Naturalness, String Landscape and Multiverse: Arthur Hebecker

Lecture Notes in Physics
Arthur Hebecker
Naturalness,
String
Landscape and
Multiverse
A Modern Introduction with Exercises
Volume 979
Founding Editors
Wolf Beiglböck, Heidelberg, Germany
Jürgen Ehlers, Potsdam, Germany
Klaus Hepp, Zürich, Switzerland
Hans-Arwed Weidenmüller, Heidelberg, Germany
Series Editors
Matthias Bartelmann, Heidelberg, Germany
Roberta Citro, Salerno, Italy
Peter Hänggi, Augsburg, Germany
Morten Hjorth-Jensen, Oslo, Norway
Maciej Lewenstein, Barcelona, Spain
Angel Rubio, Hamburg, Germany
Manfred Salmhofer, Heidelberg, Germany
Wolfgang Schleich, Ulm, Germany
Stefan Theisen, Potsdam, Germany
James D. Wells, Ann Arbor, MI, USA
Gary P. Zank, Huntsville, AL, USA
The Lecture Notes in Physics
The series Lecture Notes in Physics (LNP), founded in 1969, reports new
developments in physics research and teaching-quickly and informally, but with a
high quality and the explicit aim to summarize and communicate current knowledge
in an accessible way. Books published in this series are conceived as bridging
material between advanced graduate textbooks and the forefront of research and to
serve three purposes:
• to be a compact and modern up-to-date source of reference on a well-defined

topic;
• to serve as an accessible introduction to the field to postgraduate students and
nonspecialist researchers from related areas;
• to be a source of advanced teaching material for specialized seminars, courses
and schools.
Both monographs and multi-author volumes will be considered for publication.

Edited volumes should however consist of a very limited number of contributions
only. Proceedings will not be considered for LNP.
Volumes published in LNP are disseminated both in print and in electronic
formats, the electronic archive being available at springerlink.com. The series
content is indexed, abstracted and referenced by many abstracting and information
services, bibliographic networks, subscription agencies, library networks, and
consortia.
Proposals should be sent to a member of the Editorial Board, or directly to the
responsible editor at Springer:
Dr Lisa Scalone
Springer Nature
Physics
Tiergartenstrasse 17
69121 Heidelberg, Germany
lisa.scalone@springernature.com
More information about this series at http://www.springer.com/series/5304

Arthur Hebecker
Naturalness, String
Landscape and Multiverse
A Modern Introduction with Exercises
Arthur Hebecker
Institut für Theoretische Physik
Universität Heidelberg
Heidelberg, Germany
ISSN 0075-8450 ISSN 1616-6361 (electronic)

ISBN 978-3-030-65150-3 ISBN 978-3-030-65151-0 (eBook)
https://doi.org/10.1007/978-3-030-65151-0
© Springer Nature Switzerland AG 2021
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.
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.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To my wife, Andrea
Preface
This course intends to give a concise but technical introduction to ‘Physics Beyond
the Standard Model’ and early cosmology as seen from the perspective of string
theory. Basics of string theory will be taught as part of the course. As a central
physics theme, the two hierarchy problems (of the cosmological constant and of
the electroweak scale) will be discussed in view of ideas like supersymmetry,
string theory landscape, eternal inflation and multiverse. The presentation will
include critical points of view and alternative ideas and explanations. Problems with
solutions are also provided to facilitate the use of these notes in classroom and for
self-study.
Basic knowledge of quantum field theory (QFT), general relativity and cosmol-
ogy will be assumed. Supersymmetry, elements of supergravity and fundamentals
of string theory will be taught together with a number of geometrical concepts
needed to study string compactifications. However, given the limited scope of a
one-semester lecture series, this can clearly not replace a full string theory course or
the detailed study of string geometry.
The author has taught this course at Heidelberg University with the intention to
prepare students who have taken a two-semester QFT and a one-semester relativity
course for master’s thesis research in string phenomenology. Another goal was to
allow students who intend to do research in particle phenomenology, cosmology
or formal (mathematical) string theory to develop some basic understanding of the
possible relation of string theory to ‘real-world’ physics and its most fundamental
problems.
For students who had the privilege of enjoying a complete graduate-level
education (with full lecture courses on strings and supersymmetry/supergravity)
before embarking on research, most of the material in the first part of this course
will be familiar. Still, depending on the focus of their string and cosmology courses,
they may find useful additional information about landscape, multiverse, eternal
inflation and alternative perspectives in the second half of the course.
The detailed plan of the lecture notes is as follows: We will start in Chap. 1
with a brief tour of the Standard Model, emphasising the perspective of a low-
energy effective field theory, the coupling to gravity, and the electroweak hierarchy
and cosmological constant problems. Chapter 2 introduces supersymmetry and
supergravity which, however, offer only a partial resolution of the fine-tuning
problems of the Standard Model discussed earlier. It becomes apparent that the
vii
viii Preface
highest relevant energy scales, at the cutoff of the effective field theory, have to be
involved. This motivates the study of string theory as the best-explored candidate
quantum gravity theory in Chaps. 3 and 4. We will see that the bosonic string,
which suffers from the absence of fermions and from an unstable vacuum, has to
be promoted to the superstring. The latter provides all desired ingredients for a
consistent theory of gravity and particle physics, albeit in ten spacetime dimensions
and with far too much supersymmetry. Compactifications to four dimensions are
the subject of Chaps. 5 and 6: First, we consider pure Calabi–Yau geometries,
leading to highly supersymmetric and unrealistic 4d Minkowski-space models.
Then, the inclusion of non-perturbative objects and fluxes leads to the proper string
landscape with supersymmetric and non-supersymmetric models with non-zero
cosmological constants of either sign. The key insight is the enormous number of
such solutions (a recent analysis arriving at ∼10272,000), each corresponding to a
different 4d effective theory. We now see that, assuming the non-trivial constructions
with broken supersymmetry and positive cosmological constant stand up to further
scrutiny, the ‘fine-tuned’ or ‘unnatural’ parameters of the Standard Model may
indeed be accommodated by the string landscape. Chapter 7 deals with the important
but complicated and speculative question of how the landscape gets populated
during eternal inflation and whether statistical predictions for future observations
can be derived. A number of alternative perspectives on the hierarchy problems and
quantum gravity are discussed in Chap. 8 before, in Chap. 9, we end by summarising
the overall picture and the challenges that should have crystallised during the study
of this course.
While useful references for background material and deeper exploration will
be provided as we go along, it may not hurt to give some essential literature
right away: Good sources for the background knowledge in QFT and relativity
are [1] and [2], respectively. For more details on Standard Model and particle-
physics-related topics, Refs. [3, 4] represent useful sources. For supersymmetry and
supergravity, see [5, 6]. Two of the most complete modern string theory textbooks
are [7, 8]. Concerning string phenomenology, [9] represents a very comprehensive
monograph, which, in particular, covers the important subjects of how specific
gauge and matter sectors are realised in string compactifications—a topic that we
treat very superficially in this course. A very useful set of notes emphasising the
geometric side of how the landscape arises from string theory is [10]. For a detailed
review of string landscape physics, see [11].
Heidelberg, Germany Arthur Hebecker

October 2020
Acknowledgements
Many friends and colleagues deserve my deepest gratitude for helping me to learn
and understand the material covered in this course. Among those, I am particularly
indebted to Wilfried Buchmüller, Miriam Cvetic, Jan Louis, Dieter Lüst, John
March-Russell, Timo Weigand, and Alexander Westphal. Especially, the time during
which Timo Weigand was my colleague in Heidelberg was extremely fruitful and
enjoyable. Moreover, I am grateful to my younger colleagues Daniel Junghans and
Sascha Leonhardt who were involved in the actual teaching of this course. Special
thanks go to Daniel Junghans and Alexander Westphal for their careful reading of
the manuscript and their corrections and comments. In addition, I am grateful to
Janning Meinert for producing professional figures and to Benjamin Kellers for
helping with the index. There is a long list of colleagues and friends whose help
and advice were very important for me and eventually for these notes. They include
Andreas Braun, Felix Brümmer, Joe Conlon, Laura Covi, Gia Dvali, Ben Freivogel,
Gero von Gersdorff, Benedict von Harling, Joerg Jaeckel, Olaf Lechtenfeld, Hans-
Peter Nilles, Eran Palti, Jan Pawlowski, Tilman Plehn, Fernando Quevedo, Riccardo
Rattazzi, Michael Ratz, Fabrizio Rompineve, Micheal G. Schmidt, Gary Shiu, Pablo
Soler, Stefan Theisen, Gianmassimo Tasinato, Michele Trapletti, Enrico Trincherini,
Roberto Valandro, Johannes Walcher, James Wells, Julius Wess, Christof Wetterich,
Lukas Witkowski, Timm Wrase and many others. Each of them clearly deserves to
be mentioned more personally, and I apologise for not being able to do so for reasons
of space. Also, I ask for the forgiveness of those friends and colleagues whom I have
missed. Finally, I would like to thank my parents for encouragement throughout my
whole life and, most importantly, my wife, Andrea, for her love and for tolerating
the excessive working hours that were sometimes unavoidable in preparing these
notes. This work was supported by the Deutsche Forschungsgemeinschaft (DFG,
German Research Foundation) under Germany’s Excellence Strategy EXC-2181/1-
390900948 (the Heidelberg STRUCTURES Cluster of Excellence).
ix
Contents
1 The Standard Model and Its Hierarchy Problem(s) .. . . . . . . . . . . . . . . . . . . . 1

1.1 Standard Model—The Basic Structure . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.2 Standard Model—Parameter Count. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5
1.3 Effective Field Theories—Cutoff Perspective .. . .. . . . . . . . . . . . . . . . . . . . 8
1.4 Effective Field Theories—QFTUV vs. QFTIR . . . .. . . . . . . . . . . . . . . . . . . . 10
1.5 The Standard Model as an Effective Field Theory .. . . . . . . . . . . . . . . . . . 13
1.6 The Electroweak Hierarchy Problem . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 16
1.7 Fine-tuning .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 19
1.8 Gravity and the Cosmological Constant Problem .. . . . . . . . . . . . . . . . . . . 23
1.9 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 26
1.9.1 Electroweak Symmetry Breaking . . . . . . .. . . . . . . . . . . . . . . . . . . . 26
1.9.2 The Standard Model Is Anomaly Free . .. . . . . . . . . . . . . . . . . . . . 28
1.9.3 The Standard Model and SU (5). . . . . . . . .. . . . . . . . . . . . . . . . . . . . 32
1.9.4 Weyl Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 35
1.9.5 Covariant Expression for the 1-Loop Vacuum Energy . . . . . 38
2 Supersymmetry and Supergravity . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 41
2.1 SUSY Algebra and Superspace . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 41
2.2 Superfields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 45
2.3 Chiral Superfields .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 46
2.4 SUSY-Invariant Lagrangians .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 47
2.5 Wess-Zumino-Type Models .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 49
2.6 Real Superfields.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 50
2.7 SUSY Breaking .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 52
2.8 Supersymmetrising the Standard Model .. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 53
2.9 Supersymmetric and SUSY Breaking Masses
and Non-renormalisation . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 56
2.10 The Minimal Supersymmetric Standard Model (MSSM) . . . . . . . . . . . 58
2.11 Supergravity: Superspace Approach .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 61
2.12 Supergravity: Component Approach.. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 64
2.13 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 68
2.13.1 Simple Manipulations Within the Superspace
Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 68
2.13.2 Deriving Component Actions . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 70
xi
xii Contents
2.13.3 Fierz Identities for Weyl Spinors .. . . . . . .. . . . . . . . . . . . . . . . . . . . 74

2.13.4 SUSY in Components . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 75
2.13.5 Gauge Coupling Unification .. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 78
2.13.6 Graviton Spin (Helicity) . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 83
3 String Theory: Bosonic String .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 85
3.1 Strings: Basic Ideas .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 85
3.2 Symmetries, Equations of Motion, Gauge Choice . . . . . . . . . . . . . . . . . . . 88
3.3 Open String . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 92
3.4 Quantisation.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 94
3.5 Explicit Construction of Physical States: Open String .. . . . . . . . . . . . . . 100
3.6 Explicit Construction of Physical States: Closed String . . . . . . . . . . . . . 103
3.7 The 26d Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 105
3.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 109
3.8.1 Point Particle Action . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 109
3.8.2 Commutation Relations of Oscillator Modes .. . . . . . . . . . . . . . 110
3.8.3 Trace of the Energy-Momentum Tensor . . . . . . . . . . . . . . . . . . . . 111
3.8.4 Virasoro Algebra .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 112
3.8.5 Normal Ordering Constant as Casimir Energy . . . . . . . . . . . . . 114
3.8.6 Kalb-Ramond Field from the Worldsheet Perspective . . . . . 115
4 String Theory: Interactions and Superstring . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 117
4.1 State-Operator Correspondence .. . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 117
4.2 Scattering Amplitudes .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 119
4.3 Worldsheet Supersymmetry .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 121
4.4 Worldsheet Supergravity . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 123
4.5 Quantisation of the Superstring . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 125
4.6 GSO or Gliozzi-Scherk-Olive Projection .. . . . . . . .. . . . . . . . . . . . . . . . . . . . 129
4.7 Consistent Type II Superstring Theories . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 130
4.8 Other 10d Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 133
4.9 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 135
4.9.1 Explicit State-Operator Mapping in the Free Case . . . . . . . . . 135
4.9.2 Euler Number and Genus of Riemann Surfaces .. . . . . . . . . . . 137
4.9.3 Dilaton vs. String Coupling . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 139
4.9.4 Elementary Exercises with 2d Spinors ... . . . . . . . . . . . . . . . . . . . 140
4.9.5 SUSY Algebra in 2d . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 142
5 10d Actions and Compactification . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 145
5.1 10d Supergravities and Type IIB as an Example .. . . . . . . . . . . . . . . . . . . . 145
5.2 Kaluza–Klein Compactification.. . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 148
5.3 Towards Calabi–Yau Manifolds .. . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 151
5.4 Homology and Cohomology . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 157
5.5 Calabi–Yau Moduli Spaces. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 163
5.6 Explicit Parameterisation of Calabi–Yau Moduli Spaces .. . . . . . . . . . . 167
Contents xiii
5.7 An Aside on String Model Building: From Heterotic

Compactifications to Orientifold Models with Branes
and F-Theory.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 173
5.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 178
5.8.1 Dimensional Reduction .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 178
5.8.2 SO(2n) vs. U(n) . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 180
5.8.3 Complex Projective Spaces . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 182
6 The Flux Landscape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 187
6.1 Compact Geometries with p-Form Fluxes . . . . . . .. . . . . . . . . . . . . . . . . . . . 187
6.2 Bousso–Polchinski Model.. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 191
6.3 The Type IIB Flux Landscape (GKP) . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 194
6.4 Kahler Modulus Stabilisation and SUSY Breaking (KKLT) . . . . . . . . 197
6.5 The Anti-D3-Brane Uplift of KKLT . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 202
6.6 The Large Volume Scenario .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 207
6.7 Vacuum Statistics and the Tuning of the Cosmological Constant .. . 209
6.8 Higgs Mass and Other Landscape-Related Issues. . . . . . . . . . . . . . . . . . . . 214
6.9 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 216
6.9.1 No-Scale Kahler Potentials and KKLT .. . . . . . . . . . . . . . . . . . . . 216
6.9.2 The LVS Scalar Potential and Stabilisation Mechanism . . . 218
7 Eternal Inflation and the Measure Problem . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 223
7.1 From Slow-Roll Inflation to the Eternal Regime .. . . . . . . . . . . . . . . . . . . . 223
7.2 Eternal Inflation in the Landscape . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 226
7.3 Tunnelling Transitions in Quantum Mechanics . .. . . . . . . . . . . . . . . . . . . . 230
7.4 Tunnelling Transitions in Field Theory .. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 234
7.5 Tunnelling in Gravitational Theories . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 237
7.6 Our Universe in the Eternally Inflating Landscape . . . . . . . . . . . . . . . . . . 241
7.7 Making Statistical Predictions and the Measure Problem . . . . . . . . . . . 243
7.8 Proposed Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 244
7.9 Predictions from First Principles? . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 247
7.10 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 250
7.10.1 Coleman–De Luccia Tunnelling . . . . . . . .. . . . . . . . . . . . . . . . . . . . 250
8 Concluding Remarks and Some Alternative Perspectives . . . . . . . . . . . . . . . 255
8.1 Low-Scale SUSY Versus Technicolor . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 255
8.2 From the ‘Little Higgs’ to Large or Warped Extra Dimensions . . . . . 259
8.3 Cosmological Selection and the Relaxion .. . . . . . .. . . . . . . . . . . . . . . . . . . . 265
8.4 The Swampland Program . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 267
8.5 The Swampland and de Sitter . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 271
8.6 More Direct Approaches to Quantum Gravity . . .. . . . . . . . . . . . . . . . . . . . 275
8.7 Asymptotic Safety and the Hierarchy Problem . .. . . . . . . . . . . . . . . . . . . . 278
9 Summary .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 283
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 289
Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 309
About the Author
Arthur Hebecker after studying physics in Moscow,

Frankfurt/Main and Munich, gained his PhD from the
University of Hamburg/DESY in 1995. His Diploma
and PhD supervisors were Julius Wess (Munich)
and Wilfried Buchmüller (Hamburg). Arthur Hebecker
spent his postdoc years in Stanford, Cambridge, Hei-
delberg and at CERN. He won a Feodor-Lynen Fellow-
ship from the Humboldt-Foundation and a Heisenberg-
Fellowship from the Deutsche Forschungsgemein-
schaft. After a brief period as a staff member at
DESY/Hamburg, he became a professor of theoretical
particle physics and cosmology at the University of
Heidelberg in 2004, where he is working on physics
beyond the Standard Model, string phenomenology and
cosmological inflation. He has been a member of the
Particle Data Group since 2015, where he was a co-
author of the review on Grand Unified Theories.
xv
The Standard Model and Its Hierarchy
Problem(s) 1
As already stated in the Preface, we assume some familiarity with quantum field
theory (QFT), including basics of regularisation and renormalisation. There exists a
large number of excellent textbooks on this subject, such as [1, 12–15]. The reader
familiar with this topic will most probably also have some basic understanding of the
Standard Model of Particle Physics, although this will not be strictly necessary since
we will introduce this so-called Standard Model momentarily. It is also treated at
different levels of detail in most QFT texts, most notably in [12, 15]. Books devoted
specifically to theoretical particle physics and the Standard Model include [3, 4, 16].
A set of lecture notes covering the Standard Model and going beyond it is [17]. We
will refer to some more specialised texts as we go along.
1.1 Standard Model—The Basic Structure
A possible definition of the Standard Model is as follows: It is the most general

renormalisable field theory with gauge group
GSM = SU (3) × SU (2) × U (1) , (1.1)
three generations of fermions, and a scalar. These fields transform in the represen-
tations
(3, 2)1/6 + (3̄, 1)−2/3 + (3̄, 1)1/3 + (1, 2)−1/2 + (1, 1)1 and (1, 2)1/2
(1.2)
respectively. Here the boldface numbers specify the representations of SU (3) and
SU (2) via their dimension (in our case only singlet, fundamental or antifundamental
occur, the latter denoted by an overline) and the index gives the U (1) charge Y , also
© Springer Nature Switzerland AG 2021 1

A. Hebecker, Naturalness, String Landscape and Multiverse, Lecture Notes
in Physics 979, https://doi.org/10.1007/978-3-030-65151-0_1
2 1 The Standard Model and Its Hierarchy Problem(s)
known as hypercharge. The overall normalisation of the latter is clearly convention-

dependent.1
If one adds gravity in its simplest and essentially unique theoretical formulation
(Einstein’s general relativity), then this data offers an almost complete fundamental
description of the material world. This structural simplicity and the resulting small
number of fundamental parameters (to be specified in a moment) is very remarkable.
What is even more remarkable is the enormous underlying unification: So many very
different macroscopic and microscopic phenomena which we observe in everyday
life and in many natural sciences follow from such a (relatively) simple underlying
theory.
Clearly, important caveats have already been noted above: The description is
almost complete, the theory is relatively simple (not as simple as one would wish)
and, maybe most importantly, it is only fundamental to the extent that we can test it
at the moment. Quite possibly, more fundamental building blocks can be identified
in the future. The rest of this course is about exactly these caveats and whether,
based on those, theoretical progress is possible.
But first let us be more precise and explicit and turn the defining statements (1.1)
and (1.2) into a field-theoretic lagrangian. Given the theoretically well-understood
and experimentally tested rules of quantum field theory (QFT), this can be done
unambiguously. The structure of the lagrangian is
LSM = Lgauge + Lmat t er + LH iggs + LY ukawa . (1.3)
The gauge part is completely standard,
1 (1) (1) μν 1 1
Lgauge = − 2
Fμν F − 2 tr Fμν
(2) (2) μν
F − 2 tr Fμν
(3) (3) μν
F , (1.4)
4g1 2g2 2g3
with the upper index (i) running over U (1), SU (2), SU (3), in this order. The field
strengths are defined as Fμν = i[Dμ , Dν ] with Dμ = ∂μ − iAμ and, in the non-
abelian cases, Aμ = AA μ TA . One should also remember the standard normalisation
tr(TA TB ) = δAB /2 of the SU (N) generators in the fundamental representation.
The matter or, more precisely, the fermionic matter contribution reads

Lmat t er = / j ψj
ψ j iD with (Dj )μ = ∂μ − iRj (Aμ ) (1.5)
j
1 By this we mean the freedom to rescale the gauge potential Aμ by a constant, such that the values
of Y and of the gauge coupling change correspondingly. In contrast to the non-abelian case (see
below), there is no preferred choice intrinsic to the abelian gauge lagrangian Fμν F μν since the
latter is homogeneous in Aμ . In our conventions the electric charge is given by Q = T3 + Y , with
T3 the third SU (2) generator.
1.1 Standard Model—The Basic Structure 3
with j running over left-handed quark doublets, right-handed up- and down-type
quarks, lepton-doublet and right-handed leptons (each coming in three generations
or families):
ψj ∈ { {qLa , (uaR )c , (dRa )c , lLa , (eR ) }, a = 1, 2, 3 } .

a c
(1.6)
The five types of fermions from qL to eR c correspond precisely to the five terms
in the direct sum in (1.2). Furthermore, Rj (Aμ ) denotes the representation of

Aμ ∈ Lie(GSM ) appropriate for the fermion of type j . To make our conventions
unambiguous, we have to specify in detail how we describe the spinor fields. One
convenient choice (the one implicitly used above) is to always work with left-handed
4-component or Dirac spinors. In other words, we do not use general Dirac spinors
built from Weyl spinors according to

ψα
ψD = with α, α̇ = 1, 2 . (1.7)
χ α̇
Instead, all our 4-spinors are left-handed:

ψα
ψ= . (1.8)
0
In particular, this explains why we use the charge-conjugate of right-handed quarks

and leptons as our fundamental fields, cf.
c
(qL )α (uR )α 0
qL = vs. ucR = = . (1.9)
0 0 (uR )α̇
We see that the quantum numbers given in (1.2) can be viewed as referring to either
these left-handed fields or to the corresponding 2-component Weyl spinors. The
latter will in any case be the most useful way to describe fermions when talking
about supersymmetry below.
The scalar or Higgs lagrangian is
LH iggs = −(Dμ )† (D μ ) − V () with V () = −m2H † + λH († )2 ,

(1.10)
where is an SU (2) doublet with charge 1/2 under U (1) (the hypercharge-U (1)
or U (1)Y ). Finally, there are the Yukawa terms

LY ukawa = − λj k ψ j ψkc + h.c. , (1.11)
jk
where the sum runs over all combinations of fields for which the relevant product
of representations contains a gauge singlet. We left the group indices and their
corresponding contraction implicit.
Note that, since all our fields are l.h. 4-component spinors, we have to write
ψψ c rather than simply ψψ. The latter would be identically zero. Note also that
α̇
the 4-spinor expression ψψ c corresponds to ψ α̇ ψ in terms of the Weyl spinor ψα
contained in the 4-spinor ψ.
Crucially, the Higgs potential has a minimum with S 3 topology at || =
v 174 GeV, leading to spontaneous gauge symmetry breaking. One can choose
the VEV to be real and aligned with the lower component of , leading to the
parameterisation

0 √ .
= (1.12)
v + h/ 2
It is easy to see that the symmetry breaking pattern is SU (2) × U (1)Y → U (1)em
(see problems). Three would-be Goldstone-bosons along the S 3 directions are
‘eaten’ by three of the four vector bosons of SU (2) × U (1)Y . This leads to the
W ± and Z bosons with masses mW ± 80 GeV and mZ 90 GeV. The surviving
real Higgs scalar h is governed by
1 1
L ⊃ − (∂h)2 − m2h h2 . (1.13)
2 2
Our notation ‘⊃’ means that we are displaying only the subset of terms in L which
is most important in the present context. One can easily relate the parameters after
symmetry breaking to those of the original lagrangian:
v 2 = m2H /(2λH ) , m2h = 4λH v 2 . (1.14)

√
We have mh 125 GeV, mH = mh / 2 = 88 GeV, and λH 0.13.2
The surviving massless gauge boson is, of course, the familiar photon. Finally,
one can check that the allowed Yukawa terms suffice to give all charged fermions
a mass proportional to v. The three lightest quark masses are not directly visible to
experiment since the confinement dynamics of the SU (3) gauge theory (QCD) hide
their effect. The three upper components of the lepton-doublets—the neutrinos—
have Q = 0 and remain massless.
The reader should check explicitly which (three) Yukawa coupling terms are
allowed and that no further renormalisable operators (i.e. operators with mass
dimension ≤ 4) consistent with the gauge symmetry of the Standard Model exist.
√
2 We
√ note that a slightly different convention, v → v , with 2 = (v + h)/ 2 and hence v =
2v 246 GeV is also widely used. Ours has the advantage that mtop ≡ mt v.
1.2 Standard Model—Parameter Count 5
1.2 Standard Model—Parameter Count
The most obvious parameters are the three gauge couplings gi . Then there is of
course the Higgs quartic coupling λH and the Higgs mass parameter mH (defining
the negative quadratic term −m2H ||2 in the potential). It is not so easy to count the
independent Yukawa couplings contained in the three terms
3

λuab q La ∗ ubR + λdab q La dRb + λeab l L eR
a b
+ h.c. (1.15)
a,b=1
The reader should check that these and only these terms are GSM -invariant, given
the general Yukawa-term structure displayed in (1.11). One frequently sees the
notation (suppressing generation indices)
˜ R
q L u with ˜ α =
αβ (β )∗
(1.16)
for the first term above. Introducing this 2-component vector ˜ is necessary if one
wants to read (1.15) in terms of SU (2) matrix notation. If one simply says that
‘group indices are left implicit’ (as we do), writing ∗ is sufficient. Of course,
we could also have avoided the explicit appearance of ∗ in (1.15) altogether by
exchanging it with its complex conjugate, implicit in ‘h.c.’ This is a matter of
convention and the form given in (1.15), (1.16) is close what most authors use.
Maybe the easiest way to count the parameters in (1.15) is to think in terms
of the low-energy theory with replaced by its VEV. Then the above expression
contains three 3×3 complex mass matrices. Furthermore, these mass matrices relate
six independent sets of fermions (since the first term only contains uL and the second
only dL ). Thus, the matrices can be diagonalised using bi-unitary transformations—
i.e. a basis change of the fermion fields. We are then left with 3 × 3 = 9 mass
parameters for three sets of up- and down-type quarks and three leptons.
However, due to SU (2) gauge interactions, one of terms in

3
q aL Dq
/ La (1.17)
a=1
contains both uL and dL . It originates in the off-diagonal terms of σ 1,2 which are
/ In this uL /dL term, the unitary transformation used above does not
contained in D.
cancel and a physical 3 × 3 matrix describing ‘flavour changing charged currents’
(the CKM matrix) is left.3 Let us write the relevant term symbolically as

3
uaL γ μ U ab dLb . (1.18)
a,b=1
The matrix U arises as the product of two unitary matrices from the bi-unitary
transformations above. Hence it is unitary.
It will be useful to pause and think more generally about parameterising a unitary
n × n matrix U . First, a general complex matrix U has 2n2 real parameters. The
matrix U U † is always hermitian, so imposing the hermiticity requirement U U † = 1
imposes n2 real constraints. Then n2 parameters are left. Next, recall that orthogonal
matrices have n(n − 1)/2 real parameters or rotation angles. Thus, since unitary
matrices are a superset of orthogonal matrices, we may think of characterising
unitary matrices by n(n − 1)/2 angles and n2 − n(n − 1)/2 phases.4 Now, in
our concrete case, we are free to transform our unitary matrix (in an n-generation
Standard Model) according to
U → Du U D d , (1.19)
where Du,d are diagonal matrices made purely of phases. This is clear since we may
freely rephase the fields uaL and dLa (together with their mass partners uaR and dRa —to
keep the masses real). The re-phasing freedom of (1.19) can be used to remove 2n−1
phases from U . The ‘−1’ arises since one overall common phase of Du and Dd
cancels and hence does not affect U . So we are left with n2 −n(n−1)/2−(2n−1) =
(n − 1)(n − 2)/2 physically significant phases.
Now we return to U as part of our Standard Model lagrangian with real, diagonal
fermion mass matrix. Here n = 3 and, according to the above, the CKM matrix has
3 real “mixing angles” and one complex phase (characterising CP violation in the
weak sector of the Standard Model). For more details, see e.g. [3], Chapter 11.3.5
This brings our total parameter count to 3 + 2 + 9 + 4 = 18. However, we are
not yet done since we completely omitted a whole general type of term in gauge
theories, the so-called topological or θ -term
L ⊃ θ tr F ∧ F ∼ θ
μνρσ Fμν
a a
Fρσ ∼ θ tr F F̃ . (1.20)
3 These flavour-changing currents correspond to vertices with a (charged) W boson and two left-
handed fermions with different flavour (one up and one down, either both from the same or from
different generations).
4 This is not a proof. One needs to show that such a parameterisation in terms of angles and phases
exists. We will not touch the interesting subject of parameterisations of unitary matrices.
5 For a broader discussion of C, P, CP and its violation see e.g. [18–23] and references therein.
1.2 Standard Model—Parameter Count 7
Most naively, this adds 3 new parameters, one for each factor group. However, these
terms are total derivatives if expressed in terms of Aμ . Thus, they are invisible in
perturbation theory and do not contribute to the Feynman rules. In the non-abelian
case, there exist gauge field configurations localised in space and time (called
instantons) for which

trF ∧ F (1.21)
is non-zero. This does not clash with the total-derivative feature since no globally
defined Aμ exists for such instanton configurations. We will return to instantons in
more detail later. For the U (1), such configurations do not exist, which severely
limits the potential observability of the θ -term in U (1) gauge theories.
Furthermore, and maybe most importantly, the θ -term is precisely of the type that
the non-invariance of the fermionic path integral measure induces if chiral fermion
fields are re-phased. Thus, in the presence of charged fermions without mass terms
(or analogous Yukawa-type couplings preventing a re-phasing) such θ parameters
are unphysical. The upshot of a more detailed analysis in the Standard Model case
(where some but not all conceivable fermionic mass terms are present) is that the
SU (2) and U (1) θ -terms are unobservable (see e.g.[24]) but the QCD θ -term is
physical (for some non-trivial issues in this context see [25] and refs. therein). If
one goes beyond the Standard Model by adding more fields or even just higher-
dimension operators, the electroweak θ -terms may become physical.
A non-zero value of θQCD breaks CP. This is directly visible from the
-tensor
in the definition of the θ -term as well as from its equivalence (through re-phasing)
to complex fermion mass parameters.6 Now, let us assume that CP is broken at
the O(1) level by the theory of strong interactions and that light-quark masses are
∼GeV. Then one would expect the electric dipole moment of the neutron to be O(1)
(in GeV units). Allowing for the suppression by the tiny light-quark masses ∼ 10−3
GeV, the dipole moment should still be large enough to be detected if θQCD where
O(1). However, corresponding search experiments have so far only produced an
extremely small upper bound. The detailed analysis of this bound implies roughly
θQCD < 10−10 .
In any case, we now arrived at our final result of 19 parameters. However, the
status of these parameters is very different. Most notably, 18 of them correspond
to dimension-4 (or marginal) operators, while one—the Higgs mass term—is
dimension-2 and hence relevant. (We recall that the term ‘relevant’ refers to ‘relevant
in the IR’.)
Let us try to make the same point from a more intuitive and physical perspective:
Since the theory is renormalisable, one can imagine studying it at a very high energy
6 Recall that, at the lagrangian level, charge conjugation is related to complex conjugation. In
particular, it is broken by complex lagrangian parameters which cannot be removed by field

redefinitions.
scale, E v ∼ mH . At this scale the Higgs mass is entirely unimportant and we are
dealing with a theory of massless fields characterised by 18 dimensionless coupling
constants. Classically, this structure is scale invariant since only dimensionless
couplings are present. At the quantum level, even without the Higgs mass term,
this scale invariance is badly broken by the non-zero beta-functions, most notably
of the gauge couplings. Indeed the gauge couplings run quite significantly and, even
in the absence of the Higgs ||2 term, QCD would still confine at about 1 GeV and
break the approximate scale invariance completely in the IR.
However, this ‘high-scale’ Standard Model described above is very peculiar
in the following sense: One perfectly acceptable operator, −m2H 2 , is missing
entirely. More precisely, if we characterise the theory at a scale μ by dimensionless
couplings, e.g. gi2 (μ), λH (μ) etc., then we should include a parameter m2H (μ)/μ2 .
If we start at some very high scale (e.g. the Planck scale MP ∼ 1018 GeV—more
on this point later), then this parameter has to be chosen extremely small,
m2H (μ)/μ2 ∼ 10−32 at μ ∼ MP , (1.22)
to describe our world. Indeed, running down from that scale it keeps growing as
1/μ2 until, at about μ ∼ 100 GeV, it starts dominating the theory and completely
changes its structure. This is our first encounter with the hierarchy problem, which
we will discuss in much more detail below.
1.3 Effective Field Theories—Cutoff Perspective
In this course, we assume familiarity with basic QFT. The language of (low-energy)
effective field theory can be viewed as an important part of QFT and hence many
readers will be familiar with it. Nevertheless, since this subject is of such an
outstanding importance for what follows, we devote some space to recalling the
most fundamental ideas of effective field theory (EFT). In addition to chapters in
the various QFT books already mentioned, the reader will be able to find many sets
of lecture notes devoted specifically to the subject of EFTs, e.g. [26–31]. For a wider
perspective on effective theories (not restricted to QFT), see e.g. [32].
To begin, let us assume that our QFT is defined with some UV cutoff U V (and,
if one wants, in finite spatial volume ∼ 1/I R ), such there can be no doubt that we
are dealing with a conventional quantum mechanical system. Of course, the larger
the ratio U V /I R , the more degrees of freedom this system has. The possible IR
cutoff will not be relevant for us and we will not discuss it further. The best example
of a UV cutoff (though not very practical in perturbative calculations) is presumably
the lattice cutoff. It is e.g. well established that this leads to a good description of
gauge theories, including all perturbative as well as non-perturbative effects. Next,
it is also well-known and tested in many cases that the lattice regularisation can be
set up in such a way that Poincare-symmetry is recovered in the IR. Of course, we
could use Poincare-invariant cutoffs (e.g. dimensional regularisation, Pauli–Villars
or even string theory) from the beginning, but the lattice is conceptually simpler and
1.3 Effective Field Theories—Cutoff Perspective 9
more intuitive. Thus, we will be slightly cavalier concerning this point and assume
that we can disregard Poincare-breaking effects in the IR of our system.
As a result (and here we clearly assume a large amount of non-trivial QFT
intuition to be developed by reading standard texts) our low-energy physics can
be characterised by an action of the symbolic structure

1 θ c1 c2
S= d4 x − 2 trF 2 + trF F̃ + trF 4
+ (trF 2 2
) + · · · .
2g 8π 2 4 4
(1.23)
Here we focussed on the gauge theory case and wrote ≡ U V for our cutoff
scale. In other words, we expect that generically all terms allowed by the symmetries
are present and that, on dimensional grounds, whenever a dimensionful parameter
is needed, it is supplied by the cutoff scale . At low energies, only terms not
suppressed by powers of will be important, hence we will always encounter
renormalisable theories in the IR. The relevance of terms in the IR decreases as
their mass dimension grows. This is obvious if one thinks, e.g., in terms of the
contribution a given operator makes to a 4-gluon amplitude: The first term in (1.23)
will contribute ∼ g 2 ; the third will contribute ∼ g 4 k 4 /4 . Clearly, at small typical
momentum k, only the first term is important.7 To see this explicitly one needs to
split off the propagator from the first term and to rescale Aμ → gAμ . The lagrangian
will then contain terms of the type
A ∂ 2 A + g A2 ∂ A + g 2 A4 + g 4 (c1 /4 ) (∂A)4 + · · · , (1.24)
confirming our claim about the 4-gluon amplitude.

The numerical coefficients in (1.23) depend on the details of the regularisation
(e.g. the lattice model) or, in more physical terms, on the UV definition of our theory
at the scale . Indeed, recall our assumption that, what we perceive as a QFT at low
energies is defined as a finite quantum mechanical system at the scale . There
are in general many discrete and continuous choices hidden in this definition. They
will be reflected in the values of g, θ and the ci . Some of these terms can hence
be unusually large or small and this can to a certain extent overthrow the ordering
by dimension advertised above. However, in the mathematical limit k/ → 0, the
power of k/ wins over numerical prefactors. An exception arises if one coefficient
is exactly zero. This important possibility will be discussed below.
Let us add to our gauge theory example given above the apparently much simpler
example of a real scalar field, symmetric under φ → −φ:

1 c1
S= d4 x c0 2 φ 2 − (∂φ)2 − λφ 4 + 2 φ 6 + · · · . (1.25)
2
7 The second term is a total derivative and hence does not contribute in perturbation theory.
The key novelty is that we have a term proportional to a positive power of (a

relevant operator). In the gauge theory case, the most important operators were
merely marginal. Moreover, this term is a mass term and for c0 = O(1) the EFT
below the scale is simply empty. Thus, we must assume that a very particular UV
completion exists which allows for either c0 = 0 (for some qualitative reason) or at
least for the possibility to tune this coefficient to a very small value, c0 1. We
now see that this has some similarity to the Standard Model, where (assuming that
the Standard Model continues to be the right theory above the TeV-scale), a similar
tuning might be needed to keep m2H small.
Arguing that there is a ‘tuning’ or ‘fine-tuning problem’ based only on the above
is not very convincing. One of the reasons is that we were vague about the UV
completion at the scale . It appears possible that the right UV completion will
effortlessly allow for c0 1 or maybe even predict such a small value. Indeed, we
have to admit right away that we will not be able to rule this out during this whole
course. But we will try to explain why many researchers have remained pessimistic
concerning this option.
1.4 Effective Field Theories—QFTUV vs. QFTIR
To do so, we will now modify the use of the word effective field theory: In the above,
we assumed some finite (non-QFT) UV completion and called EFT what remains of
it in the IR. Now, we want to start with some QFT in the UV (to be itself regularised
or UV completed at even higher scales) and consider how it transits to another QFT
(which we will call EFT) in the IR. The simplest way in which this can happen is as
follows: Let our QFTU V contain a particle with mass M and focus on the physics
at k M. In other words, we ‘integrate out’ the heavy (from the IR perspective)
particle and arrive at a theory we might want to call QFTI R —our low-energy EFT.
Let us start with a particularly simple example, borrowed from [29]:
1 1 λ
L = ψi ∂/ψ − mψψ − (∂φ)2 − M 2 φ 2 + yφψψ − φ 4 . (1.26)
2 2 4!
We assume m M and we have already ignored all terms suppressed by .
The above lagrangian is then renormalisable, such that we may indeed view (1.26) as
defining our theory through some parameter choice at a very high8 scale μ1 M.
We are interested in the EFT at μ2 with m μ2 M.
The correct procedure (‘running and matching’) would be as follows: One writes
down the most general lagrangian LEF T for ψ at the scale μ2 and calculates (at
some desired loop order) a sufficiently large set of observables (e.g. mass, 4-point-
amplitude, etc.). Then one calculates the same observables using the full theory
8 We do not insist on being able to take the mathematical limit → ∞ or μ1 → ∞ since we do

not want to deal with issues like a possible Landau pole or a sign change of λ in the far UV.
1.4 Effective Field Theories—QFTUV vs. QFTIR 11
defined by (1.26). This includes tree level diagrams and loops involving φ as well as
the renormalisation group (RG) evolution. Finally, one determines the parameters
of LEF T such that the two results agree.
Our course is not primarily about EFTs and we will take a shortcut. First, we set
λ = 0 since it will not be essential in what we have to say. Second, we integrate
out φ classically: We ignore the (∂φ)2 term since we are at low energies and we
extremise the relevant part of L with respect to φ:

δ 1 y
− M 2 φ 2 + yφψψ = 0 ⇒ φ= ψψ . (1.27)
δφ 2 M2
Inserting this back into our lagrangian we obtain
y2
LEF T = ψi ∂/ψ − mψψ + (ψψ)2 + · · · . (1.28)
2M 2
Finally, we calculate loop corrections involving the heavy field φ to all operators that
potentially appear in LEF T . In this last step, the correction which is most critical
for us is the mass (or more generally the self-energy correction) for ψ, cf. Fig. 1.1.
Dropping all numerical prefactors, this gives (for details see e.g. [14])

2 −k/ + m
/) ∼ y
(p d4 k . (1.29)
(k 2 + m2 )[(k+ p)2 + M 2 ]
After summing, in the standard way, all such self-energy corrections to the
propagator, one obtains
i
. (1.30)
p
/ − m − (p
/)
This resummed propagator can be viewed as a function of the matrix-valued

/ . Its pole then determines the corrected mass mc = m + δm. This can be
argument p
made explicit by Taylor expanding (p / = mc :
/ ) around p
1
/ ) = (mc ) + (mc )(p
/ − mc ) + (m)(p
/ − mc ) + · · · .
2
(p (1.31)
2
Now the propagator takes the form
i
with mc = m + (mc ) . (1.32)
/ − mc ) − (mc )(p
(p / − mc ) + · · ·
Fig. 1.1 One-loop fermion

self-energy in the Yukawa
theory
As usual in perturbation theory we estimate
δm = (m) . (1.33)
/ = m. Introducing a cutoff , we have in

Thus, we need to evaluate (1.29) for p
total three scales: m, M and . We see right away that the (naively possible) linear
divergence arising from the term ∼ k/ will vanish on symmetry grounds as long as
our cutoff respects Lorentz symmetry. Moreover, the contribution from the term ∼ k/
in the regime k ∼ m M is suppressed by 1/M 2 . Thus, we may disregard the term
∼ k/ altogether.
We may then focus on the term ∼ m. It gets a small contribution from the
momentum region k M and is log-divergent for k M. We can finally
conclude that the leading result for the mass correction extracted from (1.29) must
be proportional to m. Any enhancement beyond this can at best be logarithmic, but
still proportional to m. For the moment this is all we need: We learn that (1.28) is
the right lagrangian after the replacement
m → mEF T ≡ mc ≡ m + δm ≡ m(1 + y 2 × O(1)) . (1.34)
Here ‘O(1)’ may include a logarithmic cutoff dependence, like in particular

ln(/M). Moreover, as noted earlier, we may define our theory at a finite scale
μ1 M. Then the log-divergence is traded for ln(μ1 /M).
We could have argued the same without even drawing any Feynman diagram:
Indeed, writing our model in terms of left- and right-handed spinors,
ψψ = ψ L ψR + ψ R ψL , (1.35)
one sees immediately that for m = 0 it possesses the Z2 symmetry
ψL → ψL , ψR → −ψR , φ → −φ . (1.36)
The mass term ∼ m breaks this symmetry. Thus, we expect that both the UV theory
and the EFT regain this symmetry in the limit m → 0. The loop correction δm of
mEF T = m + δm (1.37)
must hence itself be proportional to m. The punchline is that integrating out φ does
not clash with the lightness (or masslessness) of ψ.
It is interesting and important to develop this language further by considering the
low-energy EFT of the Standard Model below the scale of Higgs, W and Z-bosons
or the pion EFT below the confinement scale QCD . We leave it to the reader to
explore this using the vast literature.
1.5 The Standard Model as an Effective Field Theory 13
1.5 The Standard Model as an Effective Field Theory
Let us now apply the above language to the Standard Model. We first assume that
a finite cutoff TeV is present and that the Standard Model is the effective
theory valid below this cutoff. At the moment, we allow this cutoff to either be the
scale at which the framework of QFT becomes insufficient (string or some other
fundamental cutoff scale) or, alternatively, the scale at which the Standard Model is
replaced by a different, more fundamental, ultraviolet QFT. It is natural to view
as our main dimensionful parameter and organise the lagrangian as
L = L2 + L4 + L5 + L6 + · · · (1.38)
= c0 2 ||2 − |D|2 − λH ||4 + L 4 + L5 + L6 + · · · (1.39)
Here, in the first line, we have organised the lagrangian in groups of dimension-2,
dimension-4 (and so on) operators. In the second line, we have displayed L2 , which
is just the Higgs mass term, and the rest of the Higgs lagrangian explicitly. Thus, L 4
is our familiar renormalisable Standard Model lagrangian without the Higgs part.
We also have m2H = c0 2 and we note that |c0 | 1 is necessary for the Higgs
to be a dynamical field below the cutoff scale. But our discussion in the previous
section has not lead to an unambiguous conclusion about whether this should be
viewed as a problem.
Since we now think of the Standard Model as of an EFT, we included terms
of mass dimension 5, mass dimension 6, and so on. It turns out that, at mass
dimension 5, the allowed operator is essentially unique (up to the flavour structure).
It is known as the Weinberg operator. We write it down for the case of a single
family and using a two-component (Weyl) spinor notation (cf. Problem 1.9.4). The
l.h. 4-component lepton-spinor lL then takes the form

lα
lL = , with α = 1, 2 . (1.40)
0
The Weinberg operator reads

c c
L5 = (l · )2 + h.c. = liα lαj
ik
j l k l + h.c. (1.41)

Here we used the fact that two Weyl spinors can form a Lorentz invariant as
ψ α ψα =
αβ ψβ ψα , (1.42)
where the
tensor appears in its role as an invariant tensor of the Lorentz group
SL(2, C). By contrast, the
-tensors in (1.41) appear in their role as invariant tensors
of the SU (2) factor in GSM and allow us to combine two doublets (Higgs and
leptons) into a singlet.
Now, since

0 να
= and lα = , (1.43)
v eα
the low-energy effect of the above operator is to give mass to the upper component
of the lepton doublet, i.e. to the neutrino:
cv 2 α
L5 = ν να + h.c. (1.44)

Writing the neutrino as a Majorana rather than a Weyl fermion, this becomes the
familiar Majorana mass term. Introducing three families, the constant c is promoted
to a 3 × 3 matrix cab .
Given our knowledge that neutrino masses are non-zero and (without going into
the non-trivial details of the experimental situation) are of the order mν ∼ 0.1 eV,
an effective field theorist can interpret the situation as follows: The neutrino mass
measurements represent the detection of the first higher-dimension operator of the
Standard Model as an EFT. As such, they determine the scale via the relation
(assuming c = O(1))
mν ∼ v 2 / ⇒ ∼ 3 × 1014 GeV . (1.45)
On the one hand, this is discouragingly high. On the other hand, it is significantly
below the (reduced) Planck scale of MP 2.4 × 1018 GeV. It is also relatively
close to, though still significantly below, the supersymmetric Grand Unification
scale MGU T ∼ 1016 GeV to which will return later. Let us note that, without
supersymmetry, the GUT scale is less precisely defined and one may argue that the
UV scale derived from the Weinberg operator above is actually intriguingly close to
such a more general GUT scale.
It is very remarkable that the Standard Model with the Weinberg operator allows
for a simple UV completion at the scale . This so-called seesaw mechanism [33–
35] involves (we discuss the one-generation case for simplicity) the addition of just
a single massive fermion, uncharged under GSM . The relevant part of the high-scale
lagrangian is (in Weyl notation for spinors)
1
L ⊃ βlνR − MνR νR + h.c. (1.46)
2
Integrating out the extra fermion (often referred to as the right-handed neutrino νR ),
one obtains precisely the previously given Weinberg operator with
c ∼ β2 and ∼M. (1.47)

1.5 The Standard Model as an Effective Field Theory 15
In other words, the observed neutrino masses behave as
mν ∼ β 2 v 2 /M . (1.48)
As a result, we can make M (and thus ) smaller, bringing it closer to experimental

tests, at the expense of also lowering β. Of course, one has to be lucky to actually
discover νR at colliders, given that then β would have to take the rather extreme
value of 100 GeV/1014 GeV ∼ 10−6 .
An even more extreme option, which however has its own structural appeal, is
to set M to zero. This can be justified, e.g., by declaring lepton number to be a
good, global symmetry of the Standard Model (extended by r.h. neutrinos). By this
we mean the U (1) symmetry l → eiχ l, νR → e−iχ νR . Now the Standard Model
has an extra field, the fermionic singlet νR (more precisely three copies of it). The
first term in (1.46) is just another Yukawa coupling (given here in Weyl notation,
but otherwise completely analogous to e.g. the electron Yukawa term). The second
term is missing. This version of the Standard Model, extended by r.h. neutrinos, is
again a renormalisable theory and it can account for the observed neutrino masses.
The latter do not arise from the seesaw mechanism sketched above, but correspond
simply to a tiny new Yukawa coupling. In this case β ∼ mν /v ∼ 10−12, which may
be perceived as uncomfortably small. The second smallest coupling would be that
of the electron, βe ∼ 0.5 MeV /v ∼ 10−5 .
At mass dimension 6, there are many further terms that can be added to LSM .
For example, any term of L4 can simply be multiplied by ||2 . The arguably most
interesting terms are the 4-fermion-operators. They include terms like (now again in
Dirac notation)

L6 ⊃ cij kl (ψ i ψj )(ψ k ψl ) (1.49)
ij kl
as well as similar operators involving gamma matrices. Even with the restriction
by gauge invariance, there are many such terms and we will not discuss them in
any detail. Crucially, many of them are very strongly constrained experimentally.
First, if one does not impose the global symmetries of lepton and baryon number,
some of these operators induce proton decay. (We recall that with baryon number
we refer to a U (1) symmetry acting on quarks, with a prefactor 1/3 in the exponent.)
The extraordinary stability of the proton would then push up beyond 1016 GeV.
But even imposing baryon and lepton number as additional selection rules9
for (1.49), strong constraints remain. These are mostly due to so-called flavour-
9 To be precise, the two corresponding U (1) symmetries, known as U (1)B and U (1)L are so-
called accidental symmetries of the Standard Model. This means that, given just gauge symmetry
and particle content, and writing down allowed renormalisable operators, these symmetries are
automatically preserved at the classical level. It is hence not unreasonable to assume that they hold
also in certain UV completions and may constrain 4-fermion operators.
changing neutral currents or FCNCs10 (the analogues of the flavour-changing

charged currents mentioned earlier) and to lepton flavour violation (e.g. the decay
μ− → e+ + 2e− ). Such constraints push to roughly 103 TeV. Of course, the new-
physics scale can be much lower if the relevant new physics has the right ‘flavour
properties’ not to clash with data.
1.6 The Electroweak Hierarchy Problem
Now we come in more detail to what is widely considered the main problem of the
Standard Model as an effective theory: the smallness of the Higgs mass term. So far,
we have only pointed out that, in the EFT approach with cutoff , it is natural to
write
m2H ∼ c0 2 . (1.50)
We have many reasons to think that is large compared to the weak scale, implying
|c0 | 1. The main question hence appears to be whether we can invent a more
fundamental theory at scale in which |c0 | 1 can be understood.
Let us first give a very simple argument (though possibly not very strong) why
this is not easy. Namely, consider the theory as given by a classical lagrangian at
and ask for low-energy observables. The most obvious is maybe a gauge coupling,

bi
αi−1 (μ) αi−1 () + ln + O(1) , (1.51)
2π μ
where we restricted attention to the one-loop level. The relevant diagrams are just
the self-energy diagrams of the corresponding gauge boson with scalars, fermions
and (in the non-abelian case) gauge bosons running in the loop. We see that, for
μ, the correction becomes large, but it grows only logarithmically. This goes
together with the logarithmic divergence of the relevant diagrams, which is in turn
related to the vanishing mass dimension of the coupling or operator coefficient that
we are correcting. By contrast, for the Higgs mass we find [36, 37]
cH
m2H (μ) = m2H () + 2 + O(0 ) , (1.52)
16π 2
with cH a coupling-dependent dimensionless parameter to be extracted from
diagrams like those in Fig. 1.2. We see that, suppressing O(1) coefficients and
10 The name characterises processes which change flavour and have a structure that could arise from
integrating our a neutral gauge boson, like the γ or Z. It is an important fact that, in the Standard
Model, such processes or the corresponding 4-fermion-operators are extremely suppressed. They
are hence an important signal of new physics.
1.6 The Electroweak Hierarchy Problem 17
Fig. 1.2 Contributions to the

Higgs self-energy
disregarding the logarithmic running of the dimensionless couplings between μ and

, we have cH = λH + λ2t + g22 + · · · .
Thus, cH is an O(1) number and if we set = 1 TeV, only an O(1) cancellation
between the two terms on the r.h. side of (1.52) is required to get the right Higgs
mass parameter of the order of (100 GeV)2 . Things are actually a bit worse since
there is a colour factor of 3 coming with the top and other numerical factors. But,
much more importantly, we can not simply declare 1 TeV to be the scale where our
weakly coupled QFT breaks down and some totally unknown new physics (discrete
spacetime, string theory, etc.) sets in. One but not the only reason is the issue
of flavour-changing neutral currents mentioned above. If we take the (still rather
optimistic) value ∼ 10 TeV, we already require a compensation at the level of
1% or less between the two leading terms on the r.h. side of (1.52). This starts to
deserve the name fine-tuning or hierarchy problem.
A cautionary remark concerning expressions like m2H (μ) or m2H () is in
order. Such dimensionful parameters sometimes (not always) have power-divergent
loop corrections. The momentum integral implicit in the loop correction is then
dominated in the UV and changes by an O(1) factor if the regularisation procedure
changes. This is in contrast to e.g. α −1 (μ) which is, at leading order, independent
of how precisely the scale μ is defined. One can see that most easily by noting
that ln(/μ) does not change significantly in the regime /μ 1 if or μ are
multiplied by, say, a factor of 2. Thus, a possibly less misleading way to write (1.52)
is
cH
m2H = m2H, 0 + 2 + O(0 ) . (1.53)
16π 2
Here m2H is, by definition, the value of this operator in the IR and m2H, 0 is the bare
or classical value in the UV lagrangian.
Still, the fine-tuning argument is not very convincing since, in (1.53), the two
crucial terms between which a cancellation is required both depend on the cutoff
or regularisation used. For example, in dimensional regularisation with minimal
subtraction, the second term is simply zero and no cancellation appears necessary.
Now this is clearly unphysical, but one may entertain the hope that some physical
cutoff with similar features will eventually be established, defining a UV theory with
a ‘naturally’ small m2H in spite of large .
But a much more technical and stronger argument making the fine-tuning explicit
can be given. We make it using a toy model, but the relevance to the Standard
Model will be apparent. The toy model is in essence something like ‘the inverse’
of the Yukawa model of (1.26). There, we considered the mass correction a fermion
obtains when a heavy scalar is integrated out. We found that no large correction to
the small fermion mass arises. Now consider (again following [29]),
1 1 λ
L = − (∂φ)2 − m2 φ 2 − φ 4 + ψ(i ∂/ − M)ψ + yφψψ . (1.54)
2 2 4!
We literally simply renamed m ↔ M, having of course in mind that now m M.
As before, we will not go through a careful procedure of ‘running and matching’ to
derive the low-energy EFT, but take the shortcut of integrating out the heavy field
classically and adding loop corrections to the low-energy lagrangian terms.
Since the fermion appears only quadratically in the action, its equations of motion
are solved by ψ = 0 for any field configuration φ(x). Hence, the first step consists
in just dropping all terms with ψ. When considering loops, we focus only on
corrections to the scalar mass proportional to y 2 , finding
1 1 λ
LEF T = − (∂φ)2 − m2EF T φ 2 − φ 4 + · · · , (1.55)
2 2 4!
with
2
y2 tr(k/ − M)2
m2EF T m2 + k 2 d(k 2 ) . (1.56)
16π 2 0 (k 2 + M 2 )2
This integral corresponds to the second diagrams of Fig. 1.2. It is immediately clear
that both terms proportional to 2 as well as to M 2 will arise:
y2
m2EF T m2 + c 1 2
+ c 2 M 2
ln(2
/M 2
) + c 3 M 2
+ · · · . (1.57)
16π 2
See e.g. [14] for a corresponding analysis in dimensional regularisation. (Note that,
while a quadratic divergence in 4 dimensions does not show up as a pole at d = 4,
it corresponds to a logarithmic divergence in 2 dimensions and hence shows up as a
pole at d = 2.)
Crucially, we now see that if, by some ‘UV miracle’, the m2 and 2 terms
always cancel to make m2EF T very small, the tuning issue still remains: Even a
very tiny relative change of M 2 (assuming that M 2 m2EF T ), would upset this
cancellation. Of course, we cannot rule out a UV model where everything, including
masses of particles at intermediate scale (like our M with mEF T M ) are
automatically correctly adjusted to ensure the necessary cancellation in (1.57). But
now it becomes more apparent how tricky any mechanism accomplishing that would
have to be.
Concretely in the Standard Model with a seesaw mechanism for neutrino masses,
the scale M might be that of the heavy r.h. neutrino and one has, given the above,
a strong argument for fine-tuning. Alternatively, one can of course avoid any such
heavy particles (also giving up on Grand Unification—see below) and imagine that
the Standard Model directly runs into a new theory in the UV where, at some scale
1.7 Fine-tuning 19
, a massless scalar is explained without tuning. I am not aware of any sufficiently

concrete and convincing scenario of this sort. Nevertheless, we will return to a more
detailed discussion of this and related logical possibilities in Chap. 8.
For now, let us accept that, from an EFT perspective, the Standard Model with
UV scale is fine-tuned and try to quantify the problem.
1.7 Fine-tuning
Let us first emphasise that, having a small (dimensionful or dimensionless) param-

eter in an EFT is not in itself problematic or related to tuning. Indeed, the electron
mass is small, but it comes from a dimensionless Yukawa coupling which only runs
logarithmically. Thus, once small in the UV, it will stay small in the IR ‘naturally’.
Moreover, the relevant coupling of type
λe l L eR (1.58)
is forbidden by chiral symmetry transformations, e.g. eR → eiα eR . One can view

λe as a small effect in the UV lagrangian breaking this symmetry. Hence, the above
operator will only receive loop corrections proportional to this symmetry breaking
effect, i.e. to λe itself.
The same argument can be made for fermion masses even when they are viewed
as dimensionful parameters. We have seen one example in Sect. 1.3. Another
example is the Standard Model below the electroweak symmetry breaking scale,
where the electron mass term
me ee = me eL eR + h.c. (1.59)
can be forbidden by the global U (1) symmetry eR → eiα eR , as above. Hence, there
will be no loop corrections driving me up to the electroweak scale, given that the
tree-level value is small.
Small parameters with this feature are called ‘technically natural’, a notion due
to ’t Hooft [38]. More precisely, a small parameter is technically natural if, by
setting it to zero, the symmetry of the system is enhanced. The crucial point for us
is that finding such a symmetry for the Higgs mass term turns out to be difficult
if not impossible: One obvious candidate is a shift symmetry, → + α, with
α = const. But this forbids all non-derivative couplings and hence clashes with
the main role the Higgs plays in the Standard Model, most notably with the top-
Yukawa coupling, which is O(1). Nevertheless, attempts to at least alleviate the
hierarchy problem using this idea have been made and we will discuss them in
Sect. 8.2. Another option is scale invariance but, once again, the Standard Model as
a quantum theory is not scale invariant—couplings run very significantly. Moreover,
in the UV, most ideas for how the unification with gravity will work break scale
invariance completely. Again, attempts along these lines nevertheless exist and will
be mentioned. However, at our present ‘leading order’ level of discussion it is fair

to say that the smallness of m2H is probably not technically natural.
Somewhat more vaguely, one may say that the Higgs mass term is unnaturally
small. To make this statement more precise, the notion of tuning or fine-tuning
has been introduced. Roughly speaking, a theory is tuned if the parameters in the
UV theory (at the scale ) have to be adjusted very finely to realise the observed
low-energy EFT.
It is not immediately obvious how to implement this in terms of formulae since,
as just explained, e.g. the electron mass is known with high accuracy and even a
tiny change of the UV-scale Yukawa coupling will lead to drastic disagreement
with experiment. The main point one wants to make is that, as we have seen, the
smallness of the Higgs mass apparently arises from the compensation between two
terms,
cH
m2H = m2H, 0 + 2 + · · · . (1.60)
16π 2
Clearly, in such a situation, a small relative change, e.g., m2H, 0 induces a much larger
relative change of m2H .
A widely used formula implementing this is known as the Barbieri–Giudice
measure for fine-tuning [39] (see also [40] and, for modern perspectives and
further references, [41, 42]):

x ∂O ∂ ln(O)
F T = = . (1.61)
O ∂x ∂ ln(x)
Here x is the theory parameter and O the relevant observable. In our case, x = m2H, 0
and O = m2H is given by (1.60), such that
∂ ln(m2H, 0 + cH 2 /(16π 2 )) m2H, 0 2 /(16π 2 )

F T = m2H, 0 = ∼ .
∂m2H, 0 m2H, 0 + cH 2 /(16π 2 )) m2H
(1.62)
Here, in the last step, we assumed that m2H cH 2 /(16π 2 ), such that m2H, 0 ∼
cH 2 /(16π 2 ). Moreover, we have used that cH = O(1). As already noted earlier,
this just formalises what we said at the intuitive level earlier: The fine-tuning is
roughly 2 /(1 TeV)2 .
For completeness, we record the natural multi-particle generalisation of the
Barbieri–Giudice measure. In this more general context, one may call it a ‘fine-
tuning functional’, defined as a functional on the space of theories T (following
[41]):

xi ∂Oj
F T [T ] = (1.63)
O j ∂xi .
ij
1.7 Fine-tuning 21
We also note that our discussion was somewhat oversimplified and less concrete
than in [39]. There, the observable was m2Z (this is clearly tied to m2H , which is
however not directly observable). Furthermore, the UV theory was not some rather
vague cutoff-QFT, but it was a concrete model: The supersymmetric, in fact even
supergravity-extended version of the Standard Model. We will get at least a glimpse
of this below, after we have introduced supersymmetry.
Unfortunately, the above definition of fine-tuning has many problems. First, it is
clearly not reparameterisation independent. In other words, it crucially depends on
our ad hoc choice of xi as operator coefficients in a perturbative QFT and of the
Oi as, roughly speaking, particle masses. Thus, one is justified in looking for other,
possibly related, definitions. One such alternative definition is probabilistic: Choose
a (probability) measure on the set of UV theories and ask how likely it is to find
a particular low-energy observable to lie in a certain range. For example, we might
consider m2H, 0 to have a flat distribution between zero and 22 /(16π 2 ) (where we
also set cH = −1). Then we obtain a small Higgs mass only if m2H, 0 happens, by
chance, to fall very close to the center point of its allowed range.
To make this quantitative, we need to distinguish notationally between the Higgs
mass parameter m2H in our statistical set of theories and the concrete Standard
Model value of this quantity. Let us call the latter m2H, obs.. Then, we may ask
for the probability to find m2H in the interval [−m2H, obs., m2H, obs.], or equivalently
|m2H | m2H, obs.. We obtain
m2H, obs.
p(m2H, obs.) . (1.64)
2 /(16π 2 )
This is just the inverse of the Barbieri–Giudice fine-tuning value, confirming at least
at some intuitive level that the above definitions make sense. However, it becomes
even more apparent that some ad hoc assumptions have come in. In particular, we
required a measure on the space of UV theories or UV parameters.
Finally, another ambiguity of the probabilistic view on fine-tuning is related to
the choice of the allowed interval of the EFT observable. In the above, things were
rather clear since our task was to quantify the problematic smallness of the Higgs
mass relative to the cutoff. It was then natural to define all theories with |m2H |
m2H, obs. as ‘successful’. However, the Higgs mass is by now known rather precisely,
mh = 125.18 ± 0.16 GeV [43], which translates into a similarly small allowed
interval for m2H, obs.. If we had defined successful theories as those with m2H, obs.
falling into that interval, a much larger fine-tuning would result. Even worse, one
could consider the very precisely known electron mass in the same way and would
find a huge fine-tuning of the UV-scale coupling λe , in spite of the logarithmic
running and the technical naturalness.11
11 This issue can be overcome by using Bayesian inference (see e.g. [44, 45]). Here, one ascribes
certain prior probabilities to different models, each in turn predicting a certain distribution of
Thus, one has to be careful with both definitions and it may well be that the
final word about this has not yet been spoken. A suggestion for sharpening the
probabilistic perspective is as follows: Consider the manifold of UV couplings
(with some measure) and the map to the manifold of observables. On the latter,
let O0 be some qualitatively distinguished point, in our case the point of vanishing
Higgs mass term. This point is distinguished since it specifies the boundary between
two qualitatively different regimes—that with spontaneously broken and unbroken
SU (2) gauge symmetry. Let us assume that for any other point O, we can in some
way measure the distance to this special point, |O − O0 |. Now one may say that
an observed EFT, corresponding to a point Oobs. on the manifold of observables, is
fine-tuned to the extent that the probability for all theories with
|O − O0 | < |Oobs. − O0 | (1.65)
is small. In other words, we measure how unlikely it is that a randomly chosen

theory falls more closely to the special point O0 than our observed EFT.
Before closing this section, let us introduce some terminology that might be
useful to sharpen our understanding: Simply the fact that the weak scale mew
is so much smaller than the Planck scale MP may be perceived as requiring
explanation. It is common to characterise this as the large hierarchy problem. This
problem of very different energy scales in the same theory becomes an actual fine-
tuning problem if the Standard Model (or some minimal extension not affecting
the quadratic Higgs mass divergence) is valid up to MP . However, we only have
strong experimental evidence that the cutoff of the Standard Model as an EFT is
above ∼ 10 TeV. It could even be lower, but this leads to various phenomenological
problems with flavour violation and precision data. Taking such a low cutoff , one
is only faced with a mild fine-tuning ∼ m2H /(2 /16π 2 ) ∼ 10−2 for ∼ 10 TeV.
This is known as the little hierarchy problem, which is on the one hand much less
severe but on the other hand much more concrete and data-driven.
We close by recommending to the reader the lecture notes [50], which discuss
Higgs boson physics in much more detail, emphasising in particular the electroweak
hierarchy problem. See also [51] for brief, less technical discussions of the concept
of Naturalness.
Higgs masses. One may then use Bayes’ theorem to derive the posterior (i.e. after the Higgs mass
has been measured) probability of each model. The models in question may for example be the
Standard Model, with fine-tuned Higgs mass as above, and a supersymmetric model (see below)
in which the Higgs mass is naturally small. The result will of course be a much higher probability
for the ‘natural’ model, see e.g. [46–49]. Crucially, the ratio of probabilities will automatically be
insensitive to the precision with which the Higgs mass has been measured.
1.8 Gravity and the Cosmological Constant Problem 23
1.8 Gravity and the Cosmological Constant Problem
Including gravity in a minimalist approach amounts to the substitution
√ 1 √ √
LSM [ψ, ημν ] → g LSM [ψ, gμν ] + MP2 g R[gμν ] − g λ .
2
(1.66)
As a result, two essential modifications of the discussion above arise: First, we learn
that the Higgs mass problem is just one of two instances of very similar hierarchy
problems—the other being the cosmological constant problem. Second, gravity sets
an upper bound on the cutoff , in a way that sharpens the Higgs mass hierarchy
problem.
In more detail, let us start by recalling what we need to know about gravity [2,52–
55]. On the one hand, gravity changes the picture very deeply: The arena of our
Standard Model QFT changes from R4 (with flat Lorentzian metric) to a Lorentzian
manifold with dynamical metric, horizons, singularities in the cosmic past or future,
or possibly even with topology change. The causal structure, which is so crucial for
the definition of a QFT, becomes dynamical together with the metric. In particular,
if one takes the metric itself to be a dynamical quantum field, the quantisation of
this field depends on the causal structure, which follows from the (then a priori
unknown) dynamics of this field itself. Diffeomorphism invariance makes it very
hard to define what a local observable in the usual QFT sense is supposed to be.
Finally, to just mention one more issue, QFTs are most easily defined in euclidean
metric. But this is extremely problematic in gravity since, even for a topologically
trivial 4d euclidean manifold, the local value of R[gμν ] can take either sign. Thus,
fluctuations around a flat euclidean background do not necessarily suppress the
weight factor exp(−SE ) in the path integral. As a result, the whole euclidean
approach may become problematic [56].
But, on the other hand, one may also ignore most of the deep conceptual
problems above and pretend that one has added to the Standard Model QFT just
another gauge theory (see e.g. [57]). We cannot develop this approach here in any
detail but only sketch the results: One expands the metric around flat space,
gμν = ημν + hμν (1.67)
and tries to think of hμν as of a gauge potential, analogous to Aμ . The reader will
recall the covariant derivative acting on a vector field,
1 ρσ
Dμ vν = ∂μ vν − μν ρ vρ with μν ρ = g (∂μ gνσ + ∂ν gμσ − ∂σ gμν )
2
(1.68)
and the curvature tensor written as a commutator, in analogy to Fμν :
Rμνρ σ vσ = [Dμ , Dμ ] vρ , or symbolically: R ∼ [∂ − , ∂ − ] .

(1.69)
From this, it is clear that the gravitational lagrangian takes the symbolic form
(ignoring index structure and numerical factors)
MP2 [ h∂ 2 h + h(∂h)2 + h2 (∂h)2 + · · · ] . (1.70)
Defining κ ≡ 1/MP and rescaling h → κh, this becomes
h∂ 2 h + κh(∂h)2 + κ 2 h2 (∂h)2 + · · · . (1.71)
This is already quite analogous to the gauge theory structure (we are thinking of the
non-abelian case, but are suppressing the group and Lorentz indices for brevity)
A∂ 2 A + gA2 ∂A + g 2 A4 . (1.72)
The crucial differences are that g is dimensionless and the series of higher terms
terminates at the quartic vertex. By contrast, in gravity the coupling has mass
dimension −1 and the series goes on to all orders (both from R as well as from the
R 2 , R 3 terms, etc. which have to be added to the lagrangian to absorb all divergences
arising at loop level). We will not discuss the technicalities of this—suffice it to say
that the Faddeev–Popov procedure and the introduction of ghosts work, at least in
principle, as in gauge theories.
We also recall that, for any observable that we can calculate in perturbation
theory, the expansion reads
c0 + c1 κ + c2 κ 2 2 + · · · (1.73)
on dimensional grounds. From this we see that we have to expect power divergences
and that higher loops are more and more divergent, consistent with the well-known
fact that quantum gravity is perturbatively non-renormalisable.
Finally, coming closer to our main point, we remember that gμν or, in our
approach, hμν appears also in LSM [ψ, gμν ]. Since, as we know, the energy-
momentum tensor is defined as the variation of SSM with respect to gμν at the point
gμν = ημν , it is clear that the leading order coupling of h with matter is given by
L ⊃ κhμν T μν . (1.74)
This is, once again, completely analogous to the gauge theory coupling to matter via
L ⊃ gAaμ jaμ with, e.g. jaμ = ψγ μ Ta ψ . (1.75)

1.8 Gravity and the Cosmological Constant Problem 25
What is essential for us is that the cosmological constant term gives rise to an
energy-momentum tensor
T μν = −ημν λ . (1.76)
If λ is non-zero, then this corresponds to a non-zero source or ‘tadpole term’ for the
metric (gauge) field hμν :
L ⊃ −κhμν ημν λ . (1.77)
The meaning of the word tadpole in this context becomes obvious if one considers
the above as a tree-level diagrammatic effect and adds the first loop correction (due
e.g. to a scalar particle minimally coupled to gravity). This is illustrated in Fig. 1.3.
One may think of the loop diagram in Fig. 1.3 as a correction to λ, in direct
analogy to the loop corrections to the Higgs mass from integrating out heavy
particles which we discussed before. Thus, in analogy to e.g. (1.53) and renaming
our original cosmological term in the tree-level action to λ0 , we have
cλ
λ = λ0 + 4 . (1.78)
16π 2
The coefficient cλ does not include a small coupling constant but is merely
proportional to the sum of bosonic and fermionic degrees of freedom (one may
interpret this sign difference either as being due to the usual ‘minus’ for each
fermion loop or as the negative sign of the vacuum energy of the fermionic harmonic
oscillator). Related to this, one can of course interpret the divergence as a sum over
the vacuum energies of the oscillators corresponding to free field momentum modes.
Famously, if one compares the observed value of the vacuum energy,
λ (2.2 meV)4 , (1.79)
with the expectation from (1.78) based on = MP 2.4 × 1018 GeV, one
finds a mismatch (i.e. a required fine-tuning) of 10120. As in the Higgs mass
case, there are caveats to this argument: Indeed, the value of the loop correction
depends completely on the UV regularisation and one may imagine schemes where
it would simply be zero. Also, as in Higgs case, there are counterarguments to this
suggestion. Indeed, any massive particle contributes to the loop correction in a way
that depends on its mass. Thus, the observed value changes dramatically if, e.g., the
mass of the heavy r.h. neutrino needed in the seesaw mechanism changes.
Fig. 1.3 Tree level and loop effect of the cosmological constant term on the metric field hμν
To see this more explicitly, it is useful to give an explicit covariant formula for
the one-loop correction to λ (see problems for a derivation). For a single real scalar
and in euclidean signature, one has

1 d 4k
δλ = ln(k 2 + m2 ) = c0 4 + c1 2 m2 + · · · . (1.80)
2 (2π)4
We see that even the sub-leading term proportional to the mass is still also
proportional to 2 and hence huge. In fact, this is true even for the light particles
of the Standard Model. Furthermore, there are effects due to the Higgs potential,
the non-perturbative gluon-condensate of QCD and from all couplings (which enter
at the two and higher-loop level). Thus, the case for an actual fine-tuning appears
to be very strong indeed. Clearly, the amount of fine-tuning may be significantly
reduced compared to what we just estimated: We could add to the Standard Model
heavy bosons and fermions, such that above a certain mass scale M the number of
fermions and bosons is equal and at least the leading 4 term disappears.
This last idea turns out to work much better than expected. It is realised in a
systematic way in supersymmetry (SUSY) or supergravity (SUGRA). It still does
not solve the cosmological constant problem, even in principle. The reason is that
the scale of supersymmetry breaking is much too high. It does, however, solve
the Higgs mass or electroweak hierarchy problem in principle. The fact that this
solution does not work (at least not very well) in practice is due to fairly recent data,
especially from the LHC. Nevertheless, it will be important for us to study SUSY
in general and to a certain extent the SUSY version of the Standard Model. The
reasons are twofold. First, as noted, SUSY is an excellent example for how things
could work out nicely at the cutoff scale . The precise understanding of how our
apparent fine-tunings could disappear or at least be mitigated will help us to evaluate
their technical content and physical meaning. Second, if one wants quantum gravity
divergences to also be tamed at the cutoff scale, SUSY is not enough and string
theory is required. But the relation of the latter to real-world physics relies (at least
in the best understood cases) on SUSY, which we hence have to understand at least
at an introductory level.
For reviews of the cosmological constant problem, see e.g. [58–61]. In particular,
the arguments of [58] against adjustment mechanisms for a zero cosmological
constant(sometimes referred to as Weinberg’s no-go theorem) are noteworthy.
1.9 Problems
1.9.1 Electroweak Symmetry Breaking
Task Calculate W and Z boson masses as well as the electromagnetic coupling e

in terms of v and g1,2 . Derive the formula for the electric charge Q = T3 + Y , where
1.9 Problems 27
Y is the U (1) hypercharge and T3 is the so-called isospin generator of the SU (2)
gauge group.
Hints Apply the covariant derivative (for uncoloured fields)
Dμ = ∂μ − ig2 Aaμ R(T a ) − ig1 R(Y )Bμ (1.81)
to the Higgs VEV to derive the mass terms for W ± and Z. Identify the massless
field (the linear combination orthogonal to the massive vectors) as the photon and
express the covariant derivative in terms of these fields.
Solution The Higgs transforms in the fundamental representation of SU (2), hence

R(T a ) = σ a /2. It has hypercharge
√ 1/2, hence R(Y ) = 1/2. It is convenient to work
with W ± = (A1 ∓ iA2 )/ 2. Then one has
A1 σ 1 + A2 σ 2 = (A1 + iA2 )(σ 1 − iσ 2 )/2 + (A1 − iA2)(σ 1 + iσ 2 )/2

√ √
= 2W − σ − + 2W + σ + , (1.82)
where

01 00
σ + = (σ 1 + iσ 2 )/2 = , σ − = (σ 1 − iσ 2 )/2 = . (1.83)
00 10
In the symmetry-broken vacuum, one then finds:

0
0 i √ + v 3
Dμ H = Dμ =− g2 2Wμ + −g2 Aμ + g1 Bμ .
v 2 0 v
(1.84)
This gives rise to the mass term
v2 2 + 2
L ⊃ −|Dμ H |2 = − 2g2 |Wμ | + (g12 + g22 )(Zμ )2 . (1.85)
4
We have to recall that W − = (W + )∗ and the complex W boson is normalised like

a complex scalar field, i.e. without a factor 1/2 in kinetic and mass term. Moreover,
we introduced the canonically normalised massive vector
1
Zμ = (g2 A3μ − g1 Bμ ) . (1.86)
g12 + g22
Thus, the mass term is
1
L ⊃ −m2W |Wμ+ |2 − m2Z (Bμ )2 , (1.87)
2
from which we read off

√ √
mW = g2 v/ 2 and mZ = g12 + g22 · v/ 2 . (1.88)
Next, we note that the linear combination of A3 and B orthogonal to Z is

1
Aμ = (g1 A3μ + g2 Bμ ) . (1.89)
g12 + g22
It is then immediate to express A3 and B through Z and A:

1 1
A3μ = (g1 Aμ + g2 Zμ ) and Bμ = (g2 Aμ − g1 Zμ ) .
g12 + g22 g12 + g22
(1.90)
Now the covariant derivative for a general field takes the form
√
Dμ = ∂μ − ig2 2(Wμ+ R(T + ) + Wμ− R(T − )) (1.91)
i ig1 g2
− Zμ (g22 R(T 3 ) − g12 R(Y )) − Aμ (R(T 3 ) + R(Y )) . (1.92)
g12 + g22 g12 + g22
It is clear that the transition between A3 , B and Z, A may be interpreted as an SO(2)

rotation. The corresponding angle is known as the weak mixing angle or Weinberg
angle θW . It is defined by
g1
sin θW = . (1.93)
g12 + g22
In terms of this angle, the electromagnetic charge (i.e. the prefactor of Aμ in the
covariant derivative) is given by e = g2 sin θW . The group-theoretic coefficient
is Q = R(T 3 ) + R(Y ). One often keeps the necessary use of the appropriate
representation implicit, writing simply Q = T3 + Y .
1.9.2 The Standard Model Is Anomaly Free
Task Confirm this statement.
Hints Famously, in a theory with a single l.h. fermion ψ (or, equivalently, a single
Weyl fermion), the anomalous current non-conservation for
jμ ≡ ψγμ ψ (1.94)
1.9 Problems 29
reads
1
∂μ j μ = − F F̃ . (1.95)
32π 2
A classical way to derive this is to consider the corresponding amplitude relation
1
p, k|∂μ j μ (0)|0 = − p, k|
ανβρ Fαν Fβρ (0)|0 . (1.96)
32π 2
Here p, k| stands for a final state with two outgoing gauge bosons with momenta p
and k. The l.h. side of this equality is evaluated according to the diagrams in Fig. 1.4,
and the r.h. side simply by expanding the fields in terms of creation and annihilation
operators.
Given this diagrammatic understanding, it is very easy to see what the right
generalisation to the non-abelian case is: At each vertex, the abelian gauge group
generator ‘1’ has to be replaced by the corresponding non-abelian generator (Ta )ij .
As a result, one has
1 1
∂μ jaμ = − Dabc
μνρσ Fμν
b c
Fρσ with Dabc ≡ tr[Ta {Tb , Tc }] . (1.97)
32π 2 2
It should now be clear how to proceed: Consider the Standard Model fermions
as one l.h. fermion field ψ transforming the appropriate (very large) representation
of GSM . Let Ta run over all 12 generators of this group. With this interpretation of
Ta and of the trace in (1.97), one only needs to check that Dabc = 0. A lot of this is
repetitive and can be simplified. For instance, the threefold repetition due to the three
generations can be dropped—even a single generation is anomaly free. Furthermore,
rather than thinking about a complicated block-diagonal Ta , one can just sum over
the different corresponding fermions in the loop. Finally, we clearly only need to
show that Dabc = 0 for all the 3 × 3 = 9 possible different assignments of a, b and
c to the factor groups SU (3), SU (2) and U (1). Which particular generator of e.g.
SU (3) one then chooses is immaterial. As a result, the amount of work is actually
rather limited.
Fig. 1.4 Scattering amplitude interpretation of the expectation value of the axial current. The
momentum q is related by Fourier transformation to the argument x of j μ (x). In (1.96), x has
been set to zero
Solution As explained above, we need to go through all possible ways of assigning

the three generators corresponding to the three vertices of the triangle to the factors
of GSM . Thus, symbolically, we have to consider
U (1)3 , U (1)2 SU (2) , U (1) SU (2)2 , U (1) SU (2) SU (3) , · · ·

(1.98)
and so on. But the generators of SU (N) groups are all traceless, such that e.g. in the
U (1)2 SU (2) case we have (for each fermion species or, equivalently, each block)
tr[TU2 (1)TSU (2) ] = tr[TU (1) ] tr[TSU (2)] = 0 .

A 2 A
(1.99)
Thus, we only need to consider combinations where all three generators come from
the same factor or where two come from the same factor and the third from the
U (1):
U (1)3 , U (1) SU (2)2 , U (1) SU (3)2 , SU (2)3 , SU (3)3 .

(1.100)
Now let us go through this case by case. In the first case, we simply have to
sum the cubes of the charges of all fermions. The anti-commutator is, of course,
irrelevant. Using the list at the beginning of Sect. 1.1 of the notes, this gives
3 3 3 3
3×2× 1
6 + 3 × − 23 + 3 × 13 + 2 × − 12 + (1)3
= 1
36 − 8
9 + 1
9 − 1
4 +1 = 0. (1.101)
Note that the SU (3) and SU (2) representations are only relevant to determine the
multiplicities corresponding to each set of fermions.
In the second case, the anti-commutator is again irrelevant. Indeed,
tr[TU (1){TSU
a
(2), TSU (2) }] = 2 tr[TU (1) ] tr[TSU (2) TSU (2) ] ,
b a b
(1.102)
and
(2){TU (1) , TSU (2) }] = 2 tr[TU (1) ] tr[TSU (2) TSU (2) ] .
a b a b
tr[TSU (1.103)
Since the SU (2)-trace always gives δ ab /2, we just need to sum the U (1) charges of
all SU (2) doublets:
1 1
3× +0+0− +0 = 0. (1.104)
6 2
The third case is analogous: We have to sum over the U (1) charges of all
SU (3) triplets. (It does not matter whether it is a triplet or anti-triplet since
1.9 Problems 31
(3)TSU (3) ] = δ /2 holds for both). This gives

a b ab
tr[TSU
1 2 1
2× − + +0+0 = 0. (1.105)
6 3 3
(2) = σ /2 and hence

a a
In the fourth case we have TSU
1 1
(2){TSU (2) , TSU (2) }] = tr[σ a {σ b , σ c }] = tr[σ a ] 2δ bc = 0 .
a b c
tr[TSU (1.106)
8 8
Thus, we see that any theory with only fundamental representations (the antifunda-
mental is equivalent to the fundamental) of SU (2) is trivially free of the triangle
anomaly. In fact, this extends to all representations of SU (2) due to the reality-
properties of its representations.
Finally, the fifth and last case is the only one where we need to take into
consideration that different representations of the same non-abelian group appear.
We write
(3), f und. = T3 (3), ant i−f und. = T3 .

a a b a
TSU and TSU (1.107)
Now, since for a fundamental field we have
→ exp(i
T ) and ∗ → exp(−i
T ∗ ) ∗ = exp(−i
T T ) ∗ ,
(1.108)
we can conclude that
T3a = −(T3a )T . (1.109)
As a result, we find
tr[T3a {T3b , T3c }] = tr[(−T3a )T {(−T3b )T , (−T3c )T }] = −tr[T3a {T3b , T3c }] . (1.110)
Thus, we have to add the SU (3)-triplets and subtract the anti-triplets, each with its
multiplicity:
2− 1−1 = 0. (1.111)
We finally note that triangle anomalies (as considered above) involving different
gauge group factors are called ‘mixed’. Without going into details, we also record
the fact that a so-called mixed U (1)-gravitational anomaly exists. It comes from
a triangle diagram involving one gauge boson and two gravitons. To allow for a
consistent coupling of the Standard Model to gravity, this anomaly also has to
vanish. The calculation is similar to the U (1)SU (2)2 and the U (1)SU (3)2 case.
Since all fermions couple to gravity in the same way, we simply have to add all
U (1) charges:
1 2 1 1
6× −3× +3× −2× + 1 = 0. (1.112)
6 3 3 2
1.9.3 The Standard Model and SU (5)
Task Embed GSM in a natural way in SU (5) and show that the matter content of
one generation (with all its gauge charges) follows from the 10 + 5 of SU (5), where
10 stands for the antisymmetric second rank tensor and 5 for the antifundamental
representation. Consider a situation where the Standard Model follows from such
an SU (5) gauge theory (a Grand Unified Theory or GUT) realised at some higher
energy scale. Derive the tree-level prediction for the relative strength of the three
Standard Model gauge couplings.
Hints The ‘natural embedding’ corresponds, of course, to identifying the upper-left

3 × 3 block of 5 × 5 SU (5) matrices with SU (3) and the lower-right 2 × 2 block
with SU (2). The inverse would be equivalent—this is merely a convention. Hence,
when viewed as generators of SU (5), the SU (3), SU (2) and U (1) generators are
⎛ ⎞
−2
⎜ −2 ⎟
a
TSU 03×2 03×3 03×2 1 ⎜⎜
⎟
⎟
(3) 3×3 , a , √ ⎜ −2 ⎟.
02×3 02×2 02×3 TSU (2) 2×2
60 ⎜
⎝ 3 ⎠
⎟
3
(1.113)
The prefactor of the U (1) generator ensures the standard non-abelian normalisation
tr(T a T b ) = δ ab /2. With this, it is immediate to write down the branching rule
5 = (3, 1)−2 + (1, 2)3 under SU (5) → SU (3) × SU (2) × U (1) .

(1.114)
Here we have rescaled the U (1) generator in an obvious way for notational
convenience. All one now needs to do is to infer the branching rules for the 5 and 10
and to determine the gauge couplings gi of the Standard Model in the normalisation
given in the lecture. (We note that, as is probably well-known, this unification
scheme cannot work without significant loop corrections—cf. Sect. 2.13.5)
A classical introduction to group theory for physicists is [62]. For an extensive
collection of group and representation theory data see [63]. The physics of Grand
1.9 Problems 33
Unification, which builds on the technical observation discussed in this problem, is

reviewed e.g. in [64–68].
Solution The branching rule for 5 follows trivially from complex conjugation of
the above:
5 = (3, 1)2 + (1, 2)−3 . (1.115)
Here we have used the fact that 2 = 2 for SU (2). This is obvious since
Lie(SU (2)) = Lie(SO(3)) and since, as derived in quantum mechanics, SO(3)
has a unique 2-dimensional representation. It can also be demonstrated explicitly by
showing that, if
ψi → Uij ψj , and ψi∗ →→ Uij∗ ψi∗ , (1.116)
then the field χi ≡

ij ψj∗ transforms exactly as ψi . We leave that to the reader.
Formally speaking, we are claiming that the two representations 2 and its
complex conjugate, 2, are equivalent. This implies an isomorphism between the two
vector spaces which commutes with the group action. In our case, the isomorphism
is the multiplication with
. We will see a less trivial example of this below, which
we will work out and after which it will be even more clear how to finish the SU (2)
discussion.
At this point, just looking at the pure SU (2) doublet (there is only one such field
in the Standard Model!), we can already identify the U (1) charges with those of the
Standard Model. We have the covariant derivative as it follows from the GUT:
Dμ = ∂μ − igTSU
a
(2) (A2 )μ − igYGU T (A1 )μ .
a
(1.117)
According to the above,

−3
YGU T = √ . (1.118)
60
On the Standard Model side, we have
Dμ = ∂μ − ig2 TSU
a
(2) (A2 )μ − igY Y (A1 )μ
a
(1.119)
with
Y = −1/2 (1.120)
for the pure doublet (the lepton doublet). Thus, we learn that
gY2 3
g YGU T = gY Y or 2
= . (1.121)
g 5
This is the famous normalisation change between the Standard Model hypercharge
U (1) and the SU (5)-normalised U (1). Note that we call the Standard Model gauge
couplings gY , g2 and g3 at this point since, very frequently, the√name g1 is reserved
for the hypercharge coupling in GUT normalisation, i.e. g1 = 5/3 gY .
We also see that the down-type r.h. quarks have the correct charge to be the
SU (3) anti-triplet coming with this SU (2) doublet. (Their hypercharge differs by a
factor −2/3, as it follows from SU (5).)
As for the numerical outcome, we have the GUT prediction that g1 = g2 = g3 at
the GUT scale. This has to be compared to the observed values of roughly
α1−1 60 , α2−1 30 , α3−1 8 (1.122)
at the scale mZ . Here the first two values follow from αem −1 127, e = g sin θ ,
√2 W
sin θW = gY /(gY + g2 ) and sin θW 0.23 together with g1 = 5/3gY as
2 2 2 2 2
explained above. Thus, as already noted, significant loop corrections (most plausibly
from running over a large energy range) are needed for this unification scheme to
work.
Finally, three Standard Model fields are missing and we hope to get them from
the 10. To check this, let us first write down the tensor product
5 × 5 = [(3, 1)−2 + (1, 2)3 ] × [(3, 1)−2 + (1, 2)3 ] (1.123)
and anti-symmetrise:
(5 × 5)A = ((3 × 3)A , 1)−4 + (1, (2 × 2)A )6 + (3, 2)1 . (1.124)
Here the last representation only appears once since the other, equivalent term
belongs to the symmetric part of the rank-2
√ tensor. In giving the U (1) charges we
have, as before, suppressed the factor 1/ 60. Comparing Eqs. (1.118) and (1.120),
we see that the Standard Model U (1)Y charges in this normalisation follow by
multiplication with a factor (−1/2)/(−3) = 1/6. Given that (2 × 2)A is clearly
a singlet, we recognise the middle term as the r.h. electron. The last term is clearly
the l.h. quark doublet. The first term should then be the r.h. up-type quark.
All we need to establish is that
(3 × 3)A = 3 . (1.125)
To do so, we first identify the vector spaces of antisymmetric SU (3) tensor and
(anti-)vector by
ψij =
ij k ψk . (1.126)
1.9 Problems 35
Fig. 1.5 Commuting

diagram demonstrating the
equivalence of two
representations
Then we just need to show that they transform consistently, i.e., that the diagram in
Fig. 1.5 commutes. This implies
∗
Uik Uj l
klm ψm =
ij k Ukm ψm . (1.127)
To verify this equality, we remove ψm and multiply by (U T )mn :

∗
Uik Uj l Unm
klm =
ij k Ukm (U T )mn . (1.128)
On the r.h. side the two mutually inverse matrices cancel; the l.h. side is just the
epsilon tensor multiplied by the determinant of U , the latter being unity. Thus, we
are done.
1.9.4 Weyl Spinors
Tasks
(1) Define the canonical map SL(2, C) → SO(1, 3) using the vector of four sigma
matrices σμ = (1, σ1 , σ2 , σ3 ). Then go on to show that (σμ )α α̇ is an invariant
tensor of the Lorentz group. Build the Dirac spinor and gamma matrices from
Weyl spinors and sigma matrices and express the transformation of a Dirac
spinor under a Lorentz rotation in terms of a given SL(2, C) matrix M.
(2) Rewrite the Dirac spinor invariants
(1) (2) (1)† (2) (1) (2) (1) (2) (1) (2)
ψ D ψD ≡ ψD γ0 ψD , ψ D γ5 ψD , ψ D γμ ψD , ψ D γμ γ5 ψD
(1.129)
in terms of Weyl spinors. Use the upper/lower and lower/upper index summation
convention for undotted and dotted Weyl indices respectively:
ψχ ≡ ψ α χα , ψχ ≡ ψ α̇ χ α̇ . (1.130)
(3) Check the crucial identity
σ μ σν + σ ν σμ = −2ημν 1 (1.131)
and derive the Clifford algebra relation for the γ matrices from it.
Hints
(1) The first part is a direct generalisation of the construction of the map SU (2) →
SO(3) which should be familiar from quantum mechanics. The second step
is a straightforward calculation using only the fact that the indices α and
α̇ transform with SL(2, C) matrices and with complex conjugate SL(2, C)
matrices respectively. In the last step you need to use the convention that the
upper / lower two components of a Dirac spinor are given by a Weyl spinor with
lower undotted / upper dotted index.
A convenient set of conventions is that of the Appendix of the book by Wess
and Bagger [5], in particular

0 −1 0 1

αβ = ,
αβ = , such that
αβ
βγ = δα γ .
1 0 −1 0
(1.132)
This allows us to raise and lower Weyl indices with the

tensor. Of course one
needs to use the fact—please check if not obvious—that
is an invariant tensor
of SL(2, C.)
(2) This is completely straightforward. Deviating from the Wess–Bagger conven-
tions, it may be convenient to define γ 5 ∼ γ 0 γ 1 γ 2 γ 3 with a prefactor which
ensures that the l.h. projector (i.e. the projector on the undotted Weyl spinor) is
PL = (1 − γ 5 )/2.
(3) Use that
αβ = −i(σ2 )αβ together with the familiar commutation relations of
the Pauli matrices.
Solution
(1) Given a 4-vector v, define v̂ ≡ v μ σμ . The matrix v̂ is hermitian, as is the matrix
v̂ = M v̂M † , (1.133)
where M ∈ SL(2, C). Since {σμ } is a basis of hermitian 2 × 2 matrices, there

exists a unique decomposition
v̂ = v μ σμ , (1.134)
1.9 Problems 37
which defines the SL(2, C)-transformed vector v . To see that this an SO(1, 3)
transformation, it suffices to check that v 2 is preserved. This follows immedi-
ately from

v 0 + v 3 v 1 − iv 2
det v̂ = = (v 0 )2 − v 2 = −v 2 (1.135)
v 1 + iv2 v 0 − v 3
together with the obvious fact that the SL(2, C) transformation (1.133) pre-
serves the determinant.
With this, we are ready to check that (σμ )α α̇ is an invariant tensor. To do so,
let M ∈ SL(2, C) and let ∈ SO(1, 3) be its image under the map defined
above. We have
μ ν Mα β M α̇ β̇ (σν )β β̇ = μ ν (Mσν M † )α α̇ . (1.136)
Note that undotted/dotted Weyl indices by definition transform with M/M. We

also know that
Mσμ M † v μ = σμ v μ = σμ μ ν v ν (1.137)
for any v and hence
Mσν M † = σμ μ ν . (1.138)
With this, we return to (1.136) and continue the calculation according to
μ ν Mα β M α̇ β̇ (σν )β β̇ = μ ν (σρ )α α̇ ρ ν = ημσ σ τ ητ ν (σρ )α α̇ ρ ν

= ημσ ησρ (σρ )α α̇ = (σμ )α α̇ . (1.139)
Thus, we are indeed dealing with an invariant tensor.

Finally, we take

ψα
ψD = (1.140)
χ α̇
as a definition of a Dirac spinor. For covariance reasons (and up to possible

convention-dependent prefactors), we have

0 (σμ )α α̇
γμ = β̇β . (1.141)
(σ μ ) 0
The Lorentz transformation matrix is

Mα β 0
D(M) = α̇ , (1.142)
0 M β̇
α̇
where the M β̇ is obtained from Mα β by complex conjugation and rais-
ing/lowering of the indices.
(2) Using our suggestion to define γ 5 = diag(−1, 1), the result follows from the
definitions:
(1) (2) (1) (2) (1) (2) (1) (2)
ψ D ψD = χ (1) ψ (2) + ψ χ , ψ D γ5 ψD = −χ (1) ψ (2) + ψ χ (1.143)
(1) (2) (1) (1) (2) (1)
ψ D γμ ψD =ψ σ μ ψ (2) + χ (1) σμ χ (2) , ψ D γμ γ5 ψD = −ψ σ μ ψ (2) + χ (1) σμ χ (2) .
(1.144)
(3) Write
(σ μ )α̇α (σν )αβ̇ + {μ ↔ ν} =

α̇γ̇
αβ (σ μ )γ̇ β (σν )αβ̇ + {μ ↔ ν} (1.145)
= [(−iσ2 )σ μ (−iσ2 ) (σν )]

T α̇
β̇ + {μ ↔ ν} = [(σ2 )σ μ (σ2 )(σν )] α̇
β̇ + {μ ↔ ν} ,
(1.146)
where in the last two expressions σ μ and σν are assumed to have lower indices.
Now use that
σ2 σ 0 σ2 = σ0 and σ2 σ i σ2 = −σi . (1.147)
Using this minus sign, it becomes clear that the expressions with {μν} = {0i}
and {μν} = {i0} vanish after symmetrisation. The case {μν} = {00} obviously
gives the right answer. For {μν} = {ij } one needs to use σi σj + σj σi = 2δij to
find the result. The Clifford algebra for γ matrices is a direct consequence.
1.9.5 Covariant Expression for the 1-Loop Vacuum Energy

Task Derive the covariant expression (∼ d 4 k ln(k 2 +m2 )) for the vacuum energy
given in the lecture.
Hints Write down the path integral for gravity and a real scalar and integrate out the
scalar, including in particular its vacuum fluctuations. Focus only on the dependence
on rescalings of the metric, i.e. on metrics of the form gμν = αημν . This can in turn
be interpreted as a dependence on the 4-volume.
1.9 Problems 39
Solution The complete partition function (we suppress any source terms for
simplicity) reads

√ 1 2
Z= Dg Dφ exp i d 4 x g MP R − (∂φ)2 − m2 φ 2 . (1.148)
2
√
Here Dg stands for the integration over all metrics and g is the square root of the
modulus of the determinant of gμν . The φ-part of the action can be rewritten as

√
−i d 4x g φMφ with M ≡ −∂ 2 + m2 . (1.149)
For our purposes, it will be sufficient to understand how the φ part of the
partition function changes with α if gμν = αημν . Under this restriction, we can
reparameterise our spacetime such that gμν = ημν and only keep track of the
dependence on the total 4-volume V .
After Wick rotation (deformation of the x 0 integration contour from real to

imaginary axis by clockwise rotation and subsequent renaming x 0 = −ix 4 ), we
have

− d 4 x φME φ with ME ≡ −∂ 2 + m2 and ∂ 2 = δ μν ∂μ ∂ν .
V
(1.150)
Now we are dealing with a Gaussian integral with a matrix in the exponent, giving
us

1 1
Dφ exp − d x φME φ = √
4
= exp − tr ln ME . (1.151)
V det(ME ) 2
Here we have absorbed an infinite constant factor in the definition of Dφ in the first
step and applied the identity ln det = tr ln in the second step.
Now we note that in Fourier space
ME (k, p) = δ 4 (k − p) (k 2 + m2 ) , (1.152)
and hence

d 4k
tr ln ME = d 4 k δ 4 (k − k) ln(k 2 + m2 ) = V ln(k 2 + m2 ) . (1.153)
(2π)4
Here, in the first step, the δ function has remained outside the log since it only
signals that the matrix in question is diagonal. In the second step, we used

(2π)4 δ 4 (k − p) = d 4 x eix(k−p) = V for p = k. (1.154)
√
Undoing the Wick rotation and reinstating d 4x g instead of V , we find

√ 1 2 1 d 4k
Z= Dg exp i d 4 x g MP R − λ with λ= ln(k 2 + m2 ) .
2 2 (2π )4
(1.155)
Note that the intermediate transition to euclidean space could have been avoided by
regularising the oscillating Gaussian (with an i in the exponent) in some other way.
Supersymmetry and Supergravity
2
There are many motivations to learn about SUSY. Let us give a few: SUSY is the
only known symmetry relating fermions and bosons and may as such be a logical
next step in the historical road towards unification in fundamental physics. String
theory is the best-understood model of quantum gravity (or indeed the true theory
underlying quantum gravity) and its stable versions all rely on SUSY (in 2d and
in 10d). The only presently controlled roads from 10d strings to the 4d Standard
Model involve 4d SUSY theories as an intermediate step. SUSY can, at least in
principle, resolve the hierarchy problem at the scale where it becomes manifest.
In other words: If it had been discovered at the electroweak scale, we could have
found ourselves in a world without the hierarchy problem. Even if that happens
at, say, 10 TeV, the tuning would be much less severe than without SUSY. Finally,
SUSY is a central tool in formal field theory research since SUSY theories usually
involve many cancellations at the loop-level making them much better controlled.
For example, the best-understood example of the famous AdS/CFT correspondence
(to be explained later) involves an N = 4 4d super-Yang–Mills (SYM) theory. Here
N = 4 stands for four times the minimal amount of supersymmetry in the given
dimension.
The structure and notation of what follows will be strongly influenced by the
classic text [5], but there are many other useful books [69–72].
2.1 SUSY Algebra and Superspace
Recall the Poincare algebra

Pμ , Pν = 0 (2.1)

Mμν , Pρ = iημρ Pν − iηνρ Pμ (2.2)

Mμν , Mρσ = iημρ Mνσ − iηνρ Mμσ − iημσ Mνρ + iηνσ Mμρ (2.3)

42 2 Supersymmetry and Supergravity
as the symmetry algebra of R1,3 . This algebra can be represented by differential

operators acting on functions on R1,3 , e.g.

∂
Pμ = −i∂μ ∂μ = . (2.4)
∂x μ
Indeed, these operators generate translations according to
exp[i
μ Pμ ] f (x) = f (x) +
μ ∂μ f (x) + · · · = f (x +
) . (2.5)
Finite rotations in R1,3 are analogously generated by Mμν .

Any relativistic QFT has the above symmetry, but it may have additional
(‘internal’) symmetries acting on the fields. Examples are a shift symmetry φ →
φ +
or rotations in field space → exp(i
a Ta ) with ∈ CN and Ta the SU (N)
generators. Here ‘additional’ means that the full symmetry algebra is the direct sum
of Poincare and internal Lie algebra. The Coleman-Mandula theorem [73] claims
that such a direct sum structure is the only possibility for how the Poincare Algebra
can be extended to a larger symmetry of a QFT (more precisely, of the S-matrix).
This theorem can be avoided if one generalises the definition of a symmetry by
a Lie algebra: One replaces the latter by a so-called Lie superalgebra. Moreover,
the resulting extension of the Poincare algebra is (essentially) unique and is
called the supersymmetry algebra. This uniqueness is the statement of the Haag-
Lopuszanski-Sohnius theorem [74].
We will not demonstrate uniqueness but only present the result of the analysis:
The new generators to be added are (Weyl) spinors Qα and the crucial new algebra
relations are
{Qα , Qα̇ } = 2(σ μ )α α̇ Pμ , {Qα , Qβ } = 0 , {Qα̇ , Qβ̇ } = 0 . (2.6)
The main novelty is that for these generators one does not provide commutators but
anti-commutators, hence we are now dealing with a Lie superalgebra.
The object (σ μ )α α̇ is defined as
σ μ = (−1, σ 1 , σ 2 , σ 3 ) (2.7)
and is an invariant tensor of SL(2, C) just like (γ μ )ab is an invariant tensor

of SO(1, 3). In fact, these two statements are of course related since the Lie
algebras are the same and, roughly speaking, γ consists of two blocks of σ ’s
(cf. Problem 1.9.4). One can avoid σ ’s and Weyl spinors and formulate everything
using left-handed 4-spinors, but Weyl spinors are very convenient in this context.
Relations between two bosonic generators remain commutators and relations
between the new fermionic and the old bosonic commutators are also formulated
2.1 SUSY Algebra and Superspace 43
in terms of commutators:
[Pμ , Qα ] = 0 , [Mμν , Qα ] = i(σμν )α β Qβ , (2.8)
where
1
σμν ≡ − (σμ σ ν − σν σ μ ) and (σ μ )α̇α ≡
α̇β̇
αβ (σμ∗ )β̇β . (2.9)
4
We will often use an overline instead of the star (or dagger) for complex conjugation
(or the adjoint operator). The overline on σ does not specify whether upper or lower
indices are assumed. Indices can be raised or lowered using the
tensor. Given that
we need the lower-upper index version of (σμν )α β in (2.8), the expression on the
l.h. side of (2.9) should be read as defining precisely this version. Hence, it involves
an upper-upper index version of σ , which is provided on the r.h. side of (2.9).
The full SUSY algebra is defined by (2.1)–(2.3) together with (2.6) and (2.8).
Thus, we see that it consists of the Poincare algebra, the Q anti-commutators,
and the claim that the Q’s transform under the Poincare algebra as spacetime-
independent spinors. It may at this point also be useful to say more formally
what a Lie superalgebra is: It is a vector space with a Z2 grading (it splits
in an even and odd part) and with a binary operation that obeys the rules
even×even → even, even×odd → odd and odd×odd → even. Furthermore, there
are rules concerning the symmetries of these operations and Jacobi-like identities.
These are, however, automatically fulfilled if the operations are explicitly realised
through (anti)commutators, as in our case.
Next we want to represent this algebra on a larger space, called superspace. Its
coordinates are
x μ (μ = 0 · · · 3) and θ α (α = 1, 2) , (2.10)
the latter being fermionic (Grassmann variables) which form a Weyl spinor. The key
relations for our purposes are
(θ α )∗ = θ ,
α̇ α̇
{θ α , θ β } = 0 and h.c. , {θ α , θ } = 0 (2.11)
or, more explicitly,
(θ 1 )2 = (θ 2 )2 = 0 , θ 1 θ 2 = −θ 2 θ 1 , etc. (2.12)
One also defines partial derivatives
∂ ∂
∂α = ∂ α̇ = (2.13)
∂θ α ∂θ
α̇
together with the obvious rules
β̇ β̇
∂α θ β = δα β , ∂ α̇ θ = δα̇ β̇ , ∂α θ = 0 ∂ α̇ θ β = 0 . (2.14)
The reader should check that, as a result of the anticommutation relations for the
θ ’s, the ∂’s also anticommute.
The space parameterised by the x μ and θ α is called superspace, in this case
R , with 4 bosonic and 4 real fermionic (or two complex fermionic) dimensions.
4|4
Intuitively, one may want to think of R4 not as a set of points but, equivalently,
as the algebra of functions on R4 : {1, x μ , x μ x ν , · · · }.1 The generalisation to
superspace is then obvious: One simply thinks of the algebra of functions including
θ ’s, i.e. {1, x μ , θ α , x μ x ν , x μ θ α , · · · }. This is not a commutative algebra anymore
and hence it is not really the space of functions on some set of points. However, in
QFT we anyway mostly work with the space of functions on our spacetime. So the
formal generalisation from R4 to R4|4 on the basis of the respective function spaces
should not prevent us from doing all relevant manipulations.
Next, we naturally expect that a symmetry of this enlarged space will involve
some analogue of the familiar generators of translations, i.e. Qα ∼ ∂α + · · · . The
ellipsis stands for extra terms which must come in to ensure that Q’s anticommute
to give the P ’s. The correct formulae turn out to be
α̇
Qα = ∂α − i(σ μ )α α̇ θ ∂μ , Qα̇ = −∂ α̇ + iθ α (σ μ )α α̇ ∂μ . (2.15)
It is a straightforward but very important exercise to derive the essential part of the
SUSY algebra from this:
{Qα , Qα̇ } = 2i(σ μ )α α̇ ∂μ , {Qα , Qβ } = 0 , {Qα̇ , Qα̇ } = 0 . (2.16)
Let us pause for a small, technical comment: The reader will have noticed that,
with the standard identification Pμ = −i∂μ (recall that we are using a mostly-plus
metric), the algebras of (2.6) and (2.16) differ by a sign. This is nothing deep but
merely a result of two different ways of defining the operators Q and P . On the one
hand, one may think of them as acting on functions. On the other hand, as acting on
coordinates. To make this clear, one may consider the relation
(A∂ B∂ f )(x) = (B∂ f )(Ax) = f (BAx) (2.17)
between operators A, B acting on coordinates x and the corresponding differential

operators A∂ , B∂ acting on functions of x. It is immediately clear from the above
that the Lie algebras characterising the action on functions and on the space itself
1 According to the Gelfand-Naimark theorem, one may in fact (under very general circumstances)
always think of the algebra of functions on a space rather than of the space itself. These objects
encode the same information.
2.2 Superfields 45
differ by a sign. In our context, (2.16) corresponds to the action on functions and
(2.6) to that on R4|4 . The latter is also relevant for the action on the Hilbert space,
where states are e.g. of the type (x)|0, with a QFT field operator. The reader
may also want to recall that ÂB̂(x)B̂ −1 Â−1 = (ABx), with Â,B̂ operators
acting on the Hilbert space. Thus, there is no further sign change when comparing
the action on R4|4 to that on the Hilbert space.
2.2 Superfields
Now one builds a field theory on this enlarged space. A (complex) general superfield
is a function
2
F (x, θ, θ ) = f (x) + θ φ(x) + θ χ(x) + θ 2 m(x) + θ n(x) + θ σ μ θ vμ (x)
2 2
+θ 2 θ λ(x) + θ θ ψ(x) + θ 2 θ d(x) . (2.18)
Here the r.h. side is a Taylor expansion of the l.h. side where, however, all
higher terms vanish. The coefficient functions φ, χ, λ, and ψ are Weyl spinors,
anticommuting among each other and with the θ ’s.
We have started to use a very convenient shorthand notation for the product of
Weyl spinors, for example,
θ φ ≡ θ α φα =
αβ θα φβ , and analogously θ 2 = θ α θα . (2.19)
It is an essential part of this convention that suppressed undotted indices are always
summed from upper-left to lower-right. For dotted indices, the rule is inverse:
θ χ ≡ θ α̇ χ α̇ . (2.20)
This convention goes together with certain

tensor conventions:

0 −1 0 1

αβ = ,
αβ
= , with
αβ
βγ = δα γ .
1 0 −1 0
(2.21)
With this contraction one has, in spite of the anticommutation relations,
ψχ = χψ , (2.22)
as the reader should check.

It goes without saying that the Poincare algebra acts on superfields in the usual
way, e.g.
δ
F = i
μ Pμ F =
μ ∂μ F . (2.23)
By analogy, we define the SUSY transformation
δξ F = (ξ Q + ξ Q)F = [ (ξ ∂ − iξ σ μ θ ∂μ ) + h.c. ]F . (2.24)
Here by ‘h.c.’ we mean the application of a formal ∗-operation on the algebra of

functions and differential operators. In essence, this is just complex conjugation and
its obvious extension to the θ ’s. A crucial exception is the rule
(∂α )∗ = −∂ α̇ , (2.25)
which is required by consistency. The reader should check this by carefully thinking
about the possible ways to evaluate (∂α θ β )∗ .
Returning to our SUSY transformations, we note that the abstract concept of the
superfield F mainly serves the purpose of defining SUSY transformations on the
set of ‘component’ fields f , φα , etc. The latter are conventional quantum fields.
Concretely, after calculating δξ F , we expand it in a Taylor series and define δξ f ,
δξ φ, etc. as the coefficients of the various terms with growing powers of θ :
δξ F = δξ f + θ α (δξ φ)α + θ α̇ (δξ χ)α̇ · · · . (2.26)
This defines the SUSY transformation of the component fields.
2.3 Chiral Superfields
The general superfield is too large to be practically useful and it does indeed
correspond to a reducible representation of the SUSY algebra. Simpler superfields
exist and are sufficient to write down the most general SUSY lagrangian.
To define the chiral superfield, it is useful to first introduce SUSY-covariant
derivatives (in a way very similar to the Q’s):
α̇
Dα = ∂α + i(σ μ )α α̇ θ ∂μ , D α̇ = −∂ α̇ − iθ α (σ μ )α α̇ ∂μ . (2.27)
They obey
{Dα , D α̇ } = −2i(σ μ )α α̇ ∂μ , {Dα , Dβ } = 0 , {D α̇ , D α̇ } = 0 (2.28)
and, crucially, any D or D anticommutes with any Q or Q,
{Dα , Qα̇ } = 0, etc. (2.29)
This last feature implies that
D α̇ F = 0 ⇒ D α̇ δξ F = 0 . (2.30)
2.4 SUSY-Invariant Lagrangians 47
In other words, superfields fulfilling the condition D α̇ F = 0 form a subrepresen-

tation of the Lie superalgebra representation provided by general superfields. They
are called chiral superfields.
One may show that chiral superfields can always be written as
= (y, θ ) with y μ = x μ + iθ σ μ θ , (2.31)
where is an unconstrained function. It can be expanded according to

√
= A(y) + 2θ ψ(y) + θ 2 F (y) . (2.32)
As explained above for the general superfield, one obtains the SUSY transforma-
tions of the component fields by expanding δξ in the same way as . The result
reads
√
δξ A = 2ψξ
√ μ √
δξ ψ = i 2σ ξ ∂μ A + 2ξ F (2.33)
√
δξ F = i 2ξ σ μ ∂μ ψ .
We note that one can analogously define antichiral superfields, Dα = 0, and that
the conjugate of a chiral superfield is antichiral.
2.4 SUSY-Invariant Lagrangians
We state without proof that the most general such lagrangian, at the 2-derivative-
level and built from chiral superfields {1 , · · · , n } only, reads

ı
L = K(i , ) 2 + W (i ) + h.c. . (2.34)
θ 2θ θ2
Here K is a real function of a set of complex variables i . With i being chiral

ı
superfields, K becomes a general superfield. It is not chiral since both i and
are involved. The first term in L is the projection of the general superfield K on
its highest component, i.e. it is the analogue of the function d(x) appearing in the
Taylor expansion (2.18).
The function K is called the Kahler potential (for those who know this term
from complex geometry: the relevance in the present context will become clear

momentarily). The expression K 2 2 , viewed as part of the lagrangian, is called the
θ θ
D term. This name comes simply from the traditional use of the variable d(x) for
the highest component. The key point in this non-trivial way of writing a lagrangian
is, of course, its required invariance under SUSY transformations. For this, we need
to recall that the commutator of Q and Q is P . Hence the mass dimension of Q is
1/2. Since Q involves ∂/∂θ , the mass dimension of θ is −1/2 (one may think of it
very vaguely as the square root of x). Thus, in the Taylor expansion of superfields
in powers of θ the mass dimensions of components grow. As a result, due also to
the linear nature of SUSY transformations, the highest component cannot transform
into any other component—there simply is no component with a suitably high mass
dimension. The only way it can transform is into a derivative of another component.
Thus, the first term of the above lagrangian is invariant up to total derivatives, as one
would have hoped.
Similarly, W is called the superpotential and it is an analytic (or holomorphic)
function of the i . This makes W a chiral superfield. In its Taylor expansion in θ ,
with the coefficients being functions of y, the highest component is traditionally
called F (cf. (2.32)). Hence the corresponding two terms in (2.34) are sometimes
called F terms. Concretely, to get these terms one expands the chiral superfield
W (i ) in θ (with the coefficients being functions of y), extracts the coefficient of
θ 2 , and replaces y by x. The result, together with its hermitian conjugate, is the F -
term lagrangian. It is SUSY invariant up to a total derivative for the same reason as
explained in the case of the D term.
An equivalent way of writing this lagrangian is as

ı
L= d 2 θ d 2 θ K(i , ) + d 2 θ W (i ) + h.c. (2.35)
Using standard integration rules for Grassmann variables,

dθ1 θ1 = 1 and dθ1 1 = 0 , (2.36)
and the analogous formulae for θ2 , one can easily check that the integral formulation
is equivalent to the projection formulation of L. The SUSY invariance is particularly
easily seen in the integral formulation: The SUSY generator Q is a linear combi-
nation of x derivatives and θ derivatives. The x derivative of any lagrangian is, by
definition, a total derivative and thus leaves the action invariant. The θ derivative of
any expression in θ integrates to zero,

∂
dθ1 ··· = 0, (2.37)
∂θ1
as one can easily convince oneself. Thus, any action which is an integral over the full
superspace is invariant. Similarly, any action built as the integral of an expression in
θ (not θ ) and integrated over half the superspace is invariant. (Here it is important
to note that we can replace y with x by appealing to the Taylor expansion and the
irrelevance of total derivatives in x.)
2.5 Wess-Zumino-Type Models 49
2.5 Wess-Zumino-Type Models
The possibly simplest interesting SUSY model is the Wess-Zumino model [75].2 It
is defined by
m 2 λ 3
K = , W = + . (2.38)
2 3
A straightforward explicit calculation according to the rules above gives the
following component form of the lagrangian:
m
L = −|∂A|2 − iψσ μ ∂μ ψ − ψ 2 + λψ 2 A + h.c. + (mA + λA2 )F + h.c. + |F |2 .
2
(2.39)
Since F has no kinetic term (and thus does not propagate) we can integrate it out by
purely algebraic operations and without any approximation. Such fields are called
auxiliary fields. The equation of motion for F is
2
F = −mA − λA , (2.40)
and inserting this into the original lagrangian gives

m
L = −|∂A|2 − iψσ μ ∂μ ψ − ψ 2 + λψ 2 A + h.c. − V (A, A) , (2.41)
2
with the scalar potential (or F -term potential)
V (A, A) = |F |2 = |mA + λA2 |2 . (2.42)
This is easily generalised to (non-renormalisable and multi-field) models of the

type
ı
K = K(i , ) , W = W (i ) . (2.43)
We only display the purely bosonic part of the resulting component lagrangian.
More details are given in the problems. With the auxiliary fields already integrated
out, one has:
j
L = Kij (A, A) (∂Ai )(∂A ) + K ij (A, A) (∂i W (A))(∂ j W (A)) + · · · . (2.44)
2 For a veryinteresting earlier model, with supersymmetry realised non-linearly on fermionic fields
only, see [76].
Here
Kij = ∂i ∂ j K and Kij K kj = δi k , (2.45)
in other words, indices denote partial derivatives and the upper-index matrix is
defined as the inverse. To simplify notation, we have suppressed field indices in
the arguments such that, when writing e.g. W (A), we mean W (A1 , · · · , An ).
We note that the scalar components Ai parametrise a complex manifold (as
in so-called sigma-models) and, in supersymmetry, the metric on this field space
is the Kahler metric Kij , defined with the help of the Kahler potential K. The
superpotential W is locally a holomorphic function on this manifold; globally it
is a section in an appropriate complex line bundle.
2.6 Real Superfields
We have to discuss real superfields, another subrepresentation contained in that of

the general superfield, since they are needed to describe gauge theories. But we will
be very brief since, conceptually, the procedure is similar to that used in the chiral
superfield case.
A real superfield V = V (x, θ, θ ) is defined by the condition V = V . It can be
Taylor expanded in θ and θ . We will build lagrangians which are invariant under the
SUSY gauge transformation
2V → 2V + + , (2.46)
with a chiral superfield. Using this transformation, V can be brought to a form

where certain components vanish (the so-called Wess-Zumino gauge):
2 1 2
V = −θ σ μ θAμ + iθ 2 θ λ − iθ θ λ + θ 2 θ D . (2.47)
2
Moreover, applying SUSY-covariant derivatives to V , one constructs the so-called
field-strength superfield
1 2
Wα = − D Dα V , (2.48)
4
which can be shown to be chiral and gauge invariant. Its name is justified since it
does indeed contain the field strength Fμν = ∂μ Aν −∂ν Aμ in one of its components:

W = iλ(y) + D(y) + iσ μν Fμν (y) · θ + θ 2 σ μ ∂μ λ(y) , (2.49)
where 4σ μν = σ μ σ ν − σ ν σ μ . One can show that SUSY gauge transformations

contain standard gauge transformations as a subset. Hence, it is natural to look for
2.6 Real Superfields 51
SUSY-invariant and SUSY-gauge-invariant lagrangians. At the 2-derivative level,

the unique option is

1 α
α̇ 1 1 1 2
L= W W
α 2 + W α̇ W 2 = − Fμν F μν
− iλσ μ
∂μ λ + D ,
4g 2 θ θ g2 4 2
(2.50)
where λ is the gaugino and D the real auxiliary field that has already appeared
above.
It is straightforward to extend this to the non-abelian case, where V and W are
matrix-valued superfields, taking values in the Lie Algebra of the gauge group. Let
us write the corresponding lagrangian including also a charged matter superfield ,
to be thought of as a column-vector in some appropriate representation. We have

1 2 † 2V
L= tr W + h.c. + e 2 + W() 2 + h.c. . (2.51)
2g 2 θ2 θ2θ θ
Here e2V has to be taken in the representation of , and † has to be interpreted

as a row vector. The superpotential W is a holomorphic, gauge invariant function
of . We denote it by a calligraphic letter W to avoid confusion with the field
strength superfield W . This lagrangian is invariant under the non-abelian super
gauge transformations
→ e− .
†
e2V → e e2V e , (2.52)
One frequently uses the naming conventions for components
= {, ψ, F } , V = {Aμ , λ, D} . (2.53)
It is a slight abuse of notation to denote the scalar matter component by the same
name as the superfield, but this convention is widespread and it is usually clear from
the context which meaning is intended. With these conventions, the component form
of the lagrangian reads

1 1
L= tr − Fμν F μν
− 2iλσ μ
Dμ λ + D 2
(2.54)
g2 2
√
−|Dμ |2 − iψσ μ Dμ ψ + |F |2 + i 2 † λψ − ψλ + † D ,
where we have set W = 0 for simplicity. This lagrangian is called off-shell since it is
SUSY invariant without using the equations of motion. Integrating out the auxiliary
field, one arrives at the on-shell lagrangian. Concerning F , this step is trivial in the
present simple example: F is just set to zero. By contrast, integrating out D = Da Ta
induces a quartic term in the scalar fields, the so-called D-term potential.
2.7 SUSY Breaking
We have so far defined the spinor Q as a differential operator on superspace. Hence,

it is an operator on the space of superfields, hence an operator transforming different
component fields into each other. After quantisation, we will thus be able to define
a corresponding operator Q on the Hilbert space. This operator will mix bosons and
fermions and, since
[Qα , Pμ P μ ] = 0 , (2.55)
this implies that the mass of fermions and bosons (in the same superfield or mul-
tiplet) is the same. Thus, to be relevant for the real world, supersymmetry must be
spontaneously broken. In other words, while the action should be supersymmetric,
the vacuum should not be invariant under supersymmetry transformations.
At the perturbative level, this simply means that the lowest-energy field configu-
ration should not be invariant under SUSY. In the context of chiral superfields, the
r.h. side of
√
δξ A = 2ψξ
√ μ √
δξ ψ = i 2σ ξ ∂μ A + 2ξ F (2.56)
√
δξ F = i 2ξ σ μ ∂μ ψ
should hence be non-zero. Maintaining Lorentz-invariance, this can only be

achieved if F = 0 in the vacuum. This is called F -term breaking and the simplest
lagrangian with this feature is

L = 2 + c + h.c. (2.57)
θ 2θ θ2
The relevant terms in component form are
L = F F + cF + h.c. + · · · , (2.58)
which implies F = −c in the vacuum. However, while SUSY is formally broken,

the theory is free, thus F does not couple to other fields and hence the spectrum
remains supersymmetric.
This is easily remedied adding a higher-dimension operator,

1 2
L = − 2 () 2 2 + c 2 + h.c. (2.59)
M θ θ θ
Now, ignoring fermions and derivative terms, the component lagrangian reads
L = F F − 4F F /M 2 + cF + h.c. + · · · . (2.60)

2.8 Supersymmetrising the Standard Model 53
The vacuum is again at = 0 and F = −c, but now this non-zero F introduces
scalar masses and supersymmetry is broken in the spectrum of the theory.
We note that apparently simpler models which extend (2.57) by adding terms
∼2 or ∼3 to the superpotential do not work in our context. They reinstate a
SUSY-preserving vacuum, which is obvious since in such models the linear term
can be absorbed in a shift of . In fact, the simplest renormalisable model with
chiral superfields and spontaneous SUSY breaking is the O’Raifeartaigh model
with lagrangian [77]

3

i
L= i 2 + 1 (m2 + λ(3 )2 ) + μ2 3 + h.c. . (2.61)
θ 2θ θ2
i=1
It is easy to write down the F -term potential and minimise it to find spontaneous
SUSY breaking. Sometimes the name O’Raifeartaigh model is used more generally
for any model with F -term breaking.
A completely analogous story can be developed for real superfields, i.e. (abelian)
gauge theories, where SUSY breaking is signalled by a non-zero VEV of the D
term. The simplest model realising this is

1 2
L= W + 2κV 2 2, (2.62)
2g 2 θ2 θ θ
where the new term linear in V is known as Fayet-Iliopoulos or FI term [78]. At

the component level one finds
1 2
L= D + κD ⇒ D = −κg 2 = 0 . (2.63)
2g 2
As before, the model needs to be enriched to see this formally present SUSY
breaking in the spectrum. This can be achieved e.g. by adding two chiral superfields
(to avoid anomalies) with charge ±1 and mass m. One finds that the fermions remain
massless while the boson masses split according to m21,2 = m2 ± κg 2 . See e.g. [5]
for details.
2.8 Supersymmetrising the Standard Model
The Minimal Supersymmetric Standard Model or MSSM is obtained basically by

promoting all fermions and scalars of the Standard Model to chiral superfields and
all vectors to real superfields. The additional components introduced in this way
are made heavy by an appropriate mechanism of SUSY breaking, to be discussed
shortly. The only exception to this rule arises in the Higgs sector, where one now
needs two different Higgs doublets and hence two corresponding superfields: Hu
and Hd . The reason will become clear immediately. Of the many reviews of this
wider subject we refer in particular to [29, 79, 80].
After these preliminaries, we give the set of chiral superfields:
a = {Q, U, D, L, E, Hu , Hd } . (2.64)
The gauge representations are as in (1.2), up to Hu , which has opposite U (1) charge.
Our naming conventions follow (1.6) and we have suppressed the generation index
on the matter superfields Q, · · ·, E for brevity. The lagrangian can be organised in
three pieces. First,

3
1
Lgauge = tr (Wi ) (Wi )α 2 + h.c. ,
α
(2.65)
i=1
2gi2 θ
with ‘tr’ to be replaced by ‘1/2’ in the U (1) case. Second,

7

LK = †a e2V a , (2.66)
θ 2θ 2
a=1
where K stands for kinetic or Kahler potential term and where the superfield V =
V1 + V2 + V3 contains the three real superfields corresponding to three factors of
GSM . In each term, one must use the representation appropriate for a . Third, we
have the superpotential term

LW = (Wμ + WY + We ) + h.c. (2.67)
θ2
with
Wμ = μHu Hd , WY = λu QHu U + λd QHd D + λe LHd E , (2.68)
and
We = aLHu + bQLD + cU U D + dLLE . (2.69)
The structure of Lgauge and LK requires no further comments: They simply provide
the necessary kinetic terms and gauge interactions. The Standard Model Yukawa
couplings come from WY , together with new interactions that are not present in the
Standard Model. To give masses to all fermions, we were forced to introduce two
Higgs fields. Indeed, holomorphicity forbids the appearance of the ˜ variable used
in the up-type Yukawa term of the non-supersymmetric Standard Model. Hence,
a new Higgs multiplet Hu with opposite U (1) charge has to be introduced. An
independent reason for this second doublet is the need to cancel the U (1) anomaly
introduced by the fermion (the ‘Higgsino’) contained in Hd .
2.8 Supersymmetrising the Standard Model 55
Finally, there are extra terms without a Standard Model analogue, allowed due
to the enlarged field content. We have collected these terms in We but, since some
of them induce proton decay and lepton number violation, we basically want to
forbid them. We also note that we have limited ourselves to the renormalisable
level—hence W is truncated at cubic order. To see that cubic terms correspond to
marginal operators, recall that θ 2 has mass dimension −1. Hence, projection on the
θ 2 component corresponds to raising the mass dimension by one unit. Thus, mass
dimension 3 in W corresponds to mass dimension 4 in L.
To forbid We , the concept of an R-symmetry (which is crucial in SUSY
independently of phenomenology) is useful. To explain this concept, we define
standard (global) U (1) and U (1)R transformations of chiral superfields as follows:
U (1) : (y, θ ) → eim

(y, θ ) , U (1)R : (y, θ ) → ein
(y, e−i
θ ) .
(2.70)
Here m and n ≡ R() are the U (1) and U (1)R charges of , respectively. It follows
immediately that (and this is the crucial feature of an R-symmetry!) the components
transform differently, depending on their mass dimension:
A → ein
A , ψ → ei(n−1)
ψ , F → ei(n−2)
F . (2.71)
Invariance of the lagrangian requires R(K) = 0 and R(W ) = 2. The former is

2
clear since the projection on the θ 2 θ component does not change the R-charge. By
2
contrast, projection on the θ component lowers the R-charge by 2.
For our purposes, the interesting assignment is:
For Q, U, D, L, E : R = 1 and for Hu , Hd : R = 0 . (2.72)
This restricts W to the Yukawa terms. However, this is too strong since it also
forbids the so-called μ term μHu Hd . But the latter is needed since even after SUSY
breaking (see below) it is the only source for Higgsino masses. (Higgsinos are the—
so far unobserved and hence heavy—fermionic partners of the Higgs.)
A possible resolution is the breaking of U (1)R to its Z2 subgroup. By this we
mean restricting U (1) ≡ { ei
} to the two elements with
= 0 and
= π. After
this breaking to Z2 , R-charges are identified modulo 2. Indeed, superfields with
R-charge m and m + 2 now transform identically. In particular the selection rule
R(W ) = 2 for superpotential terms is modified to R(W ) = 2 mod 2 . In other
words, one now only demands R(W ) ∈ 2N. As a result, the μ term is allowed
while all terms in We are still forbidden. Moreover, the transformation rules of the
Standard Model fields and their superpartners under this so-called Z2 R-parity are
Even : Higgs scalars, fermions, gauge bosons (2.73)

Odd : Higgsinos, sfermions, gauginos . (2.74)
Here by sfermion one refers to the superpartner of a Standard Model fermion. A

sfermion is hence a boson. For example, a selectron is one of the many sleptons.
The top squark is the SUSY partner of the top-quark, etc.
The above R-parity assignments imply that any of the so-called superpartners
cannot decay into a combination of Standard Model particles. Hence the lightest
superpartner (the lightest supersymmetric particle or LSP) is absolutely stable
and provides a natural dark matter candidate. Unfortunately, with growing LHC-
bounds on its mass the abundance predicted from its so-called freeze-out in early
cosmology tends to become too high, calling for extensions of the simplest settings.
For more details see e.g. [81–84].
2.9 Supersymmetric and SUSY Breaking Masses

and Non-renormalisation
The simplest way to make the above construction realistic is to add mass terms to the
supersymmetric Standard Model such that all the superpartners of Standard Model
particles become sufficiently heavy. (Recall that the Higgsino can be made heavy by
a sufficiently large μ term.) While technically correct, such an approach of explicit
SUSY breaking is not very satisfying or illuminating concerning the resolution of
the hierarchy problem.
Hence, we will introduce somewhat more structure and try to arrive at the MSSM
using spontaneous SUSY breaking. Specifically, we introduce a hidden sector
in which SUSY is broken spontaneously.3 It will then be communicated to the
Standard Model by higher-dimension operators. To illustrate this structure, we start
with the toy model lagrangian

1
L = SS − c1 (SS)2 2 +c 2 S +h.c.+ 2+ SS 2 2. (2.75)
θ 2θ θ 2 θ 2θ M2 θ θ
We recognise a model with a chiral superfield S and with spontaneous SUSY

breaking (FS = 0). In addition, we have a free and massless chiral superfield . The
latter represents the Standard Model or, more specifically, its Higgs sector. Finally,
the last term is a higher-dimension operator, suppressed by a large mass scale M,
coupling the two sectors. All we need to know about the hidden sector is that S = 0
and FS = 0 in the vacuum. Inserting this in our lagrangian and focussing on the
-sector only, we have

1
L ⊃ 2 + A A F S FS , (2.76)
θ 2θ M2
3 In principle, one may imagine situations where SUSY is broken spontaneously in the super-
symmetrised Standard Model, without introducing any additional fields. However, it turns out that
this does not work in practice, taking into account experimental constraints on masses and the
phenomenologically required gauge symmetry breaking.
2.9 Supersymmetric and SUSY Breaking Masses and Non-renormalisation 57
where we also ignored a quartic fermionic term arising from the superfield higher-
dimension operator. We see that the result is equivalent to just having added a (‘soft’)
SUSY breaking scalar mass term to the sector
L ⊃ m2sof t |A |2 with m2sof t = |FS |2 /M 2 . (2.77)
Crucially, in our approach we see right away that this term is radiatively stable—it is
secretly a higher-dimension operator and does as such not receive power-divergent
loop corrections.4 This explains the name ‘soft’. In fact, the two sectors decouple
completely as M → ∞, making it clear that the coupling operator can only
renormalise proportionally to itself. (We see here another possibility, in addition
to symmetries, why a certain coefficient in the lagrangian may be zero in a natural
way: In its absence, the model becomes the sum of two completely independent
theories.)
Our point about the mass term not being quadratically divergent may appear
trivial—after all the sector itself is a free theory, so of course nothing renor-
malises. However, it is immediate to enrich our model by e.g. W () ∼ 3 , leading
to quartic self-interactions. Alternatively, may be charged under some gauge
group, like the Higgs in the Standard Model is. Nothing in our argument changes:
The operator ∼ 1/M 2 inducing the mass cannot have power-divergences.
However, one could clearly add a term W () ∼ mSU SY 2 to our action, in
other words, a supersymmetric mass term. We have to be sure that interactions in
the sector will not, if such a term is absent in the beginning, induce it through loop
corrections. This, as it turns out, is in fact the main point where SUSY saves us: The
superpotential does not renormalise. This so-called non-renormalisation theorem
is, at least at a superficial level and in our simple model, easy to understand [85]:
Indeed, consider the Wess-Zumino model with tree-level superpotential
m 2 λ 3
W= + . (2.78)
2 3
Introduce a U (1) and U (1)R -symmetry under which has charges (1, 1). Clearly,
this is respected by our canonical Kahler potential K, but the superpotential breaks
both symmetries. One can interpret this breaking as being due to non-zero VEVs
of superfields m and λ, the scalar components of which have acquired non-zero
VEVs. For this interpretation to work, one needs to assign to m the charges (−2, 0)
and to λ the charges (−3, −1). Now, assuming that perturbative loop corrections
4 The intuitive reason is very simple and can be thought of as the opposite of the more familiar
fact that operators with mass dimension less than four do receive power-like loop corrections.
Indeed, to correct a higher-dimension operator in an otherwise renormalisable model the loop
must produce a coefficient of negative mass dimension. This could be the result of the tree-level
coefficient multiplied by dimensionless couplings and cutoff-dependent logarithms. Any positive
power of the cutoff would have to come with a mass term in the denominator. The latter would
imply an infrared divergence, which, however, does not arise in 4d and in the present context.
break neither these U (1) symmetries nor SUSY, we expect that the effective
superpotential (relevant for the Wilsonian effective action) will still respect the two
U (1) symmetries. Using holomorphicity and the fact that each term in W must have
charges (0, 2), we have

Weff = cij k mi λj k = m2 f (λ/m) . (2.79)
ij k
In the second step, we used the fact that, under the symmetry constraints, the triple
sum collapses to a single sum, which can then be viewed as a power series in
(λ/m). This last combination of fields can appear to any power, since both its
U (1) and U (1)R charge vanish.
Now, the constant and linear term in f correspond to the terms already present
at tree level—their values are 1/2 and 1/3 by assumption. We see that higher terms
in λ, which may in principle arise from loop corrections, always come with higher
powers of and hence do not affect mass and trilinear coupling. Moreover, it is easy
to convince oneself that such higher terms in λ, as derived from (2.79), correspond
precisely to terms following from tree-level exchange of . But such tree-level
effects should not be included in Weff . They appear in the calculation of observables
if one uses only the tree-level expression for W together with the standard Feynman
rules. Including them in Weff would lead to double counting. Now, compared to tree-
level effects, loop effects always have a higher power of λ (given a certain number
of external legs). Hence such loop effects are not described by the higher-λ terms in
f . As a result, we learn that Weff = W and no loop corrections arise.
In summary, we have learned that the structure of (2.75), supplemented by a
superpotential of type (2.78), is radiatively stable. In particular, the supersymmetric
and supersymmetry breaking mass terms can both be chosen small compared to the
cutoff scale and are not subject to power-like divergences.
2.10 The Minimal Supersymmetric Standard Model (MSSM)
With this, it is straightforward to introduce SUSY breaking by a spurion5 superfield

S into the SUSY Standard Model. Without aiming at completeness, we give four
types of higher-dimension terms which are sufficient to generate all essential SUSY
breaking effects:
1 2
1
L1 = QQSS 2 2, L2 = Q SS 2 2
M2 θ θ M2 θ θ
(2.80)
1 3 1

L3 = Q S 2 + h.c., L4 = W α Wα S 2 + h.c.
M θ M θ
5 The field is spurious in that we only introduced it to parameterise a certain symmetry breaking
effect. Its actual dynamics is not important for us.

2.10 The Minimal Supersymmetric Standard Model (MSSM) 59
Here Q stands for generic Standard Model chiral superfields. The different factors
of Q in one term may also be replaced by different Standard Model fields, e.g.
Q3 → QHu U .
The effects of these different terms are easy to read off. For example,
|FS |2
L1 ⊃ |AQ |2 ≡ M02 |AQ |2 , (2.81)
M2
where we refer to M0 as the soft mass which AQ acquires. Similarly, L2 induces a
holomorphic soft mass, which due to symmetry constraints arises only in the Higgs
sector, with Q2 → Hu Hd . Furthermore, L3 induces soft trilinear or ‘A-terms’:
FS 3
L3 ⊃ A ≡ A · A3Q . (2.82)
M Q
Finally, the last term induces gaugino masses M1/2 ,
FS α
L4 ⊃ λ λα ≡ M1/2 λα λα . (2.83)
M
A standard scenario, known as ‘Gravity Mediation’ has M ∼ MP ∼ 1018 GeV,

a value which corresponds the scale at which one may expect quantum gravity to
induce all allowed higher-dimension operators.
√ Then one would need the SUSY
breaking scale in the hidden sector to be |FS | ∼ 3×1010 GeV (which is sometimes
referred to as an ‘intermediate scale’) to obtain
M0 ∼ A ∼ M1/2 ∼ 1 TeV . (2.84)
Of course, many new parameters are introduced in this way. In particular, there
are as many A-terms as there are entries in the Yukawa coupling matrices, and
the soft masses come as 3 × 3 matrices in generation space. If the scale of the
soft terms (sometimes referred to as the SUSY breaking scale) is low—e.g. in the
TeV range, then generic values for the soft terms are ruled out by flavour-changing
neutral currents (FCNCs) and other experimental signatures. Some symmetry-based
model building is needed to make this scenario realistic.
It is crucial that no renormalisable couplings between hidden and visible sector
are present. In particular,√a superpotential term SQ2 (or concretely SHu Hd ) would
induce a Higgs mass ∼ |FS |, destabilising the hierarchy. Furthermore, we need
a non-zero μ term for the Higgs, but it should not be too large, again to avoid a
hierarchy destablisation.
Thus, the task is to induce a supersymmetric μ term of the same size as the
(otherwise very similar) SUSY breaking holomorphic mass term ∼ Hu Hd (where
Hu , Hd are the scalar components, not the superfields). There is a very elegant
solution to this problem known as the Giudice-Masiero mechanism [86]. It is based
on the higher-dimension couplings

1 1
L⊃ SHu Hd + 2 SSHu Hd 2 2 . (2.85)
M M θ θ
They induce terms

|FS |2
FS
L⊃ Hu Hd 2 + Hu Hd . (2.86)
M θ M2
Here, in the first term, Hu , Hd represent superfields and, in the second term, the same
symbols are used for their scalar components. Clearly, the first term in (2.86) is the
previously discussed μ term, but with a coefficient governed by the SUSY breaking
scale, μ ∼ F S /M. The second term is the so-called Bμ term, the previously
mentioned holomorphic mass term for the Higgs:
L ⊃ Bμ Hu Hd with Bμ ∼ |FS |2 /M 2 . (2.87)
Upon integrating out the F -terms of the Higgs superfields, the μ term also
contributes to the quadratic Higgs scalar potential, which in total reads
V2 = (|μ|2 + m2Hu )|Hu |2 + (|μ|2 + m2Hd )|Hd |2 + Bμ Hu Hd + h.c.

(2.88)
†
Hu |μ|2 + m2Hu Bμ Hu
= .

H d Bμ |μ| + m2Hd
2
H d
The second line makes it apparent that we are dealing simply with a 4 × 4 complex
mass matrix, giving mass to the four scalars contained in (Hu ,
H d )T . Due to
SU (2) symmetry, this matrix has a 2 × 2 block structure and hence only two distinct
eigenvalues. Electroweak symmetry breaking requires one of the eigenvalues to be
negative.
An independent quartic Higgs interaction is not present in the SUSY Standard
Model since no cubic Higgs superpotential is allowed. However, the D term of
the SU (2) × U (1) SUSY gauge theory does the important job of creating such a
coupling:
1 2 1
V4 = (g + g22 ) (|Hu |2 + |Hd |2 )2 + g22 |Hu H d |2 . (2.89)
8 1 2
Assuming soft terms are close to the weak scale, the scalar potential V2 + V4 and
its symmetry breaking structure has been analysed in great detail, but we will not
discuss this. Suffice it to say that electroweak symmetry can be broken as required,
both Higgs doublets generically develop VEVs (the ratio being parameterised by
tan β ≡ vu /vd ), and the Higgs mass is predicted in terms of this mixing angle and
2.11 Supergravity: Superspace Approach 61
the gauge couplings. This is a great success, given in particular that all parameters
of this model are now protected from power-divergences, the SUSY breaking and
weak scale are naturally small, and the model is renormalisable and can, in principle,
be valid all the way to the Planck scale. In addition, extrapolating the Standard
Model gauge couplings to high energy scales [87–90], one finds that they meet rather
precisely at the GUT scale MGU T 1016 GeV (see Problem 2.13.5). This has been
known since about 1990 and has given a lot of credibility to the model [91].
However, the predicted Higgs mass is bounded by the Z-boson mass at tree level,
which is clearly incompatible with observations. The correction needed to bring the
Higgs mass up to its observed value of 125 GeV can be provided by loops, but this
requires a large mass of the stop quark (also ‘top squark’ or simply ‘stop’) or large
trilinear terms. This drives (again through loops) the Higgs VEV to higher values
and partially spoils the success of the hierarchy problem resolution. In addition, the
non-discovery of superpartners at the LHC has raised the lower limits for soft terms,
also limiting the success of the supersymmetric resolution of the hierarchy problem.
Thus, the phenomenological status of this model has deteriorated. From a modern
perspective, it may be appropriate to view the MSSM not as a weak-scale model but
rather as a model at a significantly higher scale, msof t mew .
This perspective implies that one integrates out all SUSY partners and the second
Higgs at msof t and is left with just the Standard Model below that scale. More
precisely, this requires that the lowest eigenvalue of the Higgs mass matrix in (2.88)
is smaller than the typical entries (which are ∼ m2sof t ). This is a fine-tuning of
the order m2ew /m2sof t which one may have to accept. This fine-tuning ensures that
m2H of the Standard Model Higgs, which sets the weak scale, is somewhat below
the SUSY breaking scale. One may refer to this as a ‘high-scale’ or ‘split’ MSSM
[92, 93], and it is not implausible that such a model (or some variant thereof) arises
in string theory (see e.g. [94, 95]).
We may here return to the terminology introduced at the end of Sect. 1.7: We
have learned that low-scale SUSY can solve the large hierarchy problem. (Here
‘low’ refers to the TeV range, including say 10 TeV or even higher.) SUSY does,
however, suffer from a little hierarchy problem. This is related to the detailed
interplay between SUSY breaking and electroweak symmetry breaking, which force
msof t to go up to 10 TeV or above. The severity of this little hierarchy problem of
low-scale SUSY depends on the details of the model and is still under debate.
2.11 Supergravity: Superspace Approach
All that was said above must, of course, be consistently embedded in a generally
relativistic framework. The resulting structure, known as supergravity, is equally
elegant and unique, though technically much more complicated than flat-space
SUSY. We can only give a brief summary of results. Since we described flat
SUSY using the superspace approach, let us start by noting that a similar (curved)
superspace approach can also be used to derive supergravity [5, 96]. For a brief
discussion of this see also [97].
One starts, as before, with coordinates
zM = (x μ , θ τ , θ τ̇ ) (2.90)
with the above indices being ‘Einstein indices’, as in conventional general relativity.
Then one introduces a vielbein, EA M (z), i.e. a basis of vectors, labelled by the
‘Lorentz indices’
A = (a, α, α̇) . (2.91)
As in general relativity, one defines a connection, introduces constraints (such as

the vanishing torsion constraint), and removes gauge redundancies. This is very
cumbersome in the present case, but it eventually leads to a supergravity superspace
action

S = d 8 z E (, ) + d 6 z ϕ 3 W () + h.c. (2.92)
Here E is the determinant of the vielbein EA M . The latter contains a real vector
superfield and an (auxiliary) chiral superfield [96]
2 2
Hμ = θ σ a θ ea μ + iθ θ ψ μ + h.c. + θ 2 θ Aμ (2.93)

ϕ = e−1 1 − 2iθ σμ ψ + Fϕ θ 2
μ
(2.94)
with e = det(ea μ ) and σμ = σa ea μ . We thus have the component fields
ea μ (x) , ψα μ (x) , Aμ (x) , Fϕ (x) . (2.95)
Here the first is the familiar vielbein of Einstein’s theory, and the last two are
auxiliaries (some authors use the notation B(x) ≡ Fϕ (x)). The crucial new feature
is a physical, propagating spin-(3/2) field ψα μ , called the gravitino, which is the
superpartner of the vielbein (or equivalently of the metric or graviton). The z
integrations are over the full or half of the Grassmann part of superspace, as in the
flat case. The argument stands for as many chiral superfields, containing matter
degrees of freedom, as one wants. The function is real.
It goes far beyond the scope of these notes to derive the component action.
However, to get a glimpse of what is going on, we can consider the flat-space limit:
ea μ = δ a μ , ψα μ = 0 , Aμ = 0 , ϕ = 1 + θ 2 Fϕ . (2.96)
2.11 Supergravity: Superspace Approach 63
Then the action takes the form

S = d 8 z ϕϕ (, ) + d 6 z ϕ 3 W () + h.c. (2.97)
From this, integrating out F and Fϕ , one can straightforwardly obtain the
supergravity scalar potential. To be specific, one finds the potential in the Brans-
Dicke frame. This is so because, in the curved case, one would have also have found

√ 1 2 (, )
S⊃ d 4x g MP R · , (2.98)
2 3
i.e. the Einstein–Hilbert term in the Brans-Dicke frame. Rescaling the metric to
absorb the factor /3, one arrives at an Einstein-frame curvature term together with
the supergravity scalar potential
2

V = eK/MP K ij (Di W )(D j W ) − 3|W |2 /MP2 (2.99)
where
Di W = ∂i W + Ki W (2.100)
and
K = −3MP2 ln[−/(3MP2 )] or = −3MP2 exp[−K/(3MP2 )] .

(2.101)
This goes together with conventional kinetic terms for the fields i , based on the
supergravity Kahler metric Kij . We have given all of the above keeping MP explicit
to make it easy to see that the flat-space limit, MP → ∞, takes us back to previous
formulae. In particular, one can see that the first term in (2.99) corresponds to the
familiar F -term scalar potential while the second term is supergravity-specific. It
is non-zero even if W is just a number and thus allows for the introduction of a
cosmological constant, albeit only a negative one. This is consistent with the fact
that the Poincare superalgebra can be generalised to Anti-de Sitter but not to de
Sitter space.
In practice, one mostly works with the above formulae in units in which MP = 1.
This is much more economical and we will do so from now on.
Let us note that, among many other terms, one has
L ⊃ −eK/2 W ψμ σ μν ψν + h.c. , (2.102)
which implies a gravitino mass
m3/2 = eK0 /2 W0 , (2.103)

where W0 and K0 are the vacuum values of W and K. We will suppress the indices
‘0’ from now on since it will be clear from the context whether the vacuum value
or some other dynamical value is meant. Supersymmetry breaking is, as before,
governed by non-zero VEVs of (some of) the F -terms,
F i = eK/2 D i W ≡ eK/2 K ij D j W . (2.104)
Realistically, we have λ = V0 0 (the non-zero meV-scale cosmological constant

value is negligible compared to particle-physics scales). Hence, the positive-definite
F -term piece and the negative |W |2 piece must compensate with high precision in
the formula for V . We thus have
|F | ∼ eK/2 |W | and hence m3/2 ∼ |F | . (2.105)
Here |F | is the length of the vector F i , calculated using the Kahler metric Kij .
We note, however, that this is in Planck units and, reinstating MP , one has m3/2 ∼
|F |/MP . Thus, if one takes the hidden sector F very low, near the weak scale (as is
in principle consistent with our SUSY breaking discussion), the gravitino can still
be very light. This, however, requires that it couples to Standard Model fields only
very weakly.
We note that the SUSY solution to the weak-scale hierarchy problem works
as before: All that we said remains valid since we are working at an EFT scale
μ MP and the rigid limit (supplemented by the gravitino, if it is sufficiently light)
can be used. The non-renormalisation theorem extends to the W of supergravity.
What is more, the presence of higher-dimension operators which was central in the
communication of SUSY breaking from hidden to visible sector can be argued to be
generic in the supergravity context: After all, the theory is non-renormalisable, so
all in principle allowed operators are expected to be present with MP -suppression.
Also, the non-linear structure of expressed in terms of K suggests such operators.
In other words, even if K = , the presence of factors like exp(K) introduces
many higher-dimension operators. The corresponding, very generic way of SUSY
breaking mediation (through Planck suppressed higher-dimension operators) is
called gravity mediation.
2.12 Supergravity: Component Approach
Before closing this section, we should note that we only discussed the superfield
approach to supergravity since it fits the previous analysis of rigid supersymmetry
best. It is not the most economical or widely used approach, which is instead
based on the component form of SUSY multiplets and (superconformal) tensor
calculus [6].
Very briefly, the story can be told as follows: In general relativity, Lorentz
symmetry becomes local. Since the SUSY parameter ξ , being a spinor, transforms
non-trivially under the Lorentz group, it would be inconsistent to consider it a global
2.12 Supergravity: Component Approach 65
object. Instead, it must be promoted to a spacetime dependent quantity,
ξ → ξ(x) , (2.106)
such that supersymmetry becomes a gauge symmetry. But now we are clearly
missing a gauge field defining the connection associated with our gauge symmetry.
By analogy to
Aμ (x) → Aμ (x) + ∂μ
(x) , (2.107)
one writes
ψμ (x) → ψμ (x) + ∂μ ξ(x) . (2.108)
The new field ψμ is a vector-spinor, also known as gravitino. We here interpret both
ξ and ψμ as 4-component spinors, specifically Majorana spinors.
The presence of the gravitino can also be motivated in a different way: Indeed,
we are clearly missing a superpartner for the graviton. As it turns out, the right
object is ψμ . To understand this better, we take a step back, forget about superfields,
and recall the SUSY algebra with its generators Q and Q (that come on top of the
Poincare generators). They have spin and hence raise or lower the spin of objects on
which they act. Indeed, developing the representation theory of the SUSY Poincare
algebra one finds multiplets including particles with different spin or, in the massless
case, helicity. We already know the multiplets
(0, 1/2) and (1/2, 1) (2.109)
corresponding to the chiral and real superfield (or the scalar and vector multiplet).
Naturally, one expects and indeed finds the multiplet
(3/2, 2) (2.110)
containing gravitino and graviton. For this to be consistent, one needs the gravitino
to contain 2 degrees of freedom on shell, to match those of the graviton. Indeed,
the general expressions for numbers of degrees of freedom of a vector-spinor,
initially and after taking into account gauge redundancy, constraints, and the on-
shell condition, are
1
d · 2[d/2] → (d − 3) · 2[d/2] . (2.111)
2
Here the exponent [d/2] (the integer fraction of d/2) characterises the dimension of
a general spinor, the reduction from d to d − 3 is associated with gauge freedom and
constraints, and the prefactor 1/2 is the usual reduction from off-shell to on-shell
degrees of freedom affecting any spinor (due to the equation of motion being first
order).
We record for completeness the underlying action and equation of motion (the
Rarita-Schwinger equation),

S=− d d x ψ μ γ μνρ ∂ν ψρ and γ μνρ ∂ν ψρ = 0 , (2.112)
although we will not have time to discuss the derivation of the physical degrees
of freedom from this dynamical description. Furthermore, we should note that the
modern way of deriving actions in this context is the so-called tensor calculus.
By this one means rules for multiplying (combining) multiplets to obtain new
multiplets. We saw an example of this when we noted that 1 (y, θ )2 (y, θ ), with
1 and 2 chiral, defines a new chiral superfield. This can be formulated without
superspace, just on the basis of the components. With this method, the full action of
supergravity, including supergravity coupled to chiral and vector multiplets, can be
derived.
More specifically, the method of choice is ‘superconformal tensor calculus’,
which first extends the theory to a conformal supergravity, then breaks scale
invariance by a VEV and removes the extra degrees of freedom by constraints.
(The non-SUSY version of this would be to replace the Planck scale by a field and
then recover usual gravity by giving this field a VEV.) In fact, this superconformal
method is also used in the superspace approach and we saw a trace of the field
whose VEV eventually breaks scaling symmetry in the chiral compensator ϕ(y, θ )
of (2.97).
Let us end with part of the general 4d supergravity action (the full action being
given e.g. in [5, 6]). The input are three functions, the (real) Kahler potential K, the
holomorphic superpotential W and the (also holomorphic) gauge-kinetic function
fab . Returning also to the Weyl description of spinors, one has:
1 1

j
√ L = R +
μνρσ ψ μ σ ν Dρ ψσ + Kij (Dμ φ i )(D μ φ ) − iχ j σ μ Dμ χ i
g 2

1 a 1
+(Refab ) − F a μν F b μν − λ σ μ Dμ λa + (Imfab )F a μν F̃ b μν
4 4

i 1
−e K/2
W ψμ σ ψν + √ (Di W )χ σ ψ μ + (Di Dj W )χ χ
μν i μ i j
2 2

+h.c.
−VF − VD + { further fermionic terms } . (2.113)

2.12 Supergravity: Component Approach 67
Here
Di W = Wi + Ki W (2.114)
Di Dj W = Wij + Kij W + Ki Dj W + Kj Di W − Ki Kj W − ij Dk W . k
We already know the F -term potential

VF = eK K ij (Di W )(D j W ) − 3|W |2 . (2.115)
The D-term potential has until now only been given implicitly and in a special case.
More generally, it reads (cf. [98] for a very compact discussion)
1
VD = [(Re f )−1 ]ab Da Db . (2.116)
2
To define the D terms, we recall that the scalars parameterise a Kahler manifold
which, to be gauged, must have some so-called (holomorphic) Killing vector fields
∂
Xa = Xai (φ) . (2.117)
∂φ i
They define the direction in which the manifold can be mapped to itself by the
gauge transformation corresponding to the index a. They also appear in the general
formula for the covariant derivatives:
(Dμ φ)i = ∂μ φ i − Aaμ Xai (φ) . (2.118)
Now, the D terms are defined as real solutions of the differential equations (the
Killing equations)
∂Da (φ, φ)
Xa i = −iK ij j
. (2.119)
∂φ
Mathematically, they are the Killing potentials. They can be given explicitly as
Da = iKi Xa i + ξa , (2.120)
where the ξa are so-called supergravity FI terms. The latter are only allowed for
abelian generators and they are believed to cause problems for a quantum gravity
UV completion.6
6 Such ‘constant’ FI terms require the supergravity to be ‘gauged’, i.e. the gravitino to be charged
under the U (1) responsible for the FI term. Concretely, there is a mixing between this gauged
U (1) and a global U (1) R-symmetry such that, in the end, a certain global U (1) survives. This
The terms we omitted when writing the action involve kinetic mixings between
matter fermions, gauginos, and gravitino (which become relevant in the presence
of gauge symmetry or SUSY breaking) as well as 4-fermion-terms and couplings
between fermions and the gauge field strength.
2.13 Problems
2.13.1 Simple Manipulations Within the Superspace Approach
Tasks
(1) Check that, with our upper-left/lower-right convention for contracting Weyl
indices, ψχ = χψ. Check that consistency requires ∂α ∂β = −∂β ∂α . Check
that, again for consistency, one must have (∂α )∗ = −∂ α̇ .
(2) Check as many of the anticommutation relations between Q, Q, D and D as
you need to feel confident.
(3) Derive the transformation rules for the components of the chiral superfield.
Hints Mostly straightforward manipulations—no hints needed. Recall that

(AB)∗ = B ∗ A∗ for an abstract algebra with a ∗-operation. When solving (3),
it is useful to first work out δξ θ , δξ θ 2 , δξ y μ and δξ f (y) for a generic function y.
Solution
(1) One immediately finds
ψχ = ψ α χα = ψ α
αβ χ β = χ β
βα ψ α = χ β ψβ = χψ . (2.121)
Furthermore,
∂1 ∂2 θ 2 θ 1 = 1 ∂2 ∂1 θ 2 θ 1 = −∂2 ∂1 θ 1 θ 2 = −1 . (2.122)
Next, one obviously has (∂α θ β )∗ = δα β . By contrast, one may also evaluate this
by first using the rules of an abstract algebra with a ‘∗’ and differentiating only
after that. In other words, consider
←−−−−−
(∂α θ β )∗ = θ (−∂ α̇ ) ,
β̇
(2.123)
is problematic as global symmetries are expected to be inconsistent with quantum gravity, see
e.g. [99, 100].
2.13 Problems 69
where the arrow indicates that the derivative still acts on the variable. Also,
we have to impose α = α̇ and β = β̇. Now, since Grassmann objects always
anticommute, we also have
←−−−−−
β̇ β̇
θ (−∂ α̇ ) = ∂ α̇ θ = δα̇ β̇ = δα β , (2.124)
as desired. Clearly, the minus sign in the action of the ‘∗’ on derivatives was
needed to get this consistent result.
(2) Using the definitions in the lecture, we have
β̇
{Qα , Qα̇ } = {∂α − i(σ μ )α β̇ θ ∂μ ,
−∂ α̇ + iθ β (σ ν )β α̇ ∂ν } = i(σ ν )α α̇ ∂ν + (−1)(−i)(σ μ )α α̇ ∂μ . (2.125)
Here we used the fact that non-zero contributions only arise from the first term
of Q acting on the second term of Q and vice versa. The resulting contributions
add up giving the overall factor of 2 in the commutator given in the lecture.
It is clear that, for two Q’s, the result will be zero since each term vanishes
separately. Also, for Q and D the result is zero on account of the sign flip in
the definition of D: the analogues of the final two terms in (2.125) cancel in this
case.
(3) We need to calculate
√
δξ (y, θ ) = [(ξ ∂ −iξ σ μ θ ∂μ )+h.c.] (A(y)+ 2θ ψ(y)+θ 2 F (y)) . (2.126)
We first note that
(δξ θ )α = ξ β ∂β θ α = ξ β or δξ θ = ξ . (2.127)
Similarly,
δξ θ 2 = ξ α ∂α θ β θβ = ξ θ + θ β ξ α ∂α θβ = ξ θ − θβ ξ α ∂α θ β = 2ξ θ . (2.128)
Furthermore, for a generic function f (y), we have
δξ f (y) = (∂μ f (y)) δξ y μ (2.129)
and
δξ y μ = [(ξ ∂ − iξ σ ν θ ∂ν ) + (−ξ ∂ + iθ σ ν ξ ∂ν )] (x μ + iθ σ μ θ )
= iξ σ μ θ − iξ σ μ θ + iθ σ μ ξ + iθ σ μ ξ = 2iθ σ μ ξ . (2.130)
Note that, to get the sign of the third term in the second line right, one needs to
take into account that
β̇ α̇ β̇ α̇ β̇ β̇
ξ ∂θ = ξ α̇ ∂ θ = −ξ ∂ α̇ θ = −ξ . (2.131)
After these preliminaries, one immediately finds

√ √
δξ = (∂μ A) (2iθ σ μ ξ ) + 2ξ ψ + 2(θ ∂μ ψ) (2iθ σ μ ξ ) + 2(ξ θ )F (2.132)
√ √ √ √ √
= 1 · ( 2ξ ψ) + 2θ ( 2iσ μ ξ ∂μ A + 2ξ F ) + θ 2 (− 2i(∂μ ψ)σ μ ξ ) .
Here, to derive the last term, we used
1
θ α θ β = −
αβ θ 2 . (2.133)
2
The second line of (2.132) is already in a form which allows√ one to directly read off
the quantities δξ A, δξ ψ and δξ F as the coefficients of 1, 2θ , and θ 2 . To match this
with the formula given in the lecture, one also needs to use the relation
ψσ μ ξ = −ξ σ μ ψ. (2.134)
This relation is easily derived using the definition of σ in the lectures through
complex conjugation. One also needs the hermiticity of Pauli matrices.
2.13.2 Deriving Component Actions
Task Consider a generic chiral superfield model defined by a Kahler potential

j
K(i , ) and a superpotential W (i ). The full component lagrangian reads
j j 1 j l
L = −gij (∂μ Ai )(∂ μ A ) − igij ψ σ μ Dμ ψ i + Rij kl ψ i ψ k ψ ψ
4
1
− (Di Dj W )ψ i ψ j + h.c. − g ij (Di W )(Dj W ) . (2.135)
2
Here
∂ ∂
∂i = , ∂ı = , gij = ∂i ∂ j K = Kij , ij k = g kl ∂i gj l ,
∂i ∂
ı
Rij kl = gml ∂j ik m ,
Di W = ∂i W , Di Dj W = ∂i (Dj W ) − ij k (Dk W ) ,

Dμ ψ i = ∂μ ψ i + j k i (∂μ Aj )ψ k . (2.136)
2.13 Problems 71
Note that and R are exactly the same Christoffel symbols and Riemann tensor
that are familiar from general relativity. The formulae only look slightly different
since we parameterise the manifold using complex coordinates and they are slightly
simpler than usual because the metric is not generic but a Kahler metric. The
covariant derivative Dμ has nothing to do with spacetime being curved (it is not)
but rather related to the fact that ψ lives in a bundle over the scalar manifold
parameterised by A. Thus, comparing ψ at two different points in x requires
knowledge of the values of A at these points.
Derive the first two and the last term in (2.135). If you wish, try also the others.
Hints You can save work by shifting x under the integral:
x μ + iθ σ μ θ , x μ − iθ σ μ θ −→ x μ , x μ − 2iθ σ μ θ . (2.137)
Independently, prove and use the formula
1 2
(θ σ μ θ )(θ σ ν θ ) = − θ 2 θ ημν . (2.138)
2
Solution We start with the last formula. It is clear that the l.h. side must be
1 2 2
proportional to θ 1 θ 2 θ θ and hence to θ 2 θ . The latter is a scalar, so it must be
multiplied by an invariant tensor with indices μ and ν, where ημν is the only choice.
Thus, one only needs to check normalisation. This is done most easily by focussing
on μ = ν = 0:

1 2 2 1 2
(θ σ 0 θ )2 = θ 1 θ + θ 2 θ = −2θ 1θ 2 θ θ . (2.139)
We also have
θ 2 = θ α
αβ θ β = 2θ 1 θ 2 (2.140)
and hence
2 1 2
θ 2 θ = −4θ 1 θ 2 θ θ . (2.141)
Recalling that we use the mostly-plus metric, the result follows.
Now we proceed to evaluate the D term of the Kahler potential. Since we are
only interested in the kinetic term of A, we can set ψ and F to zero. Thus, with the
shift of variables explained above, we have to evaluate

ı
K Ai (x), A (x − 2iθ σ θ) 2 . (2.142)
θ 2θ
This is done by first Taylor expanding A,

ı ı 2 ı
K Ai , A − 2iθ σ μ θ ∂μ A + θ 2 θ ∂ 2 A , (2.143)
where we used (2.138) to simplify the quadratic term in the expansion. Next we
Taylor expand K, keeping only what will contribute to the D term:

2 ı
K 2 = Kı (A, A) θ 2 θ ∂ 2 A
θ 2θ

1 ı j
+ Kıj (A, A) (2iθ σ μ θ ∂μ A ) (2iθ σ ν θ ∂ν A ) 2 2 . (2.144)
2 θ θ
The second term can again be simplified using (2.138), which gives

ı ı j
K 2 = Kı (A, A) ∂ 2 A + Kıj (A, A) (∂μ A ) (∂ μ A ) (2.145)
θ 2θ
ı ı j
= −∂μ (Kı (A, A) ∂ μ A + Kıj (A, A) (∂μ A ) (∂ μ A ) + total derivative
ı
= −Kj ı (∂μ Aj )(∂ μ A ) + total derivative .
This is our desired result.

To derive the last term in (2.135), we only need to consider the terms involving
2
F . It is clear that the Taylor expansion in θ 2 and θ gives

j
K 2 ⊃ Kij F i F and W ⊃ Wi F i + h.c. (2.146)
θ 2θ θ2
j
Varying w.r.t. F one finds
W j + Kij F i = 0 and hence F i = −g ij W j . (2.147)
Inserting this in the three terms of (2.146), the result
L ⊃ −g ij Wi W j (2.148)
eventually follows.
Let us finally consider the fermion kinetic term. It will be convenient to shift the
variable such that we have to deal with

K((x + 2iθ σ θ), (x)) 2 2 . (2.149)
θ θ
2.13 Problems 73
Now, suppressing the spacetime arguments and the projection on the highest com-
ponent for brevity, we expand the chiral superfields in the fermionic components:
2Kij (θ ψ) (θ ψ) . (2.150)
Then we expand ψ to linear order in the quantity 2iθ σ μ θ :
j
4Kij (θ ψ ) (θ ∂μ ψ i ) (iθ σ μ θ ) (2.151)
At this point we have to employ (2.133) and the hermitian conjugate relation
β̇ α̇ 1 2
θ θ = −
αβ θ . (2.152)
2
Thus, we have
α̇ μ β̇ 1 α̇ μ β̇
(θ α̇ ψ ) (θ α ∂μ ψα ) (iθ β σβ β̇ θ ) = − i(θ α̇ ψ ) (∂μ ψ β σβ β̇ θ ) θ 2
2
1 2
= i∂μ ψσ μ ψ θ 2 θ . (2.153)
4
Now we use the relation
ψσ μ χ = −χσ μ ψ , (2.154)
which follows from the hermiticity of σ matrices and the anticommutation of

2
spinors. Moreover, we implement the θ 2 θ projection. This gives
j
− i Kij ψ σ μ ∂μ ψ i . (2.155)
With the renaming Kij → gij this is the partial-derivative part of our kinetic term.
We still have to find the term responsible for its covariantisation. For this, we
note that we obtained the term 2iθ σ μ θ from expanding ψ. But we could equally
well have expanded A in Kij to obtain this term. The calculation proceeds precisely
as above, but in the final formula ∂μ acting on ψ is dropped. Instead, one has to
replace Kij by
∂k Kij ∂μ Ak . (2.156)
Thus, we finally have the term
j
− i∂k gij ∂μ Ak ψ σ μ ψ i . (2.157)
To see that this is what we want, we work backward from (2.135) and rewrite the
corresponding term:
j j
− igij ψ σ μ j k i (∂μ Aj )ψ k = −igij ψ σ μ g il ∂j gkl (∂μ Aj )ψ k . (2.158)
Now the agreement is apparent.
2.13.3 Fierz Identities for Weyl Spinors
Task Derive the covariant orthonormality condition for σ matrices
(σμ )α α̇ (σ ν )α̇α = −2ημν . (2.159)
Use it to simplify expressions like (σμ )α α̇ (σ μ )β β̇ and (σμ )α α̇ (σ μ )β̇β . From this,
Fierz identities like
(φσ μ χ)(ψσμ η) = −2(ψφ)(χη) and (φσ μ χ)(ησ μ ψ) = 2(ψφ)(χ η)

(2.160)
immediately follow. One can use those to replace bi-spinors within some longer
expressions according to
1
(· · · χψ · · · ) = (· · · σ μ · · · )(ψσμ χ) and
2
1
(· · · ψχ · · · ) = (· · · σ μ · · · )(χσ μ ψ) . (2.161)
2
Hints and Background Fierz identities are probably familiar in the context of
Dirac spinors, where they are also used to rewrite expressions with four spinors in
such a way that the pairs connected by index contraction (possibly through γ matri-
ces) change. The basic underlying idea making this possible is the completeness of
{1, γμ , γ5 , γμ γ5 , [γμ , γν ]} in the space of 4×4 matrices. In our context, things are
much simpler since the 4 σ -matrices already provide a basis of the space of 2 × 2
matrices.
Solution Let us start by rewriting the second matrix on the l.h. side of (2.159)
according to
(σ ν )α̇α = (σν )α α̇ =
αβ
α̇ β̇ (σν )β β̇ = [(iσ2 )σν (−iσ2 )]α α̇
= [{σ0 , −σ1 , σ2 , −σ3 }]α α̇ = [{σ0 , −σi }]α̇α . (2.162)

2.13 Problems 75
With this and the usual orthonormality relations between the Pauli matrices and the
unit matrix, the r.h. side of (2.159) immediately follows.
Now we recall that the σ matrices form a basis of 2×2 hermitian matrices. In
fact, over the complex numbers they are a basis of all 2×2 matrices. Hence we have
Mα α̇ = M μ (σμ )α α̇ (2.163)
for generic Mα α̇ and appropriate coefficients M μ . Multiplying by (σ ν )α̇α and using

(2.159), one finds
Mα α̇ (σ ν )α̇α = −2M μ ημν . (2.164)
Solving this for M μ and inserting in (2.163) gives
1
− Mβ β̇ (σ μ )β̇β (σμ )α α̇ = Mα α̇ = Mβ β̇ δα β δα̇ β̇ (2.165)
2
or, since M was generic,
(σμ )α α̇ (σ μ )β̇β = −2δα β δα̇ β̇ . (2.166)
Using the hermiticity of σ matrices and lowering the indices, one then also has
(σμ )α α̇ (σ μ )β β̇ = −2
αβ
α̇β̇ . (2.167)
From this, the identities in (2.160) follow straightforwardly by multiplication and

contraction with four spinors, where one has of course to be very careful with
the spinor ordering and signs. Finally, (2.161) provides two different ways for
reinterpreting (2.160) as a method for replacing two spinors within a longer string
of Weyl spinor expressions.
2.13.4 SUSY in Components
Task Demonstrate that the SUSY algebra is represented on the scalar (or chiral)
multiplet, without using superspace. Realise SUSY without the auxiliary field (just
on A and ψ) by allowing yourself to use the equations of motion (i.e. working on-
shell).
Hints As explained, while SUSY is very conveniently derived in superspace, it can

also be discussed entirely at the level of component fields. This is important since
in many cases (in higher dimensions, in many supergravity theories, or in situations
with more than the minimal set of Q’s, also known as N = 2 or N = 4 SUSY),
no superspace description exists or is not efficient. To discuss this component

description, one focuses on the bosonic generators
δξ = ξ Q + ξ Q . (2.168)
Their algebra, defined with commutators, is equivalent to the SUSY algebra. Start by
calculating [δξ , δη ] using the known algebra of the Q’s. Then check that the algebra
is represented on the components by using the explicit expressions for δξ A, δξ ψ and
δξ F that were given in the lecture and that have already been derived in a previous
exercise. Show also that the algebra still ‘closes’ (a common synonym for being
represented on a certain set of fields) if δξ F is dropped and, in the other expressions,
F is replaced using the equations of motion. (For simplicity, we consider the free
case and hence free equations of motion.) Note that in this latter case one has to use
equations of motion ‘to close the algebra’. One also says that the algebra is only
realised ‘on-shell’.
Use the Fierz identities and try not to get lost in the many spinors and indices,
especially when evaluating the algebra on ψ.
Solution First, one has

α̇
[ξ Q, ηQ] = ξ α Qα Qα̇ ηα̇ − ηα̇ Q Qα ξα = ξ α {Qα , Qα̇ }ηα̇ = 2ξ σ μ η Pμ
= −2i ξ σ μ η ∂μ (2.169)
and hence
[δξ , δη ] = [ξ Q, ηQ] + [ξ Q, ηQ] = [ξ Q, ηQ] − (ξ ↔ η) = −2i(ξ σ μ η − ησ μ ξ ) ∂μ .

(2.170)
To see that this explicitly holds for the scalar multiplet, we start with the scalar
component that gives this multiplet its name:
√
[δξ , δη ]A = δξ δη A − (ξ ↔ η) = δξ 2ηψ − (ξ ↔ η)
√ √ √
= 2η(i 2ησ μ ξ ∂μ A + 2ξ F ) − (ξ ↔ η)
= 2iησ μ ξ ∂μ A − (ξ ↔ η) = −2iξ σ μ η ∂μ A − (ξ ↔ η) . (2.171)
This is the desired result.
The analogous calculation for the fermion is slightly more involved:

√ √
[δξ , δη ]ψ = δξ (i 2σ μ η∂μ A + 2ηF ) − (ξ ↔ η)
√ √ √ √
= i 2σ μ η ∂μ ( 2ξ ψ) + 2η i 2 ξ σ μ ∂μ ψ − (ξ ↔ η) (2.172)
= 2i(σ μ η)(ξ ∂μ ψ) + 2iη(ξ σ μ ∂μ ψ) − (ξ ↔ η) .
2.13 Problems 77
Here in the last line we have introduced (formally superfluous) brackets to empha-
sise where the consecutive contraction of Weyl indices is interrupted. Now, using
the two Fierz-type identities in (2.161), we rewrite the terms in such a way that ξ
and η (or η and ξ ) are contracted with each other through one σ matrix:
[δξ , δη ]ψ = i(ξ σν η)(σ μ σ ν ∂μ ψ) + i(ξ σ ν η)(σ ν σ μ ∂μ ψ) − (ξ ↔ η) . (2.173)
At this point, it is convenient to make the two explicitly written terms more similar
by exchanging ξ and η in the second term (together with a sign change):
[δξ , δη ]ψ = i(ξ σν η)(σ μ σ ν ∂μ ψ) − i(ησ ν ξ )(σ ν σ μ ∂μ ψ) − (ξ ↔ η) . (2.174)
Next, employing the hermiticity of σ matrices, we may replace σ by σ in the second

term. The re-ordering of spinors which is then also necessary introduces a further
sign change:
[δξ , δη ]ψ = i(ξ σν η)(σ μ σ ν ∂μ ψ) + i(ξ σν η)(σ ν σ μ ∂μ ψ) − (ξ ↔ η) . (2.175)
Finally, using the Clifford-algebra-type relation σμ σ ν + σν σ μ = −2ημν 1 (analo-

gous to (1.131)), the desired result follows.
The calculation for the auxiliary field is again simpler:
√
[δξ , δη ]F = δξ i 2ησ μ ∂μ ψ − (ξ ↔ η)
√ √ √
= i 2ησ μ ∂μ [i 2σ ν ξ ∂ν A + 2ξ F ] − (ξ ↔ η) (2.176)
= −2ησ σ ξ ∂μ ∂ν A + 2iησ ξ ∂μ F − (ξ ↔ η) .
μ ν μ
Here, the first term in the second line simplifies if one uses the symmetry of
∂μ ∂ν to replace the product of σ matrices by −ημν 1. After this, the expression
is proportional to ξ η and vanishes upon ξ -η-antisymmetrisation. The second term
in the last line of (2.176) provides, after rewriting in terms of σ μ , our desired result.
Finally, we want to repeat the calculations for [δξ , δη ] on A and on ψ with the
auxiliary replaced according to the equations of motion. Specifically for the free
theory, that means
F = −mA , (2.177)
such that we now work with the SUSY transformation rules

√
δξ A = 2ξ ψ (2.178)
√ μ √
δξ ψ = i 2σ ξ ∂μ A − m 2ξ A . (2.179)
In the analysis of [δξ , δη ]A we do not even need the term with m which formerly
involved F . As we can see be revisiting our calculation above, this term simply
drops out under ξ -η-antisymmetrisation. By contrast, in the fermion case the last
line of (2.172) is replaced by
√
[δξ , δη ]ψ = 2i(σ μ η)(ξ ∂μ ψ) − δξ m 2ηA − (ξ ↔ η) . (2.180)
We can employ the equation of motion iσ μ ∂μ ψ + mψ = 0 to perform the rewriting

√ √
m δξ A = 2m(ξ ψ) = −i 2(ξ σ μ ∂μ ψ) , (2.181)
After that, (2.180) takes precisely the form of the last line of (2.172). But from there,
we already know how to arrive at the desired result, so we are done. In summary,
if one is prepared to use the equations of motion, one can indeed live without the
auxiliary field (on-shell SUSY).
The reader may want to continue this exercise independently by also checking
the invariance of the free lagrangian, off-shell and on-shell.
2.13.5 Gauge Coupling Unification
Task Demonstrate that precision gauge coupling unification in the SU (5) scheme
does not work well in the Standard Model but, by contrast, works extremely well
with low-scale supersymmetry.
Hints Recall that the beta function of a gauge theory with coupling g is commonly
defined as
dg b g3
β(g) = = + ··· . (2.182)
d ln μ 16π 2
Here in the last expression we gave the leading-order result with the widely used
‘beta-function-coefficient’ b encoding the numerical prefactor. For a U (1) gauge
theory one explicitly finds
q2
b= c with c = 2 / 4 / − 22 for a complex scalar/Weyl fermion/real vector
6
(2.183)
with charge q running in the loop. Here the last option is somewhat formal: Indeed,
while a charged scalar or Weyl fermion is easy to add to an abelian gauge theory,
adding a charged complex vector is somewhat artificial. More naturally, one would
view such a vector as the combination of two real vectors, each of which corresponds
to an extra U (1) gauge theory. This type of charged matter does, in turn, appear
naturally if our original U (1) is viewed as a subgroup of a non-abelian gauge group.
For this reason it is in fact useful to know the above numerical value of ‘−22’.
2.13 Problems 79
The derivation of these three values of c needs only the calculation of the log-
divergence in the familiar vacuum-polarisation or self-energy diagram and can be
found in many QFT textbooks, e.g. [1].
Obviously, the non-abelian case requires the substitution
q 2 → tr(TRa TRb ) ≡ TR δ ab (2.184)
in the relevant self-energy diagram, where R stands for the representation in which
the matter in the loop transforms. Here TR is the so-called Dynkin-index of the
representation R. The corresponding substitution in the beta function coefficient
hence reads q 2 → TR . Concretely, one has TF = 1/2 and TA = N for the
fundamental and adjoint of SU (N). One sometimes also refers to TA = T (A) =
C2 (A) as the quadratic Casmir of the adjoint representation.

It is now straightforward to obtain the values of b1,2,3 and b1,2,3 for the running of
the couplings of U (1), SU (2) and SU (3) in the Standard Model and the MSSM. It is
convenient to work with quantities like αi−1 since solving the renormalisation group
equation for these inverse squared couplings is particularly easy. Moreover, it is
useful to work with α12 ≡ α1−1 − α2−1 , etc. Also, please use SU (5)-normalisation
for the U (1) gauge coupling. Calculate the values of the mass scales M12 , M23
and M13 at which the various gauge couplings meet in the Standard Model and the
MSSM (with initial values for αi and SUSY breaking at mZ , to keep things simple).
Finally, turn the logic around and derive the predicted value of α3 at mZ as it follows
from the GUT hypothesis and the values of α1,2 at mZ .
Solution Let us start with the Standard Model and with b3 . We have contributions
from the triplets (or equivalently anti-triplets) corresponding to l.h. and r.h. up and
down-type quarks as well as from the gluons:

1 1 4
b3 = 4 · 2 · 2 · Nf · − 22 · 3 = Nf − 11 = −7 . (2.185)
6 2 3
Here, in the first term, the 4 comes from the Weyl fermion nature of our matter, the
2·2 from l.h./r.h. and up/down, the Nf = 3 from the three families, and the 1/2 from
TF = 1/2. In the second term we have the −22 from the vector nature of the gluons
and the 3 from TA = N = 3.
Next, we consider SU (2):

1 1 1 4 43 19
b2 = 4 · (3 + 1) · Nf · + 2 · − 22 · 2 = Nf − =− . (2.186)
6 2 2 3 6 6
Here, in the first term we have again a 4 from the Weyl fermion nature, a (3 + 1)
from the 3 colours of the quark doublet and the 1 lepton doublet, as well as Nf /2
as above. In the second term we have a 2 from the scalar nature of the Higgs as well
as TF = 1/2. The third term is self-explanatory, with TA = N = 2.
Finally, for U (1) we have:

2 2 2
1 1 2 2 1 1
b1 = 4 6· +3 +3 +2 + 1 Nf
2
6 6 3 3 2
2
1 3
+2 · 2
2 5
4 1 41
= Nf + = . (2.187)
3 10 10
Here the five terms inside the square bracket correspond to the contributions from
quark doublet, up and down-quark, lepton doublet and r.h. electron. The additional
contribution outside the square bracket comes from the Higgs, with a factor 2
because it is a complex scalar and another 2 because it is a doublet. Finally, the
charges are given in Standard Model hypercharge normalisation, which is corrected
by the explicit factor of 3/5 to bring us to the right normalisation for the beta
function coefficient of the U (1) as a subgroup of SU (5) (cf. the solution to
Problem 1.9.3).
The reader will not be surprised to note that the matter contribution to all bi is
the same since, as we already know, matter comes in complete SU (5) multiplets.
To get the SUSY version of the above, one needs to add the effects of gauginos,
extra Higgs and Higgsino fields, and sfermions. The gauginos give
g g 1 4 g 1
b1 = 0 , b2 = ·4·2= , b3 = · 4·3 = 2. (2.188)
6 3 6
Here the 4 comes from the gauginos being Weyl fermions and the (0, 2, 3) are the
relevant values of TA .
The Higgs contribution receives a factor of two compared to the Standard Model
because we now have two Higgs doublets. In addition, we have to replace 2 →
2 + 4, since instead of a complex scalar we now have a complex scalar and a Weyl
fermion. This amounts to a total factor of 6 or, equivalently, an additional term worth
five times the Standard Model Higgs effect. Using the Higgs part of the previous
analysis, this gives
1 5
b1h = , b2h = , b3h = 0 . (2.189)
2 6
Finally, the matter part suffers the substitution 4 → 4 + 2, i.e. an additional term
worth one half of the previous value. This means
b1m = b2m = b3m = 2 . (2.190)

2.13 Problems 81
Adding everything up and also displaying the Standard Model coefficients again
for easier reference, we now finally have

41 19 33
bi = , − , −7 and bi = , 1, −3 (2.191)
10 6 5
for the Standard Model and the MSSM, respectively.

For the rest of the exercise, our basic numerical input is
2π 2π 2π
= 370.7 , = 185.8 , = 53.2 . (2.192)
α1 α2 α3
The first two values are in fact known with much more precision than displayed
above. The last corresponds to α3 = 0.118 at mZ —by now also a very well
measured quantity. We have already discussed these numbers very roughly in
Problem 1.9.3, but here we wanted to be a bit more precise. The standard source
for such data is the Review of Particle Properties of the Particle Data Group (PDG)
[43].
On the analytic side, our main input are the three equations
bi
αi−1 (μ) = − ln(μ) + (const.)i . (2.193)
2π
Starting at some high scale M and running down to mZ this gives

bi M
αi−1 (mZ ) = αi−1 (M) + ln . (2.194)
2π mZ
Specifically, if we assume that the two couplings α1 and α2 become equal at the
scale M = M12 , then we deduce

b12 M12
α12 (mZ ) = ln , (2.195)
2π mZ
where α12 ≡ α1−1 − α2−1 and b12 ≡ b1 − b2 . We find

2πα12 (mZ ) 370.7 − 185.8
M12 = mZ exp = 90 GeV exp
b12 41/10 − (−19/6))
(2.196)
and, using analogous formulae for the other ‘unification scales’, we find
M12 = 1.0 × 1013 GeV , M23 = 9.5 × 1016 GeV , M13 = 2.4 × 1014 GeV
(2.197)
Fig. 2.1 One-loop running

of inverse gauge couplings in
the Standard Model
in the Standard Model. The running of inverse gauge couplings that corresponds
to these results is sketched in Fig. 2.1. We see that gauge couplings do not really
unify and the so-called grand unification scale MG remains somewhat vague, with
a value in the range of 1013 · · · 1017 GeV. Alternatively, one may define MG by the
unification of α1 and α2 , and attempt to predict α3 at the weak scale by running it
backwards from that point using b3 . This is illustrated in the figure by the dashed
line, and it is apparent that this prediction will not be very good.
By contrast, as one now immediately verifies using the formulae above, the same
analysis in the MSSM with SUSY breaking at mZ gives
= 2.0 × 1016 GeV ,

M12 = 2.2 × 1016 GeV ,
M23 = 2.1 × 1016 GeV .
M13
(2.198)
This has been celebrated as a great success of the SUSY-GUT idea, the scale of
which is hence quantitatively fixed: MG 2 × 1016 GeV. However, to a certain
extent this perfection is accidental, as we will explain after turning the argument
around to predict α3 (mZ ).
To derive this prediction, one combines (2.195) with its analogue for α13 , under
the assumption that M12 = M13 = MG . Eliminating MG , one finds
α12 (mZ )/b12 = α13 (mZ )/b13 (2.199)
or
b13
α3−1 (mZ ) = α1−1 (mZ ) − α12 (mZ ) , (2.200)
b12
pred.
implying the predicted value α3 (mZ ) 0.117 . The corresponding non-SUSY
prediction would be 0.071, i.e. completely off.
2.13 Problems 83
But one should not overstate the perfection of the result above: There are 2-loop
corrections to the running, which are very well-understood and lift the prediction to
pred.
α3 (mZ ) 0.129, which is about 10% too large. This becomes slightly better but
still not perfect if one takes into account that SUSY is broken not at mZ but at least
at about a TeV. Finally, there are threshold corrections both at the SUSY breaking
and the GUT scale, which also affect unification. By this we mean effects arising
because not all SUSY partners and not all new GUT scale particles are degenerate at
the respective scales msof t and MG . Thus, SUSY unification works well but not as
perfectly as the naive 1-loop analysis suggests. It does in fact become even slightly
better if the SUSY breaking scale is raised above 1 TeV. However, one has to be
honest and admit that, once one gives up on the SUSY solution of the hierarchy
problem, the SUSY breaking scale could be anywhere and one cannot really claim
any more than one predicts α3 (mZ ). A few more details and references to many
much more detailed analyses can be found in the PDG review section on Grand
Unification.
2.13.6 Graviton Spin (Helicity)
Task Show that, under transverse rotations by an angle φ, a linear superposition

of the two physical photon states can be represented by a complex number rotating
with a phase exp(iφ). Show that, analogously, the general physical graviton state
rotates twice as fast (i.e. that ‘the graviton has spin 2’).
Hints Let the photon momentum be k ∼ (1, 1, 0, 0)T . Then transversality

· k = 0
together with the gauge choice
0 = 0 leaves the two basis polarisations
⎛ ⎞ ⎛ ⎞
0 0
⎜0⎟ ⎜0⎟

(1) =⎜ ⎟
⎝1⎠ ,
(2) =⎜ ⎟
⎝0⎠ . (2.201)
0 1
The general state can be characterised by α

(1) + β
(2) or, equivalently, by

α
∈ R2 or α + iβ ∈ C . (2.202)
β
Similarly, under the constraints of transversality and tracelessness (

μν ημν = 0),
the graviton polarisation basis for k ∼ (1, 1, 0, 0)T is
⎛ ⎞ ⎛ ⎞
000 0 0 00 0
⎜0 0 0 0⎟ ⎜0 00 0⎟

(1) =⎜
⎝0 0 1 0⎠ ,
⎟
(2) =⎜
⎝0
⎟. (2.203)
00 1⎠
0 0 0 −1 0 01 0
Again, the general state is α

(1) + β
(2) and the real or complex representation is
provided by (2.202).
Solution The relevant Lorentz transformation reads

⎛ ⎞
1 00 0
⎜0 1 0 0⎟ c = cos φ

μ → μ ν
ν with =⎜
⎝0
⎟ and .
0 c −s ⎠ s = sin φ
0 0s c
(2.204)
The vector (α, β)T transforms by a φ-rotation by definition. Elementary complex

algebra then implies that
α + iβ → α + iβ = eiφ (α + iβ) . (2.205)
For the graviton, the general state can be represented by

α β
(2.206)
β −α
and the transformed state is

α β c −s α β cs cα − sβ cβ + sα cs
= =
β −α s c β −α −s c sα + cβ sβ − cα −s c

c2 α − scβ − csβ − s 2 α scα − s 2 β + c2 β + scα
=
csα + c2 β − s 2 β + csα s 2 α + scβ + scβ − c2 α

c α − s β s α + c β
= (2.207)
s α + c β −(c α − s β)

with c = c2 − s 2 = cos 2φ and s = 2sc = sin 2φ. Hence

α c −s α
= (2.208)
β s c β
and, in the complex plane,
α + iβ → α + iβ = e2iφ (α + iβ) . (2.209)

String Theory: Bosonic String
3
3.1 Strings: Basic Ideas
What we have achieved so far is not entirely satisfactory: Supersymmetry (more

precisely, the broader framework of supergravity) offers a partial solution to the
weak-scale hierarchy problem. Partial refers to the fact that SUSY partners have
not been discovered (yet?) and hence some fine-tuning is probably needed after all.
Supergravity is needed to combine this with general relativity, but it does not help
with the cosmological constant problem, which unavoidably shows up in this con-
text. Technically, the cosmological constant can be anything in supergravity: It can
be negative due to the −3|W |2 term, or positive due to a dominant |DW |2 term (with
SUSY spontaneously broken). It is also affected by UV divergences since (in spite of
the non-renormalisation theorems for W ), the Kahler potential K is loop corrected.
Moreover, the UV problems of gravity (all operators being generated at the scale
MP —i.e. formal ‘non-renormalisability’) are not resolved by the prefix ‘super’.
The string idea is illustrated in Fig. 3.1 and states simply that point-particles
should be replaced by little loops of fundamental string. This might help with UV
divergences (especially in gravity) since the interaction point is gone. Hence, when
calculating a loop, there is no way in which this loop can go to zero size by the
(e.g. two) interaction points becoming infinitely close. Some of the many standard
textbooks are [7, 8, 101–103].
But before discussing scattering, we will of course have to understand how a
single string loop moves through space (in other words, how its worldsheet is
embedded in target space, more precisely, in target spacetime), see Fig. 3.2. Before
doing so, let us consider the more familiar case of a point particle, cf. Fig. 3.3. The
embedding of the worldline γ in target space is specified by the set of functions
Xμ (τ ) and the natural action is

S = −m ds with ds 2 = −ημν dXμ dXν and dXμ = Ẋμ dτ .
γ
(3.1)

86 3 String Theory: Bosonic String
Fig. 3.1 Point particle scattering vs. string scattering
Fig. 3.2 String moving through target space
Fig. 3.3 Point particle through target space
More explicitly, this action can be written as

S = −m dτ −ημν Ẋμ Ẋν . (3.2)
One can easily check that this is reparameterisation invariant under τ → τ = τ (τ )

and that the non-relativistic limit is
m
S = dt v2 − m . (3.3)
2
Much more could be said about this simple and familiar system (see e.g. [104]), but
for now this will suffice to motivate the Nambu-Goto action for the string.
In complete analogy to the point particle, the Nambu-Goto action for the
bosonic string measures the surface area of the worldsheet embedded in target space:

SNG = −T df . (3.4)

To write this more explicitly, one parametrises the worldsheet by (cf. Fig. 3.2)
ξ ≡ (ξ 0 , ξ 1 ) ≡ (τ, σ ) . (3.5)
3.1 Strings: Basic Ideas 87
The surface area is nothing but the volume of the 2d manifold, parameterised by ξ ,
measured with the induced metric Gab . The latter is defined by
ds 2 = ημν dXμ dXν = ημν ∂a Xμ ∂b Xν dξ a dξ b ≡ Gab dξ a dξ b . (3.6)
Hence

√
SNG = −T d 2ξ −G with G ≡ det(Gab ) . (3.7)

The prefactor T specifies the string tension.

Due to the square root, the system is hard to quantise on the basis of this action.
Instead, one uses the classically equivalent Polyakov action [105–107]

T √
SP = − d 2ξ −h hab ∂a Xμ ∂b Xν ημν . (3.8)
2
Here we introduced a new degree of freedom—the worldsheet metric hab . To see

the equivalence, one integrates out hab by solving its equations of motion
√ δh √
0 = δh −hhab Gab = − √ hab Gab + −h δhab Gab . (3.9)
2 −h
One proceeds by observing that, for a generic matrix A,
δ(detA)/(detA) = δ ln(detA) = δ tr ln(A) (3.10)
and hence
δ (detA) = (detA) tr(A−1 δA) = −(detA) tr(A δA−1 ) . (3.11)
Applying this to δh, the equation of motion for hab becomes

h √
0 = δhab √ hab hcd Gcd + −hGab (3.12)
2 −h
√ √
or, using the identity h/ −h = − −h,
1
hab hcd Gcd = Gab . (3.13)
2
It is solved by hab = αGab for any α. Inserting this in the Polyakov action,

T √ T
SP = − 2
d ξ −h h Gcd cd
=− d 2ξ −α 2 G 2α −1 = SNG , (3.14)
2 2
one obtains the Nambu-Goto action.

At this point, jumping somewhat ahead, we can sketch what will follow: The
Polyakov action describes simply a 2d field theory of D free scalars, living on a
cylinder (S 1 × [Time]). This is a quantum mechanical system and its states have the
interpretation of particles living in the D-dimensional target spacetime. Consistency
will require D = 26, and the spectrum will contain a massless graviton and other
massless (as well as many heavy) fields. However, it will also contain a particle with
negative mass squared, a tachyon. Thus, the vacuum of the 26d gravitational field
theory which this bosonic string describes is unstable. This instability problem will
be cured if we move on to the superstring (based on a 2d supersymmetric worldsheet
theory). The target spacetime will then have to be 10d and contact with the real
world will be based on compactifying this 10d supergravity to 4d. The last step
means considering geometries M6 × R4 , with M6 a compact 6d manifold.
3.2 Symmetries, Equations of Motion, Gauge Choice
It is convenient to view the worldsheet theory as a 2d QFT with metric hab and D
free scalars Xμ :

T √
SP = − d 2 ξ −h (∂X)2 , (∂X)2 = hab (∂a Xμ )(∂b Xν )ημν . (3.15)
2
The three key symmetries of this theory are
1. Diffeomorphism: ξ a → ξ a (ξ 0 , ξ 1 ).
2. Poincare symmetry: Xμ → X μ = μ ν Xν + V ν with ∈ SO(1, D − 1).
3. Weyl rescalings: hab (ξ ) → h ab (ξ ) = hab (ξ ) exp[2ω(ξ )], with ω an arbitrary
real function.
The first and second are obvious and follow immediately from the structure of
our worldsheet action. It is noteworthy that target space Poincare symmetry is an
internal symmetry from the worldsheet perspective. The third is a specialty of the
string. In other words, for a similar theory of moving p-branes, parameterised by
ξ 0 , ξ 1 , · · · , ξ p , this symmetry does not exist unless p = 1.
To move on, it is convenient to use the energy-momentum tensor,
2 δS −2 δS
T MN = √ · or, equivalently TMN = √ · , (3.16)
−g δgMN −g δg MN
which takes the form TMN = diag(ρ, p, · · · , p) for an isotropic fluid. On the string
worldsheet, a slightly different normalisation is common:
−4π δSP
T ab = √ · . (3.17)
−h δhab
3.2 Symmetries, Equations of Motion, Gauge Choice 89
One easily checks that

1 1
T ab = − Gab − hab (hcd Gcd ) , (3.18)
α 2
where we also introduced the Regge slope
1
α ≡ . (3.19)
2πT
The latter is a different way to parameterise the string tension. It goes back to the
early days of string theory, when the focus was on string theory as a model of
hadronic physics. This is nicely explained in the first chapter of [101].
It follows both from our discussion in the last section as well as from the general
definition of T ab that the equation of motion of hab is
T ab = 0 . (3.20)
Moreover, tracelessness holds as an identity, i.e. independently of whether the field

configuration obeys the equations of motion:
T aa = 0 for any hab . (3.21)
The reader should convince herself that this generally follows from symmetry (3).
Finally, the equations of motion of X are
Xμ = 0 with = D a ∂a . (3.22)
It is crucial for what follows that diffeomorphisms and Weyl rescalings are (by
definition) not just symmetries but gauge redundancies. This allows one to work
in the flat gauge,
hab = diag(−1, 1) . (3.23)
Indeed, very superficially one can argue as follows: A 2d metric contains three
real functions. Diffeomorphisms and Weyl rescalings also contain 2 + 1 = 3 real
functions. Hence, it should be possible to bring hab to any desired form.
In somewhat more detail, one can explicitly check that
√ √
−h R[h ] = −h (R[h] − 2 ω) for h ab = e2ω hab . (3.24)
Now, starting from any metric h, one may try to solve the equation 2 ω = R. This
can always be achieved (in non-compact space with localised source R) since it only
requires the inversion of the Klein-Gordon operator. Without proof, we simply state
that this holds also on the cylinder, which is our case of interest. For more details,
see e.g. [8].
Once 2 ω = R is solved, one can Weyl rescale h using the solution ω. The
resulting metric will have vanishing Ricci scalar and, since in d = 2
1
Rabcd = (hac hbd − had hbc )R , (3.25)
2
it will be flat. More precisely, the worldsheet is a flat metric manifold and hence
there exist coordinates in which the metric is manifestly flat in the sense of (3.23).
Let us now focus on a flat worldsheet and on the corresponding equations of
motion
(∂τ2 − ∂σ2 )Xμ = 0 . (3.26)
It is convenient to use lightcone coordinates σ ± = τ ± σ , such that
ds 2 = −dτ 2 + dσ 2 = −dσ + dσ − and h++ = h−− = 0

1
h+− = h−+ = − (3.27)
2
and
∂
= −4∂+ ∂− with ∂± = . (3.28)
∂σ ±
The equations of motion take the form
∂− ∂+ Xμ = 0 (3.29)
and have the general solution
Xμ (σ + , σ − ) = XL (σ + ) + XR (σ − ) ,
μ μ
(3.30)
being further constrained by Xμ (τ, σ ) = Xμ (τ, σ + π), cf. Fig. 3.4. Here we have
used the reparameterisation freedom to set the circumference of the cylinder to π.
Fig. 3.4 The cylinder, on

which the X μ field theory
lives, viewed as a strip with
periodic boundary conditions.
μ μ
In this picture, XL and XR
correspond to left and
right-moving waves
3.2 Symmetries, Equations of Motion, Gauge Choice 91
This is a convention used in many string theory texts, in particular in [101] which
we mostly follow.
μ μ
Periodicity of Xμ implies periodicity of ∂+ Xμ = ∂+ XL and of ∂− Xμ = ∂− XR .
+ −
The latter depend only on σ and σ , respectively, and can therefore be represented
as Fourier series in these two variables:
+ −
fL, n e−2inσ , fL, n e−2inσ .
μ μ
∂+ XL ∼ const.L + ∂− XR ∼ const.R +
n=0 n=0
(3.31)
μ μ
Returning to XL and XR by integration, the exponentials remain exponentials and
the constants translate into linear terms. Moreover two integration constants appear.
Hence, with a certain choice of prefactors, one finds the general solution or mode
decomposition
μ 1 μ l 2 μ + il 1 μ −2inσ +
XL = x + p σ + α̃ e (3.32)
2 2 2 n n
n=0
μ 1 μ l 2 μ − il 1 μ −2inσ −
XR = x + p σ + α e . (3.33)
2 2 2 n n
n=0
√
Here we introduced l = 2α , the so-called string length. One should be aware that
the precise definition (the numerical prefactor) may vary from author to author and
from context to context.
The constants x μ /2 in the mode decomposition are chosen to be equal by
convention. It is only their sum that has physical meaning, characterising the
position of the centre of mass of the string at worldsheet time τ = 0. Note
that the coefficients of the two terms linear in σ + and σ − are forced to be
equal by the periodicity of Xμ . They describe how the position of the cen-
tre of mass changes as a function of τ . It is hence natural to identify these
coefficients, up to the proportionality factor l 2 /2, with the target space momen-
tum pμ . One could easily convince oneself at the present, classical level of
analysis that the proportionality factor has been chosen correctly for pμ to
be the standard momentum variable. But this will become clear anyway in a
moment. Reality of Xμ implies that x μ and pμ are real, consistently with their
physical meaning which we pointed out above. The oscillator modes have to
satisfy
(αnμ )∗ = α−n .
μ
(3.34)
3.3 Open String
It will later on be crucial to also consider open strings. We introduce them already
now since they are in fact a simpler version of the closed string—they basically carry
half of the degrees of freedom. Instead of a cylinder, one now has to think of a strip
(parameterised transversely by σ ∈ (0, π)) embedded in target space, cf. Fig. 3.5.
The variation of the action,
π
1 1
δS = d 2
σ (∂ 2
X) · δX − dτ dσ ∂σ (∂σ X · δX) , (3.35)
2πα 2πα 0
now includes boundary terms. Indeed, while the first term vanishes if the equations
of motion are obeyed, the second gives
σ =π
1
− dτ (∂σ Xμ ) · δXμ . (3.36)
2πα σ =0
To avoid introducing new degrees of freedom living at the boundary, we need

that expression to vanish as well. This can be achieved by two different types of
boundary conditions,
∂σ Xμ = 0 (Neumann) , δXμ = 0 (Dirichlet) . (3.37)
In the first case the string end moves freely (no momentum is lost at the end of
the string), in the second it is confined to lie in a fixed hyperplane. For example
(cf. Fig. 3.6), one can enforce Neumann boundary conditions for X0 , X2 and
Dirichlet boundary conditions for X1 . One is then dealing with an open string living
on a D-brane (where D stands for Dirichlet) filling out the X0 and X2 directions
of target spacetime. More generally, if a brane fills out p spatial dimensions, i.e.
if it is a p-dimensional object in the usual, spatial sense, one calls it a Dp-brane
Fig. 3.5 Open string
Fig. 3.6 Open string living

on a D1-brane filling out the
X 2 direction
3.3 Open String 93
Fig. 3.7 Various brane

configurations with strings
attached
(see [108] for a dedicated textbook and [109, 110] for two foundational original
papers). For target space to be stationary, branes always have to fill out the time
or X0 direction. This, of course, does not contribute to their dimensionality as a
spatial object. However, in spacetime a Dp-brane is a (p + 1)-dimensional object.
Quite generally, an appropriate combination of Neumann and Dirichlet boundary
conditions as given in (3.37) characterises open strings ending on such different
Dp-branes.
We furthermore note that configurations with various, also intersecting branes
are permitted, cf. Fig. 3.7. Jumping ahead, we record that, analogously to the closed
string states containing the target space graviton, the open-string states contain a
massless vector particle: a U (1) gauge boson. Thus, on every Dp brane one has
a localised (p + 1)-dimensional gauge theory. Moreover, one may have stacks of
branes, for example, N D-branes filling out exactly the same hyperplane, i.e. lying
on top of each other. On such a stack, there are N 2 distinct string states since each
string can begin or end on any one of these N coincident branes. This gives rise to
a U (N) gauge theory. If branes or brane stacks intersect, then the string living at
the intersection (as in the last picture in Fig. 3.7) gives rise to states (target space
particles or fields) which are charged under the two gauge groups corresponding
to the two branes. These states are confined to the intersection locus of the two
branes or brane stacks. This is how Standard Model matter fields arise in some of the
simplest phenomenologically interesting string models—the so-called intersecting
brane models (see [9] for a textbook, [111,112] for reviews and [113–115] for some
of the original papers).
What is interesting for us at the moment is that the mode decomposition of the
open string is simpler than that of the closed string. Indeed, while one needs sines
and cosines (or equivalently exponentials) to Fourier decompose a periodic function,
on an interval one can do with just sines or just cosines. Technically, one may say
(and it is easy to demonstrate this explicitly) that, for the open string, the left and
right-moving modes are identified: One arises from the other by reflection on the
boundary. Explicitly, for the case of Neumann boundary conditions, one has the
mode decomposition
1
Xμ = x μ + l 2 pμ τ + il α μ e−inτ cos(nσ ) . (3.38)
n n
n=0
Thus, it is often simpler to discuss the open string and then ‘double’ the result to go
over to the closed case.
We also note that the case of Neumann boundary conditions for all Xμ should
actually be viewed as a situation with spacetime filling branes. Thus, open strings
generally end on D-branes.
3.4 Quantisation
We will only present the old covariant approach, briefly commenting on lightcone
and modern covariant approach (also known as path integral or BRST quantisation)
at the end. The starting point is the flat-gauge Polyakov action which, breaking 2d
covariance, can be written as

1
S= d 2 σ (Ẋ2 − X 2 ) with d 2 σ = dτ dσ . (3.39)
4πα
Here we have left the index μ and its contraction implicit. Nevertheless, the above
describes D free bosons and we have to keep in mind that one of them (X0 ) has a
wrong-sign kinetic term.
The canonical variables are
∂L 1
Xμ and μ = = Ẋμ , (3.40)
∂ Ẋμ 2πα
with equal-time commutation relations
ˆ μ (τ, σ ), X̂ν (τ, σ )] = −iδ(σ − σ )δμ ν ,

[ ˆ μ,
[X̂μ , X̂ν ] = [ ˆ ν] = 0 .
(3.41)
Promoting our previous mode decomposition of Xμ (and a corresponding decom-

position of μ ) to the operator level, one finds
[p̂μ , x̂ ν ] = −iημν , [α̂m

μ
, α̂nν ] = m δm+n ημν , [α̃ˆ m
μ ˆν
, α̃n ] = m δm+n ημν ,
(3.42)
where
δm+n ≡ δm+n , 0 . (3.43)
We will drop the hats from now on, assuming that it will always be clear from
the context whether the operator or the classical variable is meant. The above
commutators make it apparent that pμ was correctly normalised to be the target
space momentum of the string.
As usual in quantum mechanics, we now need a Hilbert space representation of
our operator algebra. Given the non-trivial commutation relations of p and x, we
can only choose one of them to be diagonal. Since we are interested in a particle
3.4 Quantisation 95
interpretation of string states, it is natural to choose p and write

H= H(p) , (3.44)
p
where H(p) is the eigenspace of the operators {p̂0 , · · · , p̂D−1 } with eigenvalues
{p0 , · · · , pD−1 } ≡ p.
We now focus on the subspace corresponding to one particular value of p and
rewrite the mode-algebra acting on it:
[αm
μ
, αnν ] = m δm+n ημν → [αm
μ
, αnν † ] = |m|δm,n ημν .
(3.45)
We see that we are dealing simply with a very large set of oscillators, labelled by μ
and m > 0. We define a vacuum state and find the Fock space:

H(p) = Span αm αn · · · |0, p any number of α’s; any μ, ν, · · · ;
μ ν

any m, n, · · · > 0 . (3.46)
The situation we arrive at is very similar to the initial step of Gupta-Bleuler

quantisation of electrodynamics: There, on account of the vector index of Aμ and
the non-positive-definite metric ημν , the Fock space includes negative norm states.
They are removed by a physical state condition or constraint, related to the gauge
μ
invariance of the theory. Here, the same issue arises due to the vector index of αm .
As will become clear momentarily, the resolution is similar to case of quantum
electrodynamics (QED).
We fixed part of the gauge freedom by eliminating hab . The corresponding
equation of motion was Tab = 0, which now has to be implemented as a constraint.
It is convenient to do this in lightcone coordinates. One sees immediately that the
vanishing trace condition takes a particularly simple form:
Ta a = 0 ⇔ T+− = 0 . (3.47)
Now, since in our theory the trace vanishes identically, one only needs to enforce
the constraints
T++ = T−− = 0 . (3.48)
It is straightforward to check that
T++ = (∂+ XL ) · (∂+ XL ) and T−− = (∂− XR ) · (∂− XR ) (3.49)

and that, using the mode decomposition, the Fourier modes of these quantities read
∞
1 π 1
Lm ≡ dσ T−− e−2imσ = αm−n · αn (3.50)
4πα 0 2 n=−∞
∞
1 π 1
L̃m ≡ dσ T++ e2imσ = α̃m−n · α̃n . (3.51)
4πα 0 2 n=−∞
Here we also used the simplifying notation
μ μ l μ
α0 = α̃0 = p . (3.52)
2
For the open string, one defines
∞
1 π 1
L̃m ≡ dσ T++ e 2imσ
= α̃m−n · α̃n , (3.53)
2πα 0 2 n=−∞
with T−− being a dependent quantity. It is easy to see that
H = L0 + L̃0 (closed string) , H = L0 (open string) . (3.54)

μ
Note that α0 = lpμ for the open string.
One can check that the operators Lm (with or without tilde) satisfy the Virasoro
algebra
[Lm , Ln ] = (m − n)Lm+n + A(m)δm+n , with A(m) = (m3 − m)D/12 .

(3.55)
Here the term proportional to D is called the anomaly term and D is the central
charge. Note that this term depends on a possible additive redefinition of L0 , which
is related to the ordering ambiguity present in all the terms of type α−k αk in L0 . The
form given above assumes normal ordering, i.e. 0, 0|L0 |0, 0 = 0.
The classical part of this algebra, i.e. (3.55) without the anomaly term, is called
Witt algebra. It is satisfied by the differential operators
d
Dm = ieimθ , (3.56)
dθ
which generate diffeomorphisms on an S 1 parameterised by θ ∈ (0, 2π). These

remarks are the beginning of a long and important chapter of a proper string theory
3.4 Quantisation 97
Fig. 3.8 Illustration of a

conformal mapping of a given
field configuration to a new
one
course—2d conformal field theory. However, we are not going to discuss this, such
that a few comments will have to suffice:
When we fixed the gauge (diffeomorphisms and Weyl scalings), a residual gauge
freedom was left. It consists of diffeomorphisms under which the metric changes
only by Weyl scaling. Now it is useful to insist on the point of view that, after going
the flat gauge, we are in a fixed-background QFT and coordinate reparameterisations
are forbidden. From this perspective, the residual gauge freedom noted above corre-
sponds to spacetime dependent translations of the field configuration which preserve
angles, i.e. conformal transformations (Fig. 3.8). Our theory is invariant under those
and hence is a conformal field theory or CFT [116–121]. The Virasoro algebra
is the corresponding symmetry algebra. It is clear that conformal transformations
can be generated as spacetime dependent translations. Given that Tab generates
translations, we are not surprised to find that the Fourier modes of Tab are the desired
symmetry generators. It is also natural that the Witt algebra, as introduced above, is
the classical counterpart.
The conformal symmetry just introduced is a central tool in developing string
theory and, in particular, in deriving scattering amplitudes, loop corrections, etc.
We will have no time for this. But it may be useful to note that, when studying CFTs
in their own right, the anomaly term or, equivalently, a non-zero central charge do
not represent a problem. However, in string theory the conformal symmetry is part
of an underlying gauge symmetry and this term must vanish. It indeed does, in the
so-called critical dimensions, but to see this one needs to do the gauge fixing more
carefully, introducing Faddeev-Popov ghosts. They cancel the central charge coming
from the scalars.
Returning to our main line of development, we now want to be more explicit
about the physical state condition. As in QED, it is sufficient to demand that the
‘annihilator part’ of the constraint vanishes on physical states, i.e. Lm | phys = 0
for m ≥ 0. But it turns out that, at this point, a divergence present in the definition
of L0 has to be resolved. This has to do with operator ordering.
Indeed, our definition so far was
∞
1 1 1
(L0 )t ot = α−n αn = α02 + α−n αn . (3.57)
2 n=−∞ 2 2
n=0
We gave this quantity an index for ‘total’ since we are going to separate the normal-
ordered part from it in a moment. We also note that the ordering of the creation and
annihilation operators used above comes directly from the original definition
π
1
(L0 )t ot = Ht ot = dσ (Ẋ2 + X 2 ) . (3.58)
4πα 0
Here, for simplicity, we think of the open string or, equivalently, just the right-
moving part of the closed string.
To evaluate a constraint like (L0 )t ot | phys = 0, we want to work instead with
a normal-ordered operator. Hence, we define
∞
1 1
L0 ≡ : α−n αn := α02 + α−n αn . (3.59)
2 n=−∞ 2
n>0
Note that this supersedes our previous definition in (3.50). The two definitions differ
by a divergent normal ordering constant,
∞
1
(L0 )t ot = L0 − a with a = − (D − 2) n, (3.60)
2
n=1
following simply from (1/2)(αn αn† + αn† αn ) = αn† αn + n/2 . The prefactor (D − 2)
counts the number of oscillators that contribute. The direct calculation gives, of
course, D, but we have corrected this to (D − 2) on account of the wrong-sign
scalar X0 . This is necessary since this wrong sign-scalar is associated with negative
norm states, which are connected with the still unfixed (residual) gauge freedom.
The latter corresponds to conformal transformations or reparametrisations of type
σ+ → σ+ = σ+ (σ ) and σ− → σ− = σ− (σ ), which preserve the flat gauge. One
can fix this further gauge freedom by working in the so-called lightcone gauge. The
corresponding quantisation procedure, called lightcone quantisation, manifestly gets
rid of all oscillators except the (D − 2) transverse ones. We skip this important and
useful chapter and ask the reader to trust that (3.60) with the prefactor (D − 2) is
the correct definition of (L0 )t ot .
Alternatively, one can use the Faddeev-Popov method and introduce ghosts,
which will precisely cancel the two modes which we removed by hand. This is
known as modern covariant quantisation, another method which we will not describe
for reasons of time but which can be found in dedicated string theory textbooks.
The reader familiar with QED will immediately see that the above substitution of
the naive prefactor D with (D − 2) is analogous to the photon case: Of the formally
four degrees of freedom associated with the vector Aμ , only two transverse modes
contribute to physical quantities like Casimir effect or vacuum free energy. This
happens for exactly the same reason as here and to see it explicitly in a covariant
QED calculation one also needs ghosts.
3.4 Quantisation 99
The simplest way to explicitly calculate the normal ordering constant a is through
ζ function regularisation:
∞
! ∞
1
−s
n = lim n = lim ζ(s) = ζ(−1) = − . (3.61)
s→−1 s→−1 12
n=1 reg. n=1
This is of course quite formal and not very satisfying. Since the result is important,
we want to spend some time to explain why the normal ordering constant does in
fact have a physical and a-priori finite definition. To see this, we give the infinite
strip on which our 2d field theory lives a proper, physical width: π → πR. Then
we have
∞
1
Ht ot = α−n αn + πRλ . (3.62)
2R n=−∞
Crucially, we have here also introduced a cosmological constant counterterm.

We are now dealing with a standard QFT problem—the calculation of the total
energy of a 2d theory on a strip of width πR. The sum over zero modes has a UV
divergence, to be regularised by introducing a cutoff scale . A cutoff dependence
must also be assigned to the counterterm, λ → λ(). Its form is determined by the
requirement that the divergence for → ∞ cancels. Moreover, no finite ambiguity
will arise since we know from Weyl invariance that the renormalised cosmological
constant must vanish. Thus, we have
⎡ ! ⎤
∞
1 1 2 D−2
Ht ot = α + α−n αn + lim ⎣ n + πRλ()⎦ .
R 2 0 →∞ 2R
n>0 n−1
(3.63)
A very intuitive way of regularising this is to think in terms of physical modes with
momenta kn = n/R and to multiply the contribution of each mode by exp(−kn /).
It is then a straightforward exercise to do the summation, find the appropriate
counterterm λ(), and to obtain the finite result (cf. Problem 3.8.5)

1 1 2 D−2
Ht ot = α + α−n αn − . (3.64)
R 2 0 24R
n>0
The physical interpretation of this finite correction is clear: This is a one-loop

Casimir energy, associated with the finite size of the space on which the QFT
lives.1
1 Demonstrating that this result is in fact independent of the precise form of the regularising
function, exp(−k/) → f (k/), is not entirely trivial. See e.g. [13] for a discussion of the
corresponding 4d problem.
Returning to our stringy convention with R = 1 and to the notation L0 instead

of H , we have
1 2 D−2
(L0 )t ot = L0 − a with L0 = α + α−n αn and a= .
2 0 24
n>0
(3.65)
The physical state condition hence reads
(Lm − aδm )| phys = 0 for m≥0, (3.66)
where it is crucial to remember that L0 is, by definition, normal ordered.
3.5 Explicit Construction of Physical States: Open String
We start with the open-string worldsheet vacuum,
|0, p , defined by p̂μ |0, p = pμ |0, p . (3.67)
Our physical state conditions with m > 0 are automatically satisfied for any p,
1
Lm |0, p = αm−n αn |0, p = 0 , (3.68)
2 n
since in each term of this sum either n > 0 or m − n > 0. Thus, there is always an
annihilator involved, giving zero if applied to the vacuum.
By contrast, the m = 0 condition is non-trivial, giving

α p2 + α−n αn − a |0, p = 0 , (3.69)
n>0
√
where we used that α0 = lp = 2α p. With M 2 = −p2 , this translates into
M 2 = −a/α . (3.70)
Thus, p cannot be an arbitrary vector. Rather, it must satisfy the above mass-shell
condition.
Moving to the first excited level, we have to consider states
μ
ζμ α−1 |0, p , (3.71)
3.5 Explicit Construction of Physical States: Open String 101
with a polarisation vector ζ . The mass shell condition now reads

0 = (L0 − a)ζμ α−1 |0, p = α p2 + α−1 · α1 − a ζμ α−1 |0, p
μ μ

= α p2 + 1 − a ζμ α−1 |0, p ,
μ
(3.72)
implying
M 2 = (1 − a)/α . (3.73)
Of the Lm conditions with m > 0, now the first also becomes non-trivial:

1
0 = L1 ζ · α−1 |0, p = α1−n · αn ζ · α−1 |0, p . (3.74)
2 n
Of the various terms in the sum, only those can contribute where n ≤ 1 and 1 − n ≤
1. This occurs only for n = 0, 1, such that we find
1
0= (α1 · α0 + α0 · α1 ) ζ · α−1 |0, p = ζ · α0 |0, p = ζ · p|0, p . (3.75)
2
The implication is that the polarisation has to be transverse. We also need the norm
of the state, which is
μ
0, p|(ζμ α−1 )† (ζν α−1
ν
)|0, p = 0, p|0, pζμ ζ μ = ζ 2 . (3.76)
μ
Here we chose ζμ real, using the freedom to redefine the α−1 if necessary.
At the so-called second excited level, one has to analyse states of the form
μ μ
(
μν α−1 α1ν +
μ α−2 )|0, p , (3.77)
but we will not do so.

Instead, we summarise the results by focussing on the first excited level:
(A) For a > 1 we have M 2 < 0. This means that p is spacelike and hence timelike
vectors ζ with ζ · p = 0 exist. Thus, there are allowed states with ζ 2 < 0 and
hence negative norm. This is excluded.
The deeper reason for the problem is that a Weyl anomaly arises, which can only
be cured by considering a background which is not simply flat D-dimensional
Minkowski space. This is known as supercritical string theory and may (in
the supersymmetric case) nevertheless be relevant phenomenologically, although
this is not established (see e.g. [122–125] and the refs. therein).
(B) For a = 1, we have M 2 = −p2 = 0. This implies that the (D − 1)

independent ζ ’s which satisfy ζ · p = 0 fall into two classes:
First, there is one longitudinal polarisation, ζ ||p, corresponding to a zero-norm
state.
Second, there are (D − 2) transverse polarisations, which are spacelike and give
rise to positive-norm states. One may think, e.g. of p = (1, 1, 0, · · · , 0) and ζi =
(0, 0, · · · , 0, 1, 0, · · · , 0), with unity at position 1+i and i ∈ {1, 2, · · · , (D−2)}.
This is consistent with Gupta-Bleuler quantisation in QED. It gives the correct
description of a gauge theory with a massless vector. This case is known as
critical string theory. In the following, we will completely focus on this case
and the corresponding critical dimension D = 26.
(C) For a < 1 we have M 2 > 0 for the first excited (and all higher) levels. Thus
except possibly for the vacuum state, this case is in practice not very interesting.
It is not inconsistent at the present level of analysis (giving rise to a massive
vector with (D − 1) positive-norm states). Problems, possibly solvable, arise in
the interacting theory. This is known as the subcritical string. The Weyl anomaly
is also present, as in the supercritical case. Together, cases (A) and (C) are known
as non-critical string theory.
We close by mentioning that the overall picture in the critical case is just like
in gauge theory quantisation: We have restricted our Fock space by imposing the
physical state condition. The resulting space has no negative norm states, but the
so-called null states are still present. The actual positive-definite Hilbert space is
constructed as a quotient
H0 ≡ Hphys /Hnull . (3.78)
The mass-shell condition, originating from (L0 − 1)| phys = 0, can be written as

M 2 = −p2 = (N − 1)/α with N≡ α−n αn . (3.79)
n>0
The operator N or its expectation value is called the level. We have found a tachyon
at level 0, a massless vector at level 1, and we could have found massive string
excitations at level 2 and higher. The tachyon corresponds to the statement that
our assumed 26d Minkowski vacuum is unstable since a scalar with negative mass
squared is present. It will decay by tachyon condensation, which is an interesting
subject of research. But we will not discuss this since we use the bosonic string only
as a toy model to get ready for the superstring.
3.6 Explicit Construction of Physical States: Closed String 103
3.6 Explicit Construction of Physical States: Closed String
A repetition of the analysis of the previous section will again single out the case
a = 1 or D = 26. We focus right away on this case, recalling, however, that the
number of operators and constraints is now doubled. We rewrite
(L0 − a)| phys = 0 , (L̃0 − a)| phys = 0 (3.80)
as
(L0 − L̃0 )| phys = 0 , (L̃0 + L̃0 − 2a)| phys = 0 . (3.81)
We recall that
α02 p2
L0 = +N = α +N, (3.82)
2 4

where we used that α0 = p l/2 = p α /2 in the closed string case. Analogous
equations hold for the left-movers. With this, the physical state conditions become
(N − Ñ )| phys and (p2 α /2 + N + Ñ − 2)| phys , (3.83)
known as level matching and mass shell conditions, respectively. The latter is also
frequently given as
M 2 = 2(N + Ñ − 2)/α . (3.84)
Now one proceeds systematically, level by level, as before. At the vacuum level
one again finds a tachyon,
|0, p , M 2 = −4/α . (3.85)
At the first excited level, due to the level matching condition, both α−1 and α̃−1
have to be used:
μ ν
ξμν α−1 α̃−1 |0, p , M2 = 0 . (3.86)
Note that, as before, one really has M 2 = 2(1 + 1 − 2a)/α , such that masslessness
follows only for a = 1, i.e. in the critical dimension. At the first excited level, the
L1 and L̃1 constraints are non-trivial. They read
ξμν pμ = 0 and ξμν pν = 0 . (3.87)

It is also easy to check that the norm of our states is
phys | phys ∼ ξμν ξ μν , (3.88)
which is always non-negative if the physical state conditions are satisfied.

To classify the states, it is helpful to think of the polarisation tensor literally as of
an element in the tensor product of two copies of RD ,
μ
ξ μν = v(a) ⊗ v(b)
ν
. (3.89)
ab
In analogy to the standard treatment of the photon, one chooses a basis v(a) with
one element v(0) ∼ p, one lightlike element v(1) with non-zero product with p, and
D − 2 spacelike elements orthogonal to v(0) and v(1) . Of these, only the spacelike
vector and v(0) are allowed to appear in (3.89). Hence we have (D − 1)2 physical
basis states. Furthermore, (D −2)2 of them (those built from spacelike vectors only)
have positive norm. The rest corresponds to gauge freedom.
Choosing p ∼ v(0) ∼ (1, 1, 0, · · · , 0) and v(1) ∼ (1, −1, 0, · · · , 0), we
see explicitly how products of the (D − 2)2 transverse vectors form a basis for
the transverse polarisations ξt . They correspond to the lower-right corner of the
matrix ξ :
⎛ ⎞
0 0 0 ··· 0
⎜ 0 0 ··· 0 ⎟
⎜ 0 ⎟
⎜ ⎟
ξ =⎜ 0 0 ⎟. (3.90)
⎜ ⎟
⎝ · · ξt ⎠
0 0
The transverse physical polarisations ξt transform under SO(D − 2), the group
of rotations in the spacelike hyperplane transverse to p. This is called ‘little
group’—the subgroup of SO(1, D − 1) leaving p invariant. Our rank-2 tensor
representation of physical polarisations is not irreducible but decomposes into
symmetric, antisymmetric, and trace part. These three representations correspond
to three different fields of the D-dimensional field theory which the string describes
from the target space perspective. They are:
1. The graviton Gμν , with (D − 1)(D − 2)/2 − 1 d.o.f.s (note that for D = 4 this
correctly reproduces the known result of 2 d.o.f.s).
2. The Kalb-Ramond field or antisymmetric tensor Bμν , with (D − 2)(D − 3)/2
d.o.f.s.
3. The dilaton φ, with 1 d.o.f.
3.7 The 26d Action 105
We could go on to discuss excited states, but all we will need to know is that
there they form a tower with increasing mass and that the number of states at each
consecutive level grows extremely fast. The mass spacing is M 2 = 4/α .
3.7 The 26d Action
We are only interested in the critical case, D = 26, and we focus on the closed
string (for more details see e.g. [7]). It is immediate to write down a quadratic-
level action for the above fields (to be supplemented by the tachyon which, as we
know, has negative mass squared and makes the 26d Minkowski-space solution
unstable). Assuming that one also knows how to compute scattering amplitudes,
one can supplement this action by interaction vertices and write down the full, non-
linear expression at the 2-derivative level. It reads (suppressing the tachyon):

1 √ 1
S= d 26 x −Ge−2φ R[G] − Hμνρ H μνρ + 4(∂φ)2 , (3.91)
κ2 12
where
H = dB , (3.92)
in complete analogy with F = dA in the 1-form case.

Many important comments have to be made. First, it is apparent that the value of
κ 2 can be changed by a shift of φ. Thus, we can, for example, define κ 2 = cα 12 ,
with some numerical constant c. Then the choice of the background value of φ
determines the
√ 26d Planck mass relative to the mass of the first excited string modes,
which is 2/ α . It also governs the perturbativity of the theory, i.e. the importance
of string loops, as we will discuss further down.
Second, the apparently wrong sign of the dilaton-kinetic term is misleading.
Indeed, the above is called the string-frame action (similar to what is known as
the Brans-Dicke frame action in the non-stringy gravitational literature). One can
go to the Einstein frame by the Weyl rescaling
Gμν = G̃μν e−φ/6 . (3.93)
The result is

1 1 1
S = 2 d 26 x −G̃ R[G̃] − e−φ/3 Hμνρ H μνρ − (∂φ)2 . (3.94)
κ 12 6
In this frame, the Planck mass is manifestly fixed and the mass of the excited states
changes with varying dilaton background.
Third, this is the first (but not the last) time we encounter a higher-form gauge
theory. So it may be useful to remind the reader of some of the relevant basic
notions. (A standard summary of conventions can be found in the Appendix of

Volume II of [7].) It is convenient not to think just of an antisymmetric field or
gauge potential Aμ1 ···μp but of a differential form,
1
Ap = Aμ ···μ dx μ1 ∧ · · · ∧ dx μp . (3.95)
p! 1 p
Our present case p = 2 with Bμ1 μ2 = Aμ1 μ2 is part of the more general structure
of such gauge theories.
One should think of the dx μ as basis vectors of the dual tangent space (the
cotangent space) of a manifold, such that

∂
dx μ = δνμ . (3.96)
∂x ν
Higher p-forms take their values in the p-fold exterior product (the antisymmetric
part of the tensor product) of the cotangent space. This is symbolised by the wedge,
e.g.
dx 1 ∧ dx 2 = dx 1 ⊗ dx 2 − dx 2 ⊗ dx 1 . (3.97)
It generalises to
dx 1 ∧ · · · ∧ dx p = p! dx [1 ⊗ · · · ⊗ dx p] , (3.98)
where [· · · ] stands for antisymmetrisation. The implication is that, for example,
Ap (∂1 , · · · , ∂p ) = A[1···p] = A1···p . (3.99)
Consistently with the above, one formally defines the product of two forms
(p + q)!
(Ap ∧ Bq )μ1 ···μp+q = A[μ1 ···μp Bμp+1 ···μp+q ] . (3.100)
p!q!
Crucially, the natural map from functions (0-forms) to 1-forms,
d : f → df = ∂μ f dx μ with df (∂μ ) ≡ ∂μ f , (3.101)
has a generalisation to higher forms:
(dAp )μ1 ···μp+1 = (p + 1)∂[μ1 Aμ2 ···μp+1 ] . (3.102)
This so-called exterior derivative is a central mathematical concept. It is immediate

to convince oneself that d ◦ d = d 2 = 0.
3.7 The 26d Action 107
By its very definition, a p-form provides, at every point, a totally antisymmetric

map from the p-fold tensor product of the tangent space to the real numbers. Thus, it
can be used to define the volume of an infinitesimal parallelepiped (with orientation,
i.e. ordering of the vectors by which it is spanned) at any point of the manifold.
This gives rise to the possibility of integrating a p-form over a finite p-dimensional
submanifold:

V (Cp ) = Ap . (3.103)
Cp
After these preliminaries, it is clear how to interpret Ap as a physical gauge

potential. First, one has the gauge transformation and the gauge-invariant field
strength
Ap → Ap + dχp−1 and Fp+1 = dAp . (3.104)
The natural lagrangian is ∼ |Fp+1 |2 ≡ Fμ1 ···μp+1 F μ1 ···μp+1 and the natural
coupling to charged objects is

Smat t er ∼ Ap . (3.105)
p
This is completely analogous to electrodynamics, where the matter coupling is the

integral of A1 along the worldline of the electron. Here, it is the integral of Ap along
the p-dimensional worldvolume p of a (p − 1)-brane. (Recall the convention that
the variable p in the term Dp-brane counts only the spatial dimensions.)
In the case at hand, a charged object suitable as a source for B2 is already present
in the theory we have so far developed: It is the fundamental string itself. Thus, the
term

B2 (3.106)
2
has to be added both to our 10d action and to our worldsheet action for the string. If
B2 is non-zero, this changes our 2d theory and its quantisation.
Similarly, we see that (3.91), with Gμν = ημν , φ = 0, and B = 0, describes
the solution in the background of which our fundamental string, introduced earlier,
propagates. This is the 2d theory one is easily able to quantise. But clearly other
solutions for this 10d action exist and the string can be quantised in their background
as well. The 2d theory is then much more complicated, e.g. through

√ √
d 2 σ −hhab (∂a Xμ )(∂b Xν )ημν → d 2 σ −hhab (∂a Xμ )(∂b Xν )Gμν (X) .
(3.107)
We see that this theory now ceases to be free or quadratic in the fields. For example,
if near X = 0 we can write
Gμν = ημν + c · (X1 )2 ημν + · · · , (3.108)
we encounter a quartic interaction vertex in the worldsheet theory. Similarly, a non-

zero B2 = B2 (X) adds new terms to the worldsheet lagrangian. In particular, the X
dependence of B2 leads to new interaction terms in the theory of scalars Xμ living
on the worldsheet.
Before closing this section, we should discuss the role of the dilaton φ from the
worldsheet perspective. This field is related to the Einstein–Hilbert term,

√
d 2 σ −h R , (3.109)
of the worldsheet action. At first sight, this term is clearly allowed. It respects
all symmetries of the worldsheet, including in particular Weyl invariance. A more
careful analysis reveals, however, that it can be written as a total derivative and hence
does not affect the equations of motion. Indeed, following the standard derivation of
Einstein’s equations from the Einstein–Hilbert action, one finds

√ 1
δh d 2 σ −h R = d 2σ Rab − hab R δhab + boundary terms .
2
(3.110)
But in d = 2 one has

1
Rab − hab R = 0 (3.111)
2
as an identity. This follows from the symmetries of the Riemann tensor which,
as already noted above, can be expressed in terms of the Ricci scalar. Thus, the
worldsheet Einstein–Hilbert term does not change under continuous deformations of
the worldsheet metric. Its integral can, however, be non-zero, measuring topological
features of the worldsheet (see below).
Now, comparing the dynamics described by the target space action given above
and the role of the Einstein–Hilbert term on the worldsheet (we will see more details
of this further down), one can establish that φ has to be identified with the coefficient
of the worldsheet Einstein–Hilbert term:

1 √
SP ⊃ d 2 σ −h R φ . (3.112)
4π
As before, if φ = φ(X) is non-constant, new interactions are introduced into the

worldsheet theory.
3.8 Problems 109
As a final remark, we note that backgrounds solving the 26d equations of

motion are precisely those in which the Weyl invariance on the propagating
string worldsheet remains unbroken. In this sense, the 2d theory can be used to
directly determine the 26d dynamics, without calculating scattering amplitudes and
comparing them to 26d EFT vertices.
3.8 Problems
3.8.1 Point Particle Action
Task Guess the ‘Polyakov action’ for the point particle and derive the ‘Nambu-
Goto action’ given in the lecture. Determine the parameters to achieve consistency
which what you know from your course in special relativity.
Hints Introduce a worldline metric hτ τ ≡ h, such that ds 2 = hτ τ dτ 2 . Allow

a worldline cosmological constant term (which is forbidden in the string case by
Weyl invariance, but permitted for the point particle).
Solution The natural guess is

√
SP = c dτ h(h−1 Ẋ2 − λ) , (3.113)
from which one derives the equations of motion for h:
1 √ 1
0 = √ (h−1 Ẋ2 − λ) + h(−h−2 Ẋ2 ) = − (h−3/2 Ẋ2 + h−1/2 λ) . (3.114)
2 h 2
It follows that
h = −Ẋ2 /λ (3.115)
and hence
√ √ √
SP = c dτ h(−h−1 hλ − λ) = −2cλ dτ h = −2c λ dτ −Ẋ2 .
(3.116)
This reproduces the standard (‘Nambu-Goto-type’) relativistic point particle action

for, e.g. c = m/2 and λ = 1. Thus, the final result is

m
SP = dτ hτ τ h−1
τ τ Ẋ Ẋμ − 1 .
μ
(3.117)
2
3.8.2 Commutation Relations of Oscillator Modes
μ μ
Task Demonstrate the consistency of the commutation relations of pμ , x μ , αn , α̃n
with those of the Xμ s and μ s at equal time.
Hint It is efficient to first calculate the commutator of Xμ with μ , of Xμ with

itself, etc. using the mode expansion and then apply a Fourier transformation to
both sides.
Solution Collecting formulae from the lecture notes we have
il 1
μ −2inσ + −

Xμ (τ, σ ) = x μ + l 2 pμ τ + α̃n e + αnμ e−2inσ , (3.118)
2 n
n=0
⎧ ⎫
1 ⎨
+ −
⎬
ν (τ, σ ) = l 2 pν + l α̃nν e−2inσ + αnν e−2inσ . (3.119)
2πα ⎩ ⎭
n=0
When writing the commutator, we may right away focus on those pairings of terms
from the mode expansion which have a chance of being non-zero:
l2
[ν (τ, σ ), Xμ (τ, σ )] = − iημν
2πα
il 2 μν
−2in(σ + −σ + ) −2in(σ − −σ − )

− η e + e
4πα
n=0
i i μν
−2in(σ −σ )

= − ημν − η e + e2in(σ −σ ) .
π 2π
n=0
i i )
= − ημν − ημν e−2in(σ −σ
π π
n=0
∞

i
= − ημν e−2in(σ −σ ) . (3.120)
π n=−∞
μ
To get the sign right, it is crucial to note that [αm , αnν ]/n = mδm+n ημν /n =
−δm+n ημν . One may finish here by recognising the δ function in σ − σ on the
r.h. side.
But let us be fully explicit by finally applying a Fourier transformation in σ and
σ to both sides of our result. Using also the canonical commutation relations, the
3.8 Problems 111
l.h. side gives
π π π
2imσ
,
-
dσ e dσ e 2ikσ
− iη δ(σ − σ ) = −iη
μν μν
dσ e2i(m+k)σ
0 0 0
= −iπημν δm+k . (3.121)
Analogously, on the r.h. side one finds
i μν
− η (πδn−k )(πδm+n ) = −iπημν δm+k . (3.122)
π n
Thus, both sides agree.
The commutators [Xμ (τ, σ ), Xν (τ, σ )] (and similarly for μ ) vanish since the
relevant sums contain explicit factors of the mode indices n. For example, dropping
all prefactors and the manifestly vanishing x μ /pμ contribution, one encounters
expressions like
1
−2in(σ −σ ) 2in(σ −σ )
n e + e . (3.123)
n2
n=0
But this is zero by antisymmetry in n.
3.8.3 Trace of the Energy-Momentum Tensor
Task Use a symmetry argument to show that the trace of the energy-momentum
tensor of the string vanishes identically (no hint needed).
Solution By Weyl invariance,

√
δSP −h ab
0 = SP [hab +
hab ] − SP [hab ]
hab =
hab − T
δhab 4π
√
−h
=
− T aa . (3.124)
4π
Hence, T a a = 0.
3.8.4 Virasoro Algebra
Task Derive the classical part of the Virasoro algebra using the mode expansion
of the generators and the canonical commutation relations (or equivalently Poisson
brackets) of the oscillator modes. Then also derive the anomaly under the assump-
tion that the operator-ordering ambiguity in L0 is resolved by normal ordering, i.e.
that 0, 0|L0 |0, 0.
Hints For the first part, use the derivation or Leibniz rule for commutators:
[A, BC] = [A, B]C + B[A, C]. For the second part, argue that only expressions
with L0 on the r.h. side are affected by the ordering ambiguity. Thus, the anomaly
must take the form A(m)δm+n . Then evaluate the commutator [L1 , [Lm , L−m−1 ]]
directly and with the Jacobi identity (in derivation form). Derive from this a recur-
sive formula for the A(m). Show that the expression A(m) = am3 + bm satisfies
this relation and hence determines A(m) unambiguously up to the coefficients a
and b. Fix a, b by evaluating [Lm , Ln ] with (m, n) being (1, −1) and (2, −2) in the
zero-momentum vacuum |0, 0.
A very similar derivation can be found in [101], but try to succeed on your own
before consulting that book or our text below.
Solution We focus on D = 1 and find from the known mode expansion
1
[Lm , Ln ] = [αm−k αk , αn−l αl ] . (3.125)
4
k,l
Applying the derivation or Leibniz rule for commutators once gives
1
[Lm , Ln ] = {[αm−k αk , αn−l ] αl + αn−l [αm−k αk , αl ]} . (3.126)
4
k,l
The second application together with the standard commutation relations gives
1
[Lm , Ln ] = {αm−k [αk , αn−l ]αl + [αm−k , αn−l ]αk αl + αn−l αm−k [αk , αl ]
4
k,l
+αn−l [αm−k , αl ]αk }

1
= {kαm−k αn+k + (m − k)αk αm+n−k + kαn+k αm−k
4
k
+(m − k)αm+n−k αk } . (3.127)
Now let us shift the summation index according to k → k − n in the first and third
term. If we were in addition allowed to change the order of the α’s in the second and
3.8 Problems 113
third terms, we would obtain
1
[Lm , Ln ] = {(k − n)αm+n−k αk + (m − k)αm+n−k αk }
2
k
m−n
= αm+n−k αk = (m − n)Lm+n . (3.128)
2
k
It is clear that this operation is illegal precisely in situations with an ordering

ambiguity on the r.h. side, i.e. for m + n = 0. Hence, we have shown that
[Lm , Ln ] = (m − n)Lm+n + A(m)δm+n (3.129)
with some so far unknown function A.
Now we evaluate the commutator given in the hints directly,
[L1 , [Lm , L−m−1 ]] = (2m + 1)[L1 , L−1 ] = (2m + 1)(2L0 + A(1)) , (3.130)
and through the derivation rule,
[L1 , [Lm , L−m−1 ]] = [Lm , [L1 , L−m−1 ] + [[L1 , Lm ], L−m−1 ]

= (2 + m)[Lm , L−m ] + (1 − m)[Lm+1 , L−m−1 ]
= (2 + m)(2mL0 + A(m))
+(1 − m)((2m + 2)L0 + A(m + 1)) . (3.131)
Comparing both results gives the recursion relation
(m − 1)A(m + 1) = (2 + m)A(m) − (2m + 1)A(1) . (3.132)
Given also that A(m) = −A(−m) by its definition, it is clear that A(1) and A(2)
are sufficient to determine all A(m) unambiguously. Moreover, it is easy to check
that A(m) = am3 + bm solves the recursion:
(m−1)(a(m+1)3 +b(m+1)) = (2+m)(am3 +bm)−(2m+1)(a +b) (3.133)
for all a, b. Thus, if we can fix a, b, we have found the unique solution.
This is easy to achieve: Note first that each term in L−1 (and even more so in L1 )
contains either an annihilator or a p. Hence
0, 0|[L1 , L−1 ]|0, 0 = 0 , (3.134)

implying A(1) = 0. By contrast, L−2 contains a single term without annihilators,

hence
1 1
0, 0|[L2 , L−2 ]|0, 0 = 0, 0|L2 L−2 |0, 0 = 0, 0|α1 α1 α−1 α−1 |0, 0 = .
4 2
(3.135)
This implies A(2) = 1/2. Thus, we have to solve
1
a+b =0 and 8a + 2b = , (3.136)
2
giving a = −b = 1/12 . Clearly, if we generalise from one to D bosons, nothing
changes except that, in the very last step, one gets a factor of D/2 on the r.h. side of
(3.135). Thus, the result given in the lecture follows.
3.8.5 Normal Ordering Constant as Casimir Energy
Task Finish the calculation of the normal ordering constant a of the open string as
the Casimir energy of 2d field theory on a strip,
∞
!
D−2
− a = lim n + πR λ() .
2
(3.137)
→∞ 2
n=1
For hints see lecture notes.
Solution As explained in the lecture, we think of the sum as of a sum over modes
with physical momenta kn = n/R, suggesting a regularisation by a suppression
factor exp(−kn /). The sum S then reads
∞
∞
d −αn
S() = n e−n/R = − e ( with α = 1/R) (3.138)
dα
n=1 n=1

d 1 e−α 1
=− = , -2 =
dα 1 − e−α 1−e −α (1 − e −α )(e α − 1)
1
= + O(α)
(α − α 2 /2 + α 3 /6)(α + α 2 /2 + α 3 /6)
1
= + O(α)
α 2 (1 − α/2 + α 2 /6)(1 + α/2 + α 2 /6)
3.8 Problems 115

1 1 1 2
= 2 + O(α) = 2 1 − α + O(α)
α (1 + α 2 /12) α 12
1
= 2 R 2 − + O(1/) .
12
This gives rise to

D−2 1
− a = lim 2 R 2 − + O(1/) + πR 2 λ() . (3.139)
→∞ 2 12
The cosmological constant counterterm is unambiguously determined to be λ() =

2 (D − 2)/(2π), such that finally
D−2
a= . (3.140)
24
3.8.6 Kalb-Ramond Field from the Worldsheet Perspective
Task Work out the formal expression

B2 , (3.141)
2
such that it becomes a standard Riemann double integral in dσ 1 dσ 2 with an

integrand depending on the functions Xμ (σ ) and B2 (X).
Hints Interpret and apply the equality (in two dimensions)

dx ∧ dy = dx dy (dx ∧ dy)(∂x , ∂y ) (3.142)
between a form integral and a Riemann double integral.
Solution In analogy to the formula for translating a form integral in a Riemann

integral given in the hint, we have (suppressing the index 2 which is always
associated with our integral)

B2 = dσ 1 dσ 2 B2 (∂1 , ∂2 ) . (3.143)
Since B2 is originally defined in target space rather than on the worldsheet, we need
to push-forward the vectors ∂a to the target space using the embedding map Xμ (σ )
before we can explicitly insert them in B2 :

∂ ∂
B2 = dσ 1 dσ 2 B2 ∂1 Xμ μ
, ∂2 Xμ . (3.144)
∂X ∂Xμ
With
1
B2 = Bμν dXμ ∧ dXν (3.145)
2!
one now finds

B2 = dσ 1 dσ 2 Bμν (X(σ )) (∂1 Xμ (σ )) (∂2 Xν (σ )) , (3.146)
where σ stands for {σ 1 , σ 2 }. The factor 1/2! disappeared since we dropped a second
term, where ∂1 and ∂2 would have been exchanged.
String Theory: Interactions and Superstring
4
Before we can see what the string-theoretic UV completion of gravity has to say
about the real world, a few more formal developments are necessary. First, we want
to understand at least in principle how scattering amplitudes and loop effects are
calculated. Second, we need to introduce fermions and get rid of the tachyon. The
main textbook sources continue to be [7, 8, 101–103].
4.1 State-Operator Correspondence
Before discussing scattering amplitudes and loops, a few more words about the
worldsheet theory after gauge fixing are necessary. We learned that this is a CFT
and we will here work with the euclidean version of this theory. The symmetries of
the CFT include angle-preserving deformations of the worldsheet. For example, we
can map our fundamental cylinder corresponding to the propagation of the string to
the z-plane,
z = r eiϕ ∈ C . (4.1)
Specifically, we want the map to be such that time runs radially and circles of
constant r correspond to constant-time cuts through our cylinder (cf. Fig. 4.1). The
reader is invited to consider the explicit map z = exp(iw) and identify a strip in the
w-plane (with periodic boundary conditions, i.e. a cylinder) that is mapped to the
z-plane in the desired way.
Next, let us recall that a state in a 4d QFT may, analogously to the Schrödinger
wave function of quantum mechanics, be described by a Schrödinger wave
functional,
: φ → [φ, t] ∈ C . (4.2)

118 4 String Theory: Interactions and Superstring
Fig. 4.1 String propagation

mapped to the z-plane. The
part of the cylinder between
initial time τi and final time
τf corresponds to the annulus
(ring) between ri and rf
Here φ : x → φ(t, x) ∈ R is a field configuration at fixed time. The evolution of

such states is described by the QFT version of the Feynman path integral.
In our context, a string state at time τi is then represented, in the radial
representation, by a wave functional
i [Xi , ri ] . (4.3)
μ
Here Xi stands for any of the possible field configurations Xi (ri , ϕ). The wave
functional obtained by Hamiltonian evolution at radial time rf reads
Xf
f [Xf , rf ] = DXi DX e−S[X] i [Xi , ri ] . (4.4)
Xi
Here the labels Xi and Xf of the integral mean that we integrate over field
configurations X(r, ϕ) satisfying X(ri , ϕ) = Xi (ϕ) and X(rf , ϕ) = Xf (ϕ).
Now, consider the limit in which our evolution starts at τi = −∞, corresponding
to ri = 0 or z = 0. In this limit, we can write (4.4) as
Xf
f [Xf , rf ] = DX e−S[X] lim i [X(ri , ϕ), ri ]
ri →0
Xf
= DX e−S[X] O(z = 0) . (4.5)
Here, in the first step, we have absorbed the integral over Xi in the integral over
X, dropping the initial boundary condition. In other words, we now integrate over
functions X which are also defined inside the inner circle of radius ri (possibly
with a singularity at the origin). Nevertheless, they are weighted by the functional
i according to their values on that circle. In the limit ri → 0, this becomes a
weighting according to the local behaviour of the functions X near z = 0. Hence, in
the second step, we have introduced the operator O in the CFT, i.e. some functional
of X depending only on its local behaviour at the origin. In the simplest and most
relevant cases, this is an expression involving X(0) and its derivatives, as familiar
from a conventional local QFT operator built from a field X(z) (cf. Problem 4.9.1).
By the above procedure, we have understood how a given state, in our case the
state defined by i , specifies an operator. The opposite direction is obvious: Clearly,
4.2 Scattering Amplitudes 119
Eq. (4.5) may be interpreted as describing the evolution of some state, defined by O,
from τ = −∞ to τf .
Thus, we now know how to associate a CFT state with an operator and vice versa.
4.2 Scattering Amplitudes
After the discussion of the previous section, it should be at least intuitively clear that
the integral over fields on a cylinder can be replaced by the integral over fields on the
sphere, with appropriate operators inserted at the points which are mapped to τ =
±∞. This is illustrated in Fig. 4.2 together with an analogous map corresponding to
the 2-to-2 scattering of string states.
This leads very naturally to the following fundamental formula for n-point
scattering amplitudes in string theory, which one may view as the definition of the
theory:
∞

Dh DX
An = e−S[X,h] d 2 z1 · · · d 2 zn V1 (z1 , z1 ) · · · Vn (zn , zn ) .
VolDiff.×W eyl
g=0
(4.6)
Here the sum is over all compact oriented 2d manifolds (Riemann surfaces), as
illustrated in Fig. 4.3. The terms are labelled by the genus g of the worldsheet.
The integration is not only over scalar field configurations X but also over metrics
h. This definition is more fundamental and the gauge-fixed integral just over X
(corresponding to the CFT language) must be carefully derived from it. The reason
is that there is a non-trivial interplay between the topology of the manifold, the
position of the vertex operators and the residual gauge freedom. In this process, one
also has to divide out the infinite factor coming from gauge redundancies. This factor
becomes manifest when one uses the Faddeev-Popov method to treat the functional
integration.
The action to be used in the above is

1 √ φ √
S[X, h] =
d 2
σ −h(∂X) 2
+ d 2 σ −hR , (4.7)
4πα 4π
Fig. 4.2 Identification of non-compact worldsheets describing string propagation (left) or scat-
tering (right) with appropriate compact worldsheets (in this case spheres) with operator insertions
Fig. 4.3 Contributions of worldsheets of genus zero, one, and two to the four-point scattering
amplitude
where we suppress boundary terms relevant in the open-string case. We have

also assumed that the target space is flat and the dilaton constant. Restricting our
attention to the oriented string, one has

1 √
d 2σ −hR = χ() = 2 − 2g , (4.8)
4π
where χ is known as the Euler number. Thus, the second term on the r.h. side of
(4.7) just supplies a factor
−2+2g
gs−χ = gs with gs ≡ eφ . (4.9)
The quantity gs is known as the string coupling.

Finally, the so-called vertex operators Vi have to be chosen appropriately to
reflect the physical states in the scattering of which one is interested. They can be
derived from our understanding of the physical states of the quantised string and the
state-operator mapping. Here, we only provide as an example the vertex operator for
the tachyon of momentum k. It is basically the simplest operator that has the desired
transformation properties under translations:
μ (z,z)
V (k, z, z) = gs : eikμ X : . (4.10)
Indeed, a target space translation Xμ → Xμ +

μ gives this operator (and hence the
μ
corresponding state) an extra factor eikμ
. So this is clearly a state of momentum
k . The normalisation by gs ensures that free propagation is proportional to gs0 , tree-
μ
level 4-point-scattering to gs2 , one-loop 4-point-scattering to gs4 , and so on. Apart

from the exponential and gs prefactors, our operator is just the unit operator. It is
shown in Problem 4.9.1 that this corresponds to the vacuum state and hence the
tachyon.
From here, it would be relatively straightforward to calculate some of the
simplest amplitudes and loop diagrams (a.k.a. higher genus contributions) and
convince oneself of the promised very soft UV behaviour and loop finiteness. But
we have to move on.
4.3 Worldsheet Supersymmetry 121
4.3 Worldsheet Supersymmetry
We need to find a string theory which describes target space fermions and which has
no tachyon. Both can be achieved by supersymmetrising the worldsheet. We will
follow the presentation of [101], where the reader may also find a list of the most
important original papers. Let us only mention the key contribution of Ramond,
Neveu, and Schwarz [126, 127]. Their approach, which also underlies the following
discussion, is known as the ‘RNS superstring’. A detailed presentation can also be
found in [8].
As in 4d, we simply add fermionic worldsheet coordinates,
σa ‘+’ θα . (4.11)
The 2d Lorentz transformations are completely analogous to the familiar 4d case,
σ a → a b σ b , θα → Sα β θβ , (4.12)
with
= exp(i
ab Jab ) and S = exp(i
ab {i[γa , γb ]/4}) . (4.13)
One may easily check that

0 0 −i 1 0 i
γ = , γ = (4.14)
i 0 i0
fulfil
{γ a , γ b } = −2ηab . (4.15)
Of course, because of antisymmetry in a, b, there is only one independent generator

of type i[γa , γb ]/4 (spinor representation) or Jab (vector representation). The
commutation relations are trivially the same: After all, SO(1, 1) is a one-parameter
group. Nevertheless, the two representations are different, cf. Problem 4.9.4.
Furthermore, since S is real, one may obviously demand that the spinor is real:
∗
∗ θ− θ−
θ = = =θ. (4.16)
θ+ θ+
This is a particularly simple version of the familiar Majorana condition
T
ψ = ψc ( with ψ c ≡ C ψ ). (4.17)
The reader may want to consult the appendix of Volume 2 of [7] for a systematic
discussion of spinors in various dimensions.
Following very closely the familiar 4d procedure, one may promote the scalars
to (general) superfields,
Xμ → Y μ (σ, θ ) (4.18)
with
1
Y μ (σ, θ ) = Xμ (σ ) + θ ψ μ (σ ) + θ θ B μ (σ ) . (4.19)
2
Here θ = θ † γ 0 , as in 4d. We define SUSY generators
∂
Qα = α + i(γ a θ )α ∂a , (4.20)
∂θ
which are also Majorana spinors, and observe that they satisfy the SUSY algebra
relation
β
{Qα , Q } = −2i(γ a )α β ∂a . (4.21)
The SUSY transformation can be defined by
δξ Y (σ, θ ) = (ξ Q) Y (σ, θ ) , (4.22)
leading to
δξ X μ = ξ ψ μ (4.23)
δξ ψ = −i(γ ξ )∂a X + B ξ
μ a μ μ
(4.24)
δξ B μ = −iξ γ a ∂a ψ μ . (4.25)
To write down SUSY-invariant actions, it is sufficient to integrate any expression

in the Y μ over the full superspace. In our case this is simply a d 2 σ d 2 θ integral.
However, we are looking for a specific action which would serve as a generalisation
of the (so far flat-space) bosonic Polyakov action. For this purpose, it is convenient
to introduce the supercovariant derivative
∂
Dα = α − i(γ a θ )α ∂a . (4.26)
∂θ
The SUSY version of our bosonic action (with l = 1) can then be given as

i α
S= d 2 σ d 2 θ (D Y μ )(Dα Yμ )
4π

1 μ
=− d 2 σ (∂a Xμ ∂ a Xμ − iψ ∂/ψμ − B μ Bμ ) . (4.27)
2π
4.4 Worldsheet Supergravity 123
The auxiliary field vanishes on-shell such that, in summary, we have simply added
a free fermion ψ μ for every scalar.
4.4 Worldsheet Supergravity
The next step is to introduce gravity (more precisely, to promote the metric to a field,
since gravity in the sense of a dynamical theory does not really exist in d = 2). This
implies making SUSY local, as explained earlier.
Since our theory contains spinors, we will need a vielbein, related to the metric
by
hab = (em )a (en )b ηmn . (4.28)
Here a, b, · · · are ‘curved’ or ‘Einstein indices’ as before and m, n, · · · are ‘frame’

or ‘Lorentz indices’. Furthermore, since the Lorentz symmetry transforming the
Lorentz indices is local, we require a spin connection, to define covariant derivatives
of objects with frame indices:
∇a v m = (∂a + ωa )v m with ωa ∈ Lie(SO(1, d − 1)) , (4.29)
in our case with d = 2. It is defined by demanding covariant constancy of the

vielbein,
0 = ∇a em b = ∂a em b + (ωa )m n en b − ab c em c , (4.30)
where stands for the usual Christoffel symbols. Clearly, the object on which ∇
acts can transform in any representation of SO(1, d − 1), in which case ωa has to
be taken in that representation.
With these preliminary remarks our action becomes, in the first step,

1 μ
S2 = − d 2 σ e hab (∂a Xμ )(∂b Xμ ) − iψ γ a ∇a ψμ . (4.31)
2π
The index 2 stands for ‘quadratic order’. We want to make this invariant under a
local version of the SUSY transformations above, i.e. with ξ → ξ(σ ). In addition,
we need to define SUSY transformations of our new field, the metric or, more
appropriately, the vielbein. Working at leading order in perturbations around flat
space, em a = δ m a , one postulates
δξ em a = −2iξ γ m χa . (4.32)
Here χa is the gravitino. Its appearance on the r.h. side is natural since, as we argued
earlier, it has to come to provide the superpartner for the metric. The rest of this
relation is fixed (up to normalisation) by covariance.
The action of (4.31) is not invariant under local SUSY but, since it was invariant
under the global version, its non-invariance is controlled by the derivative of ξ . Thus,
we have

2 √
δξ S2 = d 2 σ −h (∇a ξ ) J a , (4.33)
π
where J a is by definition the Noether current corresponding to the global version of

the symmetry. Explicitly, one finds (at quadratic order in the fields)
1 b a μ
Ja = γ γ ψ ∂b Xμ , (4.34)
2
known as the supercurrent. The non-invariance of S2 can be compensated by
adding a term

2 √ 1 √
S3 = − d σ −h χ a J a = −
2
d 2σ −h χ a γ b γ a ψ μ ∂b Xμ , (4.35)
π π
and introducing the transformation law
δξ χa = ∇a ξ . (4.36)
For obvious reasons, this method of constructing supergravity actions is known as

the Noether method.
Its implementation is not yet complete: Only after a modification of δξ ψ by a
term proportional to the gravitino and the addition of a quartic term,

1 √
S4 = − d 2σ −h (ψψ) (χ a γ b γ a χb ) , (4.37)
4π
does the theory become invariant under local SUSY. We recall that the Einstein–
Hilbert term is a total derivative. This matches the fact that the gravitino kinetic
term is identically zero in d = 2 (since γ [a γ b γ c] = 0).
Finally, the theory is still Weyl invariant, with transformation laws
1 1
δω X = 0 , δω em a = ωem a , δω ψ = ωψ , δ ω χa = ωχa .
2 2
(4.38)
Due to SUSY, this symmetry now has a fermionic counterpart, parameterised by the
infinitesimal Majorana spinor η:
δη X = δη e = δη ψ = 0 , δη χa = iγa η . (4.39)
This makes our theory super-Weyl-invariant and, after gauge fixing, supercon-
formal.
4.5 Quantisation of the Superstring 125
Before closing this section, one important remark is in order: As already noted,
the action described above and its quantisation to be discussed momentarily (the
‘RNS approach’ to the superstring) are built on worldsheet supersymmetry. In
this approach, the appearance of supersymmetry in the target space theory (to
be discussed later) remains somewhat miraculous. However, there exists another
approach with a different superstring action, the so-called Green-Schwarz (GS)
superstring [128–130], which is built from the very beginning on the requirement
of target space SUSY. While it looks very different, it is equivalent to the RNS
approach. For a textbook treatment see e.g. [102].
4.5 Quantisation of the Superstring
The large gauge symmetry makes it possible to go to flat gauge: As before,

diffeomorphism and Weyl invariance are sufficient to choose a flat metric and
vielbein. The key new point is that the local transformations encoded in ξ and η
allow one to set the gravitino to zero. Thus, we have to quantise the simple action

1 μ
S=− d 2 σ (∂a Xμ )(∂ a Xμ ) − iψ γ a ∂a ψμ . (4.40)
2π
As before, the equations of motion of the fields that have been eliminated by gauge
fixing must be imposed as constraints. These are
Tab = 0 , (4.41)
as before, where we now have

i 1 i
Tab = (∂a X) · (∂b X) + ψ · γ{a ∂b} ψ − hab (∂X)2 + ψ · ∂/ψ . (4.42)
2 2 2
Here the curly brackets stand for symmetrisation. In addition, we have used local
SUSY and super-Weyl invariance (see [8] for details) to set the gravitino χ a to zero.
But, as we have just seen in the last section, its equations of motion correspond, at
leading order in χ, to the vanishing of the supercurrent:
(Ja )α = 0 . (4.43)
This is the second, new constraint.

The mode decomposition for the bosonic part is as before. To discuss the mode
decomposition of the fermionic part, write

i
S = SB + SF with SF = d 2 σ (ψ− · ∂+ ψ− + ψ+ · ∂− ψ+ ) ,
π
where γ± = γ0 ± γ1 . (4.44)
This explains our indexing convention in

ψ−
ψ= (4.45)
ψ+
since we now see that ψ+ and ψ− are left and right movers, respectively. Due to the
fermionic nature of ψ± , a sign is not detectable (observables are always built from
bilinears). Hence, the sign may not or may change when going once around the
string. As a result, two different types of boundary conditions (known as Ramond
and Neveu-Schwarz) are possible. This leads to four sectors:
ψ+ (σ + π) = +ψ+ (σ ) ; ψ− (σ + π) = +ψ− (σ ) R-R

ψ+ (σ + π) = +ψ+ (σ ) ; ψ− (σ + π) = −ψ− (σ ) R-NS
ψ+ (σ + π) = −ψ+ (σ ) ; ψ− (σ + π) = +ψ− (σ ) NS-R
ψ+ (σ + π) = −ψ+ (σ ) ; ψ− (σ + π) = −ψ− (σ ) NS-NS . (4.46)
Note that we could not have used an arbitrary phase exp(iα) instead of the sign in
the boundary conditions since our spinors are real. The mode decomposition in the
R-NS sector reads

ψ̃rμ e−2ir(τ +σ ) , ψ− = ψrμ e−2ir(τ −σ ) ,
μ μ
ψ+ = (4.47)
r∈Z r∈Z+ 12
and analogously for the other three sectors. The reality constraint translates to the
usual relation between modes with opposite frequency: (ψ̃r )∗ = ψ̃−r and (ψr )∗ =
μ μ μ
μ
ψ−r .
On the open string, one only has R-R and NS-NS sectors. To understand this,
one may think of the open string as coming from the closed string (to be viewed as
a theory on S 1 ) by ‘modding out’ a Z2 symmetry. In other words, one goes from
S 1 to S 1 /Z2 . The Z2 acts by σ → −σ on the space, turning the spatial part of the
worldsheet from a circle into an interval. Two boundaries are created at the so-called
fixed points of the action, i.e. at σ = 0 and σ = π (if we start with a 2π circle).
This Z2 action also exchanges left and right movers in terms of fields. But such an
exchange would be inconsistent in a R-NS or NS-R sector.1
Due to the above, one actually calls the two distinct open-string sectors simply R
and NS.
1 Of course, this construction can be translated into the alternative picture where the open
superstring is defined on an interval from the start. Then two consistent sets of boundary conditions
at the two boundaries σ = 0 and σ = π have to be introduced. We leave that to the reader.
4.5 Quantisation of the Superstring 127
Skipping the standard steps of canonical quantisation, we immediately display

the commutation relations of the oscillator modes, promoted to operators:
[αm
μ
, αnν ] = m δm+n ημν (4.48)

r, s ∈ Z (R)
{ψrμ , ψsν } = δr+s ημν with .
r, s ∈ Z + 1
2 (NS)
The different normalisation (manifest in the prefactor m and the missing prefactor
r) is conventional. As before, the operators responsible for the constraints are
expanded in Fourier modes,
π
√ π
1 2
Lm = dσ e imσ
T++ , Gr = dσ eirσ J+ , (4.49)
π −π π −π
with
⎧ ⎫
1 ⎨ m ⎬
Lm = : α−n · αm+n + r+ ψ−r · ψm+r : (4.50)
2 ⎩ 2 ⎭
n∈Z r∈Z+ν

0 (R)
Gr = α−n · ψr+n where ν≡ . (4.51)
1/2 (NS)
n∈Z
These operators generate the super-Virasoro algebra. More precisely, there are
two different algebras, one for the Ramond case (r, s even) and one for the Neveu-
Schwarz case (r, s odd):
[Lm , Ln ] = (m−n)Lm+n +A(m) , {Gr , Gs } = 2Lr+s +B(r)δr+s , (4.52)
[Lm , Gr ] = (m/2 − r) Gm+r (4.53)
with the anomaly terms
A(m) = D(m3 − m)/8 and B(r) = D(4r 2 − 1)/8 . (4.54)
As before, only the annihilator-part of the classical constraints is imposed quantum

mechanically:
(Lm − aδm ) |phys = 0 (m ≥ 0) , Gr |phys = 0 (r ≥ 0) ,

(4.55)
where we note that there is no normal ordering ambiguity and hence no normal
ordering constant associated with G0 .
We do not repeat the derivation but simply quote the result for the normal
ordering constant:

1 1 1 1 D−2
a = (D − 2) − = 0 (R), a = (D − 2) + = (NS) .
24 24 24 48 16
(4.56)
We see that, in the Ramond case, the fermions precisely cancel the effect of the
bosons. In the Neuveu-Schwarz case, this supersymmetric cancellation is upset by
the non-trivial boundary conditions imposed on the fermions but not on the bosons.
Let us now turn concretely to the Fock space of the open-string NS sector: We
have
μ μ
Vacuum: |0, k , Creation operators: α−m ; ψ−r (m, r > 0) . (4.57)
The mass-shell condition reads
0 = (L0 − a) |0, k = (α p2 + N α + N ψ − a) |0, k , (4.58)
where

Nα = α−m αm , Nψ = r ψ−r ψr . (4.59)
m=1,2,··· r= 12 , 32 ,···
This implies that there is a scalar at level zero,
α M 2 = −a , (4.60)
and a (target space!) vector corresponding to the physical ψ−1/2 excitations at level
1/2:
1
α M 2 =
μ

μ ψ−1/2 |0, k with −a. (4.61)
2
In analogy to the logic of the bosonic case, we expect that D = 10 (with a = 1/2)
is the critical dimension, corresponding to the vector being massless (and the scalar
a tachyon, as in the bosonic string).
Next, we turn to the open-string R-sector, which superficially differs only very
little in that

Nψ = r ψ−r · ψr = r ψ−r · ψr . (4.62)
r=0,1,2,··· r=1,2,···
μ
But this number operator leads to the very peculiar situation that the ψ0 do not
appear in L0 and hence do not affect the energy (mass squared) of a state. They do,
4.6 GSO or Gliozzi-Scherk-Olive Projection 129
however, satisfy the non-trivial algebra (D-dimensional Clifford algebra)

μ
{ψ0 , ψ0ν } = ημν . (4.63)
Hence, every mass eigenspace must carry a representation of this algebra, i.e. it must
be a target space spinor:
Vacuum: |α, k with α = 1, 2, 3, · · · , 2D/2 = 32 . (4.64)
Since a = 0, this spinor is massless. To derive the critical dimension we would need
to either consider heavier, excited states or involve ghosts and the vanishing central
charge argument. We do not do this here and only assert that the critical dimension
is still D = 10.
4.6 GSO or Gliozzi-Scherk-Olive Projection
Before constructing the 10d superstring theories which may be relevant for the real
world, we need a further technical ingredient. The underlying idea is that one may
always use a projection operator (an operator P with P 2 = P ) commuting with the
Hamiltonian H to reduce the Hilbert space H in a consistent manner. A familiar
example is the projection on symmetric and antisymmetric subspaces of H ⊗ H
to define bosons and fermions in 2-particle quantum mechanics. Another example
(from this course) is the projection of functions on S 1 to functions on S 1 /Z2 , which
corresponds to the projection to even and odd functions and hence to the projection
from closed to open string (with Dirichlet or Neumann boundary conditions). The
new Hilbert space after projection is, by definition, Image (P ).
Here, we focus on the open superstring and consider
1
P = (1 + (−1)F ) , where F ≡ Fermion number . (4.65)
2
This amounts to keeping only states with even F (note that F is only defined mod
2). Concretely, one defines
(−1)F |0, k = −|0, k (NS)

(−1)F |α, k = |β, k β α (R) where ≡ 11 ≡ 0 1 · · · 9 . (4.66)
together with
(−1)F Xμ = Xμ (−1)F and (−1)F ψ μ = −ψ μ (−1)F . (4.67)
Here the minus sign in the first line of (4.66) is a choice which eliminates the
tachyon. In the second line, the non-trivial implementation of (−1)F through the
matrix ≡ 11 (the 10d version of γ 5 ) is enforced by consistency. Indeed, (−1)F
μ μ
by its very definition anticommutes with all ψr s. This includes ψ0 , which as we
know act on the vacuum like the μ s. Thus, (−1)F must be represented by a matrix
anticommuting with all μ s. But this, by definition, is ≡ 11 .
After this projection, the tachyon is gone and the 32-component Majorana spinor
has turned into a 16-component Majorana-Weyl spinor (in 10d both conditions may
be imposed together). On shell and in terms of the appropriate representations of the
little group SO(8), one has
massless vector (8v ) + Majorana-Weyl fermion (8) . (4.68)
Here the symbol 8v stands for the (defining) vector representation of SO(8), the
symbol 8 for the chiral Majorana spinor. We will later on also need the opposite-
chirality Majorana spinor, which corresponds to an inequivalent representation. It is
denoted by 8 .
The 8v + 8 found above fit a 10d supersymmetric gauge theory. But we will not
develop this construction since it anyway must be coupled to a closed string sector.
Our purpose was only to explain the idea of this particular projection on even (or
similarly on odd) fermion number states.
4.7 Consistent Type II Superstring Theories
We now turn to the closed string case. The name ‘type II’ refers to the presence of
two supersymmetries (equivalently two gravitinos) in 10d, as will become clear in a
moment. We first recall the relevant mass-shell and level matching conditions
(L0 + L̃0 ) |phys = 0 and (L0 − L̃0 ) |phys = 0 (4.69)
with

α 2 α 2 0 (R)
L0 = p +N −ν , L̃0 = p + Ñ − ν̃ and ν/ν̃ = .
4 4 1/2 (NS)
(4.70)
Note that the spacing between the different mass levels differs by a factor of 4
compared to the open string. The lowest levels in the four possible sectors are
Sector SO(8) rep. mass

NS − 1 tachyon
NS + 8v massless
R − 8 massless
R+ 8 massless ,
4.7 Consistent Type II Superstring Theories 131
where ± refers to the eigenvalue of (−1)F on which one can potentially project and
8/8 refer to the two inequivalent spinor representations of SO(8). (Of course the
‘1’ appearing in the row of the tachyon is only intended to say that this is a scalar
with a single degree of freedom—it is strictly speaking not appropriate to classify it
using the little group of massless particles in 10d.) As a side remark, the existence of
these in total three 8-dimensional, inequivalent representations of SO(8) is related
to the Z3 symmetry of its Dynkin diagram.
When combining left and right-moving sectors, the level matching constraint
allows the (NS −) sector to be paired only with itself. The other three sectors can be
paired in any combination. This gives the unprojected spectrum
Sector SO(8) rep.

(NS −, NS −) 1
(NS +, NS +) 8v × 8v
(NS +, R −) 8v × 8
(NS +, R +) 8v × 8
(R −, NS +) 8 × 8v
··· ··· .
There are in total 10 sectors in this table and (independently of the specific
fermion-number-projector), one might imagine building a consistent theory from
any combination of them. Clearly, there are 210 possibilities to select some subset
of these sectors. But this selection cannot be random: We want it
1. Not to contain a tachyon.

2. To be modular invariant (i.e. invariant under large diffeomorphisms of a world-
sheet torus, for example, under exchange of τ and σ and hence under reinterpre-
tation of the direction of time flow, cf. Fig. 4.4)
3. To obey certain mutual consistency rules among the selected vertex operators
on the worldsheet. (There should be no leftover phase or branch cut when one
operator circles another, cf. Fig. 4.5. The operator product expansion should close
or, in other words, it should not be possible to produce a state in scattering which
we have excluded from our selection.)
Fig. 4.4 Illustration of the

intuitive meaning of modular
invariance
Fig. 4.5 In principle, branch

cuts can arise in the
correlation function between
two vertex operators. This
should, however, be forbidden
since it makes the integration
over all positions impossible
With this, it can be shown that only two inequivalent possibilities of the 210 are
left [7]. The corresponding selections are easily formulated using fermion number
constraints or projections:
Type IIA left: (−1)F = 1 right: (−1)F̃ = 1 (NS) / (−1)F̃ = −1 (R)
Type IIB left: (−1)F = 1 right: (−1)F̃ = 1 .

These general rules translate, specifically in type IIA in:
Sector SO(8) rep.
(NS+, NS+) 8v × 8v = 1 + 28 + 35 = [0]φ + [2]B2 + (2)G

(NS+, R−) 8v × 8 = 8 + 56 = spinor + vector-spinor
(R+, NS+) 8 × 8v = 8 + 56 = spinor + vector-spinor
(R+, R−) 8 × 8 = 8v + 56t = [1]C1 + [3]C3
To derive the last two columns of this table, one needs elementary representation
theory (see e.g. [63]). We will only motivate and interpret the results. We note that
SO(8) has three inequivalent 56-dimensional representations: two vector-spinors
and one antisymmetric rank-2 tensor. We used a square and round bracket for
antisymmetric and traceless symmetric tensors of a given rank. Hence e.g. [2] stands
for the familiar Kalb-Ramond field and (2) for the graviton. On the bosonic side, we
have dilaton, B2 , gμν and two form-fields, C1 and C3 . The latter are a new feature
of the superstring and the corresponding charged states are so-called D0 and D2
branes, which are non-perturbative objects (in the sense that they do not directly
follow from the perturbative analysis of worldsheet degrees of freedom). They have
to be introduced into the theory for consistency, have their own action and dynamics,
and provide potential endpoints for open strings.
We are finding a so-called N = 2 supersymmetric theory since we have

two gravitinos which are both partners of the same, unique graviton. The other
two spinors are known as dilatini. There are two independent SUSY generators
and hence SUSY transformations relating the graviton to either one or the other
4.8 Other 10d Theories 133
gravitino. However, the overall structure is more involved and all degrees of freedom
are needed to fully match fermions and bosons.
Analogously, one finds the field content of the type IIB string:
Sector SO(8) rep.
(NS+, NS+) 8v × 8v = 1 + 28 + 35 = [0]φ + [2]B2 + (2)G

(NS+, R+) 8v × 8 = 8 + 56 = spinor + vector-spinor
(R+, NS+) 8 × 8v = 8 + 56 = spinor + vector-spinor
(R+, R+) 8×8 = 1 + 28 + 35+ = [0]C0 + [2]C2 + [4]+ , C4
The key differences are that this theory is chiral (a preference is given to one of
the two different available chiralities of spinors and vector-spinors). Furthermore,
the form-field and hence the brane content is different. It is easy to remember that
type IIA and IIB theory contain odd and even p-form gauge potentials, respectively,
and hence even and odd Dp-branes. A further noteworthy specialty of the IIB theory
is the fact that the C4 theory is subject to a self-duality constraint, F5 = ∗F5 , which
halves the number of degrees of freedom (cf. the index ‘+’ of [4]+ and 35+ ).
4.8 Other 10d Theories
The name type II refers to the two supersymmetries. There is also a minimally
supersymmetric 10d superstring theory called type I with unoriented strings. It
follows by modding out worldsheet parity. By this one means introducing of an
operator which realises the classical transformation σ → −σ at the quantum
level (hence 2 = 1) and projecting on its 1-eigenspace by
1
P = (1 + ) . (4.71)
2
A detailed analysis reveals that stability (‘tadpole-cancellation’) always requires
the presence of 32 D9-branes, giving rise to gauge fields living in 10d. Due to the
projection the group is not U (32) but its ‘real subgroup’, SO(32). We will return to
this when discussing such ‘orientifold projections’ more generally in Sect. 5.7.
Furthermore, it is consistent (and allows for tachyon removal) to supersym-
metrise only the left- or right-moving half of the worldsheet theory. For obvious
reasons such theories are called heterotic and they come in two types, named after
their non-abelian gauge group (which are present in both cases): heterotic E8 × E8
and heterotic SO(32). The corresponding 10d supergravity theories are only N = 1
supersymmetric.
Not surprisingly, the SO(32) heterotic theory is related to type I by a so-
called duality. In this particular case, it is a strong-weak duality, which means
that type I at weak string coupling is identical to heterotic at strong coupling and
vice versa. In fact, all of the five 10d theories above are related to each other by
Fig. 4.6 Illustration of M-theory and its perturbative corners: the five superstring theories and 11d
SUGRA
dualities, projections or compactifications (see Fig. 4.6) and are sometimes referred
to collectively as the (perturbative corners of) M-theory.
One usually includes 11d supergravity in this set, although the fundamental
objects there appear to be membranes (specifically M2-branes) rather than strings
and the theory is much less well-understood in the ultraviolet. Occasionally, the
name M-theory is also used to refer only to 11d-supergravity rather than to the whole
set of theories. It is believed that these six theories are the calculable, perturbative
corners of a more general and not yet fully understood structure: M-theory as
‘defined’ by the inner region of the ‘amoeba’ in Fig. 4.6.
Two of the edges connecting neighbouring corners of the amoeba in Fig. 4.6 have
already been briefly mentioned in the main text. Two further edges are defined
by different possibilities for compactifying from 11d to 10d, in one case with the
introduction of two E8 gauge theories localised at the boundaries of an interval. We
now turn to the remaining two edges labelled T-duality. This is a key concept in
string theory deserving a whole section, but we have to limit ourselves to a short
qualitative explanation for reasons of space. To be specific, focus on the connection
between type IIA and IIB. In this case, the statement of T-duality amounts to the
following:
Consider type IIA with coupling gs compactified on an S 1 of radius R to 9d. At
the same time, consider type IIB with coupling gs compactified on an S 1 of radius
R . The two resulting 9d theories are exactly identical if
√
α α
R = and gs = gs . (4.72)
R R
In other words, one theory compactified on a very small (in string units) S 1 is
equivalent to the other theory compactified on a large S 1 . A simple first step in
checking this statement is to identify the spectrum of states in the two models. This
can be immediately realised with the tools already available to the reader. The key
4.9 Problems 135
point is that the tower of so-called winding states (the string wrapped around the S 1 )
on one side is mapped to the tower of Kaluza–Klein states (waves travelling around
the S 1 , cf. Sect. 5.2) on the other side. The winding number is mapped to the discrete
momentum characterising the motion in a compact dimension. A more detailed (also
10d field-theoretic) set of rules for the identification, the Buscher rules, can be found
in [131]. The procedure of T-dualisation can be iterated, such that e.g. one theory on
a small T 6 is identified with the same theory (because the number of dualisations is
even) on a large T 6 .
A key conceptual point of T-duality is that it can be viewed as a manifestation
of the fact that the stringy UV completion really abandons the concept of smooth
spacetime at sub-stringy distances: Making some compact manifold small does
not lead to a new sub-stringy geometry but rather to a familiar super-string-sized
geometry, possibly in a different theory.
4.9 Problems
4.9.1 Explicit State-Operator Mapping in the Free Case
Task Calculate explicitly the operators which, if inserted at z = 0 in the radial

description of the closed string, define the single-particle excited states created by
μ
α−m with m ≥ 1.
Hints Work with the euclidean (Wick-rotated) version of the theory, defining
e.g. (σ 1 , σ 2 ) = (σ 1 , iσ 0 ). Write w = σ 1 + iσ 2 , such that the worldsheet
cylinder corresponds to a vertical strip with width π in the complex w plane. Define
z = exp(−2iw), such that constant-time cuts of the cylinder are mapped to circles
in the w-plane. The origin of the z plane now corresponds to the infinite past of the
cylinder, σ 0 = −i∞.
Express our mode expansion of ∂− X (we suppress the index μ for brevity) in
terms of the variable z. Invert the result, expressing the oscillator modes in terms of
integrals of ∂X over a closed contour in the z plane.
Finally, use the expression obtained for a creation operator αn under a path
integral over fields on the z plane. Assuming that the fields X can be Taylor
expanded in z and z at the origin, obtain the desired expression for the vertex
operators. Start by arguing why the vacuum state with momentum p = 0
corresponds to the unit operator.
Solution We start by rewriting our formula from the lecture (with l = 1) as
1 1 i 1
XR = x + pσ− + αn e−2inσ− (4.73)
2 2 2 n
n=0
as

∂− X = αn e−2inσ− , (4.74)
n
where we also set p/2 = α0 . Using σ− = σ 0 − σ 1 = −iσ 2 − σ 1 = −w, this

becomes

− ∂w X = αn e2inw . (4.75)
n
Next, with 2w = i ln z, we have

i αn
− 2iz∂z X = αn z−n or ∂z X = . (4.76)
n
2 n zn+1
The coefficient of 1/zn+1 is extracted, using the residue theorem, by performing a

counter-clockwise contour integral with the measure dz zn /(2πi):
.
dz n
αn = −2 z ∂z X . (4.77)
2π
In the above, αn is an operator acting on a state, defined on one of the circles in

the z plane. Similarly, X is a local field operator integrated over this circle. All of this
is to be interpreted at the fixed radial time corresponding to this circle. Obviously,
such an operator identity can be used under the path integral, with some operator
inserted at z = 0 to define the initial state and with the boundary conditions at
|z| → ∞ defining the final state. The latter will not be relevant for us and we will
ignore them.
Start by inserting the unit operator at z = 0 and calling the corresponding (so far
unknown) state | :
Xf (rf ) .
dz n
lim e−(tf −t )H αn e−(t −ti )H | ∼ DX e−SP [X] z ∂z X .
ti →−∞ 2π
(4.78)
Here the r.h. side is to be interpreted as a functional depending on the boundary

conditions X = Xf at some largest circle of radius rf . The l.h. side is defined
by evolving an initial state | to the time t, corresponding to the radius r of
the contour integral on the r.h. side, applying αm , and then evolving to the final
time tf corresponding to rf . The tilde means that we are not keeping track of
normalisations.
It is immediately clear that, assuming that we integrate over well-behaved
functions X, the r.h. side vanishes for n ≥ 0 since there are no appropriate poles
4.9 Problems 137
inside the contour. But the state annihilated by all αn with non-negative n is, by
definition, the vacuum: | = |0, 0.
Next, we consider creation operators, α−n with n > 0. We also use that the
vacuum corresponds to the unit operator and repeat the step from (4.77) to (4.78)
for this case:
Xf (rf ) .
dz 1
α−n |0, 0 ∼ DX e−SP [X] n−1
∂z X , (4.79)
2π z z
where now n > 0. We have simplified the l.h. side since, as noted, we do not keep
track of the normalisation. Finally, we may Taylor expand X(z, z) keeping only the
term which will provide a non-zero contribution to the contour integral:
Xf (rf ) (∂z )n X(0)
α−n |0, 0 ∼ DX e−SP [X] . (4.80)
(n − 1)!
Thus, up to normalisation, (∂z )n X(0)/(n − 1)! is our final result for the operator
corresponding to the creation operator α−n .
4.9.2 Euler Number and Genus of Riemann Surfaces
Task Calculate explicitly the Ricci scalar R of a 2-sphere of radius R and use this
result to derive the formula
χ() = 2 − 2g (4.81)
for the Euler number

1
χ() ≡ d 2σ det(gab ) R . (4.82)
4π
Here g is the ‘number of holes’ or ‘number of handles” of the Riemann surface.
Hints Recall that the Riemann tensor in 2d is highly symmetric and that you hence
do not need to calculate all components to obtain the Ricci scalar. In the second part
of the problem, it will be sufficient if you give a ‘physicist’s derivation’, drawing
lots of pictures and taking the existence of intuitively obvious limits for granted.
Solution We will use the standard parameterisation of the unit sphere by azimuthal
and polar angle, such that
ds 2 = dθ 2 + sin2 θ dφ 2 . (4.83)
Recalling our general 2d result
1
Rabcd = (gac gbd − gad gbc )R , (4.84)
2
from the discussion of the symmetries of the bosonic string, we have
1 1
Rθφθφ = gθθ gφφ R = sin2 θ R . (4.85)
2 2
Thus,
2 2
R= 2
Rθφθφ = Rθφθ φ gφφ = 2 Rθφθ φ . (4.86)
sin θ sin2 θ
The required curvature coefficient can be obtained from the standard formula
Rabc d = −∂a bc d + ac e be d − { a ↔ b } . (4.87)
It is explicitly given by
Rθφθ φ = −∂θ φθ φ + θθ e φe φ + ∂φ θθ φ − φθ e θe φ . (4.88)
Using the standard formula
1 cd
ab c = g (∂a gbd + ∂b gad − ∂d gab ) (4.89)
2
we calculate the Christoffel symbols
1 φφ 1 cos θ
φθ φ = g (∂φ gθφ + ∂θ gφφ − ∂φ gφθ ) = g φφ ∂θ gφφ = , (4.90)
2 2 sin θ
1
θθ φ = g φφ (2∂θ gθφ − ∂φ gθθ ) = 0 , (4.91)
2
θθ θ = 0 , (4.92)
1 θθ
φθ θ = g (∂φ gθθ + ∂θ gφθ − ∂θ gφθ ) = 0 . (4.93)
2
Here the zero result in the third line is obvious since the only non-zero derivative
∂θ gφφ cannot appear.
With this, we finally obtain

cos θ cos2 θ cos2 θ cos2 θ
Rθφθ φ = −∂θ − 2
=1+ 2
− = 1, (4.94)
sin θ sin θ sin θ sin2 θ
4.9 Problems 139
hence R = 2 for the unit sphere and
R = 2/R 2 (4.95)
for a sphere of radius R.

Since the surface is 4πR 2 , we obtain the Euler number
χ(S 2 ) = 2 , (4.96)
consistent with (4.81) and the absence of handles on a sphere.

Now let us move on to the case of a torus, i.e. a sphere with one handle, g = 1.
On the one hand, it is clear that
χ(T 2 ) = 0 (4.97)
since an explicit geometry with everywhere vanishing curvature can easily be given.
On the other hand, one can deform the geometry to a ‘pancake’ with a handle
attached in its upper flat region, cf. Fig. 4.7. If there were no handle, the curvature
integral in (4.82) would give χ = 2, with the only contribution coming from the
edge of the pancake. With the handle, we know we get zero. Thus, the hatched
regions where the handle is attached give a negative contribution of −2 to the
curvature integral defining χ. Obviously, further handles will give an identical
negative contribution, demonstrating the correctness of the term −2g in (4.81).
4.9.3 Dilaton vs. String Coupling
Task Give an argument for identifying the dilaton in the exp(−2φ) prefactor of the
26d Einstein–Hilbert term with the dilaton defined as the coefficient of the Einstein–
Hilbert term on the worldsheet.
Hint Think of the loop expansion parameters in 26d-field-theory and on the

worldsheet.
Fig. 4.7 A torus deformed to a ‘pancake’ with a handle attached. The handle is realised by two
‘smokestacks’, to be identified at their edges. The only regions with non-zero curvature are at the
edge of the pancake and in the hatched areas where the handle is attached
Solution Think of graviton-graviton scattering in 26d quantum gravity as a low-

energy effective field theory. Work in the string frame and treat the dilaton as fixed
to some background VEV φ0 , plus small fluctuations which we will not be interested
in. Expanding the metric as ημν + hμν and rescaling hμν → hμν κeφ0 , we see that
3-vertices are proportional to κeφ0 and 4-vertices to κ 2 e2φ0 . Hence, for example, a 1-
loop contribution to a given process is suppressed relative to the tree level by κ 2 e2φ0 .
(Draw a few example diagrams
√ to be certain.) Moreover, we can set κ 2 ∼ Ms−24 ,

with Ms ∼ 1/ ls ∼ 1/ α the string scale, as explained in the lecture. Finally,
we assume that the string scale provides the cutoff for the UV divergent loop
diagrams. Thus, the ratio of 1-loop to tree level is
κ 2 e2φ0 24 ∼ e2φ0 ∼ gs2 , (4.98)
consistently with the expectation from the amplitude formula of the worldsheet
analysis, under the assumption that φ0 governs the worldsheet Einstein–Hilbert
term. Up to an additive redefinition of φ, this identifies the two a priori different
definitions of the dilaton.
4.9.4 Elementary Exercises with 2d Spinors
Task Make the action of SO(1,1) and SO(2) on vectors and spinors completely
explicit, paying particular attention to how the transformations of vectors and
spinors differ in the Lorentz case.
Hints Fix the normalisation of generators by analogy to the familiar 4d case. Recall
what you know from undergraduate special relativity.
Solution In the non-abelian case (SO(1, d − 1) with d > 2), the normalisation of
the generators Jab is unambiguously fixed by the non-trivial Lie algebra relations
[Jab , Jcd ] = i(ηbc Jad − ηac Jbd − ηbd Jac + ηad Jbc ) , (4.99)
which implies
(Jab )cd = i(δac δbd − δad δbd ) . (4.100)
In the SO(1, 1) (and similarly in the SO(2))) case, the Lie algebra is trivial and
does not fix the normalisation. We still use the general-d definition, such that
a
01 0 −1
(J01 )ab = i and (J01 ) a
b =i . (4.101)
−1 0 ab
−1 0 b
4.9 Problems 141
Hence a boost specified by

01 = −
10 = α/2 explicitly reads

0α cosh α sinh α
exp(i
Jab ) = exp
ab
= . (4.102)
α 0 sinh α cosh α
The last equality follows, e.g. from its obvious infinitesimal version together with
the group property, which in turn follows from the well-known formulae for
cosh(α + β) and sinh(α + β). This is where remembering undergraduate special
relativity is useful.
Moreover, it is convenient to switch from the coordinates x 0,1 underlying the

above formulae to lightcone coordinates, x ± = x 0 ± x 1 . One has x + = x 0 + x 1 =
x 0 cosh α + x 1 sinh α + x 0 sinh α + x 1 cosh α = x + exp α and similarly for x − .
Hence,

eα 0
exp(i
Jab ) =
ab
(4.103)
0 e−α
in that basis.
Next, we have

1 1 −1 0
[γ0 , γ1 ] = (4.104)
4 2 01
and hence

eα/2 0
S(α) = exp(i
{i[γ0, γ1 ]/4}) =
ab
. (4.105)
0 e−α/2
We now see explicitly how SO(1, 1), here realised as R with addition as the group
operation, is represented in two different ways on vectors and spinors.
Repeating the analysis for SO(2), we now label the coordinates by 1, 2 rather
than 0, 1 since no special role is played by x 0 = t. The lower-index version of J ,
now called (J12 )ab , remains unchanged. The upper-lower version reads
a
01
(J12 ) a
b =i (4.106)
−1 0 b
and hence

cos α sin α
exp(i
Jab ) =
ab
. (4.107)
− sin α cos α
The correct Clifford algebra is obtained if γ0 is multiplied by ‘i’, such that dilation
by eα/2 becomes a phase rotation by half of the SO(2) rotation angle:

eiα/2 0
S(α) = exp(i
ab {i[γ0, γ1 ]/4}) = . (4.108)
0 e−iα/2
This was, of course, expected.
4.9.5 SUSY Algebra in 2d
Task Check the 2d SUSY algebra given in the lecture using the explicit definitions
of Q and Q.
Hints Show that ψχ = χψ if we impose a Majorana condition on our 2d spinors.

α
Then work out explicitly what Q is in terms of θ and ∂/∂θ . It is convenient to think
of the action of bilinears like (∂/∂θα )
α . After these preliminaries, write down the
commutator of
Q and Qη, which is equivalent to the SUSY algebra (as you already
learned in 4d).
Solution Let us start by checking that ψχ = χψ for Majorana spinors. In our case,
Majorana spinors are simply spinors with real components. Hence,
T
ψ− 0 −i χ−
ψχ = ψ † γ 0 χ = = i(ψ+ χ− − ψ− χ+ )
ψ+ i 0 χ+
= i(χ+ ψ− − χ− ψ+ ) = χψ . (4.109)
One may think of this relation as of an alternative definition of what a Majorana

spinor is.
Next, we want to understand how the formal ∗-operation must act on ∂/∂θ for Q
to be Majorana. For this purpose, consider

α ∂

α (θ ψ) =
ψ (4.110)
∂θ
and

∂ β

α (ψθ ) =
α ψ δβ α = −ψ
= −
ψ . (4.111)
∂θα
We see that, with the definition

∂ ∂
=− , (4.112)
∂θ ∂θ
4.9 Problems 143
the spinor ∂/∂θ is a Majorana spinor in the sense that it obeys relations like (4.109).
Note that this peculiar minus sign is similar to the one encountered in 4d SUSY.
Also, we have
i(∂/)θ ) = −iθ † γ 0 γ 0 ∂/† γ 0 = −iθ ∂/ . (4.113)
Using the definition of Qα from the lecture, this gives
α ∂ ∂
Q =− − i(θ ∂/)α = − − i(θγ a )α ∂a . (4.114)
∂θα ∂θα
It is clear that our Q is Majorana: The first term is Majorana according to what was
said above, the second term is Majorana because θ is Majorana and both ∂a and iγ a
are real.
Now the actual calculation is easy. Using two Majorana SUSY parameters
and
η, we have

α ∂ ∂
[
Q, Qη] =
α + i
∂/θ , − ηβ − iθ ∂/η = −2i
∂/η . (4.115)
∂θ ∂θβ
Crucially, since Q and Q are not independent, there are no additional QQ or QQ

relations.
10d Actions and Compactification
5
5.1 10d Supergravities and Type IIB as an Example
The existence of a field-theoretic target space action and its fundamental relation
to the worldsheet definition of the theory has already been discussed in the context
of the bosonic string. All that was said there remains true. In particular, there are
always the 10d graviton and B2 field, coupling to the worldsheet (except in the case
of the unoriented type I string, where it falls victim to the projection taking us from
type IIB to type I). There is also always the dilaton, governing the convergence of
perturbation theory. Together, dilaton, graviton and B2 form the NS–NS sector (see
above). As a novelty, one has the Cp+1 (or RR) form fields and the corresponding
Dp-branes. These are dynamical objects, just like the string itself, but with different
dimensionalities and (at weak string coupling) larger tensions. This analysis would
be slightly different in the heterotic case, where there are no C-forms but rather
gauge fields. Crucially, there are now also fermionic partners for all the fields above.
What is very different from the bosonic case is the uniqueness status of the above
five 10d theories. In the bosonic case, we found one of many possible, similar 26d
field theories. For example, without the stringy definition, one could just add another
scalar or modify some coupling. Here, by contrast, our five 10d theories are very
special and (at the 2-derivative level) even unique, independently of their string-
theoretic underpinning. This is due to supergravity. Indeed, supersymmetry (and
supergravity) exists in various dimensions (cf. Appendix of volume II of [7]), but its
realisation becomes harder and harder as the dimension grows. This can be roughly
understood by noting that the spinor dimension grows exponentially with D, making
it more difficult to find a matching bosonic structure. Thus, it turns out that there
exist precisely four supergravity theories in 10d, and all of them can be viewed as
coming from the five superstring constructions we discussed (recall that heteretotic
SO(32) and type I give the same field theory).

146 5 10d Actions and Compactification
Let us start by noting that for even D one has
0 , · · · , D−1 and ≡ 0 · 1 · · · · · D−1 , (5.1)
allowing us to define chirality through the projector (1 + )/2. In the dimension

(D + 1), which is now odd, becomes the highest ‘usual’ gamma matrix and the
product of all gamma matrices becomes · ∼ 1. Hence, chirality cannot be
defined.
In some dimensions (see [7]), there is a Majorana spinor and, if both Weyl
and Majorana spinors exist independently, it is sometimes possible to impose both
constraints together. We have seen that this happens in D = 2, where the naive
spinor dimension is 2D/2 = 2, i.e. 4 real d.o.f., and we found spinors with one real
component.
This situation occurs again in D = 10, where the Dirac spinor has 32 components
and a 16-component real spinor exists. This spinor has 4 times the degrees of
freedom of a minimal 4d spinor; hence, the minimal 10d SUSY is referred to as
N = 4 SUSY in 4d language. One may also characterise this as 10d N = 1
SUSY. We have encountered a gauge theory with this amount of supersymmetry
when we quantised the open superstring in 10d. This gauge theory (more precisely,
its non-abelian version) can be coupled to supergravity, and it is the SUSY of the
heterotic and type I theories. It is also possible to have 10d N = 2 supergravity
(corresponding to N = 8 in 4d language). Gauge fields cannot be added to
such theories. This is the SUSY of the type II string. As already noted, SUSY
is so constraining that (under very reasonable assumptions) these four theories
can be shown to be the only 10d supergravities. This uniqueness includes the
gauge group—only E8 × E8 and SO(32) are possible on anomaly cancellation
grounds [132]. It is very intriguing that precisely these four 10d supersymmetric
field theories are realised in string theory. All these theories are also unique in the
sense that no dimensionless parameters are present. An equally unique supergravity
theory exists in 11d—it is the 11d theory linked to type IIA via compactification on
S 1 as noted earlier. No other supergravity theories above 9 dimensions are known.
All of them have been tried as starting points for a stringy description of the real
world. The landscape, i.e. a very large number of potentially suitable 4d models, has
been most convincingly established in type IIB (although there are still reasonable
doubts, to which we will come). We hence focus on this theory.
In the widely used conventions of [7, 133], the bosonic part of the string-frame
type IIB action reads

1 √ 1 1 1 1
S= d 10 x −g e−2φ R + 4(∂φ)2 − H32 − F12 − F̃32 − F̃52
2
2κ10 2 · 3! 2 2 · 3! 4 · 5!
+SCS + Sloc . (5.2)
5.1 10d Supergravities and Type IIB as an Example 147
2 = (2π)7 α 4 and
Here 2κ10
1 1
F̃3 = F3 − C0 ∧ H3 , F̃5 = F5 − C2 ∧ H3 + B2 ∧ F3 . (5.3)
2 2
The RR form field strengths with a tilde are gauge invariant (as is H3 ). This implies
special gauge transformation properties of some of the potentials, e.g.
1
C2 → C2 + dλ1 goes together with C4 → C4 + λ1 ∧ H3 . (5.4)
2
Furthermore, terms that do not involve the metric are often referred to as Chern–
Simons terms. In our case it reads

1
SCS = 2 eφ C4 ∧ H3 ∧ F3 . (5.5)
4κ10
Finally, we collect the actions of the various branes (including extended fundamental
strings) that may be present in the target space and are described by the ‘localised’
part Sloc . We just display the example of a D3-brane

1 √ 1
Sloc ⊃ SD3 = C4 − d 4ξ −g T3 with T3 = .
2π 3 α 2 D3 D3 (2π)3 α 2
(5.6)
The first part of SD3 may also be called a Chern–Simons-type term since the metric
does not enter. The coordinates ξ parameterise the world volume of the brane, and
the metric next to them is the pullback of the 10d metric. Analogous expressions for
the other odd-dimensional Dp-branes and the string have to be added. The general
formula for the tension appearing in the SDp is
e(p−3)φ/4
Tp = . (5.7)
(2π)p α (p+1)/2
This Sloc is not yet complete: It should be extended by including the brane-localised
(open-string-derived) gauge fields and their fermionic partners. The resulting so-
called DBI or Dirac–Born–Infeld action1 also includes the pullpack of B2 to the
brane. Moreover, there are further brane-localised Chern–Simons terms. For us, it is
sufficient to record that, at leading order, the gauge fields come in simply through a
brane-localised N = 4 U (1) or, in the case of a brane stack, SU (N) gauge theory
lagrangian. For many more details, see e.g. [108, 134].
1 Thereader may be surprised to see these names from the pre-stringy era of physics. It is their
work on the closely related non-linear extensions of electrodynamics which is honoured here.
We do not display the completely analogous expression for type IIA, where the
relevant RR form fields are C1 and C3 . We only note that the non-localised CS term
takes the form

B2 ∧ F4 ∧ F4 . (5.8)
The action for type I follows from that of type IIB upon a so-called orientifold
projection, i.e. a projection on states invariant under worldsheet-parity inversion. In
10d, this implies the removal of C0 , B2 and C4 . Furthermore, 32 D9-branes have
to be added, also subject to a certain projection, which restricts the gauge group
to SO(32). Thus, one basically includes the lagrangian of the corresponding 10d
super-Yang–Mills (SYM) theory.
Finally, in the heterotic case one removes the C-forms (keeping B2 ) and adds
SYM lagrangians with groups E8 ×E8 or SO(32). It is then clear that the advertised
duality between the type I and the heterotic SO(32) theory also involves the
exchange of the F3 and H3 .
We recall again that the fermionic parts also differ strongly between the various
theories, given in particular that SUSY is reduced to 10d N = 1 in all but the two
type II theories.
5.2 Kaluza–Klein Compactification
One has thus arrived at a possibly fundamental and (involving the various dualities
above) unique 10d theory. To describe a 4d world on this basis, the logical procedure
is to employ the idea of Kaluza–Klein compactification. This method of obtaining
lower from higher-dimensional theories is old and has, as we will see, some appeal
in its own right [135–140].
Let us start with what may be the simplest example: a 5d scalar field on M =
R4 × S 1 , where the S 1 has radius R (such that x 5 ∈ (0, 2πR)):

1
S= d 5 x (∂M φ)(∂ M φ) , M ∈ {0, 1, 2, 3, 5} . (5.9)
M 2
We take φ = 0 (in fact any other value, φ = const., would be equally good) as our
vacuum and parameterise fluctuations around this solution according to
∞
∞

φ(x, y) = φnc (x) cos(ny/R) + φns (x) sin(ny/R) . (5.10)
n=0 n=1
5.2 Kaluza–Klein Compactification 149
Here we renamed x 5 according to x 5 → y, and we use the argument x as x =

{x 0 , x 1 , x 2 , x 3 }. One immediately finds

1
∞
1
S = 2πR 4
d x (∂φ0 ) +
c 2
(∂φn ) + mn (φn ) + (∂φn ) + mn (φn )
c 2 2 c 2 s 2 2 s 2
,
2 4
n=1
(5.11)
with mn = n/R. Hence, our model is exactly equivalent to a 4d theory with one
massless field and a (doubly degenerate) tower of KK modes. The massless mode
parameterises a ‘flat direction’, i.e. it is not only massless but has no potential at all.
It can hence take an arbitrary constant value, which would still be a solution. Such
a field is called a modulus.
We will frequently encounter cases where the value of the modulus governs the
masses and couplings of the rest of the 4d theory. To create such a situation in our
toy model, enrich our theory by 5d fermions and introduce the 5d coupling
λφψψ . (5.12)
It is an easy exercise to derive the 4d action as above and read off explicitly how
the 4d fermion masses depend on the VEV of φ. Now φ is more like one of the
moduli we will encounter in more realistic cases below. We note that our ‘modulus’
has a serious problem: It will acquire a mass from loop corrections even in the 5d
local lagrangian. In this sense it is really not a proper modulus. We will see better
examples below.
Indeed, let us now turn to the historical example that is most directly associated
with the word Kaluza–Klein theory. Consider pure general relativity in 5d,

MP3 ,5 √
S= d 4 x dy −g5 R5 , (5.13)
2
parameterise the metric as

√
gμν + 2 2 2
√(2/MP )φ2 Aμ Aν ( 2/MP2)φ Aμ ,
(g5 )MN = (5.14)
( 2/MP )φ Aν φ
where
M, N, · · · ∈ {0, 1, 2, 3, 5} and μ, ν, · · · ∈ {0, 1, 2, 3} . (5.15)
Here MP ,5 is the 5d reduced Planck mass, while MP is at the moment just a

parameter that, together with the fields Aμ and φ, is used to characterise all
components of the 5d metric.
As above, we assume that y ∈ (0, 2πR) parameterises an S 1 , and we base our
analysis on the solution gμν = ημν , Aμ = 0 and φ 2 = g55 = 1. From on our scalar
field example, we expect that the Fourier decomposition of all fields as functions
of y will give a 4d theory with a tower of massive modes. Focussing on the zero
modes only corresponds to assuming that all fields are independent of y. Under this
assumption, it is straightforward to work out the higher-dimensional action, i.e. the
5d Ricci scalar, in terms of the ansatz (5.14). The result reads [138]

√ MP2 1 M 2 (∂φ)2
S= d x −g φ
4
R − φ 2 Fμν F μν + P , (5.16)
2 4 3 φ2
which can of course be brought to the Einstein frame by gμν → gμν /φ.
The key lessons are that the (zero modes of the) 5d metric degrees of freedom
have turned into the 4d metric, an abelian gauge field and a scalar. The appearance of
a U (1) gauge theory is not surprising since our starting point, the R4 × S 1 geometry,
clearly has a global U (1) symmetry. But, since we are in general relativity and our
starting point is diffeomorphism invariant, we are also allowed to rotate the S 1 (i.e.
shift y) differently at every point x. Hence, our symmetry must actually be a U (1)
gauge symmetry.
Moreover, we have a solution of the 5d Einstein equations for every fixed radius
R. Thus, we expect a scalar degree of freedom, corresponding to R, with an exactly
flat potential. This degree of freedom is the scalar field φ. Note that it is sometimes
convenient to parameterise the S 1 by the dimensionless variable y ∈ (0, 1) and
√
correspondingly to have φ = g55 = 2πR in the vacuum. We note that, while MP
was originally introduced as a parameter in the metric ansatz, the result (5.16) used
the identification
MP2 = 2πR MP3 ,5 . (5.17)
Before closing this generic Kaluza–Klein section, it will be useful to consider

yet another example: Let the geometry again be R4 × S 1 , and let the 5d lagrangian
contain a U (1) gauge theory. For simplicity, we will focus on the Kaluza–Klein
or dimensional reduction of this U (1), ignoring the 5d gravity part that we just
discussed. Thus, we start with

1
S = d x dy − 2 FMN F
4 MN
, (5.18)
4g5
where gMN = ηMN and y ∈ (0, 2πR). With the ansatz
AM = (Aμ , φ), (5.19)
one finds, at the zero-mode level,

1 1
S = d 4 x − 2 Fμν F μν − 2 (∂φ)2 . (5.20)
4g 2g
5.3 Towards Calabi–Yau Manifolds 151
Here 1/g 2 ≡ 2πR/g52 . The crucial lesson is that the 5d gauge field gives rise to a 4d
gauge field and a scalar, the latter being associated to A5 or, in a more geometrical
language, to the Wilson line integral
.
A= dy A5 = 2πR φ(x) . (5.21)
This Wilson line measures the phase which a charged particle acquires upon moving
once around the S 1 , just as in the Aharonov–Bohm experiment. Assuming that
the minimally charged particle (we do not display the corresponding part of the
lagrangian) has unit charge, the phase measured by φ becomes equivalent to zero for
φ = 1/R. Thus, we have found an exactly massless (at the classical level) periodic
scalar field, also known as an axion or axion-like particle or ALP.
Let us draw a lesson from the above that will also be important for string
compactifications, to be discussed below. We have seen two types of moduli arise,
one associated to the geometry of the compact space (g55 ), the other to the gauge
field configuration in the compact space (A5 ). Both have no classical potential since
in one case 5d diffeomorphism invariance, in the other case 5d gauge invariance
forbids the corresponding potential term. Moreover, due to this symmetry argument
5d loop corrections do not induce such a potential. However, in both cases 4d loop
corrections can provide a potential and, hence, give a mass to the above fields. This
is not in contradiction to the symmetry argument just stated since 4d loop effects can
in general not be written in terms of 5d local operators. However, in the presence
of enough supersymmetry in the resulting 4d theory, these loop corrections may
vanish such that the relevant moduli remain exactly massless or, more precisely,
their potential remains identically zero as an exact statement. This generally happens
in 4d N = 2 SUSY.
5.3 Towards Calabi–Yau Manifolds
We now want to explain how the 10d SUGRA theories provided by the superstring
can be compactified to 4d. There are two approaches: We could start by developing
the toy model path started in the previous section, i.e. we could consider the
geometry R9 × S 1 . This would give us a 9d theory, without too many new features
(except for supersymmetry, which would keep all moduli massless). Next, we could
consider R8 × T 2 . We would now encounter moduli corresponding to g88 , g99 and
g89 , characterising both size and the shape of the torus. Thus, we would get an
8d theory with (at least) 3 scalars corresponding to geometric moduli. Much more
could be said about compactifications on tori and related simple geometries.
However, we will take a different approach and first introduce a much more
general and powerful set of examples—the Calabi–Yau geometries. These are the
compactification spaces on which the landscape as we presently understand it is
mostly built. Later on, we will return to tori to illustrate some of the more abstract
concepts used.
Our key starting point is the desire to find a solution of the 10d equations of
motion corresponding to a 4d world. Setting all fields except the metric to zero,
this implies that we must have (R10 )MN = 0 to solve Einstein’s equations. This
condition is called Ricci flatness, and it is obviously satisfied for S 1 and the (flat)
tori T n mentioned above. The interesting and non-trivial fact is that there exists a
large class of relatively complicated compact 6d manifolds that are also Ricci flat
and, hence, represent suitable compactification spaces: the Calabi–Yau manifolds.
Before giving the definition, we need a few geometric concepts. Our treatment
will be extremely brief and hence, unfortunately, superficial. Much more material
can be found e.g. in [8, 101, 141–145].
To begin, Calabi–Yau manifolds are complex manifolds. This is a fairly straight-
forward generalisation of the familiar concept of a 2n-dimensional real differen-
tiable manifold X. The key new point is that the charts,
(Ui , φi ) , φi : Ui → φi (Ui ) ⊂ Cn , (5.22)
are now maps from open sets Ui of X to Cn , with the key compatibility condition
being that the functions φj ◦ φi−1 are holomorphic. In other words, our manifold
locally looks like Cn and coordinate changes are of the form
z i = z i (z1 , · · · , zn ) , (5.23)
with any appearance of zı in the argument of the new coordinate excluded.

On a complex manifold, it makes sense to complexify tangent and cotangent
space as well as all their higher tensor products. Thus, tensor fields are complex.
For example, local bases of tangent and cotangent space are provided by
∂ ∂
i
, and dzi , dzı , (5.24)
∂z ∂zı
with zi = x i + iy i etc. It is natural to define the tensor
∂ ∂
J = i dzi ⊗ i
− i dzı ⊗ ı , (5.25)
∂z ∂z
which may obviously be interpreted as a map Tp∗ → Tp∗ for every p ∈ X. It, roughly
speaking, corresponds to ‘multiplication by i’ in cotangent space. Its components
are

i1 0
J = (5.26)
0 −i1
in a complex basis and

0 1
J = (5.27)
−1 0
in a real basis. A crucial feature is J 2 = −1.

A real manifold with a tensor J as above is called an almost complex manifold
and J is called an almost complex structure. If such a J satisfies a certain
integrability condition (vanishing of the Nijenhuis tensor), complex coordinates can
be given and J turns into the so-called complex structure of a complex manifold.
We will only be interested in this latter case.
Even more, we will demand that our manifold has a metric (is a Riemannian
manifold) and that this metric is compatible with J . In other words, we demand
that J is covariantly constant. This turns the manifold into a Kahler manifold, a
concept which we already used when discussing field spaces of supersymmetric
field theories. We will not demonstrate this but give right away a stronger definition:
A complex manifold with a metric is called Kahler if the metric can locally be
written as
∂ 2K
gij = , (5.28)
∂zi ∂zj
with K a real function defined in every patch and with gij = gıj = 0. We note that
this last condition by itself would make the metric hermitian, but we are interested
only in the stronger Kahler condition.
We also note that the metric allows us to lower the second index of J , turning
J into a rank-2 lower-index tensor. This tensor turns out to be antisymmetric and
hence defines a 2-form, the so-called Kahler form
J = igij dzi ∧ dzj . (5.29)
We see that, given a complex structure, the 2-form J determines the metric and
vice versa. This will become important below when we will be discussing different
metrics on the same differentiable manifold.
Next, we need the concept of holonomy. We know from basic differential geom-
etry that, with a metric, one gets a unique Riemannian or Levi-Civita connection and
hence the possibility to parallel-transport tangent vectors. Given any point p ∈ X
and any closed curve C beginning and ending in p, we hence have a linear map
R(C) : Tp → Tp or R(C) ∈ SO(2n) . (5.30)
The latter statement follows if we assume orientability (for complex manifolds this
is guaranteed) and recall that the Riemannian parallel transport does not change the
length of a vector. It can be shown that the set of all R(C) forms a group and that
this group does not depend on the choice of p (assuming X is connected). This is
the holonomy group.
We are now in the position to give one (of the many equivalent) definition of a
Calabi–Yau manifold: A Calabi–Yau 3-fold (our case of interest) is a compact,
complex Kahler manifold with SU (3) holonomy. More generally, for a complex
n-fold, one demands that the holonomy is SU (n) ⊂ SO(2n). As we will argue in
a moment, this implies that some of the 10d supersymmetry is preserved in the 4d
effective field theory and that Einstein equations are solved without sources (Ricci
flatness).
Though the Einstein equations are maybe physically more important, we will
start with SUSY. Very superficially, we expect that a 4d supersymmetric effective
theory will have massless spinors. Hence spinor fields need to have zero modes. In
the simplest case this corresponds to the existence of covariantly constant spinors on
the compactification space. We will see in a moment that this covariantly constant
spinor is intimately linked to SU (3) holonomy.
But let us first give a more careful argument for why unbroken SUSY requires the
compact space to have a covariantly constant spinor: While we have not given the
supergravity transformations of the various fields in 10d, we may recall the 2d case
of worldsheet-SUGRA: Here, we have seen that the transformation of the gravitino
is proportional to the covariant derivative of the SUSY parameter, i.e. of the spinor
ξ(σ ):
δξ χa = ∇a ξ . (5.31)
This is similar in 10d. Hence, to identify a 4d SUSY parameter under which the
vacuum is invariant, one needs a covariantly constant spinor. On a curved manifold,
this is a non-trivial requirement.
To see this in more detail, we need the group-theoretic fact that SO(6) =
Spin(6)/Z2 , Spin(6) = SU (4) and that the Weyl spinor representation of Spin(6)
(i.e. a 6d spinor in euclidean signature) transforms in the 4 of SU (4), using the
previous isomorphism. Furthermore, we embed our 10d spinor in the tensor product
of 4d spinor and 6d spinor. Since 4d space is flat, the critical issue for the constancy
of our 10d spinor is the constancy of its 6d spinor part. In other words, we have to
take the 6d spinor to be covariantly constant along X. Furthermore, without loss of
generality we assume that in SU (4) notation this spinor takes the form
⎛ ⎞
0
⎜ 0 ⎟
ξ(p) = ⎜
⎝ 0 ⎠
⎟ (5.32)
ξ0 (p)
at some point p ∈ X. Since it is part of a covariantly constant spinor field, the

parallel transport will follow this field and, in particular, bring ξ(p) back to itself
for any loop C. But this clearly means that the holonomy matrices may only act on
the first 3 components, i.e. we need SU (3) holonomy.
The reverse is obvious: Given SU (3) holomomy, a covariantly constant spinor
can be constructed by parallel transporting ξ(p) given above to any point of X. The
only way in which this might fail is if the construction were ambiguous, i.e. if two
different paths from p to p gave rise to two different spinors ξ(p ). But this would
imply that a closed path starting at p exists along which the parallel transport of
ξ(p) is non-trivial. This would be in contradiction to SU (3) holonomy. Thus, we
have the equivalence between SU (3) holonomy and the existence of a covariantly
constant spinor, i.e. the survival of 4d SUSY.2
Next, we consider Ricci flatness. We first note that, on Kahler manifolds, the only
non-zero components of the Levi–Civita connection are
ij k = g kl ∂i gj l and ıj k = g kl ∂ı glj . (5.33)
This leads to significant simplifications for the Riemann tensor and the Ricci tensor,
which we do not work out. For example, the only non-vanishing Riemann tensor
components are of the form
Rij kl , (5.34)
and those related by antisymmetry in the first and second index pairs. In other words,
the first and second as well the third and fourth index pairs have to be of opposite
type (holomorphic and antiholomorphic). Moreover, the Ricci tensor can be written
as
Rij = ∂i ∂j ln det g . (5.35)
(This is standard mathematics material, see e.g. [146], and many other sources.)
As is well known, the significance of Rij α β is that, if interpreted as a matrix
with indices α and β, it describes the rotation of a covector upon parallel transport
along a loop with orientation specified by i and j . Here we use greek letters
for the second index pair to emphasise that they can take either holomorphic or
antiholomorphic values, e.g. α = k or α = k. The previously noted restrictions
on holomorphy/antiholomorphy of the second index pair mean that either (α, β) =
(k, l) or (α, β) = (k, l). This can straightforwardly be shown to imply that the
corresponding rotation matrix is in the U (n) subgroup of the general holonomy
group SO(2n). More generally, the conditions of a manifold being Kahler and
having U (n) holonomy are equivalent.
2 More precisely, 4d N = 2 SUSY in the type II case and 4d N = 1 SUSY in the type I and
heterotic case. The reason is the presence of two independent 10d SUSY generators in the former
situation.
Since U (n) = SU (n) × U (1), the spin connection of Kahler manifolds can
be thought of as the sum of an SU (n) and a U (1) connection. The latter is just
a standard U (1) connection, like in the case of an abelian gauge theory. Its field
strength Fij being non-zero characterises the holonomy not being restricted to
SU (n).
Concretely, recall that the complex structure is defined as multiplication by ‘i’
on the cotangent or tangent vector space. In components, the corresponding operator
or matrix is J α β , which is hence the generator of the U (1). The U (1) part of the
U (n) field strength encoded in Rij α β can hence be determined from the projection
on J α β . An explicit definition is
Fij ≡ tr[R̃ij J ] ≡ Rij α β J β α = iRij k k − iRij k k = 2iRij k k = −2iRi k j k = −2iRij .

(5.36)
Here the symbol R̃ is used to denote the Riemann tensor with suppressed second
index pair, as opposed to the Ricci tensor. The final manipulations leading to the
Ricci tensor require the use of the symmetry properties of the Riemann tensor
together with the Kahler property of our manifold. We leave that as a problem
(see e.g. [141]). Eventually, we see that the U (1) field strength components equal
those of the Ricci tensor up to a prefactor (note, however, the different symmetry
properties of the two tensors). Thus, SU (n) holonomy is equivalent to Ricci flatness.
A final important point concerns the definition of Calabi–Yau manifolds via
the Chern class (see e.g. [147–150]). Note first that, due to the U (n) holonomy
(or equivalently because of the special index structure of the Riemann tensor), the
tangent bundle of Kahler manifolds can be viewed as a complex vector bundle with
the curvature specified by Rij k l . In other words, one can consider the curvature
2-form
R(TX ) = dzi ∧ dzj Rij k l , (5.37)
which takes its values in Lie(U (n)). It is possible to write down the multi-form
c(X) = det(1 + R(TX )) , (5.38)
where the determinant refers to the matrix indices and multiplication relies on the
wedge product. It is then expanded according to
c(X) = 1 + c1 (X) + c2 (X) + · · · = 1 + tr R(TX )

+ tr R(TX ) ∧ R(TX ) − 2(tr R(TX )2 + · · · . (5.39)
Here ck (X) is a (2k)-form, defining the k th Chern class. Concretely, the 1st Chern
class is said to be zero if c1 is exact, which means that c1 = dω for some ω. More
formally, this means that c1 is zero in cohomology, a concept we will discuss next.
5.4 Homology and Cohomology 157
Crucially, while c1 was defined using the metric, it is invariant (up to exact pieces)
under smooth variations of the latter. It hence represents a topological invariant.
Intuitively, since c1 (X) corresponds to the U (1) field strength Fij introduced above,
one may think of a non-exact c1 characterising a non-trivial U (1) bundle associated
with Fij .
After these preliminaries, we can formulate the celebrated theorem by Yau: Let
X be a Kahler manifold and J its Kahler form. If the 1st Chern class vanishes,
then a Ricci flat metric with Kahler form J in the same cohomology class can
be given. This so-called Calabi–Yau metric is unique.
Being in the same cohomology class means that J − J is exact. The key point
is that, in practice, finding the Calabi–Yau metric is very hard (it has not been
achieved analytically in any example). However, checking the topological condition
of vanishing 1st Chern class is easy and guarantees the existence of many (explicitly
known) suitable complex manifolds on which we hence know that a Calabi–Yau
metric exists. But one will in general not find the metric for which c1 is zero as a
2-form, only one with c1 = dω.
5.4 Homology and Cohomology
We are overdue with developing a few more simple mathematical ideas concerning
in particular differential forms and topology. We start with homology and define a
p-chain as the formal sum of p-dimensional submanifolds Sp, i of some compact
manifold X:

cp = αi Sp, i . (5.40)
i
Depending on whether the coefficients αi are real, complex, integer etc., one can be
talking about homology over the real, complex or integer numbers. In the first two
cases, the p-chains form real and complex vector spaces, respectively.
One can consider the boundary of each Sp, i and hence of cp , which is a (p − 1)-
dimensional submanifold. Taking the boundary is denoted by the boundary operator
∂. A chain without boundary,
∂cp = 0 , (5.41)
is called a cycle. Crucially, ∂ 2 is zero, in other words, a boundary has itself no

boundary. A few simple examples are given in Fig. 5.1.
Given the linear operator ∂ with ∂ 2 = 0, it is natural to consider its homology
groups:
Ker(∂p ) p-cycles
Hp = = . (5.42)
Im(∂p+1 ) boundaries of (p + 1)-chains
Fig. 5.1 Some simple submanifolds and their boundaries
Fig. 5.2 Representatives of the four linearly independent homology classes in H1 (R2 ). The cycles
a and a are equivalent since their difference is a boundary. Concretely, the 1-cycle a−a represents
the boundary of the hatched 2-dimensional submanifold
The word group refers to addition, in the sense in which every vector space is an
abelian group. The index p of ∂p denotes the restriction of ∂ to the space of p-
chains. We will suppress this index when it is clear from the context on which
objects ∂ acts. As an example, we display certain 1-cycles on the genus-2 Riemann
surface R2 in Fig. 5.2. It is easy to convince oneself that, working over the real
numbers, H1 (R2 ) is 4-dimensional. Representatives a, b, c and d of the four linearly
independent homology classes (the elements of Hp ) are shown.
As another example, consider the 3-torus T 3 and convince yourself (at the
intuitive level) that dim(H0) = 1 (which corresponds to T 3 being connected),
dim(H1) = dim(H2) = 3 and dim(H3) = 1. If the torus is thought of as R3 modulo
discrete translations, representatives of H2 can be thought of as three planes, each
orthogonal to one of the three axes.
One calls the above simplicial homology.
Now we turn to p-forms as the dual objects with respect to the chains. So far, we
use the word ‘dual’ at an informal level, meaning simply that a chain cp and a form
ωp can be combined in a natural way to give a number:

ωp (cp ) = ωp = αi ωp . (5.43)
cp i Sp, i
On the space of forms, we have an operator analogous to ∂, which also squares to

zero: It is the exterior derivative d or, restricted to p-forms, dp :
dp : ωp → ωp+1 = dωp . (5.44)
Thus, it is natural to consider the cohomology groups of the de Rham cohomology:
Ker(dp ) closed p-forms

Hp = = . (5.45)
Im(dp−1 ) exact p-forms
In the last expression, we use the definition that a p-form ωp is called closed if
dωp = 0. Similarly, it is called exact if it can be written as ωp = dωp−1 .
It is easy to convince oneself that the pairing (5.43) between chains cp and forms
ωp , if restricted to cycles cp and closed forms ωp , induces a pairing between the
corresponding homology and cohomology classes. The latter are sometimes denoted
by [cp ] and [ωp ]. In other words, we claim that for a cycle cp and a closed form ωp
the integral

ωp (5.46)
cp
is independent of the representative. For example, one has

(ωp + dωp−1 ) = ωp + ωp−1 = ωp , (5.47)
cp cp ∂cp cp
since cp has no boundary. Analogously, replacing cp by cp + ∂cp+1 does not affect

the integral.
One can furthermore show that this pairing between homology and cohomology
classes is not degenerate and that hence
Hp (X) = H p (X)∗ , (5.48)
i.e. they are dual vector spaces (cf. de Rham’s theorems). In particular, their
dimensions coincide, defining the so-called Betti numbers
bp (X) = dim Hp (X) = dim H p (X) (5.49)
of the manifold X. Intuitively, they count the number of inequivalent p-cycles.

We note that, if dim X = n, there also exists a natural pairing between p-cycles
and (n − p)-cycles: the intersection number. For example, given T 3 , a 1-cycle (a
line) and a 2-cycle (a plane), one can find out whether the two intersect (intersection
number one) or do not (intersection number zero). For a Riemann surface, the
pairing is between a 1-cycle and a 1-cycle. Here the meaning of ‘intersection
number’ is obvious from Fig. 5.2. It is intuitively clear that this lifts to a paring
between homology classes.
The analogue of this on the cohomology side is

[ωp ] · [ωn−p ] = ωp ∧ ωn−p . (5.50)
This pairing is also non-degenerate and hence turns H p into the dual of the vector
space H n−p . But since we already know that Hn−p is the dual of H n−p , we have
found a canonical isomorphism
H p (X) ∼
= Hn−p (X) . (5.51)
This is known as Poincaré duality. To say this more explicitly, a p-form ωp is

Poincare dual to an (n − p)-cycle cn−p if

ωn−p = ωp ∧ ωn−p , ∀ωn−p . (5.52)
cn−p
More structure arises if a metric is present. In particular, with a metric comes the
Hodge star operator,
√
g μ1 ···μp
∗ : ωp → (∗ω)n−p with (∗ω)μp+1 ···μn = ω
μ1 ···μn .
p!
(5.53)
This gives rise to a scalar product on the space of p-forms,

(ωp , αp ) = ωp ∧ ∗αp . (5.54)
X
As a result, one can define the adjoint of d, the so-called co-differential d † . On forms
of degree p, it takes the explicit form
d † = (−1)p ∗−1 d ∗ . (5.55)
With this, one defines the Laplace operator
= d †d + d d † . (5.56)
A form is called harmonic if ω = 0. This definition gives rise to the Hodge

decomposition theorem, which states that on a compact manifold X any form has
a unique decomposition in an exact, a coexact and a harmonic piece:
ω = dα + d † β + γ with γ = 0 . (5.57)
It can furthermore be shown that β vanishes if ω is closed. As a result, any

representative of a given cohomology class has a unique decomposition in an
exact and harmonic piece. In other words, there is a unique harmonic form in any
cohomology class. Intuitively speaking, this is the constant form with the right
integral on all cycles (these integrals being fixed by the class). To give a simple
concrete example, consider T 2 being parameterised by (x, y) ∈ [(0, 1) × (0, 1)].
The harmonic one form with integral zero on the x-cycle and integral 1 on the y-
cycle is obviously given by ω = dy. A non-harmonic form in the same class would
e.g. be ω = (1 + sin(2πy)) dy.
Finally, it is possible to take the above to the realm of complex manifolds. To do
so, recall that on a complex manifold a 1-form may be written as
ω(z, z) = ω(z, z)i dzi + ω(z, z)ı dzı ≡ ω(1,0) + ω(0,1) . (5.58)
In other words, we can decompose it into its (1, 0) and (0, 1) parts. The first
corresponds to a tensor with one holomorphic and no antiholomorphic index, and
the second to a tensor with no holomorphic and one antiholomorphic index.
Such a decomposition carries over to higher forms (i.e. antisymmetric tensors)
and to cohomology classes. For example,
ω3 = ω(3,0) + ω(2,1) + ω(1,2) + ω(0,3) , (5.59)
where, e.g.
ω(2,1) = ωij k dzi ∧ dzj ∧ dzk + ωij k dzi ∧ dzj ∧ dzk + ωıj k dzı ∧ dzj ∧ dzk
= 3 ωij k dzi ∧ dzj ∧ dzk . (5.60)
To see the corresponding, refined cohomology construction more explicitly,

recall that the exterior derivative has the particularly compact definition
∂
d = dx a . (5.61)
∂x a
Here the partial derivative is supposed to act on the coefficients of any given form,
and subsequently, dx a has to be multiplied with the form using the wedge product
from the left. Let us consider specifically a manifold of complex dimension n (real
dimension 2n), such that a = 1, 2, · · · , 2n. Then it is easy to check that
∂ ∂
d = dzi i
+ dzı ı , (5.62)
∂z ∂z
or
∂ ∂
d =∂ +∂ with ∂ = dzi and ∂ = dzı . (5.63)
∂zi ∂zı
Here i = 1, 2 · · · , n. Furthermore, the holomorphic and antiholomorphic

exterior derivatives square to zero:
2
∂2 = ∂ = 0 . (5.64)
This permits the construction of a cohomology, which turns out to be independent

of whether ∂ or ∂ is used. The conventional choice is ∂. Thus, one defines the
Dolbeault cohomology
Ker(∂ p,q )
H p,q = , (5.65)
Im(∂ p,q−1 )
which contains finer information than the de Rham cohomology. One may say that
it characterises the interrelation between the non-trivial cycles and the complex
structure. We also note the so-called Hodge decomposition
H k = ⊕p+q=k H p,q . (5.66)
The dimensions of Dolbeault cohomology groups are known as Hodge numbers,
hp,q (X) ≡ dim H p,q (X) . (5.67)
They are commonly arranged in a so-called Hodge diamond. With a view to our
application to Calabi–Yau manifolds, we display the general form for the case of a
complex 3-fold:
h0,0
h1,0 h0,1
h2,0 h1,1 h0,2
h3,0 h2,1 h1,2 h0,3 . (5.68)
h3,1 h2,2 h1,3
h3,2 h2,3
h3,3
5.5 Calabi–Yau Moduli Spaces 163
5.5 Calabi–Yau Moduli Spaces
Due to SU (3) holonomy, the hodge diamond for a Calabi–Yau 3-fold is very special.
Using the same arrangement as in (5.68), it reads
1
0 0
0 h1,1 0
1 h2,1 h2,1 1 . (5.69)
0 h1,1 0
0 0
1
Here, the simplifications arising from the vertical and horizontal reflection sym-
metry of the Hodge diamond (e.g. h1,1 = h2,2 ) are generic—they hold for any
complex n-fold. Furthermore, connectedness implies h0,0 = h3,3 = 1. But some
features are specific to Calabi–Yau spaces, such as h1,0 = h2,0 = 0 and, crucially,
h3,0 = h0,3 = 1. The latter implies the existence of a unique holomorphic, harmonic
3-form, which is conventionally denoted by :
= ij k (z) dzi ∧ dzj ∧ dzk . (5.70)
Its existence can be understood on the basis of the covariantly constant spinor ψ:
ij k ∼ ψij k ψ . (5.71)
We will not argue for uniqueness. It is, however, useful to note that the existence of
a harmonic, holomorphic 3-form can be used as a defining feature for Calabi–Yau
spaces: More generally, a Calabi–Yau n-fold can be defined as a Kahler manifold
with a trivial canonical bundle. The latter is the nth exterior power of the cotangent
bundle—this is the bundle in which lives and which is trivial exactly if there is a
nowhere vanishing section—in our case the n-form .
Now, given a Calabi–Yau 3-fold, Yau’s theorem guarantees the existence of a
unique (given Kahler class and complex structure) Ricci flat metric gij . A key
question for physics is whether this metric can be deformed maintaining Ricci
flatness since this would imply the existence of moduli:
gij dzi dzj → gij dzi dzj + δgij dzi dzj + δgij dzi dzj + h.c. (5.72)
The presence of such deformations has the potential to contradict the uniqueness
part of Yau’s theorem. To avoid such a contradiction, these deformations must be
accompanied by a change of either the Kahler class or the complex structure. This
is indeed the case: A change of the metric of type δgij can be directly interpreted
as a change of (the harmonic representative of) the Kahler form J . The number of
such independent deformations, also called Kahler deformations, is counted by

h1,1 . This number is at least unity since it is always possible to simply rescale the
metric, making our manifold larger or smaller without changing its shape.
By contrast, a deformation of type δgij violates the hermiticity assumption, and
it must hence be accompanied by a change of the complex structure if one wants to
restore explicitly the Calabi–Yau situation after adding this δg to the original metric.
To count these deformations, it is useful to define a (2, 1) form
δχ = ij k δgkl dzi ∧ dzj ∧ dzl ∈ H 2,1(X) (5.73)
associated with δgkl . Here the index k of ij k has been raised using the Calabi–
Yau metric. It can be shown that this represents a one-to-one map between distinct
complex structure deformations (and hence corresponding metric deformations)
and linearly independent Dolbeault cohomology classes of type (2, 1). Here by
distinct we mean those not corresponding to reparameterisations zi → z i .
There is another way of understanding the counting of complex structure
deformations: Think of the complexified vector space of 3-cycles, with dimension
2h2,1 + 2. Two directions are distinguished by and , a feature only visible in
Dolbeault but not in de Rham cohomology. Now, the change of complex structure
is accompanied by a change of the direction of (and hence of ) in this
space. In other words, is infinitesimally rotated, and these possible rotations are
parameterized by h2,1 complex numbers. One may say that there are h2,1 complex
directions in which can develop new, infinitesimal components.
One can also invert the equations above, i.e. given a harmonic (2, 1)-form δχ,
one can explicitly write down how and the metric change:
1
δgıj = − ı kl δχklj , δ = δχ , (5.74)
||||2
with the constant

1 ij k
||||2 = ij k . (5.75)
3!
Together with the previously discussed relation
δgij = −iδJij , (5.76)
we now see explicitly how the cohomology groups H 1,1(X) and H 2,1 play a central
role in describing allowed deformations of the metric and hence the moduli space of
a Calabi–Yau manifold. Crucially, H 2,1 has to be viewed as a subspace of H 3 (X).
This subspace moves as the complex structure changes. An illustration of this has
been attempted in Fig. 5.3. In addition to the textbook literature given earlier, the
reader may want to consult [151, 152] for more details.
5.5 Calabi–Yau Moduli Spaces 165
Fig. 5.3 A visualisation attempt of how J and move in the spaces H 2 (X) and (the complex-
ification of) H 3 (X), thereby determining the metric on a Calabi–Yau manifold. Of course, the
dimensions of these spaces are in general much higher than three
Before characterising Calabi–Yau moduli spaces more quantitatively, we want to

give at least the simplest example. To do so, let us start with an important set of
examples for compact, complex Kahler manifolds: the so-called complex projective
spaces. To begin, recall that a real projective space RP n is Rn+1 \ 0 modulo the
equivalence relation x ∼ λx with λ ∈ R \ 0. Intuitively speaking, this is the set of
lines through the origin. Such a set can easily be given a differentiable structure. For
the case of RP 2 , the real projective plane, we can equivalently think of S 2 /Z2 —a
sphere with antipodal points identified.
This has a natural complex generalisation: the complex projective spaces CP n .
They are defined analogously as the set of all (n + 1)-tuples of complex numbers
(not all zero) with the equivalence relation
(z0 , · · · , zn ) ∼ (λz0 , · · · , λzn ) with λ ∈ C\0. (5.77)
For the subset Ui of all equivalence classes in which zi = 0, a chart is provided by

z0 zi−1 zi+1 zn
φi : { class of (z , · · · , z ) } →
0 n
, · · · , , , · · · , ∈ Cn .
zi zi zi zi
(5.78)
It is easy to show that these charts form an atlas and to give explicitly the
(holomorphic) transition maps. A Kahler potential in Ui is provided by
⎛ ⎞
1 ⎝
n
K (i) (x) = ln 1 + |x j |2 ⎠ , with {x 1 , · · · , x n }
2
j =1

z0 zi−1 zi+1 zn
= , · · · , , , · · · , (5.79)
zi zi zi zi
the coordinates defined above. A straightforward calculation shows that this gives
rise to a globally defined Kahler form and metric, the Fubini–Study metric. To be
very concrete, it is easy to check that CP 1 is the Riemann sphere. Crucially, all CP n
are compact.
Quite generally, submanifolds of lower dimension can often be given as zero sets
of polynomials. For example, the polynomial x 2 + y 2 − 1 defines S 1 ⊂ R2 . The
naive generalisation to (holomorphic) polynomials on Cn is not useful for us since
the resulting submanifolds are always non-compact (for n > 1). This is due to a
generalisation of the maximum modulus theorem for analytic functions. However,
starting from the compact space CP n , compact submanifolds can be defined by
polynomials. For the zero set to be well defined on the set of equivalence classes, the
polynomials have to be homogeneous. Now, it can be shown that the crucial Calabi–
Yau condition, the vanishing of the 1st Chern class, depends on the homogeneity
degree of the polynomial. If we want to get a 3-fold, we must start from CP 4 . The
Chern class vanishes if and only if the defining polynomial is of degree 5:
P5 (z) = ci1 ···i5 zi1 · · · zi5 , (5.80)
with indices running from 0 to 4 and labelling the projective coordinates of CP 4 .

The so-called quintic Calabi–Yau 3-fold is then defined as the zero set,
P5 (z) = 0 , z ∈ CP 4 . (5.81)
Here by z we mean both the set of 5 numbers {zi } and the corresponding point in the
projective space. As the coefficients of the polynomial vary, the complex structure
changes. A concrete example is given, e.g. by
P5 (z) = (z0 )5 + · · · + (z4 )5 . (5.82)
It is interesting to count the possible deformations of such a quintic hypersurface:

One first notes that the number of different monomials in a homogeneous polyno-
mial of degree d in n variables is given by the binomial coefficient3

d +n−1
. (5.83)
n−1
In our case this gives

5+5−1 9
= = 126 . (5.84)
5−1 4
3 Try to prove this!

5.6 Explicit Parameterisation of Calabi–Yau Moduli Spaces 167
From this, we have to subtract the 25 parameters of the symmetry group GL(5, C) of
CP 4 , giving us 101 parameters. Recalling what was said before about the interplay
of Dolbeault cohomology and complex structure moduli spaces, we conclude that
h2,1 (quintic) = 101. Without derivation, we also note that the Kahler form of CP 4
is unique up to scaling, such that h1,1 = 1. Thus, for the quintic the Hodge diamond
reads
1
0 0
0 1 0
1 101 101 1 (5.85)
0 1 0
0 0
1
and the real dimension of the moduli space is 2 · 101 + 1 = 203. (In fact, in string
theory the volume modulus always comes with an axionic partner, such that the
counting would be 2 · 101 + 2 = 204.)
We note that the same construction goes through for the quartic polynomial
in CP 3 , giving rise to the unique Calabi–Yau 2-fold, known as the K3-surface.
However, for 3-folds there are many more examples. First, one can generalise to
the intersection of hypersurfaces (defined by polynomials) in products of projective
spaces. This gives rise to the so-called complete-intersection CYs or CICYs. Then
one can generalise from projective spaces to weighted projective spaces. In this
case one still mods out by a rescaling with a complex parameter λ, but the different
variables scale differently, i.e. have different weights. Furthermore, one may mod
out not just by the rescaling by one such complex parameter, but by several such
scalings (with different parameters λi ). This leads to the concept of toric geometry
and toric hypersurfaces, in which Calabi–Yaus can again be defined by polynomials
of suitable degrees in the different variables (Batyrev’s construction [153]). See
e.g. [154] for a set of lecture notes starting at an elementary level and proceeding to
toric geometry. Even more general Calabi–Yau constructions exist. The total number
of known distinct examples is about half a billion: ∼ 5 × 108.
5.6 Explicit Parameterisation of Calabi–Yau Moduli Spaces
We start with an extremely simple toy model: T 2 . We can give it a complex structure
by defining it as C/Z2 . By this, we mean starting from the complex plane and
modding out a lattice of translations, generated by unity and τ ∈ C. The resulting
set of independent points, the so-called fundamental domain, is shown in Fig. 5.4.
It is parameterised, on the one hand, by z and, on the other hand, by x, y ∈ [0, 1),
Fig. 5.4 Torus defined as

C/Z2
with the relation
z = x + τy . (5.86)
The complex number τ determines the complex structure. Note that tori with
different τ are (in general) not isomorphic as complex manifolds. The holomorphic
(1, 0)-form in this case is clearly
= α dz = α dx + α τ dy , (5.87)
with α ∈ C an arbitrary constant.

Now, in analogy to the proper Calabi–Yau case, the complex structure can be
defined using the position of in the complexification of H 1 (T 2 ). For this, it is
sufficient to know the periods, i.e. the integrals of over the integral 1-cycles:
1 1
1 = = α dx = α , 2 = = α τ dy = ατ .
y= const. 0 x= const. 0
(5.88)
They can be combined in the period vector = (1 , 2 ). Since the normalisation
of is arbitrary, only ratios of these periods are meaningful. Concretely, the (in this
case single) complex structure parameter is given by τ = 2 /1 .
Next, we come to the moduli (in this case the modulus) associated with the
Kahler form. The Kahler form is harmonic and can be decomposed in a basis of
harmonic 2-forms,
J = t i ωi . (5.89)
Here the ωi are in general chosen to represent an integral 2-form basis (where by
integral we mean Poincare dual to the naturally defined integral basis of 4-cycles or,
what is the same, the dual basis to the integral 2-cycle basis). In our case there is of
course only one such 2-form:
ω1 = dx ∧ dy , such that J = t dx ∧ dy . (5.90)

At the same time, we know that
J = igij dzi ∧ dzj = igzz dz ∧ dz = igzz (dx ∧ τ dy + τ dy ∧ dx)

= −i(τ − τ ) gzz dx ∧ dy . (5.91)
Hence, we identify t as t = −i(τ − τ ) gzz . We may also write the general metric as

ds 2 = 2gzz dz dz = 2gzz dx 2 + |τ |2 dy 2 + (τ + τ ) dx dy . (5.92)
Thus, we can finally explicitly give the matrix form of the metric in terms of the
parameters t, 1 , 2 , which govern the position of J and in their respective
cohomology groups:

1 Re τ t 1 Re(2 /1 )
gab = 2gzz = .
Re τ |τ |2 Im(2 /1 ) Re(2 /1 ) |2 /1 |2
(5.93)
With somewhat more writing, one can achieve the same level of explicitness for the
toy model 3-fold T 2 × T 2 × T 2 , defined by modding out an appropriate lattice of
translations from C3 . Nevertheless, this is not a proper Calabi–Yau manifold since
its holonomy group is trivial. By contrast, a Calabi–Yau should have holonomy
group SU (3) (not just a subgroup). However, this is clearly to some extent a matter
of semantics. More importantly, T 6 is too simple for most physical applications, and
it does not give rise to the large landscape of solutions of string theory that we are
after.
Thus, we now turn to the general case of proper Calabi–Yau 3-folds, such as the
quintic and similar, even more complicated examples. The complete explicitness of
metric parameterisation that we saw above can of course not be achieved in such
cases. But our main goal for the moment will be a description in 4d supergravity
language,
j
L = Kij (∂Xi )(∂X ) + gauge, fermion, and other fields , (5.94)
where K is the Calabi–Yau metric on moduli space, parameterised by the Xi ,

which include both Kahler and complex structure moduli. This can be given rather
explicitly, even in the proper Calabi–Yau case. In principle, all that is needed is a
careful Kaluza–Klein reduction using the cycle structure of the Calabi–Yau space
(see e.g. [155–158] and refs. therein). We will only report the results.
Let us start with the Kahler moduli. As we already explained,
J = t α ωα with α = 1, · · · , h1,1 . (5.95)

Moreover, the volume of the Calabi–Yau manifold can be given as

1 1
V= J ∧J ∧J = καβγ t α t β t γ . (5.96)
6 X 6
Here one may intuitively think of components of the vector t α as measuring the
volumes of the different 2-cycles present in the Calabi–Yau. The integers καβγ are
the so-called triple intersection numbers of the 4-cycles Poincare dual to the ωα .4
The volumes of the dual 4-cycles, which are also labelled by the index α, are given
by

1 1
τα = J ∧J = καβγ t β t γ . (5.97)
2 c4α 2
Clearly, the variables t α and τα encode the same information. Using them as N = 1
SUGRA variables corresponds to choosing either of two different N = 1 sub-
algebras of the N = 2 SUSY of a Calabi–Yau compactification of type IIA or IIB
string theory. We focus (for the purpose of our later discussion of a particularly
well-understood model, called KKLT) on the IIB case and the τ variables. They are
real but, in 4d SUSY, are complexified by adding the imaginary parts

cα = C4 . (5.98)
cα
Only as a side remark, we note that, in the other SUSY, the t α would be complexified
by corresponding integrals of B2 or C2 , depending on the particular model (for many
more details, see e.g. [155–158]).
The relations (5.97) can in principle be solved for the t α :
t α = t α (τ1 , · · · , τh1,1 ) . (5.99)
With Tα = τα + icα and
1
τα = (Tα + T α ) , (5.100)
2
the volume V can be expressed as real function of the variables Tα and T α . The type
IIB Kahler moduli Kahler potential can then finally be written down as
KK = −2 ln V with V = V(Tα , T α ) . (5.101)
4 Note that two 4-cycles in a 6d manifold generically intersect in a 2d submanifold. The latter
generically intersects the third 4-cycle in points. The total number of those, with orientation, is a
function of the homology classes and is counted by the καβγ .
To describe the complex structure moduli space, we start by recalling the basis
of H1 (R2 ) as given in Fig. 5.2. We rename the relevant cycles (representatives of the
corresponding cohomology classes) as
a → A1 , b → B1 , c → A2 , d → B2 . (5.102)
It is clear that this carries over analogously to the 1-cycles of higher Riemann
surfaces, giving rise to the basis {Aa , Ba } and an intersection structure
Aa · Ab = 0 , Ba · Bb = 0 , Aa · Bb = δ a b . (5.103)
An analogous basis can be chosen for the (in this case naturally isomorphic)
vector space H1 . Such bases are called symplectic bases, on account of the
antisymmetry of the only non-vanishing intersection numbers or, on the form side,
wedge products:

ωaA ∧ ωBb = δa b = − ωBb ∧ ωaA . (5.104)
The crucial point for us is that this represents a generic feature of the so-called
middle homology or cohomology for manifolds where the dimensionalities of the
relevant cycles/forms are odd. This is true for Riemann surfaces, with which we
started, but it is equally true for complex 3-folds, our new case of interest.
Thus, now in the context of Calabi–Yaus, we choose a symplectic 3-cycle basis
as above and define the periods

za = , Gb = . (5.105)
Aa Bb
The complex parameters za with a = 0, · · · , h2,1 are sufficient to fully parameterise

the position of in H 3 (X). In fact, one of the parameters can be set to unity at the
expense of a constant, complex rescaling of , which does not induce any physical
(geometrical) change. Hence one may think of all the za s together as of ‘projective
coordinates’. Alternatively, one can set z0 = 1, with h2,1 parameters left over.
Crucially, the remaining periods Gb are not independent—they are in general
complicated functions of the za :
2,1
Gb = Gb (z0 , · · · , zh ) . (5.106)
One combines all of them in the period vector

2,1
= (z0 , · · · , zh , G0 (z), · · · , Gh2,1 (z)) . (5.107)
The explicit form of the (dependent) periods can be obtained from appropriate
differential equations (the Picard–Fuchs equations), which can be formulated on
the basis of certain topological features of the Calabi–Yau (see e.g. [159, 160]
and refs. therein). Crucially, they do not require the in general unavailable metric
information. Thus, though with much work, the periods can in principle be explicitly
obtained.
With this, we are ready to give the complex structure Kahler potential:

Kcs = − ln(i ∧ ) = − ln(−i† ) = − ln(−iza Ga (z) + iza Ga (z)) ,
X
(5.108)
where

01
= (5.109)
−1 0
is the symplectic metric. See e.g. [161] for a nice summary and explanation of these
and other related formulae.
Finally, one non-geometric modulus related to the dilaton is generally present. It
is known as the axio-dilaton (on account of the periodic scalar C0 ):
i
S = C0 + ie−φ = C0 + . (5.110)
gs
With this, the full type IIB moduli Kahler potential (corresponding to a so-called
orientifold projection with O3/O7-planes—the projection to N = 1 mentioned
earlier) reads
α
K = KK (T α , T ) + Kcs (za , za ) − ln(−i(S − S)) . (5.111)
This defines a ‘ready-to-use’ 4d supergravity model, so far without any scalar

potential. The conventions are such that MP ,4 = 1, as usual in supergravity
lagrangians, and that fields measuring distances or volumes in compact √space (in
our case the T ’s) are doing so in string units, i.e. powers of ls ≡ 2π α . Note,
however, that the 10d metric underlying our definition of the Kahler moduli is the
10d Einstein, not the string-frame metric. Thus, 4-cycle volumes in units of ls4 are
given explicitly by gs τα . Apart from this subtlety, one may roughly say that the
above is valid with ls = 1 concerning the Calabi–Yau geometry and MP ,4 = 1
concerning 4d physics.
5.7 An Aside on String Model Building: From Heterotic Compactifications to. . . 173
5.7 An Aside on String Model Building: From Heterotic

Compactifications to Orientifold Models with Branes
and F-Theory
This section serves two purposes: First, to give a very rough and entirely non-
technical overview of particle-physics model building in string theory. In other
words, we want to discuss briefly various approaches to engineering gauge group
and matter content of the Standard Model in a string compactification. But second,
as part of this discussion we will introduce the important concept of orientifold pro-
jections and orientifold planes. While our analysis will still be largely non-technical,
this part of the section is more than just informative: A clear understanding of these
ideas will be needed later on.
In the previous section, we discussed Calabi–Yaus almost entirely geometrically
—no fields other than the metric played a significant role. This view was in part
biased towards the type IIA/type IIB framework, where the gauge fields crucial
for 4d particle physics come in only in a second step, through the addition of
branes. But, historically, once Calabi–Yaus were understood, a different perspective
dominated.
The first successful attempts at semi-realistic string model building appeared in
the context of heterotic compactifications. They were based on compactifying the
10d N = 1 heterotic theory (with gauge group E8 ×E8 or SO(32)) to 4d on Calabi–
Yaus or torus orbifolds (to be explained momentarily). In this approach, it is possible
to realise fairly straightforwardly 4d N = 1 SUSY EFTs very close to the MSSM.
The gauge symmetry breaking comes from higher-dimensional gauge field strengths
and/or Wilson lines. The 4d matter content comes from 10d gauginos, which are the
only charged fermions in this context. We completely ignore this whole ‘universe’
of string-theoretic model building opportunities in the present course (see [162–164]
for the foundational papers and [101, 165–173] for reviews and more recent work).
This is due to time limitations and technical complexity as well as because the issue
of moduli stabilisation and SUSY breaking is simpler to understand in the type II
context (see, however, [174]).
With the understanding that Dp-branes with even/odd p are a natural part of
type IIA/IIB 10d supergravity (see in particular [175]), a whole new world of
constructing standard model-like EFTs opened up. Indeed, we have already learned
that, in 10d, a Dp-brane stack represents a dynamical submanifold with a certain
tension on which a (p+1)-dimensional super-Yang–Mills theory lives. The highest
dimensionality is obviously that of D9-branes, where we have a 10d gauge field
and a 10d Majorana–Weyl gaugino. The number of supercharges (and hence of
bosonic and fermionic degrees of freedom) remains the same for all p. For p < 9,
one has only a (p +1)-dimensional gauge field, and 9−p adjoint scalars carrying
the remaining bosonic degrees of freedom. The fermionic degrees of freedom are
carried by an appropriate set of lower-dimensional spinors. For example, for p = 3
one is dealing with the famous, maximally supersymmetric 4d N = 4 super-Yang–
Mills theory with four 4d gauginos.
Thus, most naively and as was already briefly advertised in Fig. 3.7, one might
think of simply compactifying either type IIA or IIB string theory on a CY to 4d and
wrap any desired number and type of D-brane stacks on the various cycles. In this
way, one should be able to generate any desired gauge sector. This is almost true,
but a crucial complication arises in the form of the necessary orientifold projections
and orientifold planes, as we now explain.
Let us pick as our first toy model a type IIB compactification from 10d to 8d
on T 2 . Then, add a D7-brane that fills out the eight non-compact dimensions and
represents a point in the T 2 . One might have hoped that this last step will only
slightly modify the previously well-defined 8d model by adding a U (1) gauge
theory. However, it turns out that the whole construction becomes inconsistent for
the following simple reason: Our brane couples to C8 , with dC8 = F9 , which
possesses a standard dual description in terms of F1 = dC0 . Thus, from the point
of view of the compact T 2 , our brane represents a point carrying an ‘electric’ gauge
charge, detected by
.
F1 = 2π . (5.112)
C1
Here C1 is a 1-cycle encircling the D7-brane. This is completely analogous to

the charge of an electron being measured by an integral of the dual, magnetic
field strength F̃2 = ∗F2 over an S 2 enclosing the electron. Thus, just like in
electrodynamics, (5.112) makes it impossible to have a non-zero charge in a
compact space: Indeed, C1 is the boundary of what is left of the torus after the disc
containing the D7 has been removed. But this 2d manifold contains no charge in
our construction, and hence, it is inconsistent that the integral of F1 = dC0 integral
over its boundary is non-zero.
The way out must build on the fact that, in a consistent compactification, the
total charge on the compact space must vanish. The most naive option of including
a D7 anti-brane is problematic since the two branes attract each other and will
eventually annihilate. Fortunately, a better route is offered by the possible presence
of the so-called orientifold planes. These are objects that, in spite of their opposite
RR-charge with respect to D-branes, are unable to annihilate them. Moreover, in
contrast to anti-branes, these O-planes break supersymmetry consistently with the
D-branes. In other words, a D-brane and a parallel O-plane break supersymmetry in
the same way as the D-brane would by itself: to half of the flat-space SUSY of the
10d theory.
To understand orientifold planes or orientifold singularities, let us first briefly
discuss the simpler, related concept of an orbifold. For this purpose, let us view
the T 2 in the above construction as C/Z2 , as explained in Sect. 5.6. Let us further
mod out the discrete group Z2 acting as z → −z, i.e. a π-rotation in the compact
space. It is easy to convince oneself that, under this transformation, four points
of the T 2 remain invariant and that the resulting quotient space T 2 /Z2 will have
four conical singularities at these loci (the so-called fixed points of the orbifolding),
cf. Fig. 5.5. We note in passing that many more options for orbifolding exist (e.g.
Fig. 5.5 Intuitive illustration of how in the procedure of modding out a Z2 symmetry a torus is
transformed into a space of half the volume and with four conical singularities at the points that are
left invariant by the Z2 action. This resulting space has the topology of a sphere but the geometry
of a ‘pillowcase’, i.e. it is flat everywhere except at the four singularities
T d /Zn ) with various d and n). In many cases, supersymmetry is partially preserved
and string theory continues to be well defined on such spaces in spite of the orbifold
singularities.
The key novelty now arises if one does not just mod out a geometric action (in our
case rotation) but combines it with worldsheet parity (for reviews and introductions,
see e.g. [108, 176]): In other words, one restricts the states of the original T 2 model
to those invariant under π-rotation together with an orientation change of the string
worldsheet.5 As a result, the fixed-point loci of the geometric Z2 action become
charged under RR and NS fields—they are the orientifold planes. Explicitly, in the
concrete case at hand each of the four singular points on the r.h. side of Fig. 5.5
(corresponding to 8d hyperplanes in the full 10d geometry) develops a C8 charge
equivalent to 4 D7 branes and a corresponding negative tension. As a result, type IIB
theory on T 2 /Z2 , where the Z2 acts as an orientifold projection, is only consistent
if 16 spacetime filling D7 branes are added. Moreover, if these are located in groups
of four at the four orientifold planes, no deformation of the flat 10d geometry arises
since neither RR nor NS fields are sourced. One says that ‘all tadpoles are cancelled
locally’.
The gauge group at each O7-plane is not, as one might have thought, U (4) but
rather SO(8). This can be understood by first placing twice the required number
of D7 branes on the T 2 in a Z2 -symmetric fashion and then modding out. In this
procedure, the branes are pairwise identified. It can be shown that, for eight branes at
one fixed point with original gauge group U (8), the appropriate projection reduced
the gauge symmetry to SO(8) = U (8)/Z2 . However, even after the projection the
D7-branes may be displaced from the O7-planes and are, in fact, free to move around
in the T 2 at zero energy cost. This breaks the gauge group generically from SO(8)4
to U (1)16. Many different intermediate gauge groups are also possible depending
on which point in this ‘D-brane’ or ‘open-string’ moduli space one is located at.
With the basic concepts explained, we may now generalise. First, as we already
noted, an orientifold projection without an accompanying spacetime action is
possible: Applied to the 10d type IIB theory, it introduces a spacetime filling O9-
plane, which requires the presence of 16 D9-branes to cancel the RR tadpole. This
5 To
be precise, one has to include (−1)FL in the projection, with FL the left-moving fermion
number.
is nothing but a construction of the type I open string through orientifolding. The
gauge group is SO(32), analogously to the SO(8) above. Next, we may start with
type IIA on S 1 and mod out a Z2 reflection in the single, real compact dimension.
The new compact space is an interval with two O8-planes at the boundaries. Each
requires the presence of eight D8 branes, leading to an SO(16)2 gauge theory.
Analogously to the O7/D7 case, this is the situation with local tadpole cancellation.
If the D8 branes are moved, the gauge group changes.
Next, we may compactify type IIA to 7d on T 3 and mod out a total reflection,
y → −y i (for i = 1, 2, 3), of the compact space together with a worldsheet
i
orientation change. This leads to 23 = 8 singularities with a ‘solid angle deficit’,

at each of which an O6-plane is located. Each O6 requires 2 D6-branes to cancel its
tadpole. The gauge theory is SO(4)8 in this simplest configuration and, as before,
can change as the branes move. The reader may complete our discussion with O5-,
O4- and O3-planes in the obvious way.
Finally, coming closer to our goal of phenomenology, one may compactify
type IIA on T 6 and mod out various Z2 symmetries together with orientation
changes, leading to the presence of intersecting O6-planes and intersecting branes.
In particular, intersections of two D6-branes or brane stacks at points in the compact
space are possible. At such points (which are in fact 4d hyperplanes filling out
all of our non-compact dimensions), open-string states ending on the different
stacks are localised, cf. the last picture in Fig. 3.7. They correspond to 4d chiral
superfields, allowing for a particularly simple and intuitive road to particle-physics
model building. These constructions are known as intersecting brane models. They
have a geometrically more challenging counterpart in the Calabi–Yau context: Here,
the Calabi–Yau that one starts with has to possess an appropriate Z2 symmetry to
allow for an orientifold projection. The branes can only be wrapped on the so-called
lagrangian submanifolds to ensure that at least N = 1 supersymmetry is preserved.
An excellent source for specifically this type of constructions (and for more details
on all subjects of this section) is [9]. See also the reviews [111,112] and the original
papers [113–115, 177].
A similar story can be developed in the type IIB context. Here, the most
promising road is that of orientifold projections introducing O3- and O7-planes. As
before, they may act on T 6 or on an appropriate Calabi–Yau. As a crucial fact, we
note that O3/O7-planes and the corresponding branes may be introduced together
while still preserving some amount of supersymmetry. Similarly, the combination
O5/O9 can be supersymmetric but is phenomenologically less interesting. By
contrast, combinations of the type Op/O(p +2), for example O5/O7, break SUSY
completely leading in general to control issues.
The gauge group of O3/O7 orientifold models comes from D3- and D7-brane
stacks. Chiral matter can live, for example, at intersections of two D7-brane stacks.
Each of those represents locally a holomorphic submanifold (a divisor), and their
intersection is in general a complex curve. Yukawa couplings can arise at loci
where such curves meet. See e.g. [178] for explicit GUT models constructed in
this approach. The whole setting of such type IIB O3/O7 Calabi–Yau or torus
orientifolds is T-dual to an appropriate type IIA model with O6-planes. To see this,
it is crucial to note that T-duality also works with branes: If a brane on the IIA side
wraps a compact S 1 , the corresponding brane on the IIB side does not wrap the
dual S 1 (and vice versa). This is consistent with the change of the dimensionality of
D-branes between IIA and IIB. It clearly allows for chains of three T-dualities that
relate certain T 6 -based models in type IIA with D6-branes to type IIB models with
D3/D7 branes.
It is even more interesting that such a duality also exists between type IIA models
on a Calabi–Yau orientifold with O6-planes and type IIB models on a Calabi–Yau
orientifold with O3-/O7-planes. For this, note first that Calabi–Yaus can in general
be viewed as T 3 fibrations (with the T 3 degenerating at various loci) over S 3 . T-
dualising the 3-torus takes one from a type-IIA model on one Calabi-Yau to a
type-IIB model on another Calabi–Yau [179]. This second Calabi-Yau is referred
to as being mirror-dual to the first one. The fact that Calabi–Yaus comes in pairs
related by mirror symmetry (which in particular exchanges h1,1 and h2,1 ) is an
important mathematical fact known independently and long before the Strominger–
Yau–Zaslow picture of ‘mirror symmetry as T-duality’ [179]. For details on mirror
symmetry, see e.g. [180–183] and refs. therein.
Finally, type IIB theory allows in its strong coupling regime (gs ∼ O(1)) for co-
dimension of two objects other than D7-branes and O7-planes. Compactifications
of this type, based on type IIB at strong coupling and including generic 7-branes
(especially with gauge groups other than SU (N) and SO(N), are known as F-
theory models [184] (see also [10, 185–187]). The various 7-branes of F-theory are
detected by the monodromy in which the axio-dilaton S = i/gs + C0 undergoes
if one encircles the brane. This is closely related to a discrete SL(2, Z) gauge
symmetry of the type IIB theory, which we, however, have no time to describe.
What is crucial for us is that this symmetry identifies certain values of S. As a
result, S does in fact not take values in the complex upper half plane as one might
naively have thought but only in the so-called fundamental domain of SL(2, Z).
Interpreting S as the complex structure parameter of a T 2 , this is precisely the region
in which it describes geometrically distinct tori (not related to each other by large
diffeomorphisms, which are in turn characterised by SL(2, Z)). As a result of all of
this, F-theory models can be characterised by T 2 fibrations (more precisely ‘elliptic
fibrations’) over a complex 3-dimensional manifold (not necessarily a Calabi–Yau).
The complex structure parameter of this ‘artificially introduced’ fibre torus specifies
how the variable S of the type IIB theory varies over the base. The monodromies
of S characterise submanifolds of complex co-dimension one (7-branes) on which
gauge theories are localised. It is quite remarkable that solutions of the type IIB
equations of motion in this setting arise precisely when the torus-fibration describes
a Calabi–Yau 4-fold.6
6 This, in turn, can be understood by starting with 11d supergravity and compactifying on this
Calabi–Yau 4-fold to 3d. One then shrinks the T 2 fibre to zero volume and uses one of its S 1 s to
go to type IIA supergravity. Subsequently, one appeals to T-duality on the other S 1 to go to type
IIB with one new non-compact dimension emerging (recall that R = α /R → ∞ at R → 0). This
takes one to a type IIB compactification to 4d, as desired.
The punchline is that F-theory models are arguably the most general and
powerful setting for string phenomenology, including through various limits and
dualities all that can be done in perturbative type IIB models with branes and much
of what can be done in the type IIA and heterotic context. It has proven a particularly
fruitful setting for constructing grand unified theories, especially because its 7-
branes allow for exceptional gauge groups and, through their breaking, for realistic
GUT models with the right Yukawa coupling structure. This has been explored
relatively recently under the name of ‘F-theory GUTs’ [188, 189] (see [186, 187]
for reviews). In addition, as will become clearer at the end of Chap. 6, F-theory
is particularly powerful in that it presumably generates the largest number of the
presently known landscape vacua.
5.8 Problems
5.8.1 Dimensional Reduction
Task Perform the KK reduction of the 5d lagrangian
L5 = i ∂/ −M (5.113)
to 4d on S 1 . Give your result in a compact, standard 4d notation as appropriate for

a theory with towers of Dirac fermions.
Now gauge the fermion in the lagrangian above, adding also a standard gauge-
kinetic term. Perform again the dimensional reduction, but disregard the higher
modes of the gauge field (to avoid dealing with towers of massive vectors, which is
interesting but not essential in our context). Show how the 4d scalar coming from A5
couples to 4d fermions. It appears naively that 5d gauge invariance, which should
be manifest as a discrete shift symmetry of A5 , is broken by such an interaction.
Resolve this puzzle!
Hints It is essential to use

−1 0
γ ≡i
5
, (5.114)
01
which differs by a prefactor from standard 4d conventions. This is clear since

otherwise the 5d Clifford algebra relations7
{γ M , γ N } = −ηMN (5.115)
7 Recall that we use the mostly plus convention.

5.8 Problems 179
would have an incorrect sign for the index choice (MN) = (55). The rest is a
straightforward analysis following the scalar case presented in the lecture. It is more
convenient to use exponentials rather than sines and cosines when dimensionally
reducing the 5d fields.
Solution Let us make the ansatz

+∞
+∞

(x, y) = ψnL (x) einy/R + ψnR (x) einy/R , (5.116)
n=−∞ n=−∞
where x ≡ {x μ } and the indices L/R denote left- and right-handed 4d fermions.
After a straightforward calculation, using in particular manipulations like
ψ n (x) e−iny/R iγ 5 ∂5 ψnR (x) einy/R = ψ n (x)(−in/R) ψnR (x) ,

L L
(5.117)
one arrives at

4 L R L
S = 2πR d x ψ 0 i ∂/ψ0L + ψ 0 i ∂/ψ0R − Mψ 0 ψ0R + h.c.
L R L
+ ψ n i ∂/ψnL + ψ n i ∂/ψnR − Mψ n ψnR + h.c.
n=0

L R
(−in/R)ψ n ψnR + (in/R)ψ n ψnL . (5.118)
We can absorb the volume factor in a field redefinition and write this as a tower of
pairs of l.h. and r.h. fermions,
+∞

L R L
S= d 4x ψ n i ∂/ψnL + ψ n i ∂/ψnR − Mn ψ n ψnR + h.c. (5.119)
n=−∞
with Dirac-type mass terms, but with complex mass parameters
Mn = M + in/R . (5.120)
Of course, the complex phases of the Mn can be absorbed in a phase rotation of,
for example, the right-handed parts. The mass parameters now become real, and the
two terms with Mn and M n can be combined in Dirac mass terms. Thus, we obtain
+∞

S= d 4x ψ n (i ∂/ − Mn )ψn (5.121)
n=−∞
with

Mn = M 2 + (n/R)2 . (5.122)
Introducing the gauging, one gets a 4d gauge theory and a real scalar coming
from A5 , as explained in the lecture. Crucially, one also finds a coupling of the
scalar to the fermions,
L
iγ 5 iA5 → −iφ ψ n ψnR + h.c. (5.123)
The fermions can again be rescaled to absorb the volume prefactor (2πR) of the
fermionic part of the action. If M n/R, it is natural to focus on the zero-mode
level of this Kaluza–Klein theory:

1 1 L
S= d 4x / ψ0L + ψ R
− 2 Fμν F μν − 2 (∂φ)2 + ψ 0 i D / ψ0R
0 iD
4g 2g

−ψ0L (M + iφ)ψ0R + h.c. . (5.124)
It is clear, however, that to make the apparently broken shift symmetry φ → φ+1/R
manifest, one needs to include higher fermion modes. Indeed, when the modulus φ
continuously changes its value from zero to 1/R, the mode with n = −1 takes the
place of the former zero mode. Thus, the model as a whole returns to a physically
equivalent situation, as it should be given that φ = 0 and φ = 1/R are related by a
gauge transformation.
5.8.2 SO(2n) vs. U(n)
Task In the lecture we used the fact that, if Rαβ γ δ is pure in the second index
pair, then the holonomy is reduced to U (n). (Here we use greek rather than latin
indices to symbolise that, e.g. α may stand for either i or ı.) This fact is in
principle obvious and does not require any demonstration. Still, to make the simple
underlying techniques more manifest, consider the following simple problem:
Let v α , wβ specify a vector pair such that v α wβ Rαβ γ δ ≡ Qγ δ describes an
infinitesimal SO(2n) rotation in the complex basis, corresponding to the appropriate
parallel transport along an infinitesimal loop. According to the pure index structure,
Q takes the form

M 0
Q= . (5.125)
0 N
In other words, Q is characterised by v α wβ Rαβ i j ≡ M i j and v α wβ Rαβ ı j ≡ N i j .

5.8 Problems 181
Which properties of M and N follow from the fact that Q corresponds to an

infinitesimal SO(2n) transformation?8
Hint Use the notation z i = M i j zj and z = x + iy such that

x
(5.126)
y
is the column vector transforming under SO(2n).
Solution The matrix Q characterises a linear transformation in the (z, z)-basis. To

translate this into the (x, y) basis, write
x i = (z i + z i )/2 = (M i j zj + N j j zj )/2
= (M i j x j + iM i j y j + N i j x j − iN i j y j )/2 (5.127)
y i = (z i − z i )/2i = (M i j zj − N j j zj )/2i
= (M i j x j + iM i j y j − N i j x j + iN i j y j )/2i. (5.128)
From this, the real-basis form Qr of the transformation Q is easily read off:

1 M + N i(M − N)
Qr = . (5.129)
2 −i(M − N) M + N
Our requirement Qr ∈ Lie(SO(2n)) implies that Qr is real antisymmetric, i.e.
M+N =M+N M T + N T = −M − N , (5.130)

M − N = −M + N MT − NT =M −N . (5.131)
Adding the first and third equations gives N = M. The other two equations imply
M T = −N. Thus,
N =M , M = −M † , (5.132)
and

M 0
Q= with M ∈ Lie(U (n)) . (5.133)
0 M
8 Of course, the transformation described by Q is in Lie(SO(2n)) by the very definition of R, such

that these properties could also be derived from elementary differential geometry. But we want a
purely algebraic derivation.
We see that Q does indeed describe an infinitesimal U (n) rotation in the complex
basis.
5.8.3 Complex Projective Spaces
Task Consider CP n with charts as defined in the lecture and obtain explicitly the
transition functions φi ◦ φj−1 . Give a general formula for the components gij of the
Fubini–Study metric in some chart φk . Show consistency between different charts.
In the special case of CP 1 ∼ = S 2 , show agreement with the round metric on the
sphere (up to normalisation).
Hints Deriving the transition functions is completely straightforward, but some

care is needed concerning the indexing of the variables in the two charts. Getting
the Fubini–Study metric in one chart requires just differentiation. To show that the
Fubini–Study metric is well-defined, it is useful to first investigate how the Kahler
potential transforms between coordinate patches. Try to make use of the (multi-
variable generalisation) of the fact that ∂z ∂ z ln(zz) = 0 for z = 0. If you get stuck
use, e.g. the lecture notes by Candelas [141] or the Wikipedia page for ‘Fubini–
Study metric’. In the last part, think of the stereographic projection.
Solution The two sets of local coordinates in φ i and φ j may be chosen as

z0 zi−1 zi+1 zn
(x 1 , · · · , x n ) = , · · · , , , · · · , (5.134)
zi zi zi zi
and

z0 zj −1 zj +1 zn
(y , · · · , y ) =
1 n
, · · · , , , · · · , . (5.135)
zj zj zj zj
The coordinate change is found by explicitly rewriting each of the x k in terms of the
y-coordinates. For definiteness, let us assume i < j . Then we find for k ≤ i:
zk−1 zk−1 zj 1
xk = i
= j · i = y k · i+1 . (5.136)
z z z y
For i + 1 ≤ k < j :
zk zk zj 1
xk = i
= j · i = y k+1 · i+1 . (5.137)
z z z y
5.8 Problems 183
Then comes a special case: For k = j ,
zj 1
xj = = i+1 . (5.138)
zi y
Finally, for j < k:
zk zk zj 1
xk = i
= j · i = y k · i+1 . (5.139)
z z z y
We may summarize all of this in the compact expression
1
(x 1 (y), · · · , x n (y)) = (y 1 , · · · , y i , y i+2 , · · · , y j , 1, y j +1 , · · · , y n ) .
y i+1
(5.140)
Obtaining the explicit form of the Fubini–Study metric is easy: Consecutive

differentiation w.r.t. x i and x j gives
(k) xı xı
2Ki = = with σ ≡ 1 + x l x l δll (5.141)
1 + x l x l δll σ
and
(k) δij xı xj
2gij ≡ 2Kij = − 2 . (5.142)
σ σ
Here summation over l and l is implicit in the first line.

To see invariance under coordinate change, recall that the two metrics in Ui and
Uj are defined as
∂ ∂ ∂ ∂
K (i) (x, x) and K (j ) (y, y) . (5.143)
∂x k ∂x l ∂y k ∂y l
This obviously defines two tensors that will by definition agree if
∂ ∂ ∂ ∂
k
K (j ) (y, y) = K (i) (x(y), x(y)) . (5.144)
∂y ∂y l ∂y k ∂y l
Note that here we also have to use the fact that the coordinate change is holomorphic,
such that holomorphic and antiholomorphic indices do not mix under reparameteri-
sation.
Now, Eq. (5.144) will clearly hold if
K (i) = K (j ) (y, y) + f (y) + f (y) . (5.145)
This is, in fact, known as a Kahler transformation—the natural way in which a

Kahler potential changes between patches on a Kahler manifold.
Showing that this holds is easy if one notes that
2K (i) (x, x) = ln σ (x, x) and 2K (j )(y, y) = ln σ (y, y) (5.146)
with σ as defined above. Moreover, let us think of a different way of labelling our
coordinates as follows:
x k ≡ zk /zi and y k ≡ zk /zj , (5.147)
such that k = 0, · · · , n, but with the caveat that x i = 1 and y j = 1, and hence these
two do not count as coordinates. In this notation, one has

n
n
σ (x, x) = |x |
k 2
and σ (y, y) = |y k |2 (5.148)
k=0 k=0
and
σ (x(y), x(y)) = σ (y, y) |zj /zi |2 = σ (y, y) |y j /y i |2 . (5.149)
Hence, we obtain the above form of a general Kahler transformation with f (y) =
ln(y j /y i ). This completes the demonstration that the metric is well-defined.
Finally, let us consider the specific case of CP 1 . In the patch U0 , we have
1 |x|2 1
2gx x = − = (5.150)
1 + |x|2 (1 + |x|2 )2 (1 + |x|2 )2
and, with x = r exp(iφ),
dr 2 + r 2 dφ 2
dsx2 = gxx dx dx + gxx dx dx = 2gxx dx dx = . (5.151)
(1 + r 2 )2
This has to be compared with the round metric on the unit sphere,
ds12 = dθ 2 + sin2 θ dφ 2 . (5.152)
Now imagine that this sphere is centred at the origin in R3 , and map it to the x-
y-plane using rays originating in the north pole and intersecting the plane and the
5.8 Problems 185
sphere (stereographic projection).9 Elementary geometry proves that the ray that
intersects the sphere at (θ, φ) will enclose an angle θ/2 with the negative vertical
axis. Hence, parameterising the plane by the complex variable x = r exp(iφ) as
above, we have r = tan(θ/2). Thus, 2dr/dθ = 1 + tan2 (θ/2) = 1 + r 2, which gives
dθ 2 tan2 (θ/2) dθ 2
dsx2 = + dφ 2
= + tan2 (θ/2) cos4 (θ/2)dφ 2
4 (1 + tan2 (θ/2))2 4
1
= (dθ 2 + sin2 θ dφ 2 ) (5.153)
4
and dsx2 = ds12 /4, as proposed. Our complex x-coordinate covers the sphere without
the north pole. The coordinate change x → 1/x takes us to the second coordinate
patch, which covers the sphere without the south pole.
9 Beware that an alternative form of the stereographic projection uses a unit sphere centred at
(0, 0, 1) ∈ R3 . This corresponds to scaling distances on the plane up by a factor of two.
The Flux Landscape
6
The general idea will be to consider compactifications with non-zero internal

components of the RR and NS field strength tensors Fp and H3 . This induces a non-
zero superpotential depending on the moduli of the supergravity models discussed
above and leads to moduli stabilisation. Moreover, the number of available distinct
models jumps from 108 to 10600 (or, in more general geometries—roughly speaking,
including D-branes—even much higher). A small fraction of them, which would still
be an enormous number, are expected to have broken supersymmetry and a positive
cosmological constant. If true, this implies a paradigm shift in fundamental physics
comparable to the Copernican revolution: Our fundamental physics parameters are
not fundamental at all but drawn from a large set of solutions—the string theory
landscape. It may even become unavoidable to invoke anthropic considerations
to come to terms with the implications [190]. We will now discuss this in detail
following the path of the Bousso–Polchinski model [191], the more realistic GKP
setting [133] and the KKLT [192] and LVS [193] proposals.1
6.1 Compact Geometries with p-Form Fluxes
Let us start with a few general comments on p-form gauge theories. Consider a
(p − 1)-form gauge theory in d dimensions, with an action of type (we disregard
purely numerical prefactors)

1
Fp ∧ ∗Fp + Ap−1 . (6.1)
g2 (p−2)-brane
1 See
[194] for a discussion of earlier references emphasising the non-uniqueness of string
compactifications and the resulting need for anthropic considerations.

188 6 The Flux Landscape
One can easily show that a dual description is provided by a theory based on the
(d − p)-form field strength F̃d−p . The latter is defined as
1
F̃d−p = ∗ Fp , (6.2)
g2
which in turn leads to the definition of a dual gauge potential via
F̃d−p = d Ãd−p−1 . (6.3)
In these new variables the action takes the form

1 1
F̃d−p ∧ ∗F̃d−p + Ãd−p−1 with g̃ = . (6.4)
g̃ 2 (d−p−2)−brane g
While the new kinetic term is just a rewriting of the old one, the charged objects
coupling to the dual potential are different. In fact, both types of charged objects are
present in the full theory. But the coupling of any one of them to the fields can only
be explicitly given on one side of the duality.2
The above is of course familiar from electrodynamics, where d = 4 and
p = 4 − p = 2, such that the tilde is really necessary to distinguish the otherwise
identical-looking dual descriptions. The charged objects on both sides are 0-branes,
i.e. particles.
Now let us consider the particularly simple case of F1 in d = 4, which is of
course nothing but a scalar (axion) field model, with A0 ≡ φ:

f 2 (∂φ)2 + φ(xi ) . (6.5)
The last term is the coupling to an instanton, a tunnelling event localised at a

point in spacetime. Clearly, such objects cannot be included in an initial field
configuration—one has to sum over them and integrate over all the xi in the path
integral.
In case this concept is unfamiliar, here is a brief excursion concerning instantons
(see [72,195–197] as well as many QFT textbooks). The ‘classical’ setting is in fact
that of a gauge theory coupled to a periodic pseudo-scalar or axion-like field or
2 The purely classical dualisation above can also be performed at the quantum level, i.e. under the
path integral. The basic idea is to implement the original Bianchi identity constraint dFp = 0 by a
Lagrange multiplier, i.e. by adding a term dFp ∧ Ad−p−1 to the action. Then one can integrate out
Fp , arriving at the dual action. In the latter, the Lagrange multiplier Ad−p−1 has become the new
dynamical field.
6.1 Compact Geometries with p-Form Fluxes 189
axion for short. For definiteness, say the gauge group is SU (2):
1 φ
L= 2
tr Fμν F μν + trF ∧ F . (6.6)
2g 8π 2
The term multiplying φ is a total derivative, but there exist field configurations
(which cannot be smoothly deformed to the vacuum) on which the integral gives
8π 2 n with n ∈ Z. Very roughly speaking, the existence of such a field configuration
is related to the fact that SU (2) ∼
= S 3 and the possibility of identifying this group-
theoretic S with the S of radial coordinates in R4 . In the euclidean path integral,
3 3
one has to sum over all such ‘bumps of energy density’ (to be interpreted as local
tunnelling events, leading from vacuum to vacuum). One also has to integrate over
all their sizes 1/M and positions. The events are suppressed by their action—
exp(−Si )—and for large Si one uses the ‘dilute gas approximation’ (cf. Fig. 6.1).
It should now be clear in which sense our model of (6.5) corresponds to instantons
of an SU (2) (more generally, SU (N)) gauge theory: The point at which the gauge
field-theoretic instanton is localised is identified with xi , and the F ∧ F term of the
lump of field strength is replaced by an approximate δ-function.
Still within our excursion about instantons, we recall that a model with a periodic
scalar like that of (6.5) can be derived by compactifying a 5d U (1) gauge theory to
4d. Interestingly, this also has instantons, but of a very different type (cf. Fig. 6.2).
We leave it as an exercise for the reader to derive the correct coupling of this type
of instanton to φ. This ends our instanton excursion.
As a side remark, the dual theory has field strength H3 = dB2 and couples to
strings (which are here unrelated to any fundamental string theory). What interests
us here is flux quantisation, which is particularly easy to understand in this case: If
our spacetime has non-trivial one-cycles, the gauge potential φ does not need to be
globally well-defined. Instead, assuming, e.g., that x3 parameterises an S 1 , it may
obey
φ(xi ) = φ(xi + 2πR) + 2πn , n ∈ Z. (6.7)
The shift must be an integer or else the exponential of the instanton coupling,
exp(iφ(xi )), would not be well-defined. Another way to formulate the same
Fig. 6.1 Instantons as

localised lumps of field
strength (figure adapted
from [198])
Fig. 6.2 Effective instanton arising from a particle–antiparticle fluctuation wrapping the compact
space of an S 1 compactification (figure adapted from [198])
condition is
.
F1 = 2πn . (6.8)
Here n is a discrete choice one has to make when defining the theory on a spacetime
with a non-trivial 1-cycle. An analogous non-trivial boundary condition would arise
if the 5d topology were trivial but the loop in (6.8) were wrapped around n strings.
The above is clearly analogous to the familiar statement
.
F2 = 2πn (6.9)
S2
for electrodynamics and an S 2 enclosing n magnetic monopoles. But this case is not
our interest at present. What we care about is flux quantisation,
.
Fp ∈ 2π Z , (6.10)
cp
which is simply a requirement of (quantum mechanical) consistency of a p-form

gauge theory (and its dual). We see that, in the absence of charges, the flux can only
be non-zero if a non-trivial p-cycle exists in the geometry. If so, it is determined by
a discrete choice one has to make for every such p-cycle.
Now let us compactify a 4d model with a 0-form gauge theory (an axion) to
3d on S 1 . The compact geometry has a single compact 1-cycle. This allows a
choice of boundary condition or, equivalently, 1-form flux on the S 1 . The freedom is
precisely that of choosing n ∈ Z in (6.7). Thus, one obtains an infinity of 3d models
with different vacuum energy: Indeed, from the perspective of the non-compact
dimensions {x 0 , x 1 , x 2 }, the gradient term (∂3 φ)2 contributes to the cosmological
constant. This already represents a small flux landscape. Moreover, the theory
possesses strings. Let us include an infinite string in our compactification. This
string is a point in the compact x 3 -direction and hence still has two dimensions—
one time and one spacelike—in the non-compact (2 + 1)-dimensional spacetime.
It is hence a domain wall in the non-compact 2d space. Once can convince oneself
that, on the two sides of this wall, the flux on the S 1 differs by one unit. Hence
6.2 Bousso–Polchinski Model 191
Fig. 6.3 Bubble nucleation

in a 4d-to-3d toy model with
1-form flux
our flux landscape is actually not just a collection of different theories, but it
possesses a dynamics allowing one to change between those: This dynamics is
bubble nucleation (cf. Fig. 6.3). The surfaces of the bubbles are the domain walls
made of the higher-dimensional charged objects. This crucial feature will survive in
the full-fledged string theory landscape.
Clearly, an analogous situation may be considered if one compactifies, for
example, a 6d gauge theory to 4d on S 2 . The S 2 may be given 2-form flux in the
sense of (6.9), giving rise to a 4d landscape of vacua labelled by n. In this case,
the flux quantisation is literally based on the same logic that forces the F2 integral
around a magnetic monopole to be quantised. One may also use the U (1) principle
bundle approach to gauge theories to think of this in terms of non-trivial fibrations
of U (1) over S 2 , which are known to be labelled by an integer, our flux number. The
case of unit flux corresponds to the famous Hopf fibration (see e.g. [147–150]).
6.2 Bousso–Polchinski Model
We have just understood that the compactification of higher-dimensional p-form

theories provides apparently rather general mechanism for creating 4d ‘landscapes’.
Now let us see how this is reflected in an effective field theory that works directly
in 4d [191] (see also [199]). Such an effective description arises naturally if
we consider the (somewhat special) case of a (d − 1)-form gauge theory in d
dimensions. Our specific interest is of course in 3-form gauge theories in d = 4:

1 2
S=− F + A3 . (6.11)
4 4 domain wall
Without sources, the equation of motion d ∗ F4 = 0 implies that F4 is constant, so

there are no propagating degrees of freedom. The only dynamics is that of domain
walls, which have some tension and hence move according to their own classical
dynamics. Moreover, they couple to A3 and hence source F4 .
With the domain wall comes a 1-form current, appearing in
1
d ∗ F4 = j1 , (6.12)
4
which is localised at the wall. As is generally the case in p-form gauge theories, the
integral of the current counts the number of charged objects. In the most familiar
case of 4d electrodynamics, the integral of the 3-form current over a spatial 3d

volume counts the number of charged particle worldlines crossing that volume.
Here, our 1-form current should integrate to unity on any line that crosses the
domain wall once. Concretely, consider a finite line, with beginning and end points
on opposite sides of the wall, such that
x 2
1 1
1= j1 = d ∗ F4 = (∗F4 ) . (6.13)
Line 4 Line 4 x1
From this, we see right away that the scalar ∗F4 jumps by 4 when crossing the
wall. The dual description, though even more exotic, is simpler:

S=− 4 F02 , (6.14)
without any meaningful ‘A−1 ’ or sources. The 0-form field strength is classically
identified with ∗F4 , it is constant in spacetime by its Bianchi identity, dF0 = 0, and
it only takes discrete values. This follows from the solution for F4 in the vicinity
of a domain wall discussed above. It can also be viewed as a degenerate version of
flux quantisation. The set of vacua following from the F0 description is displayed
in Fig. 6.4.
Now let us assume that our 4d theory possesses a large number of such 4-form
fields,

N
S=− 4i Fi,2 0 . (6.15)
i=1
This can arise, for example, if it originates from a compactification of a higher-

dimensional p-form gauge theory on a compact space with N (p +1)-cycles. The
flux on each of those cycles then corresponds to the flux number n in one of the
F0 -models in (6.15). The two-field case is illustrated in Fig. 6.5.
Fig. 6.4 Flux discretuum of

a 3-form gauge theory
6.2 Bousso–Polchinski Model 193
Fig. 6.5 Flux discretuum of

a 3-form gauge theory with
two fields
Each flux choice n = {n1 , · · · , nN } gives rise to a particular cosmological

constant

λ(n) ≡ V (n) = 4i n2i . (6.16)
i
One may ask how many different flux choices lead to λ(n) < λ0 . To simplify the
discussion, let us assume that all 4-form gauge couplings are equal: i = . The
number of flux√choices is then simply the number of lattice points n ∈ ZN inside a
ball of radius λ0 /2 . The lattice is N-dimensional, so the desired number is

K(λ0 ) ∼ ( λ0 /2 )N . (6.17)
√
If λ0 > 2 , this grows exponentially fast with N. In particular, the number
δK(λ0 , δλ) of points leading to
λ ∈ [λ0 , λ0 + δλ] (6.18)
will be extremely large for large N. This remains true if N is only moderately large
(say N = O(100), as suggested by the number of 3-cycles of the quintic). It will
also still be true if δλ is chosen very small:

δK(λ0 , δλ) ∼ ( λ0 /2 )N−1 (δλ/ λ0 2 ) . (6.19)
Note that we do not have to be afraid that regularities in the distribution of λ-values
could lead to intervals into which λ never falls: Such possible regularities will be
destroyed if we make all i different, as expected in a more realistic situation.
So far, we have a model with many solutions. These solutions give rise to a
discretuum of cosmological constants, which becomes extremely dense in the region
λ 4 (where sets the typical scale for the couplings i ). Now, by adding a
negative cosmological constant λAdS < 0, such that
N

S=− 4i Fi,2 0 + λAdS , (6.20)
i=1
we can shift this dense discretuum downwards. In this model, we are statistically
guaranteed that vacua with an extremely small cosmological constant exist. Clearly,
due to the possible bubble nucleation processes these vacua will only be metastable,
but they can be very long-lived. We will play with numbers later on to see how small
λ(n) in the model of (6.20) can really become.
6.3 The Type IIB Flux Landscape (GKP)
The key idea or observation is that, in type IIB Calabi–Yau compactifications,

the 3-form fluxes of H3 and F3 can play roughly the role of the multiple fluxes
of the Bousso–Polchinski model discussed above. The details are, however, more
complicated and in part qualitatively different, mainly due to the central role of
supersymmetry.
We start with the intuitive observation that a non-zero flux on a compact cycle
(say a 1-cycle) clearly has an energetic effect. Indeed, let us for simplicity assume
that the compact space is SA1 × SB1 and one unit of 1-form flux sits on the B-cycle.
Then

dyB F1 = 1 and hence F1 ∼ 1/RB . (6.21)
This gives rise to a contribution to the action

4 4 1 RA
S⊃− d x dyA dyB F1 ∧ ∗F1 ∼ − d x (RA RB ) · 2 ∼ − d 4x .
RB RB
(6.22)
We learn that a flux on a cycle prevents this cycle from shrinking. More generally,
if there are fluxes of various values on various cycles of a compact space, then these
fluxes tend to stabilise the shape of the manifold in a certain way. Specifically, the
ratio between the volumes of two cycles gets stabilised roughly according to the
ratio of the flux numbers on these cycles.
Concretely, we expect that 3-form fluxes will stabilise (give mass to) the complex
structure moduli, which as we know govern the ratios of 3-cycle volumes. But this
is not possible in a 4d supergravity model without superpotential since for W = 0,
no scalar potential is induced.
To make the right guess for the form of the expected flux-induced W, it is useful
to observe that (already in 10d) one can use the complex scalar field
S = C0 + ie−φ (6.23)
to define a complex 3-form flux
G3 = F3 − S H3 . (6.24)
6.3 The Type IIB Flux Landscape (GKP) 195
The kinetic terms of the two 3-form fields take the simpler form (suppressing
constant prefactors)

S⊃ d 10 x G3 ∧ ∗G3 . (6.25)
With this, one may guess the mathematically natural expression for the superpo-
tential induced by 3-form fluxes:

W = G3 ∧ 3 . (6.26)
X
This is known as (the type IIB version of) the Gukov–Vafa–Witten superpoten-
tial [200]. The latter has first been postulated and mathematically justified (in an
abstract way) for M-theory compactifications to 3d on Calabi–Yau 4-folds:

WGV W = G4 ∧ 4 . (6.27)
X4
In the famous paper by Giddings, Kachru and Polchinski (GKP) [133] (see
also [201] and [202] for a review), this superpotential was used and justified
explicitly by comparing the 4d scalar potentials derived from 4d N = 1 supergravity
and directly from 10d.
Now one can make this fully explicit by normalising the 3-form fields such that
the flux quantisation takes the form

1 1
F3 ∈ 2πZ , H3 ∈ 2πZ (6.28)
2πα 2πα
for integrals over integer cycles (see e.g. [161] for more details and examples).
Equivalently, one may decompose the fluxes in a symplectic integral form basis,
F3 = −(2π)2 α (f a ωaA +fb3 /2+b ωBb ) , H3 = −(2π)2 α (ha ωaA + hb3 /2+b ωBb ) ,
(6.29)
where the entries of the coefficient vectors f and h now have to be integer. With this,
the superpotential, given in its simplest and mathematically natural form above, can
be worked out explicitly:

W = G3 ∧ 3 = (2π)2 α (f − Sh) · (z) . (6.30)
X
The scalar potential reads, as usual,3

V = eK K ij (Di W )(Dj W ) + K αβ (Dα W )(Dβ W ) − 3|W |2 . (6.31)
Here we have, for simplicity and since they all appear in W = W (S, z), combined
2,1
the axio-dilaton S and the complex structure moduli z1 , · · · , zh in one vector:
2,1
zi = {S, z1 , · · · , zh } . (6.32)
We have furthermore redefined
Kc.s − ln(−i(S + S)) → Kc.s , (6.33)
absorbing the axio-dilaton Kahler potential into the complex structure Kahler
potential.
It is essential that W is independent of the Kahler moduli T α . Moreover, the
Kahler modulus Kahler potential takes the special form KK = − ln(V 2 ), with V 2
a homogeneous function of the T α of degree three. This constitutes a so-called no-
scale model, implying the very special result that the last two terms in (6.31) exactly
cancel [203]. This is discussed in more detail in Problem 6.9.1. In the simplest case
of a single Kahler modulus,
KK = −2 ln(V) = −2 ln((T + T )3/2 ) = −3 ln(T + T ) , (6.34)
the cancellation of the last two terms in (6.31) is easily observed. Hence, we have
V = eK K ij (Di W )(Dj W ) . (6.35)
Moreover, the equations for unbroken SUSY (the F -term conditions)
Di W = 0 for i = 1, · · · , b3 /2 (6.36)
represent b3 /2 equations for equally many complex variables. They will in general
possess a solution (or a finite set of solutions). This fixes all zi to specific values.
One may view these fields, which now have a large mass in this positive definite
potential, as being integrated out. The result is a model depending just on T (or,
more generally, all Kahler moduli) in which
V = V (T , T ) ≡ 0 . (6.37)
3 The ambiguity of the normalisation of cancels out since also appears in K. However, relative
to the conventions of [133] we have absorbed a factor of ln(2π) into the definition of K to be
consistent with our previous supergravity definition of the scalar potential.
6.4 Kahler Modulus Stabilisation and SUSY Breaking (KKLT) 197
Since
−1
T 3 (−3)
F = D T W = K T T KT W = W = −(T + T ) W = 0 ,
(T + T )2 (T + T )
(6.38)
supersymmetry is broken. The scale at which it is broken (e.g. the gravitino mass
eK/2W ) is not fixed since T is not fixed. This explains the name no-scale model.
One of the key points of [133] (known as ‘GKP’) is that they established this
vanishing potential not only (as we just did) indirectly, via 4d SUGRA arguments,
but also by explicitly providing the 10d geometry. The term ‘explicitly’ is here
interpreted as follows: One assumes that a Calabi–Yau metric is given (this is
of course not explicit but rests on the famous existence theorem). Then, given
in addition certain fluxes and other sources in the Calabi–Yau (e.g. O3-planes
and D3-branes), one is able to write down differential equations determining the
actual metric, including backreaction from fluxes. This metric corresponds to a flux
compactification to 4d Minkowski space. In fact, there is a family of such solutions,
corresponding to the flat direction characterised by the ‘no-scale modulus’ T , as
explained above.4
6.4 Kahler Modulus Stabilisation and SUSY Breaking (KKLT)
The leading proposals for the Kahler moduli stabilisation and the realisation of a
positive cosmological constant are known as KKLT [192] and LVS (Large Volume
Scenario) [193]. We will discuss them in the next three sections. For further
suggestions, see e.g. [205, 206].
We start with KKLT. As in the previous section, we focus on the simplest case
h1,1 = 1, such that
K = −3 ln(T + T ) and W = W0 = const. (6.39)
The complex structure moduli have been integrated out, and the corresponding flux
choice (together with the VEVs of the zi which it prescribes) has fixed W0 . At
leading order, we have V ≡ 0 and SUSY breaking with m3/2 = eK/2 W0 .
4 As emphasised in [204], the detailed situation is more complicated: Once W = 0, certain higher
derivative terms present in the 10d action necessarily induce a small 4d effective potential, which
in general leads to a runaway to small or large volume. However, once the non-perturbative effects
discussed in the next section are included, the volume is stabilised and it turns out that, at least in the
appropriate parametric regime, the higher-derivative corrections are not important. This viewpoint
is widely accepted but vigorously disputed in [204] on the ground that one may not proceed from
a starting point which is not a solution for all times. We here take the pragmatic EFT attitude that
a sufficiently slow runaway is as good as a static solution and correct the latter by non-perturbative
effects governed by shorter time-scales is sensible.
Various (quantum) corrections will generically lift the flatness of V , breaking the
no-scale structure. This can be α corrections (corresponding to higher-dimension
operators in 10d), loop-corrections, non-perturbative instanton effects or non-
perturbative effects from (SUSY) gauge theory confinement (also known as gaugino
condensation). The last two of these four qualitatively different effects lead to
technically similar results. In particular, W is corrected according to
W0 → W0 + A e−2πT /N , (6.40)
either by instantons (in this case N = 1) or by gaugino condensation (here, N is

the dimension of the fundamental representation of the gauge group SU (N)). This
type of corrections (see [196, 197] for reviews) is one of the basic ingredients of
the KKLT scenario for (complete) moduli stabilisation and SUSY breaking [192],
which we will describe in the rest of this section.
We will not introduce the technology necessary to analyse the case of gaugino
condensation (see [207, 208] for its original discussion in heterotic models). Suffice
it to say that, if a stack of N D7-branes is wrapped on a 4-cycle with volume
∼ Re T , the 4d theory contains a corresponding N = 1 super-Yang–Mills theory.
The latter exhibits confinement, as familiar from the non-SUSY QCD sector of the
Standard Model. In the SUSY case, confinement is characterised by a non-zero
Wnon−pert. ∼ 3 . Here is the confinement scale, and its relation to Wnon−pert.
follows on dimensional grounds. Using also the fact that the 4d gauge coupling
squared is ∼ 1/Re T by the standard logic of Kaluza–Klein reduction from 8d to 4d,
one may run from high to low energy scales and determine at which scale the gauge
coupling reaches O(1) values. This fixes in terms of Re T and, by holomorphicity,
leads to Wnon−pert. ∼ exp(−2πT /N), where N comes in through the beta-function.
Since we have already introduced some of the ideas relevant for instanton effects,
we will describe the instanton case in slightly more detail. For this, it is useful to
recall how an instanton correction in a 5d-to-4d compactification of a gauge theory
is related to the possibility of wrapping a closed electron worldline on a 1-cycle (in
this case, the unique 1-cycle S 1 ) of the compact space. This is illustrated in Fig. 6.2,
where the reader is invited to think of a particular type of e+ e− fluctuation of
the vacuum. But this time-dependent picture is not necessary—the simplest and
dominant effect corresponds to just wrapping the worldline on the minimal volume
cycle (at fixed 4d spacetime point x μ ), subsequently integrating over all x μ .
This type of instanton has an obvious analogue in compactifications of higher-
form gauge theories. The effect occurs if the compact space possesses cycles the
dimensions of which correspond to the dimensionality of the available charged
objects. Our case of interest is type IIB with its 4-form C4 and the corresponding
D3-branes. We think of the electron worldline above as of a 0-brane and, once it is
wrapped in euclidean signature to describe tunnelling, an E0-brane. Analogously,
we can think of a D3-brane (now often called an E3-brane) as being wrapped at
fixed x μ on the minimal volume 4-cycle of our Calabi–Yau. This is the origin of the
instanton correction we are after, cf. Fig. 6.6. Such instantons are called stringy or
exotic or D-brane instantons.
Fig. 6.6 An E3-brane

instanton, corresponding to a
euclidean D3-brane wrapped
on a 4-cycle of a CY over one
of the points of the
non-compact spacetime R4
At the quantitative level, we recall that our complex Kahler modulus is T =

τ + ic, where

√
τ = gCY ∼ RCY4
. (6.41)
4−cycle
The last expression is, up to the proper normalisation by the tension prefactor, the
action of the wrapped brane. Furthermore, the wrapped brane couples to C4 through

2π C4 ≡ c , (6.42)
4−cycle
√
which is just the 4d axionic scalar in T (recall that we have set 2π α = 1). Thus,
a single instanton contributes to the 4d partition function as
∼ e−2πτ e−2πic , (6.43)
where the first factor is the tunnelling suppression by the euclidean brane action. The
second factor comes from the part of the D3-brane action displayed in the previous
line. It can equivalently be viewed purely in 4d as the coupling of the 0-form gauge
field c to its 0-dimensional charged object, the instanton.
Summing over all numbers of instantons and anti-instantons (which come with
e+2πic ) leads to an exponentiation:

L4d ⊃ exp ∼ e−2πτ cos(2πc) . (6.44)
The term in the exponent is the instanton correction to the 4d effective action, and
it is precisely analogous to the possibly more familiar gauge theory case. Here, one
gets corrections ∼ e−8π/g cos(2πφ), where g is, for example, the strong gauge
2
coupling and φ the QCD axion, famously obtaining a cosine potential from this
effect.
In SUSY, such instanton corrections can enter the 4d effective action only
through either K or W :
W0 → W0 + Ae−2πT or K → K + Be−2πT + c.c. .

(6.45)
Which of the two happens depends on the geometry of the wrapped brane and will
not be discussed here [209] (see [196, 197] for reviews as well as [210, 211] and
references therein). For KKLT, we require that a correction to W arises. We also note
that the τ and c dependences are such that they combine in a holomorphic function
of T (as required by SUSY), with the proper periodicity in Im(T ). Conversely, as
shown in the problems, the evaluation of the scalar potential on the basis of W from
(6.45) leads to a term of the type of (6.44).
We can now finally proceed with the analysis of the 4d effective theory, defined
by
K = −3 ln(T + T ) and W = W0 + Ae−aT . (6.46)
It is a straightforward exercise (Problem 6.9.1) to derive the scalar potential V (τ, c),
integrate out c (by simply finding the minimum in c) and thus obtain
V = V (τ ) . (6.47)
The qualitative behaviour of this potential at W0 1 is displayed in Fig. 6.7. It is

easy to derive by analysing the standard supergravity formula for V in the regimes
e−aT W0 and e−aT W0 . One checks that V grows at small τ and approaches
zero from below at large τ . This is sufficient to conclude that the qualitative picture
is that of Fig. 6.7.
Moreover, it is easy to prove, in general, that the supergravity scalar potential
has an extremum at supersymmetric points, where the F -terms and hence the first,
positive-definite term in the supergravity potential formula vanish,
3
DW = −aAe−T − W0 + Ae−aT = 0 . (6.48)
T +T
In our case this extremum is always a minimum. Assuming c = 0, this vanishing

F -term condition is solved (implicitly in τ ) if

2
W0 = − 1 + aτ Ae−aτ (6.49)
3
Fig. 6.7 Qualitative

behaviour of the scalar
potential arising after the
inclusion of instanton or
gaugino condensation effects
in W
holds. The conclusion that W0 must be real and negative is a mere consequence of
our simplifying assumption c = 0. For a general phase of W0 (and A), we would
simply have found a non-zero value of c at the minimum. This is not important for
us.
What is important is the conclusion that W0 must be exponentially small for
parametric control, i.e. to have RCY 1. Of course, making W0 small should not
be a problem since it depends on the flux choice—it can hence be finely tuned in
the landscape. In fact, to be sure that nothing goes wrong one needs to know that
the statistical distribution of W0 in the complex W0 -plane for random flux choices
has no special feature near the origin. This crucial fact, more precisely the flatness
of the distribution of |W0 |2 values near zero, has been established with some level
of rigour in [212].
Thus, we have uncovered a landscape of supersymmetric vacua with a negative
cosmological constant, so-called SUSY AdS vacua. (Note that, in the ‘first step of
KKLT’ leading to these solutions, the broken supersymmetry of GKP is restored
in the minimum). But to describe the real world, we need a positive (even though
very tiny) cosmological constant and broken supersymmetry. Moreover, turning at
least a small fraction of the SUSY AdS vacua above into dS vacua is essential
for eternal inflation, the presently leading mechanism for populating the landscape
cosmologically (see Chap. 7).
We first give a much simplified, ‘macroscopic’ description of how dS vacua may
arise on the basis of the above (see e.g. [213–216]). Let us assume that some further
details of the model, such as branes with their gauge theories and charged matter
fields, introduce extra light degrees of freedom X and corresponding corrections to
K and W :
K → K(T T ) + δK(X, X, T , T ) , W → W (T ) + δW (X) .

(6.50)
Now, let us choose δK and δW in analogy to the one-field O’Raifeartaigh-type

model discussed in Sect. 2.7:
δK ∼ XX − (XX)2 , δW = αX . (6.51)
This will lead to DX W = 0 in the vacuum. Moreover, let us choose the parameters
such that the fluctuations of X around this SUSY breaking vacuum have a very
large mass. Then the upshot of the whole construction is that the scalar potential V
is supplemented by a so-called uplifting term
V → V + δV with δV = eK K XX |DX W |2 . (6.52)
At this generic level of analysis, the uplifting term δV could have any T
dependence, given our free choice of the T dependence of δK. In concrete string
constructions, for which the above is a toy model, δV will always be decaying
Fig. 6.8 Uplifting to a

KKLT dS vacuum
at large volume, cf. Fig. 6.8. This can be understood if one imagines that (as is
mostly the case) δK and δW are due to some local effect in the CY. Then, going
to large volume, the SUSY breaking and uplifting effects stay the same in string
units, but the Planck mass diverges. Hence, in standard supergravity conventions
with MP = 1, δV will decay with growing T .
One may expect that in the huge string theory landscape, many options for
such an uplift exist. Yet, it turns out not to be easy to construct an uplift of the
above O’Raifeartaigh type explicitly. Thus, the most explicit uplift has a somewhat
different structure: It is the anti-D3-brane uplift originally suggested by KKLT,
which arguably remains the most explicit (though nevertheless not uncontroversial5
) possibility. We turn to this construction, which requires some more technology,
next. As we will see, even though different in detail, the KKLT uplift behaves
qualitatively as explained using the O‘Raifeartaigh toy model above.
6.5 The Anti-D3-Brane Uplift of KKLT
As we explained earlier, a so-called orientifold projection reduces the supersym-

metry of a type II Calabi–Yau compactification from N = 2 to N = 1. Let us
consider the example of an O3-plane projection, which can locally be thought of as
the geometric action
(z1 , z2 , z3 ) → (−z1 , −z2 , −z3 ) , (6.53)
to be combined with worldsheet orientation change.

Locally, this projection introduces a singularity at {zi } = 0, at which (due to
the orientation change) a so-called O3-plane is localised. This is a negative-tension
5 There even exists the opinion that no uplift to a dS minimum can ever be constructed for
fundamental reasons, challenging most ideas about how string theory might be relevant to the
real world [217, 218]. We will return to this subject.
6.5 The Anti-D3-Brane Uplift of KKLT 203
object which also has opposite C4 -charge6 compared to a D3-brane. In a consistent

compactification, an O3-plane always has to come with a certain number of D3-
branes for total charge neutrality (tadpole cancellation). Concretely, the D3-charge
of an O3-plane is −1/4. The fractionality is not a problem since the compact Calabi–
Yau after orientifolding will usually have a large number, divisible by 4, of O3-
planes. For example, it is easy to check that T 6 /Z2 , with the Z2 acting as above, has
64 O3-planes.
Now, given a consistent Calabi–Yau with a number of O3-planes and a corre-
sponding number of D3-branes, it is possible to replace some or all of the D3-branes
by 3-form fluxes. This possibility arises since, through the Chern–Simons term, 3-
form fluxes contribute to the total D3 tadpole. This takes us to the realm of flux
compactifications à la GKP and, if we also allow for the non-perturbative effects
∼ e−aT introduced above, we will find ourselves in an N = 1 SUSY setting with
O3-planes, D3-branes and fluxes. The O3-planes, the D3-branes and the fluxes all
break SUSY to the same N = 1 subalgebra of the original N = 2 SUSY of the
pure Calabi–Yau model.
Next, we can think of breaking SUSY by adding a D3- and anti-D3- (for short:
D3) brane pair.7 The D3 breaks N = 2 to the opposite N = 1 subalgebra, such that
4d SUSY is now completely broken. D3 tadpole cancellation is not violated since we
added two oppositely charged objects. However, brane and anti-brane attract each
other both gravitationally and through C4 , so they will quickly find each other and
annihilate, releasing twice the energy density of the D3-brane tension. Our ‘uplift’
is thus very short-lived and not practically useful.
However, we could avoid having any D3-branes by cancelling the tadpole of
the O3-planes by flux alone. If we now add a D3-brane and increase the flux
appropriately to ensure tadpole cancellation, we appear to have the desired uplift.
Now, the D3 still breaks SUSY relative to flux and O3-planes, but there is no D3
which it could attract and annihilate.
Unfortunately, this is not yet good enough since this uplift (by twice the D3-
brane tension, which is string scale) is much too strong. Indeed, given that the
non-perturbative effects and hence the depth of the original AdS minimum are
exponentially small, the situation will be as in Fig. 6.9: The uplift is much too strong,
and no local dS minimum can be generated.
Fortunately, the key to a resolution of this problem is already contained in the
seminal work of GKP [133] discussed above. They show explicitly that the metric
on a CY orientifold threaded by 3-form flux is not of product type but warped:
ds 2 = 2 (y)ημν dx μ dx ν + gmn (y)dy mdy n . (6.54)
6 This is often referred to as D3-charge.

7 See [219–221] for the first suggestions of how to realise approximate de Sitter space using the
positive energy of a brane–anti-brane pair.
Fig. 6.9 If too high an uplift

is added to a model with
SUSY AdS vacuum, no
metastable de Sitter minimum
is generated
Fig. 6.10 Calabi–Yau with

warped throat. The
Calabi–Yau is basically
undeformed in the region of
small warping ( 1) and
strongly deformed in the
1 domain (where the
D3-brane is localised)
Here x μ (with μ = 0, · · · , 3) and y m (with m = 1 · · · 6) parameterise the non-

compact R4 and compact X6 part of our total space, respectively. This space is,
topologically and as a differentiable manifold, still of product type, R4 × X6 .
However, the metric manifold built on this basis does not share this product
structure. As we can see from the warped metric ansatz in (6.54), this breaking
of the product structure is perfectly consistent with 4d Poincare invariance as long
as y enters in the prefactor of the non-compact metric, but x does not enter in the
prefactor of the compact part of the metric. One refers to (y) as the warp factor.
Moreover, GKP show that given certain (very common8 ) features of the CY and
a particular flux choice, the compact manifold develops a strongly warped region.
This region is also known as a Klebanov–Strassler throat [223] and is graphically
often represented as in Fig. 6.10. To understand that the compact geometry is
strongly deformed at strong warping, one also needs to know that [133, 223]
gmn (y) = −2 g̃mn (y) , (6.55)
where g̃ is the Calabi–Yau metric. One says that the compact space is not ‘Calabi–
Yau’ but only ‘conformal-Calabi–Yau’. For our purposes, it is essential that strong
warping can substantially change the energy effect of the SUSY breaking D3-brane
placed in the Calabi–Yau orientifold.
8 The feature we need is a so-called conifold singularity. The latter develops when a certain type of
3-cycle shrinks to zero volume (i.e. z → 0 if z is the modulus parameterising the corresponding
period). This is in fact a generic type of 3-cycle of a CY, so such a situation arises frequently.
Conversely, the conifold singularity can be made smooth (‘deformed’) by ‘blowing up’ a 3-cycle.
For more details, see e.g. [222].
6.5 The Anti-D3-Brane Uplift of KKLT 205
To understand this central aspect, recall the Schwarzschild black hole metric
ds 2 = −f (r)dt 2 + f −1 (r)dr 2 + r 2 dω2 , (6.56)
where dω2 is the metric on the unit sphere. Clearly, f (r) bears similarity to our
2 (y). As is well-known, the vanishing of f (r) as one approaches the horizon is
responsible for the redshift effect and the force that pulls any massive object into
the black hole. The same happens in our case: The D3-brane represents a SUSY
breaking local energy density in the warped Calabi–Yau, and this brane is pulled
towards strong warping (where 1). Once there, its energetic effect as seen
from the unwarped ‘bulk’ of the Calabi–Yau is greatly reduced. In other words,
the anti-brane naturally sits at the bottom of the warped throat and uplifts the total
potential energy of the compactification only by
4min × O(1) (6.57)
in string units. The fourth power of arises since, as known from black hole
physics, f 1/2 is the redshift factor and, in our context, we are redshifting an energy
density, i.e. an object of mass dimension four.
As shown in GKP,
min ∼ exp(−2πK/3Mgs ) , (6.58)
where K and M are flux numbers associated with 3-cycles of the Klebanov–
Strassler throat geometry and gs is the string coupling constant. The latter is
governed by the modulus S stabilised by fluxes. Thus, one apparently has enough
freedom to choose fluxes in such a way that min is exponentially small.9
Before moving on, it should be mentioned that a debate about the metastability
of the anti-brane at the bottom of the throat has been going on for a number of years
(see [226–231] and references therein). Indeed, as should be clear from the above
that the D3 breaks SUSY (in the absence of any D3 brane) against the fluxes in
the throat. It can annihilate against these fluxes only at the price of overcoming an
energy barrier, making the uplifted configuration at best metastable [232]. However,
the backreaction of fluxes to the presence of the anti-brane is poorly understood, and
a barrier-free decay or outright instability has been claimed. In spite of many efforts
to show the opposite, long-lived metastability as described in [232] has remained
plausible [227, 229]. On the other hand, a better, fully backreacted understanding
of the geometry with the anti-brane included would be highly desirable but remains
challenging.
Let us now assume that the above D3 uplift does indeed provide metastable
SUSY breaking and estimate its magnitude. For simplicity, we disregard factors
of gs such that the tension of the 3-brane in either the 10d string or the Einstein
9 See [224, 225] for recent, in part, critical comments related to this point.
frame is O(1) × ls−4 ∼ O(1). Here we also, as before, use conventions in which
all dimensionful quantities in 10d are measured in units of the string scale or the
inverse string scale. If there were no warping, then, compactifying, the D3 brane
tension (more precisely, twice this number—see above) induces a 4d energy density
∼ O(1). Note that we are still using string units and our 4d Planck mass is MP2 ∼ V
(i.e. we are in a ‘Brans–Dicke frame’). Next, we Weyl rescale the 4d metric to go to
the 4d Einstein frame. This amounts to using 4d Planck units (i.e. setting the Planck
mass to unity) in the 4d effective action. Since, in this process, dimensionless ratios
of physical observables do not change, we have
ρ Einst ein ρ Brans –Dicke 1 1 1

D3 D3
∼ ∼ or Eist ein
ρD3 ∼ ∼ 3.
4
(MP )Einst ein 4
(MP )Brans –Dicke V2 V2 τ
(6.59)
Most naively, one would now like to include warping by multiplying with the
fourth power of the redshift factor 4min [192]. This is correct in principle, but
at a quantitative level a further fine point has to be taken into account [233]:
Indeed, the expression (6.58) is valid in the strongly warped region near the tip of a
Klebanov–Strassler throat. It represents correctly the dependence of the warping
on the relevant discrete flux choice. Yet, if the Calabi–Yau volume is taken to
infinity, then eventually the fluxes become so diluted that their backreaction on the
geometry is negligible and ∼ 1, even at the lowest point of the throat. This can be
quantified [233] (see also [234]) and leads to the more precise warping suppression
4min → 4min τ , (6.60)
valid only as long as 4min τ 1.

Combining everything, one arrives at

VKKLT = eK K T T |DT W |2 − 3|W |2 + Vup (τ ) , (6.61)
with
4min
K = −3 ln(T + T ) , W = W0 + Ae−aT and Vup (τ ) = c .
τ2
(6.62)
Here A, a and c are numerical O(1) factors, and W0 and min can be chosen
extremely small by an appropriate flux choice. It is easy to convince oneself
numerically or analytically that an uplifted situation with a metastable de Sitter
or Minkowski vacuum as in Fig. 6.8 can be achieved on the basis of the above
potential. The reader is invited to verify this. The key non-trivial point is that
the AdS minimum is very steep (based on the exponential behaviour of the non-
6.6 The Large Volume Scenario 207
perturbative superpotential ∼ e−aT ), while the uplift has a relatively flat, power-like
τ dependence. Hence, the local minimum survives the uplift to a value above zero.10
6.6 The Large Volume Scenario
A very promising alternative to the KKLT proposal for Kahler moduli stabilisation
in an AdS vacuum (before uplift) is provided by the Large Volume Scenario or
LVS [193, 242]. It has the disadvantage of being slightly more involved than KKLT
but the advantage that the stabilised value of the volume VLV S may be exponentially
large—a feature not available in KKLT due to the parametric behaviour VKKLT ∼
ln(1/|W0 |).
In the simplest realisation, two Kahler moduli Tb and Ts (with the indices
standing for ‘big’ and ‘small’) are required. The volume is assumed to take the
form
3/2 3/2
V(τb , τs ) ∼ τb − kτs , (6.63)
with 2τi = Ti + T i and KK = −2 ln V, as usual. Here the small 4-cycle, governed

by τs , is also known as a ‘blow-up cycle’. This is because one may say that it arises
from ‘blowing up’ (in the sense of making smooth) a singularity of a Calabi–Yau
with a single large 4-cycle. Explicit geometries leading to a Kahler potential of the
required structure and possessing other necessary features (see below) have been
studied, see e.g. [243, 244].
As in the KKLT setup of the last section, the first step is to assume that complex
structure moduli and axio-dilaton are supersymmetrically stabilised by fluxes. The
only fields we consider are, hence, Ts and Tb with the Kahler potential given above
and W0 = const. At this level, the scalar potential is identically zero because of the
no-scale structure of K ≡ KK .
Two types of correction are considered: First, instanton corrections, or a gaugino
condensate, lead to
W = W0 + Ae−aTs . (6.64)
An analogous term with an exponential suppression in τb = Re Tb will generically

also be present. It has been neglected since, as we will see below, one eventually
finds τb τs .
10 We note that a new round of criticism and defence of this construction has appeared relatively
recently, related mainly to the question whether the non-perturbative effect (in this case, gaugino
condensation) and the subsequent uplift can also be understood directly in 10d [235–240]. At this
point it appears that, yet again, the success of the KKLT construction remains plausible [238]. An
interesting novel criticism, concerned with geometric consistency, was raised in [240]. From this, a
quantitative singularity problem was derived in [241]. We will comment on these issues in slightly
more detail in Sect. 8.5.
Second, there are so-called α corrections. The original meaning of the term
is higher-dimension operators
√ in the 10d action, which are suppressed by an
appropriate power of ls ∼ α ∼ 1/Ms (see [245] for an introduction and review).
The relevant term in the present context is a particular contraction of four 10d
Riemann tensors, and the integral of which over the Calabi–Yau corrects the 4d
theory through a modification of the Kahler moduli Kahler potential [246]:
χ(CY )ζ (3)
KK = −2 ln V → KK = −2 ln(V + ξ /2) , with ξ= 3/2
.
2gs (2π)3
(6.65)
Here χ(CY ) is the Euler number of the relevant 3-fold, ζ(3) 1.202 is the
appropriate value of the Riemann zeta function and gs follows from the stabilised
value of the dilaton, as usual.
The evaluation of the F -term scalar potential using the standard supergravity
formula for the 2-field model based on Tb,s and defined by (6.64) and (6.65) gives
√
gs eKcs α 2 τs e−2aτs α|W0 |τs e−aτs 3ξ |W0 |2
V (V, τs ) − + . (6.66)
2 6kV V2 4V 3
Here α ≡ 4a|A| and Kcs is the complex structure Kahler potential with flux-
stabilised arguments. The axionic modulus Im Ts has already been integrated
out, while the Tb -axion remains massless in this approximation. We leave the
straightforward derivation to the reader (Problem 6.9.2). This scalar potential is
easily seen to have a relatively steep minimum in τs . After integrating out τs , one
finds

3gs eKcs |W0 |2 ξ k αV
V (V) · − ln 3/2
. (6.67)
2 V3 4 2a 3/2 3k|W0 |
Crucially, as observed in [193], this leads to the stabilisation of the last remaining
modulus V at an exponentially large value and a negative cosmological constant.
Supersymmetry is mainly broken by the large F -term of Tb .
To be slightly more precise, the parametric control of the final result is achieved
as follows: Eventually, the small-cycle volume is stabilised at
2/3
ξ
τs . (6.68)
2k
This needs to be at least somewhat large, due to either a large Euler number or a
small flux-stabilised values of gs . But then, due to the exponential dependence, the
parameter exp(−aτs ) controlling the instanton expansion can easily be extremely
small. Similarly, the stabilised value of V can easily turn out to be huge,
V ∼ |W0 | eaτs , (6.69)

6.7 Vacuum Statistics and the Tuning of the Cosmological Constant 209
where we have suppressed an O(1) coefficient. This is excellent news since the
volume is the main control parameter of the supergravity expansion underlying
the whole approach. Nevertheless, one has to be cautious since there is also the
(relatively) small-cycle volume τs , and hence the curvature in its vicinity does not
become exponentially small in string units. Related to this, the α correction used
in the analysis is calculated on a Calabi–Yau, while the realistic case will also have
orientifold planes and branes, with additional α corrections that are not yet fully
understood.
We leave the detailed derivation of the formulae given above to the problems.
Also, we will not discuss in any detail LVS-specific issues concerning the uplift. An
uplift is of course necessary since the mechanism discussed leads to a non-SUSY
AdS vacuum. One option is to use the same D3-uplift analysed in the context of
KKLT. In addition, specifically in LVS constructions, the so-called D-term uplift
appears to represent a promising possibility (see [247] and [98, 248] for earlier
discussions of the underlying idea).
Let us end with a comment concerning loop effects: While W is protected by
supersymmetry, KK receives corrections that, in different regimes, may be best
viewed as either 4d loop effects in a theory with KK modes, 10d loop effects in
a compactified model, or full-fledged string loop corrections. Consider first a one-
field model, where KK = −3 ln(T + T ) at tree level: A simple scaling analysis
shows that a correction δKK ∼ 1/(T + T ), if it arises, can be absorbed in a
constant shift of T under the log and hence does not affect the scalar potential [249].
This cancellation of the formally leading correction was named ‘extended no-scale
structure’ in the analysis of [250], and it continues to hold in the multi-field case
[251] and for the explicitly derived string loop corrections of simple torus-orbifold
models [252]. For the present discussion, the upshot is that the α correction with its
parametric behaviour δK ∼ 1/(T +T )3/2 produces the dominant effect in the scalar
potential (∼ 1/V 3 ). This justifies a posteriori our neglect of loop effects in the bulk
of this section. However, once several large moduli are present, their relative size
remains unstabilised in an LVS analysis using only α effects. Loop corrections are
then the leading contribution which induces a potential, at order 1/V 10/3, stabilising
all large Kahler moduli.
6.7 Vacuum Statistics and the Tuning of the Cosmological

Constant
Let us now assume that one of the moduli stabilisation and uplifting procedures
discussed in the literature (the two main examples being KKLT and LVS) or some
variant thereof works. Moreover, as we did not explain in detail but only sketched
in Sect. 5.7, there should be no problem in finding a compactification which, at
the same time, contains a standard model-like sector. Together, this implies the
existence of a landscape of realistic 4d EFTs with a certain random distribution
of operator coefficients, including, in particular, the cosmological constant λ and
the Higgs mass parameter m2H . Crucially, they would have broken supersymmetry
and, at least in part, positive λ.
Two non-trivial questions can then be asked. First, is it clear that the landscape
contains a vacuum with the apparently highly fine-tuned values of λ and m2H we
observe? Second, can we understand why we find ourselves in a world described by
such a very special vacuum?
In this section, we want to discuss, at least briefly, the first (and simpler) of
these two questions. We focus on λ and KKLT. In this case, a partial answer can
be given using a fundamental technical result of [212] (see also [94, 253, 254]).
In the analysis of [212], the focus is entirely on the flux stabilisation of complex
structure moduli (and the axio-dilaton), i.e. Kahler moduli are ignored. The setting
is (for our purposes) that of type IIB Calabi–Yau orientifolds with O3/O7-planes.
In this setting, the tadpole constraint on the flux vector (f, h) can be calculated (for
details see below) such that one knows precisely in which subset of the space of
integer vectors, this object takes its values. Each such value corresponds to a point
in complex structure moduli space at which the geometry (the variables zi and S) is
then stabilised. If the dimension of the moduli space and, hence, the vector (f, h) is
large, solving for the zi on the basis of a given flux value is practically impossible.
But, assuming that the set of relevant flux choices is large, it is possible to talk about
the resulting (approximately) statistical distribution of vacua in moduli space. In
fact, in the strict mathematical limit of a large tadpole (taking the restriction on the
length of (f, h) to infinity), this becomes a precise mathematical question.
The key answer given in [212] concerns the distribution of a particular quantity,
eK/2W0 . It was shown that, under mild assumptions, the distribution of this number
in the complex plane is flat near zero, cf. Fig. 6.11. This is not surprising since W0
is by definition a sum of many terms of varying phase and there is nothing special
about obtaining the value zero in total. We now want to include Kahler moduli (for
simplicity, a single Kahler modulus) assuming that, for a large subset of these vacua,
instanton or gaugino condensation effects are present. This leads to
W0 → W0 + Ae−aT , (6.70)
eventually giving rise to full moduli stabilisation in AdS with
λAdS ∼ −eK |W0 |2 . (6.71)
Fig. 6.11 The distribution of

eK/2 W0 in the complex plane
has no special feature near the
origin
Fig. 6.12 Distribution of the

cosmological constant before
and after uplift
The flat distribution of the complex number eK/2 W0 now implies a flat distribution
of λAdS , reaching up to zero from below (cf. the l.h. side of Fig. 6.12). After an uplift
of the type described in Sects. 6.4 or 6.5, a dense distribution of λ values including
the zero point is obtained (cf. the r.h. side of Fig. 6.12).
It is crucial in this logic that both the value of W0 and the uplift energy can be
extremely small. In the first case, the reason is the tuning in the flux discretuum,
as described above. In the second case, it is the exponential warping suppression.
Thus, a value of λ very close to zero can arise after a shallow AdS vacuum is
uplifted by a small amount. The restriction to shallow AdS vacua and small uplifts is
crucial for calculational control purposes. Specifically, small W0 implies a relatively
large volume and hence a suppression of various higher-order (α and string loop)
corrections.
Of course, it is important to quantify how dense the discretuum is and hence
how finely spaced a distribution of λ values in Fig. 6.12 one can hope for. For this,
we need to discuss tadpole cancellation for the C4 potential. By this, we mean that
the coefficient of the action term linear in C4 (the ‘tadpole’) should be zero. The
intuition behind this is best explained by an analogy to electrodynamics (see also
Sect. 5.7):
Imagine our space were not R3 but compact, say S 3 . Then by Gauss’ law, a static
solution of the Maxwell equations
d ∗ F2 = d ∗ dA1 = j3 (6.72)
clearly requires that the total number of sources add up to zero,

j3 = 0 . (6.73)
S3
Even more intuitively, the number of electrons and positrons must be the same since
there cannot be more ‘beginnings’ than ‘ends’ of electric field lines on a compact
manifold.
In our case, the Chern–Simons lagrangian

f lux
C4 ∧ F3 ∧ H3 ≡ C4 ∧ j6 (6.74)
implies that part of the sources for C4 is provided by the 3-form flux. Moreover, the
type IIB equations of motion also imply that G3 is imaginary self-dual [133],
∗6 G3 = iG3 , (6.75)
f lux
which in turn implies that CY j6 cannot get different sign contributions from
different regions of the CY. To see this, rewrite F3 ∧ H3 in terms of G3 ∧ G3 and
the latter in terms of the manifestly positive quantity G3 ∧∗G3 . The contribution of
the fluxes to the C4 source or the so-called ‘D3 tadpole’ can be written as

f lux
j6 = F3 ∧ H3 ∼ (h, f )2 , (6.76)
CY
with an appropriately defined (symplectic) product on the space of flux vectors

(h, f ).
This flux vector contribution to the D3 tadpole has to be cancelled by other
charged objects. We are discussing this before the uplift, so D3-branes are not at
our disposal. D3-branes contribute with the same sign as (supersymmetric) 3-form
fluxes. The available options are then only O3-planes or O7-planes/D7-branes. The
first contributes in an obvious way since they are charged oppositely w.r.t. D3-
branes. By contrast, the O7-plane/D7-brane contribution is indirect and involves
an integral over the curvature of the relevant co-dimension-2 submanifold in the 6d
Calabi–Yau. We are not going to spell this out explicitly but only report the results
of a more general analysis:
Type IIB compactifications in the perturbative regime find their non-perturbative
completion in the so-called F-theory models [184] (cf. the brief explanation at the
end of Sect. 5.7 and the reviews [10, 186, 187]). We only recall here that F-theory
models are based on the geometry of an elliptically fibered (roughly torus-fibered)
Calabi–Yau 4-fold. The fibre torus encodes the information that corresponds, in type
IIB language, to the variation of the axio-dilaton S. In fact, S is identified with the
complex structure parameter of the fibre torus.
In this F-theory setting, the tadpole contribution of the O7-planes and D7-branes
above is encoded in the 4-fold geometry, more precisely, in the Euler number χ4 of
the 4-fold. In fact, this result is much more general: It also includes contributions
from 7-branes other than the standard D7-branes of perturbative type IIB string
theory. In our context, the crucial constraint then becomes

χ4 h 01
N N ≤ L ≡
T
with N= and ≡ .
24 f 10
(6.77)
If this inequality is saturated, the fluxes precisely compensate the tadpole induced by
7-branes and O-planes. Otherwise, tadpole cancellation can be achieved by simply
adding an appropriate number of D3-branes.
The key geometric input is the availability of 4-folds with Euler characteristics
up to χ4 ∼ 106 (see e.g. [255]), leading to L ∼ 105. The number of vacua can then
be estimated as [212]
LK
Nvac ∼ , (6.78)
K!
where K is the number of 3-cycles of the Calabi–Yau. This number is crucial in the
present context since it determines the dimension of the flux vector N to be d = 2K.
Thus, the estimate of Nvac above
√ can roughly be understood as the volume of a 2K-
dimensional ball of radius L. This is a natural expectation since we are dealing
with a lattice with unit spacing on which the flux vector can end. Of this lattice,
only a certain subset, specified by the inequality in (6.77), is available. The details
are slightly more
√ complicated since the metric is not positive definite, such that
the ‘radius’ L does not specify a ball but the interior of a hyperboloid (the non-
compact directions of which are, however, cutoff by physical arguments and do not
lead to a divergence of Nvac ).
In the end, using the (far from maximal) numbers L = 104 for the 4-fold Euler
number and K = 300 for the number of 3-cycles of the corresponding Calabi–Yau
orientifold, one arrives at
Nvac ∼ (eL/K)K ∼ 100300 = 10600 . (6.79)
Even after appropriate reductions for the geometric constraints implied by the
gaugino condensation/instanton effect and the warped throat required for the uplift,
this is still more than sufficient to realise the desired fine-tuning for the cosmological
constant of ∼ 10−120. In fact, most naively (ignoring the reduction by geometric
constraints) one expects that of the 10600 vacua about 10480 have a cosmological
constant of the order of 10−120 or below.
At this point, a comment concerning a more recent development in the context of
vacuum counting has to be made. It concerns the number of several hundred 3-cycles
which we used and which is typical for a CY 3-fold. Clearly, an O7-orientifold of a
CY 3-fold has more moduli due to the freedom of deforming the 4 D7-branes that
originally lie on top of each O7-plane. Even more generally, similar situations can
be analysed in the F-theory context, where more types of co-dimension-2 objects
than just O7-branes and D7-planes are available. In this context, the 3-fold complex
structure and D7-brane deformation moduli are unified as complex structure moduli
of the elliptically fibered CY 4-fold. In this F-theory setting, ‘the geometry with
most flux vacua’ (as far as presently known) has recently been identified [256]. The
size of the tadpole is consistent with what was discussed above, but the number K
has now, roughly speaking, to be replaced by the number of 4-fold complex structure
moduli. In the maximal known case, this is h3,1 = 303, 148. The estimate of Nvac
based on the volume of a sphere in flux space is not a good approximation any more.
Instead, a more careful analysis leads to O(10272,000) flux vacua [256]. This exceeds
all 3-fold-based estimates by far.
6.8 Higgs Mass and Other Landscape-Related Issues
Let us now turn away from the need to fine-tune the cosmological constant and
focus on the Higgs mass and hence the electroweak hierarchy problem. The Higgs
mass parameter m2H depends both on the μ term of the MSSM (if this model
arises at an intermediate energy scale) and on the SUSY breaking soft terms.
Moreover, virtually all Standard Model parameters enter through the large loop
effects. (Here we assume that SUSY is broken at a high scale.) All these parameters
come from coefficients of operators in the landscape-derived 4d supergravity model.
In the present course, we have discussed such models including only Kahler and
complex structure moduli. However, matter fields coming from the D-brane sector
can easily be added, and the resulting structure is in principle well-understood (see
e.g. [9, 156, 158, 257–259]). The values of complex structure parameters, which are
the main entities known to ‘scan very finely’ in the landscape, enter in various ways.
In many cases, they govern the coefficients of the effective matter field lagrangian.
As a result of all of this, it is highly plausible that the fine distribution of
landscape points in the complex structure moduli space will, through several
calculational steps, translate into a fine distribution of values of m2H . Moreover, one
expects this distribution to be in no way special (e.g. more dilute) near zero. Now,
the highest SUSY breaking scale conceivable (and hence the highest natural scale
for |m2H |) is MP . Thus, to get down to the weak scale purely by tuning one has to
pay the price of a suppression factor of (100 GeV)2 /(1018 GeV)2 ∼ 10−32. Thus,
even starting with the modest 3-fold estimate of 10600, we would apparently still be
left with
10600 × 10−120 × 10−32 ∼ 10448 (6.80)
vacua with accidentally small cosmological constant and an accidentally light

Higgs in a model with very high (near MP ) SUSY breaking. Even paying some
high additional price for various model building requirements (gaugino condensate,
throats, Standard Model field content and renormalisable couplings of the right
magnitude), there appears to be no problem of finding ‘our world’ in the landscape.
There could of course be the problem that certain features (positive λ, a particular
light fermion spectrum etc.) are simply unavailable. This would clearly invalidate
the simplistic statistical reasoning we presented.
Needless to say, the problem of tuning the Higgs mass to a small value is
alleviated if we also have low-scale (or at least relatively low-scale) SUSY. Such
models with low-scale SUSY may also be available in the landscape. For example,
in the case of the LVS the volume V can be stabilised at an exponentially large value,
leading to a small gravitino mass. Moreover, the soft scale in the visible sector can
6.8 Higgs Mass and Other Landscape-Related Issues 215
be significantly smaller than m3/2 in certain settings [260]. An interesting question

is now whether we are more likely to find ourselves in a world with a purely fine-
tuned light Higgs or a light Higgs mostly due to SUSY (possibly with some extra
tuning in addition). One part of the answer can be given by asking how many of the
respective vacua are available. In other words, is it ‘cheaper’ to directly tune for a
small Higgs mass or to tune for a low SUSY breaking scale? The second option may
be preferred if one could lower the SUSY breaking scale through appropriate model
building choices rather than simply through tuning. Yet another option would be to
look for models with technicolour-like structure (see Sect. 8.1), lowering the Higgs
mass in a non-SUSY-related dynamical way. The above questions have of course
been studied from the very beginning of the string landscape to the present, but no
widely accepted answer has so far emerged. The reader may consult, e.g. [94, 261–
267] and references therein.
Of course, the discussion just started has to remain highly incomplete within the
limited scope of this course. Indeed, not just the SUSY breakingscale but all sub-
topics of string model building or string phenomenology are affected if not governed
by vacuum statistics. This includes finding the right gauge group and matter content
in the zillions of brane or gauge bundle configurations (see e.g. [168, 173, 177]),
understanding statistical aspects of the observed fermion mass patters, a possibly
necessary tuning for a flat inflationary potential (see Sect. 7.1) and many others. Let
us here only emphasise one aspect because it is particularly timely: The apparent
genericity of axions and axion-like particles in the landscape.
As we have explained, in type IIB models Kahler moduli govern brane volumes
and hence the holomorphic gauge-kinetic functions of 4d SUSY gauge theories:
f (T ) ∼ T = τ + ic . (6.81)
This leads to a 4d term ∼ c tr F F̃ , making c a so-called QCD axion in the case

of the SU (3) gauge group of the Standard Model [12, 268]. Axions are the most
plausible solution of the strong CP problem: θQCD becomes a dynamical field and
its potential automatically drives it to zero. Thus, the natural presence of axions
(they also arise if f is governed by other moduli in the IIA and heterotic context) is
a success of the stringy world view. There are also problems in that the axion mass
tends to be too high in string models [269, 270].
The string landscape adds an important aspect to this topic. Namely, it can be
shown that many, in fact possibly extremely many, periodic pseudo-scalars of the
type of c arise in some string compactifications. This idea of the so-called ‘String
Axiverse’ has been introduced in [271]. The possible existence of such axion-like
particles (ALPs) and other super-light fields has also been studied independently
of the string landscape as well as in connection with it, see e.g. [272, 273]. A
particularly interesting aspect of ALPs is that they can play the role of dark
matter, one of the central puzzles of beyond-the-standard-model physics. In this
context, string theory has more to contribute than just axions since, quite generally,
string theory compactifications tend to predict light sectors in addition to the field
content of the Standard Model. In terms of the cosmological evolution, one then
expects the so-called Remnants (see e.g. [274]). This is yet another aspect of string
phenomenology which is strongly influenced by what we think is typical in the
landscape.
However, vacuum counting alone is not sufficient to settle all the interesting
questions above. Indeed, it is possible that many more vacua with low-scale SUSY
rather than with purely fine-tuned non-SUSY light Higgs are available. But this
would become irrelevant if cosmological dynamics prefers inflation to always end
in vacua with high-scale SUSY breaking. Thus, we need to turn to the dynamics that
might be responsible for populating the landscape.
6.9 Problems
6.9.1 No-Scale Kahler Potentials and KKLT
Task Using the general supergravity formulae given earlier in the course, calculate
the scalar potential of a one-field supergravity model with

K(T , T ) = − ln (T + T )n and W = W0 = const. (6.82)
Which striking feature arises if n = 3 ? Try to generalise this special result to the case
of m variables, with e−K being a general homogeneous function of the variables
ı
(T i + T ) of degree n.
Returning to the single-modulus case, analyse the so-called ‘KKLT potential’
arising from the superpotential W = W0 + Ae−aT for n = 3. Use the notation
T = τ + ic, set A = a = 1 for simplicity and assume |W0 | 1. To draw a
qualitative plot of V (τ ), after minimising in c, it is sufficient to understand the
qualitative behaviour of V in the two regimes |e−T | |W0 | and |e−T | |W0 |.
Throughout, assume τ 1.
Hints The first part is completely straightforward. For the general case, it is useful
ı
to prove the relation (T i + T )Ki = −n and consider its derivatives.
The discussion of the KKLT potential is a straightforward exercise in paramet-
rically analysing a given function. Note that, in the second regime, you also need
to assume that the axionic variable Im T = c takes the value minimising the scalar
potential. The result is shown in Fig. 6.7.
Solution First, we have

−n n
K = −n ln(T + T ) , KT = KT = , KT T = = (K T T )−1 ,
T +T (T + T )2
(6.83)
6.9 Problems 217
and hence
V (T , T ) = eK (K T T |KT W0 |2 − 3|W0 |2 ) = eK |W0 |2 (n − 3) . (6.84)
We see that for n = 3 the potential vanishes identically, implying that T remains
a modulus in spite of W = 0. This is the simplest form of the famous no-scale
cancellation.
Now consider the multi-variable case, with
1 k
K = − ln f (T 1 + T , · · · , T k + T ) (6.85)
and
1 k 1 k
f (α(T 1 + T ), · · · , α(T k + T )) = α n f (T 1 + T , · · · , T k + T ) , (6.86)
as proposed. By Euler’s homogeneous function theorem, we have
(T i + T )∂i (e−K ) = n e−K ,

ı
(6.87)
and hence
ı
(T i + T )Ki = −n . (6.88)
j
Differentiation w.r.t. T gives
Kj + (T + T )i Kij = 0 , (6.89)
where we used Kj = Kj . Multiplying by the inverse metric, one obtains
K ij Kj + (T + T )i = 0 , (6.90)
and after further multiplication by Ki and application of (6.88),
Ki K ij Kj = n . (6.91)
Now one immediately finds the multi-variable result
V = eK (K ij (Ki W0 ) (Kj W 0 ) − 3|W0 |2 ) = eK |W0 |2 (n − 3) . (6.92)
Finally, we turn to the discussion of the model with n = 3 and superpotential

W0 + e−T . In the first regime, Re T ln(1/|W0 |), we may set W e−T . Then
the second scalar potential term, 3|W |2 , is suppressed by two powers of the large
quantity τ with respect to the F -term squared. Similarly, DT W ∂T W . Hence,
1 e−2τ
V eK K T T |∂T e−T |2 ∼ |e−T |2 ∼ . (6.93)
T +T τ
This is positive and monotonically falling.

In the second regime, Re T ln(1/|W0 |), the naively leading term is obtained
by setting W = W0 . But this vanishes by the no-scale property. Hence, we need
to consider the formally subleading terms, which involve one power of W0 and one
power of e−T . Such terms, ∼ W0 e−T , appear both in the F -term squared and in
−3|W |2 . But the second contribution suffers a relative suppression by one power
of the large quantity τ . (This is due to the enhancement of the F -term squared by
K T T , which is only partially compensated by KT .) Thus, we find
e−τ e−τ
V eK K T T (∂T e−T ) KT W 0 + h.c. ∼ 2 |W0 | ei(c+Arg W0 ) + h.c. ∼ − 2 |W0 | .
τ τ
(6.94)
In the last step, we assumed that c takes the value minimising cos(c + Arg W0 ) at
minus unity. We see that, at large τ , V is negative and approaches zero from below.
Our two results for large and ‘small’ (still much larger than unity) values of τ
guarantee the presence of a local minimum at negative value of V and with τ ∼
ln(1/|W0 |).
6.9.2 The LVS Scalar Potential and Stabilisation Mechanism
Task Derive the formula for the LVS scalar potential and the two-step stabilisation
procedure as discussed in the main text.
Hints Correct the supergravity formula for the F -term potential in two ways: First,
by the instanton effect in W (you will need to keep the leading and subleading terms)
and then by the α effect in K. From the latter correction, only the leading term is
needed, but to derive it, some algebra along the lines of the multi-field derivation
of the no-scale potential in Problem 6.9.1 is required. Basically, one has to correct
this analysis by the no-scale breaking α effect. Adding these two corrections, one
obtains the desired scalar potential for V and τs . The rest is simple algebra and
elementary parametric analysis.
Solution The starting point is the basic supergravity formula

V = eK K ij (Ki W ) (Kj W ) − 3|W |2 , (6.95)
6.9 Problems 219
with i and j labelling the Kahler moduli Tb and Ts . Let us, for the moment, ignore
the α correction and focus only on the effect of the instanton term Ae−aTs that
is added to W0 . In its absence, V would be identically zero. A non-zero result
arises only because the Ts derivative applied to W gives a non-zero value. There
are precisely three such terms, giving

δV1 = eK K ss a 2 |A|2 e−2aτs − K sj Kj aAe−aTs W 0 + c.c. , (6.96)
where we disregarded the subleading instanton correction to W in the square bracket

and made use of the fact that the derivatives of K are real in our approximation.
Employing (6.90), this becomes

δV1 = eK K ss a 2 |A|2 e−2aτs + 2τs aAe−aTs W 0 + c.c. , (6.97)
which we can now easily minimise w.r.t. the imaginary part cs of Ts = τs + ics :

δV1 = eK K ss a 2 |A|2e−2aτs − 4τs a|A||W0|e−aτs . (6.98)
To determine K ss , one has to invert the matrix

Kbb Kbs
. (6.99)
Ksb Kss
As explained in the main text, stabilisation will eventually occur in the regime of
exponentially large τb and modestly large τs . Thus, it will be justified a posteriori
that we use the relations τb τs 1 in the present analysis. One then finds that
Kss Kbb Kbs = Ksb . This implies
−1
3k
K ss (Kss )−1 √ (6.100)
8V τs
and

8 √ 2 2 −2aτs
δV1 = eK V τs a |A| e − 4τs a|A||W0 |e−aτs . (6.101)
3k
We note, in passing, that this argument can be easily rerun with several rather than
just one ‘big’ Kahler modulus. In this case one has to invert a block-diagonal matrix
with the above hierarchies characterising the different blocks.
When calculating the α correction, we may replace W by W0 since we are not
interested in quantities that are doubly small. The correction arises since the no-scale
cancellation analysed in Problem 6.9.1 fails. It is hence quantified by
, -
δV2 = eK K ij Ki Kj − 3 |W0 |2 . (6.102)
Moreover, we have
K = −2 ln(V + x) + · · · with x ≡ ξ/2 (6.103)
and with the ellipsis standing for Kahler moduli independent terms. Thus,
2Vi V x
Ki = − = Ki(0) Ki(0) 1 − , (6.104)
V +x V +x V
where K (0) is the uncorrected Kahler potential. With this, (6.88) takes the form
ı
x
(T i + T )Ki = −3 1 − . (6.105)
V
One may now follow the steps that lead to the no-scale result (6.91) with n = 3. The
corrected formula turns out to be

x ij 3x
Ki K Kj − 3 1 −
ij
= Ki K . (6.106)
V V j
The term on the r.h. side can be rewritten according to

3x 3x 3x
Ki K ij
= −Ki K ij Vj · = Ki K ij Kj · . (6.107)
V j V 2 2V
Here we have disregarded the difference between K and K (0) since the whole term is
subleading in the 1/V expansion (see [246] for more complete results). Combining
this with (6.106), one finds, again at leading order in the large volume expansion,

x ξ
Ki K ij Kj = 3 1 + = 3 1+ . (6.108)
2V 4V
This completes the determination of δV2 . It remains to check that δV1 + δV2
corresponds precisely to the scalar potential of (6.66).
Next, one may simultaneously minimise (6.66) in V and τs (see the original
papers [193, 242] and the appendix of [275], which we mainly follow in this
problem). But it may be simpler and more intuitive to adopt the EFT logic of
integrating out the heavier field τs first. For this, one may focus on the first two terms
of (6.66) only. Disregarding terms suppressed by 1/τs , one finds that the minimum
is approximately at
√
3k|W0 | τs
e−aτs . (6.109)
αV
6.9 Problems 221
A good approximation for τs is obtained by disregarding the non-exponential τs

dependence and then solving the equation:

1 αV
τs ln . (6.110)
a 3k|W0 |
The effective potential for V is now obtained by first replacing the exponentials
exp(−aτs ) in (6.66) according to (6.109) and then using (6.110) for the non-
exponential τs terms. The result is (6.67).
The approximate minimisation of (6.67) is an easy task: First, note that as
V → ∞, the second term dominates and the potential approaches zero from
below. Second, as V falls, the logarithm becomes small enough for the first term
to dominate—the potential becomes positive. Thus, the minimum occurs when the
two terms are approximately equal, i.e. at

ξ αV
a ln . (6.111)
2k 3k|W0 |
Together with (6.110), this confirms the last two relations of the LVS section of the
main text.
Eternal Inflation and the Measure Problem
7
7.1 From Slow-Roll Inflation to the Eternal Regime
The present course does, of course, assume General Relativity as a prerequisite.

Since most relativity courses include some cosmology, it appears logical to assume
that the reader will also be familiar with the most basic cosmology-related formulae.
A selection of relevant textbooks is [2, 52–55, 276–279]. We only summarise the
results to set our notation:
The cosmological principle, with excellent support from data, postulates that
space is homogeneous and isotropic on large scales. Together, these two features
imply that spacetime can be represented as a one-parameter family of homogeneous
spatial hypersurfaces Ht (with t ∈ R) which are threaded orthogonally by ‘observer
curves’. Each of those is parameterised by the observer eigentime t. In terms of the
4d metric, this means
ds 2 = −dt 2 + a 2 (t) gij dx i dx j , (7.1)
where a is the scale factor and gij is the metric on a maximally symmetric 3d space,
i.e. on a sphere, on flat R3 , or on a 3d hyperboloid.
In the simplest case, matter comes in the form of a perfect fluid,
Tμν = ρ uμ uν + p (gμν + uμ uν ) (7.2)
with density ρ and pressure p. Then the Einstein equations and the continuity
equation reduce to
3MP2 (H 2 + k/a 2 ) = ρ (7.3)

ρ̇ + 3H (ρ + p) = 0 , with the Hubble parameter H = ȧ/a . (7.4)

224 7 Eternal Inflation and the Measure Problem
Here k = +1, 0, −1 distinguishes the three cases of positive, zero and negative
spatial curvature.
A case of particular interest is that of a scalar φ with potential V (φ). Using the
standard result ρ = T + V and p = T − V (with T = φ̇ 2 /2), one then immediately
sees that (7.4) takes the form
φ̈ + 3H φ̇ + V = 0 . (7.5)
Standard slow-roll inflation [280–285] arises in the regime where the potential
V is sufficiently flat. This is conventionally quantified by requiring smallness of the
two slow-roll parameters (MP = 1 here and below):
2
1 V V

≡ 1 and η≡ 1. (7.6)
2 V V
Indeed, in this regime φ̈ can be neglected in the equation of motion for φ and ρ is
dominated by the potential energy. Thus, cosmology is described by
3H φ̇ = −V with H 2 = V /3 and a = exp(H t) . (7.7)
Here we have disregarded the curvature term k/a 2 since it anyway quickly becomes
subdominant as a grows while H remains approximately constant. This represents
a so-called quasi-de Sitter situation, exact de Sitter expansion corresponding to an
exactly constant (rather than slowly changing) H in the last equation of (7.7).
In standard inflationary cosmology one assumes that this situation lasts long
enough to explain the flatness and homogeneity of our present-day universe. But
eventually it ends since φ rolls into a region where the slow-roll conditions cease to
hold. In the simplest case, φ oscillates about its minimum and eventually decays to
Standard Model particles, reheating the universe (cf. Fig. 7.1).
Crucially, while in the slow-roll regime, φ does not only roll classically but is, at
the same time, subject to quantum fluctuations. To understand this qualitatively, it
is useful to consider the simplified case of pure de Sitter (V (φ) = const. and hence
H = const.). It is then easy to determine the inward-going geodesics in the relevant
metric (k = 0 for simplicity)
ds 2 = −dt 2 + e2H t dx 2 . (7.8)
One finds that, above some maximal radius r0 (with r 2 ≡ dx 2 ), they never reach the
origin. In other words, there exists a cosmological horizon. Its size is of the order
of the de Sitter radius 1/H . Each spatial slice falls into many so-called de Sitter
patches, which are causally disconnected. As the universe evolves, the exponential
expansion increases their number by e3 in a Hubble time tH ≡ 1/H .
At the moment, all we need to conclude from the above is that a single geodesic
observer, who by definition sits in the centre of his or her de Sitter patch, is
7.1 From Slow-Roll Inflation to the Eternal Regime 225
surrounded by a horizon. This observer sees space expand and sees objects disappear
forever as they cross the horizon. To be precise, the observer stops seeing them
before they cross the horizon due to the diverging red-shift which affects their
radiation (if emitted backward towards the central observer). We assume that the
reader has at least some rudimentary familiarity with similar horizons and similar
physical situations, either in the case of the Unruh effect (eternally accelerated
observer, seeing an Unruh horizon) or in the black hole case (static observer near a
black hole, seeing the black hole horizon). In both cases, the observer is subject
to an approximately thermal radiation coming from the horizon. Its most naive
explanation is virtual pair production, with one particle disappearing behind the
horizon and the other hitting the observer’s detector. This also happens in the de
Sitter case and, for lack of another dimensionful parameter (the Planck scale can be
taken to infinity at fixed H ), we have T ∼ H . This thermal radiation affects φ and,
again on dimensional grounds, induces a random fluctuation δφ ∼ H after each
time interval δt ∼ 1/H .
This allows us to delineate the boundary between the regimes of eternal and
slow-roll inflation for a single scalar in a flat potential. Indeed, for the classical
evolution to dominate the slow-roll displacement of φ should be larger than its
random fluctuation during an interval δt. Thus, according to (7.7)
|φ̇roll | δt ∼ V /H 2 |δφfluct. | ∼ H (7.9)
is the condition for the rolling to win over the quantum diffusion which, as we
explained, √ is an unavoidable feature of de Sitter space. We may rewrite this as
V /V V or
V . As a side remark we note that for purely quadratic
potentials, V = m2 φ 2 /2, this condition becomes 1/φ 2 m2 φ 2 . Hence, for
sufficiently small m it can be violated at large φ while still in the controlled
regime of small energy density, m2 φ 2 1. This is the famous model of ‘chaotic
inflation’ [286], which allows for eternal inflation, slow-roll inflation and reheating
all to occur in different regions of the same simple, power-law potential.
In our more contrived potential of Fig. 7.1, the above is clear even without
any calculation: Continuing the potential to the right such that it becomes more
and more flat, it is apparent that all regimes exist: in the very flat region at large
φ, one basically has pure de Sitter. The field fluctuates as described above and,
Fig. 7.1 Slow-roll inflation

ending in field oscillations
and reheating
very occasionally, enters the intermediate regime where slow-roll dominates over
fluctuations. In this region, while the universe still keeps exponentially expanding,
the field now systematically rolls to the left, eventually reheating at φ near its
minimum. As a result, a universe like our own forms in a small ‘pocket’ inside
the vast region of continued eternal expansion. In fact, infinitely many such pocket
universes will form in the underlying infinite (approximate) de Sitter space with its
fluctuating scalar φ.
The interested reader may want to consult [287] and [276, 288] for many
more details of de Sitter space and slow-roll inflation, respectively. For inflation
specifically in the stringy context, see [289, 290]. Moreover, the earlier review of
string cosmology in [291] includes the discussion of some more exotic alternative
possibilities.
7.2 Eternal Inflation in the Landscape
For the purpose of these notes, we take slow-roll inflation in our past to be an
essentially established observational fact. This may be justified on the basis of
the excellent fit of its predictions for curvature fluctuations to cosmic-microwave-
background or CMB data [292].1 The previous section described how, in a simple
single-field model, the epoch of slow-roll inflation in our universe can originate
from an eternally inflating universe. While this is appealing from the perspective of
solving (at least part of) the initial-condition problems, it is not the most obvious
or most common way in which slow-roll inflation relates to the string landscape as
it is presently understood. Indeed, in our present understanding de Sitter vacua are
rare. Solutions with a positive energy density and a very flat potential, as required
for slow roll, are more rare. Finally, solutions where the flatness is sufficient for the
eternal regime (viz. Fig. 7.1 with V becoming more and more flat at very large φ)
are the rarest of all.
However, eternal inflation may arise in a much more natural, maybe even
unavoidable way in a universe based on the string landscape. To see this, let us
step back and forget for the moment about the phenomenological requirement
of slow-roll inflation. Instead, focus on what the string landscape most naively
predicts: It contains N = 2 SUSY Minkowski vacua (e.g. from simple Calabi–Yau
compactifications) as a very well established feature. Moreover, there are N = 1
AdS vacua (like in KKLT before the ‘uplift’), a feature that I would call established,
although maybe with less mathematical rigour: One needs to combine Calabi–Yau
1 Alternatives range from modifications of the simple slow-roll dynamics described earlier (see
e.g. [293–295]) to entirely different scenarios like string-gas or pre-big-bang cosmology [296,297].
7.2 Eternal Inflation in the Landscape 227
Fig. 7.2 A simple

visualisation of the landscape
over a 1-dimensional moduli
space. Some possible down
and up-tunnelling processes
are illustrated
string geometry with instantons (or the 4d non-perturbative phenomenon of gaugino

condensation) and the fine-tuning of W0 .2
Finally, and this is the most important category of vacua for the purposes of the
present discussion, there are presumably metastable de Sitter vacua of the KKLT
or LVS type discussed earlier (or some variants thereof). Their existence had been
widely accepted since KKLT [192], but has then more recently been challenged on
fundamental grounds [217, 218, 302, 303]. This has led to a heated debate in the
framework of the ‘Swampland paradigm’, to which we will return in Sect. 8.4. For
now let us press ahead under the assumption that string theory has metastable de
Sitter vacua. In fact, while other options have been discussed (cf. Sect. 8.4), such
dS vacua are presumably our best hope for relating string theory to the real world.
Also, it has to be noted that no strong technical reasons have so far been provided
for why KKLT or some variant thereof should not work.
We will ignore N = 2 vacua since they are presumably irrelevant for the real
world. Thus, we have AdS and (with the caveats above) dS vacua, which can be
visualised as points in a high-dimensional space. This space is parameterised by
the moduli of the Calabi–Yau which, due to the non-zero scalar potential, have of
course now ceased to be moduli and are simply scalar fields. Replacing the multi-
dimensional moduli space by a single real line, the set of AdS and dS vacua may
be visualised as in Fig. 7.2. This figure or an analogous picture with a ‘mountain-
range potential’ over a 2-dimensional plane, is used widely to illustrated the string
landscape. But one should always remember that this simple scalar field theory with
a potential V and many local minima (vacua) remains only a model.
2 Most probably there also exist non-SUSY AdS solutions. It has been conjectured that those
can only be metastable [298, 299]. They would then not deserve the name vacua since, if AdS is
metastable, any observer in it will encounter the decay after a time interval comparable to the AdS
radius. Cosmologically, such ‘vacua’ can nevertheless occur, but they appear not to add anything
new to our discussion at a qualitative level and we will hence ignore them.
This may also be a suitable place to note that, if one does not insist on the AdS curvature scale
being parametrically below the KK scale, the existence of SUSY AdS vacua would be as certain as
that of SUSY Minkowski vacua. One example are compactifications of type IIB on an S 5 carrying
F5 flux to 5d AdS. This compactification may be the best established of all due to its possible
fundamental definition via AdS/CFT [300, 301]. However, we are here interested in EFTs in the
non-compact dimensions and we hence insist on the scale separation between AdS curvature and
KK scale.
Concretely, the reader should keep in mind that apart from the reduction of the
dimensionality of field space for the purpose of drawing, there are at least two
further (over)simplifications hidden in this picture:
First, the transition between two vacua has nothing to do with climbing a smooth
barrier in moduli space, at least this is not the generic case. Generically, two different
vacua are associated with different 3-form flux and are hence separated by, e.g., a
D5-brane or NS5-brane wrapping a 3-cycle of the compact space and representing
a domain wall in the non-compact 4d spacetime (cf. Sects. 6.1 and 6.2). As a result
of the flux change between the two sides of the domain wall, the values at which the
moduli are stabilised also change. Hence the picture of different minima at different
φ values is actually reasonable. Just the possibility of rolling over the smooth barrier
must be replaced by more general tunnelling transitions.
Second, the full string theory landscape does of course contain different Calabi–
Yau geometries with different topologies. Dynamical transitions between some of
them are possible and well-understood in many simple cases, (see e.g. [142]) but
it is in fact conjectured that all Calabi–Yau moduli spaces are, in a very well-
defined mathematical sense, part of a single space. This is sometimes referred
to as ‘Reid’s fantasy’ [304]. What happens to this statement in the case where
one involves orientifolding, F-theory constructions or even compactifications of
different perturbatively defined string theories together with M-theory, is less clear.
But it is expected that all of them, even compactifications to spacetime dimensions
different from four, are part of the same dynamical structure and tunnelling
transitions between all the vacua are possible (see e.g. [305] for a recent discussion).
This, of course, makes Fig. 7.2 an enormous oversimplification. In fact, one should
think of many such pictures, with different field-space dimensions of φ, glued
together.
Nevertheless, let us stick to the simple picture of scalar field minima separated by
potential barriers. Each of the minima at field values φi has a different cosmological
constant λi = V (φi ). If at least one of those minima has λi > 0 and if the probability
T for tunnelling out of this minimum (per volume and time) is smaller than the
fourth power of its expansion rate,
T H 4 (λ) , where 3MP2 H 2 (λ) = λ , (7.10)
this already gives rise to eternal inflation. Condition (7.10) may roughly be
understood as follows: It requires the density (in 4-volume) of nucleation points
of bubbles of other vacua to be smaller than the typical scale H of the underlying
dS space. There is then no danger that the loss of volume to other vacua wins over
the volume growth due to de Sitter expansion, which is governed by H .
In the generic case there is more than one dS vacuum. There are then obviously
tunnelling processes where a bubble of the energetically favoured, lower-lying de
Sitter state is nucleated within a certain higher-energy de Sitter background. This
could for example be the regions labelled ‘1’ and ‘2’ in the schematic (Penrose-
type) diagram in Fig. 7.3. We have called the presence of such tunnelling events
‘obvious’ since it corresponds to the familiar process of bubble nucleation in first-
7.2 Eternal Inflation in the Landscape 229
Fig. 7.3 Nucleation and

speed-of-light expansion of
bubbles in a ‘background’ dS
vacuum. The ‘cutoff surface’
will be discussed later
order phase transitions, where a bubble of the phase with smaller free energy
nucleates in the phase with higher free energy in which the system is started. As
a less obvious fact, in the cosmological context of tunnelling between different de
Sitter vacua the inverse process is also possible. In other words, a bubble of the
energetically disfavoured, higher-lying de Sitter can be nucleated and grow in a
low-lying background. In terms of Fig. 7.3, this may for example be the nucleation
of phase ‘1’ inside ‘3’ together with nucleation of ‘3’ inside ‘1’ on the r.h. side of
the figure. The process of ‘up-tunnelling’ is strongly suppressed, i.e. much more
rare than that of ‘down-tunnelling’. We will return to this at the quantitative level.
For now, suffice it to say that the surprising fact that up-tunnelling is possible at all
can be understood as a result of the exponential expansion of the background: While
energetically the disfavoured bubble wants to shrink, for large enough bubbles the
background expansion wins and ‘pulls’ the bubble to larger size in spite of the
apparent increase of (non-gravitational) energy associated with this. Recall that
energy conservation is anyway not (at least not in the usual, straightforward way)
a condition that can be used to constrain the allowed dynamical evolution in the
general-relativistic context.
What is crucial for us at the moment is that, due to both up and down-tunnelling,
the whole landscape gets populated once eternal inflation is running, i.e. once a
single Hubble-sized patch of any of the de Sitter vacua exists. The continued process
of the nucleation of bubbles within bubbles within bubbles sketched in Fig. 7.3
is oversimplified since there are also AdS vacua. Nucleation of corresponding
bubbles leads locally to ‘big crunches’ since the energy density imprinted in such
bubbles by the dynamics of bubble nucleation grows. This is clear from the fact
that one may think of AdS as of a contracting space. This contraction does not
interfere with empty AdS being a perfectly consistent solution of Einstein equations,
yet it unavoidably leads to a crunch if a homogeneous energy density is present.
Nevertheless, for appropriate tunnelling rates the continued appearance of such
‘terminal vacua’ does not stop the overall process of eternal inflation. Finally, we
note that the nucleation of Minkowski-space bubbles is in principle also possible.
While these do no crunch, they also decouple from the eternal inflation process since
no new de Sitter bubbles can be nucleated within them. The reason is conventional
Minkowski-space energy conservation. We have not displayed AdS and Minkowski
regions for simplicity.
7.3 Tunnelling Transitions in Quantum Mechanics
To put some meat on the largely qualitative discussion of the previous section, it is
useful to understand the calculation of tunnelling rates between the different vacua.
Let us start the discussion with tunnelling in quantum mechanics, for which there
are references at the elementary textbook level. A particularly useful analysis, taking
the reader all the way from quantum mechanics to the field theory case, is [195].
In quantum mechanics, one of the simplest cases is that of tunnelling in
the degenerate double well potential, cf. the l.h. side of Fig. 7.4. The reader is
presumably familiar with the standard WKB calculation which shows that a particle
with mass m and energy E hitting a generic barrier from the left has a non-zero
transition probability governed by an amplitude

A ∼ exp − dx 2m(V (x) − E) . (7.11)
Here the integration extends over the classically forbidden part of the barrier.
In the case of the double well, the relevant question to ask is that about the
amplitude for a particle, originally localised on the l.h. side in the ground state |a,
to be observed in the ground state |b after some time T . Assuming the ground-state
energy is small compared to the potential height, the answer is simply

A ∼ T exp − dx 2mV (x) . (7.12)
The exponential follows from (7.11) by neglecting E. One may argue for the
prefactor T using the following toy-model: Consider the ground states on the left
and right as a two state system: { |a , |b }. The Hamiltonian clearly has a small off-
diagonal term, suppressed by the small exponent in (7.12). It is then immediately
clear that, at leading order in this exponent, the transition amplitude between |a and
|b must be linear in T . (The reader is invited to think about this amplitude more
carefully for T → 0 and T → ∞, where it turns out that the result is modified. See
also the discussion below.)
Fig. 7.4 Degenerate double well potential in quantum mechanics and the corresponding inverted
potential relevant for the equation of motion of the euclidean theory
7.3 Tunnelling Transitions in Quantum Mechanics 231
Fig. 7.5 Classical solution

of the euclidean theory, which
corresponds to a solution of
the original theory with
inverted potential. The large
initial and final times −T /2
and T /2 are outside the
plotted time range
Coleman [195] presents a very elegant path integral derivation for this result by
considering the classical solution dominating the euclidean amplitude at very large
T . This euclidean amplitude reads
x(T/2)=b x(T/2)=b T /2 m
−H T −S
b| e |a ∼ Dx e ∼ Dx exp − dt ẋ 2 + V (x) .
−T /2 2
x(−T /2)=a x(−T /2)=a
(7.13)
The path integral is dominated by the extremum of the euclidean action. Quite
generally, the latter is given by the solution of the corresponding classical mechanics
problem in the inverted potential, V → −V . For infinitely large T , this solution
consists of the following: a particle starts at t = −∞ and with zero velocity on the
maximum at a. It then very slowly accelerates and rolls through the minimum at
x = 0 to the maximum at b, where it comes to rest at t = +∞.
At finite T , this process ceases to be an exact solution, but it is still an
approximate one. More precisely, there is a continuous infinity of such approximate
solutions, parameterised e.g. by the time t0 at which they cross x = 0, cf. Fig. 7.5.
Hence3
T /2
−H T
b| e |a ∼ dt0 e−Stunnel (t0 ) ∼ T e−Stunnel . (7.14)
−T /2
In going from the next-to-last to the last expression, we have disregarded the t0
dependence of the classical action St unnel . Note that the final formula manifestly
agrees with (7.12) if one evaluates the action using energy conservation.
It is an easy exercise to repeat the analysis above, allowing for any number
of ‘instanton’ transitions between a and b. Clearly, to contribute to the amplitude
3 Here we introduced an additive renormalisation of the original Hamiltonian by the oscillator

ground-state energy ω/2, where ω is the frequency for small oscillations around the minima (we
assume ωa = ωb ). Otherwise, an additional factor exp(−ωT /2) would appear on the r.h. side.
Fig. 7.6 On the left: Quantum mechanical potential allowing for a decay of a potentially long-
lived ‘ground state’ in the minimum at x = a. On the right: The corresponding inverted potential
b| e−H T |a this number must be odd, leading to the Taylor series for the hyperbolic
sine:

b| e−H T |a ∼ sinh KT e−Stunnel . (7.15)
Here K is a dimensionful prefactor of the amplitude which we suppressed in the

preceding discussion. The analytic continuation to the physically relevant real-time
case is also immediate:

b| e−iH T |a ∼ i sin KT e−Stunnel . (7.16)
Now we turn to the more interesting case where a state, originally in the
minimum at a, can decay to a negative-energy region which opens up to the right of
the point x = b (cf. Fig. 7.6). Given the double well analysis above, the most naive
guess is that the amplitude for a state to decay after a time T is again (at leading
order in T )
A ∼ T e−Stunnel . (7.17)
This time, St unnel is the action for a process where the particle starts, in the inverted
potential −V , on top of the maximum at a and then runs through the minimum
to arrive at b. Beyond b, there is no tunnelling suppression so one expects only
the potential in between the points a and b in Fig. 7.6 to contribute. This guess is
correct, but will not be very useful for going to the field theory and gravity case
later on. The problem is the abrupt and poorly defined way in which our calculation
‘ends’ at the point b, where the particle returns from tunnelling and becomes a real
particle rolling down a potential.
It is more effective to consider the euclidean classical solution in which the
particle starts at a at t = −T /2, rolls to b, and then returns to a at t = +T /2.
This process, called a bounce for obvious reasons, calculates a contribution to the
7.3 Tunnelling Transitions in Quantum Mechanics 233
amplitude
a| e−H T |a = e−E0 T + KT e−Sbounce . (7.18)
Here E0 is the leading order or perturbative energy of |a. It may, as before, be

set to zero by an appropriate redefinition of the Hamiltonian. Our interest is in the
exponentially small (non-perturbative) correction on the r.h. side of (7.18). This
correction comes from the fact that the potential has a second zero at x > 0, leading
to the existence of the classical bounce solution. The correction allows for a peculiar
interpretation. To understand this, it is important to note that the prefactor K is
determined by fluctuations of x(t) around the classical solution. More precisely, one
writes x(t) = xclass.(t) + δx(t). At leading order in δx(t), the action then becomes

m d2
S Sbounce − dt δx(t) + V (x(t)) δx(t) . (7.19)
2 dt 2
Expanding δx in the complete set of eigenfunctions of the differential operator in

(7.19), one finds
/
K∼ λi . (7.20)
i
Thus, if one of the eigenvalues λi is negative, K becomes imaginary and one sees
that exp(−Sbounce ) does in fact govern the size of an imaginary correction to E0 .
In other words, it actually governs the decay rate of the state |a, which is precisely
what we want.
To complete this argument, we need to convince ourselves that one λi is indeed
negative, in other words, that our bounce solution has a negative mode. This can be
shown rather generally, but a simple, non-rigorous argument is as follows: Consider
first the tunnelling solution between two vacua |a, |b. This cannot have a negative
mode since, intuitively, it is simply the optimal (i.e. with smallest euclidean action)
path between these vacua. Any deformation makes the action larger. By contrast,
the bounce action can clearly become smaller if one deforms x(t) appropriately. To
be specific, thinking in terms of a particle rolling in the inverted potential in Fig. 7.6,
we may slow this particle down slightly more than in the classical solution when it
approaches the turning point at b. It then never reaches b and returns prematurely,
leading to a smaller action. For more careful arguments, both in the present quantum
mechanical model and in the field theory case of the next section, see e.g. [195].
To summarise, we have learned that up to non-exponential effects, the decay rate
through a barrier as in Fig. 7.6 is
∼ exp(−Sbounce ) , (7.21)
where Sbounce is the action for a classical bounce in the inverted potential. We finally
note that Sbounce = 2St unnel and ∼ |A|2 such that the analysis based on the
bounce is fully consistent with the naive guess of (7.17).
7.4 Tunnelling Transitions in Field Theory
Before attempting to take this to the gravitational case, which is relevant for
populating the landscape, it is useful to understand the generalisation to flat-space
quantum field theory. As already in the previous section, we refer to [195] and the
original papers listed therein for a more detailed treatment.
Maybe the simplest example is that of a real scalar φ with
1
L= (∂φ)2 − V (φ) , (7.22)
2
where V is of an asymmetric double well type, as sketched in Fig. 7.7. The decay
of the metastable or false vacuum at φ = 0 to the stable or true vacuum at φ = φ1
proceeds through bubble nucleation. To understand this, we need to first understand
possible bubbles of the true vacuum inside of an infinitely extended false vacuum.
For this, let us start with the bubble wall, focussing first on the simpler case where
the two minima in Fig. 7.7 are degenerate. Moreover, let us look for a solution where
two half-spaces, say with x 1 0 and x 1 0 (and for {x 2, x 3 } ∈ R2 ) are in the false
and true vacuum, respectively. More precisely, a solution with field profile φ(x 1 )
will exist such that φ(−∞) = 0 and φ(+∞) = φ1 , with the transition occurring
mainly in the vicinity of the plane x 1 = 0. Let us define the energy per unit area of
this configuration as the bubble wall tension T . The above stationary solution will,
of course, cease to exist in the presence of an asymmetry
V ≡ V (0) − V (φ1 ) (7.23)
between the two minima. Nevertheless, there will be an alternative solution in

which the domain wall is accelerating under the pressure of the energy density
difference V between l.h. and r.h. half-space. This solution will allow for an
effective description in terms of a domain wall with a certain tension T . This latter
Fig. 7.7 Scalar field double

well potential with false
vacuum at φ = 0 and true
vacuum at φ = φ1
7.4 Tunnelling Transitions in Field Theory 235
concept will hence continue to be well-defined even in the case of an asymmetric

double well.
Next, we note that the presence of a true vacuum bubble of radius R in a false
vacuum background leads to two energetic effects: On the one hand, there is an
energy gain since the volume of the bubble is at a lower energy density. On the
other hand, the bubble wall with its positive tension comes with an energetic cost.
The total energy of a bubble of radius R is hence
4
EB (R) − πR 3 V + 4πR 2 T . (7.24)
3
Here we used the so-called thin-wall approximation, i.e. the assumption that R is
much larger than the bubble wall thickness. The latter is defined as the typical length
interval inside which most of the gradient energy of the bubble wall is localised. The
so-called critical bubble radius Rc is defined by EB (Rc ) = 0. It is clear from (7.24)
that
Rc = 3T /V . (7.25)
The critical bubble sits at the boundary between the regimes of small bubbles, which
recollapse under the bubble wall tension, and large bubbles, which grow indefinitely
under the pressure induced by V .
We note in passing that the concept of a critical bubble should be familiar
from first-order thermal phase transitions. Here, for example in an overheated fluid,
bubbles of all sizes continuously form due to thermal fluctuations. Supercritical bub-
bles then grow and lead to the emergence of extended regions of the energetically
favoured (in this case gaseous) phase. In our zero-temperature context bubbles of
different sizes form due to quantum fluctuations.
The basis for a quantitative understanding of the resulting false vacuum decay
rate has been laid in the previous section on tunnelling in quantum mechanics.
We can now easily identify the analogue of the quantum mechanical potential of
Fig. 7.6. The role of the variable x, plotted on the horizontal axis, is played by
the bubble radius R. The potential plotted vertically is now replaced by EB (R).
This function first rises since, for small R, a function ∼ R 2 is always larger than a
function ∼ R 3 . Then, of course, R 3 eventually wins and EB becomes negative as R
passes the critical radius Rc .
This allows us to describe the field-theoretic analogue of the bounce as a process
in R4 , the euclidean version of R1,3 : A small bubble emerges at some point (t, x),
grows to critical radius Rc , and then shrinks again to zero at some later euclidean
time. Topologically, this clearly corresponds to a 4-dimensional ball of true vacuum
inside a false vacuum R4 . It can be shown [306] that the solution of euclidean field
theory describing this process is in fact not only topologically such a ball. It has
perfect O(4) symmetry, i.e. it is a 4d ball also geometrically. The 3d boundary of
this ball is the bubble wall.
Let us supply a different perspective on the euclidean field theory version of the
quantum mechanical bounce of the last section. This will at the same time provide
an intuitive argument for its O(4) symmetry. In quantum mechanics, the bounce is
a solution which asymptotes (at t → ±∞) to the metastable state. The field theory
analogue is expected to be a solution asymptoting (at |x| → ∞, with x ∈ R4 ) to
the false vacuum. In quantum mechanics, the solution explores, in its centre, the
region of the potential to which the particle can decay. Thus, we expect that the field
theory bounce also explores, in its centre, the decay region. In the field-theoretic
case, this is the true vacuum and hence we expect the field theory bounce to contain
a region of true vacuum in its centre. Thus we are indeed looking for a 4d ball of
true vacuum in the false-vacuum background. This ball should be a solution of the
equations of motion following from the O(4)-symmetric euclidean action based on
(7.22). Thus, we expect O(4) symmetry. Let us then estimate the action of such
candidate field configurations, i.e. of balls of radius R filled with true vacuum and
centred on x = 0. There are contributions from the (lower-energy) volume and the
bubble wall boundary, such that
π2 4
Sball (R) − R V + 2π 2 R 3 T . (7.26)
2
This has an extremum at R = Rc , with Rc precisely the critical radius of (7.25).
As a result, we can finally write down the field-theoretic bounce action Sbounce ≡
Sball (Rc ):
27π 2 T 4
Sbounce . (7.27)
2 (V )3
To translate this in a field-theoretic decay rate, we need to pay attention to one last
important difference between the quantum mechanical and field-theoretic analyses:
The quantum mechanical rate characterises events per time, the field-theoretic rate
characterises events per time and volume. This works out in the quantum mechanical
case due to the explicit factor T that appears in the last term of (7.18). If we redo
the analysis in field theory, taking our space to be T 3 × R, we will have a continuum
of bounces since the latter can occur at any point of the spatial torus T 3 . Integrating
over all of them will give a factor Vol(T 3 ), accompanying the time factor T . Thus,
we are justified in writing
∼ e−Sbounce , (7.28)
with Sbounce calculated above and with the interpretation of as a rate of events per
time and volume.
7.5 Tunnelling in Gravitational Theories 237
7.5 Tunnelling in Gravitational Theories
First, it is clear that in a limited parametric range the analysis of the last section
continues to be valid even if our field-theoretic model is coupled to gravity. We
may restrict attention to the case where Vf alse ≡ V (0) ≥ 0 since in the opposite
case one would be starting with AdS space, which cosmologically always crunches
after a short time. Next, let us denote by Hf alse = (Vf alse /3)1/2/MP and Ht rue =
(Vt rue /3)1/2/MP the Hubble parameters of the false and true vacuum. As long as
Rc 1/Hf alse/t rue , the bubble nucleation process is occurring essentially under
flat-space conditions, such that our purely field-theoretic results for the rate continue
to hold. This covers many interesting cases.
However, these conditions can also easily be violated. For example, if V is very
small the critical bubble can be very large, invalidating the flat-space approximation.
More interestingly, in the gravitational case a qualitatively new possibility, namely
that of up-tunnelling (from smaller to larger vacuum energy density) arises. Thus, a
dedicated gravitational analysis is mandatory.
The classical paper on the subject is that of Coleman and De Luccia [307],
with a selection of subsequent analyses appearing in [308–313] and refs. therein
(cf. also the very recent discussion in [314]). To save time, we will start our
presentation by directly generalising the crucial field-theoretic bounce of Sect. 7.4
to the gravitational case. Comments concerning the intuitive interpretation will be
provided subsequently.
The bounce of Sect. 7.4 is a solution of the euclidean theory on R4 which
contains a spherical (O(4)-symmetric) domain wall with the initial-state (false)
vacuum outside and the final-state (true) vacuum inside, cf. the l.h. side of Fig. 7.8.
When generalising this to a gravitational theory, the absolute values of the respective
energy densities become relevant. We first focus on the case where both Vf alse and
Vt rue are non-negative, such that both the inside and the outside of the domain
wall become positively curved. This is displayed on the r.h. side of Fig. 7.8. To
Fig. 7.8 On the left: Sketch of the field-theoretic bounce where a ball of the final-state vacuum
is present inside flat R4 . On the right: The gravitational analogue, where this ball is present inside
of a 4-sphere (euclidean de Sitter space). Crucially, the curvature in the final-vacuum patch is in
general different from that of the surrounding S 4
understand this geometry, the key observation is that 4d euclidean de Sitter space is
simply a 4-sphere, with the curvature characterising the value of the curvature scalar
(which depends on the cosmological constant). Thus, all one needs is a geometry in
which a portion of a small-radius sphere (the false vacuum) is cut out and replaced
by a piece of a large-radius sphere (the true vacuum). Moreover this has to be a
solution of Einstein’s equations. The latter is obvious away from the domain wall,
where simply the right choice of radius has to be made. What is non-obvious and
will be discussed momentarily is the solution inside the wall, where both field value
and curvature change continuously.
Maybe the technically simplest and most straightforward approach remains that
of [307]. It starts by parameterising the metric of the euclidean geometry on the r.h.
side of Fig. 7.8 as
ds 2 = dr 2 + f (r)2 d32 , (7.29)
where d32 is the round metric on the unit-radius 3-sphere. The point r = 0 is the
centre of the true vacuum patch. This metric is accompanied by a field profile φb (r),
where the index ‘b’ stands for ‘bounce’. The function φb is approximately constant
and equal to the false vacuum value, φ = φf , for r rc . For r rc , it is again
approximately constant and equal to the true vacuum value, φ = φt . The reader may
want to recall Fig. 7.7, where φf = 0 and φt = φ1 , but with the potential shifted
upward such that both minima are de Sitter. We emphasise that rc is the value of the
coordinate r which corresponds to the location of the O(4)-symmetric domain wall.
By contrast, R(rc ) is the physical radius of the true vacuum patch. The behaviour
of φb (r) in the vicinity of r = rc depends on the precise form of the domain wall,
which in turn depends on the details of the potential. This will not be important in
the thin-wall approximation.
Let us first allow for a general O(4)-symmetric field profile φ. The euclidean
action then takes the form

1 2 1
S = d x − MP R + (∂φ) + V (φ)
4 2
2 2

2 f f 2 1 1 2
= 2π 2 3
f dr 3MP + 2 − 2 + φ +V
f f f 2
2
2 f 1 1 2
= 2π 2
f dr −3MP
3
+ 2 + φ +V , (7.30)
f2 f 2
where the prime denotes differentiation with respect to r. To obtain the expression
in the last line, we used integration by parts together with our knowledge that f (r)
is vanishing sufficiently quickly at the minimal and maximal values of r.
7.5 Tunnelling in Gravitational Theories 239
It is convenient to supplement the above with the rr-component of the Einstein

equations,
1
Rμ − Rgμν = MP2 Tμν , (7.31)
2
for the given matter action and geometry. This equation is, in fact, almost identical
to the Friedmann equation which appeared in (7.3). In the present context, it reads

f 2 k 1
3MP 2
− 2 = φ 2 − V , (7.32)
f f 2
with k = 1. The only changes are in the notation (t → r and a(t) → f (r)) and
in the relative sign between terms with and without r-derivatives. The latter arises
due to the transition from Lorentzian to euclidean signature. The combination of
the action in the form of (7.30), the equation of motion for φ following from it, and
the Einstein equation (7.32) are sufficient to calculate the exponent in the tunnelling
process we are after.
To make this explicit, we first need to modify our field-theoretic result such that it
allows for a non-zero vacuum energy in the false vacuum. This is of course essential
if we want to be able to talk about tunnelling in de Sitter space. With a view on the
l.h. side of Fig. 7.8, one immediately sees that the correct modification is
∼ e−Sbounce → ∼ e−Sbounce +Sf alse . (7.33)
Here Sbounce ≡ S[φb ] and Sf alse ≡ S[φf ], with φb the field configuration of the
bounce and φf the constant field configuration corresponding to the initial, false
vacuum state. We see that, with this generalised expression for the rate, we obtain
the correct result even if V (φf ) = 0. Indeed, any possible constant contribution to
the lagrangian simply cancels, such that only the effects of the true vacuum region
and the domain wall remain.
Now the generalisation to the gravitational case is obvious:
, -
= exp −S[φb , gb ] + S[φf , gf ] . (7.34)
Here the first term in the exponent is the action of the bounce geometry with the
corresponding field configuration (cf. the r.h. side of Fig. 7.8) and the second term is
simply the action of the sphere with constant initial-state field value and appropriate
curvature.
As noted, Eqs. (7.30) and (7.32) contain enough information to evaluate the rele-
vant actions and hence the decay rate. We leave this as an exercise (Problem 7.10.1),
which consists essentially in following the analyses of [307] and [308]. The result
can be given in a particularly compact form as [313]

27π 2 T 4
= exp(−B) , B= r(x, y) . (7.35)
2 (V )3
Here the first factor in the exponent B is the field-theoretic bounce action already
displayed in (7.27), while the second factor, r(x, y), characterises the gravitational
correction. It reads explicitly

1 + xy − 1 + 2xy + x 2
r(x, y) = 2 , (7.36)
x 2 (y 2 − 1) 1 + 2xy + x 2
where
3T 2 Vf + Vt
x= , y= with V = Vf − Vt ,
4MP2 V V
(7.37)
and MP is the reduced Planck mass.

This formula was derived for the decay of a false de Sitter vacuum to a true
de Sitter vacuum: Vf ≥ Vt > 0. Nothing changes if Vt becomes zero or even
negative: It is equally possible to glue a patch of true vacuum AdS space into a false
vacuum de Sitter sphere. The geometric situation is analogous to that displayed on
the r.h. side of Fig. 7.8. Also the derivation of the corresponding exponent B is
unchanged.
However, a new and possibly unexpected situation arises if one considers the
tunnelling from true to false-vacuum de Sitter. Let us first argue why such a process
might at all be possible: As a quantum fluctuation, any state can form, even that of
a higher-energy vacuum bubble in a lower-energy background. If the lower-energy
background is Minkowski, then such a bubble can of course never materialise. It
always has positive energy and hence would always violate energy conservation if
it were to become a real state.
By contrast, if the lower-energy vacuum is already de Sitter, then a sufficiently
large virtual bubble with false vacuum inside can be pulled to larger and larger size
by the background expansion. No energy-conservation argument forbidding such a
process can be given in the (time-dependent!) global de Sitter geometry.
The corresponding rate is easily obtained without the need for a new calcula-
tion [311, 312]. To see this, let us recall the rate for the decay of false to true de
Sitter vacuum:
f → t = exp(−Bf → t ) with Bf → t = S[φb , gb ] − S[φf , gf ] . (7.38)
As explained before, the bounce field configuration and geometry underlying

S[φb , gb ] corresponds to a patch of true vacuum φt glued into a sphere of false
vacuum, with φ = φf . Both parts of this geometry are cutouts from spheres.
Hence, it is clearly just a matter of convention which of them we call the ‘patch’
or ‘bubble’ and which the ‘background’. If we invert this interpretation, the same
7.6 Our Universe in the Eternally Inflating Landscape 241
action S[φb , gb ] may serve as a building block for the tunnelling rate from true to
false vacuum:
Bt → f = S[φb , gb ] − S[φt , gt ] . (7.39)
Crucially, we have now subtracted the true vacuum rather than the false vacuum full-
sphere action. The subtraction is justified, as before, because this action represents
the relevant background geometry. The above may be rewritten as

Bt → f = S[φb , gb ] − S[φf , gf ] + S[φf , gf ] − S[φt , gt ] . (7.40)
Now, using (7.35) as the definition of a function B = B(Vf , Vt , T ) and employing

the expression
MP4
SedS (V ) = −24π 2 (7.41)
V
for the action of euclidean de Sitter space based on a potential V , one immediately
derives

1 1
Bt → f = B(Vf , Vt , T ) + 24π 2 MP4 − . (7.42)
Vt Vf
Comparing this with our previous result
Bf → t = B(Vf , Vt , T ) (7.43)
one sees, not surprisingly, that tunnelling upwards comes at the cost of an extra
exponential suppression.
7.6 Our Universe in the Eternally Inflating Landscape
Naively one might think that we live in one of the many bubbles, and ours just
happens to have very small λ. This is roughly true, but important details are missing.
First, given how small our λ is, we naturally expect the previous vacuum’s λ to be
much larger. But a corresponding tunnelling event would have endowed our vacuum
with a large and negative spatial curvature. Our cosmological evolution would have
been governed by the FRW equation
3H 2 = ρ − 3k/a 2 with k = −1 (7.44)
and with initial conditions where the curvature term (the 2nd term on the r.h.
side) would be at least comparable to the matter term (ρ, which includes matter,
radiation and λ) from the start. In such a situation, there can be either λ domination
Fig. 7.9 Tunnelling to our vacuum, where a period of slow-roll inflation, reheating and structure
formation precedes the dS phase
Fig. 7.10 Various ‘surfaces of constant energy density’ following the initial tunnelling transition
to our bubble. This sketch is an adaptation of a figure from [317], which deals with possible bubble
collisions and their observational effects
or curvature domination succeeded by λ domination, but no extended radiation or

matter dominated epoch, as in our world. The reason is simply that, with expansion,
matter and radiation densities decay faster than curvature.
The way out is to postulate that the potential near our local minimum has
the peculiar feature of an inflationary plateau, where cosmological inflation took
place, cf. Fig. 7.9. (It does not actually have to be a plateau—any sufficiently flat
potential region would do.) Together with making the universe highly homogeneous
(i.e. solving the horizon problem), this inflationary phase dilutes the curvature
contribution (i.e. it solves the flatness problem). Inflation then ends in reheating
and structure formation, with galaxies, stars, planets, etc. Eventually, as λ gets to
dominate, the universe becomes empty and cold. (It could also crunch, if λ were
small and negative, but this is apparently not the case in our world.) Thus, observers
of ‘our kind’ actually always appear shortly after the tunnelling transition.4
Nevertheless, in a given bubble of ‘our vacuum’ their number is infinite, as are our
reheating surface and our structure formation surface (see Fig. 7.10). These surfaces
(including presumably the surface of death of all stars and hence of all civilisations)
follow the straight bubble wall surface all the way up to infinity. In the figure, these
surfaces approach the bubble wall both to the left and to the right—in reality they
approach the lightcone in R1,3 which is defined by the bubble wall. (Of course, that
could change if a collision with another bubble occurred.) One may say that the
4One may, however, speculate that civilisations can survive much longer than it takes for galaxies
and planetary systems to decay and stars to burn out [315].
7.7 Making Statistical Predictions and the Measure Problem 243
interior of any bubble is an open (infinite) Friedmann–Robertson–Walker universe.

For many more details and references see e.g. [316].
7.7 Making Statistical Predictions and the Measure Problem
Accepting the above landscape picture and eternal inflation as the process populat-
ing it, the measure problem is easy to state, at least at an intuitive level (see e.g. the
reviews [11, 318, 319]): We live in one of the vacua, but we do not know in which
one. We would like to make a statistical prediction (given that we know certain
features of our vacuum, but not all). Let us say the new observable which we are
going to measure tomorrow can take the values A or B. The most naive way to
make a statistical prediction would be to say that the ratio of probabilities is
pA /pB = NA /NB . (7.45)
Here NA/B are the numbers of observers in the multiverse who have measured all
that we have measured so far and who will, in the next measurement, find A or
B, respectively. But in eternal inflation, by definition, both numbers are infinite
and their ratio is not well-defined. What is worse, if one cuts off the infinity in
the future, the prediction becomes dependent on the precise type of cutoff. For
example, one could restrict attention to measurements before some maximal time
tmax , taking the limit tmax → ∞ in the end. But such a maximal time cutoff,
illustrated in Fig. 7.3, is not unique. This is due to the absence of a unique global
time variable in de Sitter space or, more precisely, in the more complicated geometry
of multiple de Sitter bubbles as they arise in eternal inflation. More generally
speaking, the diffeomorphism invariance of general relativity prevents the existence
of an unambiguously defined time cutoff.
Before discussing the various suggestions for how the measure problem might be
overcome, one should note one pragmatic and historically very successful approach:
One may assume that we are likely to observe what is common in the string theory
landscape, independently of the dynamics populating the latter. More precisely, this
amounts to the assumption that any bias in favour or against a certain observational
feature induced by cosmology is small compared to the bias derived simply from
counting vacua. The latter is, at least in principle, possible since to the best of our
present understanding the landscape is finite (at least if one imposes a certain IR
cutoff, excluding models with an arbitrarily low KK scale, but maybe even more
generally) [320].
The historic success of this approach is Weinberg’s prediction of the size of
the observed cosmological constant [321] (see also [322]). Crucially, the argument
comes from a period when cosmology was well enough understood to provide a firm
upper bound on our present expansion rate H . Yet, the time evolution of H was not
known precisely enough to determine whether this non-zero H was predominantly
due to matter, a cosmological constant or spatial curvature,
3H 2 = ρmat t er + λ − 3k/a 2 . (7.46)
In other words, it was known that |λ| ≤ λ0 with some positive λ0 which was
very small compared to particle physics scales. Now, let us assume that we live
in one universe in a multiverse with many different λ. Moreover, assume that these
available values of λ may be described by a statistical distribution which is smooth
and dominated by large energy scales (like the Planck, the string, or the SUSY
breaking scale). If the point λ = 0 plays no special role, then one expects that
projecting this distribution to the tiny interval (−λ0 , λ0 ) gives essentially a constant
distribution on that interval. But this projection was exactly what observations at
that time had achieved. Thus, the prediction for a future measurement of λ was to
be made using a constant distribution on the interval (−λ0 , λ0 ). This corresponds
to simply drawing a λ value from that interval. With overwhelming probability, the
result should be a value (of either sign) comparable in magnitude to λ0 . A much
smaller (absolute) value would be very unlikely. Famously, a non-zero λ ∼ λ0 was
discovered only a few years later.
We should note that a closely related but stronger and more debatable argument
predicting λ can be made. Namely, fixing all other particle-physics and cosmological
parameters (including in particular the initial curvature perturbations which have led
to the formation of structure, including stars and planets), one may argue for a so-
called anthropic prediction of λ. Indeed, if λ were much larger than λ0 , exponential
expansion would have set in earlier in the history of the universe, preventing the
formation of any structure and hence of life. By contrast, too large negative λ
would have led to a big crunch before any observers could have emerged. Thus,
the observed value of λ can be said to represent an anthropic prediction (given
the previously made assumption about the statistical distribution in the landscape),
independently of the observational status at the moment of Weinberg’s famous
paper. In fact, Weinberg’s paper emphasises this anthropic prediction rather than
the one based on the observational situation, which we explained before.
We should emphasise that such anthropic arguments based on some form of
multiverse are much older than the string theory landscape. Moreover, they can be
applied to quantities other than the cosmological constant or electroweak scale. The
reader may want to explore this line of thinking starting, e.g., with [323–326] and
references therein.
7.8 Proposed Measures
This section draws very significantly on the relatively recent and very clear
review [319]. Following this analysis, one distinguishes global and local measures.
The former count observers before some late cutoff time tmax (or in some other
way that keeps the total number of observers finite) and takes the cutoff to infinity
in the end. By contrast, local measures count observers in a way associated with a
7.8 Proposed Measures 245
single timelike geodesic, in the simplest case following a single observer on their
path through the tunnelling events of the multiverse.
Some of the oldest measures are global. For example, one may start with some
spacelike surface and define time globally using geodesics originating in its every
point. The initial surface should be finite, but this is not a problem: Global de Sitter
space has the topology of ‘Time’×S 3 , so a spacelike cut can provide the required
surface.
The most obvious choice is to use proper time [327]. But this is claimed to be
ruled out by observation on account of the ‘Youngness paradox’ [328]: Think of us
living inside a bubble5 which nucleated inside some high-λ vacuum. Due to the
very fast exponential expansion of this mother vacuum, and since there are bubbles
of our type with all kinds of ages present, we are most likely to find ourselves to be
as young as only physically possible. In other words, we should be the youngest
observer on the youngest planet in the youngest galaxy. This appears not to be
consistent with the place we occupy in our universe.
Another fairly obvious choice, one that apparently is not ruled out yet, is to use
scaling time (i.e. a scale factor cutoff). In other words, one measures time (and hence
introduces a cutoff) on the basis of the number of e-foldings [327].
Finally, there is lightcone time and correspondingly a lightcone time cutoff [329,
330]. Here, given a point in the eternally inflating spacetime, one follows its
lightcone to the future boundary. The part of the boundary inside this lightcone is
then projected back to some initial surface (using the same congruence of geodesics
discussed earlier). The size of this projection defines the time of the point.
Concerning local measures, one option is to count observers inside the causal
patch (the so-called causal diamond) of a timelike geodesic. This is not automati-
cally finite and, if it is, the late-time attractor behaviour of eternal inflation is lost.
A variant without these problems is the ‘census taker’s cutoff’, which focuses on
geodesics that end in a Minkowski vacuum6 [331]. Such geodesics are clearly
infinite, but one may restrict attention to events the future lightcone of which crosses
the central geodesic before some time t. In other words, one may count all events
inside the causal diamond with apex at t, eventually taking the limit t → ∞. The
reader may consult Fig. 7.11 for an illustration.
Other local measures count observers in a cylinder of finite physical radius
centred on a timelike geodesic (‘fat geodesic measure’ [332]) or within its apparent
horizon (‘apparent horizon measure’ [333]).
Many more possibilities and variants of the choices above are discussed in [319].
At first sight, such a proliferation of measures looks rather discouraging. However,
things are not quite as bad as they seem: It turns out that differently defined measures
may be equivalent in the sense that they give the same (or in some cases very similar)
results. In particular, there exist equivalences (referred to as ‘dualities’) between
5To be precise, near the initial bubble wall of a small-λ bubble, as explained in Sect. 7.6.
6 A.Maloney, S. Shenker, L. Susskind (unpublished); R. Bousso, B. Freivogel, S. Shenker, L.
Susskind, I. S. Yang, (unpublished).
Fig. 7.11 On the left: Penrose diagram of de Sitter space. Here the spatial S 3 is represented as an
S 2 fibred over an interval (the horizontal axis of the square). North and south pole are labelled by
N and S and the horizons of corresponding observers are indicated as dashed lines. On the right:
Upper portion of the same diagram with a de Sitter bubble (‘dS’) and a Minkowski bubble (‘M’)
added. A trajectory of an observer ending up in the Minkowski bubble is also shown
certain local and global measures. As a result of this, one may basically focus on
the following three options:
1. Lightcone time cutoff—equivalent to the causal diamond measure;

2. Scale factor cutoff—equivalent to the fat geodesic measure;
3. Apparent horizon cutoff.
A different approach is suggested in [329]. Here, one determines the abundances

of bubbles of different kinds by projecting them on some initial surface (using a
congruence of timelike geodesics). One only counts bubbles that are larger than
,
taking the limit
→ 0 in the end. Inside each bubble, one has an infinite open
FRW universe, as explained above. Thus, if a certain type of bubble allows for
observers, then the number of observers inside each specific bubble will always be
infinite. These latter infinities are not used to further weight the relative probabilities
between the different bubbles. As a result, this proposal represents a rather radical
but logically consistent deviation from all the previous discussions (where the focus
was on counting the number of observers in one way or another).
Let us end with an interesting argument against any measure based on some
geometric cutoff. It is known as the Guth–Vanchurin paradox7 and goes as
follows: Imagine an eternal de Sitter space in which observers are born at random
points in spacetime. Each observer sets an alarm clock which will, with 50%
probability, let them sleep for a short or a long period of time. Then they go to
bed. After waking up and before checking their watch, they ask themselves what
the statistical prediction is that they slept for a short or a long time. The intuitive
answer is clearly 50%. However, if they look around themselves they will (given
an appropriate choice of model parameters) clearly see many more short sleepers
than long sleepers. The reason is that the latter tend to have left the horizon. The
claim and the paradox is that this (apparently wrong) prediction will be shared by
7 A. Guth, V. Vanchurin (unpublished).

7.9 Predictions from First Principles? 247
any geometric-cutoff measure. Indeed, taking the prediction of such measures more
seriously than the intuitive answer one can even arrive at the highly counterintuitive
conclusion that ‘time will end’ [334]. Such a sudden end of time resolves the issue
by providing an objective reason for why it is more likely, after waking up, to find
out that one has only slept a short time: For the long sleepers, the probability is
higher that they run into the spacelike singularity during their nap.
7.9 Predictions from First Principles?
It is conceivable that further scrutiny of the proposed measures will show that all but
one of them are either in some way inconsistent or are in conflict with observations.
As a result, one would then find the one, correct measure for making statistical
predictions in the string multiverse. However, there is no guarantee that this will
happen. On the contrary, it is conceivable that over time even more different,
consistent measures with varying predictions will be found. In either case, it appears
highly unsatisfactory that an extra input in the form of a measure choice has to be
added to the hopefully unique quantum gravity theory from which the multiverse
follows. It would be desirable to derive a measure from first principles rather than
postulate it.
Part of the problem for achieving this is, of course, that gravity in general (even
apart from its notorious UV problems) and de Sitter space in particular lack a
satisfactory quantum mechanical understanding. One angle of attack concerning
the quantum mechanics of de Sitter space is the idea of a so-called dS/CFT
correspondence [335].
Before explaining this, it is unavoidable to devote at least a few words to the
AdS/CFT correspondence, which is much better understood and established [300,
301]. We have to restrict ourselves to stating the facts: first, (d+1)-dimensional Anti-
de Sitter space (homogeneous, negatively curved space with Lorentzian signature)
has the topology of a Bd × R. Here Bd is a d-dimensional ball, with the boundary
being at infinite distance. The real axis R represents time. The curvature of this
space has the effect that, very roughly speaking, the centre of this ‘solid cylinder’
is at lowest gravitational potential. In other words, objects tend to fall from the
boundary towards this centre.
The boundary of AdS is clearly the ‘cylinder’ S d−1 × R. The term ‘AdS/CFT
correspondence’ means the following: Any consistent gravitational theory in the (d+
1)-dimensional bulk has an equivalent description (is dual to) a d-dimensional CFT
living on the boundary. Very roughly speaking, degrees of freedom of the CFT can
be ordered according to their energy scale μ. This scale parameter μ corresponds
to the radial direction of the ‘solid cylinder’ representing the bulk. Both in the CFT
and in the bulk the path towards the UV (i.e. to towards the boundary) is infinite.
But in the IR the CFT scale μ is restricted by the compactness of its spatial volume.
The corresponding degrees of freedom of lowest energy scale are mapped to the
bulk degrees of freedom near the centre, i.e. at highest red-shift.
Now let us turn to the analogous logic in dS/CFT. First, we need to recall that (d+
1)-dimensional dS space has the topology of S d ×R. This is very different from AdS
since, in particular, spatial sections of this space are now simply spheres and as such
have no boundary. The only boundaries are now those at past and future infinity.
In contrast to AdS/CFT, these boundaries are spacelike. The idea of dS/CFT is to
map the dynamics of the bulk to a d-dimensional (euclidean) CFT that lives on the
future boundary S d . This time, the energy scale parameter μ of the CFT is expected
to correspond to the time evolution parameter of the dS bulk. The reader should be
warned that, while AdS/CFT has become one of the cornerstones of modern field
theory and gravitational research, dS/CFT remains truly conjectural.
After these lengthy preliminaries it is easy to state what a dS/CFT-based first-
principles approach to the measure problem might look like [336]: One has to
explicitly map the bulk of the eternally inflating dS spacetime to the boundary at
future infinity and hence to the CFT. If one now introduces a UV cutoff in the CFT,
this may correspond to a canonical or natural late-time cutoff in the bulk. One may
hope that this gives rise to an unambiguous way of counting observers and hence
to an a-priori definition of a measure. For an interesting toy model of the landscape
that may be related to the dS/CFT measure proposal see [337].
Another approach, suggested as an unambiguous first-principles definition of a
measure in [338], is the following: One considers the world from the perspective
of a single, abstract observer. This observer sees various tunnelling events, ‘lives’
through many big-bang cosmologies like our own, until he or she eventually ends
up in a terminal vacuum. The sequence of events which the observer witnesses is
subject to the usual uncertainty of quantum mechanics. In other words, the life of
this observer is a superposition of all the possible sequences of tunnelling events.
Thus, in a sense, the ‘many worlds’ of Everett are identified with the many worlds
of the multiverse (see also [339] for a related discussion). One now defines the
statistical prediction for any observable using the quantum mechanical expectation
values for this observable on the basis of the single-observer worldline introduced
above.
The approach just presented bears a certain similarity to the fat geodesic measure
[332] and its quantum version (the ‘quantum watcher measure’) as discussed
in [340]. It may moreover be problematic that terminal vacua, the quantum
mechanical significance of which is not understood, play a central role in the single-
observer approach just introduced. Thus, it is probably fair to say that no consensus
has so far been reached on whether dS/CFT correspondence or the single-observer
perspective or some different approach is the leading candidate for a first-principles
measure.8
Yet another perspective, which has a strong aesthetic appeal but is probably not
developed enough to be called a measure proposal, has been introduced in [342]:
8 We also note that the presentation of the single-observer approach in [341] bears some formal
similarity to the Wheeler–DeWitt equation (see below) and may hence be related to the approach
to be discussed next.
7.9 Predictions from First Principles? 249
The idea is also to focus on a single observer, but in a way very different from
counting events along the observer’s worldline. Instead, one appeals to the concept
of a wave function of the universe [343–345]. To explain this, we have to introduce
this possibly unfamiliar concept:
Let us consider the canonical quantisation of Einstein gravity, for simplicity
in a spacetime M × R, where M is a compact spacelike manifold. Now, recall
that the Hamiltonian H generates time translations t → t +
. From the
perspective of general relativity, this is just a particular diffeomorphism. However,
diffeomorphisms are gauged. Hence an operator like H generating one of them must
vanish on any physical state (since physical states are by definition gauge-invariant).
Thus, there is no time evolution but rather a so-called constraint equation,
H = 0. (7.47)
In this so-called Wheeler–DeWitt equation is a functional on the space of

metrics on M:
: Mg → [Mg ] ∈ C . (7.48)
Here Mg stands for the manifold M with metric g. Hence can be viewed as
the gravitational analogue of the Schrödinger wave function of quantum mechanics
or, better, of the Schrödinger wave functional of quantum field theory. Its physical
role differs from the latter in that it does not evolve in time. Instead, it has to be
interpreted as the wave function of the universe in the sense that it contains the
probability for observing some relevant 3-manifold at any moment in time (which
cannot be in general defined). Note that this is not incompatible with conventional
time-dependent physics [346]: One may, for example, enrich the argument of by
non-gravitational fields, [Mg ] → [Mg , φ]. Then one may consider physical
situations with a clock (made of fields φ) and ask for the probability to observe
a given metric g at a given time, encoded in the field configuration φ. Standard
physical questions about the occurrence of an event at a given time are hence
encoded in questions about the correlation between values of g and values of φ.
Given these preliminaries, the suggestion of [342] may now be roughly for-
mulated as follows: One should not ask about possible, approximately classical
histories of the universe as encoded in and try to count observers which make a
certain observation. Instead, one should adopt a coarse-grained perspective in which
one ignores all information in except that an observer makes, say, observation
A or that an identical observer makes observation B. The relative probability of
these two observations should be encoded in a hopefully well-defined and finite
form in . Crucially, in asking this question one ignores any irrelevant information
about where in the multiverse the observation occurs and which of the many
observers making identical observations one is considering. The hope is that this
coarse-graining step would make the answer well-defined. As an alternative, it
has more recently been suggested [347] to implement the idea of coarse graining
using specifically the so-called Hawking–Hartle no-boundary proposal [345] for the
definition of the wave function of the universe. It has been argued that, as a result,
the sensitivity to the exponentially large multiverse disappears if one asks the right
questions.
In summary, it should have become clear that, while a first-principles definition
of the measure is highly desirable, this subject is not settled. It is also conceivable
that the measure has indeed to be viewed as fundamental new input, in addition
to whatever the ultimate first-principles definition of string or M-theory will turn
out to be. Furthermore, it is even possible that no unique and correct measure
exists and that statistical predictions in the eternally inflating multiverse will remain
impossible as a matter of principle. A justification of this pessimistic attitude
might be the impossibility of repeating an experiment (such as the measurement
of λ) many times. Hence, the usual physicist’s definition of probabilities and of a
probabilistic prediction does not work. On the contrary, one may also defend the
possibility of a probabilistic prediction for a single observation as follows: View
the 10-fold repetition of an experiment, each of which rules out a certain theory at
99% confidence level, as one single experiment or observation. Such a viewpoint
is clearly a matter of convention and is logically perfectly acceptable. Now, this
‘10-fold’ measurement (assuming all 10 results agree) leaves only a probability of
10−20 that the theory is correct. We would clearly and rightfully dismiss such a
theory and, as just argued, we do so on the basis of a single experiment. All that
matters is a sufficiently high significance of the result.
While these are all very interesting and potentially important questions, we have
now entered a field where very little is known with certainty. So we should maybe
stop here, leaving it to the reader to explore recent, original papers and form their
own opinion.
7.10 Problems
7.10.1 Coleman–De Luccia Tunnelling
Task Fill in the calculational details leading to the expression for the exponent B in
the decay rate ∼ exp(−B) of a false de Sitter vacuum that was given in Sect. 7.5.
Hints Use the Einstein equation to bring the action to a form without r-derivatives.
Assume that the domain wall, i.e. the interval in f where most of the change of φ
occurs, is defined by fdw < f < fdw + . Moreover, assume that the thin-wall
approximation, f , is valid. Then evaluate B = S[φb , gb ] − S[φf , gf ] in three
pieces: outside the domain wall, in the domain wall region and inside the domain
wall. Since the resulting expression is, by assumption, extremal, the correct value
of fdw can be determined by extremising B w.r.t. fdw . If you get stuck, consult the
original papers [307, 308, 313].
7.10 Problems 251
Solution The desired form of the action is

3MP2
S = 4π 2 3
f (r) dr − + V (φ(r)) . (7.49)
f (r)2
Based on this general expression, we now have to evaluate
S[φb , gb ] − S[φf , gf ] . (7.50)
We split the r-integration in three pieces. First, outside the domain wall the
integrands in S[φb , gb ] and S[φf , gf ] are identical. So there is no contribution from
that region.
Second, in the domain wall region the contribution from the 1/f (r)2 -term can
be neglected. The reason is that this region is small, fdw . As one enters the
domain wall region from the outside, the function f (r), which encodes the euclidean
4d geometry, only starts to distinguish between the bounce and the S 4 false vacuum
solution. The effect is second order in the small quantity and can be neglected.
Note that, by contrast, the contribution from the V (φ) term is not small. The reason
is that V has to change by a fixed amount as one passes the wall. Hence its variation
grows as is taken to zero. We then have

rdw +δ
S[φb , gb ] − S[φf , gf ] 4π 2 f (r)3 dr V (φb (r)) − V (φf ) (7.51)
domain wall rdw
rdw +δ
4π 2 f (rdw )3 dr V (φb (r)) − Vf . (7.52)
rdw
Here the integral is over the interval in r corresponding to the domain wall, defined
by f (rdw ) = fdw and f (rdw + δ) = fdw + . Let us compare the above with the
domain wall tension, which is by definition the sum of gradient and potential energy
in the wall:
rdw +δ
1 2
T dr φ + [V (φb (r)) − Vf ] . (7.53)
rdw 2 b
Concerning the field profile in the wall, gravitational effects are subleading and we
can appeal to our understanding of a field-theoretic domain wall or bubble wall from
Sect. 7.4. The only difference is that, in (7.53), we have subtracted the false vacuum
potential energy Vf , a quantity that was set to zero by definition in our previous
field-theoretic analysis. Moreover, the dimensions parallel to the wall are irrelevant,
such that we can go back even further and think of the euclidean tunnelling solution
in quantum mechanics as discussed in Sect. 7.3 and visualised in Fig. 7.5. There,
we learned that gradient and potential energy are always equal during the transition
between the two minima (since we can think of a dynamical rolling process in the
inverted potential). As a result,

S[φb , gb ] − S[φf , gf ] 2π 2 fdw
3
T. (7.54)
domain wall
Third, inside the domain wall the field φ is constant and the Einstein equation
(7.32) can be used to change the integration variable from r to f :
0
df Vf 2
= 1− . (7.55)
dr 3MP2
Thus
1/2
fdw Vt f 2
S[φb , gb ] − S[φf , gf ] −12π 2
MP2 f df 1−
inside 0 3MP2

− Vt → Vf
⎡ 3/2 ⎤
2
12π 2 MP4 Vt fdw
= ⎣ 1− − 1⎦
Vt 3MP2

− Vt → Vf . (7.56)
Recall that fdw is nothing but the physical radius of the ball of final-state vacuum.
So let us somewhat simplify notation by writing R instead of fdw . Then, combining
the contributions from the wall and the inside region, we have
⎡ 3/2 ⎤
4 2
B(R) 6M P ⎣ V R
= T R 3
+ 1 −
t
− 1 ⎦ − Vt → Vf . (7.57)
2π 2 Vt 3MP2
The algebra involved in solving B(R) = 0 for R and inserting the result in (7.57)
is less horrible than one might expect at first sight. Indeed, the equation 0 = B (R)
may be brought to the form

0=ζ+ 1 − aζ 2 − 1 − bζ 2 with ζ = RT /2MP2 , (7.58)
and
4MP2 4MP2
a= Vf , b= Vt . (7.59)
3T 2 3T 2
7.10 Problems 253
Dividing this by ζ and rewriting the result in terms of w = 1/ζ 2 one finds
√ √
0=1+ w−a− w −b, (7.60)
which is easy to solve for w:
1
w= 1 + 2(a + b) + (a − b)2 . (7.61)
4
Next, we write the exponent B as

B(R) 8MP6 1
= ζ3 − 2 3/2
(1 − aζ ) − 1 + a → b (7.62)
2π 2 T2 a

8MP6 1 3
= ζ3 − ζ (w − a)3/2 − 1 + a → b . (7.63)
T2 a
This simplifies upon noticing that
√ 1 √ 1
w − a = [1 − (a − b)] and w − b = [1 + (a − b)] . (7.64)
2 2
One finds

B(R) 8MP6 ζ3 3 3 1 1
= 8ab + 1 + (a − b) a − 1 − (a − b) b + − .
2π 2 T2 8ab a b
(7.65)
At this point, it is useful to spell out the relation to the variables x, y introduced in
the main text:
1 y 2x y2 − 1
a −b = , a +b = , ζ = , ab = . (7.66)
x x 1 + 2xy + x 2 4x 2
Moreover, we have V = 3T 2 /4xMP2 , such that the formula for B from Sect. 7.5
(based on [308]) takes the form
B(R) 8MP6
2
= 2x 3 r(x, y) . (7.67)
2π T2
Now the agreement between (7.65) and (7.67) follows more or less immediately.
Concluding Remarks and Some Alternative
Perspectives 8
This last section is special in that it contains few equations and no exercises. It is to a
large extent a brief tour through additional topics that could and maybe should have
been covered but had to be left out for reasons of space or, more precisely, because
they will most likely not fit in the time frame of a one-semester course. Moreover,
part of the comments and ideas collected in this section lead away from the specific
‘string landscape perspective’ that was advertised and taught throughout the core
part of this course.
8.1 Low-Scale SUSY Versus Technicolor
Let us start with a brief discussion of alternatives to low-scale SUSY. Recall that
low-scale SUSY has played a central role in our course since it demonstrates that,
in principle, gauged (or otherwise interacting) scalars can be ‘naturally’ light. This
could have been a perfect explanation for the large hierarchy between the quantum
gravity scale and the electroweak scale. Maybe it still largely explains this hierarchy,
but (in part because of the non-discovery of SUSY at the LHC) this does not work
perfectly. It may still work partially and SUSY would then have to be ‘just around
the corner’ in the sense of energy scales.
However, roughly the same may be achieved by other means. The arguably main
and historically first candidate is known as Technicolor [348–353] (for reviews see
e.g. [354–359]). The term refers to a second version of ‘colour’ (as in the QCD
sector of the Standard Model), which is added to the Standard Model gauge theories
for a purely ‘technical’ reason. This reason is the creation of a technically natural
light scalar.
To understand this, recall how pions arise in the low-energy EFT of the Standard
Model (see e.g. [1, 3, 4, 12, 30]): If one neglects the small Yukawa couplings of u

256 8 Concluding Remarks and Some Alternative Perspectives
and d quarks, one can write the relevant lagrangian terms as
L ⊃ iq TL σ μ Dμ qL + iq TR σ μ Dμ qR . (8.1)
Here we have combined the pairs of Weyl fermions uL /uR and dL /dR (which make
up the up and down-quark Dirac fermions u and d) into doublets

uL uR
qL = and qR = . (8.2)
dL dR
At energies above the weak scale, the SU (2) acting on the first of these doublets
is the familiar SU (2)L gauge symmetry. But this is irrelevant for now since we are
at low energies. We also disregard U (1)em because of its small coupling. Thus, the
covariant derivatives in (8.1) refer solely to the SU (3) colour group.
Our lagrangian has, in addition to its SU (3) gauge symmetry, a global SU (2)L ×
SU (2)R symmetry:
(qL )i → (UL )i j (qL )j , (qR )i → (UR )i j (qR )j . (8.3)
Due to the strong non-perturbative effects associated with SU (3) gauge dynamics,
it is conceivable that a non-zero vacuum expectation value of fermion bilinears is
induced. Assuming that it respects the gauge symmetry and recalling that qL and qR
transform as 3 and 3 under SU (3), the only option is
(qL )i (qR )j = 0 . (8.4)
Here the contracted colour indices have not been displayed. Using a biunitary trans-
formation, this expectation value can be made diagonal. Moreover, for symmetry
reasons one expects the two eigenvalues to be equal, such that
(qL )i (qR )j = λ δij . (8.5)
This breaks SU (2)L × SU (2)R spontaneously to its diagonal subgroup, consisting

of elements of the form (U, U ) ∈ SU (2)L × SU (2)R . The generators outside the
Lie algebra of this subgroup act on the vacuum in a non-trivial way, but they clearly
cannot change the energy. Thus, we find a vacuum manifold, which can for example
be parameterised by group elements of the form
(, 1) ∈ SU (2)L × SU (2)R (8.6)
acting on (8.5):
(qL )i (qR )j → i k (qL )k (qR )j = λ i k δkj . (8.7)

8.1 Low-Scale SUSY Versus Technicolor 257
The condensation effect above is known as chiral symmetry breaking. While

it has not been rigorously derived from the QCD lagrangian, it is considered as
well established. This is based both on lattice studies and on the phenomenological
success of calculations using the above assumptions (chiral perturbation theory).
Now one writes
= exp(iπ a σa /fπ ) (8.8)
and identifies π a as the pion fields. The effective lagrangian for these massless
fields, which are clearly Goldstone bosons of the spontaneously broken global
symmetry, has to be invariant under
→ UL UR† . (8.9)
The lowest-dimensional such term is
fπ2 1
L ⊃ tr [(∂μ )(∂ μ )] = − δab (∂μ π a )(∂ μ π b ) + · · · , (8.10)
4 2
where the normalisation of π a implicit in (8.8) has been chosen such that the kinetic
term is canonical. The quantity fπ is called pion decay constant. This name can be
easily understood by promoting the partial derivatives in (8.10) to SU (2)L -covariant
derivatives and calculating the mixing of the pions with the weak bosons and hence
their decay rate (e.g. π − → W − → μ− ν μ ).
The fact that our real-world pions are massive is due to the non-zero fundamental
up and down-quark masses. These explicitly break the chiral SU (2) symmetry
which would otherwise have only been broken spontaneously by the fermion
condensate. As a result, the exactly massless bosons of the Goldstone theorem are
turned into pseudo-Goldstone bosons, which are allowed to have a small mass.
After these lengthy preliminaries, the idea of Technicolor is easy to state: Let
us assume that, in addition to the SU (3)c gauge theory with its confinement scale
QCD ∼ 0.2 GeV, there exists an SU (N)t c gauge theory confining at T C ∼
few TeV. Moreover, there are fermions (so-called techni-quarks) which are charged
under both SU (N)t c and under the electroweak gauge group SU (2)L × U (1)Y of
the Standard Model. Let us call them (QL )i and (QR )j , transforming as N and N of
SU (N)t c , respectively. The indices i and j signify some further (e.g. electroweak)
transformation properties.
In complete analogy to the well-understood case of QCD, one expects that a
condensate (QL )i (QR )j = 0 will form. This clearly has the potential to play the
role of the standard model Higgs and to give mass to W and Z bosons. For example,
to be very concrete and following the QCD-example closely, one may introduce two
techni-quark doublets

UL UR
QL = and QR = . (8.11)
DL DR
Here QL is an SU (2)L doublet, uncharged under U (1)Y . The fields UR and DR

are SU (2)L singlets with U (1)Y charge +1/2 and −1/2, respectively. Clearly, a
condensate of the expected type
(QL )i (QR )j ∼ δij (8.12)
will break the electroweak symmetry in the desired way. Indeed, it corresponds
to two Higgs doublet VEVs, both invariant under the same subgroup U (1)em ⊂
SU (2)L × U (1)Y .
Of course, this extremely simple-minded model is far from realistic, even if one
only assumes the more limited data of the pre-LHC era. One of the main reasons is
the need for Yukawa couplings, which in the present approach would most naturally
come from operators like
1
L⊃ (ψ · ψ) (Q · Q) . (8.13)
M2
In this symbolic expression ψ stands for Standard Model fermions and Q for
techni-quarks. All indices and their contractions have been suppressed. The only
point the above expression intends to make is the following: Given that such
four-fermion operators are present in the low-energy effective lagrangian and a non-
zero condensate of techni-quarks develops, Standard Model fermion masses will
in general be induced. It is also clear that now our theory will break down at the
scale M. If one desires a renormalisable quantum field theory potentially valid up
to the Planck or GUT scale, the model has to be extended. We will not study such
constructions but note that they exist in principle. However, an obvious comment is
that the largeness of the top-Yukawa coupling forces the energy scale M to be low,
leading to phenomenological problems. Moreover, similarly to the situation with
SUSY, both the well-established electroweak precision data and the more recent
non-discovery of new physics at the LHC put technicolor under pressure.
For us, the main conceptual conclusion is the following: In addition to low-scale
SUSY, technicolor offers in principle another perfectly viable, technically natural
explanation of a low electroweak scale. Here by low we mean relative to the Planck
scale. Both SUSY and technicolor share a growing ‘little hierarchy problem’. But
assuming the small amount of tuning implied by this is accepted or better models
avoiding it are found, the ‘large hierarchy problem’ remains solved. The underlying
main technical tools are very different: Non-renormalisation of the Higgs mass vs.
logarithmic running of a gauge coupling together with confinement dynamics.
While both types of model appear to fit reasonably well into what we know
about the string landscape, there is at this point a significant difference: Low-scale
SUSY comes, of course, from 4d SUSY at the compactification scale. The latter
emerges (naturally but certainly not unavoidably) from the 10d SUSY. This, in turn,
is apparently enforced on us when trying to consistently quantise the fundamental
string. As a result, one may say that the low-scale-SUSY resolution of the hierarchy
8.2 From the ‘Little Higgs’ to Large or Warped Extra Dimensions 259
problem (as well as its modern and more modest version with SUSY at about 10
TeV) is directly related to the specifically stringy approach to quantum gravity.1
By contrast, technicolor requires only the right set of gauge groups and fermionic
matter to be present at the high (e.g. compactification) scale. Achieving such a
field content in string theory looks perfectly reasonable, but there appears to be
nothing specifically stringy about it. One might want to say that the relation between
technicolor and string theory is a neutral one.
The last two paragraphs hint at a potential problem with the landscape resolution
of the electroweak hierarchy problem: Indeed, to explain that low-scale SUSY has
(so far) not been found, the string landscape has to prefer a higher SUSY breaking
scale. Given the intimate relation between SUSY breaking, compactification and
moduli stabilisation, it is fairly easy to imagine that such a preference exists and can
be quantified by a detailed study of the landscape, especially including the difficult
subject of stringy models of SUSY breaking and ‘uplifting’ from AdS to dS vacua
[94, 261–265].2
However, making the reasonable assumption that technicolor models can be
found in the landscape as well, one would expect that a (technicolor-based) low
electroweak scale should occur in a fraction of models which is not exponentially
suppressed. Such models would then be preferred relative to models with a purely
tuned small Higgs mass and even relative to low-scale SUSY models (if stringy
moduli stabilisation implies a bias against those). Thus, the world above the weak
scale should display a natural and generic variant of technicolor which, however,
it does not. A possible way out may be a bias in the string landscape against large
gauge groups with a chiral spectrum (which are needed for technicolor), but this is
pure speculation.
8.2 From the ‘Little Higgs’ to Large or Warped Extra

Dimensions
This is a good place to comment on a number of further model building ideas that
have been proposed to resolve the naturalness problem of the electroweak scale (see
[360] for an introductory review).
To begin, let us remind the reader that the conceptual reason for the lightness
of the pions in the Standard Model is the Goldstone theorem: They are Goldstone
bosons of the spontaneously broken chiral SU (2). The simplest versions of techni-
color use this idea to generate a Standard Model Higgs as a Goldstone boson of a
1 We should emphasise, however, that low-scale SUSY is certainly not a prediction of string theory.
10d stringy SUSY may be broken directly in the compactification process (e.g. through a non-
Calabi–Yau compactification) or at any energy scale between KK-scale and weak scale.
2 In fact, most work in this area is based on what is known about the statistics of the flux stabilisation
of complex structure moduli. As emphasised e.g. very recently in [267], the Kahler moduli
stabilisation is also crucial. However, at that level of detail one must also consider the statistics
of possible uplifts—a hard subject that is not well understood.
global symmetry acting on a set of techni-quarks. Moreover, technicolor separates

the scale of this symmetry breaking from some high fundamental scale through the
logarithmic running of a non-abelian gauge coupling.
Little Higgs models implement and perfect the idea of a Goldstone-boson Higgs
without the constraints and complications of the fermion condensate [361–363] (see
[364] for a review). The challenge attacked and overcome in these constructions is
to keep the Higgs unexpectedly light (hence the name ‘little’) in spite of Standard
Model gauge and Yukawa interactions. In doing so, one does not ask at the same
time for a UV completion that can be valid up to some very high scale. Things are
then simpler since one does not need to realise the Goldstone scalars as fermion
condensates. Instead, one can just start with a compact field space, like the quotient
manifold M = SU (5)/SO(5) analysed in [365]. This can be viewed as the Little
Higgs-analogue of the group manifold of SU (2) which arises in the low-energy
EFT of the Standard Model and is parameterised by our familiar three pions. Here,
by contrast, the number of scalars is larger: SU (5)/SO(5) is 14-dimensional. Some
of those scalars are removed by gauging, some become massive, but some stay light
and can play the role of the observed light Higgs doublet. However, both the gauging
by the electroweak group SU (2)L × U (1)Y and the Yukawa couplings to Standard
Model fermions have to be introduced and violate the masslessness of all of the
Goldstone bosons. The true model building challenge of Little Higgs models is then
to realise a structure in which this effect is sufficiently small, e.g. because a non-zero
potential for the Higgs degrees of freedom arises only at the 2-loop level. We will
not discuss the (very interesting) details of this.
The overall picture which eventually emerges has three basic energy scales:
They can be characterised starting from the scale f which sets the size of the
compact field space of the Goldstone bosons. This scale is completely analogous
to the pion decay constant of the Standard Model or to the axion decay constant
which defines the volume 2πf of an S 1 parameterised by an axionic scalar. If
some of the Goldstone bosons are removed by gauging, then they become massive
(together with the corresponding vectors). The relevant mass scale is f , at least if
the gauge couplings are O(1). The electroweak scale is much smaller, mEW f ,
with the hierarchy provided by loop suppression factors (1/16π 2) together with the
smallness of Standard Model couplings governing the loops. Finally, the model is
based on a compact field space, which implies the presence of higher-dimension
operators suppressed by powers of the mass scale f . This is most easily seen in
the case of standard model pions by considering (8.8) and (8.10). A loop expansion
involving such higher-dimension operators breaks down at a scale ≡ 4πf , where
the 4π come from the 4d loop suppression factor 1/16π 2 . At this scale , which
is however significantly higher than the electroweak scale, e.g. ∼ 10 TeV, a UV
completion is required. It may be field-theoretic (in the spirit of technicolor), higher-
dimensional or even stringy (see below).
A very different idea concerning the hierarchy problem has emerged in the late
1990s under the name of Large Extra Dimensions [366, 367].3 Specifically, the
scenario known as ‘ADD’ is extremely simple and builds on string-theoretic ideas,
but without explicitly using any details of the stringy UV completion. It proposes
that the world is d-dimensional, with d = 4 + n, and that the n extra dimensions
are large (in the sense of being much larger than the Planck length). The 4d Planck
scale is then given by
MP2 , 4 ∼ MP2+n n
,d R , (8.14)
with MP , d the higher-dimensional Planck scale and R a typical compactification

radius. Based on this, one can envision a situation where R is so large that MP , d
is in the TeV domain. This resolves the large hierarchy problem since, in fact,
the electroweak scale is now of the order of the fundamental UV scale MP , d .
Put differently, the quadratic Higgs mass divergence is cured by the fundamental
quantum gravity cutoff (be that superstring theory or something else). Since this
cutoff is low, no large hierarchy problem exists. However, one clearly needs a
dynamical mechanism that explains why the compact space is stabilised at R
1/MP , d .
While the large hierarchy problem may well be solved in this way, the little hier-
archy problem remains and may, in fact, be more severe than in other approaches.
Indeed, given proton-decay, flavour and electroweak-precision-data constraints, it is
certainly optimistic to assume that the fundamental quantum gravity scale can be
as low as even 10 TeV. This, of course, leaves a sizeable gap to the observed Higgs
mass of ∼100 GeV and hence a significant little hierarchy problem. Concerning this
issue, one may now imagine a combination of model building ideas where the large
hierarchy problem is overcome using the large-extra-dimensions approach and the
little hierarchy problem is solved as in the Little Higgs models discussed above.
We will return to a variant of this further down when talking about warped extra
dimensions.
Before closing this large extra dimensions or ADD discussion, we need to
comment on the number n determining the dimensionality of the compact space.
The choice n = 1 is obviously excluded since R would be way too large. Very
intriguingly, setting n = 2 and assuming MP , 6 ∼ 1 TeV gives R ∼ 1 meV−1 ∼
1 mm. But this was just about the smallest distance at which gravity was at that
time directly accessible to experiments. Hence, even such an extremely large
compactification radius appeared to be on the one hand not excluded but, on
the other, could be discovered in the foreseeable future in gravitational table-top
experiments.
Of course, the attentive reader should immediately object that the 4-dimensional
(rather than 6-dimensional) nature of the world was, even in 1998, already known
to persist up to energy scales of at least 100 GeV. However, this bound could be
3 See e.g. [368, 369] for introductory reviews.

evaded by assuming that all Standard Model particles and gauge fields are confined
on a brane, in this case a 3-brane filling out our 3 + 1 non-compact dimensions
and being point-like in the 2d compact space. With that, the ADD scenario with
n = 2 is complete. Not surprisingly, it was perceived as extremely innovative and
exciting at the time. Unfortunately, it became clear very fast that astrophysical and
cosmological constraints push the lower bound on MP , 6 way above 1 TeV. Also,
constraints on the short-distance behaviour of gravity developed fast, forcing R way
below a mm and, again, MP , 6 to values higher than TeV. This worsens the little
hierarchy problem and disfavours the n = 2 case.
For n ≥ 3, the compactification radius R is much below 1 mm even if MP , d
is kept in the TeV domain. Then so-called fifth-force experiments, testing gravity
in the sub-mm domain, provide no meaningful constraints. Also cosmological and
astrophysical bounds become less prohibitive with growing n. Such scenarios with
n ≥ 3 are still constrained but not hopeless. They do, however, become less and less
believable from the point of view of solving the hierarchy in the same way as SUSY,
technicolor, etc.: The LHC simply keeps pushing any exciting new physics to higher
and higher energies, thereby making the little hierarchy problem more severe.
The next twist in this line of thinking is the very interesting idea of warped extra
dimensions, also known as the Randall–Sundrum model [370,371] (see [372] for
the first warped scenarios). The term ‘warping’ has been discussed in quite some
detail in Sect. 6.5 and its meaning in the present context is the same as before: It
denotes compactifications where the metric in the non-compact directions depends
on the position in the compact space. More concretely, the present type of model is
based on a 5d to 4d compactification on S 1 /Z2 , i.e. on an interval. The metric reads
ds 2 = e−2ky dx 2 + dy 2 , (8.15)
with the warp factor exp(−ky) and y ∈ [0, yIR ] parameterising the extra dimension.
One refers to the interval-boundaries or ‘end-of-the-world’ branes at y = 0 and
y = yIR as the UV and IR brane, respectively. The reason is, as in Sect. 6.5, that
any mass-dimension quantity of fixed value in units of MP , 5 takes a higher or lower
value from the perspective of the 4d observer depending on whether it is located
closer to y = 0 or to y = yIR . The model can be characterised as a slice of AdS5
with two 4d-branes as boundaries. To make this geometry a solution of Einstein’s
equations, a 5d cosmological constant and appropriate 4d brane tensions have to be
added—we will not work out the details of this.
To be precise, the Randall–Sundrum model comes in two variants known as
‘RS1’[370] and ‘RS2’[371]. The brief description above refers to RS1, with RS2
corresponding essentially to the decompactification limit yIR → ∞. This variant,
while conceptually very interesting, has nothing to say about the hierarchy problem
and we will not discuss it.
The relevance of RS1 for the hierarchy problem arises as follows. Let us assume
that the Standard Model is localised at the IR brane. By this we mean, very naively,
adding a piece

√
d 4 x dy −g LSM [gμν , ψ] δ(y − yIR ) (8.16)
to the 4d Einstein–Hilbert action. The metric gμν is the pullback of the 5d metric
to the boundary and the Standard Model fields ψ are only defined at the boundary
locus. Now, due to the warping, it turns out that a fundamental Higgs mass parameter
m2H ∼ MP2 , 5 (in the local action near the IR brane) would be perceived by a 4d
observer as being much smaller than the 4d Planck scale. Parametrically, the 4d
observer finds
mH ∼ MP , 5 e−k yIR , MP2 , 4 ∼ MP3 , 5 L , where L ∼ 1/k (8.17)
is the AdS curvature radius. If k, which is governed by the 5d cosmological constant,

is chosen to be slightly smaller than MP , 5 for control purposes, and the interval
can be stabilised such that k yIR is a largish O(1) number, then the smallness of
the electroweak scale can be explained. The stabilisation problem has an elegant
solution due to Goldberger and Wise [373].
The reader may at this point be confused about how mH can be comparable to
MP , 5 in the microscopic 5d action and at the same time much smaller than MP , 5
according to the first relation in (8.17). The answer is simply that, due to warping,
MP , 5 has no unambiguous meaning in 4d. For example, a 5d Planck scale black
hole would appear to the 4d observer as 4d-Planck scale if it were located near the
UV brane. By contrast, it would be perceived as slightly above the electroweak scale
if it were found near the IR brane.
What one has achieved at this point is very similar to the ‘ADD solution’ of
the hierarchy problem: The large hierarchy between the Planck and electroweak
scale is explained, but the little hierarchy problem is as severe as ever since,
phenomenologically, we cannot afford to take MP , 5 all the way down to 100 GeV.
An enormous amount of work has gone into attempts of improving the RS1 idea in
such a way that it becomes realistic in view of precision, flavour and LHC data or,
more generally, that its little hierarchy problem is ameliorated (see [374–377] for
some of the original papers and [378–384] for a selection of reviews). One possible
way forward is to combine RS1 with the little Higgs idea explained earlier: One
may for example consider placing not just the Standard Model but its little Higgs
extension, with a UV scale in the 10 TeV range, at the IR brane. Now, having MP , 5
near 10 TeV looks much less impossible than near 100 GeV. Moreover, significant
advantages can be gained by turning the Standard Model into a mix of 4d and 5d
fields, with the latter being visible to us only through their lowest-lying KK modes.
Such KK modes can, depending on the 5d mass parameters of the underlying 5d
fields, have a y-profile localised (exponentially) near the UV or IR brane. This opens
up model building possibilities where, for example, the Higgs lives at or near the IR
brane. Heavy fermions like the top quark may also be localised near the IR brane,
naturally giving them a stronger coupling to the Higgs. The light quarks, by contrast,
live mainly in the UV. Consistently with phenomenological requirements, they are
then less affected by large higher-dimension-operators induced in the IR-part of the
model (an issue following from the low-lying local value of MP , 5 ).
The various model building ideas in the RS1 framework briefly described above
have a dual CFT interpretation. To appreciate this, the reader has to recall our very
short discussion of AdS/CFT in Sect. 7.9. There, we characterised AdS/CFT as a
map between a gravitational theory in (d+1)-dimensional AdS and a d-dimensional
CFT living on the boundary. In our case of interest, d = 4 and the global boundary
is R × S 3 . The scale invariance of the CFT allows us to take the radius of the S 3 to
infinity, considering instead an R1,3 -boundary of AdS1,4 . This may, in a first step,
be identified with the RS2 model, consisting just of an AdS space cut off by the
UV brane. The crucial difference to the pure AdS/CFT correspondence is that the
UV brane is at finite distance, which lets it play the role of a physical UV cutoff in
the CFT language. Our variable y is the analogue of the radial variable of formal
AdS/CFT, which in turn corresponds to the energy-scale variable of the CFT.
Now, turning to RS1, the following interpretation can be given in CFT language:
We start from the UV cutoff at y = 0 and move into the y-direction. This
corresponds to moving to smaller energy scales in the CFT. At some point, an IR
cutoff is encountered in the form of the IR brane at y = yIR . More precisely, in
models where the position of the IR brane is stabilised a non-trivial bulk profile
of a 5d scalar (the Goldberger-Wise scalar [373]) has to be present. This profile
determines the value of y = yIR where the IR brane will be encountered. Thus, the
4d dual of the 5d bulk theory is not a CFT but a nearly conformal theory, in which
the slow running of some coupling eventually leads to the dynamical generation of
an IR cutoff. The natural mechanism to think about here is the running of a non-
abelian gauge coupling leading to confinement at the energy scale corresponding
to y = yIR . We can now clearly appreciate that the mechanism by which RS1
explains the low-lying electroweak scale is actually the AdS-dual formulation of
the technicolor idea. The close relation between these two ideas is explored in much
of the literature on the subject cited earlier (see [385, 386] for the fundamental first
steps).
While all of this is deeply connected with string theory, it remains unclear
to which extent the RS1 approach to the little hierarchy problem can be really
implemented in string model building. The basic setting is in fact well known to
arise in the form of the Klebanov–Strassler throat glued to a compact Calabi–Yau,
cf. Sect. 6.5. Yet, the IR region of the Klebanov–Strassler throat is too simple to
house a full-fledged Standard Model. One of the problems with more complete
models (see e.g. [387, 388]) is that the strongly warped geometry is not explicitly
known.
8.3 Cosmological Selection and the Relaxion 265
8.3 Cosmological Selection and the Relaxion
Throughout this course, we have discussed two opposite ideas on how (apparent)
fine-tunings in EFTs can arise: On the one hand, a hidden mechanism (SUSY,
Technicolor, etc.) may be present, such that the fine-tuning is only apparent. On the
other hand, many parameter values may be realised in a landscape of vacua, to be
found in different parts of a multiverse. We then observe a certain parameter value
for anthropic reasons or accidentally. One may want to call this a real fine-tuning.
In this short section, we want to briefly mention a third option which may be
viewed as a compromise between the previous opposite extremes. Namely, it is
conceivable that a landscape of vacua with different parameter values exists, but
not all of them are on the same footing cosmologically. More precisely, the special
(apparently fine-tuned) value we observe may be due to the details of cosmological
dynamics. One may call this option cosmological selection.
Such an approach to the cosmological constant has been suggested long ago [199,
310, 389] on the basis of subsequent brane nucleation events. The idea is to use
a model with 4-form flux and membranes, as in (6.11), where the energy gap
between the different vacua is chosen to be tiny. If the vacuum energy without flux is
negative, one finds a dense discretuum near zero. For appropriate brane tension, the
cosmological dynamics will consist of consecutive jumps to lower and lower energy
until, just after crossing to negative values, the process stops.4 Unfortunately, this
model is not realistic: The exponential expansion in between the last jumps leads to
an unacceptable dilution of matter and radiation in the late universe.
The idea of a cosmological selection of the electroweak scale has been around
for a while [390, 391] and has more recently received much attention in the context
of the Relaxion model [392]. The key ingredient is an axion-like scalar field φ
which controls the Higgs mass. Specifically, one may assume that m2H = m2H (φ)
is a monotonically falling function of φ. This scalar rolls down a potential during
cosmological history, for example during inflation. If multiple local minima are
present, the field will eventually stop in one of them (cf. Fig. 8.1). Crucially, if such
minima are only present in the part of the field space of φ where m2H (φ) < 0, then
the observed Higgs mass parameter will be negative. Moreover, if the dynamics is
such that the field stops in one of the first minima it encounters, then |m2H | will
be much smaller than the ‘natural’ scale determined by the UV cutoff. In short,
the Higgs mass squared ‘relaxes’ cosmologically to a value which is just below the
threshold at which electroweak symmetry breaking first occurs.
4 This is due to the fact that, in AdS, transitions to lower-energy vacua are impossible if the
brane tension is too high. The reason is that, in contrast to flat space, both the volume and the
surface area of an expanding bubble in AdS grow in the same parametric way, proportionally
to R 2 . Thus, for sufficiently high surface tension the expansion of a true vacuum bubble never
becomes energetically favourable.
Fig. 8.1 Relaxion potential

(adapted from [392])
Suppressing the (canonical) kinetic terms of Higgs field H and relaxion φ, the
relevant part of the lagrangian reads
2 − gφ)|H |2 − V (φ) + 4 (H ) cos(φ/f )

L ⊃ −(MH with V0 (φ) = −κ 3 φ + · · · .
0
(8.18)
Here we recognise the φ-dependent mass squared term for the Higgs doublet
H , the perturbative φ-potential V0 and its non-perturbative correction producing
a series of minima. The latter can be generated, for example, if φ couples to a
non-abelian gauge group through the typical axionic coupling ∼ (φ/f ) trF F̃ . The
H -dependence of the prefactor of the cosine can arise if H governs the masses of
fermions charged under this group.5 The key idea is that, if an H -VEV develops,
the non-perturbative effect ∼ cos(φ/f ) turns on, leading to the desired potential of
Fig. 8.1. Here it is crucial to interpret V (φ) as resulting from (8.18) after the Higgs
has been integrated out. The reason why the cosine effect turns on only with a non-
zero H -VEV rests on well-known instanton physics: As long as H is zero, fermions
are massless and, in the presence of massless fermions, no instanton potential is
generated (see e.g. [72,195]). Crucially, the parameters of this setting can be chosen
such that the model is technically natural. A key role in this is played by the shift
symmetry of φ, which is broken only non-perturbatively and by the small parameters
g and κ.
This setting has been discussed intensely immediately after it appeared (see
e.g. [393–397]). It has also triggered a more general interest in cosmological
selection, including for the cosmological constant. The reader may want to con-
sult [398–407] and rethink the original ideas of [390, 391]. An objection one might
have is that of a certain model building complexity involved in creating precisely
the desired type of landscape. By contrast, if superstring theory is the right theory
of quantum gravity, the ‘standard’ string landscape is simply there—without any
choice. Of course, one may also try to study explicitly whether cosmological
selection arises on the basis of the string landscape [408].
5 This is clearly modelled after QCD, with φ the QCD axion and the quark masses depending on
the Higgs-VEV in the standard way. Yet, unfortunately, such a minimalist implementation does not
work phenomenologically and an extra gauge group appears to be required.
8.4 The Swampland Program 267
Fig. 8.2 On the left: String

landscape discretuum filling
out the whole 2-dimensional
plane of EFT coupling
constants λ1 and λ2 . On the
right: The region λ2 > λ1 is
forbidden
8.4 The Swampland Program
The Swampland is, by definition, the set of apparently consistent EFTs including
gravity which are not found within the string landscape [409,410] (see [411,412] for
reviews). The qualification ‘apparently consistent’ means that the EFT in question
meets all consistency requirements which a low-energy observer not concerned
with quantum gravity can impose. Thus, the Swampland program emphasises the
following remarkable point: In spite of the enormous size of the landscape, not
every field-theoretically reasonable model can be UV completed in string theory.
A popular more general definition proposes that the Swampland consists of
those low-energy EFTs which cannot be UV completed in any model of quantum
gravity, not just in string theory. The difficulty with this definition is that we have no
overview of possible quantum gravity models and that the attempts that exist outside
string theory are even less well understood than the string landscape.
At first sight, one might be very impressed with the strength of the claim that
10272,000 flux vacua [256] are not enough to realise any reasonable EFT. However,
on second thought this is obvious since the landscape is discrete.6 By contrast, the
space of EFTs is continuous due to the continuous choice of couplings or operator
coefficients. Hence, almost any EFT is in the Swampland.
In fact, the Swampland paradigm attempts to make a slightly different and far less
obvious point: It attempts to rule out whole classes of EFTs based on certain general
features. An illustration is given in Fig. 8.2: On the left, we see how the string
landscape discretuum may essentially fill the whole plane of two EFT coupling
constants λ1 and λ2 . While not every combination of λ1 and λ2 is realised, the
difference between the space of all EFTs and the landscape is clearly very hard to
probe experimentally. By contrast, the r.h. plot shows a situation where the region
λ2 > λ1 is forbidden, possibly due to a Swampland constraint. In this case, a single,
not even very precise measurement of λ2 > λ1 may in principle at once rule out
string theory as the correct UV completion of quantum gravity in the real world.
To make our discussion somewhat more concrete, let us briefly go through the
most popular swampland constraints. While some of them are extremely plausible,
it is probably fair to say that, at the moment, all of them remain conjectures.
6 At least if we count any N ≥ 2 SUSY moduli space as a single theory, which is presumably
justified since the moduli are massless, dynamical fields rather than parameters.
First, it is widely believed that no global symmetries can occur in consistent

theories of quantum gravity [413–418]. This expectation is based on the violation
of global charge in black hole evaporation and a CFT argument in perturbative
string theory [413]. There is also a proof in AdS/CFT relying on the splittability
assumption [418].
Then there is a set of conjectures about the geometry of moduli space, proposed
in the seminal papers [409, 410] that defined the term ‘Swampland’. Among those,
we emphasise the conjectures that the moduli spaces are infinite and that, as one
moves to infinity, a tower of light states becomes exponentially light (Swampland
Distance Conjecture). These statements are to be read using the mathematically
natural metric on moduli space, which is also the metric defining the kinetic term
for the corresponding scalar fields in the 4d lagrangian. Given what we know about
string compactifications, it is hard to imagine how the distance conjecture could be
false: For example, one can always take the volume modulus to infinity, in which
case the tower of KK modes becomes exponentially light. As another option, one
can take the dilaton to infinity, i.e. gs → 0, implying that the tower of string
excitations becomes light. In fact, these two options—light KK modes or light string
states—appear to represent an exhaustive list [419]. For a more quantitative (refined)
version of the distance conjecture see [420].
The swampland conjectures described above are, however, not quite as appealing
phenomenologically as envisioned in Fig. 8.2, where EFT couplings are directly
constrained. A conjecture of this latter type is the Weak Gravity Conjecture [421],
which states that gravity is always the weakest force. Concretely, the statement is
that in the presence of gravity any U (1) gauge theory must come with a light charged
particle of mass
mQ ≤ 2 g |Q| MP . (8.19)
Here g is the gauge coupling, Q ∈ Z the charge of the particle and the precise
numerical coefficient is chosen such that equality arises (for |Q| 1) precisely
if the charged object is an extremal Reissner–Nordstrom black hole. An equal sign
in (8.19) would also mean that two particles of this type are attracted by gravity
precisely as strongly as they are repelled by their common U (1) charge. In this
sense the inequality really deserves the name Weak Gravity Conjecture.
There are different motivations for this conjecture, one of the most popular being
that, if it were false, then extremal black holes would be absolutely stable. This may
be problematic [422], though no sharp argument for the Weak Gravity Conjecture
has so far emerged from this line of reasoning. Another argument for the conjecture
is that the limit g → 0 should be forbidden because it would generate a global
symmetry. The Weak Gravity Conjecture then quantifies what exactly goes wrong
with taking such a limit. Maybe most importantly, the conjecture is supported by
all controlled stringy examples, and this is at least superficially easy to understand:
Indeed, consider a D-brane model and try to take the gauge coupling to zero. The
only way in which this can be done is by either sending the brane volume or the
dilaton to infinity. But in both these cases the string scale goes to zero in Planck
8.4 The Swampland Program 269
units,7 such that one may say that (8.19) is trivially fulfilled because the 4d EFT
cutoff falls below the energy scale g MP .
Note that our statement of the conjecture in (8.19) was an enormous oversim-
plification. We have disregarded that many different versions of the Weak Gravity
Conjecture are discussed. For example, one may demand that some charged particle
satisfying the conjecture exits (mild form) or that the lightest charged particle should
do so (strong form).8 These options have already been considered in [421]. More
recently, a lot of additional effort has been devoted to the Weak Gravity Conjecture
and its extensions. This revival of the Swampland discussion (see e.g. [424–432])
has in part been triggered by an increased interest in the observational signals of
cosmological inflation, which have at that time started to constrain the magnitude
of primordial gravitational waves and, through this, the field range of the inflaton.
This, in turn, has a surprisingly direct connection to the Weak Gravity Conjecture
and the Swampland program, as we now briefly explain.
The point here is that the Weak Gravity Conjecture has a natural extension to
p-form gauge theories with p = 1. One then basically constrains the tension of the
charged (p−1)-brane in terms of coupling strengths and MP , in complete analogy
to (8.19). Specifically for an axion, viewed as a 0-form gauge theory, the coupling
strength is ∼ 1/f and the role of mQ is taken over by the instanton action Sinst . One
then has
Sinst MP /f ⇒ f MP , (8.20)
where the implication rests on the (non-trivial) assumption that Sinst 1. This
is motivated by the desire to use the dilute instanton gas approximation. One sees
that, interpreted in this way, the Weak Gravity Conjecture limits the allowed field
range of axions. This restricts the model of so-called natural inflation, which in
its simplest form relies on ‘superplanckian’ axion field ranges to realise large-field
inflation [433] (see [290] for a review).
Such possible limitations of axionic field ranges in string theory (see also [434])
may in principle be overcome by so-called axion monodromy inflation [435–438]
or its modern version, F -term axion monodromy [439–441]. The underlying idea
here is to break the axion periodicity weakly. As a result, the circular field space
turns into a spiral which rises slowly to higher and higher potential energy. Yet,
these models may turn out to be in the Swampland, either because of concrete
model building difficulties or due to generic constraints, such as a sufficiently strong
7 To be precise, one may also consider a double scaling limit where the space transverse to the
brane shrinks such that the Calabi–Yau volume does not diverge together with the brane volume.
One then needs to change duality frames to maintain control and see the emerging light states
[423].
8 The maybe most naively expected formulation that the particle with Q = 1 should satisfy the
conjecture has counterexamples. However, in all such counterexamples the lowest charge Qmin at
which the conjecture holds exceeds unity only by an O (1) factor.
version of the Swampland Distance Conjecture (see e.g. [420, 442]). Much further
interesting work has recently been done in the context of developing and connecting
various forms of the Weak Gravity and the Swampland Distance conjecture, see
e.g. [423,443,444]. Concerning the applicability to inflation, things remain unclear:
On the one hand, one may indeed hope that the tower of light states coming down at
superplanckian field excursions constrains models of inflation. On the other hand,
realistic large-field inflation needs only field ranges of the order (f ew) × MP .
Such modestly transplanckian field ranges may turn out to be consistent with all
reasonable conjectures.
Note that it is also conceivable to simply break the proposed inequality on the
r.h. side of (8.20) in a concrete model. One idea is to start with the field space
of two axions, say a T 2 with volume (2πf )2 and f < MP . All one needs is to
realise a scalar potential on this T 2 which forces the lightest effective field on a
spiralling trajectory [445]. This trajectory, while still periodic, may clearly be much
longer than 2πf . The required potential could be realised by the interplay of several
instanton-induced cosine-terms.
Even simpler, such a long spiralling or winding trajectory may be enforced by
making a certain combination of the two axions massive by a flux choice [430].
This has an interpretation as the ‘Higgsing’ of the axion or 0-form gauge theory
with the help of a (−1)-form gauge theory (as explained near (6.14)), cf. [446].
In fact, more generally, it has been pointed out that this method of Higgsing a p-
form gauge theory with a (p − 1) form gauge theory apparently represents a field-
theoretic method of breaking the strong form of the Weak Gravity Conjecture in the
IR, including in the ‘classic’ case of p = 1 [447]. Here by Higgsing we mean the
substitution
1 1 1 1
|dAp |2 + 2 |dAp−1 |2 → |dAp |2 + 2 |dAp−1 + Ap |2 .
gp2 gp−1 gp2 gp−1
(8.21)
In the case p = 1, this is clearly the standard meaning of the term Higgsing, where
A0 represents the radial direction of the conventional complex scalar which Higgses
a U (1) gauge theory. Now, starting with two gauge fields, say A(1) (2)
p and Ap , one
(1) (2)
may Higgs the linear combination Ap + NAp , where N is a large integer. It
is then easy to see that, for the surviving p-form gauge theory the Weak Gravity
Conjecture in its strong form will be broken in the IR [430, 447]. The √ reason is
basically that the effective gauge coupling is lowered from gp to gp / |N|. This
effect can even be made exponentially strong using the so-called clockwork idea
[448, 449]. If such constructions are possible in the landscape, this clearly weakens
the phenomenological relevance of the Weak Gravity Conjecture as an IR constraint.
It could also turn out that models of this type are in the Swampland and the Weak
Gravity Conjecture remains strong.
8.5 The Swampland and de Sitter 271
8.5 The Swampland and de Sitter
The Swampland program has many aspects. Proceeding largely in historical order,
the previous section emphasised its more phenomenological side: constraints on
global symmetries, weak gauge couplings and fields ranges, with a view on inflation.
There is also a more mathematical or conceptual side, focussing on what models
can arise from string theory as a matter of principle, without immediately asking for
relevance to the real world (see e.g. [412]). We will not discuss this here.
Instead, we now turn to a more recent and possibly the most important aspect
of the Swampland discussion. It is interesting conceptually but, in addition, has
very far-reaching phenomenological implications. The conjecture we are referring
to states that all de Sitter solutions, even metastable ones, are in the Swampland
[217,218,302,303].9 A particularly intense debate has initially surrounded the very
strong conjecture that [302]
|V |/V ≥ c , (8.22)
with c an O(1) constant and MP = 1. This clearly rules out de Sitter minima,
but is actually much stronger by also excluding de Sitter maxima, i.e. unstable de
Sitter solutions. This is presumably too strong since it collides with the Standard
Model Higgs potential [451, 452], the EFT of pions [453] and with relatively well
established string constructions [454].
A refined form of the de Sitter conjecture [218, 303] and an attempt of a first-
principles derivation [218] (see however [455, 456]) have subsequently appeared.
The refined formulation states that either
|V |/V ≥ c or V /V ≤ c , (8.23)
which is roughly the opposite of the slow-roll requirement. The derivation of [218]
aims only at establishing the claim at asymptotically weak coupling (basically at
large-field distance, e.g. at asymptotically large volume). It uses the Swampland
Distance Conjecture, but also relies on strong assumptions about the origin of de
Sitter entropy. Unless this argument can be made water tight, counter examples are
conceivable even asymptotically [455, 456]. But more importantly, for the actual
string landscape an asymptotic validity of (8.23) is not threatening: The landscape
as we know it certainly needs a very large set of metastable de Sitter vacua, but
it is perfectly acceptable for this set to nevertheless be finite. A series of vacua
extending to zero coupling is not required. The possibility of achieving arbitrarily
weak couplings is essential for a rigorous mathematical proof, but in physics this
may simply be too much to ask for.
Let us start the more detailed discussion by explaining why a very strong
conjecture such as (8.22) might appear appealing. Quite generally, compactifications
9 For a critical review see e.g. [450].

lead to 4d effective lagrangians of the type

(∂V)2
L ∼ V R4 − −E , (8.24)
V2
where we have suppressed O(1) coefficients for simplicity. The overall volume
factor V multiplying the 4d Ricci scalar R4 and everything else comes from the
integral over the compact space. But the volume also figures as a dynamical field,
and its kinetic term features a logarithmic derivative. Moreover, we have included a
positive energy source E (one may think of a SUSY breaking effect). After Weyl-
rescaling to the Einstein frame and introducing the canonical field φ = ln(V), one
finds

L ∼ R4 − (∂φ)2 − E e−φ . (8.25)
The potential is V (φ) ∼ e−φ , which does indeed satisfy (8.22). This will remain
the case if one goes more carefully through different types of simple explicit string
compactifications.
Let us make a slightly different but related point about creating dS space through
a compactification. In the limit V → ∞, all potential terms (from fluxes, 10d
curvature, SUSY breaking effects) tend to zero. Indeed, in this limit one approaches
uncompactified, flat 10d space, so one expects no energy density. Such an argument
has first been given for asymptotically weak string coupling rather than for large
volume [457] and is known as the Dine–Seiberg problem. Now, even if one has
two independent such terms, each of which approaches zero as φ = ln(V) → ∞,
their sum will in general not create a dS minimum (Fig. 8.3). Indeed, if both terms
are monotonically falling and have the same sign, the sum will also be monotonic.
If they have opposite sign, then depending on which term scales more strongly with
V the potential may first fall and then approach zero from below, as shown in the
figure. It may also first grow and then turn over to approach zero from above. This
Fig. 8.3 Left: The sum of two simple falling potential terms allows only for AdS, not for dS
minima. Right: Involving a third simple, monotonic term a metastable dS minimum can be created
with some tuning of coefficients
8.5 The Swampland and de Sitter 273
would give a de Sitter maximum. Thus, as long as every term has a simple scaling
behaviour with V, it requires the interplay of at least three such terms to realise a
dS solution. Moreover, these three terms, each with a different fall-off behaviour
in 1/V, need to be of the same order of magnitude to realise such a non-trivial
potential. But since the coefficients are O(1) numbers (string theory not having
free parameters), the de Sitter minimum can clearly not be at asymptotically large
volume (see also [218]).
However, as already emphasised above, phenomenology and the landscape in
general may not need asymptotically large volume or weak coupling. The scenarios
of KKLT [192] and LVS [193] show precisely how, through the interplay of three
different terms falling with 1/V, a non-trivial de Sitter minimum may in principle
arise. Thus, we are in the end faced with the, admittedly hard, problem of evaluating
the reliability of concrete proposed string constructions of 4d metastable de Sitter
EFTs. To liberate us from that challenge and ‘kill’ the dS landscape as a matter of
principle one would need harder arguments against de Sitter in string theory than
those given above.
This may be a good place to mention an older no-go argument about what
can or cannot be achieved in string compactifications. It has been shown in [458]
that perturbative compactifications involving fluxes, positive-tension branes and
warping cannot lead to 4d de Sitter. Minkowski space can only be achieved if fluxes
and warping are turned off, which is of course not interesting. The key loophole
utilised by GKP [133] is the existence of negative-tension objects, for example
O3 planes, which make warped compactifications leading to no-scale Minkowski
space possible. This is of course the basis on which then, adding non-perturbative
effects and anti-branes in warped throats, KKLT suggested what still stands up as
the simplest realistic proposal for string-derived de Sitter.
Let us note that KKLT has come under concrete criticism in the Swampland
context on a basis of arguments related to the Maldacena–Nunez no-go theorem
mentioned above [458]. It has been pointed out in [235] that, re-running the logic
of [458] while including the 10d effects of gaugino condensation apparently leads
to problems. However, a more careful analysis of the relevant action of the brane
stack on which gauginos condense [236,237] shows that a 10d description of KKLT
(usually only derived by 4d EFT methods) may be possible [238, 240] (see however
[239]). This has been further developed, again with positive outcome for KKLT
[459] (or at least for the AdS part of the construction [460]).
An interesting new KKLT criticism [240] observes that in the relevant parametric
regime the warped throat housing the D3 uplift is in general too large to fit into
an uwarped Calabi–Yau. It has been argued in [241] that the resulting strong
warping effects in the bulk of the compact space imply a ‘bulk-singularity problem’,
quantitatively ruling out at least the simplest, generic versions of KKLT. This is
clearly too recent to treat it as a conclusive statement. Moreover, by their very nature
the above issues do not extend to LVS-type models, which allow for a much larger
compact volume and hence have no problems related to strong warping.
Yet another recent set of Swampland conjectures, dealing with constraints on
AdS compactifications, may affect de Sitter constructions [461–463]. It states
(in versions of varying strength) that scale-separated AdS compactifications are

impossible. The basis is the observation that, in many well-studied string com-
pactifications to AdS, the AdS space and the compact space share the same radius.
For example, this is famously the case in AdS5 × S 5 , the geometry of the original
AdS/CFT proposal. The absence of a separation between the AdS and KK scale
implies that a proper, purely-4-dimensional EFT regime is missing. This may
imply difficulties for the so-called uplift to de Sitter, which in the simplest case
is conceived as the addition of small positive energy density to a well-defined AdS4
EFT. The absence of a 4d EFT regime does not a priori exclude an uplift, but it
forces one to analyse it in the full 10d theory.
Concerning specifically KKLT, a conjecture which is strong enough to
exclude scale-separated AdS in general does, of course, also exclude KKLT:
Recall that, before the uplift, KKLT has a 4d AdS scale ∼ |W0 | and KK scale
∼ 1/ ln1/2(1/|W0 |) with |W0 | 1. Given that this first step of KKLT is rather well
established and the possibility of fine-tuning |W0 | to parametrically small values
has been convincingly argued for [212] (see also [464]), this may speak against
the scale-separated AdS conjecture rather than against KKLT. Moreover, a class of
type-IIA compactifications to scale-separated AdS has been proposed some time
ago [465] (see also [466, 467]). These constructions are purely perturbative and do
not rely on the tuning of fluxes.
In summary, while doubts about the existence and calculational control of stringy
dS constructions are justified, they do not appear strong enough at present to
overthrow the landscape idea as it has been developed since about the turn of
the millennium. The doubts receive support from the relative complexity of the
simplest concrete models. However, as we argued, the need for such a complexity
follows simply from the parametric analysis of sums of decaying functions of the
volume (cf. Fig. 8.3). If, which is certainly possible, metastable de Sitter vacua
turn out to be inconsistent with string theory, a serious phenomenological problem
arises due to the observed cosmological dark energy. While a rolling scalar or
‘Quintessence’ [468,469] is certainly an acceptable way out [302], the realisation of
such a scenario in string theory is not without its own problems (see e.g. [470] and
[452] for a review). In particular, the technical difficulty of realising sufficiently
strong SUSY breaking in a moduli-stabilised compactification appears to be as
severe in quintessence as in de Sitter constructions [471]. As a result, one may doubt
that phenomenologically viable quintessence models can but metastable dS vacua
cannot be realised within the class of presently understood string compactifications.
Needless to say, research towards establishing or disproving known de Sitter
constructions or developing viable alternatives (quintessence or other [472]) must
go on.
8.6 More Direct Approaches to Quantum Gravity 275
8.6 More Direct Approaches to Quantum Gravity
Before coming to the end of this section and the whole set of lecture notes, we
now want to change perspective drastically. Let us recall how far from established
physics we have come in studying issues like the construction of 4d de Sitter space
through the compactification of 10d superstring theories. We were, under certain
assumptions, forced on this path by the desire to find predictivity in the UV and
hence the need to control UV divergences in quantum gravity. But promoting point
particles to strings may not be the only option for achieving this goal. Which
implications for the hierarchy problems and for low-scale physics in general would
follow if gravity could be quantised in a simpler, more direct approach?
As we already mentioned, one may treat quantum gravity as just another gauge
theory, with a spin-2 particle called the graviton being the propagating degree of
freedom [57]. In many respects this works perfectly below the scale MP , but it
proves impossible to raise the cutoff above the Planck scale. This is at least one
way to characterise the problem. An alternative perspective is that it is perfectly
acceptable to take the cutoff to infinity, but the price to pay is an infinite set of
operators, suppressed by growing powers of 1/MP2 , with unknown coefficients. All
these operators are necessary to absorb the divergences of perturbation theory and,
as the net result, one is again limited to a quantum effective theory below MP .
Now, the above impasse may clearly be a phenomenon of perturbation theory
and a non-perturbative definition may lead to a well-defined theory, potentially even
with the option of taking the limit → ∞. The preferred cutoff for a perturbative
treatment of gauge theories, dimensional regularisation as used in [57], is not
suitable for a non-perturbative definition. One obvious alternative, supported by its
success in fixed-space QFT, is the method of discretisation of spacetime, i.e. ‘the
lattice’. Of course, given that gravity makes space itself dynamical, it appears
mandatory to make the lattice dynamical rather than using a fixed (e.g. hypercubic)
lattice as is common in QFT. Thus, one should be studying the dynamics of or a
functional integral over triangulations.
This class of approaches goes back to what is known as Regge calculus [473]
(see [474] for a review and early references). The idea is clear if one visualises how
a 2d manifold can be approximated by a collection of (flat) triangles glued at their
edges. Curvature is now localised at the vertices, where 3 triangles meet, and can be
quantified by the respective deficit angles. This clearly generalises to 3d manifolds,
which can be analogously modelled by tetrahedra. The latter are now glued at faces
(triangles) and additionally meet at edges and vertices. The general terminology
would be that of simplices, in this case 3-simplices glued at 2-faces.
To describe 4d gravity, the above discretised 3-manifold has in some way to be
supplemented with time. One option would be to allow the discrete 3-manifold to
depend dynamically on a continuous time variable. Another possibility is to make
time steps discrete. A natural option in this latter case is to allow the number
of 3-simplices to change from one time-slice to another. In this way, one is in
effect triangulating a Lorentzian or (3+1)-dimensional manifold. Similarly, one
may consider discretised 4d Riemannian manifolds if one is interested in euclidean

gravity. Supplementing such a discrete description of spacetime with an action
principle, one can use this as an approximation method for classical numerical
relativity. But our interest here is based on the hope that this line of thought may
define quantum gravity. Both canonical and functional-integral quantisation can be
considered in this context. Finally, while this is not logically necessary, one may
desire to take a continuum limit, in which the typical length of an edge becomes
much smaller than the Planck length. Crucially, macroscopic dynamics should
continue to be correctly described in this limiting procedure.
A central role in present-day research on lattice or discrete quantum gravity is
played by the method of Causal Dynamical Triangulations or CDT [475] (see
e.g. [476, 477] for reviews). The key new term here is ‘causal’, by which one means
that the causal structure coming with flat Lorentz space is respected. In other words,
the topology of all spatial slices of the 4d lattice is the same. Maybe more intuitively:
The emission of baby universes is forbidden. By this we mean a process by which
our universe, say with S 3 topology, emits a small 3-sphere, such that the late-time
spatial topology is S 3 × S 3 .
This is of course quite different from what happens on the worldsheet of the
string: Here, the famous ‘trousers’ or ‘pair of pants’ geometry, corresponding to the
decay of one particle into two, is a central part of the theory. Similarly, topological
transitions are an important aspect of 10d dynamics, allowing for example the
continuous deformation of one Calabi–Yau into another. Yet, it is also clear that
with topology change comes potentially an explosion of the number of geometries
to be considered in the path integral.10 Thus, CDT is much more tractable than
an unrestricted euclidean path integral over triangulations. Still, even with this
simplification there is so far no established result about 4d Einstein gravity emerging
from a continuum limit of a discrete gravitational theory (see [485] for a very recent
discussion).
A different though not unrelated approach to the quantisation of general relativity
is Loop Quantum Gravity (LQG). The perspective adopted and the methods
used differ very much from the rest of these notes, such that even a superficial
introduction is impossible. We will only try to say a few words about the basic idea
and refer the reader to the various introductory texts on the subject, see e.g. [486–
492].
The approach is based on what is widely known as Ashtekar variables for
the canonical formulation of gravity and its subsequent quantisation [493, 494]. To
explain this, one has to look at 4d spacetime as a slicing, with each slice being a 3d
10 We note in passing that there was at some point much excitement about the phenomenological
relevance of baby universes [478, 479] (see [198] for a review). But no completely convincing
calculational approach emerged. By contrast, in the simpler 2-dimensional case, much progress
concerning the euclidean path integral treatment of quantum gravity has been made, also outside
the very special case of the critical string. This is in particular due to matrix model techniques
[480–482] (for reviews see [483, 484]). It remains unclear how much of this carries over to 4d and
we have no time to further comment on this rich field.
8.6 More Direct Approaches to Quantum Gravity 277
spatial manifold with the pullback metric. The general holonomy group on the latter
is SU (2), such that the Hilbert space may be built using the parallel transports along
loops in these spatial hypersurfaces. These are not just elements of SU (2) but, since
we are dealing with quantum mechanics, functions on SU (2). The Hilbert space of
the latter can be described using the series of all SU (2) representations, classified by
spin. The total Hilbert space of course also involves all possible loop configurations.
The canonically conjugate classical variables are related to the embedding of the
spatial surfaces in the 4d spacetime, involving in particular the extrinsic curvature.
We will not attempt to explain this and the related construction of a Hamiltonian. But
we should recall that we have already superficially met a situation of this type when
we mentioned the Wheeler–DeWitt equation and wave function of the universe in
Sect. 7.9.
In the present context, it is crucial that Ashtekar variables make the problems of
canonical quantisation of gravity more manageable, at least at some initial level. The
resulting theory of LGC is nevertheless complicated. As was already the case with
the triangulation-based approaches mentioned earlier, the crucial limiting procedure
by which an approximately flat 4d spacetime with the familiar dynamics of the
Einstein–Hilbert action should emerge remains problematic.
We recall that, not surprisingly, it is much simpler to approach quantum gravity
from the perspective of standard low-energy EFT. Here, the Hilbert space is the
Fock space of spin-2 particles, with interactions introduced in perturbation theory.
It is this approach which relates most directly to string theory where (let us say
for simplicity in 10d) the string interpretation resolves the UV problem of loop
corrections. However, the criticism that may be raised at this point is that of so-
called background dependence. In other words, in the perturbative approach and
its stringy UV completion, one starts on a given background, in the simplest case
10d Minkowski space. The string worldsheet relies on this background for its very
existence, for example because its fields, including the 2d metric, come from the
embedding in target space.11 In this sense, approaches to quantum gravity like CDT
and LQG may to some extent claim background independence as an important
merit. Yet, the price that has so far to be paid is the difficulty of connecting to
Einstein gravity at large length scales.
But one goal of this section is to arrive at another important distinction
between the canonical approaches just mentioned and string theory. It is related to
phenomenology and the hierarchy problem. Namely, if one of the former approaches
were fully successful, i.e. if it could derive a low-energy EFT from a simple
quantised model at the Planck scale, another problem is expected to arise:
On the one hand, one could arrive at a unique low-energy theory which is not
the Standard Model. This would simply be the end of the route taken. On the other
11 Of course, the quantisation of the string also provides the massless 10d graviton. Then,
condensates or coherent states of the relevant string excitations are capable of describing, at least
in principle, the full dynamics of 10d target space. We leave it to the reader to explore the relevant
literature, keeping the keywords string field theory and tachyon condensation in mind.
hand, one could discover ambiguities (such as the choice of matter fields and their
couplings) in the UV, enabling one to fit the low-energy EFT to the Standard Model.
This would in some sense be satisfactory, but it would also leave key questions about
the fundamental laws unanswered. In particular, it may then be viewed as highly
unsatisfactory that certain UV parameters would have to be tuned with the enormous
precision required to describe the small cosmological constant and Higgs mass. Of
course, we cannot rule out the third possibility that the low-energy EFT will be
unique and it will be precisely the Standard Model, with just the right apparent fine-
tuning. From what we presently know about how a UV model produces low-energy
observables, this would appear miraculous.
Indeed, visualise a set of simple formulae without free parameters predicting all
operator coefficients in an EFT at some high scale μ:
m2H (μ)/MP2 (μ) = f (μ) , λ(μ)/MP4 (μ) = g(μ) , mν,R (μ)/MP (μ) = h(μ) , ··· .
(8.26)
Here we have explicitly displayed Higgs mass parameter, cosmological constant

and right-handed neutrino mass, and we have chosen to make all these parameters
dimensionless using the Planck scale. By thinking in terms of an EFT at scale μ
we assume that a UV cutoff at or slightly above the scale μ is imposed. We may
moreover think of our theory as being compactified, for example, on a 3-torus of
radius just slightly below 1/μ.
Now, we also know that there are highly non-trivial, all-loop formulae relating the
above parameters to the eventually interesting quantities m2H (0)/MP2 (0), etc. This
last step of high-scale to low-scale evolution is entirely independent of the quantum
gravity theory producing the functions f , g, etc. It is very hard to see how, through
the disturbances of this last EFT evolution from energy scale μ to energy zero, the
required fine-tuning or apparent fine-tuning should be produced by the fundamental
theory. It is at this point where one might be tempted to prefer a fundamental theory
producing a landscape to one producing a unique field content and set of functions
f , g, etc. Clearly, in the latter case these fundamentally predicted functions would
have to be very special. For example, they must conspire to ensure that λ(0)/MP4 (0)
vanishes with very high precision. In particular, this high-precision almost-zero
result must appear after integrating out the electroweak sector and QCD.
8.7 Asymptotic Safety and the Hierarchy Problem
Finally, let us come to yet another perspective on quantum gravity which, as we

will discuss, may be viewed as being closely related to the triangulation approach
discussed above. It goes back to Weinberg [495] and can be formulated in a generic
way, viewing gravity as a quantum EFT with UV cutoff, without yet committing to
a specific technical implementation of the latter.
To explain this, recall that our best examples of well-defined quantum field
theories are, like QCD, asymptotically free: They run to a trivial RG fixed point in
8.7 Asymptotic Safety and the Hierarchy Problem 279
the UV, where the theory becomes non-interacting. One may say that it is this fixed
point which allows one to remove the cutoff completely, → ∞, thus making
the theory well-defined and predictive on all energy scales. Of course, this is only a
special and particularly simple example for a well-defined QFT: One may equally
well have a non-trivial or ‘interacting’ UV fixed point, as in the case of 4d N = 4
Super-Yang Mills theory. This theory is conformal, which means in particular that
the beta functions of all operators in the lagrangian vanish.12 To make the situation
more interesting, one may add some relevant operator, such as a mass term, to this
theory. One will then have a model with a non-trivial RG evolution and a UV-
definition in terms of an interacting fixed point. Such situations can also arise in
field theories without supersymmetry, which in some cases simply happen to flow
to a non-trivial fixed point in the UV [497].
With this is mind, one may now ask (as Weinberg did much before all of the
above examples), whether a similar situation might arise in gravity. In other words,
could gravity, which is clearly not asymptotically free, instead be asymptotically
safe by running into a non-trivial UV fixed point. More generally, this could be
a so-called fixed surface: A set of scale-invariant theories, parameterised by a
(hopefully finite) set of parameters. The low-energy theory that we observe would
then be defined by one RG trajectory taken from a continuum of such trajectories.
For example, if the fixed surface were just a fixed line, it could be that we are free to
choose the ratio of λ/MP4 in the IR. All other (higher-dimension) operators would
then be predicted. This prediction would follow from the requirement that, in the
UV, the RG trajectory of our theory of gravity hits the fixed line just described.
So far, this is general enough to include the case of a gravitational theory
defined by some form of triangulation. The set of possible continuum limits would
correspond to the above fixed surface. However, today the term asymptotic safety
is frequently used for a specific and rather different implementation of these
general physics ideas. This implementation is based on the concept of the Exact
Renormalisation Group or ERG. To explain the idea, recall the standard textbook
knowledge that a QFT is defined by a (Wilsonian) effective action S [φ]. By this we
mean that a path integral with cutoff and the above action in the exponent defines
all correlation functions. Keeping the theory unchanged, one may vary the cutoff,
in which case the functional S will vary or run with . The general idea of an RG
evolution of a whole action rather than separate coefficients can be implemented in
various explicit forms, e.g. as the Polchinski equation [498] (cf. the discussion in
[499, 500]). A related form which has proven to be particularly useful in the present
context is the Wetterich equation [501] for the so-called effective average action,
which we now explain. We follow the particularly compact and clear discussion in
[502] (see e.g. [503–505] for pedagogical introductions):
12 This is by now a standard fact. The reader may explore the original references using e.g. [496].
Let the theory be defined by some microscopic action, the interacting part of
which is denoted by Sint :

1
Z[j ] = Dφ exp − φ(−p)(p2 + m2 )φ(p) − Sint [φ] − φ · j . (8.27)
2 p
Suppressing IR fluctuations, one may write a partition function with IR cutoff k,

1
Zk [j ] = Dφ exp − φ(−p)(p2 + m2 + Rk (p2 ))φ(p) − Sint [φ] − φ · j .
2 p
(8.28)
Here the cutoff function Rk (p2 ) vanishes for p2 k 2 and diverges for p2
k 2 . Then, one may in standard fashion define an effective action k [φ] by taking
the logarithm of Zk and performing a Legendre transformation. Intuitively, one can
think of k as of a ‘coarse-grained effective action’: it encodes the information
about the theory after dynamics on length scales below 1/k has been integrated out.
The definitions imply that k approaches the microscopic UV action as k → ∞
and the standard quantum effective action, including fluctuations on all scales, as
k → 0. Thus, the dependence of k on k represents one way of characterising the
RG evolution of the theory.
It turns out that one may derive a simple evolution equation for k [501]:13
−1 d
d 1 (2) (2) δ[φ]
k = tr k + Rk Rk with k [φ](p, q) ≡ .
d ln k 2 d ln k δφ(p) δφ(q)
(8.29)
Of course, this simplicity is somewhat misleading since in practice one has to use
some form of approximation for k , for example a derivative expansion truncated
after a finite number of terms. One is then dealing with a growing system of
differential equations, depending on how many terms one wants to keep. The
observation of [506], underlying much of the ongoing work in the literature (for
a very recent review see [507]), is that a non-trivial UV fixed point arises on
the basis of the first few terms in k : the cosmological constant, the Einstein–
Hilbert term and the first higher-curvature terms. More precisely, the dimensionless
couplings λ/k 4 , MP2 /k 2 , etc. appear to run to a finite O(1) values as k → ∞. The
interesting trajectories are those where this happens together with realistic limits for
the corresponding coefficients at k → 0, such that one also finds Einstein gravity
in the IR. As described before, one would ideally hope that a finite-dimensional set
of such trajectories exists. The dimension would be that of the UV fixed surface.
All further parameters (i.e. operator coefficients) of the IR theory would then be
predicted.
13 By slight abuse of notation, Rk now denotes the diagonal matrix Rk (p2 )δ 4 (p − q).
8.7 Asymptotic Safety and the Hierarchy Problem 281
It is probably fair to say that the above picture has a number of open issues. One is
the unavoidable truncation of k . The problem is that, while some form of truncation
is technically necessary, this does not represent a controlled approximation: one is
missing a small expansion parameter in the strong coupling regime (k MP ) of
gravity. Next, the diffeomorphism invariance of the cutoff is clearly an issue (though
it may be traded for background dependence). Finally, one may be concerned at a
more conceptual level that, excluding topological and black hole fluctuations in the
UV, one is missing fundamental ingredients for a UV completion of gravity.14 To be
very concrete, one may formulate the following complaint: In the asymptotic safety
scenario, gravity in the UV is treated quite similarly to a standard CFT. This suggests
that arbitrarily small-size and hence high-energy, localised fluctuations are part of
the Hilbert space. But taking the (experimentally established) perspective of the IR
observer, those small fluctuations should collapse to black holes [509–511]. In other
words, the black hole collapse should prevent us from considering the QFT-like UV
limit on which everything was built. For further critical discussions see [512].
However, such possible criticism is not our concern here. We want for the
moment to adopt the point of view that gravity is UV completed through a non-
trivial UV fixed point and, moreover, that this also holds for gravity together with
certain matter fields, like for example the Standard Model (see e.g. [513–518]). This
allows us to ask, as we did already in the previous section, what the consequences
for the hierarchy problems might be.
The answer depends on which matter content is allowed (in the sense that the
fixed point is not lost) and what the predictions for the operator coefficients at k = 0
are. The present understanding does not suggest that the matter content is extremely
constrained. On the contrary, it appears that various theories ‘of the right type’ (with
gauge groups, fermions, scalars) may emerge in the IR. Since the fixed surface
is expected to be finite-dimensional, a certain amount of predictivity should arise.
However, it does not appear to be the case that the Standard Model parameters come
out uniquely (i.e. that there is a unique model which runs into the fixed point).15
In summary, it may be a reasonable expectation that the Standard Model with
gravity (supplemented by further fields, e.g. at the neutrino seesaw scale) runs into
a quantum gravity UV fixed point. Some of its parameters may be constrained
14 The possibility has been entertained that, at some higher energy scale, asymptotic safety comes
together with string theory [508]. In this case, such more extreme spacetime fluctuations, which
are essential in the stringy UV completion, would enter the stage after all.
15 A noteworthy exception is the argument presented in [519] that the Higgs quartic coupling
must vanish at the transition point between weakly coupled IR and UV regime . This has lead
to a successful prediction of the Higgs mass value after perturbative running to the weak scale.
However, from today’s perspective the prediction does not work perfectly any more since the
central top mass value has shifted [520].
by this requirement. However, it remains completely unclear why the two crucial,
dimensionful parameters λ(0) and m2H (0) are so incredibly tiny in Planck units.16
The problem is that, at the high scale MP , when the assumed near-fixed-point
evolution transits to the well-understood perturbative evolution of a low-energy EFT
with gravity, one would need λ(k)/k 4 and m2H (k)/k 2 to be extremely small. More
than that, they need to have just the right size to compensate for loop and non-
perturbative effects from scales between k = MP and k = 0. Achieving this remains
a challenge (see however [521]). Without such a mechanism, and assuming that tiny
values of these parameters are consistent with the UV fixed point, one arrives at the
following situation:
The hierarchy problems can be solved, but there is a price to be paid: We would
live in one of continuously many UV-consistent theories. Ours just happens to have
these peculiar parameters. One may call this a fine-tuning or refuse to use this term:
after all, there is only one theory realised in nature. Compared to the string (or any
other) landscape, one has lost the option of explaining the very special observed
parameter values using cosmology plus anthropics. But this may be just fine since,
given the absence of a natural measure on the critical surface, it is hard to claim that
the observed parameters are special in the first place.
16 In this close connection with m2H it is tempting to misread λ as the Higgs quartic coupling.
We hence remind the reader that, to distinguish it from the cutoff , we use the symbol λ for the
cosmological constant.
Summary
9
Let us recapitulate what one may have learned by going through these notes.
First, we have learned to view the Standard Model as an effective field theory
with a number of issues: too many (and too random) parameters, especially in
the Yukawa sector; no dark matter candidate; no mechanism for producing the
baryon asymmetry of the universe; no obvious inflaton candidate. Moreover, and
this was our central theme, one faces a hierarchy problem concerning the smallness
of the Higgs mass parameter m2H relative to the cutoff scale 2 . Finally, including
gravity as a low-energy effective field theory (EFT) in our framework, a very similar
second hierarchy problem between the cosmological constant λ and the relevant
cutoff scale 4 was seen to arise. We were careful to spell out these problems more
quantitatively: They are related to an enormous sensitivity of the low-energy theory
to the exact value of any new-physics parameters and mass scales that can (and
probably have to) be present between the electroweak and Planck scale.
While no widely accepted solution to the second hierarchy problem (the cosmo-
logical constant problem) exists, the electroweak hierarchy problem can at least be
drastically reduced. The essence of this resolution is to introduce an intermediate
cutoff scale, let us call it BSM , at which the Standard Model is replaced by a
theory in which a light scalar (like the Higgs) may naturally coexist with a much
higher fundamental cutoff scale . Our focus was on low-energy supersymmetry as
a concrete realisation of this idea, i.e. BSM ≡ SU SY mew , where mew stands
for the electroweak scale. The problem is that this solution works less and less well
as experiments (most recently the LHC) push SU SY significantly above mew .
We have also discussed how, in the presence of gravity, supersymmetry must be
replaced by supergravity. Here we had our first encounter with complex geometry
since the scalar fields now live on a Kahler manifold. The scalar potential is a
function on that manifold and derives (at least partially) from the superpotential.
The latter is a section in a line bundle over the aforementioned Kahler manifold. This
description of the world through quantities like the Kahler potential K and superpo-
tential W is very elegant. But it does unfortunately not help with the cosmological

284 9 Summary
constant problem: Loop corrections to K are roughly speaking as bad as those to

λ. Concerning the electroweak hierarchy, things do not improve compared to rigid
SUSY. Maybe most importantly, the infinite-dimensional continuous freedom of
choosing EFTs is not reduced—it is only repackaged in the freedom of choosing
K and W .
Given this unsatisfactory state of affairs, it is natural to keep pushing for a
more fundamental understanding at the highest scale accessible in this line of
thought: the quantum gravity scale. Here, an at first sight rather far-fetched idea
offers surprising insights: It is the suggestion to identify elementary particles with
loops of fundamental string. This approach naturally cuts off the perturbative
infinities of quantum gravity, but only at the price of 26 spacetime dimensions
and moreover, an unstable (tachyonic) vacuum. The latter problem is solved in
the supersymmetric version of the theory, which has a stable supersymmetric 10d
Minkowski vacuum. The problem of too many spatial dimensions can be overcome
by compactifying six of them (in the simplest case on Calabi–Yau manifolds).
This clearly leads to ambiguities concerning the resulting 4d EFT. Moreover, even
in 10d five different highly supersymmetric EFTs result from slightly different
ways of defining the superstring. Yet, miraculously, the concept of dualities ties all
constructions (together with a highly supersymmetric 11d model with fundamental
2-branes) into a single theory. We recall here that by ‘duality’ we mean a situation
where two different mathematical formulations define exactly the same physical
theory.
At this point, one has arrived at a possibly fundamental and essentially unique
theory (or model) of quantum gravity which has no free parameters. The only
exception is the string scale, which however simply sets the fundamental energy
scale of the world. But there is a large number of 4d solutions, related to the concrete
choice of 10d theory (one in five) and of Calabi–Yau space (one in about half a
billion). Very elegantly, the 3d complex geometry of the Calabi–Yau is governed
by the (O(100)-dimensional) complex geometry of the Calabi–Yau moduli spaces.
These are precisely the Kahler manifolds of the 4d supergravity models describing
the low-energy EFTs of these compactifications. The massless scalar fields or
moduli encode 10d metric degrees of freedom and determine the precise size and
shape of the Calabi–Yau geometry.
But this is only the beginning of the actual String Landscape: The 10d theory
possesses non-perturbative objects (such as D-branes and orientifold planes) and
allows for non-zero expectation values of p-form gauge fields (fluxes). Equipping
the Calabi–Yau with those leads to an explosion of the number of 4d solutions, a
recent estimate being as high as 10272,000. This so-called flux landscape plausibly
contains many 4d models (flux vacua) with a realistic, Standard Model-like matter
sector. It is more complicated to show explicitly but plausible that vacua with broken
supersymmetry and a positive cosmological constant are present in this large set of
solutions. The best concrete examples are known under the acronyms KKLT and
LVS.
Once this is accepted as a matter of principle (or better: rigorously established
in the future), one expects that extremely many such vacua, with the right structure
9 Summary 285
to describe the real world, exist. Thus, all EFT parameters and in particular the
cosmological constant and the Higgs mass squared are very finely scanned in the
landscape. One still has a unique fundamental theory. But, through its sheer abun-
dance of solutions, it may accommodate what the low-energy observer perceives as
extremely fine-tuned parameters. In the case of the electroweak hierarchy problem,
this tuning may bridge a small or even very large gap between SUSY breaking scale
and weak scale, i.e. both SU SY mew and SU SY mew are compatible with
this picture.
But to accommodate is not the same as to explain or even predict. Indeed, one
may feel surprised and unsatisfied by finding oneself in a ‘very special’ landscape
vacuum, with (at least) two parameters chosen in a highly non-generic way. To
quantify such a surprise or unease, one needs a measure on the landscape. Since the
landscape is discrete (under mild conditions probably even finite), a natural choice
might be that of giving each vacuum an equal weight. Then, by all that we know only
a very small fraction of vacua is closer to the special point ( λ = 0, m2H = 0 ) than
our observed Standard Model-EFT. One could say that we have hence not avoided
the fine-tuning but shown that it can be explicitly realised.
However, one may also say that finding ourselves in one of these very special
vacua is not surprising after all: Indeed, it is easy to argue that any kind of
observer (limited by what we can presently imagine) can only form if some structure
in the universe forms first. One presumably also requires some scale separation
between the energy scales of chemistry and the Planck scale. Involving this so-
called anthropic argument, one may say that we find ourselves near this particular
point in the landscape simply because other vacua have no observers.
Clearly, this is very rough and not at all quantitative. Also, the proposed
measure of counting vacua lacks justification. It would be much better to have
a theory of how the various vacua get populated cosmologically and how to ask
an observer-dependent question as we just did in a more objective manner. One
option that has been explored in some detail is that of eternal inflation during
which, starting with one of the metastable de Sitter vacua, all of them get realised
cosmologically through consecutive tunnelling processes. This eternal evolution of
‘bubbles within bubbles within bubbles’, while in principle capable of populating
the whole landscape, comes with its own issues. Maybe the most problematic is the
so-called measure problem: Namely, due to the divergence of the number of bubbles
(and hence of potential observers) at late times, one needs a cutoff. With such a
late-time cutoff imposed, one may count observers and hence derive a measure.
However, due to diffeomorphism invariance of general relativity, there appears to
be no unambiguously preferred choice of cutoff and hence no established first-
principles measure.
To illustrate what is at stake, let us imagine the landscape were understood
well-enough to know precisely which vacua with which features it contains and
what the transition rates between them are. Moreover, let us also assume an a-
priori, quantum-gravity-derived measure could then be established on this basis.
As a result, one may ask questions of the following type: Given all that we know
about our vacuum, how many observers share all these observations and live in
286 9 Summary
a vacuum with a low SUSY breaking scale? By contrast, how many observers
share all our observations and live in a vacuum with high-scale SUSY breaking?
The resulting numbers predict probabilistically what we expect to find in our EFT
at the next energy frontier. While we have only a single experiment to perform,
such a probabilistic prediction may still be meaningful since the ratio of these two
probabilities can be exponentially large. Under such circumstances we may get to
rule out a theory at many standard deviations based on a single observation.
The last two paragraphs attempted to take the string landscape idea (and
specifically its implications for the electroweak hierarchy problem) to an idealised
endpoint of purely statistical predictions about future measurements. Thinking that
far may be interesting and important, but it is also relatively speculative and far-
fetched. Many more modest and more approachable questions can be asked. First,
key aspects of the string landscape are not understood. We are very far from having
an overview of the landscape as a whole. Next, we may use the landscape not
to make statistical predictions but to draw inspiration for what could be observed
in cosmology and particle phenomenology in the future. This may be a more
immediate way in which string theory can relate to experiment. Finally, it is
interesting to investigate whether there exist consistent EFTs which we could find
realised in our universe and which do not represent any of the string landscape
vacua.
This last point deserves a more careful discussion. It is remarkable that, due
to an observation of this type, a single experiment could in principle rule out
string theory with certainty, without the need to appeal to statistics. Moreover, the
approach of studying which EFTs are not present in the landscape is very popular
at the moment of this writing. It is known as the Swampland program and its
claims about the inconsistency of certain EFTs are characterised as Swampland
conjectures. Some of them even try to exclude metastable de Sitter vacua in
complete generality. While this does not immediately rule out string theory (since
the observed exponential expansion of our universe could be due to dynamical dark
energy), it certainly clashed with most of what we thought we have learned about
string theory phenomenology in recent years. Thus, it is of immense importance to
either establish those claims or, on the contrary, to further develop our quantitative
understanding of the proposed constructions of metastable de Sitter vacua. In this
context, much more precision and explicitness is certainly desirable.
The landscape approach to thinking about and maybe resolving hierarchy
problems, especially once anthropic arguments are invoked, has received a fair
amount of criticism. This is understandable given how enormous a paradigm change
is involved. It also has to be admitted that restricting the predictivity of fundamental
scientific thought to (at least at low energies) only probabilistic statements may
be perceived as frightening. However, as we have tried to argue in the previous
subsection, the alternatives have significant shortcomings as well. Indeed, let us
assume a unique theory of quantum gravity is established which, as opposed to
string theory, does not possess a landscape of solutions. Then one clearly expects
some simple formula for e.g. the operator coefficient that we call the cosmological
constant to be provided by this theory at some energy scale μ just below MP .
9 Summary 287
But now, at least according to all that we know, it is very hard to see how such
a simple fundamental formula would combine with the known loop effects to give
the observed value λ(μ = 0) ∼ 10−120MP4 . Nevertheless, this possibility cannot be
ruled out.
Alternatively, it is conceivable that the correct theory of quantum gravity is not
unique but comes with continuous parameters. This would allow one to literally tune
those fundamental parameters to realise the hierarchies observed in our low-energy
EFTs. One may call this unsatisfactory, but it is hard to make an objective point
against this option.
At the moment, we have to let these different attitudes to the hierarchy problems
coexist and compete. In addition to studying string theory, it is certainly interesting
to look for fundamental theories capable of predicting Higgs mass parameter and
cosmological constant or to at least explicitly realise the tuning. We should also not
forget that dynamical resolutions of the hierarchy problems have not been proved to
be impossible. Just because no convincing version for the cosmological constant has
been found and because the simplest SUSY models are under pressure, one cannot
be certain that the resolution will not, after all, come from a very particular EFT
and its dynamics. Finally, recent ideas of cosmological selection, which combine
elements of EFT dynamics and a landscape, may turn out to be correct.
While all the options above are very interesting, they were of course not our main
subject. Our goal was to develop, in some technical detail, specifically the string
landscape view on the hierarchy problems. In two sentences, the result is as follows:
Through its immense number of solutions, string theory as a very concrete model
of quantum gravity may be capable of accommodating low-energy EFTs with an
extremely fine-tuned appearance. Depending on how much ‘anthropics’ and eternal-
inflation cosmology one is willing to involve, this may even be promoted to an
explanation or prediction.
The readers will make up their own mind and decide which of the above
directions to pursue or which new ideas to propose. Hopefully, these notes can
be useful for making such choices on the basis of a somewhat more technical
understanding of the string landscape.
References
1. M.E. Peskin, D.V. Schroeder, An Introduction to Quantum Field Theory (Westview Press,
Boulder, 1995)
2. R.M. Wald, General Relativity (University of Chicago, Chicago, 1984)
3. T. Cheng, L. Li, Gauge Theory of Elementary Particle Physics and Gauge Theory of
Elementary Particle Physics—Problems and Solutions (Oxford University, Oxford, 1984 and
2000)
4. J. Donoghue, E. Golowich, B.R. Holstein, Dynamics of the Standard Model. Camb. Monogr.
Part. Phys. Nucl. Phys. Cosmol. 2 (1992), 1–540
5. J. Wess, J. Bagger, Supersymmetry and Supergravity (Princeton University, Princeton, 1991)
6. D.Z. Freedman, A. Van Proeyen, Supergravity (Cambridge University, Cambridge, 2012)
7. J. Polchinski, String Theory, vol. 1 and 2 (Cambridge University, Cambridge, 2001)
8. R. Blumenhagen, D. Lüst, S. Theisen, Basic Concepts of String Theory (Springer, Berlin,
2013)
9. L.E. Ibanez, A.M. Uranga, String Theory and Particle Physics: An Introduction to String
Phenomenology (Cambridge University, Cambridge 2012)
10. F. Denef, Les Houches lectures on constructing string vacua. Les Houches 87, 483 (2008).
[arXiv:0803.1194]
11. A.N. Schellekens, Life at the interface of particle physics and string theory. Rev. Mod. Phys.
85(4), 1491 (2013). [arXiv:1306.5083 [hep-ph]]
12. S. Weinberg, The Quantum Theory of Fields, vol. 1 and 2 (Cambridge University, Cambridge,
2005)
13. C. Itzykson, J. Zuber, Quantum Field Theory (McGraw-Hill, New York, 1980)
14. M. Srednicki, Quantum Field Theory (Cambridge University, Cambridge, Cambridge, 2007)
15. M.D. Schwartz, Quantum Field Theory and the Standard Model (Cambridge University,
Cambridge, 2014)
16. O. Nachtmann, Elementary Particle Physics: Concepts and Phenomena (Springer, Berlin,
1990)
17. Schellekens: Beyond the Standard Model. https://www.nikhef.nl/~t58/lectures.html
18. M. Blanke, Introduction to flavour physics and CP violation. CERN Yellow Rep. School Proc.
1705, 71 (2017). [arXiv:1704.03753 [hep-ph]]
19. B. Grinstein, Lectures on flavor physics and CP violation (2017). [arXiv:1701.06916 [hep-
ph]]
20. Kooijman/Tuning: Lectures on CP Violation. https://www.nikhef.nl/~h71/Lectures/2015/
ppII-cpviolation-19032018.pdf
21. I.I. Bigi, A. Sanda, CP violation. Camb. Monogr. Part. Phys. Nucl. Phys. Cosmol. 9, 1–485
(2009)
22. R. Fleischer, Flavour physics and CP violation, in Lectures at the 2005 European School of
High-Energy Physics (2006). [arXiv:hep-ph/0608010 [hep-ph]]
23. G.C. Branco, L. Lavoura, J.P. Silva, CP violation. Int. Ser. Monogr. Phys. 103, 1–536 (1999)

in Physics 979, https://doi.org/10.1007/978-3-030-65151-0
290 References
24. A.A. Anselm, A.A. Johansen, Can electroweak theta term be observable? Nucl. Phys. B 412,
553 (1994). [hep-ph/9305271]
25. C. Cao, A. Zhitnitsky, Axion detection via topological Casimir effect. Phys. Rev. D 96(1),
015013 (2017). [arXiv:1702.00012 [hep-ph]]
26. H. Georgi, Effective field theory. Ann. Rev. Nucl. Part. Sci. 43, 209–252 (1993)
27. A.V. Manohar, Effective field theories. Lect. Notes Phys. 479, 311–362 (1997). [arXiv:hep-
ph/9606222 [hep-ph]]
28. A. Pich, Effective field theory: course, in Lectures at Les Houches Summer School (1998).
[arXiv:hep-ph/9806303 [hep-ph]]
29. M. Luty, TASI lectures on supersymmetry breaking (2004). hep-th/0509029
30. D.B. Kaplan, Five lectures on effective field theory. nucl-th/0510023
31. T. Cohen, As scales become separated: lectures on effective field theory. PoS TASI2018, 011
(2019). [arXiv:1903.03622 [hep-ph]]
32. J.D. Wells, Effective Theories in Physics—From Planetary Orbits to Elementary Particle
Masses (Springer, Berlin, 2012)
33. P. Minkowski, μ → eγ at a rate of one out of 109 muon decays? Phys. Lett. B 67, 421–428
(1977)
34. T. Yanagida, Horizontal gauge symmetry and masses of neutrinos. Conf. Proc. C 7902131,
95–99 (1979). KEK-79-18-95
35. M. Gell-Mann, P. Ramond, R. Slansky, Complex spinors and unified theories. Conf. Proc. C
790927, 315–321 (1979). [arXiv:1306.4669 [hep-th]]
36. E. Gildener, Gauge symmetry hierarchies. Phys. Rev. D 14, 1667 (1976)
37. M. Veltman, The infrared—ultraviolet connection. Acta Phys. Polon. B 12, 437 (1981)
38. G. ’t Hooft, C. Itzykson, A. Jaffe, H. Lehmann, P.K. Mitter, I.M. Singer, R. Stora, Recent
developments in gauge theories, in Proceedings, Nato Advanced Study Institute, Cargese,
France, August 26–September 8, 1979. NATO Science, Series B, vol. 59 (1980) pp. 1
39. R. Barbieri, G.F. Giudice, Upper bounds on supersymmetric particle masses. Nucl. Phys. B
306, 63 (1988)
40. J.R. Ellis, K. Enqvist, D.V. Nanopoulos, F. Zwirner, Observables in low-energy superstring
models. Mod. Phys. Lett. A 1, 57 (1986)
41. J.D. Wells, Naturalness, Extra-Empirical Theory Assessments, and the Implications of
Skepticism (2018). arXiv:1806.07289 and Finetuned Cancellations and Improbable Theories.
arXiv:1809.03374
42. F. Azhar, A. Loeb, Gauging fine-tuning. Phys. Rev. D 98(10), 103018 (2018).
[arXiv:1809.06220 [astro-ph.CO]]
43. M. Tanabashi et al. [Particle Data Group], Review of particle physics. Phys. Rev. D 98(3),
030001 (2018)
44. R.E. Kass, A.E. Raftery, Bayes factors. J. Am. Statist. Assoc. 90(430), 773–795 (1995)
45. R. Trotta, Bayes in the sky: Bayesian inference and model selection in cosmology. Contemp.
Phys. 49, 71–104 (2008). [arXiv:0803.4089 [astro-ph]]
46. B.C. Allanach, K. Cranmer, C.G. Lester, A.M. Weber, Natural priors, CMSSM fits and LHC
weather forecasts. JHEP 08, 023 (2007). [arXiv:0705.0487 [hep-ph]]
47. M.E. Cabrera, J.A. Casas, R. Ruiz de Austri, Bayesian approach and Naturalness in MSSM
analyses for the LHC. JHEP 03, 075 (2009). [arXiv:0812.0536 [hep-ph]]
48. S. Fichet, Quantified naturalness from Bayesian statistics. Phys. Rev. D 86, 125029 (2012).
[arXiv:1204.4940 [hep-ph]]
49. A. Fowlie, CMSSM, naturalness and the fine-tuning price of the Very Large Hadron Collider.
Phys. Rev. D 90, 015010 (2014). [arXiv:1403.3407 [hep-ph]]
50. J.D. Wells, Lectures on Higgs Boson Physics in the Standard Model and Beyond (2009).
[arXiv:0909.4541 [hep-ph]]
51. G.F. Giudice, Naturalness after LHC8. PoS EPS-HEP2013, 163 (2013). [arXiv:1307.7879]
and The Dawn of the Post-Naturalness Era. [arXiv:1710.07663]
52. S. Weinberg, Gravitation and Cosmology (Wiley, New York, 1972)
53. C.S. Misner, K.W. Thorne, J.A. Wheeler, Gravitation (Princeton University, Princeton, 2017)
References 291
54. S.M. Carroll, Spacetime and Geometry (Cambridge University, Cambridge, 2019)
55. N. Straumann, General Relativity (Springer, Berlin, 2013)
56. G. Gibbons, S. Hawking, M. Perry, Path Integrals and the Indefiniteness of the Gravitational
Action. Nucl. Phys. B 138, 141–150 (1978)
57. M. Veltman, Quantum theory of gravitation, in Methods in Field Theory, Les Houches 1975,
ed. by R. Balian, J. Zinn-Justin (North-Holland Publication, Amsterdam, The Netherlands,
1976). Course 5
58. S. Weinberg, The cosmological constant problem. Rev. Mod. Phys. 61, 1–23 (1989)
59. S. Weinberg, The Cosmological constant problems, in Talk at 4th International Symposium
Dark Matter (2000). [arXiv:astro-ph/0005265 [astro-ph]]
60. T. Padmanabhan, Cosmological constant: The Weight of the vacuum. Phys. Rept. 380, 235–
320 (2003). [arXiv:hep-th/0212290 [hep-th]]
61. T. Padmanabhan, Dark energy: mystery of the millennium. AIP Conf. Proc. 861(1), 179–196
(2006). [arXiv:astro-ph/0603114 [astro-ph]]
62. H. Georgi, Lie algebras in Particle Physics. Front. Phys. 54, 1–320 (1999)
63. R. Slansky, Group theory for unified model building. Phys. Rept. 79, 1–128 (1981)
64. G.G. Ross, Grand Unified Theories (Westview Press, New York, 2003)
65. P. Nath, P. Fileviez Perez, Proton stability in grand unified theories, in strings and in branes.
Phys. Rept. 441, 191–317 (2007). [arXiv:hep-ph/0601023 [hep-ph]]
66. S. Raby, Supersymmetric Grand Unified Theories. Lect. Notes Phys. 939, 1–308 (2017)
67. A. Hebecker, J. Hisano, Grand unified theories. Review article in Ref. [43]
68. D. Croon, T.E. Gonzalo, L. Graf, N. Kosnik, G. White, GUT Physics in the era of the LHC.
Front. Phys. 7, 76 (2019). [arXiv:1903.04977 [hep-ph]]
69. P. West, Introduction to Supersymmetry and Supergravity (World Scientific, New York, 1990)
70. S. Weinberg, Quantum Field Theory, vol. 3, (Cambridge University, Cambridge, 2005)
71. J. Terning, Modern Supersymmetry (Oxford University, Oxford, 2006)
72. M. Shifman, Advanced Topics in Quantum Field Theory (Cambridge University, Cambridge,
2012)
73. S.R. Coleman, J. Mandula, All Possible Symmetries of the S Matrix. Phys. Rev. 159, 1251
(1967)
74. R. Haag, J.T. Lopuszanski, M. Sohnius, All Possible Generators of Supersymmetries of the s
Matrix. Nucl. Phys. B 88, 257 (1975)
75. J. Wess, B. Zumino, Supergauge transformations in four-dimensions. Nucl. Phys. B 70, 39–50
(1974)
76. D. Volkov, V. Akulov, Is the Neutrino a Goldstone Particle? Phys. Lett. B 46, 109–110 (1973)
77. L. O’Raifeartaigh, Spontaneous symmetry breaking for chiral scalar superfields. Nucl. Phys.
B 96, 331–352 (1975)
78. P. Fayet, J. Iliopoulos, Spontaneously broken supergauge symmetries and goldstone spinors.
Phys. Lett. B 51, 461–464 (1974)
79. S.P. Martin, A Supersymmetry primer. Adv. Ser. Direct. High Energy Phys. 21, 1 (2010).
[hep-ph/9709356]
80. G.F. Giudice, R. Rattazzi, Theories with gauge mediated supersymmetry breaking. Phys.
Rept. 322, 419 (1999). [hep-ph/9801271]
81. M. Bauer, T. Plehn, Yet another introduction to dark matter. Lect. Notes Phys. 959 (1999).
[arXiv:1705.01987 [hep-ph]]
82. Gianfranco Bertone (ed.), Particle Dark Matter (Cambridge University, Cambridge, 2010)
83. D. Hooper, Particle dark matter, in TASI Lectures, pp. 709–764 (2010). [arXiv:0901.4090
[hep-ph]]
84. K.A. Olive, TASI lectures on dark matter (2003). [arXiv:astro-ph/0301505 [astro-ph]]
85. N. Seiberg, Naturalness versus supersymmetric nonrenormalization theorems. Phys. Lett. B
318, 469 (1993). [hep-ph/9309335]
86. G. Giudice, A. Masiero, A natural solution to the mu problem in supergravity theories. Phys.
Lett. B 206, 480–484 (1988)
292 References
87. S. Dimopoulos, S. Raby, F. Wilczek, Supersymmetry and the scale of unification. Phys. Rev.
D 24, 1681–1683 (1981)
88. S. Dimopoulos, H. Georgi, Softly broken supersymmetry and SU(5). Nucl. Phys. B 193, 150–
162 (1981)
89. L.E. Ibanez, G.G. Ross, Low-energy predictions in supersymmetric grand unified theories.
Phys. Lett. B 105, 439–442 (1981)
90. N. Sakai, Naturalness in supersymmetric guts. Z. Phys. C 11, 153 (1981)
91. U. Amaldi, W. de Boer, H. Furstenau, Comparison of grand unified theories with electroweak
and strong coupling constants measured at LEP. Phys. Lett. B 260, 447–455 (1991)
92. N. Arkani-Hamed, S. Dimopoulos, Supersymmetric unification without low energy super-
symmetry and signatures for fine-tuning at the LHC. JHEP 0506, 073 (2005). [hep-
th/0405159]
93. G.F. Giudice, A. Romanino, Split supersymmetry. Nucl. Phys. B 699, 65 (2004). Erratum:
[Nucl. Phys. B 706, 487 (2005)] [hep-ph/0406088]
94. F. Denef, M.R. Douglas, Distributions of nonsupersymmetric flux vacua. JHEP 0503, 061
(2005) [hep-th/0411183]
95. A. Hebecker, A.K. Knochel, T. Weigand, A shift symmetry in the Higgs sector: Experimental
hints and stringy realizations. JHEP 1206, 093 (2012). [arXiv:1204.2551 [hep-th]]
96. Buchbinder/Kuzenko, Ideas and Methods of Supersymmetry and Supergravity (Institute of
Physics Publishing, New York, 1995)
97. F. Quevedo, S. Krippendorf, O. Schlotterer, Cambridge Lectures on Supersymmetry and Extra
Dimensions (2010). arXiv:1011.1491 [hep-th]
98. G. Villadoro, F. Zwirner, De-Sitter vacua via consistent D-terms. Phys. Rev. Lett. 95, 231602
(2005). [arXiv:hep-th/0508167 [hep-th]]
99. Z. Komargodski, N. Seiberg, Comments on the Fayet-Iliopoulos term in field theory and
supergravity. JHEP 06, 007 (2009). [arXiv:0904.1159 [hep-th]]
100. K.R. Dienes, B. Thomas, On the inconsistency of Fayet-Iliopoulos terms in supergravity
theories. Phys. Rev. D 81, 065023 (2010). [arXiv:0911.0677 [hep-th]]
101. M.B. Green, J. Schwarz, E. Witten, Superstring Theory, vol. I and II (Cambridge University,
Cambridge, 1987)
102. K. Becker, M. Becker, J.H. Schwarz, String Theory and M-theory. (Cambridge University,
Cambridge, 2007)
103. E. Kiritsis, String Theory in a Nutshell (Princeton University, Princeton, 2007)
104. B. Zwiebach, A First Course in String Theory (Cambridge University, Cambridge, 2009)
105. S. Deser, B. Zumino, A complete action for the spinning string. Phys. Lett. B 65, 369–373
(1976)
106. L. Brink, P. Di Vecchia, P.S. Howe, A locally supersymmetric and reparametrization invariant
action for the spinning string. Phys. Lett. B 65, 471–474 (1976)
107. A.M. Polyakov, Quantum geometry of bosonic strings. Phys. Lett. B 103, 207–210 (1981)
108. C.V. Johnson, D-brane Primer. arXiv:hep-th/0007170 [hep-th]; D-branes (Cambridge Uni-
versity Press, Cambridge, 2003)
109. J. Dai, R.G. Leigh, J. Polchinski, New connections between string theories. Mod. Phys. Lett.
A 4, 2073–2083 (1989)
110. R.G. Leigh, Dirac-Born-Infeld action from Dirichlet sigma model. Mod. Phys. Lett. A 4, 2767
(1989)
111. R. Blumenhagen, M. Cvetic, P. Langacker, G. Shiu, Toward realistic intersecting D-brane
models. Ann. Rev. Nucl. Part. Sci. 55, 71–139 (2005). [arXiv:hep-th/0502005 [hep-th]]
112. R. Blumenhagen, B. Kors, D. Lust, S. Stieberger, Four-dimensional string compactifications
with D-Branes, Orientifolds and Fluxes. Phys. Rept. 445, 1–193 (2007). [arXiv:hep-
th/0610327 [hep-th]]
113. L.E. Ibanez, F. Marchesano, R. Rabadan, Getting just the standard model at intersecting
branes. JHEP 11, 002 (2001). [arXiv:hep-th/0105155 [hep-th]]
114. R. Blumenhagen, B. Kors, D. Lust, T. Ott, The standard model from stable intersecting brane
world orbifolds. Nucl. Phys. B 616, 3–33 (2001). [arXiv:hep-th/0107138 [hep-th]]
References 293
115. M. Cvetic, G. Shiu, A.M. Uranga, Chiral four-dimensional N=1 supersymmetric type 2A
orientifolds from intersecting D6 branes. Nucl. Phys. B 615, 3–32 (2001).
[arXiv:hep-th/0107166 [hep-th]]
116. A.A. Belavin, A.M. Polyakov, A.B. Zamolodchikov, Infinite conformal symmetry in two-
dimensional quantum field theory. Nucl. Phys. B 241, 333 (1984)
117. P. Di Francesco, P. Mathieu, D. Senechal, Conformal Field Theory (Springer, Berlin, 1997)
118. M. Schottenloher, A mathematical introduction to conformal field theory. Lect. Notes Phys.
759, 1–237 (2008)
119. R. Blumenhagen, E. Plauschinn, Introduction to conformal field theory. Lect. Notes Phys.
779, 1–256 (2009)
120. P.H. Ginsparg, Applied conformal field theory, in Lectures at Les Houches Summer School
(1988). hep-th/9108028
121. A.N. Schellekens, Conformal Field Theory, Lecture Notes. https://www.nikhef.nl/~t58/Site/
Lectures.html
122. I. Antoniadis, C. Bachas, J.R. Ellis, D.V. Nanopoulos, An expanding universe in string theory.
Nucl. Phys. B 328, 117–139 (1989)
123. A.A. Tseytlin, C. Vafa, Elements of string cosmology. Nucl. Phys. B 372, 443–466 (1992).
124. E. Silverstein, (A)dS backgrounds from asymmetric orientifolds. Clay Mat. Proc. 1, 179
125. S. Hellerman, I. Swanson, Cosmological solutions of supercritical string theory. Phys. Rev. D
77, 126011 (2008). [arXiv:hep-th/0611317 [hep-th]]
126. P. Ramond, Dual theory for free fermions. Phys. Rev. D 3, 2415–2418 (1971)
127. A. Neveu, J.H. Schwarz, Factorizable dual model of pions. Nucl. Phys. B 31, 86–112 (1971)
128. M.B. Green, J.H. Schwarz, Covariant description of superstrings. Phys. Lett. B 136, 367–370
(1984); Properties of the covariant formulation of superstring theories. Nucl. Phys. B 243,
285–306 (1984)
129. J.H. Schwarz, Superstring theory. Phys. Rept. 89, 223–322 (1982)
130. N. Berkovits, ICTP lectures on covariant quantization of the superstring. ICTP Lect. Notes
Ser. 13, 57–107 (2003). [arXiv:hep-th/0209059 [hep-th]]
131. T. Buscher, Path integral derivation of quantum duality in nonlinear sigma models. Phys. Lett.
B 201, 466–472 (1988);
A symmetry of the string background field equations. Phys. Lett. B 194, 59–62 (1987)
132. M.B. Green, J.H. Schwarz, Anomaly cancellation in supersymmetric D=10 gauge theory and
superstring theory. Phys. Lett. B 149, 117–122 (1984)
133. S.B. Giddings, S. Kachru, J. Polchinski, Hierarchies from fluxes in string compactifications.
Phys. Rev. D 66, 106006 (2002). [hep-th/0105097]
134. T. Ortin, Gravity and Strings (Cambridge University, Cambridge, 2015)
135. G. Nordstrom, On the possibility of unifying the electromagnetic and the gravitational fields.
Phys. Z. 15, 504 (1914). [physics/0702221 [physics.gen-ph]]
136. O. Klein, Quantum theory and five-dimensional theory of relativity. Z. Phys. 37, 895 (1926).
[Surveys High Energ. Phys. 5, 241 (1986)]
137. O. Klein, The atomicity of electricity as a quantum theory law. Nature 118, 516 (1926)
138. T. Appelquist, A. Chodos, P.G.O. Freund, Modern Kaluza-klein theories, in Frontiers in
Physics, p. 65 (Addison-Wesley, New York, 1987)
139. M.J. Duff, Kaluza-Klein Theory in Perspective. Talk at The Oskar Klein Centenary, Stockholm
(1994). hep-th/9410046
140. J.M. Overduin, P.S. Wesson, Kaluza-Klein gravity. Phys. Rept. 283, 303 (1997). [gr-
qc/9805018]
141. P. Candelas, Lectures on complex manifolds in Trieste 1987, in Proceedings, Superstrings,
pp. 1–88 (1987)
142. B.R. Greene, String theory on Calabi-Yau manifolds, in Lectures at TASI (1996). hep-
th/9702155
294 References
143. T. Hübsch, Calabi-Yau Manifolds: A Bestiary for Physicists (World Scientific, New York,
1991)
144. Y.H. He, The Calabi-Yau Landscape: From Geometry, to Physics, to Machine-Learning
(2018). [arXiv:1812.02893 [hep-th]]
145. L.B. Anderson, M. Karkheiran, TASI lectures on geometric tools for string compactifications.
PoS TASI2017, 013 (2018). [arXiv:1804.08792 [hep-th]]
146. J. Viaclovsky, Lectures on Kahler geometry, Ricci curvature, and hyperkahler metrics, in
Lecture Notes (2019). https://www.math.uci.edu/~jviaclov/lecturenotes/lecturenotes.html
147. M. Nakahara, Geometry, Topology and Physics (Institute of Physics Publishing, New York,
2003)
148. R.A. Bertlmann, Anomalies in Quantum Field Theory (Oxford University, Oxford, 1996)
149. C. Nash, S. Sen, Topology and Geometry for Physicists (Academic Press, New York, 1983)
150. M. Göckeler, T. Schücker, Differential Geometry, Gauge Theories, and Gravity (Cambridge
University, Cambridge, 1987)
151. P. Candelas, X. de la Ossa, Moduli space of Calabi-Yau manifolds. Nucl. Phys. B 355, 455
(1991); P. Candelas, Yukawa couplings between (2,1) Forms. Nucl. Phys. B 298, 458 (1988)
152. A. Font, S. Theisen, Introduction to string compactification. Lect. Notes Phys. 668, 101 (2005)
153. V.V. Batyrev, Dual polyhedra and mirror symmetry for Calabi-Yau hypersurfaces in toric
varieties. J. Alg. Geom. 3, 493–545 (1994). [arXiv:alg-geom/9310003 [math.AG]]
154. V. Bouchard, Lectures on complex geometry, Calabi-Yau manifolds and Toric geometry
155. T.W. Grimm, J. Louis, The effective action of N = 1 Calabi-Yau orientifolds. Nucl. Phys. B
699, 387 (2004). [hep-th/0403067]
156. H. Jockers, J. Louis, The effective action of D7-branes in N = 1 Calabi-Yau orientifolds. Nucl.
Phys. B 705, 167 (2005). [hep-th/0409098]
157. T.W. Grimm, J. Louis, The Effective action of type IIA Calabi-Yau orientifolds. Nucl. Phys.
B 718, 153 (2005). [hep-th/0412277]
158. M. Kerstan, T. Weigand, The effective action of D6-branes in N=1 type IIA orientifolds. JHEP
1106, 105 (2011). [arXiv:1104.2329 [hep-th]]
159. A. Klemm, S. Theisen, Considerations of one modulus Calabi-Yau compactifications: Picard-
Fuchs equations, Kahler potentials and mirror maps. Nucl. Phys. B 389, 153–180 (1993).
160. P. Fre, P. Soriani, The N=2 Wonderland: From Calabi-Yau Manifolds to Topological Field
Theories (World Scinetific, New York, 1995)
161. A. Giryavets, New attractors and area codes. JHEP 0603, 020 (2006). [hep-th/0511215]
162. P. Candelas, G.T. Horowitz, A. Strominger, E. Witten, Vacuum configurations for super-
strings. Nucl. Phys. B 258, 46–74 (1985)
163. L.J. Dixon, J.A. Harvey, C. Vafa, E. Witten, Strings on Orbifolds. Nucl. Phys. B 261, 678–686
(1985); Strings on Orbifolds 2. Nucl. Phys. B 274, 285–314 (1986)
164. L.E. Ibanez, H.P. Nilles, F. Quevedo, Orbifolds and Wilson lines. Phys. Lett. B 187, 25–32
(1987)
165. D. Bailin, A. Love, Orbifold compactifications of string theory. Phys. Rept. 315, 285–408
(1999)
166. T. Kobayashi, S. Raby, R.J. Zhang, Searching for realistic 4d string models with a Pati-Salam
symmetry: Orbifold grand unified theories from heterotic string compactification on a Z(6)
orbifold. Nucl. Phys. B 704, 3–55 (2005). [arXiv:hep-ph/0409098 [hep-ph]]
167. W. Buchmuller, K. Hamaguchi, O. Lebedev, M. Ratz, Supersymmetric standard model from
the heterotic string. Phys. Rev. Lett. 96, 121602 (2006). [arXiv:hep-ph/0511035 [hep-ph]]
168. O. Lebedev, H.P. Nilles, S. Raby, S. Ramos-Sanchez, M. Ratz, P.K. Vaudrevange,
A. Wingerter, A Mini-landscape of exact MSSM spectra in heterotic orbifolds. Phys. Lett.
B 645, 88–94 (2007). [arXiv:hep-th/0611095 [hep-th]]
169. O. Lebedev, H.P. Nilles, S. Ramos-Sanchez, M. Ratz, P.K. Vaudrevange, Heterotic mini-
landscape. (II). Completing the search for MSSM vacua in a Z(6) orbifold. Phys. Lett. B
668, 331–335 (2008). [arXiv:0807.4384 [hep-th]]
References 295
170. V. Braun, Y.H. He, B.A. Ovrut, T. Pantev, The exact MSSM spectrum from string theory.
JHEP 05, 043 (2006). [arXiv:hep-th/0512177 [hep-th]]
171. V. Bouchard, R. Donagi, An SU(5) heterotic standard model. Phys. Lett. B 633, 783–791
172. R. Blumenhagen, S. Moster, T. Weigand, Heterotic GUT and standard model vacua from
simply connected Calabi-Yau manifolds. Nucl. Phys. B 751, 186–221 (2006). [arXiv:hep-
th/0603015 [hep-th]]
173. L.B. Anderson, J. Gray, A. Lukas, E. Palti, Two Hundred Heterotic Standard Models on
Smooth Calabi-Yau Threefolds. Phys. Rev. D 84, 106005 (2011). [arXiv:1106.4804 [hep-th]]
174. L.B. Anderson, J. Gray, A. Lukas, B. Ovrut, Stabilizing all geometric Moduli in Heterotic
Calabi-Yau Vacua. Phys. Rev. D 83, 106011 (2011). [arXiv:1102.0011 [hep-th]]
175. J. Polchinski, Dirichlet Branes and Ramond-Ramond charges. Phys. Rev. Lett. 75, 4724–4727
176. J. Polchinski, S. Chaudhuri, C.V. Johnson, Notes on D-branes. Lectures Presented by
J. Polchinski (1995). [arXiv:hep-th/9602052 [hep-th]]
177. F. Gmeiner, R. Blumenhagen, G. Honecker, D. Lust, T. Weigand, One in a billion: MSSM-like
D-brane statistics. JHEP 01, 004 (2006). [arXiv:hep-th/0510170 [hep-th]]
178. R. Blumenhagen, V. Braun, T.W. Grimm, T. Weigand, GUTs in type IIB orientifold
compactifications. Nucl. Phys. B 815, 1–94 (2009). [arXiv:0811.2936 [hep-th]]
179. A. Strominger, S.T. Yau, E. Zaslow, Mirror symmetry is T duality. Nucl. Phys. B 479, 243–
180. P. Candelas, X. De La Ossa, A. Font, S.H. Katz, D.R. Morrison, Mirror symmetry for two
parameter models 1. AMS/IP Stud. Adv. Math. 1, 483–543 (1996). [arXiv:hep-th/9308083
[hep-th]]; and Mirror symmetry for two parameter models 2 Nucl. Phys. B 429, 626–674
181. S. Hosono, A. Klemm, S. Theisen, Lectures on mirror symmetry. Lect. Notes Phys. 436,
235–280 (1994). [arXiv:hep-th/9403096 [hep-th]]
182. K. Hori, Trieste lectures on mirror symmetry. ICTP Lect. Notes Ser. 13, 109–202 (2003)
183. K. Hori, S. Katz, A. Klemm, R. Pandharipande, R. Thomas, C. Vafa, R. Vakil, E. Zaslow,
Mirror Symmetry (American Mathematical Society for the Clay Mathematics Institute,
New York, 2003)
184. C. Vafa, Evidence for F theory. Nucl. Phys. B 469, 403 (1996). [hep-th/9602022]
185. A. Sen, F theory and orientifolds. Nucl. Phys. B 475, 562–578 (1996). [arXiv:hep-th/9605150
[hep-th]]
186. T. Weigand, Lectures on F-theory compactifications and model building. Class. Quant. Grav.
27, 214004 (2010). [arXiv:1009.3497 [hep-th]].
187. J.J. Heckman, Particle Physics Implications of F-theory. Ann. Rev. Nucl. Part. Sci. 60, 237
(2010). [arXiv:1001.0577 [hep-th]]
188. C. Beasley, J.J. Heckman, C. Vafa, GUTs and Exceptional Branes in F-theory-I. JHEP 01,
058 (2009). [arXiv:0802.3391 [hep-th]]
189. R. Donagi, M. Wijnholt, Model building with F-Theory. Adv. Theor. Math. Phys. 15(5), 1237–
1317 (2011). [arXiv:0802.2969 [hep-th]]
190. L. Susskind, The anthropic landscape of string theory (2003). [arXiv:hep-th/0302219 [hep-
th]]
191. R. Bousso, J. Polchinski, Quantization of four form fluxes and dynamical neutralization of
the cosmological constant. JHEP 0006, 006 (2000). [hep-th/0004134]
192. S. Kachru, R. Kallosh, A.D. Linde, S.P. Trivedi, De Sitter vacua in string theory. Phys. Rev.
D 68, 046005 (2003). [hep-th/0301240]
193. V. Balasubramanian, P. Berglund, J.P. Conlon, F. Quevedo, Systematics of moduli stabilisa-
tion in Calabi-Yau flux compactifications. JHEP 0503, 007 (2005). [hep-th/0502058]
194. A. Schellekens, The Landscape ’avant la lettre’ (2006). [arXiv:physics/0604134 [physics]]
195. S. Coleman, Aspects of Symmetry: Selected Erice Lectures. (Cambridge University,
Cambridge, 1985)
296 References
196. R. Blumenhagen, M. Cvetic, S. Kachru, T. Weigand, D-brane instantons in type II orientifolds.

Ann. Rev. Nucl. Part. Sci. 59, 269–296 (2009). [arXiv:0902.3251 [hep-th]]
197. M. Bianchi, S. Kovacs, G. Rossi, Instantons and supersymmetry. Lect. Notes Phys. 737, 303–
198. A. Hebecker, T. Mikhail, P. Soler, Euclidean wormholes, baby universes, and their impact on
particle physics and cosmology. Front. Astron. Space Sci. 5, 35 (2018). [arXiv:1807.00824
[hep-th]]
199. J.L. Feng, J. March-Russell, S. Sethi, F. Wilczek, Saltatory relaxation of the cosmological
constant. Nucl. Phys. B 602, 307–328 (2001). [arXiv:hep-th/0005276 [hep-th]]
200. S. Gukov, C. Vafa, E. Witten, CFT’s from Calabi-Yau four folds. Nucl. Phys. B 584, 69
(2000); Erratum: [Nucl. Phys. B 608, 477 (2001)]. [hep-th/9906070]
201. K. Dasgupta, G. Rajesh, S. Sethi, M theory, orientifolds and G-flux. JHEP 9908, 023 (1999).
[hep-th/9908088]
202. M. Grana, Flux compactifications in string theory: a Comprehensive review. Phys. Rept. 423,
91–158 (2006). [arXiv:hep-th/0509003 [hep-th]]
203. E. Cremmer, S. Ferrara, C. Kounnas, D.V. Nanopoulos, Naturally vanishing cosmological
constant in N=1 supergravity. Phys. Lett. B 133, 61 (1983)
204. S. Sethi, Supersymmetry breaking by fluxes. JHEP 10, 022 (2018). [arXiv:1709.03554 [hep-
th]]
205. A. Westphal, de Sitter string vacua from Kahler uplifting. JHEP 03, 102 (2007). [arXiv:hep-
th/0611332 [hep-th]]
206. V. Balasubramanian, P. Berglund, Stringy corrections to Kahler potentials, SUSY breaking,
and the cosmological constant problem. JHEP 11, 085 (2004). [arXiv:hep-th/0408054 [hep-
th]]
207. J. Derendinger, L.E. Ibanez, H.P. Nilles, On the Low-Energy d = 4, N=1 supergravity theory
extracted from the d = 10, N=1 superstring. Phys. Lett. B 155, 65–70 (1985)
208. M. Dine, R. Rohm, N. Seiberg, E. Witten, Gluino condensation in superstring models. Phys.
Lett. B 156, 55–60 (1985)
209. E. Witten, Nonperturbative superpotentials in string theory. Nucl. Phys. B 474, 343–360
210. M. Bianchi, A. Collinucci, L. Martucci, Magnetized E3-brane instantons in F-theory. JHEP
12, 045 (2011). [arXiv:1107.3732 [hep-th]]
211. E. Palti, C. Vafa, T. Weigand, Supersymmetric protection and the Swampland. JHEP 06, 168
(2020). [arXiv:2003.10452 [hep-th]]
212. F. Denef, M.R. Douglas, Distributions of flux vacua. JHEP 0405, 072 (2004). [hep-
th/0404116]
213. M.A. Luty, N. Okada, Almost no scale supergravity. JHEP 04, 050 (2003). [arXiv:hep-
th/0209178 [hep-th]]
214. K. Choi, A. Falkowski, H.P. Nilles, M. Olechowski, Soft supersymmetry breaking in KKLT
flux compactification. Nucl. Phys. B 718, 113–133 (2005). [arXiv:hep-th/0503216 [hep-th]]
215. F. Brummer, A. Hebecker, M. Trapletti, SUSY breaking mediation by throat fields. Nucl.
Phys. B 755, 186–198 (2006). [arXiv:hep-th/0605232 [hep-th]]
216. E. Dudas, C. Papineau, S. Pokorski, Moduli stabilization and uplifting with dynamically
generated F-terms. JHEP 02, 028 (2007). [arXiv:hep-th/0610297 [hep-th]]
217. U.H. Danielsson, T. Van Riet, What if string theory has no de Sitter vacua? Int. J. Mod. Phys.
D 27(12), 1830007 (2018). [arXiv:1804.01120 [hep-th]]
218. H. Ooguri, E. Palti, G. Shiu, C. Vafa, Distance and de Sitter Conjectures on the Swampland.
Phys. Lett. B 788, 180 (2019). [arXiv:1810.05506 [hep-th]]
219. S.H.S. Alexander, Inflation from D-anti-D-brane annihilation. Phys. Rev. D 65, 023507
220. G.R. Dvali, Q. Shafi, S. Solganik, D-brane inflation (2001). [arXiv:hep-th/0105203 [hep-th]]
221. C.P. Burgess, M. Majumdar, D. Nolte, F. Quevedo, G. Rajesh, R.J. Zhang, The Inflationary
brane anti-brane universe. JHEP 07, 047 (2001). [arXiv:hep-th/0105204 [hep-th]]
222. P. Candelas, X.C. de la Ossa, Comments on conifolds. Nucl. Phys. B 342, 246 (1990)
References 297
223. I.R. Klebanov, M.J. Strassler, Supergravity and a confining gauge theory: duality cascades
and chi SB resolution of naked singularities. JHEP 0008, 052 (2000). [hep-th/0007191]
224. I. Bena, E. Dudas, M. Graña, S. Lüst, Uplifting runaways. Fortsch. Phys. 67(1–2), 1800100
(2019). [arXiv:1809.06861 [hep-th]]
225. R. Blumenhagen, D. Klaewer, L. Schlechter, Swampland Variations on a Theme by KKLT
(2019). arXiv:1902.07724 [hep-th]
226. I. Bena, M. Graña, S. Kuperstein, S. Massai, Giant Tachyons in the landscape. JHEP 1502,
146 (2015). [arXiv:1410.7776 [hep-th]]
227. B. Michel, E. Mintun, J. Polchinski, A. Puhm, P. Saad, Remarks on brane and antibrane
dynamics. JHEP 1509, 021 (2015). [arXiv:1412.5702 [hep-th]]
228. D. Cohen-Maldonado, J. Diaz, T. van Riet, B. Vercnocke, Observations on fluxes near anti-
branes. JHEP 1601, 126 (2016). [arXiv:1507.01022 [hep-th]]
229. J. Polchinski, Brane/antibrane dynamics and KKLT stability (2015). arXiv:1509.05710 [hep-
th]
230. I. Bena, J. Blåbäck, D. Turton, Loop corrections to the antibrane potential. JHEP 1607, 132
(2016). [arXiv:1602.05959 [hep-th]]
231. U.H. Danielsson, F.F. Gautason, T. Van Riet, Unstoppable brane-flux decay of D6 branes.
JHEP 1703, 141 (2017). [arXiv:1609.06529 [hep-th]]
232. S. Kachru, J. Pearson, H.L. Verlinde, Brane/flux annihilation and the string dual of a
nonsupersymmetric field theory. JHEP 0206, 021 (2002). [hep-th/0112197]
233. S. Kachru, R. Kallosh, A.D. Linde, J.M. Maldacena, L.P. McAllister, S.P. Trivedi, Towards
inflation in string theory. JCAP 0310, 013 (2003). [hep-th/0308055]
234. S.B. Giddings, A. Maharana, Dynamics of warped compactifications and the shape of the
warped landscape. Phys. Rev. D 73, 126003 (2006). [arXiv:hep-th/0507158 [hep-th]]
235. J. Moritz, A. Retolaza, A. Westphal, Toward de Sitter space from ten dimensions. Phys. Rev.
D 97(4), 046010 (2018). [arXiv:1707.08678 [hep-th]]
236. Y. Hamada, A. Hebecker, G. Shiu, P. Soler, On brane gaugino condensates in 10d. JHEP 1904,
008 (2019). [arXiv:1812.06097 [hep-th]]
237. R. Kallosh, Gaugino condensation and geometry of the perfect square. Phys. Rev. D 99(6),
066003 (2019). [arXiv:1901.02023 [hep-th]]
238. Y. Hamada, A. Hebecker, G. Shiu, P. Soler, Understanding KKLT from a 10d perspective
(2019). arXiv:1902.01410 [hep-th]
239. F.F. Gautason, V. Van Hemelryck, T. Van Riet, G. Venken, A 10d view on the KKLT AdS
vacuum and uplifting (2019). arXiv:1902.01415 [hep-th]
240. F. Carta, J. Moritz, A. Westphal, Gaugino condensation and small uplifts in KKLT (2019).
arXiv:1902.01412 [hep-th]
241. X. Gao, A. Hebecker, D. Junghans, Control issues of KKLT (2020). arXiv:2009.03914
242. J.P. Conlon, F. Quevedo, K. Suruliz, Large-volume flux compactifications: Moduli spectrum
and D3/D7 soft supersymmetry breaking. JHEP 08, 007 (2005). [arXiv:hep-th/0505076 [hep-
th]]
243. M. Cicoli, S. Krippendorf, C. Mayrhofer, F. Quevedo, R. Valandro, D-Branes at del
Pezzo singularities: global embedding and moduli stabilisation. JHEP 09, 019 (2012).
[arXiv:1206.5237 [hep-th]]
244. M. Cicoli, D. Klevers, S. Krippendorf, C. Mayrhofer, F. Quevedo, R. Valandro, Explicit
de sitter flux vacua for global string models with chiral matter. JHEP 05, 001 (2014).
[arXiv:1312.0014 [hep-th]]
245. M.B. Green, Interconnections between type II superstrings, M theory and N=4 supersymmet-
ric Yang-Mills. Lect. Notes Phys. 525, 22 (1999). [arXiv:hep-th/9903124 [hep-th]]
246. K. Becker, M. Becker, M. Haack, J. Louis, Supersymmetry breaking and alpha-prime
corrections to flux induced potentials. JHEP 06, 060 (2002). [arXiv:hep-th/0204254 [hep-th]]
247. D. Cremades, M.P. Garcia del Moral, F. Quevedo, K. Suruliz, Moduli stabilisation and de
Sitter string vacua from magnetised D7 branes. JHEP 05, 100 (2007). [arXiv:hep-th/0701154
[hep-th]]
298 References
248. C. Burgess, R. Kallosh, F. Quevedo, De Sitter string vacua from supersymmetric D terms.
249. G. von Gersdorff, A. Hebecker, Kahler corrections for the volume modulus of flux compacti-
fications. Phys. Lett. B 624, 270–274 (2005). [arXiv:hep-th/0507131 [hep-th]]
250. M. Cicoli, J.P. Conlon, F. Quevedo, Systematics of string loop corrections in type IIB Calabi-
Yau flux compactifications. JHEP 01, 052 (2008). [arXiv:0708.1873 [hep-th]]
251. M. Berg, M. Haack, E. Pajer, Jumping through loops: on soft terms from large volume
compactifications. JHEP 09, 031 (2007). [arXiv:0704.0737 [hep-th]]
252. M. Berg, M. Haack, B. Kors, String loop corrections to Kahler potentials in orientifolds.
JHEP 11, 030 (2005). [arXiv:hep-th/0508043 [hep-th]]: On volume stabilization by quantum
corrections. Phys. Rev. Lett. 96, 021601 (2006). [arXiv:hep-th/0508171 [hep-th]]
253. M.R. Douglas, The statistics of string/M theory vacua. JHEP 0305, 046 (2003). [hep-
th/0303194].
254. S. Ashok, M.R. Douglas, Counting flux vacua. JHEP 0401, 060 (2004). [hep-th/0307049]
255. A. Klemm, B. Lian, S.S. Roan, S.T. Yau, Calabi-Yau fourfolds for M theory and F theory
compactifications. Nucl. Phys. B 518, 515 (1998). [hep-th/9701023]
256. W. Taylor, Y.N. Wang, The F-theory geometry with most flux vacua. JHEP 1512, 164 (2015).
[arXiv:1511.03209 [hep-th]]
257. V.S. Kaplunovsky, J. Louis, Model independent analysis of soft terms in effective supergravity
and in string theory. Phys. Lett. B 306, 269–275 (1993). [arXiv:hep-th/9303040 [hep-th]]
258. A. Brignole, L.E. Ibanez, C. Munoz, Soft supersymmetry breaking terms from supergravity
and superstring models. Adv. Ser. Direct. High Energy Phys. 18, 125–148 (1998). [arXiv:hep-
ph/9707209 [hep-ph]]
259. J.P. Conlon, D. Cremades, F. Quevedo, Kahler potentials of chiral matter fields for Calabi-Yau
string compactifications. JHEP 01, 022 (2007). [arXiv:hep-th/0609180 [hep-th]]
260. L. Aparicio, M. Cicoli, S. Krippendorf, A. Maharana, F. Muia, F. Quevedo, Sequestered de
sitter string scenarios: soft-terms. JHEP 11, 071 (2014). [arXiv:1409.1931 [hep-th]]
261. L. Susskind, Supersymmetry breaking in the anthropic landscape, in From Fields to Strings,
vol. 3 ed. by M. Shifman et al. (2004), p. 1745. [arXiv:hep-th/0405189 [hep-th]]
262. M.R. Douglas, Statistical analysis of the supersymmetry breaking scale (2004). [arXiv:hep-
th/0405279 [hep-th]]
263. G. Giudice, R. Rattazzi, Living dangerously with low-energy supersymmetry. Nucl. Phys. B
757, 19–46 (2006). [arXiv:hep-ph/0606105 [hep-ph]]
264. B.S. Acharya, K. Bobkov, G.L. Kane, J. Shao, P. Kumar, The G(2)-MSSM: an M Theory
motivated model of particle physics. Phys. Rev. D 78, 065038 (2008). [arXiv:0801.0478 [hep-
ph]]
265. H. Baer, V. Barger, D. Sengupta, Landscape solution to the SUSY flavor and CP problems.
Phys. Rev. Res. 1(3), 033179 (2019). [arXiv:1910.00090 [hep-ph]]
266. H. Baer, V. Barger, S. Salam, D. Sengupta, K. Sinha, Midi-review: status of weak scale
supersymmetry after LHC Run 2 and ton-scale noble liquid WIMP searches (2020).
[arXiv:2002.03013 [hep-ph]]
267. I. Broeckel, M. Cicoli, A. Maharana, K. Singh, K. Sinha, Moduli Stabilisation and the
Statistics of SUSY Breaking in the Landscape (2020). [arXiv:2007.04327 [hep-th]]
268. A. Ringwald, L.J. Rosenberg, G. Rybka, Axions and Other Similar Particles. Review article
in Ref. [43]
269. J.P. Conlon, The QCD axion and moduli stabilisation. JHEP 05, 078 (2006). [arXiv:hep-
th/0602233 [hep-th]]
270. P. Svrcek, E. Witten, Axions in string theory. JHEP 06, 051 (2006). [arXiv:hep-th/0605206
[hep-th]]
271. A. Arvanitaki, S. Dimopoulos, S. Dubovsky, N. Kaloper, J. March-Russell, String Axiverse.
Phys. Rev. D 81, 123530 (2010). [arXiv:0905.4720 [hep-th]]
272. J. Jaeckel, A. Ringwald, The low-energy frontier of particle physics. Ann. Rev. Nucl. Part.
Sci. 60, 405–437 (2010). [arXiv:1002.0329 [hep-ph]]
References 299
273. M. Cicoli, M. Goodsell, A. Ringwald, The type IIB string axiverse and its low-energy
phenomenology. JHEP 10, 146 (2012). [arXiv:1206.0819 [hep-th]]
274. J. Halverson, P. Langacker, TASI lectures on remnants from the string landscape. PoS
TASI2017, 019 (2018). [arXiv:1801.03503 [hep-th]]
275. A. Hebecker, S.C. Kraus, M. Kuntzler, D. Lust, T. Weigand, Fluxbranes: moduli stabilisation
and inflation. JHEP 01, 095 (2013). [arXiv:1207.2766 [hep-th]]
276. V. Mukhanov, Physical Foundations of Cosmology. (Cambridge University, Cambridge,
2005)
277. S. Weinberg, Cosmology (Oxford University, Oxford, 2008)
278. P.J.E. Peebles, Principles of Physical Cosmology (Princeton University, Princeton, 1993)
279. E.W. Kolb, M.S. Turner, The Early Universe (Westview Press, Boulder, 1990)
280. A.A. Starobinsky, A new type of isotropic cosmological models without singularity. Phys.
Lett. 91B, 99 (1980)
281. A.H. Guth, The inflationary universe: a possible solution to the Horizon and Flatness
problems. Phys. Rev. D23, 347 (1981)
282. K. Sato, First order phase transition of a vacuum and expansion of the universe. Mon. Not.
Roy. Astron. Soc. 195, 467–479 (1981)
283. A.D. Linde, A new inflationary universe scenario: a possible solution of the horizon, flatness,
homogeneity, isotropy and primordial monopole problems. Phys. Lett. 108B, 389 (1982)
284. A. Albrecht, P.J. Steinhardt, Cosmology for grand unified theories with radiatively induced
symmetry breaking. Phys. Rev. Lett. 48, 1220 (1982)
285. V.F. Mukhanov, G.V. Chibisov, Quantum fluctuations and a nonsingular universe. JETP Lett.
33, 532–535 (1981)
286. A.D. Linde, Chaotic inflation. Phys. Lett. 129B, 177 (1983); Scalar field fluctuations in
expanding universe and the new inflationary universe scenario. Phys. Lett. 116B, 335 (1982)
287. M. Spradlin, A. Strominger, A. Volovich, Les Houches lectures on de Sitter space. hep-
th/0110007
288. A. Riotto, Inflation and the theory of cosmological perturbations. ICTP Lect. Notes Ser. 14,
317 (2003). [hep-ph/0210162]
289. D. Baumann, L. McAllister, Inflation and String Theory (2014). arXiv:1404.2601 [hep-th];
Inflation and String Theory. Cambridge Monographs on Mathematical Physics (2015)
290. A. Westphal, String cosmology—Large-field inflation in string theory. Int. J. Mod. Phys. A
30(09), 1530024 (2015). [arXiv:1409.5350 [hep-th]]
291. F. Quevedo, Lectures on string/brane cosmology. Class. Quant. Grav. 19, 5721–5779 (2002).
292. Y. Akrami et al. [Planck Collaboration], Planck 2018 results. X. Constraints on inflation
(2018). arXiv:1807.06211 [astro-ph.CO]
293. A. Berera, Warm inflation. Phys. Rev. Lett. 75, 3218 (1995). [astro-ph/9509049]
294. D.H. Lyth, D. Wands, Generating the curvature perturbation without an inflaton. Phys. Lett.
B 524, 5 (2002). [hep-ph/0110002]
295. M. Alishahiha, E. Silverstein, D. Tong, DBI in the sky. Phys. Rev. D 70, 123505 (2004).
[hep-th/0404084]
296. R.H. Brandenberger, C. Vafa, Superstrings in the early universe. Nucl. Phys. B 316, 391
(1989)
297. M. Gasperini, G. Veneziano, Pre- big bang in string cosmology. Astropart. Phys. 1, 317
(1993). [hep-th/9211021]
298. H. Ooguri, C. Vafa, Non-supersymmetric AdS and the Swampland. Adv. Theor. Math. Phys.
21, 1787 (2017). [arXiv:1610.01533 [hep-th]]
299. B. Freivogel, M. Kleban, Vacua Morghulis (2016). [arXiv:1610.04564 [hep-th]]
300. J.M. Maldacena, The large N limit of superconformal field theories and supergravity. Int. J.
Theor. Phys. 38, 1113 (1999). [Adv. Theor. Math. Phys. 2, 231 (1998)]. [hep-th/9711200]
301. O. Aharony, S.S. Gubser, J.M. Maldacena, H. Ooguri, Y. Oz, Large N field theories, string
theory and gravity. Phys. Rept. 323, 183 (2000). [hep-th/9905111]
300 References
302. G. Obied, H. Ooguri, L. Spodyneiko, C. Vafa, De Sitter Space and the Swampland (2018).
arXiv:1806.08362 [hep-th]
303. S.K. Garg, C. Krishnan, Bounds on Slow Roll and the de Sitter Swampland. JHEP 11, 075
(2019). [arXiv:1807.05193 [hep-th]]
304. M. Reid, The moduli space of 3-folds with K = 0 may nevertheless be irreducible. Math.
Ann. 278, 329 (1987)
305. J. Carifio, W.J. Cunningham, J. Halverson, D. Krioukov, C. Long, B.D. Nelson, Vacuum
selection from cosmology on networks of string geometries. Phys. Rev. Lett. 121(10), 101602
(2018). [arXiv:1711.06685 [hep-th]]
306. S.R. Coleman, V. Glaser, A. Martin, Action minima among solutions to a class of euclidean
scalar field equations. Commun. Math. Phys. 58, 211 (1978)
307. S.R. Coleman, F. De Luccia, Gravitational effects on and of vacuum decay. Phys. Rev. D 21,
3305 (1980)
308. S.J. Parke, Gravity, the decay of the false vacuum and the New Inflationary Universe Scenario.
Phys. Lett. 121B, 313 (1983)
309. D. Lindley, The appearance of bubbles in de Sitter space. Nucl. Phys. B 236, 522 (1984)
310. J.D. Brown, C. Teitelboim, Dynamical neutralization of the cosmological constant. Phys. Lett.
B 195, 177 (1987); Neutralization of the cosmological constant by membrane creation. Nucl.
Phys. B 297, 787 (1988)
311. K.M. Lee, E.J. Weinberg, Decay of the true vacuum in curved space-time. Phys. Rev. D 36,
1088 (1987)
312. D. Schwartz-Perlov, A. Vilenkin, Probabilities in the Bousso-Polchinski multiverse. JCAP
0606, 010 (2006). [hep-th/0601162]
313. M.C. Johnson, Vacuum transitions and eternal inflation. PhD Thesis (University of California,
Santa Cruz, 2007). http://inspirehep.net/record/1263739/files/thesis.pdf
314. K. Eckerle, A Simple System For Coleman-De Luccia Transitions (2020). [arXiv:2003.04365
[hep-th]]
315. F.J. Dyson, Time without end: physics and biology in an open universe. Rev. Mod. Phys. 51,
447 (1979)
316. B. Freivogel, M. Kleban, M. Rodriguez Martinez, L. Susskind, Observational consequences
of a landscape. JHEP 03, 039 (2006). [arXiv:hep-th/0505232 [hep-th]]
317. M. Kleban, Cosmic bubble collisions. Class. Quant. Grav. 28, 204008 (2011).
[arXiv:1107.2593]
318. A. Vilenkin, A measure of the multiverse. J. Phys. A 40, 6777 (2007). [hep-th/0609193]
319. B. Freivogel, Making predictions in the multiverse. Class. Quant. Grav. 28, 204007 (2011).
[arXiv:1105.0244 [hep-th]]
320. B.S. Acharya, M.R. Douglas, A Finite Landscape? hep-th/0606212
321. S. Weinberg, Anthropic bound on the cosmological constant. Phys. Rev. Lett. 59, 2607 (1987)
322. A.D. Linde, The inflationary universe. Rept. Prog. Phys. 47, 925 (1984)
323. J.D. Barrow, F.J. Tipler, The Anthropic Cosmological Principle (Oxford University, Oxford,
1986)
324. C.J. Hogan, Why the universe is just so. Rev. Mod. Phys. 72, 1149–1161 (2000). [arXiv:astro-
ph/9909295 [astro-ph]]
325. M. Tegmark, A. Aguirre, M. Rees, F. Wilczek, Dimensionless constants, cosmology and other
dark matters. Phys. Rev. D 73, 023505 (2006). [arXiv:astro-ph/0511774 [astro-ph]]
326. L.J. Hall, Y. Nomura, Evidence for the Multiverse in the Standard Model and Beyond. Phys.
Rev. D 78, 035001 (2008). [arXiv:0712.2454 [hep-ph]]
327. A.D. Linde, A. Mezhlumian, Stationary universe. Phys. Lett. B 307, 25 (1993). [gr-
qc/9304015]
328. A.D. Linde, D.A. Linde, A. Mezhlumian, From the Big Bang theory to the theory of a
stationary universe. Phys. Rev. D 49, 1783 (1994). [gr-qc/9306035]
329. J. Garriga, D. Schwartz-Perlov, A. Vilenkin, S. Winitzki, Probabilities in the inflationary
multiverse. JCAP 0601, 017 (2006). [hep-th/0509184]
References 301
330. R. Bousso, Complementarity in the multiverse. Phys. Rev. D 79, 123524 (2009).
[arXiv:0901.4806 [hep-th]]
331. L. Susskind, The Census taker’s hat (2007). [arXiv:0710.1129 [hep-th]]
332. R. Bousso, B. Freivogel, I.S. Yang, Properties of the scale factor measure. Phys. Rev. D 79,
063513 (2009). [arXiv:0808.3770 [hep-th]]
333. R. Bousso, B. Freivogel, S. Leichenauer, V. Rosenhaus, Geometric origin of coincidences and
hierarchies in the landscape. Phys. Rev. D 84, 083517 (2011). [arXiv:1012.2869 [hep-th]]
334. R. Bousso, B. Freivogel, S. Leichenauer, V. Rosenhaus, Eternal inflation predicts that time
will end. Phys. Rev. D 83, 023525 (2011). [arXiv:1009.4698 [hep-th]]
335. A. Strominger, The dS / CFT correspondence. JHEP 0110, 034 (2001). [hep-th/0106113]
336. J. Garriga, A. Vilenkin, Holographic multiverse. JCAP 01, 021 (2009). [arXiv:0809.4257
[hep-th]]
337. D. Harlow, S.H. Shenker, D. Stanford, L. Susskind, Tree-like structure of eternal inflation: a
solvable model. Phys. Rev. D 85, 063516 (2012). [arXiv:1110.0496 [hep-th]]
338. Y. Nomura, Physical theories, eternal inflation, and quantum universe. JHEP 1111, 063
(2011). [arXiv:1104.2324 [hep-th]]
339. R. Bousso, L. Susskind, The multiverse interpretation of quantum mechanics. Phys. Rev. D
85, 045007 (2012). [arXiv:1105.3796 [hep-th]]
340. A. Vilenkin, A quantum measure of the multiverse. JCAP 1405, 005 (2014).
[arXiv:1312.0682]
341. Y. Nomura, The static quantum multiverse. Phys. Rev. D 86, 083505 (2012).
[arXiv:1205.5550 [hep-th]]
342. J. Hartle, T. Hertog, One Bubble to rule them all. Phys. Rev. D 95(12), 123502 (2017).
[arXiv:1604.03580 [hep-th]]
343. B.S. DeWitt, Quantum theory of gravity 1: The Canonical Theory. Phys. Rev. 160, 1113
(1967)
344. J.A. Wheeler, Superspace and the nature of quantum geometrodynamics, in Battelle ren-
contres—1967 Lectures in Mathematics and Physics (Seattle), ed. by C. DeWitt, J.A. Wheeler
(Benjamin, New York, 1968), pp. 242–307; Adv. Ser. Astroph. Cosm. 3, 27 (1987)
345. J.B. Hartle, S.W. Hawking, Wave function of the universe. Phys. Rev. D 28, 2960 (1983).
[Adv. Ser. Astrophys. Cosmol. 3, 174 (1987)]
346. T. Banks, T C P, quantum gravity, the cosmological constant and all that. . . . Nucl. Phys. B
249, 332 (1985)
347. J.J. Halliwell, J.B. Hartle, T. Hertog, What is the No-boundary wave function of the universe?
Phys. Rev. D 99(4), 043526 (2019). [arXiv:1812.01760 [hep-th]]
348. S. Weinberg, Implications of dynamical symmetry breaking. Phys. Rev. D 13, 974–996 (1976)
349. L. Susskind, Dynamics of spontaneous symmetry breaking in the Weinberg-Salam theory.
Phys. Rev. D 20, 2619–2625 (1979)
350. S. Dimopoulos, L. Susskind, Mass without scalars. Nucl. Phys. B 155, 237–252 (1979)
351. E. Eichten, K.D. Lane, Dynamical breaking of weak interaction symmetries. Phys. Lett. B 90,
125–130 (1980)
352. D.B. Kaplan, H. Georgi, SU(2)×U(1) breaking by Vacuum Misalignment. Phys. Lett. B 136,
183–186 (1984)
353. W.A. Bardeen, C.T. Hill, M. Lindner, Minimal dynamical symmetry breaking of the standard
model. Phys. Rev. D 41, 1647 (1990)
354. S.F. King, Dynamical electroweak symmetry breaking. Rept. Prog. Phys. 58, 263–310 (1995).
355. R. Chivukula, Models of electroweak symmetry breaking: course, in Lectures at NATO
Advanced Study Institute on Quantum Field Theory (1998); Les Houches Summer School in
Heoretical Physics, Session 68: Probing the Standard Model of Particle Interactions (1997).
356. K. Lane, Two Lectures on Technicolor (2002). Preprints FERMILAB-PUB-02-040-T and
BUHEP-02-15. [arXiv:hep-ph/0202255 [hep-ph]].
302 References
357. C.T. Hill, E.H. Simmons, Strong dynamics and electroweak symmetry breaking. Phys. Rept.
358. M. Piai, Lectures on walking technicolor, holography and gauge/gravity dualities. Adv. High
Energy Phys. 2010, 464302 (2010). [arXiv:1004.0176 [hep-ph]]
359. G. Cacciapaglia, C. Pica, F. Sannino, Fundamental Composite Dynamics: A Review (2020).
[arXiv:2002.04914 [hep-ph]]
360. C. Csaba, S. Lombardo, O. Telem, TASI Lectures on Non-supersymmetric BSM Models
(2018). [arXiv:1811.04279 [hep-ph]]
361. N. Arkani-Hamed, A.G. Cohen, H. Georgi, Electroweak symmetry breaking from dimen-
sional deconstruction. Phys. Lett. B 513, 232–240 (2001). [arXiv:hep-ph/0105239 [hep-ph]]
362. N. Arkani-Hamed, A.G. Cohen, T. Gregoire, J.G. Wacker, Phenomenology of electroweak
symmetry breaking from theory space. JHEP 08, 020 (2002). [arXiv:hep-ph/0202089 [hep-
ph]]
363. N. Arkani-Hamed, A. Cohen, E. Katz, A. Nelson, T. Gregoire, J.G. Wacker, The Minimal
moose for a little Higgs. JHEP 08, 021 (2002). [arXiv:hep-ph/0206020 [hep-ph]]
364. M. Schmaltz, D. Tucker-Smith, Little Higgs review. Ann. Rev. Nucl. Part. Sci. 55, 229–270
(2005). [arXiv:hep-ph/0502182 [hep-ph]]
365. N. Arkani-Hamed, A. Cohen, E. Katz, A. Nelson, The Littlest Higgs. JHEP 07, 034 (2002).
366. N. Arkani-Hamed, S. Dimopoulos, G. Dvali, The Hierarchy problem and new dimensions at
a millimeter. Phys. Lett. B 429, 263–272 (1998). [arXiv:hep-ph/9803315 [hep-ph]]
367. I. Antoniadis, N. Arkani-Hamed, S. Dimopoulos, G. Dvali, New dimensions at a millimeter to
a Fermi and superstrings at a TeV. Phys. Lett. B 436, 257–263 (1998). [arXiv:hep-ph/9804398
[hep-ph]]
368. V.A. Rubakov, Large and infinite extra dimensions: an Introduction. Phys. Usp. 44, 871–893
369. C. Csaki, TASI lectures on extra dimensions and branes, in From Fields to Strings, vol. 2, ed.
by M. Shifman et al. (2005). [arXiv:hep-ph/0404096 [hep-ph]]
370. L. Randall, R. Sundrum, A Large mass hierarchy from a small extra dimension. Phys. Rev.
Lett. 83, 3370–3373 (1999). [arXiv:hep-ph/9905221 [hep-ph]]
371. L. Randall, R. Sundrum, An Alternative to compactification. Phys. Rev. Lett. 83, 4690–4693
372. V.A. Rubakov, M.E. Shaposhnikov, Extra space-time dimensions: towards a solution to the
cosmological constant problem. Phys. Lett. B 125, 139 (1983)
373. W.D. Goldberger, M.B. Wise, Modulus stabilization with bulk fields. Phys. Rev. Lett. 83,
4922–4925 (1999). [arXiv:hep-ph/9907447 [hep-ph]]
374. T. Gherghetta, A. Pomarol, Bulk fields and supersymmetry in a slice of AdS. Nucl. Phys. B
375. R. Contino, Y. Nomura, A. Pomarol, Higgs as a holographic pseudoGoldstone boson. Nucl.
Phys. B 671, 148–174 (2003). [arXiv:hep-ph/0306259 [hep-ph]]
376. K. Agashe, A. Delgado, M.J. May, R. Sundrum, RS1, custodial isospin and precision tests.
JHEP 08, 050 (2003). [arXiv:hep-ph/0308036 [hep-ph]]
377. K. Agashe, R. Contino, A. Pomarol, The minimal composite Higgs model. Nucl. Phys. B 719,
165–187 (2005). [arXiv:hep-ph/0412089 [hep-ph]]
378. C. Csaki, J. Hubisz, P. Meade, TASI lectures on electroweak symmetry breaking from extra
dimensions (2005). [arXiv:hep-ph/0510275 [hep-ph]]
379. T. Gherghetta, Les Houches lectures on warped models and holography (2006). [arXiv:hep-
ph/0601213 [hep-ph]]
380. G.D. Kribs, TASI 2004 lectures on the phenomenology of extra dimensions (2006).
381. R. Rattazzi, Cargese lectures on extra-dimensions, in Cargese 2003, Particle Physics and
Cosmology (2003). [arXiv:hep-ph/0607055 [hep-ph]]
382. R. Contino, The Higgs as a Composite Nambu-Goldstone Boson (2010). [arXiv:1005.4269
[hep-ph]]
References 303
383. T. Gherghetta, A Holographic View of Beyond the Standard Model Physics (2010).
[arXiv:1008.2570 [hep-ph]]
384. G. von Gersdorff, Electroweak symmetry breaking in warped extra dimensions, in Proceed-
ings of the 46th Rencontres de Moriond on Electroweak Interactions and Unified Theories
(2011). [arXiv:1107.1989 [hep-ph]]
385. S.S. Gubser, AdS / CFT and gravity. Phys. Rev. D 63, 084017 (2001). [arXiv:hep-th/9912001
[hep-th]]
386. N. Arkani-Hamed, M. Porrati, L. Randall, Holography and phenomenology. JHEP 08, 017
387. J. Cascales, G. del Moral, M.P., F. Quevedo, A. Uranga, Realistic D-brane models on warped
throats: Fluxes, hierarchies and moduli stabilization. JHEP 02, 031 (2004). [arXiv:hep-
th/0312051 [hep-th]]
388. J.F. Cascales, F. Saad, A.M. Uranga, Holographic dual of the standard model on the throat.
389. L.F. Abbott, A mechanism for reducing the value of the cosmological constant. Phys. Lett. B
150, 427–430 (1985)
390. G. Dvali, A. Vilenkin, Cosmic attractors and gauge hierarchy. Phys. Rev. D 70, 063501
391. G. Dvali, Large hierarchies from attractor vacua. Phys. Rev. D 74, 025018 (2006). [arXiv:hep-
th/0410286 [hep-th]]
392. P.W. Graham, D.E. Kaplan, S. Rajendran, Cosmological relaxation of the electroweak scale.
Phys. Rev. Lett. 115(22), 221801 (2015). [arXiv:1504.07551 [hep-ph]]
393. J. Espinosa, C. Grojean, G. Panico, A. Pomarol, O. Pujolas, G. Servant, Cosmological Higgs-
Axion interplay for a naturally small electroweak scale. Phys. Rev. Lett. 115(25), 251803
(2015). [arXiv:1506.09217 [hep-ph]]
394. E. Hardy, Electroweak relaxation from finite temperature. JHEP 11, 077 (2015).
[arXiv:1507.07525 [hep-ph]]
395. S.P. Patil, P. Schwaller, Relaxing the electroweak scale: the role of broken dS symmetry. JHEP
02, 077 (2016). [arXiv:1507.08649 [hep-ph]]
396. O. Antipin, M. Redi, The Half-composite two Higgs Doublet model and the relaxion. JHEP
12, 031 (2015). [arXiv:1508.01112 [hep-ph]]
397. J. Jaeckel, V.M. Mehta, L.T. Witkowski, Musings on cosmological relaxation and the
hierarchy problem. Phys. Rev. D 93(6), 063522 (2016). [arXiv:1508.03321 [hep-ph]]
398. N. Arkani-Hamed, T. Cohen, R.T. D’Agnolo, A. Hook, H. Kim, Do D. Pinner, Solving the
hierarchy problem at reheating with a large number of Degrees of Freedom. Phys. Rev. Lett.
117(25), 251801 (2016). [arXiv:1607.06821 [hep-ph]]
399. A. Arvanitaki, S. Dimopoulos, V. Gorbenko, J. Huang, K. Van Tilburg, A small weak scale
from a small cosmological constant. JHEP 05, 071 (2017). [arXiv:1609.06320 [hep-ph]]
400. L. Alberte, P. Creminelli, A. Khmelnitsky, D. Pirtskhalava, E. Trincherini, Relaxing the
cosmological constant: a proof of concept. JHEP 12, 022 (2016). [arXiv:1608.05715 [hep-
th]]
401. M. Geller, Y. Hochberg, E. Kuflik, Inflating to the weak scale. Phys. Rev. Lett. 122(19),
191802 (2019). [arXiv:1809.07338 [hep-ph]]
402. C. Cheung, P. Saraswat, Mass Hierarchy and Vacuum Energy (2018). [arXiv:1811.12390
[hep-ph]]
403. P.W. Graham, D.E. Kaplan, S. Rajendran, Relaxation of the cosmological constant. Phys. Rev.
D 100(1), 015048 (2019). [arXiv:1902.06793 [hep-ph]]
404. A. Strumia, D. Teresi, Cosmological constant: relaxation vs multiverse. Phys. Lett. B 797,
134901 (2019). [arXiv:1904.07876 [gr-qc]]
405. G. Giudice, A. Kehagias, A. Riotto, The Selfish Higgs. JHEP 10, 199 (2019).
[arXiv:1907.05370 [hep-ph]]
406. I.M. Bloch, C. Csaki, M. Geller, T. Volansky, Crunching Away the Cosmological Constant
Problem: Dynamical Selection of a Small Λ (2019). [arXiv:1912.08840 [hep-ph]]
407. N. Kaloper, A. Westphal, A Goldilocks Higgs (2019). [arXiv:1907.05837 [hep-th]]
304 References
408. L. McAllister, P. Schwaller, G. Servant, J. Stout, A. Westphal, Runaway relaxion monodromy.

JHEP 02, 124 (2018). [arXiv:1610.05320 [hep-th]]
409. C. Vafa, The String Landscape and the Swampland (2005). [arXiv:hep-th/0509212 [hep-th]]
410. H. Ooguri, C. Vafa, On the geometry of the String Landscape and the Swampland. Nucl. Phys.
B 766, 21–33 (2007). [arXiv:hep-th/0605264 [hep-th]]
411. E. Palti, The Swampland: introduction and review. Fortsch. Phys. 67(6), 1900037 (2019).
[arXiv:1903.06239 [hep-th]]
412. T.D. Brennan, F. Carta, C. Vafa, The String Landscape, the Swampland, and the missing
corner. PoS TASI2017, 015 (2017). [arXiv:1711.00864 [hep-th]]
413. T. Banks, L.J. Dixon, Constraints on string vacua with space-time supersymmetry. Nucl. Phys.
B 307, 93–108 (1988)
414. M. Kamionkowski, J. March-Russell, Planck scale physics and the Peccei-Quinn mechanism.
Phys. Lett. B 282, 137–141 (1992). [arXiv:hep-th/9202003 [hep-th]]
415. R. Holman, S.D. Hsu, T.W. Kephart, E.W. Kolb, R. Watkins, L.M. Widrow, Solutions to the
strong CP problem in a world with gravity. Phys. Lett. B 282, 132–136 (1992). [arXiv:hep-
ph/9203206 [hep-ph]]
416. R. Kallosh, A.D. Linde, D.A. Linde, L. Susskind, Gravity and global symmetries. Phys. Rev.
D 52, 912–935 (1995). [arXiv:hep-th/9502069 [hep-th]]
417. T. Banks, N. Seiberg, Symmetries and strings in field theory and gravity. Phys. Rev. D 83,
084019 (2011). [arXiv:1011.5120 [hep-th]]
418. D. Harlow, H. Ooguri, Symmetries in Quantum Field Theory and Quantum Gravity (2018).
[arXiv:1810.05338 [hep-th]]
419. S.J. Lee, W. Lerche, T. Weigand, Emergent Strings from Infinite Distance Limits (2019).
[arXiv:1910.01135 [hep-th]]
420. D. Klaewer, E. Palti, Super-Planckian spatial field variations and quantum gravity. JHEP 01,
088 (2017). [arXiv:1610.00010 [hep-th]]
421. N. Arkani-Hamed, L. Motl, A. Nicolis, C. Vafa, The String landscape, black holes and gravity
as the weakest force. JHEP 06, 060 (2007). [arXiv:hep-th/0601001 [hep-th]]
422. L. Susskind, Trouble for remnants (1995). [arXiv:hep-th/9501106 [hep-th]]
423. S. Lee, W. Lerche, T. Weigand, Tensionless strings and the weak gravity conjecture. JHEP
10, 164 (2018). [arXiv:1808.05958 [hep-th]]
424. C. Cheung, G.N. Remmen, Naturalness and the weak gravity conjecture. Phys. Rev. Lett. 113,
051601 (2014). [arXiv:1402.2287 [hep-ph]]
425. A. de la Fuente, P. Saraswat, R. Sundrum, Natural inflation and quantum gravity. Phys. Rev.
Lett. 114(15), 151303 (2015). [arXiv:1412.3457 [hep-th]]
426. T. Rudelius, Constraints on Axion inflation from the weak gravity conjecture. JCAP 09, 020
(2015). [arXiv:1503.00795 [hep-th]]
427. M. Montero, A.M. Uranga, I. Valenzuela, Transplanckian axions!? JHEP 08, 032 (2015).
[arXiv:1503.03886 [hep-th]]
428. J. Brown, W. Cottrell, G. Shiu, P. Soler, Fencing in the swampland: quantum gravity
constraints on large field inflation. JHEP 10, 023 (2015). [arXiv:1503.04783 [hep-th]]
429. T.C. Bachlechner, C. Long, L. McAllister, Planckian axions and the weak gravity conjecture.
JHEP 01, 091 (2016). [arXiv:1503.07853 [hep-th]]
430. A. Hebecker, P. Mangat, F. Rompineve, L.T. Witkowski, Winding out of the swamp: evading
the weak gravity conjecture with f-term winding inflation? Phys. Lett. B 748, 455–462 (2015).
[arXiv:1503.07912 [hep-th]]
431. D. Junghans, Large-field inflation with multiple axions and the weak gravity conjecture. JHEP
02, 128 (2016). [arXiv:1504.03566 [hep-th]]
432. B. Heidenreich, M. Reece, T. Rudelius, Weak gravity strongly constrains large-field axion
inflation. JHEP 12, 108 (2015). [arXiv:1506.03447 [hep-th]]
433. K. Freese, J.A. Frieman, A.V. Olinto, Natural inflation with pseudo—Nambu-Goldstone
bosons. Phys. Rev. Lett. 65, 3233–3236 (1990).
434. T. Banks, M. Dine, P.J. Fox, E. Gorbatov, On the possibility of large axion decay constants.
JCAP 06, 001 (2003). [arXiv:hep-th/0303252 [hep-th]]
References 305
435. E. Silverstein, A. Westphal, Monodromy in the CMB: gravity waves and string inflation. Phys.
Rev. D 78, 106003 (2008). [arXiv:0803.3085 [hep-th]]
436. L. McAllister, E. Silverstein, A. Westphal, Gravity waves and linear inflation from axion
monodromy. Phys. Rev. D 82, 046003 (2010). [arXiv:0808.0706 [hep-th]]
437. N. Kaloper, L. Sorbo, A natural framework for chaotic inflation. Phys. Rev. Lett. 102, 121301
(2009). [arXiv:0811.1989 [hep-th]]
438. N. Kaloper, A. Lawrence, L. Sorbo, An ignoble approach to large field inflation. JCAP 03,
023 (2011). [arXiv:1101.0026 [hep-th]].
439. F. Marchesano, G. Shiu, A.M. Uranga, F-term axion monodromy inflation. JHEP 09, 184
(2014). [arXiv:1404.3040 [hep-th]]
440. R. Blumenhagen, E. Plauschinn, Towards universal axion inflation and reheating in string
theory. Phys. Lett. B 736, 482–487 (2014). [arXiv:1404.3542 [hep-th]]
441. A. Hebecker, S.C. Kraus, L.T. Witkowski, D7-Brane chaotic inflation. Phys. Lett. B 737, 16–
22 (2014). [arXiv:1404.3711 [hep-th]]
442. F. Baume, E. Palti, Backreacted axion field ranges in string theory. JHEP 08, 043 (2016).
[arXiv:1602.06517 [hep-th]]
443. L.E. Ibanez, M. Montero, A. Uranga, I. Valenzuela, Relaxion monodromy and the weak
gravity conjecture. JHEP 04, 020 (2016). [arXiv:1512.00025 [hep-th]]
444. T.W. Grimm, E. Palti, I. Valenzuela, Infinite distances in field space and massless towers of
states. JHEP 08, 143 (2018). [arXiv:1802.08264 [hep-th]]
445. J.E. Kim, H.P. Nilles, M. Peloso, Completing natural inflation. JCAP 01, 005 (2005).
446. G. Dvali, Three-form Gauging of Axion Symmetries and Gravity (2005). [arXiv:hep-
th/0507215 [hep-th]]
447. P. Saraswat, Weak gravity conjecture and effective field theory. Phys. Rev. D 95(2), 025013
(2017). [arXiv:1608.06951 [hep-th]]
448. D.E. Kaplan, R. Rattazzi, Large field excursions and approximate discrete symmetries from a
clockwork axion. Phys. Rev. D 93(8), 085007 (2016). [arXiv:1511.01827 [hep-ph]]
449. K. Choi, S. H. Im, Realizing the relaxion from multiple axions and its UV completion with
high scale supersymmetry. JHEP 01, 149 (2016). [arXiv:1511.00132 [hep-ph]].
450. Y. Akrami, R. Kallosh, A. Linde, V. Vardanyan, The Landscape, the Swampland and the Era
of precision cosmology. Fortsch. Phys. 67(1–2), 1800075 (2019). [arXiv:1808.09440 [hep-
th]].
451. F. Denef, A. Hebecker, T. Wrase, de Sitter swampland conjecture and the Higgs potential.
452. M. Cicoli, S. De Alwis, A. Maharana, F. Muia, F. Quevedo, De Sitter vs Quintessence in
string theory. Fortsch. Phys. 67(1–2), 1800079 (2019). [arXiv:1808.08967 [hep-th]]
453. K. Choi, D. Chway, C.S. Shin, The dS swampland conjecture with the electroweak symmetry
and QCD chiral symmetry breaking. JHEP 11, 142 (2018). [arXiv:1809.01475 [hep-th]]
454. J.P. Conlon, The de Sitter swampland conjecture and supersymmetric AdS vacua. Int. J. Mod.
Phys. A 33(29), 1850178 (2018). [arXiv:1808.05040 [hep-th]]
455. A. Hebecker, T. Wrase, The Asymptotic dS Swampland conjecture? a simplified derivation
and a potential loophole. Fortsch. Phys. 67(1–2), 1800097 (2019). [arXiv:1810.08182 [hep-
th]]
456. D. Junghans, Weakly coupled de Sitter vacua with fluxes and the swampland. JHEP 03, 150
(2019). [arXiv:1811.06990 [hep-th]]
457. M. Dine, N. Seiberg, Is the superstring weakly coupled? Phys. Lett. B 162, 299–302 (1985)
458. J.M. Maldacena, C. Nunez, Supergravity description of field theories on curved manifolds and
a no go theorem. Int. J. Mod. Phys. A 16, 822–855 (2001). [arXiv:hep-th/0007018 [hep-th]]
459. S. Kachru, M. Kim, L. McAllister, M. Zimet, de Sitter Vacua from Ten Dimensions (2019).
[arXiv:1908.04788 [hep-th]]
460. I. Bena, M. Grana, N. Kovensky, A. Retolaza, Kahler moduli stabilization from ten dimen-
sions. JHEP 10, 200 (2019). [arXiv:1908.01785 [hep-th]]
306 References
461. F.F. Gautason, M. Schillo, T. Van Riet, M. Williams, Remarks on scale separation in flux
vacua. JHEP 03, 061 (2016). [arXiv:1512.00457 [hep-th]]
462. F.F. Gautason, V. Van Hemelryck, T. Van Riet, The tension between 10D supergravity and dS
Uplifts. Fortsch. Phys. 67(1–2), 1800091 (2019). [arXiv:1810.08518 [hep-th]]
463. D. Lüst, E. Palti, C. Vafa, AdS and the Swampland. Phys. Lett. B 797, 134867 (2019).
[arXiv:1906.05225 [hep-th]]
464. M. Demirtas, M. Kim, L. Mcallister, J. Moritz, Vacua with Small Flux Superpotential (2019).
[arXiv:1912.10047 [hep-th]]
465. O. DeWolfe, A. Giryavets, S. Kachru, W. Taylor, Type IIA moduli stabilization. JHEP 07, 066
466. F. Marchesano, E. Palti, J. Quirant, A. Tomasiello, On Supersymmetric AdS4 Orientifold
Vacua (2020). [arXiv:2003.13578 [hep-th]]
467. D. Junghans, O-plane Backreaction and Scale Separation in Type IIA Flux Vacua (2020).
[arXiv:2003.06274 [hep-th]]
468. C. Wetterich, Cosmology and the Fate of Dilatation Symmetry. Nucl. Phys. B 302, 668–696
(1988). [arXiv:1711.03844 [hep-th]]
469. P. Peebles, B. Ratra, Cosmology with a time variable cosmological constant. Astrophys. J.
325, L17 (1988)
470. M. Cicoli, F.G. Pedro, G. Tasinato, Natural quintessence in string theory. JCAP 07, 044
(2012). [arXiv:1203.6655 [hep-th]]
471. A. Hebecker, T. Skrzypek, M. Wittner, The F -term problem and other challenges of stringy
quintessence. JHEP 11, 134 (2019). [arXiv:1909.08625 [hep-th]]
472. E. Hardy, S. Parameswaran, Thermal dark energy. Phys. Rev. D 101(2), 023503 (2020).
[arXiv:1907.10141 [hep-th]]
473. T. Regge, General relativity without coordinates. Nuovo Cim. 19, 558–571 (1961)
474. R.M. Williams, P.A. Tuckey, Regge calculus: a bibliography and brief review. Class. Quant.
Grav. 9, 1409–1422 (1992)
475. J. Ambjorn, R. Loll, Nonperturbative Lorentzian quantum gravity, causality and topology
change. Nucl. Phys. B 536, 407–434 (1998). [arXiv:hep-th/9805108 [hep-th]]
476. R. Loll, Quantum Gravity from causal dynamical triangulations: a review. Class. Quant. Grav.
37(1), 013002 (2020). [arXiv:1905.08669 [hep-th]]
477. J. Ambjorn, A. Görlich, J. Jurkiewicz, R. Loll, Nonperturbative quantum gravity. Phys. Rept.
519, 127–210 (2012). [arXiv:1203.3591 [hep-th]]
478. S.B. Giddings, A. Strominger, Axion induced topology change in quantum gravity and string
theory. Nucl. Phys. B 306, 890–907 (1988)
479. S.R. Coleman, Why there is nothing rather than something: a theory of the cosmological
constant. Nucl. Phys. B 310, 643–668 (1988)
480. V. Kazakov, A.A. Migdal, I. Kostov, Critical properties of randomly triangulated planar
random surfaces. Phys. Lett. B 157, 295–300 (1985)
481. J. Ambjorn, B. Durhuus, J. Frohlich, Diseases of triangulated random surface models, and
possible cures. Nucl. Phys. B 257, 433–449 (1985)
482. F. David, Planar diagrams, two-dimensional lattice gravity and surface models. Nucl. Phys. B
257, 45 (1985)
483. I.R. Klebanov, String theory in two-dimensions, in Trieste Spring School (1991). [arXiv:hep-
th/9108019 [hep-th]]
484. P.H. Ginsparg, G.W. Moore, Lectures on 2-D gravity and 2-D string theory, in Boulder 1992,
Proceedings, Recent Directions in Particle Theory (1992). [arXiv:hep-th/9304011 [hep-th]]
485. J. Ambjorn, J. Gizbert-Studnicki, A. Görlich, J. Jurkiewicz, R. Loll, Renormalization in
Quantum Theories of Geometry (2020). [arXiv:2002.01693 [hep-th]]
486. T. Thiemann, Lectures on loop quantum gravity. Lect. Notes Phys. 631, 41–135 (2003).
[arXiv:gr-qc/0210094 [gr-qc]].
487. H. Nicolai, K. Peeters, M. Zamaklar, Loop quantum gravity: an Outside view. Class. Quant.
Grav. 22, R193 (2005). [arXiv:hep-th/0501114 [hep-th]]
References 307
488. H. Nicolai, K. Peeters, Loop and spin foam quantum gravity: a brief guide for beginners. Lect.
Notes Phys. 721, 151–184 (2007). [arXiv:hep-th/0601129 [hep-th]]
489. A. Ashtekar, An introduction to loop quantum gravity through cosmology. Nuovo Cim. B
122, 135–155 (2007). [arXiv:gr-qc/0702030 [gr-qc]]
490. P. Dona, S. Speziale, Introductory lectures to loop quantum gravity, in Proceeding of
the 3rd School on Theoretical Physics Gravitation: Theory and Experiment, Jijel (2009).
[arXiv:1007.0402 [gr-qc]]
491. C. Rovelli, Zakopane lectures on loop gravity. PoS QGQGS2011, 003 (2011).
[arXiv:1102.3660 [gr-qc]]
492. C. Rovelli, F. Vidotto, Covariant Loop Quantum Gravity (Cambridge University, Cambridge,
2015)
493. A. Sen, Gravity as a spin system. Phys. Lett. B 119, 89–91 (1982)
494. A. Ashtekar, New variables for classical and quantum gravity. Phys. Rev. Lett. 57, 2244–2247
(1986)
495. S. Weinberg, Critical phenomena for field theorists, in Understanding the Fundamental
Constituents of Matter, ed. by A. Zichichi. The Subnuclear Series, vol. 14 (Springer, Boston,
1976)
496. L. Brink, Maximally supersymmetric Yang-Mills theory: the story of N = 4 Yang-Mills
theory. Int. J. Mod. Phys. A 31(01), 1630002 (2016). [arXiv:1511.02971 [hep-th]]
497. D.F. Litim, F. Sannino, Asymptotic safety guaranteed. JHEP 12, 178 (2014).
[arXiv:1406.2337 [hep-th]]
498. J. Polchinski, Renormalization and effective lagrangians. Nucl. Phys. B 231, 269–295 (1984)
499. U. Ellwanger, FLow equations for N point functions and bound states. Z. Phys. C 62, 503–510
500. T.R. Morris, The exact renormalization group and approximate solutions. Int. J. Mod. Phys.
A 9, 2411–2450 (1994). [arXiv:hep-ph/9308265 [hep-ph]]
501. C. Wetterich, Exact evolution equation for the effective potential. Phys. Lett. B 301, 90–94
(1993). [arXiv:1710.05815 [hep-th]]
502. D.F. Litim, Fixed points of quantum gravity and the renormalisation group. PoS QG-Ph, 024
(2007). [arXiv:0810.3675 [hep-th]]
503. M. Reuter, F. Saueressig, Quantum Gravity and the Functional Renormalization Group.
(Cambridge University, Cambridge, 2019)
504. M. Reuter, F. Saueressig, Quantum Einstein Gravity. New J. Phys. 14, 055022 (2012).
[arXiv:1202.2274 [hep-th]]
505. M. Niedermaier, The Asymptotic safety scenario in quantum gravity: an Introduction. Class.
Quant. Grav. 24, R171–230 (2007). [arXiv:gr-qc/0610018 [gr-qc]]
506. M. Reuter, Nonperturbative evolution equation for quantum gravity. Phys. Rev. D 57, 971–
507. A. Bonanno, A. Eichhorn, H. Gies, J.M. Pawlowski, R. Percacci, M. Reuter, F. Saueressig,
G.P. Vacca, Critical Reflections on Asymptotically Safe Gravity (2020). [arXiv:2004.06810
[gr-qc]]
508. S. de Alwis, A. Eichhorn, A. Held, J.M. Pawlowski, M. Schiffer, F. Versteegen, Asymptotic
safety, string theory and the weak gravity conjecture. Phys. Lett. B 798, 134991 (2019).
[arXiv:1907.07894 [hep-th]]
509. G. Dvali, C. Gomez, Self-Completeness of Einstein Gravity (2010). [arXiv:1005.3497 [hep-
th]]
510. G. Dvali, G.F. Giudice, C. Gomez, A. Kehagias, UV-completion by classicalization. JHEP 08,
108 (2011). [arXiv:1010.1415 [hep-ph]]
308 References
511. G. Dvali, C. Gomez, R.S. Isermann, D. Lüst, S. Stieberger, Black hole formation and
classicalization in ultra-Planckian 2 → N scattering. Nucl. Phys. B 893, 187–235 (2015).
[arXiv:1409.7405 [hep-th]]
512. J.F. Donoghue, A critique of the asymptotic safety program. Front. Phys. 8, 56 (2020).
[arXiv:1911.02967 [hep-th]]
513. R. Percacci, D. Perini, Asymptotic safety of gravity coupled to matter. Phys. Rev. D 68,
514. P. Dona, A. Eichhorn, R. Percacci, Matter matters in asymptotically safe quantum gravity.
515. J. Meibohm, J.M. Pawlowski, M. Reichert, Asymptotic safety of gravity-matter systems.
516. A. Eichhorn, Status of the asymptotic safety paradigm for quantum gravity and matter. Found.
Phys. 48(10), 1407–1429 (2018). [arXiv:1709.03696 [gr-qc]]
517. Y. Hamada, M. Yamada, Asymptotic safety of higher derivative quantum gravity non-
minimally coupled with a matter system. JHEP 08, 070 (2017). [arXiv:1703.09033 [hep-th]]
518. N. Christiansen, D.F. Litim, J.M. Pawlowski, M. Reichert, Asymptotic safety of gravity with
matter. Phys. Rev. D 97 (10), 106012 (2018). [arXiv:1710.04669 [hep-th]]
519. M. Shaposhnikov, C. Wetterich, Asymptotic safety of gravity and the Higgs boson mass. Phys.
Lett. B 683, 196–200 (2010). [arXiv:0912.0208 [hep-th]]
520. G. Degrassi, S. Di Vita, J. Elias-Miro, J.R. Espinosa, G.F. Giudice, G. Isidori, A. Strumia,
Higgs mass and vacuum stability in the Standard Model at NNLO. JHEP 08, 098 (2012).
[arXiv:1205.6497 [hep-ph]]
521. C. Wetterich, Quantum Scale Symmetry (2019). [arXiv:1901.04741 [hep-th]]
Index
Symbols B
A-terms, 59 Background dependence, 277
Bμ term, 60 Background independence, 277
CP violation, 6 Barbieri–Giudice measure, 20
D term, 47 Batyrev’s construction, 167
D-term potential, 51, 67 Bayesian inference, 21
D-term uplift, 209 Beta function, 78
F term, 48 Betti numbers, 159
R-parity, 55 Big crunch, 229
R-symmetry, 55 Bosonic string, 85
SU (3) holonomy, 154 Bounce, 232, 241
SU (5), 32, 79 Boundary conditions, 92
W boson, 27 Bousso–Polchinski model, 191
Z boson, 27 Branching rule, 32
α correction, 208 Brane, 107
CP n , 166 Brans–Dicke frame, 63, 105
μ term, 60 BRST quantisation, 94
tan β, 60 Bubble, 240, 242
θ-term, 6 nucleation, 191, 229
p-form fluxes, 187 wall, 234, 251
p-forms, 106 wall tension, 235
3-fold, 154 Bulk, 205
Buscher rules, 135
A
ADD, 261 C
Adjustment mechanism, 26 Calabi–Yau manifolds, 151
AdS/CFT correspondence, 247, 264 Canonical bundle, 163
AdS conjectures, 274 Casimir effect, 98
AdS vacuum, 201 Casimir energy, 99, 114
Anomaly, 28 Casmir operator, 79
Anthropic prediction, 244 Causal Dynamical Triangulations (CDT), 276
Anti-brane, 202 Census taker, 245
Ashtekar variables, 276 Central charge, 97
Asymptotic safety, 278 Chain, 158
Auxiliary field, 49 Chern class, 157
Axial current, 29 Chern–Simons term, 147
Axion, 151 Chiral compensator, 66
Axion-like particle (ALP), 151, 215 Chiral superfield, 46
Axiverse, 215 Chiral symmetry breaking, 257
in Physics 979, https://doi.org/10.1007/978-3-030-65151-0
310 Index
Chirality, 146 Duality, 133, 134, 188

CKM matrix, 6 Dynkin-index, 79
Clifford algebra, 36
Clockwork, 270
Closed, 159 E
Closed string, 92 E3-brane, 199
Co-differential, 160 Effective field theory (EFT), 8, 10
Cohomology, 157 Einstein frame, 105
Coleman–De Luccia, 237, 250 Electroweak hierarchy problem, 16
Coleman-Mandula theorem, 42 Electroweak symmetry breaking, 26
Complete-intersection Calabi-Yau (CICY), 167 Elliptic fibrations, 177
Complex manifold, 152 Energy-momentum tensor, 24, 88, 111
Complex structure deformations, 164 Eternal inflation, 225, 226
Component actions, 70 Euclidean approach, 23
Component fields, 46 Euler number, 137, 208, 213
Conformal-Calabi-Yau, 204 Exact, 159
Conformal field theory (CFT), 97, 117 Exact Renormalisation Group (ERG), 279
Cosmological constant, 23 Excited level, 100
problem, 23, 209 Exterior derivative, 106
Cosmological selection, 265 Extra dimensions
Covariantly constant spinor, 154 large, 261
Critical bubble, 235 warped, 259
Critical dimension, 102
Critical radius, 236 F
Critical string, 102 Faddeev–Popov ghosts, 97
Cutoff False vacuum, 234, 240
apparent horizon, 246 Fayet-Iliopoulos (FI) term, 53
lightcone-time, 246 Fermion condensate, 257
scale factor, 246 Fierz identities, 74
Cycle, 158 Fine tuning, 10, 19
Cylinder, 90, 118 Fixed point, 174
Fixed surface, 279
Flat gauge, 89
D Flatness problem, 242
DBI action, 147 Flavour, 5
D-brane, 92 Flavour-changing charged current, 5
de Rham cohomology, 159 Flavour-changing neutral currents (FCNCs),
de Sitter space, 224, 271 16, 59
euclidean, 237 Flux landscape, 187
patches, 224 Fock space, 95
Penrose diagram, 246 Friedmann–Robertson–Walker universe, 243
radius, 224 F-theory, 173
de Sitter vacuum, 202 Fubini–Study metric, 166, 182
Diffeomorphism, 23, 88
Differential form, 106
Dilaton, 104, 108, 139 G
Dimensional reduction, 178 Gauge coupling unification, 78
Dine–Seiberg problem, 272 Gauge-kinetic functions, 215
Dirac–Born–Infeld action, 147 Gaugino, 55
Dirac spinor, 3, 35 Gaugino condensation, 198
Dirichlet boundary conditions, 92 General superfield, 45
Dotted Weyl index, 37 Genus, 120
dS/CFT correspondence, 247 Giddings–Kachru–Polchinski (GKP) model,
dS vacuum, 202 195
Index 311
Giudice-Masiero mechanism, 59 metric, 50

Gliozzi-Scherk-Olive projection, 129 potential, 47, 170, 184
Goldberger-Wise mechanism, 263 transformation, 184
Goldstone boson, 259 Kahler Modulus Stabilisation and SUSY
Grand Unified Theory (GUT), 32 Breaking (KKLT), 197, 216
Grassmann variable, 43 Kalb-Ramond field, 104, 115
Gravitino, 62 Kaluza–Klein
Graviton, 83, 104 compactification, 148
Gravity, 23 reduction, 178
Gravity mediation, 59, 64 theory, 149, 178
Green-Schwarz (GS) superstring, 125 Killing vector, 67
GSO projection, 129 KK mode, 149
Gukov–Vafa–Witten superpotential, 195 KK reduction, 178
Gupta-Bleuler quantisation, 95 Klebanov–Strassler throat, 204
Guth–Vanchurin paradox, 246 K3-surface, 167
H L
Haag-Lopuszanski-Sohnius theorem, 42 Landscape, 214, 267
Harmonic, 160 Large diffeomorphisms, 131
Hawking–Hartle no-boundary proposal, 249 Large hierarchy problem, 22
Helicity, 83 Large Volume Scenario (LVS), 207, 218
Heterotic compactification, 173 Level matching, 103
Heterotic string theory, 133 Lie superalgebra, 42
Hidden sector, 56 Light-cone coordinates, 90
Hierarchy problem, 16 Lightest supersymmetric particle (LSP), 56
Higgsino, 54 Little group, 104
Higgs potential, 3 Little hierarchy problem, 22, 258
Higher-form gauge theory, 105 Little Higgs, 259
Hodge Loop Quantum Gravity (LQP), 276
decomposition theorem, 160 Low-scale SUSY, 255
diamond, 162
numbers, 162
star, 160 M
Holonomy, 153 Majorana mass, 14
Homology, 157 Majorana spinor, 146
Hopf fibration, 191 Maldacena–Nunez no-go theorem, 273
Hubble parameter, 223 Marginal operator, 7
Hypercharge, 2 Mass dimension, 9
Mass-shell condition, 100
M2-branes, 134
I Measure
Inflation, 223 causal diamond, 246
large-field, 270 fat geodesic, 246
Instanton, 188 geometric-cutoff, 247
Intermediate scale, 59 global, 244
Intersecting brane models, 93, 176 local, 244
Isospin, 27 Measure problem, 243
Metastable, 204, 227
Minimal Supersymmetric Standard Model
K (MSSM), 53, 58, 214
Kahler Mirror symmetry, 177
deformation, 164 Mode decomposition, 91
form, 153 Modern covariant approach, 94
manifold, 153, 184
312 Index
Modern covariant quantisation, 94 Quintic, 166

Modular invariance, 131
Moduli space, 163, 167
Moduli stabilisation, 187 R
Modulus, 149 Ramond boundary conditions, 126
M-theory, 134 Ramond-Neveu-Schwarz superstring, 121
Multiverse, 243 Randall–Sundrum model, 262
Rarita-Schwinger equation, 66
Real superfield, 50
N Regge calculus, 275
Nambu-Goto action, 86 Regge slope, 89
Naturalness, 22 Reheating surface, 242
Negative mode, 233 Reid’s fanatasy, 228
Neumann boundary conditions, 92 Relaxion, 265
Neutrino mass, 14 Relevant operator, 7
Neveu-Schwarz boundary conditions, 126 Remnants, 216
Noether method, 124 Ricci flatness, 152
No global symmetries conjecture, 268 Riemann surface, 119, 137
Non-renormalisation, 56 R-NS sector, 126
No-scale model, 196, 216 RNS superstring, 121
NS-NS sector, 126 R-R sector, 126
NS-R sector, 126 RS1, 262
RS2, 262
O Running and matching, 10
Observer, 244
trajectory, 246, 248
worldline, 246, 248 S
Old covariant approach, 94 Scattering amplitudes, 119
Old covariant quantisation, 94 Schrödinger wave functional, 117, 249
Open string, 92 Seesaw mechanism, 14
O-plane, 174, 202 Self-duality, 133
O’Raifeartaigh model, 53, 201 Self-energy, 11
Orbifold, 174 Sfermion, 55
Orientifold plane, 174, 202 Shift symmetry, 19
Oscillator modes, 91, 110 Simplices, 275
Simplicial homology, 158
Slow-roll
P inflation, 223
Particle Data Group (PDG), 81 parameters, 224
Period, 168 Soft terms, 57
Physical states, 100 Spontaneous symmetry breaking, 4
Picard–Fuchs equations, 172 Spurion, 58
Pion, 257 Standard Model, 1
Poincaré duality, 160 matter, 2
Point particle, 109 parameters, 5
Polchinski equation, 279 State-operator correspondence, 117
Polyakov action, 87 Stress-energy tensor, see Energy-momentum
Projective space, 165, 182 tensor, 24, 88, 111
String
coupling, 120, 139
Q frame, 105
Quantum gravity, 275 length, 91
Quasi-de Sitter, 224 tension, 87, 89
Quintessence, 274 Strong CP problem, 215
Index 313
Subcritical string, 102 Type IIA string theory, 132

Superconformal, 124 Type IIB string theory, 132
Supercritical string, 101
Supercurrent, 124
Superfield, 45 U
Supergravity, 61, 64 Unoriented strings, 133
10d, 145 Uplift, 202, 205, 209, 211, 259, 273
Superpotential, 48 UV parameters, 21
Superspace, 41
Superstring, 117
Supersymmetry (SUSY), 41, 215 V
breaking, 52 Vacuum decay, 234
gauge transformation, 50 Vacuum energy, 38
-GUT, 82 Vertex operators, 120
transformation, 46 Virasoro algebra, 96, 112
Super-Virasoro algebra, 127
Super-Weyl-invariant, 124
SUSY, see Supersymmetry (SUSY) W
Swampland, 267, 271 Warp factor, 204
de Sitter conjecture, 271 Wave function of the universe, 249
Distance Conjecture, 268 Weak Gravity Conjecture, 268
Symplectic bases, 171 Weak mixing angle, 28
Weinberg angle, 28
Weinberg operator, 13
T Weinberg’s no-go theorem, 26
Tachyon, 102 Wess-Zumino
Tadpole, 25, 212 gauge, 50
Target-space, 85 model, 49
T-duality, 134 Wetterich equation, 279
Technically natural, 19 Weyl
Technicolor, 255 rescaling, 88, 105
Tensor calculus, 66 spinor, 3, 35, 146
Terminal vacua, 229 Wheeler–DeWitt equation, 249
Throat, 204 Wick rotation, 39
Time cutoff, 243 Wilson line, 151
Toric geometry, 167 Winding states, 135
Triangle diagram, 31 Witt algebra, 96
Triangulations, 275 Worldsheet, 85
Trilinear term, 59 parity, 133, 175
True vacuum, 234, 240 supergravity, 123
Tunnelling, 229 supersymmetry, 121
field theory, 234
gravity, 237
quantum mechanics, 230 Y
Tunnelling transition, 227, 242 Youngness paradox, 245
Type I string theory, 133 Yukawa coupling, 3, 5
Type II superstring, 130

Naturalness, String Landscape and Multiverse: Arthur Hebecker

Uploaded by

Copyright:

Available Formats

Naturalness, String Landscape and Multiverse: Arthur Hebecker

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Naturalness, String Landscape and Multiverse: Arthur Hebecker

Uploaded by

Copyright:

Available Formats

Lecture Notes in Physics

• to be a compact and modern up-to-date source of reference on a well-defined

Both monographs and multi-author volumes will be considered for publication.

More information about this series at http://www.springer.com/series/5304

ISSN 0075-8450 ISSN 1616-6361 (electronic)

Heidelberg, Germany Arthur Hebecker

1 The Standard Model and Its Hierarchy Problem(s) .. . . . . . . . . . . . . . . . . . . . 1

2.13.3 Fierz Identities for Weyl Spinors .. . . . . . .. . . . . . . . . . . . . . . . . . . . 74

5.7 An Aside on String Model Building: From Heterotic

Arthur Hebecker after studying physics in Moscow,

1.1 Standard Model—The Basic Structure

A possible definition of the Standard Model is as follows: It is the most general

GSM = SU (3) × SU (2) × U (1) , (1.1)

© Springer Nature Switzerland AG 2021 1

known as hypercharge. The overall normalisation of the latter is clearly convention-

LSM = Lgauge + Lmat t er + LH iggs + LY ukawa . (1.3)

The gauge part is completely standard,

ψj ∈ { {qLa , (uaR )c , (dRa )c , lLa , (eR ) }, a = 1, 2, 3 } .

in the direct sum in (1.2). Furthermore, Rj (Aμ ) denotes the representation of

Instead, all our 4-spinors are left-handed:

In particular, this explains why we use the charge-conjugate of right-handed quarks

LH iggs = −(Dμ )† (D μ ) − V () with V () = −m2H †  + λH († )2 ,

v 2 = m2H /(2λH ) , m2h = 4λH v 2 . (1.14)

1.2 Standard Model—Parameter Count

particular, it is broken by complex lagrangian parameters which cannot be removed by field

m2H (μ)/μ2 ∼ 10−32 at μ ∼ MP , (1.22)

1.3 Effective Field Theories—Cutoff Perspective

A ∂ 2 A + g A2 ∂ A + g 2 A4 + g 4 (c1 /4 ) (∂A)4 + · · · , (1.24)

confirming our claim about the 4-gluon amplitude.

The key novelty is that we have a term proportional to a positive power of  (a

1.4 Effective Field Theories—QFTUV vs. QFTIR

8 We do not insist on being able to take the mathematical limit  → ∞ or μ1 → ∞ since we do

Inserting this back into our lagrangian we obtain

This resummed propagator can be viewed as a function of the matrix-valued

Fig. 1.1 One-loop fermion

As usual in perturbation theory we estimate

/ = m. Introducing a cutoff , we have in

m → mEF T ≡ mc ≡ m + δm ≡ m(1 + y 2 × O(1)) . (1.34)

Here ‘O(1)’ may include a logarithmic cutoff dependence, like in particular

one sees immediately that for m = 0 it possesses the Z2 symmetry

1.5 The Standard Model as an Effective Field Theory

The Weinberg operator reads

mν ∼ v 2 / ⇒  ∼ 3 × 1014 GeV . (1.45)

c ∼ β2 and ∼M. (1.47)

In other words, the observed neutrino masses behave as

As a result, we can make M (and thus ) smaller, bringing it closer to experimental

changing neutral currents or FCNCs10 (the analogues of the flavour-changing

1.6 The Electroweak Hierarchy Problem

Fig. 1.2 Contributions to the

disregarding the logarithmic running of the dimensionless couplings between μ and

, a massless scalar is explained without tuning. I am not aware of any sufficiently

Let us first emphasise that, having a small (dimensionful or dimensionless) param-

is forbidden by chiral symmetry transformations, e.g. eR → eiα eR . One can view

be mentioned. However, at our present ‘leading order’ level of discussion it is fair

∂ ln(m2H, 0 + cH 2 /(16π 2 )) m2H, 0 2 /(16π 2 )

|O − O0 | < |Oobs. − O0 | (1.65)

is small. In other words, we measure how unlikely it is that a randomly chosen

1.8 Gravity and the Cosmological Constant Problem

Including gravity in a minimalist approach amounts to the substitution

LH iggs = −(Dμ )† (D μ ) − V () with V () = −m2H † + λH († )2 ,

A ∂ 2 A + g A2 ∂ A + g 2 A4 + g 4 (c1 /4 ) (∂A)4 + · · · , (1.24)

The key novelty is that we have a term proportional to a positive power of (a

8 We do not insist on being able to take the mathematical limit → ∞ or μ1 → ∞ since we do

/ = m. Introducing a cutoff , we have in

mν ∼ v 2 / ⇒ ∼ 3 × 1014 GeV . (1.45)

c ∼ β2 and ∼M. (1.47)

As a result, we can make M (and thus ) smaller, bringing it closer to experimental

, a massless scalar is explained without tuning. I am not aware of any sufficiently

∂ ln(m2H, 0 + cH 2 /(16π 2 )) m2H, 0 2 /(16π 2 )

Rμνρ σ vσ = [Dμ , Dμ ] vρ , or symbolically: R ∼ [∂ − , ∂ − ] .

Now, since for a fundamental field we have

μ ν Mα β M α̇ β̇ (σν )β β̇ = μ ν (Mσν M † )α α̇ . (1.136)

μ ν Mα β M α̇ β̇ (σν )β β̇ = μ ν (σρ )α α̇ ρ ν = ημσ σ τ ητ ν (σρ )α α̇ ρ ν