluxes before and after
the periastron passage agree with the theoretical predictions [173] for the lightcurve.
Apparently, for interpretation of these results we have to invoke more sophisticated
scenarios than it was anticipated before.
Figures 49 and 50 show the variation of the VHE flux with the change of the
true anomaly and the overall differential energy spectrum, respectively. Despite the
significant changes of the absolute flux, no noticeable spectral variability was found,
in particular on time-scales of months, for the data taken between April and August
2007 [47]. The geometry of the system has an impact on the lightcurve through the
Fig. 49 Gamma-ray fluxes from PSR B1259-63 above 1 TeV as a function of the true anomaly.
The data are from the HESS observations in 2004 (daily fluxes; black points) and 2007 (monthly
fluxes; blue squares) [35, 47]. The red vertical line indicates the periastron passage. Also are shown
the level of the overall flux in 2005 and 2006 (green triangles)
Fig. 50 The differential energy spectrum of PSR B1259-63 based on 52.5 h observations between
April and August 2007. The spectrum is described by a power-law with flux normalisation F0 =
(1.1 ± 0.1stat ± 0.2sys ) × 10−12 TeV−1 cm−2 s−1 and photon index Γ = 2.8 ± 0.2stat ± 0.2sys [47]
anisotropic Compton scattering as the angle between the line of sight and the line
connecting stars changes with the orbital phase. On the other hand, several factors
can force the lightcurve of the IC signal to deviate from predictions based on the
pure geometrical treatment of the problem. A strong impact on the radiation of the
system has the variation of the magnetic field as well as non-radiative losses due to
the expansion of the shocked region. The enhanced synchrotron losses due to the
larger magnetic field close to the periastron could, in principle, explain the minimum
100 F. Aharonian
of the IC flux around the periastron passage. This should lead to an increase of the
synchrotron X-radiation which, however, is not observed. Just opposite, the X-ray
lightcurve contains a clear minimum at the periastron. It has been claimed [170] that
the dominance of adiabatic losses could be a major reason causing the tendency of
reduction of both X-ray and gamma-ray fluxes when the pulsar appears in the close
vicinity of the companion star during the periastron passage. The analysis of the
HESS data show that the TeV fluxes depend on the binary separation and hint at
a peculiar non-radiative cooling profile around the periastron [167]. The adiabatic
losses can be naturally linked to the interaction of the pulsar and stellar winds. The
modelling of hydrodynamics of such interactions [103] confirms that adiabatic losses
may indeed dominate over the radiative losses.The same interactions may lead to
formation of relativistic outflows. The detailed hydrodynamic and MHD simulations
revealed [103, 104] that already at distances comparable to the size of the binary
system the bulk Lorentz factor of the shocked flow can be as large as Γ 4. This
may result in significant anisotropy of radiation of accelerated electrons caused by
Doppler boosting. Indeed, since the Doppler factor depends on the location of the
pulsar along its orbit, it can modulate the IC gamma-ray flux [170], and thus lead to
reduction of X-ray and TeV gamma-ray signals close to the periastron.
Finally, one should mention another possible reason for suppression of the X- and
TeV gamma-ray emission; it can be caused by an “early” (sub-TeV) cutoff in the
energy spectrum of electrons due to the enhanced Compton losses around the perias-
tron [170]. An interesting feature of this interpretation is the expected anti correlation
of the GeV and TeV gamma-ray fluxes. In particular it predicts an increase of the
GeV flux in a narrow time-window around the periastron. The detection of the source
by Fermi LAT with a modest energy flux (a few times 10−11 erg/cm2 s) close to the
periastron passage [10] agrees quite well with this prediction, but unfortunately the
marginal gamma-ray signal as well as the inconsistency between spectra reported
by two independent groups [10, 226] does not allow a certain conclusion in this
regard. There is a hope that the TeV and GeV observations of the source during
the next periastron passage of the source will significantly enhance our knowledge
on temporal and spectral features of this source and correspondingly remove many
uncertainties related to the conditions in the gamma-ray production region. What
concerns the mechanism of production of gamma-rays, the arguments in favour of
its inverse Compton origin seem quite solid. The detection of TeV gamma-rays
approximately 50 days before the periastron passage in 2007 disfavours the stellar
disk scenario [47], and therefore the hadronic origin of gamma-rays, at least within
the current concept and knowledge regarding the stellar disk inclination, extension
and the density profile.
Spectacular Flare of PSR B1259-63/LS2883
As discussed in Sect. 7.2, pulsars can produce potentially detectable gamma-ray

emission also due to the bulk Comptonization of their cold ultrarelativistic winds.
While this radiation component of isolated pulsars is generally weak, unless the wind
is accelerated (relatively) close to the pulsar [100], in binary systems the radiation
is significantly enhanced because of the presence of target photons supplied by the
optical companion [88, 169, 170]. On the other hand, while IC radiation of isolated
pulsars can be readily separated from the extended synchrotron and IC emission of the
surrounding nebula with distinct spectral and spatial features, in binary pulsars the
separation of the radiation components produced before and after the termination
of the pulsar wind is a more difficult task; it requires rather careful treatment of
the spectral and temporal features of two radiation components. In this regard, the
most promising object for exploration of processes of formation, acceleration and
termination of pulsar winds in binary systems is PSR B1259-63/LS2883. While the
observations with the HESS telescopes allowed a meaningful constraint of the wind’s
Lorentz factor, Γ ≤ 106 , the main hope for detection of IC radiation of the unshocked
wind has been related to Fermi LAT [170].
The observations of PSR B1259-63/LS2883 with Fermi LAT around the periastron
passage in 2010–2011 revealed a rather complex behaviour of the system in GeV
gamma-rays [10, 226]. Close to the periastron passage, a rather weak signal has been
detected which could be related to IC radiation of the wind—both before and after its
termination [169]. Surprisingly, 30 days after the periastron passage, a spectacular
flare has been recorded. It lasted approximately two weeks with a flux enhanced by
more than an order of magnitude [10]. The flare is characterized by a sharp increase
and a smoother decay over approximately 2 weeks. During the flare, gamma-ray
luminosity has been increased to an extraordinary level, very close to the pulsar’s
spin-down luminosity L = 8×1035 erg/s. It is remarkable that such an extraordinary
event was not accompanied by any noticeable change of flux at any other wavelength!
Without any doubt, this flare represents a unique case in astrophysics when the
available energy is fully converted to non thermal high energy radiation. On the other
hand, in accordance with the current concept that the rotational energy of the pulsar is
converted to kinetic energy of cold ultrarelativistic electron–positron wind, the binary
pulsars can in principle operate as perfect (‘100 %’ efficient) gamma-ray emitters if
the density of the surrounding target photon field would be high enough for realization
of Comptonization of the pulsar wind in the saturation regime. Recently a model has
been suggested which under certain assumption can address the most outstanding
features of this unique event by radiation of the unshocked pulsar wind with a Lorentz
factor Γ0 104 . The combination of two effects both linked to the stellar disk, is
a principal element in the proposed model. The first effect is related to the impact
of the surrounding medium on the termination of the pulsar wind. Inside the disk,
the wind can suffer an “early” termination resulting in suppression of its gamma-ray
luminosity. When the pulsar escapes the disk, the conditions for termination of the
wind undergo significant changes. This would lead to a dramatic increase of the
pulsar wind zone, and thus to the proportional increase of the gamma-ray flux. On
the other hand, if the parts of the stellar disk disturbed by the pulsar can supply
infrared photons of density high enough for efficient (saturated) Comptonization of
the wind, almost the entire kinetic energy of the pulsar wind could be converted to
radiation, and thus the gamma-ray luminosity of the wind would achieve the level of
the pulsar’s spin-down luminosity.
102 F. Aharonian
Fig. 51 The sketch of the scenario explaining the gamma-ray flare of PSR B1259-63/LS2883
Fig. 52 SED of IC radiation of the unshocked pulsar wind at the epoch of 35 days after periastron
passage. Calculations for different values of the ratio of ram pressures η of the interacting pulsar and
stellar winds (η = 0.001—dotted lines; and η = 0.05—solid lines) are shown. The calculations are
performed for two target photon fields: (i) radiation of the companion optical star and (ii) radiation
of the stellar disk. Several initial pulsar wind bulk Lorentz factors indicated at the curves have been
assumed. The spectral points reported by the Fermi LAT for the post-periastron flaring episode are
shown with pentagons
The sketch of the proposed scenario is shown in Fig. 51. Calculations of spectral
energy distributions of IC radiation of the unshocked wind performed within this
scenario are presented in Fig. 52. The calculations for two key model parameters, the
Lorentz factor of the wind Γ0 = 104 and the ratio of ram pressures of the pulsar and
stellar wind pressures η = 0.05 explain quite well the measurements of Fermi LAT.
If the proposed interpretation is correct, this would be the second case, after the
Crab pulsar [62], of direct measurement of a pulsar wind’s Lorentz factor. However,
one should mention that there is another possible explanation of this extraordinary
flare, namely by the Doppler boosted radiation of the post-shock flow (see e.g.
Ref. [177]). Since the reported gamma-ray flux was close to pulsar’s spin-down
luminosity, strong boosting is a principal conditions of this interpretation Formally,
this is not an extreme assumption as it can be look at the first glance. Relativistic
flows can be formed at interactions of the pulsar and stellar winds [103], and there-
fore the broad-band radiation can be strongly affected by the Doppler boosting [128,
171]. The latter should amplify also the X- and TeV gamma ray fluxes. However,
the lack of any noticeable activity during the flare at other wavelengths makes this
interpretation less likely. Note that the apparent luminosity of the Doppler boosted
radiation can easily exceed the limit set by the pulsars spin-down luminosity, but
so far such flares have not been detected. On the other hand, the detection of such
events in future observations would rule out the origin of gamma-radiation related
to the unshocked pulsar wind.
8.3 Enigmatic TeV Binaries
Currently, the population of galactic binary systems detected in TeV gamma-rays

consists of a few objects. In addition to PSR B1259-63/LS 2883, TeV gamma-rays
have been reported from LS 5039 [36, 38] and LS I+61 303 [17, 18, 71]. Recently,
one more object joint the club of “TeV binaries”, HESS J0632+057. The source
was discovered serendipitously in the survey of the galactic plane by HESS. Because
of co-location with a variable radio and X-ray source at the position of the massive
star MWC 148, this unidentified TeV source has been suggested [155] to resemble
the TeV binary systems like LS I+61 303 or LS 5039. The subsequent gamma-ray
observations with VERITAS and HESS [188] revealed variability in the TeV gamma-
ray flux. Moreover, the X-ray observations with Swift [105] revealed a binary nature
of the source with periodicity of ≈320 days. No GeV gamma-rays have been yet
found from this source.
Except for PSR B1259-63/LS 2883, the nature of the compact stars in TeV
binaries remains highly unknown. LS I+61 303 and LS 5039 often are treated
as microquasars, i.e. objects with relativistic jets powered by accretion onto a
black hole. However, unlike the classical representatives of this source population,
e.g. GRS 1915+105 and Cyg X-1, these two objects are not univocally accepted
as microquasars. They indeed show features quite different from classical micro-
quasars. Therefore it has been advocated that they are analogues of the binary pulsar
104 F. Aharonian
PSR 1259-63 (see e.g. Ref. [126]). and therefore their non thermal activity is initiated
by pulsar winds.
The intensive searches over the last years could not, however, reveal pulsed sig-
nals from these objects. While the radio emission could not show up, in principle,
because of free-free absorption, the failure of detection of pulsed X-rays [211] can
be explained by the beam pointing away from the line of sight of observer. However,
even in this case it would be not so easy to hide the signatures of pulsars. For example,
at interaction with the pulsar wind, the stellar wind will be heated and emit thermal
X-ray emission. The lack of detection of thermal X-ray emission from LS 5039
implies a robust upper limit on the power of the pulsar wind [250] at the level close
to the luminosity of GeV gamma-rays reported by Fermi LAT [3]. This would imply
an extremely high efficiency of conversion of pulsar’s spin-down luminosity to GeV
gamma-rays. Alternatively one may assume that the gamma-radiation is strongly
Doppler-boosted which, of course, cannot be a priori excluded.
The lack of thermal X-ray emission has been also invoked as an argument against
the accretion origin of the power of compact objects in LS I+61 303 and LS 5039.
However, an interesting recent study of wind accretion on a rotating black hole [90]
in close binary systems shows that angular momentum of the accreted stellar wind
might be not sufficient for formation of accretion disk, thus the X-ray emission
would be suppressed. On the other hand, the direct wind accretion can activate
the Blanford-Znajek process, and, as a result, a powerful jet can launch without
accompanying thermal X-ray emission. It has been found that approximately 10 %
of the accretion power could be channelled into the jet and initiate observable non
thermal emission. Yet, in the case of LS 5039, one would need to invoke additional
resources, in particular Doppler boosting or significant anisotropy of emission to
explain the luminosity of reported GeV gamma-rays from this object.
The ambiguity related to the nature of the compact object in these binary systems
limits, but does not prevent detailed phenomenological studies of particle accelera-
tion and radiation processes in these complex systems. Over the last several years, the
origin of TeV gamma-ray emission of LS I+61 303 and LS 5039 has been discussed
in the context of their multi wavelength properties, under different assumptions con-
cerning the radiation and absorption processes, the sites of particle acceleration and
gamma-ray production, etc. Both sources have their specific feature, and should be
treated quite differently. In this regard, one should mention that LS I+61 303 is a
more “problematic” target for theoretical studies, despite the fact that it is one of the
deepest observed VHE sources. Due to the low fluxes, inconvenient for observations
orbital period and unpredictable behaviour at different phases, it is “much less well
characterized” [157] than LS 5039.
LS 5039 is a binary system consisting of a massive star of huge optical luminosity
( 1039 erg/s) in a 3.9 day orbit around a compact object. It shows non thermal
radio emission resolved as a bipolar-jet like structure with a speed v ≈ 0.2c [204].
TeV radiation from this source has been serendipitously discovered in 2005 during
the galactic plane survey by HESS [36]. The subsequent deep observation of the
source revealed a periodic nature of the signal—the source behaves as a perfect “TeV
gamma-ray clock” [38]. Later, periodic gamma-ray signal has been found by Fermi
Fig. 53 High and VHE gamma-ray spectra of LS 5039 measured by HESS [38] and Fermi [3] for
two broad orbital phase intervals defined as INFC 0.45 < φ ≤ 0.9 (red points) and SUPC φ ≤ 0.45
& φ > 0.9 (blue points). The black points and the dotted line represent the phase-averaged Fermi
LAT spectrum
LAT also at GeV energies [3]. Both the GeV and TeV radiation components are
modulated with 3.9 day period which coincides with the orbital period of the system
derived from the Doppler-shifted optical lines. The major fraction of gamma-ray
emission is contributed by the half of the orbit with a maximum close to the so-called
inferior conjunction (the epoch when the compact object is lined-up along our line-
of-sight in front of the optical star). Figure 53 shows the energy spectra of the source
averaged in two broad phase intervals: (i) 0.45 ≤ φ ≤ 0.9 and (ii) φ ≤ 0.45 &
φ ≥ 0.9. While the spectrum corresponding the phase interval (i) can be fitted by a
hard power-law with a photon index ≈1.85 and exponential cutoff around 9 TeV, the
interval (ii) is characterized by a quite different spectrum—a single power-law with
a photon index ≈2.5.
The time variation of the absolute flux of TeV gamma-rays can be partly explained
by the photon–photon absorption caused by the variation of the geometry and density
of the stellar radiation field [108, 125]. However, the absorption effect alone hardly
can explain the observed modulation of the energy spectrum of gamma-rays. Indeed,
while for the given temperature of radiation of the companion star kT = 3.5 eV,
the gamma-ray absorption should result in strongest modulation of the flux at low
energies, E ≤ 1 TeV, the HESS data show just opposite behaviour—a quite stable
gamma-ray flux at 0.2 TeV, but strongly variable emission at energies above 1 TeV.
Obviously, photon–photon pair production cannot be responsible for modulation of
GeV gamma-rays either; the energy of these gamma-rays is well below the interaction
threshold with optical photons. Thus, it is clear that additional processes should be
invoked to explain the observed time variation of the energy spectrum and the absolute
flux. In particular,
106 F. Aharonian
• variation of the maximum energy of accelerated electrons established in a radiation

dominated environment may have a strong impact on the temporal behaviour of
TeV gamma-ray spectra [38, 170]. While this effect may lead to a modulation of
synchrotron X-rays produced, most likely, by the same population of electrons, its
impact on the lightcurve of IC gamma-rays at GeV energies is quite small;
• the effect related to the cross-section of anisotropic Compton scattering, undoubt-
edly plays an important role in modulation of both GeV and TeV gamma-ray
signals [38, 127, 168];
• the lightcurves of both synchrotron and IC components of radiation may be affected
by hydrodynamic effects like adiabatic losses due to expansion of the production
regions [225] and Doppler boosting of radiation due to relativistic outflows formed
at the interaction of pulsar and stellar winds [128, 171].
The combination of all these effects results in rather complex correlations between
different radiation components. Figure 54 demonstrates that the TeV and X-ray
lightcurves can be described quite well by the same population of electrons under cer-
tain assumptions regarding the adiabatic and radiative losses of electrons. In Fig. 54
the phase dependence of adiabatic losses are derived phenomenologically from the
Suzaku data. Although the adiabatic losses modulate the IC gamma-ray flux in a
similar manner, the TeV gamma-rays unlike X-rays suffer significant distortion due
to photon–photon absorption. It is likely that these two additional effects related to
interactions of gamma-rays and electrons with the stellar radiation field are respon-
sible for stronger changes of VHE gamma-ray flux compared to the X-ray flux.
Within any IC model of gamma-ray emission of LS 5039, the spectrum continues
down to low energies with a peak around 10 GeV, unless one assumes a sharp low-
energy cutoff in the energy spectrum of electrons. In the case of development of
electromagnetic cascades triggered by photon–photon pair production, a maximum
in SED appears close to 30 GeV, independent of the spectrum of primary electrons
[26]. Therefore the report of the Fermi collaboration about detection of high energy
gamma-rays was not a big surprise, especially given the detected change of the signal
with orbital phase [3], which suggests a link to the VHE component of radiation.
Nevertheless, the reported high flux and the energy spectrum with a cutoff at 2 GeV
(see Fig. 53) challenge most of models suggested in the literature. The proposed in
Ref. [3] interpretation that the presence of a cutoff in the spectrum may be indicative
of magnetospheric emission, does not seems a realistic option since it cannot explain
the origin of modulation of the signal. The GeV signal challenges also the non
thermal energy budget of the source. In this regard, a possible Doppler boosting can
help significantly to reduce the energetic requirements to the source.
The lack of information about the nature of the compact object in LS 5039 prevents
us from comprehensive modelling of high energy processes in the source. However,
a number of important conclusions can be derived from a general analysis of condi-
tions concerning the acceleration of electrons and the radiation/absorption processes
of gamma-rays. For example, the sheer fact of detection of gamma-rays with energy
exceeding 10 TeV implies an extremely fast acceleration of electrons close to the
theoretical limit; the parameter η in Eq. (28) which characterizes the acceleration
(a)
(b)
(c)
(d)
(e)
Fig. 54 Light curves characterizing LS 5039 [225]. a Phase dependence of adiabatic losses derived
directly from the Suzaku data; b theoretical predictions for 1 TeV gamma-ray fluxes together
with HESS data points; c calculations for photon indices around 1 TeV together with power-law
indices reported by HESS; d calculated 1–10 keV synchrotron X-ray fluxes shown together with the
Suzaku data; e predicted IC flux integrated over the interval 1–100 GeV. The total power injected in
relativistic electrons with E −2 type spectrum is fixed at the level of 1037 erg/s. The cooled electron
spectrum is formed due to radiative (IC and synchrotron) and non-radiative (adiabatic) losses
efficiency cannot significantly exceed 10 [168]. Even so, the accelerator should be
located at periphery of the binary system, otherwise the severe energy losses of elec-
trons in high density radiation field of the companion star will prevent acceleration
to multi-TeV energies.
An independent conclusion regarding the location of the TeV gamma-ray pro-
duction region can be derived from the analysis of the allowed level of gamma-ray
absorption. If the gamma-ray emitter is located close to the compact object (i.e. deep
inside the binary system), the magnetic field of the companion star should be less
than a few Gauss. This is a necessary condition to allow pair cascades to develop
and thus to reduce effectively the photon–photon absorption opacity. However, the
magnetic field in O stars is expected to be much larger. For these (more realistic)
fields the development of the cascade is suppressed; the major fraction of the energy
108 F. Aharonian
absorbed in the stellar photon field will be emitted by secondary pairs via synchrotron
radiation in the X-ray band. In order to avoid overproduction of X-rays (compared
to the detected fluxes), the TeV gamma-ray emitter should be located at distances
≥1012 cm from the compact object. This excludes the standard pulsar scenarios sug-
gested for LS 5039 in which the emitter is located close to the pulsar (see e.g. Ref.
[127]), but does not exclude the pulsar origin of the compact object. The supersonic
outflow formed at interaction of pulsar and stellar winds, may effectively transport
the available energy to the periphery of the binary system. Alternatively the energy
of the compact object can be transported by a jet powered by radiatively inefficient
stellar wind accretion onto a rotating black hole [90].
In summary, further multi-wavelength studies are needed to shed light on the
nature of the compact object in LS 5039. The search for distinct pulsar or black-hole
signatures seems to crucial in this regard. However, we perhaps should be prepared
for discoveries of new observational features of this unusual object which could
require a dramatic revision of standard concepts of current paradigms.
9 Summary
The remarkable success of observational gamma-ray astronomy, together with recent

intensive theoretical and phenomenological studies of acceleration and radiation
processes in astrophysical environments, supply a strong rationale for the further
exploration of the sky at high and very-high energies.
In this chapter I tried to highlight the major results and advances in galactic
VHE gamma-ray astronomy. The achievements in the extragalactic VHE gamma-ray
astronomy are equally impressive with very important astrophysical and cosmolog-
ical implications.1
The recent observational results have introduced non-negligible corrections and
additions to the major scientific objections and motivations of gamma-ray astronomy.
In particular, new interesting features have been revealed which require significant
revisions of current theoretical models or even formulations of new concepts. It is
expected that over the next decade the ongoing operation of Fermi LAT will be
accompanied by observations with the next generation ground-based detectors, in
particular CTA and HAWK. The data obtained in the enormous energy range from
100 MeV to 1 PeV will provide deep insight into a number of problems of high energy
astrophysics and fundamental physics. In the area of galactic gamma-ray astronomy,
I would put an emphasis on the following major objectives:
Identification of SNRs as the major source population responsible for the pro-
duction of galactic cosmic rays. Since these sources should operate, by definition,
1 A few selected topics linked to the origin of VHE gamma-ray emission of blazers and to the
propagation of gamma-rays through the extragalactic radiation fields, have been partly covered in
my Saas-Fee lectures. However, because of the chosen format and the imposed page limit, I decided
not include them in this chapter.
as PeVatrons, the extension of gamma-ray observations up to 100 TeV and beyond

seems to be the key condition for the success. Another breakthrough is expected from
the improvement of the point spread function (PSF) of the atmospheric Cherenkov
telescope arrays. An angular resolution of about 1 arcmin seems to be crucial, espe-
cially in the case of observations of young SNRs. Fortunately, such resolution in
principle is achievable in the most important, E ≥ 10 TeV energy band. This should
allow purposeful searches for radiative signatures of PeV protons beyond the shells
of young SNRs. Since the highest energy particles escape the shell typically after
a few hundred years or even earlier, the radiative signatures of PeV particles are
expected from the regions just beyond the shell rather than from the shell itself.
The observations with an arcminute resolution should also allow localization of the
gamma-ray production regions associated with dense core clumps in the shells, as
well as with reverse shocks. Finally, the improvement of the sensitivity of the future
IACT arrays by an order of magnitude should allow an extension of the distances to
SNRs detectable in TeV gamma-rays up to 10 kpc. Thus, one may anticipate a dra-
matic increase of the number of TeV emitting young SNRs, many of which without
previously detected counterparts at other wavelengths!
Search for other principal contributors to the galactic cosmic rays. Deep mul-
tiwavelength studies of the so-called “dark” (so far not identified) TeV gamma-
ray sources may reveal a new population(s) of effective particle accelerators. We
may anticipate also discovery of a large number of new “dark” TeV emitters in
the surveys with the next generation ground-based detectors. The arrays of (rela-
tively) large-field-of-view Cherenkov telescopes designed for detection of gamma-
rays with angular resolution of order of 1 arcmin, primarily in the E ≥ 10 TeV
energy band, can be especially effective for serendipitous surveys of the galactic
plane. Together with the ultrahigh energy gamma-rays, the hard X-ray imaging tele-
scopes aboard the recently launched NuSTAR and the future ASTRO-H satellites,
should be able to conduct effective searches of currently active PeVatrons through
the synchrotron radiation of secondary electrons produced in hadronic interactions.
A complementary information will be provided by Fermi LAT at lower energies. In
particular, the gamma-ray observations from massive molecular clouds which oper-
ate as unique cosmic-ray barometers, can reveal the “smoking guns”—remnants of
ancient (already faded away) PeVatrons, as well as provide an information about
the environments which harbor these most effective cosmic ray accelerators in our
Galaxy.
Physics of pulsar winds and pulsar wind nebulae. The large number of TeV emit-
ting PWNe powered by pulsars of different age and spin-down luminosities provide
an excellent observational material for comprehensive studies of the physics of for-
mation and termination of ultrarelativistic pulsar winds. The unique feature of TeV
gamma-ray emission is the dominance of the inverse Compton scattering of electrons
on the photons of 2.7 K CMBR. This allows models-independent derivation of the
energy spectra and spatial distributions of parent electrons. Combined with the X-ray
data, accurate maps of magnetic fields can be extracted, at least for the central com-
pact regions where X-ray data are typically available. In this regard, the nebula of the
110 F. Aharonian
Vela Pulsar is of special interest because of its proximity and high spin-down lumi-
nosity. The improvement of the flux sensitivity by an order of magnitude coupled with
the angular resolution of the order of ∼1 arcmin achievable by the future atmospheric
Cherenkov telescope arrays at energies above 10 TeV, should allow unique probes of
relativistic electrons and magnetic fields inside the termination shock! Similar stud-
ies are possible also for the Crab Nebula, but invoking the temporal characteristics of
the synchrotron gamma-ray emission during the recently established quasi-regular
flares in the MeV–GeV band. The realization of these possibilities will provide a
key insight into the magnetohydrodynamics and particle acceleration processes at
the termination of pulsar winds resulting in formation of nonthermal synchrotron
and IC nebulae. On the other hand, the detailed studies of energy-dependent mor-
phologies of distant PWNe with large angular extensions (like HESS J1825-137),
will greatly contribute to the understanding of the escape of relativistic electrons
from the nebula and their diffusive propagation in the interstellar magnetic fields.
What concerns the formation of the cold ultrarelativistic pulsar winds and explo-
ration of their characteristics, in particular the location and maximum energy of the
bulk motion acceleration, a unique and powerful method in this regard could be
the combination of detailed temporal and spectroscopic studies of the characteristic
pulsed gamma-ray emission of the Crab and perhaps some other high spin-down
luminosity pulsars at energies from 10 GeV to a few TeV. The practical realization of
such studies can be conducted by CTA with significantly improved (compared to the
current telescope systems) flux sensitivities in the energy interval from tens of GeV
to a few TeV.
Solving puzzles and revealing origin(s) of VHE binaries. So far, only several
compact/variable VHE gamma-ray sources have be discovered in the Galaxy. The
nature of these objects, except for PSR B1259-63/LS 2883, remains highly uncertain.
PSR B1259-63/LS 2883 is firmly identified as a binary pulsar in which the processes
of formation of the pulsar wind, and its termination with the subsequent particle
acceleration proceed quite similarly to PWNe, but on much (orders of magnitude)
shorter timescales. Therefore, this source offers a unique opportunity to study the
complex MHD and particle acceleration processes in “on-line” regime. Although the
gamma-ray fluxes from this object are weak both at GeV and TeV energies, for com-
prehensive temporal studies with Fermi LAT and HESS, the available data obtained
during the epochs around the periastron of the system revealed significant deviations
from the early predictions. This concerns the tendency of a minimum TeV flux at
the periastron, the surprise GeV flare after several weeks of the periastron passage,
etc. Despite the attempts to explain the strange behavior of the source, it is clear
that a detailed theory can be developed only based on the higher quality gamma-ray
data. The other TeV binaries are even more enigmatic. Although originally dubbed
as microquasars, they show features quite different from the classical binary systems
containing black holes. Therefore they have been advocated to be binary pulsars.
On the other hand, no pulsed signals have been observed from these sources at any
wavelength. A more exciting scenario for these objects could be a specific regime of
accretion onto black hole—without formation of an accretion disk, but with a launch
of a powerful jet, e.g. due to the Blandford-Znajek process. One may anticipate that
intensive theoretical studies will be conducted in both directions, but it is likely that
a decisive answer on the origin of these objects in general, and on the nature of
acceleration and radiation processes in particular, will come from new observations
with the next generation of ground-based detectors. The performance of CTA con-
cerning both the broader energy coverage and higher flux sensitivity, will be adequate
(hopefully) to address all these issues.
References
1. Abdo, A.A., Ackermann, M., Ajello, M., Atwood, W.B., Axelsson, M., Baldini, L., Ballet, J.,
Barbiellini, G., Baring, M.G., Bastieri, D., et al.: The First Fermi Large Area Telescope
Catalog of Gamma-ray Pulsars. ApJS 187, 460–494 (2010)
2. Abdo, A.A., et al. (Fermi collaboration): A Population of Gamma-Ray Millisecond Pulsars
Seen with the Fermi Large Area Telescope. Science 325, 848– (2009)
3. Abdo, A.A., et al. (Fermi collaboration): Fermi/LAT observations of LS 5039. ApJ 706,
L56–L61 (2009)
4. Abdo, A.A., et al. (Fermi collaboration): Modulated High-Energy Gamma-Ray Emission
from the Microquasar Cygnus X-3. Science 326, 1512– (2009)
5. Abdo, A.A., et al. (Fermi collaboration): Fermi Large Area Telescope Observations of the
Crab Pulsar And Nebula. ApJ 708, 1254–1267 (2010)
6. Abdo, A.A., et al. (Fermi collaboration): Fermi Large Area Telescope Observations of the
Supernova Remnant W28 (G6.4-0.1). ApJ 718, 348–356 (2010)
7. Abdo, A.A., et al. (Fermi collaboration): Fermi-Lat Discovery of GeV Gamma-Ray Emission
from the Young Supernova Remnant Cassiopeia A. ApJ 710, L92–L97 (2010)
8. Abdo, A.A., et al. (Fermi collaboration): Gamma-Ray Emission from the Shell of Supernova
Remnant W44 Revealed by the Fermi LAT. Science 327, 1103– (2010)
9. Abdo, A.A., et al. (Fermi collaboration): Observation of Supernova Remnant IC 443 with the
Fermi Large Area Telescope. ApJ 712, 459–468 (2010)
10. Abdo, A.A., et al. (Fermi collaboration): Discovery of High-energy Gamma-ray Emission
from the Binary System PSR B1259-63/LS 2883 around Periastron with Fermi. ApJ 736, L11
(2011)
11. Abdo, A.A., et al. (Fermi collaboration): Gamma-Ray Flares from the Crab Nebula. Science
331, 739– (2011)
12. Abdo, A.A., et al. (Fermi collaboration): Observations of the Young Supernova Remnant RX
J1713.7-3946 with the Fermi Large Area Telescope. ApJ 734, 28 (2011)
13. Abdo, A.A., et al. (Milagro Collaboration): TeV Gamma-Ray Sources from a Survey of the
Galactic Plane with Milagro. ApJ 664, L91–L94 (2007)
14. Abramowski, A., et al. (HESS collaboration): A new SNR with TeV shell-type morphology:
HESS J1731–347. A&A 531, A81 (2011)
15. Acciari, V.A., et al. (VERITAS collaboration): Observation of Extended Very High Energy
Emission from the Supernova Remnant IC 443 with VERITAS. ApJ 698, L133–L137 (2009)
16. Acciari, V.A., et al. (VERITAS collaboration): Discovery of TeV Gamma-ray Emission from
Tycho’s Supernova Remnant. ApJ 730, L20 (2011)
17. Acciari, V.A., et al. (VERITAS collaboration): VERITAS Observations of the TeV Binary
LSI + 61303 During 2008–2010. ApJ 738, 3 (2011)
18. Acciari, V.A., et al. (VERTAS collaboration): Multiwavelength Observations of LSI + 61303
with Veritas, Swift, and RXTE. ApJ 700, 1034–1041 (2009)
19. Acciari, V.A., et al. (VERTAS collaboration): Observations of the Shell-type Supernova Rem-
nant Cassiopeia A at TeV Energies with VERITAS. ApJ 714, 163–169 (2010)
112 F. Aharonian
20. Acero, F., et al. (HESS collaboration): First detection of VHE γ-rays from SN 1006 by HESS.
A&A 516, A62 (2010)
21. Acero, F., et al. (HESS Collaboration): Localizing the VHE γ-ray source at the Galactic
Centre. MNRAS 402, 1877–1882 (2010)
22. Achterberg, A., Gallant, Y.A., Kirk, J.G., Guthmann, A.W.: Particle acceleration by ultrarel-
ativistic shocks: theory and simulations. MNRAS 328, 393–408 (2001)
23. Actis, M., et al. (CTA consortium): Design concepts for the Cherenkov Telescope Array
CTA: an advanced facility for ground-based high-energy gamma-ray astronomy. Experimental
Astronomy 32, 193–316 (2011)
24. Adriani, O., Barbarino, G.C., Bazilevskaya, G.A., et al. (Pamela collaboration): An anomalous
positron abundance in cosmic rays with energies 1.5–100 GeV. Nature 458, 607–609 (2009)
25. Adriani, O., Barbarino, G.C., Bazilevskaya, G.A., et al. (Pamela collaboration): PAMELA
Measurements of Cosmic-Ray Proton and Helium Spectra. Science 332, 69–71 (2011)
26. Aharonian, F., Anchordoqui, L., Khangulyan, D., Montaruli, T.: Microquasar LS 5039: a TeV
gamma-ray emitter and a potential TeV neutrino source. Journal of Physics Conference Series
39, 408–415 (2006)
27. Aharonian, F., Buckley, J., Kifune, T., Sinnis, G.: High energy astrophysics with ground-based
gamma ray detectors. Reports on Progress in Physics 71(9), 096,901 (2008)
28. Aharonian, F., et al. (HEGRA collaboration): Evidence for TeV gamma ray emission from
Cassiopeia A. A&A 370, 112–120 (2001)
29. Aharonian, F., et al. (HEGRA collaboration): An unidentified TeV source in the vicinity of
Cygnus OB2. A&A 393, L37–L40 (2002)
30. Aharonian, F., et al. (HEGRA Collaboration): Variations of the TeV energy spectrum at
different flux levels of Mkn 421 observed with the HEGRA system of Cherenkov telescopes.
A&A 393, 89–99 (2002)
31. Aharonian, F., et al. (HEGRA Collaboration): The Crab Nebula and Pulsar between 500 GeV
and 80 TeV: Observations with the HEGRA Stereoscopic Air Cerenkov Telescopes. ApJ 614,
897–913 (2004)
32. Aharonian, F., et al. (HESS collaboration): Very high energy gamma rays from the direction
of Sagittarius A∗ . A&A 425, L13–L17 (2004)
33. Aharonian, F., et al. (HESS collaboration): Detection of TeV γ-ray emission from the shell-
type supernova remnant RX J0852.0-4622 with HESS. A&A 437, L7–L10 (2005)
34. Aharonian, F., et al. (HESS collaboration): Discovery of extended VHE gamma-ray emission
from the asymmetric pulsar wind nebula in MSH 15–52 with HESS. A&A 435, L17–L20
(2005)
35. Aharonian, F., et al. (HESS collaboration): Discovery of the binary pulsar PSR B1259-63 in
very-high-energy gamma rays around periastron with HESS. A&A 442, 1–10 (2005)
36. Aharonian, F., et al. (HESS collaboration): Discovery of Very High Energy Gamma Rays
Associated with an X-ray Binary. Science 309, 746–749 (2005)
37. Aharonian, F., et al. (HESS collaboration): Very high energy gamma rays from the composite
SNR G0.9 + 0.1. A&A 432, L25–L29 (2005)
38. Aharonian, F., et al. (HESS collaboration): 3.9 day orbital modulation in the TeV γ-ray flux
and spectrum from the X-ray binary LS 5039. A&A 460, 743–749 (2006)
39. Aharonian, F., et al. (HESS Collaboration): Discovery of very-high-energy γ-rays from the
Galactic Centre ridge. Nature 439, 695–698 (2006)
40. Aharonian, F., et al. (HESS collaboration): Energy dependent γ-ray morphology in the pulsar
wind nebula HESS J1825-137. A&A 460, 365–374 (2006)
41. Aharonian, F., et al. (HESS collaboration): First detection of a VHE gamma-ray spectral
maximum from a cosmic source: HESS discovery of the Vela X nebula. A&A 448, L43–L47
(2006)
42. Aharonian, F., et al. (HESS collaboration): Observations of the Crab nebula with HESS. A&A
457, 899–915 (2006)
43. Aharonian, F., et al. (HESS collaboration): Primary particle acceleration above 100 TeV in
the shell-type supernova remnant RX J1713.7-394 with deep HESS observations. A&A 464,
235–243 (2007)
44. Aharonian, F., et al. (HESS collaboration): Discovery of very high energy gamma-ray emission
coincident with molecular clouds in the W 28 (G6.4-0.1) field. A&A 481, 401–410 (2008)
45. Aharonian, F., et al. (HESS collaboration): Discovery of Gamma-Ray Emission From the
Shell-Type Supernova Remnant RCW 86 With Hess. ApJ 692, 1500–1505 (2009)
46. Aharonian, F., et al. (HESS collaboration): Spectrum and variability of the Galactic center
VHE γ-ray source HESS J1745-290. A&A 503, 817–825 (2009)
47. Aharonian, F., et al. (HESS collaboration): Very high energy γ-ray observations of the binary
PSR B1259-63/SS2883 around the 2007 Periastron. A&A 507, 389–396 (2009)
48. Aharonian, F., Neronov, A.: High-Energy Gamma Rays from the Massive Black Hole in the
Galactic Center. ApJ 619, 306–313 (2005)
49. Aharonian, F., Neronov, A.: TeV Gamma Rays From the Galactic Center Direct and Indirect
Links to the Massive Black Hole in Sgr A. Ap&SS 300, 255–265 (2005)
50. Aharonian, F., Taylor, A.M.: Limitations on the photo-disintegration process as a source of
VHE photons. Astroparticle Physics 34, 258–266 (2010)
51. Aharonian, F.A.: TeV gamma rays from BL Lac objects due to synchrotron radiation of
extremely high energy protons. New Astr. 5, 377–395 (2000)
52. Aharonian, F.A.: Very high energy cosmic gamma radiation: a crucial window on the extreme
Universe. River Edge, NJ: World Scientific Publishing (2004)
53. Aharonian, F.A., Akerlof, C.W.: Gamma-Ray Astronomy with Imaging Atmospheric
Cerenkov Telescopes. Annual Review of Nuclear and Particle Science 47, 273–314 (1997)
54. Aharonian, F.A., Atoyan, A.M.: Compton scattering of relativistic electrons in compact X-ray
sources. Ap&SS 79, 321–336 (1981)
55. Aharonian, F.A., Atoyan, A.M.: Cosmic gamma-rays associated with annihilation of relativis-
tic e+ –e− pairs. Physics Letters B 99, 301–304 (1981)
56. Aharonian, F.A., Atoyan, A.M.: On the emissivity of π 0̂ˆ-decay gamma radiation in the vicinity
of accelerators of galactic cosmic rays. A&A 309, 917–928 (1996)
57. Aharonian, F.A., Atoyan, A.M.: On the origin of TeV radiation of SN 1006. A&A 351, 330–
340 (1999)
58. Aharonian, F.A., Atoyan, A.M.: Broad-band diffuse gamma ray emission of the galactic disk.
A&A 362, 937–952 (2000)
59. Aharonian, F.A., Atoyan, A.M., Kifune, T.: Inverse Compton gamma radiation of faint syn-
chrotron X-ray nebulae around pulsars. MNRAS 291, 162–176 (1997)
60. Aharonian, F.A., Atoyan, A.M., Nagapetyan, A.M.: Photoproduction of Electron–Positron
Pairs in Compact X-Ray Sources. Astrophysics 19, 187 (1983)
61. Aharonian, F.A., Belyanin, A.A., Derishev, E.V., Kocharovsky, V.V., Kocharovsky, V.V.: Con-
straints on the extremely high-energy cosmic ray accelerators from classical electrodynamics.
Phys. Rev. D 66(2), 023005 (2002)
62. Aharonian, F.A., Bogovalov, S.V., Khangulyan, D.: Abrupt acceleration of a ‘cold’ ultrarela-
tivistic wind from the Crab pulsar. Nature 482, 507–509 (2012)
63. Aharonian, F.A., Drury, L.O., Voelk, H.J.: GeV/TeV gamma-ray emission from dense mole-
cular clouds overtaken by supernova shells. A&A 285, 645–647 (1994)
64. Aharonian, F.A., et al. (HESS collaboration): High-energy particle acceleration in the shell
of a supernova remnant. Nature 432, 75–77 (2004)
65. Aharonian, F.A., Kelner, S.R., Prosekin, A.Y.: Angular, spectral, and time distributions of
highest energy protons and associated secondary gamma rays and neutrinos propagating
through extragalactic magnetic and radiation fields. Phys. Rev. D 82(4), 043002 (2010)
66. Aharonian, F.A., Konopelko, A.K., Völk, H.J., Quintana, H.: 5@5—a 5 GeV energy threshold
array of imaging atmospheric Cherenkov telescopes at 5 km altitude. Astroparticle Physics
15, 335–356 (2001)
67. Aharonian, F.A., Kririllov-Ugriumov, V.G., Vardanian, V.V.: Formation of relativistic
electron–photon showers in compact X-ray sources. Ap&SS 115, 201–225 (1985)
68. Aharonian, F.A., Plyasheshnikov, A.V.: Similarities and differences between relativistic
electron–photon cascades developed in matter, photon gas and magnetic field. Astroparti-
cle Physics 19, 525–548 (2003)
114 F. Aharonian
69. Aharonian, F.A., Sunyaev, R.A.: Gamma-ray line emission, nuclear destruction and neu-
tron production in hot astrophysical plasmas—The deuterium boiler as a gamma-ray source.
MNRAS 210, 257–277 (1984)
70. Akhiezer, A.I., Merenkov, N.P., Rekalo, A.P.: On a kinetic theory of electromagnetic showers
in strong magnetic fields. Journal of Physics G Nuclear Physics 20, 1499–1514 (1994)
71. Albert, J., et al. (MAGIC collaboration): Variable Very-High-Energy Gamma-Ray Emission
from the Microquasar LS I +61 303. Science 312, 1771–1773 (2006)
72. Albert, J., et al. (MAGIC collaboration): Discovery of Very High Energy Gamma Radiation
from IC 443 with the MAGIC Telescope. ApJ 664, L87–L90 (2007)
73. Albert, J., et al. (MAGIC collaboration): Observation of VHE γ-rays from Cassiopeia A with
the MAGIC telescope. A&A 474, 937–940 (2007)
74. Albert, J., et al. (MAGIC collaboration): Very High Energy Gamma-Ray Radiation from the
Stellar Mass Black Hole Binary Cygnus X-1. ApJ 665, L51–L54 (2007)
75. Aleksic, J., et al. (MAGIC collaboration): Phase-resolved energy spectra of the Crab pulsar
in the range of 50–400 GeV measured with the MAGIC telescopes. A&A 540, A69 (2012)
76. Aliu, E., et al. (VERITAS collaboration): Detection of Pulsed Gamma Rays Above 100 GeV
from the Crab Pulsar. Science 334, 69– (2011)
77. Anchordoqui, L.A., Beacom, J.F., Goldberg, H., Palomares-Ruiz, S., Weiler, T.J.: TeV γ rays
and neutrinos from photodisintegration of nuclei in Cygnus OB2. Phys. Rev. D 75(6), 063001
(2007)
78. Anguelov, V., Vankov, H.: Electromagnetic showers in a strong magnetic field. Journal of
Physics G Nuclear Physics 25, 1755–1764 (1999)
79. Atoyan, A., Buckley, J., Krawczynski, H.: A Gamma-Ray Burst Remnant in Our Galaxy:
HESS J1303-631. ApJ 642, L153–L156 (2006)
80. Atoyan, A., Dermer, C.: Gamma Rays from the Tycho Supernova Remnant: Leptonic or
Hadronic? ArXiv e-prints (2011)
81. Atoyan, A., Dermer, C.D.: TeV Emission from the Galactic Center Black Hole Plerion. ApJ
617, L123–L126 (2004)
82. Atoyan, A.M.: Radio spectrum of the Crab nebula as an evidence for fast initial spin of its
pulsar. A&A 346, L49–L52 (1999)
83. Atoyan, A.M., Aharonian, F.A.: On the mechanisms of gamma radiation in the Crab Nebula.
MNRAS 278, 525–541 (1996)
84. Atoyan, A.M., Aharonian, F.A.: Modelling of the non-thermal flares in the Galactic micro-
quasar GRS 1915 + 105. MNRAS 302, 253–276 (1999)
85. Atoyan, A.M., Aharonian, F.A., Tuffs, R.J., Völk, H.J.: On the gamma-ray fluxes expected
from Cassiopeia A. A&A 355, 211–220 (2000)
86. Atwood, W.B., Abdo, A.A., Ackermann, M., Althouse, W., Anderson, B., Axelsson, M.,
Baldini, L., Ballet, J., Band, D.L., Barbiellini, G., et al.: The Large Area Telescope on the
Fermi Gamma-Ray Space Telescope Mission. ApJ 697, 1071–1102 (2009)
87. Balbo, M., Walter, R., Ferrigno, C., Bordas, P.: Twelve-hour spikes from the Crab Pevatron.
A&A 527, L4 (2011)
88. Ball, L., Kirk, J.G.: Probing pulsar winds using inverse compton scattering. Astroparticle
Physics 12, 335–349 (2000)
89. Baring, M.G.: Synchrotron pair cascades in strong magnetic fields. A&A 225, 260–276 (1989)
90. Barkov, M.V., Khangulyan, D.V.: Direct wind accretion and jet launch in binary systems.
MNRAS 421, 1351–1359 (2012). DOI LSI + 61303 or LS 5039
91. Bednarek, W.: Gamma-rays from synchrotron pair cascades in blazars? MNRAS 285, 69–81
(1997)
92. Bednarek, W.: High energy processes in pulsar wind nebulae. Ap&SS 309, 179–187 (2007)
93. Bednarek, W., Idec, W.: On the variability of the GeV and multi-TeV gamma-ray emission
from the Crab nebula. MNRAS 414, 2229–2234 (2011)
94. Berezhko, E.G., Ksenofontov, L.T., Völk, H.J.: Confirmation of strong magnetic field ampli-
fication and nuclear cosmic ray acceleration in SN 1006. A&A 412, L11–L14 (2003)
95. Berezhko, E.G., Völk, H.J.: Hadronic versus leptonic origin of the gamma-ray emission from
supernova remnant RX J1713.7-3946. A&A 492, 695–701 (2008)
96. Berezinskii, V.S., Bulanov, S.V., Dogiel, V.A., Ptuskin, V.S.: Astrophysics of cosmic rays.
Amsterdam: North-Holland (1990)
97. Beskin, V.S., Rafikov, R.R.: On the particle acceleration near the light surface of radio pulsars.
MNRAS 313, 433–444 (2000)
98. Blumenthal, G.R., Gould, R.J.: Bremsstrahlung, Synchrotron Radiation, and Compton Scat-
tering of High-Energy Electrons Traversing Dilute Gases. Reviews of Modern Physics 42,
237–271 (1970)
99. Boettcher, M., Schlickeiser, R.: The pair production spectrum from photon–photon annihila-
tion. A&A 325, 866–870 (1997)
100. Bogovalov, S.V., Aharonian, F.A.: Very-high-energy gamma radiation associated with the
unshocked wind of the Crab pulsar. MNRAS 313, 504–514 (2000)
101. Bogovalov, S.V., Chechetkin, V.M., Koldoba, A.V., Ustyugova, G.V.: Interaction of pulsar
winds with interstellar medium: numerical simulation. MNRAS 358, 705–715 (2005)
102. Bogovalov, S.V., Khangoulian, D.V.: On the origin of the torus and jet-like structures in the
centre of the Crab Nebula. MNRAS 336, L53–L55 (2002)
103. Bogovalov, S.V., Khangulyan, D., Koldoba, A.V., Ustyugova, G.V., Aharonian, F.A.: Mod-
elling the interaction between relativistic and non-relativistic winds in the binary system
PSR B1259-63/SS2883—II. Impact of the magnetization and anisotropy of the pulsar wind.
MNRAS 419, 3426–3432 (2012)
104. Bogovalov, S.V., Khangulyan, D.V., Koldoba, A.V., Ustyugova, G.V., Aharonian, F.A.: Mod-
elling interaction of relativistic and non-relativistic winds in binary system PSR B1259-
63/SS2883—I. Hydrodynamical limit. MNRAS 387, 63–72 (2008)
105. Bongiorno, S.D., Falcone, A.D., Stroh, M., Holder, J., Skilton, J.L., Hinton, J.A., Gehrels, N.,
Grube, J.: A New TeV Binary: The Discovery of an Orbital Period in HESS J0632+057. ApJ
737, L11 (2011)
106. Bonometto, S., Rees, M.J.: On possible observable effects of electron pair-production in
QSOs. MNRAS 152, 21 (1971)
107. Bosch-Ramon, V.: Theoretical overview on high-energy emission in microquasars. Ap&SS
309, 321–331 (2007)
108. Böttcher, M., Dermer, C.D.: Photon–Photon Absorption of Very High Energy Gamma Rays
from Microquasars: Application to LS 5039. ApJ 634, L81–L84 (2005)
109. Bucciantini, N., Arons, J., Amato, E.: Modelling spectral evolution of pulsar wind nebulae
inside supernova remnants. MNRAS 410, 381–398 (2011)
110. Buehler, R., Scargle, J.D., Blandford, R.D., Baldini, L., Baring, M.G., Belfiore, A., Charles,
E., Chiang, J., D’Ammando, F., Dermer, C.D., Funk, S., Grove, J.E., Harding, A.K., Hays, E.,
Kerr, M., Massaro, F., Mazziotta, M.N., Romani, R.W., Saz Parkinson, P.M., Tennant, A.F.,
Weisskopf, M.C.: Gamma-Ray Activity in the Crab Nebula: The Exceptional Flare of 2011
April. ApJ 749, 26 (2012)
111. Bulik, T., Rudak, B., Dyks, J.: Spectral features in gamma-rays expected from millisecond
pulsars. MNRAS 317, 97–104 (2000)
112. Bykov, A.M., Pavlov, G.G., Artemyev, A.V., Uvarov, Y.A.: Twinkling pulsar wind nebulae
in the synchrotron cut-off regime and the γ-ray flares in the Crab Nebula. MNRAS 421,
L67–L71 (2012)
113. Casanova, S., Aharonian, F.A., Fukui, Y., Gabici, S., Jones, D.I., Kawamura, A., Onishi, T.,
Rowell, G., Sano, H., Torii, K., Yamamoto, H.: Molecular Clouds as Cosmic-Ray Barometers.
PASJ 62, 769– (2010)
114. Casanova, S., Jones, D.I., Aharonian, F.A., Fukui, Y., Gabici, S., Kawamura, A., Onishi, T.,
Rowell, G., Sano, H., Torii, K., Yamamoto, H.: Modeling the Gamma-Ray Emission Produced
by Runaway Cosmic Rays in the Environment of RX J1713.7-3946. PASJ 62, 1127–1134
(2010)
115. Cheng, K.S., Chernyshov, D.O., Dogiel, V.A., Ko, C.M., Ip, W.H.: Origin of the Fermi Bubble.
ApJ 731, L17 (2011)
116 F. Aharonian
116. Cheng, K.S., Chernyshov, D.O., Dogiel, V.A., Ko, C.M., Ip, W.H., Wang, Y.: The Fermi
Bubble as a Source of Cosmic Rays in the Energy Range ≥ 1015 eV. ApJ 746, 116 (2012)
117. Chernyakova, M., Malyshev, D., Aharonian, F.A., Crocker, R.M., Jones, D.I.: The High-
energy, Arcminute-scale Galactic Center Gamma-ray Source. ApJ 726, 60 (2011)
118. Coppi, P.S., Blandford, R.D.: Reaction rates and energy distributions for elementary processes
in relativistic pair plasmas. MNRAS 245, 453–507 (1990)
119. Crocker, R.M., Aharonian, F.: Fermi Bubbles: Giant, Multibillion-Year-Old Reservoirs of
Galactic Center Cosmic Rays. Physical Review Letters 106(10), 101102 (2011)
120. Crusius, A., Schlickeiser, R.: Synchrotron radiation in random magnetic fields. A&A 164,
L16–L18 (1986)
121. de Jager, O.C., Harding, A.K.: The expected high-energy to ultra-high-energy gamma-ray
spectrum of the Crab Nebula. ApJ 396, 161–172 (1992)
122. Del Zanna, L., Amato, E., Bucciantini, N.: Axially symmetric relativistic MHD simulations
of Pulsar Wind Nebulae in Supernova Remnants. On the origin of torus and jet-like features.
A&A 421, 1063–1073 (2004)
123. Derishev, E.V., Aharonian, F.A., Kocharovsky, V.V., Kocharovsky, V.V.: Particle acceleration
through multiple conversions from a charged into a neutral state and back. Phys. Rev. D 68(4),
043003 (2003)
124. Drury, L.O., Aharonian, F.A., Voelk, H.J.: The gamma-ray visibility of supernova remnants.
A test of cosmic ray origin. A&A 287, 959–971 (1994)
125. Dubus, G.: Gamma-ray absorption in massive X-ray binaries. A&A 451, 9–18 (2006)
126. Dubus, G.: Gamma-ray binaries: pulsars in disguise? A&A 456, 801–817 (2006)
127. Dubus, G., Cerutti, B., Henri, G.: The modulation of the gamma-ray emission from the binary
LS 5039. A&A 477, 691–700 (2008)
128. Dubus, G., Cerutti, B., Henri, G.: Relativistic Doppler-boosted emission in gamma-ray bina-
ries. A&A 516, A18 (2010)
129. Eichler, D., Wiita, P.J.: Neutron beams in active galactic nuclei. Nature 274, 38 (1978)
130. Ellison, D.C., Bykov, A.M.: Gamma-ray Emission of Accelerated Particles Escaping a Super-
nova Remnant in a Molecular Cloud. ApJ 731, 87 (2011)
131. Ellison, D.C., Patnaude, D.J., Slane, P., Raymond, J.: Efficient Cosmic Ray Acceleration,
Hydrodynamics, and Self-Consistent Thermal X-Ray Emission Applied to Supernova Rem-
nant RX J1713.7-3946. ApJ 712, 287–293 (2010)
132. Enomoto, R., et al. (CANGAROO collaboration): The acceleration of cosmic-ray protons in
the supernova remnant RX J1713.7-3946. Nature 416, 823–826 (2002)
133. Finkbeiner, D.P.: Microwave Interstellar Medium Emission Observed by the Wilkinson
Microwave Anisotropy Probe. ApJ 614, 186–193 (2004)
134. Freire, P.C.C., et al. (Fermi collaboration): Fermi Detection of a Luminous γ-Ray Pulsar in a
Globular Cluster. Science 334, 1107– (2011)
135. Fritz, K.D.: Synchrotron emission spectra from shockwaves in active galactic nuclei—an
energy and space dependent diffusion coefficient. A&A 214, 14–28 (1989)
136. Fukui, Y., Sano, H., Sato, J., et al.: A Detailed Study of the Molecular and Atomic Gas
Toward the gamma-ray SNR RX J1713.7-3946: Spatial TeV gamma-ray and ISM Gas Cor-
respondence. ArXiv e-prints (2011)
137. Gabici, S., Aharonian, F.A., Casanova, S.: Broad-band non-thermal emission from molecular
clouds illuminated by cosmic rays from nearby supernova remnants. MNRAS 396, 1629–1639
(2009)
138. Gaensler, B.M., Slane, P.O.: The Evolution and Structure of Pulsar Wind Nebulae. ARA&A
44, 17–47 (2006)
139. Gaisser, T.K.: Cosmic rays and particle, physics (1990)
140. Ginzburg, V.L., Syrovatskii, S.I.: The Origin of Cosmic Rays. New York: Macmillan (1964)
141. Giordano, F., Naumann-Godo, M., Ballet, J., Bechtol, K., Funk, S., Lande, J., Mazziotta,
M.N., Rainò, S., Tanaka, T., Tibolla, O., Uchiyama, Y.: Fermi-LAT Detection of the Young
SuperNova Remnant Tycho. ArXiv e-prints (2011)
142. Giuliani, A., et al. (AGILE collaboration): AGILE detection of GeV γ-ray emission from the
SNR W28. A&A 516, L11–L14 (2010)
143. Giuliani, A., et al. (AGILE collaboration): Neutral Pion Emission from Accelerated Protons
in the Supernova Remnant W44. ApJ 742, L30–L34 (2011)
144. Gold, T.: Rotating Neutron Stars as the Origin of the Pulsating Radio Sources. Nature 218,
731–732 (1968)
145. Gould, R.J., Schréder, G.P.: Opacity of the Universe to High-Energy Photons. Physical Review
155, 1408–1411 (1967)
146. Grondin, M.H., Funk, S., Lemoine-Goumard, M., Van Etten, A., Hinton, J.A., Camilo, F.,
Cognard, I., Espinoza, C.M., Freire, P.C.C., Grove, J.E., Guillemot, L., Johnston, S., Kramer,
M., Lande, J., Michelson, P., Possenti, A., Romani, R.W., Skilton, J.L., Theureau, G., Wel-
tevrede, P.: Detection of the Pulsar Wind Nebula HESS J1825-137 with the Fermi Large Area
Telescope. ApJ 738, 42 (2011)
147. Guilbert, P.W., Fabian, A.C., Rees, M.J.: Spectral and variability constraints on compact
sources. MNRAS 205, 593–603 (1983)
148. Harding, A.K., Muslimov, A.G.: Pulsar Pair Cascades in a Distorted Magnetic Dipole Field.
ApJ 726, L10 (2011)
149. Heitler, W.: Quantum theory of radiation (1954)
150. Hewish, A., Bell, S.J., Pilkington, J.D.H., Scott, P.F., Collins, R.A.: Observation of a Rapidly
Pulsating Radio Source. Nature 217, 709–713 (1968)
151. Higdon, J.C., Lingenfelter, R.E.: OB Associations, Supernova-generated Superbubbles, and
the Source of Cosmic Rays. ApJ 628, 738–749 (2005)
152. Hillas, A.M.: TOPICAL REVIEW: Can diffusive shock acceleration in supernova remnants
account for high-energy galactic cosmic rays? Journal of Physics G Nuclear Physics 31,
95–131 (2005)
153. Hinton, J.A., Aharonian, F.A.: Inverse Compton Scenarios for the TeV Gamma-Ray Emission
of the Galactic Center. ApJ 657, 302–307 (2007)
154. Hinton, J.A., Hofmann, W.: Teraelectronvolt Astronomy. ARA&A 47, 523–565 (2009)
155. Hinton, J.A., Skilton, J.L., Funk, S., Brucker, J., Aharonian, F.A., Dubus, G., Fiasson, A.,
Gallant, Y., Hofmann, W., Marcowith, A., Reimer, O.: HESS J0632+057: A New Gamma-
Ray Binary? ApJ 690, L101–L104 (2009)
156. Hofmann, W.: Performance Limits for Cherenkov Instruments. ArXiv Astrophysics e-prints
astro-ph/0603076 (2006)
157. Holder, J.: TeV Gamma-ray Astronomy: A Summary. ArXiv e-prints (2012)
158. Horns, D., Aharonian, F., Santangelo, A., Hoffmann, A.I.D., Masterson, C.: Nucleonic
gamma-ray production in Vela X. A&A 451, L51–L54 (2006)
159. Inoue, T., Yamazaki, R., Inutsuka, S.i., Fukui, Y.: Toward Understanding the Cosmic-Ray
Acceleration at Young Supernova Remnants Interacting with Interstellar Clouds: Possible
Applications to RX J1713.7-3946. ApJ 744, 71 (2012)
160. Iyudin, A.F., Diehl, R., Bloemen, H., Hermsen, W., Lichti, G.G., Morris, D., Ryan, J., Schoen-
felder, V., Steinle, H., Varendorff, M., de Vries, C., Winkler, C.: COMPTEL observations of
Ti-44 gamma-ray line emission from CAS A. A&A 284, L1–L4 (1994)
161. Jones, F.C.: Calculated Spectrum of Inverse-Compton-Scattered Photons. Physical Review
167, 1159–1169 (1968)
162. Kamae, T., Karlsson, N., Mizuno, T., Abe, T., Koi, T.: Parameterization of γ, e+/− , and
Neutrino Spectra Produced by p-p Interaction in Astronomical Environments. ApJ 647, 692–
708 (2006)
163. Katz, B., Waxman, E.: In which shell-type SNRs should we look for gamma-rays and neutrinos
from P-P collisions? JCAP 1, 18 (2008)
164. Kelner, S.R., Aharonian, F.A.: Energy spectra of gamma rays, electrons, and neutrinos pro-
duced at interactions of relativistic protons with low energy radiation. Phys. Rev. D 78(3),
034013 (2008)
165. Kelner, S.R., Aharonian, F.A., Bugayov, V.V.: Energy spectra of gamma rays, electrons, and
neutrinos produced at proton–proton interactions in the very high energy regime. Phys. Rev.
D 74(3), 034,018 (2006)
118 F. Aharonian
166. Kennel, C.F., Coroniti, F.V.: Magnetohydrodynamic model of Crab nebula radiation. ApJ 283,
710–730 (1984)
167. Kerschhaggl, M.: TeV flux modulation in PSR B1259-63/LS2883. A&A 525, A80 (2011)
168. Khangulyan, D., Aharonian, F., Bosch-Ramon, V.: On the formation of TeV radiation in LS
5039. MNRAS 383, 467–478 (2008)
169. Khangulyan, D., Aharonian, F.A., Bogovalov, S.V., Ribo, M.: Post-Periastron Gamma Ray
Flare from PSR B1259-63/LS 2883 as a Result of Comptonization of the Cold Pulsar Wind.
ArXiv e-prints (2011)
170. Khangulyan, D., Hnatic, S., Aharonian, F., Bogovalov, S.: TeV light curve of PSR B1259-
63/SS2883. MNRAS 380, 320–330 (2007)
171. Khangulyan, D.V., Aharonian, F.A., Bogovalov, S.V., Koldoba, A.V., Ustyugova, G.V.: Hydro-
dynamics of Interaction of Pulsar and Stellar Winds and its Impact on the High Energy Radi-
ation of Binary Pulsar Systems. International Journal of Modern Physics D 17, 1909–1915
(2008)
172. Kirk, J.G.: Particle Acceleration in Relativistic Current Sheets. Physical Review Letters
92(18), 181101 (2004)
173. Kirk, J.G., Ball, L., Skjaeraasen, O.: Inverse Compton emission of TeV gamma rays from
PSR B1259-63. Astroparticle Physics 10, 31–45 (1999)
174. Kirk, J.G., Lyubarsky, Y., Petri, J.: The Theory of Pulsar Winds and Nebulae. In: W. Becker
(ed.) Astrophysics and Space Science Library, Astrophysics and Space Science Library, vol.
357, p. 421 (2009)
175. Komissarov, S.S., Lyubarsky, Y.E.: Synchrotron nebulae created by anisotropic magnetized
pulsar winds. MNRAS 349, 779–792 (2004)
176. Komissarov, S.S., Lyutikov, M.: On the origin of variable gamma-ray emission from the Crab
nebula. MNRAS 414, 2017–2028 (2011)
177. Kong, S.W., Cheng, K.S., Huang, Y.F.: Modelling the light curves of PSR B1259-63/LS
2883-II. The effects of anisotropic pulsar wind and Doppler-boosting. ArXiv e-prints (2012)
178. Kozlovsky, B., Murphy, R.J., Ramaty, R.: Nuclear Deexcitation Gamma-Ray Lines from
Accelerated Particle Interactions. ApJS 141, 523–541 (2002)
179. Kusunose, M., Takahara, F.: A Leptonic Model of Steady High-energy Gamma-Ray Emission
from Sgr A*. ApJ 748, 34 (2012)
180. Landau, L., Rumer, G.: The Cascade Theory of Electronic Showers. Royal Society of London
Proceedings Series A 166, 213–228 (1938)
181. Levinson, A.: Particle Acceleration and Curvature TeV Emission by Rotating, Supermassive
Black Holes. Physical Review Letters 85, 912–915 (2000)
182. Levinson, A., Blandford, R.: On the Jets Associated with Galactic Superluminal Sources. ApJ
456, L29 (1996)
183. Linden, T., Lovegrove, E., Profumo, S.: The Morphology of Hadronic Emission Models for
the Gamma-Ray Source at the Galactic Center. ArXiv e-prints (2012)
184. Liu, S., Melia, F., Petrosian, V., Fatuzzo, M.: Stochastic Acceleration in the Galactic Center
HESS Source. ApJ 647, 1099–1105 (2006)
185. Lobanov, A.P., Horns, D., Muxlow, T.W.B.: VLBI imaging of a flare in the Crab nebula: more
than just a spot. A&A 533, A10 (2011)
186. Lunardini, C., Razzaque, S.: High Energy Neutrinos from the Fermi Bubbles. ArXiv e-prints
(2011)
187. Lyutikov, M., Balsara, D., Matthews, C.: Crab GeV flares from the corrugated termination
shock. MNRAS p. 2769 (2012)
188. Maier, G., for the VERITAS Collaboration, Skilton, J., for the HESS Collaboration: VHE
Observations of the Binary Candidate HESS J0632 + 057
189. Malkov, M.A., Diamond, P.H., Sagdeev, R.Z.: On the Gamma-Ray Spectra Radiated by Pro-
tons Accelerated in Supernova Remnant Shocks near Molecular Clouds: The case of Super-
nova Remnant RX J1713.7-3946. ApJ 624, L37–L40 (2005)
190. Malkov, M.A., O’C Drury, L.: Nonlinear theory of diffusive acceleration of particles by shock
waves. Reports on Progress in Physics 64, 429–481 (2001)
191. Markoff, S., Falcke, H., Fender, R.: A jet model for the broadband spectrum of XTE J1118 +
480
192. Mertsch, P., Sarkar, S.: Fermi Gamma-Ray “Bubbles” from Stochastic Acceleration of Elec-
trons. Physical Review Letters 107(9), 091101 (2011)
193. Mirabel, I.F., Rodríguez, L.F.: A superluminal source in the Galaxy. Nature 371, 46–48 (1994)
194. Montmerle, T.: On gamma-ray sources, supernova remnants, OB associations, and the origin
of cosmic rays. ApJ 231, 95–110 (1979)
195. Morlino, G., Amato, E., Blasi, P.: Gamma-ray emission from SNR RX J1713.7-3946 and the
origin of galactic cosmic rays. MNRAS 392, 240–250 (2009)
196. Morlino, G., Caprioli, D.: Strong evidences of hadron acceleration in Tycho’s Supernova
Remnant. ArXiv e-prints (2011)
197. Mücke, A., Engel, R., Rachen, J.P., Protheroe, R.J., Stanev, T.: Monte Carlo simulations of
photohadronic processes in astrophysics. Computer Physics Communications 124, 290–314
(2000)
198. Mücke, A., Protheroe, R.J., Engel, R., Rachen, J.P., Stanev, T.: BL Lac objects in the syn-
chrotron proton blazar model. Astroparticle Physics 18, 593–613 (2003)
199. Murase, K., Beacom, J.F.: Very-high-energy gamma-ray signal from nuclear photodisinte-
gration as a probe of extragalactic sources of ultrahigh-energy nuclei. Phys. Rev. D 82(4),
043008 (2010)
200. Nolan, P.L., Abdo, A.A., Ackermann, M., Ajello, M., Allafort, A., Antolini, E., Atwood,
W.B., Axelsson, M., Baldini, L., Ballet, J., et al.: Fermi Large Area Telescope Second Source
Catalog. ApJS 199, 31 (2012)
201. Ochelkov, I.P., Usov, V.V.: Curvature radiation of relativistic particles in the magnetosphere
of pulsars. I—Theory. Ap&SS 69, 439–460 (1980)
202. Ostrowski, M., Bednarz, J.: Comment on the first-order Fermi acceleration at ultra-relativistic
shocks. A&A 394, 1141–1144 (2002)
203. Pacini, F.: Rotating Neutron Stars, Pulsars and Supernova Remnants. Nature 219, 145–146
(1968)
204. Paredes, J.M., Martí, J., Ribó, M., Massi, M.: Discovery of a High-Energy Gamma-Ray-
Emitting Persistent Microquasar. Science 288, 2340–2342 (2000)
205. Parizot, E., Marcowith, A., van der Swaluw, E., Bykov, A.M., Tatischeff, V.: Superbubbles
and energetic particles in the Galaxy. I. Collective effects of particle acceleration. A&A 424,
747–760 (2004)
206. Piddington, J.H.: The Crab Nebula and the Origin of Interstellar Magnetic Fields. Australian
Journal of Physics 10, 530 (1957)
207. Pittori, C., et al. (AGILE Collaboration): First AGILE catalog of high-confidence gamma-ray
sources. A&A 506, 1563–1574 (2009)
208. Plyasheshnikov, A.V., Aharonian, F.A.: Characteristics of air showers produced by extremely
high energy gamma-rays. Journal of Physics G Nuclear Physics 28, 267–288 (2002)
209. Plyasheshnikov, A.V., Aharonian, F.A., Völk, H.J.: On the potential of the imaging
atmospheric Cherenkov technique for study of cosmic gamma rays in the energy range 10–100
TeV. Journal of Physics G Nuclear Physics 26, 183–201 (2000)
210. Prantzos, N., Boehm, C., Bykov, A.M., Diehl, R., Ferrière, K., Guessoum, N., Jean, P.,
Knoedlseder, J., Marcowith, A., Moskalenko, I.V., Strong, A., Weidenspointner, G.: The
511 keV emission from positron annihilation in the Galaxy. Reviews of Modern Physics 83,
1001–1056 (2011)
211. Rea, N., Torres, D.F., Caliandro, G.A., Hadasch, D., van der Klis, M., Jonker, P.G., Méndez,
M., Sierpowska-Bartosik, A.: Deep Chandra observations of TeV binaries—II. LS 5039.
MNRAS 416, 1514–1521 (2011)
212. Rees, M.J., Gunn, J.E.: The origin of the magnetic field and relativistic particles in the Crab
Nebula. MNRAS 167, 1–12 (1974)
213. Renaud, M., Vink, J., Decourchelle, A., Lebrun, F., den Hartog, P.R., Terrier, R., Couvreur,
C., Knödlseder, J., Martin, P., Prantzos, N., Bykov, A.M., Bloemen, H.: The Signature of 44 Ti
in Cassiopeia A Revealed by IBIS/ISGRI on INTEGRAL. ApJ 647, L41–L44 (2006)
120 F. Aharonian
214. Romani, R.W.: Gamma-Ray Pulsars: Radiation Processes in the Outer Magnetosphere. ApJ
470, 469 (1996)
215. Romero, G.E., Kaufman Bernadó, M.M., Mirabel, I.F.: Recurrent microblazar activity in
Cygnus X-1? A&A 393, L61–L64 (2002)
216. Sabatini, S., et al. (AGILE collaboration): Episodic Transient Gamma-ray Emission from the
Microquasar Cygnus X-1. ApJ 712, L10–L15 (2010)
217. Sano, H., Sato, J., Horachi, H., et al.: Star-forming Dense Cloud Cores in the TeV Gamma-ray
SNR RX J1713.7-3946. ApJ 724, 59–68 (2010)
218. Snowden, S.L., Egger, R., Freyberg, M.J., McCammon, D., Plucinsky, P.P., Sanders, W.T.,
Schmitt, J.H.M.M., Truemper, J., Voges, W.: ROSAT Survey Diffuse X-Ray Background
Maps. II. ApJ 485, 125 (1997)
219. Stecker, F.W.: Photodisintegration of Ultrahigh-Energy Cosmic Rays by the Universal Radi-
ation Field. Physical Review 180, 1264–1266 (1969)
220. Stecker, F.W.: Cosmic gamma rays. NASA Special, Publication 249 (1971)
221. Striani, E., et al. (AGILE collaboration): The Crab Nebula Super-flare in 2011 April:
Extremely Fast Particle Acceleration and Gamma-Ray Emission. ApJ 741, L5 (2011)
222. Su, M., Slatyer, T.R., Finkbeiner, D.P.: Giant Gamma-ray Bubbles from Fermi-LAT: Active
Galactic Nucleus Activity or Bipolar Galactic Wind? ApJ 724, 1044–1082 (2010)
223. Svensson, R.: Steady mildly relativistic thermal plasmas—Processes and properties. MNRAS
209, 175–208 (1984)
224. Svensson, R.: Non-thermal pair production in compact X-ray sources—First-order Compton
cascades is soft radiation fields. MNRAS 227, 403–451 (1987)
225. Takahashi, T., Kishishita, T., Uchiyama, Y., Tanaka, T., Yamaoka, K., Khangulyan, D., Aha-
ronian, F.A., Bosch-Ramon, V., Hinton, J.A.: Study of the Spectral and Temporal Characteris-
tics of X-Ray Emission of the Gamma-Ray Binary LS 5039 with Suzaku. ApJ 697, 592–600
(2009)
226. Tam, P.H.T., Huang, R.H.H., Takata, J., Hui, C.Y., Kong, A.K.H., Cheng, K.S.: Discovery of
GeV γ-ray Emission from PSR B1259-63/LS 2883. ApJ 736, L10 (2011)
227. Tanaka, S.J., Takahara, F.: A Model of the Spectral Evolution of Pulsar Wind Nebulae. ApJ
715, 1248–1257 (2010)
228. Tanaka, T., Allafort, A., Ballet, J., Funk, S., Giordano, F., Hewitt, J., Lemoine-Goumard, M.,
Tajima, H., Tibolla, O., Uchiyama, Y.: Gamma-Ray Observations of the Supernova Remnant
RX J0852.0-4622 with the Fermi Large Area Telescope. ApJ 740, L51 (2011)
229. Tanaka, T., Uchiyama, Y., Aharonian, F.A., Takahashi, T., et al.: Study of Nonthermal Emission
from SNR RX J1713.7-3946 with Suzaku. ApJ 685, 988–1004 (2008)
230. Tanimori, T., et al. (CANGAROO collaboration): Discovery of TeV Gamma Rays from SN
1006: Further Evidence for the Supernova Remnant Origin of Cosmic Rays. ApJ 497, L25
(1998)
231. Tavani, M.: The variable Crab Nebula. ArXiv e-prints (2011)
232. Tavani, M., Arons, J.: Theory of High-Energy Emission from the Pulsar/Be Star System PSR
1259-63. I. Radiation Mechanisms and Interaction Geometry. ApJ 477, 439 (1997)
233. Tavani, M., et al. (AGILE collaboration): Extreme particle acceleration in the microquasar
CygnusX-3. Nature 462, 620–623 (2009)
234. Tavani, M., et al. (AGILE collaboration): Direct Evidence for Hadronic Cosmic-Ray Accel-
eration in the Supernova Remnant IC 443. ApJ 710, L151–L155 (2010)
235. Tavani, M., et al. (AGILE collaboration): Discovery of Powerful Gamma-Ray Flares from the
Crab Nebula. Science 331, 736– (2011)
236. Timokhin, A.N.: Time-dependent pair cascades in magnetospheres of neutron stars—I.
Dynamics of the polar cap cascade with no particle supply from the neutron star surface.
MNRAS 408, 2092–2114 (2010)
237. Torres, D.F., Marrero, A.Y.R., de Cea Del Pozo, E.: The GeV to TeV connection in the
environment of SNR IC 443. MNRAS 408, 1257–1266 (2010)
238. Uchiyama, H., Matsumoto, H., Tsuru, T.G., Koyama, K., Bamba, A.: Suzaku Observation of
HESS J1825-137: Discovery of Largely-Extended X-Rays from PSR J1826-1334. PASJ 61,
189 (2009)
239. Uchiyama, Y., Funk, S., Katagiri, H., Katsuta, J., Lemoine-Goumard, M., Tajima, H., Tanaka,
T., Torres, D.: Fermi-LAT Discovery of GeV Gamma-ray Emission from the Vicinity of SNR
W44. ArXiv e-prints (2012)
240. Uchiyama, Y., on behalf of the Fermi LAT collaboration: GeV Gamma Rays from Supernova
Remnants Interacting with Molecular Clouds. ArXiv e-prints (2011)
241. Uzdensky, D.A., Cerutti, B., Begelman, M.C.: Reconnection-powered Linear Accelerator and
Gamma-Ray Flares in the Crab Nebula. ApJ 737, L40 (2011)
242. Van Etten, A., Romani, R.W.: Multi-zone Modeling of the Pulsar Wind Nebula HESS J1825-
137. ApJ 742, 62 (2011)
243. Vissani, F., Aharonian, F., Sahakyan, N.: On the detectability of high-energy galactic neutrino
sources. Astroparticle Physics 34, 778–783 (2011)
244. Völk, H.J., Berezhko, E.G., Ksenofontov, L.T.: Internal dynamics and particle acceleration in
Tycho’s SNR. A&A 483, 529–535 (2008)
245. Volpi, D., Del Zanna, L., Amato, E., Bucciantini, N.: Non-thermal emission from relativistic
MHD simulations of pulsar wind nebulae: from synchrotron to inverse Compton. A&A 485,
337–349 (2008)
246. Wang, Q.D., Lu, F.J., Gotthelf, E.V.: G359.95-0.04: an energetic pulsar candidate near Sgr
A*. MNRAS 367, 937–944 (2006)
247. Weekes, T.C.: TeV Radiation from Binary Stars. In: J. Sahade, G.E. McCluskey, Y. Kondo
(eds.) The Realm of Interacting Binary Stars, Astrophysics and Space Science Library, vol.
177, p. 297 (1992)
248. Weekes, T.C.: Very high energy gamma-ray astronomy. The Institute of Physics Publishing
(2003)
249. Yuan, Q., Yin, P.F., Wu, X.F., Bi, X.J., Liu, S., Zhang, B.: A Statistical Model for the γ-ray
Variability of the Crab Nebula. ApJ 730, L15 (2011)
250. Zabalza, V., Bosch-Ramon, V., Paredes, J.M.: Thermal X-Ray Emission from the Shocked
Stellar Wind of Pulsar Gamma-Ray Binaries. ApJ 743, 7 (2011)
251. Zdziarski, A.A., Sikora, M., Dubus, G., Yuan, F., Cerutti, B., Ogorzałek, A.: The gamma-ray
emitting region of the jet in Cyg X-3. MNRAS 421, 2956–2968 (2012)
252. Zenitani, S., Hoshino, M.: The Generation of Nonthermal Particles in the Relativistic Magnetic
Reconnection of Pair Plasmas. ApJ 562, L63–L66 (2001)
253. Zirakashvili, V.N., Aharonian, F.: Analytical solutions for energy spectra of electrons accel-
erated by nonrelativistic shock-waves in shell type supernova remnants. A&A 465, 695–702
(2007)
254. Zirakashvili, V.N., Aharonian, F.A.: Nonthermal Radiation of Young Supernova Remnants:
The Case of RX J1713.7-3946. ApJ 708, 965–980 (2010)
255. Zirakashvili, V.N., Aharonian, F.A.: Radioactivity and electron acceleration in supernova
remnants. Phys. Rev. D 84(8), 083,010 (2011)
Multi-Messenger Astronomy
and Dark Matter
Lars Bergström
1 Preamble
Astrophysics, and more specifically astroparticle physics, has been going through
tremendous progress during the last two decades. Still, one of the main problems,
that of the nature of the dark matter, remains unsolved. With the help of accelerator
experiments (at CERN’s Large Hadron Collider (LHC) in particular, which started
operation in 2010 and which is currently gathering an impressive integrated lumi-
nosity) we could soon hope to get a first indication of the mass scale for the new
physics that is associated with dark matter. However, to actually prove that a particle
discovered at accelerators has the right properties to constitute the astrophysical dark
matter, complementary methods are needed. The fact that a candidate for dark matter
is electrically neutral (as not to emit nor absorb light—that is what we mean with
the term “dark”) can plausibly be determined at accelerators. However, the coupling
of the dark matter particles to other matter needs to be weak, and the lifetime of the
dark matter particle needs to be at least of the order of the age of the universe. This
cannot be tested at accelerators—the dark matter particles would leave the detector
in some 100 ns. There could be very useful information still gathered at the LHC,
as possibly decays of more massive states in the “dark sector” would be observable,
and the missing energy could be estimated.
Fortunately, through observations of various types of messengers—radio waves,
microwaves, IR, optical and UV radiation, X-rays, γ-rays and neutrinos, there is
great hope that we could get an independent indication of the mass scale of dark
matter. This variety of possible methods of indirect detection methods is a part of
multi-messenger astronomy, and it is the second way by which we approach the
dark matter problem. In particular, for models where the dark matter particles are
L. Bergström
Department of Physics, The Oskar Klein Centre, AlbaNova, Stockholm University,
SE-106 91 Stockholm, Sweden

126 L. Bergström
involved in breaking the electroweak symmetry of the Standard Model, so-called

WIMP models (for weakly interacting massive particles), prospects of detection in
the near future look promising. We will look in some detail on the properties of
WIMP candidates, where the fact that they are massive means that they move non-
relativistically in galactic halos, and form so-called cold dark matter (CDM). One
thought earlier that neutrinos could be the dark matter, but they would constitute hot
dark matter (HDM), which is not favoured by observations. Due to free-streaming
motion, they would only form very large structures first, which then fragment into
smaller scales, like galaxies. This scenario does not agree with observations, as it
gives too little power on small scales. Of course, one may also consider an in-between
scenario, warm dark matter, usually consisting of having a sterile neutrino (i.e., with
no direct Standard Model couplings) in the keV mass region. These may perhaps
have some virtue of explaining possible anomalies in dark matter distribution on the
very smallest scales, but reliable methods are so far lacking to probe the dark matter
distribution, and its couplings to baryons, on these scales.
As a third approach, ingenious experiments for direct detection employing solid
state devices, liquid noble gases, etc. can be used to tell us about other important
properties of dark matter, like the spin-dependent or spin-independent cross section
of dark matter particle scattering on nucleons. Once signals start to be found (and
there are some, however not undisputed ones, already), an exciting puzzle will present
itself, putting all these pieces of information together. For indirect detection, astro-
physical backgrounds that could mask or mimic dark matter signatures will often be
a great challenge to overcome. It should therefore be useful to the reader to study also
the accompanying articles by Felix Aharonian and Chuck Dermer in this volume—
not the least to understand the very interesting aspects of those processes in their
own right.
In this set of lectures, I will treat all of the dark matter-related aspects in some
detail, and also cover some other current problems of astroparticle physics and cos-
mology. The sections in these lectures correspond roughly to the lectures at the
Saas-Fee Course in Les Diablerets in March, 2010, i.e.,
• The particle universe: introduction, cosmological parameters.
• Basic cross sections for neutrinos and γ-rays; IceCube.
• Density of relic particles from the early universe.
• Dark matter: direct and indirect detection methods; the galactic centre & other
promising DM sources.
• Neutrinos and antimatter from dark matter, Sommerfeld enhancement.
• Supersymmetric dark matter, DarkSUSY.
• Particular dark matter candidates (WIMPS, Kaluza-Klein particles, sterile neutri-
nos . . .).
• Diffuse extragalactic γ-rays, Primordial black holes, Hawking radiation.
• Gravitational waves.
The order has been slightly changed (cf. the Table of Contents), and in many cases
I have updated the material since the time of the lectures, referring to important
developments (actually, quite a number of them) that have appeared after the time
Multi-Messenger Astronomy and Dark Matter 127
of the Course. This is of course mandatory in a field that evolves so rapidly. For the
more basic parts of this review, I have relied heavily on the textbook by Ariel Goobar
and myself [1]. Also material from various reviews I have written over the last few
years [2–5] has come to use, but also a lot of new material. With these lecture notes,
I hope to convey at least some of the excitement I feel for this topic, which relates to
some of the outstanding questions still with us in particle physics and cosmology.
2 The Particle Universe: Introduction
2.1 Introduction
One of the most impressive achievements of science is the development of a quite

detailed understanding of the physical properties of the universe, even at its earliest
stages. Thanks to the fruitful interplay between theoretical analysis, astronomical
observations and laboratory experiments we have today very successful ‘Standard
Models’ of both particle physics and cosmology. The Standard Model of particle
physics involves matter particles: quarks which always form bound states such as
neutrons and protons, and leptons like the electron which is charged and therefore can
make up neutral matter when bound to nuclei formed by neutrons and protons. There
are also neutral leptons, neutrinos, which do not form bound states but which play a
very important role in cosmology and particle astrophysics as we will see throughout
these lecture notes. The other important ingredients in the Standard Model of particle
physics are the particles which mediate the fundamental forces: the photon, the gluons
and the W and Z bosons.
The Standard Model of cosmology is the hot model, which states that the universe
is not infinitely old but rather came into existence some 13.7 billion years ago.
There may have been a short period with extremely rapid expansion, inflation, which
diluted all matter, radiation and other structures (like magnetic monopoles) that might
have existed before inflation. When inflation ended, there was a rapid heating (or,
thus, rather re-heating) which meant a re-start of expansion, now governed by the
relativistic degrees of freedom of our universe, i.e., radiation. The expansion started
out in a state which after this small fraction of a second was enormously compressed
and very hot (the relation between the density and the temperature can be determined
by near-equilibrium thermodynamics at this epoch, when the expansion was “slow”
and adiabatic). No bound states could exist because of the intense heat which caused
immediate dissociation even of protons and neutrons into quarks if they were formed
in the quark-gluon plasma. Subsequently, the universe expanded and cooled, making
possible the formation of a sequence of ever more complex objects: protons and
neutrons, nuclei, atoms, molecules, clouds, stars, planets,. . . . As we will see, the
observational support for the big bang model is overwhelming, but it contains new
elements, of dark matter and dark energy, that were not entirely expected. The key
observations are:
128 L. Bergström
• The present expansion of the universe.

• The existence of the cosmic microwave background radiation (CMBR), i.e. the
relic radiation from the hot stage of the early universe, and measurements of the
temperature variations therein.
• The presence of related structure in the late-time distribution of galaxies, so-called
“baryon acoustic oscillations” (BAO).
• Supernova cosmology that measures the expansion history, with the surprising
result that the cosmic expansion is accelerating.
• The successful calculations of the relative abundance of light elements in the
universe, which accurately agrees with what would be synthesized in an initially
hot, expanding universe.
• The concept of cosmological inflation, which successfully predicted the geometric
flatness of the universe, (thus that the average density is near the critical density,
i.e., Ωtot = 1 to an excellent approximation) and gave an explanation of the form
of the nearly scale invariant, gaussian temperature fluctuations.
• The discovery of dark matter, pioneered by Zwicky in the 1930s, has stood the test
of time and is now an established piece of the cosmological standard model. Dark
energy, in its simplest form just a constant vacuum energy, is the other part which
explains Ωtot = 1 and the accelerated expansion of the universe.
Several of these observations have been awarded the Nobel Prize, the latest thus
being the prize for the discovery of the accelerated expansion of the universe through
supernova observations (S. Perlmutter, B. Schmidt and A. Riess, 2011).
As another piece of evidence in favour of the big bang scenario, can be taken the
fact that the oldest objects found in the universe—globular clusters of stars and some
radioactive isotopes—do not seem to exceed an age around 13 billion years. This
gives strong evidence for a universe with a finite age, such as the big bang model
predicts.
In some areas, there are new pieces of information to await. For instance, one of
the main objectives of the Planck satellite, which will present cosmological data in
early 2013, is to search for non-gaussian features, which could tell us more about
the mechanism of inflation.
Although there are still many puzzles and interesting details to fill in, both in the
Standard Model of particle physics and in the big bang model, they do remarkably
well in describing a majority of all phenomena we can observe in nature. Combined,
they allow us to follow the history of our universe back to only about 10−10 s after the
big bang using established physical laws that have been checked in the laboratory.
Extending the models, there are inflationary scenarios that describe the evolution
back to 10−43 s after the big bang!
Behind this remarkable success are the theories of General Relativity and Quan-
tum Field Theory, which we use in these lecture notes. However, many fundamental
aspects of the laws of nature remain uncertain and are the subject of present-day
research. The key problem is, as it has been for many decades, to find a valid descrip-
tion of quantized gravity, something which is needed to push our limit of knowledge
even closer to (and maybe eventually explaining?) the big bang itself.
In this section we will review some of the most striking observational facts about
our universe.
2.2 Basic Assumptions
A basic concept in modern cosmology is that of the “Copernican principle”, i.e. the
supposition that the universe on the average is homogeneous and isotropic. Although
this is definitely not true on galactic scales and smaller, the distribution of matter
seems to become more and more smooth on large scales, and on the largest scales we
can observe, probed by the CMBR, isotropy and homogeneity seems to be fulfilled.
The inhomogeneities seem to be 10−5 or smaller, apart from a dipole component
in the CMBR, which however has a natural interpretation in terms of motion of our
galaxy towards other massive galaxies. Given isotropy and homogeneity, the most
general line element is the one found by Friedmann, Lemaître, Robertson and Walker
(FLRW),
dr 2
ds = dt − a (t)
2 2 2
+ r dθ + r sin θdφ .
2 2 2 2 2
(1)
1 − kr 2
Measurements on the cosmic microwave background gives (and inflationary theories

predicted) k = 0, i.e., a geometrically flat universe on large scales, to good accu-
racy. (There have been suggestions that some of the features of the homogeneous
and isotropic model can be alternatively explained if we live in an inhomogeneous
universe with large “bubbles” of atypical density. Although a logical possibility,
combined constraints from galaxy surveys, supernova data, and the CMBR mean
that we would have to live at a fine-tuned location near the centre of such a bubble
[6]. We will thus not consider these scenarios.)
The scale factor a(t) follows equations first derived by Friedmann from Einstein’s
equations in general relativity:
2
ȧ 8πG N
H (t) ≡
2
= ρtot . (2)
a 3
Here G N is Newton’s gravitational constant, and ρtot is the total average energy
density of the universe. The time-dependent Hubble parameter H (t), has a value
today which is known as the Hubble constant,
H (t0 ) ≡ H0 = h · 100 km s−1 Mpc−1 . (3)
This defines the dimensionless quantity h ∼ 0.7, which has to be given by measure-
ment.
The equation which determines the acceleration of the scale factor is also derived
from Einstein’s equations:
130 L. Bergström
2
2ä ȧ
+ = −8πG N p, (4)
a a
with p being the total pressure.
2.3 Energy and Pressure
In general, there are several components contributing to the energy density, at least
matter, radiation and dark energy, where the simplest possibility is a constant vacuum
energy—the modern version of Einstein’s cosmological constant:
ρtot = ρm + ρrad + ρΛ . (5)
For an isotropic and homogeneous model, the relevant elements of the energy-
momentum tensor are
T i j = pδi j (6)
T i0
=0 (7)
T 00
= ρtot (8)
and there is for each component contributing to p and ρtot a relation
pi = wi · ρi (9)
called the equation of state, which enables one to make predictions for the time
evolution of the expansion of the universe and for the relative weights of the different
energy components. For non-relativistic matter, the pressure is proportional to (v/c)2 ,
and therefore negligible, p = 0, i.e. w M = 0. For radiation on the other hand,
p = ρ/3, so w R = 1/3. What is the equation of state for vacuum energy? This is
easy to motivate from symmetry reasons (as was done already by Lemaître in the
1930s). The energy momentum tensor has to be proportional to the only available
rank-2 tensor in empty space-time, namely the Minkowski metric tensor in the cosmic
rest frame:
⎛ ⎞ ⎛ ⎞
1 0 0 0 ρΛ 0 0 0
μν ⎜ 0 −1 0 0 ⎟ ⎜ 0 −ρΛ 0 0 ⎟
TΛ = ρΛ ⎜ ⎟ ⎜
⎝ 0 0 −1 0 ⎠ = ⎝ 0 0 −ρΛ 0 ⎠ .
⎟ (10)
0 0 0 −1 0 0 0 −ρΛ
This form is thus dictated by the requirement of Lorentz invariance. Comparing with
the general form of the energy-momentum tensor which has ρ and p in the diagonal,
we thus see that the equation of state is p = −ρ, i.e., wΛ = −1. The vacuum energy
thus acts as a fluid with negative pressure.
2.4 Contributions to Vacuum Energy
How do we describe the contents of the universe, including vacuum energy? Based
on its success in particle physics, we try to do it by using quantum field theory, with
its particles and fields. A field is a dynamical quantity which is defined in all points
of space and at all times. Particles are the lowest excitations of the fields. A particle
is characterized by the mass m, spin s, charge Q, and maybe other internal quantum
numbers.
The lowest excitations of the field, carrying energy E and three-momentum p can
be quantized as harmonic oscillators fulfilling, in the cosmic rest frame (the reference
frame where the CMBR looks maximally isotropic), the mass shell condition
pμ p μ = m 2 , (11)
where the four momentum

p μ = (E, p) (12)
and
pμ = (E, −p). (13)
For each possible momentum mode, there will, as for the simple quantum mechanical
harmonic oscillator, be a zero-point energy

1 1 1
E i = ω( pi ) n + = pi2 + m 2 n + = pi2 + m 2 . (14)
2 n=0 2 n=0 2
However, for a given field, these have to be summed for all modes, meaning that
there will be a huge zero-energy density
1 1
ρΛ = d 3 p p2 + m 2 . (15)
2 (2π)3
The highly divergent integral has to be cut-off at some large energy scale, and the
first guess is the Planck mass, m Pl ∼ 1019 GeV, thus
1 1 m Pl
ρΛ = d 3 p p 2 + m 2 ∼ m 4Pl . (16)
2 (2π)3
Unfortunately, this is orders of magnitude too large, and is the same disastrous result
one would get by using simple dimensional analysis. Namely, what is the natural
132 L. Bergström
scale of ρΛ ? We see here that it is governed by the cut-off mass scale when new
physics appears, and dimensional analysis gives that in units where c = 1 so that
length is proportional to an inverse mass, and thus energy per unit volume becomes
[ρΛ ] = [M 4 ]. The only mass scale in gravity is m Pl , thus
Λ ∼ m Pl .
ρth 4
(17)
Unlike other guesses in physics based on dimensional analysis, this is a terrible

prediction. The present-day vacuum energy density of the universe is given by mea-
surements of supernovae and the CMBR and is (using k = 0)
−122 4
Λ ∼ 10
ρobs m Pl m 4Pl ∼ ρth
Λ. (18)
To go back to our field theory result, the zero-point energy is really a conse-
quence of the quantum mechanical commutator between the field and its canonical
momentum. However, for fermions, anticommutators are used, meaning the sign
of the vacuum energy is changed. So, introducing the fermion number F = 1 for
fermions, F = 0 for bosons, one gets
1 1 m Pl
ρΛ = (−1) F d 3 p p2 + m 2 . (19)
2 (2π)3
Remarkably, if there are as many fermionic degrees of freedom as bosonic, and they
pairwise have the same mass, the vacuum energy would vanish. Examples of theo-
ries having this property are supersymmetric theories, with unbroken supersymmetry.
However, since we do not see 0.511 MeV scalar electrons (selectrons), supersym-
metry has to be broken. Therefore large effects of the zero-point energy remain, and
ρΛ ∼ m 4SU SY with m SU SY (1000 GeV, say) the scale of SUSY breaking. Better, but
still enormously much “too high”.
In summary, we have encountered one of the most severe problems of cosmol-
ogy and particle astrophysics: Why is the cosmological constant so small, but still
not zero? (By the way, nobody found a good reason that it should be exactly zero,
anyway. . .) Supersymmetry alleviates the problem somewhat, but since supersym-
metry is broken there remains large mass terms still giving a value some 50–60 orders
of magnitude different from the observed value.
In cosmology the cosmological constant has a dramatic effect. Since it is related
to the energy density of the vacuum, and the vacuum is growing in size due to the
expansion, it will eventually dominate completely. Matter is on the other hand more
and more diluted and becomes less and less important, and radiation is also diluted
plus red-shifted:
ρm ∼ (1 + z)3 , ρr ∼ (1 + z)4 , ρΛ ∼ (1 + z)0 . (20)
We see that in the early universe (large redshifts), vacuum energy was irrelevant.
Today matter and vacuum energy are almost equal (why now?). In the future, the
expansion rate will grow exponentially, as we will see in the section on inflation,
Sect. 3.2.
To explain the smallness of Λ some people resort to (almost) non-scientific rea-
soning: the anthropic principle, or the landscape of string theory vacua. There the
argument goes roughly like this: There exist an amazingly large number of different
vacua, i.e., ground states, of the theory, and maybe all of these are realized some-
where in nature. But of course, those with very large values of Λ would accelerate
so quickly that structure would not form in the universe and therefore no life could
appear. But since we exist, we have to do so in one of the very few universes where
life did evolve. Of course, this sounds more like post-dicting the properties of our
universe rather than predicting them, which perhaps just shows the desperation in
front of the problem of the size of the cosmological constant.
Let us have another look at Planck-mass phenomena. Consider the scattering of
a photon on an electron, Compton scattering (we will treat this in detail later, see
Fig. 3). The relation between the incident and outgoing wavelength as a function of
scattering angle is given by
2π 2π
λ − λ = (1 − cos θ) = (1 − cos θ) ≡ λc (1 − cos θ) . (21)
mec me
Here λc ≡ 2π/m e is called the Compton wavelength (or radius) of the parti-
cle (the electron in this case). Note that we use here and onwards units such that
c = = 1. This implies that time and length have the same dimension which is
inversely proportional to the dimension of mass.
We will see in Sect. 9 the expression for the Schwarzschild radius (the radius
which marks the limit of where light can leave the black hole)
2G N M
rS = = 2G N M (22)
c2
Thus, the Compton radius decreases with mass, but the Schwarzschild radius
increases with mass. When are the two equal, i.e., how big must the mass be for
the Compton radius to be smaller than the Schwarzschild radius? This is when quan-
tum gravity should be important. (All details of particle properties are smeared out by
quantum fluctuations on the order of the Compton wavelength or less, so for λc > rs
the black hole properties should be unnoticeable.) We see
λc π m 2Pl
= ∼ . (23)
rS G N M2 M2
Thus, when the mass of an elementary particle is larger than the Planck mass, its
Compton radius is smaller than its Schwarzschild radius, which implies that we need
quantum gravity! None exists yet, but perhaps string theory is the best bet for such a
fundamental theory at the Planck scale? For an electron, λc /r S ∼ 1045 , so quantum
134 L. Bergström
gravity effects are completely negligible at the particle level. The same is true for all
other Standard Model particles.
2.5 Summary of Observations
To end this section where the main theoretical lines for describing the universe have
been laid out, we summarize what we know about the cosmological parameters of
the universe from the impressive recent measurements. Analyses combining high-
redshift supernova luminosity distances, microwave background fluctuations (from
the satellite WMAP) and baryon acoustic oscillations (BAO) in the galaxy distribu-
tion give tight constraints [7] on the present mass density of matter in the universe.
This is usually expressed in the ratio
Ω M = ρ M /ρc , (24)
normalized to the critical density,
ρc = 3H02 /(8πG N ) = h 2 × 1.9 · 10−29 g cm−3 . (25)
The value obtained for the 7-year WMAP data [7] for cold dark matter for the
(unknown) particle X making up the dark matter is Ω X h 2 = 0.113 ± 0.004, which is
around 5 times higher than the value obtained for baryons, Ω B h 2 = 0.0226 ± 0.0005.
Here h = 0.704 ± 0.014 is the derived [7] present value of the Hubble constant in
units of 100 km s−1 Mpc−1 . In addition, the WMAP data is consistent within a
percent with a flat universe (Ωtot = 1) and a value for the dark energy component,
e.g. the cosmological constant Λ, of ΩΛ = 0.73 ± 0.02.
One of the main problems for cosmology and particle physics is to explain the
measured density of dark matter, and to give candidates for the identity of the dark
matter particles. The fact that dark matter is definitely needed on the largest scales
(probed by WMAP), on galaxy cluster scales (as pointed out by Zwicky already in
1933 [8], and verified by gravitational lensing and the temperature distribution of
X-ray emitting gas) all the way down to the smallest dwarf galaxies, means that
solutions based on changing the laws of gravity seem less natural. In particular, the
direct empirical proof of the existence of dark matter given by the “Bullet Cluster”
[9] is very difficult to circumvent, as the X-ray signal from the baryonic matter and
the gravitational lensing signal from dark matter are clearly separated.
Although the existence of a non-zero cosmological constant (or some similar
form of dark energy) in the present-day universe came as a big surprise to most
cosmologists and particle physicists, the most successful models of evolution in the
universe do make use of a similar effect in models of inflation, as we will see in
Sect. 3.2.
3 Relic Density of Particles
There are several important examples of freeze-out in the early universe, for instance
at the synthesis of light elements one second to a few minutes after the big bang,
and the microwave photons from the “surface of last scattering” several hundred
thousand years later. Before we calculate freeze-out, it is convenient to introduce
a formalism which considers freeze-out in general: that is, what happens when a
particle species goes out of equilibrium. A rigorous treatment has to be based on
the Boltzmann transport equation in an expanding background, but here we give a
simplified treatment (see, for example [1] for a more complete discussion).
There are several different contributions to Ω = ρ/ρc , like radiation Ω R , matter
Ω M and vacuum energy ΩΛ .
The equations of motion for the matter in the universe are given by the vanishing
of the covariant divergence of the energy-momentum tensor
αβ
T;β = 0 (26)
This gives, for the FLRW metric,
d d
(ρa 3 ) = − p a 3 (27)
dt dt
which shows that the change of energy in a comoving volume element is equal to
minus the pressure times the change in volume. This can be rewritten as
dp d
a3 = [a 3 (ρ + p)] (28)
dt dt
which can be interpreted as a conservation law for the entropy in a volume a 3 (T ).

For radiation, where p = ρ/3, (27) gives ρ ∼ a −4 . Note that all particles fulfilling
mc2 k B T have the equation of state of radiation.
The Friedmann equation is
8πG N ρ
H 2 (t) = (29)
3
where as a good approximation only the relativistic species contribute appreciably
to ρ. Note that the Hubble parameter H (t) has units of 1/(time). This means in our
units that it has dimensions of mass. The age of the universe at a given time t is simply
of the order of H −1 (t), at least when the scale factor increases as a power of t.
We now treat schematically the thermodynamics of the expanding universe. We
assume, which is true if reactions between different species of particles are rapid
enough, that we can use the thermodynamical quantities, temperature T , pressure p,
entropy density s, and other quantities, at each time t to describe the state of the
universe. The constituents have number density n and typical relative velocities v,
136 L. Bergström
and scattering or annihilation cross-section σ, meaning that the interaction rate per
particle Γ is given by
Γ = nσv. (30)
The condition that the interactions maintain equilibrium is that the interaction rate
is larger than the expansion rate of the universe:
Γ H (31)
Typically, the number density of particles decreases faster with temperature and
therefore with time than the Hubble parameter does. This means that at certain
epochs some of the particle species will leave thermodynamic equilibrium. Their
number density will be “frozen” at some particular value which then only changes
through the general dilution due to the expansion. This “freeze-out” of particles is an
important mechanism which explains the particle content of the universe we observe
today.
Using relativistic statistical mechanics in the cosmic rest frame, the distribution
function f i (p) for particle species of type i is
1
f i (p) = (E i −μi )
(32)
e T ±1
with E i = p2 + m i2 the energy, μi is the chemical potential and T the temperature

(we put k B = 1). The minus sign is for particles that obey Bose-Einstein statis-
tics (bosons) and the plus sign is for particles obeying the exclusion principle and
therefore Fermi-Dirac statistics (fermions). To a good approximation the chemical
potentials can be neglected in the very early universe.
We denote by gi the number of internal degrees of freedom of particle i. The
photon has two polarization states and therefore gγ = 2. The neutrinos only have
one polarization state, giving gν = 1, electrons and muons have ge,μ = 2 (and the
same numbers for the antiparticles).
With these definitions, the number density for species i is
gi
ni = f i (p)d 3 p, (33)
(2π)3
and its energy density is
gi
ρi = E i (p) f i (p)d 3 p. (34)
(2π)3
The expression for the pressure is
gi |p|2
pi = f i (p)d 3 p. (35)
(2π)3 3E i (p)
In the nonrelativistic limit T /m 1 we can solve the integrals analytically, and

the result both for Fermi-Dirac and Bose-Einstein particles is
3
mT 2 m
n N R = gi e− T , (36)
2π
ρN R = m · n N R , (37)
and
pN R = T · n N R ρN R (38)
For nonrelativistic matter, E = m + 3T /2.

In the ultrarelativistic approximation, T /m 1, the integrals can also be per-
formed with the results
⎧ 2
⎪ π
⎨ 30 gi T ,
4 Bose-Einstein
gi ∞ 3
E dE
ρR = = (39)
6π 2 0 e TE ± 1 ⎪ ⎩ 78 π30 gi T 4 , Fermi-Dirac,
2
⎧
ζ(3)
⎪
⎨ π2 gi T ,
3 Bose-Einstein
nR = (40)
⎪
⎩ 43 ζ(3) g T 3 , Fermi-Dirac,
π2 i
with ζ(x) is the Riemann zeta function, ζ(3) = 1.20206. . . The average energy ρ/n
for a relativistic particle is
E B E ∼ 2.7T (41)
and
E F D ∼ 3.15T (42)
For photons, with the mass m γ = 0, and gγ = 2, the expression for ργ (T ) ∼ T 4 is

the famous Stefan Boltzmann law for electromagnetic black-body radiation.
The total contribution to the energy and number density of all kinds of particles in
the early universe is to a good approximation (since the energy and number density
of a nonrelativistic species is exponentially suppressed),
π2
ρ R (T ) = geff (T )T 4 (43)
30
1 π2
p R (T ) = ρ R (T ) = geff (T )T 4 (44)
3 90
138 L. Bergström
where geff (T ) counts the total number of internal degrees of freedom (such as spin,
colour, etc.) of the particles whose mass fulfill m T , and which are in thermo-
dynamic equilibrium with the “primordial cosmic soup” of particles in the early
universe. The expression for geff (T ) has the factor 7/8 for fermions.
As an example, we calculate geff (T ) for a temperature of, say, 1 TeV when all
the particles of the Standard Model were relativistic and in thermal equilibrium. The
total number of internal degrees of freedom of the fermions is 90 and for the gauge
and Higgs bosons 28, so the total expression for geff is
7
geff (T ∼ 1 TeV) = 28 + · 90 = 106.75 (45)
8
If we insert the expression for the energy density into the Friedmann equation
(29) we get for the radiation-dominated epoch in the early universe
8πG N 8πG N π 2 geff T 4

H2 = ρR = geff T 4 = 2.76 2 (46)
3 3 30 m Pl
or
√ T2
H = 1.66 geff (47)
m Pl
This is a very important formula governing the physics of the early universe.
For radiation domination, it can be shown that
√
a(t) ∼ t (48)
deriving from the equation of state p = ρ/3. For matter domination, that is, for
p ∼ 0, one has
2
a(t) ∼ t 3 . (49)
So for radiation domination,
ȧ 1
H= = (50)
a 2t
and the time-temperature relation becomes
2
m Pl 1 MeV
t = 0.30 √ ∼ s (51)
geff T 2 T
We now have to determine which particles are in thermal equilibrium at a given

temperature, so that we can calculate geff (T ). The entropy S(V, T ) is introduced
through
1
d S(V, T ) = [d(ρ(T )V ) + p(T )d V ] (52)
T
this gives (see [1])

V
S(V, T ) = (ρ(T ) + p(T )) (53)
T
and from the conservation of the energy-momentum tensor follows

d a3
[ρ(T ) + p(T )] = 0. (54)
dt T
Identifying the volume V with a 3 (t) and comparing with (53) we find the law of
conservation of entropy in the volume a 3 (t). Sometimes it is more useful to work with
the entropy density s(T ) rather than the total entropy S(V, T ) within the volume V .
The definition is thus:
S(V, T ) ρ(T ) + p(T )
s(T ) ≡ = (55)
V T
In the early universe, both the energy density and the pressure were dominated
by relativistic particles with the equation of state p = ρ/3. Using (55) and the
relativistic expressions for the energy density and the pressure (Eqs. (43) and (44)),
gives density s
2π 2 s 3
s= g T (56)
45 eff
where geff
s is defined in a similar way as g .
eff
Since s and n γ both vary as T 3 there is a simple relationship between them. With
2ζ(3) 3
nγ = T (57)
π2
from Eq. (40), we find
π4
s= g s n γ ∼ 1.8geff
s
nγ (58)
45ζ(3) eff
Following [1] we now consider a case of great interest for the dark matter problem.
Suppose that there exists some kind of unknown particle χ, with antiparticle χ̄, that
can annihilate each other and be pair created through processes χ + χ̄ ↔ X + X̄ ,
where X stands for any type of particle to which the χs can annihilate.1 We further
assume that the X particles have zero chemical potential and that they are kept in
thermal equilibrium with the photons and the other light particles in the early universe
(the X particles can be quarks, leptons etc.)
1 The supersymmetric neutralino is actually its own antiparticle (just as the photon is its own
antiparticle). The formalism is very similar in this case. In particular, a neutralino can annihilate
with another neutralino giving other, non-supersymmetric particles in the final state.
140 L. Bergström
How will the number density n χ evolve with time (and therefore with temper-
ature)? It is clear that in exact thermal equilibrium the number of χ particles in a
EQ
comoving volume Nχ = a 3 n χ will be given by the equilibrium value n χ (T ) (see
(40)). (In exact thermal equilibrium the rate for the process χ + χ̄ ↔ X + X̄ is
the same in both directions.) If the actual number density n χ (T ) is larger than the
equilibrium density the reaction will go faster to the right: that is, the χ particles will
annihilate faster than they are created. The depletion rate of χ should be proportional
to σχχ̄→X X̄ |v|n 2χ (quadratic in the density, since it should be proportional to the
product of n χ and n χ̄ , and these are equal). However, χ particles are also created by
EQ
the inverse process, with a rate proportional to (n χ )2 . We have thus ‘derived’ the
basic equation that governs the departure from equilibrium for the species χ:
dn χ
+ 3H n χ = −σχχ̄→X X̄ |v|[n 2χ − (n χE Q )2 ]. (59)
dt
The left-hand side comes from a13 dt d

[n χ a 3 ]; the term proportional to 3H just
expresses the dilution that automatically comes from the Hubble expansion. The
expression σχχ̄→X X̄ |v| stands for the thermally averaged cross section times veloc-
ity. This averaging is necessary, since the annihilating particles have random thermal
velocities and directions. Summing over all possible annihilation channels gives
dn χ
+ 3H n χ = −σ A |v|[n 2χ − (n χE Q )2 ], (60)
dt
where σ A is the total annihilation cross section.
Using the time-temperature relation of Eq. (51) (for radiation dominance)
m Pl
t = 0.30 √ (61)
T geff
2
this can be converted to an evolution equation for n χ as a function of temperature.

Introducing the dimensionless variable x ≡ m χ /T , and normalizing n χ to the entropy
density:
nχ
Yχ = (62)
s
gives after some intermediate steps

dYχ m χ m Pl ceff π
=− σ A |v|(Yχ2 − (YχE Q )2 ) (63)
dx x2 45
where
gs
ceff = √ eff (64)
geff
or, after some rearrangement,

2
x dYχ ΓA Yχ
EQ
=− EQ
−1 , (65)
Yχ dx H Yχ
EQ
where Γ A = n χ σ A |v|. This equation can be solved numerically with the boundary
condition that for small x, Yχ ∼ YχE Q (since at high temperature the χ particles were
in thermal equilibrium with the other particles). We see from (65) that the evolution
is governed by the factor Γ A /H , the interaction rate divided by the Hubble expansion
rate.
The solutions to these equations have to be obtained numerically in the general
case to find the temperature T f and therefore the value of x f of freeze-out and the
asymptotic value Yχ (∞) of the relic abundance of the species χ. There are, however,
EQ
some simple limiting cases. If the species χ is relativistic at freeze-out, then Yχ is
not changing with time during the period of freeze-out, and the resulting Yχ (∞) is
just the equilibrium value at freeze-out,
45ζ(3) geff
Yχ (∞) = YχE Q (x f ) = (66)
2π 4 geff
s (x )
f
where geff = g for bosons and 3g/4 for fermions. A particle that was relativistic at
freeze-out is called a hot relic. A typical example is the neutrino. The present mass
density of a hot relic with mass m is

−2 geff mχ
Ωχ h = 7.8 · 10
2
(67)
geff
s (x ) 1 eV
f
Note that today the motion of a particle with mass greater than the small number
T0 = 2.73 K = 2.4 · 10−4 eV is of course non-relativistic and therefore the contribu-
tion to the energy density is dominated by its rest mass energy. A Standard Model neu-
trino has geff = 2·3/4 = 1.5 and decoupled at a few MeV when geff s =g
eff = 10.75.
We find
i m νi
Ων ν̄ h 2 = . (68)
(93 eV)
As we will see, present estimates of the neutrinos masses, based on the observation
of neutrino oscillations, give a sum much less than 1 eV, which means that neutrinos
are not the main form of dark matter. On the other hand, we are now rather certain
that they do contribute a small fraction of non-baryonic dark matter!
This analysis has been valid for hot relics, or hot dark matter. For cold relics (parti-
cles that were non-relativistic at freeze-out) the Eq. (65) has to be solved numerically.
There one finds that for massive particles in the mass range between, say, 10 GeV
and a few TeV, x f ∼ 1/20, and moreover to a good approximation the relic density
only depends on the cross section times velocity,
142 L. Bergström
2.8 · 10−26 cm3 s−1

ΩX h2 0.11 × . (69)
σ A |v|
Another striking result is that, if one gives typical gauge couplings to the particle X ,
and a mass of typical weak interaction magnitude (100–300 GeV, say), then σ A v
is such that the resulting Ω X h 2 ∼ 0.11. This is the essence of what is sometimes
called the “WIMP miracle”.
As can be understood, the value of x f when Yχ leaves the equilibrium curve
is lower for a smaller cross section σ A . This is because of the fact that in thermal
equilibrium, massive particles will eventually be exponentially suppressed. That is,
more weakly interacting particles decouple earlier, and since the equilibrium curve
for a non relativistic species drops fast with increasing x, more weakly coupled
particles will have a higher relic abundance.
Going through the numerical analysis one finds that a hypothetical neutrino with
mass m ν ∼ 3 GeV would also have about the right mass to close the universe. On
the other hand, the range between 90 eV and 3 GeV is cosmologically disallowed
for a stable neutrino. There are arguments from large-scale structure formation that
favour cold relics over hot relics, so such a neutrino would be a good dark matter
candidate. Data from the LEP accelerator at CERN have, however, excluded any
ordinary neutrino with a mass in the GeV range.
3.1 Coannihilations
There are instances when the simple treatment discussed here has to be improved.
One example is for instance the possibility that entropy may be generated by other
particles than those of the Standard Model, before, at, or after decoupling. Another
example, which for instance appears in some supersymmetric models, is that of
coannihilations. This was first discussed in [10], here we follow the more detailed
treatment in [11].
We will here outline the procedure developed in [2, 11] which is used in Dark-
SUSY [144]. For more details, see [2, 11]. DarkSUSY is a free Fortran package
that can be used to compute a variety of dark matter related quantities, such the relic
density and the scattering and annihilation rates to many different channels. It was
developed for computations in the Minimal Supersymmetric extension to the Stan-
dard Model (MSSM), but it is modular and can be adapted to most WIMP models.
We consider annihilation of N particles with mass m i and internal degrees of
freedom gi . For convenience, we may order them such that m 1 ≤ m 2 ≤ · · · ≤
m N −1 ≤ m N . For the lightest particle (which is the dark matter candidate, if a sym-
metry is guaranteeing the stability, like what is called R-parity for supersymmetry,
see Sect. 5), we use both the notation m 1 and m χ .
All heavier particles will eventually decay to the lightest, stable, and therefore we
add the number densities up,
N
n= ni .
i=1
The scattering rate of particles on particles in the thermal background “soup” is gen-
erally much faster than the annihilation rate, since the background particle densities
of Standard Model particles, n S M is much larger than each of the particle densities
in the dark sector n i . The important SM particles are, as we have seen, those that
are relativistic and cold dark matter particles (WIMPs) are nonrelativistic, and thus
suppressed by the Boltzmann factor. Thus, the n i distributions remain in thermal
equilibrium during their (“chemical”) freeze-out.
We then get
dn
= −3H n − σeff v n 2 − n 2eq (70)
dt
where eq
eq
ni n j
σeff v = σi j vi j . (71)
n eq n eq
ij
with
( pi · p j )2 − m i2 m 2j
vi j = . (72)
Ei E j
Using the Maxwell-Boltzmann approximation for the velocity distributions one

can derive the following expression for the thermally averaged annihilation cross
section [11] √
∞ 2 s
0 dpeff peff Weff K 1 T
σeff v = . (73)
4
gi m i2 mi 2
m1T i g1 m 2 K 2 T
1
where K 1 (K 2 ) is the modified Bessel function of the second kind of order 1 (2), T is
the temperature, s is the usual Mandelstam variable and
pi j gi g j
Weff = Wi j
peff g12
ij

[s − (m i − m j )2 ][s − (m i + m j )2 ] gi g j
= Wi j . (74)
ij
s(s − 4m 21 ) g12
Here,
1/2 1/2
s − (m i + m j )2 s − (m i − m j )2
pi j = √ , (75)
2 s
144 L. Bergström
and the invariant annihilation rate is

√
Wi j = 4 pi j sσi j = 4σi j ( pi · p j )2 − m i2 m 2j = 4E i E j σi j vi j (76)
and, finally, the effective momentum
1
peff = p11 = s − 4m 21 . (77)
2
Since Wi j (s) = 0 for s ≤ (m i + m j )2 , the terms in (74) are non-negative. For a

two-body final state, Wi j is given by the expression
2−body |k|
Wi j = √ |M|2 dΩ, (78)
16π 2 gi g j S f s
internal d.o.f.
that after some manipulations leads to (63). Here k is the final center-of-mass momen-
tum, S f is a symmetry factor equal to 2 for identical final particles.
So, what could the dark matter be? It turns out that in particle physics, there are
hypothetical particles, like supersymmetric partners of ordinary particles, that have
the right interaction strength and mass range to be promising dark matter candidates.
In particular, the neutralino has all the properties of a good dark matter candidate.
Since it is electrically neutral it does not emit or absorb radiation which makes it
‘dark’ (invisible matter is thus a better term than dark matter). The couplings of neu-
tralinos are generally of weak interaction strength, but the large number of possible
annihilation channels, which depends on the unknown supersymmetry breaking para-
meters, makes an exact prediction of mass and relic abundance uncertain. Scans of
parameter space show, however, that a neutralino in the mass range between 30 GeV
and a few TeV could give a relic density close to the critical density. We will later in
these notes have much more to say about this.
3.2 Inflation
An important ingredient in today’s cosmology is, as mentioned, the concept of infla-

tion, which was introduced by Alan Guth in the early 1980s. Here we use the near
t
constancy of H to write a2 /a1 = exp( t12 H (t)dt): that is, Einstein’s equations
including a cosmological constant read
1
Rμν − gμν R = 8πGTμν + Λgμν . (79)
2
Here we have put the Λ term on the right hand side, which shows that a cosmological
term acts as a stress-energy tensor, albeit with the unusual equation of state pvac =
−ρvac . (We have already used that one may trivially include vacuum energy in
the term proportional to G, with ρΛ = Λ/(8πG).) The equation of state implies

that the entropy density according to (55) is s ∼ ρ + p = 0. This means that,
when vacuum energy dominates, the entropy vanishes. This can be understood from
statistical mechanics. Entropy is related to the total number of degrees of freedom,
and the vacuum (at least if it is unique) is just one state, that is only one degree
of freedom. Of course, the entropy that was in a patch before inflation will still be
there after inflation—but it will be diluted by an exponentially large factor due to the
expansion.
In the situation when the constant vacuum energy dominates the expansion, the
Friedmann equation (2) becomes very simple:
2
ȧ Λ
H2 = = (80)
a 3
or
ȧ Λ
H= = = const (81)
a 3
with the (de Sitter) solution

a ∼ eHt . (82)
In inflation, the expansion rate is constant, which causes an exponential growth of

the scale factor.
In many models of inflation, the phase transition involving a scalar field, the
inflation field, took place at temperatures around the hypothetical Grand Unification
scale TGU T ∼ 1015 GeV, at the corresponding Hubble time H −1 ∼ 10−34 s. If the
universe stayed in the inflationary state for a short time, say 10−32 s, many e-folds of
inflation took place. When inflation stopped, the huge vacuum energy of the inflation
field went into creation of ordinary particles, and a reheating of the universe took
place. The reheating temperature is of the order of the temperature of the phase
transition, TR H ∼ 1015 GeV if the inflation is strongly enough coupled to ordinary
matter, as it is in many successful models of inflation.
Let us see what happened to a small region with radius of for example 10−23 cm
before inflation. The entropy within that volume was only around 1014 , but after
inflation the volume of the region has increased by a factor given by the cube of
the scale factor, (e100 )3 = 10130 . Thus, after the entropy generated by reheating,
the total entropy within the inflated region had grown to around 10144 . Entropy was
generated because the equation of state changed from p = −ρ to p = ρ/3, meaning
that the entropy density s ∼ p + ρ increased dramatically.
This huge entropy increase solves many problems of cosmology. The “horizon
problem”—i.e., how came that regions of the universe that are too far from each
other to be causally connected today, still have exactly the same microwave back-
ground temperature—is solved since our whole observable universe arose from a
very small thermalized volume before inflation, and the smooth region after inflation
had sufficient entropy to encompass our observable universe.
146 L. Bergström
During inflation the energy density, and the negative pressure, of the universe
were constant, whereas the scale factor a increased exponentially. This means that
the total Ω after inflation was exponentially close to unity. (Like a balloon which
would inflate to be as big as the Earth would locally look very flat.) Thus, the present
value should also be equal to unity with an accuracy of many decimal places, perhaps
the most important successful prediction of inflation.
Even if Ω = 1 is predicted, there is nothing telling us the subdivision of Ω into
contributions from radiation, matter and vacuum energy. As we have noted, however,
the ‘natural’ contribution of ΩΛ is either extremely small or extremely large. Only
during very brief epochs can ΩΛ be of similar magnitude as the matter contribution
Ω M . This is actually a still unsolved problem, why is it that the energy density in
matter ρ M is about the same as ρΛ today?
The period of inflation and reheating is strongly non-adiabatic, since there was an
enormous generation of entropy at reheating. After the end of inflation, the universe
‘restarted’ in an adiabatic phase with the standard conservation of aT , and it is
because the universe automatically restarts from very special initial conditions given
by inflation that the horizon and flatness problems are solved.
It is instructive to see how inflation can be produced in field theory. A Lagrangian
density of the form
1
L = ∂ μ φ∂μ φ − V (φ) (83)
2
can be shown to give a contribution to the energy-momentum tensor T μν of the form
T μν = ∂ μ φ∂ ν φ − Lg μν . (84)
For a homogeneous state, the spatial gradient terms vanish, meaning that T μν
becomes that of the perfect fluid type. If one would keep the gradient terms, one sees
that they are divided by a(t)2 , which means that after a short period of inflation they
are exponentially suppressed. The resulting ρ and p are
φ̇2
ρ= + V (φ) (85)
2
and
φ̇2
p= − V (φ), (86)
2
and we see that the equation of state ρ = − p will be fulfilled if we can neglect the
kinetic term ∼ φ̇2 (this is called “slow-roll” inflation).
The exact equations of motion of φ can be derived from the condition of vanishing
μν
covariant divergence of the energy-momentum tensor, Tν = 0, which gives
φ̈ + 3H φ̇ + V (φ) = 0 (87)
This is similar to the equation of motion of a ball in a potential well with friction
∼3H φ̇, and can be solved by elementary methods. We assume that at very high
temperatures, φ = 0 gives the locations of the minimum of the potential. Temperature
dependent terms in the effective potential then generate another minimum for φ =
φvac = 0, an example of what is called spontaneous symmetry breakdown. To
produce a long enough period of inflation and a rapid reheating after inflation, the
potential V (φ) has as mentioned to be of the “slow-roll” type, with the field spending
a long time on the nearly flat, horizontal part of the potential. In the beginning, on
the almost horizontal slow ‘roll’ towards a deep potential well, φ̈ can be neglected,
and the slow-roll equation of motion is
3H φ̇ + V (φ) = 0, (88)
together with the Friedmann equation

8πG N 1 2
H2 = φ̇ + V (φ) , (89)
3 2
which during slow roll, when φ̇2 is small, can be approximated by
8πG N
H2 = V (φ). (90)
3
One can from this get an expression for the number Nφ of e-folds of the scale factor,
φ2
a2 V (φ)
Nφ ≡ log = H dt ∼ dφ. (91)
a1 φ1 V (φ)
Thus, for a large growth of the scale factor, V (φ) has to be very flat (V (φ) ∼ 0).
This may be unnatural except perhaps in some supersymmetric theories where ‘flat
directions’ can occur because of the pattern of supersymmetry breaking. In a situation
of such a slow roll of the inflation field, the exact form of the potential does not matter
so much, and the relevant physics can be expressed in terms of the so-called slow-roll
parameters
2
Ḣ φ̇2 1 V
ε = − 2 = 4πG N 2 = (92)
H H 16πG N V

1 V V
η= = (93)
8πG N V 3H 2
where the second equation in (92) comes from taking the derivative of (89) and
inserting into (87). The variable ε is a measure of the change of the Hubble expansion
during inflation; for inflation to happen at all, ε < 1 is needed.
148 L. Bergström
In the picture of the rolling ball, reheating corresponds to oscillations in the

potential well. Thus, for enough entropy to be generated the well has to be rather
steep. The problem of constructing a suitable potential is to simultaneously have it
flat near φ = 0 and steep near φ = φmin .
A way to avoid a phase transition, and in fact the simplest model of inflation is
the chaotic inflation model of Andrei Linde [12]. It relies on the fact that the key
ingredient for inflation to occur is that the field is rolling slowly, so that the energy
density is nearly constant during a sufficient number of e-foldings of the scale factor.
Since the rolling is damped by the presence of the term proportional to H in (87),
and H according to the Friedmann equation is given by the height of the potential (if
kinetic terms can be neglected), inflation will be possible for any positive, power-law
potential V (φ), for example the simplest V (φ) = 21 m 2 φ2 , as long as the field values
start out large. As Linde has argued, this may not be unreasonable since these initial
values may be given stochastically (“chaotically”) at the Planck epoch, and those
regions where the field values are large start to inflate rapidly dominating the volume
of the universe. There are also constructions relying on the existence of more than
one scalar field, keeping the same general features but with changes in the details.
Since the value of the total energy density Ω = 1 is found observationally in
current measurements of the CMBR anisotropy which yield Ω = 1.003 ± 0.010,
the most natural explanation seems to be that the universe has indeed gone through
a period of inflation. An important test of inflation may be produced by the upcom-
ing measurements from the Planck satellite of the detailed pattern of temperature
fluctuations in the CMBR. Inflation predicts a nearly but not perfect scale-invariant
spectrum of fluctuations (which is when the index of scalar fluctuations n s = 1), and
present measurements from WMAP give n s ∼ 0.96, in excellent agreement. Inflation
could also have generated gravitational (tensor) waves during the phase transitions
which would give a particular pattern (“B-modes”) in the microwave sky. However,
the amplitude of tensor to scalar fluctuations depends rather strongly on the model.
It will be interesting to see whether the Planck satellite, when cosmological data are
released in early 2013, will detect such a B-mode pattern.
4 Basic Cross Sections for Neutrinos and γ-Rays
Among the various messengers from the Galaxy and structures further away, neu-
trinos and γ-rays have the advantage that they follow straight lines (or to be more
exact, geodesics; the deviations from straight lines can in almost all cases be safely
neglected—exceptions are given for particles originating or travelling very near black
holes). This is in contrast to all other cosmic rays, electrons, protons, nuclei, and
antimatter (positrons, antiprotons and some antinuclei like antideuterons). Neutrons
would in principle also travel rectilinearly apart from effects of their magnetic
moment. However, their finite lifetime (of the order of 10 min in the rest frame)
means that for energies less than a few TeV which is the energy range we will
generally be concerned with, they cannot travel over astrophysical distances. They
β-decay to a proton, an electron and an (anti-)neutrino.
Although neutrinos and γ-rays (high-energy photons) are both encompassed in
the Standard Model of particle physics and therefore in principle should interact with
similar strengths given by gauge couplings, this is in practice not so. The reason is
the difference that the photon is a massless, spin-1 gauge particle, i.e., a mediator of a
force (the electromagnetic force, i.e., it couples to electrons and protons, and all other
particles with electric charge) while the neutrino is a spin-1/2 matter particle which
in turn interacts through the weak force mediated by the heavy W and Z bosons. The
large, important difference of masses between weak bosons and the photon is due to
the hitherto experimentally unverified, but hopefully soon to be verified mechanism,
the Higgs mechanism. This breaks the gauge group of the Standard Model, leaving
only the electromagnetic U (1)em unbroken and therefore the photon massless. It
means that for energies up to 1 TeV or so, neutrinos have very small cross section,
which however rises with energy, until the centre-of-mass energy is of the order of
the W and Z masses, at which point neutrinos start to react roughly as strongly as
photons. Let us now discuss in some more detail how some simple particle cross
sections are computed.
4.1 Estimates of Cross Sections
The calculation of collision and annihilation cross sections, and decay rates of par-
ticles, is an important task in particle physics. Here we will present only a brief
outline of how this is done, and focus on ‘quick-and-dirty’ estimates which may be
very useful in cosmology and astrophysics. For the local microphysics in the FLRW
model, only three interactions—electromagnetic, weak and strong—between parti-
cles need to be considered. The gravitational force is completely negligible between
individual elementary particles—for instance, the gravitational force between the
proton and the electron in a hydrogen atom is around 1040 times weaker than the
electromagnetic force. However, gravity, due to its coherence over long range, still
needs to be taken into account through its influence on the metric. This means that
the dilution of number densities due to the time dependence of the scale factor a(t)
has to be taken into account. In the next section we will see how this is done.
Let us begin with the interaction strengths. The strength of the electromagnetic
interaction is governed by the electromagnetic coupling constant gem , which is simply
the electric charge. As usual, we take the proton charge e as the basic unit and can
thus write
gem = Qe (94)
where Q is the charge of the particle in units of the proton charge (for a u-quark, for
example, Q u = +2/3). In our system of units,
150 L. Bergström
e2
≡ αem (95)
4π
where αem is the so-called fine structure constant which has the value of around
1/137 at low energies.2 The weak coupling constant is of similar magnitude:
e
gw = (96)
sin θW
with θW the weak interaction (or Weinberg) angle, which has the numerical value
sin2 θW ∼ 0.23. The fact that the weak and electromagnetic coupling constants are
of the same order of magnitude is of course related to the fact that they are unified
in the Standard Model to the ‘electroweak’ interaction.
The coupling constant of the strong interaction, gs , is somewhat higher. Also, it
runs faster (it decreases) with energy than the electromagnetic coupling. At energies
of a few GeV,
g2
αs ≡ s ∼ 0.3 (97)
4π
Let us look at the Feynman diagram for a simple process like e+ e− → μ+ μ− (Fig. 1).
The amplitude will be proportional to the coupling constants at both vertices, which
in this case are both equal to e. The cross section, being proportional to the square
of the amplitude, is thus proportional to e4 ∝ α2 .
The total energy of the e+ e− pair in the centre of momentum frame is E√ +
cm (e ) +
E cm (e− ), which is conventionally noted with the Mandelstam variable as√ s. Since
the total momentum in this frame is zero, the four-momentum
√ p μ = ( s, 0, 0, 0)
is identical to that of a massive particle of mass M = s which is at rest. Energy
and momentum conservation then tells us that the photon in the intermediate state
has this four-momentum. However, a freely propagating photon is massless, which
means that the intermediate photon is virtual by a large amount. In quantum field √
theory one can show that the appearance of an intermediate state of virtual mass s
for a particle with real rest mass Mi is suppressed in amplitude by a factor (called
the propagator factor)
Fig. 1 A Feynman diagram

representing the annihilation e- -
of an electron and a positron
to a muon pair
+ +
e time
2This coupling constant usually denoted just α without the subscript, as all others, depends on the
energy scale, for example, the energy transfer, of the process. At 100 GeV energy αem is ∼1/128.
P(s) = 1/(s − m i2 ) (98)
In this case (m γ = 0), we have a factor of 1/s. (If one does this rigorously, one should
insert a small imaginary part in the denominator, which defines how the singularity
on the mass shell is treated.) The outgoing particles (in this case the muons) have
a large number of possible final states to enter (for example, all different scattering
angles in the centre of momentum frame). This is accounted for by the so-called
phase space factor φ, which generally grows as s for large energies. For the cross
section σ 2
α
σ(e+ e− → μ+ μ− ) ∝ φ (99)
s2
with φ the phase space factor. If s is large compared to m 2e and m 2μ , φ ∝ s, and
α2
σ(e+ e− → μ+ μ− ) ∼ (100)
s
This is not an exact expression. A careful calculation (see next section) gives
4πα2 /(3s), but it is surprisingly accurate and often accurate enough for the esti-
mates we need in big bang cosmology.
Since the weak coupling strength is similar to the electromagnetic strength, the
same formula is valid for, e.g., νe + e → νμ + μ which goes through W exchange
(see Fig. 2). The only replacement we need is 1/s → 1/(s − m 2W ) for the propagator,
thus
α2 s
σ(ν̄e + e− → ν̄μ + μ− ) ∼ (101)
(s − m 2W )2
When s m 2W , this gives σw ∼ α2 s/m 4W , which is a very small cross section, e.g.,
for MeV energies. One should notice, however, the fast rise with energy due to the
factor s. This is the historical reason for the name ‘weak interaction’, which is really
not appropriate at high energies (much larger than m W ), where the two types of cross
sections become of similar size.
Fig. 2 A Feynman diagram representing the annihilation of an electron neutrino and a positron to
a muon neutrino and a muon
152 L. Bergström
e e
Fig. 3 A Feynman diagram representing the γe → γe process. In the classical limit, this is called
Thomson scattering. The quantum version is called Compton scattering, and in the relativistic
regime, the result is given by the Klein-Nishina formula
Note that once one remembers the factors of coupling constants and the prop-
agators, the magnitude of cross sections can often be estimated by simple dimen-
sional analysis. A cross section has the dimension of area, which in our units means
(mass)−2 . It is very useful to check that the expressions (100) and (101) have the
correct dimensions.
A decay rate Γ can be estimated in a similar way. If a collection of identical
unstable particles decay, their number density decreases as e−Γ t which means that
Γ has the dimensions of mass.
A fermion has a propagator that behaves as 1/m (instead of 1/m 2 ) at low energies.
This means that the Thomson cross section σ(γe → γe) at low energies E γ m e
can be estimated to be (see Fig. 3)
α2
σT ≡ σ(γe → γe) ∼ (102)
m 2e
4.2 Examples of Cross Section Calculations
The estimates we have given are in many cases sufficient for cosmological and
astrophysical applications. However, there are cases when one would like to have a
more accurate formula. We now provide only a couple of examples and summarize
the general framework for calculation and the main results.
4.3 Definition of the Cross Section
The differential cross section dσ/dt for 2 → 2 scattering a + b → c + d is given

by the expression
dσ |2
|T
= (103)
dt 16πλ s, m a2 , m 2b
where the Lorentz invariant Mandelstam variables are s = ( pa + pb )2 , t = ( pa −

|2 is the polarization-summed and squared quantum
pc )2 and u = ( pa − pd )2 . |T
mechanical transition amplitude, while the function λ is defined as
λ(x, y, z) ≡ x 2 + y 2 + z 2 − 2x y − 2x z − 2yz . (104)
For a 2 → 2 process, the kinematically allowed region in s is
s > (m c + m d )2 (105)
which can be understood

√ from energy conservation: In the centre of momentum
system (cms), where s corresponds to the total energy, at least the rest mass energy
m c + m d has to be provided.
cms | ≤ 1,
The kinematical limits for t can be obtained from the condition | cos θac
with
s(t − u) + (m a2 − m 2b )(m 2c − m 2d )
cos θac
cms
= . (106)
λ(s, m a2 , m 2b ) λ(s, m 2c , m 2d )
A typical calculation (following the treatment of [1]) involves computing the

matrix element in terms of s and t and carrying out the t integration to obtain the
total cross section.
In the one-photon exchange approximation, the cross section for the annihilation
process e+ e− → μ+ μ− is

+ − 2πα2
+ − β2
σ(e e → μ μ ) = β 1− (107)
s 3
where the only approximation made is to neglect m e (this is allowed, since m 2e /

m 2μ 1). Here β is the velocity of one of the outgoing muons in the centre of momen-
tum system, β = 1 − 4m 2μ /s. In the relativistic limit of s m 2μ , (β → 1), this
becomes
4πα2
σ e+ e− → μ+ μ− large s = (108)
3s
in agreement with our simple estimate (100).
154 L. Bergström
Fig. 4 The same Feynman diagram can, after rotation of the external legs, describe both eγ →
eγ, e+ e− → γγ, and γγ → e+ e− . Here time is as usual flowing from left to right
4.4 The γγee System
By different permutations of the incoming and outgoing particles, the basic γγee
interaction (shown in Fig. 3) can describe all of the astrophysically important
processes (see the contributions by F. Aharonian and C. Dermer in this volume)
γγ → e+ e− , e+ e− → γγ, and γe± → γe± , see Fig. 4.
For γγ → e+ e− the result is
πα2 1 + β
σ γγ → e+ e− = 1 − β 2
3 − β 4
ln + 2β β 2
− 2
2m 2e 1−β
(109)
where β now is the velocity of one of the produced electrons in the centre-
of-momentum frame, β = 1 − 4m 2e /s. Near threshold, i.e. for small β, the expres-
sion in square brackets can be series expanded to 2β + O(β 2 ), and thus
πα2
σ γγ → e+ e− small β (110)
m 2e
e+ e- system
1e-23
Cross section [cm2]
1e-24
1e-25
1e-26
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Fig. 5 The cross sections (in cm2 ) for photon–photon annihilation (dotted line), e+ e− → γγ
(dashed line) and Compton scattering (solid line) as a function of the cms velocity β of the electron
In the other extreme, β → 1,

√
4πα2 s
σ γγ → e+ e− s4m 2 ln −1 (111)
e s me
One could in fact have guessed most of this to a fair amount of accuracy by the
simple dimensional and vertex-counting rules. At low energy, the only available mass
scale is m e , so the factor α2 /m 2e could have been guessed for that reason. The factor
β could also have been inferred with some more knowledge of non-relativistic partial
wave amplitudes. At low energy, the = 0 (S-wave) amplitude should dominate, and
this contributes to the cross section proportionally to β. A partial wave contributes
to the total cross section with a term proportional to β 2+1 . We see from Eq. (107)
that in the case of e+ e− → μ+ μ− the S-wave dominates at low energy, but when
β → 1, the P-wave contribution is 1/3. At high energy, when m e can be neglected,
the dimensions have to be carried by s. Only the logarithmic correction factor in
(111) could not have been easily guessed.
These formulas show that the γγ → e+ e− cross section rises from threshold to
a maximum at intermediate energies and then drops roughly as 1/s at higher β, i.e.,
higher cms energy in the process (see Fig. 5).
The results for the reverse process e+ e− → γγ are of course extremely similar.
Now, the process is automatically always above threshold. For β → 0 (with β now the
velocity
of one of the incoming particles in the cm-system, still given by the formula
β = 1 − 4m 2e /s), the flux factor ∼1/β implicit in Eq. (103) diverges. Since the
156 L. Bergström
outgoing photons move away with β = c = 1 there is no partial-wave suppression

factor, and we can thus expect the cross section at low energy to behave as
α2
σ e+ e− → γγ low energy ∼ (112)
βm 2e
and the high-energy behavior by the same formula, with m 2e replaced by s (and possi-
bly a logarithmic factor). These expectations are borne out by the actual calculation,
which gives

+ − πα2 1 − β 2 3 − β 4 1+β
σ e e → γγ = ln −2+β 2
(113)
2βm 2e 2β 1−β
Note the similarity with Eq. (109). The 1/β behavior of the cross section (see the
dashed curve in Fig. 5) was noted by Arnold Sommerfeld in the 1930s, and he showed
how one can make an improved calculation valid at very small velocities by not only
treating the annihilating particles as plane waves, but using wave functions appro-
priate for the attractive Coulomb interaction between the electron and positron. He
thereby described a generic mechanism, the so-called Sommerfeld enhancement
mechanism, which recently has played an important role for dark matter calcula-
tions, as we will see later in Sect. 6.3.
Compton and Inverse Compton Scattering
As the final example, we consider Compton scattering γ +e− → γ +e− . Historically,

this was first computed for an incoming beam of photons of energy ω which hit
electrons at rest. Later on, the related process of a very high energy electron or positron
colliding with a low-energy photon (such as coming from the cosmic microwave
background, or from infrared or optical radiation created in stellar processes) and
upscattering that photon to high, maybe GeV energy or higher, has been found to
be very important in astrophysics. Despite being really one and the same process,
the latter situation is often referred to as the inverse Compton or IC process. In fact,
the inverse Compton process is one purely leptonic process of creating high-energy
γ-rays, and could be important for the emission γ-rays in several cases, such as
AGNs, GRBs and various supernova remnants. However, to completely prove such
a scenario, it is important to search for, or give upper limits on, neutrino emission. In
competing hadronic models of emission, where γ-rays mainly come from π 0 decays,
one should also have essentially the same amount of charged pions which decay into
a leptons and neutrinos. Also for some “leptophilic” models of dark matter, where
electrons and muons are main annihilation products, inverse Compton processes may
be quite important, e.g., near the galactic centre where radiation fields are large.
For scattering of a photon by an angle θ with respect to the incident photon
direction, the outgoing photon energy ω is given by energy-momentum conservation
meω
ω = (114)
m e + ω (1 − cos θ)
In this frame, the unpolarized differential cross section, the Klein-Nishina formula
as it was first computed by Klein and Nishina shortly after Dirac had presented his
equation describing relativistic electrons (and positrons), is
2
dσ α2 ω ω ω
= + − sin2 θ (115)
dΩ 2m 2e ω ω ω
Integrated over all possible scattering angles this gives the total cross section
πα2 (1 − β)
σ(γ + e → γ + e) =
m 2e β 3
1 + β 2β 3 (1 + 2β)
4β
× + β 2 + 2β − 2 ln − (116)
1+β 1−β (1 + β)2
where β is now the incoming electron velocity in the centre of momentum frame,
β = (s − m 2e )/(s + m 2e ). If one expands this result around β = 0, one recovers the
Thomson scattering result
8πα2
σThomson = ∼ 6.65 · 10−25 cm2 (117)
3m 2e
and the large-s, so-called Klein Nishina regime gives

2πα2 s 1
σKN = ln + (118)
s m 2e 2
We see that for photon energies much larger than m e —that is, in the Klein-Nishina
regime—the Compton cross section falls quite rapidly.
In the classical Compton scattering situation, the outgoing photon energy is always
less than the incoming one. Thus, energetic photons traveling through a gas of cold
electrons will be ‘cooled’ by Compton scattering. In the IC case (for example for the
cosmic microwave background radiation passing through a galaxy cluster with hot
gas) energetic electrons may instead transfer energy to photons, thereby ‘heating’
them. For CMBR this is called the Sunyaev-Z’eldovich effect, and has a large range
of applicability (for instance, it has recently been used to find galaxy clusters).
When computing actual numbers for the cross sections (which should have the
dimensions of area) in our units, a useful conversion factor is
1 GeV−2 = 0.389 · 10−27 cm2 (119)

158 L. Bergström
In Fig. 5 the numerical results are summarized. The cross sections are shown (in
cm2 ) for γγ → ee, ee → γγ and γe → γe as a function of the cms velocity β of
the electron. We see in the figure the different behaviour at low cms velocity already
discussed, but that they show a similar decrease at high energy.
Another process of great astrophysical importance is bremsstrahlung. By this
is meant the emission of photons from charged particles which are accelerated or
decelerated. If this acceleration is due to circular motion in a magnetic field, the term
synchrotron radiation is used. Through these processes (unlike Compton scattering)
the number of photons can change. This is needed, for instance in the early universe,
if thermal equilibrium is to be maintained, since the number density of photons has
to vary, as it depends strongly on temperature. Most of the produced photons have
very low energy (long wavelength). If fast electrons pass through a region where
synchrotron radiation and bremsstrahlung occur, these low-energy photons may be
upscattered in energy through the inverse Compton process. This may for example
explain the observations of very high-energy photons in active galactic nuclei.
For a detailed discussion of these and other quantum electrodynamic (QED)
processes, see standard textbooks in quantum field theory, for example, [13], or
a simplified treatment along the lines given here, in [1]. And, of course, for good
examples of the use of these processes in astrophysics, see the accompanying lectures
by F. Aharonian and C. Dermer in this volume.
4.5 Processes Involving Hadrons
Since protons and neutrons belong to the most common particles in the universe, it
is of course of great interest to compute processes where these and other hadrons
(such as pions) are involved. This is, however, not easy to do from first principles.
The reason that in the previous section we could compute so accurately weak and
electromagnetic processes is that we could use perturbation theory (as summarized,
for example, in Feynman diagrams). The expansion parameter, the electroweak gauge
coupling constant g or rather αew = g 2 /(4π) ∼ 10−2 , is small enough that a lowest-
order calculation is enough to obtain very accurate results.
In quantum chromodynamics (QCD), which also is a gauge theory just as QED,
we also have a coupling constant αs . Due to the fact that the gauge group of QCD
is SU(3), which involves self-interactions of the 8 spin-1 gluons, there are important
differences. We say that QCD is a non-abelian gauge theory whereas QED is based on
the abelian group U(1) with only one spin-1 gauge field, the photon. One consequence
of this difference is that QCD has what is called asymptotic freedom meaning that the
coupling strength which is of order unity at a few hundred MeV, “runs” to smaller
values for large energies. The energy scale is set, for example, by the energy or
momentum transfer Q (Q 2 ≡ −t with t the usual Mandelstam variable) in the
process. Thus, for processes with large Q 2 , we should be able to use low-order
perturbative QCD, although with lower accuracy than for QED due to the possible
importance of higher-order corrections. At low energies when the QCD coupling
becomes of the order unity perturbation theory breaks down. In the nonperturbative
regime we have to rely on empirical methods, such as “QCD sum rules” [14] or large
computer simulations, where one tries to solve QCD by formulating it as a field theory
on a lattice. Although the problem is notoriously difficult, the agreement between
the spectrum of hadrons, i.e., the masses and quantum numbers of the lowest-lying
states with experimentally measured quantities, is quite satisfactory for the most
recent numerical simulations [15].
For processes like proton-proton scattering at low energies, the picture of strong
interactions being due to the exchange of gluons breaks down. Instead one may
approximate the exchange force as being due to pions and other low-mass mesons
with surprisingly good results (this is in fact what motivated Yukawa to predict
the existence of pions). If one wants to make crude approximations of the strong
interaction cross section in this regime, σs ∼ 1/m 2π is a good estimate.
In the perturbative regime at high Q 2 , the scattering, for example, of an electron off
a proton (‘deep inelastic scattering’) can be treated by the successful parton model.
Here, the momentum of a hadron at high energy is shared between its different
constituents. Among the constituents are of course the quarks that make up the
hadron (two u and one d quarks in the case of the proton), i.e., the valence quarks.
In addition, there may be pairs of quarks and antiquarks produced through quantum
fluctuations at any given “snapshot” of the hadron. The incoming exchange photon
sent out from an electron in ep scattering may hit these “sea quarks”, which will
therefore contribute to the scattering process.
Since the partons interact with each other, they can share the momentum of the
proton in many ways. Thus, there will be a probability distribution, f i (x, Q 2 ), for a
parton of type i (where i denotes any quark, antiquark or gluon) to carry a fraction x
of the proton momentum. These functions cannot be calculated from first principles.
However, once determined (by guess or by experimental information from various
scattering and annihilation processes) at a particular value of Q 20 , the evolution of
the structure functions with Q 2 can be predicted. This analysis, as first convincingly
performed by Altarelli and Parisi [16], gives rise to a predicted variation of the deep
inelastic scattering probability with Q 2 (so-called scaling violations) which has been
successfully compared with experimental data. The success of the perturbative QCD
program, including the running of αs in agreement with the asymptotic freedom
prediction, and the agreement of scaling violations in several processes with data,
resulted in a Nobel Prize for Gross, Wilczek and Politzer in 2004.
With the successful QCD parton model, we can now compute many electromag-
netic and weak processes, including those when hadrons are involved. For instance,
the neutrino-proton scattering cross section is be given by the scattering of a neu-
trino on a single quark or antiquark. This calculation is easily done in a way similar
to how we computed the ν̄e + e− → ν̄μ + μ− cross section. The only change is
that the contributions from all partons have to be summed over, and an integral of x
performed.
As an example, we give the expression for the electromagnetic cross section
p + p → μ+ + μ− , which is called the Drell-Yan process, in the QCD parton
model. The fundamental process must involve charged partons, i.e., quarks (since
160 L. Bergström
we assume that strong interactions dominate and thus neglect the weak contribution),
q + q̄ → γ ∗ → μ+ + μ− , with the (valence) quark taken from one of the protons and
the (sea) antiquark from the other. The momentum transfer in the process is Q 2 = ŝ,
where ŝ = ( pμ+ + pμ− )2 . We know from (108) that the parton level cross section
is 4παeq2 /3ŝ (where we have to take into account that the quark charge eq is not the
unit charge). Since the parton from proton 1 carries the fraction x1 and that from
proton 2 x2 of the respective parent proton, ŝ = x1 x2 s, with s = ( p1 + p2 )2 . The
total cross section for producing a muon pair of momentum transfer ŝ is thus
dσ 4πα2 1 1
= kc ei2 d x1 d x2
d ŝ 3ŝ q 0 0

× f q (x1 , ŝ) f q̄ (x2 , ŝ) + f q̄ (x1 , ŝ) f q (x2 , ŝ) δ(ŝ − x1 x2 s)
Here kc is a colour factor, which takes into account that for a given quark of a given
colour, the probability to find in the other proton an antiquark with a matching (anti-)
colour is 1/3. Thus, in this case kc = 1/3. In the reverse process, μ+ + μ− → q + q̄,
all the quark colours in the final state have to be summed over (each contributes to
the cross section), so in that case kc = 3.
4.6 Neutrinos
Neutrinos are the neutral counterparts of the charged leptons: e, μ and τ . There are
therefore three types of “active” neutrinos in the Standard Model of particle physics:
νe , νμ and ντ . Neutrinos are fermions, i.e., spin- 21 particles. Apart from their possible
gravitational interactions, they interact with matter only through the exchange of the
mediators of the weak force, the W and Z bosons. They are fundamental particles
without constituents, as far as is known, have extremely small masses and lack electric
charge. Among the most spectacular events in astrophysics are supernova explosions.
In a few seconds, more energy is released in neutrinos from the forming neutron star
than all the electromagnetic emission from an entire galaxy over a decade.
Neutrinos would provide an important contribution to the total energy density
of the universe if they had a mass in the eV range. Present-day analyses of the
microwave background and the matter distribution favour as we mentioned, cold
dark matter over hot dark matter which eV-scale Standard Model neutrinos would
constitute. However, neutrinos are plausibly part of dark matter, as one of the most
important discoveries of the last 15 years has been the definitive demonstration that
neutrinos are not massless. It has been shown that neutrinos species can transform
into each other through quantum mechanical mixing, which means that in principle
they could also decay into each other, e.g., by emitting a photon. The huge mass
scale of the W and Z bosons means, however, that the lifetime even for the heaviest
neutrino is much longer than the age of the universe, so they are effectively stable.
Their production through thermal processes in the early universe would then mean
that their number density is of the same order of magnitude today as that of the
microwave background photons, and since they are electrically neutral they qualify
as dark matter. This was in fact one of the most studied candidates for dark matter in
the 1970s, when one did not know how many different neutrino types there were (this
was fixed in the 1990s to 3 standard species, thanks to the CERN LEP experiments),
neither very much about their mass. Today’s knowledge give estimates of far less
than a percent for their contribution to Ω.
There is of course a possibility that there may exist neutrinos with weaker cou-
plings than standard model ones, and if they have mass in the keV range they would
qualify as “warm” dark matter, that would still be allowed from structure formation.
However, this involves some fine-tuning to get the correct relic density and to avoid
other experimental bounds [17].
Neutrinos are also very important information carriers from violent astrophysical
processes, which is why neutrino astrophysics is a very active field of research at the
present time. An attractive property of neutrinos is in fact their feeble interactions at
low energies, which means that they may penetrate regions with dense matter, e.g.,
the solar interior, or near remnants of supernovae, without being absorbed. Where
other particles become trapped or can only propagate through very slow diffusive
processes (for instance, it takes on the order of a million years for a photon created
near the centre of the sun to diffuse out to the solar surface), neutrinos are able to
escape. Neutrinos can thus connect regions of matter that would otherwise be isolated
from each other. Because they are almost massless, they move effectively with the
speed of light, which makes energy transfer (i.e., radiative heat conduction) very
efficient, e.g., from the interior of the sun. Unfortunately, the fact that neutrinos are
so weakly interacting, also means that they are extremely difficult to detect. As of
today, the only neutrinos of astrophysical interest that have been detected are those
created in the fusion processes in the solar interior, and the exceptional supernova in
1987, where a handful of neutrinos was detected a few hours before it was spotted
optically in the Southern sky.
Neutrino interactions with matter are divided into two kinds, neutral current (NC)
interactions mediated by the neutral Z bosons, and charged current (CC) interactions
involving the exchange of W+ and W− bosons. NC interactions are responsible for
annihilation reactions involving neutrinos,
e+ + e− → νμ + ν̄μ
or for example, and elastic scattering interactions such as
ν μ + e − → νμ + e − .
In CC interactions there is a change of fermion type, of “flavour”. For example,

an antineutrino can be absorbed by a proton, producing a neutron and a positron in
the final state. This comes about because at the parton level a u-quark in the proton is
changed into a d-quark, which means it is transformed to a neutron. In this process
162 L. Bergström
charge is transferred, both for the leptons as the neutrino becomes a charged lepton,
and for the hadrons as the positively charged proton becomes a neutron.
4.7 Neutrino Interactions
For the neutrino process (the flavour-changing charged current interaction) ν̄e e− →
ν̄μ μ− the cross section at low energies (but still high enough to produce the heavier
muon) is
gw
4s
σ ν̄e e− → ν̄μ μ− ∼ (120)
96πm 4W
Before it was known that W bosons existed, Enrico Fermi3 had written a phe-
nomenological theory for weak interactions with a dimensionful constant (the Fermi
constant) G F . The relation between Fermi’s constant and the gauge theory quanti-
ties is
GF g2
√ = w2 1.166 · 10−5 GeV−2 (121)
2 8m W
Using the Fermi constant, the cross section can now be written
G2 s
σ ν̄e e− → ν̄μ μ− = F . (122)
3π
We note that the cross section rises with s 2E ν m e and thus linearly with
neutrino energy. When s starts to approach m 2W , the W propagator 1/(s − m 2W )
has to be treated more carefully. It can be improved by writing it in the so-called
Breit-Wigner form
1 1
→ (123)
s − m 2W s − m 2W + iΓ m W
where Γ is the total decay width (around 2 GeV) of the W . We see from this that
a substantial enhancement of the cross section is possible for s m 2W . This is an
example of a resonant enhancement in the s-channel. For a target electron at rest,
this resonance occurs at around 6.3 PeV and is sometimes called the Glashow reso-
nance. If astrophysical sources exist which produce electron antineutrinos with such
high energies, the prospects of detecting them would be correspondingly enhanced.
However, well above the resonance, the cross section will again start to decrease like
1/s, just as in the electromagnetic case, e+ e− → μ+ μ− .
It should be noted that the latter process, e+ e− → μ+ μ− , also receives a con-
tribution from an intermediate Z boson. At low energies this is negligible, but due
3 Enrico Fermi is well-known today as the one who has been honoured by giving the γ-ray satellite,
FERMI, its name. This has to do with his description of acceleration processes in astrophysics.
to the resonant enhancement it will dominate near s m 2Z . This is the principle

behind the Z studies performed at the LEP accelerator at CERN (where all other
fermion-antifermion pairs of the Standard Model were also produced except for t t¯,
which was not kinematically allowed). In a full calculation, the two contributions
have to be added coherently and may in fact interfere in interesting ways, producing
for example, a backward-forward asymmetry between the outgoing muons.
A detailed calculation for neutrino energies above around 5 MeV shows that the
total cross section for the reaction ν X e− → ν X e− is well approximated by [18]:

−45 Eν
σνe = C X · 9.5 · 10 · cm2 (124)
1 MeV
where the flavour-dependent constants C X are
Ce = 1 (125)
and
1
Cμ = Cτ = (126)
6.2
The cross section is larger for electron neutrinos as they can, unlike the other
neutrino species, couple to the electrons in the target through both NC and CC
interactions.
Laboratory experiments have, so far, not succeeded in directly measuring the mass
of any neutrino. Instead, the negative results have been expressed in the form of upper
limits, due to the finite resolution of the experiments. The best (lowest) upper limits
on the mass of the electron neutrino come from the studies of the electron energy
spectrum in tritium decay:
3
H →3 He + e− + ν̄e
As the minimum amount of energy taken by the νe is its mass, the end-point energy
of the emitted electron is a measurement of m νe . According to these experiments the
mass of the electron neutrino is lower than 3 eV at the 95 % confidence level [19].
With the KATRIN experiment being constructed in Karlsruhe, Germany, one hopes
to decrease this upper limit (or find a non-zero value) by an order of magnitude [20].
The discovery of the tau neutrino was announced in 2000 by the DONUT col-
laboration at Fermilab, through appearance in charm meson decays in photographic
emulsion.
Mixing of neutrino species is a very interesting quantum mechanical effect which
may occur if the weak-interaction eigenstates νe , νμ and ντ are not the mass eigen-
states that propagate in vacuum. We can then express a flavour or weak-interaction
neutrino eigenstate, ν f , as a linear superposition of orthogonal mass eigenstates, νm :
|νf = c f m | νm .
m
164 L. Bergström
Of course, all three neutrinos may mix, but it is instructive to see what happens
if just two of the neutrinos mix, e.g, νμ ↔ νe mixing with mixing angle θ. The
time evolution of a muon neutrino wave function, produced e.g. in pion decays, with
momentum p is then
| νe (t) = − sin θe−i E 1 t | ν1 + cos θe−i E 2 t | ν2 (127)
where E 1 and E 2 are the energies of the two mass eigenstates. Two energy levels
arise if ν1 and ν2 have different masses, for the same momentum, p. Then, for small
neutrino masses m i E i ,
m2
Ei = p + i (128)
2p
The probability P(νe → νe ) = |νe | νe |2 , that an electron neutrino remains a νe

after a time t then becomes

1
P(νe → νe ) = 1 − sin2 (2θ) sin2 (E 2 − E 1 )t (129)
2
For very small neutrino masses, using (128),

! "
m 22 − m 21
P(νe → νe ) = 1 − sin2 (2θ) sin2 t (130)
4E
where E is the energy of νe .

Thus the probability the electron neutrino transforms to a muon neutrino at a time
t is 2
2 Δm
P(νe → νμ ) = sin (2θ) sin
2
t (131)
4E
where Δm 2 = |m 22 − m 21 |.
From (131) it is seen that the probability function for flavour change oscillates,
with an amplitude given by sin2 (2θ) and oscillation frequency ∼Δm 2 /E. This is now
the generally accepted reason for the deficit of solar electron neutrinos, as deduced
by combining data from the Super-Kamiokande experiment in Japan (most recently
[21]), which sees the deficit of electron neutrinos, with SNO in Canada, which has
measured the neutral current cross section, which shows no deficit [22]. As the neutral
current has the same strength for all three neutrinos, this is strong evidence that the
total flux is unchanged, but the flux of electron neutrinos has decreased due to mixing.
Numerically, the oscillation length becomes

E 1 eV2
L ν = 1.27 m. (132)
1 MeV Δm 2
In fact, a direct proof that oscillations occur in the (anti-) neutrino sector is given
by recent results from the KamLAND experiment [23], where reactor antineutrinos
have been shown to oscillate over more than one period of oscillation in L/E.
4.8 Atmospheric Neutrinos
Neutrinos are copiously produced in the atmosphere by hadronic and muonic decays
following the interaction of cosmic rays with atomic nuclei N ,
⎧
⎨ p/n + N → π + /K + + · · ·
π + /K + → μ+ + νμ
⎩
μ+ → e+ + ν̄μ + νe ,
⎧
⎨ p/n + N → π − /K − + · · ·
π − /K − → μ− + ν̄μ (133)
⎩
μ− → e− + νμ + ν̄e
Studying the end result of these reactions one expects that there are about twice
as many muon neutrinos than electron neutrinos produced in the atmosphere:
φνμ + φν̄μ
=2 (134)
φνe + φν̄e
This expectation holds at low energies. At higher energies, additional effects

have to be taken into account: for example, the competition between scattering and
decay of the produced pions, and also time dilation. As the energy spectrum of the
primary nuclei reaches out to ∼1020 eV, one expects neutrinos to be produced up to
comparable energies.
Due to the complicated chain of reactions in the cascade, computer simulations
are needed to find the differential spectrum of atmospheric neutrinos. One finds that
there is a broad peak around 0.1 GeV (∼1 cm−2 s−1 ) and at very high energies, E ν
much larger than TeV, a falling flux ∼E −3.7 .
The cross section for neutrino-nucleon interactions in a target can be calculated
by inserting the nucleon mass instead of m e in our previous example. In the region
of the maximum flux of atmospheric neutrinos the cross section is σν N ∼ 10−39
cm2 . The Super-K experiment showed that also atmospheric neutrinos oscillate, and
that part of the muon neutrinos disappear due to νμ → ντ mixing taking place. (Due
to the high τ ± lepton mass (1.8 GeV) the ντ s generated by mixing will not have
enough energy on average to make the charged current interaction ντ + N → τ + X
kinematically possible. Their contribution to the neutral current events is too small
to be easily detected.) The Super-K data on the angular and energy dependence of
the oscillation is in excellent agreement with the E/L ratio given by (132).
166 L. Bergström
4.9 Neutrinos as Tracers of Particle Acceleration
A kiloton-size detector is necessary to observe neutrinos from sources as close as

the Earth’s atmosphere, or the Sun (this actually gave a Nobel Prize to Davies and
Koshiba, in 2002). To be able to search other astrophysical objects, the required
detector mass becomes very large, megaton or even gigaton (or volume of order
km3 ).
Consider the νμ → μ charged current weak interaction in a medium,
νμ + N → μ + · · · ,
where N is a nucleon in the medium in or surrounding the detector. The muon range
rises with energy, and around 1 TeV (1012 eV) it is more than 1 km. The detection area
is therefore greatly enhanced at high energies. In water or ice, a good approximation
of the muon range as a function of energy is given by

Eμ
Rμ ≈ 2.5 ln 2 · + 1 km (135)
1 TeV
The produced muon conserves, on average, the direction of the incoming neutrino.
The average of the square of the νμ -μ angle is approximately (see [1])
1
1 TeV 2
θ2 ≈ 2 deg. (136)
Eν
The cross section for neutrino interaction with a fixed target rises linearly with
energy. Neutrino telescopes for very high energies become efficient at a few GeV,
where the product of the neutrino-matter cross section and the muon range rises
approximately as E ν2 . Above 1 GeV, the induced flux of muons from atmospheric
neutrinos, for example, is about 1 m−2 year−1 .
This detection scheme does not work as well for other types of neutrinos. Electrons
(from νe +N → e+· · · ) have a very short range as they lose energy through radiative
processes, due to their small mass. On the other hand τ leptons, the heaviest known
charged leptons with m τ = 1.78 GeV, are produced in charged current interactions
of ντ , but they are very short lived (tτ ∼ 10−13 s). Therefore they are not suitable
for detection, except for the fraction of times where the τ decays into μν̄μ ντ , which
happens in roughly 20 % of the cases. However, in large neutrino detectors such as the
IceCube, one may perhaps detect ultra-high-energy electron and τ neutrino events
by the intense cascade of light that is produced by secondary electrons, positrons and
photons. In the case of τ neutrinos, special relativity may help to produce a good
signature. If sources of PeV (1015 eV) τ neutrinos exist, the produced charged τ
lepton would have a relativistic γ factor as large as
Eν
γ∼ ∼ 106 (137)
mτ
which means, thanks to time dilation, that in the detector reference frame the τ lepton
will travel a distance γctτ ∼ 10 m. The “double bang” created by the charged current
interaction and the subsequent decay of the τ lepton, separated by 10 m, would be
the signature of PeV τ neutrinos.
Since neutrinos oscillate, very energetic τ neutrinos could be produced by mixing
with muon neutrinos created in high-energy pion decays in cosmic accelerators. This
is, e.g., the basis for the experiments OPERA [24] at Gran Sasso and MINOS [25]
at Fermilab, where a few events of produced τ leptons have in fact been reported.
In present detectors, only neutrino-induced muons moving upwards in the detec-
tors (or downwards but near the horizon) are safe tracers of neutrino interactions.
Most muons moving downwards have their origin in cosmic-ray nuclei interacting
with the Earth’s atmosphere and produce a very difficult background. At the surface
of the Earth, the flux of downward-going muons produced in the atmosphere is about
106 times larger than the flux of neutrino-induced upward-moving muons.
By going underground, the material (rock, water, ice, etc.) above the detector
attenuates the flux of atmospheric muons considerably. In addition, if it is experi-
mentally possible to select events where a muon is moving upwards the Earth itself
acts as a filter since only neutrino-induced muons can be produced upward-going
close to the detector.
4.10 AMANDA, IceCube and Direct Detection of WIMPs
Neutrinos may give clues to the dark matter problem in another way than just being
a small part of the dark matter due to their tiny mass. If the dark matter has a
component that is massive and weakly coupled (electrically neutral) it will be non-
relativistic at freeze-out, which is of course the WIMP paradigm of cold dark matter.
A good template for a dark matter WIMP candidate is as we mentioned the lightest
supersymmetric particle—plausibly the neutralino χ (see Sect. 5 for more details).
Neutralinos (or other WIMPs) have interactions with ordinary matter which are
equally as small as those of neutrinos. However, since they move with non-relativistic
velocity there is a chance that they become gravitationally trapped inside, for exam-
ple, the Sun or the Earth. A neutralino scattering e.g., in the Sun will lose energy and
fall further inside the solar volume, and successive scatterings in the solar medium
will soon make it lose more and more energy. In the end, neutralinos will assemble
near the centre. As they are their own antiparticles (they are Majorana fermions),
they can annihilate with each other, resulting in ordinary particles (quarks, leptons,
gauge particles).
As the annihilation rate is proportional to the scattering rate, and the interior of the
Earth is almost entirely spin-0 nuclei, constraints on the spin-independent scattering
rate from experiments described below in Sect. 8 mean that neutrinos from the center
168 L. Bergström
of the Earth are not a very promising signal for canonical WIMPs. However, as the
Sun consists to some 75 % of single protons (i.e., hydrogen nuclei) with spin-1/2,
spin-dependent scattering is important and searching for neutrinos from the Sun
stands well in competition with other experiments. We will return also to this later.
Most of the annihilation products in the Sun create no measurable effect; they are
just stopped and contribute somewhat to the energy generation. However, neutrinos
have the unique property that they can penetrate the whole Sun without being much
absorbed, at least for WIMPs less massive than a few hundred GeV. An annihilating
neutralino pair of mass m χ would thus give rise to high-energy neutrinos of energy
around m χ /3 or so (the reason that E ν = m χ is that other particles created in the
annihilation process share the energy). The signal of high-energy neutrinos (tens to
hundreds of GeV—to be compared with the ‘ordinary’ MeV solar neutrinos) from
the centre of the Sun would be an unmistakable signature of WIMP annihilation.
The detection of muons in IceCube, for instance, relies on the Cherenkov effect.
This coherent emission of light follows a characteristic angle given by the Mach
relation
1
cos θ =
βn
where β is the speed of the particle traversing the medium in units of the speed of
light and n is the index of refraction of the medium. The Cherenkov effect takes
place when
1
β> .
n
Cherenkov radiation constitutes a very small fraction of the total energy loss of a
charged particle as it crosses a medium. The superluminal condition is fulfilled only
between the UV and near-infrared region of the electromagnetic spectrum. In water
or ice, for example, where the index of refraction for UV and optical wavelengths
averages around 1.3, the Cherenkov radiation cut-off in the UV region is around
70 nm. The differential energy loss into Cherenkov photons in water or ice is a few
percent of the total differential energy loss of a charged track moving with a speed
very close to c.
4.11 Water and Ice Cherenkov Telescopes
Neutrinos can thus be detected indirectly by the Cherenkov radiation from charged
leptons and hadrons produced in neutrino interactions with matter. The extremely
large detector volumes needed to detect neutrinos from distances beyond our Sun
makes the use of any other material than water or ice very difficult.
A typical detector consists of an array of light sensors (photomultipliers, PM)
with good time resolution (∼1 ns) distributed in the medium. The pattern of the hit
PMs, and relative arrival times, are then used to fit the direction of the particle that
generated the Cherenkov radiation. The correlation between the original direction
of the neutrino and the produced charged lepton means that one may reconstruct the
direction of the incoming neutrino.
Antares is a good prototype, for a larger detector being planned with the working
name KM3NET, near Toulon in the Mediterranean. The AMANDA experiment at the
South Pole was similarly an excellent working prototype, where the disadvantages
related to the remote location of the telescope were compensated by the virtues of the
glacier ice, found to be the clearest natural solid on Earth. The Cherenkov photons
emitted along the path of a muon at some wavelengths can be selected hundreds of
metres away from the muon track.
The AMANDA detector was a great success, but was too small and has recently
been abandoned, replaced by a much larger detector, the IceCube, with 80 strings
encompassing roughly a cubic kilometer of ice. Construction was finished in 2010,
and at that time also a smaller and denser inset, the DeepCore detector, was completed.
This allows a lower detection energy threshold which is particularly beneficial for the
WIMP search. Unfortunately, despite the heroic effort to build the first large neutrino
detector in this remote location, no astrophysical neutrino source including WIMPs
has yet been detected, but it is only a year that data have been collected (for a recent
review of dark matter detection in neutrino telescopes, see [26]).
5 Supersymmetric Dark Matter
As we have mentioned several times already, one of the prime candidates for the non-
baryonic cold dark matter particle is provided by the lightest supersymmetric particle,
most likely the lightest neutralino χ. Even if it would be that supersymmetry were
not realized in nature, the neutralino is still important as a nice, calculable template
for a generic WIMP.
In most versions of the low-energy theory which results from the largely unknown
mechanism of supersymmetry breaking, there is a conserved multiplicative quantum
number, R-parity:
R = (−1)3(B−L)+2S , (138)
where B is the baryon number, L the lepton number and S the spin of the particle.
This implies that R = +1 for ordinary particles and R = −1 for supersymmetric
particles. In fact, for phenomenological reasons, this symmetry is required, as its
conservation protects the proton from decaying through the mediation of R-symmetry
breaking interactions. The R-symmetry means that supersymmetric particles can
only be created or annihilated in pairs in reactions of ordinary particles. It also means
that a single supersymmetric particle can only decay into final states containing
an odd number of supersymmetric particles. In particular, this makes the lightest
supersymmetric particle stable, since there is no kinematically allowed state with
negative R-parity which it can decay to. This is of course of utmost importance
for the dark matter problem. Also other WIMP models of dark matter needs some
170 L. Bergström
mechanism to prevent decay, and the simplest mechanism is a discrete symmetry

like the double-valued (Z 2 ) R-symmetry. (Another reason for stability could be the
quantum numbers of the particles in the theory. There are in fact models with high
spin or isospin multiplets which also have a stable particle which could act as dark
matter [27].)
Pair-produced neutralinos in the early universe which left thermal equilibrium as
the universe kept expanding should, due to their stability, have a non-zero relic abun-
dance today. If the scale of supersymmetry breaking is related to that of electroweak
breaking, the neutralino will act as a WIMP and therefore a dark matter candidate
with a relic density of the same order of magnitude as the value implied by the WMAP
measurements. This is a very elegant and economical method to solve two of the most
outstanding problems in fundamental science, dark matter and the unification of the
basic forces, if they have a common element of solution—supersymmetry.
5.1 Supersymmetric Dark Matter Particles
If R-parity is conserved, the lightest supersymmetric particle should be stable. The

most plausible candidate is the lightest neutralino χ. As we will see in Sect. 5.2 it is a
mixture of the supersymmetric partners of the photon, the Z and the two neutral C P-
even Higgs bosons present in the minimal extension of the supersymmetric standard
model (MSSM). It is electrically neutral and thus neither absorbs nor emits light, and
is stable, surviving since earliest epoch after the big bang. Its gauge couplings and
mass means that for a large range of parameters in the supersymmetric sector a relic
density is predicted in the required range to explain the observed Ωχ h 2 ∼ 0.11. Its
electroweak couplings to ordinary matter also means that its existence as dark matter
in our galaxy’s halo may be experimentally tested.
Unfortunately, very little is known about how supersymmetry is broken (for a
discussion, see [28]), and therefore any given supersymmetric model contains a large
number of unknown parameters (of the order of 100). Such a large parameter space
is virtually impossible to explore by present-day numerical methods, and therefore
simplifying assumptions are needed. Fortunately, most of the unknown parameters
such as CP violating phases influence the properties relevant for cosmology, and for
detection, very little.
Usually, when scanning the large space of a priori unknown parameters in super-
symmetry, one thus makes reasonable simplifying assumptions and accepts solutions
as cosmologically appropriate if they give a neutralino relic density in the range
0.09 < 2 <

∼ Ωχ h ∼ 0.12 (139)
Recently, there has been a number of analyses where the relic density, and other
parameters or experimental quantities known within some error bounds are allowed
to vary. By using so-called Markov Chain Monte Carlo methods (MCMC), one can
get a “global fit” of the best-fit models using statistical methods [29, 30]. Usually,
one employs what is called a Bayesian method which needs some assumption about
the prior distribution of probabilities. In the case of mass parameters one may, for
instance, choose linear or logarithmic scans. If experimental data are good enough,
it can be shown that the choice of priors is not crucial. However, so far there has been
a lack of experimental information, meaning that the predicted most likely regions in
parameter space may depend quite sensitively on priors (see, e.g., [31]). Hopefully,
the situation may soon change with new results from the LHC. A drawback of the
method of global fits is that it is very computer intensive, meaning that only very
simplified models of supersymmetry have been fully investigated so far.
Besides its interesting implications for cosmology, the motivation from parti-
cle physics for supersymmetric particles at the electroweak mass scale has become
stronger due to the apparent need for 100 GeV–10 TeV scale supersymmetry to
achieve unification of the gauge couplings in view of LEP results. (For an extensive
review of the literature on supersymmetric dark matter up to mid-1995, see Ref. [32].
More recent reviews are [33] and [34]).
A great virtue of supersymmetry at the phenomenological level is that it gives an
attractive solution to the so-called hierarchy problem, which is to understand why
the electroweak scale at a few hundred GeV is so much smaller than the Planck scale
∼1019 GeV despite the fact that there is nothing in non-supersymmetric theories to
cancel the severe quadratic divergences of loop-induced mass terms. In supersym-
metric theories, the partners of differing spin would exactly cancel those divergencies
(if supersymmetry were unbroken). Of course, supersymmetric models are not guar-
anteed to contain good dark matter candidates, but in the simplest models R-parity
is conserved and the neutralino naturally appears as a good candidate.
The MSSM
The minimal supersymmetric extension of the standard model (MSSM) is defined

by the particle content and gauge couplings required by supersymmetry and a
gauge-invariant superpotential. Thus, to each particle degree of freedom in the non-
supersymmetric Standard Model, there appears a supersymmetric partner with the
same charge, colour etc, but with the spin differing by half a unit. The only addition
to this doubling of the particle spectrum of the Standard Model concerns the Higgs
sector. It turns out that the single scalar Higgs doublet is not enough to give masses to
both the u- and d-like quarks and their superpartners (since supersymmetry forbids
using both a complex Higgs field and its complex conjugate at the same time, which
one does in the non-supersymmetric Standard Model). Thus, two complex Higgs
doublets have to be introduced. After the usual Higgs mechanism, three of these
states disappear as the longitudinal components of the weak gauge bosons leaving
five physical states: two neutral scalar Higgs particles H1 and H2 (where by con-
vention H2 is the lighter state), one neutral pseudoscalar state A, and two charged
scalars H ± .
The Z boson mass gets a contribution from the vacuum expectation values vi of
both of the doublets,
172 L. Bergström
# $ # $
H11 = v1 , H22 = v2 , (140)
with g 2 (v12 + v22 ) = 2m 2W . One assumes that vacuum expectation values (VEVs) of
all other scalar fields (in particular, squark and sleptons) vanish, as this avoids color
and/or charge breaking vacua.
The supersymmetric theory also contains the supersymmetric partners of the spin-
0 Higgs doublets. In particular, two Majorana fermion states, higgsinos, appear as the
supersymmetric partners of the electrically neutral parts of the H1 and H2 doublets.
These can mix quantum mechanically with each other and with two other neutral
Majorana states, the supersymmetric partners of the photon (the photino) and the
Z (the zino). When diagonalizing the mass matrix of these four neutral Majorana
spinor fields (neutralinos), the lightest physical state becomes an excellent candidate
for cold dark matter.
The one-loop effective potential for the Higgs fields has to be used to obtain
realistic Higgs mass estimates. The minimization conditions of the potential allow
one to trade two of the Higgs potential parameters for the Z boson mass m 2Z =
2
2 (g + g )(v1 + v2 ) (where g = e/ sin θW , g = e/ cos θW ) and the ratio of VEVs,
1 2 2 2
tan β. This ratio of VEVs

v2
tan β ≡ (141)
v1
always enters as a free parameter in the MSSM, although it seems unlikely to be

outside the range between around 1.1 and 60, with some preference for the higher
values. The third parameter can further be re-expressed in terms of the mass of one
of the physical Higgs bosons, for example m A .
5.2 Higgs and Supersymmetry
At the ATLAS and CMS experiments at the CERN Large Hadron Collider (LHC),
a discovery of the Higgs particle has not yet been claimed (by the end of 2011), as
the statistical significance is still below the wanted 5 σ (standard deviations). How-
ever, there are intriguing indications showing up at more than 3 σ at a mass value
around 125 GeV. If this would stand when more statistics is gathered in 2012, it
could means that the Standard Model of particles and fields would be completed
with a most wanted spin-0 boson, the Higgs particle. Moreover, a mass below 130
GeV is a firm prediction of supersymmetry, so it may also show the way to a whole
new phenomenology, including a very interesting dark matter candidate—the light-
est supersymmetric particle, generally thought to be the neutralino. As mentioned in
Sect. 5.1, this is a quantum mechanical mixture of the supersymmetric partner of the
photon, the neutral weak gauge boson Z and the neutral spin-1/2 partners of each
of the two Higgs doublets which are needed by supersymmetry. In supersymmetric
theories, the most likely dark matter candidate is a quantum mechanical superposi-
tion, called the neutralino χ of electrically neutral supersymmetric fermions.
Of course, if the 125 GeV Higgs also signals the presence of supersymmetry, then
a rich spectrum of particles, several of which may be in reach kinematically at the
LHC, is expected. Even if supersymmetry is not realized in nature, it will continue
to play a role as an important template for dark matter, as the neutralino is a very
attractive, calculable candidate for a generic WIMP. We will return to this later.
5.3 The Neutralino Sector
The neutralinos χ̃i0 , of which the lightest is the dark matter candidate, are linear
combination of the neutral gauge bosons B̃, W̃3 (or equivalently γ̃, Z̃ ) and of the
neutral higgsinos H̃10 , H̃20 . In this basis, their mass matrix
⎛ ⎞
M1 0 − g√v1 + g√v2
2 2
⎜ ⎟
⎜ 0 M2 gv
+ 1
√ gv
− 2
√ ⎟
⎜ ⎟
M=⎜ ⎟
2 2
⎜ g√v1 gv
⎟ (142)
⎜ − 2 + √21 0 −μ ⎟
⎝ ⎠

+ g√v2 − gv
√ 2 −μ 0
2 2
can be diagonalized to give four neutral Majorana states,
χ̃i0 = ai1 B̃ + ai2 W̃ 3 + ai3 H̃10 + ai4 H̃20 (143)
(i = 1, 2, 3, 4) the lightest of which, χ01 or simply χ, is then the candidate for the
particle making up the dark matter in the universe.
The coefficients in (143) are conveniently normalized such that for the neutralino
4
|a1 j |2 = 1. (144)
j=1
The properties of the neutralino are quite different depending on whether it consists
mainly of gaugino ( j = 1, 2) or higgsino ( j = 3, 4) components. We therefore
define a parameter, Z g , which tells the size of the gaugino fraction:
2
Zg = |a1 j |2 . (145)
j=1
174 L. Bergström
A neutralino is often said to be gaugino-like if Z g > <

∼ 0.99, higgsino-like if Z g ∼ 0.01,
and mixed otherwise.
To simplify, one often makes a diagonal ansatz for the soft supersymmetry-
breaking parameters in the sfermion sector. This allows the squark mass matrices to
be diagonalized analytically. Such an ansatz implies the absence of tree-level flavor
changing neutral currents (FCNC) in all sectors of the model. In models inspired
by low-energy supergravity with a universal scalar mass at the grand-unification
(or Planck) scale the running of the scalar masses down to the electroweak scale
generates off-diagonal terms and tree-level FCNC’s in the squark sector.
In the estimates of detection rates here, we will adhere to a purely phenomeno-
logical approach, where the simplest unification and scalar sector constraints are
assumed, and no CP violating phases outside those of the Standard Model, but no
supergravity relations are used. This reduces the number of free parameters to be
scanned over in numerical calculations to seven: tan β, M1 , μ, m A , and three para-
meters related to the sfermion sector (the exact values of the latter are usually not
very important). In fact, on can reduce the number of parameters further by choos-
ing, e.g., explicit supergravity models, but this only corresponds to a restriction to
a subspace of our larger scan of parameter space. In fact, data from the LHC have
already excluded large sectors of the simplified models.
The non-minimal character of the Higgs sector may well be the first experimental
hint at accelerators of supersymmetry. At tree level, the H20 mass is smaller than m Z ,
but radiative (loop) corrections are important and shift this bound by a considerable
amount. However, even after allowing for such radiative corrections it can hardly
be larger than around 130 GeV. When there were some weak indications of a Higgs
signature at 140 GeV in LHC data reported in mid-2011, this looked like bad news
for the MSSM. However, with further data, the preferred mass is now around 125
GeV, which is easily accommodated.
5.4 Experimental Limits
The successful operation of the CERN accelerator LHC at centre of mass energies
above 7 TeV without observing any supersymmetric particles, in particular squarks
of gluinos, puts important constraints on the parameters of the MSSM. However, it
may be that the mass scale of neutralinos is decoupled from the other supersymmetric
particle masses (e.g, in “split susy” models [35]).
It has proven to be very difficult, however, to put very tight lower limits on the
mass of the lightest neutralino, because of the multitude of couplings and decay
modes of the next-to-lightest supersymmetric particle. The lightest neutralino can in
general only be detected indirectly in accelerator experiments through the missing
energy and momentum it would carry away from the interaction region.
The upper limit of dark matter neutralino masses in the MSSM is of the order of
7 TeV [11]. Above that mass, which is still far from the unitarity bound of 340 TeV
[36], the relic density becomes larger than the allowed WMAP upper limit. To get
values for the lightest neutralino mass larger than a few hundred GeV, however, some
degree of “finetuning” is necessary. (On the other hand, we have seen that for the
other important unknown part of the energy density of the universe, the cosmological
constant Λ, a “finetuning” of many orders of magnitude seems also to be necessary.)
By making additional well-motivated but not mandatory restrictions on the para-
meter space, such as in supergravity-inspired models or in simplified constrained
MSSM models (CMSSM), one gets in general masses below 600 GeV [37, 38] for
the lightest neutralino, but as mentioned these models are feeling some tension from
early LHC data.
5.5 Supersymmetry Breaking
Supersymmetry is a mathematically beautiful theory, and would give rise to a very

predictive scenario, if it were not broken in an unknown way which unfortunately
introduces a large number of unknown parameters.
Breaking of supersymmetry has to be present since no supersymmetric particle has
as yet been detected, and unbroken supersymmetry requires particles and sparticles
to have the same mass. This breaking can be achieved in the MSSM by a soft potential
which does not reintroduce large radiative mass shifts (and which indicates that the
lightest supersymmetric particles should perhaps not be too much heavier than the
250 GeV electroweak breaking scale). The origin of the effective low-energy potential
need not be specified, but it is natural to believe that it is induced through explicit
breaking in a hidden sector of the theory at a high mass scale. The supersymmetry
breaking terms could then be transmitted to the visible sector through gravitational
interactions.
Another possibility is that supersymmetry breaking is achieved through gauge
interactions at relatively low energy in the hidden sector. This is then transferred
to the visible sector through some messenger fields which transform non-trivially
under the Standard Model gauge group. However, we shall assume the “canonical”
scenario in most of the following.
Since one of the virtues of supersymmetry is that it may establish grand unification
of the gauge interactions at a common mass scale, a simplifying assumption is often
used for the gaugino mass parameters,
5
M1 = tan2 θW M2 ∼
− 0.5M2 , (146)
3
and αem
M2 = M3 ∼
− 0.3M3 , (147)
sin θW αs
2
where θW is the weak mixing angle, sin2 θW ≈ 0.22.

176 L. Bergström
When using the minimal supersymmetric standard model in calculations of relic

dark matter density, one should make sure that all accelerator constraints on super-
symmetric particles and couplings are imposed. In addition to the significant restric-
tions on parameters given by LEP and LHC, the measurement of the b → sγ
quark flavour changing process is providing important bounds, since supersymmetric
virtual particles may contribute significantly to this loop-induced decay. There are
also constraints arising if one wants to attribute the slightly abnormal value of (g −2)
for the muon [39] to supersymmetric contributions from virtual particles. The relic
density calculation in the MSSM for a given set of parameters is nowadays accurate
to a few percent or so [2].
5.6 Other Supersymmetric Candidates
Although the neutralino is considered by most workers in the field to be the preferred
supersymmetric dark matter candidate, we mention briefly here also some other
options.
If the axion, the spin-0 pseudoscalar field which solves the strong CP problem
exists, and if the underlying theory is supersymmetric, there should also exist a spin-
1/2 partner, the axino. If this is the lightest supersymmetric particle and is in the
multi-GeV mass range, it could compose the cold dark matter of the universe (for a
review, see [40]).
A completely different type of supersymmetric dark matter candidate is provided
by so-called Q-balls [41], non-topological solitons predicted to be present in many
versions of the theory. These are produced in a non-thermal way and may have a large
lepton or baryon number. They could produce unusual ionization signals in neutrino
telescopes, for example. However, the unknown properties of their precise formation
mechanism means that their relic density may be far below the level of observability,
and a value around the observationally favoured Ω Q ∼ 0.22 may seem fortuitous
(for a recent review of the physics of Q-balls, see [42]).
Of course, the possibility of dark matter being non-supersymmetric WIMPs still
remains. However, the interaction cross sections should then be quite similar as for
supersymmetric particles. Since, the rates in the MSSM are completely calculable
once the supersymmetry parameters are fixed, these particles, in particular neutrali-
nos, serve as important templates for reasonable dark matter candidates when it
comes to designing experiments with the purpose of detecting dark matter WIMPs.
6 Detection Methods for Neutralino Dark Matter
The ideal situation would appear if supersymmetry were discovered at accelerators,

so that direct measurements of the mass of the lightest supersymmetric particle,
its couplings and other properties could be performed. This would give a way to
check from very basic principle if this particle is a good dark matter candidate—
if it is electrically neutral and has the appropriate mass and couplings to give the
required relic density to provide Ωχ h 2 ∼ 0.11. So far, no signal of supersymmetry
has been found at either LEP, Fermilab, or LHC. An indirect piece of evidence for
supersymmetry would be the discovery of a Higgs particle below around 130 GeV,
since this is the maximal value of the lightest Higgs mass after radiative corrections,
in the MSSM. In the non-supersymmetric Standard Model the Higgs could be much
heavier. It is indeed encouraging that the first signs of the Higgs at LHC seems to
correspond to a mass of 125 GeV.
If we assume a local neutralino halo density of ρχ = ρ ∼ 0.4 GeV/cm3 [43],
and a typical galactic velocity of neutralinos of v/c ∼ 10−3 , the flux of particles
of mass 100 GeV at the location of a detector at the Earth is roughly 109 m−2 s−1 .
Although this may seem as a high flux, the interaction rate has to be quite small,
since the correct magnitude of Ωχ h 2 ∼ 0.11 is only achieved if the annihilation
cross section, and therefore by expected crossing symmetry also the scattering cross
section, is of weak interaction strength.
The rate for direct detection of galactic neutralinos, integrated over deposited
energy assuming no energy threshold, is
R= Ni n χ σiχ v, (148)

i
where Ni is the number of nuclei of species i in the detector, n χ is the local galactic
neutralino number density, σiχ is the neutralino-nucleus elastic cross section, and
the angular brackets denote an average over v, the neutralino speed relative to the
detector.
The most important non-vanishing contributions for neutralino-nucleon scattering
are the scalar-scalar coupling giving a spin-independent (SI) effective interaction, and
the spin-dependent (SD) axial-axial interaction,

Leff = f S I (χ̄χ) N̄ N + f S D χ̄γ μ γ 5 χ N̄ γμ γ 5 N . (149)
Usually, it is the spin-independent interaction that gives the most important contribu-
tion in realistic target materials (such as Na, Cs, Ge, I, or Xe), due to the enhancement
caused by the coherence of all nucleons in the target nucleus.
The neutralino-nucleus elastic cross section can be written as
2 v2
4m iχ
1
σiχ = dq 2 G iχ
2
(q 2 ), (150)
4πv 2 0
where m iχ is the neutralino-nucleus reduced mass, q is the momentum transfer and

G iχ (q 2 ) is the effective neutralino-nucleus vertex. One may write
2
G iχ (q 2 ) = Ai2 FS2I (q 2 )G 2S I + 4λi2 J (J + 1)FS2D (q 2 )G 2S D , (151)
178 L. Bergström
which shows the coherent enhancement factor Ai2 for the spin-independent cross
section. A reasonable approximation for the gaussian scalar and axial nuclear form
factors is
FS I (q 2 ) = FS D (q 2 ) = exp(−q 2 Ri2 /62 ), (152)
1/3
Ri = (0.3 + 0.89Ai ) fm, (153)
which gives good approximation to the integrated detection rate [44] (but is less
accurate for the differential rate [45]). Here λi is related to the average spin of the
nucleons making up the nucleus. For the relation between G S I , G S D and f S I , f S D
as well as a discussion of the several Feynman diagrams which contribute to these
couplings, see e.g. [46–48]. One should be aware that these expressions are at best
approximate. A more sophisticated treatment (see discussion and references in [32])
would, however, plausibly change the values by much less than the spread due to the
unknown supersymmetric parameters.
For a target consisting of Ni nuclei the differential scattering rate per unit time
and unit recoil energy E R is given by
dR ρχ dσiχ
S0 (E R ) = = Ni d 3 v f (v) v (v, E R ). (154)
d ER mχ d ER
The nuclear recoil energy E R is given by
2 v 2 (1 − cos θ ∗ )
m iχ
ER = (155)
mi
where θ∗ is the scattering angle in the center of mass frame. The range and slope of
the recoil energy spectrum is essentially given by non-relativistic kinematics. For a
low-mass χ, the spectrum is steeply falling with E R ; interaction with a high-mass χ
gives a flatter spectrum with higher cutoff in E R .
The total predicted rate integrated over recoil energy above a given (generally
detector-dependent) threshold can be compared with upper limits coming from var-
ious direct detection experiments. In this way, limits on the χ-nucleon cross section
can be obtained as a function of the mass m χ [49–52]. The cross section on neu-
trons is usually very similar to that on protons, so in general only the latter is dis-
played below. Major steps forward have been taken in recent years. For example, the
CDMS-II experiment [53] and XENON100 [54] have been pushing the limits down
by a large factor, reaching now 10−44 cm2 for masses around 50 GeV. This together
with a larger detector mass (for XENON, 1 tonne is presently being installed) and
other improvements will enable a thorough search well beyond the present range of
WIMP-nucleon cross sections. In Europe there are several other ambitious endeav-
ours underway, such as DARWIN, a large liquid noble gas detector, and EURECA,
a solid state detector.
The rate in (154) is strongly dependent on the velocity v of the neutralino with
respect to the target nucleus. Therefore an annual modulation of the counting rate is
in principle possible, due to the motion of the Earth around the Sun [55]. One can
thus write
S(E R , t) = S0 (E R ) + Sm (E R ) cos [ω(t − t0 )] , (156)
where ω = 2π/365 days−1 . Starting to count time in days from January 1st, the phase
is t0 = 153 days since the maximal signal occurs when the direction of motion of
the Earth around the Sun and the Sun around the galactic center coincide maximally,
which happens on June 2nd every year [55]. Similarly, the counting rate is expected
to be the lowest December 2nd every year. Here S0 (E R ) is the average differential
scattering rate in Eq. (154) and Sm (E R ) is the modulation amplitude of the rate. The
relative size of Sm (E R ) and S0 (E R ) depends on the target and neutralino mass as
well as on E R . Typically Sm (E R ) is of the order of a few percent of S0 (E R ), but
may approach 10 % for small m χ (below, say, 50 GeV) and small E R (below some
10 keV).
Since the basic couplings in the MSSM are between neutralinos and quarks,
there are uncertainties related to the hadronic physics step which relates quarks and
gluons with nucleons, as well the step from nucleons to nuclei. These uncertainties
are substantial, and can plague all estimates of scattering rates by at least a factor
of 2, maybe even by an order of magnitude [56, 57]. The largest rates, which as
first shown in [46] could be already ruled out by contemporary experiments, are
generally obtained for mixed neutralinos, i.e. with Z g neither very near 0 nor very
near 1, and for relatively light Higgs masses (since Higgs bosons mediate a scalar,
spin-independent exchange interaction).
The experimental situation is becoming interesting as several direct detection
experiments after many years of continuing sophistication are starting to probe inter-
esting parts of the parameter space of the MSSM, given reasonable, central values
of the astrophysical and nuclear physics parameters. Perhaps most striking is the 8 σ
evidence for an annual modulation effect claimed to be seen in the NaI experiment
DAMA/LIBRA [58] (see Sect. 8 where present data are summarized).
Many of the present day detectors are severely hampered by a large background of
various types of ambient radioactivity or cosmic-ray induced activity (neutrons are
a particularly severe problem since they may produce recoils which are very similar
to the expected signal). A great improvement in sensitivity would be acquired if one
could use directional information about the recoils. There are some very interesting
developments also along this line, but a full-scale detector is yet to be built.
Direction-sensitive detectors would have an even bigger advantage over pure
counting experiments if the dark matter velocity distribution is less trivial than the
commonly assumed maxwellian.
180 L. Bergström
6.1 Indirect Searches
Besides these possibilities of direct detection of supersymmetric dark matter (with

even present indications of the existence of a signal [58]), one also has the possibility
of indirect detection through neutralino annihilation in the galactic halo. This is
becoming a promising method thanks to very powerful new detectors for cosmic
gamma rays and neutrinos planned and under construction. Also, with time more has
become known about the distribution of dark matter thanks to very ambitious N-body
simulations [59–61], and a large amount of substructure has been found. This would
enhance indirect detection, as it is proportional to the line-of-sight integral of the
square of the number density.
There has been a balloon-borne detection experiment [62], with increased sensi-
tivity to eventual positrons from neutralino annihilation, where an excess of positrons
over that expected from ordinary sources was found. However, due to the rather poor
quality of the data, it was not very conclusive.
In 2008, however, this changed completely when the data on the positron to
electron ratio, rising with energy, from the satellite PAMELA was presented [63].
Somewhat later, FERMI-LAT reported a rise above the expectation from secondary
production (by cosmic rays) also for the sum of positrons and electrons [64].
An unexpectedly high ratio of positrons over electrons was measured by PAMELA,
in particular in the region between 10 and 100 GeV, where previously only weak indi-
cations of an excess had been seen [65–69]. This new precision measurement of the
cosmic ray positron flux, which definitely disagrees with a standard background
[70] has opened up a whole new field of speculations about the possible cause of
this positron excess. Simultaneously, other data from PAMELA indicate that the
antiproton flux is in agreement with standard expectations [71].
There are a variety of astrophysical models proposed for the needed extra primary
component of positrons, mainly based on having nearby pulsars as a source [72, 73].
Although pulsars with the required properties like distance, age, and energy output
are known to exist, it turns out not to be trivial to fit both FERMI and PAMELA data
with these models (see, for example, [74, 75]). For this and other reasons, the dark
matter interpretation, which already had been applied to the much more uncertain
HEAT data [62] was one of the leading hypotheses.
It was clear from the outset that to fit the PAMELA positron data and FERMI’s
sum of positrons and electrons with a dark matter model a high mass is needed (on
the order of 600 GeV to several TeV). However, since the local average dark matter
density is well-known to be around 0.4 GeV/cm3 [43], the number density decreases
as 1/M X and therefore the annihilation rate goes as 1/M X2 with M X the mass of
the annihilating particle. This means that with σv = 3 · 10−26 cm3 /s, which is the
standard value of the annihilation rate in the halo for thermally produced WIMPs
(see Eq. (69)), the rate of positrons, even for a contrived model which annihilates to
e+ e− with unit branching ratio is much too small to explain the measured result.
To a good approximation, the local electron plus positron flux for such a model
is given by, assuming an energy loss of 10−16 E 2 GeVs−1 (with E in GeV) from
inverse Compton and synchrotron radiation,

2
dφ 1 TeV
E3 = 6 · 10−4 E θ(M X − E)Btot m−2 s−1 sr −1 GeV2 , (157)
dE MX
which means that the boost factor Btot must be of about 200 for a 600 GeV
particle, that may otherwise explain the positron excess. Similar boost factors
seem to be generic, also for supersymmetric models giving e+ e− through internal
bremsstrahlung [76].
Such a boost factor can in principle be given by a large inhomogeneity in the
DM distribution which has to be very local, since positrons and electrons of sev-
eral hundred GeV do not diffuse very far before losing essentially all their energy.
Although not excluded, this would seem to be extremely unlikely in most struc-
ture formation scenarios. Therefore, most models rely on the Sommerfeld enhance-
ment factor (see Sect. 6.3). This means a non-negligible amount of fine-tuning of
the mass spectrum, in particular also for the degeneracy between the lightest and
next-to-lightest particle in the new sector. For a detailed discussion of the required
model-building, see [77]. Similar fine-tuning is needed for the decaying dark matter
scenario, where the decay rate has to be precisely tuned to give the measured flux.
Since the antiproton ratio seems to be normal according to the PAMELA measure-
ments [71], the final states should be mainly leptons (with perhaps intermediate light
new particles decaying into leptons). For an interesting such model, which may in
fact contain an almost standard axion, see [78].
It seems that at present it is possible to construct models of the Sommerfeld
enhanced type [79] which do marginally not contradict present data. However, con-
straints are getting severe and the dark matter solution to the positron excess is
currently not as fashionable as a couple of years ago. It will be interesting, however,
to see the first results from the AMS-02 experiment [80] on the International Space
Station, which should appear in the summer of 2012.
A very rare process in proton-proton collisions, antideuteron production, may
be less rare in neutralino annihilation [81]. However, the fluxes are so small that
the possibility of detection seems marginal even in the AMS-02 experiment, and
probably a dedicated space probe has to be employed [82].
6.2 Indirect Detection by γ-Rays from the Halo
With the problem of a lack of clear signature of positrons and antiprotons, one would
expect that the situation of gamma rays and neutrinos is similar, if they only arise
from secondary decays in the annihilation process. For instance, the gamma ray
spectrum arising from the fragmentation of fermion and gauge boson final states
is quite featureless and gives the bulk of the gamma rays at low energy where the
cosmic gamma ray background is severe. However, an advantage is the directional
182 L. Bergström
information that photons carry in contrast to charged particles which random walk
through the magnetic fields of the Galaxy [83].
For annihilation into quark-antiquark pairs, or W and Z bosons, the continuous
energy spectrum one gets after fragmentation into SM particles can rather well and
conveniently be parametrized as
d Ncont (E γ )/d E γ = (0.42/m χ )e−8x /(x 1.5 + 0.00014), (158)
where m χ is the WIMP mass and x = E γ /m χ . For more detailed spectra, one may
for instance use standard particle physics codes like PYTHIA [84] (as is done in [2]).
One should note that for τ leptons in the final state (158) is not a good approximation,
as this gives a harder spectrum.
Gamma-Ray Lines
An early idea was to look for a spectral feature, a line, in the radiative annihilation
process to a charm-anticharm bound state χχ → (c̄c)bound + γ [85]. However, as the
experimental lower bound on the lightest neutralino became higher it was shown that
form factor suppression rapidly makes this process unfeasible [86]. The surprising
discovery was made that the loop-induced annihilations χχ → γγ [86–88] and
χχ → Z γ [89–91] do not suffer from any form factor suppression.
The rates of these processes are difficult to estimate because of uncertainties in
the supersymmetric parameters, cross sections and halo density profile. However, in
contrast to the other proposed detection methods they have the virtue of giving very
distinct, “smoking gun” signals of monoenergetic photons with energy E γ = m χ (for
χχ → γγ) or E γ = m χ (1−m 2Z /4m 2χ ) (for χχ → Z γ) emanating from annihilations
in the halo.
The detection probability of a gamma-ray signal, either continuous or line, will of
course depend sensitively on the density profile of the dark matter halo. To illustrate
this point, let us consider the characteristic angular dependence of the γ-ray line
intensity from neutralino annihilation χχ → γγ in the galactic halo. Annihilation
of neutralinos in an isothermal halo with core radius a leads to a γ-ray flux along the
line-of-sight (l.o.s.) direction n̂ of
dF σγγ v ρχ 2
n̂ (0.94 × 10−13 cm−2 s−1 sr −1 )
dΩ 10−29 cm−3 s−1 0.3GeV cm−3
2
100GeV R
× J (n̂) (159)
mχ 8.5kpc
where σγγ v is the annihilation rate, ρχ is the local neutralino halo density and R is
the distance to the galactic center. The integral J (n̂) is given by
1
J (n̂) = ρ2 ()d(n̂), (160)
Rρ2χ l.o.s.
and is evidently very sensitive to local density variations along the line-of-sight path
of integration. In the case of a smooth halo, its value ranges from a few at high
galactic latitudes to several thousand for a small angle average towards the galactic
center in the Navarro, Frenck and White (NFW) [105] model [92].
Since the neutralino velocities in the halo are of the order of 10−3 of the velocity
of light, the annihilation can be considered to be at rest. The resulting gamma ray
spectrum is a line at E γ = m χ of relative linewidth 10−3 (coming from the Doppler
effect caused by the motion of the WIMP) which in favourable cases will stand out
against background.
Detection of a γ-rate line signal would need a detector with very good energy
resolution, like 1 % or better. This may be achieved by FERMI (although only upper
limits have been presented so far [93]). However, the Russian satellite GAMMA-400
[94] seems to have very promising characteristics for this type of dark matter search,
when it is launched by the end of this decade. This could be a very interesting new
instrument in the search for γ-ray lines from annihilation (or decay) of dark matter.
The calculation of the χχ → γγ cross section is technically quite involved with a
large number of loop diagrams contributing. A full calculation in the MSSM was per-
formed in [95, 96]. Since the different contributions all have to be added coherently,
there may be cancellations or enhancements, depending on the supersymmetric para-
meters. The process χχ → Z γ is treated analogously and has a similar rate [89–91].
An important contribution, especially for neutralinos that contain a fair fraction of
a higgsino component, is from virtual W + W − intermediate states. This is true both
for the γγ and Z γ final state for very massive neutralinos [89–91]. In fact, thanks to
the effects of coannihilations [11], neutralinos as heavy as several TeV are allowed
without giving a too large Ω. These extremely heavy dark matter candidates (which,
however, would require quite a degree of finetuning in most supersymmetric models)
are predominantly higgsinos and have a remarkably large branching ratio into the
loop-induced γγ and Z γ final states (the sum of these can be as large as 30 %). If
there would exist such heavy, stable neutralinos, the gamma ray line annihilation
process may be the only one which could reveal their existence in the foreseeable
future (since not even LHC would be sensitive to supersymmetry if the lightest
supersymmetric particle weighs several TeV). In fact the high branching ratio for
higgsino annihilation to 2γ was the reason that Hisano et al. [97] took a closer look
at the process and discovered the effect of Sommerfeld enhancement.
Internal Bremsstrahlung
The γγ process appears in a closed loop meaning that it is suppressed by powers of the
electromagnetic coupling constant. An amusing effect appears, however, for Majo-
rana fermions at even lower order. It was early realized that there could be important
spectral features [98], and recently it has been shown that internal bremsstrahlung
184 L. Bergström
(IB) from produced charged particles in the annihilations could yield a detectable
“bump” near the highest energy for heavy gauginos or Higgsinos annihilating into
W boson pairs, such as expected in split supersymmetry models [99]. In [100], it was
furthermore pointed out that IB often can be estimated by simple, universal formulas
and often gives rise to a very prominent step in the spectrum at photon energies
of E γ = m χ (such as in lightest Kaluza-Klein particle (LKP) models [101] (see
Sect. 7.8)). The IB process was thoroughly treated in [102], and here we summarize
the main results.
In [98] it was shown that the radiative process χ0 χ0 → f f¯γ may circumvent
the chiral suppression, i.e., the annihilation rate being proportional to m 2f . This is
normally what one would get for annihilation into a fermion pair from an S-wave
initial state [103], as is the case in lowest order for non-relativistic dark matter
Majorana particles in the Galactic halo (see also [104]). Since this enhancement
mechanism is most prominent in cases where the neutralino is close to degenerate
with charged sleptons, it is of special importance in the so-called stau coannihilation
region.
A fermion final state containing an additional photon, f f¯γ, is thus not subject
to a helicity suppression. The full analytical expressions are lengthy, but simplify in
the limit of m f → 0. Then one finds
% &
dN f + f − 4x 2x μ(μ − 2x) μ
=λ× − − log , (161)
dx μ(μ − 2x) (μ − x)2 (μ − x)3 μ − 2x
with
|g̃ R |4 + |g̃ L |4 2 −1
λ = (1 − x)αem Q 2f m χ σvχχ→ f f¯ .
64π 2
where μ ≡ m 2˜ /m 2χ + 1 = m 2˜ /m 2χ + 1 and g̃ R PL (g̃ L PR ) is the coupling between

fR fL
neutralino, fermion and right-handed (left-handed) sfermion. This confirmed the
result found in [98] for photino annihilation. Note the large enhancement factor
m 2χ /m 2f due to the lifted helicity suppression (from σvχχ→ f f¯ ∝ m 2f m −4 χ ), and
another large enhancement that appears at high photon energies for sfermions degen-
erate with the neutralino.
Internal bremsstrahlung from the various possible final states of neutralino anni-
hilations is included in DarkSUSY [2]. The total γ-ray spectrum is given by
γ,sec γ,IB γ,line !

d N γ,tot dNf dNf dNf
= Bf + + , (162)
dx dx dx dx
f
where B f denotes the branching ratio into the annihilation channel f . The last term
in the above equation gives the contribution from the direct annihilation into photons,
γγ or Z γ, which result in a sharp line feature [89–91, 95, 96]. The first term is the
contribution from secondary photons from the fragmentation of the fermion pair.
0.1
0.01
0.001
dN/dE
0.0001
1e-05
1e-06
0 0.2 0.4 0.6 0.8

x
Fig. 6 The distribution of γ-rays from the internal bremsstrahlung process χ0 χ0 → f f¯γ is shown
as the solid line, and compared to the standard case, (158) (dashed line). As can be seen, the internal
bremsstrahlung process gives a very hard spectrum, which may counteract the fact that radiation of
a photon always is suppressed by a factor of ∼αem /π
This “standard” part of the total γ-ray yield from dark matter annihilations shows a
feature-less spectrum with a rather soft cutoff at E γ = m χ .
In Fig. 6 an example of the energy distribution of photons given by (161) is shown.
Density Profile and γ-Ray Detection
To compute J (n̂) in (160), a model of the dark matter halo has to be chosen. The
universal halo profile found in simulations by Navarro, Frenk and White (NFW)
[105] has a rather significant enhancement ∝ 1/r near the halo centre,
c
ρNFW = , (163)
r (a + r )
where c is a concentration parameter and a a typical length scale for the halo. In fact,
more detailed later simulations have given a slightly different shape, the so-called
Einasto profile, r α
ρEinasto = ρs e− α ( α ) −1 ,
2
(164)
with α ∼ 0.17 for the Milky Way. Except near r = 0, this profile is actually quite
similar to the NFW profile, and it has slightly higher density outside the very center.
The local dark matter density near the solar system can be quite well determined [43]
and is ρ0 0.4 GeV/cm3 . If these forms of the density can be applied to the Milky
Way, this would lead to a much enhanced annihilation rate towards the galactic centre,
186 L. Bergström
10 −10
10 −12
10 −14
dN/dE [arb. units]
10 −16
10 −18
10 −20
10 −22
10 0 10 1 10 2 10 3 10 4
E (GeV)
Fig. 7 The energy distribution of γ-rays from WIMP dark matter annihilation into a bb̄ pair, for a
dark matter particle mass of 100, 200, 500, 1000, 1500, and 3000 GeV, respectively. One can see
that the bulk of the signal is at low energies. (Here the line signals from γγ and Z γ have not been
included)
and also to a very characteristic angular dependence of the line signal. This would
be very beneficial when discriminating against the extragalactic γ ray background,
and Imaging Air Cherenkov Telescope Arrays (IACTAs) are be eminently suited to
look for these signals since they have an angular acceptance which is well matched
to the angular size of the Galactic central region where a cusp is likely to be. Both
H.E.S.S. [106], MAGIC [107] and Whipple [108] have searched for a signal at the
galactic center or in other dark matter concentrations, but are still a couple of orders
of magnitude above the flux which would correspond to the canonical WIMP flux,
Eq. (69). Maybe with the planned CTA project [109] one may get to the interesting
region of parameter space for supersymmetric or other WIMPs.
Also the energy threshold of present-day IACTAs is too high (of the order of
50 GeV or higher) to be very useful for WIMPs of 100 GeV or lighter. There have
been discussions about a high-altitude detector with lower threshold, perhaps as low
as 5 GeV [110], which would be very beneficial for dark matter detection, see Fig. 7.
Space-borne gamma-ray detectors, like the FERMI satellite have a much smaller
area (on the order of 1 m2 instead of 104 −105 m2 for IACTAs), but a correspondingly
larger angular acceptance so that the integrated sensitivity is in fact similar. This is
at least true if the Galactic center does not have a very large dark matter density
enhancement which would favour IACTAs. The total rate expected in FERMI can
be computed with much less uncertainty because of the angular integration [111].
Directional information is obtained and can be used to discriminate against the diffuse
extragalactic background. A line signal can be searched for with high precision, since
the energy resolution of FERMI is at the few percent level.
Indirect Detection Through Neutrinos
The density of neutralinos in the halo is not large enough to give a measurable flux
of secondary neutrinos, unless the dark matter halo is very clumpy [112–114]. In
particular, the central Galactic black hole may have interacted with the dissipationless
dark matter of the halo so that a spike of very high dark matter density may exist
right at the Galactic centre [115]. However, the existence of these different forms
of density enhancements are very uncertain and depend extremely sensitively on
presently completely unknown aspects of the formation history of the Milky Way.
More model-independent predictions (where essentially only the relatively well-
determined local halo dark matter density is of importance) can be made for neutrinos
from the centre of the Sun or Earth, where neutralinos may have been gravitationally
trapped and therefore their density enhanced. As they annihilate, many of the possible
final states (in particular, τ + τ − lepton pairs, heavy quark–antiquark pairs and, if
kinematically allowed, W ± H ∓ , Z 0 Hi0 , W + W − or Z 0 Z 0 pairs) give after decays
and perhaps hadronization energetic neutrinos which will propagate out from the
interior of the Sun or Earth. (For neutrinos from the Sun, energy loss of the hadrons
in the solar medium and the energy loss of neutrinos have to be considered [116,
117]). In particular, the muon neutrinos are useful for indirect detection of neutralino
annihilation processes, since muons have a quite long range in a suitable detector
medium like ice or water. Therefore they can be detected through their Cherenkov
radiation after having been produced at or near the detector, through the action of a
charged current weak interaction νμ + A → μ + X .
Detection of neutralino annihilation into neutrinos is one of the most promising
indirect detection methods, and will be subject to extensive experimental investi-
gations in view of the new neutrino telescopes (IceCube, ANTARES, KM3NET)
planned or under construction [118]. The advantage shared with gamma rays is that
neutrinos keep their original direction. A high-energy neutrino signal in the direc-
tion of the centre of the Sun or Earth is therefore an excellent experimental signature
which may stand up against the background of neutrinos generated by cosmic-ray
interactions in the Earth’s atmosphere.
The differential neutrino flux from neutralino annihilation is
f
d Nν ΓA d Nν
= Bχf (165)
d Eν 4π D 2 d Eν
f
where Γ A is the annihilation rate, D is the distance of the detector from the source
(the central region of the Earth or the Sun), f is the neutralino pair annihilation
f f
final states, and Bχ are the branching ratios into the final state f . d Nν /d E ν are the
energy distributions of neutrinos generated by the final state f . Detailed calculations
188 L. Bergström
of these spectra can be made using Monte Carlo methods [117, 119–121]. Effects of
neutrino oscillations have also been included [122].
The neutrino-induced muon flux may be detected in a neutrino telescope by mea-
suring the muons that come from the direction of the centre of the Sun or Earth.
For a shallow detector, this usually has to be done in the case of the Sun by look-
ing (as always the case for the Earth) at upward-going muons, since there is a huge
background of downward-going muons created by cosmic-ray interactions in the
atmosphere. The flux of muons at the detector is given by
d Nμ ∞ ∞ Eν dσν (E ν , E μ ) d Nν
= NA d Eν dλ d E μ P(E μ , E μ ; λ) ,
d Eμ E μth 0 Eμ d E μ d Eν
(166)
where λ is the muon range in the medium (ice or water for the large detectors in
the ocean or at the South Pole, or rock which surrounds the smaller underground
detectors), dσν (E ν , E μ )/d E μ is the weak interaction cross section for production
of a muon of energy E μ from a parent neutrino of energy E ν , and P(E μ , E μ ; λ)
is the probability for a muon of initial energy E μ to have a final energy E μ after
passing a path-length λ inside the detector medium. E μth is the detector threshold
energy, which for “small” neutrino telescopes like Baksan, MACRO and Super-
Kamiokande is around 1 GeV. Large area neutrino telescopes in the ocean or in
Antarctic ice typically have thresholds of the order of tens of GeV, which makes
them sensitive mainly to heavy neutralinos (above 100 GeV) [123]. Convenient
approximation formulas relating the observable muon flux to the neutrino flux at a
given energy exist [124].
The integrand in (166) is weighted towards high neutrino energies, both because
the cross section σν rises approximately linearly with energy and because the average
muon energy, and therefore the range λ, also grow approximately linearly with E ν .
Therefore, final states which give a hard neutrino spectrum (such as heavy quarks,
τ leptons and W or Z bosons) are usually more important than the soft spectrum
arising from light quarks and gluons.
The rate of change of the number of neutralinos Nχ in the Sun or Earth is governed
by the equation
Ṅχ = CC − C A Nχ2 (167)
where CC is the capture rate and C A is related to the annihilation rate Γ A , Γ A =

C A Nχ2 . This has the solution

CC t
ΓA = tanh2 , (168)
2 τ
√
where the equilibration time scale τ = 1/ CC C A . In most cases for the Sun, and
in the cases of observable fluxes for the Earth, τ is much smaller than a few billion
years, and therefore equilibrium is often a good approximation ( Ṅχ = 0 in (167)).
This means that it is the capture rate which is the important quantity that determines
the neutrino flux.
The capture rate induced by scalar (spin-independent) interactions between the
neutralinos and the nuclei in the interior of the Earth or Sun is the most difficult
one to compute, since it depends sensitively on Higgs mass, form factors, and other
poorly known quantities. However, this spin-independent capture rate calculation is
the same as for direct detection. Therefore, there is a strong correlation between the
neutrino flux expected from the Earth (which is mainly composed of spin-less nuclei)
and the signal predicted in direct detection experiments [123, 125]. It seems that
even the large (kilometer-scale) neutrino telescopes planned will not be competitive
with the next generation of direct detection experiments when it comes to detecting
neutralino dark matter, searching for annihilations from the Earth. However, the
situation concerning the Sun is more favourable. Due to the low counting rates
for the spin-dependent interactions in terrestrial detectors, high-energy neutrinos
from the Sun constitute a competitive and complementary neutralino dark matter
search. Of course, even if a neutralino is found through direct detection, it will
be extremely important to confirm its identity and investigate its properties through
indirect detection. In particular, the mass can be determined with reasonable accuracy
by looking at the angular distribution of the detected muons [126, 127].
For the Sun, dominated by hydrogen, the axial (spin-dependent) cross section is
important and relatively easy to compute. A good approximation is given by [32]
!
sd
C ρχ 100 GeV σ sd 270 km/s

pχ
= ,
(1.3 · 1023 s−1 ) 0.3 GeV cm−3 mχ 10−40 cm2 v̄
(169)
where σ sd
pχ is the cross section for neutralino-proton elastic scattering via the axial-
vector interaction, v̄ is the dark-matter velocity dispersion, and ρχ is the local dark
matter density. The capture rate in the Earth is dominated by scalar interactions,
where there may be kinematic and other enhancements, in particular if the mass of
the neutralino almost matches one of the heavy elements in the Earth. For this case,
a more detailed analysis is called for, but convenient approximations are available
[32]. In fact, also for the Sun the spin-dependent contribution can be important, in
particular iron may contribute non-negligibly.
A neutrino telescope of area around 1 km2 , which is roughly the size of IceCube,
has discovery potential for a range of supersymmetric models, which cannot easily
be probed using other methods, see [123].
6.3 Antimatter Detection of Dark Matter
Antimatter does not seem to be present in large quantities in the universe, as can be
inferred from the absence of γ-ray radiation that would have been created in large
amounts if astrophysical anti-objects would annihilate on their matter counterparts
190 L. Bergström
(this would also cause deviations from the pure black-body form of the cosmic
microwave background, something which is very severely limited by WMAP data and
will be further probed by the Planck satellite). In fact, both the analysis of primordial
nucleosynthesis and the CMB, give a non-zero number around 10−10 for the baryon-
antibaryon asymmetry, which means that matter dominated over antimatter already
in the very early universe. On the other hand, dark matter annihilation in almost
all models occurs from a matter-antimatter symmetric initial state and thus equal
amounts of matter and antimatter is created. This leads to an interesting possible new
primary source of positrons and antiprotons (i.e. the stable anti-particles of protons) in
the cosmic rays of dark matter halos, including the one where the Milky Way resides.
(There is always a small amount of antimatter produced as secondary particles in
collisions with galactic gas and dust by ordinary cosmic rays, of course.) As discussed
extensively at conferences in 2009 (see, e.g., [128]) this was an extremely hot topic
then. This was due to the PAMELA and FERMI collaborations just having discovered
an anomalously high ratio of positrons over electrons up to 100 GeV [63], and sum
of positrons and electrons up to 1 TeV [64], respectively. During the last two years,
this anomaly, although possible to explain by dark matter annihilation, needs such
large boost factors (e.g., from Sommerfeld enhancement to be discussed below),
and somewhat contrived, leptophilic models, that these models are feeling severe
pressure from other detection methods, e.g, γ-rays from the central parts of the
Galaxy [129]. Alternative astrophysical explanations are on the other hand possible
with quite standard assumptions. One cannot say that the dark matter explanation is
yet completely ruled out, but it is in strong tension from other measurements.
Returning to more standard WIMP models, there have recently been improve-
ments in the computations of the annihilation rate at low velocity as is the case in
galaxies, where v/c ∼ 10−3 . An amusing effect is caused due to the suppression
of the 3 S1 for an initial state of two Majorana spinors (such as neutralinos) at zero
velocity, due to the requirement of Fermi statistics. Namely, one cannot have two
identical fermions in the same spin state. This means that annihilation only occurs
from the pseudoscalar 1 S0 state where one of the particles has spin up, the other
spin down. This causes for instance the annihilation amplitude into a light fermion-
antifermion pair, like e+ e− , to be suppressed by an explicit helicity factor of the
fermion mass (as in the limit of zero mass, the vertices are helicity-preserving, and
to cause a spin flip a mass term is needed). Direct annihilation into e+ e− was thus
thought to be very subdominant. However, it was realized [130, 131] (building on an
old idea [132]), that a spin-flip by one of the Majorana fermions caused by emitting
a photon could first of all relieve the helicity suppression of the process to a mere
α/π ordinary radiative factor. And, in addition, the spectral shape of the emitted
photon is very favorable for detection, causing a shoulder which peaks close to the
dark matter particle mass. In particular, for heavy (TeV-scale) WIMPs this could be
quite important, and using the radiative peak would help extracting the signal over
background [133]. Recently, these radiative processes have been generalized also to
emission of other gauge bosons, and have been shown to be quite important generally
[134, 135].
The Sommerfeld Effect
The possibility of an enhanced annihilation rate due to DM halo substructure has

been realized for a long time [112–114]. However, it seems hard to produce a boost
factor of the order of a few hundred to a thousand in the solar neighborhood, as
would be needed to explain the PAMELA and FERMI excesses. This is because
substructure survives in numerical simulations mostly in the outer portions of the
halo, due to tidal stripping in the inner part.
Another potentially very important effect, Sommerfeld enhancement, which may
explain the large boost had been found a few years earlier. This effect, was com-
puted for electromagnetism by Arnold Sommerfeld many years ago [136], but it was
rediscovered [77, 97] in the quantum field theory of very heavy dark matter particles
in the limit when the gauge particles, γ, Z 0 and W ± are essentially massless, or at
least have a Compton wavelength that is sufficiently large compared to the would-be
bound state caused by the attractive gauge forces. (Of course, a bound state is never
really formed due to the fast time scale of annihilation.)
In the quantum mechanical calculation of electron scattering and e+ e− annihila-
tion, Sommerfeld enhancement is caused by the distortion of the plane wave describ-
ing the relative motion of the annihilating particle pair through the near formation
of a bound state caused by photon exchange. In the so-called ladder approximation
for QED (where one sums only certain types of Feynman diagrams), one obtains
this Sommerfeld effect, and the square of the wave function at the origin in relative
coordinates r1 − r2 , which enters into the probability for the short-distance process
of annihilation, is increased by the factor [77]

πα
|Ψ (0)|2 β
S= = , (170)
|Ψ(0) (0)|2 − πα
1−e β
with α the fine-structure constant, and β the relative velocity. This can be expanded
to S Q E D = πα/β for small relative velocities. In the Milky Way halo, velocities
are typically β ∼ 10−3 , so this limit is certainly relevant. For smaller galaxies or
DM substructure, velocities (as measured by velocity dispersions) are even smaller.
Of course, there is no direct photon exchange between DM particles, since they are
electrically neutral. However, if there are charged states nearby in mass, the neutral
pair may momentarily, before annihilation, transform into a charged pair which in
turn may exchange a photon between then. These are the basic processes that have to
be summed to all orders in the ladder approximation, and which lead to Sommerfeld
enhancement (see Fig. 8).
One could of course also have a Yukawa-like particle (i.e., spinless) of mass m Y ,
mediating a weak attractive force with coupling constant αY between DM particles
of mass m χ . The small velocity limit of the enhancement then becomes
αY m χ
SY ∝ . (171)
mY
192 L. Bergström
Fig. 8 Diagrams illustrating the field-theoretical reason for the Sommerfeld enhancement. The
figure is drawn for a supersymmetric neutralino (which is the case where the effect was first found
in dark matter physics [97]), but similar diagrams apply for any dark matter candidate which first
of all is heavy compared to the exchanged particle in the t-channel (i.e. in the “ladder”), and where
there is a near degeneracy between the neutral state being the dark matter and the virtual states (in
this case charged particles, charginos). On the left is shown the lowest order contribution, which
gets very important for large masses, and which is further enhanced by the ladder diagrams of the
type shown on the right. The net result could be an “explosive annihilation”, to quote [97]
In some cases, depending on the detailed nature of the mediating particles, the
enhancement factor S can indeed be as high as several hundred to a few thou-
sand, depending on the exact parameters. The effect is generally strongly velocity-
dependent, depending on velocity as 1/β or even (near resonance) 1/β 2 but in the
Yukawa case the 1/β scaling is valid only for β > m Y /m χ . At smaller velocities
and outside resonances, the effect saturates at m Y /m χ [137].
Important bounds come from γ-rays, but also from the non-observation of energy
distortions in the cosmic microwave background. It may still be possible to (mar-
ginally) fit the PAMELA/FERMI excess, if one takes astrophysical uncertainties into
account [79].
It should be noted that the Sommerfeld effect has a solid theoretical backing and
is important, if the mass and coupling parameters are in the right range. For super-
symmetric models, however, it occurs only for very heavy neutralinos (generally
higgsinos) and the phenomenology has only been partly investigated [138].
To conclude this section on detection methods of WIMPs, we have seen that
supersymmetric particles, which are the theoretically most plausible WIMPs have
many interesting features which may make them detectable in the not too distant
future. Supersymmetry, in particular MSSM, invented already in the 1970s, and
obtained as a phenomenological manifestation of the most realistic string theories,
has since the early 1980s, when the C D M paradigm first won universal acclaim,
been the prime template for a WIMP [103, 139].
Even in the MSSM, however, there are in principle more than a hundred free
parameters, meaning that for practical reasons the templates, for instance used at
the LHC experiments, are drastically simplified versions, like constrained MSSM
(CMSSM) or the even more constrained minimal super gravity (mSUGRA), which
do not, in contrast to the full MSSM, correspond very well to more recent thinking
about supersymmetry breaking [140]. This has to be kept in mind when discussing the
impressive LHC limits. Even in still simplified versions, like the 19 to 24-parameter
“phenomenological MSSM”, pMSSM [141, 142], the bounds on particle masses
given, e.g., by fulfilling the WMAP relic density, are not very constraining at the
moment [143]. Of course, the outlook for the MSSM would be much bleaker if a
light Higgs (with mass below roughly 130 GeV) were not to be established by the
end of the 7 TeV run, in 2012.
With the freely available [144] DarkSUSY package [2], one can compute in detail
the relic density, not only for supersymmetric models, but since the package has a
modular design, one can insert any favourite model one has for WIMP-like dark
matter. Of course, DarkSUSY is mostly used for the supersymmetric case, and it
has been originally set up for a general pMSSM model, with large freedom in the
choice of parameters.
7 Particular Dark Matter Candidates
7.1 WIMP Models
The particle physics connection is particularly striking in the WIMP scenario, namely
that for typical gauge couplings and a mass at the weak interaction scale of a few
hundred GeV, consistent with the relic density computed using standard big bang
thermodynamics, as we saw in Sect. 3. This is rather well tested by the calculation
of the abundances of hydrogen and helium in the early universe, through big bang
nucleosynthesis. The calculation of these abundances turns out to be in amazingly
good agreement with the measured ones. Using the same early universe thermody-
namics and solving the Boltzmann equation for hypothetical dark matter particles of
mass m χ , we found that the annihilation rate σv needed to explain Ωχ h 2 ∼ 0.11
(as determined by WMAP), naturally appears for ordinary gauge couplings and a
mass between around 20 GeV to a few TeV—a WIMP.
Although this is not a completely convincing argument for WIMP dark matter—
it may perhaps be a coincidence—it nevertheless gives WIMP candidates a flavour
of naturalness. For non-WIMP candidates there is, on the other hand, usually a fine
tuning involved, or use of non-standard cosmology, to obtain the correct relic density.
Even limiting oneself to WIMP models for dark matter, the literature is extensive,
and among some recent developments, which cannot be discussed in this review in
any detail, can be mentioned:
194 L. Bergström
7.2 Dark Stars
Since cosmological structure in WIMP models occurs hierarchically, starting from

scales as small as 10−12 –10−6 M [145], the idea has been put forward that the earliest
dense, small structures created by dark matter may play a role in star formation and
if the dark matter particles annihilate within the stars, unusual stellar evolution may
result [146–149].
7.3 Inelastic Dark Matter
These are dark matter candidates which may be excited to a state with slightly higher
mass and therefore cause a higher than usual direct detection rate [150–155], and
also relieve the tension between the different direct detection experiments.
7.4 Dynamical Dark Matter
As it is not obvious that there is only one type of particle making up the dark matter
(neutrinos should, for example contribute up to a few percent), an extreme solution
could be to have a very large number, with different spins, masses, etc. [156].
7.5 Leptophilic Dark Matter
As we have mentioned, there was an almost explosion of suggestions of this kind of

models in 2009, when the dark matter interpretation of the anomalous positron ratio
measured by PAMELA [63] and FERMI [64] was proposed to be explained by dark
matter annihilation. Leptophilic means that these dark matter particles annihilate
mainly to leptons, for example by proceeding through axion-like particles below the
pion mass [157–163]. Although the original motivation for these models has become
somewhat weaker, the concept has established itself in the dark matter community.
7.6 Supersymmetric Models Beyond the MSSM
Of course, even though the minimal supersymmetric version of the standard model,
the MSSM, has more than 100 free parameters, models having, e.g., motivation from
new scenarios of supersymmetry breaking, are of course logically possible. These
“beyond the MSSM” or BMSSM models [164–167] may among other things give
a higher Higgs mass than the limit of 130 GeV given by minimal SUSY models. In
the summer of 2011, this was perhaps a favoured scenario, as the first indications
of the Higgs mass was around 140 GeV. However, with more data, the preferred
range (not yet significant enough by the end of 2011 to be called a discovery) is now
124–126 GeV which is more easily encompassed in the MSSM.
7.7 Asymmetric Dark Matter
This is a class of dark matter models which may also explain the baryon (or lepton)
asymmetry of the universe [168–173]. This generally only works for masses around or
below 10 GeV, and this mass range has been in focus recently due to a (possible) signal
in direct detection experiments [58, 174–176], and maybe also in γ-ray detection in
the direction near the Galactic centre [177]. However, it remains to see whether these
indications will stand the test of time. A similar model is “emergent dark matter”.
This is a recent version of asymmetric DM with larger possible parameter range,
such as a DM mass up to 100 GeV [168–173].
7.8 Kaluza-Klein Models
A candidate for dark matter, the so-called LKP (for lightest Kaluza-Klein particle)
has been identified. This appears in theories with extra dimensions, and has a rich
phenomenology which we will not enter into here (for a review, see [178]). The main
difference with supersymmetry is that the dark matter candidate has spin-1, and can
give the correct relic density for a mass in the range from 600 GeV to 1 TeV.
7.9 Inert Higgs Doublet
Interesting are also versions of the Standard Model with an enlarged Higgs sector. If
there would be, for instance, a second Higgs doublet which does not couple directly
to Standard Model particles (an “inert doublet”), there turns out to be a stable spin-0
state which then would be the dark matter particle (see [179], and references therein).
7.10 Non-WIMP Models
WIMPs are arguably the leading candidates for Dark Matter, due to lack of fine-
tuning to get correct relic density. In most models, the annihilation cross section
which sets the relic density also implies observable rates in various DM detection
experiments.
A word of caution is in place here, however. There are many non-WIMP models
that also have good particle physics motivation, and may be detectable, like: axions,
196 L. Bergström
gravitinos, superWIMPS, non-thermal dark matter, decaying dark matter, sterile Neu-
trinos, Q-balls. . . The literature is extensive, but a good summary of both WIMP and
non-WIMP models has recently appeared, namely, a 700-page book giving details
of most dark matter scenarios [180].
7.11 The Axion
Another, rather different candidate [181] for dark matter is provided by the axion, a
hypothetical light boson which was introduced for theoretical reasons to explain the
absence of C P violation in the strong interactions (as far as we know, C P violations
only take place in the weak interactions). It turns out that for a mass range between
10−6 and 10−3 eV, the axion could give a sizable contribution to Ω M . It couples
very weakly to ordinary matter, but it may be converted into a photon in a cavity
containing a strong magnetic field (the basic coupling is to two photons, but here
the magnetic field takes the role of one photon). Experiments in the USA and Japan
are currently probing parts of the interesting mass region. A section about the axion
should always be inserted when describing dark matter candidates, since the axion
has, as has the lightest supersymmetric particle, a good particle physics motivation
for its existence.
8 Dark Matter Detection: Status
As we have mentioned, there are basically three different, and complementary meth-
ods for detecting WIMPs. First, the dark matter particle may be directly produced at
accelerators, in particular at the LHC, which today is the only high-energy accelerator
running (although data from Fermilab’s Tevatron collider will still be analyzed and
may give surprises in the coming year or so). Of course, it is not clear that the particle
will be kinematically allowed, and even if it is produced, one will not know that the
lifetime is of the required cosmological order of magnitude. Anyway, detecting a
candidate and determining its mass would be a great gain when combining with the
other two search methods of dark matter, namely direct and indirect detection. In
particular, direct detection experiments have seen an impressive gain of sensitivity
during the last few years. The idea is to register rare events giving a combination of
scintillation, ionization and nuclear recoil signals in chunks of matter shielded from
cosmic rays in underground sites.
In indirect detection, one rather registers products of dark matter annihilation
from regions in the surrounding universe with a high dark matter density like the
galactic centre, dwarf spheroidal galaxies, or the interior of the Earth or the Sun. An
interesting feature of indirect detection is that the expression for the local annihilation
rate of a pair of DM particles χ (here assumed, like in supersymmetry, to be self-
charge-conjugate, of relative velocity vr el
Γann ∝ n 2χ σann (vr el )vr el (172)
is the dependence on the square of the number density. Also, the cross section may
depend in non-trivial ways on the relative velocity. In particular, for low velocities
the rate may be much higher than at high velocity, for models containing an attractive
force between the annihilating particles. This is in particular true for models with
so-called Sommerfeld enhancement [97], a resonant enhancement by in some cases
orders of magnitude (see Sect. 6.3). This means that dwarf galaxies (dark matter
subhalos) may be particularly interesting objects to study, as they are completely
dark matter dominated with low rate of cosmic-ray induced γ-rays, and their low
mass means a relatively low velocity dispersion, meaning higher possible rates if
Sommerfeld enhancement is active.
So far, indirect methods have not been as competitive as direct detection, but
recently the FERMI collaboration has started to probe the interesting WIMP region
by stacking data from several dwarf galaxies [182].
For non-WIMP dark matter, like sterile neutrinos (warm DM), the production rate
in the early universe generally has to be tuned to give the observed relic density, but
phenomenologically warm DM is possible, and according to some analyses even
preferred in cosmological data [183]. However, the significance is weak and may be
influenced by statistical bias [184]. Ordinary, active neutrinos have too small mass
to contribute significantly to the dark matter density, although in the extreme case
may contribute a couple of percent to the critical density today.
A very interesting effect for direct detection of dark matter WIMPs in terrestrial
detectors comes about due to the motion of the solar system in the Galaxy [55]. This
circular speed is around 200 km/s, and the direction of the “wind” of dark matter
particles varies in between seasons. This is due to the detector following the Earth’s
motion around the Sun and sometimes (actually around June 2) having “headwind”
of WIMPs and sometimes (December 2) “tailwind”. As the cross section between a
WIMP and the detector target depends strongly on their relative velocity, this causes
a few percent annual modulation of the detection rate, something that is a very
distinct signature. The DAMA/LIBRA experiment in the Gran Sasso tunnel [58]
has in fact seen an annual modulation, which has a statistical significance of more
than 8 standard deviations. However, since no other experiment has found the same
effect (see Table 1), the effect can still not be taken as an established detection of dark
matter. There have been attempts to interpret the DAMA signal as possibly being due
to a neutralino of the MSSM [185, 186]. It seems premature, however, to draw strong
conclusions from this experiment alone. Besides some cloudy experimental issues,
the implied scattering rate seems somewhat too high for the MSSM or any other
canonical WIMP, given the strong Higgs mass bounds from LEP and LHC unless
one stretches the astrophysical and nuclear physics quantities. Also, it is disturbing
that neither XENON100 nor CDMS-II see an effect despite their nominally higher
sensitivity. Clearly, even more sensitive experiments, preferably also using NaI, seem
to be needed to settle this issue. An interesting idea, DM-Ice [187], uses the IceCube
site to deploy crystals of NaI with ice as a very calm surrounding medium. If an
annual modulation could be measured also there one could check whether it has the
198 L. Bergström
Table 1 Some of the recent experimental claims for possible dark matter detection, and a comment
on the present status
Experiment Status of claim
DAMA/LIBRA annual modulation [58] Unexplained at the moment; not confirmed by
other experiments [54, 188]
CoGeNT excess events and annual Tension with other data [54, 188]
modulation [174, 175]
EGRET excess of GeV photons [189, Due to instrument error (?)—not confirmed by
190] FERMI [191]
INTEGRAL 511 keV γ-line from galactic Does not seem to have spherical
centre region [192] symmetry—shows an asymmetry which
follows the disk (?) [193]
PAMELA: Anomalous ratio of cosmic ray May be due to DM [194, 195], or pulsars [72,
positrons/electrons [63] 73]—energy signature not unique for DM
FERMI positrons + electrons excess [64] May be due to DM [194, 195], or pulsars [72,
73]—energy signature not unique for DM
FERMI γ-ray excess towards galactic Unexplained at the moment—astrophysical
centre [196, 197] explanations possible [198, 199], no
statement from the FERMI collaboration
WMAP radio “haze” [200] Has a correspondence in “FERMI bubbles”
[201]—probably caused by outflow from the
galactic center
same phase as that of DAMA, or if it rather follows the seasons (which are opposite
on the southern hemisphere).
There have recently been a number of claimed possible detections of dark matter,
see Table 1. Of the items in Table 1, it seems that only the positron excess at high
energy (20 GeV–1 TeV) and the γ-ray excess towards the galactic center, inferred by
an analysis of FERMI public data [196, 197], can be due to dark matter annihilation
without tension from other data. However, they may both perhaps more naturally
be explained by ordinary astrophysical processes. In addition, the DM explanation
of the PAMELA and FERMI data as we have seen needs a leptophilic particle of
TeV-scale mass and a very much boosted cross section. Although this may perhaps
be obtained, stretching all uncertainties involved [202], and employing Sommerfeld
enhancement [79], the remaining window seems quite tight.
The DAMA/LIBRA annual modulation is a statistically very strong signal (signif-
icance of the order of 8σ), however the lack of supporting data from other experiments
is disturbing. The annual modulation hinted at by CoGeNT [174, 175] is statistically
much weaker, and the purported excess unmodulated signal may in fact be incompat-
ible with the level of modulation reported. Also, it seems that the DAMA/LIBRA and
GoGeNT signals, if interpreted as being due to dark matter, may be in tension with
each other, even if one uses freedom in isospin violation, inelastic scattering, and
non-standard halo properties [203–206]. At the moment this is one of the unsolved,
frequently debated issues in the dark matter community.
The recent improvement of the upper limits on the WIMP-nucleon scattering
cross section reported by CDMS II [188] and, in particular, XENON100 [54] are
truly impressive. Not only does it cast some doubt on other reported experimental
results, the sensitivity is also good enough to start probing the parameter space of
supersymmetric models [2]. The new calibration of the sensitivity to low-energy
recoils of Xenon adds to the credibility of the new limits. The very good news is
also that the installation of the next stage, a 1 ton liquid Xenon detector, has already
started in the Gran Sasso experimental halls in Italy.
Of course, a much more decisive claim of detection of dark matter would result
if any of the other methods, like a suitable new particle candidate being detected
at the LHC, or a signature in γ-rays from the Galactic dark matter halo would be
discovered.
In the first runs at LHC, no signs of a Higgs particle, nor supersymmetry or any
other of the prime candidates for dark matter, have been discovered. On the other
hand, the mass region 115–130 GeV, interesting for the lightest Higgs boson in the
simplest versions of supersymmetry, was yet to be investigated, and in fact a weak
indication around 125 GeV seem to have been found.
One possible scenario might be that such a Higgs particle is indeed found, but the
particles carrying non-trivial R-parity all have masses beyond reach with the LHC.
This is not impossible, depending on the amount of fine-tuning one is willing to
tolerate. In fact, if one puts no prior constraints on the supersymmetric parameter
space other than one should have the WMAP-measured relic density, and fulfill all
other experimental constraints (cf. [143]), a mass for the lightest supersymmetric
neutralino in the TeV region is generic. For such heavy dark matter neutralinos, the
rate for direct detection will also be small, and it would seem impossible to test such
a scenario. However, for this particular case indirect detection through gamma rays
turns out to have an interesting advantage, as the new imaging air Cherenkov arrays
like CTA will have their peak sensitivity in the energy range between a few hundred
GeV to a few TeV [207].
Depending on the particular model realized in nature, Sommerfeld enhancement
of indirect detection may also be operative. However, these large arrays will be
served by a large astrophysical community which will be very much interested in
transient or periodic events, meaning that a “boring” search for a stationary dark
matter spectral signature during hundreds or even thousands of hours seem out of
question. One may therefore consider a dedicated particle physics experiment, the
“Dark Matter Array”, DMA [208] only used for dark matter search. This would have
great, and complementary, potential to the large direct detection experiments that are
presently being planned. In fact, we mentioned, and you heard at the lectures by F.
Aharonian (see Sect. 2.3 in his part), that there are ideas [110] on how to decrease
the lower threshold for detection, something that could increase the sensitivity for
DM detection considerably (see Fig. 7). If such a working prototype could be built,
this idea may materialize in the next decade as a new way to search for phenomena
beyond the Standard Model—with an expensive dedicated detector, still far below
the cost of a new high-energy accelerator.
Of course, LHC data has already started to exclude some regions of supersymmet-
ric parameter space, although not very much. This may be surprising, but is in fact
due to the relative independence of the squark and gluino sector of supersymmetry,
200 L. Bergström
Fig. 9 Scan of the MSSM L. Bergström, 2011

parameter space showing the
direct detection cross section 10-6
versus indirect detection
through gamma rays. The
uppermost points are excluded 10-7
Spin-independent cross section, pb

by XENON100, and points
which survive also the LHC
-8
2011 data are shown in black 10
10 -9
10-10
-11
10
-12
10
-3 -2 -1
10 10 10 1 10
2 -29 3 -1 -2
S = N(E › 1 GeV) v cont /M (10 cm s GeV )
and the neutralino sector, which hosts the dark matter candidate. In fact, as men-
tioned, there are so-called split supersymmetry models, which have this dichotomy
explicitly postulated [35].
The complementarity of direct and indirect detection is shown in Fig. 9, where
also the effects on the parameter space caused by the XENON100 bounds and LHC
2011 bounds, respectively, are shown.
9 A Detailed Calculation: The Saas-Fee WIMP
An interesting question came up during the Saas-Fee Course: Could there be a cos-
mological contribution to the γ-ray spectrum making up the deficit in the diffuse
γ-ray emission measured by FERMI? As we heard, this is not readily explained
by adding the well-known sources like AGNs, millisecond pulsars and star-forming
galaxies described by C. Dermer.
Here we will outline the simple steps in making the dark matter prediction for this
flux, based on [209] (see [210] for a much more thorough treatment). We will see how
that could lead us to predict a several hundred GeV dark matter particle—the Saas-
Fee particle as we named it at the Saas-Fee Course in 2010. This was only published
in online slides from my talk,4 and should not be taken too seriously. However, as a
4 Available on the website of the Course: http://isdc.unige.ch/sf2010/.

pedagogical example of a surprising effect of the accumulated dark matter structure

in the universe it is quite instructive.
As we have seen, in the presently most successful model for structure formation,
ΛCDM, most of the matter is in the form of non-relativistic cold dark matter (CDM),
but with a contribution to the present-day energy density also from a cosmological
constant (Λ). As shown by detailed N -body simulations (see, e.g., [59–61] and ref-
erences therein), in such a picture large structures form by the successive merging
of small substructures, with smaller objects generally being denser. N -body simu-
lations also show that the dark matter density profile in clusters of galaxies and in
single galaxies develops an enhancement of the density near the centre, although it is
at present unclear how steep this increase is, and whether it even shows a cusp near
the center like in the popular parametrization of NFW, Eq. (163), ρCDM (r ) ∼ r −α
with α close to 1 [105] (a very similar profile, the Einasto profile, Eq. (164), does
not show a singularity, but behaves rather similarly on moderate length scales).
At present, it is not clear whether these N -body predictions are in agreement
or not with all available data (one recently acknowledged problem is, for exam-
ple, the apparent lack of a halo mass threshold for dwarf galaxies [211]). On large
scales, however, the ΛCDM scenario gives excellent agreement with observations.
On smaller scales, the dynamical range of the simulations is not enough, and one of
the main puzzles is how to properly include the non-linearities induced by baryonic
matter in the form of supernova explosions and other feedback mechanisms.
Let us assume that the ΛCDM picture is basically correct and that structure forms
hierarchically, with the number density of halos of mass M being distributed as
d N /d M ∝ M −β with β ∼ 1.9 − 2, as predicted by the Press-Schechter theory [212]
and also verified in N -body simulations. Furthermore, the concentration of halos
grows in these simulations with decreasing mass.
It is interesting that the averaging involved in computing the integrated signal
of annihilation γ-rays of unresolved cosmological dark matter gives results which
are more robust to changes in the details of how the dark matter is distributed on
small scales. (The same is actually also true for all sources which are completely
encompassed by the angular resolution cone of a given γ-ray experiment, for the
obvious source of the galactic centre, the prediction of fluxes differ by up to 4 orders
of magnitude for different models: in particular they are very sensitive to the presence
or not of a central cusp.)
Let us consider annihilation of a WIMP such as the lightest neutralino χ of the
MSSM, as a template. The mass range is from around 20 GeV up to several TeV [2].
For the sake of pedagogy, let us start with the unrealistic case of all the dark matter
is smoothly distributed with the density distribution being just a number, the average
density, at all redshifts. The idea is that since the dark matter was more dense in the
early universe, one may get a large (red-shifted) flux by integrating over the line of
sight from 0 to very high redshifts. Actually, in papers preceding [209] this was the
only effect considered. We will soon modify the results by introducing the effects of
structure, which indeed increases the flux by many orders of magnitude.
202 L. Bergström
9.1 The Flux in a Smooth Universe
The comoving number density n c of WIMPS, after decoupling from chemical equi-
librium (“freeze-out”) at very large temperatures (T ∼ m χ /20) is depleted only
slightly due to self-annihilations, governed by the Boltzmann equation
ṅ c = −σv(1 + z)3 n 2c , (173)
where σv is the thermal- and angular-averaged annihilation rate, which, to an

excellent approximation after freeze-out, is velocity independent, since the neutrali-
nos move non-relativistically, and are always in a dominant S-wave component (at
least for our supersymmetric WIMP templates).
Each pair of χ particles that disappears through annihilation give rise to Nγ pho-
tons on the average, with an energy distribution in the rest frame of the annihilation
pair,
d Nγ (E) d Ncont
= (E) + bγγ δ m χ − E . (174)
dE dE
Here the first term gives the average continuum gamma ray distribution per annihi-
lating χ and we have also added a term for the possible γγ line contribution, with bγγ
being the branching ratio into γγ (one could also have an additional Z γ channel).
A γ-ray observed today, at redshift z = 0, of energy E 0 would correspond to an
energy at the emission redshift z of E = (1 + z)E 0 . We can now track, using the
Boltzmann equation, the number of WIMPs that have disappeared from redshift z
until now, and fold in the energy distribution (174). Thus we get a first estimate of the
level of the diffuse extragalactic γ-ray flux. As usual, H0 is the Hubble parameter,
and we use the relation between time and redshift (see, e.g., [1]) d/dt = −H0 (1 +
z)h(z)d/dz with

h(z) = Ω M (1 + z)3 + Ω K (1 + z)2 + ΩΛ ∼ Ω M (1 + z)3 + ΩΛ . (175)
Here Ω M , ΩΛ and Ω K = 1 − Ω M − ΩΛ are the present fractions of the critical

density given by matter, vacuum energy and curvature. We can here use the result
from Sect. 2 that the universe to an excellent approximation is flat, Ω K = 0. We then
obtain
dn c (z) σv (1 + z)2
= n c (z)2 . (176)
dz H0 h(z)
The differential energy spectrum of the number density n γ of photons generated

by WIMP annihilations is then given by
dn γ dn c mχ d Nγ (E) dn c
= Nγ = d E. (177)
dz dz 0 dE dz
Here, dn c /dz can be computed directly from (176) to excellent accuracy, replacing
the exact solution n c (z) by the present average number density of neutralinos n 0 on
the right hand side. This we can do since the comoving number density does not
change appreciably after freeze-out.
Neglecting the baryon contribution (as we will see, factors of order unity will
not make a difference here, due to much larger uncertainties in structure formation),
Ωχ ∼ Ω M , we obtain
n 0 = ρχ /m χ = ρc Ω M /m χ . (178)
Here ρc = 1.06 · 10−5 h 2 GeV/cm3 and h as before is the scaled Hubble parameter
in units of 100 km s−1 Mpc−1 , h ∼ 0.7. There are a few more effects we have to
include. We have to use the fact that all photons move with velocity c and that the
average flux is isotropic from each volume element where annihilation takes place,
giving a factor 1/4π per steradian. The cross section times velocity average should,
for Majorana particles, also be divided by 2, something which was missing in the
original derivation [209], but added in [210] (see the published version). Some of the
photons will be absorbed after travelling over cosmological distances. This can to
the level of our approximate calculation be handled by introducing a simple energy-
and redshift-dependent factor e−z/z max (or the more detailed calculation in [210] a
more complicated factor depending on z and E 0 ).
The resulting γ-ray flux at the detector is then given by:
c dn γ
φγ = = 4.2 · 10−14 cm−2 s−1 sr −1 GeV−1
8π d E 0
Γ26 Ω M2 h3 z up (1 + z)3 e−z/z max d Nγ (E 0 (1 + z))
× 2
dz . (179)
m 100 0 h(z) dE
where we defined Γ26 = σv/(10−26 cm3 s−1 ) and m 100 the mass in units of
100 GeV.
For the energies we are interested in, 1 GeV < E 0 < 500 GeV, it is the starlight
and (more poorly known) infrared background radiation which is most important,
whereas the CMBR may be important for higher energies. An optical depth of order
unity is reached for a redshift which in [210] was approximated by z max old (E ) ∼
0
3.3(E 0 /10 GeV) −0.8 , which represented older results. However, the newer data dis-
cussed in the lectures by C. Dermer (see Sect. 7.3 in his part), indicate much less
absorption. As a representative of the more recent evaluation of this absorption [213],
new (E ) ∼ 2.3(E /50 GeV)−1.1 .
we take instead the simple approximation z max 0 0
The exponential form is a good approximation for small values of z max as is dom-
inant in most of the cases we study here. The upper limit of integration is given by
kinematics, z up = m χ /E 0 − 1, as the maximum rest frame energy of a photon in an
annihilation event is E = m χ . The gamma line contribution to (179) is particularly
simple, just picking out the integrand at z + 1 = m χ /E 0 ; it has the very distinctive
and potentially observable signature of being asymmetrically smeared to lower ener-
gies (due to the redshift) and of suddenly dropping just above m χ . Unfortunately,
204 L. Bergström
for most models the branching ratio for this channel is too small to be measurable
with present-day energy resolution, and we will drop it from now on. (This may
however change when the high-resolution instrument GAMMA-400 [94] is opera-
tional towards the end of this decade. This is specified to have an energy resolution
of 1 %, which will be a perfect instrument for searching for γ lines from annihilation,
and also from models where dark matter decays radiatively [214].) The continuum
emission will produce a characteristic, although less conspicuous feature, a smooth
“bump” below around one tenth of the neutralino mass, and may be more difficult to
detect. One should notice that there are particular models where radiative corrections
(“internal bremsstrahlung”) may give a significantly harder spectrum near E γ = m χ ,
facilitating discrimination against most backgrounds [130, 131].
9.2 Including Effects of Cosmic Structure
To give an example of the results (which in [209] contained both obsolete SUSY
models and not very accurate data from the old EGRET experiment), we take a
generic model with mass 600 GeV, and the canonical WIMP averaged cross section
times velocity of σc = 3 · 10−26 cm3 s−1 , in the “concordance” cosmology
Ω M = 0.3, ΩΛ = 0.7, h = 0.7. The continuous γ-ray rest frame energy distri-
bution per annihilating particle comes mainly from hadronization and decay of π 0 s
and is conveniently parametrized to reasonable accuracy as (see Eq. (158))
d Ncont (E)/d E = (0.42/m χ )e−8x /(x 1.5 + 0.00014)
where x = E/m χ . This is valid for most quark jet final states, except for the top.
Also, τ lepton decays give a somewhat harder γ-ray spectrum, and as mentioned
internal bremsstrahlung may be important for certain types of models.
The most difficult, but also most important and interesting part of the calculation is
to include the effects of structure formation. Following [209], we consider first a halo
of mass M whose radial density profile can be described by ρ D M (r ) = ρD M f (r/a),
with ρD M being a characteristic density and a a length scale. These are found in
N -body simulations not to be independent parameters, as smaller halos are generally
associated with higher densities.
As a simple first model for structure formation, assume that the halo of mass M
accreted from a spherical volume of radius R M , determined by requiring that the
average cosmological density times that volume is equal to M, ρ0 · 4π R 3M /3 = M
(with ρ0 ∼ 1.3 · 10−6 GeV/cm3 ). The increase of average squared overdensity per
halo, which is what enters the annihilation rate, is given by:
' 2 (
ρD M ρD M I2
Δ ≡2
= , (180)
ρ0 ρ0 I1
r <R M
R /a
where In ≡ 0 M y 2 dy( f (y))n with y ≡ r/a. Here the dependence on the limits
of integration is rather weak, at least for profiles less cuspy than the NFW profile
[105] (see Eq. (163)).
Computing I2 /I1 numerically, and using values of ρD M /ρ0 as determined for
Milky Way size halos we find values of Δ2 of 1.5 · 104 for the NFW profile, and
7 · 103 for a cored, modified isothermal profile (modified so that the density falls as
1/r 3 at large radii). The flux ratio, 2:1 for these two models should be compared
with the ratios roughly 100:1 obtained within a few-degree cone encompassing the
galactic center, showing the announced relative insensitivity to halo density profiles.
We should now also take into account that the number density of halos is scaling
like ∼1/M 1.9 , and that small-mass halos are denser. We can resort to the highest-
resolution N -body simulations available to date [215]. Fitting the concentration para-
meter of halos by
c ∼ 100 (Mvir / h −1 M )−0.08 , (181)
one finds to a good approximation

−0.2
Δ2 ∼ 2 · 105 M12 , (182)
where M12 is the halo mass in units of 1012 solar masses. This means that the total
flux from a halo of mass M scales as M 0.8 . Since the number density of halos goes
as M −1.9 , the fraction of flux coming from halos of mass M scales as M −1.1 . Thus
the γ-ray flux will dominantly come from the smallest CDM halos. In simulations,
substructure has been found on all scales (being limited only by numerical resolution).
For very small dark matter clumps, however, no gain in overdensity is expected, since
once the matter power spectrum enters the k −4 region a constant density is expected.
There are arguments [216] that structure is present in cold dark matter models all
the way down to 10−6 or smaller [217]. We conservatively set 1M as the minimal
scale. In a more detailed treatment, one should also include effects of clumps within
clumps, which increase the enhancement. However, destruction of DM clumps near
large central densities of halos should also be included.
Finally, regarding redshift dependencies, we assumed in [209] a constant enhance-
ment factor Δ2 out to z ∼ 1, and somewhat arbitrarily imposed quadratic growth
in the enhancement factor from z ∼ 10 to the fully developed epoch z = 1. (The
computed flux is not very sensitive to this assumption.) Furthermore, in (179) we
make the replacement (1 + z)3 → 1, reflecting the fact that the we are now dealing
with a clumped, rather than a smooth distribution with density scaling ∼(1 + z)3 .
We thus arrive at the following expression for the flux including structure forma-
tion
c dn γ
φγ = = 4.2 · 10−14 cm−2 s−1 sr −1 GeV−1
8π d E 0
Γ26 Ω M2 h3 z up Δ2 (z)e−z/z max d Nγ (E 0 (1 + z))
× dz . (183)
m 2100 0 h(z) dE
206 L. Bergström
0.001
E dN/dE [MeV cm sr s ]
-1 -1
-2 0.0001
1e-05
2
1e-06
1e-07
100 1000 10000 1e+05
Photon Energy (MeV)
Fig. 10 The predicted diffuse extragalactic γ-ray flux computed using the methods described in
the text, for a 600 GeV WIMP with a cross section compatible with the WMAP-inferred relic
density and with different assumptions for the effects of structure. The diffuse extragalactic data
was measured by FERMI-LAT [191]
9.3 The Saas-Fee WIMP
We find using (183) (see also [218]) that the flux from small halo structure is enhanced
by roughly a factor (4 − 10) · 107 compared to the smooth case, giving in the upper
range observability for the annihilation parameters as used above. The uncertainties
reside mainly in the still poorly known factor Δ2 (z) and its extrapolation to small
halo masses (and also the effects of DM clumps within clumps, for instance).
In Fig. 10, we show the results for this 600 GeV WIMP model. The results are
compared with the measurements from FERMI-LAT [191], and despite the fact that
there is this uncertainty in the absolute rates, it is amusing, as I discussed at the
Saas-Fee course, that the possible break in the FERMI data may be caused by a
new contribution from 500–600 GeV mass annihilating dark matter (“The Saas-Fee
WIMP”, of which there would be one per 2 litres in the lecture hall at Les Diablerets
as in all other places on Earth) setting in. It will obviously be interesting to follow the
development of this data set during the next few years, to see if this model survives
or even becomes more convincing.
It has of course to be remembered that the strength of the annihilation signal can
be much lower than the proof-of-existence example chosen for Fig. 10 in which case
a detection would be correspondingly more difficult. On the other hand, there may
be particle physics effects (such as Sommerfeld enhancement) which could give a
higher flux.
As a recent illustration of the importance of adding up all structure present, e.g.,
in galaxy clusters, can be mentioned the results of [219–223] and [224], where it was
shown that by choosing particularly favourable, not too distant clusters, one is very
close to the current observational limits from FERMI-LAT. Indeed, there may even
be (weak) indications from FERMI data of a signal [225].
A related type of analysis for the diffuse extragalactic case is performed in a
similar way as when analyzing the angular fluctuations in the CMB. Also using this
method of analysis, the conclusion is that with FERMI-LAT data one may be very
near detection [226–229].
10 Primordial Black Holes as Dark Matter?
Sometimes one gets the question from lay persons, when telling that we are interested
in the problem of dark matter: Could it be black holes? Black holes are in some sense
dark: they do not emit light from within the event horizon, so the question is not
completely out of context. However, the only black holes which we are relatively
certain to exist are those around 2–20 solar masses, and the supermassive ones like
the one residing at the Galactic center (of mass a few times 106 M ). We also know
of even more massive ones (of mass up to a few times 109 M ) making up active
galactic nuclei (AGNs). However, the galactic halos in which even the most massive,
“supermassive”, black holes reside have a factor of at least 1000 more total mass.
Thus their influence on the cosmic energy balance is rather marginal. Also the solar
mass type black holes which are produced as end results of stellar evolution constitute
a small fraction, by far too small to explain the abundant and omnipresent dark matter.
Finally, most black holes are in practice not very dark, as their concentrated mass
has an effect on surrounding matter, causing various types of radiation processes, as
ordinary matter is “eaten” by the black hole.
However, very massive black holes may be important for dark matter detection:
if left undisturbed for a long time, they may adiabatically attract surrounding dark
matter, changing the NFW-type distribution to a much more spiky cusp [230]. As the
annihilation rate grows with the square of the dark matter density, this could give a
dramatically increased rate of γ-rays, and in particular neutrinos (which are unlikely
to be absorbed by surrounding matter). More extreme versions of this scenario are in
fact ruled out already, due to the lack of unambiguous dark matter signals from the
galactic centre. An interesting possibility is that intermediate mass black holes exist,
where this type of signal could be close to detection, e.g., with FERMI-LAT [231].
10.1 Primordial Black Holes
There is also a small, be definite probability that small mass black holes would have
been formed in the early universe. These “primordial” black holes would have to
have been formed with a rather peculiar mass spectrum not to overclose the universe,
and not to create too much radiation due to the interesting spontaneous emission of
Hawking radiation, named after its discoverer (in theory).
208 L. Bergström
Let us first remind ourselves of the metric surrounding a point mass M,

rS 2 1
ds 2 = 1 − dt − r S dr − r
2 2
dθ2 + sin2 θdϕ2 , (184)
r 1− r
where r S is the Schwarzschild radius, r S = 2GM (here G is Newton’s constant, and

as usual we set c = 1). A radial photon trajectory in this metric is given by ds 2 = 0,
which gives
dr
dt = , (185)
1 − rrS
for θ = φ = const. The time for the photon to travel from ri to r f is thus

rf dr r f − rS
t f − ti = = r f − r i + r S ln . (186)
ri 1 − rrS ri − r S
This diverges as ri → r S , so that a light ray which starts at r < r S will never reach
an outside observer: we have a black hole! If we define the sphere at r = r S as the
areas ABH of the black hole, we find
ABH = 4πr S2 = 4π(2GM)2 = 16πG 2 M 2 . (187)
It is interesting to contrast this with the behaviour of a solid sphere in classical

1 2
physics, where M ∼ R 3 which gives R ∼ M 3 , so that Aclass ∼ M 3 . This difference
is due to the strong curvature of space-time near the black hole.
As we noted, known black hole candidates either have a mass of a few solar
masses (probably remnants of stellar collapse, as the maximal mass of a neutron star
is somewhere between 1.4 solar masses—the Chandrasekhar mass—and a few solar
masses), or a few million solar masses (Milky Way centre) to billions of solar masses
(AGNs). There is no known present formation mechanism for BHs of mass less than
a solar mass, so these, if they exist, must be primordial (PBHs), i.e. produced in the
early universe, e.g. at some phase transition after inflation. There are various limits
restricting formation scenarios, in general one has to “cook up” a power spectrum of
density fluctuations which peaks at a particular mass length scale. When the horizon
passes that scale, copious production of BHs may occur in such a scenario. An
example can be found in a recent paper [232] where one tries to explain all of dark
matter with PBHs, by having a power spectrum with a huge peak (δρ/ρ ∼ 0.1) at a
scale corresponding to a black hole mass of 10−7 M .
10.2 Hawking Radiation
If PBHs exist, one may detect them through Hawking radiation, as Hawking dis-
covered in a remarkable 1975 paper [233] that a black hole should emit thermal
radiation. This can be explained as a tunneling phenomenon [234].
Let us make an extremely simplified heuristic version of the derivation. Let us say
that we have an isolated black hole. We can then for certain say that it is inside the
Schwarzschild radius r S . This uncertainty in the position of the radiation gives an
uncertainty of the time Δt ∼ r S /c = 2GM/c3 , but the uncertainty relation between
time and energy, ΔEΔt ∼ /2 gives
c3 c3 1
ΔE ∼ ∼ E th = k B T → k B T ∼ = . (188)
4GM 4GM 4GM
Thus, in our units, where also k B = 1, the temperature T = 1/(4GM). This is only
a factor of 2π different from Hawking’s result:
1
TH = . (189)
8πGM
Of course, Hawking’s derivation is much more beautiful, by explaining the radiation
by mixing of positive and negative energy states due to the strong space-time curva-
ture near the black hole. Another way to understand the process is that for a virtual
particle pair created just at the horizon, one of the particles will be dragged into the
black hole, releasing gravitational binding energy to the other particle, which can
then appear as a real propagating state outside the horizon.
An interesting consequence of Hawking radiation and the equivalence principle
is that an uniformly accelerated observer, with acceleration a, in empty space should
see a thermal distribution of particles—the Unruh effect. The Unruh temperature is
T = a/2π. Attempts have been made to measure the Unruh effect at accelerators,
but present-day accelerations are not large enough. It has been argued, however, that
the so-called Sokolov-Ternov effect (depolarization of electrons in a storage ring due
to synchrotron radiation) really has the same origin as the Unruh effect—and it has
been experimentally verified [235].
10.3 Thermodynamics of Black Holes
If we regard the Hawking temperature as a true thermodynamical temperature

T (M) = T (E), there should also be an entropy (Bekenstein entropy) associated
with the BH:
210 L. Bergström
1 dE M 1 ABH
T (E) = ; dS = →S= 8πGEdE = 4πGM 2 =.
8πGE T (E) 0 4 G
(190)
If we remember that G = 1/M Pl 2 = l 2 , we see that each “Planck area” of
Pl
the surface of the BH contributes one quarter unit of entropy, and one gets huge
numbers. This is still mysterious—what are the degrees of freedom describing the
black hole, and why does ordinary matter that forms a BH suddenly increase its
entropy enormously?
To describe black hole evaporation, it is useful to remember the form of a thermal
distribution for a particle species
1 1
f i (p) = E i −μi
= E i −μi
. (191)
e kB T ±1 e T ±1
This means that for the rate of mass loss we can analogously write [236]
dM 1 ∞ Ed E
=− Γj = . . . = −5 · 1025 f (M)M −2 gs −1 . (192)
dt 2π mj e8πGME ± 1
j
Here Γ j is the absorption rate for particle of type j and the sum is over all particle-
antiparticle pairs. This gives the evaporation time
3
Mmax M2 6 · 10−27 Mi
τevap ∼ dM ∼ s. (193)
Mmin f (M) f (Mi ) 1g
Thus, only BHs with mass >1015 g are stable on cosmological time scales (so
don’t worry about BHs produced at LHC—they would evaporate immediately—if
they exist!) Upper limits of γ-ray emission from EGRET and FERMI-LAT gives the
approximate bound for light PBHs:
ΩPBH (M < 1015 g) < 10−8 . (194)
Actually, since the temperature increases with decreasing mass, all particles, even
more massive than those presently produced at accelerators, may give a contribution
in the final “Hawking explosion”. In particular, if supersymmetry is realized in nature,
the end-time evolution may have interesting differences from the scenario with only
Standard Model particles [237].
10.4 Formation of Primordial Back Holes
Let us discuss how primordial black holes formed in the early universe (see [238]).
The relevant length scale is the particle horizon length, so that

t
M = γ Mparticle horizon = 2 · 105 γ M , (195)
1s
where γ ∼ 0.2 depends on the detailed formation scenario. We can now compute
the fraction of total energy density in black holes at formation:
ρPBH (ti ) g 1 M 23 n
−29 1 i 4 PBH (t0 )
β(M) ≡ = 8 · 10 √ . (196)
ρ(ti ) γ 106.75 M 1 Gpc−3
This means a contribution to Ω today of

− 1
Mn PBH (t0 ) β(M) √ gi − 14 M 2
ΩPBH = = γ . (197)
ρc 1.2 · 10−8 106.75 M
The WMAP bound (the PBHs would behave gravitationally as cold dark matter)
Ω < 0.25, gives
1
−18 1 gi 41 M 2
β(M) < 2 · 10 √ . (198)
γ 106.75 1015 g
(This is valid for BHs that have not evaporated today, i.e., for M > 1015 g.) It is
convenient to divide out the cosmology/formation factors and consider the simpler
expression for the energy density limit from WMAP:
1
√ gi − 14 M 2
β (M) < γ β(M) = 2 · 10−18 . (199)
106.75 1015 g
Limits on β (M) can be obtained from a variety of data, from BBN and CMB in the
early universe to the galactic and diffuse extragalactic γ-ray emission, gravitational
lensing data and large scale structure. The limits we just computed on ΩPBH is also
important in the region M ∼ 1015 –1027 g (for a detailed summary of the situation,
see [238]).
To conclude: PBHs of mass less than around 1015 g cannot be the dark matter
due to important constraints from the absence of Hawking radiation in 1–100 MeV
γ-rays, but may still be a subdominant component. It is worthwhile to look for γ-ray
signatures—a discovery of Hawking radiation would be truly wonderful!
At all masses, there are important bounds from a variety of methods. In principle,
there are mass ranges where PBHs can still be the dark matter—all of dark matter,
but one needs contrived production mechanisms such as a strongly peaked, and fine-
tuned, density power spectrum.
212 L. Bergström
11 Gravitational Waves
We will now, in view of the multi-messenger aspects of this lecture series, discuss
one more type of radiation which is deeply linked to the theory of general relativity
on which modern cosmology rests: gravitational radiation.
Due to the nonlinearity of Einstein’s equations, it is virtually impossible to find
exact solutions to the metric tensor g μν (r, t) corresponding to the dynamics, for
example, of a massive star which collapses to a black hole near the strong gravitational
field of the star (using supercomputers, numerical studies can, however, be made).
Far from the source of the gravitational field, it is on the other hand reasonable to
use a first-order approximation. The gravitational deformation of space-time near
celestial bodies like the Earth or the Sun due to conceivable astrophysical processes
happening elsewhere in the Galaxy is indeed as we will see extremely tiny, which
justifies such a perturbative approach. The same smallness of the effect unfortunately
also make detection of gravitational radiation very challenging.
11.1 The Gauge Choice for Electromagnetism
Recall the way one derives the existence of electromagnetic waves in Maxwell’s
theory. One inserts the vector potential Aμ in the equations of motion for a vanishing
current j μ (since we are dealing with propagation in vacuum) to obtain

Aμ − ∂ μ ∂ν Aν = 0 (200)
Through the use of the gauge freedom Aμ → Aμ + ∂ μ f , we can choose Aμ to fulfill

A0 = 0 and also the so-called Lorentz condition ∂ν Aν = 0. This leads to the wave
equation
Aμ = 0 (201)
with solutions of the form

μx
Aμ (r, t) = μ e±i(ωt−k·r) = μ e±ik μ (202)
where k μ kμ = 0 (light-like propagation) and the gauge conditions A0 = 0 and

∂ν Aν = ∇ · A = 0 translate into 0 = 0 and k · = 0. This means that the two
physical degrees of freedom are transverse to the direction of propagation, and there
is no time-like mode of propagation (this is deeply connected to the masslessness of
the photon).
11.2 Gauge Choice for the Metric Perturbation
In the case of gravity waves in Einstein’s theory of general relativity, we can similarly
make a first-order expansion of the dynamical degrees of freedom, which are the
components of the metric tensor field gμν , around the constant Minkowski metric
ημν :
gμν = ημν + h μν , (203)
and work to the first non-vanishing order in h μν .

Now we have a spin-2 field h μν instead of the vector quantity Aμ , but again
we can use a gauge-like invariance (which in this case rather is re-parametrization
invariance)
xμ → xμ + ξμ (x) (204)
translating into
h μν → h μν − ∂μ ξν − ∂ν ξμ . (205)
Using this we may impose the so-called traceless gauge condition
h μμ = 0 (206)
The analogy of A0 = 0 is
h 0ν = h ν0 = 0, (207)
and of the transversality condition
∇ ·A=0 (208)
is
∇i h iν = ∇i h νi = 0. (209)
The Einstein equation (neglecting back-reaction, i.e. the contribution to the

energy-momentum tensor by the gravitational field itself) becomes simply
h μν = 0. (210)
11.3 Solutions to the Wave Equation
Exactly like for photons we can write for the wave solutions to Einstein’s equation
h μν = E μν e±i(ωt−k·x) (211)
214 L. Bergström
with k 2 = ω 2 , i.e. massless propagation, with the speed of light. (There have been
brave attempts to replace Einstein’s gravity with a massive theory, with the extra
component having extremely small mass. This would lead to many interesting dif-
ferences, perhaps even explaining the small value of the cosmological constant. So
far, there has not appeared any generally accepted way to do this, however.)
We can represent E μν by a 4 × 4 matrix, which, exactly like for Aμ , should
reflect the gauge choice. We know already that the E 0ν row and E μ0 columns are
zero. Also E has to be symmetric in the two indices (since the metric is). Further,
k i E iν = k j E μ j = 0, meaning that also the elements of the E 3ν column and E μ3 row
are zero for a wave propagating in the z-direction. So, we really just have zeros for
our perturbative solution apart from a symmetric, traceless 2 × 2 matrix. A general
such matrix is a linear combination of
⎛ ⎞
00 0 0
⎜0 1 0 0⎟
+
E μν =⎜ ⎟
⎝ 0 0 −1 0 ⎠ (212)
00 0 0
and ⎛ ⎞
0 0 0 0
⎜0 0 1 0⎟
×
E μν =⎜
⎝0
⎟ (213)
1 0 0⎠
0 0 0 0
For a given value of the 3-component z, and at time t, we can then write
+ ×
E μν (t) = h + (t)E μν + h × (t)E μν . (214)
Look at the case

h + = 0, h × = 0. (215)
At a given time, we have in the unperturbed case
Δs 2 = ημν Δx μ Δx ν = (Δt)2 − (Δx i )2 = − (Δx i )2 (216)

i i
For two diametrically opposed points on the unit circle,
Δx i = (2 cos θ, 2 sin θ, 0) (217)
and their distance is

d0 = −ηi j (Δx i )(Δx j ) = 2 sin2 θ + cos2 θ = 2. (218)
In the perturbed case (i.e., if a gravity wave passes)

(a) (b)
Fig. 11 a The deformation of the unit circle caused by gravity waves proportional to the polarization
amplitude h + . Shown are the unperturbed circle and the maximally stretched configurations along
the two axes of symmetry, the x and y axes. b The corresponding pattern for the orthogonal
polarization state described by the amplitude h × . Note that the axes along which stretching and
compression occur form 45-degree angles to the x and y axes
d+ = −(ηi j + h + E i+j )Δx i Δx j = 4 − h + (t)4(cos2 θ − sin2 θ) (219)
2 − h + (t)(cos2 θ − sin2 θ) = 2 − h + (t) cos 2θ. (220)
For simplicity, we may work with real h by combining as usual the waves with the
two signs in the exponential, giving
h+ + +
μν = E μν h + (t) = E μν cos (ωt − k · r) (221)
and we see that the unit circle will be successively “compressed” or “squeezed”
depending on the sign of the last factor (see Fig. 11, where the corresponding defor-
mation caused by h × is also shown).
These are the two independent quadrupole deformations of a circle. This means
that the source of the gravitational field giving gravity waves has to have a quadrupole
moment, Q. From dimensional reasoning,
216 L. Bergström
G N Q̈ 4G N E kin
h∼ ∼ , (222)
d d
which is obtained by the crude estimate
Q ∼ Ml 2 ⇒ Q̇ = M2l l˙ = 2Mlv ⇒ Q̈ ∼ 2Mv 2 = 4E kin . (223)
For objects in the Milky Way, typically d ∼ 10 kpc, and with E kin ∼ M we find
h ∼ 10−17 . (224)
On the other hand, for the distance appropriate for the Virgo galaxy cluster,
h ∼ 10−20 . (225)
These extremely tiny deformations is the reason for the non-detection so far of
gravitational radiation, although there are promising objects like coalescing neutron
stars which should have amplitudes nearing the experimental upper limits.
In a sense, gravity waves have already been indirectly detected, however, by
comparing the slowing-down of the rotation rate of the binary pulsar system PSR
1913–16 by Hulse and Taylor (Nobel Prize of 1993):
dP
= (−2.4225 ± 0.0056) · 10−12 , (226)
dt
with the general relativistic calculation (with energy loss due to gravitational radia-
tion):
d PG R
= −2.40 · 10−12 . (227)
dt
This excellent agreement has put severe limits on possible modifications of Einstein
gravity. But effects of gravity waves have so far never been detected directly on
Earth, despite an impressive increase in sensitivity of the LIGO experiment in the
US, and VIRGO in Italy. Actually by combining several experiments and searching
for time-coincident effects, one may both decrease various noise sources and increase
the sensitivity for a signal. This is presently done by LIGO, VIRGO and GEO600 in
Germany. All three detector are presently being upgraded to more advanced versions.
However, it may be that a space experiment, LISA, will be needed to detect a sig-
nificant signal. Its status is, however, at present unclear due to the difficult financial
situation in most countries of the world.
We finally remind that there is also a possibility of detecting gravitational waves
that are relics of dramatic processes in the early universe, such as during the epoch of
inflation or during the formation of cosmic strings, if such exist. In that case, the most
promising method is through analyzing the imprints they have made in the CMBR.
As gravitational waves carry a quadrupole moment it is possible to distinguish their
effects through studies of CMBR polarization. With the Planck satellite it is possible
to search for gravitational waves of very long wavelength generated through these
hypothetical processes. Results are expected in early 2013.
12 Conclusions
This finishes our trip through the universe, looking at fundamental processes of
similar interest to particle physicists, cosmologists, astrophysicists and astroparticle
physicists alike. As hopefully has becomes clear, by combining information from all
messengers that we have available: photons of all wavelengths, neutrinos, antimatter
and perhaps gravitational waves, we may study from the Earth some of the most
energetic and interesting processes in the universe. If we are lucky, we may even
solve the problem of the nature of the dark matter, which has been with us since the
times of Fritz Zwicky. Let us remind ourselves of his prophetic words from 1933
[8], after observing the rapid movement of the galaxy members of the Coma cluster,
which pointed to an overdensity of matter in the cluster:
If this over-density is confirmed we would arrive at the astonishing conclusion
that dark matter is present with a much greater density than luminous matter. . .
Acknowledgments The author is grateful to several colleagues, including G. Bertone, T. Bringmann,

J. Conrad, J. Edsjö, P. Gondolo, A. Goobar, P.O. Hulth, E. Mörtsell, F. Aharonian and C. Dermer as
well as the participants of the 2010 Saas-Fee Course for many useful suggestions and discussions. A
particular thanks to the organizers of the School, chaired by Roland Walter, for making the School
such an enjoyable and memorable event, and to Marc Türler for a careful reading of the manuscript.
References
1. L. Bergström, A. Goobar, Particle Astrophysics and Cosmology, 2nd edn. (Springer, Dor-
drecht, 2006). ISBN-13: 978-3540329244
2. P. Gondolo, J. Edsjö, P. Ullio, L. Bergström, M. Schelke, E.A. Baltz, JCAP 0407, 008 (2004)
. [astro-ph/0406204]
3. L. Bergström, G. Bertone, in Particle Dark Matter, ed. by Bertone, G. (Cambridge University
Press, Cambridge, 2010) pp. 491–506. ISBN-13: 978-0521763684
4. L. Bergström, invited talk at the EPS-HEP Conference, to appear in the Proceedings (Grenoble,
2011)
5. L. Bergström, Dark Matter and Imaging Air Cherenkov Arrays, to appear in Astroparticle
Physics, special issue for the Cherenkov Telescope Array (2012)
6. M. Blomqvist, E. Mortsell, JCAP 1005, 006 (2010). [arXiv:0909.4723 [astro-ph.CO]]
7. E. Komatsu et al. WMAP Collaboration. Astrophys. J. Suppl. 192, 18 (2011).
[arXiv:1001.4538 [astro-ph.CO]]
8. F. Zwicky, Helv. Phys. Acta 6, 110 (1933)
9. D. Clowe, M. Bradac, A.H. Gonzalez, M. Markevitch, S.W. Randall, C. Jones, D. Zaritsky,
Astrophys. J. 648, L109–L113 (2006). [astro-ph/0608407]
10. P. Binetruy, G. Girardi, P. Salati, Nucl. Phys. B 237, 285 (1984)
11. J. Edsjö, P. Gondolo, Phys. Rev. D 56, 1879 (1997). [hep-ph/9704361]
12. A. Linde, Phys. Lett. B129, 177 (1983)
13. C. Itzykson, J.B. Zuber, Quantum Field Theory, (McGraw-Hill, New York, 1980)
14. L.J. Reinders, H. Rubinstein, S. Yazaki, Phys. Rept. 127, 1 (1985)
218 L. Bergström
15. J.B. Kogut, Rev. Mod. Phys. 55, 775 (1983)

16. G. Altarelli, G. Parisi, Nucl. Phys. B 126, 298 (1977)
17. M. Shaposhnikov, I. Tkachev, Phys. Lett. B 639, 414 (2006). [hep-ph/0604236]
18. G.G. Raffelt, Stars as Laboratories for Fundamental Physics, (University of Chicago Press,
Chicago, 1996)
19. K. Nakamura et al., Particle data group collaboration. J. Phys. G G 37, 075021 (2010)
20. A. Osipowicz et al., KATRIN Collaboration. hep-ex/0109033
21. K. Abe et al., Super-Kamiokande collaboration. Phys. Rev. D 83, 052010
(2011).[arXiv:1010.0118 [hep-ex]]
22. B. Aharmim et al., SNO collaboration. Phys. Rev. C 81, 055504 (2010). [arXiv:0910.2984
[nucl-ex]]
23. S. Abe et al., KamLAND collaboration. Phys. Rev. Lett. 100, 221803 (2008).
[arXiv:0801.4589 [hep-ex]]
24. R. Acquafredda et al., OPERA collaboration. New J. Phys. 8, 303 (2006). [hep-ex/0611023]
25. P. Adamson et al., MINOS collaboration. Phys. Rev. Lett. 107, 011802 (2011).
[arXiv:1104.3922 [hep-ex]]
26. F. Halzen, D. Hooper, New J. Phys. 11, 105019 (2009). [arXiv:0910.4513 [astro-ph.HE]]
27. M. Cirelli, A. Strumia, M. Tamburini, Nucl. Phys. B 787, 152 (2007). [arXiv:0706.4071
[hep-ph]]
28. K.A. Intriligator, N. Seiberg, Class. Quant. Grav. 24, S741 (2007). [hep-ph/0702069]
29. R.R. de Austri, R. Trotta, L. Roszkowski, JHEP 0605, 002 (2006). [hep-ph/0602028]
30. P. Scott, J. Conrad, J. Edsjö, L. Bergström, C. Farnier, Y. Akrami, JCAP 1001, 031 (2010).
31. C. Strege, G. Bertone, D.G. Cerdeno, M. Fornasa, R.R. de Austri, R. Trotta, [arXiv:1112.4192
[hep-ph]]
32. G. Jungman, M. Kamionkowski, K. Griest, Phys. Rept. 267, 195 (1996). [hep-ph/9506380]
33. G. Bertone, D. Hooper, J. Silk, Phys. Rept. 405, 279 (2005) [hep-ph/0404175]
34. L. Bergström, New J. Phys. 11, 105006 (2009) [arXiv:0903.4849 [hep-ph]]
35. G. F. Giudice, A. Romanino, Nucl. Phys. B 699, 65 (2004). [Erratum-ibid. B 706 (2005) 65]
[hep-ph/0406088]
36. K. Griest, M. Kamionkowski, Phys. Rev. Lett. 64, 615 (1990)
37. J. Ellis, T. Falk, K.A. Olive, Phys. Lett. B444, 367 (1998)
38. J. R. Ellis, T. Falk, G. Ganis, K.A. Olive, M. Srednicki, Phys. Lett. B 510, 236 (2001).
[hep-ph/0102098]
39. G.W. Bennett et al., Muon G-2 collaboration. Phys. Rev. D 73, 072003 (2006). [hep-
ex/0602035]
40. L. Covi, H. -B. Kim, J.E. Kim, L. Roszkowski, JHEP 0105, 033 (2001). [hep-ph/0101009]
41. A. Kusenko, M.E. Shaposhnikov, Phys. Lett. B 418, 46 (1998). [hep-ph/9709492]
42. M.I. Tsumagari, [arXiv:0910.3845 [hep-th]]
43. R. Catena, P. Ullio, JCAP 1008, 004 (2010). [arXiv:0907.0018 [astro-ph.CO]]
44. J. Ellis, R. Flores, Phys. Lett. B263, 259 (1991)
45. J. Engel, Phys. Lett. B264, 114 (1991)
46. L. Bergström, P. Gondolo, Astropart. Phys. 5, 263 (1996)
47. M. Drees, M. Nojiri, Phys. Rev. D48, 3483 (1993)
48. D. Bailin, G.V. Kraniotis, A. Love, Nucl. Phys. B556, 23 (1999)
49. L. Baudis et al., Heidelberg-Moscow collaboration. Phys. Rev. D59, 022001 (1999)
50. R.W. Schnee et al., CDMS collaboration, Phys. Rep. 307, 283 (1998)
51. P. Belli et al., DAMA collaboration. Phys. Rep. 307, 269 (1998)
52. P. Smith et al., UKDM collaboration. Phys. Rep. 307, 275 (1998)
53. Z. Ahmed et al., CDMS-II collaboration. Phys. Rev. Lett. 106, 131302
(2011).[arXiv:1011.2482 [astro-ph.CO]]
54. E. Aprile et al., XENON100 collaboration. Phys. Rev. Lett. 105, 131302 (2010).
55. A.K. Drukier, K. Freese, D.N. Spergel, Phys. Rev. D 33, 3495 (1986)
56. A. Bottino, F. Donato, N. Fornengo, S. Scopel, Astropart. Phys. 13, 215 (2000). [hep-
ph/9909228]
57. S. Andreas, T. Hambye, M.H.G. Tytgat, JCAP 0810, 034 (2008). [arXiv:0808.0255 [hep-ph]]
58. R. Bernabei et al., DAMA collaboration. Eur. Phys. J. C56, 333–355 (2008). [arXiv:0804.2741
[astro-ph]]
59. V. Springel, J. Wang, M. Vogelsberger, A. Ludlow, A. Jenkins, A. Helmi, J.F. Navarro, C.S.
Frenk et al., Mon. Not. Roy. Astron. Soc. 391, 1685 (2008). [arXiv:0809.0898 [astro-ph]]
60. V. Springel, S.D.M. White, C.S. Frenk, J.F. Navarro, A. Jenkins, M. Vogelsberger, J. Wang,
A. Ludlow et al., Nature 456N7218, 73–80 (2008)
61. J. Diemand, M. Kuhlen, P. Madau, M. Zemp, B. Moore, D. Potter, J. Stadel, Nature 454, 735
(2008). [arXiv:0805.1244 [astro-ph]]
62. S.W. Barwick et al., Phys. Rev. Lett. 75, 390–393 (1995)
63. O. Adriani et al., PAMELA collaboration. Nature 458, 607 (2009). [arXiv:0810.4995 [astro-
ph]]
64. A.A. Abdo et al., The fermi LAT collaboration. Phys. Rev. Lett. 102, 181101
(2009).[arXiv:0905.0025 [astro-ph.HE]]
65. D. Müller, K.K. Tang, Astrophys. J. 312, 183 (1987)
66. R.L. Golden et al., Astrophys. J. 457, L103 (1996)
67. H. Gast, J. Olzem, S. Schael, Proceedings of XLIst Rencontres de Moriond. Electroweak
Interactions and Unified Theories, (La. Thuile, Italy, 2006), pp. 421–428
68. S.W. Barwick et al., Astrophys. J. 482, L191 (1997)
69. J.J. Beatty et al., Phys. Rev. Lett. 93, 241102 (2004)
70. I.V. Moskalenko, A.W. Strong, Astrophys. J. 493, 694 (1998). [arXiv:astro-ph/9710124]
71. O. Adriani et al., Phys. Rev. Lett. 102, 051101 (2009). [arXiv:0810.4994 [astro-ph]]
72. F.A. Aharonian, A.M. Atoyan, H. J. Völk, Astron. Astrophys. 294, L41–L44 (1995)
73. D. Grasso et al., FERMI-LAT collaboration. Astropart. Phys. 32, 140–151 (2009).
[arXiv:0905.0636 [astro-ph.HE]]
74. S. Profumo, Cent. Eur. J. Phys. 10, 1 (2011). [arXiv:0812.4457 [astro-ph]]
75. D. Malyshev, I. Cholis, J. Gelfand, Phys. Rev. D 80, 063005 (2009). [arXiv:0903.1310 [astro-
ph.HE]]
76. L. Bergström, T. Bringmann, J. Edsjö, Phys. Rev. D 78, 103520 (2008).[arXiv:0808.3725
[astro-ph]]
77. N. Arkani-Hamed, D.P. Finkbeiner, T.R. Slatyer, N. Weiner, Phys. Rev. D79, 015014 (2009)
[arXiv:0810.0713 [hep-ph]]
78. Y. Nomura, J. Thaler, Phys. Rev. D 79, 075008 (2009). [arXiv:0810.5397 [hep-ph]]
79. D.P. Finkbeiner, L. Goodenough, T.R. Slatyer, M. Vogelsberger, N. Weiner, JCAP 1105, 002
(2011) [arXiv:1011.3082 [hep-ph]]
80. K. Lubelsmeyer, A. Schultz von Dratzig, M. Wlochal, G. Ambrosi, P. Azzarello, R. Battiston,
R. Becker, U. Becker et al., Nucl. Instrum. Meth. A 654, 639 (2011)
81. F. Donato, N. Fornengo, P. Salati, Phys. Rev. D 62, 043003 (2000). [hep-ph/9904481]
82. J.E. Koglin, T. Aramaki, S.E. Boggs, W.W. Craig, H. Fuke, F. Gahbauer, C.J. Hailey, N.
Madden et al., J. Phys. Conf. Ser. 120, 042011 (2008)
83. H.U. Bengtsson, P. Salati, J. Silk, Nucl. Phys. B346, 129 (1990)
84. T. Sjostrand, S. Mrenna, P.Z. Skands, JHEP 0605, 026 (2006). [hep-ph/0603175]
85. M. Srednicki, S. Theisen, J. Silk, Phys. Rev. Lett. 56, 263 (1986). Erratum-ibid 56, (1986)
1883
86. L. Bergström, H. Snellman, Phys. Rev. D37, 3737 (1988)
87. A. Bouquet, P. Salati, J. Silk, Phys. Rev. D40, 3168 (1989)
88. G. Jungman, M. Kamionkowski, Phys. Rev. D51, 3121 (1995)
89. M. Urban et al., Phys. Lett. B293, 149 (1992)
90. L. Bergström, J. Kaplan, Astropart, Phys. 2, 261 (1994)
91. P. Ullio, L. Bergström, Phys. Rev. D57, 1962 (1998)
92. L. Bergström, P. Ullio, J.H. Buckley, Astropart. Phys. 9, 137–162 (1998). [astro-ph/9712318]
220 L. Bergström
93. A.A. Abdo, M. Ackermann, M. Ajello, W.B. Atwood, L. Baldini, J. Ballet, G. Barbiellini, D.
Bastieri et al., Phys. Rev. Lett. 104, 091302 (2010). [arXiv:1001.4836 [astro-ph.HE]]
94. A.M. Galper, R.L. Aptekar, I.V. Arkhangelskaya, M. Boezio, V. Bonvicini, B.A. Dolgoshein,
M.O. Farber, M.I. Fradkin et al., Astrophys. Space Sci. Trans. 7, 75 (2011)
95. L. Bergström, P. Ullio, Nucl. Phys. B504, 27 (1997)
96. Z. Bern, P. Gondolo, M. Perelstein, Phys. Lett. B411, 86 (1997)
97. J. Hisano, S. Matsumoto, M.M. Nojiri, O. Saito, Phys. Rev. D71, 063528 (2005). [hep-
ph/0412403]
98. L. Bergström, Phys. Lett. B 225, 372 (1089)
99. L. Bergström, T. Bringmann, M. Eriksson, M. Gustafsson, Phys. Rev. Lett. 95, 241301 (2005).
[arXiv:hep-ph/0507229]
100. A. Birkedal, K.T. Matchev, M. Perelstein, A. Spray, [arXiv:hep-ph/0507194]
101. L. Bergström, T. Bringmann, M. Eriksson, M. Gustafsson, Phys. Rev. Lett. 94, 131301 (2995).
[arXiv:astro-ph/0410359]
102. T. Bringmann, L. Bergström, J. Edsjö, JHEP 0801, 049 (2008). [arXiv:0710.3169 [hep-ph]]
103. H. Goldberg, Phys. Rev. Lett. 50, 1419 (1983)
104. E.A. Baltz, L. Bergström, Phys. Rev. D 67, 043516 (2003). [arXiv:hep-ph/0211325]
105. J.F. Navarro, C.S. Frenk, S.D.M. White, Astrophys. J. 462, 563 (1996)
106. A. Abramowski et al., H.E.S.S. collaboration. Phys. Rev. Lett. 106, 161301
(2011).[arXiv:1103.3266 [astro-ph.HE]]
107. J. Aleksic et al., The MAGIC collaboration. JCAP 1106, 035 (2011). [arXiv:1103.0477 [astro-
ph.HE]]
108. M. Wood, G. Blaylock, S.M. Bradbury, J.H. Buckley, K.L. Byrum, Y.C.K. Chow, W. Cui,
I.D.l.C. Perez et al., [arXiv:0801.1708 [astro-ph]]
109. T. Bringmann, M. Doro, M. Fornasa, JCAP 0901, 016 (2009). [arXiv:0809.2269 [astro-ph]]
110. F.A. Aharonian, A.K. Konopelko, H.J. Volk, H. Quintana, Astropart. Phys. 15, 335–356
(2001). [astro-ph/0006163]
111. E. A. Baltz, B. Berenji, G. Bertone, L. Bergström, E. Bloom, T. Bringmann, J. Chiang, J.
Cohen-Tanugi et al., JCAP 0807, 013 (2008). [arXiv:0806.2911 [astro-ph]]
112. J. Silk, A. Stebbins, Astrophys. J. 411, 439 (1993)
113. L. Bergström, J. Edsjö, P. Gondolo, P. Ullio, Phys. Rev. D 59, 043506 (1999). [astro-
ph/9806072]
114. B. Moore, S. Ghigna, F. Governato, G. Lake, T.R. Quinn, J. Stadel, P. Tozzi, Astrophys. J.
524, L19 (1999). [astro-ph/9907411]
115. P. Gondolo, J. Silk, Phys. Rev. Lett. 83, 1719 (1999)
116. S. Ritz, D. Seckel, Nucl. Phys. 304, 877 (1988)
117. J. Edsjö, Ph.D. Thesis, Uppsala University, hep-ph/9504205
118. F. Halzen, Comments Nucl. Part. Phys. 22, 155 (1997)
119. A. Bottino, N. Fornengo, G. Mignola, L. Moscoso, Astropart. Phys. 3, 65 (1995)
120. V. Berezinsky et al., Astropart. Phys. 5, 333 (1996)
121. L. Bergström, J. Edsjö, P. Gondolo, Phys. Rev. D55, 1765 (1997)
122. M. Blennow, J. Edsjö, T. Ohlsson, JCAP 0801, 021 (2008). [arXiv:0709.3898 [hep-ph]]
123. L. Bergström, J. Edsjö, P. Gondolo, Phys. Rev. D58, 103519 (1998)
124. T.K. Gaisser, F. Halzen, T. Stanev, Phys. Rep. 258, 173 (1995)
125. M. Kamionkowski, G. Jungman, K. Griest, B. Sadoulet, Phys. Rev. Lett. 74, 5174 (1995)
126. J. Edsjö, P. Gondolo, Phys. Lett. B357, 595 (1995)
127. L. Bergström, J. Edsjö, M. Kamionkowski, Astropart. Phys. 7, 147 (1997)
128. A. Strumia, PoS EPS-HEP2009, 012 (2009)
129. L. Bergström, G. Bertone, T. Bringmann, J. Edsjö, M. Taoso, Phys. Rev. D 79, 081303 (2009).
[arXiv:0812.3895 [astro-ph]]
130. T. Bringmann, L. Bergström, J. Edsjö, JHEP 0801, 049 (2008). [arXiv:0710.3169 [hep-ph]]
131. L. Bergström, T. Bringmann, J. Edsjö, Phys. Rev. D78, 103520 (2008). [arXiv:0808.3725
[astro-ph]]
132. L. Bergström, Phys. Lett. B225, 372 (1989)
133. T. Bringmann, F. Calore, G. Vertongen, C. Weniger, [arXiv:1106.1874 [hep-ph]]

134. N.F. Bell, J.B. Dent, A.J. Galea, T.D. Jacques, L.M. Krauss, T.J. Weiler, [arXiv:1104.3823
[hep-ph]]
135. P. Ciafaloni, M. Cirelli, D. Comelli, A. De Simone, A. Riotto, A. Urbano, JCAP 1110, 034
(2011). [arXiv:1107.4453 [hep-ph]]
136. A. Sommerfeld, Annalen der Physik 403, 257 (1931)
137. M. Kamionkowski, S. Profumo, Phys. Rev. Lett. 101, 261301 (2008). [arXiv:0810.3233]
138. A. Hryczuk, R. Iengo, P. Ullio, JHEP 1103, 069 (2011). [arXiv:1010.2172 [hep-ph]]
139. J.R. Ellis, J.S. Hagelin, D.V. Nanopoulos, K.A. Olive, M. Srednicki, Nucl. Phys. B238, 453–
476 (1984)
140. P. Meade, N. Seiberg, D. Shih, Prog. Theor. Phys. Suppl. 177, 143–158 (2009).
[arXiv:0801.3278 [hep-ph]]
141. C.F. Berger, J.S. Gainer, J.L. Hewett, T.G. Rizzo, JHEP 0902, 023 (2009). [arXiv:0812.0980
[hep-ph]]
142. J.A. Conley, J.S. Gainer, J.L. Hewett, M.P. Le, T.G. Rizzo, [arXiv:1103.1697 [hep-ph]]
143. L. Bergström, Phys. Lett. B225, 372 (1989)
144. For downloading DarkSUSY, visit http://www.fysik.su.se/~edsjö/ds
145. T. Bringmann, New J. Phys. 11, 105027 (2009a). [arXiv:0903.0189 [astro-ph.CO]]
146. D. Spolyar, K. Freese, P. Gondolo, Phys. Rev. Lett. 100, 051101 (2008). [arXiv:0705.0521
[astro-ph]]
147. F. Iocco, Astrophys. J. 677, L1–L4 (2008). [arXiv:0802.0941 [astro-ph]]
148. P. Scott, M. Fairbairn, J. Edsjö, Mon. Not. Roy. Astron. Soc. 394, 82 (2009). [arXiv:0809.1871
[astro-ph]]
149. S. Sivertsson, P. Gondolo, Astrophys. J. 729, 51 (2011). [arXiv:1006.0025 [astro-ph.CO]]
150. D. Tucker-Smith, N. Weiner, Phys. Rev. D64, 043502 (2001). [hep-ph/0101138]
151. J. March-Russell, C. McCabe, M. McCullough, JHEP 0905, 071 (2009). [arXiv:0812.1931
[astro-ph]]
152. Y. Cui, D.E. Morrissey, D. Poland, L. Randall, JHEP 0905, 076 (2009). [arXiv:0901.0557
[hep-ph]]
153. D.P. Finkbeiner, T. Lin, N. Weiner, Phys. Rev. D80, 115008 (2009). [arXiv:0906.0002 [astro-
ph.CO]]
154. D.S.M. Alves, M. Lisanti, J.G. Wacker, Phys. Rev. D82, 031901 (2010). [arXiv:1005.5421
[hep-ph]]
155. M.T. Frandsen, F. Kahlhoefer, J. March-Russell, C. McCabe, M. McCullough, K. Schmidt-
Hoberg, Phys. Rev. D84, 041301 (2011). [arXiv:1105.3734 [hep-ph]]
156. K.R. Dienes, B. Thomas, [arXiv:1106.4546 [hep-ph]]
157. Y. Nomura, J. Thaler, Phys. Rev. D79, 075008 (2009). [arXiv:0810.5397 [hep-ph]]
158. P.J. Fox, E. Poppitz, Phys. Rev. D79, 083528 (2009). [arXiv:0811.0399 [hep-ph]]
159. B. Kyae, JCAP 0907, 028 (2009). [arXiv:0902.0071 [hep-ph]]
160. D. Spolyar, M.R. Buckley, K. Freese, D. Hooper, H. Murayama, [arXiv:0905.4764 [astro-
ph.CO]]
161. A. Ibarra, A. Ringwald, D. Tran, C. Weniger, JCAP 0908, 017 (2009). [arXiv:0903.3625
[hep-ph]]
162. T. Cohen, K.M. Zurek, Phys. Rev. Lett. 104, 101301 (2010). [arXiv:0909.2035 [hep-ph]]
163. N. Haba, Y. Kajiyama, S. Matsumoto, H. Okada, K. Yoshioka, Phys. Lett. B695, 476–481
(2011). [arXiv:1008.4777 [hep-ph]]
164. A. Menon, D.E. Morrissey, C.E.M. Wagner, Phys. Rev. D70, 035005 (2004). [hep-
ph/0404184]
165. M. Dine, N. Seiberg, S. Thomas, Phys. Rev. D76, 095004 (2007). [arXiv:0707.0005 [hep-ph]]
166. N. Bernal, A. Goudelis, JCAP 1003, 007 (2010). [arXiv:0912.3905 [hep-ph]]
167. M. Berg, J. Edsjö, P. Gondolo, E. Lundström, S. Sjors, JCAP 0908, 035 (2009).
[arXiv:0906.0583 [hep-ph]]
168. D.B. Kaplan, Phys. Rev. Lett. 68, 741–743 (1992)
222 L. Bergström
169. D.E. Kaplan, M.A. Luty, K.M. Zurek, Phys. Rev. D79, 115016 (2009). [arXiv:0901.4117
[hep-ph]]
170. A.L. Fitzpatrick, D. Hooper, K.M. Zurek, Phys. Rev. D81, 115005 (2010).
[arXiv:1003.0014[hep-ph]
171. M.T. Frandsen, S. Sarkar, Phys. Rev. Lett. 105, 011301 (2010). [arXiv:1003.4505 [hep-ph]]
172. M.L. Graesser, I.M. Shoemaker, L. Vecchi, [arXiv:1103.2771 [hep-ph]]
173. S. Profumo, L. Ubaldi, JCAP 1108, 020 (2011). [arXiv:1106.4568 [hep-ph]]
174. C. E. Aalseth et al., CoGeNT collaboration. Phys. Rev. Lett. 106, 131301 (2011).
175. C.E. Aalseth, P.S. Barbeau, J. Colaresi, J.I. Collar, J. Diaz Leon, J.E. Fast, N. Fields, T.W.
Hossbach et al., Phys. Rev. Lett. 107, 141301 (2011). [arXiv:1106.0650 [astro-ph.CO]]
176. G. Angloher, M. Bauer, I. Bavykina, A. Bento, C. Bucci, C. Ciemniak, G. Deuter, F. von
Feilitzsch et al., [arXiv:1109.0702 [astro-ph.CO]]
177. D. Hooper, T. Linden, Phys. Rev. D 84, 123005 (2011). [arXiv:1110.0006 [astro-ph.HE]]
178. D. Hooper, S. Profumo, Phys. Rept. 453, 29 (2007). [hep-ph/0701197]
179. M. Gustafsson, E. Lundstrom, L. Bergström, J. Edsjö, Phys. Rev. Lett. 99, 041301 (2007).
[astro-ph/0703512 [ASTRO-PH]]
180. G.F. Bertone (ed.), Particle Dark Matter, (Cambridge University Press, Cambridge, 2010)
181. S.J. Asztalos et al., The ADMX collaboration. Phys. Rev. Lett. 104, 041301 (2010).
182. M.L. Garde, J. Conrad, J. Cohen-Tanugi, M. Kaplinghat, G. Martinez, For the Fermi-LAT
collaboration. [arXiv:1111.0320 [astro-ph.HE]]
183. J. Hamann, S. Hannestad, G.G. Raffelt, I. Tamborra, Y.Y.Y. Wong, Phys. Rev. Lett. 105,
181301 (2010). [arXiv:1006.5276 [hep-ph]]
184. A.X. Gonzalez-Morales, R. Poltis, B.D. Sherwin, L. Verde, [arXiv:1106.5052 [astro-ph.CO]]
185. A. Bottino, F. Donato, N. Fornengo, S. Scopel, Phys. Lett. B423, 109 (1998)
186. R. Arnowitt, P. Nath, Phys. Rev. D60, 044002 (1999)
187. J. Cherwinka, R. Co, D.F. Cowen, D. Grant, F. Halzen, K.M. Heeger, L. Hsu, A. Karle et al.,
188. Z. Ahmed et al., The CDMS-II collaboration. Science 327, 1619–1621 (2010).
189. S.D. Hunter, D.L. Bertsch, J.R. Catelli, T.M. Digel, S.W. Dingus, J.A. Esposito, C.E. Fichtel,
C.E. Hartman et al., Astrophys. J. 481, 205–240 (1997)
190. W. de Boer, C. Sander, V. Zhukov, A.V. Gladyshev, D.I. Kazakov, Astron. Astrophys. 444, 51
(2005). [astro-ph/0508617]
191. A.A. Abdo et al., The Fermi-LAT collaboration. Phys. Rev. Lett. 104, 101101 (2010).
192. J. Knodlseder, P. Jean, V. Lonjou, G. Weidenspointner, N. Guessoum, W. Gillard, G. Skinner,
P. von Ballmoos et al., Astron. Astrophys. 441, 513–532 (2005). [astro-ph/0506026]
193. G. Weidenspointner, G. Skinner, P. Jean, J. Knodlseder, P. von Ballmoos, G. Bignami, R.
Diehl, A.W. Strong et al., Nature 451, 159–162 (2008)
194. P. Meade, M. Papucci, A. Strumia, T. Volansky, Nucl. Phys. B831, 178–203 (2010).
[arXiv:0905.0480 [hep-ph]]
195. L. Bergström, J. Edsjö, G. Zaharijas, Phys. Rev. Lett. 103, 031103 (2009). [arXiv:0905.0333
[astro-ph.HE]]
196. D. Hooper, L. Goodenough, Phys. Lett. B697, 412–428 (2011). [arXiv:1010.2752 [hep-ph]]
197. D. Hooper, T. Linden, [arXiv:1110.0006 [astro-ph.HE]]
198. M. Chernyakova, D. Malyshev, F.A. Aharonian, R.M. Crocker, D.I. Jones, Astrophys. J. 726,
60 (2011). [arXiv:1009.2630 [astro-ph.HE]]
199. A. Boyarsky, D. Malyshev, O. Ruchayskiy, [arXiv:1012.5839 [hep-ph]]
200. D. Hooper, D.P. Finkbeiner, G. Dobler, Phys. Rev. D76, 083012 (2007). [arXiv:0705.3655
[astro-ph]]
201. M. Su, T.R. Slatyer, D.P. Finkbeiner, Astrophys. J. 724, 1044–1082 (2010). [arXiv:1005.5480
[astro-ph.HE]]
202. G. Bertone, M. Cirelli, A. Strumia, M. Taoso, JCAP 0903, 009 (2009). [arXiv:0811.3744
[astro-ph]]
203. T. Schwetz, J. Zupan, JCAP 1108, 008 (2011). [arXiv:1106.6241 [hep-ph]]
204. P.J. Fox, J. Kopp, M. Lisanti, N. Weiner, [arXiv:1107.0717 [hep-ph]]
205. M. Farina, D. Pappadopulo, A. Strumia, T. Volansky, [arXiv:1107.0715 [hep-ph]]
206. M.T. Frandsen, F. Kahlhoefer, C. McCabe, S. Sarkar, K. Schmidt-Hoberg, [arXiv:1111.0292
[hep-ph]]
207. Cherenkov Telescope Array Consortium, [arXiv:1008.3703 [astro-ph.IM]]
208. L. Bergström, T. Bringmann, J. Edsjö, Phys. Rev. D83, 045024 (2011). [arXiv:1011.4514
[hep-ph]]
209. L. Bergström, J. Edsjö, P. Ullio, Phys. Rev. Lett. 87, 251301 (2001). [astro-ph/0105048]
210. P. Ullio, L. Bergström, J. Edsjö, C.G. Lacey, Phys. Rev. D 66, 123502 (2002). [astro-
ph/0207125]
211. I. Ferrero, M.G. Abadi, J.F. Navarro, L.V. Sales, S. Gurovich, [arXiv:1111.6609 [astro-
ph.CO]]
212. W.H. Press, P. Schechter, Astrophys. J. 187, 425 (1974)
213. J.D. Finke, S. Razzaque, C.D. Dermer, Astrophys. J. 712, 238 (2010) [arXiv:0905.1115 [astro-
ph.HE]]
214. W. Buchmuller, A. Ibarra, T. Shindou, F. Takayama, D. Tran, JCAP 0909, 021 (2009).
[arXiv:0906.1187 [hep-ph]]
215. M. Boylan-Kolchin, V. Springel, S.D.M. White, A. Jenkins, G. Lemson, Mon. Not. Roy.
Astron. Soc. 398, 1150 (2009). [arXiv:0903.3041 [astro-ph.CO]]
216. S. Hofmann, D.J. Schwarz, H. Stoecker, Phys. Rev. D 64, 083507 (2001). [astro-ph/0104173]
217. T. Bringmann, New J. Phys. 11, 105027 (2009). [arXiv:0903.0189 [astro-ph.CO]]
218. A.A. Abdo et al., Fermi-LAT collaboration. JCAP 1004, 014 (2010). [arXiv:1002.4415 [astro-
ph.CO]]
219. A. Pinzke, C. Pfrommer, L. Bergström, Phys. Rev. Lett. 103, 181302 (2009).[arXiv:0905.1948
[astro-ph.HE]]
220. T.E. Jeltema, J. Kehayias, S. Profumo, Phys. Rev. D80, 023005 (2009). [arXiv:0812.0597
[astro-ph]]
221. M.A. Sanchez-Conde, M. Cannoni, F. Zandanel, M.E. Gomez, F. Prada, [arXiv:1104.3530
[astro-ph.HE]]
222. A. Pinzke, C. Pfrommer, L. Bergström, [arXiv:1105.3240 [astro-ph.HE]]
223. L. Gao, C.S. Frenk, A. Jenkins, V. Springel, S.D.M. White, [arXiv:1107.1916 [astro-ph.CO]]
224. L. Gao, C.S. Frenk, A. Jenkins, V. Springel, S.D.M. White, Mon. Not. Roy. Astron. Soc. 419,
1721 (2012). [arXiv:1107.1916 [astro-ph.CO]]
225. J. Han, C.S. Frenk, V.R. Eke, L. Gao, S.D.M. White, [arXiv:1201.1003 [astro-ph.HE]]
226. S. ’i. Ando, E. Komatsu, Phys. Rev. D 73, 023521 (2006). [astro-ph/0512217]
227. P.D. Serpico, E. Sefusatti, M. Gustafsson, G. Zaharijas, [arXiv:1109.0095 [astro-ph.CO]]
228. A. Cuoco, T. Linden, M.N. Mazziotta, J.M. Siegal-Gaskins, V. Vitale, E. Komatsu, f. t. F. -L.
collaboration. [arXiv:1110.1047 [astro-ph.HE]]
229. J. Zavala, M. Vogelsberger, T.R. Slatyer, A. Loeb, V. Springel, Phys. Rev. D 83, 123513
(2011). [arXiv:1103.0776 [astro-ph.CO]]
230. P. Gondolo, J. Silk, Phys. Rev. Lett. 83, 1719 (1999). [astro-ph/9906391]
231. G. Bertone, A.R. Zentner, J. Silk, Phys. Rev. D 72, 103517 (2005). [astro-ph/0509565]
232. P.H. Frampton, M. Kawasaki, F. Takahashi, T.T. Yanagida, JCAP 1004, 023 (2010).
[arXiv:1001.2308 [hep-ph]]
233. S.W. Hawking, Commun. Math. Phys. 43, 199 (1975). [Erratum-ibid. 46 (1976) 206]
234. M.K. Parikh, F. Wilczek, Phys. Rev. Lett. 85, 5042 (2000). [hep-th/9907001]
235. E.T. Akhmedov, D. Singleton, Int. J. Mod. Phys. A 22, 4797 (2007). [hep-ph/0610391]
236. J.H. MacGibbon, B.J. Carr, Astrophys. J. 371, 447 (1991)
237. M. Sahlen, [astro-ph/0304478]
238. B.J. Carr, K. Kohri, Y. Sendouda, J. ’i. Yokoyama, Phys. Rev. D 81, 104019 (2010).
Sources of GeV Photons and the Fermi Results
Charles D. Dermer
The Fermi Gamma-ray Space Telescope, now in its fourth year after launch, con-
tinues to make important discoveries and establish new results in all directions:
pulsar astronomy, cosmic-ray physics, AGN and black-hole astrophysics, galactic
astronomy, gamma-ray bursts (GRBs), limits on dark matter and Lorentz invari-
ance violations, γ-ray astronomy of the Sun, moon, and Earth, etc. In this chapter,
I survey results at medium energy γ rays, from some tens of MeV (at energies
above nuclear de-excitation γ-ray lines) to ≈100 GeV where the ground-based γ-ray
Cherenkov detector arrays become more sensitive. As shorthand, Fermi and medium-
energy γ-ray astronomy is referred to here as “GeV astronomy,” and ground-based
Cherenkov γ-ray astronomy 100 GeV as “VHE astronomy.”
The Fermi results already provided considerably more material than could be
presented in the nine lectures that I gave on this subject at the Saas-Fee school on
“Astrophysics at Very High Energies,” held 15–20 March 2010 in Les Diablerets,
Switzerland. Happily, though, Professor Lars Bergström gave a brilliant series of
lectures that covered dark matter, so the absence here of extended discussion on
dark matter and new physics in GeV astronomy reflects Prof. Bergström’s better
capabilities to address this subject. My lectures and this book chapter are therefore
restricted to astrophysical and astroparticle sources of GeV radiation rather than to
γ rays with origins in exotic particle physics and dark matter.
Even while the school was in progress, news appeared of a new type of γ-ray
emitter of GeV photons that was identified with the symbiotic binary Nova V407
Cygni. This extraordinary system reveals an explosive shock evolving on a timescale
of days to weeks, rather than the hundreds of years for supernova remnants (SNRs).
I showed the 2010 March 18 ATEL [2] announcing the Fermi result in my last lecture,
capturing the real-time recognition of a new type of galactic γ-ray source triggered
by a thermonuclear explosion on a white dwarf fed by its binary red giant’s wind.
C. D. Dermer (B)
United States Naval Research Laboratory, Code 7653, 4555 Overlook Ave.,
Washington, DC 20375-5352, USA

228 C. D. Dermer
The discovery has been published by the Fermi Collaboration [1], and is already
triggering a new line of research strongly tied to MeV line astronomy and white-
dwarf physics (cf. citations to [1]). So these lectures and this write-up capture an
early glimpse of the state of knowledge of astrophysical sources of 100 MeV and
GeV radiation obtained with the Large Area Telescope (LAT) on Fermi at ≈2–3 years
into the mission, weighted by the extragalactic interests of the author.
Alongside the LAT on Fermi is the Gamma-ray Burst Monitor (GBM), sensitive to
GRBs and bright transients in the 10 keV–30 MeV range. This review can only briefly
mention important GRB results made with the GBM—by itself and with the LAT—
and related GRB science employing Swift, INTEGRAL, and other detectors. Indeed,
multiwavelength science is value-added science, and the possibilities to uncover
the underlying physics of the powerful compact systems that are at the heart of
high-energy astronomy are multiplied by radio/microwave/sub-mm/IR/optical/UV/
X-ray/MeV/TeV/neutrino/gravitational-wave data correlated with the GeV window,
now observed with unprecedented clarity due to the LAT on Fermi.
The GeV field is in full discovery mode, not only due to Fermi but also thanks
to AGILE, an EGRET-like sentinel of bright γ-ray transients, and to ground-based
VHE observatories, particularly HESS, VERITAS, and MAGIC.1 The ν Fν spectral
energy distributions (SEDs) based on simultaneous and overlapping data sets are
providing valuable information about Galactic sources, blazars and radio galaxies,
and starburst and normal galaxies. GeV astronomy with Fermi is still in the midst of
an active phase as Fermi accumulates data and increasing time makes faint sources
visible and detection of rare cosmic transients more likely.
This chapter will be divided into sections that follow the course of lectures deliv-
ered at the Saas-Fee course. Though now somewhat out-of-date, these lectures can be
found on my website in ppt format.2 The topics of the lectures and the corresponding
sections of this chapter are:
1. GeV instrumentation and the GeV sky with the Fermi Gamma-ray Space Tele-
scope
2. First Fermi Catalog of Gamma Ray Sources and the Fermi Pulsar Catalog
3. First Fermi AGN Catalog
4. Relativistic jet physics and blazars
5. γ rays from cosmic rays in the Galaxy
6. γ rays from star-forming galaxies and clusters of galaxies, and the diffuse extra-
galactic γ-ray background
7. Microquasars, radio galaxies, and the extragalactic background light
8. Fermi observations of GRBs
9. Fermi acceleration, ultra-high energy cosmic rays, and black holes
1 AGILE: Astro-rivelatore Gamma a Immagini LEggero (Gamma-ray Light Imaging Detector);

HESS: High Energy Stereoscopic System, in Namibia. VERITAS: Very Energetic Radiation
Imaging Telescope Array System, in Arizona. MAGIC: Major Atmospheric Gamma-ray Imaging
Cherenkov Telescope in La Palma, Canary Islands.
2 http://heseweb.nrl.navy.mil/gamma/~dermer/default.htm
Sources of GeV Photons and the Fermi Results 229
Besides a discussion of the results of the Fermi Gamma-ray Telescope, I also

include here some high-energy astrophysical theory essential for analysis of γ-ray
data that builds on the research presented in my book with Govind Menon, Ref. [4]:
“High Energy Radiation from Black Holes: Gamma Rays, Cosmic Rays, and Neutri-
nos,” published by Princeton University Press in 2009. The book itself is focused on
theory rather than observation, and develops the hypothesis that the most energetic
and powerful radiations in nature are made by particles energized through Fermi
acceleration processes in black-hole jets powered by rotation.
It is not possible, even at this early stage of the Fermi mission, to adequately sum-
marize all the results from Fermi. But together with the accompanying lectures and
book, this article provides a broad overview of some recent astrophysical advances
in the Fermi era of γ-ray astronomy.
1 GeV Instrumentation and the GeV Sky with the Fermi

Gamma-Ray Space Telescope
1.1 Historical Introduction
The year 2010 represents a highpoint in high-energy astronomy. Astronomical obser-

vatories at multiwavelength electromagnetic, neutrino, cosmic-ray, and gravitational
wavebands are operating and collecting data. X-ray astronomy has sensitive point-
ing X-ray telescopes, Chandra and XMM-Newton, deployed in space to observe
catalogued sources. Broadband X-ray and soft γ-ray observatories like Swift, INTE-
GRAL and Suzaku are available to measure the X/γ spectra of compact objects.
The Rossi X-ray Timing Explorer (RXTE) continues to operate and provide a mon-
itor of the brightest X-ray sources in the sky. Already mentioned were AGILE and
the ground-based γ-ray air Cherenkov telescopes. The Cherenkov Telescope Array
(CTA) consortium [3] is planning to start building as early as 2014 with array com-
pletion as early as 2018. Lacking at the moment, however, is an operating all-sky
water-based Cherenkov telescope successor to MILAGRO in the 1–100 TeV range.
This gap will soon be filled by the High Altitude Water Cherenkov (HAWC) exper-
iment [5] on the Sierra Negra mountain near Puebla, Mexico, which uses 300 tanks
rather than a single pond as utilized by MILAGRO.
The Pierre Auger Observatory, located in the Mendoza province of Argentina
and covering an area the size of Rhode Island, had its third data release in 2010 [6],
giving the spectrum and composition of ultra-high energy cosmic rays (UHECRs)
with energies E 1018 eV. The IceCube Neutrino Observatory, most sensitive to
astrophysical neutrinos with energies of 100 TeV–100 PeV (≈1014 –1018 eV), has
deployed all 86 of its strings in the latest austral summer (December, 2010), and has
developed the DeepCore subarray that is sensitive to lower energy, ≈10–100 GeV,
neutrinos. LIGO, the Laser Interferometry Gravitational-wave Observatory, hoping
to detect gravitational radiation from coalescing compact objects, is operating at
230 C. D. Dermer
design sensitivity. Development of an order-of-magnitude more sensitive Advanced

LIGO has been approved, with completion expected for 2017 or thereafter. The
NASA Laser Interferometry Space Antenna (LISA) is supported in the recent Astro-
2010 study, though ESA is developing a separate space-based gravitational wave
facility. Likewise, Constellation-X has evolved into IXO/ATHENA.3 The large-area
X-ray timing mission RXTE will soon be ended, with the ESA Large Observatory
for X-ray Timing (LOFT) mission taking its place. Here in the US, a focusing hard
X-ray telescope, NuSTAR (Nuclear Spectroscopic Telescope Array), in the 5–80 keV
range, and the Gravity and Extreme Magnetism Small Explorer, GEMS, a NASA
mission to study X-ray polarization of astrophysical sources in the 2–10 keV range,
will soon be launched.
Brief History of GeV Astronomy
The progress of GeV astronomy in the range from ≈10 MeV to ≈100 GeV followed
a period of remarkable advances starting over 40 years ago that culminated with the
launch of Fermi. Prior to the Fermi-LAT, the most important detectors and some of
their achievements in the development of medium-energy γ-ray astronomy are the
following:
• 1967–1968, OSO-3, the Third Orbiting Solar Observatory, carried a Cherenkov
counter experiment sensitive to >50 MeV γ rays, of which 621 were detected [19].
It discovered one source, the extended γ-ray emission of the Milky Way.
• 1972–1973, SAS-2, the Small Astronomy Satellite-2 [16], a spark chamber exper-
iment sensitive to γ rays with energies between ≈30 and 200 MeV (and an integral
flux above 200 MeV). It detected ≈8000 celestial γ-rays, making the first γ-ray
identifications of the Crab and Vela pulsars, Geminga (γ195 + 5, then unidenti-
fied), and the Cygnus region, and an association with Cygnus X-3 was suggested.
A north-south asymmetry in the Galactic γ-ray plane emission was noted and
attributed to the massive stars in the Gould belt. An isotropic γ-ray background
radiation was first reported from analysis of SAS-2 data [15, 33].
• 1975–1982, COS-B, the Cosmic ray Satellite (option B), a magnetic-core, wire-
matrix spark chamber sensitive to γ rays with energies from ≈30 MeV to ≈5 GeV,
with an effective area of ≈50 cm2 at 400 MeV [29]. Its orbit resulted in a large and
variable background of charged particles. During its lifetime, it detected ≈200000
γ-rays, with the COS-B Caravane Collaboration announcing the discovery of 25
sources, most along the Galactic plane. These included 2CG 135+01, now iden-
tified with LSI +61◦ 303, and the first extragalactic source of 100 MeV γ rays,
3C 273 [18].
• 1991–2000, EGRET, the Energetic Gamma Ray Experiment Telescope on the
Compton Gamma Ray Observatory, was a spark chamber experiment with large,
1200 cm2 effective area between 200 MeV and 1 GeV, accompanied by excellent
3 ESA’s Cosmic Vision International X-ray Observatory/Advanced Telescope for High ENergy
Astrophysics, sensitive to photons with energies 0.1–40 keV.

background rejection [32]. During its 9 year mission lifetime, it collected >2×106
γ rays and discovered that blazars and GRBs are luminous γ-ray sources. Because
of the importance of this experiment in the development of GeV astronomy, we
describe EGRET in more detail below.
• 2007–, AGILE, with a Gamma Ray Imaging Detector (GRID) and small calorime-
ter giving γ-ray sensitivity from 30 MeV to 50 GeV, roughly comparable to
EGRET. Along with the GRID is the accompanying 18–60 keV super-AGILE
hard X-ray survey instrument [14]. The first AGILE GRID catalog consists of
47 >100 MeV γ-ray sources with significance >4σ from data taken between
July 2007 and June 2008 [25]. The brightest sources in the catalog, e.g., Vela,
Crab and, during periods of outburst, 3C 454.3, exceed integral photon fluxes
of 2 × 10−6 ph(>100 MeV)/cm2 -s, while the weakest are fainter than 50 ×
10−8 ph(>100 MeV)/cm2 -s.
The EGRET Experiment on CGRO
The Compton Gamma Ray Observatory, or CGRO, was a pioneering γ-ray space
observatory (5 April 1991–4 June 2000) consisting of 4 instruments, OSSE, the Ori-
ented Space Scintillator Experiment (sensitive from ≈50 keV to 10 MeV); BATSE,
the Burst And Transient Source Experiment (≈20 keV–few MeV); COMPTEL, the
Compton Telescope (≈800 keV–30 MeV); and EGRET, the Energetic Gamma Ray
Experiment Telescope (≈30 MeV–10 GeV). EGRET’s detection method utilized a
gas-filled spark chamber that tracked the γ ray after it converted to an electron-
positron pair by pair-production interactions with nuclei in the thin Ta foils in the
gas-filled spark chamber. Directional information was obtained by time-of-flight
coincidence with a scintillator array in the lower spark chamber assembly. Charged
particles were vetoed by an anticoincidence shield and total γ-ray energy was mea-
sured with EGRET’s total absorption scintillator counter, the TASC, consisting of
crystal scintillators and photomultipliers [20].
The Third EGRET (3EG) catalog [17], made from data accumulated between
1991 April 22 and 1995 October 3, consists of 271 sources of >100 MeV emission,
including a Solar flare in 1991, the Large Magellanic Cloud, five pulsars, one probable
radio galaxy, namely Centaurus A, and 66 high-confidence detections of blazars. The
catalog also lists 27 lower confidence potential blazar detections, and contains 170
unidentified EGRET sources lacking associated sources. Five GRBs were detected
with the spark chamber on EGRET. Figure 1 shows a skymap of the 3EG sources.
The EGRET Field of View (FoV), defined roughly by the solid angle within which
the effective area is greater than 1/2 of the on-axis effective area, was ≈0.5 sr, or
≈1/24th of the full sky. EGRET operated in a pointing mode, and targeted one region
of the sky for two weeks, representing ∼106 s after time for Earth occultation and
time spent in the South Atlantic Anomaly are subtracted. The Point Spread Function
(PSF) at 100 MeV was ≈5.7◦ , with the PSF improving roughly as ≈E −1/2 [31, 32].
The first 18 months of the mission were devoted to a full-sky survey.
232 C. D. Dermer
+90
Third EGRET Catalog
E > 100 MeV
+180 -180
LMC
Pulsars
Solar Flare
Active Galactic Nuclei
Unidentified EGRET Sources -90
Fig. 1 EGRET all-sky map of γ-ray sources [17]
Important results from analysis of the full 6 years of data (degradation of spark
chamber gas led to smaller apertures after 4 years into the mission) include the
following:
1. Diffuse extragalactic background with intensity ≈1.5 keV/cm2 -s-sr, correspond-
ing to an energy density u γ ≈ 10−17 erg/cm3 ;
2. The galactic flux is as much as 50× brighter than the extragalactic flux, and
much softer, but not as soft as expected if the γ rays are formed by secondary
nuclear production by Galactic cosmic rays with the same spectrum as observed
locally (the so-called “EGRET excess”);
3. Typical fluxes of EGRET sources are between ∼(10−7 and 10−6 ) ph(>100 MeV)
/cm2 -s, with a typical 2-week on-axis limiting flux at ≈(15–25) × 10−8 ph(>
100 MeV)/cm2 -s;
4. Galactic sources, including young radio pulsars;
5. Sources with significant flux variability.
Temporal variability is an essential characteristic of GRBs and blazars, Solar flares,
and now V407 Cyg. The large FoVs of the LAT and Swift, and larger still with
BATSE and GBM, are crucial for study of γ-ray transients.
Point Source Sensitivity of EGRET
The best description of results leads to the question of units in GeV astronomy.
Because medium-energy and high-energy γ-ray astronomy is challenged by lim-
ited signal counts, an integral photon flux is the most natural unit. For the EGRET
experiment, units of 10−6 ph(>100 MeV)/cm2 -s are suitable, as this value roughly
separates signal-dominated and noise-dominated detection in EGRET, as we now
show. Units 10−8 ph(>100 MeV)/cm2 -s are more suitable for Fermi sources.
As measured with EGRET, the diffuse, or at least unresolved intensity of photons

at 100 MeV, is measured to be I ≈ 1.5 keV/cm 2 -s-sr at E ≡ m e c2 ≈ 100 MeV.
Writing

2−Γγ keV
I = 1.5 , (1)
100 cm2 -s-sr
with photon number index Γγ and 100 = 100 MeV/m e c2 , implies

∞ 1−Γγ
dN(>) I 1.5 × 10−5 ph(>100 MeV)
= d = . (2)
dAdtdΩ m e c2 2 Γγ − 1 100 cm2 -s-sr
EGRET measured a hard spectrum, with Γγ ≈ 2.1 [26], whereas LAT measures a
softer spectrum, with Γγ ≈ 2.4 [7], and with a 100 MeV intensity only 60 % as large
as the EGRET intensity.4 For the EGRET intensity, therefore, dN(>100 )/dAdtdΩ ≈
1.5 × 10−5 ph(>100 MeV)/cm2 -s-sr.
The EGRET PSF at 100 MeV, as previously noted, is ≈5.7◦ . In comparison, the
LAT PSF (68 % containment radius) at 100 MeV is ≈3.5◦ , with the PSF dropping to
≈0.6◦ for 1 GeV photons, and ≈0.15◦ for 10 GeV photons (for conversion in the thin
layers) [10]. The EGRET PSF at 100 MeV represents about π(5.7◦ )2 ≈ 100 square
degrees, or ≈1/400th of the full sky. Thus the flux from each patch corresponding
to the EGRET PSF is ≈5 × 10−7 ph (>100 MeV)/cm2 -s. A γ-ray source is signal
dominated for EGRET when its flux is 10−6 ph (>100 MeV)/cm2 -s. By contrast,
it is noise dominated when its flux is 10−6 ph (>100 MeV)/cm2 -s.
The time Δt needed to accumulate 100 photons with EGRET, consisting √ of 50
signal S and 50 background B photons, and to give a detection at the ≈S/ 2B ∼ = 5σ
level, is found through an expression for the integral photon flux F(>) of the source.
For a source at the level of
F(>100 )
F−8 ≡ −8
, (3)
10 ph(>100 MeV)/cm2 -s
the number of detected photons is ≈10−8 F−8 × Δt × 1000 cm2 , so that ≈50 ph
can be detected from a bright source at the level of F−8 ∼ 102 during an EGRET
observation period of Δt ≈ one day (during half this time, the Earth is occulted). A
nominal 2-week observation period with Δt ≈ 106 s on source gives a 5σ limiting
sensitivity for sources with F(>) ≈ 2 × 10−7 ph(>100 MeV)/cm2 -s.
These sorts of arguments can be used to estimate the time needed to make a
detection and resolve temporal variability with EGRET, LAT, and counter detec-
tors with broad FoVs, including neutrino telescopes. The Fermi LAT becomes noise
dominated at much lower flux levels than EGRET, with sources regularly detected
4 A number index Γγ = 2.41 ± 0.05 and intensity normalization I (>100 MeV) = (1.03 ± 0.17) ×
10−5 /cm2 -s-sr for the intensity spectrum of the isotropic diffuse γ-ray background are measured
from the first year Fermi data [7]. The lower diffuse extragalactic flux measured with Fermi compared
to EGRET is partially but not entirely due to resolving more point sources out.
234 C. D. Dermer
at F−8 < 1. Note that the background is smaller at higher energies as a result of
the smaller PSF, but the smaller flux usually reduces significance for a given obser-
vation time except for hard-spectrum sources. A better approach for characterizing
detection significance is likelihood analysis [21] (described below), but the detection
significance can be simply estimated as outlined above [12].
The above estimates apply to high galactic latitude, |b| > 10◦ sources where
the γ-ray sky is dominated by extragalactic sources and unresolved isotropic γ-ray
background. At lower galactic latitudes, the diffuse γ-ray emission from cosmic-ray
interactions with gas and dust makes source detection more difficult. Subtraction of
the diffuse emission and nonuniform and variable background requires a Galactic
model for cosmic-ray/gas interactions. The sensitivity of source detection to back-
ground model is seen in the alternate 3EGR analysis of the EGRET data by Casandjian
& Grenier [11]. They do not confirm 107 3EG sources, most in the vicinity of the
Gould Belt, and find 30 new sources in the full 9-year data set.
Long after EGRET’s effective lifetime had expired (though BATSE, OSSE, and
COMPTEL were still collecting valuable data), CGRO was deorbited into the Pacific
Ocean in 2000. The launch of INTEGRAL in October 2002, Swift on November 20,
2004, and AGILE on April 23, 2007, helped fill the γ-ray gap leading to Fermi.
1.2 Fermi Gamma-Ray Space Telescope
The Fermi Gamma-ray Space Telescope consists of two major detector systems,
namely
1. The Large Area Telescope (LAT), a pair-conversion telescope sensitive between
≈20 MeV and 300 GeV, with the higher energy limit a result of the vanishing
small detection rate given Fermi’s ∼1 m2 aperture. The Fermi-LAT has opened the
previously unexplored ≈10–100 GeV window, as self-vetoing due to particle
shower backsplash in the anti-coincidence detector may have reduced EGRET’s
effective area above ≈5 GeV [27]. Fermi nominally operates in a survey mode,
and given its large FoV, scans the entire sky every 3 h.
2. The Gamma-ray Burst Monitor (GBM), sensitive in the 8 keV–40 MeV range,
consisting of 12 NaI detectors sensitive between 8 keV and 1 MeV, and 2 BGO
detectors sensitive in the 0.15–40 MeV range. The scintillator detectors surround
the base of the LAT, and view the entire unocculted sky which, at the nominal
565 km altitude of Fermi, represents ≈2/3rd of the full sky.
Some of the science questions that Fermi was designed to answer are:
• How do supermassive black holes in Active Galactic Nuclei (AGN) create powerful
jets of material moving at nearly light speed? What are the jets made of?
• What are the mechanisms that produce GRB explosions? What is the energy bud-
get?
• What is the origin of the cosmic rays that pervade the Galaxy?
• How does the Sun generate high-energy γ rays in flares?

• How has the amount of starlight in the universe changed over cosmic time?
• What are the unidentified γ-ray sources found by EGRET?
GLAST Becomes Fermi
After a number of delays, a Delta II 7920-H(eavy) rocket carrying the Gamma-ray

Large Area Space Telescope—GLAST—payload was launched from Cape Canaveral
Air Station on 2008 June 11 at 12:05 pm EDT into a 565 km altitude circular orbit
with 25.6◦ inclination and a 96 min period. GLAST completed a 60 day checkout
period by early August, and released its first-light image on 26 August 2008, when
it was renamed the Fermi Gamma-ray Space Telescope, after the Italian-American
physicist Enrico Fermi (1901–1954).
The first-light image, based on only 4 days of observation, already reveals dozens
of sources to the eye, with the blazar 3C 454.3 almost as bright as the Vela pulsar,
which is the brightest persistent γ-ray source in the sky (F−8 ∼ = 1060). Limiting
fluxes of EGRET sources are at the level of F−8 ≈ 20–30 for a two-week observation,
with a corresponding all-sky flux limit of F−8 ≈ 15–30 for a year of observing (given
the FoV of EGRET). For a source with a flat ν Fν SED, or a number index Γγ = −2,
Fermi reaches F−8 ∼ 1 in one year over the entire high Galactic latitude |b| > 10◦
sky. Due to its energy-dependent effective area and PSF, limiting fluxes are strongly
dependent on both source and background spectrum. Hard-spectrum sources with
number index ≈ − 1.5 are detected with comparable significance as soft spectrum
sources, but at integral photon fluxes as low as F−8 ≈ 0.1.
The LAT operates in a nominal scanning mode whereby the spacecraft rocks about
the zenith. The rocking angle was equal to 39◦ in the first part of the Fermi mission,
and then increased to 50◦ after 3 September 2009. The larger rocking angle gives
a more uniform exposure, but with the loss of data due the increased fluorescence
γ-ray emission from cosmic-ray bombardment of the Earth’s atmosphere.5 The LAT
observes the entire sky every two orbits, or ≈3 h, by rocking north and south of
the zenith on alternate orbits, with each point in the sky receiving ≈30 min exposure
during this time. Onboard triggering has also been enabled for the Fermi spacecraft to
autonomously slew. As of October 2011, more than 30 autonomous repoint requests
(ARRs) have taken place, resulting in 5-h pointing mode observations in response
to bright GRBs detected with the GBM (the length of the ARR repoint was reduced
in the third year of the mission). Several dedicated Targets of Opportunity (ToO)
pointings have been executed, including one to 3C 454.3, two to the flaring Crab,
one to Cyg X-3, and two Solar pointings.6 A regular schedule of nadir pointings to
look for γ rays from terrestrial γ-ray flashes, or TGFs, is underway.
The Fermi/GLAST LAT Collaboration is an international organization originally
comprised of institutions in France, Italy, Japan, Sweden, and the United States.
5 This is also called, colloquially and inaccurately, “γ-ray albedo,” or just “albedo”.
6 http://fermi.gsfc.nasa.gov/ssc/observations/timeline/posting/
236 C. D. Dermer
There are about 120 full Collaboration members, and a total of about 200 sci-
entific members, including affiliated scientists, postdocs, and students. The prin-
cipal investigator of the project is Professor Peter Michelson of Stanford, and the
project scientist during the development, commissioning, and early science phase was
Dr. Steven Ritz, now professor at UC Santa Cruz. Dr. Julie McEnery of NASA’s
Goddard Space Flight Center is now Fermi LAT project scientist. A review of the
first-year LAT results is given by Michelson, Atwood, and Ritz [22], and the detailed
instrument description can be found in the paper by Atwood et al. (2009) [10].
In recognition of the importance of the Fermi mission to high-energy astronomy,
Peter Michelson, Bill Atwood, and the Fermi Gamma Ray Space Telescope LAT team
were awarded the 2011 Rossi Prize of the High Energy Astrophysics Division of the
American Physical Society, “for enabling, through the development of the Large
Area Telescope, new insights into neutron stars, supernova remnants, cosmic rays,
binary systems, active galactic nuclei, and gamma-ray bursts.” UCSC physicist Bill
Atwood was also recently awarded the 2012 W. K. H. Panofsky Prize in Experimental
Particle Physics by the American Physical Society.
LAT Instrument Description
The LAT is a pair-conversion telescope composed of 16 tracker-converter towers,

a calorimeter, an anticoincidence shield, and an electronics system to filter signal
from background. The conversion of γ-ray photons to electron-positron pairs takes
place when a γ ray produces a pair in one of the 16 high-Z tungsten planes in each
tower. The photon direction is measured with a precision silicon-strip tracker with a
228 µ pitch (separation between strips), totaling 8.8 × 105 channels distributed over
18 tracker planes in each of 16 tracker modules. The Si tracker follows the e+ -e−
pairs to give direction, and the photon energy is measured in a by a cesium-iodide
calorimeter in a hodoscopic assembly that images the shower, allowing for additional
rejection of non-photon-like events. A segmented anticoincidence detector (ACD)
rejects background of charged cosmic rays. Segmentation of the ACD helps prevent
self-vetoing at high energies (Fig. 2).
Table 1 gives the performance characteristics of the Fermi LAT. Depending on the
source type, different classes of events are adopted to maximize sensitivity. Three
hierarchical analysis classes based upon analysis cuts and event selections are gen-
erally used for Fermi analysis. For the most stringent diffuse event class used to
identify γ rays, the on-axis effective area increases from ≈1500 cm2 at 100 MeV to
≈8000 cm2 at 1 GeV, and is roughly constant at higher energies. The transient event
class is used for the study of GRBs and brief transients where the amount of back-
ground can be greater due to the smaller time over which data is taken. The source
class, intermediate to the other two, is where the residual instrumental plus envi-
ronmental background is comparable to the extragalactic diffuse emission measured
with EGRET.
Fig. 2 Schematic of the

Large Area Telescope on
Fermi (from Atwood et al.
2009 [10])
Table 1 Performance
Parameter Value or range
characteristics of the Fermi
LAT [10] Energy range 20 MeV–300 GeV
Effective area at normal incidence (60◦
off-axis)
100 MeV 3700 (700) cm2
1 GeV ≤8300 (2900) cm2
10 GeV ≤8400 (3100) cm2
Energy resolution (1σ, on-axis)
100 MeV–1 GeV 15 %–9 %
1 GeV–10 GeV 15 %–9 %
10 GeV–300 GeV 8.5 %–18 %
Single photon angular resolution (68 %
containment radius)
100 MeV 3.5◦
1 GeV 0.6◦
>10 GeV <0.15◦
Field of View (FoV) 2.4 sr
Timing accuracy 300 ns
Event readout time (deadtime) 26.5 µs
LAT Instrument Response
Even though two tracker towers and three calorimeter modules of the LAT were
beam tested, the original science tools provided for instrument response were the
Pass 6_v3 response functions based on extensive GEANT4 Monte Carlo simulations
of the satellite that followed secondary particles and photons through the digitization
and software filters to determine—assuming tracker azimuthal symmetry—effective
238 C. D. Dermer
area, energy uncertainty, and PSF (or containment radius). The event reconstruction,
filters, and different event classes are described in [10].
The P6_v11 corrections to the P6_v3 response functions using a few months of
on-orbit data [8] showed that the earlier response functions gave smaller angles for
68 % and 95 % containment radii above ≈5 GeV than given by the on-orbit calibra-
tion. The PSF is best fit with a King model profile of the form
−g
1 1 θ2
f King (θ, σ, g) = 1− 1+ 2 , (4)
2πσ 2 g 2σ g
where θ is the angle between the incident, “true” photon direction and the recon-
structed direction, and g is a fitting parameter. The wings of the PSF follow a power-
law rather than exponential behavior, and are well fit with the sum of two King
functions.
The first-year analyses and the 1FGL, and second-year analyses leading to the
2FGL, generally employ the P6 and P7 instrument response functions, respectively.
Updated instrument performance that improves on pre-flight Monte Carlo and muon
calibrations by using inflight corrections to the instrument response functions are
found at7 .
Utilization of Fermi-LAT to energies as low as 30 MeV is possible with the LAT
Low-Energy (LLE) technique [24]. This approach was developed for transients, such
as Solar flares and GRBs, where a direction is known and a short time window can be
defined. During this period, less discriminating event selections are used, the analysis
includes all photons with reconstructed arrival directions within 20◦ of the target,
and the criteria for >100 keV vetoing in the ACD is relaxed.
1.3 Energy, Flux, and Luminosity
Consistent with the notation of [4], we define the dimensionless photon energy, in
units of electron rest mass, as
hν Eγ
= 2
= . (5)
mec m e c2
Flux density Fν is usually reported in units of Jansky (1 Jy = 10−23 erg/cm2 -s-

Hz), so that the quantity ν Fν is an energy flux F (units of erg/cm2 -s, or Jy-Hz,
noting that 1010 Jy-Hz = 10−13 erg/cm2 -s). The luminosity distance d L for a steady,
isotropically emitting source is defined so that the energy flux F is related to the
source luminosity L ∗ (erg/s) according to the Euclidean expression
7 www.slac.stanford.edu/exp/glast/groups/canda/lat_Performance.htm and http://fermi.gsfc.nasa.

gov/ssc/data/analysis/documentation/Cicerone/Cicerone_LAT_IRFs
L∗
F= . (6)
4πd L2
If φ() is the measured spectral photon flux (units of photons cm−2 s−1 −1 ), then
ν Fν = m e c2 2 φ(). Henceforth we use the notation
f = ν Fν (7)
for the ν Fν flux. From the definitions of F and f ,

∞ f
F= d . (8)
0
Considering Eq. (6), the luminosity radiated by a source between measured photon
energies 1 and 2 , or between source frame photon energies 1 (1 + z) and 2 (1 + z),
is therefore given by
2 ln 2
L ∗ [1 (1 + z), 2 (1 + z)] = 4πd L2 m e c2 d φ() = 4πd L2 d(ln ) f . (9)
1 ln 1
Equation (9) shows that if the ν Fν spectrum is flat with value f 0 , corresponding to
a photon flux φ() ∝ −2 , then the apparent power of the source over one decade of
energy is ≈(ln 10)L 0 ≈ 2.30L 0 , where L 0 = 4πd L2 f 0 .
From these relations, one obtains the mean apparent isotropic γ-ray energy release
4πd L2 F Δt
E∗ = , (10)
(1 + z)
for a source at redshift z that releases average energy flux F during observing
timescale Δt. The apparent versus the absolute energy releases and luminosities
depends on the jet structure and variability behavior. For a steady, two-sided top-hat
jet with uniform emission within angle θ ≤ θ j of the jet axis, the absolute luminosity
L abs = f b L iso , (11)

1
where the beaming factor f b = 2 × 2π μ j dμ j /4π = 1 − μ j , and μ j = cos θ j .
For θ j = 0.1 (5.7◦ ), f b ∼
= 1/200, whereas if θ j = 0.01 (0.57◦ ), f b ∼
= 1/20000 =
−5
5 × 10 . What is reported is the apparent isotropic luminosity L ∗,iso = 4πd L2 F,
with absolute luminosity implied by arguments for the jet opening angle and beaming
factor.
240 C. D. Dermer
Variability Information
The Schwarzschild radius of a black hole of mass M is

2G M ∼ M
RS = = 3.0 × 10 5
cm ∼= 10−4 M9 pc, (12)
c2 M
defining M9 = M/(109 M ). Variations in the source flux by a large factor (2) over
a time scale Δt must, from causality arguments for a stationary source, originate from
an emission region of size R cΔt/(1 + z). Incoherent superpositions of emission
from larger size scales and from regions that are not in causal contact would usually
(though not always) wash out large-amplitude fluctuations. For high-quality data
from bright flares, large amplitude variations in source flux on timescale Δt would,
from this argument, imply a black-hole mass
(Δt/104 s)
M9 . (13)
1+z
This relation is even preserved in systems with relativistic outflows, unless specific
model-dependent provisions are made on the method of wind dissipation. Variability
timescales far shorter than the light-crossing time across an ≈109 M black hole
have been measured in the TeV BL Lac objects PKS 2155-304, Mrk 501, and Mrk
421, discussed in more detail below. Another interesting fiducial is the ratio of the
Eddington luminosity to the light-crossing time across a Schwarzschild black hole
[13], namely
L Edd m p c4 ∼
= 4π = 2.5 × 1043 erg s−2 . (14)
tS σT
We take this criterion as separating the extreme universe from the moderate universe.
During intense 3C 454.3 flaring activity [9], the ratio of the apparent isotropic lumi-
nosity and the source variability timescale strongly violate this limit, making this an
extreme event.
Extragalactic Background Light (EBL)
In intergalactic space, the energy density of the EBL is dominated by that of

the cosmic microwave background radiation (CMBR), with present temperature
TC M B = 2.72 K and energy density u C M B R ∼
= 0.25 eV/cm3 ∼ = 4 × 10−13 erg/cm3 .
The intensity of the EBL at infrared frequencies is difficult to measure directly
because of foreground radiations, such as Galactic electron synchrotron radiation
and zodiacal light scattered by dust in our Solar system. Photons observed at
10–100 GeV interact primarily with EBL photons at optical and UV energies. The
energy density of the dust and stellar components of the EBL is ≈10 % of the CMBR
energy density at the present epoch.
Fig. 3 Spectral energy densities in intergalactic space of various radiations, including the CMB,
the infrared (IR) and optical, X-ray, γ ray, and the extragalactic cosmic ray energy density. Also
shown is the energy density of cosmic rays measured near Earth; the transition energies between
the modulted and unmodulated, and galactic and extragalactic components remains uncertain
Figure 3 shows the energy density of photons in intergalactic space and cosmic
rays in outer space in the Solar cavity near Earth. Fermi measurements [7] give a lower
intensity for the diffuse extragalactic γ-ray background (EGB) than EGRET data [26,
28], probably due to EGRET miscalibration above ≈5 GeV [27].8 The cosmic-ray
particle spectrum is modulated at low, ∼GeV/nucleon energies by changing Solar
activity. Cosmic-ray source origin changes from a Galactic to extragalactic origin at
an energy scale currently under debate. The case for an extragalactic origin of the
UHECRs can be made by comparing Galactic size with the Larmor radius

E A m p c2 γ ∼ E/1020 eV
rL = = = 100 kpc (16)
QB Z eB Z BµG
of an ion with energy E, Lorentz factor γ, atomic charge Z and atomic mass A. The
characteristic magnetic field B = 10−6 BµG G of the Milky Way is a few µG in the
disk and probably much less in the Galactic halo.
8 The spectral energy density of the isotropic background radiation field at 100 MeV as measured
with EGRET is
4π 4π keV ∼ −18
u() = I ∼
= = 10 erg cm−3 . (15)
c c cm2 -s-sr
Because the I spectrum is essentially flat, the total energy density over the 100 MeV–100 GeV
energy range gives a factor ln 103 ∼
= 6.9; thus the EGRET energy density of the EGB is nearly 5
orders of magnitude below the present CMBR energy density.
242 C. D. Dermer
1.4 Limits to the Extreme Universe
The largest luminosity is obtained if the entire rest mass energy of the object with
mass M is transformed into energy on a light-crossing time scale for the gravitational
radius of the object, so that
Mc2 c5
L max = = = 3.6 × 1059 erg s−1 (17)
G M/c3 G
[23, 30]. The most luminous blazar source yet detected with the Fermi LAT is 3C
454.3, with apparent isotropic γ ray luminosity reaching L γ ≈ 2 × 1050 erg/s. The
most luminous GRBs detected with the Fermi-LAT, namely GRB 080916C and GRB
090510A, reached L γ ≈ 1053 erg/s.
2 Fermi Gamma-Ray Source Catalogs and Fermi Pulsars
Now we take a global view, and look at the Fermi Large Area Telescope First Source
Catalog (1FGL) [43] as a catalog, or systematic collection of objects, before focusing
on γ-ray emission from neutron stars and the first Fermi pulsar catalog [44]. The
1FGL catalog, taken with 11 months of data between 2008 August 4 and 2009
July 4, expands the 3 month bright source list (BSL, or 0FGL) [35], just as the First
LAT AGN Catalog (1LAC) [45], the subject of the next lecture, expands the LBAS
(LAT Bright AGN Sample) [36]. The 2FGL and 2LAC build on the 1FGL and 1LAC,
respectively. The 1FGL and 2 FGL catalogs are described in this section.
Source detection significance for the 1FGL and 2FGL catalogs is based on the
likelihood test statistic (T S). The likelihood ratio compares the likelihood of a null
hypothesis with the likelihood of the model. According to Wijk’s theorem, −2 times
the logarithm of this ratio approaches χ2 for large χ2 . The likelihood L is the prob-
ability of observing n i j counts in pixel i, j given the expected number of counts λi j
in pixel i, j. Poisson statistics gives
n
λi ji j exp(−λi j )
L= pi j = . (18)
ni j !
ij ij
Dropping a term independent of model, the logarithm of the likelihood is

log L = n i j log(λi j ) − λi j , (19)
ij ij
and T S = −2 log(L/L0 ) → χ2 , where L0 is the likelihood of the null hypothesis

for the given data set. Equation (19) gives a prescription for calculating T S when
analyzing γ-ray data. Mattox and colleagues [21] originally applied this method to
EGRET data.
The Fermi catalogs are based on a limiting test statistic, which is T S > 25 in
the 1FGL, 2FGL, 1LAC and 2LAC, and T S > 100 in the BSL and the LBAS. This
introduces systematic effects that should be taken into account when interpreting the
1FGL. For example, a cut on T S produces integral flux thresholds strongly dependent
on the source spectral hardness. Because of the steeply falling Galactic background
at high galactic latitudes, high-latitude hard spectrum sources will have a large T S
for much smaller integral fluxes than a soft-spectrum source detected at the same
significance level.
2.1 First Fermi Catalog of Gamma-Ray Sources: 1FGL
The 1FGL comprises 1451 sources with T S > 25, corresponding to a significance
4σ, and represents 21.22 Ms (or ≈73.3 % livetime) of data. Most of the deadtime
loss is due to passage through the SAA South Atlantic Anomaly (≈13 %) and to
readout deadtime (9.2 %). The exposure is uniform to within a factor ≈1.25 between
north and south and the detections are based on integrated data rather than shorter
bright periods or flares. An improved background and calibration model is used in
the 1FGL compared to the 0FGL.
The 1FGL is not a flux-limited survey. Although the exposure is relatively
uniform—an excellent feature of the GLAST design in nominal scanning mode—
there are large differences in integral photon fluxes of sources depending on source
location and the strong Galactic diffuse emission. Consequently, the 1LAC catalog is
drawn from sources above 10◦ latitude where the Galactic diffuse intensity is low.9
Any conclusions are weakened when source identifications are not complete, for
example due to the limits of a counterpart catalog used to make associations, or to
uncertainty in distance or lack of redshift, as this incompleteness can bias results.
The 1FGL catalog gives source location regions and association criteria defined
in terms of elliptical fits to the 95 % confidence regions, and power-law spectral
fits as well as flux measurements in five energy bands for each source. In addition,
monthly light curves are provided. Firm identifications with sources found in other
astronomical catalogs are based on correlated variability, e.g., rotation or orbital
period in the case of pulsars and X-ray binaries, or morphological similarities with
counterparts at other wavelengths, as in the case of SNRs.
For the catalogs and association criteria used, 630 of the sources in the 1FGL are
unassociated. Due to the sensitivity of the results to the model of interstellar diffuse
γ-ray emission used to model the bright foreground, 161 sources at low Galactic
latitudes towards bright local interstellar clouds are flagged as having properties that
are strongly dependent on the Milky Way gas model.
9 The Galactic diffuse flux at 10◦ ≤ |b| ≤ 20◦ , averaged over longitude, is still a factor of 2–3
greater than the extragalactic diffuse at higher latitudes [42].
244 C. D. Dermer
The principal source classes found in the 1FGL and listed in Table 2 are
• 3 high mass X-ray Binaries
• Rotation-powered and millisecond pulsars
• Supernova remnants
• Globular clusters
• >600 blazars
• 28 Non-blazar AGNs and Radio Galaxies
• Dozens of AGNs of uncertain type
• 2 starburst Galaxies
In addition, the Sun, moon, and Earth, hundreds of GBM GRBs, and more than a
dozen LAT GRBs had been detected by the time of publication of the 1FGL.
Association depends on the underlying catalog and the degree of confidence
assigned in the probability of association. For blazars, four catalogs with substantial
overlap were used, giving 689 1FGL associations with sources found in at least one
of these catalogs. A fuller discussion of the AGN catalog is deferred to the next lec-
ture. Though no extended radio lobes of radio galaxies were reported in the 1FGL,
the lobes of Centaurus A have since been imaged [46].
Table 2 LAT 1FGL and 2FGL source classes [43, 70]

Description Designator Assoc. (ID) Assoc. (ID)
1FGL 2FGL
Pulsar, X-ray or radio, identified by pulsations psr (PSR) 7 (56) 0 (83)
Pulsar, radio quiet (LAT PSR, subset of above) PSR 24– 25–
Pulsar wind nebula pwn (PWN) 2 (3) 0 (3)
Supernova remnant snra (SNR) 41 (3) 62c (6)
Globular cluster glc (GLC) 8 (0) 11 (0)
High Mass X-ray binary hxb (HXB) 0 (2) 0 (4)
Micro-quasar object: X-ray binary mqo (MQO) 0 (1) 0 (1)
(black hole or neutron star) with radio jet
Nova nov(NOV) 0 (1)
BL Lac type of blazar bzb (BZB) 295 (0) 428 (7)
FSRQ type of blazar bzq (BZQ) 274 (4) 353 (17)
Non-blazar active galaxy agn (AGN) 28 (0) 10 (1)
Radio galaxy 10 (2)
Active galaxy of uncertain type agu (AGU) 92 (0) 257 (0)
Normal galaxy gal (GAL) 6 (0) 4 (2)
Starburst galaxy sbg (SBG) 2 (0) 4 (0)
Seyfert galaxy sey (SEY) 5 (1)
Unassociated 630 576 + 1d
Total 1478b 1873e
a Indicatesa potential association with a SNR or PWN
b 779+630+(69) = 1409+(69) = 1478. Greater than 1451 because of multiple class assignments
c Some of the 62 sources may also be associated with PWNe
d 576 unassociated plus one with uncertain class
e 1169 + 577 + (127) = 1746 + (126) = 1873
Regarding other source classes in the 1FGL, 41 Fermi sources are associated with
SNRs or non-pulsed γ-ray emission, and three are sufficiently physically extended
that their morphology counts as an identification (W44 W51C, and IC 443; see
Lecture 5). Fermi reported three high-mass X-ray binary systems in the 1FGL, namely
LS 5039, LSI +61◦ 303, and Cyg X-3 (Lecture 7), and a fourth was discovered later. In
the 1FGL, it was unclear whether γ rays were made by massive O stars, but since then
η Carinae has been been established as a γ-ray source [49]. Ten BSL candidates out of
205 in the 0FGL, do not show up in the 1FGL, illustrating the highly variable nature
of some γ-ray sources (though note that all these sources are found in the Galactic
ridge, |l| < 60◦ , where background modeling is especially important). On top of
this, 630 sources were unassociated in the 1FGL. The unidentified Fermi sources
are clustered towards the plane of the Galaxy where diffuse background Galactic
emission and higher source density makes associations more tentative, though there
are still many high-latitude unidentified Fermi sources.
2.2 Second Fermi Catalog of Gamma-Ray Sources: 2FGL
The 2FGL catalog [70] was released at the time of writing. This source catalog was
derived from data taken during the first 24 months of the science phase of the mission,
which began on 2008 August 4. Source detection is based on the average flux over
the 24-month period. The 2FGL includes source location regions and fits to model
spectral forms. Also included are flux measurements in 5 energy bands and light
curves on monthly intervals for each source. Twelve sources in the 2FGL are found
to be spatially extended.
The 2FGL contains 1873 sources detected and characterized in the 100 MeV–
100 GeV range, of which 127 are firmly identified and 1170 are reliably associated
with counterparts of known or likely γ-ray source classes. Although the diffuse
Galactic and isotropic models used in the 2FGL analysis are improved compared to
the 1FGL catalog, caution flags for 162 sources indicate possible confusion, given
the uncertainty in the underlying diffuse model. Table 2 lists the number of sources
of various types in the 2FGL.
Some important improvements compared to the 1FGL catalog are:
1. The 2FGL catalog is based on data from 24 months of observations.
2. The data and Instrument Response Functions use Pass 7 event selections, rather
than the Pass 6 event selections used in the 1FGL.
3. The 2FGL employs a new, higher-resolution model of the diffuse Galactic and
isotropic emissions.
4. Spatially extended sources and sources with spectra other than power laws are
incorporated into the analysis.
5. The source association process has been refined and expanded.
246 C. D. Dermer
2.3 Fermi Pulsars
In the 1FGL, 56 pulsars are identified by their γ-ray pulsations, and another 7 associ-
ations are based on expectations from pulsar catalogs, e.g., Ė/d 2 ranking. Six 1FGL
sources are associated with pulsar wind nebulae (PWNe) that lack known pulsars,
and 8 1FGL sources are associated with globular clusters illuminated, most likely,
by the superposition of millisecond pulsar (MSP) γ-ray emissions. The number of
Geminga-like pulsars that lack detectable radio emission, has grown by a large factor.
There are 83 γ-ray pulsars, now all identified, in the 2FGL.
A catalog of 46 γ-ray pulsars is presented in the First LAT Pulsar Catalog,
Ref. [44]. This catalog includes 16 “blind-search” pulsars discovered by searching for
pulsed γ-ray emission at the position of bright LAT sources. Pulsed γ-ray emission
from 24 known pulsars were discovered using ephemerides derived from monitoring
radio pulsars, of which 8 are MSPs. The remaining 6 γ-ray pulsars were known
previously.
EGRET Pulsars
The Crab and Vela pulsars were known prior to CGRO, and Geminga was known
as a bright γ-ray point source. The pulsar nature of Geminga was only established
by detection of X-ray pulsations in ROSAT data early in the EGRET era [64]. The
EGRET pulsars, with EGRET fluxes above 100 MeV and 2FGL fluxes between
1 and 100 GeV, ordered by brightness, are listed in Table 3.
PSR 1951+32, located in the confusing Cygnus arm region, is associated with
the SNR CTB 80 discovered by Kulkarni et al. [68]. It has a phase-averaged ν Fν
spectrum rising ∝ 0.2 and peaking at ≈2 GeV, with a slower than exponential decline
at higher energies [51].
Table 3 γ-ray fluxes of EGRET pulsars [17, 44, 70]

Pulsar Period (ms) Age (kyr) EGRET Pulsar catalog 2FGL
P P/2 Ṗ F−8 F−8 F(1–100 GeV)a
0833-45, Vela 89.3 11.3 834.3 ± 11.2 1061 ± 7.0 135.8 ± 0.4
J0633+1746, Geminga 237 340 352.9 ± 5.7 305.3 ± 3.5 72.9 ± 0.3
0531+21,b Crab 33 1.25 226.2 ± 11.2 209 ± 4 18.3 ± 0.15
1706-44 102 17.6 111.2 ± 6.2 149.8 ± 4.1 19.1 ± 1.7
1055-52 197 540 33.3 ± 3.82 30.45 ± 1.7 5.0 ± 0.09
1951+32c 39.5 110 16 ± 2 17.6 ± 1.9 2.1 ± 0.07
1509-58,c Circinus 88.9 150 – 8.7 ± 1.4 1.45 ± 0.08
a Also in units of 10−8 ph cm−2 s−1
b Associated with SN 1054
c Pulsars not reported in the 3EG [71]
The Circinus pulsar was detected with COMPTEL but not EGRET, and is unusual
in having an inferred polar magnetic field B p = 15.77 TG, compared to ∼TG
(1012 G) fields for the others. Geminga is unusually close, at d ∼
= 160 pc [57],
whereas the others are more likely to be at ∼kpc distances.
Elementary Pulsar Physics
Over 1900 pulsars are known today, mostly through radio surveys. Pulsars are rapidly
rotating, highly magnetized neutrons star. The neutron stars themselves are formed by
explosions made by the collapsing cores of massive stars (core-collapse supernovae),
or through accretion-induced collapse of white dwarfs. Misalignment of the magnetic
axis and rotation axis makes in the simplest configuration a rotating dipole-field
geometry. Emission beamed along favorably defined field lines naturally makes a
wide range of pulse structures for different obliqueness and inclination and gap sizes.
The additional range of parameters associated with period and period derivative,
along with poorly understand radiation physics, allows for purely empirical and
kinematic pulse profile fitting.
Neutron stars are predicted to have masses near 1.4M and radii ∼15 km. The
two properties of the pulsar that can be very precisely measured are the period P and
period derivative Ṗ. Besides these observables, theory gives the mass of the neutron
star, M N S , and its radius R N S [77]. The uncertain equation of state of nuclear matter
determines whether neutron stars with given masses can exist, whether collapse to a
black hole occurs, or whether other degenerate quark phases exist.
Electron degeneracy pressure in a white dwarf cannot support the gravitational
force of a degenerate core exceeding the Chandrasekhar mass MC = 1.4M . Either
through nuclear burning or accretion, neutron stars can be formed. Neutron stars with
masses M > MC can test black-hole formation theory, but the uncertainty in orbital
inclination usually makes for large uncertainties in the neutron-star mass. By means
of radio timing measurements of the Shapiro delay10 of the binary millisecond pulsar
J1614-2230, a 1.97 ± 0.04M pulsar mass was recently measured [58], which rules
out some exotic hyperon and boson condensate equations of state.
Noting that velocity v = Ω × R, and the angular speed Ω = 2π/P, then the
light-cylinder radius at which the speed v = c is
R LC = Pc/2π. (20)
Even the rotation of a simple misaligned dipole field leads to unusual geometric
features, for example, the footprint on the neutron star surface of the field lines that
open into the light cylinder [60].
An expression for the surface polar magnetic field can be obtained by equating
the rotational spindown energy loss rate with the magnetic dipole radiation power.
10 The Shapiro delay is a regular general-relativistic change in the light travel time from the pulsar
as the radio photon travels through the gravitational field of the binary system; see [81].
248 C. D. Dermer
The former is
dE r ot d 1 4π 2
− = I Ω2 = I Ṗ, (21)
dt dt 2 P3
with the moment of inertia I depending on the mass and equation of state of neutron
star matter. This implies a characteristic age
τ = P/2 Ṗ, (22)
for P much longer than the initial spin period. The magnetic dipole power can be
estimated from the Poynting flux of a dipole radiation at the light cylinder, namely,
dE md B 2 (R LC ) 1 R 6N S B N2 S
− = 4π R 2LC c = B N2 S c∝ . (23)
dt 8π 2 R 4LC P4
Thus
BN S ∝ P Ṗ. (24)
The magnetic field B LC at R LC is therefore

1/2
24π 4 I Ṗ
B LC = . (25)
c3 P 5
We can imagine a time-independent configuration for a spinning, magnetized

neutron star if we wait long enough until all current flows have asymptotically relaxed
to their steady-state values. This defines the condition of a force-free magnetosphere,
where the vanishing of the Lorentz force F = Q[(v/c) × B + E)] = 0 implies the
existence of a magnetosphere filled with plasma with Goldreich-Julian [62] density
Ω·B
ρGJ = − . (26)
2π
Pulsar models build on these elementary concepts. For generic polar cap mod-
els, charge depletion in the polar field lines that open to the light cylinder, compared
with the Goldreich-Julian density, generates strong electric fields that induce vacuum
breakdown and current flow. In outer gap models, strong fields are generated at the
gaps created by the surface defined by the condition Ω · B = 0. The primary γ-ray
production and attenuation processes are curvature radiation, Compton radiation, and
magnetic pair production. Slot gap models study a more sophisticated realization of
the pair-starved region and how a slot gap forms along field lines with different accel-
eration potentials. Besides discriminating between magnetospheric models, pulsar
studies help disentangle the geometry of a pulsar. For more on pulsar models, see
[56, 60, 76].
Properties of Fermi Pulsars
Two different approaches are taken to discover pulsars, either by period folding with
data from a pulsar established at other frequencies, or to perform a blind search for
the pulsation. In the first method, the timing parameters of pulsars already known
at radio or X-ray frequencies are used to search for evidence of γ-ray pulsations
in the γ-ray data. For the blind-search technique, spectral power is calculated from
time-tagged γ-ray photon event data by searching through likely values of P and Ṗ
values. This is extremely computationally intensity, and not feasible for MSPs found
in binary systems. More computationally efficient techniques compute the arrival
time differences of γ rays, which cluster at values corresponding to multiples of
the pulse period [55]. For γ-ray astronomy, with limited photon statistics, only the
time-differencing technique for blind searches is effective at finding pulsars.
The first blind-search detection with Fermi was a pulsar with period P = 0.317 s in
the shell of the SNR CTA1, coincident with unidentified X-ray and 3EG J0010+7309
γ-ray sources [34]. The spin-down power of this pulsar is ≈4.5 × 1035 erg s−1 , and
the pulsar’s inferred age is 14000 years, which is consistent with the SNR age. Of
the 46 Fermi pulsars reported in the First Fermi Pulsar Catalog, 16 of them were
found using blind-search techniques in sources that are positionally coincident with
unidentified EGRET sources and supernova remnants [37]. These 16 pulsars are all
young and highly magnetized, with inferred magnetic fields between ≈1–10 TG.
The rotational energy-loss rates of pulsars in the First Fermi Pulsar Catalog range
from ∼3×1033 erg s−1 to 5×1038 erg s−1 , with the young, highly magnetized pulsars
typically having rotational energy-loss rates exceeding 1035 erg s−1 . Comparing
with the phase-averaged apparent isotropic γ-ray luminosity implies efficiencies
for the conversion of spin-down power to γ-ray energy in the range from ∼0.1 %
to ≈100 %. About 75 % of the γ-ray pulses have two peaks, but a third emission
structure, P3, shows up in the Vela pulse profile and moves to later phases with
increasing photon energy [41]. As Fig. 4 shows, the main peak, P1, which is dominant
from at least UV to GeV energies, becomes less intense compared to P2 at multi-GeV
energies.
The Fermi LAT phase-averaged spectrum of γ-ray pulsars can be well fit by a
generalized exponentially cutoff power law, given by
N (E) ∝ E −Γγ exp[−(E/E c )b ], (27)
with hard photon number indices Γγ generally in the range 1 Γγ 2, and

cutoff energies E c between ≈1 and 5 GeV. This form has been found to apply to all
types of pulsars, whether radio or γ-ray selected, or normal or millisecond pulsars.
For Vela itself, as seen in Fig. 5, Γγ = 1.38 ± 0.03, E c = 1.36 ± 0.15 GeV, and
b = 0.69 ± 0.02 (quoting statistical errors only) [50]. The sub-exponential (b < 1)
rather than super-exponential (b > 1) cutoff that would be produced if the γ rays were
made deep within the polar cap of the neutron star magnetosphere, and the detection
of emission at tens of GeV, is evidence against a polar-cap model. A phase-averaged
spectrum fitted with Eq. (27) is the composite phase-resolved spectra that individually
250 C. D. Dermer
Fig. 4 Vela light curves at optical, X-ray, and γ-ray energies [41], binned to 0.01 of the pulsar
phase. The main peaks P1, P2 and P3 are labeled in the top right panel. The bottom left panel shows
the 8–16 keV RXTE light curve [65] along with the radio pulse profile (dashed lines). At lower
right, the 4.1–6.5 eV HST/STIS NUV light curve [75] is shown
can be fit by simple exponential behaviours [50, 51]. Significant variations of cutoff
energy with phase are needed to reproduce the phase-averaged spectra.
Millisecond Pulsars and Globular Clusters
Pulsed γ-ray emission was detected with the Fermi LAT from J0030+0451, making
it the first firm detection of a MSP in γ rays [40], although there was a marginal
EGRET detection of PSR J0218+4232 [67] that has been confirmed with the LAT
[40]. Nine months into science operations, the Fermi-LAT Collaboration reported 8
γ-ray MSPs [39], establishing a new population of γ-ray emitters. As noted above,
it is not computationally feasible to perform blind searches for binary MSPs, which
introduces too many possibilities into the timing solutions. This makes it harder to
Fig. 5 Phase-averaged spectrum for 0.1 < E(GeV) < 60 [50]
know what fraction of the unidentified sources are MSPs, or if there are any radio-
quiet MSPs.
Globular clusters are quasi-spherical stellar systems consisting of hundreds of
thousands of old (ages of ≈1010 years) metal-poor stars. The high-stellar density at
the centers of these systems makes for unusual multi-body encounters. According to
the most common scenario, binary neutron-star systems in low-mass X-ray binaries
can be spin up by accretion over long time spans to periods of a few ms, provided that
the polar surface magnetic field is 109 G, which is ∼3 orders of magnitude smaller
than the typical TG fields of young γ-ray pulsars. Even though they have very different
magnetic field and rotation rates, the γ-ray spectra of MSPs and young pulsars are
similar, and well-represented by a hard power-law with a modified exponential cutoff
given by Eq. (27).
The Fermi-LAT collaboration has reported the detection of the famous globular
cluster 47 Tucanae at a significance of 17σ [38]. The γ-ray spectral shape is consistent
with a large populations of MSPs. With a typical luminosity of 3 × 1033 erg s−1
in γ-rays, ∼20–40 MSPs are required to give the measured γ-ray luminosity. Using
587 days of data, the number of Fermi sources associated with globular clusters has
grown to 8 [47], and now 11 in the 2FGL [70].
Both γ-ray detected “garden-variety” pulsars and MSPs have the highest values
of magnetic field at the light cylinder, B LC , suggesting that similar emission mech-
anisms operate. The γ-ray luminosity
√ grows with spin-down energy L γ ∝ Ė at
Ė 10 erg s and L γ ∝ Ė at Ė 1035 erg s−1 , with large scatter in this
35 −1
relation due, at least, to distance uncertainties (Fig. 6 in [44]).
Pulsar Wind Nebulae
Given the accuracy of timing parameters for a pulsar, the rotational energy-loss rate
is known to the uncertainty of the moment of inertia of the neutron star. The spin
252 C. D. Dermer
energy not coming out as γ rays, which usually accounts for no more than ∼10 %
of the spin-down energy, must come out in a different form, e.g., as field energy
in the form of a relativistic Poynting wind with such small baryon loading. The
wind Lorentz factors attains values of ∼108 . The interaction of the outflowing MHD
wind with the ISM makes a termination shock where additional particle acceleration
can take place. The termination shock separates the cold upstream MHD wind with
a turbulent flow downstream into the ISM. The boundary between the wind and
ISM is highly structured because of various fluid instabilities. Cold particles can be
directly injected into the ISM with the wind Lorentz factor, or accelerated at the
termination shock.
Identification of pulsar wind nebulae (PWNe) with the Fermi LAT depends on
finding steady extended GeV radiation. Three PWNe were reported with Fermi LAT
using 16 months of survey data, namely the Crab nebula, Vela-X, and the PWN
inside MSH 15-52 [53]. Searching in the off-pulse emission of the pulsar yields a
candidate PWN related to PSR J1023-5746 and coincident with HESS J1023-575.
The sources with GeV-detected PWNe have the highest spin-down flux and tend to
have the highest spin-down power, with Ėr ot 1037 erg s−1 .
Crab Nebula and Flares
The Crab pulsar and nebula is associated with a supernova that took place in 1054 CE
at a distance of ≈2 kpc. It is the best example of a center-filled plerionic SNR, where
the interior is illuminated by synchrotron emission powered by the wind of a young,
33 ms neutron star. The spin-down energy of the Crab pulsar is 4.6 × 1038 erg s−1 ,
of which ≈30 % is converted into the radiant power of the nebula. The mean spectral
index of the nebula in the LAT range is ∼ = −4 between ≈100 and 400 MeV, hardening
to an index of ∼ = − 1.64 at higher energies [48] (see Fig. 6). The harder, high energy
component is the synchrotron self-Compton component (predicted by Gould [63]),
and well-explained as SSC if the mean magnetic field is ≈200 µG [48, 54].
Year-scale variations by a several percent are found in Crab nebular fluxes mea-
sured with RXTE, Swift BAT, and INTEGRAL. The fluxes also declined during
the first two years of the Fermi mission, consistent with measurements of an ≈7 %
decrease between 12 and 500 keV as measured with the GBM [82]. In the period
between 2008 August and 2009 April, no variations in the 100 MeV–30 GeV flux
were found with the LAT [48].
Although some evidence for flux variations had been noted during the EGRET
era [59], Fermi-LAT detected 100 MeV flaring episodes from the Crab lasting for
16 days in February 2009 and 4 days in September 2010 [52]. The September 2010
flare was first announced by the AGILE team [78], and they also reported an earlier
October 2007 flare [79]. Typical γ-ray flare luminosities are ≈4 × 1036 erg/s, and
reach an integral photon number flux above 100 MeV that is ≈10× brighter than the
mean flux.
The PSF of front events of the Fermi-LAT is ≈0.12◦ –0.15◦ at E 10 GeV,
so imaging can be achieved to ≈0.12◦ /8 ∼ 1 (at 2 kpc, 1 and 1 correspond to
Fig. 6 COMPTEL 0.7–30 MeV data (gray filled squares) and long-term average Fermi-LAT
100 MeV–300 GeV γ-ray data (open circles), and February 2009 (open squares) and September
2010 (filled squares) flares of the Crab nebula [52]
≈0.6 pc and ≈3 × 1016 cm, respectively). The extent of the radio nebula is ≈10 ,
and the extent of the Chandra X-ray torus from which perpendicular jets emerge is
≈1.3 . A bright inner X-ray ring in the Chandra image is ≈0.5 in extent. Zooming
into the central region with both Chandra and HST11 reveals a dynamic wind, wisps,
and ∼1 knots that brighten and dim. For comparison, one light day corresponds
to ≈3 × 1015 cm, far smaller than the smallest resolvable knots, yet the γ-ray flares
radiate ≈1 % of the spin-down energy.
The crucial feature of the Crab spectrum may well be that the quiescent spectrum
displays a strong softening at ≈20 MeV. This value is a factor ≈10 smaller than the
maximum synchrotron frequency at ≈200 MeV obtained by balancing the timescale
for electron synchrotron losses with the Larmor timescale on which particles gain
energy through Fermi particle acceleration mechanisms (see Sect. 5.5), for a non-
relativistic flow. A mildly relativistic flow can enhance the emission by large factors
and could be compatible with changes in the flow profile at the termination shock
when the ultra-relativistic wind slows to mildly relativistic speeds. Considerable
theoretical interest has focused on perturbations of the flow induced by statistical
fluctuations [83], by the generation of large amplitude MHD waves [69], and by
outflowing knots formed at an oblique termination shock [66]. No evidence for the
flare energy generation is found in the pulsar spin-down behavior [52].
11 http://chandra.harvard.edu/photo/2002/0052/animations.html
254 C. D. Dermer
Pulsar Physics with Fermi
Three years after launch, over 100 γ-ray pulsars are known, and γ-ray emission is
associated with 8 globular clusters counting 47 Tuc [47]. In NGC 6624, a single
MSP dominates the γ-ray energy output [61]. The γ-ray pulsars are divided into 31
radio-selected MSPs, 37 blind search γ-ray pulsars, and 38 young radio-selected γ-
ray pulsars. Of the 37 blind search pulsars, extensive follow-up observations reveal
pulsed radio emission in only 3 cases. Radio followup of Fermi pulsars is described,
e.g., in [72].
Some of the interesting open questions relate to the field geometry and radiation
mechanisms for pulsar γ rays. The significant fraction of Geminga-like blind-search
pulsars is generally interpreted in terms of a larger γ-ray than radio cone. The detec-
tion of pulsed emission in the Crab to ≈125 GeV [80] means that this emission has
to be made high in the magnetosphere. Emission to these energies requires extreme
parameters for a curvature radiation origin, and may imply Compton-scattering in
the Crab’s magnetosphere. (The curvature radiation mechanism remains a feasible
mechanism for pulsars older than the Crab.) The L γ ∝ Ė a behaviour is not well-
explained from the lowest MSP powers to the highest pulsar powers. Pulsar γ-ray
emission can make a significant, tens of %, fraction of the Galactic diffuse emission.
This fraction could be very different in early-type galaxies with only MSPs now
active.
For reviews of Fermi pulsars, see [73] and [74].
3 Fermi AGN Catalogs
The discovery of the γ-ray blazar class is one of EGRET’s lasting legacies. Of the 66
high-confidence sources associated with AGNs in the 3EG catalog [17], all except
one—the nearby radio galaxy Centaurus A—were associated with blazars.
Blazars are sources that exhibit violent optical variability (e.g., changing in flux
by ∼50 % in one day), high optical polarization (exceeding several percent), and flat
radio spectra with radio spectral index αr < 0.5 at GHz frequencies (Fν ∝ ν −αr ).
Superluminal motion at radio frequencies and highly luminous and variable γ-ray
emissions are also typical blazar properties. Blazars themselves are interpreted to
be relativistic jet sources powered by a supermassive black hole, like radio galaxies,
though with the observer looking nearly along the jet axis. The variety of multi-
wavelength SEDs displayed by blazars and their misaligned populations can often,
though not exclusively, be attributed to orientation effects amplified by the Doppler
boosting of the jetted radiation (Sect. 4).
3.1 LAT Bright AGN Sample (LBAS) and First LAT AGN
Catalog (1LAC)
Three lists of AGNs detected with Fermi have now been published by the Fermi
Collaboration. These are the LAT Bright AGN Sample (LBAS) [36], and the First
and Second LAT AGN Catalogs, 1LAC [45] and 2LAC [92], respectively.
The LBAS is based on 3 months of science observations taking place between
2008 August 4 and 2008 October 30, and consists of 106 high Galactic latitude (|b| >
10◦ ) sources associated with AGNs. These sources have a test statistic T S > 100,
corresponding to 10σ significance, and are a subset of the 205 sources listed in the
BSL [35].
By comparison, the 3EG [17] and the EGR [11] list 31 sources with significance
>10σ, of which 10 are at high latitude. Remarkably, 5 of the >10σ EGRET sources
are not found in the BSL. These are the flaring blazars NRAO 190, NRAO 530,
1611+343, 1406-076 and 1622-297, the latter of which is the most luminous blazar
detected with EGRET [108].
The 1LAC [45] is a subset of the 1451 sources in the 1FGL [43] derived from
analysis of the first 11 months of LAT science data. There are 1043 1FGL sources at
high latitudes, of which 671 are associated with 709 AGNs, with the larger number
of AGNs than 1LAC sources due to multiple associations. Associations are made
by comparing the localization contours with counterparts in various source catalogs,
for example, the flat-spectrum 8.4 GHz CRATES (Combined Radio All-Sky Targeted
Eight GHz Survey; [104]) and the Roma BZCAT blazar catalog [109]. The probability
of association is calculated by comparing the likelihood of chance associations with
catalog sources if randomly distributed.
Of the 671 associations, 663 are considered “high-confidence” associations due
to more secure positional coincidences. The “clean” sample is a subset of the high-
confidence associations consisting of 599 AGNs with no multiple associations or
other analysis flags, for example, evidence for extended emission. As listed in
Table 4, these subdivide into 275 BL Lac objects, 248 flat spectrum radio quasars,
26 other AGNs, and 50 AGNs of unknown types. The “New Classes” category
contains non-blazar AGNs, including starburst galaxies and various types of radio
galaxies, e.g., narrow line and broad line. An AGN is classified as an “unknown”
type either because it lacks an optical spectrum, or the optical spectrum has insuf-
ficient statistics to determine if it is a BL Lac objects or a flat spectrum radio
quasar (FSRQ). In comparison with the 671 AGNs in the 1LAC, EGRET found
66 high-confidence (>5σ) and another 27 lower-confidence detections with sig-
nificance between 4σ and 5σ, as noted earlier. Thus the 1LAC already repre-
sents an order-of-magnitude increase in the number of AGNs over EGRET. There
are ≈300 unassociated and therefore unidentified high-latitude Fermi sources in
the 1LAC.
256 C. D. Dermer
Table 4 Classes of γ-ray emitting AGNs and galaxies in the 1LAC and 2LAC “clean” samples
Class Number in 1LAC Characteristics Prominent members
(2LAC)
All 599 (885)
BL Lac objects 275 (395) Weak emission lines AO 0235+164
syn
…LSP 64 (61) ν pk < 1014 Hz BL Lacertae
syn
…ISP 44 (81) 1014 Hz < ν pk < 1015 Hz 3C 66A, W Comae
syn
…HSP 114 (160) ν pk > 10 Hz
15 PKS 2155-304, Mrk 501
FSRQs 248 (310) Strong emission lines 3C 279, 3C 354.3
…LSP 171 (221) PKS 1510-089
…ISP 1 (3)
…HSP 1 (0)
New Classesa 26 (24)
…Starburst 3 (2) Active star formation M82, NGC 253
…MAGN 7 (8) Steep radio spectrum AGNs M87, Cen A, NGC 6251
…RL-NLS1s 4 (4) Strong FeII, narrow PMN J0948+0022
permitted lines
…NLRGs 4 (–)c Narrow line radio galaxy 4C +15.05
…other sourcesb 9 (11)
Unknown 50 (156)
a Total adds to 27, because the RL-NLS1 source PMN J0948+0022 is also classified as FSRQ in
the 1LAC
b Includes PKS 0336-177, BZU J0645+6024, B3 0920+416, CRATES J1203+6031, CRATES
J1640+1144, CGRaBS J1647+4950, B2 1722+40, 3C 407, and 4C +04.77 in 1LAC clean sample
c Class designation deprecated in 2LAC
3.2 Classification of Radio-Emitting AGNs and Unification
Different classes of extragalactic AGNs are defined according to observing frequency.

We have already noted the association of Fermi sources with BL Lac objects and
FSRQs, which are based on an optical classification. The precise definition used by
the Fermi team is that an AGN is a BL Lac object if the equivalent width of the
strongest optical emission line is <5 Å, and the optical spectrum shows a Ca II H/K
break ratio <0.4 in order to ensure that the radiation is predominantly nonthermal
(the Ca II break arises from old stars in elliptical galaxies). The wavelength coverage
of the spectrum must satisfy (λmax − λmin )/λmax > 1.7 in order that at least one
strong emission line would have been detected if present. This helps guard against
biasing the classification for AGNs at different redshifts where the emission lines
could be redshifted out of the relevant wavelength range. For sources exhibiting BL
Lac or FSRQ characteristics at different times, the criterion adopted is that if the
optical spectrum conforms to BL Lac properties at any time, then it is classified as a
BL object.
The criterion for classification of radio galaxies according to their radio prop-
erties stems from the remarkable correlation between radio morphology and radio
luminosity [101]. The twin-jet morphology of radio galaxies is seen in low-power
radio galaxies, whereas the lobe and edge-brightened morphology is found in high-
power radio galaxies, with a dividing line at ≈2 × 1025 W/Hz-sr at 178 MHz, or at
a radio luminosity of ≈2 × 1041 erg s−1 for the current cosmology. Besides a radio-
morphology/radio-power classification, radio spectral hardness can also be used to
characterize sources as flat-spectrum and steep-spectrum sources. Furthermore, radio
galaxies can be subdivided according to the widths of the optical emission lines into
broad- and narrow-line radio galaxies. Correlations between radio-core dominance
and γ-ray luminosity supports a scenario where the jet γ-ray flux is greatest along
the jet direction [107].
Blazars and radio galaxies can also be classified according to their broadband SED
when there is sufficient multiwavelength coverage to reconstruct a spectrum from
the radio through the optical and X-ray bands and identify a ν Fν peak frequency
syn
ν pk of the lower energy, nonthermal synchrotron component of the spectrum (see
syn
Sect. 4). When the peak frequency ν pk of the synchrotron component of the spectrum
is <1014 Hz, then a source is called low synchrotron-peaked (LSP), whereas if the
syn
SED has ν pk > 1015 Hz, then it is referred to as high synchrotron-peaked (HSP).
syn
Intermediate synchrotron-peaked (ISP) objects have 1014 Hz < ν pk < 1015 Hz.
SEDs of the bright Fermi LBAS sources are constructed in [86]. Essentially all
FSRQs are LSP blazars, whereas BL Lac objects have large numbers in all LSP, ISP,
and HSP subclasses.
According to the standard unification scenario for radio-loud AGNs [117], radio
galaxies are misaligned blazars, and FR1 and FR2 radio galaxies are the parent
populations of BL Lac objects and FSRQs, respectively. To establish this relationship
requires a census of the various classes of sources that takes into account the different
beaming properties for the Doppler-boosted radiation of blazars. Even if analysis of
data of radio galaxies and blazars supports the unification hypothesis, this paradigm
still does not explain the reasons for the differences between radio-quiet and radio-
loud AGNs, or between BL Lac objects and FSRQs.
New classes of extragalactic Fermi sources found in the 1LAC include starburst
galaxies (Sect. 6.1), narrow line radio galaxies (NLRGs), radio-loud narrow-line
Seyfert 1s (RL-NLS1s), and radio-quiet AGNs. Five NLRGs are reported in the
1LAC. These objects have narrow emission lines in their optical spectrum, suggesting
that they are observed at large angles with respect to the jet direction, with the
surrounding dust torus obscuring the broad line region (BLR).
RL-NLS1s have also been recently established as a γ-ray source class [84].
These objects show narrow Hβ lines with FWHM line widths 1500 km s−1 , weak
forbidden lines ([O I I I ]/H β < 3) and an Fe II bump [111]. By comparison with the
∼109 M black holes in blazars, the host galaxies of RL-NLS1s are spirals that have
nuclear black holes with relatively small (∼106 –108 M ) mass that accrete at a high
Eddington ratio. The detection of these objects challenges scenarios (e.g., [114])
where radio-loud AGNs are hosted by elliptical galaxies that form as a consequence
of galaxy mergers.
The 1LAC includes 10 associations with radio-quiet AGNs. In 8 of these cases,
at least one blazar, radio galaxy, or CRATES source is also found close to the γ-ray
258 C. D. Dermer
source. In the remaining two cases, the association probabilities are weak. Thus none
appear in the 1LAC “clean” sample. In some of these candidate radio-quiet γ-ray
sources, such as the Sy 2 galaxies NGC 4945 or NGC 1068 [106], which are also
starburst galaxies, the γ rays could be made by cosmic-ray processes rather than from
a radio-quiet Sy nucleus. More extensive searches for γ rays from Swift-BAT AGNs
has not established, however, that radio-quiet AGNS are GeV γ-ray emitters [94].
3.3 Properties of Fermi AGNs
Various correlations are found by comparing γ-ray properties of Fermi AGNs

according to their radio, optical, or SED classification. Probably the most pro-
nounced correlation is between the >100 MeV γ-ray spectral index Γγ and opti-
cal AGN type. FSRQs have significantly softer spectra than BL Lac objects, with
Γγ ∼= 2.40 ± 0.17 for FSRQs and Γγ ∼ = 1.99 ± 0.22 for BL Lac objects in
the LBAS [36]. The SED classification shows that the mean γ-ray spectral index
Γγ ∼= 2.48, 2.28, 2.13, and 1.96 when the class varies from FSRQs to LSP-BL
Lacs, ISP-BL Lacs, and HSP-BL Lacs, respectively. The progressive hardening from
FSRQs to BL Lac objects can be seen in Fig. 7, which also compares with values for
radio galaxies and star-forming galaxies [89].
Fermi data reveal complex γ-ray blazar spectra. FSRQs and LSP-BL Lac objects,
and most ISP blazars with sufficiently good statistics, show breaks in the ≈1–10 GeV
range [88]. This was already apparent from the first observations of the bright blazar
3C 454.3 [84], to be discussed in more detail below. The HSP blazars, though, are
generally well-described by a flat or rising ν Fν SED in the GeV range, with ν Fν
peak frequencies between ≈100 GeV–TeV energies implied by VHE data.
Only 121 out of 291 BL Lac objects had measured redshifts at the time of publi-
cation of the 1LAC. For sources with measured redshift, BL Lac objects are mostly
found at low redshift, z 0.4, with only a few HSP BL Lac objects at higher red-
shifts. By contrast, the FSRQs span a wide range from z ≈ 0.2 to the highest redshift
1LAC blazar with z = 3.10.
This significant redshift incompleteness hampers interpretation of AGN proper-
ties, in particular, Fig. 7, which can only display sources with known redshifts. For
these sources, the hard-spectrum BL Lac objects typically have much lower L γ than
the FSRQs. This divide has been interpreted as a change in the accretion regime at
≈1 % of the Eddington luminosity [100]. In addition, the nearby radio galaxies with
z 0.1 inhabit a separate portion of the Γγ versus L γ plane, and are characterized by
lower γ-ray luminosities than their parent populations. The two more distant steep
spectrum radio sources, 3C 207 (z = 0.681) and 3C 380 (z = 0.692), and the FR2
radio galaxy PKS 0943-76 (z = 0.27) fall, however, within the range of γ-ray lumi-
nosities measured from FSRQs. Indeed, steep spectrum radio sources are thought to
be slightly misaligned FSRQs.
Fig. 7 Gamma-ray spectral slope Γγ of BL Lac objects (open blue circles), FSRQs (open black
squares), FR1 radio galaxies (red circles), FR2 radio sources (green squares), and star-forming
galaxies (magenta diamonds), are plotted as a function of their 100 MeV–5 GeV γ-ray luminosity L γ
3C 454.3 and FSRQs
The FSRQ 3C 454.3, at z = 0.859, underwent giant flares and became the brightest
γ-ray source in the sky for a week in 2009 December and 2010 April [91], and again in
2010 November [9]. The outburst in 2010 April triggered a pointed-mode observation
by Fermi. During the December outburst, its daily flux reached F−8 = 2200(±100),
corresponding to an apparent isotropic luminosity of L iso ≈ 3×1049 erg s−1 , making
it the most luminous blazar yet detected with Fermi. In its 2010 November outburst,
it reached F−8 ∼ = 6000 and L iso ≈ 1050 erg s−1 , becoming ≈5 times brighter than
the Vela pulsar.
Figure 8 shows the light curve of 3C 454.3 [9], which can also be found at the
public website of Fermi monitored sources.12 The fluxes are plotted in durations
of one day over the course of the Fermi mission. The flux rises to a plateau level
preceding γ-ray outbursts, with the 2008 July outburst showing strong resemblance
to those in 2009 August and December, and 2010 April and November. Intense
flaring occurs during periods of enhanced activity, which is to say that the flares are
not isolated events, but seem to occur during periods of enhanced accretion activity.
As noted already in the initial Fermi report [84], the spectrum of 3C 454.3 breaks
strongly by 1.2(±0.3) units at E br ≈ 2 GeV. Such a break is inconsistent with
12 http://fermi.gsfc.nasa.gov/ssc/data/access/lat/msl_lc/
260 C. D. Dermer
Flux [E>100MeV] (10-5 ph cm-2 s-1)
Flux [E>100MeV] (10-5 ph cm-2 s-1)

6
1
10-1
4
3 10-2
48000 49000 50000 54200 54400 54600 54800 55000 55200 55400 55600 55800
Time (MJD)
0
54600 54800 55000 55200 55400 55600
Time (MJD)
Fig. 8 Fermi LAT daily light curve of 3C 454.3 [9], showing the giant flares in December 2009,
April 2010, and November 2010 (MJD 55200 corresponds to 4 January 2010). Inset shows the light
curve on a logarithmic scale, with black points from EGRET [17], and blue points from AGILE
[115]
simple radiative cooling scenarios, which predict a break by 0.5 units. The more
recent analyses of data from the major outbursts of 3C 454.3 [9, 91] confirm the
strong spectral break and finds that E br is very weakly dependent on the flux state,
even when the flux changes by more than an order of magnitude (see Fig. 9). No
consistent pattern expected in acceleration and cooling scenarios [105] is found in
the spectral index/flux plane.
The origin of the spectral break in 3C 454.3 bears on several important issues in
FSRQs: the location of the γ-ray emission site; the source of soft target photons in
Compton-scattering models; and the relation of FSRQs and BL Lac objects in view
of the disappearance of such breaks in ISP and HSP blazars. Such a break would be
readily understood if the target field was sufficiently strong to attenuate the blazar
radiation by γγ absorption processes, but the intense Ly α radiation field at 10.2 eV
observed with GALEX [98] would make a break at E br 30 GeV [113]. Models
employing photon attenuation deep within the BLR by He II recombination and Ly
α photons with E > 54.4 eV [112] have been proposed to explain this break, as have
nonthermal electron scattering models with accretion-disk and BLR photon targets
[103]. The spectral break could also be due to Klein-Nishina effects in scattering,
as has been proposed to explain the SED of PKS 1510-089 [87]. The KN break due
to upscattered Ly α radiation occurs at a few GeV, and the observed break energy is
insensitive to the Doppler factor. This scattering problem is treated in Sect. 4.
-8
10
ν Fν (erg cm-2 s-1 ) -9

10
14
break cutoff
-10 12
10 Hist. flares
E break or Ecutoff (GeV)
10 Nov. 2010
0
-11 0 5 10 15 20 25 30 35 40 45
10 -6
Flux [E>100 MeV] (10 ph cm-2s-1)
2 3 4
10 10 10
E (MeV)
Fig. 9 Fermi LAT ν Fν SEDs for four different time periods corresponding to preflare (magenta
lower spectrum), plateau (green third from top), flare (blue top spectrum), and post-flare (red second
from top) periods, along with the fitted functional forms, including a broken power law (solid), log-
parabola (dashed), and power-law with exponential cutoff (dashed-triple-dotted) functions [9]. The
inset displays E break (red) and E cutoff energies (blue) as a function of flux for the different periods
(filled symbols) and for historical flares (open symbols)
PKS 2155-304, Mrk 501, and BL Lac Objects
PKS 2155-304, an X-ray selected BL Lac object at z = 0.116 and an EGRET source
[118], is one of the most prominent representatives of the HSP blazar population.
During a period of extraordinary flaring on 2006 July 28, PKS 2155-304 exhibited
a succession of γ-ray flares varying on time scales as short as ≈5 min with apparent
luminosities >1046 erg/s—larger still when EBL absorption is included [95]. This is
an extreme, hypervariable source in the sense of Eq. (14).
An 11 day campaign between 2008 August 25 and 2008 September 6 was orga-
nized early in the Fermi mission to measure the SED at optical (ATOM; Automatic
Telescope for Optical Monitoring), X-ray (RXTE and Swift), and γ-ray (Fermi and
HESS) frequencies. It turned out to be in a low state, well fit by a standard one-zone
synchrotron/SSC model, with Doppler factor δD = 32, magnetic field B = 0.018 G,
and comoving size R = 1.5×1017 cm (corresponding to a variability time of 2 days;
[96]).13 When flaring, however, the one-zone synchrotron SSC model for PKS 2155-
304 completely fails, or at least requires δD 100 [102].
13 The reader is assumed to be familiar with the synchrotron/SSC model; see, e.g., [4, 102, 99, 116].
262 C. D. Dermer
10-9
SMA
VLBA(BP143) Swift/BAT
VLBA_core(BP143) RXTE/PCA
VLBA_core_ellipsefit(BP143) Swift/XRT
-10 VLBA(BK150)
10
VLBA_core(BK150)
Noto
Metsahovi
Medicina
VLBA(MOJAVE)
ν Fν [erg cm-2 s-1 ]
10-11 VLBA_Core(MOJAVE)
OVRO
RATAN
UMRAO
Effelsberg
10-12
MAGIC
Swift/UVOT
BradfordRoboticTelescope VERITAS
GASP VERITAS_flare
GRT
Fermi
-13 MITSuME
10
ROVOR
WIRO
OAGH
CampoImperatore
10-14
10 16 18 20 26 28
10 1012 1014 10 10 10 1022 1024 10 10
ν [Hz]
Fig. 10 SED for Mrk 501 averaged over all observations taken during the multifrequency campaign
performed between 2009 March 15 (MJD 54905) and 2009 August 1 (MJD 55044) [90]. The legend
reports the correspondence between the instruments and the measured fluxes
One-zone synchrotron/SSC models with δD 10 give good fits to the long-term

average spectra of other HSP BL Lac objects such as Mrk 421 and Mrk 501. The
spectacular multiwavelength SED of Mrk 421 shown in Fig. 10 represents one of
the most detailed multiwavelength blazar SEDs yet assembled [90]. The optical and
X-ray data have been corrected for Galactic extinction, but the host galaxy, which is
clearly visible at IR/optical frequencies, has not been subtracted. The MAGIC and
VERITAS data have been corrected for EBL absorption (Sect. 7.3). A single-zone
synchrotron/SSC model with 12 δD 22, B ≈ 15–30 mG, and R ≈ 8–50 lt-
day, gives a good fit to the nonflaring SED shown in Fig. 10. By comparison, radio
galaxies are fit by synchrotron/SSC models with Doppler factors of order unity. We
return to this point in Sect. 7.
3.4 Second LAT AGN Catalog (2LAC)
The 2LAC [92] is based on the first two years of scientific data taken with the
Fermi Gamma ray Space Telescope. It contains 1016 |b| > 10◦ sources that are
associated at high confidence with AGNs (1319 of the 1873 1FGL sources are at
|b| > 10◦ ). The 2LAC clean sample comprises 885 sources, consisting of 395
BL Lac objects, 310 FSRQs, 156 blazars of unknown type, 8 misaligned AGNs,
4 RL-NLSy1 galaxies, 10 AGNs of other types, and 2 starburst galaxies (NGC

4945 has fallen out of the list); see Table 4. The photon index distribution of the
blazars of unknown type suggests that they comprise roughly equal numbers of
BL Lacs and FSRQs. Of the 395 BL Lac objects, 220 (∼55 %) lack measured
redshifts, and this fraction is roughly the same for LSP, ISP, and HSP BL Lac
objects.
As shown in the 2LAC, threshold sensitivity in terms of photon flux is strongly
dependent on source spectral index, whereas energy flux is not, and detection of
FSRQs and BL Lac objects is complete to an energy flux of ≈5 × 10−12 erg/cm2 -s.
Because more and deeper surveys have taken place in the northern hemisphere, the
smaller fraction of southern-hemisphere BL Lac objects indicates that some 60 unas-
sociated sources at negative declination are expected to be BL Lac objects.
Besides validating the results of the 1LAC with additional data and improved
analysis techniques, the 2LAC shows that the mean fractional variability of FSRQs
is greater than that of BL Lac objects. The duty cycle, as defined by the fraction of
time when the average flux exceeds 1.5 standard deviations above the mean flux,
is ≈5–10 % for bright blazars. The number of BL Lacs and FSRQ have increased
by 44 % and 35 % from the 1LAC to 2LAC clean samples, respectively, with the
future increases in the number of FSRQs expected to be even more modest. This is
due to cosmological effects, as reflected in the flattening in the log N -log S distrib-
ution, and K-corrections that pushes the GeV peak of the ν Fν spectra to lower ener-
gies where the Fermi-LAT effective area rapidly declines and sensitivity to FSRQs
gets worse.
Blazars make up 97 % of the extragalactic γ-ray sources detected with the
Fermi-LAT. The number of misaligned sources has stalled, with the same num-
ber of radio galaxies—eleven—in the 2LAC as in the misaligned AGN paper. But
these are not the same objects. The radio galaxies 3C 78, 3C 111, and 3C 120 are
not found in the 2LAC, evidently because a variable jetted emission component that
contributed to the γ-ray emission in the past has gone quiet. Now found in the 2LAC
are the FRI radio galaxies Fornax A and Cen B, and the head-tail radio galaxies
IC 310 [110], the latter of which is also a MAGIC source [97]. The radio/γ-ray
connection [93] and GeV–TeV synergy [85] resulting from the Fermi-LAT for AGN
studies, let alone Galactic sources, cannot be adequately reviewed here; see [92]
and [70].
4 Relativistic Jet Physics
Here we develop and apply relativistic jet radiation physics [4] to some puzzles
arising from the Fermi observations.
264 C. D. Dermer
4.1 GeV Spectral Break in LSP Blazars
As already described in Sect. 3.3, an important new result in blazar physics not
anticipated before the launch of Fermi is the prevalence of a spectral softening in the
γ-ray spectra of low synchrotron-peaked (LSP) blazars, including both FSRQs and
LSP BL Lac objects. The spectral softening occurs generally between 1 and 10 GeV
[36, 45]. For 3C 454.3 [91, 119], the break is 1 unit, and the energy of the break
is rather insensitive to flux (Fig. 9). A softer spectral break, possibly consistent with
radiative cooling, is found at multi-GeV energies in the Fermi/MAGIC spectrum
of FSRQ 4C +21.35 [175]. If a GeV spectral softening and significant GeV flux
variability accompanied by modest spectral variability is typical of blazars, then
some physical process should be able to explain this behaviour.
γ Rays from External Compton Scattering
We treat the problem that blazar γ radiation arises from the scattering of a quasi-
isotropic target photon field [172] in the Klein-Nishina regime. Compton scat-
tering takes place in the Thomson limit when 4γ 1 (Eq. (6.123) in [4]),
where γ and are the electron Lorentz factor and photon energy in the comoving
frame.
Consider an external isotropic, monochromatic photon field with photon energy
(in m e c2 units). The average energy in the comoving frame is ∼ = Γ .
The upscattered photon energy in the Thomson regime is C ∼ (4/3)γ 2 . Hence
=
= (4/3)γ 2 2 = (4γ )2 /12. Thus scattering is in the Thomson regime provided
C
12 C 1, or (δ /Γ )/[12 (1 + z)], implying E (GeV) 12/[E (eV)]
D C C
for δD ∼ = Γ . Away from the endpoints, therefore, the scattered photon energy mark-
ing the transition to the KN regime is very weakly dependent on Doppler factor in
highly beamed relativistic jets.
The spectrum of an isotropic nonthermal power-law electron distribution, when
transformed to the stationary frame, remains a power-law with the same index as
in the comoving frame, but with a strong angular dependence. The endpoints of
the distribution are the low-and high-energy electron Lorentz cutoffs, γ1 and γ2
respectively, boosted by δD [147]. Consequently, the shape of the scattered spec-
trum cannot depend on δD provided that γ1 = δD γ1 γKN ∼ = (4∗ )−1 δD

γ2 = γ2 .
If the break energy observed in 3C 454.3 at ≈2 GeV is due to the transition to
scattering in the KN regime, then the underlying target photon energy E ∗ ≈ 6 eV. This
is close to the energy of the Ly α photon at 10.2 eV, or Hβ at 2.55 eV. If the origin
of the spectral break is due to scattering of nearly mono-energetic line radiation,
then this would (1) place the scattering site within the BLR, and (2) explain the near
constancy of the spectral break, independent of flux state.
We use the method of Georganopoulos et al. (2001) [147], applied to isotropic
radiation fields in the stationary frame of the supermassive black hole and BLR. The
differential Compton-scattered spectrum when isotropic, monoenergetic electrons

Compton upscatter isotropic, monochromatic target photons, is given for the full
Compton cross section in the head-on approximation by
2
3 cσT u 0 s Ne (γ, Ωs )dγ γ
df C
= FC (q, Γe )H − s (28)
4 d L2 ∗ γ 2 1 + (1/Γe )
(Eq. (6.129), [4]), with
N (γ )
Ne (γ, Ωs ) = δD
3
, γ = γ/δD (29)
4π
(Eq. (6.124)),
s = (1 + z) ≡ z , Ωs = Ω∗ , (30)
1 (Γe q)2
FC (q, Γe ) = 2q ln q + (1 + 2q)(1 − q) + (1 − q), (31)
2 (1 + Γe q)
s /γ
q≡ , Γe ≡ 4γ∗ , (32)
Γe (1 − s /γ)
and q is restricted to the range (4γ 2 )−1 ≤ q ≤ 1 [134, 154] (Eqs. (6.74)–(6.76),
(6.125)–(6.127), [4]). Restriction to the Thomson regime occurs when Γe 1 or
4γ∗ 1. The final term in Eq. (31) dominates for scattering in the KN regime.
(A simpler form for analytic work is the isotropic Thomson cross section,
2
FT,iso (ˆ) = 1 − ˆ (33)
3
(Eq. (6.71), [4]), with ˆ = s /4γ 2 (Eq. (6.70)), which assumes isotropic scattering
in the electron rest frame.)
Thus an isotropically distributed nonthermal electron distribution in the comoving
frame gives a Compton-scattered spectrum
∞ ∞
3 cσT 2s 3 u ∗ (∗ ) Ne (γ/δD )
f C,iso = δ d∗ dγ FC (q, Γe ), (34)
4 4πd L2 D 0 2∗ γmin γ2
(Eq. (6.123); considering only upscattering), with

s 1
γmin = 1+ 1+ (35)
2 s ∗
266 C. D. Dermer
Fig. 11 Spectrum of Compton-scattered line radiation. Left Dependence on Doppler factor δD for
sources at different redshifts. Right Dependence on endpoints. Within the endpoints, the spectral
shape of the function is independent of δD
(Eq. (6.90)). Substituting Eq. (29) for an isotropic power-law electron spectrum in the
comoving frame, so that N (γ ) = Neo γ − p H (γ ; γ1 , γ2 ),14 scattering an external
monochromatic line spectrum u ∗ (∗ ) = u ∗ δ(∗ − ∗o ) gives
γ2
3 cσT u ∗ s 2 3+ p
fC,iso
= δD Neo dγ γ −( p+2) FC (q, Γe ). (36)
4 4πd L2 ∗ min[γmin ,γ1 ]
This is numerically solved to give the results shown in Fig. 11, which were compared
in [91] with the rapidly falling spectrum of 3C 454.3.
Figure 12 shows a model where the γ-ray continuum observed from 3C 454.3
with the Fermi LAT [91] are formed by power-law nonthermal jet electrons, with
number index p = 3.17, that Compton-scatter Ly α line photons. The model is
insensitive to values of lower and upper comoving electron Lorentz factors γmin and
γmax provided γmin 102 and γmax 104 . The Klein-Nishina softening from a
power-law electron distribution gives a poor fit to the data, but if one assumes that the
underlying electron distribution is curved, as in the case of a log parabola function
[160], for example, it may be possible to obtain a good fit to the 3C 454.3 data for
this model.
For integrations over blackbody spectra, substitute
2m e c2 3∗
u ∗,bb (∗ ) = (37)
λ3C exp(∗ /Θ) − 1
into Eq. (34). The dimensionless temperature of the radiation field is Θ = kB T /m e c2 .

The CMBR can be approximated as an isotropic, monochromatic radiation field
14The Heaviside functions H (x − a) and H (x; a, b) are defined such that H (x − a) = 1 if x > a
and H (x − a) = 0 otherwise, and H (x; a, b) = 1 if a ≤ x ≤ b and H (x; a, b) = 0 otherwise.
Fig. 12 Model for the

γ-ray spectrum of 3C 454.3
when isotropic power-law
electrons with index p = 3.17
in a relativistic jet Compton
scatter Ly α photons [91]
u ∗ (∗ ) = u 0 δ(∗ − 0 ), with u 0 = 4 × 10−13 (1 + z)4 ergs/cm3 and 0 =

2.70ΘC M B R ∼= 1.24 × 10−9 (1 + z). The condition q < 1 implies s < γ/(1 + Γe−1 ).
Compton Emissivity and Electron Energy-Loss Rate
In the calculation of time-dependent blazar spectra with evolving electron distrib-

utions, or in pair cascade calculations, it is necessary to consider the full Compton
cross section in scattering and electron-energy loss calculations. When γ 1, the
head-on approximation, where the incident photon is assumed to be directed oppo-
site (“head-on” to) the direction of the electron when transformed to the electron rest
frame, can be employed.
For general treatments, the time-dependent emissivity s jC (s , Ωs ; r, t) is cal-
culated. For external isotropic photon spectra, the angle-dependent emissivity when
jetted nonthermal, isostropic electrons with electron spectrum n e (γ, Ωe ) Compton-
scatter photons of an isotropic monochromatic photon field is given, in the head-on
approximation, by the expression
2 ∞
3 s n e (γ, Ωe ) 1
s jC (s , Ωs ) = cσT u 0 dγ FC (q, Γe )H q; 2 , 1 (38)
4 1 γ2 4γ
(Eq. (6.74)). Here the electron spectrum is written in the stationary frame where the
radiation field is isotropic. The ν L ν Compton luminosity for a single electron is
given by

3 cσT u 0 s 2 γ
s L C (s ) = FC (q, Γe ) H s ; , . (39)
4 γ2 1 + /γ 1 + Γe−1
268 C. D. Dermer
The electron energy loss rate by Compton scattering is given by

∞ γ/(1+Γe−1 )
3 cσT u 0
−γ̇C = ds L C (s ) = ds s FC (q, Γe )
0 4 m e c2 γ 2 2
1
12cσT u 0 γ 2 q
= dq FC (q, Γe ). (40)
mec 2
0 (1 + Γe q)3
Using the isotropic Thomson kernel, Eq. (33) becomes in the Thomson regime,
Γe 1,

cσT u 0 s 2 3 x2 x γ
s L C (s ) ∼
= 1 + 1 − H ;
s , .
2γ 2 4 (1 − x) Γe (1 − x) 1 + Γe−1
(41)
Here x = s /γ so that q = x/[Γe (1 − x)]. To lowest order,
Γe
cσT u 0 x cσT u 0 Γe2 4
− γ̇T ∼
= dx x 1− = = cσT u 0 γ 2 , (42)
22 0 Γe 122 3
recovering the familiar result in the limit γ 1, /γ 1. Expanding the energy-

loss rate expression, again using the isotropic Thomson kernel, gives the expansion

4 3 163 2
− γ̇C,i=1 → cσT u 0 γ 2 1 − Γe + Γ + O[Γe3 ] (43)
3 2 40 e
correct to O(Γe2 ).
4.2 Leptonic Jet Models
A photon flux of F−8 = 2200 from 3C 454.3 at z = 0.859 implies an apparent

>100 MeV isotropic γ-ray luminosity of Lγ = 3.3 × 1049 erg s−1 [91]. The flux
reached a factor ≈6 larger in the November 2010 outburst [9], corresponding to an
apparent luminosity of ≈2 × 1050 erg/s, which is the record-holder for all blazars,
including PKS 1622-297 from the EGRET era [108].
Using the measured flux and a one-day variability timescale at the time that the
most energetic photon with E ≈ 20 GeV was detected implies a minimum Doppler
factor of δD,min ≈ 13, as we show. Assuming that the outflow Lorentz factor Γ ≈ 20,
consistent with the inferred value of δD,min and with radio observations at a different
epoch [155], then simple arguments suggests a location R cΓ 2 tvar /(1 + z) ≈
0.2(Γ /20)2 (tvar /day) pc, which is at the outer boundary of the BLR. Flux variability
on time scales as short as 3 h was measured at another bright flux state [9], which
suggests that the γ-ray emission site would be even deeper in the BLR. This stands in
contrast to inferences based on coherent optical polarization changes over timescales
of weeks prior to a γ-ray flare in 3C 279 [121], which places the emission site at
much larger distances. The MAGIC detection of VHE emission from PKS 1222+21
(4C +21.35) [128] is even stronger evidence for γ-ray production at the pc scale or
farther from the central engine.
Jet Doppler Factor
Emission from bulk magnetized plasma in relativistic motion provides the best expla-
nation for the large apparent isotropic luminosities L iso , energy releases Eiso , and
short variability times tvar from cosmological γ-ray sources. The comoving emission-
region size scale R is related to tvar through the expression R cΓ tvar /(1 + z).
Combined Fermi LAT and GBM observations give the best GRB data from which to
determine the minimum bulk outflow Lorentz factor Γmin through γγ opacity argu-
ments. For blazars, multiwavelength X-ray measurements combined with Fermi-LAT
data are important for inferring the minimum Doppler factor.
It is simple to derive Γmin in a blast-wave formulation, noting that the internal
photon energy density
4πd L2 Φ (1 + z)2 d L2 Φ
u γ ≈ ≈ , (44)
Γ 2 4π R 2 c Γ 6 c3 tvar
2
using R ≈ Γ 2 ctvar /(1+z). The optical depth for γγ → e+ e− processes is τγγ (1 ) ∼ =
R σT (1 /2)u γ (2/1 )/(m e c2 ), where R = R/Γ and = 2/1 from the threshold
condition. The condition τγγ (1 ) < 1 with the relation Γ 1 /(1 + z) = 1 implies
1/6
σT d L2 (1 + z)2 f ˆ 1 2Γ 2
Γ Γmin ∼
= , ˆ = (45)
4tvar m e c4 (1 + z)2 1
[124]. Here f is the ν Fν flux at photon energy m e c2 , which is evaluated at = ˆ

due to the peaking of the γγ cross section near threshold.
The minimum Doppler factor δmin (δ D = [Γ (1 − β cos θ)]−1 ) defined by the
condition τγγ (1 ) = 1, can be estimated to ≈10 % accuracy compared to results of
more detailed numerical calculations through the expression
1/6
σT d L2 (1 + z)2 f ˆ 1
δmin ∼
= , (46)
4tv m e c4
where f is the ν Fν spectrum measured at frequency ν = m e c2 / h, tv is the variabil-

ity time, and E 1 = m e c2 1 is the maximum photon energy during the period in which
f is measured. The ν Fν flux f in Eq. (46) is evaluated at = ˆ = 2δD
2 /(1 + z)2
1
from the pair-production threshold condition.
270 C. D. Dermer
For 3C 454.3, writing f = 10−10 f −10 erg cm−2 s−1 in Eq. (46) gives δmin ≈
1/6
16.5 f −10 E 1 (10 GeV)/tv (6 h) and Ê(keV) ∼= 3.4(δmin /15)2 /E 1 (10 GeV).
Contemporaneous Swift XRT observations give the ν Fν flux between 0.2 and 10 keV.
From Swift public data [98], ν Fν (4 keV) ≈ 6 × 10−11 erg/cm2 -s, so that δmin ≈ 15.
Because δD > δmin , we write δ = 20δ20 and δ20 ≈ 1. This value compares favor-
ably with δD = 24.6 ± 4.5, bulk Lorentz factor Γ = 15.6 ± 2.2, and observing
angle θ(◦ ) = 0.8 ± 0.2 obtained from superluminal observations [155], derived from
measurements made at a different time.
Variability Time Scale
In the case of 3C 454.3, the emission region for one of its major flares is constrained in
a colliding shell geometry to be at distance R 2cΓ 2 tv /(1 + z) ≈ 0.1Γ20 2 t (6 h) pc,
v
assuming Γ20 ≡ Γ /20 ∼ = 1. This is within the BLR [176]. Ways to avoid this
conclusion are to assume that Γ 20, which would be associated with large jet
powers, or for only a small portion of the emission region to be active, which would
lower the jet’s radiative efficiency given that only a small fraction of the emitting
surface is radiating. This might be implausible in view of the already large apparent
isotropic γ-ray luminosity of 3C 454.3. Recollimation shocks (e.g., [135]) at the pc
scale might reduce the characteristic size of the emission region, though the contrast
(cδtv /[(1 + z)R pc (pc)] ∼ 2 × 10−3 tv pc/R pc ) between the size of the emission
region and location R = R pc pc seems unexpectedly small even for a narrow jet.
Alternately, flaring episodes with short variability times might take place within the
BLR, whereas the more slowly varying emissions could be radiated by jet plasma
at larger distances, but then we would expect large spectral variations due to the
different target photon sources.
The uses of variability studies to infer properties of blazars, GRBs, or other sources
is an ongoing concern, because flow properties are inferred from the variability
timescale. The auto-correlation function can be used to infer a characteristic vari-
ability timescale, as can breaks in the power density spectrum of sources [122]. Other
techniques for variability analysis of the extensive compilation of blazar light curves
are currently in development.
Equipartition Field and Jet Power
Here we perform an equipartition power analysis for the giant outbursts from 3C
454.3 [9, 91]. In the blob framework, where a spherical emitting region with radius
R in the comoving frame moves with Lorentz factor Γ at an angle θ to the line
of sight, the absolute jet power is [136, 148] (Sect. 5.5, [4]), including the photon
power,
2 Γ 2 (δ D Beq )2 2 4 Γ2
L j = 2πr βc 2 χ + 3/2 + 4πd L2 4 Φ. (47)
δD 8π 3χ 2δ D
Here Φ = 10−9 Φ−9 erg cm−2 s−1 is the measured bolometric photon energy flux
(4πd L2 Φ = 3.6 × 1048 erg s−1 for Φ−9 = 1; Φ−9 ∼ = 10 and 50 during the December
2009 and November 2010 outbursts, respectively), and the factor Γ 2 /2δ 4D in the last
term of this expression recovers the increased energy requirements due to debeaming
when θ 1/Γ , or δ D Γ , in the limit that the jet opening angle is 1. The factor
χ is the deviation from equipartition, so that δ D B = χ(δ D Beq ) and
2/7
9πd L2 f pk m e c2 (1 + ζ pe ) ln(2 /1 )
δ D Beq = Bcr 2 cσ V
(48)
4 (1 + z) pk Ucr T b
(Eq. (7.80), [4]). Here f pk is the ν Fν peak of the synchrotron spectrum, Ucr =
2 /8π, B = 4.414 × 1013 G, and ζ is the relative energy in protons to nonther-
Bcr cr pe
mal electrons which emit synchrotron radiation with a spectrum α = 0.5 between
observed photon energies 1 and 2 .
Letting (1 + ζ pe ) ln(2 /1 ) = 102 Λ2 , the peak synchrotron frequency ν pk =
10 ν13 Hz, and f pk = 10−10 f −10 erg cm−2 s−1 , substitution of parameters for 3C
13
454.3 gives
( f −10 Λ2 )2/7
Beq (G) = 3.25 13/7 1/7 6/7 (49)
δ20 ν13 tv (days)
i.e., an equipartition field of a few G. The corresponding jet power is

2/7 δ 2/7 t 2/7 (days)
Γ 46 ( f −10 Λ2 )
2 4
−1 ∼ 20 v
L j (erg s )= 2.5 × 10 χ2
+ (50)
δ 2D 1/7
ν13 3χ3/2

Φ −9
+ 5 × 1045 2 .
δ20
The Eddington luminosityfor a 109 M9 Solar mass black hole is 1.26 × 1047 erg s−1 .
Estimates for the black-hole mass in 3C 454.3 range from M9 ≈ 0.5 [98] to M9 ≈ 4
[150]. Thus we see that the jet power from 3C 454.3 would be super-Eddington if χ
departs from its equipartition value by more than a factor ≈4. Moreover, the system
cannot be severely debeamed, though this is already unlikely from other arguments,
e.g., core dominance parameter and superluminalmotion observations [155].
By taking the ratio of the synchrotron and Compton ν Fν peaks, using the relations
pk,syn ∼
= (3/2)γ 2pk (B/Bcr )δ/(1 + z) and pk,C ∼
= (4/3)γ 2pk Γ δ/(1 + z), we have
∼
= ( pk,C /Γ pk,syn )(B/Bcr ) or
E pk,C (100 MeV)B(3 G)

E (eV) ∼
= 4.6 . (51)
ν13 Γ20
272 C. D. Dermer
This suggests that the soft photon energy scattered to make the GeV emission is a
few eV, which would correspond to either atomic line radiation or scattered Shakura-
Sunyaev accretion-disk radiation.
The energy density of the external radiation field can also be estimated from
the 3C 454.3 observations. Let AC represent the ratio of the energy fluxes in the
Compton and synchrotron components. If u denotes the energy density of the tar-
get radiation field in the stationary frame, and u B denotes the comoving magnetic-
field energy density, then for Γ 1, 4u Γ 2 /3 ≈ AC U B , or u (erg cm−3 ) ∼ =
0.007(AC /10)B 2 (3 G)/Γ20 2 . This can be compared to energy densities in the BLR
using relations between the BLR radius and line luminosity [98]. The implied energy
densities are an order of magnitude larger, suggesting that either B > 3 G or Γ < 20.
4.3 Hadronic Jet Models
Measurements of air showers induced by cosmic rays impacting the atmosphere

give the best evidence for the existence of powerful accelerators of UHECRs with
E 1018 eV, with Larmor radii so large that they almost certainly originate from
beyond the Galaxy.
Adiabatic Losses, and Photopair and Photopion Losses on the CMBR
The radial scale factor

R0
R= (52)
1+z
at redshift z. In an adiabatic expansion process, dQ = 0 = dU + pdV and the

energy content U = uV , where u is the energy density and V is the volume. Thus
udV + pdV = −V du, or (u + p)dV = −V du, so
dV dx dy dz dR du
− = + + =3 = . (53)
V x y z R u+p
For a relativistic gas with p = u/3,
3dR du 3d ln u
− = 4u = −3d ln R = , (54)
R (3) 4
which implies u ∝ R −4 . Because the energy density U = uV ∝ R −1 , γ ∝ R −1 , so

−γ̇ ∝ −(1/R 2 )dR/dt ∗ (see Sect. 9.4 in [4]).
With R given by Eq. (52),

dz
−1

− γ̇ = (1 + z) γ = H0 Ωm (1 + z)3 + ΩΛ , (55)
dt
∗
for a flat cold dark matter cosmology with cosmological constant Λ. Thus

d ln γ 1 dz
− = = H0 , (56)
dt (1 + z) dt ∗
The mean-free path for energy losses due to photopair production is given in
Sect. 9.3.2 of [4] by the expression
cE c
rφe (E 20 ) = =
|dE/dt| |d ln γ/dt|
√
∼ 1.0[E 18 (1 + z)]1/2 y
= Gpc ≡ Gpc, (57)
(1 + z) Fφe [E 18 (1 + z)]
3 (1 + z) Fφe (y)
3
where
0.69 ln(y/2)
Fφe (y) = 0.74 + 1.78 ln(y/2) + , (58)
(y/2)3/2
taking the asymptote y 1 or E 18 (1 + z)−1 . If the ankle in the spectrum

is a consequence of photopair losses of ultra-high energy protons, as advocated by
Berezinskii and colleagues [132], then the energy of the ankle would reflect the
redshift epoch of greatest source production of the UHECRs, noting that rφe ∝
(1 + z)−3 at the maximum energy loss of protons due to photopair losses on the
CMB.
Approximately, then

d ln E (1 + z)3 [0.74 + 1.78 ln(y/2)]
=− √ , (59)
dt φe y (Gpc/c)
when y = E 18 (1 + z) 1. The conversion (Gpc/c) = 1.03 × 1017 s ∼ = 3.2 ×

109 years.
The mean-free-path for a proton to lose energy through photopion losses on the
CMBR is given by Ref. [4], Sect. 9.2.4, by
c ∼ 13.7 exp[4.0/E 20 (1 + z)]

rφπ (E 20 ) = = Mpc, (60)
|d ln E/dt|φπ (1 + z)3 [1 + 4.0/E 20 (1 + z)]
which, though derived in the limit E 20 4/(1 + z), gives a good approximation
even at higher energies.
274 C. D. Dermer
Photopion Production Efficiency
From the relations between time elements in the stationary frame (starred), the
comoving frame (primed), and the observer frame (unscripted), R = βct∗ = βcΓ t ,
so t ∼ = (1+z)R/βΓ 2 c, since t ∼
= R/βΓ c for relativistic flows, and t ∼ = (1+z)t /Γ .

The comoving dynamical time scale is therefore given by tdyn = R /c = R/Γ c.
The energy flux
dE L∗ dE∗
2 dE
Φ= = , L ∗ = = Γ , (61)
dAdt 4πd L2 dt ∗ dt
noting Γ E = E∗ and dt ∗ = dt /Γ . Therefore L() = 4πd L2 f = Γ 2 L( ), and
L() L ( )
n() = , so n ( ) = . (62)
4π R cm e c
2 2 4πΓ 2 R 2 m e c3
The causality constraint for rapidly variable emissions from a relativistic blast wave is
Δr Δt Δt∗ ∼ Γ Δt
Δr ∼
= = = , (63)
Γ c Γc (1 + z)c
and tvar ∼
= (1 + z)R/Γ 2 c.
We write the target ν Fν photon spectrum as
a b

f = f pk H ( pk − ) + H ( − pk ) . (64)
pk pk
The threshold for photopion (φπ) processes is γ p m π /m e ∼ = 400, using the

approximation of [131] that the product of the photopion inelasticity and cross section
is K pγ σφπ = 70 µb above comoving photon energy ∼ = 400. Because γ p =
E p /Γ m p c2 and = (1 + z)/Γ , the threshold escaping proton energy is
∼ 400m p c Γ and E ppk ∼

2 2 400m p c2 Γ 2 ∼ 3.7 × 1017 Γ32
E thr
p = = = eV (65)
(1 + z) (1 + z) pk (1 + z) pk
where
Γ
Γ3 ≡ .
1000
The timescale for energy loss due to photohadronic processes is
∞
−1
t pγ (E p ) ∼
= c(K pγ σ pγ ) d n ( ). (66)
thr
Thus
c(K pγ σ pγ )d L2 ∞ f
−1
t pγ (E p ) ∼
= d . (67)
R 2 m e c3 Γ 2 thr 2
Using Eq. (64) gives

(K pγ σ pγ )d L2 f pk
−1
Ep ∼
pk
t pγ = , (68)
R 2 m e c2 Γ pk (1 − b)(1 + z)
The photopion production efficiency

tdyn
η pγ (E p ) = (E )
. (69)
t pγ p
pk
At E p = E p ,
(K pγ σ pγ )d L2 f pk
pk pk
η pγ = η pγ E p = . (70)
Γ 4 m e c4 tvar pk (1 − b)
Thus ⎧ 1−b
⎪
⎪ Epkp pk
⎪
⎪ , Ep < Ep
⎨ ⎧E p

η pγ (E p ) ∼
pk
= η pγ ⎨ E p 1−a . (71)
⎪
⎪ , a < 1 , E pk < E
⎪
⎪
pk
⎩⎩ E p p p
1, a>1
To illustrate these results, we apply them to Fermi observations of GRBs. For

GRB090510 at z = 0.903 and d L ∼ = 1.8 × 1028 cm, E p ∼
pk
= 2 × 1017 (Γ32 / pk ) eV
and η pγ ∼
pk
= 0.03 f −5 /Γ34 t−2 pk (1 − b), and is thus at the ∼1–10 % level. For GRB
080916C at z = 4.35 and d L ∼ = 1.25 × 1029 cm [120], E p ∼
pk
= 7 × 1016 (Γ32 / pk ) eV
pk ∼
and η pγ = 0.0015 f −6 /Γ3 tvar (s) pk (1 − b).
4
For GRB 090926A at z = 2.1062 and d L = 16.54 Gpc, the estimate above gives
Ep ∼ = 1017 (Γ32 / pk ) eV and η pγ ∼
pk pk
= 0.01( f −6 /6)/Γ34 tˆvar pk (1 − b). The spectral
parameters to derive the photopion efficiency are given for the Band function in
Table 5 [125]. Here the variability time scale tvar is scaled to 0.15tˆvar s, which is
the characteristic FWHM time of the well-resolved pulse. One percent efficiency
Table 5 Band function fits to GRB 090926A

Time interval f −6 (10−6 erg cm−2 s−1 ) a b pk
+0.07
(a): 0.0–3.3 s 3.5 1.58 ± 0.03 −0.64−0.09 0.66
(b): 3.3–9.7 s 6 1.45 −0.46 0.56
(c): 9.7–10.5 s 5 1.41 −1.69 0.41
(d): 10.5–21.6 s 1 0.30 −0.80 0.36
Intervals (c) and (d) are best fit with additional cut-off power-law/power-law component
276 C. D. Dermer
can easily become 100 % if Γ3 ≈ 0.3, and more for higher energy protons, making
escape difficult. If long GRBs accelerate UHECRs, then the blast-wave Lorentz
factor must be close to that given in the simple γγ opacity limit, otherwise the GRB
would be highly radiative by hadronic processes. With the strong dependence on
Γ , photohadronic interactions become 100 % efficient when Γ ≈ 300, so if GRBs
accelerate UHECRs, an accompanying photohadronic γ-ray spectral component is
predicted [140, 141, 162, 169].
Proton Models
The comoving synchrotron cooling timescale of an ion with atomic mass A and
charge Z is given by (Eqs. (7.49)–(7.50), [4])
3
A3 mp 6πm e c
tsyn = 4 . (72)
Z me σT B 2 γ
Balancing the synchrotron energy-loss timescale with the shortest energy-gain

timescale allowed in standard first- or second-order Fermi acceleration processes
gives the relation

A2 m p 2 6πm e c
B γ 2 = 3 , (73)
Z me σT φ
where φ 1 is the number of radians over which a particle gains ∼e of its original
energy. The observed peak synchrotron photon energy is

Γ syn 3Γ Z me B 2
syn ∼
= ∼
= γ
(1 + z) 2(1 + z) A mp Bcr
∼ Γ 27 m p 108 Γ3 A
= ≈ . (74)
(1 + z)φ 8α f m e (1 + z)(φ/10) Z 2
Overlooking other limitations on particle acceleration, Eq. (74) shows that proton
synchrotron radiation from GRB jets and blazars can reach ≈50 TeV (Γ ≈ 103 ) and
≈5 TeV (Γ ≈ 100), respectively. Proton synchrotron models have been developed
in Refs. [127, 161] for γ-ray blazars, and in [171] for GRBs. We consider application
of this type of model to these two source classes.
GRBs
Eliminating γ from these equations using the observer time tsyn = (1 + z)tsyn
/Γ
2 2
for the emission to be radiated at measured energy m e c syn = Γ m e c syn /(1+z) =
100E 100 MeV implies a comoving magnetic field [171, 179]
1/3
3 A5/3 1+z 8πm e c 2/3
B (G) =
25/3 Z 7/3 Γ Bcr syn σT tsyn

∼ 1.3 × 105 A5/3 1 + z 1/3
= 2/3 , (75)
tsyn (s) Z 7/3 Γ3 E 100
and an isotropic jet power, dominated by magnetic-field energy, of
2 B 2 10/3
1 2 2 2 ∼ 1 c3 Γ 6 tvar 32 A10/3 c3 mp
L ∗ > L ∗,B ∼ = R cΓ B = =
2 2 (1 + z)2 213/3 Z 14/3 (1 + z)4/3 m e
16/3 2
Γ 16/3 tvar
2 8πm e c 4/3 ∼ A10/3 Γ3 tvar (s)
× 2/3 2/3 = 2.5 × 1059 14/3 erg/s,
Bcr syn tsyn σT Z 2/3
[(1 + z)tsyn (s)]4/3 E 100
(76)
letting the blast-wave radius R ∼

= Γ 2 ctvar /(1 + z). For Fe ( A = 56, Z = 26), the
power requirements are reduced by a factor ≈0.17.
Equation (76) shows that L B ∝ Γ 16/3 . The absolute energy release is
16/3
Δt 2.5 × 1059 A10/3 Γ3 tvar 2 (s)
E∗,abs = fb L ∗ Δt (s) f b erg, (77)
1+z (1 + z)7/3 Z 14/3 E 2/3 tsyn
4/3
(s)
100
where f b is the beaming factor (compare Eq. (11)). Using a 100 kG magnetic field
as a fiducial, so B = 105 B5 G, then the absolute GRB energy release is E∗,abs ∼ =
1.3 × 1059 B52 Δt (s)tvar
2 (s) f Γ 6 /(1 + z)2 erg. Such large energy requirements pose
b 3
a problem for highly magnetized GRB models.
In the model of Razzaque et al. (2010) [171], taking a lower limit Γ3 ∼ = 0.5
implied by naive γγ opacity arguments, gives the absolute energy requirements for
GRB 080916C [120] of
−2/3
E∗,abs ∼
5/3
= 2 × 1052 E 100 (Γ3 /0.5)16/3 (θ j /1 deg)2 tsyn (s) erg,
after multiplying by a two-sided jet beaming factor f b = (1 − cos θ j ) ∼ = 1.5 ×

10−4 (θ j /1 deg)2 in the limit θ j 1, from Eq. (11). The precise values of tvar , tsyn
and Δt depend on model interpretation. For the external shock model [139] applied
to GRB 080916C, the rapid variability may be made by irregularities in the complex
surrounding medium, so tvar ∼ = 0.1 s, the proton synchrotron cooling timescale to
make the delayed onset is tsyn ∼ = 4 s, and the GRB duration Δt ∼ = 10–50 s.
From Eq. (77), strong proton/ion synchrotron radiations can be emitted in the
Fermi range near 1 GeV or even 100 MeV due to strong cooling in an intense mag-
netic field, but the absolute energies are large without a very small beaming fac-
tor. The jet break time with apparent isotropic energy release ≈2 × 1057 ergs is
−1/3
tbr ∼
= 0.3(θ j /1 deg)16/3 n 0 days. For such a narrow jet, the jet break would have
taken place before Swift slewed to GRB 080916C at ≈T0 + 17.0 h [170].
278 C. D. Dermer
More complicated geometries might, however, relax the bulk Lorentz factor
requirement further [149]. If the inner engine makes the prompt MeV radiation
and residual shell collisions at larger radii make LAT γ-ray photons, then Γ could be
as low as ∼300 [159]. In this case, the absolute energy release from GRB 080916C
could easily be 1052 erg.
For a proton synchrotron model of GRBs to be viable [171], a narrow jet opening
angle of order 1◦ along with a value of Γ 0.5 gives E∗,abs ∼ = few ×1052 ergs,
within ranges implied by interpretation of radio and γ-ray observations and beaming
[137]. The detection of distinct components in GRB spectra suggests that a cascading
interpretation be more carefully considered [130]. The local rate density of Type Ib/c
supernova progenitors of long GRBs like GRB 080916C or low-luminosity GRBs
like GRB 980425 or GRB 060218 cannot exceed ≈300 Gpc−3 year −1 [173], and the
local beaming-corrected rate density of long duration GRBs is ≈10–50 Gpc−3 year −1
[151, 158]. The local Type 1b/c rate is 9+3 −5 × 10 Gpc
3 −3 year −1 [173]. This agrees
with the star-forming galaxy SN Ib/c rate of ≈0.28 century−1 per 300 Mpc3 per L ∗
galaxy ≈104 Gpc−3 year−1 . These imply a beaming rate of ≈(10–50)/9000 ≈ 10−3 –
0.006. With the ≈1◦ opening angle required to explain GRB 080916C, then all
Type 1b/c SNe would have to make GRBs, presenting another challenge for strong
magnetic-field jet models of GRBs.
Blazars
Proton synchrotron and photohadronic models for blazars have been developed,
as noted, in [127, 161]. The protons and ions make a γ-ray component that originates
from pion-induced cascades, and the synchrotron radiations of pions and muons make
additional γ-ray emissions. Primary electron synchrotron radiation is still usually
required to make the nonthermal radio/optical/X-ray synchrotron blazar emission.
Anita Reimer gives [123] a detailed fit to the famous HBL Mrk 421 at z = 0.031 at
luminosity distance d L = 134 Mpc. Mrk 421 is one of the first Whipple TeV blazars
and the first one with sub-hour [146] variability detected. For the composite SED of
Mrk 421 averaged between 2009 January 9 (MJD 54850) to 2009 June 1, Reimer
makes a proton fit with the parameters shown in Table 6.
The Hillas criterion [152] for a single-zone blob model requires the comoving
Larmor radius to be smaller than the comoving blob radius, that is,
rL < rb . (78)
Table 6 Proton blazar model for Mrk 421 [123]

Parameter Value
B 50 G
δD ≈ Γ 12
rb 4 × 1014 cm
E max eVa
7.2 × 1019 Z Γ12 B50
aΓ = B /50 G, δ = δ /12
= Γ /12, B50
12 12 D
With Eq. (16) giving the maximum escaping particle energy E max ≈ Γ E max ≈

Z eB rb , we see that this model blazar can accelerate super-GZK (E 5 × 1019 eV)
particles if charged ions, but this is only possible for protons with slightly larger
magnetic fields. These protons have maximum comoving Lorentz factors
γ p,max ≈ 6.4 × 109 B50

.
For particles well-trapped by the magnetic field of the plasma, γ p,max = 109 γ9 with
γ9 1.
For the 2-sided jet power as defined in Eq. (47), taking δD ≈ Γ gives

L j = 2πrb2 βcΓ 2 u i , i = p, e, B, (79)
i
and the total Poynting power dominated by magnetic field energy is
1
L j,B = rb2 βcΓ 2 B 2 ∼ 2 2
= 4.3 × 1044 Γ12 B50 erg/s, (80)
4
which is not excessive, considering that the Eddington luminosity is L Edd ≈
1047 M9 erg/s, and M9 ≈ 0.2–0.9 for the supermassive black hole power Mrk 421.
We still need to calculate the particle power and the efficiency for hadronic pro-
duction of γ-ray emission. In the case of proton synchrotron radiation amounting to
total energy flux Φ (erg/cm2 -s), the particle power is

N p0 m p c2 γ p,max 3 m p c3 N p,0 γ p,max Γ 2
L j, p = 2πrb2 βcΓ 2 = , (81)
Vb 2 rb
for a spherical comoving emission region.

According to our standard approach [4],
L
ν Fν = f ∼
= δD
4
∼ Φ.
4πd L2
The proton-synchrotron energy loss rate is

2
dE 4 me
− = cσT U B γ 2 , (82)
dt p,syn 3 mp
where U B = B 2 /8π.
For a large range of proton spectra, the γ-ray power from photohadronic processes
can be approximated by the number N p,0 of protons with Lorentz factor γ ≈ γ p,max ,
from which follows that
280 C. D. Dermer
2
mp 3πd L2 Φ
N p,0 = 4 cσ U γ 2
(83)
me δD T B p,max
and
2
mp 9πm p c3 d L2 Γ 2 Φ Γ12
2Φ
−10
L j, p = ≈ 3.2 × 1044 erg/s, (84)
me 2rb δD
4 cσ U γ
T B p,max
4 B 2 γ
δD 50 9
where the bolometric γ-ray energy flux supplied by the proton synchrotron process
is 10−10 Φ−10 erg/cm2 -s. The particle power is reasonable, even if γ9 → 0.1. The
proton synchrotron frequency, from Eq. (74),
3δD m e B 2
p,syn ∼
= γ ∼
= 1.1 × 104 γ92 B50

δ12 , (85)
2(1 + z) m p Bcr p,max
so
E p,syn ∼
= 5.6γ92 B50

δ12 GeV. (86)
The equipartition field is, from Eq. (49), or Eq. (7.81) [4]
4/7 2/7
d28 (1 + z)5/7 f −10 Λ2/7
Beq (G) ∼
= 130 6/7 13/7 1/7
. (87)
td δ D ν13
In the case of Mrk 421, ν13 = 104 ν17 , Λ = 10Λ10 , td = (1+z)rb /cδD = 1200/δ12 s,
implying that
2/7 2/7
f −10 Λ10
Beq (G) ∼ = 4.5 1/7
, (88)
δ12 ν17
so that this model is magnetically dominated with a magnetic field ≈10 × Beq .
The minimum bulk Lorentz factor from γγ constraints, using Eq. (45), is Γmin =
20.3 f −10 [E 1 /(10 TeV)]1/6 = 13.8 f −10 [E 1 /(TeV)]1/6 , for a maximum measured
photon energy E 1 and using a variability timescale of ≈103 s. Note that f −10 rep-
resents the ν Fν flux for the target photons and, though both have the same units, is
a differential quantity compared to the bolometric energy flux Φ. The condition for
γ-ray transparency is somewhat inconsistent with the detection of multi-TeV photons
from Mrk 421.
The cooling regime for ultra-high energy protons through synchrotron cooling is
/t , with strong synchrotron cooling when ρ 1
determined by the ratio ρ ≡ tsyn dyn
and weak cooling when ρ 1. From Eq. (72) we have
3(m p /m e )3 140
ρ= → . (89)
4σT U B γ p rb γ9
Thus this proton synchrotron model for Mrk 421 is in the weak cooling regime.
Alternately, we can use the expression [171] for the saturation Lorentz factor

mp Bcr 9 3 × 1010
γsat = = , (90)
me B 4α f
B50
for which particles with larger Lorentz factors are in the strong cooling regime (here
α f = e2 /c ≈ 1/137 is the fine structure constant).
We can use Eq. (71) to determine the photopion production efficiency for this
model. For pk = 1017 ν17 /1.24 × 1020 = 8.1 × 10−4 ν17 , a ∼ = 0 and b ∼
= −1, we
obtain
5.6 × 10−7 f −10 f −10
→ 4.3 × 10−5
pk
η pγ = , (91)
t (days)(Γ /12) ν17
4 ν17
taking tv = (1 + z)rb /cδD = 0.013/δ12 days. The comoving Lorentz factor of

protons interacting with target photons with energy pk is γ pk ∼
= 6 × 106 . Protons
∼
with larger Lorentz factors interact primarily with the a = 0 portion of the ν Fν
spectrum with frequencies less than ν pk . The efficiency of proton energy loss with
γ γ pk approaches

Ep ∼ f −10 γ9
= 7.4 × 10−3
pk
η pγ (E p ) ≈ η pγ . (92)
E pk ν17
For photopion losses (as for proton synchrotron losses), the energy loss is most
efficient for the highest energy particles, reaching ∼1 % for γ9 ∼ 1. Somewhat
larger efficiencies are allowed—provided these allowances are consistent with the
variability data—if the protons are trapped on long times compared with the crossing
time, and the emission region is slow to expand.
In conclusion, hadronic models for blazars and GRBs face no insurmountable
objections based on power or energetics.
4.4 Cascade Halos and the Intergalactic Magnetic Field (IGMF)
The magnetic field BIGM in the IGM is bounded by a primordial field generated by
quantum fluctuations in the early universe or decoupling transitions of the funda-
mental forces [144]. Dynamo processes amplify the seed fields. Gamma ray astron-
omy provides a method for measuring the intergalactic magnetic field (IGMF) by
exploiting the effects of the γγ → e+ e− pair-production process on γ rays from
extragalactic TeV sources interacting with soft photons of the EBL.
When VHE γ rays interact with ambient radiation fields and the EBL, then blazars,
and therefore radio galaxies (i.e., misaligned blazars) are surrounded by anisotropic
282 C. D. Dermer
Table 7 Derived limits on

1ES 0229+200 θj (rad) BIGM (G)
BIGM for the source 1ES
0229+200 assuming Reference [166] π 3 × 10−16
persistent emission Reference [177] 0.1 5 × 10−15
Reference [178] 0.03 2 × 10−15
Reference [143] 0.1 5 × 10−15
jets of relativistic leptons made when the γ rays materialize into energetic e+ e− pairs.
These leptons spiral in the ambient magnetic field to make extended synchrotron
radiation halos [126], and GeV γ rays by Compton-scattering photons of the CMBR
[138]. Arrival time information in pair cascades generated by impulsive sources of
high-energy, multi-TeV γ radiation can also be used to infer the strength of the IGMF
[168, 164].
Because the emission of 1–10 TeV photons from a source at redshift z 1 is
attenuated by, primarily, the IR EBL, then the cascade spectrum can be calculated
for a given spectral model of the EBL and properties of BIGM . For sufficiently weak
magnetic fields, the pairs travel rectilinearly while Comptonizing CMB photons to γ
ray energies. The absence of this cascade signature in joint Fermi-HESS observations
of candidate TeV blazars implies a lower limit on BIGM [138] under the assumption
that the blazar persistently emits high-energy radiation for arbitrarily long time.
Neronov & Vovk [166] and Tavecchio et al. [177] argued that nondetection of the
TeV blazars 1ES 1101-232 (z = 0.186), 1ES 0229+200 (z = 0.14), 1ES 0347-
121 (z = 0.188), and H 2356-309 (z = 0.165) by Fermi implies a lower limit
BIGM 3 × 10−16 G (Table 7).
The high-energy electrons and positrons also undergo deflections in the ambient
magnetic field, so the emissions arriving latest generally come from leptons that
have been most severely deflected. This will cause steady sources to be surrounded
by an extended angular halo formed by leptons deflected back into the line of sight.
Ando & Kusenko [129] claim that ∼30 halos are found in stacked Fermi data for
170 AGNs that are bright at 10–100 GeV. By fitting the angular distribution of the
halo, they deduce that BIGM ≈ 10−15 (λcoh /kpc)−1/2 , for magnetic-field correlation
length λcoh ∼ 10–100 kpc. This claim is disputed in [167], in part because of the use
of P6_v3 response functions which are known to have inaccurate PSF from on-orbit
calibration data [8].
More realistic limits on B I G M F based on evidence that the blazar was operating
only during the time frame over which it was observed, was proposed in [142], and
independently, in [143].
Cascade Radiation from EBL Attenuation of TeV Photons
Consider a source and observer separated by distance d, as shown in Fig. 13. Photons
with dimensionless energy 1 = hν1 /m e c2 emitted at angle θ1 to the line of sight
between the source and observer, travel an average distance λγγ = λγγ (1 , z) before
λ (E )
psf GeV
λ (E ) θ = θBw
dfl
γγ TeV
θ
θ psf
j θ
TeV Source d Observer
Fig. 13 Sketch of the geometry of the event. A photon with energy E TeV TeV emitted at angle
θ1 ≤ θj to the line of sight, where θj is the jet half-opening angle, interacts with an EBL photon to
create an electron-positron pair. The electron is deflected by angle θdfl and scatters a CMB photon
to energy E GeV GeV and angle θ, which is detected as a source photon by the Fermi LAT when θ
falls within the angular point spread function θ ps f (GeV) from the source direction
materializing into an electron-positron pair via γγ absorption with photons of the

EBL. After production, the pairs cool by scattering CMB radiation, which is detected
at an angle θ to the line of sight to the source when the secondary electrons and
positrons (hereafter referred to as electrons) are deflected by an angle θdfl . Thus
θ1 = θdfl − θ. For the purposes of the argument, we neglect redshift effects for the
TeV blazar sources under consideration, which limits the treatment to sources at
z 0.2 (see [165] for redshift corrections).
The time delay Δt between the reception of photons directed towards the observer
and those which undergo the process described above is given through the expression
d sin(θdfl − θ)
cΔt = λγγ + x − d = λγγ + −d
sin θdfl
= λγγ (1 − cos θdfl ) − d(1 − cos θ), (93)
noting that x = d sin θ1 / sin θdfl and λγγ = d sin θ/ sin θdfl . In the limit of small
observing and deflection angles, Eq. (93) implies

∼ λγγ 2 λγγ
Δt = θ 1− . (94)
2c dfl d
Moreover, this time delay is observed by photons received at
λγγ
θ∼
= θdfl (95)
d
to the line of sight.
The deflection angle depends on the Lorentz factor γ = 106 γ6 of the produced
electrons, and can be written in terms of the received photon energy E = E GeV GeV.
The average photon energy of the CMB at low redshift is 0 ≈ 1.24 × 10−9 in m e c2
units, so that the mean Thomson-scattered photon energy is T ≈ (4/3)0 γ 2 . Thus,
an electron with Lorentz factor γ scatters CMB radiation to photon energy E when
284 C. D. Dermer
√
γ6 ∼
= 1.1 E GeV . The characteristic length scale for energy losses due to Thomson
scattering is
3m e c2 0.75
λT = = Mpc, (96)
4σT u CMB γ γ6
where u CMB ∼ = 4 × 10−13 erg cm−3 is the CMB energy density at low redshifts.
While losing energy, the electron is deflected by an angle θB ∼ = λT /rL in a uni-
form magnetic field of strength BIGM = 10−15 B−15 G oriented perpendicular to the
direction of motion of the electron, where the Larmor radius rL = m e c2 γ/eB ∼ =
0.55(γ6 /B−15 ) Mpc. Thus the deflection angle for an electron losing energy by scat-
tering CMB photons to energy E in a uniform field is θ B = λT /rL ∼ = 1.1B−15 /E GeV .
Introducing a coherence length λcoh that characterizes the typical distance over which
the magnetic field direction changes by ≈π/2, then the deflection angle

1,
i f λT < λcoh
θdfl ∼
= θB λcoh ≡ wθB , (97)
λT , i f λT > λcoh .
√
with w ≡ H (λcoh − λT ) + λcoh /λT H (λT − λcoh ).
The EBL model of [145] for sources at z = 0.14 gives λγγ (E) ∼ = 200 Mpc,
125 Mpc, and 70 Mpc at E = 1, 3, and 10 TeV, respectively. A low EBL based on
galaxy counts [156] gives values of λγγ (E) ∼ = 280 Mpc, 150 Mpc, and 85 Mpc at
E = 1, 3, and 10 TeV, respectively. Thus we write λγγ = 100λ100 Mpc, so that
λ100 ≈ 1 gives a minimum value of Δt and θ for the reprocessed TeV radiation.
For a source at distance d = dGpc Gpc, with dGpc ∼ 1 corresponding to z ∼ 0.2,
the time delay for emission observed at angle

λ100 B−15 w
θ∼
= 0.01 (98)
dGpc E/10 GeV
from the line of sight is given from Eq. (94) by

2
B−15 w λ100
Δt (year) ∼
= 2 × 106 λ100 1 − 0.1 (99)
E/10 GeV dGpc
Equation (99) shows that small time delays are implied when λγγ is small or
λγγ ≈ d. When λγγ λT , an additional delay ≈λT θdfl 2 /c arises during the time
that the electrons are losing energy and being deflected by the IGMF [153, 169].
Such small values of λγγ ∼ 1 Mpc are only relevant at low redshifts to 100 TeV
photons pair producing within ≈1 Mpc of their source which, however, may be in
the galaxy cluster environments where the magnetic field is not representative of the
dominant volume of the voids. Thus we can dismiss such an origin of a short time
delay without assuming special properties of the TeV sources. The case λγγ ∼ d
formally gives short time delays, but this corresponds to the case when the mean
free path for γγ pair production is about equal to the source distance, which occurs
for γ rays with energies of several hundred GeV when d ∼ 1 Gpc. In this case, the
secondary electrons which take half the energy of the high-energy photon scatter
CMB photons to 100 MeV. Even so, the attenuation of the high-energy photons
takes place over the entire distance d, so that the upscattered photons with short delay
time comprise only a very small fraction of the incident flux attenuated close to the
observer.
The remaining alternative to avoid assuming that TeV blazars are steady on
timescales of millions of years is to suppose that either BIGM 10−15 G or that
λcoh 1 Mpc. If λcoh ∼ 1 Mpc, this contradicts the claim that the IGMF has been
measured to be BIGM 10−15 G. If λcoh 1 Mpc, then the field must be even
larger in order that the electrons are deflected away from the direction of the photon
source. Thus TeV blazars must be assumed to be steady emitters on long timescales.
Here we relax this assumption, and suppose most cautiously that the blazar has been
operative over the last few years that the they have been observed. This reduces the
implied lower limit by several orders of magnitude, as will be seen. Here we treat the
blazar as a point source; see [142] when the Fermi-LAT PSF constraint is included
(Fig. 13).
Derivation of the Cascade Spectrum
Consider a source that emits TeV-scale photons from within a photon beam of half-
angle θ j , and an apparent isotropic spectral luminosity ∗ L ∗ (∗ ) within the beam
(“flat-topped” jet). If the TeV photons can escape from the nuclear environments
without γγ absorption, then they may still be attenuated by interactions with EBL
photons. After γγ attenuation, e+ -e− pairs lose energy, primarily by Compton scat-
tering the CMBR. This upscattered radiation will be detected by an observer until
the cooling pairs are deflected out of the beam.
We use the notation of Ref. [4]. From Eq. (7), denote the ν Fν spectrum by
∗ L ∗ (∗ )
f = ν Fν = , = hν/m e c2 , ∗ = (1 + z).
4πd L2
The correction due to EBL attenuation is
∗ L ∗ (∗ ) m e c2 2 Ṅ ()
f = exp[−τ γγ (; z)] = exp[−τγγ (, z)]. (100)
4πd L2 4πd L2
At low redshifts, z 1, ∗ ≈ , and L() = m e c2 2 Ṅ (), where Ṅ () is

the photon injection function. The number of photons surviving to the observer is
Ṅ () exp[−τγγ (; z)], so the number of absorbed photons is Ṅabs () = Ṅ ()[1 −
exp[−τγγ (; z)], and this also represents the electron injection function Ṅin j (γi ),
with normalization
286 C. D. Dermer
∞ ∞
dγi Ṅin j (γi ) = 2 d Ṅabs (), (101)
1 0
because each photon makes two leptons with γi ∼ = /2. So Ṅin j (γi ) = 4 Ṅabs () =
4 Ṅ (){exp[τγγ (, z) − 1]}. Using Eq. (100) gives the injection function
16πd L2 f
Ṅin j (γi ) = {exp[τγγ (, z)] − 1}, = 2γi . (102)
m e c2 2
This injection source of e+ -e− loses energy by Compton-scattering photons of the

CMBR to GeV energies, and this cascade GeV component is considerably dimmed
when the leptons are deflected out of the beam.
Photons with energy ≈1 TeV make leptons with γ ≈ 106 ,15 which scatter the
CMBR to γ 2 0 ∼ 103 , or to photon energies ∼500 MeV. Scattering is in the Thomson
regime for 4γ0 1 or γ 2 × 108 , that is, electrons with energies 100 TeV.
The Thomson energy-loss rate
dγ 4 u0 2
− γ̇T = − |T = cσT γ ≡ νT γ 2 , (103)
dt 3 m e c2
with u 0 /m e c2 = 4.9 × 10−7 cm−3 . The solution to the steady-state electron conti-
nuity equation is ∞
1
N (γ) = dγ Ṅ (γ ). (104)
νT γ 2 γ
Limiting γ > γd f l , where γd f l is the deflection Lorentz factor where the lepton is
deflected out of the jet opening angle by the IGMF. Taking the luminosity spectrum
from Compton scattering [4], Eqs. (102) and (28) give
s L C (s ) = 4πd L2 f s
2 ∞
12πd 2 cσT u 0 s FC (q, Γe ) ∞ f {exp[τγγ (, z)] − 1}
= dγ dγi ,
νT m e c 2 0 γlow γ 4
γ 2
(105)
with = 2γi . The interior integral is the injection function from high-energy γ
rays absorbed by photons of the EBL, and depends on a model of the optical-depth
function τγγ (, z) for a photon with detected energy that was emitted by a source
at z. In general, one uses the expression Eq. (31) for the Compton kernel FC (q, Γe ),
though the isotropic Thomson kernel, Eq. (33), is sufficiently accurate our purposes
here.
15 At high energies, most of the energy is taken by one of the leptons. See, e.g., [157].
The lower limit γlow in Eq. (105) is the maximum of various constraints given
by kinematic factors, engine duration, and γd f l . The kinematic
√ Lorentz factor γknm
given by Eq. (35). In the Thomson regime, γknm → (1/2) s /0 .
The deflection Lorentz factor is determined by the condition that the energy-loss
timescale is equal to the timescale for e+ -e− pairs to be deflected out of the beam. The
Thomson energy loss timescale −γ̇T = νT γ 2 implies tT = 1/νT γ. The deflection
timescale
θ j rL m e cγ
td f l = = θj , when λT < λcoh , (106)
c eB
√
and td f l = θ j (m e cγ/eB) λT /λcoh when λT > λcoh . Solving gives
⎧
⎪
⎨ θ j m e cνT ,
eB
γd f l > c
νT λcoh
γd f l = 1/3 2/3 . (107)
⎪
⎩ cλνcoh eB
, γd f l < c
T m e c2 θ j νT λcoh
This constraint implies γd f l = 3.7 × 106 B−15 /θ−1 for γ > 7.5 × 105 /λcoh (Mpc),
1/3
and γd f l = 6.2 × 106 λcoh (Mpc)(B−15 /θ−1 )2/3 for γ > 7.5 × 105 /λcoh (Mpc).
To avoid solving a time-dependent electron continuity equation, we introduce
electron Lorentz factor limits γ(Δt) to define the time the engine was operating.
Following Eq. (34), noting that f s = s L(s )/4πd 2 , gives

3 s 2 ∞ −4 s
f s = dγ γ 1− 2 ,
2 0 max 40 ,γd f l ,γ(Δt)
s 4γ 0
∞
f {exp[τγγ (, z)] − 1}
× dγi , (108)
γ 2
with = 2γi .
The minimum Lorentz factor γ related to period Δt of activity of central engine is
determined by equating the time delay with the extra pathlength followed by photons.
Thus
∼ λγγ (1 ) + λT (γ) θ2 ,
Δt = (109)
dfl
2c
and λtot = λγγ (1 ) + λT (γ) = 100λ100 Mpc, θd f l = wθ B , θ B = λT /rL . From this
we derive
γ(Δt)
⎧ 1/4
⎪
⎪ λtot ∼
1/4 1/2
9.9×109 λ100 B−15 7.5×105
⎨ meB
e cνT 2cΔt = [Δt (s)]1/4
, λcoh (Mpc) < γ
= 1/3 . (110)
⎪
⎪ 2/3
λtot λcoh ∼
1/3 2/3 1/3
2.3×1011 λ100 B−15 λcoh (Mpc)
⎩ eB 2 = ,γ< 7.5×105
me c 2νT Δt [Δt (s)]
1/3 λcoh (Mpc)
288 C. D. Dermer
Note that there is no separate constraint that the electrons have cooled long enough
to scatter significant emission in the given band. In the stationary frame, one can
define the cooling electron Lorentz factor γcool giving the characteristic Lorentz
factor of electrons that have cooled in time Δt, expressed in terms of the GeV photon
energy E GeV to which CMBR photons are Compton-scattered CMB photons. At low
redshifts,
√
3 3 0 ∼ 7.7 × 1019 s ∼ 2.2 Myr
tT = Δt = = = = √ , (111)
4cσT u 0 γcool (Δt) 2cσT u 0 γ E GeV
we see that it takes nearly a million years for the generated electrons and positrons to
cool and begin to make their strongest emissions in the GeV–10 GeV band of Fermi.
But this applies for pair halo emissions made at large angles to the jet axis. The
cooling timescale compared to the engine starting time, as measured by an observer
within the jetted emission cone, is instead given by Eq. (109), which we write as
2
B−22
Δt (s) ∼
= 60 λ100 2
w2 . (112)
E GeV
The lack of distinct GeV echoes from impulsive or flaring high-energy sources can
limit parts of the BIGM -λcoh parameter space. Better yet, the discovery of such echoes
in GRB light curves, as originally proposed by Plaga (1995) [168], would finally
reveal the primordial magnetic field. From Eq. (111), the Thomson cooling length of
relativistic electrons scattering the CMBR is, at redshift z, given by
2.3 × 1030 cm 0.7 Mpc

λT = √ , (113)
γ(1 + z) 4 E GeV
where the last expression applies at low redshifts.

Figure 14 shows calculations with this semi-analytic model [142]. Here we use
a photon-energy dependent expression for λγγ and perform an integration over the
CMBR spectrum. If the jet is persistent on long time scales, then the jet opening angle
is limited to be 0.4, as can be seen in the middle panels of this figure. Restricting
TeV activity of 1ES 0229+200 to ≈3–4 years, during which the source has been
observed, leads to a more robust lower limit of BIGM 10−18 G, which can be
larger by an order of magnitude if the intrinsic source flux above ≈5–10 TeV from
1ES 0229+200 is strong.
If there were no intergalactic magnetic field at all, then the pairs made from a
source at distance d, minus the distance λγγ over which they are made, will have
cooled to electron Lorentz factors γ given by d − λγγ = λT ∼ = 750 Mpc/(γ/103 ).
It is interesting to think that the pair injection process will have seeded the voids of
intergalactic space with an ultra-relativistic nonthermal electron-positron component
that has already cooled to low energies. Given how active TeV sources are in our
(a) (d)
(b) (e)
(c) (f)
Fig. 14 Model of cascade radiation spectrum, Eq. (100), applied to HESS, VERITAS, and Fermi
observations of 1ES 0229+200, using model spectra (solid curves) and EBL model of [145] to give
attenuated source spectra. a Cascade spectra for 1ES 0229+200 assuming persistent TeV emission at
the level observed with HESS and VERITAS, for different values of BIGM and λcoh = 1 Mpc (solid)
or λcoh = 100 kpc (dot-dashed) for a jet opening angle θ j = 0.1. The intrinsic TeV source spectrum
is given by power-law with ν Fν index = 4/5 with super-exponential cutoff ∝ exp[−(E/5 TeV)2 ].
The PSF constraint for the λcoh = 1 Mpc case is shown by the dot-dashed curves. d Same as for a,
except that θ j = 0.3. e Same as a, except that θ j = 1.0. b, c, d Here the TeV engine operates for
3 years, 3 years, and 100 years, respectively, but the intrinsic TeV source spectrum differs. In b, it
is the same as a. In c, it has an exponential cutoff ∝ exp(−(E/10 TeV). In f, the intrinsic source
spectrum use parameters of Dolag et al. (2011) [143]. In (b), (c), and (d), λcoh = 1 Mpc and cascade
spectra are calculated for different values of BIGM , as labeled
290 C. D. Dermer
present declining phase of the universe (in terms of star formation activity), TeV
sources must have been vigorous operating in the early universe.
Without making severe assumptions about the γ-ray duty cycle and radiative
behavior of blazars like 1ES 1101-232 and 1ES 0229+200 on long timescales, the
best limit to the strength of the IGMF is 10−18 G for λcoh 1 Mpc [142, 143].
Constraints on the value of BIGM 3 × 10−19 G can be obtained from a search for
pair echos in the analysis of GRB data [163, 174], so if the larger field is correct,
then no GeV echo radiation is predicted from GRBs.
5 γ Rays from Cosmic Rays in the Galaxy
Cosmic rays are the most energetic particles in the universe, and sources of
• the light elements Li, Be, B;
• the Galactic radio emission;
• the Galactic γ-ray emission;
• Galactic pressure;
• collisional excitation of atoms and molecules;
• terrestrial 14 C and 10 Be, with half lifes of ≈5700 years and ≈1.5 Myr, respectively;
and
• astrobiological effects.
Cosmic rays are composed mainly of protons and ions, but also include energetic elec-
trons, positrons, and antiprotons.16 They make up an important particle background in
the space radiation environment and contribute to the space weather. Cosmic-ray elec-
trons emit radio and X-ray synchrotron radiation, X-ray and γ-ray bremsstrahlung
radiation, and X-rays and γ rays from (“inverse”) Compton scattering [207]. It is
believed with good reason that Galactic GeV–PeV cosmic rays are accelerated by
supernova remnants [221, 225]. The origin of the UHECRs, and its relation to the
origin of the cosmic rays, is an open question; the hypothesis that their origin involves
rotating black holes is developed in [4].
The difficulty to solve the problem of cosmic-ray origin for the ∼GeV cosmic-
ray protons and ions that carry the bulk of the cosmic-ray energy content is that,
being charged particles, they do not point back to their sources as a consequence of
intervening magnetic fields that deflect them in transit. Sites of high-energy particle
interactions are identified by γ rays, but ascertaining whether the emission is made
by hadronic cosmic rays is complicated by the possible leptonic origin of most of the
γ rays. GeV–PeV neutrinos, by comparison with charged cosmic rays, unambigu-
ously point to the sources of the cosmic-ray hadronic interactions, but are faint and
difficult to detect.
Cosmic-ray models use strong nuclear interaction cross sections and acceleration,
loss and transport physics to derive the cosmic ray intensity and spectrum, whether in
16 High-energy neutrinos could also, depending on definition, be included.

supernova remnants, diffuse and extended clouds of gas, galaxy clusters, the Galactic
Center region, or wherever there is a significant column of gas with an illuminating
cosmic-ray flux. The Sun and Solar flares are especially instructive for study of
particle acceleration, transport, and radiation physics from γ-ray observations.
5.1 γ Rays from Solar System Objects
Besides direct observations, the effects of cosmic rays in the Solar system are traced
via the cosmic-ray induced γ-ray flux of the Sun, moon, and Earth. In the Solar
cavity, the cosmic-ray intensity is modulated by the outflowing Solar wind plasma,
making an anti-correlated decrease and lag in the 10 GeV/nucleon cosmic-ray flux
reaching in the Solar cavity with a period of the 11-year sunspot cycle (one-half
the 22 year Solar cycle). Interaction of these cosmic rays with Solar system objects
make GeV γ-ray flux that varies on these timescales. With 3 years of data taken with
Fermi, this only amounts to 30 % of the sunspot cycle, the first 2.5 years of which
were taken with the Sun in a deep and extended Solar minimum.
Due to its proximity, the Earth is the strongest γ-ray source for the Fermi-
LAT, which is why zenith-angle cuts are made on source spectral reconstruc-
tion to eliminate the interfering effects of the cosmic-ray induced γ-ray “albedo”
emission (recall footnote 5). The γ-ray spectrum of the Earth albedo depends on
angle to the nadir. At Fermi-LAT’s h = 565 km orbit, it views the horizon at
θndr ≈ arcsin(1 + h/rE )−1 ∼ = 66.5◦ ; the Earth’s radius is rE = 6378.1 km. Fermi is
therefore exposed to (1/2)(1 − cos(π − θndr )) ≈ 70 % of the full sky, with the Earth
occulting ≈30 % of the sky.
Observations of Earth albedo flux reveals a number of interesting things. One
is the exposure bias toward the North Ecliptic Pole due to favored rocking to the
North. Recall that Fermi rocks about zenith, early in the mission by 39◦ , which was
increased to 50◦ later. Because of the intensity of the Earth albedo, a standard analysis
cut is to accept photons only within 105◦ of zenith. For studies of the albedo, then,
an acceptance of θndr 80◦ gathers mostly cosmic-ray induced terrestrial γ-ray
emission, i.e., albedo. Knowing the albedo spectrum gives, empirically, a better
characterization of the γ-ray background, and is useful for Fermi-LAT calibration.
Deconvolving the γ-ray spectrum with an Earth atmosphere model and knowledge
of the interaction cross sections would give the primary cosmic-ray spectrum and
information about the deflection of primary cosmic rays by the geomagnetic field.
In [180], two data sets are gathered: one during the first 90 days of the Fermi
mission, and a second for a two-orbit pointing at the Earth’s limb. The analysis
covers a total of 6.4 × 106 events giving the albedo spectrum in the 100 MeV–TeV
range, with 218 ph(>100 GeV), and 16 ph(>500 GeV).
From two-dimensional intensity maps with increasing energy range, a beautiful
high-energy ring forms above 3–10 GeV due not to Fermi’s energy-dependent PSF
but to beaming of the emission during formation (see Fig. 15). The bright limb at
292 C. D. Dermer
Fig. 15 Exposure maps (top) and intensity maps (bottom) of terrestrial γ-ray albedo emission [180]
the Earth’s horizon are γ-ray light showers made by grazing incidence cosmic rays
coming directly towards the LAT.
The next interesting feature of albedo emission is the angle-dependent spectrum.
At the nadir, the spectrum is soft, and generated by γ-rays that are backscattered by
large angles, which is considerably less likely due to threshold and beaming effects
at 1–10 GeV/nucleon, where the particle Lorentz factor γ a few. The deflection
of cosmic rays by the geomagnetic field makes an east-west effect that fades out
at high energies. The spectral intensity rises and hardens until the Earth’s limb is
reached. See [180] for details.
For the moon, without an atmosphere, a steep secondary nuclear production spec-
trum with flux F−8 = 110 ± 20 [220] is made as cosmic-ray GeV protons and ions
impact the surface of the moon, confirming the EGRET detection [256]. A search
for γ-ray emission from asteroid populations and other Solar system rocks and dust
[238, 239] is currently in progress.
The γ-ray emission from the Sun consists of two components: cosmic rays impact-
ing the surface of the Sun to make an albedo-type emission, and cosmic-ray electrons
Compton-scattering solar photons to γ-ray energies [237, 240, 241]. Both compo-
nents of emission are sensitive to the phase of the Solar cycle, and both components
have been detected. In analysis of 18 months of data [189], the solar disk emission is
found at the level of F−8 ≈ 46. The measured integral flux of the extended non-disk
emission from a region of 20◦ radius centered on the Sun is F−8 ≈ 70. So the Sun and
moon are really bright γ-ray sources, and one has to be alert to data contamination
and spurious variations when the Sun or moon drift past.
Fig. 16 Fermi-LAT 2FGL 100 MeV–10 GeV all-sky map, using 2 years of sky survey data (2008
Aug 4–2010 Aug 1) [70]. Scale in units of 10−7 erg/cm2 -s-sr. The 2FGL consists of 1873 sources,
with 1170 associations and 127 identifications, of which 1319 are at |b| > 10◦
The Fermi-LAT measurement of the Solar emission confirms Seckel’s model

[247], but at a much higher (by a factor ≈7) flux. The disk emission is practically flat
in a ν Fν spectrum up to ≈8 GeV, where it begins to fall off. The observed spectrum
and angular profile is in good agreement with theoretical predictions for the quiet
Sun emission [237].
The first Fermi-LAT as well as nuclear γ-ray line flare in Solar Cycle 24 is the
2010 June 12 M3 flare [195]. The LAT emission lasted for only ≈50 s, compared
to past long-duration Solar γ-ray flares. Combined Fermi GBM and LAT analysis
reveals a rich γ-ray line spectrum superimposed on a continuum with a hard-energy
tail that is consistent with either a separate nonthermal bremsstrahlung component,
or pion production in energetic nuclear events. The hard bremsstrahlung X-rays and
the ≈300 MeV γ rays flare up within 3 s of each other, placing strong requirements
on acceleration theory and target properties. As we enter the active portion of the
Solar cycle, a wealth of new Fermi data on particle acceleration physics of large
Solar flares is anticipated.
Also interesting and timely is the use of Fermi for the study of terrestrial γ-ray
flashes [216] that accompany thunderstorms found mainly in the sub-tropics and
tropics. Experiments in nadir-pointing modes to increase sensitivity to TGFs are
currently underway. Interesting analysis effects having to do with the shortness of
the pulse duration and the strong soft X-ray emission affecting the ACD, which is
also important for Solar flares, reveals Fermi’s capability for this science.
294 C. D. Dermer
5.2 GeV Photons from Cosmic Rays
The cosmic-ray induced γ-ray glow of the Milky Way is the most pronounced and
distinctive feature of the Fermi sky (Fig. 16). A dominant fraction of the Galactic
γ-ray emission is believed to be truly diffuse, and made by cosmic-ray bombardment
of gas and dust in the interstellar medium (ISM). The most important hadronic process
for γ-ray production is secondary nuclear production from the collisions of cosmic-
ray protons and ions with ISM particles. The most abundant secondaries are pions
(others are kaons and heavier baryons and baryonic resonances). The pions decay
according to the scheme
π 0 → 2γ, π + → μ+ + νμ , π − → μ− + ν̄μ , (114)
μ+ → e+ + ν̄μ + νe , μ− → e− + νμ + ν̄e . (115)
For the purposes of Galactic cosmic-ray physics, the decays are essentially instan-
taneous, and result in γ-ray, electron, and neutrino injection emissivities (as defined
below, Eq. (118)) proportional to the cosmic-ray intensity and gas density at that loca-
tion. The secondary electrons and positrons, as well as those accelerated at cosmic-ray
sources, lose energy during propagation mainly by synchrotron and Compton losses
at high energies, bremsstrahlung losses at intermediate energies, and ionization and
Coulomb losses at low energies, and can also in principle be reaccelerated.
The diffuse Galactic γ-ray glow is the superposition of all the radiations made
by π 0 -decay γ rays, γ rays from cosmic-ray electrons that Compton-scattered the
available radiation fields, and electron bremsstrahlung γ rays. Electron synchrotron
radiation would only contribute to the Galactic γ-ray emission from localized sources,
such as pulsar wind nebula.
A simple expression for the demodulated cosmic-ray proton intensity in the
local interstellar space inferred from measurements of the near-Earth cosmic-ray
intensity is
J p (E p , Ω p ) = 2.2E −2.75
p CR p/cm2 - s-GeV-sr (116)
[211], so that the GeV cosmic-ray flux represents tens of cosmic-ray protons
per cm2 per s. The cosmic-ray kinetic-energy density is dominated by the kinetic
energy T p (= E p − m p c2 ) of ∼GeV protons, and is given by
∞
4π
uC R K E = dT p T p J p (E p , Ω p ) ∼
= 0.7 eV/cm3 , (117)
c 0
to which ions contribute another factor ∼30–50 %. The dominant elementary hadronic
process is p + p → π ±,0 , so astrophysical studies of secondary nuclear production
focus on this process. For more detail on radiative processes and cosmic-ray propa-
gation, see [205, 219, 230, 251].
Background Modeling
Reliability of source detection is improved if the intensity of the diffuse background

radiation is known, because a clumpy gaseous structure illuminated by cosmic-ray
induced γ-ray fluorescence could be mistaken for a point source. Moreover, knowl-
edge of the background is required to recognize dim sources, because the background
dominates for all but the brightest sources. The background model for Fermi analy-
sis takes into account diffuse Galactic γ rays from interactions of cosmic rays with
material found in the various phases of the ISM, including the neutral hydrogen, HI,
molecular hydrogen, H2 , ionized hydrogen, HII or H+ , and the dark gas phase [224].
The distribution of neutral atomic hydrogen (HI) is traced by 21 cm line surveys. The
molecular hydrogen (H2 ) distribution is derived indirectly, most commonly by using
2.6 mm line observations of carbon monoxide (CO). (The ratio of H2 to CO—the
so-called ‘X ’-factor, X = N (H2 )/WCO 17 is derived from γ-ray observations.) The
total atomic and molecular gas column density can also be traced indirectly from
extinction and reddening by dust, depending on the relative fraction of dust and gas.
Cosmic rays also interact with ionized hydrogen. The low-density ionized gas can
be inferred from dispersion measures of pulsar signals in the radio band.
Cosmic rays move in large-scale galactic magnetic field and diffuse by scattering
off magnetic turbulence. In the thick disk of the Galaxy, where the bulk of the
diffuse γ-ray emission is made at |b| 5◦ –10◦ , the cosmic-ray intensity changes
with galactocentric distance since the source distribution is peaked at star-forming
arms at 4–6 kpc. The cosmic-ray intensity also changes with the distance from the
Galactic plane because of escape, and because the diffusion coefficient is energy
dependent, with high-energy particles diffusing faster through the Galaxy. In addition
to the CMBR, the different distributions of background optical and IR fields, and
location-dependent Galactic magnetic fields, means that cosmic-ray electrons suffer
position-dependent energy losses.
For the electronic component, gradients in the particle distribution can be severe
and contribute a Galactic background contribution to the γ-ray emission. The
GALPROP (GALactic cosmic ray PROPagation) model [242, 248–250], started in
1996 and developed independently of Fermi, determines cosmic-ray diffusion coef-
ficients from fits to cosmic-ray data. The spatial and momentum diffusion equations
for cosmic-ray transport are solved, taking into account source injection, energy and
fragmentation losses, and energy changes for cosmic-ray protons, ions, and electrons.
The stellar optical field, assorted IR and PAH lines in the 10 µ valley, and a FIR dust
peak at ∼100 µ [242] provide target photons to be Compton scattered by relativistic
electrons. The GALPROP model is constrained by the energy dependence of the B/C
and 9 Be/10 Be ratio, from which predictions for the e+ , e− , p̄, and γ-ray spectra and
intensity can be made.
Template modeling of γ-ray emission from Gould belt clouds in Cassiopeia and
Cepheus [184] shows a weak Galactocentric gradient from the Gould belt to the
17N (H2 ) is the column density of molecular hydrogen, and WCO is the brightness temperature of
CO integrated over velocity [233].
296 C. D. Dermer
Perseus arm, and an increase in the X -factor from ∼

=0.87 × 1020 cm−2 (K km s−1 )−1
in the Gould belt clouds to ∼
=1.9 × 10 cm (K km s−1 )−1 in the Perseus arm. The
20 −2
dark gas represents ≈50 % of the mass traced by CO.
Diffuse Galactic γ Rays from Cosmic Rays
There can be little doubt that cosmic-ray interactions make a large fraction of the
γ rays observed with Fermi. This is established most clearly by Fermi-LAT obser-
vations [182] towards a region in the third quadrant between Galactic longitude
200◦ –260◦ and latitude 22◦ –60◦ that contains no known molecular clouds. After
subtracting point sources and Compton emission, the residual 100 MeV–10 GeV γ-
ray intensity exhibits a very strong linear correlation with atomic gas column density.
According to the model of Cordes & Lazio (2002) [209], the N(HII) column
density is only (1–2) × 1020 cm−2 and fairly smooth in the third quadrant regions
analyzed in Ref. [182]. The contribution from ionized gas has a small effect on the
measured emissivity, as the N(HI) column density ranges from (1–13) × 1020 cm−2 .
The measured integral γ-ray emissivity measured with the Fermi LAT is 1.63 ±
0.05 × 10−26 photons/s-sr-H-atom and 0.66 ± 0.02 × 10−26 photons/s-sr-H-atom
above 100 MeV and above 300 MeV, respectively, with an additional systematic error
of ∼10 %.
These numbers are explained in the first approximation if cosmic rays pervade
the gaseous disk of the Milky Way with the same intensity, Eq. (116), as observed
locally. The pion production rate per unit volume, or differential photon emissivity, is
∞
dσ p H →π0 (T p )
ṅ p H →π0 (Tπ ) = 4πn H dT p J p (T p , Ω p ) , (118)
0 dT π
and the bolometric π 0 γ-ray production rate per H atom is, therefore
∞
dN γ
qγ ≡ /n H ∼
= 2ζ dT p J p (T p , Ω p )σ pp→π0 (T p ), (119)
dtdV dΩ T p,thr
including a factor for two photons per π 0 and a metallicity correction ζ ≈ 1.5.
Between E p ≈ 1.3 GeV, just above threshold, and E p ≈ 10 GeV, the inclusive π 0
production cross section is approximately linear, and can be written as
σ pp→π0 (mb) ∼
= 4 mb(E p /1.3 GeV),
from which, with Eq. (116), we find that

ζ
qγ ∼= 2.2 × 10−26 [s-sr-H-atom]−1 , (120)
1.5
roughly agreeing with the Fermi-LAT measurements [182] quoted above.

Fig. 17 Galactic diffuse emission intensity as measured in the latitude range 10◦ ≤ |b| ≤ 20◦
[42]. (left) LAT Galactic diffuse γ-ray intensity data are given by red dots and red cross-hatched
error regions; EGRET data are given by the blue crosses and blue shaded regions. (right) LAT
data compared to a model consisting of π 0 decay γ rays (red); bremsstrahlung from primary and
secondary electrons and positrons (magenta); Compton-scattered soft photons to γ-ray energies by
cosmic-ray electrons (green); unidentified background consisting of point sources, isotropic diffuse
Galactic, and the isotropic extragalactic γ-ray background
The intermediate latitude, 10◦ ≤ |b| ≤ 20◦ , Galactic diffuse γ-ray emission
intensity spectra as measured with the Fermi-LAT and EGRET are shown in the left
panel of Fig. 17 [42]. The prominence of the π 0 decay feature is apparent, as also
are the large systematic differences at both low, 300 MeV, and high, 300 MeV
energies. In the right panel of Fig. 17, the LAT spectrum shown in the left panel is
compared with spectra from an a priori model (based only on local cosmic-ray data)
for the diffuse Galactic γ-ray emission, updated from GALPROP. As the Fermi LAT
accumulates data, and analysis becomes even more accurate below 100 MeV for
the diffuse class, we can foresee using Fermi data to determine the best interstellar
cosmic-ray proton and He ion spectra rather than the other way around.
Other Fermi LAT Cosmic Ray Results
CR e spectrum: The Fermi LAT can be used as a cosmic-ray electron detector by using
its anticoincidence dome to identify incoming charged particles, and distinguishing
between leptons and ions from the tracks in the tracker and the calorimeter. The
Fermi telescope has an acceptance of >2 m2 -sr for combined cosmic ray e+ and e−
(CR e), since it cannot distinguish charge directly.
The extraordinary statistics of the Fermi enabled a precise measurement of the
CR e spectrum, showing a featureless spectrum consistent with a power law of
number index ∼ =−3.04 between ≈25 and 900 GeV [183]. The CR e spectrum is
298 C. D. Dermer
harder than a GALPROP model prediction where the diffusion coefficients tuned to
the low-energy electron spectra are extrapolated to higher energies. The measurement
disagrees with the ATIC report [208]. The PAMELA measurement of CR e+ [197]
and the cosmic-ray electron spectrum inferred at TeV energies with HESS data [201],
in addition to the Fermi measurement of the cosmic-ray electron spectrum, place
constraints on cosmic-ray electron models that require either (or a combination of)
modifications to the propagation characteristics; local sources, most likely pulsars
[259]; or contributions from dark matter (see lectures by Prof. Bergström and [243]).
The 2010 Fermi analysis gives the CR e spectrum from ≈7 GeV to 1 TeV, with a
slight hardening above 100 GeV and a softening above 500 GeV [192].
EGRET excess: This term refers to EGRET measurements of the diffuse Galactic
γ-ray intensity that were found, irrespective of direction, to be in excess of that
predicted using the local demodulated cosmic ray spectrum and measured target gas
mass [227]. Possible explanations included an
1. unusual location and local cosmic-ray spectrum measured here at Earth;
2. nuclear physics wrong;
3. addition of γ ray signal from annihilating dark matter; or
4. EGRET miscalibration.
With the launch of the Fermi Gamma ray Space Telescope, measurements [42] of
the diffuse Galactic emission have been found to favor the latter hypothesis, namely,
that EGRET was poorly calibrated above ≈5 GeV. The differences in the EGRET
and Fermi LAT diffuse Galactic γ-ray emission are shown in the left panel of Fig. 17.
The GALPROP model decomposes the Galactic plane emission into pionic, elec-
tron bremsstrahlung and Compton fluxes, point sources, and an isotropic diffuse
background. The π 0 → 2γ signature is clearly seen in the spectrum of the diffuse
Galactic γ radiation [42]. Discrepancies between EGRET and Fermi spectra still
remain at 100 MeV, where systematics effects become severe for both the EGRET
spark chamber and the LAT tracker; see Fig. 17. The latest analysis of diffuse Galactic
γ radiation in the Third Galactic Quadrant finds weak evidence at best for a Galac-
tocentric gradient in the cosmic-ray intensity [193].
Fermi has also measured the positron flux and fraction using the Earth as a mag-
netic field [196].
5.3 Fermi Bubbles
It is not possible to cover, or even mention, all Fermi-LAT diffuse studies in this short
review, e.g., the Galactic center region, the Cygnus Loop [228], and γ-ray emission
from massive star clusters and OB associations like the cosmic-ray filled cocoon
region in Cygnus [194], but we would be remiss not to mention the Fermi bubbles
that D. Finkbeiner and colleagues find [212, 254]. In their analysis, backgrounds
determined by template fitting are subtracted from the γ-ray data, leaving (apparently)
well-defined structures—the Fermi bubbles—at ≈1–10 GeV, and which probably
extend to much higher energies. They are symmetrically arranged north and south
of the Galactic Center with a width of ≈40◦ , and extend some ≈50◦ in height. They
are correlated with the WMAP haze, which is excess 20–40 GHz emission found
after correcting for the dipole anisotropy and subtracting out emission traced by Hα
and soft synchrotron radiation extrapolated from the 408 MHz Haslam survey [215].
The 1–100 GeV luminosity of both bubbles, if emanating from the Galactic Center
region, is ≈2 × 1037 erg/s. With a minimum lifetime of ≈10 kpc/c, its minimum
energy content is 4 × 1049 erg. The bubble luminosity represents ∼5 % of the
100 MeV–100 GeV luminosity of the Galaxy [252].
The haze itself, if not an artifact of template fitting or mis-extrapolation of the
radio emission [234], could be formed by dust, spinning dust, or dark matter, but
a synchrotron origin seems favored, especially given the Fermi bubbles. Indeed, a
leptonic model for the bolometric SED of the bubbles, with synchrotron radiation
for the WMAP haze and Compton-scattered CMB γ rays for the Fermi-LAT emis-
sion, can be arranged by tailoring the electron distribution [254]. A detailed model
including inverse Compton scattering off the CMB, FIR, and optical/UV radiation
fields fits the average LAT γ-ray spectrum [235], which is practically flat between
1 and 100 GeV. There is furthermore no energetics problem in a leptonic model, but
it requires reacceleration and a low-energy cutoff in the electron spectrum. The sharp
edges reported for the Fermi bubbles, to be confirmed in ongoing Fermi-LAT analy-
ses, stand in contrast to the WMAP haze. A specific hadronic model [210] suffers
from weak radiative efficiency in the dilute wind from the Galactic center, and lateral
diffusion would smear the edges of the bubbles.
Of even greater interest is what the existence of the Fermi bubble says about our
Galaxy. Was it a giant Galactic explosion, the residuum of a more dynamic period in
the history of our Galaxy, a superwind from a past starburst episode, or something
else?
5.4 γ-Ray Supernova Remnants
The cosmic-ray power PC R can be estimated as

1 eV/cm3 1.6 × 10−12 erg/cm3 × 4 × 1066 cm3
PC R ∼ VM W ∼
tesc 20 Myr
∼ 1040 erg/cm3 , (121)
using a residence time of ≈20 Myr from analysis of 10 Be abundance in cosmic rays
[219], and the Milky Way volume VM W ∼ π(200 pc)(15 kpc)2 ∼ 4 × 1066 cm3 .
Thus the sources of the Galactic cosmic rays need to supply 1040 erg/s in the
form of nonthermal particle power throughout the Galaxy. One Galactic supernova
every thirty years supplying ≈1051 erg in outflowing kinetic energy will inject with
300 C. D. Dermer
10 % efficiency ≈1041 erg/s of cosmic-ray power, so significant inefficiency can be

accommodated if SNRs accelerate the cosmic rays. For the cosmic-ray power implied
by Fermi-LAT γ-ray observations of the Milky Way for leptonic and hadronic cosmic-
ray models, see [252].
The substantial time-averaged kinetic powers of SNRs make them the favored
candidate source population for the hadronic cosmic rays in our Galaxy. This power
is greater than the time-averaged power available from, for example, pulsars, galactic
X-ray binaries, stellar winds, or novae. In the electromagnetic window, the confirming
signature of enhanced cosmic-ray activity is the π 0 γ-ray bump [221, 225] peaking
at m π0 /2 ≈ 70 MeV in a number spectrum, and at several hundred MeV (depending
on the spectrum) in a ν Fν representation. The confirming signature of π 0 production
at 200 MeV competes with systematics and inaccuracies in the background model
which rise due to the large LAT PSF in the lowest decade of its energy range (Table 1).
Nearly 300 SNRs are known, mostly through radio detections. The Chandra
X-ray catalog contains nearly 100 SNRs,18 and ROSAT Galactic SNRs must number
several dozen. The poor EGRET PSF made it impossible to identify specific SNRs
with sites of enhanced >100 MeV emission, though several associations could be
made, including IC 443, W28, and W44 [214, 253, 257].
In the 1FGL, 41 SNRs are associated with LAT γ-ray sources, and 3 are, through
morphological features, identified with SNRs (Table 2), namely W51C, W44, and IC
443. By the time of 2FGL, 62 associations and 9 (morphological) identifications of
SNRs and PWN with Fermi LAT and other wavelengths have been made (Table 2).
The identifications represent 6 SNRs, adding also W28, W30, and the Cygnus Loop
[228], and three PWNe, namely J0835.3-4510 (Vela X), J1509.6-5850 (MSH 15-52;
possibly the remnant of a supernova in 185 CE), and J1826.1-1256 (HESS J1825-
137). A fourth PWN, MSH 15-52, was found in a search of off-pulse emission of
PSR B1509-58 [188].
Some important questions to be answered are
• Are cosmic-ray protons and electrons accelerated at SNR shocks, and with what
relative efficiencies?
• With what efficiency is shock kinetic energy converted to cosmic-ray energy?
• What is the spectrum of accelerated particles escaping from the shock acceleration
site?
• What is the maximum particle energy, and do SNRs accelerate CRs up to or beyond
the knee of the cosmic-ray spectrum?
• Does magnetic field amplification take place in the vicinity of SNRs?
Fermi has now detected historical/young (3000 years; see Table 8) SNRs,
namely Cas A (1680 CE), Tycho (1572 CE), and RX J1713.7-3946 (393 CE),
intermediate age (≈104 years) SNRs, e.g., IC 443 and W51C, and middle-aged
(2 × 104 years) SNRs, including W44, W28, the Cygnus Loop, and G349.7+0.2.
All SNRs detected with Fermi, other than Cas A, show evidence for molecular cloud
interactions.
18 http://hea-www.harvard.edu/ChandraSNR/
Table 8 Historical supernovae [223]

SN Typea Distance Sizeb γ-ray detectionc Pulsard
(kpc) GeV/VHE
1987A II 55 0.8
1680; Cas Ae II 3.4 5.9 × 5.5 Y/y
1604; Kepler Ia 5 4.5 × 3.7
1572; Tycho Ia 2.4 8.7 × 8.6 Y/y
1181; 3C 58 (?) II 2.6 6 × 3 J0205+6449
1054; Crab II 2 2.3 × 2.1 y/Y Crab
1006 Ia 1.6 32 × 26 ?/Y
393; RX J1713.7-3946 II (?) 1 ≈40 × 40 y/Y
386?
369?
185; MSH 15-52 (?) II y/Y 1509-58
a [231,232]
b Chandra X-ray sizes16 , except for SN 1987A (HST), RX J1713.7-3946 [200], and 3C 58 (radio;
[244])
c Flux measured at GeV or VHE energies; Y: larger ν F flux, y: smaller ν F flux
ν ν
d See Table 3
e Detection by Flamsteed disfavored [223]
Detecting the π 0 decay feature in Fermi-LAT SNR spectra depends on good back-
ground modeling to expose the low-energy (200 MeV) flux. Shock compression
of pre-existing cosmic rays could be argued to account for apparent cosmic-ray
production at SNR shocks, but this seems less plausible now that we have mea-
sured GeV γ-ray luminosities ranging from ≈1033 erg/s for the Cygnus loop, seen
as an ≈1◦ radial loop and the largest resolved γ-ray SNR, to ≈1034 –1035 erg/s
for young SNRs, and ≈1035 –few ×1036 erg/s for SNRs showing strong molecu-
lar cloud interactions [258]. This implies cosmic-ray kinetic energy on the order of
EC R (erg) ∼ 1036 L 36 t pp ∼
= 1051 L 36 /n(cm−3 ), in accord with the SNR hypothesis
for cosmic-ray origin. This estimate is not, however, so strong, given that distance
uncertainties can easily lead to factor of 4 or more uncertainty in L.
For Cas A, one of the youngest SNRs detected with Fermi [187], combined VHE
(MAGIC, VERITAS, and HEGRA) data show a ν Fν peak at a few GeV, with ambigu-
ous evidence for a low-energy cutoff, but clear evidence for a TeV steepening (see
Fig. 18). Cas A, with an extent of ≈6 , appears almost as a point source for the
LAT. The somewhat older SNR RX J1713.7-3946 is famous for being the first SNR
for which a VHE (HESS) map was made [199]. The TeV emission is extremely
well-correlated with X-ray maps, e.g., HESS and Suzaku [255]. The Fermi-LAT
map [190] displays an asymmetry towards the north, in the direction of a molecular
cloud where the X-ray and TeV emission is enhanced. The GeV spectrum itself is
remarkably hard, with spectral index = 1.5 ± 0.1. The ν Fν spectrum of this remnant
peaks between ≈200 GeV and 1 or 2 TeV, with a rapid cutoff at higher energies.
Nevertheless, the spectrum of RX J1713.7-3946 extends to ≈100 TeV, making it the
source of the most energetic γ rays yet detected. Its hard spectrum cannot be made
302 C. D. Dermer
Fig. 18 Combined Fermi-LAT and TeV spectra of selected SNRs [218]. The double arrow is the
Fermi energy range
by conventional hadronic models. One-zone leptonic models can fit the X-ray/γ-ray
spectra of RX J1713.7-3946, as can two-zone models.
The prototypical Type Ia Tycho SNR has also been recently reported as a GeV
source with a GeV–TeV γ-ray spectrum implying, if hadronic, a proton spec-
trum with a −2.3 number index [222]. Claims have been made that the combined
LAT/VERITAS spectrum requires hadrons [236], but a two-zone leptonic model
with both bremsstrahlung and Compton processes can fit the spectrum as well [202].
Because of the weakness of the GeV/TeV γ-ray flux, the ν Fν peak of Tycho can be
constrained to be few GeV, and the spectrum is not well-resolved in the crucial
energy range 300 MeV where the π 0 bump might be evident. For this purpose, a
better SNR might be IC 443 [185], a 3–4 kyr old core collapse SN (like Cas A and
RX J1713) surrounded by clouds of gas. With an 0.75◦ radio radius, it is resolvable
with Fermi. The SED of IC 443 is a flat ν Fν spectrum between ≈300 MeV and
5 GeV with a peak at 1 or 2 GeV. The spectrum extends to VERITAS/MAGIC VHE
energies like a power-law with ν Fν slope equal to ≈−0.6.
The middle-aged shell SNR W44 [186] has an age of ≈20 kyr. The γ-ray shape
closely following the 4.5 µ Spitzer IR image which traces shocked H2 . Its spectrum
peaks at 1 GeV or less, and it falls off sharply at higher energies, so is not detected
with the VHE telescopes. Its physical extent precludes the γ-ray emission from
originating from a pulsar, though a pulsar, B1853+01 with age ≈20 kyr, is found in
this SNR and could be associated with W44. If so, W44 is probably the result of a
core-collapse SNR, and therefore Type II.
As a final example of Fermi-LAT observations of SNRs, consider W51C, the
first resolved GeV SNR at GeV energies. It is an ≈10 kyr old remnant with extended
GeV emission compatible with the location and shape of the ROSAT emission [181].
The spectrum of W51C is similar to that of W44, peaking below ≈1 or 2 GeV, and
falling off steeply at higher energies. HESS weakly detects emission at about 1 TeV
from W51C, (whereas W44 is not detected at VHE energies), implying a ν Fν spectral
index ≈ − 0.5 between 10 GeV and 1 TeV.
In terms of cosmic-ray origin theory, a few remarks can be made. First, almost
all Fermi-LAT SNRs, other than Cas A, show evidence for interactions with mole-
cular clouds, in particular, those with OH (1720 MHz) maser emission from the
OH hydroxyl molecule tracing dense, shocked gas [217, 226]. Out of 24 known
maser SNRs reported in [226], 10 have GeV and/or TeV associations, and 6 have
both. Detection of illuminated molecular cloud complexes in front of the SNR shock
could reveal the existence of runaway cosmic rays that more likely would be protons
than electrons.
Second, nonthermal cosmic-ray protons lose energy on the secondary nuclear
production timescale t pp = (nσ pp c)−1 ≈ 35 Myr/n(cm−3 ), whereas nonthermal
electrons lose energy through bremsstrahlung on the free-free timescale t f f =
(nσ f f c)−1 ≈ 6 × 35 Myr/n(cm−3 ), where σ f f ≈ α f σT ∼ = σ pp /6. In either case, the
process is radiatively inefficient unless the target density n 1 cm−3 , but the non-
thermal electron bremsstrahlung is nearly as efficient as nuclear production processes.
The nonthermal bremsstrahlung or secondary nuclear production model face ener-
getics problems if they requires 1050 erg in cosmic rays.
Third, ionization/Coulomb losses can harden a nonthermal electron spectrum,
making γ rays from a leptonic bremsstrahlung model masquerade as a π 0 feature. A
full spectral model for multiwavelength production from SNRs requires, most gener-
ally, particle acceleration at the forward and reverse shocks, zones of different mag-
netic field strength at which particle acceleration can occur, leptonic bremsstrahlung
and Compton scattering, and secondary nuclear production.
Perhaps there is a trend in the SEDs of SNRs that can reveal the likely
γ-ray production mechanism (Fig. 18). For leptonic models, both (i) nonthermal
bremsstrahlung and (ii) Compton scattering, are potentially feasible. For hadronic
models, only (iii) secondary nuclear/particle process is effective in the SNR environ-
ment. Confrontation of these three nonthermal processes with multiwavelength SNR
data imply power and spectral and morphological constraints that can in principle
identify the dominant radiation process, though in practice, this has not been so sim-
ple. But establishing a trend from hard to soft γ-ray spectra with age, even scaled to
the Sedov age, is oversimplified, as is evident from the rather soft spectrum of Tycho.
For useful studies, a plot of the VHE (100 GeV–10 TeV) flux divided by the GeV
(300 MeV–30 GeV) flux versus SNR age (to avoid uncertainties in distance measure-
ments) for Type Ia and Type II SNe might provide some insight on the evolution of
particle acceleration with remnant age. At present, it still seems premature to claim
that the problem of the origin of the Galactic cosmic rays is solved.
304 C. D. Dermer
Fig. 19 Shock geometry in Upstream (-) Downstream (+)

rest frame of shock. In this u
- n
+
figure, δ is the shock width
Flow
n u
-
δ
+
5.5 Nonrelativistic Shock Acceleration of Electrons
First-order Fermi acceleration is highly developed as a mechanism to accelerate

cosmic rays [206, 213]. Here we illustrate this process by calculating the nonthermal
spectrum of test particles accelerated at a shock-discontinuity in density and velocity.
We use a continuity equation approach,19 which complements the approach to shock
acceleration using the convection-diffusion equation and probability arguments [4].
The geometry we consider is particle acceleration at a discontinuity in velocity
and density, as illustrated in Fig. 19. The upstream unshocked material has speed
u in a frame where the supersonic flow intercepts gas at rest. For a strong nonrel-
ativistic shock with compression ratio χ = 4, the upstream (−) flow approaches
with speed u − = β− c = 4βc/3 and the downstream (+) flow recedes with speed
u + = β+ c = βc/3 in the comoving primed frame stationary with respect to the
shock. Consequently u = u − − u + = βc. Here we treat the acceleration of relativis-
tic nonthermal particles with Lorentz factor γ 1 and speed β par c ≈ c.
In the general case, the compression ratio
u−
χ= . (122)
u+
For a lossless adiabatic shock wave,
γ̂ + 1 → 4
χ= γ̂ = 5/3 (123)
γ̂ − 1 + 2/M21 1 + 3/M21
(Eq. (13.8), [4]), where γ̂ is the adiabatic index, and γ̂ = 5/3 for a nonrelativistic
monatomic gas. When M1 1 and γ̂ = 5/3, χ → 4. It follows from Eq. (122) and
the condition u = u − − u + that
19The treatment of [191] assumes energy-independent escape timescale representing some second-
order Fermi acceleration scenarios.
χu u
u− = , u+ = , (124)
χ−1 χ−1
so that in the limit M 1, u − → 4u/3 and u + → u/3 in the frame comoving with
the shock.
The change of energy of a relativistic particle with dimensionless momentum
p = βγ = γ 2 − 1 → 1, when γ 1, in a complete cycle of Fermi acceleration
is, from Eq. (12.10) [4], given by

Δγ ∼ 4
= β, (125)
γ F1 3
where u = βc = u − − u + . For second-order Fermi acceleration,

Δγ ∼ 4
= β2, (126)
γ F2 3
and both expressions receive an added boost ∝ Γ 2 , at least in the first cycle of
acceleration, for shock acceleration by a relativistic flow with speed 1 − 1/Γ 2 c.
The rate of energy gain, or acceleration rate, in nonrelativistic first-order Fermi
acceleration is given for particles
Δγ
γ̇FI ∼
= , (127)
tcyc
where the cycle time tcyc is given by

4κ− κ+ 4
tcyc = + = (κ− + χκ+ )
vu− u+ vu −

mc2 4p χ − 1 B− ∼ mc2 4γ χ − 1
= η− + χη+ = ∝ γ. (128)
Q B− 3u χ B+ Q B− 3u χ
[229], where v ∼
= c is the particle speed, χ = u − /u + = ρ+ /ρ− is the compression
ratio, and ρ = mn is the mass density. The spatial diffusion coefficient
1 1 1
κ± = λ± v = η±rL± v = η±rL±
o
pv,
3 3 3
here writing the diffusion coefficients in terms of the parameters η± that give the
particle mean-free-paths scaled to the values in the local magnetic field. In the Bohm
diffusion approximation, the diffusion mean-free path is set equal to the Larmor
radius, so η± = 1. Thus the Bohm approximation is
1
κB± = rL± v,
3
306 C. D. Dermer
where the Larmor radius

mc2
rL± = rL±
o
p= p. (129)
Q B±
The acceleration rate of nonthermal relativistic particles by first-order shock accel-

eration can therefore be expressed as
4β 4 u
γ̇acc = γ= γ. (130)
3tcyc 3 ctcyc
The acceleration rate is independent of p or γ because tcyc ∝ γ from Eq. (128).

In nonrelativistic shock acceleration, the escape probability per cycle is
u+ 4u
Pesc = 4β+ = 4 = . (131)
c c(χ − 1)
[203, 204]. The escape time is therefore
tcyc tcyc
tesc = = . (132)
Pesc 4β+
The steady-state particle continuity equation with gains, losses, and escape takes
the form (Eq. (C10), [4])
∂ n(γ)
[γ̇ n(γ)] + = ṅ(γ). (133)
∂γ tesc (γ)
Equation (133) has solution, for an energy-gain process γ̇ > 0, given by (Eq. (C11),
[4]) γ γ
dγ
n(γ) = γ̇ −1 dγ ṅ(γ ) exp −
. (134)
1 γ tesc (γ )γ̇(γ )
Equations (130) and (132) imply
β χ−1
tesc (γ)γ̇acc (γ) = γ= γ, (135)
3β+ 3
so −3/(χ−1)
3ṅ 0 γ
n(γ) = tcyc . (136)
4βγ γ0
noting that this represents the differential number density spectrum of particle accel-
erated at the shock discontinuity.
Taking a δ-function source injection ṅ(γ) = ṅ 0 δ(γ − γ0 ), the injection spectrum
downstream where the particles are no longer subject to acceleration at the shock is
−At p
n(γ) 4β+ 3ṅ 0 γ
ṅ ds (γ) = = n(γ) = , (137)
tesc tcyc (χ − 1)γ0 γ0
where
u − + 2u + 2+χ
At p = = (138)
u− − u+ χ−1
is the test-particle number index well known in studies of nonrelativistic shock accel-
eration [206].
Now consider shock acceleration with the addition of Thomson/synchrotron radia-
tive losses, so
4 β
γ̇ = γ̇acc + γ̇rad = γ − νγ 2 . (139)
3 tcyc
Following the approach above gives

γ
ṅ 0 3 dx
n(γ) = exp − , (140)
4 βγ
− νγ 2 χ−1 γ0 x(1 − kx 2 )
3 tcyc
where
3νT m e c2 4 χ−1 B−
k= , and T = η− + χη+ . (141)
4β Q B− 3u χ B+
Therefore
−At p
n(γ) 3ṅ 0 γ
ṅ ds (γ) = = . (142)
tesc (χ − 1)(1 − kγ 2 )3/2 1 − kγ 2
This is a pileup spectrum for χ > 5/2, that is, when At p is harder than 3. The
maximum electron energy is given by 1 − kγ 2 = 0, or

1 4β
γmax =√ = . (143)
k 3νT
Even though the spectrum piles up, the number of particles and the energy of the
accelerated particles is convergent.
The synchrotron emission made by this electron distribution has a maximum value
at dimensionless photon energy

3 B 2 9πe χ B− B+ /B 2
syn,max = γ = β
2
(144)
2 Bcr Bcr σT χ−1 1+
u ph
(η− + χη+ BB−+ )
uB
(cf. [198]). Thus

308 C. D. Dermer

27 χ B− /B+
syn,max = β2 . (145)
8α f χ−1 1+
u ph
(η− + χη+ BB−+ )
uB
The apparent divergence due to the term χ − 1 in the denominator of Eq. (145) can be
seen not to arise, considering that χu 2 /(χ − 1) = u − u. The leading term, syn,max ∼
=
27/8α f ∼ = 462, or E syn,max ∼
= 236 MeV, represents a bound for nonrelativistic shock
acceleration, remarkably close in value to the cutoff energy of the Crab pulsar wind
nebula [48]. This indicates that the pulsar wind nebula of the Crab is formed by a
wind termination shock moving out at mildly relativistic velocities. Small changes
in the bulk speed from a knot in the Crab pulsar wind, as imaged with Chandra,
could produce the Fermi and AGILE flares from the Crab [52]. Better imaging at
≈100 MeV would localize the emission source, but is hardly possible with Fermi.
Addition of a diffusion term will produce smoothed, realistic pile-up electron
injection distributions formed in first-order shock acceleration. The addition of a
diffusion term is under study in work with P. Becker. Note that these pileup functions
differ from the Schlickeiser pile-ups [245, 246] where escape is always independent
of particle energy. At relativistic energies and with relativistic flows, the maximum
dimensionless synchrotron energy is syn,max ∼ = (27/8α f )Γ /(1 + z), which is rele-
vant when interpreting the maximum photon energies in GRBs as a consequence of
synchrotron emission formed by particle acceleration at an external shock.
6 γ Rays from Star-Forming Galaxies and Clusters of Galaxies,

and the Diffuse Extragalactic γ-Ray Background
6.1 γ Rays from Star-Forming Galaxies
Galaxies with ongoing star formation, most notably the Milky Way in which we live,
are illuminated at γ-ray energies by secondary products of cosmic-ray interactions
with gas and dust. The characteristic 100 MeV γ-ray luminosities of normal star-
forming galaxies are ∼1038 –1040 erg s−1 , some factors of ∼103 –1010 × smaller than
those of active galaxies. Yet the star-forming galaxies vastly outnumber the AGNs.
For example, the space density of a typical L∗ spiral galaxy like the Milky Way is
≈3 × 106 –107 Gpc−3 , by comparison with the space density of FR II radio galaxies,
which is ≈2000 Gpc−3 [278]. (The volume of 1 Gpc3 extends to about z = 0.15
from the present epoch.)
Besides the Milky Way, the Large Magellanic Cloud, detected earlier with EGRET
with flux F−8 ∼ = 19, is now measured with Fermi LAT at the level of F−8 ∼ = 26 ± 2
[260]. At a distance of ≈50 kpc, the LMC has about 10 % of the mass and a supernova
rate ≈20 % of the Milky Way. The LAT resolves the LMC, and finds that 30 Doradus,
its major star-forming region, is a bright source of γ rays that does not consist
of significant point source contributions. The γ-ray spectrum is consistent with an
origin in cosmic-ray production. Its γ-ray emission correlates well with massive
star forming regions and an ionized H+ template, but more poorly with neutral or
molecular gas distribution. The γ-ray emission morphology is surprisingly compact,
and indicates that cosmic rays are accelerated in star-forming regions that are not
very diffusive, thus accounting for the bright compact emission centered around
30 Doradus.
Our other notable dwarf companion galaxy, the Small Magellanic Cloud, is also
for the first time detected in γ rays. The LAT measures a flux of F−8 = 3.7 ± 0.7
[261] from an extended, ∼3◦ region. Unlike the LMC, the γ-ray emission from the
SMC is not clearly correlated with the distribution of massive stars or supernova
remnants, though the emission may trace supergiant shells.
Though not unexpected [284], the discovery of starburst galaxies at GeV [262] and
TeV energies is important to test predictions of cosmic-ray origin. The two nearest
starburst galaxies, M82 and NGC 253, each at a distance of ≈4 Mpc, were detected
at a level of F−8 ∼= 1.6 and F−8 ∼ = 0.6, respectively [262], with the LAT, soon after
being reported as VERITAS [266] and HESS [267] sources, respectively. Properties
of the brightest star-forming galaxies at γ-ray energies are shown in Table 9. Although
correlations with mass, SN rate, and their product can be extracted, the situation, as
exhibited by the distribution of γ-ray emission in the LMC, is far more complex.
As can be seen from Fig. 7, starburst galaxies define a separate track in the spectral
index versus luminosity plane.
Now that the LAT sensitivity for star-forming galaxies has reached beyond our
neighboring galaxies, we can expect the number to grow. The Andromeda galaxy,
M31, has been detected at a level of F−8 ∼ = 0.9 [265]. Other starburst and infrared
luminous galaxy detections from sources like Arp 220 [287] or Mrk 273 are keenly
anticipated as the LAT photon statistics accumulate.
Table 9 Properties of star-forming galaxies

Galaxy d RSN MGas F−8 a 4πd 2 Fγ Lbγ Index
(kpc) (century−1 ) (109 M ) (1041 ph/s) (1039 erg/s)
MW – 2.0 ± 1.0 6.5 ± 2.0 – 11.8 ± 3.4c 1.2 ± 0.5 2.2 ± 0.15
LMC 52 ± 2 0.5 ± 0.2 0.67 ± 0.08 26.3 ± 2.0 0.78 ± 0.08 0.041 ± 0.007 2.26 ± 0.11
SMC 61 ± 3 ≈0.12d ≈0.45 3.7 ± 0.7 0.16 ± 0.04 0.008 ± 0.003 2.23 ± 0.12
M31 780 ± 30 1.1 ± 0.2 7.7 ± 2.3 0.9 ± 0.2 6.6 ± 1.4 0.43 ± 0.09 2.1 ± 0.22
M82 3600 ± 300 20 ± 10 2.5 ± 0.7 1.6 ± 0.5 250 ± 90 13 ± 5 2.2 ± 0.2
N253 3900 ± 400 20 ± 10 2.5 ± 0.6 0.6 ± 0.4 110 ± 70 7.2 ± 4.7 1.95 ± 0.4
References [252, 261, 262]
a F = 10−8 F ph (>100 MeV)/cm 2 -s
γ −8
b Fluxes and luminosities in 100 MeV–5 GeV range
c Value is strongly dependent on assumed size of Galactic halo used to model MW; [252] find
0.6 F−8 1.0

d Reference [286]
310 C. D. Dermer
6.2 γ Rays from Clusters of Galaxies
Clusters and superclusters of galaxies are the largest manifestations of the ongoing
structure formation. Regions collapse by gravitation, driven primarily by dark matter
fluctuations. The magnitude of the density fluctuation determines the formation time,
insofar as larger structures form by accreting smaller clumps in a hierarchical merging
scenario. The result, as seen in numerical simulation codes, is lumpy, continuous
accretion forming filaments and webs of enhanced density and magnetic field.
Clusters of galaxies are the most energetic events in the universe since the big bang,
releasing as much as G M1 M2 /R ∼ 1063 –1064 erg, here taking M1 ∼ 1015 M ∼
10M2 and R ∼ 1 Mpc. Yet they are hardly the most luminous, since the energy is
released over a Hubble timescale H0−1 ∼ 4.3 × 1017 s ∼ 13.6 Gyr, implying a power
of ∼2 × 1045 –2 × 1046 erg/s. This energy, carried primarily in the gravitational
potential of the dark matter, goes into heating, turbulence, and gas motions. In the
events of cluster mergers, turbulence is generated in wakes, and shocks are formed
in the collision. Nonthermal particles are accelerated by shocks and turbulence.
Evidence for nonthermal particle acceleration in merging clusters of galaxies is
already known from the existence of radio halos and relics. The relativistic elec-
trons and positrons making this radio emission are either accelerated directly or
produced as secondary pairs by cosmic-ray protons and ions colliding with particles
in the intracluster medium. The unpolarized central radio halos, ∼ Mpc size, with
a morphology similar to the X-ray bremsstrahlung, reflect particle acceleration at
the merger shocks between two merging clusters, intermittent AGN activity, and
dynamical friction of galaxies in the hot, ∼keV, thermal plasma. By contrast, radio
relics, also about ∼1 Mpc in size, which lie on the cluster outskirts, display elongated
morphologies, and are up to ∼50 % polarized, may reflect a different origin from the
radio halos. Most likely, they represent ongoing accretion of pristine gas from the
big bang.
The halo radio emission of, e.g., the Coma cluster of galaxies, is very soft above
≈1 GHz, and can be modeled in a merging cluster framework if the system is
observed soon after the cluster and subcluster have merged [270]. By scaling non-
thermal cosmic-ray proton and ion energy to the electron energy required to make
the observed radiation, predictions for the γ-ray emission can be made. The iden-
tification of a nonthermal hard X-ray feature as Compton-scattered CMBR would
mean a larger nonthermal electron total energy, giving more optimistic predictions
for γ-ray emission from clusters of galaxies. In any case, clusters of galaxies act
as storage volumes for the cosmic-ray protons [271], which have radiative lifetimes
exceeding the Hubble time. Because of the long, ∼10 Gyr, crossing time of elec-
trons, versus a radiative lifetime of ∼0.1 Myr, in-situ acceleration of pre-existing
relativistic electrons by turbulence [272, 273] is now favored to explain radio halos
and relics.
The Fermi LAT has many high-priority cluster candidates, including not only
Coma and Abell 2256, but Hydra, Centaurus, and Fornax. The Perseus cluster is
dominated by the AGN 3C 84 in the central elliptical galaxy NGC 1275, and is
detected with the Fermi LAT [264]. But this γ-ray emission is from a variable blazar
core (though there may be steady emission at the level of F−8 4). Radiation
mechanisms involving the energetic leptons together with the decay of neutral pions
produced by hadronic interactions have the potential to produce abundant GeV pho-
tons. Using data from 2008 August to 2010 February, upper limits of 33 galaxy
clusters, selected according to their proximity, mass, X-ray flux, temperature, and
non-thermal activity were reported by the Fermi collaboration [268]. The flux upper
limits, in the 0.2–100 GeV range, are typically at the level of F−8 ∼ = 0.1–5. These
results limit the relativistic-hadron-to-thermal energy density ratio to be 5–10 % in
several clusters [274, 281].
As disappointing as it has been that γ rays have not yet been detected from galaxy
clusters, we are still learning about the injection conditions in the low Mach number
merger shocks, the high Mach number accretion shocks, and the injection conditions
of turbulence models. A purely secondary production model where all the radio
emission results from secondary electrons formed in secondary nuclear production
is likely ruled out [268, 275], and dark matter annihilation cross sections are further
constrained.
6.3 Extragalactic γ-Ray Background and Populations
Figure 20 shows the spectrum of the diffuse extragalactic γ-ray background (EGB)
obtained in an analysis of the first 10 months of Fermi-LAT science data in the
range 200 MeV–100 GeV based on the GALPROP model for the Galactic emission
[7], alongside the EGRET EGB [26] and the EGRET EGB based on a GALPROP
analysis of the Galactic diffuse EGRET emission [28]. The Fermi-LAT intensity
extrapolated to 100 MeV based on the power-law fit I (>100 MeV) = (1.03±0.17)×
10−5 /cm2 -s-sr is significantly lower than that obtained from EGRET data, namely
IEGRET (>100 MeV) = (1.45±0.05)×10−5 /cm2 -s-sr [26]. The GALPROP analysis
of the EGRET data [28] agrees, however, with the Fermi flux extrapolated to 100 MeV.
Furthermore, the Fermi-LAT spectrum is compatible with a featureless power law
with index −2.41 ± 0.05 [7]. This is significantly softer than the EGRET spectrum,
with index −2.13 ± 0.03 [26]. To check that the different spectra are not due to the
thr = 10, comparable to the
instrumental point-source sensitivities, a threshold flux F−8
average EGRET sensitivity, and an isotropic γ-ray intensity like that quoted above is
measured. Therefore, the discrepancy cannot be attributed to a lower threshold for
resolving point sources. Most likely the relatively hard spectral slope for the EGRET
EGB was due to underestimation of the EGRET effective area above ≈ few GeV
[27], or due to lack of a model for Compton γ rays, which can make a significant
contribution to the EGB at high latitudes.
The origin of the EGB is an interesting open question. Because of the preponder-
ance of blazars in the Third EGRET catalog [17], a large fraction of this emission
was expected to be contributed by blazars, though some pre-Fermi studies [12,
280] found that blazars made only a small fraction of this emission. The detection
312 C. D. Dermer
Fig. 20 Fermi measurement of the diffuse extragalactic gamma-ray background (EGB) [7] (purple
filled circles). Also plotted is the EGRET result [26] (filled black circles) and background intensity
from GALPROP analysis of EGRET data [28] (open red circles)
of ∼1000 high Galactic latitude sources [70], most of which are blazars [92], has
allowed a precise characterization of the contribution of blazars to the EGB. Such
studies have shown that unresolved blazars contribute 30 % of the EGB emission
[263, 269].
If not blazars, then other source classes must be invoked to account for the EGB.
Possibilities include emission from star-forming galaxies, misaligned blazars (i.e.,
radio galaxies), γ rays made by particles accelerated by structure-formation shocks
in clusters of galaxies, [277, 282, 283], and dark matter; see references in [263].
Star-forming galaxies such as our own Milky Way are known to be γ-ray emitters
due to the interaction of cosmic rays with interstellar gas and radiation fields [262].
The superposition of the numerous but individually γ-ray weak star-forming galaxies
can contribute a greater fraction of the EGB than the rare, individually bright γ-ray
blazars [276, 285]. Fermi detection of several star-forming galaxies, including NGC
253, M82, and the LMC [262], supports this possibility, though the exact percentage
remains highly uncertain.
Included in the EGB are pulsar contributions, including millisecond pulsars.
Because millisecond pulsars (MSPs) are also found in early-type galaxies, they might
contribute a significant fraction of the EGB. A way to determine their integrated
diffuse emission would be through the integration of their flux (or luminosity) distri-
bution in γ-rays. The typical spectrum of MSPs is hard (e.g., photon index of ∼1.5)
and shows an exponential cut-off around a few GeV [50]. Thus the diffuse emis-
sion arising from MSPs should show a bump around a few GeV, similar to a feature
found in the Galactic diffuse emission that has been ascribed to annihilating dark
matter [279]. The left-hand panel of Fig. 21 illustrates a possible decomposition of the
EGB into different source classes, including a contribution from WIMP dark-matter
annihilation.
Fermi EGB
FSRQs
BL Lacs
Star forming galaxies
-3 Milliseconf Pulsars
10 559 GeV WIMP, b b
Total
E2 dN/dE [cm-2 s-1 sr -1 MeV]
10-7 LogLγ =45.6 -- 46.9
LogLγ =46.9 -- 47.5
10-4 10-8
γ]
Φ(L ,z) [Mpc-3 L-1
LogLγ =47.5 -- 47.9
-9 LogLγ =47.9 -- 49.4

10
-5
10 10-10
γ
10-11
-6
10
10-12
3 5
102 10 104 10
0 0.5 1 1.5 2 2.5 3 3.5
Energy [MeV] z
Fig. 21 Left panel Fermi-LAT measurement of the extragalactic diffuse background spectrum (data
points; see [7]), and model contributions from different source classes. The curves represent the inte-
grated contributions of: (1) FSRQs (thin solid line), (2) BL Lac objects (thin short dashed line), (3)
star-forming galaxies, (4) MSPs (dot-long dashed), (5) WIMP bb̄ annihilation (long dashed) and (6)
sum of all the previous (hatched). Right panel Luminosity function of γ-ray selected FSRQs [269]
A preliminary γ-ray luminosity function for FSRQs detected with the Fermi-LAT
is shown in the right-hand panel of Fig. 21 [269]. The luminosity function shows
the change in the space density of FSRQs as a function of redshift and for different
luminosity classes. What is apparent from Fig. 21 is that the space density of powerful
FSRQs peaked at redshifts ≈2, and then declined to the space density now observed.
The peak of maximum growth occurs at different epochs for different luminosity
classes, with the more luminous sources reaching their maximum space density
earlier in the history of the universe, while the bulk of the population (the lower
luminosity blazars) are more abundant at present times. This cosmic downsizing
behavior cannot persist to very large (z 1) redshifts and early times, as may be
reflected by the sudden turnoff of high-z γ-ray blazars. The largest redshift γ-ray
blazar in both the 1LAC and 2LAC is at z = 3.1, despite many candidate higher-
redshift blazars.
7 Microquasars, Radio Galaxies, and the EBL
7.1 γ-Ray Binaries
The literature on the subject of Galactic binaries is vast; a good X-ray review remains
Ref. [322]. Here we highlight only a few of those aspects that make for an appreciation
of GeV–TeV γ-ray binaries in the Galaxy, of which 6 are well-established, namely
1. LS 5039 (P = 3.9 days);
2. LSI +61◦ 303 (P = 26 days);
314 C. D. Dermer
3. PSR B1259-63 (P = 3.5 years);

4. Cygnus X-3 (P = 4.8 h);
5. 1FGL J1018.6-5856 (P = 16.6 days) [300]; and
6. Cygnus X-1 (P = 5.6 days).
These sources are exclusively identified with high-mass X-ray binary (HMXB) sys-
tems and are identified, not merely associated, with the named system by period
folding or time differencing with the orbital period. With its long orbital period, this
is harder for PSR B1259-53, but γ-ray flaring near periastron strengthens the identifi-
cation. PSR B1259-63 is an example of a Be X-ray binary system, where the compact
object, a young pulsar with a spin period of 48 s, is in a highly eccentric orbit around
a luminous B star that displays emission lines in its optical spectrum (thus the des-
ignation “Be”). Flares take place near periastron at all frequencies, when the pulsar
passes through the equatorial stellar wind [318], and now also at GeV energies [294].
High mass and low mass refer to the companion star, not the compact object.
HMXBs are X-ray binaries composed of a compact object—white dwarf, neutron
star, or black hole—and a young massive O or B main-sequence or Wolf-Rayet star.
The low-mass X-ray binaries (LMXBs) typically have older, less massive compan-
ions 2M . If the companion compact object is a pulsar, it can vary in age from
young and energetic to old and slowed down (and/or spun up). The fuel illuminating
galactic X-ray binaries is accretion onto the compact object, which is transferred,
primarily, through stellar wind in HMXBs and Roche-lobe overflow in LMXBs.
X-ray variability is modulated by orbital effects, the most important being the change
in the binary orbital separation distance. In only the case of PSR B1259-63 are
X-ray pulsations at the pulsar period seen, which are also found for intermediate
mass X-ray binaries like Her X-1 and other HMXBs at X-ray energies.
Microquasars are X-ray binaries with radio jets. HMXBs number around 7, includ-
ing SS 433 and V4641 Sgr to the list above. There are 9 LMXB microquasars in the
Paredes list from 2005 [331]. Some of the most famous are GX 339-4, Circinus X-1
and Scorpius X-1.
Like the radio structures seen in extragalactic AGNs (e.g., 3C 84), jetted structures
have been found in unusual X-ray binary sources, e.g., the Hertz-Grindlay source
1E 1740.7-2942 lying ∼0.7◦ away from the Galactic center [326]. The analogy with
extragalactic radio sources was strengthened when this source [327] and other LMXB
sources such as GRS 1915+107 [328] were shown to exhibit superluminal effects.
No LMXBs are yet detected at GeV or TeV energies.
Prior to Fermi, three Galactic binaries were detected at TeV energies, namely LSI
+61◦ 303 with MAGIC and VERITAS, and LS 5039 and PSR B1259-63 with HESS
[295]. An isolated flare from Cyg X-1 was observed with MAGIC. No solid EGRET
identification with a Galactic binary source had been made, though both LS 5039
[332] and LSI +61◦ 303, even in the COS-B days [18], were suspected to be γ-ray
sources.
A central debate regarding γ-ray binaries is whether they are scaled-down quasars,
as suggested by the microquasar label, or colliding stellar wind/pulsar wind systems.
Leptonic microquasar models [297, 302, 330] invoke electrons or hadrons in jets
accelerated up to TeV energies with mildly relativistic outflows, accounting for the
extended radio emission. Confirming evidence would be VHE emission from X-ray
binaries with definite black holes, e.g., Cyg X-1, V 4641, or GRS 1915+105.
The other class of pulsar/star model invokes the rotational energy of the neutron
star and the interactions of the pulsar and stellar winds [305, 323]. This model surely
applies to PSR B1259-63. Insofar as LS 5039 and LSI +61◦ 303 have compact
objects with M < 4M , the question of whether the compact object is a black hole
or neutron star is unresolved. The confirming evidence for a neutron star is, of course,
detection of pulsations.
When considering the different sources, it is useful to have a theoretical picture in
mind. Most developed and perhaps most likely is a leptonic model, where electrons,
which are accelerated in the extreme environment near the compact object, Compton
scatter the photons of the hot star. Those γ rays are in turn attenuated by the stellar
radiation. To first order, a source at superior conjunction should be GeV luminous
(more head-on collisions for Compton scattering) and less TeV luminous (due to
γγ attenuation). It should anti-correlate with the TeV emission, which has least
γγ opacity and therefore should be brightest at inferior conjunction.
LS 5039
Though sitting on the edge of the Galactic plane, LS 5039 is well-detected and
identified by phase-folding [290]. Its Fermi-LAT light curve is remarkably anti-
correlated with the HESS TeV emission. At superior conjunction, when the TeV flux
is in a low power-law state, the Fermi GeV spectrum is in a high state. At inferior
conjunction, the roles are reversed. Only quasi-simultaneity is possible, because
Fermi will necessarily integrate over longer times to achieve the same sensitivity as
a TeV telescope. The joint spectra are also remarkable, with the GeV high state fit
by an exponentially cut-off power law, and taking place at the same phase as the
low-state power-law HESS spectrum. In reverse, the low-state GeV power law takes
place when the TeV flux is elevated.
LSI +61◦ 303
The Fermi LAT light curve of this source is a rather sharp-peaked sinusoid, with
peaks of emission around periastron [291]. The GeV radiation is modulated with a
maximum close to periastron, and minimum close to apastron. TeV emission peaks
at different orbital phase, revealing a flux anti-correlation with the MAGIC TeV
emission that might be expected in a leptonic model. The phase-averaged Fermi
spectrum is well described by an exponentially cutoff power-law with index ≈2.2
and cutoff energy ≈6.3 GeV. This spectrum is not found to vary strongly with phase.
The non-simultaneous MAGIC and VERITAS data corresponding to different phases
look, if they correspond to the time of Fermi observations, like a distinct spectral
component, with the hard component extending to TeV energies with a spectral index
≈1.9 overtaking the exponentially decaying GeV spectrum.
316 C. D. Dermer
Cygnus X-3
Cyg X-3 is a HMXB microquasar with an orbital period of 4.8 h—short for an
HMXB—at a distance of ≈7 kpc from the Earth. It is found in a complicated region,
the Cygnus arm, but its period readily identifies it in γ rays. Its companion is the
Wolf-Rayet star V1521 Cyg. It exhibits bright radio flares and goes through active
states. It is not known whether it contains a black hole or neutron star binary. Four
episodes of flares were observed with AGILE [341] a few days before radio/X-ray
flares. The Fermi LAT light curve exhibits periods of scattered activity, and periods
of quiescence that appear to be anti-correlated with the X-ray emission [289]. The
radio and γ-ray activity are also correlated.
1FGL J1018.6-5856
Period-folding analysis has resulted in the recent discovery of a new high-mass

X-ray/γ-ray binary, 1FGL J1018.6-5856 [300, 317]. The 100 MeV–200 GeV Fermi-
LAT data is modulated with the 16.58 ± 0.04 day orbital period. The Swift XRT
finds an X-ray source coincident with the position of the Fermi source. The digitized
sky survey (DSS2) image of this region shows a star with spectrum very similar to
that of the gamma-ray binary LS 5039, and the Australia Telescope Compact Array
detects a radio source at 9 GHz. Chandra and XMM-Newton observations [333] find
a coincident X-ray source from which the column density and reddening can be
determined. With 2MASS optical observations, this implies a star of type of O6V at
a distance of 6–12 kpc, large by comparison with the distances of LS 5039 and LSI.
PSR B1259-63
PSR B1259-63 may help us understand these systems best by virtue of the compact
object being unambiguously a neutron star. The pulsar has a 47.76 ms rotation period
and a massive, ≈10M Be star companion, SS 2883. The stars orbit with a 3.5 year
orbital period and eccentricity = 0.87, guaranteeing episodes of strong interactions
near periastron passage. Because of its inclination and the geometry of the equatorial
stellar wind, the pre- and post-periastron passages are not symmetrical. EGRET on
CGRO failed to detect it during 7 weeks of observation around the 1994 periastron
passage [340], but flaring at radio, optical, X-ray, and HESS TeV [295] has been
monitored in previous passages.
So it was with great anticipation that observers at all wavelengths waited for the
2010 December passage. PSR B1259-63 did not fail to perform: the light curves
measured by an array of instruments at different wavelengths and epochs are shown
in Fig. 22, and reveal the complexity of the encounter [294]. It was only weakly
detected with the Fermi LAT on the approach to periastron and up to 20 days after
(MJD 55517-55563), during which the pre- and near-periastron spectrum in Fig. 22
was measured. The system flared at GeV energies starting at ≈30 day post-periastron,
Fig. 22 Left Radio, X-ray, GeV, and TeV light curves of PSR B1259-63 at different epochs around
the 2011 perastron passage on 2010 December 15 (MJD 55545), as labeled [294]. The 120 days
shown represents ≈13 % of the orbital period. Right Fermi spectra of the system in the pre- and
near-periastron period (blue circles) and in the post-periastron flare (red squares)
with the Fermi LAT measuring during period MJD 55575-55624 shown as the post-
periastron spectrum in Fig. 22. The reason that the GeV flare seems to lag behind the
peak of the radio and X-ray emissions at ≈20 day post-periastron is not understood.
Though the time of 2010 periastron passage was not ideal for TeV observations
due to moonlight conditions, additional post-periastron TeV data for the most recent
passage will help illuminate this system. We can reliably predict interesting events
in the future.
7.2 Misaligned Blazars and Radio Galaxies
The misaligned AGN (MAGN) population detected with Fermi using the first
15 months of science survey data consists of 11 sources (Fig. 7; [89]), including
7 FRI and 2 FRII radio galaxies, which are thought to be the parent population of BL
Lac objects and FSRQs, respectively, in addition to 2 steep spectrum radio quasars
that are believed to be slightly misaligned FSRQs. The MAGNs are associated with
objects in the Third Cambridge (Revised) and Molonglo radio catalogs. The MAGN
sources in the 3CRR catalog have large core dominance parameters compared to the
general 3CRR source population, implying that the beamed component makes an
318 C. D. Dermer
appreciable contribution to the γ-ray flux. This is furthermore supported by the fact
that three of these sources—3C 78, 3C 111, and 3C 120—do not appear in 2LAC,
evidently due to variability.
Why the ratio of measured γ-ray luminosities of FRI galaxies and BL Lac objects
span a much larger range than the comparable ratio for FRII radio galaxies and
FSRQs is an open question, if in fact not due to statistics. The simplest possibility
is that there are a lot more nearby FRIs, and limitations on Fermi sensitivity will
therefore favor detection of these nearby sources. At z 1, the luminosity distance
d L ≈ 4200z(1 + z) Mpc. As noted earlier, the Fermi-LAT reaches a limiting energy-
flux sensitivity of ≈5 × 10−12 erg cm−2 s−1 for two years of observations that is,
unlike integral photon flux, only weakly dependent on source spectral index [92].
Fermi can thus only detect sources with luminosity L γ 1046 z 2 (1 + 2z) erg s−1 .
The number of nearby FRI and FRII galaxies depends on the space density of
these objects which, in the study by Gendre, Best, and Wall [278], is inferred from
1.4 GHz NVSS-FIRST radio observations (recall that the Fanaroff-Riley dichotomy
is based on 178 MHz luminosities P178 MHz ). From their Fig. 12, FRI and FRII radio
galaxies have a local (z < 0.3) space density of ≈35000 Gpc−3 and ≈2000 Gpc−3 ,
respectively, at P1.4 GHz > 1022 W Hz−1 sr−1 or L r 1040 erg/s, implying that the
local space density of FRIs exceeds that of FRIIs by ≈20.
The volume of the universe within z < 0.15 (d L < 700 Mpc) is ≈1 Gpc3 , so
there are thousands of FRIs and FRIIs within z ∼ = 0.1, yet other than 3C 111,
all LAT-detected radio galaxies within this volume are FRI sources (see Fig. 29;
[303]). Sensitivity limitations and the abundance of nearby FRIs could explain the
MAGN population statistics, but low apparent-luminosity off-axis FRIIs, which are
far more radio luminous than BL Lac objects and FRI galaxies, are lacking. With
only a few thousand randomly aligned sources within z = 0.1, a narrower γ-ray
beaming cone in FSRQs, with a more rapid fall-off in off-axis flux, makes detection
of these nearby sources far less likely than a broader γ-ray emission cone in BL
Lac objects, as expected for the different beaming factors from SSC emission and
external Compton processes [301, 311]. This could also reflect differences in jet
structure between FSRQs and BL Lac objects [299, 315], or extended jet or lobe
emission in FRIs [296] that is missing in FRII galaxies.
Both core and lobes can significantly contribute to the measured γ-ray luminosi-
ties. In the case of Centaurus A, the values of L γ of the core and lobes are comparable,
with the lobe emission primarily attributed to Compton-scattered CMBR [46]. The
significant or dominant lobe component means that the core luminosity of misaligned
AGNs can be less than the measured L γ unless the lobe emission is resolved.
One-zone synchrotron/SSC models with δD 10 give good fits to the long-term
average spectra of other HSP BL Lac objects such as Mrk 421 and Mrk 501. The
SEDs of radio galaxies, in contrast, are fit with much lower Doppler factors. The
SED of the core of Cen A, for instance, can be fit with δD ≈ 1 and bulk Lorentz
factors Γ ≈ few [292]. Likewise, the SEDs of the FR1 radio galaxies M87 and NGC
1275 [264] and M87 [288] are well fit with δD ≈ 2 and Γ ∼ 4.
The much larger values of Γ for BL Lac objects than for their putative par-
ent population, the misaligned FR1 radio galaxies, is contrary to simple unification
expectations. Moreover, even though the γ-ray luminosities from FR1 radio galax-
ies are much smaller than that of BL Lac objects (Fig. 7), they are still larger than
expected by debeaming the radiation of BL Lac objects with Γ 20. Additional soft
target photons that can be Compton scattered to high energies result in a reduction of
the value of δD compared to those implied by the one-zone synchrotron/SSC model.
These target photons can be produced in a structured jet, as in the spine and sheath
model [299]. Another soft photon source arises if blazar flows decelerate from the
inner jet to the pc scale [312], seemingly in accord with the mildly relativistic flows
at the sub-pc scale found in radio observations of Mrk 421 and Mrk 501.
7.3 The EBL
In the most general sense, the EBL refers to the cosmic microwave background
radiation (CMBR), the radiations from all the past stars and black holes, the glow from
annihilating and decaying dark matter, and any residual emissions from exotic particle
decays. The EBL at infrared, optical and UV frequencies originates predominantly
from stellar emissions, either directly or after being absorbed and reradiated by dust.
The present EBL is a consequence of the star formation history of the universe, light
absorption and re-emission by dust, and different types of dust extinction in various
classes of galaxies throughout cosmic time. The IR and optical/UV EBL attenuates
TeV and 10–100 GeV γ-rays, respectively. Knowledge of the absorption due to the
EBL is therefore necessary to infer the intrinsic, escaping spectra of extragalactic
γ-ray sources.
Directly measuring the EBL at IR and optical wavelengths is difficult because of
the interfering foreground zodiacal light and Galactic synchrotron radiation. Many
attempts have been taken to predict the EBL intensity. Empirical methods [309, 336,
338] sum optical/IR emissions from galaxy-counts using, e.g., luminosity-dependent
galaxy SEDs, and extrapolating to high redshift where data is lacking. Semi-analytic
models of the EBL [304, 316, 334] model galaxy formation following mergers of
dark matter halos, with effects of supernova feedback, dust attenuation, and metal
production considered. Other models [145, 319, 320] are based on integrating stellar
light with dust absorption, employing arguments for the star formation rate and
the initial mass function of the stars. A lower bound to the EBL intensity can be
determined by galaxy counts at different frequencies and redshifts. The upper panel
in Fig. 23 shows a comparison of observations with models, and inferred constraints.
Curves are model C from [145] (solid black curve); best fit model from [320] (red
dotted curve); [309] (dashed green curve); the fiducial model of [316] (dashed blue
curve); and the fast evolution and baseline models from [338] (upper and lower dot-
dashed violet curves, respectively). Orange curves and inverted triangles are upper
limits derived from blazar observations by [324] and [310], respectively.
Gamma-ray astronomy offers a special technique to probe the EBL at IR through
UV frequencies, because photons of the EBL attenuate γ rays via pair production
through the γγ → e+ e− process. The threshold condition implies that 400 GeV
320 C. D. Dermer
λ [μm]
3 2 1 0
10 10 10 10
2
10
-14
10
E I(E) [nW m sr ]
-1
E u(E) [erg cm ]
-3
-2
1
10
-15
10
Finke et al. (2010)
0
Kneiske et al. (2004) 10
Franceschini et al. (2008)
Gilmore et al. (2009)
Stecker et al. (2006)
-16
10 -3 -2 -1 0 1
10 10 10 10 10
E [eV]
Fig. 23 Upper EBL observations, models, and constraints [145]. See text for references and details.
Lower Highest energy photons observed from blazars and GRBs with the Fermi-LAT as a function
of the source redshift. Curves giving the τγγ = 3 optical depth for EBL attenuation are shown for
different models as labeled [293]
photons can pair produce on soft EBL photons with 1 µ wavelength, and 4 TeV
photons with 10 µ EBL photons, etc., so that
E1
λ(μ) ⇔ .
0.4 TeV
Analysis of the effects of this process have already ruled out high-intensity models
of the EBL at IR/optical frequencies [293]. Ruling out other EBL models will require
longer integration times to improve implementation of tests like the following:
1. Flux ratio method. By assuming that intrinsic blazar spectra are independent of
redshift, then the ratio of high to low energy fluxes will decrease with redshift
due to enhanced EBL attenuation [298]. It is necessary to use separate blazar
populations in this test since Fermi has established significant spectral differences
between FSRQs and BL Lacs, but this test is not yet sensitive enough to identify
any trend of spectral softenings due to EBL absorption.
2. Deabsorbed γ-ray spectrum. By extrapolating the Fermi-LAT spectrum of an indi-
vidual source to high energies, and assuming that the spectrum does not harden
with energy, then deabsorbing the spectrum limits the range of EBL models
that are compatible with this underlying assumption [314]. Upper limits are also
placed on the allowed EBL intensity by assuming that the intrinsic source spec-
trum is limited in hardness by some theoretically determined value [324, 337],
which depends on details of the particle acceleration mechanism [339]. Specific
approaches determine the likelihood of different EBL models to be compatible
with the measured γ-ray spectrum by assuming an intrinsic (usually power-law)
source spectrum.
3. Distribution of highest energy photons. The probability of detecting high-energy
photons depends on the γγ optical depth associated with a given EBL model. The
detection of such photons, though rare in number, can place severe demands on the
apparent power budget of a source. If the absolute power or energy if separately
determined, e.g., by using optical breaks to infer the jet opening angle in GRBs,
then the required energy release may be extreme and call the EBL model into
question (see the lower panel in Fig. 23).
4. Detection of lobes of radio galaxies. Photons of the EBL are Compton scattered
to GeV energies by relativistic electrons in the radio lobes of nearby extended
radio galaxies. Thus measurements of the GeV lobe flux can be used to determine
the level of the EBL [313]. The first radio galaxy with radio lobes resolved by
Fermi is Cen A [46]. Due to the low significance of the Fermi detection and
the lack of data at energies ∼2 GeV, the method cannot yet be applied. Longer
exposures should reduce the error bars below 2 GeV and detect photons in the
critical regime above ∼2 GeV where the emission from the up-scattered EBL is
expected to dominate. Besides Cen A, Cen B and Fornax A—two other radio
galaxies with lobes—have recently been detected [92]. These objects, as well as
NGC 6251 [325], are potential targets for resolving extended lobe emission to
measure or constrain the EBL.
Implicit in the above methods is the assumption that the γ rays are made at
the source. This need not be the case if blazars or GRBs are sources of UHECRs.
UHECR protons that are able to escape from the blazar and structured regions without
deflection can deposit energy in transit through the IGM via photopair production.
The cascade radiation can produce persistent TeV emission that would confuse the
interpretation of the attenuation due to the EBL. This is less likely for UHECRs
accelerated by GRBs, where deflection of the UHECRs would suppress the emission
generated by the UHECRs. This is currently an active area of research; see, e.g.,
[157, 306–308, 329, 335].
322 C. D. Dermer
8 Fermi Observations of Gamma Ray Bursts
The effective lifetime for GRB studies using EGRET’s spark chamber on the Comp-
ton Gamma-ray Observatory ended ≈4.5 years into mission, after 1996 [20].20 The
depletion of spark-chamber gas was mitigated through the introduction of a narrow-
field mode suitable for pointed observations. This made the chance of catching a
GRB, proportional to EGRET’s FoV, too improbable without rapid automated slew-
ing, which was not possible for CGRO. Consequently EGRET only detected a total
of five spark-chamber GRBs, all early in the mission [352]. These are GRB 910503,
GRB 910601, the superbowl GRB 930131, the famous long-lived GRB 940217 [358],
and GRB 940301. In the wide-field mode, EGRET was sensitive to ≈1/25th of the
full sky, which is ≈1/5th as large as the FoV of Fermi [10].
8.1 Fermi LAT Observations of GRBs
Since Fermi operations began, 13 GRBs were reported as significantly detected in

the LAT by the Fermi Collaboration from early August 2008 through calendar year
2009.21 The year 2010 saw a dearth of bright LAT GRBs, but the rate has picked up
in 2011 with the detection of a few remarkable events like GRB 110721A and GRB
110731A, the former of which displays classic “fast-rise, exponential decay”-type
GBM light curves with prompt LAT emission [369, 371], and the latter of which
is the first long-duration GRB jointly detected with Fermi-LAT and Swift from the
prompt phase into the afterglow [348, 364]. In the meantime, the development of the
LLE technique ([24]; Sect. 1.2) has found LAT emission from several GRBs in the
30–100 MeV range. Given the addition of these GRBs, the rate of GRB detection
with the LAT is about 1 per month, and the number of Fermi-LAT GRBs has reached
nearly 30. Because analysis is in progress on 2011 GRBs, and there were so few
GRBs in 2010, here we review GRB observations only through 2009.
All LAT GRBs during this period are also GBM GRBs and comprise the brighter
GBM GRBs, as already expected from a comparison between EGRET and BATSE
GRBs in terms of fluence [360]. The 13 Fermi LAT GRBs observed through 2009
include 11 long GRBs and 2 short bursts, namely GRB 081024B and GRB 090510
(z = 0.903). The most studied—because they are brightest—GRBs are GRB
090902B (z = 1.822) [342], which provides the first strong evidence for a hard
spectral component in long GRBs; GRB 080916C (z = 4.35) [120], the first bright
long GRB; and GRB 090926A (z = 2.106), a burst with a narrow spike from the
lowest to highest energies in an SED that requires both a Band function and a hard
20 Very bright GRBs like GRB 990123 [349] could still be detected far off the COMPTEL and OSSE
axes while making a signal in the Energetic Gamma Ray Experiment Telescope’s Total Absorption
Shower Counter. EGRET TASC and BATSE data were used to made the discovery of the additional
hard component in GRB 941017 [355].
21 See http://fermi.gsfc.nasa.gov/ssc/observations/types/grbs.
-4
10
GRB
100 MeV - 10 GeV Fluence (erg cm )

10%
-2
080916c
GRB
GRB 090902B
090510 1%
-5 GRB GRB
10 930131 940217
GRB 100%
081024B GRB
080825C
GRB
-6 090217 0.1%
10
GRB 940301 910503
910601
-7
10
-6 -5 -4 -3
10 10 10 10
-2
20 keV - 2 MeV Fluence (erg cm )
Fig. 24 Fluence-fluence diagram showing 6 Fermi [345] GRBs (red data points) and 5 EGRET
spark-chamber [360] GRBs (black data points). The EGRET fluence is measured from 100 MeV
to 5 GeV, whereas the Fermi LAT fluence is measured from 100 MeV to 10 GeV. Short hard GRBs
are circled
power-law component to fit. GRB 090926A also reveals an extraordinary spectral

softening at 1 GeV in its time-integrated spectrum when the hard LAT spectral com-
ponent is bright. These long GRBs had GBM fluences F in the 20–2000 keV range
10−4 erg cm−2 (Fig. 24). The bright, short GRB 090510, with F ∼ = 10−5 erg cm−2 ,
also shows (like GRB 090902B and GRB 090926A) a distinct high-energy power-
law spectral component in addition to a Band component [124]. Its short duration,
large distance, and the detection of a 31 GeV photon permit strict tests on quantum
gravity theories that predict a dependence of the speed of light in vacuo that is linear
with energy [343].
Besides these notable GRBs are the less well-known and also less fluent long
duration GRBs 090323 (z = 3.57), 090626 and 090328 (z = 0.736), the widely
off-LAT-axis GRB 081215A, the first LAT GRB 080825C [344], the unusual GRB
090217 [346] showing none of the typical properties of LAT GRBs, and the unre-
markable LAT GRBs 091003A and GRB 091031. The weakest fluence GRB of the
sample is the first short GRB detected at LAT energies, GRB 081024B [345], with
F ≈ 4 × 10−7 erg cm−2 . The weakness of this GRB could be related to the
high E pk ≈ 2–3 MeV of its Band-function component, but the time-averaged
E pk ∼= 4 MeV for GRB 090510 between 0.5 and 1 s after trigger is even higher [124].
For those GBM GRBs occurring within the LAT FoV, detection of GRBs with
the LAT is almost guaranteed when F 10−4 erg cm−2 . The detection rate slips to
less than 50 % when F ≈ 3 × 10−5 erg cm−2 , and becomes highly improbable for
F 10−5 erg cm−2 . This behavior undoubtedly reflects a distribution in the ratios
of 100 MeV LAT to GBM energy fluence [356].
324 C. D. Dermer
In the first 16 months of Fermi science operations, 1 GRB per month was
detected with the Fermi LAT, or ≈9 GRB/year, with LAT detecting short GRBs at
≈10–20 % of the rate of long GRBs. GRBs are detected with the GBM at a rate
of 250 GRB/year, or ≈500 GRB/year (full-sky). When corrected for FoV, EGRET
detected ≈25 GRB/year (full sky), while the Fermi LAT detects ≈50 GRB/year
(full sky). Given the much larger effective area of Fermi than EGRET, by a factor
≈6 [≈(8000–9000 cm2 )/(1200–1500 cm2 )], this small rate increase is something
of a surprise, compounded by the ongoing sparse period of Fermi LAT detections
of GRBs in 2010. Part of this difference is the stronger detection criteria of Fermi
LAT than EGRET. But an improvement in flux sensitivity by a factor ≈6, with an
accompanying rate increase by only a factor ≈2–3 suggests that LAT GRBs are being
sampled in a portion of their log N −log F distribution that is flattened by cosmolog-
ical effects. This is consistent with the known redshifts of LAT GRBs, which range
from ≈0.7 to z = 4.35, with a very rough average redshift of z = 2 for long GRBs
and z ≈ 1 for short GRBs (based only on GRB 090510). If typical, both classes of
GeV-emitting GRBs would be subject to strong cosmological effects on the fluence
and flux distributions.
Fluence-Fluence Diagram
Figure 24 shows the fluence-fluence diagram for the 5 EGRET spark-chamber [360]
and for 6 Fermi [345] GRBs. Most GBM GRBs have F 10−5 erg cm−2 , and are
only rarely detected with the LAT. Because of the small number of LAT GRBs, it
is not yet clear whether there is a systematic difference between fluence ratios of
EGRET and Fermi LAT-detected GRBs. The weakest Fermi LAT GRBs in terms of
GBM fluence are both short duration GRBs. This could indicate a preference for short
GRBs to have a larger ratio of LAT to GBM fluences than long GRBs, depending on
possible triggering biases, e.g., increased LAT background for long GRBs.
The apparent isotropic energies Eiso of GBM and LAT emission for LAT GRBs
with known redshifts are in several cases 1054 erg. For GRB 080916C, Eiso ≈
1055 erg. The LAT GRBs tend to have the largest energies of all measured GRBs,
and as a result are good for radio studies [137].
Fermi LAT GRB Phenomenology
Some distinct behaviors have been identified in Fermi GRBs, namely:

• Extended (long-lived) LAT (100 MeV–GeV) emission compared to the GBM
(20 keV–2 MeV) emission, known already from EGRET observations, especially
GRB 940217 [358].
• Delayed onset of the LAT emission compared to the GBM emission, in both long
and short classes of GRBs.
• Power-law temporal decay profiles of the LAT extended emission, decaying typi-
cally ∝ t −1.5 [354].
• Appearance of separate power-law spectral components with photon number index
harder than −2.
• Delayed onset of the lowest energy GBM emission at ≈10 keV, seen for example
in GRB 090902B and GRB 090926A.
• Quasi-thermal Band function components with steep Band β found, e.g., in GRB
090902B at E 1 MeV [342].
The onsets of the >100 MeV emission appear to be delayed by ∼0.1t90 com-
pared to the 100 keV–MeV emission (with t90 measured, e.g., in the 50–300 keV
GBM/BATSE range). This is one of the key and unanticipated results on GRBs
from Fermi, and it appears to operate equally for both the long- and short-duration
LAT GRBs. There have as yet been no LAT detections of members of the low-
luminosity/sub-energetic class of GRBs that includes GRB 980425 and GRB 030329,
nor have any X-ray flashes or X-ray luminous GRBs been detected with the LAT.
Because GBM’s primary triggering modes are similar to BATSE, high E pk , relatively
low-z GRBs are more likely to be detected compared to Swift.
The luminosity function can be constructed from observations of GRBs. This
function gives information about the rate density, the opening angle, and the total
nonthermal energy radiated by GRBs. A description of the luminosity function for
GRBs, and how the luminosity density can be evaluated for different GRB models
is now given.
8.2 GRB Luminosity Function
The luminosity function Y (L) describes the distribution of apparent isotropic lumi-
nosities of members of a common source class. For GRBs, one can construct the
luminosity function for, at least, the long duration (classical) GRBs, the short hard
GRBs, and the X-ray flashes and sub-energetic (low-luminosity) GRBs. As defined
by [361], the luminosity function
α β −1
dN dN L L
YLZVD (L) = L = = Y0 + H (L; L l , L u ), (146)
dL d ln L Lb Lb
with normalization L u /L b
Y0−1 = Lb dx(x α + x β )−1 .
L l /L b
326 C. D. Dermer
As defined by [151], the luminosity function

⎧ α
⎪
⎪
⎨ L ∗ , L ∗ /Δ1 < L < L ∗
L
YGPW (L) = c0 β , (147)

⎪
⎪
⎩ LL , L ∗ < L < Δ L
2 ∗
∗
with normalization
β
c0−1 = α−1 (1 − Δ−α −1
1 ) + β (Δ2 − 1).
Generally, the luminosity function depends on z, so that Y (L) = Y (L; z).
Luminosity Density from the Luminosity Function
The luminosity density = dE/dVdt at redshift z, referred to comoving volume, is

given by ∞
(z) = ṅ(z) Δt d(ln L) L Y (L; z), (148)
0
where ṅ(z) is the comoving rate density and the integration over dL or d ln L depends
on the definition of the luminosity function Y (L). The local luminosity density, which
depends on the low-redshift luminosity function and the local rate density, is
∞
0 = ṅ 0 Δt d(ln L) L Y (L). (149)
0
The relation of mean duration Δt to luminosity is normalized so that the fluence
F = L Δt.
For the model of Ref. [151], Y (L) is dimensionless (per unit ln L), and the expres-
sion for the local luminosity density is analytic, given by
−(1+α) 1+β
GPW = ṅ 0 c0 ΔtL ∗ 1 − Δ1 /(1 + α) + Δ2 − 1 /(1 + β) , (150)
−β
with c0−1 = α−1 (1 − Δ−α −1
1 ) + β (1 − Δ1 ).
For the model of Ref. [361], Y (L) has dimensions of L −1 , and the local luminosity
density requires a simple numerical integration of the expression
∞ α β −1
ṅ 0 Y0 Δt L L
LZVD = dL L + , (151)
L∗ 0 L∗ L∗
with −1
L u /L b
Y0−1 = LB dx (x α1
+x ) α2 −1
L low /L b
Elementary Cosmology: Following, e.g., [361], we write the differential rate of

GRBs with redshift z between z and z + dz and luminosity L in the range L to
L + dL by
dN ṅ(z) dV (z)
= Y (L). (152)
dtdzdL 1 + z dz
The comoving volume element is
dV (z) c 4πd L2
= . (153)
dz H0 (1 + z)2 Ωm (1 + z)3 + ΩΛ
To derive this, note that dV ∗ = cdt ∗ dA∗ , so dV ∗ /dz = c|dt ∗ /dz|dA∗ . With dA∗ =
(R∗ χ)2 dΩ∗ = d L2 (z)dΩ/(1 + z)4 (Ref. [4], Eq. (4.43); see below). The cosmology
of the universe we inhabit is a flat ΛCDM universe, that is, a dark-matter dominated
flat universe with nonzero cosmological constant, well-described by
dt ∗ −1
= . (154)
dz H0 (1 + z) Ωm (1 + z)3 + ΩΛ
This defines, noting that V = (1 + z)3 V∗ , Eq. (153). Our standard values are H0 =
72 km s−1 Mpc−1 , Ωm = 0.73 = 1 − ΩΛ .
The proper distance, though not directly measurable, is the distance between two
objects that would be measured at the same time t. The proper distance at the present
epoch is just the comoving coordinate, so

χ/c z dt ∗
d pr op = χ = ct = c dt = c dz (1 + z ). (155)
0 0 dz
The energy flux for a source isotropically radiating luminosity L = dE/dt at proper
distance d pr op is related to the energy flux from a source with isotropic luminosity
L ∗ = dE∗ /dt ∗ at luminosity distance d L through the relation
dE dE/dt dE∗ /dt ∗

Φ= = = . (156)
dAdt 4πd 2pr op 4πd L2
t
The fluence F = t12 dt Φ(t) measured over some time interval t1 –t2 is therefore
related to the apparent isotropic energy release through the expression
4πd L2 F
E∗ = . (157)
1+z
328 C. D. Dermer
For an expanding universe, dE∗ = ∗ dN = (1 + z)dN = (1 + z)dE, and dt ∗ =

dt/(1 + z). From the definition of d L , the energy flux

dE L∗ dA dE dt ∗ dE (1 + z)2 dA
Φ= = = = Φ. (158)
dAdt 2
4πd L 4πd L2 dE dt dAdt 4πd L2
∗
Therefore dA = R 2 χ2 dΩ = 4πd L2 /(1 + z)2 . At emission time t∗ , dA∗ = R∗2 χ2 dΩ∗ ,

and dA = R 2 χ2 dΩ, so that dA∗ /dA = 1/(1 + z). This implies from the def-
inition of the luminosity distance and the angular diameter R∗ χ, with dE/dt =
(1 + z)−2 (dE∗ /dt ∗ ), and letting dΩ = dΩ∗ ,
2
R
d L = d L (z) = (1 + z)d prop = (R∗ χ) = (1 + z)2 d A
R∗
z
c 1 cz 3Ωm
= (1 + z) dz → 1+ 1− z + O(z 2 ) . . . (159)
H0 0 Ωm (1 + z )3 + ΩΛ H0 4
∼
= 4170z(1 + 0.8z) Mpc for a flat universe.
Compare this derivation from Ref. [4]. The directional event rate, or event rate per
sr, is
∞
d Ṅ 1 dt ∗ dt ∗ (R∗ χ)2 ṅ ∗ (z)
=
dV ∗ ṅ ∗ (t∗ ) =c
dz , (160)
dΩ 4π dt 0 dz (1 + z)
where the burst emissivity ṅ ∗ (z) gives the rate density of events at redshift z. Volume
density in comoving and proper coordinates is related by the expression ṅ ∗ (∗ ; z) =
ṅ co (∗ ; z)(1 + z)3 . Assuming separability of source emission properties and the rate
density, then ṅ co,i (z) = ṅ 0i Σi (z), where Σi (z) is the structure formation history
(SFH) of sources of type i, defined so that Σi (z 1) = 1, and ṅ 0i is the local
(z 1) rate density of bursting sources of type i. Thus
∞
d Ṅ c d L2 (z)ṅ i Σi (z)
= dz . (161)
dΩ H0 0 (1 + z)3 Ωm (1 + z)3 + ΩΛ
This expression can be easily generalized to accommodate spectral behavior.

Star formation rate functions: The redshift-dependent rate density by Porciani
& Madau (2001; SFR2) that has roughly constant star formation at z > 2 [365] is
described by
exp(3.4z)
ṅ PM (z) = 23ṅ 0 . (162)
exp(3.4z) + 22
The analytic function
1 + a1
ΣSFR (z) = , (163)
(1 + z)−a2 + a1 (1 + z)a3
Fig. 25 Star formation rate 1000

4
factors [158, 357, 374] Σ(z)=(1+z)
SFR IR (n = 8)
Sanders (2004)
100
Porciani & USFR
Σ (z) Madau (2001)
Le & Dermer (2007)
i
10
Hopkins & Beacom (2006)
SFR HB
Yuksel
et al. (2008)
1 LSFR
Wick et al. (2004)
0 1 2 3 4 5 6 7 8 9 10
Redshift z
[374], with a1 = 0.005, a2 = 3.3, and a3 = 3.0 for the lower star formation rate
(LSFR) and a1 = 0.0001, a2 = 4.0, and a3 = 3.0 for the upper star formation rate
(USFR) describes extreme ranges of optical/UV measurements without and with
dust extinction corrections, respectively.
The SFR history (SFR3) of Hopkins and Beacom (2006) [357], which is interme-
diate between the LSFR and USFR, is given by their analytic fitting profile
1 + (a2 z/a1 )
ΣHB (z) = , (164)
1 + (z/a3 )a4
where a1 = 0.015, a2 = 0.10, a3 = 3.4, and a4 = 5.5 are best fit parameters
(Fig. 25). An update of the SFH fit of [357] is contained in Ref. [375] in the form of
a continuous broken power law,
bη cη 1/η
1+z 1+z
ρ̇∗ (z) = ρ̇0 (1 + z)aη
+ + , (165)
B C
where a = 3.4 (b = −0.3, c = −3.5) is the logarithmic slope of the first (middle,
last) piece breaking at z 1 = 1 and z 2 = 4, and the normalization is ρ̇0 =
0.02M Mpc−3 year−1 . Using η = −10 smooths the transitions, while η → ∞
recovers the kinkiness of the original form. Here B = (1 + z 1 )1−a/b ∼ = 5000, and
C = (1 + z 1 )(b−a)/c (1 + z 2 )1−b/c ∼
= 9.
Figure 25 shows star formation rate factors used in different models for GRBs.
For data, see [367, 374]. An enhanced rate of GRBs compared to the global star
formation rate is required to fit Swift statistical data [158], as confirmed in other
studies [372, 375].
330 C. D. Dermer
Luminosity Function in Le & Dermer Model
We approximate the temporally-evolving SED of, for example, a GRB or blazar by

the expression
ν Fν ≡ f (t) ∼
= f pk S(x)H (μ; μ j , 1)H (t; 0, Δt) , (166)
and
∗
S(x) = x a H (1 − x) + x b H (x − 1), x ≡ = = . (167)
∗, pk pk pk
The bolometric fluence of the model GRB for observers with θ ≤ θ j is

∞ ∞ f (t)
F= dt d = λb f pk Δt, (168)
−∞ 0
where λb is a bolometric correction to the peak measured ν Fν flux. If the SED is

described by Eq. (167), then λb = (a −1 − b−1 ), and is independent of pk . The
beaming-corrected γ-ray energy release E∗γ for a two-sided jet is
F
E∗γ = 4πd L2 (1 − μ j ) , (169)
1+z
where the luminosity distance is given in Eq. (159). Substituting Eq. (168) for F into
Eq. (169) gives the peak flux
E∗γ
f pk = . (170)
4πd L2 (z)(1 − μ j )Δt∗ λb
Finally, by substituting Eq. (170) into Eq. (166), the energy flux becomes
f (t) = f pk H (μ; μ j , 1) H (t; 0, Δt)S(x). (171)
The observed directional event rate for bursting sources with ν Fν spectral flux
greater than fˆ¯ at observed photon energy ¯ is given by

d Ṅ (> fˆ¯ ) c ∞
∞ ∞ ∞ μjmax 1
= df dE∗γ d pk∗ d(Δt∗ ) dμ j dμ
dΩ H0 fˆ¯ 0 0 0 μjmin μj
∞
d L2 (z)ṅ co (z)Y (E∗γ , μ j , pk∗ , Δt∗ )
× dz δ f − f pk S(x) . (172)
0 (1 + z)3 Ωm (1 + z)3 + ΩΛ
For discrete values of E∗γ , pk∗ , and Δt∗ , the property distribution function
Y (E∗γ , μ j , pk∗ , Δt∗ ) = g(μ j ) δ(E∗γ −E∗γ0 ) δ( pk∗ − pk∗0 ) δ(Δt∗ −Δt∗0 ) , (173)
where g(μ j ) is the jet opening angle distribution, and fˆ¯ is the instrument’s detector
sensitivity. The detector threshold for Swift and pre-Swift GRB detectors is taken to
be ∼10−8 and ∼10−7 ergs cm−2 s−1 , respectively.
In the model of Le & Dermer (2007) [158], the form for the jet opening angle
g(μ j ) is chosen to be
g(μ j ) = g0 (1 − μ j )s H (μ j ; μjmin , μjmax ), (174)
where s is the jet opening angle power-law index; for a two-sided jet, μjmin ≥ 0.
Normalized to unity, Eq. (174) gives
1+s
g0 = . (175)
(1 − μjmin )1+s − (1 − μjmax )1+s
Integrating over μ in Eq. (172) gives the factor (1 − μ j ) describing the rate reduction
due to the finite jet opening angle. Hence, Eq. (172) becomes

d Ṅ (> fˆ¯ ) c ∞
μjmax
= df dμ j g(μ j )(1 − μ j )
dΩ H0 fˆ¯ μjmin

∞ d L2 (z) ṅ co (z) δ f − f pk S(x)
× dz , (176)
0 (1 + z)3 Ωm (1 + z)3 + ΩΛ
where f pk is given by Eq. (170). The redshift, size, and opening angle distributions
can then be calculated for comparison with data.
Fitting to redshift and opening-angle distributions of pre-Swift and Swift GRBs
using this model gives a good fit, assuming a flat ν Fν spectrum with bolometric
factor λb = 5, for the following parameters:
Δt∗ = 10 s, E∗γ = 4 × 1050 erg s−1 , s = −1.3, μmin = 0.765, μmax = 0.99875,
(177)
i.e., θmin = 0.05 (2.9◦ ), θmax = 0.7 (40◦ ). The “true” event rate ṅ G R B = (7.5–
9.6) Gpc−3 year−1 . This can be compared with beaming corrected rates of ∼=75×0.5–
40 Gpc−3 year−1 [151]. The local emissivity in all forms of electromagnetic radiation
from long GRBs is automatically implied in this model, namely
LD = ṅ G R B E∗γ ≈ (3−4) × 1043 erg s−1 Mpc−1 (178)

332 C. D. Dermer
Luminosity function: We derive the luminosity function for the Le & Dermer
(2007) model, which can be checked in the limit ∞z 1 to agree with Eq. (178). The
isotropic luminosity L ∗,iso = 4πd L2 Φ = 4πd L2 0 f /. From Eqs. (166) and (170),
f pk = L ∗,iso (t)S(x)/λb ,
where
L ∗,iso (t) = L ∗,iso H (μ; μ j , 1)H (t; 0, Δt)

E∗γ
= H (μ; μ j , 1)H (t; 0, Δt)
Δt∗ (1 − μ j )
is the apparent isotropic luminosity that would be measured within the beaming
cone during time t from 0 to Δt. Thus the relation between the apparent luminosity
L = L ∗,iso and l∗ = E∗γ /Δt∗ = 4 × 1040 erg s−1 is, obviously, L = l∗ /(1 − μ j ) for
a two-sided jet (0 < μ j ≤ 1). Thus dL/dμ j = L/(1 − μ j ), and
dN dN
dL = dμ j (1 − μ j ), (179)
dV dtdL dV dtdμ j
noting the extra factor (1 − μ j ) which accounts for the smaller opening angle of the
more apparently luminous jets.
From the formulation of the model,
dN
= ṅ i g(μ j ) = ṅ i g0 (1 − μ j )s H (μ j ; μmin
j , μ j ).
max
(180)
dV dtdμ j
Therefore

dN dμ j dN (1 − μ j )s+2
=
(1 − μ j ) = ṅ i g0 H (μ j ; μmin
j , μj )
max
dV dtdL dL dV dtdμ j L

g0 ṅ i l∗ s+3 l∗ 1
= H L; , . (181)
l∗ L 1 − μj min 1 − μmax
j
The luminosity function in this model is

s+3
dN g0 l∗
ΦLD (L) = = H L; 1.7 × 1051 erg s−1 , 3.2 × 1053 erg s−1 .
dL l∗ L
(182)
Therefore L 2 Φ(L) ∝ L −1−s ∝ L 0.3 . From Eq. (175), g0 = 0.051.
Integrating Φ(L) over L gives the emissivity according to Eq. (149). We find
that integration of Eq. (182) recovers the luminosity density given by Eq. (178). The
luminosity function in the form ṅ i L 2 ΦLD (L) is plotted in Fig. 26. The model is
overconstrained, which problem can easily be relaxed by changing the functional

dependence of the jet opening angle.
Local Luminosity Density of GRBs
The “ν Fν ” intensity I (erg/cm2 -s-sr) = m e c2 2 φ()/4π for unbeamed sources,

where φ() is the differential photon flux, is given by

m e c3 ∞ dt ∗ 2∗ ṅ co (∗ ; z)
I =
dz . (183)
4π 0 dz 1+z
(Eq. (4.57) in Ref. [4]). This equation also applies to ∞beamed sources provided that
the apparent comoving event rate density ṅ co (z) = 0 d∗ ṅ co (∗ ; z) is increased by
the inverse of the beaming factor, f b , to get the true source rate density (see Eq. (11)).
The photon luminosity density at redshift z is given by
∞
(z) = m e c2 d∗ ∗ ṅ co (∗ ; z) (184)
0
and the local luminosity density 0 ≡ (z 1). Assuming that the only property
of long-duration GRBs that is redshift dependent is the rate density, then (z) =
∞
0 Σ(z). The integrated intensity I = 0 dI is related to the mean photon flux
Φ = dE/dAdt through the relation I = Φ/4π, so
H0 Φ
0 =k , (185)
c
where ∞ Σ(z)
k −1 ≡ dz . (186)
0 (1 + z)2 Ωm (1 + z)3 + ΩΛ
Table 10 gives values for k obtained by numerically integrating Eq. (186), using the
SFR factors shown in Fig. 25 with the standard ΛC D M cosmology with Hubble
constant H0 = 72 km/s-Mpc, Ωm = 0.27 and ΩΛ = 0.73. The integration was
truncated at a redshift z max = 10.
Table 10 Constant k, from

SFR k
Eq. (186), for different star
formation rate factors Constant comoving 1.85
LSFR [374] 0.52
USFR [374] 0.13
HB [357] 0.37
LD [158] 0.33–0.37
334 C. D. Dermer
The mean flux Φ over the full sky can be estimated from the total fluence per year
from GRBs, given the number of GRBs per year. For 1293 GRBs in 4B catalog the
total fluence (20–300 keV) is 63.80 × 10−4 erg/cm2 and the total fluence >20 keV
is 153.63 × 10−4 erg/cm2 (M. Gonzalez, private communication, 2003). For the
4th BATSE Catalog, considering 666 Burst/year full sky with 1293 bursts implies
1293/(666 GRB/year) = 1.94 years = 61225300 s, so that the mean 20–300 keV
flux 1.04 × 10−10 erg/cm2 -s or 3.3 × 10−3 erg/cm2 -year. Likewise, for the total
>20 keV fluence was, we obtain 2.54 × 10−10 erg/cm2 -s or 7.9 × 10−3 erg/cm2 -year
Band (2002) [347] obtains 550 GRBs/year for BATSE exposure, which brings the
flux down to 6.5 × 10−3 erg/cm2 -year. Subtracting 10–20 % for short GRBs gives
(5–6) × 10−3 erg/cm2 -year.
From Eq. (185), writing Φ−2 = Φ/(10−2 erg/cm2 -year) we find 0 = 2.3 ×
10 kΦ−2 erg/Mpc3 -year. With Φ−2 ≈ 0.5 and k ∼
43
= 0.35, then

k Φ−2
0 = 0.5 × 1043 erg/Mpc3 -year, (187)
0.35 0.6
in agreement with the results of Eichler et al. (2010) [353].

This value is at least a factor of 5 smaller than values obtained through statistical
treatments of GRB data [151, 158, 361], as can be seen from Table 11. Here we give
the local luminosity densities for long GRB models derived either numerically or
analytically from the luminosity functions derived in these models.
The reason for the discrepancy between the local emissivity of long duration GRBs
derived from BATSE data and inferred from Swift data is suggested by Fig. 26. This
figure compares different models for the local GRB luminosity functions, plotted
in the form of a differential emissivity function ṅ 0 Δt L Y (L), where ṅ 0 is the local
rate density, and the characteristic duration of a GRB in the explosion frame is
Δt = 10 s. Because the treatments of Schmidt (2001) [368] and Guetta, Piran, &
Waxman (2005) [151] are based on BATSE observations in the 50–300 keV range, a
bolometric correction factor =6.3 is applied to normalize them to the 1 keV–10 MeV
range (D. Wanderman, private communication, 2010). In this case, the implied lumi-
nosity density from the treatments of the luminosity function agree within factors
of ≈3.
Table 11 Local luminosity density of long GRBs for different models

Model 0 Photon energy range
(1044 erg/Mpc3 -year)
Schmidt (2001) [368] 0.03 50–300 keV
Guetta et al. (2005) [151] 0.05–0.08 50–300 keV
Le & Dermer (2007) [158] 0.30–0.38 All photon energies
Liang et al. (2007) [361] 2.0 1 keV–10 MeV
Wanderman & Piran (2010) [372] 1.3a 1 keV–10 MeV
a Normalizing to the Swift durations brings this value down by a factor of ≈2
Le & Dermer (2007)

Guetta et al. (2007)
Liang et al. (2007)
10 3 Wanderman & Piran (2010)
Schmidt (2001)
n L Φ(L) (Δt/10 s)
[10 erg Gpc yr ]
-1
-3
1 02
50
(2)
0
10 1
10 0 10 1 10 2 10 3 10 4
50 -1
L(10 erg s )
Fig. 26 Long duration GRB luminosity function, plotted in the form of a differential luminosity
density = ṅ 0 L Y (L)Δt, for various models
This figure show that the integrated luminosity density depends sensitively on
the maximum luminosity taken in the construction of the luminosity function. When
redshift information is available, very luminous GRBs are found with apparent lumi-
nosities reaching 1053 erg/s. To fit these luminous GRBs, the luminosity function
must extend to large luminosities, and these very rare, very luminous GRBs make
a significant contribution to the integrated luminosity density. For such luminosity
functions, the large fluence GRBs make up a large, even dominant part of the total
GRB fluence, but such high fluence events are so rare that the chances to now of such
a GRB with large, 10−2 erg/cm2 , fluence having been detected is low.
The local electromagnetic emissivity of long GRBs, given in Table 11, is therefore
found to be 1044 erg/year-Mpc3 [350]. This can be compared with the minimum
local emissivity of U H EC R ≈ 4 × 1044 erg/s-Mpc required to power UHECRs, as
seen in Fig. 27, and discussed in more detail in the next section. Related estimates
can be made for other source classes [361, 363], including the short hard GRB
class, the X-ray flashes, the sub-energetic GRBs, and engine-driven supernovae with
relativistic outflows that lack GRB-type emissions. These classes, though they have
much larger local rate densities than long-duration GRBs, are not much preferred
over long-duration GRBs on the basis of their local luminosity density. The so-called
low-luminosity GRBs, for example, have a large luminosity density in ejecta kinetic
energy, 1046 erg Mpc−3 year−1 [362, 373], but smaller or comparable nonthermal
γ-ray luminosity densities than long-duration GRBs.
336 C. D. Dermer
Fig. 27 Local luminosity density requirements on different models for UHECR production from
long-duration GRBs with rate density following various star-formation-rate functions shown in
Fig. 25
8.3 Closure Relations
The relativistic blast wave theory for the afterglow has become the industry standard
model. Familiarity with the derivation of the closure relations is required for basic
knowledge of GRB physics. The reader may see22 for the derivation, [4] for review,
and [366] for the essential treatment.
Interpreting the delayed onset of the Fermi-LAT radiation as due to external shock
emission, one approach is to suppose that the blast wave decelerates adiabatically in a
uniform surrounding medium [359], with closure relation ν Fν ∝ t (2−3 p)/4 ν (2− p)/2 ,
where p is the electron injection index. A value of p ≈ 2.5 gives a reasonable fit
to the data. Another regime to consider is a radiative GRB blast wave [354], where
the comparable closure relation is ν Fν ∝ t (2−6 p)/7 ν (2− p)/2 , with p ≈ 2 giving a
plausible fit to the data. The adiabatic model requires unusually low densities and
magnetic fields for GRB 080916C, and the radiative model supposes pair loading
can help achieve strong cooling.
Alternate leptonic models for Fermi LAT GRBs include photospheric models with
the photospheric emission passing through shocked plasma in the colliding shells
or external shocks of the GRB outflow [370]. A joint Fermi-Swift paper examines
leptonic afterglow models for GRB 090510 [351].
22 http://heseweb.nrl.navy.mil/gamma/~dermer/notes/index.htm
9 Fermi Acceleration, Ultra-High Energy Cosmic Rays,

and Black Holes
Fermi acceleration is intrinsically coupled to the presence of a magnetic field that

governs particle motions. The energy is gained stochastically in second-order Fermi
scenarios, for example, via resonant pitch-angle scattering, or the particle energy
is gained systematically by those few particles that diffuse back and forth across a
shock front multiple times before convecting downstream (Sect. 5.5). In either case,
the energy gain rate is related to the Larmor rate v/rL . The magnetic field and the
size scale set a fundamental limit on the maximum particle energy through the Hillas
criterion, Eq. (78).
9.1 Maximum Particle and Synchrotron Photon Energy
In Fermi acceleration scenarios, the maximum particle energy is obtained by bal-

ancing acceleration rate with the energy-loss rate. For √
relativistic electrons, the max-
imum comoving electron Lorentz factor is γmax = 6πe/φσT B . The maximum
synchrotron photon energy for a relativistic jet source at redshift z is, from Eq. (145),
27 Γ
syn,max ∼
= , (188)
8α f φ(1 + z)
or
(Γ /1000)
E syn,max ∼
= 240 GeV. (189)
φ(1 + z)
The delayed appearances of the highest energy photons from Fermi LAT GRBs,
which typically happens at times after the GBM t90 , calls into question a synchrotron
origin for these photons, unless Γ remains large at the end of the prompt phase of
GRBs.
The maximum energy for protons and ions is obtained from the Hillas criterion,
written as rL = E /β Q B < R . The maximum escaping particle energy E max =

Γ E max < Γ E = Z βeB R, noting that R = R/Γ from length contraction of
the stationary frame size scale as measured in the comoving frame. Relating the
magnetic field energy density by a factor B < 1 times the proper frame energy
density associated with the wind luminosity L, then B 2 = 2 B L/(βc R 2 Γ 2 ), and

Ze 2β B L
E max < , (190)
Γ c
338 C. D. Dermer
implying
3 × 1045
Lγ Γ 2 E 20
2
erg/s (191)
Z 2β
[303, 384, 399], noting that the apparent γ-ray luminosity L γ < L.
Equations (72) and (74) imply

∼ A Γ mp 27 ∼ A(Γ /1000)
E ionsyn, max = m e c2 = 0.44 PeV. (192)
φZ 1 + z m e
2 8α f φZ 2 (1 + z)
Thus proton or ion synchrotron can make a γ-ray component in jet sources provided
the ion power and energy is sufficiently great.
9.2 L-Γ Diagram
Using the minimum Lorentz or Doppler factors implied from γ-ray opacity argu-
ments, Eqs. (45) and (46), along with the apparent γ-ray luminosity corresponding
to the time that Γmin is measured, we can construct an L-Γ diagram.
Figure 28 shows such a plot for blazars and GRBs [303]. Note that GRB 090510A
is a short hard GRB, whereas GRB 080916C and GRB 090926A are long-duration
GRBs. Except for PKS 2155-304, which uses HESS data for the giant outbursts in
2006, all the data were measured with the Fermi LAT. As can be seen, the powerful
58
Blazars GRBs
1. GRB 090510A
56 2. GRB 080916C
3. GRB 090926A
4. 3C 454.3 2.
54 3.
5. PKS 2155-304 1.
6. NGC 1275
[L(erg s )]
52
-1
7. Cen A
50
4.
10
48
Z = 26
log
Z=1 5.
(protons) (Fe)
46
6.
44 7.
42 Cooling vs. Acceleration

Cooling vs. Available Time
40
3 4
0.1 1 10 100 10 10
Fig. 28 Data show apparent isotropic L versus Γmin for blazars, radio galaxies, and GRBs. Solid
and dot-dashed curves plot the constraint given by Eq. (191)
Fermi GRBs have more than adequate luminosity to accelerate either protons or
ions to ultra-high energies, even with Lorentz factors Γmin ∼ 103 as inferred from
γγ opacity arguments. By comparison, the blazars and radio galaxies have smaller
apparent luminosities and also smaller Γmin . Acceleration of high-Z ions like Fe
to ultra-high energies is possible for these sources on the basis of Eq. (190), but
acceleration of protons appears unlikely, except possibly during flaring episodes.
9.3 Luminosity Density of Extragalactic γ-Ray Jet Sources
Maintaining the intensity of UHECRs against photopion losses with the CMB
requires a luminosity density in UHECRs of ∼1044 erg/Mpc3 -year [398]. This value
can easily be obtained by dividing the energy density of UHECRs at ≈1020 eV, which
is ∼10−21 erg/cm3 , by the photopion loss timescale t pπ ∼ 100 Mpc/c. Because t pπ
increases rapidly with decreasing energy, the required luminosity density remains at
this level even when fitting to the ankle of the cosmic-ray spectrum at E ≈ 4×1018 eV
[381], though precise values of the local luminosity density depends in detail on the
assumed cosmic star-formation rate factor applicable to GRBs.
The standard argument for inferring the luminosity density of a source class is
to assume that the nonthermal γ-ray luminosity represents, to an uncertain factor
of order unity, the luminosity in UHECRs. This seems reasonable because some
radiative losses into γ rays are likely to occur during acceleration. If those losses
are small, then the UHECR power could exceed the γ-ray luminosity, but if the
particles experience severe radiative losses during acceleration, then they could not
be the sources of the UHECRs [398]. This estimate is furthermore independent of
beaming factor, because for every beamed source we detect, a proportional number of
misdirected sources will point away from us. The argument suffers, however, from the
likelihood that a large percentage of the γ radiation is produced by ultra-relativistic
leptons (especially in the case of blazars where correlated variability with lower
energy bands can be monitored). Moreover, it is not certain that MeV γ rays from long
duration GRBs, which comprise the bulk of the energy output, is entirely nonthermal,
inasmuch as a thermal/photospheric interpretation can potentially resolve the line-of-
death problem that plagues nonthermal synchrotron interpretations of GRBs [396].
Thus the nonthermal LAT emission might be a more appropriate luminosity with
which to define the nonthermal luminosity density of GRBs [353].
The luminosity density of GRBs has been evaluated in detail in Sect. 8.2. The
luminosity density of radio galaxies and blazars, based on the average 100 MeV–
100 GeV flux measured in the 1LAC [45], is shown in Fig. 29.
340 C. D. Dermer
Fig. 29 γ-ray luminosity density inferred from Fermi observations of various classes of γ-ray
emitting AGNs [303], compared with the local luminosity density of long-duration GRBs (Table 11),
and UHECR source requirements
9.4 Origin of UHECRs
The sources of the UHECRs are unknown. With the anisotropy of the UHECRs
observed by Auger becoming less significant than originally reported, leaving the
cluster of UHECRs in the vicinity of Cen A as the only outstanding hotspot [6], even
Galactic sources have been considered as plausible UHECR candidates [380]. Indeed,
a wide variety of possible source classes could contribute to the UHECRs, including
magnetars and young pulsars [376, 386], particle acceleration in structure-formation
shocks [395], and new physics candidates. Criteria that the sources of UHECRs
should meet are [382]
• extragalactic origin;
• mechanism to accelerate to ultra-high energies;
• adequate luminosity density;
• sources within the GZK radius; and
• UHECR survival during acceleration, escape, and transport.
On this basis, we [4] suggest that the very luminous, very energetic extragalactic
blazar and GRB jet sources are the most plausible candidates, with the BL Lac
objects and FR1 radio galaxies perhaps favored [303]. Electromagnetic signatures of
ultra-relativistic hadrons in GRBs and blazars are not clearcut. High-energy neutrino
detections from transient or bursting γ-ray sources will be crucial to finally solve
this puzzle.
9.5 Black Holes, Jets, and the Extreme Universe
Black Hole Physics
The Kerr metric for an uncharged rotating black hole, written in the Boyer-Lindquist
coordinates {t, r, θ, ϕ} useful for asymptotic analysis in the exterior region of the
horizon of the black hole, takes the form [397]
ds 2 = (β 2 − α2 ) dt 2 + 2 βϕ dϕ dt + γrr dr 2 + γθθ dθ2 + γϕϕ dϕ2 , (193)
where the metric coefficients are given by
2Mr
β 2 − α2 = gtt = −1 + ,
ρ2
−2Mra sin2 θ ρ2
βϕ ≡ gtϕ = , γrr = ,
ρ2 Δ
Σ sin θ
2 2
γθθ = ρ2 , and γϕϕ = .
ρ2
Here, α is the lapse function and β is the shift vector,
ρ2 = r 2 + a 2 cos2 θ, Δ = r 2 − 2Mr + a 2 , and Σ 2 = (r 2 + a 2 )2 − Δ a 2 sin2 θ.
Additionally
ρ2 Δ βϕ2 √ √
α2 = , β2 = , and −g = α γ = ρ2 sin θ.
Σ2 γϕϕ
The parameter a is the angular momentum per unit mass of the Kerr black hole,
so a M is the
√ angular momentum of the black hole. The horizons are located at
r± = M ± M 2 − a 2 , as shown in Fig. 30.
The set of points described by gtt = 0 defines the ergosphere
rer g (θ) = M + M 2 − a 2 cos2 θ. (194)
Note that gtt > 0 in the region
r+ < r < rer g (θ), (195)
and consequently t˜ ≡ ∂t becomes spacelike in the above region. When r > rer g , t˜
is timelike and future directed. See Fig. 30.
In elementary electrodynamics [387], the Poynting vector in flat space far away
from a gravitating object is given by
342 C. D. Dermer
Fig. 30 Sketch of locations of the horizons r+ and r− and the ergosphere rer g (θ) in the Kerr
geometry
c
S= (E × H ), (196)
4π
and
d d
Etot = (E f ield + Emech ), (197)
dt dt
where
1
E f ield = d 3 x (E 2 + B 2 ), and (198)
8π V

d
Emech = d3x J · E (199)
dt V
give the field and mechanical energies, respectively. Thus we see that J · E is an
emissivity, so conservation of energy requires
∂u
+ ∇ · S = −J · E, (200)
∂t
where the field energy density is u = (E · D + B · H )/8π.

Electromagnetic energy extraction through the Blandford-Znajek process involv-

ing Penrose processes in curved space can be expressed in the case of a stationary
axisymmetric force-free magnetosphere as
d 2E √ √
|BZ = Sr γrr = −Hϕ Ω B r γrr (201)
dAdt
[388, 391]. The fields must satisfy the Znajek regularity condition [400]

sin2 θ r
Hϕ r+ = B (2Mr Ω − a) (202)
α r+
at the event horizon. We can suppose that the poloidal function Ω defining the fields
is radially independent, so
π
dE √ sin2 θ
|BZ = 4π Mr+ dθ γ Ω(Ω H − Ω)B+
r2
, (203)
dt 0 α
r is the radial magnetic field at the event horizon and
where B+
a
ΩH = (204)
2Mr+
is the angular velocity of the event horizon [389].

The form of Ω in Eq. (203) is found by solving the constraint equation, originally
derived by Blandford and Znajek [378], governing fields and currents in a black-hole
magnetosphere under the force-free condition. The constraint equation in the 3 + 1
formalism takes the form [391]
2
1 dH ϕ
= α(ρc Ωγϕϕ − Jϕ ), (205)
2Λ dΩ
where the charge density ρc and current density J is expressed in terms of Ω and
metric coefficients. By satisfying the Znajek regularity condition, Eq. (202), we [391]
obtained solutions a
Ω± ≡ , (206)
2Mr+ ± ρ2+
(ρ2+ = r+2 + a 2 cos2 θ) by considering the behavior of the constraint equation in the
far-field limit for radially-independent Ω. Thus

a
Ω+ ≡ , (207)
2Mr+ + ρ2+
344 C. D. Dermer
and
1
Ω− ≡ . (208)
a sin2 θ
The Ω+ solution implies a rate of electromagnetic energy extraction of
⎧

8π 2 2 ⎨ 2M ,
a 2
dE π Q 20 a a a/2M 1
|BZ = arctan − → B0 M ×
dt ar+ r+ 2M 3 ⎩ 2 π − 1 , a → M,
9 4 2
(209)
after relating Q 0 and B0 [4]. The Ω+ solution generalizes the Blandford-Znajek
[378] split monopole solution for all values of 0 < a < M. Inspection of this
solution shows that Poynting flux outflow is greatest along the equator, so the energy
flux has a pancake geometry. The jet geometry, as found ubiquitously in nature, is
not described by this solution.
The exact Ω− solution implies an electromagnetic energy flux peaking towards the
poles of the rotating black hole, but describes an inward energy flux, so is unphysical.
The Ω− solution has an additional freedom in the assignment of an arbitrary poloidal
field Λ in the prescription for the poloidal magnetic field
Λ
B P = √ (−Ω,θ ∂r + Ω,r ∂θ ) (210)
γ
[392]. Energy extraction is accomplished not only by Poynting flux outflow, but
also by the matter currents that sustain the black-hole magnetosphere. Analysis of
the timelike geodesics consistent with the Znajek regularity condition [393], and
symmetry transformation of the Ω− solution yields a dual class of exact solutions
with positive energy extraction of Poynting and matter outflows. The radial magnetic
field for this dual class of solutions takes the form
2 cos θ
Br = Λ√ . (211)
a γ sin3 θ
The extra freedom in the solutions with general θ-dependent Λ provides formal
solutions devoid of strict physical content, and with arbitrary a dependence.
Performing dimensional analysis of energy extraction using Eq. (203) (see [4,
391, 394] for detailed treatments) yields an estimate of the BZ power given by
√ √
dE γ γ 2 r2
|BZ ≈ 4π Mr+ Ω(Ω H − Ω)B+ ≈ π Mr+
r2
Ω B , (212)
dt α α H +
where B+r is the radial component of the magnetic field threading the event horizon.
Note that the expression peaks at Ω = Ω H /2 [390]. Along the equatorial direction,
√
= π/2, γ/α ∼
θ ∼ = Σ 2 /Δ. Supposing the expression is evaluated on a size scale
somewhat larger than the event horizon, to cancel the divergence, we have
dE a 2 a 2
|BZ ≈ πc 2
r+ r2
B+ ≈ 1047 M92 B42 erg/s, (213)
dt M M
where M9 is the mass of the black hole expressed in units of 109 M , and the magnetic
field is given in units of 104 G.
As expressed, and often found in the literature, the value of 104 G is artificial. The
magnetic field threading the ergosphere can be scaled to the Eddington luminosity
L Edd = 1.26 × 1047 M9 erg/s (214)
by the expression
B2 L Edd
= B Edd 2c
, (215)
8π 4πr+
where the parameter Edd reflects not only different accretion rates but different
radiative efficiencies, including the reduction in radiative efficiencies in the low
Eddington ratio, advection-dominated regime. Thus
dE
dt |BZ
≈ B Edd 1. (216)
L Edd
Jets and the Extreme Universe
The time-averaged luminosities of FSRQs extend to values in excess of L γ ≈

1049 erg s−1 (Fig. 7). In the extraordinary 2010 November flare of 3C 454.3 [9],
L γ reached apparent isotropic luminosity of (2.1 ± 0.2) × 1050 erg s−1 over a period
of a few hours, making it the most luminous blazar yet observed. Black-hole mass
estimates for 3C 454.3 are in the range 0.5 M9 4 [98], where 109 M9 M is the
mass of the black hole powering this AGN. For this range of masses, the Eddington
luminosity therefore ranges from ≈6 × 1046 erg s−1 to ≈5 × 1047 erg s−1 . During
this extreme outburst, the apparent luminosity of 3C 454.3 was more than a factor
of ≈400 greater than its Eddington luminosity. Even its time-averaged luminosity of
L γ ≈ 1049 erg s−1 is super-Eddington by a factor of ≈20.
Assuming that the Eddington condition does limit accretion flow onto the black
hole, which is likely to be the case for the long-term average luminosity if not for
the flaring luminosity, then the absolute radiant luminosity is limited to a value of
L abs 5 × 1047 erg s−1 . This is consistent with the large apparent luminosities if
the emission is highly beamed. For a simple top-hat jet beaming factor, a jet opening
angle θ j implies a beaming factor f b = θ2j /2 for a two-sided jet with θ j 1, from
Eq. (11). A mechanism for collimation is, however, required. Should this arise from
the Blandford-Znajek process, then we are still restricted to values of the absolute
Blandford-Znajek power, Eq. (213). As we have seen, by scaling the energy density
of the magnetic field to the energy density of accreted matter near the event horizon
shows that the Blandford-Znajek power is likewise Eddington-limited. Making the
346 C. D. Dermer
Fig. 31 Luminosity versus characteristic timescales for variability or duration for various types
of astronomical sources and cosmic events. Upper limit is from Eq. (17) and the horizontal line
separating the extreme and moderate universe is defined by Eq. (14)
hypothesis [383] that the extraction of energy through black-hole rotation

√ collimates
the jet outflow with cos θ j ≈ a/M, then f b ∼
= 1 − (a/M) and θ j ∼ = 2(1 − a/M),
implying a/M > 1 − (L Edd /L γ ). If the jet opening angle is a consequence of the
bulk Lorentz factor Γ of the outflow, then Γ L γ /2L Edd .
For the case of 3C 454.3 in its flaring state, when L γ /L Edd 103 , this hypoth-
esis then implies that a/M > 0.999 and Γ 23, which is consistent with the
value Γmin ≈ 14 from γγ opacity arguments [9]. In the most conservative case
with M9 = 4, a/M ∼ = 0.998 and Γ 15. This is marginally consistent with the
limiting maximum value a/M ∼ = 0.996 suggested by Aschenbach from analyses
of microquasars and the Galactic Center black hole [377]. The smaller black-hole
mass estimate, M9 = 0.5 [98], implies a value of a/M that violates this limit by a
large margin. Much work, both numerical and theoretical, has been devoted to jet
formation from the Blandford-Znajek process, and it is unclear if jet collimation can
be described by the guess that cos θ j ≈ a/M, but it is interesting to suggest this
possibility, which leads to values of Γ consistent with separate inferences regarding
the outflow Lorentz factor.
Figure 31 summarizes some various source classes and their residence in a dia-
gram of L versus time. The extreme universe, as contrasted with the moderate, or
relaxed universe, is defined by Eq. (14), when the ratio of the apparent luminosity and
variability timescale exceeds the Eddington luminosity divided by the light-crossing
time for the Schwarzschild radius, L Edd /tS = 1.26 × 1043 erg/s2 . “BH*” stands
for events where normal main sequence stars are tidally captured by a black hole,
as in the 28 March 2011 event [379], and “EMRI” stands for extreme mass-ratio
inspiral events [385]. The luminosity limit of the extreme universe is given by
Eq. (17), namely c5 /G = 3.63 × 1059 erg/s. Through observations of blazars and
GRBs, γ-ray astronomy is pushing toward this limit. A coordinated assault involving
gravitational wave observatories, the Fermi γ-ray telescope, neutrino observatories
and multiwavelength campaigns might over the next decade reach this limit of the
extreme universe.
Acknowledgments The results presented in these lectures would not have been possible without
the dedicated efforts of hundreds of Fermi Collaboration members. I would like to specifically thank
the following persons for correspondence and use of presentation material:
1. GeV instrumentation and the GeV sky with the Fermi Gamma-ray Space Telescope
Thanks to S. Ritz, P. Michelson, J. McEnery, S. Digel, D. Thompson, W. Atwood
2. First Fermi Catalog of Gamma Ray Sources and the Fermi Pulsar Catalog
Thanks to L. Guillemot, S. Digel, P. Ray
3. First Fermi AGN Catalog
Thanks to B. Lott, P. Giommi
4. Relativistic jet physics and blazars
Thanks to C.C. Cheung, J. Finke
5. γ rays from cosmic rays in the Galaxy
Thanks to S. Funk, A. Strong, I. Moskalenko, N. Giglietto, W. Atwood, S. Digel
6. γ rays from star-forming galaxies and clusters of galaxies, and the diffuse extragalactic γ-ray
background
Thanks to K. Bechtol, J. Knödlseder, P. Martin, M. Ackermann
7. Microquasars, radio galaxies, and the extragalactic background light
Thanks to R. Dubois, C.C. Cheung, J. Finke, S. Razzaque
8. Fermi Observations of Gamma Ray Bursts
Thanks to A. von Kienlin, V. Connaughton, K. Hurley, C. Fryer, C. Kouveliotou, N. Omodei,

S. Razzaque
9. Fermi acceleration, ultra-high energy cosmic rays, and Fermi at 2
Thanks to A. Levinson, M. Barkov, G. Menon
I would especially like to thank Paola Grandi, Benoit Lott, David Paneque, Stefan Funk, Marco
Ajello, and Aous Abdo and Damien Parent for providing, and in some cases modifying from the
original, Figs. 7, 8 and 9, 10, 18, 21, and 22, respectively. I would also like to thank Justin Finke for
the use of Fig. 23, Soeb Razzaque for the use of Fig. 29, and Jean-Marc Casandjian for correspon-
dence. Detailed comments by Tyrel Johnson and Igor Moskalenko are gratefully acknowledged. In
addition, I would like to thank Marco Ajello, Armen Atoyan, Kohta Murase, and Markus Böttcher
348 C. D. Dermer
for collaboration, and Roland Walter and Marc Türler for the opportunity to lecture in the Saas-Fee
course.
This work was supported by the Office of Naval Research and NASA.
References
1. Abdo et al. 2010, Science, 329, 817

2. Cheung, C. C., Donato, D., Wallace, E., Corbet, R., Dubus, G., Sokolovsky, K., & Takahashi,
H. 2010, The Astronomer’s Telegram, 2487, 1
3. CTA Consortium, Actis, M., et al. 2011, Experimental Astronomy, 32, 193 (arXiv:1008.3703)
4. Dermer, C. D., & Menon, G. 2009, High Energy Radiation from Black Holes: Gamma Rays,
Cosmic Rays, and Neutrinos (Princeton University Press, Princeton)
5. Sandoval, A., et al. 2009, arXiv:0912.3329
6. The Pierre AUGER Collaboration, Abreu, P., Aglietta, M., et al. 2010, Astroparticle Physics,
34, 314
Lecture 1: GeV Instrumentation and the GeV Sky with the Fermi Gamma-ray Space
Telescope
7. Abdo, A. A., et al. 2010, Phys. Rev. Lett., 104, 101101
8. Abdo, A. A., Ackermann, M., Ajello, M., et al. 2009, Astroparticle Physics, 32, 193
9. Abdo, A. A., et al. 2011, ApJ, 733, L26
10. Atwood, W. B., et al. 2009, ApJ, 697, 1071
11. Casandjian, J.-M., & Grenier, I. A. 2008, A&A, 489, 849
12. Dermer, C. D. 2007, ApJ, 659, 958
13. Elliot, J. L., & Shapiro, S. L. 1974, ApJ, 192, L3
14. Feroci, M. et al. 2010, A&A, 510, A9
15. Fichtel, C. E., Kniffen, D. A., & Hartman, R. C. 1973, ApJ, 186, L99
16. Fichtel, C. E., Hartman, R. C., Kniffen, D. A., Thompson, D. J., Ogelman, H., Ozel, M. E.,
Tumer, T., & Bignami, G. F. 1975, ApJ, 198, 163
17. Hartman, R. C., et al. 1999, ApJS, 123, 79
18. Hermsen, W., et al. 1977, Nature, 269, 494
19. Kraushaar, W. L., Clark, G. W., Garmire, G. P., Borken, R., Higbie, P., Leong, V., & Thorsos,
T. 1972, ApJ, 177, 341
20. Kurfess, J. D., Bertsch, D. L., Fishman, G. J., & Schönfelder, V. 1997, Proceedings of the
Fourth Compton Symposium, 410, 509
21. Mattox, J. R., et al. 1996, ApJ, 461, 396
22. Michelson, P. F., Atwood, W. B., & Ritz, S. 2010, Reports on Progress in Physics, 73, 074901
23. Misner, C. W., Thorne, K. S., & Wheeler, J. A. 1973, Gravitation (San Francisco: W.H.
Freeman and Co.)
24. Pelassa, V., Preece, R., Piron, F., Omodei, N., and Guiriec, S. 2010, arXiv:1002.2617
25. Pittori, C., et al. 2009, A&A, 506, 1563
26. Sreekumar, P., et al. 1998, ApJ, 494, 523
27. Stecker, F. W., Hunter, S. D., & Kniffen, D. A. 2008, Astroparticle Physics, 29, 25
28. Strong, A. W., Moskalenko, I. V., & Reimer, O. 2004, ApJ, 613, 956
29. Swanenburg, B. N., et al. 1981, ApJ, 243, L69
30. Thompson, T. A., & Lacki, B. C. 2011, arXiv:1103.1886
31. Thompson, D. J. 1986, Nucl. Instrum. Methods Phys. Res. A, 251, 390
32. Thompson, D. J., et al. 1993, ApJS, 86, 629
33. Thompson, D. J., & Fichtel, C. E. 1982, A&A, 109, 352
Lecture 2: Fermi Gamma-ray Source Catalogs and Fermi Pulsars

34. Abdo, A. A., et al. 2008, Science, 322, 1218
35. Abdo, A. A., et al. 2009, ApJS, 183, 46
36. Abdo, A. A., et al. 2009, ApJ, 700, 597
40. Abdo, A. A., et al. 2009, ApJ, 699, 1171
41. Abdo, A. A., Ackermann, M., Atwood, W. B., et al. 2009, ApJ, 696, 1084
43. Abdo, A. A., et al. 2010, ApJS, 188, 405
44. Abdo, A. A., et al. 2010, ApJS, 187, 460; (e) 2011, 193, 22A
45. Abdo, A. A., et al. 2010, ApJ, 715, 429
47. Abdo, A. A., et al. 2010, A&A, 524, A75
48. Abdo, A. A., et al. 2010, ApJ, 708, 1254
49. Abdo, A. A., Ackermann, M., Ajello, M., et al. 2010, ApJ, 723, 649
51. Abdo, A. A., Ajello, M., Antolini, E., et al. 2010, ApJ, 720, 26
53. Ackermann, M., et al. 2011, ApJ, 726, 35
54. Atoyan, A. M., & Aharonian, F. A. 1996, MNRAS, 278, 525
55. Atwood, W. B., Ziegler, M., Johnson, R. P., & Baughman, B. M. 2006, ApJ, 652, L49
56. Baring, M. G. 2001, astro-ph/0106161
57. Caraveo, P. A., Bignami, G. F., Mignani, R., & Taff, L. G. 1996, ApJ, 461, L91
58. Demorest, P. B., Pennucci, T., Ransom, S. M., Roberts, M. S. E., & Hessels, J. W. T. 2010,
Nature, 467, 1081
59. de Jager, O. C., Harding, A. K., Michelson, P. F., Nel, H. I., Nolan, P. L., Sreekumar, P., &
Thompson, D. J. 1996, ApJ, 457, 253
60. Dyks, J., Harding, A. K., & Rudak, B. 2004, ApJ, 606, 1125
61. Fermi-LAT Collaboration, Freiere, P. C. C., et al. 2011, Science, 334, 1107
62. Goldreich, P., & Julian, W. H. 1969, ApJ, 157, 869
63. Gould, R. J. 1965, Physical Review Letters, 15, 577
64. Halpern, J. P., & Holt, S. S. 1992, Nature, 357, 222
65. Harding, A. K., Strickman, M. S., Gwinn, C., et al. 2002, ApJ, 576, 376
66. Komissarov, S. S., & Lyutikov, M. 2011, MNRAS, 414, 2017
67. Kuiper, L., Hermsen, W., Verbunt, F., et al. 2000, A&A, 359, 615
68. Kulkarni, S. R., Clifton, T. C., Backer, D. C., Foster, R. S., & Fruchter, A. S. 1988, Nature,
331, 50
69. Lyutikov, M., Balsara, D., & Matthews, C. 2012, MNRAS, 422, 3118
70. Nolan, P. L., Abdo, A. A., Ackermann, M., et al. 2012, ApJS, 199, 31
71. Ramanamurthy, P. V., Bertsch, D. L., Dingus, B. L., et al. 1995, ApJ, 447, L109
72. Ray, P. S., Kerr, M., Parent, D., et al. 2011, ApJS, 194, 17
73. Ray, P. S., & Parkinson, P. M. S. 2011, High-Energy Emission from Pulsars and their Systems,
37 (arXiv:1007.2183)
74. Romani, R. W. 2011, American Institute of Physics Conference Series, 1357, 205
75. Romani, R. W., Kargaltsev, O., & Pavlov, G. G. 2005, ApJ, 627, 383
76. Romani, R. W., & Watters, K. P. 2010, ApJ, 714, 810
77. Shapiro, S. L., & Teukolsky, S. A. 1983, Black Holes, White Dwarfs, and Neutron Stars: The
Physics of Compact Objects (New York, Wiley-Interscience)
78. Tavani, M., Striani, E., Bulgarelli, A., et al. 2010, The Astronomer’s Telegram, 2855, 1
79. Tavani, M., Bulgarelli, A., Vittorini, V., et al. 2011, Science, 331, 736
80. VERITAS Collaboration, Aliu, E., Arlen, T., et al. 2011, Science, 334, 69
350 C. D. Dermer
81. Weinberg, S. 1972, Gravitation and Cosmology: Principles and Applications of the General
Theory of Relativity (Wiley: New York)
82. Wilson-Hodge, C. A., Cherry, M. L., Case, G. L., et al. 2011, ApJ, 727, L40
83. Yuan, Q., Yin, P.-F., Wu, X.-F., et al. 2011, ApJ, 730, L15
Lecture 3: Fermi AGN Catalogs
84. Abdo, A. A., et al. 2009, ApJ, 707, L142
86. Abdo, A. A., et al. 2010, ApJ, 716, 30
87. Abdo, A. A., et al. 2010, ApJ, 721, 1425
88. Abdo, A. A., et al. 2010d, ApJ, 710, 1271
89. Abdo, A. A., et al. 2010e, ApJ, 720, 912
90. Abdo, A. A., et al. 2011, ApJ, 727, 129
92. Ackermann, M., Ajello, M., Allafort, A., et al. 2011, ApJ, 743, 171; arXiv:1108.1420
93. Ackermann, M., Ajello, M., Allafort, A., et al. 2011, ApJ, 741, 30
95. Aharonian, F., et al. 2007, ApJ, 664, L71
96. Aharonian, F., et al. 2009, ApJ, 696, L150
97. Aleksić, J., Antonelli, L. A., Antoranz, P., et al. 2010, ApJ, 723, L207
98. Bonnoli, G., Ghisellini, G., Foschini, L., Tavecchio, F., & Ghirlanda, G. 2010, MNRAS, 410,
368
99. Ghisellini, G., Maraschi, L., & Dondi, L. 1996, A&A Suppl., 120, 503
100. Ghisellini, G., Maraschi, L., & Tavecchio, F. 2009, MNRAS, 396, L105
101. Fanaroff, B. L., & Riley, J. M. 1974, MNRAS, 167, 31P
102. Finke, J. D., Dermer, C. D., Böttcher, M. 2008, ApJ, 686, 181
103. Finke, J. D., & Dermer, C. D. 2010, ApJ, 714, L303
104. Healey, S. E., Romani, R. W., Taylor, G. B., Sadler, E. M., Ricci, R., Murphy, T., Ulvestad, J.
S., & Winn, J. N. 2007, ApJS, 171, 61
105. Kirk, J. G., Rieger, F. M., & Mastichiadis, A. 1998, A&A, 333, 452
106. Lenain, J.-P., Ricci, C., Türler, M., Dorner, D., & Walter, R. 2010, A&A, 524, A72
107. Lister, M. L., & Homan, D. C. 2005, AJ, 130, 1389
108. Mattox, J. R., Wagner, S. J., Malkan, M., McGlynn, T. A., Schachter, J. F., Grove, J. E.,
Johnson, W. N., & Kurfess, J. D. 1997, ApJ, 476, 692
109. Massaro, E., Giommi, P., Leto, C., Marchegiani, P., Maselli, A., Perri, M., Piranomonte, S.,
& Sclavi, S. 2009, A&A, 495, 691
110. Neronov, A., Semikoz, D., & Vovk, I. 2010, A&A, 519, L6
111. Pogge, R. W. 2000, New Astronomy Reviews, 44, 381
112. Poutanen, J., & Stern, B. 2010, ApJ, 717, L118
113. Reimer, A. 2007, ApJ, 665, 1023
114. Sanders, D. B., Soifer, B. T., Elias, J. H., et al. 1988, ApJ, 325, 74
115. Striani, E., et al. 2010, ApJ, 718, 455
116. Tavecchio, F., Maraschi, L., & Ghisellini, G. 1998, ApJ, 509, 608
117. Urry, C. M., & Padovani, P. 1995, PASP, 107, 803
118. Vestrand, W. T., Stacy, J. G., & Sreekumar, P. 1995, ApJ, 454, L93
Lecture 4: Relativistic Jet Physics
119. Abdo, A. A., et al. 2009, ApJ, 699, 817
120. Abdo, A., et al., 2009, Science, 323, 1688
121. Abdo, A. A., et al. 2010, Nature, 463, 919
124. Ackermann, M. et al., ApJ, Ackermann, M., et al. 2010, ApJ, 716, 1178
126. Aharonian, F. A., Coppi, P. S., & Völk, H. J. 1994, ApJ, 423, L5
127. Aharonian, F. A. 2000, New Astronomy, 5, 377

128. Aleksić, J., Antonelli, L. A., Antoranz, P., et al. 2011, ApJ, 730, L8
129. Ando, S., & Kusenko, A. 2010, ApJ, 722, L39
130. Asano, K., Guiriec, S., & Mészáros, P. 2009, ApJ, 705, L191
131. Atoyan, A. M., & Dermer, C. D. 2003, ApJ, 586, 79
132. Berezinsky, V., Gazizov, A. Z., & Grigorieva, S. I. 2005, Physics Letters B, 612, 147
133. Berezinsky, V. 2008, Advances in Space Research, 41, 2071
134. Blumenthal, G. R., & Gould, R. J. 1970, Reviews of Modern Physics, 42, 237
135. Bromberg, O., & Levinson, A. 2009, ApJ, 699, 1274
136. Celotti, A., & Fabian, A. C. 1993, MNRAS, 264, 228
137. Cenko, S. B., et al. 2010, ApJ, 711, 641
138. D’Avezac, P., Dubus, G., & Giebels, B. 2007, A&A, 469, 857
139. Dermer, C. D., & Mitman, K. E. 1999, ApJ, 513, L5
140. Dermer, C. D. 2007, ApJ, 664, 384
141. Dermer, C. D., Ramirez-Ruiz, E., & Le, T. 2007, ApJ, 664, L67
142. Dermer, C.D., Cavadini, M., Razzaque, S., Finke, J.D., Chiang, J. & Lott, B., 2011 ApJ, 733,
L21
143. Dolag, K., Kachelriess, M., Ostapchenko, S., & Tomàs, R. 2011, ApJ, 727, L4
144. Elyiv, A., Neronov, A., & Semikoz, D. V. 2009, Phys. Rev. D, 80, 023010
145. Finke, J. D., Razzaque, S., & Dermer, C. D. 2010, ApJ, 712, 238
146. Gaidos, J. A., et al. 1996, Nature, 383, 319
147. Georganopoulos, M., Kirk, J. G., & Mastichiadis, A. 2001, ApJ, 561, 111; (e) 2004, ApJ, 604,
479
148. Ghisellini, G., Tavecchio, F., Foschini, L., et al. 2010, MNRAS, 402, 497
149. Granot, J., Cohen-Tanugi, J., & do Couto e Silva, E. 2008, ApJ, 677, 92
150. Gu, M., Cao, X., & Jiang, D. R. 2001, MNRAS, 327, 1111
151. Guetta, D., Piran, T., & Waxman, E. 2005, ApJ, 619, 412
152. Hillas, A. M. 1984, Ann. Rev. of Astronomy and Astrophysics, 22, 425
153. Ichiki, K., Inoue, S., & Takahashi, K. 2008, ApJ, 682, 127
154. Jones, F. C. 1968, Physical Review, 167, 1159
155. Jorstad, S. G., et al. 2005, AJ, 130, 1418
156. Kneiske, T. M., & Dole, H. 2010, A&A, 515, A19
157. Kotera, K., Allard, D., & Lemoine, M. 2011, A&A, 527, A54
158. Le, T., Dermer, C. D. 2007, ApJ, 661, 394
159. Li, Z. 2010, ApJ, 709, 525
160. Massaro, F., Tramacere, A., Cavaliere, A., Perri, M., & Giommi, P. 2008, A&A, 478, 395
161. Mücke, A., & Protheroe, R. J. 2001, Astroparticle Physics, 15, 121
162. Murase, K., & Nagataki, S. 2006, Phys. Rev. D, 73, 063002
163. Murase, K., Zhang, B., Takahashi, K., & Nagataki, S. 2009, MNRAS, 396, 1825
164. Murase, K., Takahashi, K., Inoue, S., Ichiki, K., & Nagataki, S. 2008, ApJ, 686, L67
165. Neronov, A., & Semikoz, D. V. 2009, Phys. Rev. D, 80, 123012
166. Neronov, A., & Vovk, I. 2010, Science, 328, 73
167. Neronov, A., Semikoz, D. V., Tinyakov, P. G., & Tkachev, I. I. 2011, A&A, 526, A90
168. Plaga, R. 1995, Nature, 374, 430
169. Razzaque, S., Mészáros, P., & Zhang, B. 2004, ApJ, 613, 1072
170. Razzaque, S. 2010, ApJ, 724, L109
171. Razzaque, S., Dermer, C. D., & Finke, J. D. 2010, Open, Astronomy, 3, 150 (arXiv:0908.0513)
172. Sikora, M., Begelman, M. C., & Rees, M. J. 1994, ApJ, 421, 153
173. Soderberg, A. M., et al. 2006, Nature, 442, 1014
174. Takahashi, K., Murase, K., Ichiki, K., Inoue, S., & Nagataki, S. 2008, ApJ, 687, L5
175. Tanaka, Y. T., Stawarz, Ł., Thompson, D. J., et al. 2011, ApJ, 733, 19
176. Tavecchio, F., Ghisellini, G., Bonnoli, G., & Ghirlanda, G. 2010, MNRAS, 405, L94
177. Tavecchio, F., Ghisellini, G., Foschini, L., Bonnoli, G., Ghirlanda, G., & Coppi, P. 2010,
MNRAS, 406, L70
352 C. D. Dermer
178. Tavecchio, F., Ghisellini, G., Bonnoli, G., & Foschini, L. 2011, MNRAS, 414, 3566
179. Wang, X.-Y., Li, Z., Dai, Z.-G., & Mészáros, P. 2009, ApJ, 698, L98
Lecture 5: γ Rays from Cosmic Rays in the Galaxy
180. Abdo, A. A., Ackermann, M., Ajello, M., et al. 2009, Phys. Rev. D, 80, 122004
181. Abdo, A. A., Ackermann, M., Ajello, M., et al. 2009, ApJ, 706, L1
182. Abdo, A., et al. 2009 ApJ, 703, 1249
186. Abdo, A. A., Ackermann, M., Ajello, M., et al. 2010, Science, 327, 1103
191. Achterberg, A. 1979, A&A, 76, 276
192. Ackermann, M., et al. 2010, Phys. Rev. D, 82, 092004
194. Ackermann, M., Ajello, M., Allafort, A., et al. 2011, Science, 334, 1103
196. Ackermann, M., Ajello, M., Allafort, A., et al. 2012, Physical Review Letters, 108, 011103
197. Adriani, O., et al. 2009 Nature, 458, 607
198. Aharonian, F. A., & Atoyan, A. M. 1999, A&A, 351, 330
199. Aharonian, F. A., Akhperjanian, A. G., Aye, K.-M., et al. 2004, Nature, 432, 75
200. Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et al. 2006, A&A, 449, 223
201. Aharonian, F., et al. 2008, Phys. Rev. Lett., 101, 261104
202. Atoyan, A., & Dermer, C. D. 2012, ApJ, 749, L26
203. Bell, A. R. 1978, MNRAS, 182, 147
204. Bell, A. R. 1978, MNRAS, 182, 443
205. Berezinskii, V. S., Bulanov, S. V., Dogiel, V. A., & Ptuskin, V. S. 1990, Astrophysics of Cosmic
Rays (Amsterdam: North-Holland)
206. Blandford, R. D., and Eichler, D. 1987, Physics Reports, 154, 1
207. Bouchet, L., Strong, A. W., Porter, T. A., Moskalenko, I. V., Jourdain, E., & Roques, J. 2011,
ApJ, 739, 29
208. Chang, J., et al. 2008, Nature, 456, 362
209. Cordes, J. M., & Lazio, T. J. W. 2002, arXiv:astro-ph/0207156
210. Crocker, R. M., & Aharonian, F. 2011, Phys. Rev. Lett., 106, 101102
211. Dermer, C. D. 1986, A&A, 157, 223
212. Dobler, G., Finkbeiner, D. P., Cholis, I., Slatyer, T., & Weiner, N. 2010, ApJ, 717, 825
213. Drury, L. O’C. 1983, Reports on Progress in Physics, 46, 973
214. Esposito, J. A., Hunter, S. D., Kanbach, G., & Sreekumar, P. 1996, ApJ, 461, 820
215. Finkbeiner, D. P. 2004, ApJ, 614, 186
216. Fishman, G. J., Bhat, P. N., Mallozzi, R., et al. 1994, Science, 264, 1313
217. Frail, D. A. 2011, arXiv:1108.4137
218. Funk, S. 2010, http://meetings.aps.org/Meeting/APR10/Event/116093, APS April Meeting
Abstracts, 3002
219. Gaisser, T. K. 1990, Cosmic Rays and Particle Physics (Cambridge: Cambridge University
Press)
220. Giglietto, N., for the Fermi Large Area Telescope Collaboration 2009, arXiv:0912.3734
221. Ginzburg, V. L., & Syrovatskii, S. I. 1964, The Origin of Cosmic Rays (New York: Macmillan)
222. Giordano, F., Naumann-Godo, M., Ballet, J., et al. 2011, 2012, ApJ, 744, L2
223. Green, D. A., & Stephenson, F. R. 2003, Supernovae and Gamma-Ray Bursters, 598, 7
224. Grenier, I. A., Casandjian, J.-M., & Terrier, R. 2005, Science, 307, 1292
225. Hayakawa, S. 1969, Cosmic ray physics. Nuclear and astrophysical aspects, Interscience
Monographs and Texts in Physics and Astronomy, (New York: Wiley-Interscience)
226. Hewitt, J. W., Yusef-Zadeh, F., & Wardle, M. 2009, ApJ, 706, L270
227. Hunter, S. D., et al. 1997, ApJ, 481, 205
228. Katagiri, H., Tibaldo, L., Ballet, J., et al. 2011, ApJ, 741, 44
229. Lagage, P. O., & Cesarsky, C. J. 1983, A&A, 118, 223
230. Longair, M. S. 1994, High Energy Astrophysics (Cambridge: Cambridge University Press),
2nd ed.
231. Lopez, L. A., Ramirez-Ruiz, E., Badenes, C., et al. 2009, ApJ, 706, L106
232. Lopez, L. A., Ramirez-Ruiz, E., Huppenkothen, D., Badenes, C., & Pooley, D. A. 2011, ApJ,
732, 114
233. Lebrun, F., Bennett, K., Bignami, G. F., et al. 1983, ApJ, 274, 231; (e) 1984, ApJ, 282, 359
234. Mertsch, P., & Sarkar, S. 2010, JCAP, 10, 19
235. Mertsch, P., & Sarkar, S. 2011, Phys. Rev. Lett., 107, 091101
236. Morlino, G., & Caprioli, D. 2012, A&A, 538, A81
237. Moskalenko, I. V., Porter, T. A., & Digel, S. W. 2006, ApJ, 652, L65; (e) 2007 ApJ, 664, L143
238. Moskalenko, I. V., Porter, T. A., Digel, S. W., Michelson, P. F., & Ormes, J. F. 2008, ApJ,
681, 1708
239. Moskalenko, I. V., & Porter, T. A. 2009, ApJ, 692, L54
240. Orlando, E., & Strong, A. W. 2007, Astrophysics & Space Science, 309, 359 (astro-ph/
0607563)
241. Orlando, E., & Strong, A. W. 2008, A&A, 480, 847
242. Porter, T. A., Moskalenko, I. V., Strong, A. W., Orlando, E., & Bouchet, L. 2008, ApJ, 682,
400
243. Porter, T. A., Johnson, R. P., & Graham, P. W. 2011, ARA&A, 49, 155
244. Reynolds, S. P. 1988, IAU Colloq. 101: Supernova Remnants and the Interstellar Medium,
331
245. Schlickeiser, R. 1984, A&A, 136, 227
246. Schlickeiser, R. 1985, A&A, 143, 431
247. Seckel, D., Stanev, T., & Gaisser, T. K. 1991, ApJ, 382, 652
248. Strong, A. W., & Moskalenko, I. V. 1998, ApJ, 509, 212
251. Strong, A. W., Moskalenko, I. V., & Ptuskin, V. S. 2007, Ann. Rev. of Nuclear and Particle
Science, 57, 285
252. Strong, A. W., Porter, T. A., Digel, S. W., Jóhannesson, G., Martin, P., Moskalenko, I. V.,
Murphy, E. J., & Orlando, E. 2010, ApJ, 722, L58
253. Sturner, S. J., & Dermer, C. D. 1995, A&A, 293, L17
254. Su, M., Slatyer, T. R., & Finkbeiner, D. P. 2010, ApJ, 724, 1044
255. Tanaka, T., Uchiyama, Y., Aharonian, F. A., et al. 2008, ApJ, 685, 988
256. Thompson, D. J., Bertsch, D. L., Morris, D. J., & Mukherjee, R. 1997, JGR, 1021, 14735
257. Torres, D. F., Romero, G. E., Dame, T. M., Combi, J. A., & Butt, Y. M. 2003, Phys. Reports,
382, 303
258. Uchiyama, Y., 2011, arXiv:1104.1197
259. Yüksel, H., Kistler, M. D., & Stanev, T. 2009, Phys. Rev. Lett., 103, 051101
Lecture 6: γ Rays from Star-Forming Galaxies and Clusters of Galaxies, and the Diffuse
γ-Ray Background
260. Abdo, A. A., et al. 2010, A&A, 512, A7
261. Abdo, A. A., et al. 2010, A&A, 523, A46
262. Abdo, A. A., et al. 2010, ApJ, 709, L152
264. Abdo, A. A., et al. 2009, ApJ, 699, 31
265. Abdo, A. A., et al. 2010, A&A, 523, L2
266. Acciari, V. A., et al. 2009, Nature, 462, 770
267. Acero, F., et al. 2009, Science, 326, 1080
354 C. D. Dermer
268. Ackermann, M., Ajello, M., Allafort, A., et al. 2010, ApJ, 717, L71
269. Ajello, M., Shaw, M. S., Romani, R. W., et al. 2012, ApJ, 751, 108
270. Berrington, R. C., & Dermer, C. D. 2003, ApJ, 594, 709
271. Berezinsky, V. S., Blasi, P., & Ptuskin, V. S. 1997, ApJ, 487, 529
272. Brunetti, G., Blasi, P., Cassano, R., & Gabici, S. 2004, MNRAS, 350, 1174
273. Brunetti, G., & Blasi, P. 2005, MNRAS, 363, 1173
274. Brunetti, G., & Lazarian, A. 2011, MNRAS, 410, 127
275. Brunetti, G. 2011, Journal of Astrophysics and Astronomy, 32, 437
276. Fields, B. D., Pavlidou, V., & Prodanović, T. 2010, ApJ, 722, L199
277. Gabici, S., & Blasi, P. 2003, Astroparticle Physics, 19, 679
278. Gendre, M. A., Best, P. N., & Wall, J. V. 2010, MNRAS, 404, 1719
279. Goodenough, L., & Hooper, D. 2009, arXiv:0910.2998
280. Inoue, Y., & Totani, T. 2009, ApJ, 702, 523
281. Kushnir, D., & Waxman, E. 2009, JCAP, 8, 2
282. Loeb, A., & Waxman, E. 2000, Nature, 405, 156
283. Miniati, F., Ryu, D., Kang, H., et al. 2000, ApJ, 542, 608
284. Pavlidou, V., & Fields, B. D. 2001, ApJ, 558, 63
285. Stecker, F. W., & Venters, T. M. 2011, ApJ, 736, 40
286. Tammann, G. A., Loeffler, W., & Schroeder, A. 1994, ApJS, 92, 487
287. Torres, D. F. 2004, ApJ, 617, 966
Lecture 7: Microquasars, Radio Galaxies, and the EBL
288. Abdo, A. A., et al. 2009, ApJ, 707, 55
289. Abdo, A., et al. 2009, Science, 326, 1512
290. Abdo, A. A., et al. 2009, ApJ, 706, L56
291. Abdo, A. A., et al. 2009, ApJ, 701, L123
292. Abdo, A. A., et al. 2010, ApJ, 719, 1433
295. Aharonian, F., et al. 2005, A&A, 442, 1
296. Böttcher, M., Dermer, C. D., & Finke, J. D. 2008, ApJ, 679, L9
297. Bosch-Ramon, V., Romero, G. E., & Paredes, J. M. 2006, A&A, 447, 263
298. Chen, A., Reyes, L. C., & Ritz, S. 2004, ApJ, 608, 686
299. Chiaberge, M., Celotti, A., Capetti, A., & Ghisellini, G. 2000, A&A, 358, 104
300. Corbet, R.H.D., et al. 2011, The Astronomer’s Telegram #: 3221
301. Dermer, C. D. 1995, ApJ, 446, L63
302. Dermer, C. D., & Böttcher, M. 2006, ApJ, 643, 1081
303. Dermer, C. D., & Razzaque, S. 2010, ApJ, 724, 1366
304. Domínguez, A., Primack, J. R., Rosario, D. J., et al. 2011, MNRAS, 410, 2556
305. Dubus, G. 2006, A&A, 456, 801
306. Essey, W., & Kusenko, A. 2010, Astroparticle Physics, 33, 81
307. Essey, W., Kalashev, O. E., Kusenko, A., & Beacom, J. F. 2010, Phys. Rev. Lett., 104, 141102
308. Essey, W., Kalashev, O., Kusenko, A., & Beacom, J. F. 2011, ApJ, 731, 51
309. Franceschini, A., Rodighiero, G., & Vaccari, M. 2008, A&A, 487, 837
310. Finke, J. D., & Razzaque, S. 2009, ApJ, 698, 1761
311. Georganopoulos, M., Kirk, J. G., & Mastichiadis, A. 2001, ApJ, 561, 111
312. Georganopoulos, M., & Kazanas, D. 2003, ApJ, 594, L27
313. Georganopoulos, M., Sambruna, R. M., Kazanas, D., Cillis, A. N., Cheung, C. C., Perlman,
E. S., Blundell, K. M., & Davis, D. S., 2008, ApJ, 686, L5
314. Georganopoulos, M., Finke, J. D., & Reyes, L. C. 2010, ApJ, 714, L157
315. Ghisellini, G., Tavecchio, F., & Chiaberge, M. 2005, A&A, 432, 401
316. Gilmore, R. C., Madau, P., Primack, J. R., Somerville, R. S., & Haardt, F. 2009, MNRAS,
399, 1694
317. Hill, A. B., Szostek, A., Corbel, S., et al. 2011, MNRAS, 415, 235
318. Johnston, S., Manchester, R. N., Lyne, A. G., D’Amico, N., Bailes, M., Gaensler, B. M., &
Nicastro, L. 1996, MNRAS, 279, 1026
319. Kneiske, T. M., Mannheim, K., & Hartmann, D. H. 2002, A&A, 386, 1
320. Kneiske, T. M., Bretz, T., Mannheim, K., & Hartmann, D. H. 2004, A&A, 413, 807
321. Kotera, K.K., Allard, D., & Lemoine, M. 2011, A&A, 527, A54
322. Lewin, W., van Paradijs, J., & van den Heuvel, E. P. J., eds., 1995, X-ray Binaries (Cambridge
Univ. Press, Cambridge)
323. Maraschi, L., & Treves, A. 1981, MNRAS, 194, 1P
324. Mazin, D., & Raue, M. 2007, A&A, 471, 439
325. Migliori, G., Grandi, P., Torresi, E., et al. 2011, A&A, 533, A72
326. Mirabel, I. F., Rodriguez, L. F., Cordier, B., Paul, J., & Lebrun, F. 1992, Nature, 358, 215
327. Mirabel, I. F., & Rodríguez, L. F. 1994, Nature, 371, 46
328. Mirabel, I. F., & Rodríguez, L. F. 1999, ARA&A, 37, 409
329. Murase, K., Dermer, C. D., Takami, H., & Migliori, G. 2012, ApJ, 749, 63
330. Orellana, M., & Romero, G. E. 2007, Astrophysics and Space Science, 309, 333
331. Paredes, J. M. 2005, Chinese Journal of Astronomy and Astrophysics Supplement, 5, 121
332. Paredes, J. M., Martí, J., Ribó, M., & Massi, M. 2000, Science, 288, 2340
333. Pavlov, G. G., Misanovic, Z., Kargaltsev, O., & Garmire, G. P. 2011, The Astronomer’s
Telegram, 3228, 1
334. Primack, J. R., Bullock, J. S., & Somerville, R. S. 2005, High Energy Gamma-Ray Astronomy,
745, 23
335. Razzaque, S., Dermer, C. D., & Finke, J. D. 2012, ApJ, 745, 196
336. Salamon, M. H., & Stecker, F. W. 1998, ApJ, 493, 547
337. Schroedter, M. 2005, ApJ, 628, 617
338. Stecker, F. W., Malkan, M. A., & Scully, S. T. 2006, ApJ, 648, 774
339. Stecker, F. W., Baring, M. G., & Summerlin, E. J. 2007, ApJ, 667, L29
340. Tavani, M., et al. 1996, Astronomy & Astrophysics Supp., 120, 221
341. Tavani, M., et al. 2009, Nature, 462, 620
Lecture 8: Fermi Observations of Gamma Ray Bursts
342. Abdo, A. A., et al. 2009, ApJ, 706, L138
343. Abdo, A. A., et al. 2009, Nature, 462, 331
344. Abdo, A. A., et al. 2009, ApJ, 707, 580
345. Abdo, A. A., et al. 2010, ApJ, 712, 558
346. Ackermann, M., et al. 2010, ApJ, 717, L127
347. Band, D. L. 2002, ApJ, 578, 806
348. Bregeon, J., McEnery, J., & Ohno, M. 2011, GRB Coordinates Network, 12218, 1
349. Briggs, M. S., et al. 1999, ApJ, 524, 82
350. Chakraborti, S., Ray, A., Soderberg, A. M., Loeb, A., & Chandra, P. 2011, Nature Communi-
cations, 2
351. De Pasquale, M., et al. 2010, ApJ, 709, L146
352. Dingus, B. L. 1995, Astrophys. Space Sci., 231, 187
353. Eichler, D., Guetta, D., & Pohl, M. 2010, ApJ, 722, 543
354. Ghisellini, G., Ghirlanda, G., Nava, L., & Celotti, A. 2010, MNRAS, 403, 926
355. González, M. M., Dingus, B. L., Kaneko, Y., Preece, R. D., Dermer, C. D., & Briggs, M. S.
2003, Nature, 424, 749
356. Guetta, D., Pian, E., & Waxman, E. 2011, A&A, 525, A53
357. Hopkins, A. M., & Beacom, J. F. 2006, ApJ, 651, 142
358. Hurley, K., et al. 1994, Nature, 372, 652
359. Kumar, P., & Barniol Duran, R. 2009, MNRAS, 400, L75
360. Le, T., & Dermer, C. D. 2009, ApJ, 700, 1026
361. Liang, E., Zhang, B., Virgili, F., & Dai, Z. G. 2007, ApJ, 662, 1111
362. Murase, K., Ioka, K., Nagataki, S., & Nakamura, T. 2006, ApJ, 651, L5
363. Murase, K., & Takami, H. 2009, ApJ, 690, L14
356 C. D. Dermer
364. Oates, S. R., Beardmore, A. P., Holland, S. T., et al. 2011, GRB Coordinates Network,
12215, 1
365. Porciani, C., & Madau, P. 2001, ApJ, 548, 522
366. Sari, R., Piran, T., & Narayan, R. 1998, ApJ, 497, L17
367. Sanders, D. B. 2004, Advances in Space Research, 34, 535
368. Schmidt, M. 2001, ApJ, 559, L79
369. Tierney, D., & von Kienlin, A. 2011, GRB Coordinates Network, 12187, 1
370. Toma, K., Wu, X.-F., & Mészáros, P. 2011, MNRAS, 415, 1663
371. Vasileiou, V., Piron, F., Tierney, D., et al. 2011, GRB Coordinates Network, 12188, 1
372. Wanderman, D., & Piran, T. 2010, MNRAS, 406, 1944
373. Wang, X.-Y., Razzaque, S., & Mészáros, P. 2008, ApJ, 677, 432
374. Wick, S. D., Dermer, C. D., & Atoyan, A. 2004, Astroparticle Physics, 21, 125
375. Yüksel, H., Kistler, M. D., Beacom, J. F., & Hopkins, A. M. 2008, ApJ, 683, L5
Lecture 9: Fermi Acceleration and Ultra-High Energy Cosmic Rays
376. Arons, J. 2003, ApJ, 589, 871
377. Aschenbach, B. 2004, A&A, 425, 1075
378. Blandford, R. D., & Znajek, R. L. 1977, MNRAS, 179, 433
379. Bloom, J. S., Giannios, D., Metzger, B. D., et al. 2011, Science, 333, 203
380. Calvez, A., Kusenko, A., & Nagataki, S. 2010, Phys. Rev. Lett., 105, 091101
381. Dermer, C. D. 2007, arXiv:0711.2804
382. Dermer, C. D. 2011, American, Institute of Physics Conference Series, 1358, 355,
arXiv:1103.4834
383. Dermer, C., & Lott, B. 2012, Journal of Physics Conference Series, 355, 012010,
arXiv:1110.3739
384. Farrar, G. R., & Gruzinov, A. 2009, ApJ, 693, 329
385. Gair, J. R. 2009, Classical and Quantum Gravity, 26, 094034
386. Ghisellini, G., Ghirlanda, G., Tavecchio, F., Fraternali, F., & Pareschi, G. 2008, MNRAS,
390, L88
387. Jackson, J. D. 1962, Classical Electrodynamics (New York: Wiley)
388. Komissarov, S. S. 2004, MNRAS, 350, 427
389. Levinson, A. 2006, Trends in Black Hole Research, 119 (arXiv:astro-ph/0502346)
390. Levinson, A. 2006, International Journal of Modern Physics A, 21, 6015
391. Menon, G., & Dermer, C. D. 2005, ApJ, 635, 1197
392. Menon, G., & Dermer, C. D. 2007, General Relativity and Gravitation, 39, 785
393. Menon, G., & Dermer, C. D. 2009, Physical Review D, 79, 123005
394. Menon, G., & Dermer, C. D. 2011, MNRAS, 417, 1098
395. Murase, K., Inoue, S., & Nagataki, S. 2008, ApJ, 689, L105
396. Ryde, F., & Pe’er, A. 2009, ApJ, 702, 1211
397. Wald, R. M. General Relativity, 1984 (Chicago, University of Chicago Press)
398. Waxman, E., & Bahcall, J. 1999, Phys. Rev. D, 59, 023002
399. Waxman, E. 2004, New Journal of Physics, 6, 140
400. Znajek, R. L. 1977, MNRAS, 179, 457
Index
Greek Letters Big bang, 127, 128

p0decay, 8, 22, 32, 62 Blackbody, 266
c-ray Black hole, 4, 240, 314, 337
binary, 96, 110, 313 Kerr, 341
cross section, 148 physics, 341
detection, 185 primordial, 207, 210
lines, 4, 182 Schwarzschild, 240
3C 454.3, 240, 259, 266, 270, 346 thermodynamics, 209
Blandford–Znajek process, 111, 343, 346
Blazar, 7, 10, 21, 228, 231, 244, 254, 257, 259,
A 263, 278
Active galactic nuclei. See AGN high synchrotron-peaked (HSP), 257
ADAF, 70 low synchrotron-peaked (LSP), 257, 264
AGILE, 6, 228, 231, 316 misaligned, 317
AGN, 156, 200, 234, 244, 254–256, 258 BLR, 257, 264, 272
blazar. See Blazar Bohm approximation, 305
BL Lac object, 261 Boltzmann equation, 202
FSRQ, 259, 312, 345 Bose–Einstein statistics, 136
Altarelli Parisi evolution, 159 Branching ratio, 184
AMANDA, 167, 169 Bremsstrahlung, 4, 22, 28, 158, 184, 298, 303
AMS, 181 cross section, 29, 30
ANTARES, 187 internal, 183
Antimatter, 148, 189 Broad line region. See BLR
ASTRO-H, 63, 65, 109
Axino, 176
Axion, 195, 196 C
Cascade
electromagnetic, 24–26, 42, 48
B hadronic, 48
Baksan, 188 radiation, 168, 282, 285
BAO, 128, 134 Central molecular zone, 71
Baryon acoustic oscillations. See BAO Chandra, 19, 62, 229, 253, 300
Baryon–antibaryon asymmetry, 190 Chandrasekhar mass, 247
BATSE, 334 Cherenkov
Beaming factor, 239, 277 radiation, 11

Saas-Fee Advanced Course 40, DOI: 10.1007/978-3-642-36134-0,
Ó Springer-Verlag Berlin Heidelberg 2013
358 Index
C (cont.) EGB, 241, 311

telescope. See IACT EGRET, 5, 204, 228, 230–232, 241, 246, 298,
Cherenkov Telescope Array. See CTA 311
Cluster of galaxies. See Galaxy cluster Einstein equation, 129, 144, 212, 213
CMB, 49, 91, 128, 134, 148, 156, 190, 192, Electroweak interaction. See Interaction
241, 266, 272, 273, 283 Energy density, 130, 148
polarization, 216 Energy–momentum tensor, 130, 135, 139, 146
Coannihilation, 142, 184 Entropy, 135, 139, 142, 145
CoGeNT, 198 Equation of state, 130, 138, 144, 145
Compton Equipartition, 270, 280
cross section, 265, 267 Extensive air shower, 17
radiation, 156 Extragalactic c-ray background, See EGB
scattering, 68, 152, 156, 260, 264 Extragalactic background light. See EBL
COS-B, 230, 314
Cosmic microwave background. See CMB
Cosmic Ray, 4, 197, 227, 297, 299 F
Galactic, 50, 51, 64, 108, 109, 232, 290, Fermi
294, 296 acceleration, 276, 304, 337
spectrum, 241 bubbles, 73, 74, 298
ultra-high energy. See UHECR catalog, 243, 245, 254, 255, 262
Cosmological constant, 130, 132, 144, 273 GBM, 234
Coupling constant, 150 LAT, 8, 77, 234, 236, 237, 255, 262, 322
CP violation, 196 satellite, 5, 21, 180, 186, 190, 192, 194,
Crab Nebula, 77, 82, 87, 252, 308 197, 207, 227, 234, 235, 246, 249, 254,
Cross section, 26, 148, 152 258, 291
estimate, 149 Fermi-Dirac statistics, 136
CTA, 9, 20, 56, 65, 93, 199, 229 Feynman diagram, 150, 152, 154
Curvature radiation, 23 Field theory, 145
Cutoff power law, 61, 78, 249, 315 Fine structure constant, 149
Freeze-out, 141, 143
Friedmann equation, 135, 145
D
DAMA/LIBRA, 179, 198
Dark matter, 8, 125, 139, 144, 176, 196 G
KCDM, 201, 327 Galactic center, 14, 15, 66, 70, 71, 186
annihilation, 311 Galaxy cluster, 8, 19, 134, 157, 217, 310
asymmetric, 195 GAMMA-400, 183, 204
candidates, 193 Gamma-ray. See c-ray
cold, 167, 172, 173 Gamma-ray burst. See GRB
density, 176 Gauge choice, 212
hot, 141 Gaugino, 173
particle, 186 GEANT4, 237
Dimensional analysis, 152 GEMS, 230
DMA, 199 General relativity, 129, 212, 213
Doppler GEO600, 216
boosting, 254 Giant molecular cloud, 10, 14, 19
factor, 269 Glashow resonance, 162
GLAST, 235
Globular cluster, 244, 250
E Gluino, 174, 199
EBL, 8, 240, 262, 282, 313, 319 Gluon, 127, 159
Eddington luminosity, 240, 271, 345 Gould belt, 295
Index 359
Grand unification, 145 J

Gravitational Jet model
lensing, 134 hadronic, 272
wave, 212 leptonic, 268
Gravitino, 196 Jet physics, 263
GRB, 4, 10, 21, 156, 227, 231, 276–278, 322,
333, 339
luminosity function, 325 K
Kaluza-Klein, 195
Klein-Nishina
H effect, 60, 260
Hadron, 158 formula, 152, 157
Hadronic model, 55, 59, 156, 278 regime, 43, 45, 80, 87, 157, 264
Halo, 177, 181, 183, 185, 205, 281 KM3NET, 169, 187
HAWK, 17, 229
Hawking radiation, 209
HEAT, 180 L
Heaviside function, 266 Larmor radius, 241, 273, 278, 284, 306
HEGRA, 13 Lein-Nishina
HESS, 7, 12, 56, 71, 186, 228, 289 regime, 70
Higgs, 171, 172, 179, 194, 195 LEP, 171
Higgsino, 172–174, 183, 192 Lepton, 127
Hillas criterion, 278, 337 Leptonic model, 55, 59, 68, 70
Horizon problem, 145 Leptophilic model, 156, 194
HST, 253 LHAASO, 17
Hubble LHC, 171, 172, 174, 183, 200
constant, 129, 134 Light cylinder, 248
expansion, 140, 147 LIGO, 216, 229
parameter, 129, 135, 136, 202, 203 LISA, 216, 230
time, 145 LMC, 308
Lorentz
factor, 70, 80, 81, 100, 101, 103, 238, 287,
I 337
IACT, 168, 186, 199 gauge, 212
technique, 11, 14, 21 invariance, 130
IceCube, 167, 169, 187, 189, 197, 229 Luminosity density, 333
Imaging air Cherenkov telescope, See IACT Luminosity distance, 328
Inflation, 127, 144, 146, 147
field, 145
INTEGRAL, 19, 229 M
Interaction MACRO, 188
electroweak, 150 MAGIC, 7, 12, 79, 186, 228
neutrino, 162 Magnetobremsstrahlung, 23
strength of, 149 Magnetosphere, 248
strong, 150, 159 Majorana, 172, 173, 190, 203
weak, 151 Mandelstam variables, 153
Intergalactic magnetic field, 281 Maxwell’s equations, 212
Inverse-Compton, 7 MCMC, 170
cross section, 42 Mean-free path, 273
scattering, 22, 42, 45, 47, 84 Microquasar, 21, 97, 313, 314, 346
Inverse-Compton scattering, 299 Microwave background. See CMB
360 Index
MILAGRO, 7, 17, 229 Propagator, 150

Milky Way, 187, 216, 308 PSF, 231, 233, 238, 252
Minkowski metric, 130 Pulsar, 77, 81, 246, 249
MSSM, 171, 172, 174, 179, 192 binary, 98
Muon, 176 magnetosphere, 77, 80
millisecond, 78, 200, 244, 246, 250, 312
physics, 247, 254
N Pulsar wind nebula. See PWN
N-body, 201, 205 PWN, 7, 10, 70, 77, 81, 87, 92, 94, 109, 246,
Neutralino, 139, 144, 167, 169, 172–177, 183, 251, 308
187, 192, 203 PYTHIA, 182
Neutrino, 8, 32, 65, 127, 141, 142, 160, 162,
187
atmospheric, 165 Q
cross section, 148 Q-balls, 176, 196
Neutron star, 4, 78, 247, 314 QCD
Nobel Prize, 128, 159, 166, 216 parton model, 159
Nucleosynthesis, 4, 190, 193 running coupling constant, 150, 158
NuSTAR, 63, 65, 109, 230 Quantum chromodynamics. See QCD
Quark, 127, 159
colour, 160
O Quasar. See AGN
Oscillation length, 164
OSO-3, 230
R
Radio galaxies, 10, 244, 257, 313, 317
P ROSAT, 300
Pair annihilation, 4, 31, 150
Pair production
Bethe-Heitler, 23, 24, 28, 47, 50 S
cross section, 29, 30, 45 Saas-Fee particle, 200, 206
magnetic, 40 SAS-2, 230
photon–photon, 42, 45, 47 Scalar electrons, selectrons, 132
PAMELA, 180, 181, 190, 192, 194, 298 Schwarzschild radius, 133, 208, 240, 346
Particle acceleration, 166 SED, 10, 28, 37, 80, 83, 85, 92, 94, 228, 254,
Parton model. See QCD 303
PeVatron, 7, 53, 58, 60, 65 Sfermion, 174
Photino, 172 Sgr A*, 15, 66, 68, 69, 72
Photo-disintegration, 23, 24, 47, 48 Shakura-Sunyaev accretion disk, 272
Photomeson, 24, 42, 48, 67 Shock acceleration, 26, 39, 48, 51, 85, 300,
Photon. See c 304, 307
Photopair, 272, 273 SNR, 7, 10, 19, 51, 52, 56, 59, 63, 108, 156,
Photopion, 272–275, 281, 339 227, 244, 299, 300
Pierre Auger Observatory, 229 Solar flares, 4
Pion, 156 Solar system, 291
decay. See p0 decay Sommerfeld enhancement, 156, 183, 191, 197,
production, 296 199
Planck South Atlantic Anomaly, 231, 243
mass, 131, 133 Spallation, 5
satellite, 148, 190 Sparticle, 175
Plerion, 19, 70 Spectral energy distribution. See SED
Point spread function. See PSF Squark, 174
Polarization, 254 SSC, 261, 262, 318
p–p interaction, 32, 56, 67 Standard Model, 127, 171, 199
Index 361
Star formation rate, 328 Vacuum expectation values, 172

Star-forming galaxies, 200, 308 Variability time scale, 270
Starburst galaxies, 10, 244, 257, 309 VERITAS, 7, 12, 79, 228, 289
Strong interaction. See Interaction VIRGO, 216
Structure formation, 142, 161, 203, 204 Virtual particle, 150, 176, 209
Structure functions, 159
Sunyaev–Z’eldovich effect, 157
Super–Kamiokande, 188 W
Superbubble, 65 W boson, 127, 162, 182
Superluminal, 168, 254, 270, 271, 314 Weak coupling constant, 150
Supernova remnant. See SNR Whipple, 13, 186
Supersymmetry, 132, 147, 169–172, 175 White dwarf, 227, 247, 314
SUSY. See supersymmetry WIMP, 74, 142, 143, 167, 169, 312
Suzaku, 19, 229 annihilation, 201, 202
Swift, 229, 331 model, 193
Synchrotron, 23, 36, 86, 181, 261, 276, 278, WMAP, 134, 148, 190, 199, 211, 299
307, 337
Synchrotron self-Compton. See SSC
X
X-ray binary, 7
T high-mass, 244, 314
Test statistic (TS), 242 low-mass, 314
Thomson XENON100, 200
cross section, 46, 152 XMM-Newton, 19, 62, 229
regime, 43, 45, 80, 264, 265, 286
scattering, 152, 284
Y
Yukawa, 159, 191
U
UHECR, 229, 241, 321, 337, 340
Unruh effect, 209 Z
Z boson, 127, 171, 182
Zino, 172
V Zwicky, 217
Vacuum energy, 130–132, 145

Astrophysics at Very High Energies - Saas-Fee Advanced Course 40. Swiss Society For Astrophysics and Astronomy (PDFDrive)

Uploaded by

Copyright:

Available Formats

Astrophysics at Very High Energies - Saas-Fee Advanced Course 40. Swiss Society For Astrophysics and Astronomy (PDFDrive)

Uploaded by

Document Information

Original Description:

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

Astrophysics at Very High Energies - Saas-Fee Advanced Course 40. Swiss Society For Astrophysics and Astronomy (PDFDrive)

Uploaded by

Copyright:

Available Formats

Saas-Fee Advanced Course 40

Swiss Society for Astrophysics and Astronomy

For further volumes:

Astrophysics at Very High

Saas-Fee Advanced Course 40

Swiss Society for Astrophysics and Astronomy

ISSN 1861-7980 ISSN 1861-8227 (electronic)

Library of Congress Control Number: 2013932553

Ó Springer-Verlag Berlin Heidelberg 2013

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

string cosmology. He is collaborating in various experiments to search for evi-

Versoix, October 2012 Roland Walter

Gamma Rays at Very High Energies . . . . . . . . . . . . . . . .......... 3

8.3 Enigmatic TeV Binaries . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

Multi-Messenger Astronomy and Dark Matter . . . . . . . . . . . . ...... 125

7.5 Leptophilic Dark Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

Sources of GeV Photons and the Fermi Results . . . . . . . . . . . . . .... 227

4.2 Leptonic Jet Models . . . . . . . . . . . . . . . . . . . . . . . . . ..... 268

F. Aharonian et al., Astrophysics at Very High Energies, 3

essentially different detection methods and approaches applicable to different energy

1.1 Status of Observational Gamma Ray Astronomy

clocks”—the Pulsars—constitute the largest population of identified galactic GeV

1.2 Links to Other Disciplines

On the GeV–TeV Links

2 Astrophysical Potential of Ground-Based Detectors

Earth’s atmosphere is not transparent to gamma-rays of any energy. Therefore, their

showers (initiated by interactions of gamma-ray) either directly or through their

Fig. 7 The energy-flux

2.1 IACT Arrays

Figures 7, 8, 9, and 10 represent three characteristic examples demonstrating the

2.2 Potential of EAS Arrays

Mkn 421 (2000/2001)

2.3 Future IACT Arrays

Planning of the next generation of Imaging Atmospheric Cherenkov Telescope

efficiency of detection/identification of electromagnetic showers, (ii) good accuracy

Fig. 12 Layout of the future Cherenkov telescope array

us to probe the gamma-ray luminosities of potential TeV emitters at the levels of

The general tendency of decreasing gamma-ray fluxes with energy becomes

Sub-100 GeV Regime

The energy threshold εth of IACTs is generally defined as a characteristic energy

Multi-GeV Regime: Gamma-Ray Timing Explorers

The remarkable success of observational gamma-ray astronomy, together with

3.1 General Comments

Any interpretation of an astrophysical observation starts with identification of

Because of universal presence of dense radiation fields, ultrarelativistic electrons

cosmic gamma-ray production. In any case, the photo-disintegration reaction rates

Relativistic Electron–Photon Cascades

The interactions of relativistic electrons, either accelerated directly or being secondary

(the counterpart of bremsstrahlung), in photon gas (the counterpart of inverse Comp-

Calculations of emissivity of gamma rays resulting from interactions of accelerated

synchrotron radiation or inverse Compton scattering (IC) in the Thomson regime,

3.2 Brief Overview of Important Processes

Interactions with Matter

0 0.2 0.4 0.6 0.8 1

energies ε ≥ εcr . At lower energies electrons dissipate their energy by ionization

Correspondingly, the lifetime of electrons is

where n is the number density of the ambient gas.

Here κ = 4E p /m 2p c4 and x = E γ /E p .