*291*
*52*
*9MB*

*English*
*Pages 625*
*Year 2017*

- Author / Uploaded
- Ni W.-T (ed.)

This page intentionally left blank

Published by :RUOG6FLHQWL¿F3XEOLVKLQJ&R3WH/WG 7RK7XFN/LQN6LQJDSRUH 86$R৽FH:DUUHQ6WUHHW6XLWH+DFNHQVDFN1- 8.R৽FH6KHOWRQ6WUHHW&RYHQW*DUGHQ/RQGRQ:&++(

/LEUDU\RI&RQJUHVV&DWDORJLQJLQ3XEOLFDWLRQ'DWD 1DPHV1L:HL7RX±HGLWRU 7LWOH2QHKXQGUHG\HDUVRIJHQHUDOUHODWLYLW\IURPJHQHVLVDQGHPSLULFDO IRXQGDWLRQVWRJUDYLWDWLRQDOZDYHVFRVPRORJ\DQGTXDQWXPJUDYLW\ HGLWRU:HL7RX1L1DWLRQDO7VLQJ+XD8QLYHUVLW\+VLQFKX 'HVFULSWLRQ6LQJDSRUH+DFNHQVDFN1-:RUOG6FLHQWL¿F>@_,QFOXGHV ELEOLRJUDSKLFDOUHIHUHQFHV ,GHQWL¿HUV/&&1_,6%1VHWDONSDSHU _ ,6%1;VHWDONSDSHU _,6%1YDONSDSHU _ ,6%1YDONSDSHU _,6%1YDONSDSHU _ ,6%1;YDONSDSHU 6XEMHFWV/&6+*HQHUDOUHODWLYLW\3K\VLFV +LVWRU\_*UDYLWDWLRQDOZDYHV_ &RVPRORJ\_4XDQWXPJUDYLW\ &ODVVL¿FDWLRQ/&&4&2_''&GF /&UHFRUGDYDLODEOHDWKWWSOFFQORFJRY

%ULWLVK/LEUDU\&DWDORJXLQJLQ3XEOLFDWLRQ'DWD $FDWDORJXHUHFRUGIRUWKLVERRNLVDYDLODEOHIURPWKH%ULWLVK/LEUDU\

7KHFRYHU¿JXUHLVDGDSWHGIURPD¿JXUHE\'DYLG-&KDPSLRQ03,IRU5DGLRDVWURQRP\ 7KHDUWLFOHVLQWKLVWZRYROXPHVHWZHUHSUHYLRXVO\SXEOLVKHGLQYDULRXVLVVXHVRI ,QWHUQDWLRQDO-RXUQDORI0RGHUQ3K\VLFV'

&RS\ULJKWE\:RUOG6FLHQWL¿F3XEOLVKLQJ&R3WH/WG $OOULJKWVUHVHUYHG7KLVERRNRUSDUWVWKHUHRIPD\QRWEHUHSURGXFHGLQDQ\IRUPRUE\DQ\PHDQVHOHFWURQLFRU PHFKDQLFDOLQFOXGLQJSKRWRFRS\LQJUHFRUGLQJRUDQ\LQIRUPDWLRQVWRUDJHDQGUHWULHYDOV\VWHPQRZNQRZQRUWR EHLQYHQWHGZLWKRXWZULWWHQSHUPLVVLRQIURPWKHSXEOLVKHU

)RUSKRWRFRS\LQJRIPDWHULDOLQWKLVYROXPHSOHDVHSD\DFRS\LQJIHHWKURXJKWKH&RS\ULJKW&OHDUDQFH&HQWHU ,QF5RVHZRRG'ULYH'DQYHUV0$86$,QWKLVFDVHSHUPLVVLRQWRSKRWRFRS\LVQRWUHTXLUHGIURP WKHSXEOLVKHU

'HVN(GLWRU1J.DK)HH 7\SHVHWE\6WDOOLRQ3UHVV (PDLOHQTXLULHV#VWDOOLRQSUHVVFRP 3ULQWHGLQ6LQJDSRUH

Foreword

General Relativity (GR) is founded on the observation of Mercury perihelion precession anomaly discovered by Le Verrier and improved by Newcomb, the Michelson– Morley experiment, and the precision E¨ otv¨ os experiment. Its theoretical basis is based on Special Relativity (previously called the restricted theory of relativity), Einstein Equivalence Principle (EEP) and the realization that the metric is the dynamical quantity of gravity together with the Principle of General Covariance and the absence of other dynamical quantities. The establishment of GR in 1915 is a community eﬀort with Albert Einstein clearly playing the dominant role. For one hundred years, its applicability through solar system to cosmology is prevailing. If one includes the cosmological constant (proposed in 1917 by Einstein) in GR, there have not been any fully established non-applicable places. The only possible potential exception is the missing mass (dark matter)-deﬁcient acceleration issue. Dark energy and quantum gravity are needed in the present theoretical foundation of physics; however, more experimental clues are needed. The framework applicability of GR is already demonstrated in theoretical inﬂation models with quantum ﬂuctuations leading to structure formation with experimentally observed spectrum. To celebrate the GR centennial, we solicit the writing of 23 chapters in these two volumes consisting of ﬁve parts: Part I. Part II. Part III. Part IV. Part V.

Genesis, Solutions and Energy. Empirical Foundations. Gravitational Waves. Cosmology. Quantum Gravity.

Volume 1 consists of Part I, Part II and Part III; Volume 2 consists of Part IV and Part V. In Part I, Valerie Messager and Christophe Letellier start in Chapter 1 with a genesis of special relativity to set the stage. They rely on the original literature to make the development clear and connected. Thanks to many thorough researches in the last 50 years, the path to general relativity is clear. A concise exposition of the path is presented in Chapter 2. In Chapter 3, Christian Heinicke and Friedrich Hehl present the historical development and detailed properties of the basic and fundamental spherical Schwarzschild and axisymmetric Kerr solutions. In Chapter 4, Chiang-Mei Chen, James Nester and Roh-Sung Tung expound the important and useful concept of energy with its many facets and various applications. v

vi

Foreword

In Part II, the empirical foundations of GR are examined. First, the cornerstone Einstein Equivalence Principle (EEP) is explored. Ever since about 100 ps (the time of electroweak phase transition or the equivalent/substitute) at the quasiequilibrium Higgs/intermediate boson energy scale from the Big Bang (or equivalent/substitute), photons and charged particles are abundant. With the premetric formulation of electrodynamics, we examine the tests of EEP via the metric-induced spacetime constitutive tensor density. The non-birefringence of the cosmic electromagnetic wave propagation in spacetime is observed to ultrahigh precision. This constrains the spacetime constitutive tensor density to Maxwell–Lorentz (metric) form plus a scalar (dilaton) degree of freedom and a pseudoscalar (axion) degree of freedom to high precision. The accurate agreement of cosmic microwave background spectrum with the Planck spectrum constrains the fractional change of the cosmic dilaton to be less than 8 × 10−4 . The Galileo weak equivalence principle (WEP) experiments (E¨ otv¨ os-type experiments) constrain the fractional dilatonic change in the solar system to be less than 10−10 . Accompanying the axion degree of freedom is the rotation of linear polarization in the cosmic propagation of electromagnetic waves called cosmic polarization rotation (CPR). Sperello di Serego Alighieri reviews the constraints from radio galaxy observations and CMB polarization observations to give a general constraint of 0.02 rad for the mean (uniform) CPR and also a constraint of 0.02 rad for the CPR ﬂuctuations. In many inﬂation models dilatons and axions play important roles; these investigations are crucial to give clues or constraints on the models. Frequency and time are the most precise metrological quantities. Their uses in gravity experiments are unavoidable. The use of GR in time synchronization and in GPS, GLONNESS, Galileo and Beidou becomes a folk talk. There are two good ways to compare precision clocks: (i) ﬁber ´ links; (ii) space optical links using laser ranging. Etienne Samain expounds the space optical link approach and addresses the laser ranging missions T2L2 (Time transfer by Laser link), LRO (Lunar Reconnaissance Orbiter) and LTT (Laser Time Transfer) together with future space mission proposals for fundamental physics, solar system science/navigation in which laser links are of prime importance. Solarsystem observation provides the original impetus and the ﬁrst conﬁrmation of GR. Chapter 8 summarizes the progress of classical solar system tests and explores its potential in the future. Improvement of three or more orders of magnitude is still possible. Perhaps the most dramatic development in testing relativistic gravity and in improving the dynamical foundations of general relativity is the discovery and observation of pulsars, binary pulsars, millisecond pulsars and double pulsars since 1967, 1974, 1982 and 2003 respectively. Richard Manchester reviews the pulsar observation in its relation with gravity in Chapter 9 with a brief introduction to basic pulsar properties and pulsar timing. He presents a rather thorough account of dynamical tests of GR and the strong equivalence principle together with a lucid but in-depth account of GW detection using pulsar timing arrays (PTAs). See front cover for an illustrative schematic of a PTA.

Foreword

vii

In 1916 Einstein predicted gravitational waves (GWs) in GR almost immediately after his founding of it. The existence of gravitational waves is the direct consequence of general relativity and unavoidable consequences of all relativistic gravity theories with ﬁnite velocity of propagation. Their importance in GR is like that electromagnetic waves in Maxwell–Lorentz theory of electromagnetism. Einstein’s general relativity and relativistic gravity theories predict the existence of gravitational waves. Gravitational waves propagate in spacetime forming ripples of spacetime geometry. In the introductory chapter of Part III, Kazuaki Kuroda, Wei-Ping Pan and I review and summarize the complete GW spectrum, the methods of detection, and the detection sensitivities in various frequency bands with a brief introduction to GW sources. At the time Einstein predicted GWs in GR, he estimated that GWs were experimentally not detectable due to feeble strengths. However, thanks to one hundred years of development of experimental methods and technology together with the discovery of various astrophysical compact objects and cosmological sources, GWs are now on the verge of detection in three frequency bands. The very low frequency band (10 fHz–300 pHz) GWs are on the verge of detection by the PTAs; Richard Manchester covers this part in his chapter on pulsars and gravity in Part II. As mentioned in Chapter 10, the observation of PTAs has already constrained the isotropic GW background to a level excluding most current models of supermassive black hole formation. This is a strong signal that PTA observation is on the verge of detecting GWs. The high frequency band (10 Hz–100 kHz) GWs are on the verge of detection by ground-based interferometers; Kazuaki Kuroda addresses the detection methods and the sources in the second chapter of Part III. The extremely low (Hubble) frequency band (1 aHz–10 fHz) GWs may also be on the verge of detection by CMB polarization observations; the present status is brieﬂy reviewed in the introduction chapter of Part III. The low frequency band (100 nHz–100 mHz) and the middle frequency band detections will have the greatest S/N ratios according to the present expectation. We review the sources, goal sensitivities, various mission proposals together with the current supporting activities in the third chapter of Part III. The GW quadrupole radiation formula has already been veriﬁed by the binary pulsar observations. In the next hundred years we will see great discoveries and immense focused activities toward the establishment and ﬂourish of GW astronomy and GW cosmology. GW physics and GW astronomy will become a precision discipline in the coming century. The development of cosmology is most dramatic during the last hundred years. From Kapteyn universe in 1915 of observed disk star system of 10 kpc diameter and 2 kpc thickness with the Sun near its center to full-ﬂedged precision cosmology now is monumental in the human history. It is fortunate that the development of observational cosmology has GR theory as a theoretical basis and goes hand-in-hand with the development of general relativity. This is fortunate both for observational cosmology and for GR. Using the Cosmological Principle Einstein looked into cosmological solutions in GR in 1917. The fast development of observational distance ladder around that time soon extends the reach of astronomy to modern cosmos.

viii

Foreword

Studies in the fundamental issues on the origins of cosmos lead to anthropological principle, cosmic inﬂation, and cosmic landscape scenarios. The cosmos is believed to be open in (extended beyond) the Hubble distance scale. Part III consists of seven chapters: Martin Bucher and I present some introductory remarks with a discussion of missing mass-deﬁcient acceleration issue in the ﬁrst chapter; Marc Davis reviews the observation and evolution of cosmic structure; Martin Bucher give a rather comprehensive exposition of the physics (almost on every aspect of cosmology) of CMB; Xiangcun Meng, Yan Gao and Zhanwen Han review the SNIa as a standardizable distance candle, its nature, its progenitors and its role in the cosmology together with related current issues; Toshifumi Futamase on the gravitational lensing in cosmology; K. Sato and Juni’ichi on cosmic inﬂation with a brief historical exposition on the development in Japan and Russia; David Chernoﬀ and Henry Tye on inﬂation and cosmic strings from the point of view of string theory. The quest for a satisfactory quantum description of gravity began very early. Einstein thought that quantum eﬀects must modify general relativity in his ﬁrst paper on GWs in 1916. Klein argued that the quantum theory must ultimately modify the role of spatiotemporal concepts in fundamental physics in 1927. Part V on Quantum Gravity consists of 4 chapters. Chapter 20 gives a bird’s-eye survey on the development of fundamental ideas of quantum gravity together with possible observations of quantum gravitational eﬀects in the foreseeable future. The classical age (1958–1969; according to the chronological classiﬁcation of Rovelli) started with ADM canonical formalism and concluded with DeWitt–Wheeler equation and DeWitt’s derivation of Feynman rules for perturbative GR. In the middle ages (1970–1983), the discovery of black hole thermodynamics and Hawking’s derivation of black hole radiation radically aﬀected our understanding of general relativity. In the renaissance period (1984–1994), there are two inﬂuential developments. From the covariant approach, attempts to get rid of inﬁnities merge into string theory. The use of strings and branes extends the theoretical framework of quantum ﬁeld theory. From the canonical approach, background-independent loop quantum gravity emerged 20 years after DeWitt–Wheeler equation. In Chapter 21, Richard Woodard starts with experiences of two personal academic careers through the classical and middle ages, advocates that the cosmological data from the epoch of primordial inﬂation is catalyzing the maturation of quantum gravity from speculation into a hard science, explains why quantum gravitational eﬀects from primordial inﬂation are observable, reviews what has been done in perturbative quantum gravity, tells us what the future holds both theoretically and observationally, and discusses what this tells us about quantum gravity. In Chapter 22, Steven Carlip reviews the discovery of black hole thermodynamics and summarizes the many independent ways of obtaining the thermodynamic and statistical mechanical properties of black holes. This has oﬀered us some early hints about the nature of quantum gravity. Steven then describes some of the remaining puzzles, including the nature of the quantum microstates, the problem of universality, and the information loss paradox. In the last chapter, Dah-Wei Chiou gives us a rather self-contained introductory review

Foreword

ix

on loop quantum gravity — a background-independent nonperturbative approach to a consistent quantum theory of gravity placing emphasis on the fundamental ideas and their signiﬁcance. The review presents the canonical formulation of loop quantum gravity as the central topic and covers brieﬂy the spin foam theory, the relation to black hole thermodynamics and the loop quantum cosmology with current directions and open issues summarized. Although we do not yet have a consistent calculable quantum gravity theory which has a good degree of completeness like quantum electrodynamics or quantum chromodynamics, the eﬀorts to ﬁnd one already led to the consistent renormalization of the gauge theory in 1960’s. The new development since 1980’s together with more understanding and further development of perturbation theory may give clues to a consistent theory. During these endeavors, the quest for a well-developed quantum gravity phenomenology including the quest to ﬁnd a correct inﬂationary (or non-inﬂationary) scenario may play a signiﬁcant role. The hope is that we will have one within a generation. This book is written and assembled for graduate students and general scientiﬁc-oriented readers alike. Each chapter is basically a review article. The ﬁve Parts are interconnected. Diﬀerent combinations can be designed for special topics for graduate students and advanced undergraduates. For example, following combinations are suitable for each topic named: (i) Basics (Selected Topics in GR): Part I, Chapters 8, 9, 10, 13, 20; (ii) Empirical Foundations (Empirical Foundations of Relativistic Gravity): Chapter 2, Part II, Chapters 10, 11, 13, 14, 16, 20; (iii) Gravitational Waves: Chapters 2, 9, Part III, Chapters 15, 18, 19; (iv) Cosmology: Chapters 5, 6, 10, 12, Part IV, Chapter 20; (v) Quantum Gravity: Chapters 3, 4, 10, 18, 19, Part V. There can be various other combinations too. We are grateful to all contributors for agreeing to write comprehensive reviews to make this publication possible. We would also like to thank all the referees for their valuable comments and suggestions: Martin Bucher, Stephen Carlip, DahWei Chiou, Sperello di Serego Alighieri, Angela Di Virgilio, John Eldridge, Jeremy Gray, Friedrich Hehl, Jim Hough, Ekaterina Koptelova, Ettore Majorana, James Nester, Ulrich Schreiber, Alexei Starobinsky, David Tanner, Richard Woodard, AnMing Wu, Masahide Yamaguchi. We thank the World Scientiﬁc staﬀ, especially Dr. K. K. Phua and Kah Fee Ng for their generous support in completing the book. We dedicate this two-volume GR centennial book to the founders of GR and various communities who have contributed to this dramatic century of development and applications of GR. Wei-Tou Ni November, 2015

x

Foreword

Note Added in Proof After the foreword was written, LIGO Scientiﬁc and Virgo Collaborations announced in February 2016 and in June 2016 the ﬁrst direct detections of gravitational waves (GWs) by LIGO Hanford and LIGO Livingston detectors in September 2015 and in December 2015. With the LIGO discovery announcements, two important things are veriﬁed: (i) GWs are directly detected in the solar-system; (ii) Black holes (BHs), binary BHs and BH coalescences are discovered and measured experimentally and directly with the distances reached more than 1 billion light years. These discoveries constitute the best celebration of the centennial of the genesis of general relativity. We refer the readers to Refs. 1 and 2 for the discovery and Refs. 3 and 4 for a brief history of gravitational wave research. A web page will be set up for updates of the reviews of these two volumes. Please see http://astrod.wikispaces.com/ for announcement. References 1. B. P. Abbott et al. (LIGO Scientiﬁc and Virgo Collaborations), Observation of gravitational waves from a binary black hole merger, Phys. Rev. Lett. 116 (2016) 061102. 2. B. P. Abbott et al. (LIGO Scientiﬁc and Virgo Collaborations), GW151226: Observation of gravitational waves from a 22-solar-mass binary black hole coalescence, Phys. Rev. Lett. 116 (2016) 241103. 3. J. L. Cervantes-Cota, S. Galindo-Uribarri and G.F. Smoot, A brief history of gravitational waves, Universe 2(22) (2016) 09400. 4. C.-M. Chen, J. M. Nester, W.-T. Ni, A brief history of gravitational wave research, Chinese Journal of Physics 55 (2017) 142–169.

Contents

Volume 1 Foreword

v

Color plates

I-CP1

Part I. Genesis, Solutions and Energy 1. A genesis of special relativity Val´erie Messager and Christophe Letellier IJMPD 24 (2015) 1530024 1. Introduction 2. The Ether: From Celestial Body Motion to Light Propagation 2.1. Its origin 2.2. The luminiferous ether 3. Galileo’s Composition Law for Velocities 4. Questioning the Nature of Light: Waves or Corpuscles? 5. From Electrodynamics to Light 5.1. Amp`ere’s law 5.2. Maxwell’s electromagnetic waves as light 5.3. Helmholtz’s theory 5.4. Hertzs experiments for validating Maxwell’s theory 6. Invariance of the Field Equations from a Frame to Another One 6.1. Hertz’s electrodynamic theory 6.2. Voigt’s wave equation 6.3. Lorentz’s electrodynamical theory 6.4. Larmor’s theory 7. Poincar´e’s Contribution 8. Einstein’s 1905 Contribution 9. Conclusion Appendices A. 1. Fizeau’s experiments xi

I-1 I-3

I-3 I-5 I-5 I-8 I-11 I-15 I-24 I-24 I-28 I-32 I-33 I-37 I-37 I-41 I-42 I-50 I-51 I-72 I-76 I-77 I-77

xii

Contents

A. 2. Michelson and Morley’s experiments 2. Genesis of general relativity — A concise exposition Wei-Tou Ni IJMPD 25 (2016) 1630004

I-77 I-85

1. Prelude — Before 1905

I-86

2. The Period of Searching for Directions and New Ingredients: 1905–1910

I-91

3. The Period of Various Trial Theories: 1911–1914 4. The Synthesis and Consolidation: 1915–1916

I-96 I-100

5. Epilogue

I-103

3. Schwarzschild and Kerr solutions of Einstein’s ﬁeld equation: An Introduction Christian Heinicke and Friederich W. Hehl IJMPD 24 (2015) 1530006 1. Prelude 1.1. Newtonian gravity 1.2. Minkowski space

I-109

I-109 I-109 I-114

1.2.1. Null coordinates

I-115

1.2.2. Penrose diagram 1.3. Einstein’s ﬁeld equation

I-115 I-118

2. The Schwarzschild Metric (1916)

I-120

2.1. Historical remarks 2.2. Approaching the Schwarzschild metric

I-120 I-122

2.3. Six classical representations of the Schwarzschild metric

I-126

2.4. The concept of a Schwarzschild black hole 2.4.1. Event horizon

I-126 I-128

2.4.2. Killing horizon

I-130

2.4.3. Surface gravity 2.4.4. Inﬁnite redshift

I-131 I-131

2.5. Using light rays as coordinate lines 2.5.1. Eddington–Finkelstein coordinates 2.5.2. Kruskal–Szekeres coordinates 2.6. Penrose–Kruskal diagram 2.7. Adding electric charge and the cosmological constant: Reissner–Nordstr¨ om 2.8. The interior Schwarzschild solution and the TOV equation

I-131 I-132 I-133 I-135 I-136 I-137

Contents

xiii

3. The Kerr Metric (1963)

I-141

3.1. Historical remarks

I-141

3.2. Approaching the Kerr metric

I-144

3.2.1. Papapetrou line element and vacuum ﬁeld equation

I-144

3.2.2. Ernst equation (1968)

I-147

3.2.3. From Ernst back to Kerr

I-148

3.3. Three classical representations of the Kerr metric

I-149

3.4. The concept of a Kerr black hole

I-151

3.4.1. Depicting Kerr geometry 3.5. The ergoregion

I-152 I-155

3.5.1. Constrained rotation

I-155

3.5.2. Rotation of the event horizon

I-156

3.5.2. Penrose process and black hole thermodynamics

I-156

3.6. Beyond the horizons 3.6.1. Using light rays as coordinate lines

I-157 I-158

3.7. Penrose–Carter diagram and Cauchy horizon

I-160

3.8. Gravitoelectromagnetism, multipole moments

I-161

3.8.1. Gravitoelectromagnetic ﬁeld strength

I-163

3.8.2. Quadratic invariants

I-165

3.8.3. Gravitomagnetic clock eﬀect of Mashhoon, Cohen et al.

I-166

3.8.4. Multipole moments: Gravitoelectric and gravitomagnetic ones

I-167

3.9. Adding electric charge and the cosmological constant: Kerr–Newman metric

I-168

3.10. On the uniqueness of the Kerr black hole

I-170

3.11. On interior solutions with material sources

I-171

4. Kerr Beyond Einstein

I-172

4.1. Kerr metric accompanied by a propagating linear connection

I-172

4.2. Kerr metric in higher dimensions and in string theory

I-174

Appendix

I-175

A.1. Exterior calculus and computer algebra

I-175

xiv

Contents

4. Gravitational energy for GR and Poincar´e gauge theories: A covariant Hamiltonian approach Chiang-Mei Chen, James Nester and Roh-Suan Tung IJMPD 24 (2015) 1530026

I-187

1. Introduction

I-188

2. Background

I-189

2.1. Some brief early history

I-189

2.2. From Einstein’s correspondence

I-190

2.3. Noether’s contribution

I-192

2.4. Noether’s result

I-193

3. The Noether Energy–Momentum Current Ambiguity 4. Pseudotensors

I-194 I-196

4.1. Einstein, Klein and superpotentials

I-197

4.2. Other GR pseudotensors

I-198

4.3. Pseudotensors and the Hamiltonian

I-200

5. The Quasi-Local View

I-201

6. Currents as Generators

I-201

7. Gauge and Geometry

I-202

8. Dynamical Spacetime Geometry and the Hamiltonian 8.1. Pseudotensors and the Hamiltonian

I-203 I-204

8.2. Some comments 9. Diﬀerential Forms 10. Variational Principle for Form Fields

I-204 I-204 I-206

10.1. Hamiltons principle

I-207

10.2. Compact representation

I-207

11. Some Simple Examples of the Noether Theorems

I-208

11.1. Noether’s ﬁrst theorem: Energy–momentum 11.2. Noether’s second theorem: Gauge ﬁelds

I-208 I-209

11.3. Field equations with local gauge theory

I-211

12. First-Order Formulation 13. The Hamiltonian and the 3 + 1 Spacetime Split

I-213 I-214

13.1. Canonical Hamiltonian formalism

I-215

13.2. The diﬀerential form of the spacetime decomposition

I-215

13.3. Spacetime decomposition of the variational formalism

I-217

Contents

14. The Hamiltonian and Its Boundary Term

xv

I-218

14.1. The translational Noether current

I-219

14.2. The Hamiltonian formulation

I-220

14.3. Boundary terms: The boundary condition and reference

I-221

14.4. Covariant-symplectic Hamiltonian boundary terms

I-222

15. Standard Asymptotics

I-223

15.1. Spatial inﬁnity

I-224

15.2. Null inﬁnity

I-224

15.3. Energy ﬂux

I-225

16. Application to Electromagnetism

I-225

17. Geometry: Covariant Diﬀerential Formulation

I-227

17.1. Metric and connection

I-228

17.2. Riemann–Cartan geometry

I-229

17.3. Regarding geometry and gauge

I-230

17.4. On the aﬃne connection and gauge theory

I-230

18. Variational Principles for Dynamic Spacetime Geometry

I-232

18.1. The Lagrangian and its variation

I-232

18.2. Local gauge symmetries, Noether currents and diﬀerential identities

I-233

18.3. Interpretation of the diﬀerential identities

I-238

19. First-Order Form and the Hamiltonian

I-240

19.1. First-order Lagrangian and local gauge symmetries

I-240

19.2. Generalized Hamiltonian and diﬀerential identities

I-241

19.3. General geometric Hamiltonian boundary terms

I-244

19.4. Quasi-local boundary terms

I-245

19.5. A preferred choice

I-245

19.6. Einstein’s GR

I-246

19.7. Preferred boundary term for GR

I-247

20. A “Best Matched” Reference

I-248

20.1. The choice of reference

I-249

20.2. Isometric matching of the 2-surface

I-250

20.3. Complete 4D isometric matching

I-251

xvi

Contents

20.4. Complete 4D isometric matching 21. Concluding Discussion Part II. Empirical Foundations 5. Equivalence principles, spacetime structure and the cosmic connection Wei-Tou Ni IJMPD 25 (2016) 1630002

I-251 I-252 I-263 I-265

1. Introduction

I-265

2. Meaning of Various Equivalence Principles

I-270

2.1. Ancient concepts of inequivalence

I-271

2.2. Macroscopic equivalence principles

I-271

2.3. Equivalence principles for photons (wave packets of light)

I-273

2.4. Microscopic equivalence principles

I-273

2.5. Equivalence principles including gravity (Strong equivalence principles)

I-276

2.6. Inequivalence and interrelations of various equivalence principles

I-277

3. Gravitational Coupling to Electromagnetism and the Structure of Spacetime

I-278

3.1. Premetric electrodynamics as a framework to study gravitational coupling to electromagnetism

I-278

3.2. Wave propagation and the dispersion relation

I-279

3.2.1. The condition of vanishing of B(1) and B(2) for all directions of wave propagation

I-282

3.2.2. The condition of (Sk) B(1) = (P)B(1) = 0 and A(1) = A(2) for all directions of wave propagation

I-284

3.3. Nonbirefringence condition for the skewonless case

I-284

3.4. Wave propagation and the dispersion relation in dilaton ﬁeld and axion ﬁeld

I-288

3.5. No ampliﬁcation/no attenuation and no polarization rotation constraints on cosmic dilaton ﬁeld and cosmic axion ﬁeld

I-292

3.6. Spacetime constitutive relation including skewons

I-293

Contents

xvii

3.7. Constitutive tensor from asymmetric metric and Fresnel equation

I-297

3.8. Empirical foundation of the closure relation for skewonless case

I-300

4. From Galileo Equivalence Principle to Einstein Equivalence Principle

I-303

5. EEP and Universal Metrology

I-305

6. Gyrogravitational Ratio

I-307

7. An Update of Search for Long Range/Intermediate Range Spin–Spin, Spin–Monopole and Spin–Cosmos Interactions

I-308

8. Prospects

I-309

6. Cosmic polarization rotation: An astrophysical test of fundamental physics Sperello di Serego Alighieri IJMPD 24 (2015) 1530016

I-317

1. Introduction

I-317

2. Impact of CPR on Fundamental Physics

I-318

3. Constraints from the Radio Polarization of RGs

I-319

4. Constraints from the UV Polarization of RGs

I-320

5. Constraints from the Polarization of the CMB Radiation

I-321

6. Other Constraints

I-325

7. Discussion

I-326

8. Outlook

I-327

7. Clock comparison based on laser ranging technologies ´ Etienne Samain IJMPD 24 (2015) 1530021

I-331

1. Introduction

I-331

2. Scientiﬁc Objectives

I-335

2.1. Time and frequency metrology

I-335

2.2. Fundamental physics

I-338

2.3. Solar System science

I-340

2.4. Solar System navigation based on clock comparison

I-341

3. Time Transfer by Laser Link: T2L2 on Jason-2

I-341

3.1. Principle

I-341

3.2. Laser station ground segment

I-342

xviii

Contents

3.3. Space instrument

I-344

3.4. Time equation

I-347

3.5. Error budget

I-349

3.6. Link budget

I-351

3.7. Exploitation

I-352

4. One-Way Lunar Laser Link on LRO Spacecraft

I-357

5. Prospective

I-361

6. Conclusion and Outlook

I-364

8. Solar-system tests of relativistic gravity Wei-Tou Ni

I-371

IJMPD 25 (2016) 1630003 1. Introduction and Summary

I-371

2. Post-Newtonian Approximation, PPN Framework, Shapiro Time Delay and Light Deﬂection

I-374

2.1. Post-Newtonian approximation

I-375

2.2. PPN framework

I-377

2.3. Shapiro time delay

I-380

2.4. Light deﬂection

I-381

3. Solar System Ephemerides

I-382

4. Solar System Tests

I-385

5. Outlook — On Going and Next-Generation Tests 9. Pulsars and gravity R. N. Manchester

I-393 I-407

IJMPD 24 (2015) 1530018 1. Introduction 1.1. Pulsar timing 2. Tests of Relativistic Gravity 2.1. Tests of general relativity with double-neutron-star systems 2.1.1. The Hulse–Taylor binary, PSR B1913 + 16 2.1.2. PSR B1534 + 12 2.1.3. The double pulsar, PSR J0737 − 3039A/B 2.1.4. Measured post-Keplerian parameters 2.2. Tests of equivalence principles and alternative theories of gravitation 2.2.1. Limits on PPN parameters

I-407 I-410 I-412 I-412 I-412 I-415 I-417 I-421 I-421 I-423

Contents

2.2.2. Dipolar gravitational waves and the constancy of G 2.2.3. General scalar–tensor and scalar–vector–tensor theories 2.3. Future prospects 3. The Quest for Gravitational-Wave Detection 3.1. Pulsar timing arrays 3.2. Nanohertz gravitational-wave sources 3.2.1. Massive black-hole binary systems

xix

I-427 I-429 I-431 I-432 I-432 I-435 I-435

3.2.2. Cosmic strings and the early universe

I-439

3.2.3. Transient or burst GW sources

I-440

3.3. Pulsar timing arrays and current results 3.3.1. Existing PTAs 3.3.2. Limits on the nanohertz GW background 3.3.3. Limits on GW emission from individual black-hole binary systems 3.4. Future prospects 4. Summary and Conclusion Part III. Gravitational Waves 10. Gravitational waves: Classiﬁcation, methods of detection, sensitivities, and sources Kazuaki Kuroda, Wei-Tou Ni and Wei-Ping Pan IJMPD 24 (2015) 1530031

I-443 I-444 I-445 I-446 I-450 I-452 I-459 I-461

1. Introduction and Classiﬁcation

I-461

2. GWs in GR

I-464

3. Methods of GW Detection, and Their Sensitivities 3.1. Sensitivities

I-470 I-471

3.2. Very high frequency band (100 kHz–1 THz) and ultrahigh frequency band (above 1 THz)

I-477

3.3. High frequency band (10 Hz–100 kHz)

I-478

3.4. Doppler tracking of spacecraft (1 μHz–1 mHz in the low-frequency band)

I-480

3.5. Space interferometers (low-frequency band, 100 nHz–100 mHz; middle-frequency band, 100 mHz–10 Hz)

I-481

3.6. Very-low-frequency band (300 pHz–100 nHz)

I-486

3.7. Ultra-low-frequency band (10 fHz–300 pHz)

I-488

xx

Contents

3.8. Extremely-low (Hubble)-frequency band (1 aHz–10 fHz) 4. Sources of GWs

I-489 I-491

4.1. GWs from compact binaries

I-491

4.2. GWs from supernovae

I-492

4.3. GWs from massive black holes and their coevolution with galaxies

I-493

4.4. GWs from extreme mass ratio inspirals (EMRIs)

I-495

4.5. Primordial/inﬂationary/relic GWs

I-495

4.6. Very-high-frequency and ultra-high-frequency GW sources

I-496

4.7. Other possible sources

I-496

5. Discussion and Outlook 11. Ground-based gravitational-wave detectors Kazuaki Kuroda IJMPD 24 (2015) 1530032 1. Introduction to Ground-Based Gravitational-Wave Detectors 1.1. Gravitational-wave sources

I-497 I-505

I-505 I-506

1.1.1. Achieved sensitivities of large projects

I-506

1.1.2. Coalescences of binary neutron stars

I-508

1.1.3. Coalescences of binary black holes

I-508

1.1.4. Supernova explosion

I-509

1.1.5. Quasi-normal mode oscillation at the birth of black hole

I-509

1.1.6. Unstable fast rotating neutron star

I-510

1.2. Acceleration due to a gravitational wave

I-510

1.3. Response of a resonant antenna

I-512

1.4. Response of a resonant antenna

I-515

1.4.1. Directivity

I-516

1.4.1. Positioning

I-518

1.5. Comparison of a resonant antenna and an interferometer 2. Resonant Antennae

I-519 I-519

2.1. Development of resonant antennae

I-520

2.2. Dynamical model of a resonant antenna with two modes

I-523

2.3. Signal-to-noise ratio and noise temperature

I-525

Contents

2.4. Comparison of ﬁve resonant antennae 3. Interferometers 3.1. First stage against technical noises in prototype interferometers

xxi

I-526 I-527 I-528

3.1.1. 3 m-Garching interferometer 3.1.2. 30 m-Garching interferometer

I-528 I-530

3.1.3. Glasgow 10 m-Fabry–Perot Michelson interferometer

I-533

3.1.4. Caltech 40 m-Fabry–Perot Michelson interferometer 3.1.5. ISAS 10 m and 100 m delay-line interferometer

I-535 I-536

3.2. Further R&D eﬀorts in the ﬁrst-generation detectors 3.2.1. Power recycling

I-536 I-537

3.2.2. Signal recycling and resonant side-band extraction 3.3. Fighting with thermal noise of the second stage

I-538 I-539

3.3.1. Mirror and suspension thermal noise 3.3.2. Thermal noise of optical coating 3.4. Fighting against quantum noises and squeezing 3.4.1. Radiation pressure noise 3.4.2. Squeezing 4. Large Scale Projects

I-540 I-542 I-543 I-543 I-544 I-546

4.1. LIGO project 4.2. Virgo project

I-546 I-548

4.3. GEO project 4.4. TAMA/CLIO/LCGT(KAGRA) project

I-552 I-555

4.4.1. TAMA

I-555

4.4.2. CLIO 4.4.3. LCGT (KAGRA)

I-558 I-561

4.4.4. Einstein telescope 5. Summary

I-565 I-566

Appendix A. Thermal Noise

I-567

A.1. Nyquist theorem A.2. Thermal noise of a harmonic oscillator

I-567 I-568

Appendix B. Modulation Appendix C. Fabry–Perot Interferometer

I-569 I-571

xxii

Contents

C.1. Fabry–Perot cavity

I-571

C.2. Frequency response of a Fabry–Perot Michelson interferometer Appendix D. Newtonian Noise

I-572 I-573

12. Gravitational wave detection in space Wei-Tou Ni IJMPD 25 (2016) 1630001

I-579

1. Introduction

I-579

2. Gravity and Orbit Observations/Experiments in the Solar System 3. Doppler Tracking of Spacecraft

I-586 I-589

4. Interferometric Space Missions

I-591

5. Frequency Sensitivity Spectrum 6. Scientiﬁc Goals

I-596 I-601

6.1. Massive black holes and their co-evolution with galaxies

I-601

6.2. Extreme mass ratio inspirals 6.3. Testing relativistic gravity

I-603 I-603

6.4. Dark energy and cosmology

I-603

6.5. Compact binaries 6.6. Relic GWs

I-604 I-604

7. Basic Orbit Conﬁguration, Angular Resolution and Multi-Formation Conﬁgurations 7.1. Basic LISA-like orbit conﬁguration

I-605 I-605

7.2. Basic ASTROD orbit conﬁguration

I-607

7.3. Angular resolution 7.4. Six/twelve spacecraft formation

I-611 I-612

8. Orbit Design and Orbit Optimization Using Ephemerides

I-612

8.1. CGC ephemeris 8.2. Numerical orbit design and orbit optimization for eLISA/NGO

I-613

8.3. Orbit optimization for ASTROD-GW 8.3.1. CGC 2.7.1 ephemeris

I-616 I-616

8.3.2. Initial choice of spacecraft initial conditions 8.3.3. Method of optimization 9. Deployment of Formation in Earthlike Solar Orbit 10. Time Delay Interferometry

I-614

I-616 I-617 I-619 I-619

Contents

xxiii

11. Payload Concept

I-622

12. Outlook

I-624

Subject Index

I

Author Index

XIII

Volume 2 Foreword

v

Color plates

II-CP1

Part IV. Cosmology 13. General Relativity and Cosmology Martin Bucher and Wei-Tou Ni IJMPD 24 (2015) 1530030 14. Cosmic Structure Marc Davis

II-1 II-3

II-19

IJMPD 23 (2014) 1430021 1. History of Cosmic Discovery

II-19

2. Measurement of the Galaxy Correlation Function

II-22

2.1. Before 1980

II-22

2.2. After 1980

II-23

2.3. Remarkable large-scale structure in simulations

II-25

2.4. Measurement of the BAO eﬀect

II-26

2.5. Further measurements of the power spectrum

II-28

2.6. Lyman-α clouds

II-29

3. Large Scale Flows

II-31

4. Dwarf Galaxies as a Probe of Dark Matter

II-34

5. Gravitational Lensing

II-38

5.1. Double images

II-38

5.2. Bullet cluster

II-38

5.3. Substructure of gravitational lenses

II-38

6. Conclusion 15. Physics of the cosmic microwave background anisotropy Martin Bucher IJMPD 24 (2015) 1530004

II-40 II-43

1. Observing the Microwave Sky: A Short History and Observational Overview

II-43

2. Brief Thermal History of the Universe

II-54

xxiv

Contents

3. Cosmological Perturbation Theory: Describing a Nearly Perfect Universe Using General Relativity

II-58

4. Characterizing the Primordial Power Spectrum

II-61

5. Recombination, Blackbody Spectrum, and Spectral Distortions

II-62

6. Sachs–Wolfe Formula and More Exact Anisotropy Calculations

II-63

7. What Can We Learn From the CMB Temperature and Polarization Anisotropies?

II-69

7.1. Character of primordial perturbations: Adiabatic growing mode versus ﬁeld ordering

II-69

7.2. Boltzmann hierarchy evolution

II-71

7.3. Angular diameter distance

II-76

7.4. Integrated Sachs–Wolfe eﬀect

II-77

7.5. Reionization

II-78

7.6. What we have not mentioned

II-83

8. Gravitational Lensing of the CMB

II-84

9. CMB Statistics

II-86

9.1. Gaussianity, non-Gaussianity, and all that

II-86

9.2. Non-Gaussian alternatives

II-92

10. Bispectral Non-Gaussianity

II-92

11. B Modes: A New Probe of Inﬂation

II-94

11.1. Suborbital searches for primordial B modes

II-95

11.2. Space based searches for primordial B modes

II-96

12. CMB Anomalies 13. Sunyaev–Zeldovich Eﬀects 14. Experimental Aspects of CMB Observations

II-96 II-98 II-100

14.1. Intrinsic photon counting noise: Ideal detector behavior

II-102

14.2. CMB detector technology

II-104

14.3. Special techniques for polarization

II-106

15. CMB Statistics Revisited: Dealing with Realistic Observations

II-110

16. Galactic Synchrotron Emission

II-112

17. Free–Free Emission

II-113

18. Thermal Dust Emission

II-114

19. Dust Polarization and Grain Alignment

II-116

19.1. Why do dust grains spin?

II-117

Contents

19.2. 19.3. 19.4. 19.5.

About which axis do dust grains spin? A stochastic diﬀerential equation for L(t) Suprathermal rotation Dust grain dynamics and the galactic magnetic ﬁeld 19.5.1. Origin of a magnetic moment along L 19.6. Magnetic precession 19.6.1. Barnett dissipation 19.7. Davis–Greenstein magnetic dissipation 19.8. Alignment along B without Davis–Greenstein dissipation 19.9. Radiative torques 19.10. Small dust grains and anomalous microwave emission (AME) 20. Compact Sources 20.1. Radio galaxies 20.2. Infrared galaxies 21. Other Eﬀects 21.1. Patchy reionization 21.2. Molecular lines 21.3. Zodiacal emission 22. Extracting the Primordial CMB Anisotropies 23. Concluding Remarks 16. SNe Ia as a cosmological probe Xiangcun Meng, Yan Gao and Zhanwen Han IJMPD 24 (2015) 1530029 1. Introduction 2. SNe Ia as a Standardizable Distance Candle 3. Progenitors of SNe Ia 4. Eﬀect of SN Ia Populations on Their Brightness 5. SN Ia’s Role in Cosmology 6. Issues and Prospects 17. Gravitational Lensing in Cosmology Toshifumi Futamase IJMPD 24 (2015) 1530011 1. Introduction and History 2. Basic Properties for Lens Equation 2.1. Derivation of the cosmological lens equation

xxv

II-118 II-118 II-119 II-120 II-121 II-122 II-122 II-124 II-125 II-126 II-128 II-130 II-131 II-132 II-132 II-132 II-132 II-133 II-133 II-134 II-151

II-151 II-152 II-157 II-160 II-163 II-167 II-173

II-173 II-176 II-176

xxvi

Contents

2.2. Properties of lens mapping 2.3. Caustic and critical curves 2.3.1. Circular lenses 2.3.2. The Einstein radius and radial arcs 2.3.3. Non-circular lenses 3. Strong Lensing 3.1. Methods of solving the lens equation: LTM and non-LTM 3.2. Image magniﬁcation 3.3. Time delays 3.4. Comparison of lens model software 3.4.1. Non-light traces mass software 3.4.2. Light traces mass software 3.5. Lens statistics 4. Weak Lensing 4.1. Basic method 4.1.1. Shape measurements 4.2. E/B decomposition 4.3. Magniﬁcation bias 4.3.1. Simulation test 4.3.2. Higher-order weak lensing-ﬂexion and HOLICs 4.4. Cluster mass reconstruction 4.4.1. Density proﬁle 4.4.2. Dark matter subhalos in the coma cluster 4.5. Cosmic shear 4.5.1. How to measure the cosmic density ﬁeld 5. Conclusion and Future 18. Inﬂationary cosmology: First 30+ years Katsuhiko Sato and Jun’ichi Yokoyama IJMPD 24 (2015) 1530025 1. Introduction 1.1. Developments in Japan 1.2. Developments in Russia 1.3. Inﬂation paradigm 2. Resolution of Fundamental Problems 3. Realization of Inﬂation

II-179 II-183 II-184 II-187 II-189 II-190 II-190 II-191 II-191 II-194 II-194 II-194 II-195 II-196 II-197 II-199 II-203 II-206 II-206 II-207 II-208 II-211 II-212 II-214 II-217 II-219 II-225

II-225 II-227 II-228 II-230 II-231 II-233

Contents

xxvii

3.1. Three mechanisms 3.2. Inﬂation scenario 4. Slow-Roll Inﬂation Models 4.1. Large-ﬁeld models 4.2. Small-ﬁeld model 4.3. Hybrid inﬂation 5. Reheating 6. Generation of Quantum Fluctuations that Eventually Behave Classically 7. Cosmological Perturbation 8. Generation of Curvature Fluctuations in Inﬂationary Cosmology 9. Tensor Perturbation 10. The Most General Single-Field Inﬂation 10.1. Homogeneous background equations 10.2. Kinetically driven G-inﬂation 10.3. Potential-driven slow-roll G-inﬂation 11. Power Spectrum of Perturbations in Generalized G-inﬂation 11.1. Tensor perturbations 11.2. Scalar perturbations 12. Inﬂationary Cosmology and Observations 12.1. Large-ﬁeld models 12.2. Small-ﬁeld model 12.3. Hybrid inﬂation model 12.4. Noncanonical models and multi-ﬁeld models 13. Conclusion 19. Inﬂation, string theory and cosmic strings David F. Chernoﬀ and S.-H. Henry Tye IJMPD 24 (2015) 1530010

II-233 II-234 II-236 II-236 II-237 II-238 II-239

1. Introduction 2. The Inﬂationary Universe 3. String Theory and Inﬂation 3.1. String theory and ﬂux compactiﬁcation 3.2. Inﬂation in string theory 4. Small r Scenarios 4.1. Brane inﬂation ¯ 4.1.1. D3-D3-brane inﬂation

II-273 II-277 II-280 II-281 II-282 II-283 II-284 II-285

II-242 II-244 II-246 II-249 II-250 II-251 II-253 II-254 II-255 II-255 II-258 II-261 II-264 II-265 II-266 II-266 II-267 II-273

xxviii

Contents

4.1.2. Inﬂection point inﬂation 4.1.3. DBI model 4.1.4. D3-D7-brane inﬂation 4.2. K¨ ahler moduli inﬂation 5. Large r Scenarios 5.1. The Kim–Nilles–Peloso mechanism 5.1.1. Natural inﬂation 5.1.2. N-ﬂation 5.1.3. Helical inﬂation 5.2. Axion monodromy 5.3. Discussions 6. Relics: Low Tension Cosmic Strings 6.1. Strings in brane world cosmology 6.2. Current bounds on string tension Gμ and probability of intercommutation p 7. Scaling, Slowing, Clustering and Evaporating 7.1. Large-scale string distribution 7.2. Local string distribution 8. Detection 8.1. Detection via Microlensing 8.2. WFIRST microlensing rates 8.3. Gravitational waves 9. Summary Part V. Quantum Gravity 20. Quantum gravity: A brief history of ideas and some outlooks Steven Carlip, Dah-Wei Chiou, Wei-Tou Ni and Richard Woodard IJMPD 24 (2015) 1530028 1. Prelude 2. Perturbative Quantum Gravity 3. String Theory 4. Loop Quantum Gravity 5. Black Hole Thermodynamics 6. Quantum Gravity Phenomenology 21. Perturbative quantum gravity comes of age R. P. Woodard IJMPD 23 (2014) 1430020

II-286 II-286 II-287 II-287 II-288 II-288 II-288 II-288 II-290 II-291 II-292 II-293 II-296 II-297 II-299 II-302 II-305 II-307 II-307 II-307 II-311 II-314 II-323 II-325

II-325 II-327 II-328 II-332 II-334 II-337 II-349

Contents

1. Introduction 2. Why Quantum Gravitational Eﬀects from Primordial Inﬂation are Observable 2.1. The background geometry 2.2. Inﬂationary particle production 3. Tree Order Power Spectra 3.1. The background for single-scalar inﬂation 3.2. Gauge-ﬁxed, constrained action 3.3. Tree order power spectra 3.4. The controversy over adiabatic regularization 3.5. Why these are quantum gravitational eﬀects 4. Loop Corrections to the Power Spectra 4.1. How to make computations 4.2. -Suppression and late-time growth 4.3. Nonlinear extensions 4.4. The promise of 21 cm radiation 5. Other Quantum Gravitational Eﬀects 5.1. Linearized eﬀective ﬁeld equations 5.2. Propagators and tensor 1PI functions 5.3. Results and open problems 5.4. Back-Reaction 6. Conclusions 22. Black hole thermodynamics S. Carlip IJMPD 23 (2014) 1430023 1. Introduction 2. Prehistory: Black Hole Mechanics and Wheeler’s Cup of Tea 3. Hawking Radiation 3.1. Quantum ﬁeld theory in curved spacetime 3.2. Hawking’s calculation 4. Back-of-the-Envelope Estimates 4.1. Entropy 4.2. Temperature 5. The Many Derivations of Black Hole Thermodynamics 5.1. Other settings 5.2. Unruh radiation

xxix

II-349 II-351 II-351 II-355 II-358 II-359 II-360 II-363 II-369 II-369 II-371 II-372 II-376 II-380 II-382 II-384 II-384 II-386 II-395 II-399 II-402 II-415

II-415 II-416 II-418 II-419 II-420 II-422 II-422 II-423 II-424 II-425 II-425

xxx

Contents

5.3. 5.4. 5.5. 5.6. 5.7. 5.8. 5.9.

6.

7.

8. 9.

10.

Particle detectors Tunneling Hawking radiation from anomalies Periodic Greens functions Periodic Gravitational partition function Periodic Pair production of black holes Periodic Quantum ﬁeld theory and the eternal black hole 5.10. Periodic Quantized gravity and classical matter 5.11. Periodic Other approaches Thermodynamic Properties of Black Holes 6.1. Periodic Black hole evaporation 6.2. Periodic Heat capacity 6.3. Periodic Phase transitions 6.4. Periodic Thermodynamic volume 6.5. Periodic Lorentz violation and perpetual motion machines Approaches to Black Hole Statistical Mechanics 7.1. Periodic “Phenomenology” 7.2. Periodic Entanglement entropy 7.3. Periodic String theory 7.3.1. Weakly coupled strings and branes 7.3.2. Fuzzballs 7.3.3. The AdS/CFT correspondence 7.4. Loop quantum gravity 7.4.1. Microcanonical approach 7.4.2. Microcanonical approach 7.5. Other ensembles 7.6. Induced gravity 7.7. Logarithmic corrections The Holographic Conjecture The Problem of Universality 9.1. State-counting in conformal ﬁeld theory 9.2. Application to black holes 9.3. Eﬀective descriptions The Information Loss Problem 10.1. Nonunitary evolution

II-426 II-426 II-427 II-428 II-429 II-431 II-431 II-432 II-433 II-433 II-434 II-434 II-435 II-435 II-436 II-437 II-437 II-438 II-440 II-440 II-441 II-441 II-442 II-442 II-444 II-445 II-445 II-446 II-446 II-448 II-449 II-450 II-451 II-451 II-452

Contents

xxxi

10.2. No black holes

II-452

10.3. Remnants and baby universes

II-453

10.4. Hawking radiation as a pure state

II-454

11. Conclusion

II-455

Appendix A. Classical Black Holes

II-456

23. Loop quantum gravity Dah-Wei Chiou IJMPD 24 (2015) 1530005

II-467

1. Introduction

II-467

2. Motivations

II-469

2.1. Why quantum gravity?

II-469

2.2. Diﬃculties of quantum gravity

II-470

2.3. Background-independent approach

II-470

3. Connection Theories of General Relativity 3.1. Connection dynamics

II-471 II-471

3.2. Canonical (Hamiltonian) formulation

II-473

3.3. Remarks on connection theories

II-476

4. Quantum Kinematics 4.1. Quantization scheme

II-478 II-478

4.2. Cylindrical functions

II-479

4.3. Spin networks

II-481

4.4. S-knots

II-483

5. Operators and Quantum Geometry 5.1. Holonomy operator

II-486 II-486

5.2. Area operator

II-487

5.3. Volume operator

II-489

5.4. Quantum geometry

II-490

6. Scalar Constraint and Quantum Dynamics

II-492

6.1. Regulated classical scalar constraint

II-492

6.2. Quantum scalar constraint

II-495

6.3. Solutions to the scalar constraint

II-498

6.4. Quantum dynamics

II-500

7. Inclusion of Matter Fields

II-503

7.1. Yang–Mills ﬁelds

II-503

7.2. Fermions

II-504

7.3. Scalar ﬁelds

II-505

7.4. S-knots of geometry and matter

II-506

xxxii

Contents

8. Low-Energy Physics 8.1. Weave states 8.2. Loop states versus Fock states 8.3. Holomorphic coherent states 9. Spin Foam Theory 9.1. From s-knots to spin foams 9.2. Spin foam formalism 10. Black Hole Thermodynamics 10.1. Statistical ensemble 10.2. Bekenstein–Hawking entropy 10.3. More on black hole entropy 11. Loop Quantum Cosmology 11.1. Symmetry reduction 11.2. Quantum kinematics 11.3. Quantum constraint operator 11.4. Physical Hilbert space 11.5. Quantum dynamics 11.6. Other models 12. Current Directions and Open Issues 12.1. The master constraint program 12.2. Algebraic quantum gravity 12.3. Reduced phase space quantization 12.4. Oﬀ-shell closure of quantum constraints 12.5. Loop quantum gravity versus spin foam theory 12.6. Covariant loop quantum gravity 12.7. Spin foam cosmology 12.8. Quantum reduced loop gravity 12.9. Cosmological perturbations in the Planck era 12.10. Spherically symmetric loop gravity 12.11. Planck stars and black hole ﬁreworks 12.12. Information loss problem 12.13. Quantum gravity phenomenology 12.14. Supersymmetry and other dimensions Subject Index Author Index

II-507 II-507 II-508 II-508 II-511 II-511 II-514 II-515 II-516 II-517 II-519 II-520 II-520 II-522 II-524 II-526 II-527 II-528 II-529 II-529 II-530 II-530 II-532 II-533 II-533 II-534 II-534 II-534 II-535 II-535 II-536 II-537 II-537 I XIII

Chapter 13, Fig. 2; Chapter 14, Fig. 13.

Chapter 13, Fig. 3.

II-CP1

II-CP2

Chapter 13, Fig. 4.

Chapter 14, Fig. 1.

II-CP3

Chapter 14, Fig. 5.

Chapter 14, Fig. 6.

II-CP4

Chapter 14, Fig. 8.

Chapter 14, Fig. 9.

II-CP5

Chapter 15, Fig. 2.

Chapter 15, Fig. 3.

Chapter 15, Fig. 4.

II-CP6

Chapter 15, Fig. 6.

II-CP7

Chapter 15, Fig. 7.

II-CP8

Chapter 15, Fig. 8.

II-CP9

Chapter 15, Fig. 9.

Chapter 15, Fig. 17.

II-CP10

Chapter 16, Fig. 1.

Chapter 16, Fig. 5.

II-CP11

Chapter 16, Fig. 6.

Chapter 16, Fig. 7.

II-CP12

Chapter 17, Fig. 8.

Chapter 17, Fig. 9.

Chapter 17, Fig. 13.

II-CP13

Chapter 17, Fig. 14.

Chapter 17, Fig. 15.

II-CP14

Chapter 19, Fig. 1.

Chapter 19, Fig. 2.

Chapter 19, Fig. 3.

II-CP15

Chapter 19, Fig. 4.

G c2 10 Log10 n kpc

15

10

8

3

5

R kpc 5

10

15

5

10

Chapter 19, Fig. 6.

Chapter 19, Fig. 7.

20

25

II-CP16

Flux

Time Chapter 19, Fig. 8.

Log10 4

2

0

14

12

2

4

Chapter 19, Fig. 9.

10

8

Log10

May 4, 2017 9:55 b2167-ch13

Chapter 13 General relativity and cosmology

Martin Bucher†,‡ and Wei-Tou Ni§ †Laboratoire

APC, Universit´ e Paris 7/CNRS, Bˆ atiment Condorcet, Case 7020, 75205 Paris Cedex 13, France [email protected]

‡Astrophysics and Cosmology Research Unit and School of Mathematics, Statistics and Computer Science, University of KwaZulu-Natal, Durban 4041, South Africa §Center for Gravitation and Cosmology, Department of Physics, National Tsing Hua University, Hsinchu, Taiwan 30013, Republic of China

[email protected]

This year marks the 100th anniversary of Einstein’s 1915 landmark paper “Die Feldgleichungen der Gravitation” in which the ﬁeld equations of general relativity were correctly formulated for the ﬁrst time, thus rendering general relativity a complete theory. Over the subsequent hundred years, physicists and astronomers have struggled with uncovering the consequences and applications of these equations. This paper, which was written as an introduction to six chapters dealing with the connection between general relativity and cosmology that will appear in the two-volume book One Hundred Years of General Relativity: From Genesis and Empirical Foundations to Gravitational Waves, Cosmology and Quantum Gravity, endeavors to provide a historical overview of the connection between general relativity and cosmology, two areas whose development has been closely intertwined. Keywords: General relativity; cosmology; centennial connection. PACS Number(s): 04, 98.80.−k

One hundred years ago, the best model of the universe summarizing the state of the observations at that time was the Kapteyn universe, which consists of a system of stars distributed more or less uniformly within a disk about 10 kpc in diameter and 2 kpc thick. In this model, the Sun is situated near the center of the disk. In 1917, Einstein1 proposed a static cosmological model based on the “cosmological principle,” a generalization of the Copernican principle postulating that the homogeneity and isotropy of space in the large should be extended to include the time dimension as well. Using distance determinations to about 100 globular clusters, Shapley in 1918 pushed back the boundaries of the measured universe and concluded that the Sun lies near the edge of this distribution. As a II-3

May 4, 2017 9:55 b2167-ch13

II-4

M. Bucher and W.-T. Ni

result of rapid improvements in astronomical observations, the size of our observed universe soon grew to extend almost to our causal horizon. This next group of chapters comprising Part IV of GR100 book deals with Cosmology, applying Einstein’s theory of general relativity to the universe as a whole. Cosmology considers the universe on very large scales and evolving over very long times, comparable to the age of the universe. Cosmology today is often described as being a “precision” science and the insistence on this term reﬂects that cosmology had not always been seen as such. When Einstein formulated his general theory of relativity, little was known about the universe beyond our own galaxy, and although distant galaxies had been observed as “nebulous” unresolved spiral blotches, it was not at all clear that these “nebulae” consisted of numerous stars much like our own galaxy. A glimpse of the state of aﬀairs slightly after the formulation of general relativity may be gained by reading the written summaries2 of the “Great Debate” held at the Smithsonian institution in 1920, where Harlow Shapley and Heber Curtis sparred over the question of the “scale of the universe.” The former argued that the “nebulae” were simply clouds lying on the periphery of our own galaxy, whereas the latter maintained that the observations at the time suggested that the “nebulae” were distant “island universes” much like our own galaxy. A key observation that helped settle this question in favor of the latter point of view was the discovery by Edwin Hubble of Cepheid variable stars in the Andromeda galaxy, which established that Andromeda lies far beyond the conﬁnes of our own galaxy. On the observational side, as telescopes and other observational techniques greatly improved, our view of the universe progressively expanded to greater and greater distances. Once it had been established that the universe was populated by galaxies of diﬀerent sizes, a recurrent question became whether the universe was homogeneous and isotropic on the largest observable scales, and it is only recently that galaxy surveys suﬃciently deep and with suﬃcient statistics became available to settle this question deﬁnitively. An excellent historical account of these early debates and their role of the development of modern cosmology can be found in Peebles’ 1980 book3 (see also Ref. 4). The theory underlying the hot big bang model as we know it today was developed before the observational issues mentioned above had been settled. The geometry and time evolution of the universe as predicted by Einstein’s theory are given by what is now known as the Friedmann–Lemaˆıtre–Robertson–Walker (FLRW) model, which describes the solutions to Einstein’s ﬁeld equations for a spatially homogeneous and isotropic universe whose scale factor varies with time. This solution to the Einstein ﬁeld equations was ﬁrst put forward by Alexander Friedmann in 19225,6 and later independently by Georges Lamaˆıtre7 (see also Ref. 8). Robertson9 and Walker10 subsequently showed that this was the only solution to the ﬁeld equations consistent with spatial homogeneity and isotropy. As discussed above, in 1917 Einstein1 had put forth a theory of a static universe — a solution of the general relativity ﬁeld equations that is not only homogeneous and isotropic

May 4, 2017 9:55 b2167-ch13

General relativity and cosmology

II-5

in the three spatial dimensions but also homogeneous in time. Given the lack of compelling observational evidence to the contrary at the time, Einstein believed that an eternal universe, in which the Copernican principle held not only in three spatial dimensions but also in time, was more elegant and hence more plausible. In order to satisfy the gravitational ﬁeld equations, Einstein had to introduce a cosmological constant term, denoted by Λ, a proposal that according to George Gamow, Einstein had later once described as his “greatest blunder,” although the authenticity of this quote is doubted by some. Of the following six chapters of this volume dealing with the connection between cosmology and general relativity, the ﬁrst three chapters deal primarily with the observations underpinning our modern conception of the cosmos. The ﬁrst chapter of this group “Cosmic structure” provides a comprehensive overview of galaxy surveys and the information provided by them regarding the large-scale distribution of mass in the universe.11 The second chapter discusses the physics of the 2.725 K cosmic microwave background (CMB) radiation.12 The CMB conﬁrms the big bang story in two ways. First, the excellent agreement between the observed frequency spectrum of the microwave sky and a perfect blackbody spectrum at T = 2.725 K provides strong conﬁrmation of the expanding universe scenario, in which the universe at early times was much smaller and hotter, and thus once in a state extremely close to thermal equilibrium. Second, the measurement of the small departures from homogeneity and isotropy conﬁrms the gravitational instability hypothesis for the origin of structure. This chapter describes in detail how the precise mapping of the anisotropies of the blackbody temperature in both intensity and polarization can be used to test theories of the very early universe based on new physics far beyond the standard model and how these ﬂuctuations provide the initial conditions for the subsequent evolution leading to the formation of structure. The third chapter deals with the SN Ia distance scale and the discovery of the accelerated expansion of the universe.13 The fourth chapter “Gravitational lensing in cosmology” details the present state of gravitational lensing probes.14 The observation of the deﬂection of light rays as predicted by general relativity of course dates back to the famous expedition led by Arthur Eddington in 1919 to the island Princip´e oﬀ the west coast of Africa in order to measure the bending of light by the Sun during a total eclipse. Since that celebrated and crucial experiment conﬁrming the theory of general relativity, gravitational lensing has evolved into a powerful tool for mapping the inhomogeneity in the distribution of mass. As pointed out by Robert Dicke in the late 1960s,15,16 a universe that is governed by general relativity (through the FLRW solution) and ﬁlled with ordinary matter (i.e., a combination of nonrelativistic and relativistic particles with w = p/ρ ranging from 0 to 1/3) provides a model that is unsatisfactory in several respects including (i) the so-called “horizon problem,” whereby distant regions in opposite directions in the sky appear similar but could never have been in causal contact subsequent to the apparent initial singularity17,18 and (ii) the “ﬂatness

May 4, 2017 9:55 b2167-ch13

II-6

M. Bucher and W.-T. Ni

problem,” under which the relation between the expansion rate and mean density at very early times usually expressed in terms of the dimensionless density parameter Ω(t) = 8πGρmean (t)/3H 2 (t) initially had to be tuned near one with incredible precision to avoid a universe that today is either nearly empty or already re-collapsed.15,16 Although Dicke did not propose a concrete alternative, he suggested that the big bang model as formulated at the time might not be the whole story. “Cosmic inﬂation,” reviewed in the ﬁfth chapter,19 is a theory postulating that the very early universe underwent a period of quasi-exponential inﬂation during which whatever inhomogeneities may have existed prior to inﬂation are eﬀectively erased and replaced with quantum ﬂuctuations of the quantum ﬁelds relevant during inﬂation. The ﬁnal chapter “Inﬂation, string theory and cosmic strings”20 discusses possible connections between cosmic inﬂation and ideas from superstring theory concerning how inﬂation might be realized as part of a theory unifying all the fundamental interactions. Inﬂationary cosmology as developed in the early 1980s oﬀered an attractive proposal, solving many of the problems that one would face if one naively extrapolated a conventional matter-radiation equation of state backward in time all the way to the putative big bang singularity. Moreover, inﬂation oﬀers a predictive mechanism for how the primordial perturbations from a perfectly homogeneous and isotropic FLRW background solution are generated. But inﬂation is an incomplete theory whose predictions are not completely deﬁned without specifying a broader particle physics model within which inﬂation is realized. This ﬁnal chapter reviews work on how inﬂation might be realized within the framework of superstring theory. These chapters serve to summarize the role of general relativity in modern cosmology. Two key issues in modern cosmology are the dark matter problem and the dark energy problem. Since we have not included separate chapters for these issues, we give a brief review here. Both problems lead to postulating a new contribution to the stress–energy tensor because the known components combined with gravity as we understand it cannot account for the observations. In general, matter and objects in the universe have been discovered through nongravitational means, for example, through the electromagnetic radiation that they emit or scatter, but the history of objects ﬁrst discovered through their gravitational pull on other objects dates quite far back, to the discovery of new outer planets. Since Herschel’s discovery of the planet Uranus in 1781, its observed orbit persistently deviated from its predicted Newtonian trajectory according to the ephemeris calculations at the time. Hussey suggested in 1834 that this disagreement could be explained by perturbations arising from an undiscovered planet. In 1846, Le Verrier predicted the position of this new planet. On September 25, 1846, Galle and d’Arrest discovered this new planet, now known as Neptune, to within a degree of Le Verrier’s prediction. This discovery was a great triumph for Newton’s gravitational theory and was the ﬁrst example where deviations of an observed orbit from the predicted orbit led to the discovery of missing mass.21

May 4, 2017 9:55 b2167-ch13

General relativity and cosmology

II-7

When suﬃcient data from meridian and transit observations of Mercury had been accumulated, in 1859 Le Verrier discovered a discrepancy between the observations and Newtonian gravity. This discrepancy may be described as the anomalous perihelion advance of Mercury,a at the rate of 38" per century. Using improved calculations and data sets, Newcomb in 1882 measured this discrepancy more precisely, obtaining 42".95 per century. A more recent value is 42".98 ± 0.04 per century.22 In the last half of the 19th century, eﬀorts to account for the anomalous perihelion advance of Mercury explored two general directions: (i) Searching for a putative planet “Vulcan” or other matter inside Mercury’s orbit and (ii) postulating an ad hoc modiﬁed gravitational force law. Both these directions proved unsuccessful. Proposed modiﬁcations of the gravitational law included Clairaut’s force law (of the form A/r2 + B/r4 ), Hall’s hypothesis (that the gravitational attraction is proportional to the inverse of distance to the (2 + δ) power instead of the square) and velocity-dependent force laws. (The reader is referred Ref. 23 for an in depth history of the measurement and understanding of Mercury’s perihelion advance.) A compelling solution to this problem had to await the development of general relativity. When general relativity is taken as the correct theory for predicting corrections to Newton’s theory, we understand why when the observations reached an accuracy of the order of 1" per century (transit observations), a discrepancy would be seen. Over a century, Mercury orbits around the Sun 400 times, amounting to a total angle of 5 × 108 arcsec. The fractional relativistic correction (perihelion advance anomaly) of Mercury’s orbit is of order ξGM Sun /dc2 , (i.e. 8 × 10−8 ) with ξ = 3 and d being the distance of Mercury to the Sun. Therefore, the relativistic correction for perihelion advance is about 40 arcsec per century. As the orbit determination of Mercury reached an accuracy of order 10−8 , the relativistic corrections to Newtonian gravity became manifest. We thus see how gravitational anomalies can lead either to the discovery of missing matter or to a modiﬁcation of the fundamental theory for gravity. But there is also a third more mundane possibility. The secular contribution to the change of Moon’s orbit owing to the back reaction of lunar tides of Earth is such an example.b When such orbital anomalies were found, their solution involved neither missing mass nor modiﬁed Newtonian dynamics. Moreover, the discovery of Pluto might be an example of an accidental discovery resulting from the interplay between theory and experiment. Observations of Neptune in the late 19th century made astronomers suspect that there could be another planet besides Neptune perturbing Uranus’ orbit. In the early 20 century, Lowell searched in vain for such a planet at the Lowell observatory, which he founded in Flagstaﬀ. In 1931, Tombaugh a Here,

“anomalous” means after the much larger corrections from perturbations of other planets have been subtracted. Only this residual, or anomalous part, constitutes a sign of something new. b The action on the Earth is to slow down its rotation so that after some time leap seconds need to be inserted.

May 4, 2017 9:55 b2167-ch13

II-8

M. Bucher and W.-T. Ni

at Lowell observatory discovered Pluto 6◦ oﬀ its predicted position. However, its mass was much smaller than the predicted mass, so its inﬂuence on the orbits of Uranus and Neptune is negligible given the precision at that time. Pluto thus became the ninth major planet. However subsequently, as a result of a redeﬁnition by the IAU of what constitutes a major planet, Pluto was downgraded to become reclassiﬁed as a “trans-Neptunian” object. In 1992, the next trans-Neptunian object was discovered and since then more than 1200 such objects have been discovered with Eris (discovered in 2005) more massive than Pluto. We now turn to dark matter, which was ﬁrst introduced to account for the dynamics of clusters and the rotation curves of spiral galaxies, because the ordinary visible baryonic matter in the form of stars assuming plausible mass-to-light ratios, and also of the gas in the case of galaxy clusters, was unable to account for the observations if the correctness of ordinary Newtonian gravity is assumed. (In this context, corrections from special and general relativity are negligible.) The ﬁrst hints of the need to postulate a dark matter component date back to the 1930s, when Zwicky found that the virial mass of galaxy clusters (i.e., the mass deduced by applying the virial theorem to the observed internal velocities of their member galaxies) greatly exceeds the total mass inferred from the luminous matter based on plausible mass-to-light ratios.24 X-ray observations in the 1970s and 1980s alleviated this discrepancy without completely resolving it. The need to postulate an additional dark matter component of some sort also arose from studies of the dynamics of spiral galaxies. In a series of papers in the 1970s, Rubin, Ford, and Thonnard measured the rotation curves of a number of disk galaxies and found that rotation speeds were larger than would be expected from the gravitational attraction arising from the visible mass distribution.25−27 Moreover, the shape of the rotation curves inferred from the visible mass did not agree with the observations, which found a rotation velocity almost constant with varying radius. The authors interpreted their ﬁndings as evidence for a new dark matter component. Logically this conﬂict, known as the missing mass problem, could arise from a mass discrepancy, an acceleration discrepancy, or possibly even both. Many people believe that the missing mass is the dark matter, whose presence is made manifest only through its gravitational interaction with the visible baryonic matter. Others however explored the possibility that Newtonian gravitational dynamics should be modiﬁed. In 1982, Milgrom proposed the phenomenological modiﬁed Newtonian dynamics (MOND) law for small accelerations.28 Under this hypothesis, the gravitational dynamics become modiﬁed when the acceleration is smaller than a0 ∼ 10−10 ms−2 (Fig. 1). It was later shown possible to explain such a phenomenological Ansatz in the framework of a relativistic theory of gravity with additional degrees of freedom; however, none of these theories are yet satisfactory from a theoretical standpoint.31

May 4, 2017 9:55 b2167-ch13

General relativity and cosmology

II-9

Fig. 1. Critical acceleration from galactic rotation curves. The ratio of dynamical to baryonic mass is shown at each point along rotation curves as a function of the centripetal acceleration at that point. The panel shows data for 74 galaxies (see Ref. 29). The presence of missing or “dark” matter is the conventional explanation for why in spiral galaxies the measured rotation curves do not agree, neither in magnitude nor in shape, with the rotation curves predicted assuming a reasonable mass-to-light ratio and taking into account only the contribution from the disk. However, if the MOND hypothesis is adopted under which the gravitational dynamics becomes modiﬁed when the predicted acceleration is smaller than a0 (∼ 10−10 ms−2 ), the rotation curves can be ﬁt assuming a constant mass-to-light ratio and no halo dark matter component. (Note: Figure reprinted with permission from Ref. 30.)

Perhaps today one of the strongest arguments for the need for a dark matter component arises from CMB anisotropy measurements as explained in the chapter on the CMB.12 The standard six-parameter model, which includes a dark matter component comprising ∼ 24% of the critical mass, provides an adequate ﬁt to the observations and it is not possible to account for the observations with a model having only baryonic matter. Compared to other probes, the CMB is a particularly clean probe of cosmological models and parameters because its interpretation is based on linear theory except for small and calculable nonlinear corrections. Except for a few degeneracies (which can be lifted by combining with a few other reliable ancillary data sets, such as BAO), the errors on the cosmological parameters as determined using the CMB are typically of order 1% and characterized by a well understood error budget. See Refs. 32 and 33 for a detailed discussion of the current status of CMB constraints. Another dramatic though less quantitative recent illustration of the need for dark matter arises from analyzing the Bullet Cluster as observed by gravitational lensing (which provides a clean probe of the mass) and as observed in X-rays. The Bullet Cluster, shown in Fig. 2,34 is in fact the merger of two galaxy clusters. The X-ray image highlights the intracluster gas, which has been shocked and thus heated up as the result of the collision. This intracluster gas provides the dominant contribution to the baryonic mass, the mass from stars being subdominant. If there were no dark matter, the mass of the merged system would be concentrated in the center of the collision and would roughly coincide with the X-ray hot spot. But this is not what is observed in the gravitational lensing reconstruction of the projected mass density, which includes all types of mass whether visible or not. Instead, the

May 4, 2017 9:55 b2167-ch13

II-10

M. Bucher and W.-T. Ni

Fig. 2. Chandra X-ray image of the “Bullet Cluster” (red) superposed with lensing image (blue). In this merger seen in the X-ray, we see the hot baryonic intracluster gas that was shocked at the center of the collision. However in the gravitational lensing reconstruction we see two presumably dark matter halos comprising the majority of the mass which have been able to pass through each other without interaction. (Credit: NASA, see Ref. 34 for a detailed explanation.) (For color version, see page II-CP1.)

lensing reconstruction shows two halos which have hardly been disrupted by the collision apparently simply having gone through each other, as one might expect from a weakly interacting dark matter component. While striking, this special system should not be overinterpreted, because it is diﬃcult to render quantitative this qualitative interpretation. As the case for a dark matter component strengthened, more and more experiments to search for and ultimately characterize the dark matter have come on line. These experiments fall into two classes: direct and indirect. The direct dark matter search experiments look for the recoil of nuclei in the detector due to scattering of nuclei by dark matter particles. Backgrounds, for example from cosmic ray muons or from radioactive decays within or near the detector, obviously constitute a formidable obstacle. Indirect dark matter experiments search for an overproduction (or anomaly) of particles of various kinds from dark matter decays or annihilations. The current status of direct dark matter searches is summarized in Fig. 3.35 Indirect dark matter search experiments or observations may be subject to conﬂicting interpretations. The data from the ATIC and PAMELA experiments, for example, of the electron and positron ﬂuxes have been interpreted as evidence for dark matter annihilation, although other explanations have been put forth. See the Ref. 36 and references therein for a recent discussion of the methodology and

May 4, 2017 9:55 b2167-ch13

General relativity and cosmology

II-11

Fig. 3. Particle dark matter searches: Current status of constraints on WIMP dark matter from direct detection. Here spin-independent couplings have been assumed. (Note: Figure reprinted with permission from Ref. 35). (For color version, see page II-CP1.)

current status of indirect searches. To date, no conﬁrmed dark matter candidates have been found. However, these experiments have made considerable progress in constraining models and in demonstrating and perfecting experimental techniques. Search experiments for example have already excluded a large part of the cMSSMpreLHC model parameter space. The discovery of ﬂat spiral galaxy rotation curves and the subsequent dark matter search takes us to the early 1990s when the favored cosmological model includes radiation (contributing negligibly to the mean density today), ordinary matter, and a weakly interacting dark matter component with a contribution such that the sum of the components yields a spatially ﬂat universe. Despite the apparent beauty of a spatially ﬂat universe with a vanishing cosmological constant, there were a few wrinkles to this story given the evidence at that time including (i) the inconsistency between the high measured value of the Hubble constant and the ages of the oldest known objects in the universe, (ii) the inconsistency between cluster baryon fraction and nucleosynthesis predictions, and (iii) the inconsistency with the value of Ωmatter inferred from large-scale ﬂows. (See, for example, Refs. 37 and 38 and references therein.) This model was simpler than the six-parameter minimal model presently in vogue and by the end of the 1990s, it became clear that another component was needed to explain the current acceleration of the universe. This could be either a cosmological constant or some other form of stress–energy with a similar equation of state (i.e., a component with a large negative pressure) that subsequently became known as “dark energy.”

May 4, 2017 9:55 b2167-ch13

M. Bucher and W.-T. Ni

II-12

In 1970s, when astrophysical and cosmological observational data accumulated, the improvement in the determination of the age of globular clusters,39,40 the Hubble constant, and the abundance of elements led to tensions in ﬁtting FLRW cosmological models using only visible and dark matter. There were arguments that these measurements might already indicate a nonzero cosmological constant; however, the evidence was not yet compelling (see, e.g. Refs. 41–43). In 1998, when the light curve correction to the intrinsic variability in absolute luminosities of the type Ia supernova had been perfected and a suﬃciently large sample of such supernovae at intermediate redshift had been accumulated, it was shown that the expansion of the universe was accelerating. Acceleration of the scale factor was inconsistent with the minimal Ωmatter = 1 cosmological model in vogue at the time and could be explained by a nonzero cosmological constant or other equally radical extension of the then accepted cosmological model. See the chapter SNe Ia as a cosmological probe for a detailed account of this development.31 A key challenge of contemporary observational cosmology is to characterize the nature of the dark energy. If we ignore how the dark energy responds to cosmological perturbations, the problem can be expressed as measuring the run of the dimensionless parameter w = p/ρ through cosmic history, where p is the pressure and ρ is the density of the dark energy. If the dark energy were a cosmological constant, it follows that w = −1 at all times, but a host of other scenarios have been proposed where w is not exactly −1. Observations of the cosmic microwave background are ill-suited to characterizing the nature of the dark energy because the eﬀect of the dark energy on the CMB anisotropies can for the most part be encapsulated into a single number — the angular diameter distance to the surface of last scatter of the CMB photons at z ≈ 1100. What is needed is a survey of the geometry of the universe out to large redshifts. From cosmological observations, it has been inferred that our universe is very close to being spatially ﬂat. In a FLRW universe, the luminosity distance as a function of redshift is given by the integral z − 1 −1 dz Ωm (1 + z )3 + ΩDE (1 + z )3(1+w) 2 , dL (z) = (1 + z)(H0 ) 0

where ΩDE is the present dark energy density parameter and the equation of state of the dark energy w is assumed to be constant. For nonconstant w or a nonﬂat FLRW universe, the expression is similar but slightly more complicated. To determine w(z), luminosity distances and angular diameter distances need to be measured at many redshifts where dark energy plays a role. These can be measured for example by using (i) type Ia supernovae (as standard candles), (ii) baryon acoustic oscillations (BAO) in the matter power spectrum (as standard rulers), (iii) gamma ray bursts (as standards candle, although at present this method remains speculative) and (iv) gravitational waves (GW) from compact binaries and SMBHs (as standard “sirens” or candles). All these methods suﬀer from dispersion and bias due to gravitational lensing of distant objects, which can substantially alter their apparent luminosities and sizes. GW methods have the potential for high precision,

May 4, 2017 9:55 b2167-ch13

General relativity and cosmology

II-13

but one will likely have to wait about 20 years for a space-based GW observatory (e.g., eLISA44 ) to be put in place (see Chap. 12 for a review on GW mission proposal studies [45]). GW methods are also limited by gravitational lensing. However, since the intrinsic measurement uncertainty can be made very small (better than 0.1% with enough events), lensing uncertainty can be reduced by accumulating enough events for a detailed statistical analysis.46 Assuming ΛCDM, we can investigate whether the parameter values found using the CMB observations are consistent with the parameter choices indicated by completely independent probes such as observations of type Ia supernovae and observations of the scale of the baryonic acoustic oscillations at various redshifts, as shown in Fig. 4. The fact that the three ellipsoids overlap indicates that ΛCDM is telling a consistent story.

Fig. 4. Concordance of CMB, BAO and SNIa observations. The respective error ellipsoids are shown on the Ωm −ΩΛ plane. The supernova data are from the updated Union2.1 compilation of 580 Supernovae. (Note: Figure reprinted with permission from Ref. 47.) (For color version, see page II-CP2.)

May 4, 2017 9:55 b2167-ch13

II-14

M. Bucher and W.-T. Ni

On smaller scales and at later times when nonlinear eﬀects diﬃcult to model are dominant, the success of the ΛCDM is less compelling. This is not because there is a clear contradiction between the predictions and what is observed, but rather because the predictions of the theory are notoriously diﬃcult to calculate because of the presence of nonlinear physics and poorly understood processes such as star formation and galaxy formation for which much of the modeling is at present still largely ad hoc. It has been argued by some that alternative models of gravity are able to account for the observations without positing an additional dark matter component. For example, in Fig. 1, it is argued that MOND theory oﬀers a better explanation of galactic rotation curves. In MOND proposed by Milgrom28 in 1983, Newtonian theory remains valid for large accelerations, but becomes modiﬁed for centripetal accelerations smaller than the critical acceleration a0 ∼ 10−10 ms−2 . Numerically, the critical acceleration is related to the cosmological constant Λ : a0 ∼ Λ1/2 in natural units. Gravitational acceleration, of course, is not an invariant in Einstein’s theory, so in order to incorporate MOND into general relativity, some additional ﬁeld would be required to deﬁne a preferred time direction or equivalently a foliation of spacetime into three-dimensional hypersurfaces. Stratiﬁed theories of gravity and preferred-frame theories of gravity introduce a preferred foliation. With a preferred foliation and extra degrees of freedom other than the physical metric, the theories still need to satisfy the strong equivalence principle to experimental precision to be viable. This is one reason why stratiﬁed theories or preferred-frame theories are not easy to construct in an empirically viable manner, so that they are consistent with experiments and observations on various noncosmological scales and with local experiments. Such a theory was constructed in 197348,49 and also constructed for MOND phenomenology in 2004.31 Nevertheless, both these theories have encountered restrictions as empirical evidence is accumulated. For example, pulse timing observations on the relativistic pulsar–white dwarf binary PSR J1738+ 0333 provide stringent tests of these theories.50 However, GR is covariant, but cosmology is stratiﬁed in the large. In studies of the microscopic origin of gravity and quantum gravity, the question arises whether Lorentz invariance is fundamental or derived, especially in the canonical formulation.51 The histories of cosmology and relativity theory over the last 100 years have been closely intertwined. On the one hand, models of the universe rely on general relativity theory for part of their dynamics. On the other hand, the universe provides a testing ground for general relativity where Einstein’s theory can be confronted with observation under the most extreme conditions: on the largest length scales, over the longest time intervals, and at the highest energy scales, close to the putative big bang singularity. We close by identifying four areas where despite the existence of plausible hypotheses, it is likely that the last word has not been said and surprises may be forthcoming: (i) The dark matter acceleration discrepancy problem, (ii) the dark energy problem, (iii) the inﬂationary epoch, and (iv) a possible quantum gravity phase. It is likely that over the coming 100 years, part of the story told here may be substantially rewritten.

May 4, 2017 9:55 b2167-ch13

General relativity and cosmology

II-15

References 1. A. Einstein, “Kosmologische Betrachtungen zur allgemeinen Relativit¨ atstheorie,” Sitzungsberichte der Preussischen Akademie der Wissenschaften (Berlin) (8 February 1917), p. 142, reprinted in English in H. Lorentz, A. Einstein, H. Minkowski and H. Weyl, The Principle of Relativity: A Collection of Original Memoirs, 1923, Methuen, London. Paperback reprint, Dover, New York. 2. H. Shapley and H. Curtis, “The scale of the Universe,” Bull. Natl. Res. Counc. 2 (1921) 171, https://archive.org/ details/scaleofuniverse00shap. 3. P. J. E. Peebles, The Large-Scale Structure of the Universe (Princeton University Press, Princeton, 1980). 4. P. J. E. Peebles, Principles of Physical Cosmology (Princeton University Press, Princeton, 1993). ¨ 5. A. Friedmann, “Uber die Kr¨ ummung des Raumes,” Z. Phys. A 10 (1922) 377. ¨ 6. A. Friedmann, “Uber die M¨ oglichkeit einer Welt mit konstanter negativer Kr¨ ummung des Raumes,” Z. Phys. A 21 (1924) 326. 7. G. Lemaˆıtre, “Un Univers homog`ene de masse constante et de rayon croissant rendant compte de la vitesse radiale des n´ebuleuses extragalactiques,” Ann. Soc. Sci. Brux. 47 (1927) 49. 8. A. S. Eddington, “On the instability of Einstein’s spherical world,” Mon. Not. R. Astron. Soc. 90 (1930) 668. 9. H. P. Robertson, “Kinematics and world structure,” Astrophys. J. 82 (1935) 284. 10. A. G. Walker, “On Milne’s theory of world structure,” Proc. Lond. Math. Soc. s2–42 (1937) 90. 11. M. Davis, “Cosmic structure”, in One Hundred Years of General Relativity: From Genesis and Empirical Foundations to Gravitational Waves, Cosmology and Quantum Gravity, ed. W.-T. Ni (World Scientiﬁc, Singapore, 2015); Int. J. Mod. Phys. D 23 (2014) 1430011. 12. M. Bucher, “Physics of the cosmic microwave background anisotropy”, in One Hundred Years of General Relativity: From Genesis and Empirical Foundations to Gravitational Waves, Cosmology and Quantum Gravity, ed. W.-T. Ni (World Scientiﬁc, Singapore, 2015); Int. J. Mod. Phys. D 24 (2015) 1530004. 13. X. Meng, Y. Gao and Z. Han, “SNe Ia as a cosmological probe”, in One Hundred Years of General Relativity: From Genesis and Empirical Foundations to Gravitational Waves, Cosmology and Quantum Gravity, ed. W.-T. Ni (World Scientiﬁc, Singapore, 2015); Int. J. Mod. Phys. D 24 (2015) 1530029. 14. T. Futamase, “Gravitational lensing in cosmology”, in One Hundred Years of General Relativity: From Genesis and Empirical Foundations to Gravitational Waves, Cosmology and Quantum Gravity, ed. W.-T. Ni (World Scientiﬁc, Singapore, 2015); Int. J. Mod. Phys. D 24 (2015) 1530011. 15. R. H. Dicke, Gravitation and the Universe, Jayne Lectures for 1969 (American Philosophical Society, Philadelphia, 1970). 16. R. H. Dicke and P. J. E. Peebles, “The Big Bang cosmology: Enigmas and nostrums”, in General Relativity: An Einstein Centenary Survey, eds. S. Hawking and W. Israel (Cambridge University Press, Cambridge, 1979). 17. W. Rindler, “Visual horizons in world models,” Mon. Not. R. Astron. Soc. 116 (1956) 663. 18. C. W. Misner, “Mixmaster Universe,” Phys. Rev. Lett. 22 (1969) 1071. 19. K. Sato and J. Yokoyama, “Inﬂationary cosmology: First 30+ Years,” in One Hundred Years of General Relativity: From Genesis and Empirical Foundations to Gravitational Waves, Cosmology and Quantum Gravity, ed. W.-T. Ni (World Scientiﬁc, Singapore, 2015); Int. J. Mod. Phys. D 24 (2015) 1530025.

May 4, 2017 9:55 b2167-ch13

II-16

M. Bucher and W.-T. Ni

20. D. Chernoﬀ and H. Tye, “Inﬂation, string theory and cosmic strings”, in One Hundred Years of General Relativity: From Genesis and Empirical Foundations to Gravitational Waves, Cosmology and Quantum Gravity, ed. W.-T. Ni (World Scientiﬁc, Singapore, 2015); Int. J. Mod. Phys. D 24 (2015) 1530010. 21. P. Moore, The Story of Astronomy, 5th edn. (Grosset & Dunlap Publishing, New York, 1977). 22. I. I. Shapiro, “Solar system tests of GR: Recent results and present plans”, in General Relativity and Gravitation, eds. N. Ashby, D. F. Bartlett and W. Wyss (Cambridge University Press, Cambridge, 1990), p. 313. 23. N. T. Roseveare, Mercury’s Perihelion from Le Verrier to Einstein (Clarendon Press, Oxford, 1982). 24. F. Zwicky, “On the masses of nebulae and clusters of nebulae,” Astrophys. J. 86 (1937) 217. 25. V. Rubin and W. K. Ford Jr., “Rotation of the Andromeda Nebula from a spectroscopic survey of emission regions,” Astrophys. J. 159 (1970) 379. 26. V. C. Rubin, N. Thonnard and W. K. Ford Jr., “Extended rotation curves of highluminosity spiral galaxies. IV — Systematic dynamical properties, Sa through Sc,” Astrophys. J. 225 (1978) L107. 27. V. Rubin, N. Thonnard and W. K. Ford Jr., “Rotational Properties of 21 Sc Galaxies with a Large Range of Luminosities and Radii from NGC 4605 (R = 4 kpc) to UGC 2885 (R = 122 kpc),” Astrophys. J. 238 (1980) 471. 28. M. Milgrom, “A modiﬁcation of the Newtonian dynamics as a possible alternative to the hidden mass hypothesis,” Astrophys. J. 270 (1983) 365. 29. S. S. McGaugh, “The mass discrepancy–acceleration relation: Disk mass and the dark matter distribution,” Astrophys. J. 609 (2004) 652, arXiv:astro-ph/0403610. 30. S. S. McGaugh, “A tale of two paradigms: The mutual incommensurability of ΛCDM and MOND,” Can. J. Phys. 93 (2015) 250, arXiv:1404.7525. 31. J. D. Bekenstein, “Relativistic gravitation theory for the modiﬁed Newtonian dynamics paradigm,” Phys. Rev. D 70 (2004) 083509. 32. Planck Collab. (P. A. R. Ade et al.), “Planck 2013 results. XVI. Cosmological parameters,” Astron. Astrophys. 571 (2014) A16, arXiv:1303.5076. 33. Planck Collab. (P. A. R. Ade et al.), “Planck 2015 results. XIII. Cosmological parameters,” arXiv:1502.01589. 34. M. Markevitch et al., “Direct constraints on the dark matter self-interaction crosssection from the merging galaxy cluster 1E0657-56,” Astrophys. J. 606 (2004) 819, arXiv:astro-ph/0309303. 35. K. A. Olive et al., “Review of Particle Physics,” Chin. Phys. C 38 (2014) 090001, http://pdg.lbl.gov/2014/reviews/rpp2014-rev-dark-matter.pdf. 36. M. Cirelli, “Indirect searches for dark matter: A status review,” Pramana 79 (2012) 1021, astro-ph/1202.1454. 37. M. Turner, “The case for ΛCDM”, in Critical Dialogues in Cosmology, ed. N. Turok (World Scientiﬁc, Singapore, 1997). 38. J. P. Ostriker and P. J. Steinhardt, “The observational case for a low-density Universe with a non-zero cosmological constant,” Nature 377 (1995) 600. 39. P. Demarque and R. D. McClure, in The Evolution of Galaxies and Stellar Populations, eds. B. M. Tinsley and R. B. Larson (Yale University Observatory, New Haven, 1977). 40. H. Saio, “Ages of globular clusters. II,” Astrophys. Space Sci. 50 (1977) 93. 41. V. Petrosian, “Confrontation of Lemaˆıtre models and the cosmological constant with observations,” in Confrontation of Cosmological Theories with Observational Data, ed. M. S. Longair (Reidel, Dordrecht, 1974).

May 4, 2017 9:55 b2167-ch13

General relativity and cosmology

II-17

42. J. E. Gunn and B. M. Tinsley, “An accelerating Universe,” Nature 257 (1975) 454. 43. B. M. Tinsley, “Accelerating Universe revisited,” Nature 273 (1978) 208. 44. http://www.esa.int/Our Activities/Space Science/ESA s new vision to study the invisible Universe. 45. W.-T. Ni, “Gravitational wave detection in space,” in One Hundred Years of General Relativity: From Genesis and Empirical Foundations to Gravitational Waves, Cosmology and Quantum Gravity, ed. W.-T. Ni (World Scientiﬁc, Singapore, 2015); Int. J. Mod. Phys. D 24 (2016) 1630001. 46. T. D. Saini, S. K. Sethi and V. Sahni, “Possible use of self-calibration to reduce systematic uncertainties in determining distance-redshift relation via gravitational radiation from merging binaries,” Phys. Rev. D 81 (2010) 103009, arXiv:1005.4489. 47. The Supernova Cosmology Project (N. Suzuki et al.), “The Hubble Space Telescope Cluster Supernova Survey: V. Improving the Dark Energy Constraints Above z > 1 and Building an Early-Type-Hosted Supernova Sample,” Astrophys. J. 746 (2012) 85, arXiv:1105.3470. 48. W.-T. Ni, “A new theory of gravity,” Phys. Rev. D 7 (1973) 2880. 49. News and Views, “Preponderance of theories of gravity,” Nature 244 (1973) 537. 50. P. C. C. Freire et al., “The relativistic pulsar–white dwarf binary PSR J1738+0333 – II. The most stringent test of scalar-tensor gravity,” Mon. Not. R. Astron. Soc. 423 (2012) 3328, arXiv:1205.1450. 51. C. Soo and H.-L. Yu, “General Relativity without paradigm of space-time covariance, and resolution of the problem of time,” Prog. Theor. Exp. Phys. 2014 (2014) 013E01, arXiv:1201.3164.

May 4, 2017 9:55 b2167-ch14

Chapter 14 Cosmic structure

Marc Davis Astronomy Department, University of California at Berkeley, Berkeley, CA 94720-3411, United States [email protected]

The history of cosmic structure goes back to the time of Einstein’s youth, although few scientists actually thought of the problem of galaxy and cluster formation. The data and ideas were collected slowly as astronomers slowly realized the nature of the problem of large-scale structure. This paper will review several of the key episodes in the history of the ﬁeld. Starting with the discovery of dark matter in the 30s, the CMBR discovery in the 1960s to the idea of an early episode of inﬂation in the 1980s, the ﬁeld has had an acceleration of discovery. In the 80s it was realized that the initial conditions of the universe were speciﬁed by the cold dark matter (CDM). Now initial conditions for the formation of structure could be speciﬁed for any type of dark matter. With the advent of computing resources, highly nonlinear phases of galaxy formation could be simulated and scientists could ask whether cold dark matter was the correct theory, even on the scale of dwarf spheroidal galaxies, or do the properties of the dwarfs require a diﬀerent type of dark matter? In an idiosyncratic list, we review several of the key events of the history of cosmic structure, including the ﬁrst measurements of ξ(r), then the remarkable success of Λ CDM explanations of the large-scale universe. We next turn to velocity ﬁelds, the large-scale ﬂow problem, a ﬁeld which was so promising 20 years ago, and to the baryon acoustic oscillations, a ﬁeld of remarkable promise today. We review the problem of dwarf galaxies and Lyman-α absorption systems, asking whether the evidence is pointing toward a major switch in our understanding of the nature of dark matter. Finally, we discuss ﬂux anomalies in multiply-lensed systems, which set constraints on the number of dwarf galaxies associated with the lensing galaxy, a topic that is now very interesting since simulations have indicated there should be hundreds of dwarfs orbiting the Milky Way, rather than the 10 that are known. It is quite remarkable that many of the today’s results are dependent on techniques ﬁrst used by Einstein. Keywords: Cosmology; large scale structure. PACS Number(s): 95.35.+d, 98.52.−b, 98.65.−r, 98.80.−k

1. History of Cosmic Discovery On the birthdate of general relativity, the discovery of the Hubble law,1 that the large luminous nebulae were nearby galaxies receding from the Milky Way with velocity proportional to their distance, was more than 10 years in the future. That the galaxies were clustered was realized in the 1920’s but was ﬁrst discussed by Shapley and Ames2 upon the completion of their uniform plate survey of the 1000 II-19

May 4, 2017 9:55 b2167-ch14

II-20

M. Davis

brightest galaxies over the entire sky. The thought that the stars seen in the heavens were but a small fraction of the mass of the universe was not realized until 1933. The idea that the nature of the clustering would tell us anything about the nature of the dark matter was far into the future.3 The existence of dark matter had been known for the last 70 years, although at ﬁrst few acknowledged it. Zwicky,4 after 10 years of observations and comments on the strange results he was ﬁnding for such nearby clusters as Coma, concluded that the velocity dispersion within clusters of galaxies was so high, ∼ 1500 km/s, that they could not exist for longer than 1 billion years unless there was unseen mass holding them together. It was not until the 1980s that reconstruction of the mass proﬁle by data from the Uhuru X-ray satellite observed the Coma cluster of galaxies and reported that its mass proﬁle was consistent with Zwicky’s earlier estimates.5 Since then X-ray observations of clusters show this thermal Bremstrallung emission to always be present in galaxy clusters. Now, strong lensing of background galaxies by rich foreground clusters unambiguously shows the presence of dark matter in the foreground. Gravitational lensing is another tool which shows that the clusters contain enormous amounts of dark matter, just like Zwicky argued 70 years ago. An example is shown in Fig. 1.

Fig. 1. Gravitational lensing of Abell A2218 taken by the HST (Ref. 6). This beautiful image would certainly impress Einstein! The remarkable circular images of faint background galaxies can only be produced if there is 50 times the visible mass within the cluster, exactly as Zwicky had argued. (For color version, see page II-CP2.)

May 4, 2017 9:55 b2167-ch14

Cosmic structure

II-21

Evidence for dark matter in individual galaxies began to accumulate with Rubin et al.’s observation of emission line regions in the Andromeda galaxy,7 which they observed to have a ﬂat rotation curve, showing that something exterior to the starlight dominates its mass distribution. The evidence that the rotation curves of many galaxies was ﬂat could not be gathered until image tubes were available for spectrographs,8 and with such a device they found that every galaxy in their study had a ﬂat rotation curve (Fig. 2) rather than the expected ρ(r) ∝ r−1/2 , which immediately tells us the dark matter in galaxies has a mass proﬁle ρ(r) ∝ r−2 , indicating a large amount of dark matter in the outer regions of galaxies. This came as a shock to the cosmological community, but they already suspected the galaxies and clusters to be hiding vast amounts of dark matter.

Fig. 2. Rotation curves of 21 spiral galaxies, spectra taken by Rubin et al. (Ref. 8). Many of the galaxies show remarkably ﬂat rotation, indicative of dark matter in the outer portions of the galaxy.

May 4, 2017 9:55 b2167-ch14

II-22

M. Davis

The outrageous realization that our universe is dominated by an unseen dark matter has led to alternative theories, chief among them the so-called MOND theory (Modiﬁed Newtonian Dynamics).9 MOND modiﬁes Newton’s second law by the addition of a constant, a0 , a rather draconian solution to the dark matter problem. But it is healthy for the ﬁeld to see alternatives and to see how much of the cosmological phenomenology can be reproduced by a “simple” but radical modiﬁcation. Of course, this must ultimately be explained by a modiﬁed ﬁeld theory.10 However, later work showed that MOND does not even match the linear theory result seen on larger scales,11 which is a fatal error, and therefore it shall not be covered in this paper. The notion that our universe underwent an inﬂationary phase12,13 was an incredible motivator of an enormous body of work to ﬂesh out details. Finally, we had an answer to the question of why the universe should be approximately homogeneous and isotropic, and perhaps there would eventually emerge an explanation of the amplitude of the ﬂuctuations. Inﬂation ﬁnally gave us explanations for these and why the density of the universe was close to critical. It furthermore motivated the experimental search for the dark matter. Now, cosmologists realized that the large-scale structure of the universe was telling them the nature of the ﬂuctuation growth at a very early time, and publications exploded. This paper, will be an idiosyncratic review of key aspects of the historical development of cosmic structure. The ﬁeld is large and life is short. I apologize if I do not not discuss your particular ﬁeld. 2. Measurement of the Galaxy Correlation Function 2.1. Before 1980 Peebles14 wrote a series of 12 papers in which he applied power spectra analysis to all extragalactic catalogs then in existence. The utility of its transform, the covariance function ξ(r), provided a powerful and discriminating cosmological test, a tool to determine the veracity of a given theory. First, a few deﬁnitions given by Peebles which are now the standard. In a catalog listing only the right ascension and declination of galaxies, deﬁne solid angle dΩ and catalog density N . From any galaxy in the catalog, the probability that a neighboring cell is occupied is deﬁned as dP = (N dΩ)[1 + ω(θ)], where ω(θ) is the angular correlation function, with θ the angle between the two galaxies. If the galaxies are unclustered then ω(θ) = 0. This of course is sensitive to the distances of galaxies in the catalog. More interesting is the three-dimensional deﬁnition of correlations, dP = ndV [1 + ξ(r)], where n is the density and dV is the volume element and ξ(r) is the deviation from uniformity as a function of the distance r between the two galaxies. Of course, the galaxy correlations were presumed to be closely related to the underlying mass correlations, an assumption that has held up well. It was shown later that the galaxy ﬂuctuations δg can be generally represented as δg = bδdm ,

May 4, 2017 9:55 b2167-ch14

Cosmic structure

II-23

where δdm are the ﬂuctuations in dark matter and b is the galaxy bias, b ∼ 1.3, although if it is examined using a suﬃciently large catalog of galaxies,15 one can see it to be dependent on the galaxy’s mass and color, eﬀects which were known but ignored in Peebles earlier work. Peebles showed that the spatial correlation function, ξ(r), of diﬀerent catalogs obeyed simple scaling laws, indicating a reproducibility that must be telling us something quite important about the nature of structure formation. An example of the angular correlation scaling Peebles emphasized is shown in Fig. 3, where the angular correlations of the Zwicky catalog is plotted against the Shane–Wirtanen correlations and the Jagellonian ﬁeld correlations. The nonlinear correlation function has a power law slope of γ = 1.8, that is ξ(r) = (r/r0 )−γ with r0 = 5.4 h−1 Mpc. Once the catalogs are adjusted for their diﬀering depths, the correlation functions are seen to match. What was the shape of the correlation function telling us? 2.2. After 1980 The ﬁrst redshift survey of galaxies was published in 1982,17 and it opened our eyes, giving cosmologists a new, deﬁnitive look at what the universe contained, with enormous ﬁlaments of galaxies, huge voids, and rich clusters at the intersections of the ﬁlaments. It was clear that the universe was not the result of power-law initial conditions throughout the scales of galaxies and clusters, as had been advocated prior to 1980. Power-law initial conditions were simply a holding pattern in the absence of any better theory, but the birth of the inﬂationary models12,13 changed everything.

Fig. 3. Left: Angular covariance function for the Jagellonian catalog (squares), Shane–Wirtanen catalog (circles), and Zwicky catalog (triangles). Right: Test of the scaling relationship, where the w values have been scaled to the Zwicky catalog and the abscissas adjusted for diﬀerent depths (Ref. 16).

May 4, 2017 9:55 b2167-ch14

M. Davis

II-24

Until the advent of the inﬂationary model, it was not clear how to interpret the covariance function and the large-scale structure that cosmologists were beginning to see.17 At this time the data was clearly ahead of theory. Detailed calculations of the temperature ﬂuctuations in the cosmic background radiation for universes dominated by massive collisionless relics of the big bang were just becoming available19,20 as well as analytic approximations for the transfer function in the early universe, T (k). A ﬁrst analytic approximation of the density perturbation spectrum was |δk |2 =

Ak , (1 + αk + βk 3/2 + γk 2 )2

with α = 1.7 Γ, β = 9.0 Γ3/2 , γ = 1.0 Γ2 and Γ = (Ωm h2 ), with Ωm the matter density of the universe and h = H0 /100. The natural expectation was that Γ = 0.5, for Ωm = 1 and h = 0.7. With this power spectrum one could begin to run n-body codes to see if any of this made sense and could match the data.21 After careful calibration of the APM catalog of the full southern sky, Efstathiou and collaborators18 were able to measure the angular correlation function, w(θ), over large enough scales to deﬁnitively show that Γ = 0.2, and not Γ = 0.5 as most of the cosmological community believed. Figure 4 is what convinced the community of the problem. Conventionally it had been assumed that the cosmological density parameter Ωm = 1 and that the cosmological constant Λ is zero. But the data

Fig. 4. Data points from the APM catalog (Ref. 18). The heavy black line is the ﬁt for Γ = 0.2 while the light dotted line is for Γ = 0.5 (Ref. 18). Note that it falls well below the data for scales greater than 2 degrees, thus ruling out the standard model at the time. The thin solid and dashed lines show the results of the linear theory for the scale invariant CDM models with h = 1 and 0.75, respectively. This dotted line indicates that something is wrong with the Ωm = 1 CDM model.

May 4, 2017 9:55 b2167-ch14

Cosmic structure

II-25

favored a low-density model, either an open cosmology or one dominated by the cosmological constant, with more large-scale power than in the standard Ωm = 1 CDM model. A Λ cold matter model (ΛCDM) was immediately accepted, since an open cosmology, Ωm < 1, ΩΛ = 0, was even more abhorrent.

2.3. Remarkable large-scale structure in simulations Measurement of the detailed large-scale galaxy correlations would have to await the completion of extremely large redshift surveys such as the SDSS and 2dFGRS. Beautiful redshift space maps from both of then, is shown in the top and left portions of Fig. 5. These projects were only possible with teams of astronomers using dedicated telescope time and specialized instrumentation, principally multiple ﬁbers allowing the spectroscopic observation of hundreds of galaxies per exposure, as opposed to one object at a time in the ﬁrst redshift surveys. Now, the total

Fig. 5. Redshift space slices of the CfA2 stickman and the deeper walls seen with the SDSS and the 2dFGRS, as well as equivalent slices taken from a simulation of a ΛCDM cosmogony (Ref. 22). The galaxies are colored blue, while the simulation points are colored red. (For color version, see page II-CP3.)

May 4, 2017 9:55 b2167-ch14

II-26

M. Davis

surveys were several hundred thousand galaxies each. The advance of technology made this possible at a reasonable price and telescope time. Computer simulations were essential to understanding the models predicted by inﬂation. The most powerful computers available for cosmological simulations in the 80’s were made by Digital Equipment Corporation and featured all of 4 MB of memory. The earliest models21 did show that CDM models could produce the remarkable ﬁlaments and voids that had recently been discovered. The models were very small, 323 particles, but they were suﬃcient to make the point. As computing power increased, the models grew with them and cosmological simulations used a substantial fraction of available computing resources. The Millennium Run22 is one of the largest simulations of the formation of structure within the ΛCDM cosmogony so far carried out. It uses 1010 particles to follow the dark matter distribution in a cubic region 500 h−1 Mpc on a side, and has a spatial resolution of 5 h−1 kpc. Equivalent slices from the n-body simulations are shown as red points in Fig. 5. Note how well the Millennium simulation matches the statistics of the 2dFRGS (left and right). While the simulation does not have a rich cluster like Coma in the foreground (top), the statistics are still extremely similar in the bottom images. Note the very long stretch of connected ﬁlaments in the simulation, matching those of the observations. Pictures such as this demonstrate that ΛCDM ﬁts the data extremely well and contains a large grain of truth. The observations show the presence of extremely large ﬁlaments, and the clusters reside at the ﬁlament intersections. Note also the elongation of galaxies pointing toward the observer, but these are redshif t space maps. For example, in the center of the CfA2 “stick man,” the cluster is elongated, pointing to us. This is the Coma cluster, showing the high virialized velocity dispersion of the galaxies in the cluster, just as Zwicky observed 70 years ago. Now with these huge redshift surveys it is possible to measure galaxy correlations to extremely large scales. Furthermore, they have shown that the clustering strength of the galaxies is stronger for those objects of higher luminosity and red color, as expected in detailed models that were becoming available. 2.4. Measurement of the BAO eﬀect Since 2000, the acoustic peaks in the cosmic microwave background anisotropy power spectrum have clearly become one of the strongest cosmological probes.23 They measure the contents and curvature of the universe, demonstrate that the cosmic perturbations are generated early (z ∼ 1100) and are dominantly adiabatic, and by their mere existence largely validate the simple theory used to support their interpretation. The acoustic peaks occur because the cosmological perturbations are sound waves in the fully ionized relativistic plasma of the early universe. The recombination to a neutral gas at redshift z ∼ 1100 abruptly decreases the sound speed and eﬀectively ends the wave propagation of the CMBR and entrained baryons.

May 4, 2017 9:55 b2167-ch14

Cosmic structure

II-27

These ﬂuctuations give rise to Doppler peaks of the CMBR, as well as to ﬂuctuations known as the baryon acoustic oscillations (BAO), which have size of the ﬁrst doppler peak. Because the universe has a signiﬁcant ratio of baryons to DM, they will be imprinted onto the late-time power spectrum of the nonrelativistic matter. The baryonic perturbations continue to move after recombination, reaching a scale of 150 Mpc, where they begin to attract the dark matter. The resulting peak is perfect for performing the classical angular diameter test to redshift z 1 with galaxy samples from the BOSS survey, or to redshifts z ∼ 3 with absorption lines from the quasar sample of the BOSS survey.24 For example, Fig. 6 is a ﬁrst measurement of the BAO eﬀect, from an early analysis of 47,000 galaxies from the SDSS. The bump seen in the correlations is weak but it tells a remarkable story. First, recall the deﬁnition of angular distances. The comoving sizes, in a universe with Λ and matter (Ωm ), of an object or a feature at redshift z in line-of-sight (r ) and transverse (r⊥ ) directions are related to the observed sizes ∆z and ∆θ by H(z) and DA (z): r =

c∆z , H(z)

r⊥ = (1 + z)DA (z)∆θ.

The angular distance is given by c DA (z) = 1+z

0

z

dz . H(z)

The Hubble parameter is given by

H(z) = h Ωm (1 + z) + (1 − Ωm )exp 3 3

0

z

1 + w(z) dz 1+z

1/2

Fig. 6. A ﬁrst detection of the BAO. The plot of s2 ξ(s) from the SDSS survey (Ref. 23). The magenta line is a pure CDM model without a BAO, and the other lines are diﬀerent values of Ωm h2 . (For color version, see page II-CP3.)

May 4, 2017 9:55 b2167-ch14

II-28

M. Davis

Fig. 7. Plot of distance via the BAO technique versus redshift for four separate teams. A ΛCDM model, normalized by data from the Planck satellite, is a perfect ﬁt to the data. This indicates that the equation of state parameter w must be close to constancy, as a cosmological constant model (Ref. 25).

and w, the equation of state of the dark energy is p w(z) = . ρ z Since the measurement involves an integral of H(z), it is most informative to measure the BAO as a function of redshift by combining the results of diﬀerent surveys. An example is given in Fig. 7 showing that the data is exactly consistent with an interpretation of w = −1 as a constant, as with the cosmological constant Λ. The BAO measurements are proving to be an excellent means of constraining w(z). The growth of studies of the BAO is stunning to this old observer. In the CfA1 redshift survey17 of bright (mB < 14.5) galaxies, we were at the cutting edge of technology and galaxies were measured one at a time. The shot noise was too large and the volume too small to allow ξ(r) to be measured for any scale r > 20 Mpc. Now the realistic goal of the observing teams is to measure galaxies ∼ 1000 at a time with a target of 106 galaxies, or more, giving them the ability to measure ξ(r) reliably to scales r ∼ 150 Mpc. Their motivation is to measure the scale of the BAO peak to better than 1%, and then to use this as a “standard ruler” for a precise cosmological test, setting strong constraints on possible models of dark energy, in particular the equation state parameter w. The measurements are going very well and the constraint on w is becoming tighter and tighter. This is a worthwhile goal as deviations from the standard DE model might be very modest. 2.5. Further measurements of the power spectrum Figure 8 is a plot of the observed power spectrum of structure with wavelength spanning four orders of magnitude, where the solid line is ΛCDM. Asymptotically the curve varies as k 1 on large-scale and as k −3 on small-scale. Data ranges from measurements of the cosmic microwave background on large-scale to Lyman-α forest

May 4, 2017 9:55 b2167-ch14

Cosmic structure

II-29

Fig. 8. A power spectra summary covering four orders of magnitude from a variety of cosmological studies, ranging from CMB measurements to constraints from Ly α clouds (Refs. 26 and 27). The agreement with the curve, a ΛCDM model, is remarkable. (For color version, see page II-CP4.)

clouds on small scales. A cold dark matter (CDM) particle is nonrelativistic as diﬀerent scales come within the horizon. Scales on the left-hand side, where the curve is rising, came within their horizon after the universe was dominated by nonrelativistic matter, that is matter dominated, while scales on the right-hand side of the ﬁgure came within their horizon while the universe was radiation dominated. If the dark matter is relativistic it will free-stream until the temperature of the CMB drops below its mass, and thus perturbations of that scale are erased. This is why the perturbations of a light particle, e.g. a particle of rest mass less than 1 keV, produce a cut-oﬀ in the observable range. If the dark matter is, for example, a warm dark matter candidate (WDM), the power spectrum on small scales, high values of k, will show a cut-oﬀ when the temperature of the universe was equal was equal to the rest mass of the particle. The precise slope of the small-scale measurements is sensitive to the masses of the neutrino background particle, which will cause the slope to bend down from the small scale asymptotic slope of −3. It is quite remarkable that the full spectrum of cosmic structure ﬁts the ΛCDM model so well. A tremendous amount of work went into these tests in the last 30 years, and they all set constraints on cosmic structure as deﬁned by ΛCDM. 2.6. Lyman-α clouds Observations of redshifted Lyman α forest absorption at z ∼ 2−4 in the spectra of QSOs provide a sensitive probe of the distribution of gaseous matter in the universe,

May 4, 2017 9:55 b2167-ch14

II-30

M. Davis

Fig. 9. A simulation of the Lyman-α forest showing the typical spectrum of a high redshift quasar (Ref. 28). The optically thin Ly-α lines are the crossings of the line of sight through the large scale ﬁlaments. A representation of large scale structure of the ﬁlaments is shown as well. (For color version, see page II-CP4.)

indicating it to be highly ionized. The understanding of the origin of the Lyman-α clouds became clear only after n-body simulations in the 1980s demonstrated the ubiquity of the ﬁlamentery clustering throughout the intergalactic medium in a CDM universe. A model of the Lyman-α forest is given in Fig. 9 where it is seen that the quasar line of sight pierces a ﬁlament is the site of a cloud. Prior to this realization the community did not have any realistic models for the forest, but this model appears to satisfy all the constraints. The precise measurements of the clustering of Lyman-α forest absorption lines provides further support for cosmological models based on inﬂation, CDM and vacuum energy, as seen in Fig. 8. The Lyman-α forest, due to its spectral nature, probes the matter power spectrum in velocity space. But the measurement is also sensitive to the mass of the dark matter particle which would generate a small-scale cut-oﬀ in the linear power spectrum, with a cut-oﬀ that scales approximately inversely with the particle mass and is a result of free streaming of the dark matter. If this mass is of order 1 keV, the cut-oﬀ occurs on the scale of dwarf galaxies, and the dark matter particle would be in the class of warm dark matter (WDM).29 If the WDM is a thermal relic of mass 1 keV, the cut-oﬀ occurs for k(h/Mpc) > 10, which is slightly beyond the reach of the power spectrum plotted in Fig. 8. However, a further probe is to measure the individual line shapes of Lyman-α clouds, which will be broadened by the thermal velocity, and this can be observed with ground-based spectrographs on large telescopes. By requiring that there be enough small-scale power in the linear power spectrum to reproduce the observed properties of the Lyman-α forest in quasar spectra, Narayanan et al.30 arrive at a lower limit to the mass of the WDM particle of 0.75 keV. They argue that any model that suppresses the CDM linear power spectra

May 4, 2017 9:55 b2167-ch14

Cosmic structure

II-31

more severely than a 0.75 keV particle cannot reproduce the Lyman-α forest. Recent progress in this ﬁeld is reviewed by Viel29 who cites upper limits of 4.5 keV. Improving these observations is important because a cut-oﬀ in small-scale power of the dark matter would corroborate the possible cut-oﬀ as indicated by dwarf galaxies, discussed below. 3. Large Scale Flows Soon after the discovery of the CMBR, the search began for large-scale inhomogeneities in the CMB, and the dipole anisotropy was detected in 1976 in competition between groups at Princeton and Berkeley.31,32 With the advent of the COBE satellite in 1989 we ﬁnally measured the ﬁrst ∼ 20 spherical harmonics, including a beautiful picture of the dipole anisotropy. The Milky Way and local group of galaxies are moving at 640 km/s, 40◦ away from the Virgo local supercluster, which had been thought to be the attractive center. The question immediately raised by this observation was what was responsible for our motion? Was the ﬂow generated gravitationally, or does it result from some nongravitational eﬀect? Expectation of the CDM models was that the dipole anisotropy should be generated by local perturbations, with essentially nothing generated from great distances, z > 0.15. This would be the same type of motion which is apparent in n-body models, where galaxies and large-scale structure are moving together, simply expressing the local gravitational ﬁeld. Because the gravity ﬁeld is the vector sum of all the perturbations on every scale, it makes no sense to break down our motion into components generated by named clusters structures, since they do not dominate our gravitational acceleration. Rather, the gravity is controlled by the total mass distribution which is dominated by the voids and ﬁlaments. About all that can be done is to plot the net acceleration of the local group as a function of distance. It was recognized that the measurement of large-scale ﬂow could serve as a check on Ωm , as it is a measure of the gravity on scales even larger than clusters, ∼ 50 Mpc. One presumes that the galaxy distribution, multiplied by a bias factor “b,” is a substitute for the large-scale mass ﬁeld. It need not be precise, merely tracing the rough outline of the mass distribution. Because of the long decay time from the time, z ≈ 1100, when the gravity and velocity ﬁelds were released from the radiation, any random velocity ﬁeld has decayed, eliminating rotational and other decaying modes. The velocity ﬁeld present now has an especially simple relationship with the responsible gravitational ﬁeld. Those components of the velocity ﬁeld which are not coherent with the density ﬂuctuations have decayed as the universe expands, and so at late times one expects the velocity ﬁeld to be aligned with the gravity ﬁeld, at least in the limit of small amplitude ﬂuctuations. In the linear regime, this relation implies a simple proportionality between the gravity ﬁeld g and the velocity ﬁeld vg , namely vg ∝ gt where the only possible time t is the Hubble time. The exact expression depends on the

May 4, 2017 9:55 b2167-ch14

M. Davis

II-32

mean cosmological mass density Ωm and is given by34 vg (r) =

2f (Ωm ) g(r), 3H0 Ωm

where f (Ω) is the linear growth rate. Given complete knowledge of the mass ﬂuctuation ﬁeld, δρ (r), over all space, the gravity ﬁeld g(r) is r − r g(r) = G¯ ρ d3 r δρ (r ) , |r − r|3 where ρ¯ is the mean mass density of the universe. Replacing the integral over space with a sum over the galaxies in a catalog, we have H0 β 1 ri − r H0 β vg (r) = + r, 4π¯ n i φ(ri ) |ri − r|3 3 where n ¯ is the true mean galaxy density in the sample, β ≡ f (Ω)/b with f ≈ Ω0.55 the linear growth factor with Λ included, and with radial selection function φ(r).a The work is greatly aided by supplementing it with secondary distance indicators of as many galaxies as possible. These secondary distance indicators can be for example, galaxies with measured magnitudes and the amplitude of its asymptotic rotation curve, which is suﬃcient to predict a galaxy’s distance to approximately 20%.35 Of course with the relatively poor distance indicator, it will take many galaxies to produce a meaningful signal, but the samples now include approximately 5000 galaxies drawn from the full sky. The techniques for ﬁnding the best ﬂow-ﬁeld have undergone considerable evolution, a recent ﬂow-ﬁeld is show in Fig. 10.33 The velocity ﬁeld is determined from the galaxies having secondary distance indicators by a technique known as the inverse Tully–Fisher method, to which approximately 40 orthogonal modes are ﬁt. The gravity ﬁeld is determined by a full-scale sample of 45,000 galaxies in the 2MRS catalog, which is useful to a distance of z < 0.04. The ﬁgures all show the ﬂow as seen from the local group frame, but in order to see the deviations beyond the dipole, 400 km/s of dipole anisotropy is subtracted from the LG motion of all the ﬁgures. The top ﬁgure is the distribution of galaxies less than cz < 2000 km/s where the red is outﬂowing from our perspective and the blue is inﬂowing, with the size of the symbol being proportional to the velocity. The second row is for galaxies with 2000 < cz < 4000 km/s, the third row is for galaxies with 4000 < cz < 6000 km/s, and the fourth row is for galaxies with 6000 < cz < 10000 km/s. The ﬁgures on the right are the gravity ﬁeld computed from the full sky distribution of galaxies, with the same change of coordinates as for the ﬁgures on the left. This gravity ﬁeld was generated with Ωm = 0.3. Note that the velocity ﬁeld is not a simple dipole, as the local velocity is responding to all the multipoles of the local gravity. Note also the quite amazing agreement between the curves on the a φ(r)

is deﬁned as the fraction of the luminosity distribution function observable at distance r for a given ﬂux limit.

May 4, 2017 9:55 b2167-ch14

Cosmic structure

II-33

Fig. 10. The velocity ﬁeld of local galaxies (left) and the gravity ﬁeld (right) derived from the 2MRS sample (Ref. 33). The stars are outﬂowing from the local group frame of reference and the circles are inﬂowing. The size of the symbol is proportional to the velocity, with a caption of sizes drawn at the bottom of the ﬁgure. To better see the ﬁelds, a 400 km/s dipole has been subtracted from each plot. The four rows divide the samples by distance. It is important to note that the two ﬁelds were independently derived and show essentially perfect agreement, just as linear theory predicts.

left with those on the right. This tells us that for the most part linear perturbation theory does describe the motion of galaxies in the universe, a reassuring check. It is apparent that the source of the bulk of the 640 km/s dipole pattern seen in the local CMBR has 500 km/s due to the galaxy distribution up to a distance of 100 h−1 Mpc, with the rest generated by larger scale structure, as expected in ΛCDM models. There is no need for exotic explanations of our dipole velocity ﬁeld.

May 4, 2017 9:55 b2167-ch14

II-34

M. Davis

The agreement between the velocity ﬁeld and the gravitational ﬁeld is particularly reassuring for our understanding of cosmological theory. Such a conclusion is a comfort for linear perturbation theory in an expanding universe and was certainly expected, because the eﬀective smoothing of the velocity and gravity ﬁelds has left only linear perturbations, which appears to be adequate for the large scales tested by this method. We see no evidence that the dark matter does not follow the galaxy distribution, and it is consistent with constant bias on large scales. There is no evidence for a nonlinear bias in the local ﬂows. A smooth, unclustered component to the dark matter in the universe can only be tested by the agreement between β measured in the large-scale ﬂows and Ωm measured by the CMBR. To the accuracy of the measurement, the two do agree. Furthermore, the velocity–gravity comparison measures the acceleration on scales up to 30–50 Mpc and since we derived a similar value of β as for clusters of galaxies, we conclude that dark matter appears to fully participate in the clustering on scales of a few megaparsec and larger. It is interesting that the counts of galaxies give the best possible gravity ﬁeld, reinforcing the old idea that the mass of the halo around a galaxy is not very well correlated with the luminosity of that galaxy. We ﬁnd no evidence for large-scale ﬂows beyond those indicated by the gravity ﬁeld. Note that our analysis has not used the CMBR dipole, but at the end we see a velocity ﬁeld that is consistent with it. We see no evidence that the dipole in the CMBR is produced by anything other than our motion in the universe. 4. Dwarf Galaxies as a Probe of Dark Matter Astronomers did not realize that there was a missing satellite problem for the Milky Way until large simulations showed an enormous number of substructures to every large galaxy. Simulated galaxies retain a large amount of substructure formed by earlier collapses on smaller scales, predicting hundreds or thousands of subhalos in contrast to the ∼ 10 dwarf spheroidal satellites of the Milky Way. N -body simulations of the formation of dark matter halos robustly predict that approximately 10% of the mass of any galactic dark halo should be in the form of sub-halos with numbers that rise steeply toward lower masses. For the past 10 years, astronomers have been aware of this disagreement between observed dwarf spheroidal galaxies and their numerical simulations of normal galaxies.36 Boylin-Kolchin et al.37 used the aquarius simulations to show that the most massive subhaloes in detailed ΛCDM galaxy-mass simulations of dark matter haloes are grossly inconsistent with the dynamics of the brightest Milky Way dwarf spheroidal galaxies. The best-ﬁtting hosts of the MW dwarf spheroidals in all CDM simulations predict at least 10 haloes with rotation curves double that amount observed in MW dwarfs. Another question is what has happened to the missing 90% of condensed halo’s with the mass of MW dwarfs? The dwarf spheroidals have lower dispersion and therefore lower central density or lower mass than predicted for the most massive satellites37 as shown in Fig. 11.

May 4, 2017 9:55 b2167-ch14

Cosmic structure

II-35

Fig. 11. Observed Vcirc values of the nine bright, LV > 105 L , dwarf spheroidals in the MW along with rotation curves corresponding to dwarfs with Vmax = (12, 18, 24, 40) km/s (Ref. 37). The absence of any dwarfs in the Milky Way with potential of ∼ 40 km/s is a serious problem for the CDM cosmogeny as there are far more than 10 dwarfs in the simulations with potential this deep.

These dwarfs, and the absence of the abundant sub-halos, do not seem to match the expectations of ΛCDM, and this is the ﬁrst prediction of the theory that fails to match the observations. However, the dark matter predictions hold only if the velocity dispersion of the dark matter particle is very low, which would be the case if the dark matter mass is suﬃciently high, 100 keV. We know so little about the nature of the dark matter. Is this a clue to its mass or its history? All we know is that the DM is collisionless and has a rest mass of at least 1 keV, with guesses that the DM is the supersymmetric partner to an ordinary particle with a mass in the range 10–100 GeV. After 30 years of eﬀort, we still have no idea if dark matter is a supersymmetric particle with mass greater than a baryon. The missing satellite problem seems like it could be solved fairly easily by baryonic physics, which is not included in the simulations above. For example, the velocity threshold at which subhalo satellites and dwarf satellite counts diverge is close to 30 km/s, a value that is reached by the heating of the ultraviolet photoionizing background which suppresses gas accretion in small halos with low velocity dispersion. Perhaps these halos are dark matter only, and the baryons are simply too hot to condense?

May 4, 2017 9:55 b2167-ch14

II-36

M. Davis

Even if the baryons condense, the diﬃculty of understanding galaxies is complex because the gas will form stars and supernovae, which can reheat the gas, causing it to outﬂow, possibly changing the orbits of the dark matter particles so as to reduce the DM central density.38 Whether this is possible for the very low mass, ∼ 107 M dwarfs, is not clear. Analytical models39 suggest that with so few stars there is not enough energy in supernovae alone to create suﬃciently large dark matter cores. Alternatively, supernova and stellar winds from the ﬁrst generation of stars could drive remaining gas out of the shallow potential wells of these low mass halos making them impossible to see. But it has been known for some time that physics including more than gravitation was essential to the evolution of the baryons if the simulations are to make galaxies with observed properties. The addition of supernova feedback and hot winds from early type stars, in addition to heating by AGNs, reduces the density of baryons and DM in the galactic center and are now realized as critical in galaxy formation.40 Thus, it is necessary to understand these additional heating mechanisms very well before concluding that there is a problem with standard CDM. The situation is at present controversial, with diﬀerent simulators producing diﬀerent results.41 Dark matter halos oﬀer a unique method of determining the mass of DM. The transfer function of ﬂuctuations on galaxy sizes varies as k −3 (k is the wave number), provided the DM particle is collisionless and eﬀectively at rest, with small free streaming length compared to the size of the ﬂuctuation. WDM is a simple modiﬁcation of CDM in which the dark matter particles have initial velocities due either to their having decoupled as thermal relics or to their having been formed via nonequilibrium decay. The free streaming length of the DM, the distance a dark matter particle will travel in the early universe, erases structure of that mass and smaller. An important check on the rest mass of the particle is the smallest astronomical structure that is dominated by dark matter. If the rest mass of the dark matter is suﬃciently high then the dark matter has ever more negligible peculiar velocities, but for rest masses of order 5 keV the particles today have a primordial velocity dispersion of approximately 20 km/s, and this is suﬃcient to play an important role in the formation of dwarf galaxies. The structure of galaxies in CDM is self-similar because the free streaming length is extremely small so there is no scale deﬁned by the collapsing structure. In WDM, the smallest galaxies would not be self-similar, but their DM would present a scale, the free streaming scale, of order the drift distance a particle would travel in the age of the universe. The WDM cosmogony thus erases suﬃciently small perturbations. CDM cannot pick a scale at the present epoch and any DM particle of rest mass greater than ∼ 100 keV therefore is a CDM candidate. Could it be that WDM erased the small-scale perturbations? No CDM-based model of the satellite population of the Milky Way explains this result. The problem lies in the densities: The number density of galaxies in the model is typically set by matching halos to the galaxy counts in the sky and in so

May 4, 2017 9:55 b2167-ch14

Cosmic structure

II-37

doing requires the dwarf spheroidals to have DM haloes that are a factor of 5 more massive than is observed. Thus, the problem of the shallow potential wells remains. Rather than erasing small-scale perturbations, as in WDM models, another method of smoothing the core of dwarf galaxies is to have the dark matter be selfinteracting (SIDM).42 This model has been discussed several times in the literature starting with Spergel and Steinhardt.43 If the dark matter scatters it will form a core in the center of galaxies, making the potential well shallower, reducing the velocities of the simulation. The central density is reduced as shown in Fig. 12, bringing the simulations into closer agreement with the data. But the dark matter would need a strongly interacting cross-section which seems unlikely. Rocha et al.42 argue that SIDM with σ/m 0.1 cm2 /g appears capable of reproducing reported core sizes and central densities of dwarfs. (1 cm2 /g = 1 barn/GeV, a nuclear-scale cross-section.) They have also shown that self-interaction of that strength does not force the galaxies to become spherical,44 which previously was an argument against SIDM models. SIDM models usually have assumed the simplest scattering cross-section. The particles are like billiard balls with no velocity dependence in their cross-section. In the absence of a theory of the dark matter, it is best to keep the simulations simplest. However, a few theorists are braver,45,46 who use a nonpower-law velocity dependence of the cross-section which is motivated by a Yukawa-like term. As in the other SIDM models, they ﬁnd the density proﬁle of dwarfs is unchanged except for the constant density core. These models oﬀer an interesting alternative to the standard cold dark matter models that can reduce the discrepancy between the brightest Milky Way satellites and the dense subhaloes found in CDM simulations. The understanding of the nature of the dwarf spheroidal galaxies is clearly a point of tension with the CDM models, a problem which must be solved, either

May 4, 2017 9:55 b2167-ch14

II-38

M. Davis

through astrophysics or particle physics. Perhaps some clariﬁcation will emerge from detailed high-resolution study of Lyman-α clouds, or from a realistic assessment of the eﬀects of the baryonic component? It is quite remarkable that the dwarf spheroidals might actually set an extremely important particle physics parameter. 5. Gravitational Lensing Einstein knew all about gravitational lensing, and it is appropriate to celebrate the 100th year of GR with a section detailing how gravitational lensing is used today. This section shall discuss three separate projects where gravitational lensing was essential. This is not a complete list but represents the ﬁrst projects I would desire to discuss with Einstein. Reviews of lensing are given by Blandford and Narayan.47 5.1. Double images In 1979, the astronomical community was greatly surprised when Walsh et al. announced the discovery of a gravitational lens.48 Two images of a quasar at z = 1.4, separated by 5.7 arc seconds, were the ﬁrst documented case of a gravitational lens. The lens was a cluster of galaxies plus an elliptical galaxy in a cluster at an intermediate redshift. The separation of the lens is twice that expected for most galaxies and the quasars are oﬀ-center from the galaxy, suggesting the galaxy cluster is partly responsible for the lensing. Only a few scientists realized the power of gravitational lensing but soon the ﬁeld exploded. An example of the spectacular use of lensing is that of Markevitch,49 see Fig. 13. Now gravitational lensing has become a wonderful, unexpected tool. 5.2. Bullet cluster The bullet cluster is an example of two clusters that have slammed through each other and in another ∼ 109 years will merge into one, a process that is completely normal in the growth of structure, as shown in the beautiful n-body movies. The baryons emitting the X-rays are colored red in Fig. 13, while the weak gravitationallensing results showing the existence of dark matter is colored blue. Galaxies are mixed with the dark matter, as expected since their cross-section is small, but the collisional hot gas shows a bow shock created when the two clouds collided. The discovery of this remarkable “bullet cluster”49 has shown quite clearly that the dark matter is collisionless while the hot gas behaves like normal baryons. This has emerged as a very illustrative example that is a favorite to discuss in class with students. The problems with the speed of the merging clusters strikes me as a minor issue. A MOND explanation for the bullet cluster is a bit far-fetched. 5.3. Substructure of gravitational lenses Another way to search for the missing dark matter substructure is to monitor the ﬂux of multiply-imaged quasars behind foreground galaxies. If the galaxy is

May 4, 2017 9:55 b2167-ch14

Cosmic structure

II-39

Fig. 13. Chandra X-ray image of the “bullet cluster” (red) superposed with lensing image of two clusters (blue) (Ref. 49). (For color version, see page II-CP1.)

a smooth potential, then the relative brightness and positions of the quasars can be predicted, but if there are multiple dwarf galaxies within the foreground, then the microscopic details of the quasars will be anomalous. Kochanek and Dalal50 detail results of seven cases and conclude that lowmass subhalos remain the best explanation of the galaxy’s potential. They conclude that a large fraction (∼ 1%– 2%) of the projected mass at the Einstein radius, ∼ 5 kpc, must be in the form of local substructure. The number of anomalies is far higher than the 10 dwarf ellipticals in the Milky Way. When this result was ﬁrst announced, the astronomical community was skeptical, but Dalal and Kochanek have brought together data from the radio and the optical which eliminates foreground dust as the explanation. This is strong evidence in favor of the ΛCDM model but the uncertainties of the measurement are quite large because of the small sample size. Analysis of the SDSS sample will ﬁnd more multiply-lensed objects, but they will have small angular size, making them diﬃcult for detailed followup. The multiply-lensed quasars are giving results that appear consistent with the standard model but the number of suitable candidates are insuﬃcient to give details. A new technique for detecting dwarfs in other galaxies involves ALMA imaging of galaxies that are strongly lensing background galaxies containing dust.51 The lensed dusty galaxies can be found in the data from the CMB experiments at the South Pole. Most galaxies have dust in their nuclear regions of size ∼ 1 kpc, corresponding to an angular scale of 0.2, a size large enough to prevent lensing by a single star in the foreground galaxy. But it will be lensed by dwarf galaxies of order 108 M◦ , which have an Einstein radius of the same angular extent. The dusty nuclear regions

May 4, 2017 9:55 b2167-ch14

II-40

M. Davis

typically show strong velocity gradients, some indicating fast-rotating disks and others indicating major mergers. The spatially resolved spectroscopy of ALMA thus will permit the enhancement of the spatial resolution. Simulations of the procedure appear promising, and they estimate as many as one dwarf object per target. This is a brilliant technique for utilizing the magniﬁcation of gravitational lenses. 6. Conclusion It is remarkable how developed the ΛCDM story has become in a relatively short time, a beautiful tale that we would love to share with Einstein. It is a new spin on his “greatest blunder.” And Einstein would certainly love the uses to which gravitational lensing, his tool of 100 years ago, has been applied. The ﬁeld has developed in the last century in ways that were unimaginable while Einstein was alive. The search for the dark matter is still outstanding after 30 years of eﬀort, but hopefully it will soon be discovered either in an underground experiment or at CERN’s LHC. The inability to detect the dark matter as regions of phase space are steadily excluded will become a problem for the WIMP explanation of dark matter unless the next generation of experiments has a positive detection. The possible hints of problems with the ΛCDM model on the scale of dwarfs are extremely puzzling and will hopefully be resolved in a few years. Detecting fainter dwarfs is certainly worthwhile. The prospect of using ALMA to observe strong galactic lenses that are magnifying background dusty galaxies is an excellent idea and we hope the experiment works. Or perhaps the dwarf data is telling us that the notion of self-interacting dark matter should be taken seriously? The next few years will, with the new telescopes under development and further study of the remarkable recent advances in the inﬂationary early universe, lead to new, tighter cosmological constraints that will have implications for particle physics that I cannot imagine today. References 1. E. Hubble, Contributions from the Mount Wilson Observatory, Vol. 3 (1927), pp. 23– 28. 2. H. Shapley and A. Ames, Harvard College Obs. Bull. 887 (1932) 1–6. 3. A. G. Doroshkevich, Y. Zeldovich and I. D. Novikov, Pis’ma Zh. Ehksp. Teor. Fiz. 8 (1968) 90–95. 4. F. Zwicky, Helvetica Phys. Acta 6 (1933) 110–127. 5. K. I. Tanaka, Y. Fujishima and M. Fujimoto, Publ. Astron. Soc. Jpn. 34 (1982) 147. 6. J. P. Kneib, R. S. Ellis, I. Smail, W. J. Couch and R. M. Sharples, Astrophys. J. 471 (1996) 643. 7. V. C. Rubin and W. K. Ford, Astrophys. J. 159 (1970) 379. 8. V. C. Rubin, W. K. Ford and N. Thonnard, Astrophys. J. 238 (1980) 471–487. 9. M. Milgrom, Astrophys. J. 270 (1983) 365–370. 10. J. D. Bekenstein, Phys. Rev. D 70 (2004) 3509. 11. A. Nusser, Mon. Not. R. Astron. Soc. 331 (2002) 909–916.

May 4, 2017 9:55 b2167-ch14

Cosmic structure

12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46.

II-41

A. H. Guth, Phys. Rev. D 23 (1981) 347–356. A. D. Linde, Phys. Lett. B 100 (1981) 37–40. P. J. E. Peebles, Astrophys. J. 185 (1973) 413–440. I. Zehavi, Z. Zheng, D. H. Weinberg, J. A. Frieman and the SDSS team, Astrophys. J. 630 (2005) 1–27. P. J. E. Peebles, Astrophys. J. 196 (1975) 647–652. M. Davis, J. Huchra, D. W. Latham and J. Tonry, Astrophys. J. 253 (1983) 423. G. Efstathiou, W. J. Sutherland and S. J. Maddox, Nature 348 (1990) 705–707. G. R. Blumenthal and J. R. Primack, in Fourth Workshop on Grand Unification, eds. H. A. Weldon, P. Langacker and P. J. Steinhardt (Birkhauser, Boston, 1983), p. 256. J. R. Bond and G. Efstathiou, Astrophys. J. 285 (1984) L45–L48. M. Davis, G. Efstathiou, C. Frenk and S. D. M. White, Astrophys. J. 292 (1983) 371. G. Lemson and the Virgo Consortium, arXiv:astro-ph/0608019. D. Eisenstein, I. Zehavi, D. W. Hogg, R. Scoccimarro, M. R. Blanton and SDSS team, arXiv:astro-ph:/0501171. D. Eisenstein, F. Beutler, A. S. Bolton, A. Burden and BOSS team, Am. Astron. Soc. 223 (2014) 245–246. L. Anderson, E. Aubourg, S. Bailey and BOSS team, arXiv:astro-ph:/1312.4877v2. M. Tegmark and M. Zaldarriaga, arXiv:astro-ph/0207047. R. Croft, D. Weinberg, M. Bolte, S. Burles, L. Hernquist, N. Katz, D. Kirkman and D. Tytler (2002), arXiv:astro-ph/0012324. C. Stark, Private communication (2014). M. Viel, G. Becker, J. Bolton and M. Haehnelt, arXiv:1306.2314 [astro-ph.CO]. V. K. Narayanan, D. N. Spergel, R. Dave and C. P. Ma, Astrophys. J. 543 (2000) L103–L106. B. E. Corey and D. T. Wilkinson, Bull. AAS 8 (1978) 351. G. F. Smoot, M. V. Gorenstein and R. A. Muller, Phys. Rev. Lett. 39 (1978) 898. M. Davis, A. Nusser, K. L. Masters, C. Springob, J. P. Huchra and G. Lemson, Mon. Not. R. Astron. Soc. 413 (2011) 2906. P. J. E. Peebles, The Large-Scale Structure of the Universe (Princeton University Press, Princeton, 1980). C. M. Springob, K. L. Masters, M. P. Haynes, R. Giovanelli and C. Marinoni, Astrophys. J. Suppl. 172 (2007) 599. P. Bode, J. P. Ostriker and N. Turok, Astrophys. J. 556 (2001) 93–107. M. Boylan-Kolchin, J. S. Bullock and M. Kaplinghat, Mon. Not. R. Astron. Soc. 422 (2012) 1203–1218. D. H. Weinberg, J. S. Bullock, F. Governato, R. K. de Naray and A. H. G. Peter, arXiv:1306.0913 [astro-ph.CO]. J. Penarrubia, A. Pontzen, M. G. Walker and S. E. Koposov, Astrophys. J. Lett. 759 (2012) L42. A. Pontzen and F. Governato, Nature 506 (2014) 171–178. A. Brooks, Annalen der Physik (2014) arXiv: 1407.7544. M. Rocha, A. Peter, J. Bullock, M. Kaplinghat, S. Garrison-Kimmel, J. Onorbe and L. Moustakas, Mon. Not. R. Astron. Soc. 430 (2013) 81. D. N. Spergel and P. J. Steinhardt, Phys. Rev. Lett. 84 (2000) 3760. A. H. G. Peter, M. Rocha, J. S. Bullock and M. Kaplinghat, Mon. Not. R. Astron. Soc. 430 (2013) 105–120. A. Loeb and N. Weiner, arXiv:1011.6374 [astro-ph.CO]. M. Vogelsberger, J. Zavala and A. Loeb, Mon. Not. R. Astron. Soc. 423 (2012) 3740– 3752.

May 4, 2017 9:55 b2167-ch14

II-42

47. 48. 49. 50. 51.

M. Davis

R. D. Blandford and R. Narayan, Ann. Rev. Astron. Astrophys. 30 (1992) 311–358. D. Walsh, R. F. Carswell and R. J. Weymann, Nature 279 (1979) 381–384. M. Markevitch et al., Astrophys. J. 606 (2004) 819. C. S. Kochanek and N. Dalal, Astrophys. J. 610 (2004) 69–79. Y. Hezaveh, N. Dalal, G. Holder, M. Kuhlen, D. Marrone, N. Murray and J. Vieira, Astrophys. J. 767 (2013) 9–18.

May 4, 2017 9:55 b2167-ch15

Chapter 15 Physics of the cosmic microwave background anisotropy

Martin Bucher Laboratoire APC, Universit´ e Paris 7/CNRS, Bˆ atiment Condorcet, Case 7020, 75205 Paris Cedex 13, France Astrophysics and Cosmology Research Unit and School of Mathematics, Statistics and Computer Science, University of KwaZulu-Natal, Durban 4041, South Africa [email protected]

Observations of the cosmic microwave background (CMB), especially of its frequency spectrum and its anisotropies, both in temperature and in polarization, have played a key role in the development of modern cosmology and of our understanding of the very early universe. We review the underlying physics of the CMB and how the primordial temperature and polarization anisotropies were imprinted. Possibilities for distinguishing competing cosmological models are emphasized. The current status of CMB experiments and experimental techniques with an emphasis toward future observations, particularly in polarization, is reviewed. The physics of foreground emissions, especially of polarized dust, is discussed in detail, since this area is likely to become crucial for measurements of the B modes of the CMB polarization at ever greater sensitivity.

1. Observing the Microwave Sky: A Short History and Observational Overview In their 1965 landmark paper A. Penzias and R. Wilson,153 who were investigating the origin of radio interference at what at the time were considered high frequencies, reported a 3.5 K signal from the sky at 4 GHz that was “isotropic, unpolarized, and free from seasonal variations” within the limits of their observations. Their apparatus was a 20 foot horn directed at the zenith coupled to a maser ampliﬁer and a radiometer (see Fig. 1). The maser ampliﬁer and radiometer were switched between the sky and a liquid helium cooled reference source used for comparison. Alternative explanations such as ground pick-up from the sidelobes of their antenna were ruled out in their analysis, and they noted that known radio sources would contribute negligibly at this frequency because their apparent temperature falls rapidly with frequency.

II-43

May 4, 2017 9:55 b2167-ch15

II-44

M. Bucher

Fig. 1. Horn antenna used in 1964 by Penzias and Wilson to discover the CMB. (Credit: NASA image)

In a companion paper published in the same issue of the Astrophysical Journal, Dicke et al.46 proposed the explanation that the isotropic sky signal seen by Penzias and Wilson was in fact emanating from a hot big bang, as had been suggested in the 1948 paper of Alpher et al.6 suggesting the presence of the photon blackbody component having a temperature of approximately a few K. Their prediction was based on considering the conditions required for successful nucleosynthesis in an expanding universe — that is, to create an appreciable fraction of primordial helium from the neutrons that are decaying as the universe is expanding. The importance of this discovery became almost immediately apparent, and others set out to better characterize this excess emission, which later would become known as the Cosmic Microwave Background (CMB), or sometimes in the older literature the Cosmic Microwave Background Radiation (CMBR). The two principal questions were: (i) To what extent is this background isotropic? (ii) How close is the spectrum to a perfect Planckian blackbody spectrum? The main obstacle to answering these questions was the lack of adequate instrumentation, and this was the main reason why the ﬁrst detection of the CMB anisotropy had to wait until 1992, when the COBE team announced their observation of a statistically signiﬁcant anisotropy of primordial origin after the dipole due to our motion with respect to the CMB had been subtracted205 (see Figs. 2 and 3). The COBE satellite, in a low-Earth orbit, carried three instruments: the diﬀerential microwave radiometer (DMR),204 the far infrared absolute spectrophotometer (FIRAS),133 and the diﬀuse infrared background experiment (DIRBE). By today’s standards, the measurement of the cosmic microwave background anisotropy was crude. The angular resolution was low — the width of the beam was 7◦ (FWHM),

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-45

Fig. 2. The microwave sky as seen by the COBE DMR (diﬀerential microwave radiometer) instrument. The top panel shows the microwave sky as seen on a linear temperature scale including zero. No anisotropies are visible in this image, because the CMB monopole at 2.725 K dominates. In the middle panel, the monopole component has been subtracted. Apart from some slight contamination from the galaxy near the equator (corresponding to the plane of our Galaxy), one sees only a nearly perfect dipole pattern, owing to our peculiar motion with respect to the rest frame deﬁned by the CMB. In the bottom panel, both the monopole and dipole components have been removed. Except for the galactic contamination around the equator, one sees the cosmic microwave background anisotropy along with some noise. (Credit: NASA/COBE Science Team) (For color version, see page II-CP5.)

and the sky map used for the analysis was smoothed to 10◦ to suppress beam artefacts. The contribution from instrument noise was large by contemporary standards. The COBE noise was 43, 16, and 22 mK Hz−1/2 for the 31, 53, and 90 GHz channels, respectively.21,a Nevertheless, COBE did provide a convincing ﬁrst detection of the CMB anisotropy, and most importantly established the overall level of the primordial cosmological density perturbations, which played a crucial role in determining the viability of the cosmological models in vogue at that time, which were much more numerous and varied than today. comparison, for WMAP the detector sensitivities were 0.7, 0.7, 0.9, 1.1, and 1.5 mK · Hz−1/2 for the K, Ka, Q, V, and W bands, respectively.98 a By

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-46

Fig. 3. COBE DMR individual frequency maps (with monopole and dipole components removed). Taking data at several frequencies is key to proving the primordial origin of the signal and removing galactic contaminants. Here are the three frequency maps from the COBE DMR observations. (Credit: NASA/COBE Science Team) (For color version, see page II-CP5.)

The COBE detection was followed by an intense eﬀort to characterize the CMB anisotropy at greater sensitivity and on smaller angular scales. There were numerous experiments from the ground at locations where the column density of water in the atmosphere is particularly low, such as Saskatoon, the Atacama Desert in Chile, the South Pole, and the Canary Islands, as well as from stratospheric balloons. Figure 4 shows the state of play about four years after COBE, with two competing theoretical models plotted together with the data points available at that time. In Sec. 6 we shall present the physics of the CMB systematically, but jumping ahead a bit, we give here a few words of explanation for understanding this plot. In most theoretical models, the CMB is generated by an isotropic Gaussian stochastic process, in which case all the available information concerning the underlying theoretical model can be extracted by measuring the angular power spectrum of the CMB anisotropies. Because of isotropy, one may expand the map in spherical harmonics to extract its angular power spectrum, deﬁned as CT T,obs =

+ 1 |aTm |2 . (2 + 1) m=−

(1)

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-47

Fig. 4. State of CMB observations in 1998. After the COBE DMR detection at very large angular scales at the so-called Sachs–Wolfe plateau, numerous groups sought to discover the acoustic peaks, or to ﬁnd their absence. This compilation from 1998 shows the state of play at that time. While an unmistakable rise toward the ﬁrst acoustic peak is apparent, it is not so apparent what happens toward higher . Two theories broadly compatible with the data at that time are plotted, one with Ωk = 0 and another with Ωk ≈ 0.7, which predicts a higher amplitude of the primordial perturbations and an acoustic peak shifted to smaller angular scales, as explained in Sec. 7.3. (Credit: Max Tegmark) (For color version, see page II-CP5.)

It is customary to plot the quantity ( + 1)C /(2π), which would be constant for a scale invariant pattern on the sky.b This is the quantity plotted in Fig. 4, where the question posed at the time was whether there is a rise to a ﬁrst acoustic peak followed by several decaying secondary acoustic peaks, as indicated by the solid theoretical curves. This structure is a prediction of simple inﬂationary models — or more precisely, of models where only the adiabatic growing mode is excited with an approximately scale invariant primordial spectrum. In this plot, one sees fairly convincing evidence for a rise in the angular power, but the continuation of the curve is unclear. Several experiments contributed to providing the ﬁrst clear detection of the acoustic oscillations, namely TOCO,134 MAXIMA,77 and Boomerang.15,140 speaking, scale invariance on the sky can be deﬁned only asymptotically in the → ∞ limit, because the curvature of the celestial sphere breaks scale invariance. The use of ( + 1) rather than 2 here is an historical convention. b Strictly

May 4, 2017 9:55 b2167-ch15

II-48

M. Bucher

Fig. 5. Boomerang observation of the acoustic oscillations. The CT T power spectrum as measured by Boomerang is shown on the left without a ﬁt to a theoretical model and on the right with the theoretical predictions for a spatially ﬂat cosmological model with an exactly scale invariant primordial power spectrum for the adiabatic growing mode. (Credit: Boomerang Collaboration)

Figure 5 shows the power spectrum as measured by one of these experiments, namely Boomerang, where a series of well deﬁned acoustic oscillations is clearly visible. In the meantime, parallel eﬀorts were underway in Europe and in the United States to prepare for another CMB space mission to follow on COBE at higher sensitivity and angular resolution. The COBE beam, which was 7◦ FWHM, did not use a telescope but rather microwave feed horns pointed directly at the sky. While not allowing for a high angular resolution, the feed horns had the advantage of producing well deﬁned beams with rapidly falling sidelobes. The US NASA WMAP satellite, launched in 2001, delivered its one-year data release in 2003 (including TT and TE),216–234 and its ﬁrst polarization data (including also EE) in 2006235–238 (see Figs. 6 and 7). WMAP continued taking data for nine years, and released installments of papers based on the ﬁve-, seven-, and nine-year data releases, in which the results were further reﬁned beneﬁtting from a longer integration time √ (which nominally would shrink error bars in proportion to 1/ tobs ) as well as improved instrument modeling.239–251 WMAP used horns pointed at a 1.4 m×1.6 m oﬀ-axis Gregorian mirror to obtain diﬀraction limited beams (although the mirror was under-illuminated somewhat to reduce far sidelobes). The 20 detectors were based on coherent ampliﬁcation using HEMTs (high electron mobility transistors) — the state of the art in coherent low-noise ampliﬁcation at the time. The coherent detection technology used by WMAP had the advantage that the electronics could be passively cooled. The competing incoherent bolometric detection technology permits superior sensitivity, which operates almost at the quantum noise

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-49

Fig. 6. WMAP single temperature frequency maps. WMAP observed in ﬁve frequency bands. (Credit: NASA/WMAP Science Team) (For color version, see page II-CP6.)

limit of the incident photons, but requires cooling to approximately 100 mK. The cryogenic system required was judged to present substantial risk for a space mission at the time, and this was likely one of the reasons why WMAP was selected over one of its competitor missions, which was similar to the European Planck HFI proposal. The successor to WMAP was the European Space Agency (ESA) Planck satellite, which was launched in May 2009.155–181 Planck consisted of two instruments: (i) a low-frequency instrument with three channels (at 30, 44, and 70 GHz) using a coherent detection technology, and (ii) a high-frequency instrument using cryogenically-cooled bolometric detectors observing in six bands (100, 143, 217, 353, 545, and 857 GHz. All bands except the highest two bands were polarization sensitive and had a diﬀraction limited angular resolution, starting at 33 arcmin for the 30 GHz channel and going down to 5.5 arcmin for the 217 GHz channel. (The top three channels have an angular resolution of approximately 5’.) Planck HFI took data from August 2009 to January 2012, when the coolant for its high-frequency instrument ran out. The Planck Collaboration reported its ﬁrst results for cosmology in March 2013 based on its temperature anisotropy data. The ﬁrst results for cosmology using the polarization data collected by Planck were delivered in February 2015.

May 4, 2017 9:55 b2167-ch15

II-50

M. Bucher

Fig. 7. WMAP single frequency polarization maps. The lines (which may be thought of as doubleheaded vectors) show the amplitude and orientation of the linear polarization of the measured p CMB anisotropy in the indicated bands. The amplitude P = Q2 + U 2 is also shown using the indicated color scale. (Credit: NASA/WMAP Science Team) (For color version, see page II-CP7.)

Figure 8 shows the Planck full-sky maps of the intensity of the microwave emission in nine frequency bands ranging from 30 GHz to 857 GHz. Figure 9 shows a cleaned full-sky map where a linear combination of the single band maps has been taken in order to isolate the primordial cosmic microwave background signal. While the ﬂuctuations in the cleaned map in Fig. 9 do not appear to single out any particular direction in the sky and appear consistent with an isotropic Gaussian random process, the maps in Fig. 8 show a clear excess in the galactic plane. These full-sky maps use a Molleweide projection in galactic coordinates, with the galactic center at the center of the projection. Even though the galactic contamination depends largely on the angle to the galactic plane, with a lesser tendency to increase toward the galactic center, considerable structure can be seen in the galactic emission over a broad range of angular scales. The central bands include the least amount of galactic contamination, which visibly is much larger for the lowest and the highest frequencies shown. The CMB dipole amplitude is ∆T = 3.365 ± 0.027 mK and directed toward (l, b) = (264.4◦ ± 0.3◦ , 48.4 ± 0.5◦ ) in galactic coordinates.104 The maps have been processed to remove the 2.725 K CMB monopole component as well as the smaller

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-51

Fig. 8. Planck single frequency temperature maps. The ESA Planck satellite surveyed the sky in nine broad [i.e., (∆ν)/ν ≈ 0.3] frequency bands, centered at 30, 44, 70, 100, 143, 217, 353, 545, and 857 GHz, shown in galactic coordinates. The units are CMB thermodynamic temperature [see Eq. (3)]. The nonlinear scale avoids saturation in regions of high galactic emission. (Credit: ESA/Planck Collaboration) (For color version, see page II-CP8.)

CMB dipole having a peak-to-peak amplitude of approximately 6.73 mK, so that only the spherical harmonic multipoles with ≥ 2 are included. CMB angular power is typically expressed in terms of D = ( + 1)C /2π, which in the ﬂat sky approximation corresponds to rms power (in µK2 ) per unit logarithmic interval in the spatial frequency. A scale invariant temperature spectrum on the celestial sphere would correspond to 2 C being constant. For comparison we give here a

May 4, 2017 9:55 b2167-ch15

II-52

M. Bucher

Fig. 9. Planck CMB map obtained from combining the single-frequency maps. A linear combination of the Planck single frequency maps (shown in Fig. 8) is taken. This linear combination is optimized to ﬁlter out any unwanted contaminants based on their diﬀering frequency dependence. The success of this procedure to remove the galactic contaminants is evident from the disappearance of excess power along the equator corresponding to the galactic plane. (For more details, see Ref. 161.) (Credit: ESA/Planck Collaboration) (For color version, see page II-CP9.)

few ballpark numbers characterizing the strength of the primordial CMB temperature anisotropy. Figure 10 plots the CMB power spectrum as observed by Planck together with the predictions of the ﬁt to a six-parameter theoretical model, which we will discuss further below. The magnitude at low , before the rise to the ﬁrst acoustic, or Doppler, peak at ≈ 220, is D ≈ 103 µK2 , which would correspond to an rms temperature in the neighborhood of 30 µK. The rise to the ﬁrst acoustic peak increases the power by about a factor of six, and then after approximately ﬁve oscillations, damping eﬀects take over, making the oscillations less apparent and causing the spectrum to suﬀer a quasi-exponential decay. Although the detectors (to the extent that their response is ideal or linear) measure the intensity expressed as the “spectral radiance” or “speciﬁc intensity” Iν (having units of erg s−1 cm−2 Hz−1 str−1 ) averaged over a frequency band deﬁned by the detector, it is convenient to re-express these intensities in terms of an eﬀective temperature. There are two types of eﬀective temperature: the Rayleigh– Jeans (R–J) temperature and the thermodynamic temperature. It is important to keep in mind the distinction, especially at high frequencies [in this context ν > ∼ νCMB = h−1 (kB TCMB ) = 57 GHz] where the factor relating the two becomes large. In the R–J limit, where (hν/kB T ) 1, the blackbody spectral radiance may be approximated as Iν = Bν (T ) ≈ 2(ν/c)2 (kB T ), as one would obtain classically from the density of states of the radiation ﬁeld and assigning an energy (kB T ) to each

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-53

Fig. 10. CMB angular power spectrum as measured by Planck. The binned CT T power spectrum as given in the Planck 2013 release is plotted with error bars that combine uncertainties from cosmic variance (dominant at low ) and instrumental noise (dominant at high ). The solid curve indicates the theoretical predictions of the six-parameter concordance model with the 1σ cosmic variance for the adopted binning scheme. The lower panel shows the residuals. For more details, see Ref. 166. The exquisite ﬁt, seen here up to about ≈ 2200, into the damping tail, has been shown to extend to smaller angular scales by ACT58 and SPT.102 (Credit: ESA/Planck Collaboration)

harmonic oscillator degree of freedom. If we invert, assuming the R–J limit above (regardless of whether this limit is valid), we obtain the following deﬁnition for the R–J temperature 1 1 c 2 TR−J (ν) = Iν . (2) 2 kB ν The “thermodynamic” temperature corresponding to a speciﬁc intensity Iν at a certain frequency, on the other hand, is obtained by inverting the unapproximated blackbody expression Iν = Bν (T ) = (2hν 3 /c2 )[exp(hν/kB T ) − 1]−1 . For the CMB, where the variations about the average CMB temperature are small (i.e., ∆T /T ≈ 10−5 ), linearized perturbation theory is valid and one can invert this equation, obtaining δTCMB =

(ex − 1)2 δTR−J , x2 ex

(3)

where x = hν/kB T. This factor is approximately one for x < ∼ 1, but rises exponen> tially for x ∼ 1. It should be remembered that for adding optically thin emissions, it is the intensities, and not the thermodynamic temperatures, that add. Moreover, foreground

May 4, 2017 9:55 b2167-ch15

II-54

M. Bucher

emission power laws know nothing about TCMB and thus are naturally expressed using either the speciﬁc intensity, or equivalently the R–J temperature. Other important milestones of CMB observation include the DASI discovery of the polarization of the CMB in 2002111 and the observations at small angular scales carried out by the ACT and SPT teams, which at the time of writing constitute the best measurements of the microwave anisotropies on large patches of the sky at high angular resolution, measuring the power spectrum up to ≈ 10,000. The DASI experiment was carried out using several feed horns pointed directly at the sky. Correlations were taken of the signals from pairs of horns, thus measuring the sky signal interferometrically. The 6 m diameter Atacama Cosmology Telescope (ACT) situated in the Atacama desert in Chile58,109 and the 10 m SPT (South Pole Telescope) located at the Amundsen–Scott South Pole Station in Antarctica102,190,194 probe the microwave sky on very small angular scales, beyond the angular resolution of Planck. CMB observations are almost always diﬀraction limited, so that the angular resolution is inversely proportional to the telescope diameter. Ground based instruments, despite all their handicaps (i.e., atmospheric interference, lack of stability over long timescales, ground pickup from far sidelobes) will always outperform space based experiments in angular resolution. There are dozens of other suborbital experiments not mentioned here but which paved the road for contemporary CMB observation. These experiments were important not only because of their observations, but also because of their role in technology development and in the development of new data analysis techniques. More information on these experiments can be found in other earlier reviews.c 2. Brief Thermal History of the Universe The big bang model of the universe is an unﬁnished story. Certain aspects of the big bang model are well established observationally, while other aspects are more provisional and represent our present best bet speculation. In the account below, we endeavor to distinguish what is relatively certain and what is more speculative. If we assume: (i) the correctness of general relativity, (ii) that the universe is homogeneous and isotropic on large scales (at least up to the size of that part of the universe presently observable to us), and (iii) that for calculating the behavior of the universe, small-scale anisotropies can be averaged over, we obtain the Friedmann–Lamaˆıtre–Robertson–Walker (FLRW) solutions to the Einstein ﬁeld equations, which we now describe. The metric for this family of solutions takes the following form: ds2 = −dt2 + a2 (t)γij dxi dxj ,

(4)

c See http://lambda.gsfc.nasa.gov/links/experimental sites.cfm for an extensive compilation of CMB experiments with links to their websites. The book147 provides an insightful account emphasizing the early history of CMB observations with contributions from many of the major participants.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-55

where (i, j = 1, 2, 3) and the line element d2 = γij dxi dxj describes a maximally symmetric three-dimensional space that may be Euclidean (ﬂat), hyperbolic, or spherical.d For the ﬂat case, which is the most important and most often discussed, Eq. (4) takes the form ds2 = −dt2 + a2 (t)[dx21 + dx2 2 + dx3 2 ].

(5)

Here a(t) represents the scale factor of the universe. Under the assumption of homogeneity and isotropy, the stress–energy tensor (expressed as a mixed tensor with one contravariant and one covariant index) must take the form ρ(t) 0 0 0 0 p(t) 0 0 . Tµ ν = (6) 0 0 p(t) 0 0 0 0 p(t) The Einstein ﬁeld equations Gµν = Rµν − (1/2)gµν R = (8πG)Tµν may be reduced to the two equations H 2 (t) ≡

a˙ 2 (t) 8πGN k = ρ(t) − 2 2 a (t) 3 a (t)

(7)

and a ¨(t) 4πG =− (ρ + 3p). a(t) 3

(8)

(For more details, see for example the discussion in the books by Weinberg221 or Wald.218 ) The above discussion, which included only general relativity, is incomplete because no details concerning the dynamics of ρ(t) and p(t) have been given. The only constraint imposed by general relativity is stress–energy conservation, which in the most general case is expressed as Tµν ;ν = 0, but in the special case above takes the form dρ(t) = −3H(t)[ρ(t) + p(t)]. dt d More

(9)

generally, the three-dimensional line element ds2 = γij dxi dxj = „

dx1 2 + dx2 2 + dx3 2 „ « «2 , k 1+ (x1 2 + x2 2 + x3 2 ) 4

where k = +1, 0, or −1 corresponds to a spherical, ﬂat (Euclidean), or hyperbolic threedimensional geometry, respectively. Apart from an overall change of scale, these are the only geometries satisfying the hypotheses of spatial homogeneity and isotropy. By substituting tan(χ/2) = r/2 or tanh(χ/2) = r/2 for the cases k = +1 or k = −1, we may obtain the more familiar representations ds2 = dχ2 + sin χ2 dΩ(2) 2 or ds2 = dχ2 + sinh χ2 dΩ(2) 2 , respectively, where dΩ(2) 2 = dθ 2 + sin θ 2 dφ2 .

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-56

The relationship between ρ and p is known as the equation of state, and for the special case of a perfect ﬂuid p depends only on ρ and on no other variables. Two special cases are of particular interest: a nonrelativistic ﬂuid, for which p/ρ = 0, and an ultrarelativistic ﬂuid, for which p/ρ = 1/3. Until the mid-90s it was commonly believed that except at the very beginning of the universe, a two-component ﬂuid consisting of a radiation component and matter component suﬃces to account for the stress–energy ﬁlling the universe. In such a model

ρ(t) = ρm,0

a0 a(t)

3 + ρr,0

a0 a(t)

4 .

(10)

Here the subscript 0 refers to the present time. For the radiation component, there is obviously the contribution from the CMB photons, whose discovery was recounted above in Sec. 1. If we assume that the universe started as a plasma at some very large redshift with a certain baryon number density, most conveniently parameterized as a baryon-to-photon or baryon-to-entropy ratio, we would conclude that there was statistical equilibrium early on between the various species, and from this assumption we can calculate the neutrino density today. It turns out that the neutrinos are colder than the photons because the neutrinos fell out of statistical equilibrium before the electrons and positrons annihilated. The e+ e− annihilation had the eﬀect of heating the photons relative to the neutrinos, because all the entropy that was in the electrons and positrons was dumped into the photons. Consequently, assuming a minimal neutrino sector, we can calculate the number of eﬀective bosonic degrees of freedom due to the cosmic neutrino background, which according to big bang theory must be present but has never been detected. Measuring the number of baryons in the universe today is more diﬃcult, and is the subject of an extensive literature. The realization of a need for some additional nonbaryonic “dark” matter dates back to Fritz Zwicky’s study of the Coma cluster in 1933 and is a long story that we do not have time to go into. Today “cold dark matter” is part of the concordance model, where the name simply means that whatever the details of this extra component may be (the lightest supersymmetric partner of the ordinary particles in the standard electroweak model, the axion, or yet something else), for analyzing structure formation in the universe, we can treat this component as nonrelativistic (cold) particles that can be idealized as interacting only through gravitation. In the mid-1990s it was realized that a model where the stress–energy included only matter and radiation components could not account for the observations. A crucial observation for many researchers was the measurement of the apparent luminosities of Type Ia supernovae as a function of redshift,154,192 although at that time there were already several other discordant observations indicating that the universe with only matter and radiation could not account for the observations.e To e See

for example some of the contributions in Ref. 213 and references therein.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-57

reconcile the observations with theory, another component having a large negative pressure, such as would arise from a cosmological constant, was required. A cosmological constant makes a contribution to the stress–energy Tµν proportional to the metric tensor gµν meaning that p = −ρ exactly. This relation inserted into the right-hand side of Eq. (9) implies that any expansion (or contraction) does not change the contribution to the density from the cosmological constant. With a cosmological constant, the density of the universe scales in the following manner:

0

3 4 a0 a0 a0 + ρm,0 ρ(t) = ρΛ,0 + ρr,0 . (11) a(t) a(t) a(t) The cosmological constant is most important at late times, when its density rapidly comes to dominate over the contribution from nonrelativistic matter (consisting according to our best current understanding of baryons and a so-called cold dark matter component). It is customary to express the contributions of each component as a fraction of the critical density ρcrit = 3H 2 /8πG, which would equal the total density if the universe were exactly spatially ﬂat. For the ith component, we deﬁne Ωi = (ρi /ρcrit). From the present data, we do not know whether the new component (or components) with a large negative pressure is a cosmological constant, although current observations indicate that w = p/ρ for this component must lie somewhere near w = −1. A key question of many of the major initiatives of contemporary observational cosmology is to measure w (and its evolution in time) in order to detect a deviation of the behavior of the dark energy from that predicted by a cosmological constant. The earliest probe of the hot big bang model arises from comparing the primordial light element abundances as inferred from observations to the theory of primordial nucleosynthesis. In a hot big bang scenario, there is only one free parameter the baryon-to-photon ratio ηB , or equivalently, the baryon-to-entropy ratio ηS , which is related to the former by a constant factor. As discussed above, in the hot big bang theory at early times, in the limit as a(t) → 0, we have ρ → ∞, and T → ∞. For calculating the primordial light element abundances, we do not need to extrapolate all the way back to the initial singularity, where these quantities diverge. Rather it suﬃces to begin the calculation at a temperature suﬃciently low so that all the baryon number of the universe is concentrated in nucleons (rather than in free quarks), but high enough so that there are only protons and neutrons rather than bound nuclei consisting of several nucleons, and so that the reaction p + e− n + νe ,

(12)

(as well as related reactions) proceeds at a rate faster than the expansion rate of the universe H. Under this assumption, the ratio of protons to neutrons is determined

May 4, 2017 9:55 b2167-ch15

II-58

M. Bucher

by an equilibrium condition expressed in terms of chemical potentials µ(p) + µ(e− ) = µ(n) + µ(νe ) + ∆M,

(13)

where ∆M = MN − MP − Me = 939.56 MeV − 938.27 MeV − 511 KeV = 0.78 MeV. (14) At ﬁrst, when T (∆M ), the protons and neutrons have almost precisely equal abundances, but then, as the universe cools down and expands, the neutrons become less abundant than the protons owing to the neutron’s slightly greater mass. Then the reactions (12) freeze out — that is, their rate becomes small compared to H — and the neutron–proton ratio becomes frozen in. Subsequently protons and neutrons fuse to form deuterium, almost all of which combines to form 4 He, either directly from a pair of deuterons or through ﬁrst forming 3 He, which subsequently fuses with a free neutron. Almost all the neutrons that do not decay end up in the form 4 He, because of its large binding energy relative to other low A nuclei, but trace amounts of other light elements are also produced. Even though other heavier nuclei have a larger binding energy per nucleon and their production would be favored by equilibrium considerations, the rates for their production are too small, principally because of the large Coulomb barrier that must be overcome. In primordial big bang nucleosynthesis (BBN) the only free parameter is ηB . Since nucleosynthesis takes place far into the epoch of radiation domination, the density as a function of temperature is determined by the equation ρ(T ) = (π 2 /30)Nrel T 4 where Nrel is the eﬀective number of bosonic degrees of freedom. The expansion rate then is important because it determines for example how many neutrons are able to be integrated into nuclei heavier than hydrogen before decaying. Based on the particles known to us that are ultrarelativistic during nucleosynthesis, the photon, electron, positron, neutrinos, and antineutrinos, we think that we know what value Nrel should have. But nucleosynthesis can be used to test for the presence of extra relativistic degrees of freedom. We shall see later that the CMB can also be used to constrain Nrel . For a review with more details about primordial nucleosynthesis, see Ref. 219. 3. Cosmological Perturbation Theory: Describing a Nearly Perfect Universe Using General Relativity The broad brush account of the history of the universe presented in the previous section tells an average story, where spatial homogeneity and isotropy have been assumed. As we shall see, early on, at large z, this story is not far from the truth, especially on large scales. But a universe that is exactly homogeneous and isotropic would be quite unlike our own, in that there would be no clustering of matter, observed very concretely in the form of galaxies, clusters of galaxies, and so forth.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-59

In this section, we present a brief sketch of the theory of cosmological perturbations. For excellent reviews with a thorough discussion of the theory of cosmological perturbations in the framework of general relativity, see Refs. 103 and 139. For the “scalar” perturbations,f we may write the line element for the metric with its linearized perturbations Φ(x, η) and Ψ(x, η) in the following form: ds2 = a2 (η)[−(1 + 2Φ)dη 2 + (1 − 2Ψ)dx2 ].

(15)

The functions Ψ(x, η) and Φ(x, η) are the Newtonian gravitational potentials. At low velocities Φ(x, η) is most relevant, but Ψ(x, η) can be probed observationally using light or some other type of ultrarelativistic particles. When there are no anisotropic stresses (i.e., all the partial pressures of the stress–energy tensor are equal), Φ = Ψ. For a universe ﬁlled with a perfect ﬂuid with an equation of state p = p(ρ), the evolution of the Newtonian gravitational potential Φ(x, η) is governed by the equation (see Ref. 139 for a derivation and more details) Φ + 3H(1 + cs 2 )Φ − cs 2 ∇2 Φ + [2H + (1 + 3cs 2 )H2 ]Φ = 0,

(16)

where cs 2 = dp/dρ, H = a /a, and = d/dη. Here η is the conformal time, related to the more physical proper time by the relation dt = a(η)dη. If the spatial derivative term (i.e., ∇2 Φ) is neglected, which is an approximation valid on superhorizon scales, then the derived quantity10,83 ζ =Φ+

2 (Φ + H−1 Φ ) 3 (1 + w)

(17)

(where w = p/ρ) is conserved on superhorizon scales. This approximate invariant, which can be related to the spatial curvature of the surfaces of constant density, is invaluable for relating perturbations at the end of inﬂation (or some other very early noninﬂationary epoch) to the later times of interest in this chapter. This property is particularly useful given our ignorance of the intervening epoch, the period of “reheating” or “entropy generation,” the details of which are unknown and highly model dependent. To understand the qualitative behavior of the above equation, it is important to focus on the role of the Hubble parameter H(t) = a(t)/a(t), ˙ which at time t deﬁnes the dividing line between superhorizon modes [for which |kphys | < ∼ H(t)] and subg H(t)]. This distinction relates to the relevance horizon modes [for which |kphys | > ∼ f In

the theory of cosmological perturbations, the terms “scalar,” “vector,” and “tensor” are given a special meaning. A tensor of any rank that can be constructed using derivatives acting on a scalar function and the Kronecker delta multiplying a scalar function is regarded as a “scalar”. A quantity that is not a “scalar” but can be constructed in the same way from a 3-vector ﬁeld is regarded as “vector,” and a “tensor” is a second-rank tensor with its “scalar” and “vector” components removed. Under these deﬁnitions, the linearized perturbation equations reduce to a block diagonal form in which the “scalar,” “vector,” and “tensor” sectors do not talk to each other. g Here “horizon” means “apparent horizon,” which depends only on the instantaneous expansion rate, unlike the “causal horizon,” which depends on the entire previous expansion history.

May 4, 2017 9:55 b2167-ch15

II-60

M. Bucher

of the spatial gradient terms in the evolution equations for the cosmological perturbations. This same distinction may be re-expressed in terms of the conformal time ˙ η and the co-moving wavenumber k, where H = a /a takes the place of H = a/a, > H(t). Here with superhorizon meaning |k| < H(t) and subhorizon meaning |k| ∼ ∼ dots denote derivatives with respect to proper time whereas primes denote derivatives with respect to conformal time. It turns out that for calculations, working with conformal time and co-moving wavenumbers is much more convenient than working with physical variables. What happens to the perturbations as the universe expands depends crucially on whether the co-moving horizon size is expanding or contracting. During inﬂation, or more generally when w < −1/3, the co-moving horizon size is contracting. This means that the dynamics of Fourier modes, which initially hardly felt the expansion of the universe, become more and more aﬀected by the expansion of the universe. Terms proportional to H and H 2 in the evolution equations are becoming increasingly relevant whereas the spatial gradient terms are becoming increasingly irrelevant as the mode “exits” the horizon, ﬁnally to become completely “frozen in” on superhorizon scales. During inﬂation a mode starts far within the horizon and crosses the horizon at some moment during inﬂation. At the end of inﬂation, all the modes of interest lie frozen in, far outside the horizon. During inﬂation, w = p/ρ was slightly more positive than −1. Formally inﬂation ends when w crosses the milestone w = −1/3, which is when the co-moving horizon stops shrinking and starts to expand. This moment roughly corresponds to the onset of “reheating” or “entropy generation,” when the vacuum energy of the inﬂaton ﬁeld gets converted into radiation, or equivalently ultrarelativistic particles, which afterward presumably constitute the dominant contribution to the stress–energy Tµ ν , so that w ≈ 1/3. During the radiation epoch following “entropy generation,” modes successively re-enter the horizon again becoming dynamical. The degrees of freedom describing the modes are not the same as previously during inﬂation. This is at present our best-bet story of what likely happened in the very early universe. In this chapter, we shall be interested in the later history of the universe, when modes are re-entering the horizon. We shall not enter into the details of how the primordial perturbations are generated from quantum ﬂuctuations of the vacuum during inﬂation, instead referring the reader to the chapter by K. Sato and J. Yokoyama on Inﬂationary Cosmology for this early part of the story. Nor shall we discuss alternatives to inﬂation. The original papers in which the scalar perturbations generated from inﬂation were ﬁrst calculated include Mukhanov,136–138 Hawking,86 Guth,75 Starobinsky,209 and Bardeen et al.11 For early discussion of the generation of gravity waves during inﬂation and their subsequent imprint on the CMB, see Refs. 258–260, 3, and 261. A more pedagogical account may be found for example in the books by Liddle and Lyth84 and by Peacock,146 as well as in several review articles, including for example Ref. 126.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-61

We emphasize the simplest type of cosmological perturbations, known as primordial “adiabatic” perturbations. The fact that the perturbations are “primordial” means that at the late times of interest to us, only the growing adiabatic mode is present, because whatever the amplitude of the decaying mode may initially have been, suﬃcient time has passed for its amplitude to decay away and thus become irrelevant. The term “adiabatic” deserves some explanation because its customary meaning in thermodynamics does not precisely correspond to how it is used in the context of cosmological perturbations. In a universe with only adiabatic perturbations, at early times all the components initially shared a common equation of state and common velocity ﬁeld, and moreover on surfaces of constant total density, the partial densities of the components are also spatially constant. This is the state of aﬀairs initially, on superhorizon scales, but subsequently after the modes enter the horizon, the diﬀerent components can and do separate. Adiabatic modes are the simplest possibility for the primordial cosmological perturbations, but they are not the only possibility. Isocurvature perturbations where the equation of state varies spatially are also possible, as discussed in detail for example in Refs. 22, 28 and 257 references therein. For isocurvature perturbations, the ratios of components vary on a constant density surface, on which the spatial curvature is constant as well. However the ratios between the components are allowed to vary. As the universe expands, these variations in the equation of state eventually also generate curvature perturbations. It might be argued that the isocurvature modes should generically be expected to have been excited with some nonvanishing amplitude in multi-ﬁeld models of inﬂation. For recent constraints on isocurvature perturbations from the CMB, see Planck Inﬂation 2013171 and the extensive list of references therein. 4. Characterizing the Primordial Power Spectrum In the previous section, we deﬁned the dimensionless invariant ζ(x), conserved on superhorizon scales, which can be used to characterize the primordial cosmological perturbations at some convenient moment in the early universe when all the relevant scales of interest lie far outside the horizon. Because this quantity is conserved on superhorizon scales, the precise moment chosen is arbitrary within a certain range. The existence of a quantity conserved on superhorizon scales allows us to make precise predictions despite our nearly complete ignorance of the details of reheating. Assuming a spatially ﬂat universe, we may expand ζ(x) into Fourier modes, so that d3 k ζ(k) exp[ik · x]. (18) ζ(x) = (2π)3 If spatial homogeneity and isotropy are assumed, the correlation function of the Fourier coeﬃcients must take the form ζ(k)ζ(k ) = (2π)3 k −3 P(k)δ 3 (k − k ),

(19)

May 4, 2017 9:55 b2167-ch15

II-62

M. Bucher

which also serves as the deﬁnition of the function P(k) known as the primordial power spectrum. The k −3 factor is present so that P(k) is likewise dimensionless. An exactly scale invariant power spectrum corresponds to P(k) ∼ k 0 . The meaning of scale invariance can be understood by writing the mean square ﬂuctuation at a point as the following integral over wavenumber: +∞ d[ln(k)]P(k). (20) ζ 2 (x = 0) = −∞

Changes of scale correspond to rigid translations in ln(k), and the only choice for P(k) that does not single out a particular scale is P(k) = (constant). 5. Recombination, Blackbody Spectrum, and Spectral Distortions It is sometimes said that the CMB is the best blackbody in the universe. This is not true. Observationally the only way to test how close the CMB is from a perfect blackbody is to construct a possibly better artiﬁcial blackbody and perform a diﬀerential measurement between the emission from the sky and this artiﬁcial blackbody as a function of frequency. In our discussion of the CMB, we stressed that theory predicted a frequency spectrum having a blackbody form to a very high accuracy. This indeed was one of the striking predictions of the big bang theory, and the lack of any measurable deviation from the perfect blackbody spectral shape over a broad range of frequencies was probably one of the most important observational facts that led to the demise of the alternative steady state cosmological model. To date, the best measurements of the frequency spectrum of the CMB are still those made by the COBE FIRAS instrument.65,131,132 FIRAS133 made diﬀerential measurements comparing the CMB frequency spectrum on the sky to an artiﬁcial blackbody. Despite the many years that have gone by since COBE and the importance of measuring the absolute spectrum of the CMB, no better measurement has been carried out because improving on FIRAS would require going to space in order to avoid the spectral imprint of the Earth’s atmosphere. However a space mission concept called PIXIE has been proposed that would essentially redo FIRAS but with over two orders of magnitude better sensitivity and with polarization sensitivity included.105 While diﬀerences in temperature between diﬀerent directions in the sky, particularly on small angular scales, can be measured from the ground, albeit with great diﬃculty, the same is not possible for measurements of the absolute spectrum. From a theoretical point of view, even in the simplest minimal cosmological model with no new physics, deviations from a perfect blackbody spectrum are predicted that could be measured with improved observations of the absolute spectrum having a sensitivity signiﬁcantly beyond that of FIRAS. It is often believed that when we look at the CMB, we are probing conditions in the universe at around last scattering — that is, around z ≈ 1100 — and that the CMB photons are

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-63

unaﬀected by what happened during signiﬁcantly earlier epochs. This is not true. Although when z > ∼ 1100, photons are frequently scattered by electrons randomizing their direction and causing them to move diﬀusively, this frequent scattering is ineﬃcient at equilibrating the kinetic temperature of the electrons to the energy spectrum of the photons and vice versa owing to the smallness of the dimensionless parameter Eγ /me , where Eγ is a typical photon energy. This ratio gives the order of magnitude of the fractional energy exchange due to electron recoil for a typical collision. Given that in this limit the approach to a blackbody spectrum is a diﬀusive process, we would estimate that (Eγ /me )2 collisions correspond to one decay time of the deviation of the photon energy spectrum from the equilibrium spectrum. Plugging in Eγ ≈ kB Tz=1100 ≈ 0.3 eV, we get (Eγ /me )2 ≈ 4 × 1010 ! Thomson scattering, however, cannot equilibrate the photon number density with the available energy density, so with Thomson scattering alone, an initially out of equilibrium photon energy distribution, for example as the result of some sort of energy injection, would generically settle down to a photon phase space distribution having a positive chemical potential. Processes such as Bremsstrahlung and its inverse or double Compton scattering are required to make the chemical potential decay to zero, and these processes freeze out at z ≈ 106 .97 Therefore any energy injection between z ≈ 106 and z ≈ 103 will leave its mark on the absolute spectrum of the CMB photons, either in the form of a so-called µ distortion, or for energy injected at later times, in the form of a more complicated energy dependence. A broad range of interesting early universe science can be explored by searching for deviations from a perfect blackbody spectrum. Some of the spectral distortions are nearly certain to be present, while others are of a more speculative nature. On the one hand, y-distortions in the ﬁeld (that is, away from galaxy clusters where the y-distortion is localized and particularly large) constitute a nearly certain signal, as do the spectral lines from cosmological recombination. But sources like decaying dark matter31 may or may not be present. Another interesting source of energy injection arise from Silk damping91,201 of the primordial power spectrum on small scales, beyond the range of scales where the primordial power spectrum can be probed by other means. If a space-based instrument is deployed with the requisite sensitivity, a formidable challenge will be to distinguish these signals from each other, and also from the more mundane galactic and extragalactic backgrounds. From this discussion one might conclude, borrowing terminology from stellar astrophysics, that the CMB “photosphere” extends back to around z ≈ 106 . 6. Sachs–Wolfe Formula and More Exact Anisotropy Calculations At early times (z < ∼ 1100) the universe is almost completely ionized and the photons scatter frequently with free charged particles, primarily with electrons because the proton–photon scattering cross-section is suppressed by a factor of (me /mp ). During this period the plasma consisting of baryons, leptons, and photons may be

May 4, 2017 9:55 b2167-ch15

II-64

M. Bucher

regarded as a single tightly-coupled ﬂuid component almost inviscid but with a viscosity relevant on the shortest scales of interest for the calculation of the CMB temperature and polarization anisotropies. Later, as the universe cools the electrons and protons (and other nuclei) combine to form neutral atoms whose size greatly exceeds the thermal wavelength of the photons, rendering the universe almost transparent. Subsequent to this process, known as “recombination” (despite the fact that the universe had never previously been neutral), the photons cease to rescatter and free stream toward us today.149 This last statement is only 90% accurate, because the late-time reionization starting at about z ≈ 6–7 causes about 10% of the CMB photons to be rescattered. The eﬀect of this rescattering on the CMB anisotropies will be discussed in detail in Sec. 7.5. It is during this transitional epoch, situated around z ≈ 1100, that the CMB temperature and polarization anisotropies were imprinted. Thus when we observe the microwave anisotropies on the sky today, we are probing the physical conditions on a sphere approximately 47 billion light years in radius today,h or more precisely the intersection of our past light cone at the surface of constant cosmic time at z ≈ 1100. This sphere is known as the surface of last scatter or the last scattering surface. The transition from a universe that is completely opaque to the blackbody photons, in which the tight-coupling approximation holds for the baryon–lepton plasma because the photon mean free path is negligible, to a universe that is virtually transparent is not instantaneous. This fact implies that the last scattering surface (LSS) is not inﬁnitely thin, rather having a ﬁnite width that must be taken into account for calculating the small angle CMB anisotropies, because this proﬁle of ﬁnite thickness smears out the small-scale three-dimensional inhomogeneities as they are projected onto the two-dimensional celestial sphere. Figure 11 shows a plot of the ionization history of the universe (not taking into account the late-time reionization at z > ∼ 6 mentioned above). (For a more detailed discussion of the early ionization history of the universe and the physics by which it is determined, see for example Refs. 87, 196, 197, 211 and 30). The ﬁrst calculation of the CMB temperature anisotropies predicted in a universe with linearized cosmological perturbations was given by Sachs and Wolfe in their classic 1967 paper195 (see also Ref. 150). In their treatment, the LSS surface is idealized as the surface of a three-dimensional sphere — in other words, the transition from tight-coupling to transparency is idealized to be instantaneous. Locally, on this surface the photon–baryon ﬂuid is subject to two kinds of perturbations: (i) perturbations in density δγ−b and (ii) velocity perturbations vb . The former translate into ﬂuctuations in the photon blackbody temperature Tγ , into “intrinsic” temperature ﬂuctuations at the last scattering surface with δTγ /T¯γ and the second translate into a Doppler shift of the CMB temperature. If there were no

h The radius of this sphere in terms of today’s comoving units is larger than the age of the universe (approximately 13.8 Gyr) converted into a distance owing to the expansion of the universe.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-65

Fig. 11. Ionization fraction as function of redshift with eﬀect of late-time ionizing radiation ignored. Calculated using the code CosmoRec by Jens Chluba (for details see, Ref. 262).

metric perturbations, we would simply have ˆ ˆ δTf (Ω) δTi (Ω) 1 ˆ = − δvγ · Ω, ¯ ¯ c Tf Ti

(21)

ˆ is the unit outward normal on the last scattering surface. But there are also where Ω additional terms due to the metric perturbations, and careful calculations along the perturbed geodesics yield the following modiﬁcation to the above equation: ηf ˆ ˆ δTf (Ω) δTi (Ω) 1 ˆ = − δv · Ω + Φ + dη(Φ + Ψ ), (22) γ c T¯f T¯i ηi where the metric perturbations are parametrized as in Eq. (15) using conformal Newtonian gauge.i This equation is known as the Sachs–Wolfe formula. Because of the approximations involved, the Sachs–Wolfe formula is not used for accurate calculations (especially on small angular scales). However, it oﬀers a good approximation to the CMB anisotropies on large angular scales and provides invaluable intuition that is lacking in more precise treatments. A few words about gauge dependence,j an issue that renders linearized perturbation theory within the framework of general relativity somewhat messy. While the ˆ T¯f is the same in all coordinate systems (except ﬁnal observed anisotropy δTf (Ω)/ i The Sachs–Wolfe formula as given here includes only the “scalar” perturbations. The formula can be generalized to include “vector” and “tensor” contributions, for which the only contributions are from the Doppler and ISW (integrated Sachs-Wolfe) terms. j “Gauge dependence” here means invariance under general coordinate transformations, which (consistent with the linear approximation above) may be truncated at linear order.

May 4, 2017 9:55 b2167-ch15

II-66

M. Bucher

for the monopole and dipole terms), the attribution of the total anisotropy among the various terms (i.e., intrinsic temperature ﬂuctuation, Doppler, gravitational redshift, integrated Sachs–Wolfe) depends on the choice of coordinates. We may for example choose coordinates so that the photon temperature serves as the time coordinate, making the ﬁrst term disappear, or alternatively, we may make the Doppler term disappear by making our co-moving observers move with the local photon rest frame. Students of cosmology are sometimes misled into believing the solution to these ambiguities is to use conformal Newtonian gauge, or equivalently the “gauge invariant” formalism, which is equivalent to transforming to conformal Newtonian gauge. It is believed that this gauge choice is somehow more “physical” because at least to linear order, the gauge conditions lead to a unique choice of gauge. It is certainly nice to have a gauge condition leading to a unique choice of coordinates, but this uniqueness comes at a price. The Newtonian gauge condition is “nonlocal” because of its reliance on the decomposition of hµν into “scalar”, “vector”, and “tensor” components, which requires information extending all the way out to spatial inﬁnity. But no one has ever seen all the way out to spatial inﬁnity. For this reason, the Newtonian potentials Ψ and Φ are unphysical because they cannot be measured. Their determination would require information extending beyond our horizon. We may further simplify Eq. (22) assuming that only the adiabatic growing mode has been excited. Using the fact that δTi and v are not independent of Φ, we obtain the often cited result δT Φ ˆ = − − v · Ω + dη(Φ + Ψ ), (23) T 3 although the −1/3 factor is not quite right because a matter-dominated evolution for a(t) is assumed rather than a more careful treatment taking into account matterradiation equality occurring at around zeq ≈ 3280. (For a discussion see Ref. 224.) Using this result and ignoring the integrated Sachs–Wolfe term, we obtain Φ δT ˆ ˆ (Ω) = − − v · Ω. T 3

(24)

On large angular scales — that is, large compared to the angle subtended by the horizon on the last scattering surface — the ﬁrst term dominates. The modes labeled by k obey an oscillatory system of coupled ODEs, and at the putative big bang each mode starts with a deﬁnite sharp phase corresponding to the part of the cycle where v = 0. It is only at horizon crossing that the phase has evolved suﬃciently for the velocity term to contribute appreciably to the CMB anisotropy. Figure 12 shows the calculated form of the CMB temperature power spectrum. For < ∼ 100, ∆T /T ≈ −Φ/3 provides an adequate explanation for this leftmost plateau of the curve, but at higher a rise to a ﬁrst acoustic peak, situated at about ≈ 220, is observed followed by a series of decaying secondary oscillations.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-67

Fig. 12. Theoretical CMB spectrum. CMB temperature power spectrum predicted for a model with only the adiabatic growing mode excited, as in the standard concordance cosmological model, is shown. The theoretical model here assumes the best-ﬁt cosmological parameters taken from the Planck 2013 Results for Cosmology.166

A more accurate integration of the evolution of the adiabatic mode up to last scattering can be obtained in the ﬂuid approximation. This approximation assumes that the stress–energy content of the universe can be described by a ﬂuid description consisting of two components coupled to each other only by gravity.199 However more accurate calculations must go beyond the ﬂuid approximation using a Boltzmann formalism that includes higher order moments. However, before describing the Boltzmann formalism, capable of describing this intermediate regime for the photons, we note one ﬁnal shortcoming of the approximation with tight-coupling and instantaneous recombination — namely, the treatment of polarization. We present here an approximate, heuristic treatment in order to provide intuition. The scattering of photons by electrons is polarization dependent, and this eﬀect leads to a polarization of the CMB anisotropy when the fact that recombination does not occur instantaneously is taken into account. The Thomson scattering cross-section of a photon oﬀ an electron is given by dσ = dΩ

e2 mc2

2 (ˆ i · ˆf )2 ,

(25)

where ˆi and ˆf are the initial and ﬁnal polarization vectors, respectively. To see qualitatively how the polarization of the CMB comes about, let us for the moment assume that the photons coming from next-to-last scattering are

May 4, 2017 9:55 b2167-ch15

II-68

M. Bucher

Fig. 13. Anisotropy of polarized Thomson scattering and the origin of the CMB polarization. We show two examples of photon scattering from (NLS) (next-to-last scattering) to (LS) (last scattering) and ﬁnally to the observer (at O). The three segments in the diagram have been chosen mutually at right angles in order to maximize the eﬀect, and for simplicity we assume that the radiation emanating from (NLS)1 and (NLS)2 is unpolarized. All the radiation that scatters coming from (NLS)1 to (LS) and then towards O is completely linearly polarized in the ˆ2 direction, because the other polarization is parallel to (LS)O and thus cannot be scattered in that direction. Likewise, the scattered photons emanating from (NLS)2 are completely linearly polarized in the orthogonal direction for the same reason. Consequently if there were only two sources as above, measuring the Stokes parameter Q at O amounts to measuring the diﬀerence in intensity between these two sources.

unpolarized and calculate the polarization introduced at last scattering, as indicated in Fig. 13. Measuring photons with one linear polarization selects photons that propagated from next-to-last to last scattering at a small angle to the axis of linear polarization, while photons with the other linear polarization tend to come from a direction from next-to-last to last scattering with a small angle with the other axis. Consequently, measuring the polarization amounts to measuring the temperature quadrupole as seen by the electron of last scattering. If we make the simplifying assumption that the radiation emanating from next-to-last scattering is unpolarized, we obtain the following expression for the Stokes parameters from the linear polarization,

ρ=ρmax Q ˆ sin2 θ cos φ = d(−exp[−τ (rl.s. + ρ)]) dΩ sin φ ρ=0 U l.s. × T (ρ sin θ cos φ, ρ sin θ sin φ, rl.s. + ρ cos θ).

(26)

Here we assume that the line of sight is along the z direction. The expression gives two of the ﬁve components of quadrupole moment of the temperature as seen by the electron at last scattering, and after integration along the line of sight would give the total linear polarization with the polarization at next-to-last scattering neglected.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-69

7. What Can We Learn From the CMB Temperature and Polarization Anisotropies? The previous section showed how starting from Gaussian isotropic and homogeneous initial conditions on superhorizon scales, the predicted CMB temperature and polarization anisotropies are calculated, describing in detail all the relevant physical processes at play. In this section, we turn to the question of what we can learn about the universe from these observations. We focus on how to exploit the temperature and polarization two-point functions, which under the assumption of Gaussianity would summarize all the available information characterizing the underlying stochastic process. Gaussianity is a hypothesis to be tested using the data, as discussed separately in Sec. 9. 7.1. Character of primordial perturbations: Adiabatic growing mode versus ﬁeld ordering In the aftermath of the COBE/DMR announcement of the detection of the CMB temperature anisotropy, two paradigms oﬀered competing explanations for the origin of structure in the early universe. On the one hand, there was cosmic inﬂation, which in its simplest incarnations predicts homogeneous and isotropic Gaussian initial perturbations where only the adiabatic growing mode is excited. On the other hand, there was also another class of models in which the universe is postulated initially to have been perfectly homogeneous and isotropic. However, subsequently a symmetry breaking phase transition takes place, with the order parameter ﬁeld taking uncorrelated values in causally disconnected regions of spacetime. Then, as the universe expands and the co-moving size of the horizon grows, the ﬁeld aligns itself over domains of increasingly large co-moving size, generally in a self-similar way. In many of these models, topological defects of varying co-dimension — such as monopoles, cosmic strings, and domain walls — arise after the phase transition, but textures where the spatial gradient energy is more diﬀusely distributed are also possible.k In these models the contribution of the ﬁeld ordering sector to the total energy is always subdominant, and cosmological perturbations are generated in a continuous manner extending to the present time. The stress–energy Θµν from this sector sources perturbations in the metric perturbation hµν . These metric perturbations in turn generate perturbations in the dominant components contributing to the stress–energy: the baryons, photons, neutrinos, and cold dark matter. In these models, the cosmological perturbations are not primordial, but rather are continuously generated so that the perturbations on the scale k are generated primarily as the mode k enters the apparent horizon.152 This property implies that the decaying mode as well as the growing mode of the adiabatic perturbations is excited, and this fact has a spectacular eﬀect on the shape of the predicted CMB power k For

the physics of topological defects, see Ref. 184, and for a comprehensive account emphasizing the connection to cosmology, see Ref. 217.

May 4, 2017 9:55 b2167-ch15

II-70

M. Bucher

spectrum. While inﬂationary models, which excite only the growing mode, predict a series of sharp, well-deﬁned acoustic oscillations, ﬁeld ordering models predict broad, washed out oscillations or no oscillations at all in the angular CMB power spectrum.4,36,151 Heuristically one can understand this behavior by arguing that the positions of the peaks reﬂect the phase of the oscillations. Therefore, if both the growing and decaying modes are present in exactly equal proportion, there should be no oscillations in the angular spectrum. However, precise predictions require diﬃcult numerical simulations and depend on the precise model for the topological defect or ﬁeld ordering sector. Figure 14 shows the shape of the predicted CMB TT power spectra for some ﬁeld ordering models, to be compared with the predictions for minimal inﬂation shown in Fig. 12. After COBE, a big question was whether improved degree-scale CMB observations would unveil the ﬁrst Doppler peak followed by a succession of decaying secondary peaks at smaller angular scales, as predicted by an approximately scale invariant primordial power spectrum for the growing adiabatic mode, or whether some other shape would be observed, perhaps the one predicted by a ﬁeld ordering model.

Fig. 14. Structure of acoustic oscillations: ﬁeld ordering predictions. These acoustic oscillation predictions from four ﬁeld-ordering models are to be contrasted with the sharp, well-deﬁned peaks in Fig. 12 as predicted by inﬂation. (Reprinted with permission from Ref. 151.) (Credit: Pen, Seljak and Turok)

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-71

We already presented (in Sec. 1) the experimental results for the ﬁrst observations of the acoustic oscillations and their subsequent precise mapping (see in particular Figs. 4, 5 and 10.) The data clearly favor simple models of inﬂation producing adiabatic growing mode perturbations with an approximately scale invariant power spectrum and exclude scenarios where topological defects serve as the primary source for the initial density perturbations. However, models including a small admixture of defect induced perturbations cannot be ruled out.174 7.2. Boltzmann hierarchy evolution In the previous section, we derived the Sachs–Wolfe formula, which provides an intuitive understanding of how the CMB anisotropies are imprinted. The Sachs– Wolfe treatment provides an approximate calculation of the CMB temperature anisotropies, correctly capturing their qualitative features. However to calculate predictions accurate at the sub-percent level, as is required for confronting current observations to theoretical models, the transition from the tight-coupling regime to transparency, or the free-streaming regime, must be modeled more realistically. This requires going beyond the ﬂuid approximation, where the photon gas can be described by its ﬁrst few moments. Modeling this transition faithfully requires a formalism with an inﬁnite number of moments — that is, a formalism in which the fundamental dynamical variable is the photon phase space distribution function ˆ , ˆ) where ν is the photon frequency, and the unit vectors n ˆ and ˆ denote f (x, t, ν, n the photon propagation direction and electric ﬁeld polarization, respectively. The perturbations about a perfect blackbody spectrum are small, so truncating at linear order provides an adequate approximation. Moreover, Thomson scattering is independent of frequency, and thus (to linear order) does not alter the blackbody character of the spectrum. Consequently, a description in terms of a perturbation of the blackbody temperature that depends on spacetime position, propagation direction, and polarization direction can be used in place of the full six-dimensional phase space density, thus reducing the number of dimensions of the phase space by one.l For a calculation correct to second order, a description in terms of a perturbed temperature would not suﬃce. As before we consider a single “scalar” Fourier mode k of ﬂat three-dimensional space, so that an exp[ik · x] spatial dependence is implied. A complication in describing polarization arises in specifying a basis, which preferably is accomplished ˆ using only scalar quantities. We describe the polarization using the unit vectors n ˆ to deﬁne the unit vectors and k ˆ ˆ (ˆ ˆ n · k) ˆ = k−n , θ ˆ−n ˆ ˆ (ˆ

k n · k)

ˆ ˆ×k ˆ= n φ , ˆ

ˆ n × k

(27)

l Although written as a continuum variable, ˆ is in reality a discrete variable corresponding to the Stokes parameters (I, Q and U ).

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-72

ˆ form an orthonormal basis. which together with photon propagation direction n ˆ as We thus describe the Stokes parameters of the radiation propagating along n follows: ˆ ab + Q(ˆ ˆ θ ˆ⊗θ ˆ−φ ˆ ⊗ φ] ˆ ab . n, ν)Eb (ˆ n, ν) = I(ˆ n, ν; k)δ n, ν; k)[ Ea (ˆ

(28)

For the “scalar” mode, the Stokes parameters U and V vanish in this basis. The Thomson scattering cross-section does not vary with frequency, a property that provides substantial simpliﬁcation. Suppose that at some initial time the photon distribution function has a blackbody dependence on frequency, with a blackbody temperature suﬀering only small linearized perturbations about an average temperature, and that these temperature perturbations depend on the spacetime position, propagation direction, and linear polarization direction. The frequency independence of Thomson scattering ensures that a photon distribution function initially having this special form will retain this special form during its subsequent time evolution. Initially, when tight-coupling is an excellent approximation, any deviation in the frequency dependence from a perfect blackbody spectrum decays rapidly. Moreover, any initial polarization is erased by the frequent scattering. Therefore a blackbody form described by a perturbation in the spectral radiance of the form ∂Bν (T ) n, ˆ, x, t) = T¯γ ∆ab (ˆ n, x, t)ˆ a ˆb δIν (ˆ ∂T

(29)

is well motivated. Here

Bν (T ) =

2hν 3 c2

−1 hν exp −1 kB T

(30)

is the spectral radiance for a blackbody at temperature T . The unperturbed spectral radiance is Iν,0 (ˆ n, ˆ, x, t) = Bν (Tγ (t)), where Tγ (t) is the photon temperature at cosmic time t in the unperturbed background expanding FLRW spacetime. (In Sec. 5, however, we discuss some interesting possible caveats to this assumption.) For the “scalar” mode of wavenumber k, we may decompose (scl)

n, x, t) = ∆T ∆ab (ˆ

(scl)

(ˆ n, x, t)δab + ∆P

(ˆ n, x, t)(θˆa θˆb − φˆa φˆb ). (scl)

(31)

(scl)

We now turn to the evolution of the variables ∆T (ˆ n, t) and ∆P (ˆ n, t). It is ˆ to expand convenient, exploiting the symmetry under rotations about k, (scl)

∆T

(ˆ n, t; k) =

∞

(scl) ˆ ·n ˆ ), ∆T, (t; k)(−i) (2 + 1)P (k

=0 (scl) ∆P (ˆ n, t; k)

=

∞ =0

(32) (scl) ∆P, (t; k)(−i) (2

ˆ ·n ˆ ). + 1)P (k

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy (scl)

II-73

(scl)

The evolution of ∆T, and ∆P, is governed by the following inﬁnite system of equations127 : (scl)

∆T 0

(scl)

∆T 1

(scl)

∆T 2

(scl)

∆T

(scl)

∆P

= −k∆T 1 + φ , (scl)

v k (scl) b (scl) (scl) (∆T 0 − 2∆T 2 + ψ) + a(t)ne (t)σT − ∆T 1 , 3 3

k Π (scl) (scl) (scl) = (2∆T 1 − 3∆T 3 ) + a(t)ne (t)σT + ∆T 2 , 5 10

=

= =

k (scl) (scl) (scl) {∆T (−1) − ( + 1)∆T (+1) } − a(t)ne (t)σT ∆T , (2 + 1) k (scl) (scl) (scl) {∆P (−1) − ( + 1)∆P (+1) } − a(t)ne (t)σT ∆P (2 + 1)

1 1 + a(t)ne (t)σT Π δ,0 + δ,2 , for ≥ 0, 2 5

for ≥ 3,

(33)

where (scl)

(scl)

(scl)

Π = ∆T 0 + ∆P 0 + ∆P 2 .

(34)

These equations are known as the Boltzmann hierarchy for the photon phase space distribution function. In the last line of Eq. (33) δab denotes the Kronecker delta function. For practical numerical calculations, this inﬁnite set of coupled equations must be truncated at some max . In principle, ignoring questions of computational eﬃciency, we could truncate the system of equations in (33) at some large max , suﬃciently above the maximum multipole number of interest today, and integrate the coupled system from some suﬃciently early initial time to the present time t0 to ﬁnd the “scalar” temperature and polarization anisotropies today, which would be given by the following integrals over plane wave modes: (scl) (scl) n, t0 ) = d3 k∆T (ˆ n, t0 ; k), ∆T (ˆ (scl) ∆P (ˆ n, t0 )

=

(35) d

3

(scl) k∆P (ˆ n, t0 ; k).

This was the approach used in the COSMICS code,16 one of the ﬁrst codes providing accurate computations of the CMB angular power spectrum. However in this approach much computational eﬀort is expended at late times when there is virtually no scattering. A computationally simpler but mathematically completely equivalent approach known as the line of sight formalism was introduced in Ref. 200. The crucial idea is to express the anisotropies today (at x = 0 and η = η0 ) in terms of a line of sight

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-74

integral, so that

(S)

∆T (ˆ n) =

(S) ∆P (ˆ n)

dη exp[−τ (η)] exp[ik(η − η0 )]

0

× and

η0

dτ 1 − ∆T 0 + iµvB + P2 (µ)Π + (φ − ikµψ) dη 2

(36)

dτ dη exp[−τ (η)] exp[ik(η − η0 )] − (1 − P2 (µ))Π dη

(37)

η0

= 0

(S)

(S)

(S)

where Π = ∆T 2 +∆P 2 +∆P 0 . We may think of exp[−τ (η)](−dτ /dη)dη = V(η)dη as a measure along the line of sight weighted according to where last scattering takes place. We shall call V(η) the visibility function. In Eq. (36), the ﬁrst term represents the nonpolarized contribution at last scatter and the second term represents the contribution from the integrated Sachs–Wolfe eﬀect. In Eq. (37) for the polarized anisotropy, the only contribution is from last scatter. It is convenient to rewrite the above integrals using integration by parts so that all spatial derivatives along the line of sight and occurrences of µ are eliminated, and in the process we also omit the monopole surface term at the endpoint corresponding to the observer as this contribution is not measurable. After this integration by parts, the integrals above take the form η0 (S) (S) n) = dη exp[ikµ(η0 − η)]ST,P ((η0 − η)ˆ n, η), (38) ∆T,P (ˆ 0

where the scalar source functions are (S)

ST

v 3Π = exp[−τ (η)](φ + ψ ) + V ∆T 0 + ψ + b + 2 k 4k

vb 3Π 3Π + V + 2 + V 2 , k 4k 4k

(39)

3 (V(k 2 Π + Π ) + 2V Π + V Π), 4k 2 and the primes denote derivatives with respect to conformal time η. Now that the spatial derivatives have been eliminated, it is no longer necessary to use the plane waves exp[ikµ(η0 − η)] as the eigenfunctions of the Laplacian operator with eigenvalue −k 2 . It is convenient instead to use a spherical wave expansion with j (k(η0 − η)) η0 (S) (S) ∆;A = dηj k(η0 − η) SA (η0 − η)ˆ n, η , (40) (S)

SP

=

0

where A = T, P. The line of sight formulation results in a signiﬁcant reduction of the computational eﬀort required for computing the high coeﬃcients for two reasons. First, the Boltzmann hierarchy can be truncated at low , because only moments up to

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-75

= 2 appear in the integral. Second, when the universe has become transparent, there is no longer any contribution from the Thomson scattering. We express the initial state for the growing adiabatic mode at some early reference time whose exact value must be chosen after the epoch of entropy generation following inﬂation but before any of the relevant modes re-entered the horizon. We use the variable ζ(x), whose value is conserved on superhorizon scales, to characterize the primordial perturbations in the adiabatic growing mode. We may expand, either using plane wave Fourier modes or a spherical wave expansion, so that (41) ζ(x) = d3 kζ(k) exp[ik · x], or ζ(x) =

+ ∞

∞

dkζ,m (k)j (kr)Ym (θ, φ).

(42)

0

=0 m=−

We characterize the homogeneous and isotropic Gaussian statistical ensemble for ζ(x) by means of a power spectrum P(k), which suﬃces to completely deﬁne the statistical process generating ζ(x). We deﬁne P(k) according to the following expectation values: ζ(k)ζ(k ) = (2π)3 δ 3 (k − k )P(k),

(43)

ζm (k)ζ m (k ) = δ, δm,m δ(k − k )P(k).

(44)

or equivalently

It follows that (S),T am

∞

=

η0

dk

0

0

(S) dηj k(η0 − η) ST (η, k)ζm (k),

(45)

so that (S),T T C

=

(S),T (S),T [am ]∗ am

= 0

∞

(S),T

dkP(k)∆

(S),T

(k)∆

(k),

(46)

where (S),T ∆ (k)

= 0

η0

(S) dηj k(η0 − η) ST η, k .

(47)

The above approach was ﬁrst implemented in the publicly available code CMBFAST200 and later in the code CAMB.121 m More details about the computational issues may be found in these papers. m For

a download and information on CAMB, see www.camb.info.

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-76

7.3. Angular diameter distance The radius of the last scattering surface expressed in terms of co-moving units today (for example in light-years or Mpc) is given by the integral xls = H0 −1

1

als

da 1 √ 2 −4 a Ωr a + Ωm a−3 + ΩΛ

(48)

for a universe whose stress–energy content includes radiation, matter, and a cosmological constant. Here the scale factor is given by als = (zls + 1)−1 . For other equations of state the argument of the square root above is modiﬁed accordingly. A physical scale d at last scattering is converted to a co-moving scale today x according to x = (zls + 1)d, and under the assumption of a ﬂat spatial geometry, the angle in radians subtended by an arc of length d on the last scattering sphere is θ = (d/dls ). Except for the integrated Sachs–Wolfe eﬀect (discussed below in Sec. 7.4), which is very hard to measure because the S/N is negligible but for the very ﬁrst multipoles, the projection eﬀect described above is the only way in which late-time physics enters into determining the angular power spectrum. Possible spatial curvature, parameterized through Ωk , the details of a possible quintessence ﬁeld, or similar new physics that alter the expansion history at late time enters into the CMB power spectra only through a rescaling of the relation between angular scales, on the one hand, and physical scales, on the other, around last scattering, and this rescaling can be encapsulated into a single variable dls . If the family of FLRW solutions is extended to admit spatial manifolds of constant positive or negative curvature, the expression for the angular size must be modiﬁed to the following for a negatively curved (hyperbolic) universe 1/2

θ=

Ωk dls 1/2 sinh[Ωk dls ]

d , dls

(49)

or in the case of positive curvature (i.e., a spherical universe) θ=

(−Ωk )1/2 dls d . sin[(−Ωk )1/2 dls ] dls

(50)

In the former case, the additional prefactor giving the contribution for the nonEuclidean character of the geometry has a demagnifying eﬀect and increases the angular diameter distance, whereas for the spherical case the eﬀect is the opposite. In the early 1990s, many theorists working in early universe cosmology fervently believed on grounds of simplicity in a cosmological model with Ωk = 0 (supposedly an inexorable prediction of inﬂation) and no cosmological constant (regarded as extremely ﬁnely tuned and “unnatural”) containing only matter and

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-77

radiation so that Ωr + Ωm = 1 exactly. This belief was maintained despite a number of discordant astronomical observations. Many in the astronomical community, however, rather adopted the attitude “I believe only what I see,” and thus favored some kind of a low-density universe, either a negatively curved universe where Ωk = 1 − Ωm − Ωr or a universe with a nonzero cosmological constant, where ΩΛ = 1 − Ωm − Ωr .n Although it was widely claimed that spatial ﬂatness was a prediction of inﬂation, the basic idea of how to produce a negatively curved universe within the framework of inﬂation had already been proposed in 1982 by Gott and Statler.72,73 However, since they had not calculated the perturbations predicted in such a model, it was not clear that such a model would work. In 1995 Bucher et al.26,27 and Yamamoto et al.252 calculated the perturbations for single bubble open inﬂation, showing them to be consistent with all the observations available at the time. Present constraints from the CMB limit |Ωk − 1| to a few percent at most, the exact number depending on the precise parametric model assumed.166,171 7.4. Integrated Sachs–Wolfe eﬀect The discussion above claimed that the impact of late-time physics on the CMB power spectrum could be completely encapsulated into a single parameter: the angular diameter distance to last scattering dls . This is only approximately true, because the geometric argument above assumed that all the CMB anisotropy can be localized on the last scattering surface and ignored the integrated Sachs–Wolfe term in Eq. (22), which gives the part of the linearized CMB anisotropy imprinted after last scattering. Intuitively, the integrated Sachs–Wolfe term can be understood as follows. The Newtonian gravitational potential blueshifts the CMB photons as they fall into potential wells and similarly redshifts them as they again climb out. If the depth of the potential well does not change with time, the two eﬀects cancel. In this case, there is no integrated Sachs–Wolfe contribution because the integral can be reduced to the sum of two surface terms. But if the depth of the potential well changes with time, in particular if the overall scale of the potential is decaying, the two eﬀects no longer cancel and an integrated Sachs–Wolfe contribution is imprinted. In the linear theory, the gravitational potential at late times may be factorized in the following way: Φ(x, t) = Φ(x)T (t),

(51)

and we may normalize T (t) to one when the universe has become matter dominated, but before the cosmological constant — or some other dark energy sector — has started to alter the expansion history. n For a snapshot of the debates in cosmology in the mid-1990s, see the proceedings of the conference Critical dialogues in cosmology held in 1995 at Princeton University.213

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-78

From the CMB anisotropies alone, it is hard to determine the contribution of the integrated Sachs–Wolfe eﬀect to the total CMB anisotropy on large angular scales. However one can try to isolate its contribution by measuring the cross-correlation of the CMB temperature anisotropy with the large-scale structure using a broad radial window function extending to large redshifts as was ﬁrst proposed in Ref. 37, whose aim was to ﬁnd evidence for a nonzero cosmological constant. The evolution equation for the linearized density contrast of a single pressureless matter component is148 3 δ¨m (t) + 2H(t)δ˙m (t) − H 2 (t)Ωm (t)δm (t) = 0, 2

(52)

where the dots denote derivatives with respect to proper time and H(t) = a(t)/a(t). ˙ Here we treat the baryons and the CDM (cold dark matter) as a single component ignoring any relative velocity between them. This is not a bad approximation at late times, particularly on very large scales. For an Einstein–de Sitter universe (for which Ωr is negligible and ΩΛ = 0), which is a good approximation for our universe at intermediate redshifts, because 4 for z < ∼ zeq ≈ 3.4 × 10 radiation is unimportant and after the decoupling of the photons and baryons, it is valid to set the velocity of sound equal to zero as above. Let zmΛ = 1/(1 + amΛ ) be the redshift where the nonrelativistic matter and the putative cosmological constant contribute equally to the mean density of the universe, so that Ωm = amΛ 3 /(1 + amΛ 3 ) and ΩΛ = 1/(1 + amΛ 3 ). For ΩΛ = 0.68, for example, we would have zmΛ = 0.3 and amΛ = 0.78. For the Einstein–de Sitter case, a(t) = t2/3 and the growing and decaying solution to Eq. (52) are t2/3 and t−1 , respectively. (1/a2 )∇2 φ = (3/2)H 2 Ωm δm , and for the growing mode Φ(x) is independent of time, implying that there is no integrated Sachs–Wolfe contribution at late times. However, as the cosmological constant begins to take over and ΩΛ starts to rise toward one, Φ starts to decay, as sketched in Fig. 15. As explained in detail in Ref. 37, the maximum signal-to-noise (S/N) that can be extracted from the ISW correlation, even with an ideal measurement of the large-scale structure anisotropy within our past light cone, is modest because of the “noise” from the primary CMB anisotropies emanating from the last scattering surface. This cosmic variance noise cuts oﬀ the available S/N because on small angular scales, the relative contribution from the ISW term plummets. The S/N is centered around ≈ 20 and integrates to about S/N ≈ 6.5 (assuming ΩΛ0 = 0.7). For an early detection of a correlation, see Ref. 25, and for the most recent work in this area see Ref. 168 and references therein.

7.5. Reionization According to our best current understanding of the ionization history of the universe, the early ionization fraction around recombination is accurately modeled assuming homogeneity and using atomic rate and radiative transfer equations. As

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-79

Fig. 15. Decay of the gravitational potential for the late-time integrated Sachs–Wolfe eﬀect. We sketch the time dependence of the gravitational potential for a universe with ΩΛ0 = 0.9, 0.8, 0.7, and 0.6 (from top to bottom on the right) with a = 1 today. The qualitative behavior will be similar for other kinds of dark energy.

a rough estimate, one may assume the reionization fraction predicted by equilibrium thermodynamics given by the Saha equation, in which there is an exponential sensitivity to changes in temperature. After x ≈ 12 , which occurs at about z ≈ 1270, the ionization fraction x(z) rapidly decays to almost zero, except for trace amounts of ionized hydrogen (denoted as HII in the astrophysical literature) and of free electrons, due to the ineﬃciency of recombination in the presence of trace concentrations. If it were not for inhomogeneities and structure formation, this nearly perfect neutral phase would persist to the present day. However, the initially small primordial inhomogeneities grow and subsequently evolve nonlinearly. The ﬁrst highly nonlinear, gravitationally collapsed regions give rise to the ﬁrst generation of stars and quasars, which serve as a source of UV radiation that reionizes the neutral hydrogen almost completely, except in a minority of regions of extremely high gas density, where recombination occurs suﬃciently frequently to counteract the ionizing ﬂux of UV photons. The ﬁrst observational evidence for such nearly complete reionization came from observations of the spectra of quasars at large redshift in the interval 10.4 eV/(1 + zquasar ) < Eγ < 10.4 eV — that is, blueward of the redshifted Lyman α line.74 If the majority of the hydrogen along the line of sight to the quasar were neutral, resonant Lyα (2p → 1s) scattering would deﬂect the photons in this frequency range out of the line of sight, and this part of the spectrum would be devoid of photons. The fact that the spectrum is not completely blocked in this range indicates that the universe along the line of sight was reionized. Almost complete absorption by neutral hydrogen — that is, the so-called Gunn–Peterson trough — was not observed in

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-80

quasar spectra until 2001, when spectroscopic follow-up of several high-redshift quasars discovered by the Sloan Digital Sky Survey (SDSS) detected such a trough from z ≈ 5.7 to z ≈ 6.3, where the continuum emission blueward of the Lyα appears to have been completely absorbed by neutral hydrogen.14 The big question here is exactly when did the universe ﬁrst become reionized. Or said another way: When did the ﬁrst generation of stars and quasars form and become capable of generating enough ionizing radiation to convert the neutral gas to its present ionized state? This is a vast subject of very current research, sometimes described as the exploration of the “Dark Ages” of the universe. While we believe that we know a lot about the conditions around the time when the CMB anisotropies were imprinted, and we also know a lot about the recent universe extending to moderate redshift, little is known with any degree of certainty concerning this intermediate epoch. Good reviews of this subject include Ref. 12. Future observations of the 21 cm hyperﬁne transition of atomic hydrogen (HI) by the next generation of radio telescopes, in particular the Square Kilometer Array (SKA),o promise to provide three-dimensional maps of the neutral hydrogen in our past light cone. Here we limit ourselves to discussing the impact of reionization on the CMB anisotropies. The most simplistic model — or caricature — of reionization postulates that the ionization fraction x(z) changes instantaneously from x = 0 for z > zreion to x = 1 for z < zreion. Under this assumption, the optical depth for rescattering by the reionized electrons is zreion d dzσT ne (z) . τ= (53) dz 0 The Thomson scattering optical depth from redshift z to the present day is given by the integral τ (z) =

=

cσT H0

1

a(z)

cσT ne0 H0

da ne (a) −4 a (Ωr a + Ωm a−3 + ΩΛ )1/2 1

a(z)

da xe (a) , a4 (Ωr a−4 + Ωm a−3 + ΩΛ )1/2

(54)

where a(z) = 1/(z+1), σT is the Thomson scattering cross-section, H0 is the present value of the Hubble constant (in units of inverse length), and ne0 is the electron density today, assuming xe = 1. The functions ne (a) and xe (a) are the electron densityp and the ionization fraction at redshift z = 1/a − 1, respectively. o https://www.skatelescope.org/.

is customary to deﬁne xe as ne /nB , so that xe slightly exceeds one when the helium is completely ionized as well.

p It

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-81

On small angular scales, the eﬀect of reionization on the predicted CMB temperature and polarization anisotropy spectra is straightforward to calculate. Of the CMB photons emanating from the LSS, a fraction exp(−τ ) does not undergo rescattering by the reionized electrons and their fractional temperature perturbations are preserved. However, for the remaining fraction (1 − exp(−τ )) ≈ τ that is rescattered, the small-angle CMB anisotropies are completely erased. The reionized electrons nearly completely wash out small-scale detail much like a window pane of ground glass. The net eﬀect is that the small-scale CMB anisotropies are attenuated by a factor of exp(−τ ) in amplitude and by a factor of exp(−2τ ) in the power spectrum. On larger angular scales, the attenuation of the anisotropy from the rescattered photon component is incomplete, and a more detailed calculation is needed to determine the detailed form of the -dependent attenuation factor. The order of magnitude of the angular scale where the attenuation starts to become incomplete is given by the ratio of the co-moving distance travelled from next-tolast to last scattering by a rescattered photon dreion to the radius of the LSS dLS . We thus obtain an angular scale θreion ≈ dreion /dLS for this transition. For determining the cosmological model from the CTT power spectrum alone, the reionization optical depth is highly degenerate with the overall amplitude of the cosmological perturbations A. This is because except on very large scales, where cosmic variance introduces substantial uncertainty, it is the combination A exp(−2τ ) that determines the observed amplitude of the CMB angular power spectrum. The polarization power spectra, meaning C EE and also C TE , help to break this degeneracy by providing a rather model independent measurement of τreion . Recall our order of magnitude estimate for the polarization introduced by the ﬁnite thickness of the LSS ∂2T P ∼ d2 2 , (55) ∂x where d and x are expressed in co-moving units and d is the mean co-moving distance that the photon travels from next-to-last scattering to last scattering. The second derivative is that of the temperature anisotropy evaluated at the position of a typical electron of last scatter. Let us compare the polarization imprinted during recombination to that imprinted later for the fraction of photons (1 − exp(−τ )) ≈ τ rescattered by the free electrons from reionization, for the moment assuming that the second derivative factors are comparable for the two cases. However, drec dreion ; therefore, we may expect that Preion /Prec ≈ τ (drec /dreion )2 1. This polarization from reionization, however, is concentrated almost exclusively in the ﬁrst few multipoles, because directions separated by a small angle see the same quadrupole anisotropy from last scattering from the vantage point of a typical free electron after reionization. Figure 16 shows the scalar power spectra for four values of τ keeping the other cosmological parameters ﬁxed. We shall see that the same ampliﬁcation applies to the tensor modes, or primordial gravitational waves, presumably generated during inﬂation.

May 4, 2017 9:55 b2167-ch15

II-82

M. Bucher

Fig. 16. CMB power spectra for four values of the reionization optical depth. We plot the indicated power spectra for τ = 0.00, 0.04, 0.08, and 0.12 assuming the same concordance values for the other cosmological parameters. We multiply the spectra by exp(2τ ) in order to remove the trivial dependence on τ at small angular scales and thus highlight what occurs at low .

We now turn to observations allowing us to ﬁx the value of τ using the low C spectra and C TE spectra based on the qualitative theoretical arguments above but made more precise by calculating τ within the framework of the previously described six-parameter concordance model. While the ﬁrst observations of the polarization of the CMB were reported by the DASI experiment,111 it was the WMAP large-angle polarization results, ﬁrst the TE correlations reported in the one-year release, later followed by the EE correlation results from the third-year results, that provided the ﬁrst determination of τreion using the CMB, and subsequently reﬁned in later updates to the WMAP results with more integration time. As part of their ﬁrst-year results,231 the WMAP team reported a value for the reionization optical depth of τ = 0.17 ± 0.04 at 68% conﬁdence level, which would correspond to 11 < zr < 30 at 95% conﬁdence where a step function proﬁle for the ionization fraction is assumed. This determination was based on exploiting only the measurement of the TE cross-correlation power spectrum, because for this ﬁrst release the analysis of the polarized sky maps was not suﬃciently advanced to include a reliable EE auto-correlation power spectrum measurement. The reported value of τ was larger than expected. In their third-year release,237 when WMAP presented a full analysis of the polarization with EE included, the determination of τ shifted downward, namely to τ = 0.10 ± 0.03 (using only EE data) and τ = 0.09 ± 0.03 with (TT, TE and EE all included). The ﬁnal nine-year WMAP release251 cites a value of τ ≈ 0.089 ± 0.014 using WMAP alone. Including other external data sets results in small shifts about this value. EE

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-83

The Planck 2013 Results for Cosmology161,166,171 release did not include an analysis of the Planck polarization data. Therefore the likelihoods used to determine the cosmological parameters included the WMAP determination of τ as a prior, and in some cases the WMAP low- polarization likelihood was used in order to break the degeneracy between τ and A described above that arises when only temperature data is used. It has been pointed out that reionization histories, which are completely characterized by function xe (z), cannot be reduced to a single number. Some work has been done to characterize what further information can be extracted from the low- CMB concerning the reionization history, as described in Refs. 95 and 101, but the conclusion is that from the CMB data alone at most a few nonzero numbers can be extracted at S/N > ∼ 1 given the large cosmic variance at low . In the above discussion we have assumed that reionization occurs homogeneously in space, an approximation adequate for calculating its eﬀect on the anisotropies at large and intermediate angular scales. However, according to our best understanding, the universe becomes reionized by the formation of bubbles, or Str¨ omgren spheres, surrounding the ﬁrst sources of ionizing radiation, which grow and become more numerous, and eventually percolate. While this basic picture is highly plausible, the details of modeling exactly how this happens remain speculative. The eﬀect on the CMB is signiﬁcant only at very small angular scales, far into the damping tail of the primordial anisotropies. Power from the low- anisotropies is transformed into power at very high . This can be understood by considering the power in Fourier space resulting from the rather sharp bubble walls, which in projection look almost like jump discontinuities. Some simulations of this eﬀect, along with other eﬀects at large , can be found for example in Ref. 253. 7.6. What we have not mentioned The preceding subsections provide a sampling of what can be learned from the CMB TT and polarization power spectra with each point discussed in some detail. For lack of space, we cannot cover all the parameters and extensions of the simple six-parameter concordance model that can be explored, and here we provide a few words regarding what we have not covered giving some general references for more details. The structure of the acoustic oscillations and the damping tail (relative heights, positions, etc.) of the CMB power spectra depends on the cosmological parameters, as forecast and explained pedagogically in Refs. 92, 99, and most recently applied to the Planck 2013 data for determining the cosmological parameters.166 Some of the constraints are obtained in the framework of the standard six-parameter concordance model, but many extensions to this model can be constrained, such as theories with varying αQED , nonminimal numbers of neutrinos, varying neutrino masses, isocurvature modes, nonstandard recombination, to name just a few examples. Moreover the question can be addressed whether there exists statistically

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-84

signiﬁcant evidence in favor of extending the basic six-parameter model to a model having additional parameters. See Refs. 166 and 171 for an extensive set of references. 8. Gravitational Lensing of the CMB The previous sections presented a simpliﬁed view of the CMB where the CMB anisotropies are imprinted on the so-called last scattering surface, situated around z ≈ 1100. It is assumed that we look back to this surface along straight line geodesics in the unperturbed coordinate system. This is not a bad approximation, but it is not the whole story. Inhomogeneities in the distribution of matter, particularly at late times when the clustering of matter has become nonlinear, act to curve the photon, or geometric optics, trajectories. These nonuniform deﬂections distort the appearance of the surface of last scatter, much like a fun house mirror. Gravitational lensing of the CMB is a small eﬀect but large enough so that it has to be taken into account in order to compare models of the primordial universe with the observations correctly.264 In one sense, gravitational lensing may be viewed as a nuisance eﬀect to be removed. But gravitational lensing of the CMB may also be regarded as an invaluable tool for probing the inhomogeneities of the matter distribution between us and the last scattering surface. Gravitational lensing is presently a very active area of astronomy, and the CMB is only one of the many types of objects whose gravitational lensing may be measured in order to probe the underlying mass inhomogeneities in the universe. Gravitational lensing of the CMB may be contrasted with competing weak lensing probes in that: (i) for the CMB, the sources (i.e., the objects being lensed) are the most distant possible, situated at z ≈ 1100, (ii) because of the above, CMB lensing is sensitive to clustering at larger redshifts, and thus less sensitive to nonlinear corrections, and (iii) unlike observations of the correlations of the observed ellipticities of galaxies, where there are systematic errors due to intrinsic alignments, for CMB lensing there is no such problem. In the linear theory (which is a good approximation, especially for a heuristic discussion), the deﬂection due to lensing may be described by means of a lensing ˆ on the celestial sphere. Let Ω ˆ unlens potential, deﬁned as a function of position Ω coordinatize the last scattering surface as it would look in the absence of lensing, ˆ lens to be the position on the last scattering surface with lensing taken and deﬁne Ω into account. We may express ˆ unlens + ∇Φlens (Ω). ˆ ˆ lens = Ω Ω

(56)

Tlens = Tunlens − (∇Φlens ) · (∇Tunlens ),

(57)

It follows that

and similar expressions may be derived for the polarization. As in weak lensing, we may derive a projected Poisson equation for the lensing potential xsource ˆ = ˆ x), dx W (x; xsource ) δρ(Ω, (58) ∇2 Φlens (Ω) 0

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-85

where ∇2 is the Laplacian operator on the sphere. More details about the lensing of the CMB and what can be learned from it may be found in the review123 and references therein. The ﬁrst discovery of gravitational lensing of the CMB involved detecting a statistically signiﬁcant nonzero cross-correlation between a lensing map reconstructed using the WMAP data and infrared galaxies.203 The use of a cross-correlation allows one to detect a small signal within a noisy map. Figure 17 shows the gravitational lensing spectrum as measured by Planck.167 Measurements of gravitational lensing have also been made by ACT39 and SPT,214 and the Polarbear experiment has recently reported the discovery of B modes due to lensing.182 The lensing ﬁeld described above (which can be expressed either as a potential, as a deﬂection ﬁeld, or as a dilatation and shear ﬁeld) can in principle be recovered from distortions in the small-scale power spectrum due to the lensing ﬁeld on larger scales. As a general rule, the higher the resolution of the survey, the better the reconstruction. Hu and Okamoto143,93 proposed a reconstruction based on Fourier modes, and later many other estimators have been proposed, which are all quite similar because they are all exploiting the same signal. In the future, as observations at higher sensitivity and in particular due to B modes at small scales become available, lensing promises to become a powerful probe of the clustering of matter and halo structure. It is hoped that lensing will eventually be able to determine the absolute neutrino masses, even if they are as low as allowed by present neutrino oscillation data.120 Lensing

Fig. 17. Power spectrum of CMB gravitational lensing potential as measured by Planck. The angular power spectrum for the lensing potential as reported in the Planck 2013 results is shown above. The overall statistical signiﬁcance for a lensing detection is greater than 25σ. The solid curve indicates the expectation based on the concordance model whose six parameters have been ﬁxed by the Planck data combined with ancillary data. For more details, see Ref. 167. (Credit: ESA/Planck Collaboration) (For color version, see page II-CP9.)

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-86

detected through B modes is particularly advantageous because the fact that the B modes should be zero in the absence of lensing means that cosmic variance does not intervene and that only the instrument noise is an issue.q

9. CMB Statistics 9.1. Gaussianity, non-Gaussianity, and all that For the simplest models analyzed at linear order, the primordial CMB signal is the outcome of an isotropic Gaussian random process on the celestial sphere. While any particular realization of this random process produces a sky temperature map that is not isotropic, the hypothesis is that the underlying stochastic process is isotropic. Concretely the hypotheses of isotropy and Gaussianity imply that the ˆ whose multipole expansion in terms of probability of obtaining a sky map T (Ω), spherical harmonic coeﬃcients is given by ˆ = T (Ω)

∞ +

ˆ am Ym (Ω),

(59)

=0 m=−

is as follows: P ({am }) =

1 1 |am |2 exp − . 2 Cth 2πCth =0 m=− + ∞

(60)

ˆ = (θ, φ) denotes a position on the celestial sphere and the set of posiHere Ω tive coeﬃcients {Cth } represents the angular power spectrum of the underlying stochastic process, or the theoretical power spectrum. If we had postulated Gaussianity alone without the additional hypothesis of isotropy, we would instead have obtained a more general probability distribution function having the form 1 −1/2 (2πCth ) exp− (61) am ∗ (Cth )−1 m, m a m . P ({am }) = det 2 ,m ,m

Here Cth is the covariance matrix for the spherical harmonic multipole coeﬃcients. This Ansatz is too general to be useful because there are many more parameters than observables, and in cosmology we can observe only a single sky — or said another way, only a single realization of the stochastic process deﬁned in Eq. (61). The assumption of isotropy is very strong, greatly restricting the number of degrees conclusion hardly depends on r whatever its ﬁnal value may turn out to be, because the angular scales best for probing lensing lie beyond the recombination bump of the B modes.

q This

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-87

of freedom by setting all the oﬀ-diagonal elements to zero and requiring that the diagonal elements depend only on . Happily, the present data, subject to some caveats (such as gravitational lensing, discussed above in Sec. 8), seem consistent with the hypotheses of isotropy and Gaussianity, notwithstanding some “anomalies” at moderate statistical signiﬁcance (discussed in detail below in Sec. 12). These anomalies could be a sign of something new, but given their limited statistical signiﬁcance, the argument that they are simply statistical “ﬂukes” cannot be rejected. For completeness, we now indicate how the above formalism is extended to include the polarization of the primordial CMB signal. There are several formalisms for describing the polarization of the CMB, for example, using spin-weighted spherical harmonics, or using the Stokes parameters I, Q, U, V .23 (The polarization of the CMB and its description are discussed in Refs. 35, 96, and 108.) We may deﬁne ˆ = Ea (Ω) ˆ ∗ Eb (Ω) ˆ time where Ea (Ω) ˆ is the electromagnetic ﬁeld component Tab (Ω) ˆ propagating from the direction Ω and units are chosen so that for a unit electric ˆ a εb represents the CMB thermodynamic temperature polarization vector ε, Tab (Ω)ε when by means of a linear polarizer only the component along ε is measured. We may decompose ˆ = T (Ω)δ ˆ ab + Pab (Ω), ˆ Tab (Ω)

(62)

where the polarization tensor (containing the Stokes parameters Q, U , and V ) is trace-free and Hermitian (and also symmetric in the absence of circular polarization). We must now choose a convenient basis for the polarization. If we considered a single point on the celestial sphere, we could using Stokes parameters set Q = T11 − T22 and U = 2T12 , but no natural choice for the directions of the orthonormal ˆ can be singled out as preferred. unit vectors eˆ1 and eˆ2 in the tangent space at Ω Had we instead chosen

cos θ sin θ eˆ1 eˆ1 = , (63) −sin θ cos θ eˆ2 eˆ2 we would instead have the Stokes parameters

Q U

=

cos 2θ

sin 2θ

−sin 2θ

cos 2θ

Q U

.

(64)

The transformation law in Eq. (64) indicates that the traceless part of the polarization is of spin-2. If the sky were ﬂat with the geometry of R2 instead of S 2 and we had to include homogeneous modes as well, there would be no way out of this quandary. But there is a celebrated result of diﬀerential geometry that a fur on a sphere of even dimension cannot be combed without somewhere leaving a bald spot

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-88

or singularity.r A traceless tensor ﬁeld on the sphere cannot have an everywhere vanishing covariant derivative. We can generate a complete basis for the polarization on the celestial sphere by ˆ as means of derivative operators acting on the spherical harmonics. Using Ym (Ω) a starting point, we may deﬁne an E mode basis vector as follows:

ˆ = N E,B ∇a ∇b − 1 δab ∇2 Ym (Ω), ˆ Eab;m (Ω) (65) 2 where NE,B is a normalization factor. Here a, b = 1, 2 indicate components with respect to an arbitrary orthonormal basis on S 2 . Without introducing more structure, this is the only way to produce a second-rank traceless tensor on the sphere ˆ in a way that does not break isotropy, with derivative operators acting on Ym (Ω) ˆ a feature that ensures that Eab;m (Ω) transforms under rotations according to the quantum numbers m. However, there is another polarization, rotated with respect to the E mode by 45◦ , and in order to write down a basis for this other polarization, known as the B modes, it is necessary to introduce an orientation on the sphere through the unit antisymmetric tensor with 12 = 1. This choice of orientation, or volume element in the language of diﬀerential geometry, deﬁnes a handedness, or direction in which the 45◦ rotation is made. We may deﬁne Bab =

1 (δaa + aa )(δbb + bb )Ea b , 2

(66)

or in terms of derivative operators Bab =

1 E,B ˆ N (ac δbd + bc δad )∇c ∇d Ym (Ω). 2

(67)

The diﬀerence between E modes and B modes is illustrated in Fig. 18. We want to normalize so that ˆ ∗ (Ω) ˆ = dΩB ˆ ab,m (Ω)B ˆ ∗ (Ω) ˆ = δ, δm,m (68) ˆ ab,m (Ω)E dΩE ab, m ab, m and

ˆ ab,m (Ω)B ˆ ∗ (Ω) ˆ = 0, dΩE ab, m

(69)

so we require that100 NE,B

r This

=

2 . ( + 2)( + 1)( − 1)

is a consequence of the “hairy ball” theorem.60

(70)

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-89

Fig. 18. E and B modes of the CMB polarization. The polarization pattern in the left panel can be expressed as the double derivative acting on a potential with its trace removed (in this case a Gaussian potential) and thus is an E mode. The B mode pattern on the right, however, cannot be represented in this way, but can be represented by a pseudo-scalar potential whose double derivative with the trace removed followed by a 45◦ rotation gives the indicated pattern.

Although the above discussion emphasized the role of spherical harmonics, the distinction between E and B modes does not rely on spherical harmonics, and much confusion has resulted from not fully appreciating this point. The absence of an E mode component or a B mode component may be formulated in a completely local manner using covariant derivative operators acting on the polarization tensor ˆ If ∇a Pab = 0, then the E mode is absent, and similarly if ab ∇a Pbc = 0, Pab (Ω). then the B mode is absent. Consequently, if we are willing to diﬀerentiate the data, we could detect B modes locally with absolutely no leakage. However, the above does not imply that a unique decomposition of Pab into E and B on an incomplete sky is possible because nonsingular vector ﬁelds (akin to harmonic functions) can be found that satisfy both conditions, although no nontrivial such vector ﬁelds exist deﬁned on the full sky. Where global topology does enter is in eliminating conﬁgurations that satisfy both the above diﬀerential constraints. Having deﬁned the E and B modes, we now indicate how the statistical distribution deﬁned in Eq. (60) is modiﬁed to include polarization. We modify our notation somewhat to avoid unnecessary clutter. When polarization is included, T E B our vector {aT m } is generalized to become {am , am , am }, which we shall denote as the vector (t, e, b). For each we have a positive deﬁnite covariance matrix of the form

CTT

C = CET

CBT

CTE

CTB

CEE

CEB .

CBE

CBB

(71)

If we assume that the underlying physics imprinting the cosmological perturbations is invariant under spatial inversion, then the above covariance matrix

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-90

simpliﬁes to TT C C = CET 0

CTE CEE 0

0

0 ,

(72)

CBB

because under spatial inversion CTB → −CTB ,

CEB → −CEB ,

(73)

whereas the other power spectra coeﬃcients preserve their sign. If parity is a symmetry of the physical processes generating the primordial perturbations and imprinting the CMB anisotropies, the expectation values CTB,th and CEB,th must vanish, although for any particular sky realization CTB,obs and CEB,obs will not vanish because of cosmic variance. (Recall that CTB,th and CEB,th is the average over a ﬁctitious ensemble including an inﬁnite number of sky realizations.) Studying CTB,obs and CEB,obs to search for statistically signiﬁcant deviations from zero is a way to search for parity violation in the very early universe. However, in practice adjusting experimental parameters to minimize CTB,obs and CEB,obs is often used to calibrate detector angles in B mode polarization experiments. The propagation of light, or electromagnetic radiation, at least historically and conceptually, occupies a privileged place in the theories of special and general relativity, because of the constancy of the speed of light in vacuum and the fact that at least in the eikonal approximation, light travels along null geodesics, with its electric polarization vector propagated along by parallel transport. It is by thinking of bundles of light rays that we construct for ourselves a physical picture of the causal structure of spacetime. A modiﬁcation of the propagation of light (for example, from dilaton and axion couplings,141 a possible birefringence of spacetime,32,62,82,206 or alternative theories of electromagnetism70,71,142 ) would not necessarily undermine the foundations of relativity theory, but would probably merely lead to a more complicated theory, possibly only in the electromagnetic sector. The CMB may be considered an extreme environment because of the exceptionally long travel time of the photons, and one can search for new eﬀects beyond the known interactions with matter (e.g., Thomson and other scattering, collective plasma eﬀects such as dispersion and Faraday rotation). Some such eﬀects would lead to spectral distortions, for example through missing photons. The rotation of the polarization vector during this long journey would lead to mixing of E and B modes, and since the primordial E modes are so much larger, leads to interesting constraints on birefringence, which will greatly improve as more data comes in from searches for primordial B modes. The statistical description in Eqs. (71) and (72) suﬃces for confronting the predictions of theory with idealized measurements of the microwave sky with complete sky coverage and no instrument noise. We must further assume that there are no

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-91

secondary anisotropies, nor galactic foregrounds, both of which have been perfectly removed. Even under these idealized assumptions, it is not possible to pin down the theory completely because of a phenomenon known as cosmic variance. The problem is that while we would like to characterize the covariance of the underlying theory CAB,th (where A, B = T, E, B), we have only a single realization of the microwave sky, and CAB,obs is only an estimate of CAB,th with ﬂuctuations about its expectation value, which is equal to CAB,th . If we consider just the temperature ﬂuctuations, we may deﬁne the sample variance for the th multipole CTT,obs =

+ 1 |am |2 . (2 + 1)

(74)

m=−

The random variable CTT,obs obeys a χ2 -distribution with (2 + 1) degrees of free dom and thus has fractional variance of 2/(2 + 1). If we consider a microwave sky map bandlimited up to max , there are approximately max 2 degrees of freedom, so for example the overall amplitude of the cosmological perturbations may be determined with a fractional accuracy of approximately 1/max . Most analyses of the CMB data (possibly combined with other data sets) assume a model where the theoretical power spectrum depends on a number of cosmological parameters, which we may abstractly denote as α1 , α2 , . . . , αD , also written more compactly as the vector α. For example, in the six-parameter minimal cosmological model claimed suﬃcient in the 2013 Planck Analysis166 to explain the present data, the parameters comprising this vector were AS , ns , H0 , ωb , ωCDM , and τ. Whether one adopts a frequentist or Bayesian analysis to analyze the data, the input is always the relative likelihood of the competing models given the observations, or a ratio of the form: P (observations | α1 ) . P (observations | α2 )

(75)

In this ratio, the measure for the probability density of the outcome, represented by a continuous variable, cancels. In this abridged discussion, we shall follow a Bayesian analysis where a prior probability Pprior (α) is assumed on the space of models. In Bayesian statistics, Bayes’ theorem is used to tell us how we should rationally update our prior beliefs (or prejudices) in light of the data or observations to yield a posterior probability to characterize our updated beliefs given by the formula Pposterior (α) =

P (observations | α)Pprior (α) . D d α P (observations | α )Pprior (α )

(76)

There is no reason why the posterior distribution should lend itself to a simple analytic form, and a common procedure is to explore the form of the posterior

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-92

distribution using Markov Chain Monte Carlo (MCMC) methods,57,122 which are particularly well suited to explore distributions of large dimensionality. However, it often occurs that the likelihood and the posterior distribution can be adequately represented as a Gaussian by expanding about its maximum likelihood value. In this case, the maximum of the log of the probability can be found using a numerical optimization routine, and then by computing second derivatives using numerical ﬁnite diﬀerences, the Gaussian approximation to the posterior can be found. This inﬁnitesimal analysis is often referred to as a Fisher analysis.

9.2. Non-Gaussian alternatives It is incredibly diﬃcult to test the Gaussianity of the primordial microwave sky without some guidance from theory as to what non-Gaussian alternatives are well motivated. The space of non-Gaussian stochastic models is dauntingly vast, with the space of Gaussian models occupying by any measure an almost negligible fraction of the model space. If one tests enough models, one is guaranteed to produce a result beyond any given threshold of signiﬁcance. Consequently, one of the questions lurking behind any claim of a detection of non-Gaussianity is how many similar models did one test.

10. Bispectral Non-Gaussianity Although it was recognized that inﬂation, in its simplest form described by Einstein gravity and a scalar ﬁeld minimally coupled to gravity, would have some nonlinear corrections, early analyses of the cosmological perturbations generated by inﬂation linearized about a homogeneous solution and calculated the perturbations in the framework of a linearized (free) ﬁeld theory. According to the lore at the time, the nonlinear corrections to this approximation would be too small to be observed. The ﬁrst calculations of the leading nonlinear corrections were given by Maldacena128 and Aquaviva et al.,7 who calculated the bispectrum or three-point correlations of the primordial cosmological perturbations within the framework of single-ﬁeld inﬂation. Subsequent work indicated that in models of multi-ﬁeld inﬂation, bispectral nonGaussianity of an amplitude much larger than for the minimal single-ﬁeld inﬂationary models can be obtained. The predictions of many of these multi-ﬁeld models were within the range that would be detectable by the Planck satellite. For many of these models, the non-Gaussianity, owing to the fact that it is generated by dynamics on superhorizon scales where derivative terms are unimportant, is well described by the “local” Ansatz for bispectral non-Gaussianity, under which the non-Gaussian ﬁeld ζNL is generated from an underlying Gaussian ﬁeld ζL according to the rule 2 ζNL (x) = ζL (x) + ζL (x) ,

(77)

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-93

where fNL is a dimensionless parameter that quantiﬁes the degree of nonGaussianity. Here fNL is independent of wavenumber, but a dependence on wavenumber can also be contemplated, so that the above relation is generalized to the following (now working in wavenumber space): ζNL (k) =

d k1

d3 k2 fNL (k; k1 ; k2 )δ 3 (k − k1 − k2 )ζL (k1 )ζL (k2 ).

3

(78)

Here as a consequence of the hypothesis of isotropy, we have expressed fNL as a function depending only on the lengths of the vectors. Any two triangles (where the three vectors close) that may be mapped into each other by an isometry are assigned the same fNL . For bispectral non-Gaussianity of the local form, as deﬁned in Eq. (77), we brieﬂy sketch how the order of magnitude of (S/N )2 for a detection of fNL from a CMB map extending up to max may be estimated. We simplify the estimate by employing the ﬂat sky approximation, allowing us to replace the unintuitive discrete sums and Wigner 3j symbols with the more transparent continuum integrals and a two-dimensional δ-function. Approximating C ≈ C0 −2 (in other words, we ignore acoustic oscillations, damping tails, and all that, and we assume that the CMB is scale invariant on the sky, as would occur if one could extrapolate the simpliﬁed Sachs–Wolfe formula (∆T )/T = −Φ/3 to arbitrarily small scales), we may write

S N

2

= O(1)

d2 (2π)2

d2 2 (2π)2

d2 3 (2π)2 δ 2 (1 + 2 + 3 ) (2π)2

fNL 2 C0 4 (1 −2 2 −2 + 2 −2 3 −2 + 3 −2 1 −2 )2 C0 3 1 −2 2 −2 3 −2

max 2 = O(1)(fNL ) C0 max 2 ln . min ×

(79)

The quantity in the ﬁrst line simply approximates the sum over distinguishable triangles, the combinatorial factors being absorbed in the O(1) factor, and the second line gives the signal-to-noise squared of an individual triangle, as calculated diagrammatically. More exact calculations taking into account more details conﬁrm this order of magnitude estimate. Numerical calculations are needed to determine the O(1) factor and deal with max more carefully. Nevertheless this rough estimate reveals several characteristics of local non-Gaussianity, namely: (i) the presence of the ln(max /min) factor indicates that the bulk of the signal arises from the coupling of large angle modes and small angle modes (i.e., the modulation of the smallest scale power near the resolution limit of the survey by the very large angle modes (i.e., = 2, 3, . . .)), and (ii) the cosmic variance of the estimator derives from the inability to measure the small-scale power accurately, and thus decreases

May 4, 2017 9:55 b2167-ch15

II-94

M. Bucher

in proportion to the eﬀective number of resolution elements of the survey. Cosmic variance on large scales is not an issue. As long as we know the particular realization on large scales accurately (which is not subject to cosmic variance), we know what kind of modulation to look for in the small-scale anisotropies. Theoretical models for other shapes for the bispectral anisotropy can be motivated by models for fundamental physics (see for example Refs. 9 and 107 and references therein) and have been searched for as well. The Planck 2013 Results gave equil local ortho = 2.7 ± 5.8, fNL = −42 ± 75, and fNL = −25 ± 39. the following limits173 : fNL This result rules out many of the models developed to explain a result at low statistical signiﬁcance from some analyses of the WMAP data and was a profound disappointment to those who hoped that Planck would turn up striking evidence against the simplest inﬂationary models. One may anticipate that the results from Planck 2014 will improve modestly on the 2013 results by including more modes because of the use of polarization. If there is a next-generation all-sky CMB polarization satellite having good angular resolution so that S/N≈ 1 maps may be obtained up to max ≈ 3000, an improvement on the limits on fNL by a factor of a few may be envisaged. However, beyond ≈ 3000, the primary CMB anisotropies become a sideshow. As one pushes upward in , other nonprimordial sources of anisotropy take over, such as gravitational lensing, the Sunyaev–Zeldovich eﬀect, and point sources of various sorts. These contaminants have angular power spectra that rise steeply with while the CMB damping tail is falling rapidly, in fact almost exponentially. For including more modes, polarization may be helpful. On the one hand, polarization measurements present formidable instrumental challenges because of their lower amplitude compared to the temperature anisotropies. However, for the polarization, the ratio of foregrounds to the primordial CMB signal is favorable up to higher than for temperature. For searching for non-Gaussianity, the primordial CMB is highly linear, unlike other tracers of the primordial cosmological perturbations. However given that the basic observables are two-dimensional maps, the small number of modes compared to three-dimensional tracers of the primordial perturbations is a handicap.

11. B Modes: A New Probe of Inﬂation As discussed above, some limits on the possible contribution from tensor modes, or primordial gravitational waves, can be obtained from analyzing the C T T power spectrum. However, given that r is not large, these limits become highly model dependent, and because of cosmic variance as well as uncertainty as to the correct parametric model, cannot be improved to any substantial extent. This diﬃculty is illustrated by the results from the Planck 2013 analysis,171 which did not include the data on polarized anisotropies collected by Planck but only the TT power spectrum. If one assumes the minimal six-parameter model plus r added as an extra parameter, a limit of r < 0.11 (95% C.L.) is obtained. However, if the primordial

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-95

power spectrum in this model (where a form P (k) ∼ k ns −1 is assumed) is generalized to P (k) ∼ (k/k ∗ )ns −1 exp(α(ln(k/k ∗ ))2 ), the limits on r loosen by about a factor of two to become r < 0.2 (95% conﬁdence level). This lack of robustness to the assumptions of the parametric model highlights the fragility of this approach and demonstrates that the statistical error bars are not to be trusted unless one has strong reason to trust the underlying parametric model. However, given our current understanding of inﬂationary potentials and dynamics, we have no reason to trust any of the parametric models, which serve more as ﬁtting functions for summarizing the current state of the observations. Searching for tensor modes using B modes, on the other hand, allows us to probe much lower values of r in a manner that is substantially model independent because in the linear theory “scalar” perturbations cannot generate any B modes, and this conclusion is independent of the parametric model assumed. 11.1. Suborbital searches for primordial B modes In March 2014, the BICEP2 Collaboration17,18 claimed a detection of primordial gravitational waves based primarily on observations at a single frequency (150 GHz) over a small patch of the sky (1% of the sky) that was believed to be of particularly low polarized dust emission. Using a telescope based at the South Pole, the BICEP2 team created a map at 150 GHz of the polarized sky emission from which the B mode contribution was extracted. The BICEP2 analysis included many null tests to estimate and exclude systematic errors as well as to estimate the statistical noise in the measurement. However, in order to claim a detection of primordial gravitational waves, the BICEP2 team had to exclude a nonprimordial origin for the observed B mode signal, for which the main suspect would be polarized thermal dust emission. An independent analysis66 called into question the BICEP2 estimate of the likely dust contribution in their ﬁeld, claiming that all the observed signal could be attributed to dust. A subsequent Planck analysis estimating the dust contribution in the BICEP2 ﬁeld by extrapolating polarized dust maps at higher frequencies where dust is dominant down to 150 GHz conﬁrmed this ﬁnding.159 While this analysis does not necessarily exclude a nonzero primordial contribution to the B modes observed by BICEP2 at 150 GHz, the argument crucial to the BICEP2 claim that a dominant contribution from dust can be excluded collapses in light of this new ﬁnding from Planck concerning the expected contribution from thermal dust emission. A cross-correlation study is now underway as a joint project of the two teams that will combine the BICEP2 map at 150 GHz with the Planck polarization maps at higher frequencies (which are dominated by dust). It remains to be seen whether this eﬀort can establish a limit on r better than what is currently available or whether this combined analysis might even result in a detection of primordial B modes albeit at a lower level of r. In any case, a number of experiments are now underway involving observations at multiple frequencies that will push down the limits on r using B modes if not

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-96

result in a ﬁrst detection if r is not too small. The BICEP2 team has a series of upgrades to their experiment expanding their frequency coverage as well as massively increasing the number of detectors and hence their sensitivity. Other competing eﬀorts include POLARBEAR,8 SPIDER34 and QUBIC,13 as well as ACTPol and SPTPol. Moreover, farther into the future, a more ambitious upgrade to ground based eﬀorts is contemplated through the US Stage 4 (S4)1,2 CMB experiment currently under consideration by the US DOE. S4 contemplates deploying approximately a total of 2×105 detectors from the ground in order to achieve a massive improvement in raw sensitivity. Although it is claimed that an improvement in the control of systematic errors to a comparable level can be achieved, it remains to be seen whether S4 can realize its projected performance. 11.2. Space based searches for primordial B modes Several groups around the world have proposed space missions speciﬁcally dedicated to mapping the microwave polarization over the full sky with a sensitivity that would permit the near ultimate measurement of the CMB B modes. In Europe, three ESA missions had been proposed: BPol in 2007,43 COrE33 in 2010, both of which were medium-class, and the large-class mission PRISM185 in 2013. None of these missions was selected. A mission proposal called COrE+ is currently in the process of being submitted. In Japan a mission called LiteBird81 has been developed, but ﬁnal approval is still pending. In the US, CMBPol/EPIC61 has proposed a number of options for a dedicated CMB polarization space mission,s but none of these has been successful in securing funding. All the above concepts involve a focal plane including several thousand single-mode detectors in order to achieve the required sensitivity. Another proposal endeavors to achieve the needed sensitivity by another means: multi-moded detectors. In the Pixie105 proposal, a Martin–Pupplet Fourier spectrometer comparing the sky signal to that of an artiﬁcial blackbody is proposed. The setup is similar to the COBE FIRAS instrument, except that polarization sensitive bolometers and a more modern technology are used, which will allow an improvement in the measurement of the absolute spectrum about two orders of magnitude better than FIRAS. 12. CMB Anomalies The story recounted so far has emphasized the agreement of the observations accumulated over the years with the six-parameter so-called “concordance” model. But this would not be a fair and accurate account without reporting a few wrinkles to this remarkable success story. These wrinkles may either be taken as hints of new physics, or discounted as statistical ﬂukes, perhaps using the pejorative term s See

http://cmbpol.uchicago.edu/papers.php for a list of papers on this eﬀort.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-97

a posteriori statistics. On the one hand, one can argue that if one looks at enough models that are in some sense independent, one is bound to turn up something at high statistical signiﬁcance, and by most standards the signiﬁcance of these anomalies is not high. Which interpretation is preferred is presently under debate. Ultimately, despite all the fancy statistical terminology used in this discourse, what one concludes inevitably relies on a subjective judgment of theoretical plausibility. One anomaly explored over the years, known under several names including “hemispherical asymmetry,” “dipolar modulation,” and “bipolar disorder,” asks the question whether the local angular power spectrum is identical when compared between opposite directions on the celestial sphere. There is no unique way to pose this question precisely. The statistical signiﬁcance, amplitude, and direction of the asymmetry depend somewhat on the precise formulation chosen. However, the WMAP ﬁnding of a dipolar modulation with an amplitude of ≈ 7% and a signiﬁcance ranging within about 2−4σ, and pointing approximately toward (l, b) = (237◦ , 20◦ ) in galactic coordinates is broadly conﬁrmed by the Planck 2013 Results. This conﬁrmation by Planck, owing to the additional high frequency coverage, renders less plausible a nonprimordial explanation based on foreground residuals. The sensitivity to dipolar modulation on these angular scales is primarily limited by cosmic variance for both WMAP and Planck, so future experiments cannot hope to improve substantially on the signiﬁcance of these results. Planck, however, because of its superior angular resolution, was able to probe for dipolar modulation pushing to smaller scales, where cosmic variance is less of a problem and a possible modulation can be constrained more tightly. [The measurement uncertainty in the scale-dependent dipole modulation of the amplitude of the perturbations ∆A() (assuming broad binning) scales as 1/ as long as cosmic variance is the limiting factor.] The Planck 2013 data ﬁnd no evidence of dipolar modulation at the same amplitude extending to small angular scales, suggesting that if the dipolar modulation observed by WMAP is not a statistical ﬂuke, a scale-dependent theoretical mechanism for dipolar modulation is needed. Moreover, Planck does not see evidence of modulation associated with higher multipoles (e.g., quadrupolar and higher order disorders). This last point is important for constraining theoretical models producing statistical anisotropy by means of extra ﬁelds disordered during inﬂation. Another anomaly is the so-called “cold spot.” Rather than limiting ourselves to two-point statistics, or to three-point statistics, searching for a bispectral signal using theoretically motivated templates, as described in Sec. 10, we may ask whether the most extreme values of an appropriately ﬁltered pure CMB map lie within the range that may be expected assuming an isotropic Gaussian stochastic process, or whether their p-values (i.e., probability to exceed) render the Gaussian explanation implausible. There are many ways to formulate questions of this sort, a fact that renders the assessment of statistical signiﬁcance diﬃcult. One approach that has been applied to the data is to ﬁlter the sky maps with a spherical Mexican hat wavelet (SMHW) ﬁlter, which on the sphere would have the proﬁle of the Laplacian

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-98

operator applied to a Gaussian kernel of a given width σ, giving a broad [i.e., (∆)/ = O(1)] two-dimensional spatial bandpass ﬁlter. (See Ref. 215 for more details concerning the methodology and Ref. 216 for the original claim of a cold spot detection. See also Ref. 248 for an assessment by the WMAP team as well as Ref. 172 for the analysis from Planck 2013.) The ﬁnal anomaly concerns possible alignments of the low- multipoles, imaginatively named the “axis of evil” by Land and Magueijo.114 A Gaussian isotropic theory for the temperature anisotropyt predicts that the multipole vectors a = {Am }+ m=− are each Gaussian and independently distributed. If we add three or more such vectors together according to the tensor product of representations of SO(3) L1 ⊗ L2 ⊗ · · · ⊗ LN ,

(80)

which may be decomposed into a direct sum of irreducible representations, provided that the triangle inequality is satisﬁed, we may extract one or more scalars (transforming according to L = 0) from the above tensor product. The mathematics is simply that of the usual addition of angular momentum. For each of these scalars, we may ask whether the observed value lies within the range expected from cosmic variance in the framework of a Gaussian theory taking into account the uncertainty in the determination of the underlying power spectrum. The approach just described n)| was not the approach of Land and Magueijo, who maximized maxm∈[−,+] |am (ˆ ˆ where am (ˆ n) is the expansion coeﬃcient in the coordinate sysas a function of n ˆ pointing to the north pole. In this way, for each generic value of the tem with n multipole number a unique double headed vector, or axis, ±ˆ n may be extracted, and the alignments between these axes may be assessed. Land and Magueijo report an alignment of the = 3, = 4, and = 5 axes with a p-value less than 10−3 in the Gaussian theory. Accurate p values may be obtained for each of these invariant quantities by resorting to MC simulations to account for the practicalities of dealing with a cut sky, etc. However, there still remains a subjective element to assessing statistical signiﬁcance. One may ask how many invariants were tried before arriving at a reportable statistically signiﬁcant anomaly. 13. Sunyaev–Zeldovich Eﬀects In the simpliﬁed discussion of reionization presented in Sec. 7.5, it was assumed that the electrons responsible for the rescattering are at rest with respect to the cosmic rest frame. This approximation treats the reionized gas as a cold plasma having a vanishing peculiar velocity ﬁeld. As pointed out by Sunyaev and Zeldovich,208 corrections to this approximation arise in two ways: (i) In the so-called thermal Sunyaev–Zeldovich eﬀect (tSZ), the random thermal motions of the free electrons t The discussion can readily be generalized to polarization, but we shall stick to the temperature anisotropy alone in our discussion.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-99

Fig. 19. Frequency dependence of thermal Sunyaev–Zeldovich spectral distortions. The solid and dashed curves show the fraction spectral distortion in terms of the CMB thermodynamic and R–J temperatures, respectively. For low frequencies (ν < 217 GHz), there is a decrement in temperature, corresponding to the fact that low frequency photons are on the average Doppler shifted to higher frequencies by the hot electrons. On the other hand, for ν > 217 GHz the net eﬀect is to increase the photon phase space density.

alter the spectrum of the rescattered CMB photons through the Doppler eﬀect and Compton recoil. Because v = 0, this eﬀect is second order in the velocity, or linear in the electron temperature, and results in a spectral distortion not respecting the frequency dependence of a blackbody spectrum with a perturbed temperature. (See Fig. 19.) (ii) In the kinetic Sunyaev–Zeldovich (kSZ) eﬀect the peculiar velocity of the gas results in a shift in temperature proportional to the component of the peculiar velocity along the line of sight. The kSZ eﬀect has the same frequency dependence as the underlying primary CMB perturbations, thus making it hard to detect given its small magnitude. There is also a much smaller eﬀect where the transverse peculiar velocity imparts a linear polarization to the scattered photons for which the polarization tensor Pij is proportional to (v⊥ ⊗ v⊥ − 12 v⊥ 2 δ ⊥ ), where v⊥ is the peculiar velocity perpendicular to the line of sight and δ ⊥ is the Kronecker delta function in the plane perpendicular to the line of sight. Galaxy clusters are ﬁlled with hot gas [Tgas = O(10 KeV)] that emits primarily at X-ray frequencies. The hot electrons of this fully ionized gas, or plasma, scatter CMB photons by Thomson, or Compton, scattering, shifting their frequency by a factor of approximately (1 + β cos θ) where β = kB T /me c2 and we have ignored higher order corrections in β. The y-distortion parameter along a line of sight is given by the integral y=

dτ

kB Te = mc2

dσT ne

kB Te , mc2

(81)

May 4, 2017 9:55 b2167-ch15

II-100

M. Bucher

and the fractional perturbation in the CMB thermodynamic temperature in the nonrelativistic approximation is given by

ˆ δTCMB (ν, Ω) x(ex + 1) ˆ = − 4 y(Ω), (82) ex − 1 T¯CMB ˆ is the y-distortion map as deﬁned where x = hν/kB TCMB = ν/(57 GHz) and y(Ω) by the line of sight integral in Eq. (81). On the one hand, the thermal Sunyaev–Zeldovich eﬀect provides a powerful probe of the dynamics of galaxy clusters, which can be used to discover new clusters and to probe the structure of known clusters away from their central core. On the other hand, for observing the primordial CMB, the thermal Sunyaev–Zeldovich eﬀect constitutes a contaminant that must either be removed or modelled. The review by Birkinshaw19 recounts the early history of tSZ measurements. See also the reviews by Carlstrom et al.29 and by Rephaeli.191 For the observation of the kSZ, see Ref. 79. For more recent measurements, see the SZ survey papers from the ACT,130 SPT,223 and Planck169,170 collaborations. See Ref. 185 for a discussion of what might ultimately be possible from space in the future. 14. Experimental Aspects of CMB Observations Like astronomical observations at other wavelengths, most modern microwave experiments consist of a telescope and a number of detectors situated at its focal plane. The COBE experiment was an exception, as it used single-moded feed horns pointed directly at the sky to deﬁne its beam, which was relatively broad (7◦ FWHM). However for measurements at higher angular resolution, it is diﬃcult to construct a feed horn forming a suﬃciently narrow beam on the sky without the help of intermediate optics, unless one resorts to interferometry as for example in the DASI experiment. Observations of the sky at microwave frequencies present a number of challenges unique to this frequency range. Unlike at other frequencies, the microwave sky is remarkably isotropic. The CMB monopole moment, and to a much lesser extent the dipole moment, dominates. In the bands where other contaminants contribute least—that is, roughly in the range 70–150 GHz—virtually all the photons collected result from the isotropic 2.725 K background. The brightest feature superimposed on this uniform background is the CMB dipole, resulting from our proper motion with respect to the cosmic rest frame, at the level of 0.1%. When this dipole component is removed, the dominant residual is the primordial CMB anisotropy, whose amplitude on large angular scales is about 35 µK, or roughly 10−5 . This situation is to be contrasted for example with X-ray astronomy where only some 50 photons need to be collected to make an adequate galaxy cluster detection. For the CMB about 1010 photons must be collected in each pixel just to obtain the temperature anisotropy at a signal-to-noise ratio of one, and many more photons are needed for measuring polarization at the same marginal signal-to-noise. The challenge is to

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-101

measure minute diﬀerences in temperature between diﬀerent points on the celestial sphere without introducing spurious eﬀects. These can arise from a variety of sources: drifts in the zero point of the detector, detector noise, hot objects in the far sidelobes of the beam, and diﬀraction around the edges of the mirrors, to name some of the most common problems. For this reason, almost all measurements of the microwave sky are in some sense diﬀerential. Drifts in the zero point of the detector (known as 1/f noise) can be mitigated by rapidly switching either between the sky and an artiﬁcial cold load with a long time constant or between diﬀerent parts of the sky. This can be accomplished by moving the beam rapidly across the sky, so that in one way or the other the basic observable becomes diﬀerences in intensity of pixels whose measurements are closely separated in time. The point to take away is that one absolutely avoids making direct rather than diﬀerential measurements of the sky temperature. Another speciﬁcity of astronomical observations at microwave frequencies is that extremely stringent requirements must be imposed limiting the magnitude of the far sidelobes of the beam. Given the smallness of the diﬀerences in temperature of interest, contamination from the ground, the earth, the sun, and moon, and also from the galaxy near the galactic plane, can easily introduce spurious signals if the beam does not fall suﬃciently rapidly away from its central peak. Ground pickup (from a spherical angle of 2π) is particularly challenging to shield, and this is why the WMAP and Planck satellites were situated at L2 (the second Lagrange point, where the earth and the moon have a tiny angular diameter) rather than in a lowearth orbit, which would equally well avoid atmospheric emission, which is the other major motivation for going to space. The second earth-sun Lagrange point (L2) is 1.5×106 km from the earth — that is, about 4 times the earth–moon distance. From L2 the earth subtends an angle of only 16 arcminutes. More importantly from L2, the sun, earth, and moon all appear at approximately the same position in the sky making it easier to keep away from the far sidelobes of the beam. A formidable challenge for suborbital observations is emission from the atmosphere, primarily but not exclusively from water vapor. The emission from the atmosphere has a complicated frequency dependence, as shown in Fig. 20, where the optical depth as a function of frequency is shown. This optical depth must be multiplied by the atmospheric temperature (i.e., ≈ 270 K) to obtain the sky brightness. The structure in frequency of the regions of high opacity implies that bands must be carefully chosen to avoid emission lines and bands, giving less ﬂexibility for wide frequency coverage than from space. The problems arising from atmospheric emission can be mitigated to some extent by choosing a site on the ground with a particularly small water column density, such as the South Pole or the Atacama Desert in Chile, or by observing from a stratospheric balloon. Atmospheric emission is problematic in two respects. First, if the atmospheric emission were stable in time, it would merely introduce additional thermal loading on the detectors, which could be mitigated simply by deploying more detectors or by observing over

May 4, 2017 9:55 b2167-ch15

II-102

M. Bucher

Fig. 20. Atmospheric optical depth as a function of frequency for 1 mm of precipitable water vapor. Water vapor is an important but not the only source of atmospheric contamination. See Ref. 145 for more details. A column density of 0.5–1 mm of precipitable water vapor corresponds to good conditions in the Atacama Desert. Conditions at the South Pole are often better by a factor of a few. The black curve shows the opacity, or optical depth, and the red curve shows the opacity multiplied by 1/20 in order to highlight the structure around the resonances. The limited transmission 0 < t < 1 [where t = exp(−τ )], on the one hand, attenuates the desired sky signal. But more serious is the superimposed additive noise, having a brightness temperature (1 − t)Tatm . Here Tatm is the eﬀective temperature of the atmosphere. (Credit: J. Cernicharo)

a longer time. If the contribution from the sky is Tsky , then the ﬂuctuations in intensity at the detector (assumed perfect for the moment) would increase by a factor of (Tsky + TCMB )/TCMB , meaning that we could measure the sky temperature with the same error by increasing either the number of detectors or the total observation time by a factor of ((Tsky + TCMB )/TCMB )2 . At 150 GHz from the South Pole, for example, to choose a band where the sky brightness temperature is not very high, one has Tsky ≈ 16 K. This means that under the idealized assumptions made here, almost 40 times more detectors would be required than for a space-based experiment. But the situation is worse than this, because the atmosphere contamination is not stable in time but rather ﬂuctuates in a complicated way characterized by a wide range of time scales (owing to the underlying turbulence). Moreover, the atmospheric load varies with zenith angle, roughly according to a secant law. Of course, observations from the ground are attractive because of their low cost, the fact that telescopes can be improved from season to season to address problems as they arise, and larger telescopes with many more detectors can be deployed. As experiments increase in sensitivity, requirements for mitigating these problems will become more stringent. At present it is not known what is the limit on the quality of measurements that can be achieved from the ground or from balloons.

14.1. Intrinsic photon counting noise: Ideal detector behavior In radio astronomy (where the frequency of observations is typically lower than for CMB observations) the following expression, known as the radio astronomers’ equation,45,110 gives the fractional error of an observation under slightly idealized

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-103

assumptions: 2 δI Tsky + Tsys 1 = . I Tsky (∆B)tobs

(83)

Here tobs is the time of observation of a pixel and (∆B) is the bandwidth. Tsky is the sky brightness temperature averaged over the pixel and Tsys represents the additional noise introduced within the receiver. This expression can be understood by considering the source in the sky as having thermal statistics — that is, ﬂuctuating in intensity as a Gaussian random ﬁeld. In radio astronomy, detection is coherent and each detector picks up only a single transverse mode of the electromagnetic ﬁeld incident from the sky, which is converted by an antenna or a feed horn into an electrical signal, which depends only on time. This signal is coherently ampliﬁed and nowadays digitized. Mathematically, in the absence of additional detector noise, the sky signal can be thought of as a signal whose complex amplitude ﬂuctuates according to a Gaussian distribution with a variance proportional to Tsky . The coherence time is 1/(∆B), so in a time interval tobs , Nsamp = 2(∆B)tobs independent realizations of this Gaussian random process whose variance we are trying to measure are collected, leading to a fractional error in the determination of the variance, or Tsky , of 1/ Nsamp . For a more realistic measurement noise is also introduced, and in the above formula the noise is idealized as an independent additive Gaussian random signal, characterized in terms of a system temperature Tsys , to be added to Tsky to obtain the variance actually measured. The additive noise lumped together and known as the “system” temperature includes thermal emission from the atmosphere, Johnson noise from dielectric losses in the feedlines, noise in the detector itself — in other words, everything other than the sky temperature that one would measure from space with an ideal measuring device. The above formula was derived treating the incoming electromagnetic ﬁeld as entirely classical, and for a thermal source this treatment is valid in the R– J part of the spectrum — that is, when Tsource hν/kB . For the CMB, with TCMB = 2.725 K, this requires that ν < ∼ νCMB = kB TCMB /h = 57 GHz, meaning that quantum eﬀects introduce signiﬁcant corrections to the above formula, but one is still far from the regime where photons arrive in a nearly uncorrelated way, obeying Poissonian counting statistics, which is what happens for observations in the extreme Wien tail of the blackbody distribution. Before introducing the quantum corrections to Eq. (83), let us ﬁnish our discussion of this result based on classical theory extracting all the lessons to be learnt from this result. In the R–J regime, where N 1, N being the photon occupation number, roughly speaking, photons arrive in bunches of N photons, so the discreteness of the photons is not an issue. As long as a fraction greater than about 1/N of the photons is captured, little additional noise is introduced. The intrinsic noise of the incoming electromagnetic ﬁeld is almost entirely classical in origin and can be modeled faithfully using stochastic

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-104

classical electromagnetic formalism. (For a more detailed discussion of these issues, see Ref. 129.) From the above formula, we learn that noise can be reduced by the following measures: (i) choosing as wide a bandwidth as possible, (ii) increasing the observation time, and (iii) increasing the number of detectors. We also learn that although lowering the internal noise of the detection system increases the accuracy of the measurements, once the system temperature is approximately equal to the brightness temperature of the source, lowering the system temperature further leads only to marginal improvement in sensitivity. We now derive the quantum corrected version of the radio astronomer’s equation, using a Planck distribution instead of the classical description based on a deterministic stochastic Gaussian ﬁeld. For the Planck distribution, where pN = xN /(1 − x) with x = exp(−hν/kB T ), ¯2 + N ¯. (δN )2 = N 2 + N = N

(84)

¯ 1, the ﬁrst term dominates, reproducing the result In the R–J regime, where N obtained using a classical random ﬁeld as described above. But as the frequency is increased and one starts to pass toward the Wien tail of the distribution, the second term increases in importance. In the extreme Wien tail, where the second term dominates, we observe the ﬂuctuations characterized by Poisson statistics, where the (rare) arrivals of photons are completely uncorrelated. This means that the fractional error for the number of photons counted must be increased by a factor of

hν 1 , (85) 1 + ¯ = exp 2kB (Tsky + Tsys ) N and the quantum-corrected version of Eq. (83) becomes113 δI = I

Tsky + Tsys Tsky

2

1 hν exp . 2kB (Tsky + Tsys ) (∆B)tobs

(86)

14.2. CMB detector technology Having analyzed the performance of an ideal detector limited only by the intrinsic ﬂuctuations of the incoming electromagnetic ﬁeld, we now review the state of the art of existing detector technologies, which may be divided into two broad classes: (i) coherent detectors, and (ii) incoherent detectors. Coherent detectors transfer the signal from a feed horn or antenna onto a transmission line and then coherently amplify the signal using a low-noise ampliﬁer, generally using high electron mobility transistors (HEMTs) and switching with a cold load or between diﬀerent points on the sky in order to mitigate 1/f noise.183 Coherent detection is an older technology. Coherent detectors have the advantage

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-105

that they can operate at much higher temperatures than the competing incoherent bolometric detectors. Avoiding active cooling through a cryogenic system is a deﬁnite plus, especially in space, where reducing risk of failure is important. However, unlike their incoherent competitors, at the high frequencies of interest for CMB observations, coherent detectors have noise levels far above the quantum noise limit, and until now coherent detectors have not been able to reach frequencies above circa 100 GHz, although some believe that this situation may improve. It is not presently known how coherent ampliﬁcation would function in the Wien part of the blackbody spectrum. Coherent detectors were used for the COBE DMR experiment, for WMAP (including the frequencies 20, 30, 40, 60, and 90 GHz), and the Planck LFI (low frequency instrument), which included the frequencies 30, 45, and 70 GHz. One advantage of coherent detectors is their insensitivity to cosmic rays. Another property of coherent detectors is that for a single mode measurement, all four Stokes parameters (i.e., I, Q, U, V) can be measured simultaneously. For incoherent detectors one can measure only two Stokes parameters simultaneously. However for measurements outside the R–J part of the spectrum, this advantage of coherent detectors disappears, because in the Wien part of the spectrum the photons are not bunched to any appreciable degree. On the other hand, for coherent detectors, it is not possible to construct multi-moded detectors, where the number of detectors can be greatly reduced by sacriﬁcing angular resolution, but not at the cost of less sensitivity on large angular scales. Coherent detectors are always single-moded, sampling at the diﬀraction limit. Incoherent detectors, unlike their coherent counterparts, do not attempt to amplify the incident electromagnetic wave. Rather the incident electromagnetic wave, or equivalently the stream of photons, is directly converted into heat, changing the temperature of a small part of the detector of small heat capacity, whose temperature is monitored using a thermistor of some sort. Typically the detector is cooled to an average temperature much lower than that of the incident radiation. The bolometers of the Planck high frequency instrument (HFI) were cooled to 0.1 K. Consequently each CMB photon produces many phonons in the detector, and therefore the counting statistics of the phonons do not to any appreciable extent limit the measurement. Heat ﬂows from the sky onto the detector and is then conducted to a base plate of high heat capacity and thermal stability. A thermal circuitry (which can be modelled as a sort of RC circuit with as many loops as time constants) is chosen with a carefully optimized time response. Several technologies are used for the thermistor. One of the most popular nowadays is transition edge sensors (TES),117 where a superconducting thin ﬁlm is kept at the edge of the superconducting transition, so that minute changes in temperature give rise to large changes in resistivity. A feedback circuitry with heating is used to maintain the assembly near a ﬁxed location on the normal-superconducting transition edge. One disadvantage of TES detectors is their low electrical impedance,

May 4, 2017 9:55 b2167-ch15

II-106

M. Bucher

which requires that SQUIDs (superconducting quantum interference devices), which need careful magnetic shielding, be used for the readout. Several new promising technologies are being developed and perfected for the next generation of bolometric detectors. One such technology is kinetic inductance detectors (also known as KIDs).41,47 For these detectors, a resonant electric circuit is constructed by depositing a pattern of superconducting ﬁlm on a thick dielectric. Photons incident on the superconducting ﬁlm break Cooper pairs causing its surface impedance to vary with the incident photon ﬂux, which in turn alters the resonant frequency. The shifts in frequency of the resonator allow the KID to be read out in a simple way without using SQUIDs. The surface impedance of the superconducting ﬁlm is largely determined by the inertia of the superconducting electrons, hence the name “kinetic inductance.” Another possibly promising technology is the cold electron bolometer,112 involving a small metal ﬁlament connected to a superconductor on both sides by means of a junction with a thin insulator in between. A bias current that passes through the assembly acts to cool the electrons in the thin wire through a Peltier-like eﬀect to a temperature far below that of the substrate. CEBs are hoped to be more resistant to cosmic ray spikes than other detector technologies. Cosmic rays have proved to be more problematic than anticipated in the Planck experiment, and substantial eﬀort was needed to remove the contamination from cosmic ray events in the raw time stream.181 Some of the data was vetoed and corrections were applied to the long-time tails of the larger magnitude events. While experiments from the ground beneﬁt from substantial shielding of cosmic rays by the Earth’s atmosphere, how best to minimize interference from cosmic rays will be a major challenge for the next generation of experiments from space, which target sensitivities more than two orders of magnitude beyond the Planck HFI. Another challenge will be multiplexing, which is necessary to reduce the cooling requirements when the number of detectors is greatly expanded.

14.3. Special techniques for polarization If it were not for the fact that the polarized CMB anisotropy (i.e., the Q and U Stokes parameters) is much smaller than the anisotropy in the I Stokes parameter, measuring polarization would not pose any particular problems, because most detectors are polarization sensitive or can easily be made so, and there would be no need to discuss any techniques particular to polarization measurements. An inspection of the relative amplitudes of the spectra in Fig. 21 highlights the problems encountered. In particular if one wants to search for B modes at the level of r ≈ 10−3 or better, the requirements for preventing leakages of various types become exceedingly stringent, as we now describe. Let us characterize the problem for the most diﬃcult polarization observation: measuring the primordial B modes. We may classify the leakages to be avoided

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-107

Fig. 21. Summary of T,E, and B anisotropy power spectra from scalar and tensor modes. The green curves (from top to bottom) represent the TT, TE and EE CMB power spectra for the “scalar” mode, while the blue curves represent the CMB isotropies for the tensor mode. The broken blue curves (from top to bottom) and the top solid blue curve represent the TT, TE, EE and BB anisotropies assuming a value of the tensor-to-scalar ratio of r = 0.1. The lower two solid curves represent the predicted BB anisotropy for r = 0.01 and r = 0.001. The red curve shows the BB power spectrum arising nonlinearly from the scalar mode as a result of gravitational lensing. (Credit: M. Bucher)

by ordering them from most to least severe: (i) leakage of the CMB temperature monopole to the B mode, (ii) leakage of the T anisotropies to the B mode, and (iii) leakage of the E mode to the B mode. One of the solutions to many of these problems is polarization modulation. The idea is to place an element such as a rotating half-wave plate, whose orientation we shall denote by the angle θ, between the sky and the detector, preferably as the ﬁrst element of the optical chain, as sketched in Fig. 22. Half-wave plates can be constructed using anisotropic crystals such as sapphire or using a mirror with a layer of wires a certain distance above it so that one linear polarization reﬂects oﬀ the wires while the other reﬂects oﬀ the mirror, resulting in a diﬀerence in path length. Mathematically, we have

(out) Ex (out)

Ey

=

=

+ cos(θ)

−sin(θ)

+ sin(θ)

+ cos(θ)

+ cos(2θ)

+ sin(2θ)

−sin(2θ)

−cos(2θ)

1

0

0

−1

(in)

Ex

(in)

Ey

+ cos(θ)

+ sin(θ)

−sin(θ)

+ cos(θ)

(in)

Ex

(in)

Ey

,

(87)

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-108

Fig. 22. Polarization modulation by means of a rotating half-wave plate. We show a transmissive rotating half-wave plate (HWP), here placed in front of a microwave feed horn pointing directly at the sky. However for modern CMB experiments, the horn would generally be replaced with a telescope having intermediate optical elements and horns on the focal plane. The rotating HWP is birefringent, with its fast polarization axis represented by the line OO . As explained in the main text, by measuring only that part of the signal varying at an angular frequency 4Ω (where Ω is the angular velocity of the rotating HWP), one can prevent leakage from the much larger Stokes parameter I into the linearly polarized components Q and U. The presence of the HWP, while complicating the instrument design somewhat, allows many hardware requirements to be relaxed substantially compared to what would be needed without polarization modulation.

so that in terms of the Stokes parameters, expressed below as the matrix-valued expectation value Ei Ej∗ , so that

I + Q U + iV U − iV I − Q det

I +Q cos 2Ωt sin 2Ωt = U − iV sin 2Ωt −cos 2Ωt

=

U + iV I −Q

sky

cos 2Ωt

sin 2Ωt

sin 2Ωt

−cos 2Ωt

I + cos(4Ωt)Q + sin(4Ωt)U

−cos(4Ωt)U + sin(4Ωt)Q − iV

−cos(4Ωt)U + sin(4Ωt)Q + iV

I − cos(4Ωt)Q − sin(4Ωt)U

. (88)

By measuring only the component that varies with an angular frequency 4Ω, we can measure the polarization without making two independent measurements of large numbers that are then subtracted from each other. The much larger Stokes intensity I does not mix because it remains constant in time. To see how polarization modulation helps, let us suppose that the vector (ideal) (ideal) (ideal) (ideal) (ideal) (actual) Pdet = (Idet , Qdet , Udet , Vdet ) is replaced by the vector Pdet , with (actual) (actual) (actual) (actual) (actual) (ideal) Pdet = (Idet , Qdet , Udet , Vdet ) = (I + )Pdet , where I is the

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-109

Fig. 23. Planck polarization sensitive bolometers. Shown above is a photograph of one of the Planck polarization sensitive bolometers consisting of two grids of parallel wires oriented orthogonally and placed on top of each other but separated by a small distance. Each grid has its own thermistor, allowing the two selected components of the linear polarizations to be read out separately. (Credit: ESA/Planck Collaboration)

identity matrix and represents a hopefully small but unknown uncorrected residual error in the actual linear response of the detector pair. We discuss the setup for measuring polarization used in Planck as described above (see Fig. 23). We describe some of the possible systematic eﬀects and how polarization modulation by means of a rotating half-wave plate would remove these systematic eﬀects. Ideally the two orthogonal grids of wires would correspond but for the direction of the linear polarization to exactly the same circular beam on the sky and be perfectly calibrated, with no oﬀset between the beam centers, nor any time-dependent drift of the diﬀerential gain. But in practice the beams are not identical and the gain and oﬀset of the electronics drift in time. The centers of the beams do not coincide, and without corrections this would cause a local gradient of the T to be mistaken for E or B polarization. Diﬀerential ellipticity would cause the second derivative of T to leak into E and B. Moreover, the actual beams are ˆ Q(Ω), ˆ and U (Ω)] ˆ more complicated and require three functions [i.e., maps of I(Ω), for a complete description. All the above eﬀects disappear, at least ideally, with the polarization modulation scheme described above. Polarization modulation can resolve or mitigate a lot of problems but is not a panacea. For example, polarization modulation cannot prevent E → B leakages. If there are any errors in the calibration of the angles of the direction of the linear polarization in the sky, these errors act

May 4, 2017 9:55 b2167-ch15

II-110

M. Bucher

to rotate E into B and vice versa. Moreover polarization modulation has been diﬃcult to realize in practice for a number of reasons such as spurious signals due to microphonic coupling. Both continuously rotating and discretely stepped halfwave plates are options for polarization modulation. 15. CMB Statistics Revisited: Dealing with Realistic Observations The analysis of a CMB experiment going from the raw time stream of data taken, to sky maps at a given frequency, then to a clean sky map containing only the CMB, and ﬁnally to a likelihood function, whose input is a complete theoretical power spectrum, involves many steps and details. Space does not permit a complete discussion. The following Planck papers and references therein describe this analysis for the Planck experiment from the Planck 2013 Results series. See in particular the following papers: Overview of products and results,161 Low frequency instrument data processing,179 High frequency instrument data processing,180 Component separation,162 and HFI energetic particle eﬀects.181 Similar descriptions can be found for other experiments. In this section, we limit ourselves to sketching a few issues and to describing an idealized statistical analysis. One of the main steps in CMB analysis is constructing a likelihood for the data given a particular theoretical model. Relative likelihood lies at the heart of almost all statistical inference, whether one uses frequentist methods, Bayesian methods, or an ecumenical approach borrowing from both doctrines. Here we shall discuss only how to formulate the likelihood, as how to exploit the likelihood is less speciﬁc to the CMB and details can be found in standard treatises on modern statistics. (See for example Ref. 125 for a brief overview and Ref. 210 for an authoritative treatment emphasizing the frequentist approach. For Bayesian sampling as applied to the CMB, see Refs. 122 and 57 and references therein.) Constructing a likelihood is straightforward for an idealized survey for which the noise in the sky maps is Gaussian and isotropic. For the simplest likelihood function, the input argument is a model for the CMB sky signal, or more precisely the parameters deﬁning this model. This theoretical model deﬁnes an isotropic Gaussian stochastic process whose parameters one is trying to infer. In order to introduce some of the complications that arise, we start by writing down the likelihood L for the data given the model, or rather the variable −2 ln[L], which in some ways resembles χ2 in the Gaussian approximation. We have TT,(sky) ∞ TT,(th) + N B C + N C (2 + 1) ln −2 ln(L) = + −1 . TT,(sky) TT,(th) C + N B C + N =2 (89) Here the parameters N represent the noise of the measurement. In the approximation of white instrument noise — that is, with no unequal time correlations — we have N = N0 . The factor B represents the smearing of the sky signal due to the

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-111

ﬁnite beam width. For the beam proﬁle approximated as having a Gaussian shape, where the sky intensity map is convolved with the Gaussian funcB = exp[−σ 2 2 ]√ tion K(θ) = (1/ 2πσ 2 ) exp[−θ2 /(2σ 2 )].u Note that we ignore constant oﬀsets in the log likelihood because for most statistical analysis only diﬀerences in the log likelihood are relevant. The presence of the normalization of the Gaussian implies that a quadratic representation of the log likelihood is not exact and higher order corrections are needed to avoid bias, especially at low . To generalize the above expression to include polarization, we deﬁne the matrix

CTE CTT + B −1 NTT , (90) C = CTE CEE + B −1 NEE so that −2 ln(L) =

∞

(th) (obs) (obs) (th) (2 + 1) ln[C (C )−1 ] + tr[C (C )−1 ] − 2 .

(91)

=2

In order to generalize to less idealized (i.e., more realistic) situations, it is useful to rewrite the above results in a more abstract matrix notation, in which the expressions can be interpreted simultaneously as expressions in real (i.e., angular) space and as expressions in harmonic space, so that −2 ln[L] = det(Cth + N) + tT (Cth + N)−1 t.

(92)

Here the vector t represents the observed sky map, Cth is the covariance matrix of the underlying sky signal (in general depending on a number of parameters whose values one is trying to infer), and N represents the covariance matrix of the instrument (and other) noise. In the idealized case where everything is isotropic, this is simply a rewriting of Eq. (89), but in less idealized cases the likelihood in Eq. (92) is correct whereas Eqs. (89) and (91) are not applicable to the more general case. Let us now consider the eﬀect of partial sky coverage, which may result from a survey that does not cover the whole sky or from masking portions of the sky such as the galaxy and bright point sources where contamination of the primordial signal cannot be corrected for in a reliable way. If the vectors and matrices in Eq. (92) are understood as representations in pixel space, at least formally there is no diﬃculty in evaluating this expression in the subspace of pixel space representing the truncated sky. If the dimensionality of the pixelized maps were small, there would be no diﬃculty in simply evaluating this expression by brute force. Another complication arises from inhomogeneous noise. Most surveys do not cover the sky uniformly because requirements such as avoiding the sun, the earth, and the planets as well as instrumental considerations lead to nonuniform sky coverage. This leads to a u Note that in the literature, beam widths are generally speciﬁed in terms of FWHM rather than using the Gaussian deﬁnition above.

May 4, 2017 9:55 b2167-ch15

II-112

M. Bucher

noise matrix that has a simple representation in pixel space (under the assumption of white noise), but not in harmonic space. Unfortunately evaluating likelihoods such as in Eq. (92) exactly is not possible for full-sky surveys at high angular resolution such as the Planck survey for which pixelized maps include up to about Npix = 5 × 107 pixels. While Cth is simple (i.e., sparse) in harmonic space, it is dense in pixel space, and the opposite holds for N (ignoring correlated noise), so no representation can be found to simplify the calculation in which all the matrices are sparse. Since even enumerating the matrices involves O(Npix 2 ) operations, and inverting these matrices or taking their determinant requires O(Npix 3 ) operations, it is immediately apparent that brute force will not work, and more clever, approximate techniques are required. Pixel based likelihoods are however used at low , where other approximations have diﬃculty representing the likelihood, in large part because the cosmic variance is non-Gaussian. While the likelihoods above assumed an isotropic Gaussian stochastic process for the underlying theoretical model generating the anisotropy pattern in the sky, these likelihoods can be generalized to include nonlinear eﬀects, such as nonGaussianity through fNL , gravitational lensing, and certain models of weak statistical anisotropy, to name just a few examples. Using the likelihood as a starting point, possibly subject to some approximations, has the virtue that one is guaranteed that an optimal statistical analysis will result without any special ingenuity. In the above simpliﬁed discussion, we have not touched on a number of very important issues, of which we list just a few examples: asymmetric beams, far sidelobe corrections, bandpass mismatch, errors from component separation, estimating noise from the data, and null tests (also known as “jackknives”). The reader is referred to the references at the beginning of this Section for a more complete discussion and references to relevant papers. 16. Galactic Synchrotron Emission An inspection of the single frequency temperature maps from the Planck space mission (see Fig. 8), shown here in galactic coordinates so that the equator corresponds to the galactic plane, suggests that the 70 GHz frequency map is the cleanest, as the excess emission around the galactic plane is the narrowest at this frequency. But at lower frequencies, the region where the galactic contamination dominates over the primordial CMB temperature anisotropies widens. This contamination at low frequencies is primarily due to galactic synchrotron emission, resulting from ultrarelativistic electrons spiralling in the galactic magnetic ﬁeld, whose strength is of order a few microgauss. (See for example Ref. 78.) This emission is generally described as nonthermal because the energy spectrum of charged particles is non-Maxwellian, especially at high energies. The electrons in question reside in the high energy tail of the electron energy spectrum, described empirically as a falling power law spectrum. Models of cosmic ray acceleration, for example arising about

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-113

a shock from an expanding supernova remnant, are able to explain spectra having such a power law form.20 If the radiating charged particles were nonrelativistic, their radiation would be emitted at the cyclotron frequency ωc = eB/mc, with very little emission in higher harmonics, and given the measured values of the galactic magnetic ﬁeld, one could not thus explain the observed high frequency emission. However, if the charged particles are highly relativistic, due to beaming eﬀects the bulk of the radiation is emitted in the very high order harmonics, allowing the observed emission to be explained for reasonable values of the galactic magnetic ﬁeld.69 Although the data concerning the three-dimensional structure of the galactic magnetic ﬁeld is at present in a very rudimentary state, and little is known observationally about the cosmic ray spectrum outside our immediate solar neighborhood, the physical mechanism underlying galactic synchrotron emission is well understood physics. One important point is that no matter what the details of the cosmic ray energy spectrum may be, the kernel of the integral transform relating this distribution to the frequency spectrum of the observed synchrotron emission applies considerable smoothing. Thus we can assume with a high degree of conﬁdence that the synchrotron spectrum is smooth in frequency. Theory also predicts that this synchrotron emission will be highly polarized, at least to the extent that the component of the galactic magnetic ﬁeld coherent over large scales is not negligible. This expectation is borne out by observations. It is customary to describe the synchrotron spectrum using a power law, so that its R–J brightness temperature may be ﬁt by an Ansatz of the form

α ˆ ν) = TR−J (Ω, ˆ ν0 ) ν , (93) TR−J (Ω, ν0 where empirically it has been established that α ≈ 2.7–3.1. The same holds for the polarization — that is, the Stokes Q and U (and also V ) parameters. In the ultrarelativistic approximation this emission is primarily linearly polarized, but there is also a smaller circularly polarized component suppressed by a factor of 1/γ. Variations in the spectral index depending on position in the sky have been observed.67 A description of the WMAP full-sky observations of synchrotron emission appears in the ﬁrst-year WMAP foregrounds paper229 and the polarization observations are described in the three-year WMAP paper on foreground polarization.235 A greater lever arm, in particular for studying the synchrotron spectral index, can be obtained by including the 408 MHz Haslam map covering almost the entire sky.80 17. Free–Free Emission Another source of low-frequency galactic contamination is free–free emission arising from Bremsstrahlung photons emitted from electron–electron collisions in the interstellar medium, and to a lesser extent from electron-ion collisions. Like galactic synchrotron emission, the free–free emission brightness temperature falls with

May 4, 2017 9:55 b2167-ch15

II-114

M. Bucher

increasing frequency, but the fall-oﬀ is slower than that of the galactic synchrotron emission. Unlike synchrotron emission, free–free emission is not polarized. Hα (n = 3 → n = 2) emission resulting from recombination of ionized atomic hydrogen can be used as a template for removing the free–free component, because the Hα emission likewise is proportional to the square of the electron density. However, because the ΓHα (T ) and Γfree–free (T, ν) do not have an identical temperature dependence, free– free removal using an Hα template has some intrinsic error. ΓHα (T ) is deﬁned as the constant of proportionality in the emissivity relation: Hα = ΓHα (T )ne 2 where Hα is a bolometric emissivity (power per unit volume) and ne is the density of free electrons. Likewise free–free(ν) = Γfree–free (T, ν)ne 2 where the emissivity has units of power per unit frequency per unit volume per unit solid angle. 18. Thermal Dust Emission In Sec. 16 we discussed galactic synchrotron emission, whose R–J brightness temperature rises with decreasing frequency, making it exceedingly diﬃcult to measure CMB anisotropies at frequencies below ≈ 20 GHz, especially close to the galactic plane where the galactic synchrotron emission is most intense. We also saw how the fact that the brightness temperature increases with decreasing frequency can be understood as a consequence of the synchrotron optical depth being much smaller than unity and increasing with decreasing frequency. Eventually when the frequency is low enough so that the optical depth is near one, the brightness temperature stops rising and approaches a temperature related to the cosmic ray electron kinetic energy. This section considers another component contributing to the microwave sky: the thermal emission from interstellar dust, whose brightness temperature increases in the opposite way — that is, the brightness temperature increases with increasing frequency. This behavior results because at microwave frequencies, the wavelength greatly exceeds the typical size of a dust grain. At long wavelengths, the crosssection for absorbing or elastically scattering electromagnetic radiation is much smaller than the geometric cross-section σgeom ≈ a2 where a is the eﬀective grain radius. The dust grains can be approximated as dipole radiators in this frequency range because higher order multipoles are irrelevant. The dependence of the crosssection on frequency can be qualitatively understood by modeling the electric dipole moment of the grain as a damped harmonic oscillator where the resonant frequency is much higher than the frequencies of interest. In this low frequency approximation a 2 a2 (94) σabs ∼ λ for absorption, and for the elastic, or Rayleigh, scattering component, which becomes subdominant at very low frequencies, a 4 σelastic ∼ a2 . (95) λ

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-115

This is a simplistic approximation that we later shall see is not accurate, but it does provide intuition about the qualitative behavior expected. Our knowledge of interstellar dust derives from combining diﬀerent observations spanning a broad range of frequencies, from the radio to the UV and even X-ray bands, in order to put together a consistent theoretical model able to account simultaneously for all the observations. (For a nice recent overview, see for example Refs. 48 and 49.) In this section, we restrict ourselves to discussing the thermal microwave emission properties of the dust. An empirical law commonly used to model thermal dust emission at low frequencies is a Planck blackbody spectrum modulated by a power law emissivity, or graybody factor,

β ν B(ν, Tdust ) , (96) Idust (ν) = Idust (ν0 ) ν0 B(ν0 , Tdust ) where the Planck spectrum is given by B(ν; T ) =

2hν 3 1

. hν c2 exp −1 kB T

(97)

The physical basis for this modiﬁcation to the Planck spectrum can be partially motivated by the following argument based on the linear electric susceptibility of the dust grain, following a line of reasoning ﬁrst suggested by EM Purcell in 1969.188 The simple argument, subject to a few caveats, suggests that β = 2. However the observational data do not bear out this prediction and instead are better ﬁt by a power law with β ≈ 1.4 − 1.6.144,158 Let χ(ω) represent the linear susceptibility of a dust grain, which as a consequence of the Kramers–Kronig dispersion relations must include both nonvanishing real and imaginary parts. Here d(ω) = χ(ω)E(ω), where d(ω) and E(ω) are the grain electric dipole moment and the surrounding electric ﬁeld, respectively. As a consequence of causality, χ(ω) is analytic on the lower half-plane. Its poles in the upper half-plane represent those decaying mode excitations of the grain that are coupled to the electromagnetic ﬁeld. Moreover, on the real axis χ(−ω) = [χ(+ω)]∗ . If we assume analyticity in a neighborhood of the origin, we may expand as a power series about the origin, so that χ(ω) = ω0 + iχ1 ω + χ2 ω 2 + · · · ,

(98)

and χ0 , χ1 , χ2 , . . . are all real. It follows that the energy dissipated is characterized by the absorptive cross-section σabs (ω) = and at low frequencies

4πωIm[χ(ω)] , c

σabs (ω) ≈

4πχ1 c

(99)

ω2,

(100)

May 4, 2017 9:55 b2167-ch15

II-116

M. Bucher

which corresponds to an emissivity index β = 2 as ω → 0. The perhaps questionable assumption here is that there exists a circle of nonzero radius about the origin of the ω-plane where χ(ω) has no singularities. If we postulate that there is an inﬁnite number of poles of suitably decreasing strength whose accumulation point is ω = 0, we may evade the conclusion that asymptotically as ω → 0, β → 2. For an insightful discussion of how such low-frequency poles could arise, see C. Meny et al.135 19. Dust Polarization and Grain Alignment The fact that interstellar dust grains are aligned was ﬁrst seen through the polarization of starlight at optical frequencies in 1949.76,89,90 Since this discovery an abundant literature has been amassed elucidating numerous aspects of dust grain dynamics. The best bet solution to the alignment problem has evolved over the years as additional relevant physical eﬀects were pointed out. It is unclear whether the current understanding will remain the last word on this subject. The most naive solution to the alignment problem would be some sort of compass needle type mechanism with some dissipation. However, it was discovered early on that in view of the plausible magnitudes for the galactic magnetic ﬁeld, such an explanation cannot work.207 The theoretical explanation that has evolved involves detailed modeling of the rotational dynamics of the grains, which as we shall see below almost certainly rotate suprathermally. For microwave observations of the primordial polarization of the CMB at ever greater sensitivity, the removal of contamination from polarized thermal dust emission is of the utmost importance. This fact is highlighted by the discussion of how to interpret the recent BICEP2 claim (discussed in Sec. 11.1) of a detection of a primordial B mode signal based on essentially a single frequency channel. This interpretation however seems unlikely given subsequent Planck estimates of the likely dust contribution.159 The problem is that little data is presently available on polarized dust emission at microwave frequencies. Current models of polarized dust involve a large degree of extrapolation and are in large part guided by simplicity. From a theoretical point of view, polarized dust could be much more complicated than the simplest empirical modeling suggests. For example there is no a priori reason to believe that polarized emission from dust must follow the same frequency spectrum as its emission in intensity. It could well be that the weighting over grain types gives a diﬀerent frequency dependence, or that as more precise dust modeling is called for, the need to include several species of dust with spatially nonuniform proportions will become apparent. Moreover, based on theoretical considerations, it is not clear that there should be a simple proportionality between polarization and the strength of the magnetic ﬁeld. If the BICEP2 B mode signal is conﬁrmed as a dust artefact, we could see ourselves in the coming years trying to detect values of the tensor-to-scalar ratio around r ≈ 10−3 , which would require exceedingly accurate polarized dust removal from the individual frequency maps, and before this happens a wealth of new observational data on polarized dust emission will

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-117

become available. If, on the other hand, a detection of B modes at a large value of r is made, as one tries to map out the precise primordial B mode spectrum, for example to measure nT and thus check the consistency between PS (k) and PT (k), the need for accurate dust removal will likewise be great, but for slightly diﬀerent reasons. Since the observational inputs are likely to change, here we particularly emphasize the underlying physics of the alignment, which may be updated, but is likely to change less. The theoretical issues of dust alignment have so far not been widely discussed among those not specialized in the ISM (interstellar medium). Nevertheless, we believe that the questions raised will become increasingly relevant for CMB observations in the future, and this is why a thorough discussion is presented here. We shall ﬁnd that the alignment of interstellar dust grains is closely linked to the rotational dynamics of a dust grain. We therefore start with a study of the rotation of a dust grain. Mathematically, the simplest case to analyze is a prolate (needle-like) or oblate (ﬂattened spherical) dust grain with an axis of azimuthal symmetry, which constitutes a symmetric top. In this case, the motions can be solved in terms of elementary functions. For the more general case of an asymmetric grain (where none of the three principal values of the moment of inertia tensor are equal) integration of the equations of motion in the absence of torques is more diﬃcult, but can be achieved using Jacobian elliptic functions.225 In the noninertial frame rotating with the dust grain, which is convenient because in this frame the moment of inertia tensor is constant in time, two types of inertial forces arise: (i) a centrifugal force, described by the potential Vctr (x)/m = −Ω2 (t)x2 + (Ω(t) · x)2 , and (ii) a Coriolis force Fcor /m = −2Ω(t) × v, which for our purposes is unimportant because the velocities are negligible in the co-rotating frame. In order to understand the alignment process, we must model the solid as being elastic with dissipation included. 19.1. Why do dust grains spin? A ﬂippant answer to the above question might be: Why not? If we can associate a temperature with the three kinetic degrees of freedom of the dust grain, we would conclude that 3 1 2 Iω ∼ kB Trot . (101) 2 2 Early discussions of rotating dust grains supposed that the rotational degrees of freedom are subject to random torques, so that Γ = 0.

(102)

Later it was realized that suprathermal rotation is plausible, where the implied temperature of the rotation can greatly exceed the eﬀective temperature of any of the other relevant degrees of freedom. The question now arises what temperature one should use for Trot . A natural candidate might be the kinetic temperature of the ambient gas Tkin . Indeed one

May 4, 2017 9:55 b2167-ch15

II-118

M. Bucher

expects molecules of the ambient gas to collide with the dust grains, and such collisions do provide random torques, causing the angular momentum to undergo a random walk accompanied by a dissipation mechanism, as must be present to prevent the angular velocity from diverging with time. 19.2. About which axis do dust grains spin? This question has two parts. First, we may ask what is the alignment of the corotating internal (body) coordinates with respect to the angular momentum vector. Second, we may ask how the angular momentum aligns itself with respect to the inertial coordinates of the ambient space. In this section, we consider only the ﬁrst question, postponing the second question to a later section. But for exceptional situations where two or more eigenvalues of the moment of inertia tensor coincide, the principal axes of moment of inertia tensor provide a natural set of axes for the dust grain, and we may order the eigenvalues so that I1 < I2 < I3 . We argue that the dust grain tends to align itself so that its I3 axis is parallel to the angular momentum. This orientation of the grain minimizes the rotational energy subject to the constraint that the total angular momentum remains constant. If the grain starts in a state of random orientation, the grain wobbles. While L remains constant, Ω(t) = I−1 (t)L (expressed in inertial coordinates) does not, but rather ﬂuctuates, exciting internal vibrational modes that dissipate. As the excess rotational energy is dissipated, an alignment as described above is achieved.v The relevant question is how this time scale compares to other time scales of the grain rotation dynamics. It turns out that this time scale, which we denote τI−L , is short. In the equations that follow, we exploit the shortness of τI−L compared to the other relevant time scales. This hierarchy of time scales allows us to simplify the equations by excluding the degrees of freedom that describe the lack of alignment of the principal axis of maximum moment of inertia from L, leading to a considerable simpliﬁcation of the formalism. 19.3. A stochastic diﬀerential equation for L(t) Mathematically, we may describe the combined eﬀects of the torque from random collisions and the associated damping through the coherent torque using a stochastic diﬀerential equation of the form

˙ ¯ L(t) + Γran (t). (103) L(t) = −α L(t) − L |L(t)| v Much to the dismay of NASA engineers, this is exactly what happened to the Explorer 1 rocket, the ﬁrst US satellite launched in 1958 following Sputnik, due primarily to dissipation from its whisker antennas. The elongated streamlined cylinder was initially set in rotation about its long axis, but subsequently, after an initial wobble phase, ended up spinning about an axis at right angles in the body coordinates. Since then rotational stability, also relevant in ballistics, has become a carefully considered design issue for spacecraft.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-119

In this equation, the torque described by the ﬁrst term is coherent with respect to the body coordinates of the dust grain. Later we shall also consider torques that are coherent with respect to the ambient inertial coordinates, which must be modeled in a diﬀerent way. Here the stochastic Gaussian forcing term has the expectation values Γran,i (t) = 0, Γran,i (t)Γran,j (t ) = µL δij δ(t − t ),

(104)

where the angular momentum diﬀusion coeﬃcient has units [µL ] = (Torque)2 /(Time). Let us analyze the stationary ensemble deﬁned by the above equation — in other words, the probability function p(L) toward which the biased random walk deﬁned by the above equation evolves after a long period of time. In the low¯ with negligible fractional ﬂuctuations, and the direction temperature limit, |L| ≈ L ˆ L = L(t)/|L(t)| follows a random walk on the of the angular momentum vector n ¯ 2 . In the opposite high surface of the sphere with diﬀusion coeﬃcient µnˆ = µL /L ¯ temperature limit, the L term is a sideshow or small correction, and the main eﬀect is the competition between the ﬁrst and second terms leading to a thermal ensemble, which for an isotropic moment of inertial tensor would take the form

3/2 β p(L) = exp[−βL2 ], (105) 2π where β = α/µL . Without other sources of dissipation or random torques, we expect that β = 1/(2kB Tkin Imax ) where Imax is the largest eigenvalue of the moment of inertia tensor of the particular grain being modelled and Tkin is the kinetic temperature of the ambient gas. It is straightforward to generalize Eq. (105) to the intermediate case. It was however later noted that a number of mechanisms exist for which Γ = 0, which can give rise to what is known as “suprathermal rotation”189 — that is, rotation at an angular velocity larger than any temperature associated with the other degrees of freedom interacting with the dust grain. 19.4. Suprathermal rotation To see how such suprathermal rotation might come about, consider the caricature of a dust grain as sketched in Fig. 24. Let us assume that the dust grain emits radiation isotropically in the infrared, but that in the visible and UV bands radiation is absorbed or reﬂected by the surfaces indicated in the ﬁgure. Let us further assume an isotropic visible or UV illumination. The blackened (perfectly absorbing) faces suﬀer a radiation pressure equal to half that of the mirrored surfaces, giving a coherent torque in the direction indicated in the ﬁgure. The presence of two distinct temperatures, reminiscent of a thermal engine, is crucial to avoid thermal equilibrium for the rotational degree of freedom. As in the thermodynamics of an engine, there is a heat ﬂow from a hotter temperature to a colder temperature with some

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-120

Fig. 24.

Coherent torques from grain radiation pressure.

of the energy being siphoned oﬀ in the form of work exerted on the three rotational degrees of freedom of the dust grain. Other possibilities for coherent torques result from photoelectron emission at privileged sites and the catalysis of molecular hydrogen on favored sites on the grain surface. In the former case, the incident UV photon is almost pure energy due to its masslessness, but the emitted electron carries substantial momentum for which there is a recoil. Molecular hydrogen forms almost exclusively through catalysis on a grain surface. Unlike in ordinary chemistry where truly two-body collisions are an exception and activation barriers prevent reactions from establishing equilibrium, for molecular hydrogen formation in the cold interstellar medium, the bottleneck is the lack of three-body collisions and the lack of an eﬃcient mechanism for carrying away the excess energy, so that the hydrogen atoms combine rather than simply bouncing oﬀ each other. In principle, the energy can be lost radiatively, but the rate for this is too slow. Given that 4.2 eV is liberated, the recoil at selected sites can produce a substantial coherent torque on the grain.

19.5. Dust grain dynamics and the galactic magnetic ﬁeld Much of the dust grain rotational dynamics does not single out any preferred direction. To explain the polarization of starlight and of the thermal dust emission at microwave frequencies, some mechanism is needed that causes the dust grains to align themselves locally at least along a common direction such as that of the magnetic ﬁeld.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-121

19.5.1. Origin of a magnetic moment along L Two eﬀects generate a magnetic moment along the axis of L: (i) Inhomogeneous electric charge distribution in the dust grain. The second moment of the electric charge density about the principal axis of the largest eigenvalue of the moment of inertia tensor (denoted z) induces a magnetic dipole moment as the grain rotates about this axis, according to µ=ˆ zΩz d3 rρe (r)(x2 + y 2 ), (106) where ρe (r) is the electrostatic charge density of the dust grain. In general, and particularly when the grain structure is amorphous, the net charge density is nonvanishing, so the integral above does not vanish. This eﬀect is known as the Rowland eﬀect. (ii) The Barnett eﬀect. Minimizing the free energy tends to align the free spins of a rotating body with the direction of rotation. In nonferromagnetic substances this eﬀect is usually negligible, but rotating dust grains present an extreme environment because of their minute size and isolation from their environment. Interstellar dust grains are among the fastest rotating macroscopic objects known. To understand the origin of the Barnett eﬀect, consider an isolated rotating body, whose conserved total angular momentum may be decomposed into two parts Ltot = Lbulk + Sfree .

(107)

We consider the energetics of the exchange of angular momentum between the bulk and spin degrees of freedom. We assume the spins to be free neglecting any interactions between them. In this case, the Hamiltonian for the system is H=

(Ltot − S)2 L2bulk = , 2I 2I

(108)

and retaining only the term linear in the spin, we obtain H = −Ω · S,

(109)

where Ω(t) = I−1 (t)L is the angular rotation velocity. The spins want to align themselves with the rotation to lower the rotational energy.115 So far we have not included any terms opposing complete alignment, so at T = 0 complete alignment would result with S = Savailable Ω/|Ω|. But at ﬁnite temperature entropic considerations oppose such alignment, and it is the free energy F = H − T S(S) or F (S; Ω) = −Ω · S + aT S2

(110)

that is minimized. Far from saturation, when the degree of alignment is small, this quadratic term provides a good approximation to the entropy of the spin system.

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-122

19.6. Magnetic precession Having established that a spinning dust grain has a magnetic moment parallel to its angular momentum vector, which is aligned with the principal axis of the largest eigenvalue of the moment of inertia tensor, we now consider the eﬀect of the torque resulting from the interaction with the ambient magnetic ﬁeld. The interaction Hamiltonian is H = −µ · B,

(111)

L˙ = µ × B = γm L × B.

(112)

from which it follows that

In other words, the angular momentum vector precesses at a frequency ωprec = γm |B| around the magnetic ﬁeld. Under this precession, the angle between L and µ remains constant, meaning that this eﬀect neither aligns nor disaligns the grain with respect to the local magnetic ﬁeld direction. We estimate the time scale of this precession. Let S denote the spin polarization density (angular momentum per unit volume), which under saturation (i.e., all free spins aligned) would have a magnitude Ssat , whose order of magnitude can be estimated by assuming approximately Avogadro’s number NA of spins having a magnitude in a cubic centimeter volume, giving Ssat ≈ O(1)(4×10−4 )erg · s · cm−3 . We obtain the following free energy density (per unit volume):

2 S 1 F0 (S; T ) = (kB T )Nspin . (113) 2 Ssat Here for a paramagnetic substance the order of magnitude for Nspin might be around 1024 cm−3 but for a ferromagnetic substance could be much smaller, corresponding to the density of domains. Minimizing the free energy with respect to S gives the spin polarization density S=

2 Ssat = IS Ω. (kB T )Nspin

(114)

We express the constant of proportionality as if it were a moment of inertia because it has the same units. IS Nfree 2 ≈ IL kB T ρa2 ≈ (3 × 10

−5

)

Nfree 1024 cm−3

100 K T

3g cm−3 ρ

10 nm a

2 .

(115)

19.6.1. Barnett dissipation The coupling in the Barnett eﬀect between the bulk and spin angular momenta does not act instantaneously. Rather as the instantaneous angular velocity Ω(t)

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-123

changes, there is a lag in the response of S(t). In his landmark 1979 paper,189 Purcell pointed out that this lag would help align the body coordinates so that the axis of maximum moment of inertia is parallel to the angular momentum vector. Purcell coined the name Barnett dissipation for this eﬀect, whose magnitude is greater than the mechanical dissipation mentioned above. Two types of free spins participate in the Barnett eﬀect: the electronic spins and the nuclear spins. The electronic spins are much more relevant for establishing a magnetic moment along L as discussed above in Sec. 19.5.1 because electronic magnetic moments are larger than nuclear magnetic moments by a factor of about (mp /me ). But for dissipation, this strong coupling is a liability. The nuclear spins contribute predominantly to Barnett dissipation because of their longer lag time. Mathematically, the coupled equations ˙ S(t) = −γ(S(t) − αΩ(t)), ˙ L(t) = +γ(S(t) − αΩ(t)),

(116)

describe this eﬀect. We may integrate out S(t) to obtain the following equation in inertial coordinates for the torque on L resulting from the exchange of angular momentum between the bulk and spin degrees of freedom ∞ ˙ L(t) = −αγΩ(t) + αγ 2 dτ exp[−γτ ]Ω(t − τ ). (117) 0

To calculate how fast this exchange of angular momentum aligns the axis of maximum moment of inertia with the angular momentum vector, we calculate the averaged rotational energy loss rate of the bulk degrees of freedom retaining only secular contributions. This calculation is simpler than keeping track of torques. The degree of alignment of Ω(t) with the axis of maximum moment of inertia is a function of the rotational energy in the bulk rotational degrees of freedom. The instantaneous power transferred from the bulk rotation to the spin degrees of freedom is given by ˙ ˙ = −Ω(t) · S(t) Ploss (t) = Ω(t) · L(t) = −γIs Ω(t) − γ Ω(t) − γ

∞

! dt exp[−γt ]Ω(t − t ) ,

(118)

0

and to obtain the averaged or secular contribution, we decompose Ω(t) into harmonic components of amplitude Ωa and frequency ωa , so that !# " ∞ 1 2 ωa 2 dt exp[−γt ]Ω(t − t ) = Ωa 2 Ω(t) · Ω(t) − γ . (119) 2 a γ + ωa 2 0 It follows that 1 2 ωa 2 Ploss = − IS Ωa 2 . 2 γ + ωa 2 a

(120)

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-124

We see that the stationary component (with ωa = 0) does not contribute to the loss but the precessing components dissipate. We obtain an order of magnitude for the alignment rate by assuming that (δI)/I ≈ O(1) and that except for the stationary component ωa ≈ ωmax = L/Imax , so that the excess energy is approximately Eexcess ≈ O(1)Imax ωmax 2 sin2 θ

(121)

where θ is the disalignment angle, and the rate of dissipation is approximately 2 P¯loss ≈ O(1)γIS ωmax

γ2

ωmax 2 sin2 θ, + ωmax 2

(122)

so that γalign ≈ γ

IS ωmax 2 . I3 γ 2 + ωmax 2

(123)

The action of the imperfect coupling of the spins to the bulk angular momentum and bulk rotational velocity is somewhat analogous to that of a harmonic balancer on a rotating shaft (such as a crankshaft), a common device in mechanical engineering. Here however the weakly and dissipatively coupled ﬂywheel is replaced by the internal spin degrees of freedom. 19.7. Davis–Greenstein magnetic dissipation If a grain rotates with an angular velocity along the direction of the magnetic ﬁeld, in the frame co-rotating with the grain, the magnetic ﬁeld is time independent. However the component of the angular velocity in the plane normal to the magnetic ﬁeld Ω⊥ gives rise to an oscillating magnetic ﬁeld of angular frequency Ω⊥ . In general, the magnetic susceptibility at nonzero frequency has a nonzero imaginary part χm,i , where χm (ω) = χm,r (ω) + iχm,i (ω).

(124)

The lag in the response of the magnetization to the applied driving ﬁeld gives rise to dissipation accompanied by a torque that tends to align (or antialign) the spin with the magnetic ﬁeld. This eﬀect is known as the Davis–Greenstein mechanism.40 The alignment tendency of the Davis–Greenstein mechanism may be described by the pair of equations 1 L˙ ⊥ = − L⊥ , τDG

L˙ = 0,

(125)

where the decay time τDG indicates the time scale of the Davis–Greenstein relaxation process. In deriving this simple form, we have assumed that τprec τDG , so that we may average over many precessions. We also assume that the alignment

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-125

with the axis of maximum moment of inertia likewise is very fast. If these assumptions do not apply, an averaged equation of the simple form above does not hold, and these eﬀects need to be analyzed simultaneously. A key issue in evaluating the viability of the Davis–Greenstein mechanism is the frequency of reversals of the direction of the coherent torque, termed “crossover” events in the literature. If we consider for example coherent torques from the formation of molecular hydrogen at a small number of catalysis sites on the grain surface, the orientation of this torque depends on the structure of the outermost surface atomic monolayer, and its structure is likely to change as new layers are deposited or as old layers are eroded away. Consequently, the long term coherence time of the “coherent” torque is uncertain. Under the assumption that the alignment of the body coordinates with the angular momentum is fast, it is the evolution of the coherent torque projected onto the axis of maximum moment of inertia that matters. While the grain rotates suprathermally, its direction is relatively stable against random perturbing torques, but when its rotation velocity is in the process of changing sign and of the same order as the would-be “thermal” rotation velocity, the grain is particularly susceptible to changing orientation as a result of random torques as discussed in detail in Refs. 262 and 263. In deciding whether τDG is too long for producing a signiﬁcant degree of alignment, we must consider competing eﬀects that tend to disalign the grains, which modify the above equation to become a stochastic diﬀerential equation of the form ¯ % 1 $ 1 ¯ eL + L n(t) (L − ˆ eB (ˆ eB · L)) − L − Lˆ L˙ = − 1/2 τDG τrel τrand

(126)

eL = L/|L|, and n(t) is a normalized white noise source whose where ˆ eB = B/|B|, ˆ expectation value is given by ni (t) nj (t ) = δ(t − t ).

(127)

If τrel < ∼ τrand , in the absence of a magnetic ﬁeld, L undergoes a random walk on the ¯ If τDG τrand the magnetic dissipation is strong surface of a sphere of radius L. enough to bring about almost complete alignment. However if, on the other hand, τDG τrand , the magnetic dissipation leads to a negligible amount of alignment. 19.8. Alignment along B without Davis–Greenstein dissipation Because of the magnetic precession discussed in Sec. 19.6, an average alignment with the magnetic ﬁeld direction (or more precisely with a direction very close to the magnetic ﬁeld direction) can be obtained if there is a tendency toward alignment with some other direction characterized by a time scale talign where talign tprec . The eﬀect of precession about an axis Ω (presumably along ±B) in competition with alignment about another direction n ˆ is described by the ordinary diﬀerential

May 4, 2017 9:55 b2167-ch15

M. Bucher

II-126

equation ¯ L˙ = Ω × L − γ(L − n ˆ L).

(128)

This equation has two time scales: Ω−1 and γ −1 . We are primarily interested in fast precession (i.e., Ω γ). Equation (128) is a linear equation whose solution decomposes into a stationary part and transient spiralling in to approach the stationary attractor. We solve −γ +Ω 0 L⊥1 n ˆ ⊥1 ¯ 0 L⊥2 = −γ L ˆ ⊥2 , (129) −Ω −γ n 0

0

−γ

L

n ˆ

so that L⊥1 =

γ2 ¯ γΩ ¯ Lˆ n⊥2 − 2 Lˆ n⊥1 , γ 2 + Ω2 γ + Ω2

L⊥2 =

γ2 ¯ γΩ ¯ Lˆ n⊥1 + 2 Lˆ n⊥2 , γ 2 + Ω2 γ + Ω2

(130)

¯ n . L = Lˆ ¯ In the fast precession regime of most interest to us, L = LΩ(Ω ·n ˆ )/Ω2 , with corrections suppressed by a factor (γ/Ω). The precession has no alignment tendency of its own, but if the precession period is shorter than the time scale of the alternative alignment mechanism, the precession acts to suppress alignment in the plane normal to Ω. If it turns out that radiative torques combined with the fast precession described above is the solution to the alignment problem, a number of interesting observational consequences follow. Blind data analysis might suggest that one should be able to predict polarization from the thermal dust emission intensity combined with a map of the galactic magnetic ﬁeld, perhaps deduced from rotation measures and synchrotron emission. However, if the above mechanism is correct, this procedure will not work, even with a three-dimensional model of the unpolarized dust emission and galactic magnetic ﬁeld. In the above equations, provided that the galactic magnetic ﬁeld is strong enough, its direction serves to deﬁne the direction of the grain alignment but its precise magnitude has little relevance. The above mechanism can also accommodate a large amount of small-scale power in the polarized dust signal, even if both the galactic magnetic ﬁeld and the dust density are smooth. One ¯ n(x) to might expect the ﬁeld characterizing the radiative alignment tendency L(x)ˆ include an appreciable component varying on small scales. 19.9. Radiative torques Electromagnetic radiation scattered or absorbed by the dust grain in general exerts a torque on the dust grain, as already modeled in Sec. 19.4 for isotropic illumination,

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-127

which is simpler because the torque direction is constant in the co-rotating body coordinates. Here we generalize to anisotropic illumination. Let Γ(ν, n ˆ ) expressed in body coordinates be the torque exerted on a dust grain by monochromatic electromagnetic radiation of frequency ν and unit spectral radiance propagating in the n ˆ direction. For an illumination described by the spectral radiance I(ν, n ˆ ), the torque exerted on the dust grain in inertial coordinates is given by ∞ dν dˆ nΓ(ν, R−1 (t)ˆ n)I(ν, n ˆ ). (131) Γ(t) = 0

S2

The complication here arises from the need to include both inertial and co-rotating body coordinates, related by the transformation xiner = R(t)xbody . Assuming that the time scale for radiative torques is long compared to the moment of inertia alignment time scale τI−L , we may take Ω to be aligned with L ¯ n ˆ L ), deﬁned as and deﬁne an averaged kernel Γ(ν, ˆI ; n 1 ¯ n Γ(ν, ˆI ; n ˆ L) = 2π

0

2π

dχΓ(ν, R(χ,ˆ ˆ I ), nL ) n

(132)

where R(χ,ˆ ˆ L axis. Apart from a few nL ) represents a rotation by χ about the n general observations, little can be said about the form and properties of the radiative torque kernel, which diﬀer from grain to grain. At very long wavelengths — that is, long compared to the size of the dust grain — it is a good approximation to treat the grain as a dipole (as one does for Rayleigh scattering). But in the dipole approximation there are no torques. Therefore, (a/λ) serves as an expansion parameter for which the torque approaches zero as this parameter tends toward zero. Likewise a spherical grain gives a vanishing torque at any frequency as the result of symmetry arguments. Convolving the incident radiation intensity I(ν, n ˆ I ) with this kernel, we obtain the angle-averaged torque Γtot,rad (ˆ nL ). If there were no magnetic ﬁeld, we could incorporate the torque above into the previously derived stochastic diﬀerential equation for the evolution of L [Eq. (103)] to obtain

˙L(t) = −βL(t) + β L ¯ L(t) + Γtot,rad L(t) + nran (t), (133) |L(t)| |L(t)| where a coherent torque in body coordinates has been included as well. The above equation describes a random walk L(t) in three-dimensional angular momentum space. Owing to the stochastic random torque noise term, the rotational degree of freedom loses memory of its initial state after some characteristic decay time tran , after which its state can be described by a probability distribution p(L) calculable based on the above equation. The stochastic equation above has the general form L˙ = v(L) − n(t),

(134)

May 4, 2017 9:55 b2167-ch15

II-128

M. Bucher

for which the associated Fokker–Planck equation describing the evolution of the probability density is ∂p(L) = ∇2 p(L) + ∇ · (p(L)v(L)). (135) ∂t With the magnetic ﬁeld included, we add the magnetic precession term to obtain

˙L(t) = −βL(t) + β L ¯ L(t) + Γtot,rad L(t) + γmag L × B + nran (t), (136) |L(t)| |L(t)| where the constant γmag relates the magnetic moment to the angular momentum according to µB = γmag L. When tran , trad tprec , it is possible to reduce the dimension of the stochastic diﬀerential equation to two dimensions, namely the angular velocity ω and the angle θBL between the vectors L and B, obtaining θ˙BL = F (θBL , ω) + nθ (t),

ω˙ = G(θBL , ω) + nω (t).

(137)

Unfortunately, very little can be said about the form of the functions F (θBL , ω) and G(θBL , ω), which depends on the detailed grain shapes and other properties. Draine and Weingartner53,54,222 studied three possible irregular grain shapes (shown in Fig. 25), calculating numerically the radiative torque as a function of illumination direction and frequency. Then assuming a certain anisotropic illumination, they calculated the functions F and G in Eq. (137) and studied the solutions to this equation, numerically establishing the orbits in the absence of noise. Their simulations conclude that radiative torques can provide a plausible alignment mechanism. The proposal that radiative torques resulting from anisotropic illumination align the interstellar dust grains with the magnetic ﬁeld oﬀers a promising alternative to the Davis–Greenstein magnetic dissipation mechanism whose characteristic time scale may be too long to provide the needed degree of alignment. However, its viability is hard to evaluate given our ignorance concerning the shapes of the dust grains. Without such shape information, the best case that can be put forward in favor of this scenario based on radiative torques and anisotropic illumination is to show that some candidate shapes exist that give the right orders of magnitude combined with the argument that the Davis–Greenstein scenario cannot account for the data. It is discouraging that it is not possible to formulate precise predictions for the radiative torque scenario. The origin of grain alignment seems to remain an open question. 19.10. Small dust grains and anomalous microwave emission (AME ) The thermal emission of interstellar dust grains discussed above was based on the assumption that its thermal state may be characterized by a mean temperature (around 20 K) and that ﬂuctuations about this temperature subsequent to the absorption of an UV photon constitute a negligible correction. If a dust grain is

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-129

Fig. 25. A selection of grain shapes for the study of radiative torques. In Ref. 53, Draine and Weingartner calculated the radiative torque on a dust grain as a function of frequency ν and ˆ for the three grain shape illustrated above. (Credit: B. Draine and J. incident photon direction n Weingartner)

suﬃciently large, this is a good approximation, but for small molecules an FUV (far ultraviolet) photon can raise the grain temperature momentarily by an order of magnitude or more, allowing emission in bands where there would be almost no emission for a grain at its average temperature. The bulk of the energy absorbed from the UV photon is then rapidly re-emitted through IR ﬂuorescence as the molecule rapidly cools back down to its mean temperature. Hints of such emission requiring a temperature signiﬁcantly above the mean temperature were noted through the observation of aromatic infrared emission features consistent with C– H bonds on the edge of an aromatic molecule — that is, around 3.3 and 11.3 microns.55,118 Later, analysis of the IRAS dust maps showed that a signiﬁcant fraction of the UV photons absorbed through extinction were re-emitted in bands of wavelength shorter than 60 µ, where under the assumption of a reasonable average temperature and a modiﬁed blackbody law there should be no emission.186 As a consequence, the dust models were revised to extend the size distribution to include small grains thought to be PAHs as a result of the infrared emission features described above. PAHs (polycyclic aromatic hydrocarbons) are essentially small monolayer planar sheets of graphite with hydrogen frills on their edges. A template for the small grain population can be constructed using maps of infrared emission at λ < ∼ 60µ where the thermal emission from large grains does not

May 4, 2017 9:55 b2167-ch15

II-130

M. Bucher

contribute. Since the short wavelength emission from the dust grains is proportional to the local FUV ﬂux density, these templates can in principle be improved by dividing by the local FUV ﬂux density to obtain a better map of the projected small dust grain column density, which in turn can be correlated with the AME (anomalous microwave emission) at low frequencies. It was also pointed out that such a population of very small dust grains would be expected to rotate extremely rapidly, giving rise to electric dipole emission at its rotation frequency and possibly at higher harmonics, since Pe ∼ µE ω 4 and ω 2 ≈ (kB Trot )/ma2 in the absence of suprathermal rotation.5,50,51 The possibility of magnetic dipole emission from small rotating grains has also been investigated.52 Observationally, emission at low frequencies, where diﬀuse thermal dust emission should be minuscule, was found correlated with high frequency dust maps.42,106,119 Such emission would be consistent with dipole emission from small rapidly rotating dust grains, as described above. As an alternative explanation for such a correlation, it was proposed that regions of free–free emission could be correlated with regions of thermal dust emission. However, this low frequency excess was found not to be correlated with Hα line emission, which serves as a tracer of free–free emission. One possibility to weaken such a correlation would be to postulate an extremely high temperature, so that Γfree–free (T )/ΓHα is large. However, it was found that such a high temperature would require an implausible energy injection rate to prevent this phase from cooling. The spectrum of anomalous microwave emission, presumably from spinning dust, has been measured by taking spectra in targeted regions having a large contribution from AME and subtracting the expected synchrotron, free–free, thermal dust, and primordial CMB signals.88,220 Figure 26 shows the spectrum thus obtained. The observed fall-oﬀ at low ν disfavors the hypothesis of free–free emission correlated with thermal dust emission as the origin of the anomalous emission, because in that case, the spectrum of the residual should rise as ν is lowered. Most theoretical studies have assumed that the small grains do not become aligned with the galactic magnetic ﬁeld, but alignment mechanisms for this grain population have not yet been investigated in detail. Observational limits on the anomalous microwave emission constrain the degree of polarization to less than a few per cent.193 20. Compact Sources The above processes pertain primarily to diﬀuse galactic emission, which contaminates the primordial CMB signal most severely at low multipole number . On smaller scales, however, compact sources become more of a concern and eventually become the dominant contribution at all frequencies. We use the term “compact” to denote both unresolved point sources and partially resolved, localized sources. The angular power spectrum of point sources distributed according to a Poissonian distribution would scale with multipole number as 0 . If ignoring the structure

May 10, 2017 11:11

II-131

100

Du st

1000

(V ibr ati on al)

Haslam et al Reich & Reich Andersson et al COSMOSOMAS WMAP DIRBE

Dust (rotaional)

10

S (Jy)

104

105

Physics of the cosmic microwave background anisotropy

1

Free–Free

1

10

100

1000

(GHz) Fig. 26. Consistency of anomalous microwave emission with spinning dust. The multi-frequency observations shown here were taken for a molecular cloud believed to have strong anomalous dust emission. We observe that a linear combination of thermal dust and free–free emission cannot explain the observed shape. However, if an anomalous dust component with the spectral shape predicted by Draine and Lazarian for electric dipole emission from spinning dust is included (dotted curve), a reasonable ﬁt may be obtained. For more details, see Ref. 220. (Credit: COSMOMAS Collaboration)

of the acoustic oscillations, one approximates the primordial CMB spectrum as having the shape −2 , one would conclude that compact sources rapidly take over at high . Because of Silk damping combined with the smoothing from the ﬁnite width of the last scattering surface, the fall-oﬀ of the CMB at large is even more rapid. The fact that point sources are clustered rather than distributed in Poissonian manner causes the point source power spectrum to be clustered, leading to an excess of power at small compared to an 0 power law.

20.1. Radio galaxies Section 16 described the physical process of synchrotron emission, emphasizing its contribution to the diﬀuse emission from our own galaxy. However the same process is also at work in galaxies other than our own, and in particular in the rare but highly luminous radio galaxies at high redshift, extreme processes fueled by a central black hole give rise to intense emission in many bands, including synchrotron emission in the radio and microwave bands. For our purposes, the most salient and relevant property of these sources is their inverted spectrum. For a detailed survey of radio sources, see Ref. 212. Since each source diﬀers slightly in its spectral

May 4, 2017 9:55 b2167-ch15

II-132

M. Bucher

properties, the strategy for dealing with these sources is to mask out the brightest sources and then model the residual arising from the remaining unmasked sources. 20.2. Infrared galaxies In Sec. 18, we discussed at length the contamination from thermal emission from the interstellar dust in our galaxy. Thermal dust emission, however, is not restricted to our galaxy. All galaxies contribute some thermal dust emission. However, the galaxies at high redshift in which star formation is occurring at a rapid rate contribute predominantly to the extragalactic thermal dust emission in the microwave bands of interest to us. In fact, most of the energy of the radiation from these galaxies in which intense star formation is taking place is emitted in the infrared, because the light and UV emitted is absorbed by dust, which becomes heated and in turn re-emits in the infrared. Some of these infrared galaxies are resolved by current observation. Others remain unresolved by present day observations and these are attributed to the diﬀuse infrared background. These infrared galaxies do not all emit with the same frequency spectrum, in part because they are spread over a wide range of redshifts. For the early discovery papers on the CIB (cosmic infrared background) from COBE see Refs. 187, 85 and 63. For more recent observations from Planck, see Ref. 155. 21. Other Eﬀects 21.1. Patchy reionization In Sec. 7.5, we idealized the reionized hot electron gas extending from intermediate redshifts to today as being isotropically distributed so that the free electron density depends only on redshift. Such a reionization scenario is implausible but provides a good approximation for the eﬀect of reionization on the CMB power spectrum at low . In a more realistic reionization scenario, the universe becomes reionized in an inhomogeneous manner. As the ﬁrst stars and quasars form, their UV radiation creates around them spheres of almost completely ionized hydrogen (i.e., HII) that grow and eventually coalesce. These spheres have sharp edges that act to transform large-angle power from the primordial CMB into small-scale power. For some estimates of the impact due to patchy reionization of the angular CMB spectrum, see Ref. 253. 21.2. Molecular lines The sources of diﬀuse galactic emission discussed so far all have a broad continuum emission spectrum. But there also exist molecular lines, mostly from rotational transitions of molecules, that emit in the microwave bands of interest here. The brightest of these are the CO lines,38 of which the lowest is situated at 115.3 GHz.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-133

Since CO is a linear molecule, there are also transitions at integer multiples of this frequency. Removing CO emission from the Planck data has proven to be a challenge.163 For more sensitive future experiments, it is likely that other lines will also present concerns. 21.3. Zodiacal emission Besides the planets and asteroids, our solar system is ﬁlled with large dust grains, known as zodiacal dust, which on the average are much larger than the interstellar dust grains described above. These grains manifest themselves by a variety of means. In the visible they scatter light from the sun, and this background must be subtracted when analyzing diﬀuse emission in the optical and near infrared bands. The zodiacal dust is concentrated around the ecliptic plane. It is a good but not perfect approximation to model the distribution of zodiacal dust assuming azimuthal symmetry. The temperature of the zodiacal dust is much above that of the interstellar dust and roughly proportional to the inverse square of the distance from the sun. Subtracting zodiacal emission is challenging because its contribution to a given point on the celestial sphere depends on the time of year when the observation is made, because observations at diﬀerent times of the year look through diﬀerent columns of dust, even when azimuthal symmetry is assumed. The ideal observing program for characterizing the zodiacal dust would cover the whole sky many times over the year in order to accumulate as many lines of sight to the same point on the celestial sphere as possible and to carry out a sort of tomography. For a discussion of zodiacal emission at long wavelengths as measured by Planck, see Ref. 164. For previous work on zodiacal emission in the far infrared, see Ref. 64. 22. Extracting the Primordial CMB Anisotropies Most of the contaminants described above, with the exception of gravitational lensing and the kinetic SZ eﬀect, have a frequency dependence that diﬀers from that of the primary CMB anisotropies. This property can be exploited to remove these nonprimordial contaminants and lies at the heart of all component separation techniques. If there are N distinct components, each of which is characterized by a ﬁxed frequency spectrum, in the absence of measurement noise, a perfect separation of these components is possible using sky maps in N frequency bands. With more bands, the separation is overdetermined allowing validation of the model. The component separation strategy sketched above, which may be described as a linear method, is based on a linear model, but this is not the only way to clean maps. More sophisticated nonlinear models may be formulated, often based on or inspired by Bayesian statistics that exploit other properties such as the positivity of certain components, spatial variation of spectral indices, and so on. Some methods are characterized as being “blind,” meaning that the number of components is indicated by data itself, while other methods are more physically based.

May 4, 2017 9:55 b2167-ch15

II-134

M. Bucher

Because of limitations of space, we do not go into the details or practicalities of component separation, instead referring the reader to a few excellent reviews for more details. For an early discussion of the component separation problem and various approaches, see Refs. 24 and 44. Reference 116 includes a comparison of methods used for the Planck analysis. Reference 56 describes foreground removal challenges for future CMB B mode polarization experiments and includes extensive references to the literature. 23. Concluding Remarks Prior to the discovery of the CMB anisotropy in 1992 by the COBE team, cosmology was a ﬁeld with a questionable reputation. Many from other ﬁelds described cosmology as a ﬁeld where theorists were free to speculate with almost no observational or experimental constraints to contradict them. Much was made of the supposed distinction between an “observation” and an “experiment.”w However, between 1992 and today cosmology has transformed into a precision science where it has become increasingly diﬃcult to propose new models without contradicting present observation. Ironically, much science having only a tenuous connection to cosmology is now being explained to the public as serving to probe the big bang. In this review, we have tried to sketch the most important aspects of the physics of the CMB in a self-contained way. Other reviews with a diﬀerent emphasis include Refs. 94, 255 and 256. We have seen how the CMB provides a snapshot of the state of the universe at around z ≈ 1100, when the primordial perturbations were still very nearly linear, and how the precise angular pattern on the sky is also sensitive to a number of other cosmological parameters, which can be determined in many cases to the percent level or better. We have also explored tests of non-Gaussianity. At present no non-Gaussianity of a primordial origin has been observed, ruling out many nonminimal models of inﬂation, in particular multi-ﬁeld models developed to explain hints of non-Gaussianity in the WMAP data, which subsequently were excluded by Planck. We also saw how better measurements of the CMB absolute spectrum could reveal deviations from a perfect blackbody spectrum, some of which are to be expected and others of which would constitute signs of novel physics, such as energy injection at high-redshift. The absolute spectrum remains a promising frontier of CMB science. The other promising area of CMB research is the characterization of the polarized sky, in particular to search for primordial B modes from inﬂation, which constitute the most unique prediction of inﬂation. Maps of B modes on smaller angular scales would also allow a characterization of the gravitational lensing ﬁeld at much w It

is not clear whether proton–antiproton colliders would be classiﬁed as constituting an “experiment” under this criterion, because the experimenter has virtually no control over parameters of the hard parton–parton collisions of interest due to the broad spread in x of the parton distribution functions.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-135

greater precision than is possible from observations of the temperature anisotropy. Such data would provide an important cross check of weak lensing surveys using the ellipticities of galaxies and provide a unique probe of absolute neutrino masses, perhaps allowing one to distinguish the normal from the inverted mass hierarchy. Since this volume is dedicated to Albert Einstein and the hundredth anniversary of his celebrated paper presenting the general theory of relativity, we close with a few remarks about Einstein, and in particular about his relation to the subject of this chapter. Einstein is remembered by today’s generation of physicists mostly for his role in the development of the special and general theories of relativity. This part of Einstein’s work sometimes overshadows Einstein’s numerous achievements in other areas and the extreme breadth of his scientiﬁc interests. Einstein is often viewed today as the paragon of the abstruse, highly mathematical theorist. Many physicists today would be surprised to learn that Einstein patented a design for an ammonia–butane refrigerator.59 Others might be surprised to learn that Einstein is the author of one experimental paper for the work after which the Einstein–de Haas eﬀect is named. This eﬀect may be thought of as the converse of the Barnett eﬀect, which we saw in Sec. 19.5.1 plays a crucial role in the alignment of interstellar dust grains. Moreover, the stochastic diﬀerential equations in that section bear a close relation to Einstein’s work on Brownian motion. General relativity was once regarded as an abstruse, highly mathematical subﬁeld of physics. Within astrophysics, relativity was ﬁrst introduced through the emergence of a subﬁeld known as “relativistic astrophysics.” But nowadays general relativity pervades almost all areas of astrophysics and has become one of the elements of the underlying fundamental physics that every astrophysicist must know. One can no longer neatly separate “relativistic astrophysics” from the rest of astrophysics. This chapter told the story of the big bang cosmology from a point of view emphasizing the observations of the relic blackbody radiation. Einstein’s theory of general relativity culminating in his celebrated 1915 formulation of the gravitational ﬁeld equationsx,y enters as a key ingredient to this story in two ways. First there are the unperturbed spacetime solutions now known as the FLRW solutions. This family of spatially homogeneous and isotropic solutions to the Einstein equations was ﬁrst found by Alexandre Friedmann in 1922 and independently by Georges Lemaˆıtre in 1927. At that time, the solutions were surprising because of their prediction of a non-eternal dynamical universe whose natural state was either expansion or contraction, an idea that was before its time. It was not until 1929 that Hubble discovered that the recessional velocities of distant galaxies except for some noise vary approximately linearly with distance (although in its ﬁrst version x A.

Einstein, “Die Feldgleichungen der Gravitation,” Sitzungsberichte der K¨ onigliche Preussiche Akademie der Wissenschaften, p. 844 (25 November, 1915). y For an authoritative account from a scientiﬁc perspective of the path involving over a dozen papers and several reversals that ﬁnally led Einstein to the gravitational ﬁeld equations as we know them today (i.e., Gµν = Rµν − 12 gµν R = (8πGN )Tµν ), see Abraham Pais, Subtle is the Lord (New York, Oxford University Press, 1982).

May 4, 2017 9:55 b2167-ch15

II-136

M. Bucher

the data was not statistically compelling). Einstein’s general theory was deduced by drawing mainly on considerations of mathematical simplicity and elegance and on the equivalence principle, but remarkably applied to the expanding universe through the FLRW solution and its linearized perturbations. General relativity remains the basis of cosmology despite the great improvement in the data that has taken place between the 1920s and now. Today there is much talk about “modiﬁed gravity,” largely motivated by the discovery of the so-called “dark energy,” which no one yet really understands. It can be argued whether the dark energy belongs to the left-hand side of the Einstein ﬁeld equation, in which case it would constitute a revision to Einstein’s theory, or on the right-hand side, in which case it would simply constitute a new form of exotic stress–energy, leaving the gravitational sector unchanged. But so far Einstein’s theory of relativity as a classical ﬁeld theory has survived without change. To be sure, we believe that gravitational theory as formulated by Einstein is an incomplete story. Where quantum eﬀects become relevant, we know that there must be a more complete theory of which classical relativity is but a limiting case. But we do not yet know what that larger theory is. In the classical regime, Einstein’s theory has survived many stringent tests, and perhaps more importantly has become an indispensable foundation for our current understanding of the universe and its origins.

Acknowledgments MB would like to thank Ken Ganga, Kavilan Moodley, Heather Prince, Doris Rojas, and Andrea Tartari for many invaluable comments, corrections, and suggestions, and Jean-Luc Robert for making several of the ﬁgures. MB thanks the University of KwaZulu-Natal for its hospitality where a large part of this review was written. Note added in Proof: Since the writing of this chapter, the Planck 2015 Cosmological Results have appeared in the form of a series of papers of which a complete listing can be found in the overview article.265 Apart from an improved analysis of the temperature anisotropies, this release contains for the ﬁrst time a detailed analysis of the Planck TE and EE polarization results (except at the lowest multipole numbers, which will be analyzed as part of the 2016 ﬁnal release). The joint BICEP/Keck Array–Planck analysis of the B mode signal in the BICEP2 patch covering about a square degree of the sky resulted in a new upper limit on the tensorto-scalar ratio of (r < 0.12) at 95% conﬁdence.266 This joint analysis exploited the Planck high-frequency maps dominated by polarized dust emission to clean the dust contribution from the BICEP2/Keck Array maps at 150 GHz. This new limit marks the ﬁrst time that a limit on r from B modes alone is competitive with limits obtained from the temperature anisotropy. It should be stressed that the limits from B mode polarization alone are much more robust against model assumptions than limits obtained using the temperature anisotropy for the reasons explained in the main text.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-137

References 1. K. N. Abazajian et al., Neutrino physics from the cosmic microwave background and large scale structure, Astropart. Phys. 63 (2015) 66, arXiv:astro-ph/1309.5383. 2. K. N. Abazajian et al., Inﬂation physics from the cosmic microwave background and large scale structure, Astropart. Phys. 63 (2015) 55, arXiv:astro-ph/1309.5381. 3. L. F. Abbott and M. B. Wise, Constraints on generalized inﬂationary cosmologies, Nucl. Phys. B 244 (1984) 541. 4. A. Albrecht, D. Coulson, P. Ferreira and J. Magueijo, Causality, randomness, and the microwave background, Phys. Rev. Lett. 76 (1996) 1413. 5. Y. Ali-Ha¨ımoud, C. M. Hirata and C. Dickinson, A reﬁned model for spinning dust radiation, Mon. Not. R. Astron. Soc. 395 (2009) 1055. 6. R. A. Alpher, H. Bethe and G. Gamov, The origin of chemical elements, Phys. Rev. 73 (1948) 803. 7. V. Acquaviva, N. Bartolo, S. Matarrese and A. Riotto, Gauge-invariant second-order perturbations and non-Gaussianity from inﬂation, Nucl. Phys. B 667 (2002) 119, arXiv:astro-ph/0209156. 8. K. Arnold et al., The POLARBEAR CMB polarization experiment, Proc. SPIE 7741 (2010) 77411E. 9. D. Babich, P. Creminelli and M. Zaldarriaga, The shape of non-Gaussianities, J. Cosmol. Astropart. Phys. 08 (2004) 009, arXiv:astro-ph/0405356. 10. J. M. Bardeen, Gauge-invariant cosmological perturbations, Phys. Rev. D 22 (1980) 1882. 11. J. M. Bardeen, P. J. Steinhardt and M. S. Turner, Spontaneous creation of almost scale-free density perturbations in an inﬂationary universe, Phys. Rev. D 28 (1983) 679. 12. R. Barkana and A. Loeb, In the beginning: The ﬁrst sources of light and the reionization of the universe, Phys. Rep. 349 (2001) 125, arXiv:astro-ph/0010468. 13. E. Battistellie et al., QUBIC: The QU bolometric interferometer for cosmology, Astropart. Phys. 34 (2011) 705, arXiv:astro-ph/1010.0645. 14. R. H. Becker et al., Evidence for reionization at z ∼ 6: Detection of a Gunn–Peterson trough in a z = 6.28 quasar, Astron. J. 122 (2001) 2850, arXiv:astro-ph/0108097. 15. P. de Bernardis et al., A ﬂat universe from high-resolution maps of the cosmic microwave background radiation, Nature 404 (2000) 955, arXiv:astro-ph/0004404. 16. E. Bertschinger, COSMICS: Cosmological initial conditions and microwave anisotropy codes, arXiv:astro-ph/9506070. 17. BICEP2 Collab. (P. A. R. Ade et al.), Detection of B-mode polarization at degree angular scales by BICEP2, Phys. Rev. Lett. 112 (2014) 241101. 18. BICEP2 Collab. (P. A. R. Ade et al.), BICEP2. II. Experiment and three-year data set, Astrophys. J. 792 (2014) 62, arXiv:astro-ph/1403.4302. 19. M. Birkinshaw, The Sunyaev–Zel’dovich eﬀect, Phys. Rep. 310 (1999) 97, arXiv:astro-ph/9808050. 20. R. Blandford and D. Eichler, Particle acceleration at astrophysical shocks: A theory of cosmic ray origin, Phys. Rep. 154 (1987) 1. 21. N. W. Boggess et al., The COBE mission: Its design and performance two years after launch, Astrophys. J. 397 (1992) 420. 22. J. R. Bond and G. Efstathiou, Cosmic background radiation anisotropies in universes dominated by nonbaryonic dark matterm, Astrophys. J. 285 (1984) L45. 23. M. Born and E. Wolf, Principles of Optics (Pergamon, New York, 1970). 24. F. R. Bouchet and R. Gispert, Foregrounds and CMB experiments: I. Semi-analytical estimates of contamination, New Astron. 4 (1999) 443, arXiv:astro-ph/9903176.

May 4, 2017 9:55 b2167-ch15

II-138

M. Bucher

25. S. Boughn and R. Crittenden, A correlation between the cosmic microwave background and large-scale structure in the universe, Nature 427 (2004) 45, arXiv:astroph/0305001. 26. M. Bucher, A. S. Goldhaber and N. Turok, An open universe from inﬂation, Phys. Rev. D 52 (1995) 3314, arXiv:hep-ph/9411206. 27. M. Bucher and N. Turok, Open inﬂation with an arbitrary false vacuum mass, Phys. Rev. D 52 (1995) 5538. 28. M. Bucher, K. Moodley and N. Turok, General primordial cosmic perturbation, Phys. Rev. D 62 (2000) 083508, arXiv:astro-ph/9904231. 29. J. E. Carlstrom, G. P. Holder and E. D. Reese, Cosmology with the Sunyaev– Zel’dovich eﬀect, Ann. Rev. Astron. Astrophys. 40 (2002) 643, arXiv:astroph/0208192. 30. J. Chluba and R. M. Thomas, Towards a complete treatment of the cosmological recombination problem, Mon. Not. R. Astron. Soc. 412 (2011) 748, arXiv:astroph/1010.3631. 31. J. Chluba and R. A. Sunyaev, Pre-recombinational energy release and narrow features in the CMB spectrum, Astron. Astrophys. 501 (2009) 29, arXiv:astroph/0803.3584. 32. C. R. Contaldi, J. Magueijo and L. Smolin, Anomalous CMB polarization and gravitational chirality, Phys. Rev. Lett. 101 (2008) 141101, arXiv: astro-ph/0806.3082. 33. COrE Collab. (F. Bouchet et al.), COrE (Cosmic Origins Explorer): A White Paper, arXiv:1102.2181. 34. B. P. Crill et al., SPIDER: A balloon-borne large-scale CMB polarimeter, Proc. SPIE 7010, Space Telescopes and Instrumentation 2008 : Optical, Infrared, and Millimeter, 70102P (July 12, 2008). 35. R. Crittenden, R. L. Davis and P. J. Steinhardt, Polarization of the microwave background due to primordial gravitational waves, Astrophys. J. 417 (1993) L13, arXiv:astro-ph/9306027. 36. R. G. Crittenden and N. Turok, Doppler peaks from cosmic texture, Phys. Rev. Lett. 75 (1995) 2642. 37. R. G. Crittenden and N. Turok, Looking for a cosmological constant with the ReesSciama eﬀect, Phys. Rev. Lett. 76 (1996) 575, arXiv:astro-ph/9510072. 38. T. M. Dame et al., The milky way in molecular clouds: A new complete CO survey, Astrophys. J. 547 (2001) 792. 39. S. Das et al., Detection of the power spectrum of cosmic microwave background lensing by the Atacama Cosmology Telescope, Phys. Rev. Lett. 107 (2011) 021301, arXiv:astro-ph/1103.2124. 40. L. Davis and J. Greenstein, The polarization of starlight by aligned dust grains, Astrophys. J. 114 (1951) 206. 41. P. Day et al., A broadband superconducting detector suitable for use in large arrays, Nature 425 (2003) 817. 42. A. de Oliveira-Costa et al., Galactic microwave emission at degree angular scales, Astrophys. J. 482 (1997) L17, arXiv:astro-ph/9702172. 43. P. de Bernardis, M. Bucher, C. Burigana and L. Piccirillo, B-Pol: Detecting primordial gravitational waves generated during inﬂation, Exp. Astron. 23 (2009) 5, arXiv:0808.1881. 44. J. Delabrouille, J.-F. Cardoso and G. Patanchon, Multi-detector multi-component spectral matching and applications for CMB data analysis, Mon. Not. R. Astron. Soc. 346 (2003) 1089, arXiv:astro-ph/0211504.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-139

45. R. H. Dicke, The measurement of thermal radiation at microwave frequencies, Rev. Sci. Instrum. 17 (1946) 238. 46. R. H. Dicke, P. J. E. Peebles, P. G. Roll and D. T. Wilkinson, Cosmic black-body radiation, Astrophys. J. 142 (1965) 414. 47. S. Doyle, P. Mauskopf, J. Naylon, A. Porch and C. Duncombe, Lumped element kinetic inductance detectors, J. Low Temp. Phys. 151 (2008) 530. 48. B. T. Draine, Interstellar dust grains, Ann. Rev. Astron. Astrophys. 41 (2003) 241, arXiv:astro-ph/0304489. 49. B. T. Draine, Astrophysics of dust in cold clouds, in The Cold Universe, Saas-Fee Advanced Course 32, ed. D. Pfenniger (Springer-Verlag, Berlin, 2004), pp. 213–303, arXiv:astro-ph/0304488. 50. B. T. Draine and A. Lazarian, Electric dipole radiation from spinning dust grains, Astrophys. J. 508 (1998) 157, arXiv:astro-ph/9802239. 51. B. T. Draine and A. Lazarian, Diﬀuse galactic emission from spinning dust grains, Astrophys. J. 494 (1998) L19. 52. B. T. Draine and A. Lazarian, Magnetic dipole microwave emission from dust grains, Astrophys. J. 512 (1999) 740, arXiv:astro-ph/9807009. 53. B. T. Draine and J. C. Weingartner, Radiative torques on interstellar grains. I. Superthermal spin-up, Astrophys. J. 470 (1996) 551. 54. B. T. Draine, J. C. Weingartner and C. Joseph, Radiative torques on interstellar grains. II. Grain alignment, Astrophys. J. 480 (1997) 633. 55. W. W. Duley and D. A. Williams, The infrared spectrum of interstellar dust: Surface functional groups on carbon, Mon. Not. R. Astron. Soc. 196 (1981) 269. 56. J. Dunkley et al., CMBPol mission concept study: Prospects for polarized foreground removal, AIP Conf. Proc. 1141 (2009) 222, arXiv:astro-ph/0811.3915. 57. J. Dunkley, M. Bucher, P. G. Ferreira, K. Moodley and C. Skordis, Fast and reliable MCMC for cosmological parameter estimation, Mon. Not. R. Astron. Soc. 356 (2005) 925, arXiv:astro-ph/0405462. 58. J. Dunkley et al., The Atacama Cosmology Telescope: Cosmological parameters from the 2008 power spectrum, Astrophys. J. 739 (2011) 52, arXiv:astro-ph/1009.0866. 59. A. Einstein and L. Szil´ ard, Refrigeration (Appl: 16 December 1927; Priority: Germany, 16 December 1926) U. S. Patent 1,781,541, 11 November 1930. 60. M. Eisenberg and R. Guy, A proof of the hairy ball theorem, Am. Math. Mon. 86 (1979) 571. 61. EPIC Collab. (J. Bock et al.), Study of the experimental probe of inﬂationary cosmology EPIC-Intemediate mission for NASA’s einstein inﬂation probe, arXiv:0906.1188. 62. A. Fert´e and J. Grain, Detecting chiral gravity with the pure pseudospectrum reconstruction of the cosmic microwave background polarized anisotropies, Phys. Rev. D 89 (2014) 103516, arXiv:astro-ph/1404.6660. 63. D. J. Fixsen et al., The spectrum of the extragalactic far-infrared background from the COBE FIRAS observations, Astrophys. J. 508 (1998) 123. 64. D. J. Fixsen and E. Dwek, The zodiacal emission spectrum as determined by COBE and its implications, Astrophys. J. 578 (2002) 1009. 65. D. J. Fixsen et al., The cosmic microwave background spectrum from the full COBE FIRAS data set, Astrophys. J. 473 (1996) 576. 66. R. Flauger, J. C. Hill and D. N. Spergel, Toward an understanding of foreground emission in the BICEP2 region, arXiv:astro-ph/1405.7351. 67. U. Fuskeland, I. K. Wehus, H. K. Eriksen and S. K. Naess, Spatial variations in the spectral index of polarized synchrotron emission in the 9 yr WMAP sky maps, Astrophys. J. 790 (2014) 104, arXiv:astro-ph/1404.5323.

May 4, 2017 9:55 b2167-ch15

II-140

M. Bucher

68. M. Giard, G. Serra, E. Caux, F. Pajot and J. M. Lamarre, First detection of the aromatic 3.3-micron feature in the diﬀuse emission of the Galactic disk, Astron. Astrophys. 201 (1988) L1. 69. V. L. Ginzburg and S. I. Syrovatsk, Developments in the theory of synchrotron radiation and its reabsorption, Ann. Rev. Astron. Astrophys. 7 (1969) 375. 70. A. S. Goldhaber and M. Nieto, Terrestrial and extraterrestrial limits on the photon mass, Rev. Mod. Phys. 43 (1971) 277. 71. A. S. Goldhaber and M. Nieto, Photon and graviton mass limits, Rev. Mod. Phys. 82 (2010) 939, arXiv:hep-ph/0809.1003. 72. J. R. Gott III, Creation of open universes from de Sitter space, Nature 295 (1982) 304. 73. J. Richard Gott III and T. S. Statler, Constraints on the formation of bubble universes, Phys. Lett. B 136 (1984) 157. 74. J. E. Gunn and B. A. Peterson, On the density of neutral hydrogen in intergalactic space, Astrophys. J. 142 (1965) 1633. 75. A. H. Guth and S. Y. Pi, Fluctuations in the new inﬂationary universe, Phys. Rev. Lett. 49 (1982) 1110. 76. J. S. Hall, Observations of the polarized light from stars, Science 109 (1949) 166. 77. S. Hanany et al., MAXIMA-1: A measurement of the cosmic microwave background anisotropy on angular scales of 10’-5◦ , Astrophys. J. 545 (2000) L5, arXiv:astroph/0005123. 78. J. L. Han et al., Pulsar rotation measures and the large-scale structure of the galactic magnetic ﬁeld, Astrophys. J. 642 (2006) 868, arXiv:astro-ph/0601357. 79. N. Hand et al., Evidence of galaxy cluster motions with the kinematic Sunyaev– Zel’dovich eﬀect, Phys. Rev. Lett. 109 (2012) 041101, arXiv:astro-ph/1203.4219. 80. C. G. T. Haslam et al., A 408 MHz all-sky continuum survey. I — Observations at southern declinations and for the north polar region, Astron. Astrophys. 100 (1981) 209. 81. M. Hazumi et al., LiteBIRD: A small satellite for the study of B-mode polarization and inﬂation from cosmic background radiation detection, Proc. SPIE 8442, Space Telescopes and Instrumentation 2012 : Optical, Infrared, and Millimeter Wave, 844219 (August 22, 2012). 82. J. Kaufman, B. Keating and B. Johnson, Precision tests of parity violation over cosmological distances, arXiv:astro-ph/1409.8242. 83. D. Lyth, Large-scale energy-density perturbations and inﬂation, Phys. Rev. D 31 (1985) 1792. 84. A. R. Liddle and D. H. Lyth, Cosmological Inflation and Large-Scale Structure (Cambridge University Press, Cambridge, 2009). 85. M. G. Hauser et al., The COBE diﬀuse infrared background experiment search for the cosmic infrared background. I. Limits and detections, Astrophys. J. 508 (1998) 25. 86. S. W. Hawking, The development of irregularities in a single bubble inﬂationary universe, Phys. Lett. B 115 (1982) 295. 87. C. M. Hirata and E. R. Switzer, Primordial helium recombination. II. Two-photon processes, Phys. Rev. D 77 (2008) 083007, arXiv:astro-ph/0702144. 88. S. R. Hildebrandt, R. Rebolo, J. A. Rubi˜ no-Mart´ın, R. A. Watson, C. M. Guti´errez, R. J. Hoyland and E. S. Battistelli, COSMOSOMAS observations of the cosmic microwave background and Galactic foregrounds at 11 GHz: Evidence for anomalous microwave emission at high Galactic latitude, Mon. Not. R. Astron. Soc. 382 (2007) 594, arXiv:0706.1873.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-141

89. W. A. Hiltner, Polarization of light from distant stars by interstellar medium, Science 109 (1949) 165. 90. W. A. Hiltner, On the presence of polarization in the continuous radiation of stars. II., Astrophys. J. 109 (1949) 471. 91. W. Hu, D. Scott and J. Silk, Power spectrum constraints from spectral distortions in the cosmic microwave background, Astrophys. J. 430 (1994) L5, arXiv:astroph/9402045. 92. W. Hu and M. White, Acoustic signatures in the cosmic microwave background, Astrophys. J. 471 (1996) 30, arXiv:astro-ph/9602019. 93. W. Hu and T. Okamoto, Mass reconstruction with CMB polarization, Astrophys. J. 574 (2002) 566, arXiv:astro-ph/0111606. 94. W. Hu and S. Dodelson, Cosmic microwave background anisotropies, Ann. Rev. Astron. Astrophys. 40 (2002) 171–216, arXiv:astro-ph/0110414. 95. W. Hu and G. P. Holder, Model-independent reionization observables in the CMB, Phys. Rev. D 68 (2003) 023001, arXiv:astro-ph/0303400. 96. W. Hu and M. White, A CMB polarization primer, New Astron. 2 (1997) 323, arXiv:astro-ph/9706147. 97. A. F. Illarionov and R. A. Siuniaev, Comptonization, the background-radiation spectrum, and the thermal history of the universe, Sov. Astron. 18 (1975) 691. 98. N. Jarosik et al., First year Wilkinson microwave anisotropy probe (WMAP) observations: On-orbit radiometer characterization, Astrophys. J. 148 (2003) 29, arXiv:astro-ph/0302224. 99. G. Jungman, M. Kamionkowski, A. Kosowsky and D. N. Spergel, Cosmological parameter determination with microwave background maps, Phys. Rev. D 54 (1996) 1332, arXiv:astro-ph/9512139. 100. M. Kamionkowski, A. Kosowsky and A. Stebbins, Statistics of cosmic microwave background polarization, Phys. Rev. D 55 (1997) 7368, arXiv:astro-ph/9611125. 101. M. Kaplinghat, M. Chu, Z. Haiman, G. Holder, L. Knox and C. Skordis, Probing the reionization history of the universe using the cosmic microwave background polarization, Astrophys. J. 583 (2003) 24, arXiv:astro-ph/0207591. 102. R. Keisler et al., A measurement of the damping tail of the cosmic microwave background power spectrum with the South Pole Telescope, Astrophys. J. 743 (2011) 28, arXiv:astro-ph/1105.3182. 103. H. Kodama and M. Sasaki, Cosmological perturbation theory, Prog. Theor. Phys. Suppl. 78 (1984) 1. 104. A. Kogut et al., Dipole anisotropy in the COBE DMR ﬁrst-year sky maps, Astrophys. J. 419 (1993) 1, arXiv:astro-ph/9312056. 105. A. Kogut et al., The primordial inﬂation explorer (PIXIE): A nulling polarimeter for cosmic microwave background observations, J. Cosmol. Astropart. Phys. 1107 (2011) 025, arXiv:astro-ph/1105.2044. 106. A. Kogut, A. J. Banday, C. L. Bennett, K. M. Gorski, G. Hinshaw and W. T. Reach, High-latitude galactic emission in the COBE diﬀerential microwave radiometer 2 year sky maps, Astrophys. J. 460 (1996) 1. 107. E. Komatsu and D. Spergel, Acoustic signatures of the primary microwave background bispectrum, Phys. Rev. D 63 (2001) 063002, arXiv:astro-ph/000503. 108. A. Kosowsky, Cosmic microwave background polarization, Ann. Phys. 246 (1996) 49, arXiv:astro-ph/9501045. 109. A. Kosowsky, The Atacama Cosmology Telescope, New Astron. Rev. 47 (2003) 939, arXiv:astro-ph/0402234. 110. J. Kraus, Radio Astronomy (McGraw-Hill, New York, 1966).

May 4, 2017 9:55 b2167-ch15

II-142

M. Bucher

111. J. Kovac et al., Detection of polarization in the cosmic microwave background using DASI, Nature 420 (2002) 772, arXiv:astro-ph/0209478. 112. L. Kuzmin, Ultimate cold-electron bolometer with strong electrothermal feedback, Proc. SPIE Conf., Millimeter and Submillimeter Detectors for Astronomy II, Vol. 5498 (2004), pp. 349–361. 113. J. M. Lamarre, Photon noise in photometric instruments at far infrared and submillimeter wavelengths, Appl. Opt. 25 (1986) 870. 114. K. Land and J. Magueijo, Examination of evidence for a preferred axis in the cosmic radiation anisotropy, Phys. Rev. Lett. 95 (2005) 071301, arXiv:astro-ph/0502237. 115. L. D. Landau and E. M. Lifshitz, Electrodynamics of Continuous Media: Course of Theoretical Physics, Vol. 8 (First English edition) (Pergamon Press, New York, 1960). 116. S. M. Leach et al., Component separation methods for the Planck mission, Astron. Astrophys. 491 (2008) 597, arXiv:astro-ph/0805.0269. 117. A. T. Lee, P. L. Richards, S. W. Nam, B. Cabrera and K. D. Irwin, A superconducting bolometer with strong electrothermal feedback, Appl. Phys. Lett. 69 (1996) 1801. 118. A. Leger and J. L. Puget, Identiﬁcation of the “unidentiﬁed” IR emission features of interstellar dust? Astron. Astrophys. 137 (1984) L5. 119. E. M. Leitch, A. C. S. Readhead, T. J. Pearson and S. T. Myers, An anomalous component of galactic emission, Astrophys. J. 486 (1997) L23, arXiv:astro-ph/9705241. 120. J. Lesgourgues and S. Pastor, Massive neutrinos and cosmology, Phys. Rep. 429 (2006) 307, arXiv:astro-ph/0603494. 121. A. Lewis, A. Challinor and A. Lasenby, Eﬃcient computation of cosmic microwave background anisotropies in closed Friedmann–Robertson–Walker models, Astrophys. J. 538 (2000) 473. 122. A. Lewis and S. Bridle, Cosmological parameters from CMB and other data: A Monte–Carlo approach, Phys. Rev. D 66 (2002) 103511, arXiv:astro-ph/0205436. 123. A. Lewis and A. Challinor, Weak gravitational lensing of the CMB, Phys. Rep. 429 (2006) 1, arXiv:astro-ph/0601594. 124. E. M. Lifshitz and I. M. Khalatnikov, Investigations in relativistic cosmology, Adv. Phys. 12 (1963) 185. 125. R. Lupton, Statistics in Theory and Practice (Princeton University Press, Princeton, 1993). 126. D. H. Lyth and A. Riotto, Particle physics models of inﬂation and the cosmological density perturbation, Phys. Rep. 314 (1999) 1, arXiv:hep-ph/9807278. 127. C.-P. Ma and E. Bertschinger, Cosmological perturbation theory in the synchronous and conformal newtonian gauges, Astrophys. J. 455 (1995) 7, arXiv:astroph/9506072. 128. J. Maldacena, Non-Gaussian features of primordial ﬂuctuations in single ﬁeld inﬂationary models, J. High Energy Phys. 05 (2003) 013, arXiv:astro-ph/0210603. 129. L. Mandel and E. Wolf, Optical Coherence and Quantum Optics (Cambridge University Press, Cambridge, 1995). 130. T. A. Marriage et al., The Atacama Cosmology Telescope: Sunyaev–Zel’dovich — Selected galaxy clusters at 148 GHz in the 2008 survey, Astrophys. J. 737 (2011) 61, arXiv:astro-ph/1010.1065. 131. J. C. Mather et al., A preliminary measurement of the cosmic microwave background spectrum by the cosmic background explorer (COBE) satellite, Astrophys. J. 354 (1990) L37. 132. J. C. Mather et al., Measurement of the cosmic microwave background spectrum by the COBE FIRAS instrument, Astrophys. J. 420 (1994) 439.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-143

133. J. C. Mather, D. J. Fixsen and R. A. Shafer, Design for the COBE far-infrared absolute spectrophotometer, Proc. SPIE 2019, Infrared Spaceborne Remote Sensing (1993) 168. 134. A. D. Miller et al., A measurement of the angular power spectrum of the CMB from = 100 to 400, Astrophys. J. 524 (1999) L1, arXiv:astro-ph/9906421. 135. C. Meny et al., Far-infrared to millimeter astrophysical dust emission I: A model based on physical properties of amorphous solids, Astron. Astrophys. 468 (2007) 171, arXiv:astro-ph/0701226. 136. V. F. Mukhanov, Gravitational instability of the universe ﬁlled with a scalar ﬁeld, JETP Lett. 41 (1985) 493. 137. V. F. Mukhanov and G. V. Chibisov, Quantum ﬂuctuation and nonsingular universe (in Russian), JETP Lett. 33 (1981) 532. 138. V. F. Mukhanov and G. V. Chibisov, The vacuum energy and large scale structure of the universe, Sov. Phys. JETP 56 (1982) 258. 139. V. F. Mukhanov, H. A. Feldman and R. H. Brandenberger, Theory of cosmological perturbations, Phys. Rep. 215 (1992) 203. 140. C. B. Netterﬁeld et al., A measurement by BOOMERANG of multiple peaks in the angular power spectrum of the cosmic microwave background, Astrophys. J. 571 (2002) 604, arXiv:astro-ph/0104460. 141. W.-T. Ni, Dilaton ﬁeld and cosmic wave propagation, Phys. Lett. A 378 (2014) 3413, arXiv:1410.0126. 142. W.-T. Ni, Searches for the role of spin and polarization in gravity, Rep. Prog. Phys. 73 (2010) 056901, arXiv:0912.5057. 143. T. Okamoto and W. Hu, CMB lensing reconstruction on the full sky, Phys. Rev. D 67 (2003) 083002, arXiv:astro-ph/0301031. 144. D. Paradis et al., Variations of the spectral index of dust emissivity from HiGAL observations of the galactic plane, Astron. Astrophys. 520 (2010) L8, arXiv:1009.2779. 145. J. R. Pardo, J. Cernicharo and E. Serabyn, Atmospheric transmission at microwaves (ATM): An improved model for millimeter/submillimeter applications, IEEE Trans. Antennas Propagation 49 (2001) 1683. 146. J. A. Peacock, Cosmological Physics (Cambridge University Press, Cambridge, 1998). 147. P. J. E. Peebles, Lyman A. Page, Jr. and R. B. Partridge (eds.), Finding the Big Bang (Cambridge University Press, Cambridge, England, 2009). 148. P. J. E. Peebles, The Large-Scale Structure of the Universe (Princeton University Press, Princeton, 1980). 149. P. J. E. Peebles, Recombination of the primeval plasma, Astrophys. J. 153 (1968) 1. 150. P. J. E. Peebles and J. T. Yu, Primeval adiabatic perturbation in an expanding universe, Astrophys. J. 162 (1970) 815. 151. U.-L. Pen, U. Seljak and N. Turok, Power spectra in global defect theories of cosmic structure formation, Phys. Rev. Lett. 79 (1997) 1611. 152. U.-L. Pen, D. N. Spergel and N. Turok, Cosmic structure formation and microwave anisotropies from global ﬁeld ordering, Phys. Rev. D 49 (1994) 692. 153. A. A. Penzias and R. W. Wilson, A measurement of excess antenna temperature at 4080 Mc/s, Astrophys. J. 142 (1965) 421. 154. S. Perlmutter et al., Measurements of Ω and Λ from 42 High-Redshift Supernovae, Astrophys. J. 517 (1999) 565, arXiv:astro-ph/9812133. 155. Planck Collab. (P. A. R. Ade et al.), Planck early results XVIII: The power spectrum of cosmic infrared background anisotropies, Astron. Astrophys. 536 (2011) A18, arXiv:astro-ph/1101.2028.

May 4, 2017 9:55 b2167-ch15

II-144

M. Bucher

156. Planck Collab. (P. A. R. Ade et al.), Planck early results. XX. New light on anomalous microwave emission from spinning dust grains, Astron. Astrophys. 536 (2011) A20, arXiv:1101.2031. 157. Planck Collab. (P. A. R. Ade et al.), Planck intermediate results. XV. A study of anomalous microwave emission in Galactic clouds, Astron. Astrophys. 565 (2014) A103, arXiv:1309.1357. 158. Planck Collab. (P. Ade et al.), Planck intermediate results. XIV. Dust emission at millimetre wavelengths in the Galactic plane, Astron. Astrophys. 564 (2014) A45, arXiv:1307.6815. 159. Planck Collab. (R. Adam et al.), Planck intermediate results. XXX. The angular power spectrum of polarized dust emission at intermediate and high galactic latitudes, arXiv:astro-ph/1409.5738. 160. Planck Collab. (P. Ade et al.), Planck 2013 results. XI. All-sky model of thermal dust emission, Astron. Astrophys. 571 (2014), arXiv:astro-ph/1312.1300. 161. Planck Collab. (P. Ade et al.), Planck 2013 results: I. Overview of products and results, Astron. Astrophys. 571 (2014) A1, arXiv:astro-ph/1303.5062. 162. Planck Collab. (P. Ade et al.), Planck 2013 results: XII. Diﬀuse component separation, Astron. Astrophys. 571 (2014) A12, arXiv:astro-ph/1303.5072. 163. Planck Collab. (P. Ade et al.), Planck 2013 results: XIII. Galactic CO emission, Astron. Astrophys. 571 (2014) A13, arXiv:astro-ph/1303.5073. 164. Planck Collab. (P. Ade et al.), Planck 2013 results: XIV. Zodiacal emission, Astron. Astrophys. 571 (2014) A14, arXiv:astro-ph/1303.5074. 165. Planck Collab. (P. Ade et al.), Planck 2013 results: XV. CMB power spectra and likelihood, Astron. Astrophys. 571 (2014) A15, arXiv:astro-ph/1303.5075. 166. Planck Collab. (P. Ade et al.), Planck 2013 results: XVI. Cosmological parameters, Astron. Astrophys. 571 (2014) A16, arXiv:astro-ph/1303.5076. 167. Planck Collab. (P. Ade et al.), Planck 2013 results: XVII. Gravitational lensing by large-scale structure, Astron. Astrophys. 571 (2014) A17, arXiv:astro-ph/1303.5077. 168. Planck Collab. (P. Ade et al.), Planck 2013 results: XIX. The integrated Sachs–Wolfe eﬀect, Astron. Astrophys. 571 (2014) A19, arXiv:astro-ph/1303.5079. 169. Planck Collab. (P. Ade et al.), Planck 2013 results: XX. Cosmology from Sunyaev–Zeldovich cluster counts, Astron. Astrophys. 571 (2014) A20, arXiv:astroph/1303.5080. 170. Planck Collab. (P. Ade et al.), Planck 2013 results: XXI. All-sky Compton-parameter map and characterization, Astron. Astrophys. 571 (2014) A21, arXiv:astroph/1303.5081. 171. Planck Collab. (P. Ade et al.), Planck 2013 results: XXII. Constraints on inﬂation, Astron. Astrophys. 571 (2014) A22, arXiv:astro-ph/1303.5082. 172. Planck Collab. (P. Ade et al.), Planck 2013 results: XXIII. Isotropy and statistics of the CMB, Astron. Astrophys. 571 (2014) A23, arXiv:astro-ph/1303.5083. 173. Planck Collab. (P. Ade et al.), Planck 2013 results: XXIV. Constraints on primordial non-Gaussianity, Astron. Astrophys. 571 (2014) A24, arXiv:astro-ph/1303.5084. 174. Planck Collab. (P. Ade et al.), Planck 2013 results: XXV. Searches for cosmic strings and other topological defects, Astron. Astrophys. 571 (2014) A25, arXiv:astroph/1303.5085. 175. Planck Collab. (P. Ade et al.), Planck 2013 results: XXVI. Background geometry and topology of the Universe, Astron. Astrophys. 571 (2014) A26, arXiv:astroph/1303.5086.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-145

176. Planck Collab. (P. Ade et al.), Planck 2013 results: XXVII. Doppler boosting of the CMB: Eppur si muove, Astron. Astrophys. 571 (2014) A27, arXiv:astroph/1303.5087. 177. Planck Collab. (P. Ade et al.), Planck 2013 results: XXIX. The Planck catalogue of Sunyaev–Zeldovich sources, Astron. Astrophys. 571 (2014) A29, arXiv:astroph/1303.5089. 178. Planck Collab. (P. Ade et al.), Planck 2013 results. XXVIII. The Planck catalogue of compact sources, Astron. Astrophys. 571 (2014) A28, arXiv:astro-ph/1303.5088. 179. Planck Collab. (N. Aghanim et al.), Planck 2013 results. II. Low frequency instrument data processing, Astron. Astrophys. 571 (2014) A2, arXiv:1303.5063. 180. Planck Collab. (P. A. Ade et al.), Planck 2013 results. VI. High frequency instrument data processing, Astron. Astrophys. 571 (2014) A6, arXiv:1303.5067. 181. Planck Collab. (P. A. Ade et al.), Planck 2013 results. X. HFI energetic particle eﬀects, Astron. Astrophys. 571 (2014) A10, arXiv:1303.5071. 182. Polarbear Collab. (P. A. R. Ade et al.), A measurement of the cosmic microwave background B-mode polarization power spectrum at sub-degree scales with POLARBEAR, Astrophys. J. 794 (2014) 171, arXiv:1403.2369. 183. M. Pospieszalski and E. J. Wollack, Ultralow noise InP ﬁeld eﬀect transistor radio astronomy receivers: State of the art, 13 Int. Conf. Microwaves, Radar and Wireless Communications, MIKON-2000. 184. J. Preskill, Vortices and monopoles, in Architecture of Fundamental Interactions at Short Distances, ed. P. Ramond (North-Holland, Amsterdam, 1987). 185. PRISM (Polarized Radiation Imaging and Spectroscopy Mission): An Extended White Paper PRISM Collab. (P. Andr´e et al., J. Cosmol. Astropart. Phys. 02 (2014) 006, arXiv:1310.1554. 186. J. L. Puget and A. Leger, A new component of the interstellar matter: Small grains and large aromatic molecules, Ann. Rev. Astron. Astrophys. 27 (1989) 161. 187. J.-L. Puget et al., Tentative detection of a cosmic far infrared background with COBE, Astron. Astrophys. 308 (1996) L5. 188. E. M. Purcell, On the absorption and emission of light by interstellar grains, Astrophys. J. 158 (1969) 433. 189. E. M. Purcell, Suprathermal rotation of interstellar grains, Astrophys. J. 231 (1979) 404. 190. C. L. Reichardt et al., A measurement of secondary cosmic microwave background anisotropies with two years of South Pole Telescope Observations, Astrophys. J. 755 (2012) 70, arXiv:astro-ph/1111.0932. 191. Y. Rephaeli, Comptonization of the cosmic microwave background: The Sunyaev– Zeldovich eﬀect, Ann. Rev. Astron. Astrophys. 33 (1995) 541. 192. A. G. Riess et al., Observational evidence from supernovae for an accelerating universe and a cosmological constant, Astron. J. 116 (1998) 1009, arXiv:astroph/9805201. 193. J. A. Rubi˜ no-Mart´ın, C. H. L´ opez-Caraballo, R. G´enova-Santos and R. Rebolo, Observations of the polarisation of the anomalous microwave emission: A review, Advances in Astronomy 2012 (2012) 351836. 194. J. Ruhl et al., The South Pole Telescope, Proc. SPIE 5498 (2004) 11, arXiv:astroph/0411122. 195. R. K. Sachs and A. M. Wolfe, Perturbations of a cosmological model and angular variations of the microwave background, Astrophys. J. 147 (1967) 73. 196. S. Seager, D, Sasselov and D. Scott, A new calculation of the recombination epoch, Astrophys. J. 523 (1999) L1, arXiv:astro-ph/9909275.

May 4, 2017 9:55 b2167-ch15

II-146

M. Bucher

197. S. Seager, D. Sasselov and D. Scott, How exactly did the universe become neutral? Astrophys. J. Suppl. 128 (2000) 407, arXiv:astro-ph/9912182. 198. K. Sellgren, The near-infrared continuum emission of visual reﬂection nebulae, Astrophys. J. 277 (1984) 623. 199. U. Seljak, A two-ﬂuid approximation for calculating the cosmic microwave background anisotropies, Astrophys. J. 435 (1994) L87. 200. U. Seljak and M. Zaldarriaga, A line-of-sight integration approach to cosmic microwave background anisotropies, Astrophys. J. 469 (1996) 437, arXiv:astroph/9603033. 201. J. Silk, Cosmic black-body radiation and galaxy formation, Astrophys. J. 151 (1968) 459. 202. J. Silk and J. Chluba, Next steps for cosmology, Science 344 (2014) 586. 203. K. M. Smith, O. Zahn and O. Dor´e, Detection of gravitational lensing in the cosmic microwave background, Phys. Rev. D 76 (2007) 043510, arXiv:astro-ph/0705.3980. 204. G. Smoot et al., COBE diﬀerential microwave radiometers: Instrument design and implementation, Astrophys. J. 360 (1990) 685S. 205. G. F. Smoot et al., Structure in the COBE diﬀerential microwave radiometer ﬁrstyear maps, Astrophys. J. 396 (1992) L1. 206. L. Sorbo, Parity violation in the cosmic microwave background from a pseudoscalar inﬂaton, J. Cosmol. Astropart. Phys. 1106 (2011) 003, arXiv:astro-ph/1101. 1525. 207. L. Spitzer, Jr. and J. W. Tukey, A theory of interstellar polarization, Astrophys. J. 114 (1951) 187S. 208. R. A. Sunyaev and Y. B. Zeldovich, Small-scale ﬂuctuations of relic radiation, Astrophys. Space Sci. 7 (1970) 3. 209. A. A. Starobinsky, Dynamics of phase transition in the new inﬂationary universe scenario and generation of perturbations, Phys. Lett. B 117 (1982) 175. 210. A. Stuart and K. Ord, Kendall’s Advanced Theory of Statistics, Classical Inference, Vol. 2 (Oxford University Press, New York, 1991). 211. E. R. Switzer and C. M. Hirata, Primordial helium recombination. I. Feedback, line transfer, and continuum opacity, Phys. Rev. D 77 (2008) 083006, arXiv:astroph/0702143. 212. A. Taylor, K. Grainge, M. Jones, G. Pooley, R. Saunders and E. Waldram, The radio source counts at 15 GHz and their implications for cm-wave CMB imaging, Mon. Not. R. Astron. Soc. 327 (2001) L1, arXiv:astro-ph/0102497. 213. N. Turok (ed.), Critical Dialogues in Cosmology (World Scientiﬁc Publishing, Singapore, 1997). 214. A. van Engelen et al., A measurement of gravitational lensing of the microwave background using South Pole Telescope data, Astrophys. J. 756 (2012) 142, arXiv:1202.0546. 215. P. Vielva, E. Mart´ınez-Gonz´ alez, M. Cruz, R. B. Barreiro and M. Tucci, Cosmic microwave background polarization as a probe of the anomalous nature of the cold spot, Mon. Not. R. Astron. Soc. 410 (2011) 33. 216. P. Vielva, E. Mart´ınez-Gonz´ alez, R. B. Barreiro, J. L. Sanz and L. Cay´ on, Detection of non-gaussianity in the Wilkinson Microwave Anisotropy Probe ﬁrst-year data using spherical wavelets, Astrophys. J. 609 (2004) 22. 217. A. Vilenkin and E. P. S. Shellard, Cosmic Strings and Other Topological Defects (Cambridge University Press, Cambridge, 1994). 218. Wald, General Relativity (University of Chicago Press, Chicago, 1984).

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-147

219. T. P. Walker, G. Steigman, H.-S. Kang, D. M. Schramm and K. A. Olive, Primordial nucleosynthesis redux, Astrophys. J. 376 (1991) 51. 220. R. A. Watson et al., Detection of anomalous microwave emission in the perseus molecular cloud with the COSMOSOMAS experiment, Astrophys. J. 624 (2005) L89, arXiv:astro-ph/0503714. 221. S. Weinberg, Gravitation and Cosmology: Principles and Applications of the General Theory of Relativity (J. Wiley & Sons, New York, 1972). 222. J. C. Weingartner and B. T. Draine, Radiative torques on interstellar grains. III. Dynamics with thermal relaxation, Astrophys. J. 589 (2003) 289. 223. R. Williamson et al., A Sunyaev–Zel’dovich-selected sample of the most massive galaxy clusters in the 2500 deg2 South Pole Telescope survey, Astrophys. J. 738 (2011) 139, arXiv:astro-ph/1101.1290. 224. M. White and W. Hu, The Sachs–Wolfe eﬀects, Astron. Astrophys. 321 (1997) 8–9, arXiv:astro-ph/9609105. 225. E. T. Whittaker, A Treatise on the Analytical Dynamics of Particles and Rigid Bodies (Cambridge University Press, Cambridge, 1917). 226. WMAP Collab. (C. L. Bennett et al.), First year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Preliminary maps and basic results, Astrophys. J. Suppl. 148 (2003) 1. 227. WMAP Collab. (G. Hinshaw et al.), First year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Data processing methods and systematic error limits, Astrophys. J. Suppl. 148 (2003) 63. 228. WMAP Collab. (D. N. Spergel et al.), First year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Determination of cosmological parameters, Astrophys. J. Suppl. 148 (2003) 175. 229. WMAP Collab. (C. L. Bennett et al.), First year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Foreground emission, Astrophys. J. Suppl. 148 (2003) 97. 230. WMAP Collab. (H. V. Peiris et al.), First year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Implications for inﬂation, Astrophys. J. Suppl. 148 (2003) 213. 231. WMAP Collab. (L. Page et al.), First year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Interpretation of the TT and TE angular power spectrum peaks, Astrophys. J. Suppl. 148 (2003) 233. 232. WMAP Collab. (A. Kogut et al.), First year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Temperature-polarization correlation, Astrophys. J. Suppl. 148 (2003) 161. 233. WMAP Collab. (E. Komatsu et al.), First year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Tests of gaussianity, Astrophys. J. Suppl. 148 (2003) 119. 234. WMAP Collab. (G. Hinshaw et al.), First year Wilkinson Microwave Anisotropy Probe (WMAP) observations: The angular power spectrum, Astrophys. J. Suppl. 148 (2003) 135. 235. WMAP Collab. (A. Kogut et al.), Three-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Foreground polarization, Astrophys. J. Suppl. 665 (2007) 355. 236. WMAP Collab. (G. Hinshaw et al.), Three-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Temperature analysis, Astrophys. J. Suppl. 170 (2007) 288.

May 4, 2017 9:55 b2167-ch15

II-148

M. Bucher

237. WMAP Collab. (L. Page et al.), Three-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Polarization analysis, Astrophys. J. Suppl. 170 (2007) 335. 238. WMAP Collab. (D. N. Spergel et al.), Three-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Implications for cosmology, Astrophys. J. Suppl. 170 (2007) 377. 239. WMAP Collab. (J. Dunkley et al.), Five-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Bayesian estimation of CMB polarization maps, Astrophys. J. Suppl. 701 (2009) 1804–1813. 240. WMAP Collab. (G. Hinshaw et al.), Five-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Data processing, sky maps, and basic results, Astrophys. J. Suppl. 180 (2009) 225–245. 241. WMAP Collab. (B. Gold et al.), Five-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Galactic foreground emission, Astrophys. J. Suppl. 180 (2009) 265–282. 242. WMAP Collab. (M. Nolta et al.), Five-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Angular power spectra, Astrophys. J. Suppl. 180 (2009) 296– 305. 243. WMAP Collab. (J. Dunkley et al.), Five-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Likelihoods and parameters from WMAP Data, Astrophys. J. Suppl. 180 (2009) 306–329. 244. WMAP Collab. (E. Komatsu et al.), Five-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Cosmological interpretation, Astrophys. J. Suppl. 180 (2009) 330–376. 245. WMAP Collab. (N. Jarosik et al.), Seven-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Sky maps, systematic errors, and basic results, Astrophys. J. Suppl. 192 (2011) 14J. 246. WMAP Collab. (B. Gold et al.), Seven-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Galactic foreground emission, Astrophys. J. Suppl. 192 (2011) 15G. 247. WMAP Collab. (D. Larson et al.), Seven-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Power spectra and WMAP-derived parameters, Astrophys. J. Suppl. 192 (2011) 16L. 248. WMAP Collab. (C. Bennett et al.), Seven-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Are there cosmic microwave background anomalies? Astrophys. J. Suppl. 192 (2011) 17. 249. WMAP Collab. (E. Komatsu et al.), Seven-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Cosmological interpretation, Astrophys. J. Suppl. 192 (20011) 18K. 250. WMAP Collab. (C. L. Bennett et al.), Nine-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Final maps and results, Astrophys. J. Suppl. 208 (2013) 20B. 251. WMAP Collab. (G. F. Hinshaw et al.), Nine-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Cosmology results, Astrophys. J. Suppl. 208 (2013) 19H. 252. K. Yamamoto, M. Sasaki and T. Tanaka, Large angle CMB anisotropy in an open universe in the one-bubble inﬂationary scenario, Astrophys. J. 455 (1995) 412, arXiv:astro-ph/9501109. 253. O. Zahn, M. Zaldarriaga, L. Hernquist and M. McQuinn, The inﬂuence of nonuniform reionization on the CMB, Astrophys. J. 630 (2005) 657, arXiv:astro-ph/0503166.

May 4, 2017 9:55 b2167-ch15

Physics of the cosmic microwave background anisotropy

II-149

254. M. Zaldarriaga and U. Seljak, All-sky analysis of polarization in the microwave background, Phys. Rev. D 55 (1997) 1830, arXiv:astro-ph/9609170. 255. P. J. E. Peebles, Cosmic background temperature anisotropy in a minimal isocurvature model for galaxy formation, Astrophys. J. 315 (1987) L73. 256. D. Samtleben, S. Staggs and B. Winstein, The cosmic microwave background for pedestrians: A review for particle and nuclear physicists, Ann. Rev. Nucl. Part. Sci. 57 (2007) 245, arXiv:0803.0834 [astro-ph]. 257. M. Kamionkowski and A. Kosowsky, The cosmic microwave background and particle physics, Ann. Rev. Nucl. Part. Sci. 49 (1999) 77, arXiv:astro-ph/9904108. 258. R. Fabbri and M. Pollock, The eﬀect of primordially produced gravitons upon the anisotropy of the cosmological microwave background radiation, Phys. Lett. B 125 (1983) 445. 259. V. Rubakov, M. Sazhin and A. Veryaskin, Graviton creation in the inﬂationary universe and the grand uniﬁcation scale, Phys. Lett. B 115 (1982) 189. 260. A. Starobinsky, Spectrum of relict gravitational radiation and the early state of the universe, JETP. Lett. 30 (1979) 682. 261. A. Starobinsky, Cosmic background anisotropy induced by isotropic ﬂat-spectrum gravitatinal-wave perturbations, Sov. Astron. Lett. 11 (1985) 133. 262. L. Spitzer Jr. and T. A. McGlynn, Disorientation of interstellar grains in suprathermal rotation, Ap. J. 231 (1979) 417. 263. A. Lazarian and B. Draine, Thermal ﬂipping and thermal trapping: New elements in grain dynamics, Ap. J. 516 (1999) 37. 264. A. Blanchard and J. Schneider, Gravitational lensing eﬀect on the ﬂuctuations of the cosmic background radiation, Astron. Astrophys. 184 (1987) 1. 265. Planck Collab. (R. Adam et al.), Planck 2015 results. I. Overview of products and scientiﬁc results, arXiv:1502.01582. 266. BICEP2/Keck and Planck Collab. (P. A. R. Ade et al.), A joint analysis of BICEP2/Keck Array and Planck data, Phys. Rev. Lett. 114 (2015) 101301, arXiv:1502.00612.

May 4, 2017 10:3 b2167-ch16

Chapter 16 SNe Ia as a cosmological probe∗

Xiangcun Meng†, Yan Gao‡ and Zhanwen Han§ BPS Group, Yunnan Observatories, Chinese Academy of Sciences, Kunming 650216, P. R. China †[email protected] ‡[email protected] §[email protected]

Type Ia supernovae (SNe Ia) luminosities can be corrected in order to render them useful as standard candles that are able to probe the expansion history of the universe. This technique was successfully applied to discover the present acceleration of the universe. As the number of SNe Ia observed at high redshift increases and analysis techniques are perfected, people aim to use this technique to probe the equation-of-state of the dark energy (EOSDE). Nevertheless, the nature of SNe Ia progenitors remains controversial and concerns persist about possible evolution eﬀects that may be larger and harder to characterize than the more obvious statistical uncertainties. Keywords: SNe Ia; dark energy. PACS Number(s): 97.60.Bw, 98.80.−k, 95.36.+x

1. Introduction “Guest stars”, the name given to supernovae by ancient Chinese astronomers, have been recorded for several thousands of years, but the ﬁrst modern study of supernovae began on 31 August 1885. On that day, Hartwig discovered a “nova” in M31, which disappeared 18 months later. In 1919, Lundmark1 noticed that the distance to M31 is about 0.7 million ly (later studies found it to be actually 2.5 million ly), which implies that the Hartwig “nova” is brighter than other novae by three orders of magnitude. The term “supernova” was coined shortly after. Later, it was realized that supernovae can be a powerful tool for measuring the distance to extragalactic sources. However, for want of detailed enough observed spectra, supernovae were not classiﬁed until 1940.2 With the advent of consistently accurate observed spectra, it was ﬁnally realized that there are two physically distinct classes of supernovae: thermonuclear supernovae, i.e. Type Ia supernovae (SNe Ia), and core collapse supernovae, including SN IIP, IIL, IIn, IIb, Ib and Ic.3 Type I supernovae are distinguished by the absence of hydrogen lines in their optical spectra. SNe Ia are deﬁned by a deep absorption trough around 6150 ˚ A, which is now known to be due to silicon, additionally.3 II-151

May 4, 2017 10:3 b2167-ch16

II-152

X. Meng, Y. Gao and Z. Han

Due to the homogeneity of their properties, SNe Ia are often used to measure distance. Here, we will review the methods commonly used to standardize SNe Ia for this purpose in Sec. 2, and the progenitor problem in Sec. 3. In Sec. 4, we discuss the eﬀect of diﬀerent populations of SNe Ia on their brightness. In Sec. 5, we will present the application of SNe Ia in cosmology, i.e. the discovery of accelerating cosmic expansion, and the eﬀorts to probe the equation-of-state of the dark energy (EOSDE). Finally, we summarize the future promise for SNe Ia as a probe to precision cosmology in Sec. 6. 2. SNe Ia as a Standardizable Distance Candle Generally, for two nearby objects (z 1) with astronomy apparent magnitudes m1 and m2 , an Euclidean spatial geometry can be assumed, the ratio of their apparent brightness E1 /E2 , is related by the following equation: m2 −m1 E1 = 100 5 , E2

(1)

which indicates that if the apparent magnitude of one object is lower than that of another one by 5 magnitudes, the object is brighter than the other by 100 times. For the same object, its apparent magnitude at 10 pc is deﬁned as the absolute magnitude, which may reﬂect its intrinsic luminosity. Assuming the absence of extinction by dust, its apparent magnitude, m, and absolute magnitude, M , follow the equation: m−M d2 E10 = 100 5 = 2 , (2) E 10 where E10 is the apparent brightness at 10 pc, E the observed brightness, and d the distance to the object in units of parsecs. Equation (2) basically expresses the inverse-square law between apparent brightness and distance. Therefore, apparent magnitude, absolute magnitude and distance are related by the following equation:

M = m + 5 − 5 log d,

(3)

and m − M = 5 log d − 5 is called distance modulus. As we shall discuss later in Sec. 5, at larger redshifts, for which z 1 does not hold, the relation between apparent and absolute magnitudes is no longer so simple and depends on the details of the cosmological model assumed. If there is a method to independently obtain the absolute magnitude of an object, the distance may be obtained by measuring its apparent magnitude, like Eq. (3), and then this object may be taken as a standard candle to measure distance. SNe Ia are excellent candidates for standard candles after being corrected by some methods. Among all the subclasses of diﬀerent supernovae,3 SNe Ia are the most homogeneous, with many practically identical properties, e.g. lightcurve shape, maximum luminosity, spectrum, and as a result, SNe Ia were taken to be perfect standard candles.4 However, 1991 was a fateful year for SNe Ia studies, two peculiar SNe Ia were

May 4, 2017 10:3 b2167-ch16

SNe Ia as a cosmological probe

II-153

Fig. 1. The B-band lightcurves for diﬀerent SNe Ia observed. (For color version, see page IICP10.) Source: From Ref. 9.

found, SN 1991bg and 1991T. SN 1991bg was fainter than normal SNe Ia by about 2 magnitudes in the V-band,5,6 while SN 1991T was brighter than normal SNe Ia by 0.4 magnitudes.7,8 The discovery of these two peculiar SNe Ia implies a distance error of about 30% if SNe Ia are assumed to be perfect standard candles. As shown by the collection of diﬀerent supernova lightcurves in Fig. 1, diﬀerent SNe Ia have diﬀerent peak brightnesses. It became a matter of critical importance to ﬁnd a way to reduce the distance error originating from this heterogeneity of SNe Ia.106 In 1993, Phillips10 discovered that the absolute magnitude at maximum light of SNe Ia and the speed at which the luminosity fades in the B-band (blue light) over the ﬁrst 15 days after the maximum light are related, as shown in Fig. 2. This relation implies that the brightness of SNe Ia is mainly dominated by one parameter, and it is widely agreed that this parameter is the amount of 56 Ni produced during the supernova explosion that determines its maximum luminosity. Actually, one may arrive at another conclusion from Fig. 2, that the intrinsic magnitude dispersion of SNe Ia in the I-band is smaller than those in the B and V bands, i.e. the infrared measurement of SNe Ia may yield a more precise distance. However, several years later, after increasingly dim SNe Ia were included, the linear relation was found to be a quadratic or an exponential relation as shown in Fig. 3.11,12 Although this relation is widely accredited to Phillips, it was originally discovered by Rust13 and Pskovskii,14,15 who noticed the correlation between the maximum light of SNe Ia and their light decline rate.

May 4, 2017 10:3 b2167-ch16

II-154

X. Meng, Y. Gao and Z. Han

Fig. 2. The correlation between the absolute magnitudes of SNe Ia in B, V and I bands and ∆m15 (B) which measures the B-band lightcurve drops during the ﬁrst 15 days after the maximum light. Source: From Ref. 10.

Following the discovery by Phillips,10 several groups developed further improved methods to reduce the distance error, and SNe Ia were not a standard candle any more, but a distance indicator. Most notably, Riess et al.16,17 developed the multicolor lightcurve shapes (MLCS) method, which is known as the most mathematically sound method for inferring a distance. The method accounts for the nonuniform error on the distance from the eﬀects of diﬀerent bands and diﬀerent supernova colors. In this method, a “training set” of several Type Ia supernova (SN Ia) lightcurves was constructed, and a complete model of the lightcurves is dependent on three parameters, i.e. a distance modulus, a brightness oﬀset which is called “luminosity correction” and an extinction value. To calibrate the brightness of the training set, the distance modulus and the extinction value are obtained by independent methods. Actually, the brightness oﬀset reﬂects that an SN Ia with a broader lightcurve has a higher peak luminosity, i.e. the Phillips relation. The reason that a correction was made for color is that redder SNe Ia are generally less luminous, both in intrinsic terms and for dust reddening considerations.16,18

May 4, 2017 10:3 b2167-ch16

SNe Ia as a cosmological probe

II-155

Fig. 3. Modiﬁed Phillips’ relation with the so-called peculiar SNe added (ﬁlled points). The solid line is an exponential ﬁt for all the data. Source: From Ref. 12.

Compared to the Phillips relation, although the MLCS method does not signiﬁcantly increase the precision of distance measurements, the method may be the most complete and rigorous one mathematically. Figure 4 shows the typical dispersions of light and color curves after correction by the MLCS method, and from the ﬁgure, we can see that the SNe Ia can act as very good distance indicators, because the dispersion at peak brightness is very small. Almost at the same time, Perlmutter et al.19,20 developed another tool, named the stretch factor method. This method has a similar distance precision compared with the MLCS method, as shown in Fig. 5, where the data are the same to those in Fig. 1. Again, SNe Ia are proven to be potent distance indicators. It is worth noting that the MLCS method and the stretch factor method essentially take advantage of the same underlying phenomena as that underlying the Phillips relation, i.e. the slower the lightcurve evolves, the brighter the SN Ia. To obtain the parameters crucial to the implementation of these methods, various algorithms have been developed to ﬁt the lightcurves of SNe Ia data, such as BATM, MLCS2k2, SALT2 and SiFTO.21–26

May 4, 2017 10:3 b2167-ch16

II-156

X. Meng, Y. Gao and Z. Han

Fig. 4. Dispersions of noise-free light and color curves after correction by MLCS method, where the “gray snake” present 1σ-conﬁdence region. Source: From Ref. 16.

Other research groups have also tried to ﬁnd other methods to obtain more accurate SNe Ia distance measurements. For example, in 1995, Nugent et al.27 noticed a linear relation between the maximum luminosity of an SN Ia and the ratio of the fractional depths of the bluer to that of the redder absorption troughs of its silicon absorption line, or the ﬂux ratio over the Ca II, H and K absorption features in a spectrum near maximum light (see also Refs. 28 and 29). SNe Ia were therefore calibrated to be practical standard candles, and have been applied successfully for cosmological purposes, ultimately leading to the discovery of the accelerating expansion of the universe.17,20 The above methods generally decrease the distance error to 5–10%. However, for a well-observed sample, the typical dispersion on a Hubble diagram may be as small as 0.12 magnitudes, indicating a typical distance accuracy of 5% for a single SN Ia. In 2005, Wang et al.30 found that the peak brightness of SNe Ia and

May 4, 2017 10:3 b2167-ch16

SNe Ia as a cosmological probe

II-157

Fig. 5. The lightcurves for diﬀerent SNe Ia after correction by stretch factor method, where the data are the same to that in Fig. 1. (For color version, see page II-CP10.) Source: From Ref. 9.

their B–V colors at 12 days after peak B-band luminosity were strongly correlated. Calibrating SNe using this correlation, the typical error of the SNe Ia on a Hubble diagram can be reduced to less than 0.12 magnitudes. For a typical sample of SNe Ia with little reddening from host galaxies, the scatter may drop further to only 0.07 magnitudes, or 3–4% in distance.30,31 Here, it must be emphasized that for most cases, the observational error is not mainly due to imperfections in the observational apparatus, but rather from a lack of uninterrupted telescope time, i.e. it is diﬃcult to obtain a complete lightcurve for each single object. Recently, in addition to the lightcurves and the colors of the SNe Ia, some spectra properties were found to be helpful for improving the accuracy of derived distances to SNe Ia.28,29,32,33 For precision cosmology, within most redshift intervals, systematic errors from the above correlations dominate over the statistical errors, the latter of which can be dealt with by increasing the size of an observational sample.34–36 The dominant systematic uncertainties for SNe Ia are survey-dependent. As far as what the uncertainties are and how they can be improved are concerned, one may refer to relevant detailed reviews.37–39 3. Progenitors of SNe Ia In 1960, Hoyle and Fowler40 suggested that a degenerate white dwarf (WD) may ignite a runaway thermonuclear fusion in its center, and the released energy may disrupt the whole WD, resulting in an SN Ia explosion. Owing to a series of breakthroughs on the subject since then, it has been indisputably proven that all SNe Ia arise from such explosions of carbon–oxygen white dwarfs (CO WDs) in binary

May 4, 2017 10:3 b2167-ch16

II-158

X. Meng, Y. Gao and Z. Han

systems, of which the most substantial evidence is derived from the observation of SN 2011fe.41 For a CO WD to undergo such an explosion, it must reach or exceed a mass of about 1.4 M , henceforth referred to as the Chandrasekhar mass limit. The limit was named after Subrahmanyan Chandrasekhar, who was the ﬁrst person to work out that there was a maximum mass for a WD to be supported by electron degeneracy pressure. He did this by combining quantum theory with relativity, and it was the ﬁrst time to make a quantitative prediction about the maximum mass of WDs. In 1983, he was awarded the Nobel Prize in Physics mainly due to the discovery of the mass limit of WDs. However, the mass of a CO WD at birth cannot be more than 1.1 M ,42,43 according to canonical stellar evolutionary theory, and therefore the CO WD must accrete additional material somehow (i.e. from its companion in a binary system), and gradually increase its mass to the aforementioned maximum stable mass before it can achieve this state and undergo an SN explosion. The details of this accretion process will be discussed later. During an SN Ia explosion, nearly half of the mass involved is depleted and synthesized into radioactive 56 Ni, the radioactive energy of which is then injected into the supernova ejecta, heating it to the point where it becomes very luminous, ultimately resulting in the emissions which we observe.44 The amount of 56 Ni resulting from an SN explosion is the dominant determining factor behind the maximum luminosity of the SN Ia.45 However, despite all that we know, there still remain quite a few unresolved problems where SNe Ia are concerned. To name but a few: what are the progenitors of SNe Ia, or how does a CO WD increase its mass to the Chandrasekhar limit? How exactly does the WD explode? Do the basic properties of SNe Ia remain constant with redshift? Among all these basic problems, the most urgent is the one regarding the progenitors of SNe Ia, which can potentially aﬀect the precision of distance measurement via SNe Ia, and hinder the development of precision cosmology. More speciﬁcally, the progenitor problem is important, for cosmology as well as for other areas of astrophysics, in the following ways. (i) When measuring the cosmological parameters with SNe Ia, one needs to know of any possible evolution of the luminosity and birth rate of SNe Ia as a function of redshift, which is mainly determined by the progenitor model. (ii) The progenitor model provides basic input parameters for explosion models of SNe Ia. (iii) Galactic chemical evolution models require some basic parameter input, such as the stellar feedback from SNe Ia, both chemical and radiative. These basic input parameters are closely related to the progenitor models of SNe Ia. (iv) The identiﬁcation of the progenitor systems of SNe Ia may constrain binary evolution theory.46–49 Over the course of the last three decades, several progenitor models were discussed by many research groups. These models may be categorized based on the nature of the proposed companion of the CO WD, and we list the most notable ones below. (i) The single degenerate (SD) scenario50,51 : In this scenario, the companion may be a main sequence star, a subgiant, a red giant or a helium star.52–57 The accreted hydrogen-rich or helium-rich material undergoes stable burning on the

May 4, 2017 10:3 b2167-ch16

SNe Ia as a cosmological probe

II-159

surface of a CO WD, yields carbon and oxygen, which is then deposited onto the WD. When the WD mass reaches some value approaching the Chandrasekhar mass limit, an SN Ia is produced. The companion survives after the supernova explosion. This scenario is widely accepted and is the leading scenario. (ii) The double degenerate (DD) scenario: In this scenario, a binary system comprising two CO WDs loses its orbital angular momentum by means of gravitational radiation, and merges as a consequence. If the total mass of the merger exceeds the Chandrasekhar mass limit, an SN Ia may be expected.58–61 In this scenario, the merger is disrupted completely and the companion does not survive. Note, however, it is possible that the merger product may experience a core collapse rather than a thermonuclear explosion.62,63 The SD and DD scenarios are the two main competing scenarios in the SN Ia community. Generally, to form an SD system leading to an SN Ia, the primordial binary system needs to experience one common envelope phase, while to form a DD system, the situation is more complicated and one extra common envelope phase can be necessary. For example, in Fig. 6, three subchannels lead to a WD + main sequence system ending in an SN Ia explosion, and for every subchannel, the system experiences a single common envelope phase only (see details in Wang and Han48 ).

Fig. 6. Schematic for the subchannels producing WD + MS systems, which may lead to SNe Ia. For every subchannel, the binary system experiences a common envelope phase. (For color version, see page II-CP11.) Source: From Ref. 48.

May 4, 2017 10:3 b2167-ch16

II-160

X. Meng, Y. Gao and Z. Han

In addition to the models mentioned above, the double-detonation model (subChandrasekhar mass model) is also frequently discussed. In this model, the companion of the CO WD is a helium WD or a helium star, which ﬁlls its Roche lobe and initiates a stable mass transfer. If the mass-transfer rate is not high enough, the helium material may not burn stably, and as a consequence gradually accumulates on the CO WD. When the helium layer is massive enough, a detonation, which is a supersonic blast wave, is ignited at the bottom of the helium layer, in which the critical mass to ignite a detonation is generally dependent on the WD mass and the mass-transfer rate.64 At the moment of ignition, WD is generally less massive than the Chandrasekhar mass limit. The resulting detonation wave compresses the inner CO material, leading to a second detonation near the core of the CO WD.65–67 After the supernova explosion, a low mass helium star or a low mass helium WD may survive to form a hypervelocity star, whose velocity is high enough to exceed the escape velocity of its parent galaxy.68 Remarkable progress has been made on this model recently, though many problems still exist.69,70 If this model does indeed produce SNe Ia, it would contribute quite signiﬁcantly to SNe Ia rates.71 Besides the SD and DD scenarios and the double detonation model, other similar models have been proposed to explain the diversity generally observed among SNe Ia, such as the super-Chandrasekhar mass model, the single star model, the spinup/spin-down model, the delayed dynamical instability model, the core-degenerate model, the model of a collision between two WDs, and the model of WDs near black holes, etc. (See Wang and Han48 for a review.) At present, it is premature to exclude any of these models conclusively, and no single model can explain all the properties of SNe Ia alone. It is possible that many of these models, or at least both the SD and DD scenarios, contribute to SNe Ia.48,49,72,73 4. Eﬀect of SN Ia Populations on Their Brightness When applying SNe Ia as cosmological standard candles, it is usually assumed that their properties and the above calibrated relations are invariant with redshift, i.e. the demographics of SNe Ia does not evolve with redshift, or the diﬀerence between SNe Ia arising from diﬀerent stellar populations is negligible. This assumption is crucial for precision cosmology, since it may lead to systematic errors in the measurements of cosmological parameters. One way to test this assumption is to search for correlations between the properties of SNe Ia and those of their environments, since the environments of SNe Ia can represent the host population which gives rise to SNe Ia to a great extent. By the merit of these studies, it is now known that SN lightcurves and peak luminosities are correlated with host galaxy star formation rate, host population age, galaxy mass and metallicity.74–77 Even before SNe Ia were applied to measure cosmological parameters, it was generally agreed upon that the brightest SNe Ia always occur in late-type galaxies.78 It should be noted, however, that both late- and early-type galaxies can host dimmer

May 4, 2017 10:3 b2167-ch16

SNe Ia as a cosmological probe

II-161

SNe Ia, and subluminous SNe Ia tend to be discovered in galaxies with a signiﬁcant old population,79 which leads to a lower mean peak luminosity for SNe Ia in earlytype galaxies than in their late-type counterparts.78,80 In addition, the mean peak brightness of SNe Ia in a galaxy is more homogeneous in the outer than in the inner regions.81,82 Also, both the birth rate and luminosity of SNe Ia trace the star formation rate of host galaxies.74,77 All these observations seem to imply that age is one of the factors which aﬀect the luminosity of SNe Ia, and that dimmer SNe Ia can arise from progenitors spanning a wide range of ages.75,76,83–85 Considering that the typical galactic star formation rates increase by a factor of ∼ 10 up to z = 1.5, one may expect that the luminosities of SNe Ia will generally increase with redshift, and it has been shown that the mean “stretch factor” increases as well.86,87 According to these observations, we may conclude that the average value of the maximum luminosities of SNe Ia, as well as the range over which they span, both decrease with their delay times. It can also be concluded that, should the maximum luminosity be determined by only one parameter, as postulated by the Phillips relation,10 it can be expected that the possible range of the parameter must decrease, and that its mean value must either increase or decrease with the age of SNe Ia. At present, although it has been widely accepted that the peak luminosity of SNe Ia is dictated by the quantity of 56 Ni produced during their explosion, it is still controversial which parameter physically determines the 56 Ni production.88 The dimmer brightness for an old SN Ia could be either due to a high ignition density for a long cooling time before accretion,89,90 or due to a lower carbon abundance requirement for a massive WD with a less massive secondary under the frame of the SD model.91–93 In addition, the total mass of the DD systems was suggested to the origin of the brightness variation of SNe Ia,37,94 but it seems not to ﬁt the age constraints on the brightness of SNe Ia from observations.95 Metallicity is another factor aﬀecting the luminosity of SNe Ia.75–77,96,97 Theoretically, high metallicity progenitors would produce subluminous SNe Ia: A high 22 Ne abundance in high metallicity WDs generates more neutrons to feed the explosion nucleosynthesis, yielding more stable 58 Ni, a process which consumes the radioactive 56 Ni powering the lightcurve of SNe Ia.98 The spectra of high-z SNe Ia, whose metallicity is on an average lower than that of local SNe Ia, contain less intermediate mass elements, which is consistent with the idea that these SNe Ia produce more 56 Ni to power their luminosity.87 Metallicity may lead to an ∼ 25% variation on the peak luminosity of SNe Ia, but observation has shown that metallicity might only have an ∼ 10% eﬀect on this luminosity.75 Such a diﬀerence between theory and observation could be the result of a metallicity–age–carbon abundance degeneracy, or an unknown metallicity eﬀect. A high metallicity may lead to a relatively shorter delay time,95,99,100 regardless of what the progenitor model is, and hence a slightly brighter SNe Ia. On the other hand, a star of given mass with a high metallicity will produce a relatively less massive WD42,43 that has a higher carbon abundance,92 which was suggested to produce a brighter SNe Ia.92,93 Therefore, the

May 4, 2017 10:3 b2167-ch16

II-162

X. Meng, Y. Gao and Z. Han

eﬀect of metallicity on the peak luminosity of SNe Ia could be more complex than what one would imagine at ﬁrst glance. The mass of a host galaxy may also signiﬁcantly aﬀect the properties of SNe Ia. For example, many observations have shown that the birth rate of SNe Ia depends heavily on the mass of their host galaxies.101–103 Compared with the age and metallicity, the eﬀect of the mass of host galaxies could be more signiﬁcant.77,84,96,104,105 More massive galaxies tend to give rise to dimmer SNe Ia,77,105 a trend which seems to be a combined result of the eﬀects of age and metallicity on the brightness of SNe Ia, i.e. higher mass galaxies have the tendency to retain more metals due to gravitational eﬀects, and at the same time, tend to have more old stars, supergiant elliptical galaxies in particular. The above facts imply that the peak luminosities of SNe Ia evolve with redshift. Despite this, if the absolute magnitude of SNe Ia corrected by their lightcurve shape and color does not change with redshift, it is not necessarily problematic to take SNe Ia as standard candles for cosmology. However, people have to face an embarrassing situation that the age, metallicity and host galaxy mass have a systemic eﬀect on the brightness of SNe Ia even after their brightness was corrected by the lightcurve shape and color.77,84,96,97,104,105 For example, as shown in Fig. 7, Sullivan et al.77 found that SNe Ia originating in massive host galaxies and galaxies with low speciﬁc star-formation rates, of the

Fig. 7. Residuals from the best-ﬁtting cosmology for a low-redshift sample as a function of host galaxy mass after correction for lightcurve shape and color. The red circles are the mean residuals in bins of host galaxy mass. The histograms on the right-hand the distribution of the residuals for SNe Ia in low (dashed histogram) and high (gray histogram) mass hosts, where the mass boundary is 1010 M , and the horizontal lines show the average residuals for the subsamples in low and high mass hosts. (For color version, see page II-CP11.) Source: From Ref. 77.

May 4, 2017 10:3 b2167-ch16

SNe Ia as a cosmological probe

II-163

same lightcurve shape and color, are generally 0.08 magnitudes brighter than those found anywhere else. This result does not depend on any assumed cosmological model or SN lightcurve width. Such a trend may lead to a systematic error on the measurement of the EOSDE, if it is not corrected. Therefore, besides the lightcurve shape and the color, at least a third parameter to correct the brightness of SNe Ia, which is correlated to one or more properties of the host galaxies, must be incorporated to avoid systematic Hubble diagram residuals. 5. SN Ia’s Role in Cosmology The previous sections have discussed the properties of SNe Ia with an emphasis on the corrections that can be applied to compensate for the variations in the absolute luminosities and also in their spectra.a In this section, we focus on how Eq. (3) must be modiﬁed when the condition z 1 no longer holds. We emphasize how information concerning the cosmological model can be extracted from the observed redshift-apparent magnitude relation. To this end, we must ﬁrst reﬁne our deﬁnition of distance, which is complicated by the expansion of the universe. The “co-moving” distance dCM to an object at redshift z is given by the integral z H0 dz dCM = cH0 −1 , (4) H(z ) 0 where H is the Hubble constant, which is a function of z, and H0 is its present value. For a universe ﬁlled with radiation, matter and a cosmological constant: H0 = [ΩR (1 + z )4 + ΩM (1 + z )3 + Ωk (1 + z )2 + ΩΛ ]−1/2 , H(z )

(5)

where Ωk = (1 − ΩR − ΩM − ΩΛ ). In the case of a “quintessence” or “dark energy” component of constant w = p/ρc2 instead of a cosmological constant (with w = −1), we would replace the ΩΛ term with ΩDE (1 + z )3(1+w) . For Ωk = 0 (a spatially ﬂat universe), the “co-moving” distance as deﬁned above coincides with the “angulardiameter” distance dAD , but in the presence of a nonﬂat spatial geometry, H −1 (Ωk )−1/2 sinh[H0 (Ωk )1/2 dCM (z)], if Ωk > 0, 0 dAD (z) = dCM (z), (6) if Ωk = 0, H −1 (−Ω )−1/2 sin[H (−Ω )1/2 d (z)], if Ω < 0. 0

k

0

k

CM

k

Finally, the “luminosity” distance dL is deﬁned in terms of the “angular-diameter” distance according to dL (z) = (1 + z)dAD (z), a When

(7)

the apparent luminosities of objects at diﬀering redshifts are being compared, care must be taken to correct for the redshifting of the spectral bands. This is known as the so-called “Kcorrection”.

May 4, 2017 10:3 b2167-ch16

II-164

X. Meng, Y. Gao and Z. Han

where the (1 + z) factor has been inserted in order to account for two eﬀects: (i) As a photon propagates from redshift z to us today, its energy is redshifted, or diminished, by a factor of (1 + z)−1 , and (ii) the rate of emission of photons is lowered by a factor of (1 + z)−1 in terms of today’s time, rather than the time at the instant of emission. It follows that using the formula =

L , dL 2

(8)

where L is the absolute (bolometric) luminosity and is the apparent (bolometric) luminosity holds for all redshifts when the luminosity distance dL as deﬁned above is used. The object of cosmological supernova studies is to measure the function dL (z) beyond its ﬁrst-order linear term whose coeﬃcient is cH0 −1 . By measuring the higher-order corrections (quadratic order and beyond), we uncover the equation-ofstate of what is driving the present expansion of the universe, and through the cubic and higher terms discover how this equation-of-state evolved with redshift. The SN Ia’s role in cosmology has been reviewed by many authors and further details on this technique can be found in their reviews.38,39,106,107 Here, we summarize some historic events in this ﬁeld. In 1927, Lemaˆıtre108 deduced that the velocity of recession of distant objects relative to an observer can be approximated to be proportional to distance, should Einstein’s general relativity equation be assumed to be true. In 1929, Hubble109 noticed that a linear relation actually exists between an object’s recession velocity and its distance, judging by 22 external galaxies. This relation was henceforth named Hubble’s law, and hallmarks the discovery of the expansion of the universe. After this discovery, two questions arise naturally, i.e. what is the future of the universe? and what are the energy sources driving the expansion of the universe? To answer these questions, one needs to measure the recession velocity variation as a function of time or redshift, i.e. the deceleration parameter of the expansion q0 . In modern cosmology, including Einstein’s cosmological constant,110 the cosmic deceleration parameter is deﬁned as ΩM − ΩΛ , (9) 2 where ΩM is the ratio between the actual mass density of the universe and the critical density above which the universe would collapse again someday if ΩΛ = 0, which is only dependent on the actual mass density of the universe, and ΩΛ describes how the cosmological constant aﬀects the expansion of the universe.107 For a universe of ΩΛ = 0, the behavior of the universe is determined by ΩM and ΩΛ together. Here, q0 = 0 leads to a constant expansion velocity, q0 < 0 leads to an accelerating expansion and q0 > 0 leads to a decelerating expansion. It can be clearly seen in the equation that the expansion of the universe is accelerating if ΩΛ > (ΩM /2). For a universe of ΩΛ = 0, q = (ΩM /2), and therefore, the universe is q0 =

May 4, 2017 10:3 b2167-ch16

SNe Ia as a cosmological probe

II-165

Fig. 8. Joint conﬁdence level for (ΩM , ΩΛ ) from SNe Ia corrected by MLSC method. Source: From Ref. 17.

always decelerating, where q > 12 , q = 12 and q < 12 corresponds to the closed, ﬂat and open universe, respectively (please see Fig. 8). Although it was known that SNe Ia are very good standard candles,4 signiﬁcant progress on this subject had yet to be made until the 1990s, after the advent of CCD imagers.111,112 In the 1990s, two teams focused on the measurement of the distance to high-z supernovae, i.e. the Supernova Cosmology Project (SCP) and the High-z team (HZT). Both teams built up large samples, including those measured by the Hubble Space Telescope (HST), and then individually published their twin studies.17,20 In the twin studies, they adjusted the parameters of an (ΩM , ΩΛ ) universe to ﬁt their data, as shown in Fig. 8 which is an example from Riess et al.,17 and obtained a similar conclusion. The surprising result is that high-z SNe Ia were observed to be fainter than expected from their low-z counterparts in

May 4, 2017 10:3 b2167-ch16

II-166

X. Meng, Y. Gao and Z. Han

a matter-dominated universe. A matter-dominated closed universe (i.e. ΩM = 1) was unambiguously ruled out at a conﬁdence level of greater than 7σ. Neither does an open, Λ = 0 cosmology ﬁt the data well; the cosmological constant is nonzero and positive, with a conﬁdence level of P (Λ > 0) = 99%. For a ﬂat universe (ΩM + ΩΛ =1), a positive ΩΛ was found to be required at a conﬁdence level of 7σ. Actually, the probability of an accelerating expansion of the universe with (ΩM > 0, ΩΛ > 0, q0 < 0) was modest, typically 3σ or less, i.e. the conﬁdence level for a Λ dominated accelerating expansion was less than 3σ. For a ﬂat universe, the best-ﬁt parameters are ΩM ≈ 0.3 and ΩΛ ≈ 0.7, which supported an accelerated expansion as derived from Eq. (9). The twin studies are regarded as the original evidences for the discovery of the accelerating-expansion universe, and in 2011, Saul Perlmutter, Brian Schmidt and Adam Riess were awarded the Nobel Prize in Physics for the discovery. It should be noticed that the expansion of the universe is actually evolving. At z ≈ 0.45, the density of matter and dark energy was equal, and matter dominated at early epochs, i.e. the expansion of the universe was decelerating at z > 0.45.39 The discovery of accelerating expansion today implies that our universe is dominated by some unknown form of energy, which is often characterized by the ratio of pressure to density, i.e. the dark energy equation-of-state, w = P/ρc2 . For some arbitrary length scale a, the density of the universe evolves as ρ ∝ a−3(1+w) .

(10)

From this equation, it is easily seen that w = 0 for normal matter, i.e. the density decreases as the universe expands. If w = −1, which corresponds to Einstein’s cosmological constant, its density is a constant, and does not dilute with the expansion of the universe. A value of w < −1 is permitted although it is debatable theoretically. In this situation, the dark energy density will increase with time, and may ultimately lead to the destruction of galaxies.37,113 If −1 < w < 0, the dark energy density will decrease with time, and it will become problematic whether the expansion of the universe will continue to accelerate if the value of w is large enough. If ΩM is assumed to be negligible, a value of w < − 31 for dark energy would be suﬃcient for the indeﬁnite acceleration of its expansion. However, considering that it is very likely for ΩM to have a value of larger than 0.2,17,20 the dark energy has to have an even smaller value of w for this to happen. It is veriﬁed that the critical point for the expansion of the universe to continue accelerating is in fact w = − 12 .114 However, it is possible that w is not a constant, but rather a variable which also evolves with time in the general form w(z) = w0 + wa z/(1 + z), where w will be a constant w0 if wa = 0. Following the discovery of dark energy, the EOSDE and the evolution of EOSDE with time became burning issues. Although the measurement of w is still consistent with an accelerating expansion of the universe, the uncertainty of w is still very large in those early observations. The value of w was generally located between −0.7 and −1.6 at a typical conﬁdence level of 95%.22,114,115 Today, more

May 4, 2017 10:3 b2167-ch16

SNe Ia as a cosmological probe

II-167

Fig. 9. Conﬁdence level contours in the plane of cosmological parameters (ΩM , w0 , wa ). The horizontal dashed lines indicate a cosmological constant, i.e. w0 = −1 and wa = 0. Source: From Ref. 120.

and more evidences based on SNe Ia observations from SNe Ia groups and other observational constraints, such as those from Supernova Legacy Survey (SNLS), Harvard-Smithsonian Center for Astrophysics (CfA), Sloan Digital Sky Survey (SDSS) and SCP, support a ﬂat universe with a constant w = −1, i.e. at least rapidly evolving forms of dark energy were ruled out although the current observational constraints on wa are weak.34–36,116–120 For example, in Fig. 9, combined with full Wilkinson Microwave Anisotropy Probe 7-year (WMAP7) power spectrum, SNLS3 and SDSS DR7 data, ΩM = 0.271 ± 0.015, w0 = −0.905 ± 0.196 +1.094 were obtained, which are consistent with a ﬂat, w = −1 and wa = −0.984−1.097 120 For most of these studies, the HST played a vital role in determining universe. the cosmological parameters, and provides many high-quality, high-z observations which remain one of our most eﬃcient tools for understanding the properties of dark energy.115,116,121 An event at z = 1.914 was even discovered recently by HST, whose magnitude is consistent with that expected from ΛCDM cosmology.122 6. Issues and Prospects SNe Ia are now applied to constrain the properties of the dark energy and the evolution of the properties with time. Although SNe Ia may provide more strict constraints than some other surveys, such as baryon acoustic oscillations (BAO) and weak gravitational lensing (WL), the combination of the results from more than one survey is often applied to constrain the EOSDE parameter.113,120 For dark energy, a change of 1% in w only leads to a change of 3% in dark energy

May 4, 2017 10:3 b2167-ch16

II-168

X. Meng, Y. Gao and Z. Han

density at redshift z = 2 and a change of 0.2% at z = 1−2.113 Therefore, to study the properties of the dark energy quantitatively, a precision of better than 2% to measure distance is required.123 However, even for the most well-observed sample, the present distance accuracy seems not to be good enough. At present, a wellobserved low redshift sample may fulﬁll the requirement of 2% precision, but the distance error increases to ∼ 5% at redshifts greater than 1.106,127 As we know, SNe Ia with redshifts greater than 1 provide the best constraints on the EOSDE. For precision cosmology, there are still several diﬃculties with regard to SNe Ia. First, not all SNe Ia may be taken to be distance indicators, such as 2002cx-like supernovae.124 Such SNe Ia conform to no relations which regular SNe Ia do, such as the Phillips relation, and therefore cannot be standardized. Another subclass of SNe Ia that do not adhere to the Phillips relation are the so-called superluminous SNe Ia.125 If these peculiar supernovae are not excluded from an SNe Ia sample, the conﬁdence level of the resulting distance measurement would decrease greatly. Second, new lightcurve analysis methods which have been developed tend to be based on new intrinsic parameters, which are assumed either to be invariant with redshift, or at least to vary with redshift in such a way that does not introduce further systematic uncertainty. Despite the reduction of the luminosity uncertainty from 0.15 magnitudes to less than 0.10 magnitudes, or better than 5% in terms of distance uncertainty,33,126 which these methods serve to achieve, such assumptions ought to be checked carefully, since it has been shown that systematic-uncertaintyinducing evolution may exist,86,87 and besides the lightcurve shape and the color, a third parameter correlated to one or some properties of the host galaxy is necessary. Finally, it is crucial to increase the detection eﬃciency of SNe Ia at high redshifts for further studies, especially at redshifts greater than 1. Two factors may aﬀect this detection eﬃciency. One is that only luminous SNe Ia are likely to be detected due to Malmquist bias, but luminous SNe Ia are relatively rare. The other is that most of the radiation may be shifted into infrared bands for the SNe Ia at high redshifts, which makes it diﬃcult to detect SNe Ia by optical telescopes. Infrared space telescopes, in this context, may be an alternative for future SNe Ia science.38 For example, the recently installed Wide-Field Camera 3 on HST has granted SN surveys an unprecedented depth, i.e. it is possible to detect a typical SN Ia at z ≈ 2.5 by HST. In addition to the diﬃculties mentioned above, for high-z supernovae, the conﬁdence level for classiﬁcation by spectrum is not as high as that for low-z supernovae. For a high-z SNe Ia sample, although practical studies of photometric identiﬁcation have developed methods which achieve a purity of ∼ 95% for SNe Ia,128–130 there exists a possibility that such samples may be contaminated by a few Type Ib/Ic supernovae.122,131 Therefore, techniques to remove such sample impurities need to be developed. However, if an appropriate observational strategy is adopted, the eﬀect of time dilation could provide a great opportunity for high-z SN observation. If the cosmological redshift is derived from the expansion of the universe, the observed time interval of an event will be dilated by a factor of (1 + z), which means that

May 4, 2017 10:3 b2167-ch16

SNe Ia as a cosmological probe

II-169

the lightcurve of a high-z SN Ia will present a broadening peak and a more shallow slope for the later exponential decline phase. Therefore, the eﬀect of time dilation increases the probability to detect the high-z SNe Ia, although it becomes relatively diﬃcult to obtain a complete light curve. Acknowledgments We are very grateful to the anonymous referees for their kind comments that improved this manuscript greatly. This work was partly supported by NSFC (Grant Nos. 11473063, 11522327, 11390374 and 11521303), the Western Light Key Project of the Chinese Academy of Sciences and Key Laboratory for the Structure and Evolution of Celestial Objects, Chinese Academy of Sciences. Z.H. thanks the support by the Strategic Priority Research Program “The Emergence of Cosmological Structures” of the Chinese Academy of Sciences, Grant No. XDB09010202, and Science and Technology Innovation Talent Programme of the Yunnan Province (Grant No. 2013HA005). References 1. 2. 3. 4. 5. 6. 7. 8. 9.

10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26.

K. Lundmark, K. Sven. Vetenskapsakad. Handl. 60 (1920) 1. R. Minkowski, Publ. Astron. Soc. Pac. 52 (1940) 206. A. V. Filippenko, Annu. Rev. Astron. Astrophys. 35 (1997) 309. D. Branch and G. A. Tammann, Annu. Rev. Astron. Astrophys. 30 (1992) 359. A. V. Filippenko et al., Astron. J. 104 (1992) 1543. B. Leibundgut et al., Astron. J. 105 (1993) 301. A. V. Filippenko et al., Astrophys. J. 384 (1992) L15. M. M. Phillips et al., Astron. J. 103 (1992) 1632. A. Kim, Stretched and non-stretched B-band Supernova lightcurves, LBNL Report No. LBNL-56164 Lawrence Berkeley National Laboratory, US Department of Energy, Berkeley, California (2004) M. M. Phillips, Astrophys. J. Lett. 413 (1993) L105. M. M. Phillips et al., Astron. J. 118 (1999) 1766. P. M. Garnavich et al., Astrophys. J. 613 (2004) 1120. B. W. Rust, The use of supernovae lightcurves for testing the expansion hypothesis and other cosmological relations, Ph.D. thesis, University of Illinois (1974). Y. P. Pskovskii, Sov. Astron. 21 (1977) 675. Y. P. Pskovskii, Sov. Astron. 28 (1984) 658. A. Riess et al., Astrophys. J. 473 (1996) 88. A. Riess et al., Astron. J. 116 (1998) 1009. A. Conley et al., Astrophys. J. 664 (2007) L13. S. Perlmutter et al., Astrophys. J. 483 (1997) 565. S. Perlmutter et al., Astrophys. J. 517 (1999) 565. G. Contardo et al., Astron. Astrophys. 359 (2000) 876. J. L. Tonry et al., Astrophys. J. 594 (2003) 1. S. Jha, A. G. Riess and R. P. Kirshner, Astrophys. J. 659 (2007) 122. J. Guy et al., Astron. Astrophys. 466 (2007) 11. A. Conley et al., Astrophys. J. 681 (2008) 482. S. Dado and A. Dar, Astrophys. J. 809 (2015) 39, arXiv:1301.3333 [astro-ph.HE].

May 4, 2017 10:3 b2167-ch16

II-170

X. Meng, Y. Gao and Z. Han

27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46.

P. Nugent et al., Astrophys. J. Lett. 455 (1995) L147. S. Bailey et al., Astron. Astrophys. 500 (2011) L17. J. M. Silverman et al., Mon. Not. R. Astron. Soc. 425 (2012) 1889. X. F. Wang et al., Astrophys. J. Lett. 620 (2005) L87. X. F. Wang et al., Astrophys. J. 645 (2006) 488. K. Maeda et al., Mon. Not. R. Astron. Soc. 413 (2011) 3075. A. G. Kim et al., Astrophys. J. 766 (2013) 84. R. Kessler et al., Astrophys. J. Suppl. Ser. 185 (2009) 32. M. Hicken et al., Astrophys. J. 700 (2009) 1097. R. Amanullah et al., Astrophys. J. 716 (2010) 712. D. A. Howell, Nat. Commun. 2 (2011) 350. B. Leibundgut, Annu. Rev. Astron. Astrophys. 39 (2001) 67. P. Astier, Rep. Prog. Phys. 75 (2012) 116901. F. Hoyle and W. A. Fowler, Astrophys. J. 132 (1960) 565. P. E. Nugent et al., Nature 480 (2011) 344. H. Umeda et al., Astrophys. J. 513 (1999) 861. X. Meng, X. Chen and Z. Han, Astron. Astrophys. 487 (2008) 625. D. Branch, Nature 431 (2004) 1044. W. D. Arnett, Astrophys. J. 253 (1982) 785. M. Livio, in Type Ia Supernovae: Theory and Cosmology, eds. J. Truran and J. C. Niemeyer (Cambridge University Press, New york, 1999), p. 33. B. Leibundgut, Astron. Astrophys. Rev. 10 (2000) 179. B. Wang and Z. Han, New Astron. Rev. 56 (2012) 122. D. Maoz, F. Mannucci and G. Nelemans, Annu. Rev. Astron. Astrophys. 52 (2014) 107. J. Whelan and I. Iben, Astrophys. J. 186 (1973) 1007. K. Nomoto, F.-K. Thielemann and K. Yokoi, Astrophys. J. 286 (1984) 644. X. D. Li and E. P. J. van den Heuvel, Astron. Astrophys. 322 (1997) L9. I. Hachisu, M. Kato, K. Nomoto and H. Umeda, Astrophys. J. 519 (1999) 314. I. Hachisu, M. Kato and K. Nomoto, Astrophys. J. 522 (1999) 487. N. Langer et al., Astron. Astrophys. 362 (2000) 1046. Z. Han and Ph. Podsiadlowski, Mon. Not. R. Astron. Soc. 350 (2004) 1301. B. Wang, X. Meng, X. Chen and Z. Han, Mon. Not. R. Astron. Soc. 395 (2009) 847. I. Iben and A. V. Tutukov, Astrophys. J. Suppl. Ser. 54 (1984) 335. R. F. Webbink, Astrophys. J. 277 (1984) 355. R. Pakmor et al., Nature 463 (2010) 61. M. H. van Kerkwijk, P. Chang and S. Justham, Astrophys. J. 722 (2010) L157. H. Saio and K. Nomoto, Astron. Astrophys. 150 (1985) L21. H. Saio and K. Nomoto, Astrophys. J. 500 (1998) 388. I. Iben and A. V. Tutukov, Astrophys. J. 342 (1989) 430. S. E. Woosley and T. A. Weaver, Astrophys. J. 423 (1994) 371. E. Livne and D. Arnett, Astrophys. J. 452 (1995) 62. K. J. Shen, J. Guillochon and R. J. Foley, Astrophys. J. 770 (2013) L35. S. Geier et al., Science 347 (2015) 1126. K. J. Shen and L. Bildsten, Astrophys. J. 699 (2009) 1365. S. A. Sim et al., Mon. Not. R. Astron. Soc. 420 (2012) 3003. A. J. Ruiter et al., Mon. Not. R. Astron. Soc. 417 (2011) 408. P. Ruiz-Lapuente, New Astron. Rev. 62 (2014) 15. X. F. Wang et al., Science 340 (2013) 170. M. Sullivan et al., Astrophys. J. 648 (2006) 868.

47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74.

May 4, 2017 10:3 b2167-ch16

SNe Ia as a cosmological probe

75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121.

II-171

D. A. Howell et al., Astrophys. J. 691 (2009) 661. J. S. Gallagher et al., Astrophys. J. 685 (2008) 752. M. Sullivan et al., Mon. Not. R. Astron. Soc. 406 (2010) 782. M. Hamuy et al., Astron. J. 112 (1996) 2391. D. A. Howell, Astrophys. J. 554 (2001) L193. T. D. Brandt et al., Astron. J. 140 (2010) 804. L. Wang, P. H¨ oflich and J. C. Wheeler, Astrophys. J. Lett. 483 (1997) L29. A. G. Riess et al., Astron. J. 117 (1999) 707. J. D. Neill et al., Astrophys. J. 707 (2009) 1449. R. R. Gupta et al., Astrophys. J. 740 (2011) 92. M. Rigault et al., Astrophys. J. 802 (2015) 20. D. A. Howell, M. Sullivan, A. Conley and R. Carlberg, Astrophys. J. Lett. 667 (2007) L37. M. Sullivan et al., Astrophys. J. Lett. 693 (2009) L76. Ph. Podsiadlowski et al., New Astron. Rev. 52 (2008) 381. B. K. Krueger et al., Astrophys. J. Lett. 719 (2010) L5. X. Chen, Z. Han and X. Meng, Mon. Not. R. Astron. Soc. 438 (2014) 3358. X. Meng and W. Yang, Astrophys. J. 710 (2010) 1310. H. Umeda et al., Astrophys. J. Lett. 522 (1999) L43. X. Meng and W. Yang, Astron. Astrophys. 531 (2011) A94. D. Maoz and F. Mannucci, Publ. Astron. Soc. Aust. 29 (2012) 447. X. Meng and W. Yang, Astron. Astrophys. 543 (2012) A137. M. Childress et al., Astrophys. J. 770 (2013) 108. B. T. Hayden et al., Astrophys. J. 764 (2013) 191. F. X. Timmes, E. F. Brown and J. W. Truran, Astrophys. J. Lett. 590 (2003) L83. X. Meng, X. Chen and Z. Han, Mon. Not. R. Astron. Soc. 395 (2009) 2103. X. Meng, Z. Li and W. Yang, Publ. Astron. Soc. Jpn. 63 (2011) L31. F. Mannucci, Chin. J. Astron. Astrophys. Suppl. 8 (2008) 143. W. Li et al., Mon. Not. R. Astron. Soc. 412 (2011) 1473. Y. Gao and C. Pritchet, Astron. J. 145 (2013) 83. J. Johansson et al., Mon. Not. R. Astron. Soc. 435 (2013) 1680. M. Rigault et al., Astron. Astrophys. 560 (2013) A66. A. G. Kim et al., Astropart. Phys. 63 (2015) 2. S. M. Carroll, W. H. Press and E. L. Turner, Annu. Rev. Astron. Astrophys. 30 (1992) 499. G. Lemaˆıtre, Ann. Soc. Sci. Brux. A 47 (1927) 49. E. Hubble, Contrib. Mt. Wilson Obs. 3 (1929) 23. A. Einstein, Sitz. der K¨ onigl. Preuß. Akad. der Wiss. (1917) 142. M. Hamuy et al., Astron. J. 112 (1996) 2408. M. Hamuy et al., Astron. J. 112 (1996) 2398. J. A. Frieman, M. S. Turner and D. Huterer, Annu. Rev. Astron. Astrophys. 46 (2008) 385. R. A. Knop et al., Astrophys. J. 598 (2003) 102. A. G. Riess et al., Astrophys. J. 607 (2004) 665. A. G. Riess et al., Astrophys. J. 659 (2007) 98. G. Miknaitis et al., Astrophys. J. 666 (2007) 674. W. M. Wood-Vasey et al., Astrophys. J. 666 (2007) 694. A. Conley et al., Astrophys. J. Suppl. Ser. 192 (2011) 1. M. Sullivan et al., Astrophys. J. 737 (2011) 102. P. M. Garnavich et al., Astrophys. J. Lett. 493 (1998) L53.

May 4, 2017 10:3 b2167-ch16

II-172

122. 123. 124. 125. 126. 127. 128. 129. 130. 131.

X. Meng, Y. Gao and Z. Han

D. O. Jones et al., Astrophys. J. 768 (2013) 166. P. H¨ oflich et al., Astrophys. J. 710 (2010) 444. W. D. Li et al., Publ. Astron. Soc. Pac. 115 (2003) 453. D. A. Howell et al. Nature 443 (2006) 308. K. S. Mandel, G. Narayan and R. P. Kirshner, Astrophys. J. 731 (2011) 120. P. Astier et al., Astron. Astrophys. 572 (2014) A80. R. Kessler et al., Publ. Astron. Soc. Pac. 122 (2010) 1415. M. Sako et al., Astrophys. J. 738 (2011) 162. G. Bazin et al., Astron. Astrophys. 534 (2011) A43. S. A. Rodney et al., Astrophys. J. 746 (2012) 5.

May 4, 2017 9:54 b2167-ch17

Chapter 17 Gravitational lensing in cosmology

Toshifumi Futamase Astronomical Institute, Tohoku University, Sendai 9808578, Japan [email protected]

Gravitational lensing is a unique and direct probe of mass in the universe. It depends only on the law of gravity and does not depend on the dynamical state nor the composition of matter. Thus it is used to study the distribution of the dark matter in the lensing object. Combined with the traditional observations such as optical and X-ray, it gives us useful information on the structure formation in the universe. The lensing observables depend also on the global geometry as well as on the large scale structure of the universe. Therefore it is possible to derive useful constraints on cosmological parameters once the distribution of lensing mass is accurately known. Since the ﬁrst discovery of a lensing event by a galaxy in 1979, various kinds of lensing phenomena caused by stars, galaxies, clusters of galaxies and large scale structure have been observed and are used to study the mass distribution at various scales and cosmology. Gravitational lensing is now regarded as an indispensable research tool in observational cosmology. In this article we give a pedagogical introduction to gravitational lensing and its applications to cosmology. Keywords: Cosmology; gravitational lensing; strong lensing; weak lensing. PACS Number: 98.80.Es

1. Introduction and History Propagation of light rays from a distant source to the observer is governed by the gravitational ﬁeld of the intervening mass distribution. This fact was realized even before Einstein formulated the general theory of relativity. It seems that Cavendish and Michell are the ﬁrst who presented the idea that the bending of a light ray occurs near a massive body around 1780 (see Valls-Gabaud2 for the detailed history). Then German astronomer, Soldner, calculated the bending angle of a light ray around a massive object in 1801 using Newtonian gravity and obtained one half of the correct value.3 The correct value can be only calculated by knowing the fact that spacetime is curved by a massive object. General relativity provides a method to calculate the spatial curvature around a massive object. Einstein calculated the bending angle of a light ray grazing the surface of the sun correctly as 1.76 arcsec (Einstein, 1915). In 1919, a team led by A. Eddington succeeded to measure the change in position of stars around the sun during an eclipse and obtained results consistent with the prediction by general relativity. II-173

May 4, 2017 9:54 b2167-ch17

II-174

T. Futamase

Eddington explained the gravitational deﬂection of light by the analogy of lensing eﬀect in both 1911 and 1915 predictions of the bending angle. In 1924 the Russian scientist Chwolson pointed out the possibility of producing fake double stars by gravitational lensing and furthermore without detailed calculation he pointed out a possible existence of perfect ring image if the source is located exactly on the line of sight of the lens.4 In 1936 a Czech amateur scientist, Mandl, visited Einstein at Princeton asking about the possible existence of multiple images of a distant star due to the gravitational force of a star. Stimulated by the discussion, Einstein wrote a paper on gravitational lensing by a star with detailed calculation and showed the existence of the so-called Einstein ring, predicted by Chwolson.5 He concluded however, that gravitational lensing by a star is unlikely to occur because of the extremely small probability of lensing. It should be noted that Link studied similar lensing event before Einstein. He calculated not only the position of the image but also magniﬁcation.6,7 Zwicky estimated the probability of lensing of one galaxy by another galaxy as ∼ 10−3 (Ref. 8) which is much larger than that of a star and tried to discover lensing phenomena in galaxy clusters. The attempt failed mainly because the lensed images are too faint to be detected by the apparatus available at that time. After that there were no active researches on gravitational lensing except some important papers. In 1960 Russian astronomers Idlis and Gridneva wrote a paper which is regarded as the precursor of the weak lensing idea. In 1964 Refsdal9 and Liebes10 argued that the gravitational lensing eﬀect may be used to measure the mass of a lensing object. Refsdal also pointed out that if a source’s luminosity varies in time, there will be a time lag of variability between the lensed images which can be used to estimate the Hubble parameter. The ﬁrst lensing system QSO 0957+561A, B was discovered in early 1979 by Walsh, Carswell, and Weymann.11 They made photometric observations of two images of a single quasar at z = 1.41. The lensing galaxy was also observed at redshift z = 0.355 in a cluster of galaxies. In 1997 the time lag of the variability was observed as 417 ± 3 days between images A and B.12 Since then, other systems with multiple images have been observed, and the list has grown to approximately 200 by now. Time lags have been observed in more than 10 lensed systems as well. This list is expected to be on the order of 100 in the coming decade. In the late 1980s, high resolution observations by the Hubble Space Telescope (HST) became available and discovered a very elongated image called a giant luminous arc (GLA) in the cluster Abell 370 in 1986.13 A multitude of arcs and distorted images in massive clusters have been described by Soucail et al.14 An Einstein ringlike image was also found in the system MG 1131+0456 in 1988.15 Gravitational lensing is classiﬁed into three regimes: strong, weak and micro lensing. Figure 1 shows a schematic representation of strong and weak lensing phenomena. Strong lensing results in multiple images and/or highly deformed arclike images, called giant luminous arcs. It is used not only to measure the spatial dark matter distribution in the lensing object (e.g. Kneib et al.16 ), but also to constrain

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-175

the cosmological parameters. It has been shown that a good reconstruction of a lensing system may constrain the density parameter and the cosmological constant (Futamase and Yoshida, 1999; Yamamoto et al., 2001; Soucail, Kneib and Golse, 2004; Meneghetti et al., 2005; Jullo et al.17 ). An important application of strong lensing is to use it as a natural telescope. Strong lensing magniﬁes small and faint objects to make them observable, objects that are otherwise not observable. In particular, strong lensing by clusters has been used to study distant galaxies (Ellis et al.18 ; Kneib et al.19 ; Richard et al.20 ; Bouwens et al.21 ; Coe et al.22 ). For this study, a highly accurate determination of mass distribution is necessary, and this becomes possible using a large number of multiple-image and arcs discovered by the HST proﬁtting from progress in lensmodeling techniques. It is claimed that the mass in cluster cores is measured to a few percent accuracy (e.g. Bradac et al.23,24 ; Jullo et al.25 ; Jullo and Kneib26 ). In the case of clusters observed by the Hubble Space Telescope Frontier Fields galaxies at z > 10 and as faint as 32 AB magnitude (about magnitudes fainter than the limit of Hubble Ultra Deep Field) are expected to be detected (Richard et al.27 ). Future prospects for this study seem very promising. Another application of strong lensing is lens statistics (Turner, Ostriker and Gott28 ). It has been shown that lens statistics are a good indicator of cosmological parameters, in particular, the cosmological constant (Turner29 ; Fukugita, Futamase and Kasai30 ). Weak lensing results only in weakly deformed images and the lensing signal is obtained only by averaging over background galaxies. While weak lensing is not spectacular compared with strong lensing, it is very important and useful. For example, strong lensing images are only observed in the central part of a cluster, which constitutes only a tiny part of the whole mass distribution of the cluster. Information from weak lensing is necessary to reveal the whole mass distribution of the cluster. The method of mass reconstruction by measuring the weak lensing eﬀect was developed and applied to Abell 1689 and Cl 1409+53 by Tyson et al.31 More sophisticated methods of weak lensing mass reconstruction were developed by Kaiser and Squires32 and Kaiser, Squires and Broadhurst (now known as the KSB method33 ). This method and its reﬁnements have been applied to many clusters of galaxies. Weak lensing by large scale structure known as cosmic shear depends on the growth rate of the structure and thus on the cosmic expansion rate as well as the law of gravity. Thus, the accurate measurement of cosmic shear is expected to constrain the nature of dark energy and the deviation from general relativity. The ﬁrst detection of cosmic shear was made in 2000 by several groups,35–40 and is one of the most active area of research in gravitational lensing. Microlensing is a type of strong lensing by a stellar size object but whose deﬂection angle is on the order of micro-arcsec. Thus no separate images are observed, but only the magniﬁcation of the image is observed. It is used to discover MAssive Compact Halo Objects (MACHOs)41 and planets outside of the solar system. We will not consider microlensing in this review.

May 4, 2017 9:54 b2167-ch17

II-176

T. Futamase

For general review of gravitational lensing, we refer the reader to Schneider, Ehlers and Falco,42 Blandford and Narayan,43 Refsdal and Surdej44 and Narayan and Bartelmann.45 2. Basic Properties for Lens Equation The global geometry and overall evolution of the universe is well described as homogeneous and isotropic by the Friedman–Robertson–Walker (FRW) metric. However, the matter distribution is highly inhomogeneous, up to 100 Mpc as we observe. A light ray propagates through these inhomogeneities feeling the local gravitational potential. How the averaged FRW universe appears from the local inhomogeneous universe within the framework of general relativity is a complicated and diﬃcult problem. Fortunately, the gravitational lensing phenomena considered in this review are caused by weak gravitational ﬁelds and we may regard the ﬁeld as locally stationary on the time scale of the deﬂection of a light ray. This allows us to use a simple treatment of light ray propagation. We will derive the gravitational lensing equation in the cosmological situation starting from the geodesic equation and then summarize the basic concepts and properties of this equation. 2.1. Derivation of the cosmological lens equation To study the propagation of a light ray, we need to specify the background geometry. Since we can assume that the gravitational ﬁeld is weak everywhere and locally quasi-stationary, it is suﬃcient to consider the following metric: (1) ds2 = a2 −(1 + 2Ψ)dη 2 + (1 − 2Ψ) dχ2 + r2 (χ)dΩ2 , where dΩ2 = dθ2 + sin2 θdφ2 is the line element on the 2-dimensional unit sphere t and η = dt /a(t ) is the conformal time. r(χ) is the comoving angular diameter distance deﬁned by

χ −1/2 K sin √ K >0 K K =0 r(χ) = χ (2)

χ (−K)−1/2 sinh √ K < 1. −K The curvature parameter K is expressed by the total density parameter at present Ω0 as K = (Ω0 − 1)H02 . Although the above metric is the ﬁrst-order deviation from the FRW background, it may still apply in the situation with a fully nonlinear matter distribution. This can be understood by noticing that the Newtonian potential is determined by the cosmological Poisson equation. ∇2 Ψ = 4πGa2 ρb δm ,

(3)

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-177

where ρb is the background density and δm is the matter density contrast deﬁned as δm = (ρm − ρb )/ρb . This equation gives

−2 δm 1 |Ψ|, a Ωm,0 LH

(4)

where and LH = H0−1 denote the characteristic comoving scale of the density ﬂuctuation and the Hubble radius, respectively. Thus one can safely assume |Ψ| 1 even if δm 1 as long as LH (Futamase46 ) which is always assumed in lensing studies. The propagation of a light ray is described by a null geodesic equation. dk µ + Γµαβ k α k β = 0, dλ

(5)

gµν k µ k ν = 0.

(6)

One can solve the above equations perturbatively with the 0th order being the solution in the homogeneous and isotropic background k µ (λ) = k (b)µ (λ) + δk µ (λ).

(7)

By choosing the origin of the coordinate at the observer (us) and assuming a small deﬂection angle everywhere, we can set the path of the background solution as follows: k (b)µ = (−1, 1, 0, 0),

(8)

where we choose polar coordinate (t, r, θ, φ) and carry out backward ray-tracing from the observer at λ = 0. The integral curve is chosen as x(b)µ = (−λ, λ, θI , 0),

(9)

where |θI | 1 is interpreted as the image position angle. Since we observe the projection eﬀect on the sky, we are interested in the perpendicular components to the line of sight of the above equation. The perpendicular components of the ﬁrst order solution may be written as δk a (λ) = −

2 2 r (λ)

λ

dλ ∂ a Ψ(λ ),

(10)

0

where a = θ, φ and ∂ a Φ = (Φ,θ , Φ,φ / sin2 θ). The Newtonian calculation gives the same expression with a factor of 1 instead of 2. The diﬀerence is due to the eﬀect of spatial curvature. This may be integrated to give:

θa (λ) = θIa − 2 0

λ

dλ

r(λ − λ) a ∂ Ψ(λ ). r(λ )r(λ)

(11)

May 4, 2017 9:54 b2167-ch17

II-178

T. Futamase

Since lensing phenomena are always observed in a small patch of the sky, we can safely ignore the curvature of the sky (ﬂat sky approximation) and use locally Cartesian coordinates θ = (θ1 , θ2 ) around the line of sight k (b)µ . Deﬁning the source plane coordinate as β = θ(λs ) where λS is the location of the source and the image plane coordinate as θ = θ I , one can write the cosmological lens equation as follows: β = θ − α(θ),

(12)

where the projected bending angle is expressed by the Newtonian potential Ψ as follows:

χs r(λs − λ) (3) dλ α=2 ∇ Ψ(X(λ)), (13) r(λs ) 0 where ∇(3) is the 3-dimensional gradient operator and ∂x = ∂θ /r(λ). This suggests that the bending angle may be expressed by the 2-dimensional gradient ∇θ of a 2-dimensional potential ψ(θ1 , θ2 ) α = ∇θ ψ,

(14)

where ∇θ = (∂θ1 , ∂θ2 ). In particular, if the lensing object lies within a suﬃciently small region between λL ± ∆λ (∆λ λL ) such as clusters and galaxies compared with the relevant cosmological distances, one can assume that the light ray is deﬂected at λ = λL . The projected bending angle takes the form

DLS λL +∆λ α=2 dλ ∇(3) Ψ(x(λ)), (15) DS λL −∆λ where DS = a(λs )r(λs ) and DLS = a(λs )r(λs − λL ) are the angular diameter distances from the observer to the source, and from the lensing object to the source, respectively. This is called the thin lens approximation. Figure 3 shows the conﬁguration of the lensing equation in the thin lens approximation. Using the above expression for α, one ﬁnds 2κ ≡ ∇θ · α = ∆θ ψ =

DL DLS 8πGΣ . DS c2

(16)

We also deﬁned the convergence κ, which is the normalized surface mass density κ(θ) ≡

Σ(θ) , Σcr

where the critical surface mass density is given by

dS c2 D S −1 = 1.16 × 10 h Σcr ≡ g/cm−2 , 4πGDLS DL dL dLS where dX denotes the angular diameter distance DX in units of c/H0 .

(17)

(18)

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

Observer

II-179

Cluster of Galaxies

Background Galaxy

Non-Linear

Multiple Images

Arclets

Optical Path Wave Front

Weak Shear

Multiple Images Area

Linear

Fig. 1.

The illustration of the gravitational lensing mechanism (see Ref. 1).

A region of a mass distribution with Σ > Σcr can produce multiple images for some source positions and is referred to as being super-critical. One can write the bending angle in the following more tractable form using the surface mass density Σ, or the convergence κ: α=

4G DL DLS c2 D S

d2 θ Σ(θ )

θ − θ 1 = |θ − θ |2 π

d2 θ κ(θ )

θ − θ . |θ − θ |2

(19)

2.2. Properties of lens mapping One may regard the lens equation as mapping from the source plane to the image plane. It is not a one-to-one mapping, in general. The Jacobian of the mapping characterizes the local distortion of the mapping: Aij =

∂βi = δij − αij = δij − ψ,ij . ∂θj

(20)

This 2 by 2 symmetric matrix may be decomposed into a trace and traceless part as follows: A(θ) =

1 − κ − γ1

−γ2

−γ2

1 − κ + γ1

= (1 − κ)

1

0

0

1

−

γ1

γ2

γ2

−γ1

,

(21)

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-180

Fig. 2. Giant luminous arc in galaxy cluster Abell 383. Source: http://www.spacetelescope.org/images/heic1106a/.

Fig. 3.

Geometry of the gravitational lensing.

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-181

where γ ≡ γ1 + iγ2 is the gravitational complex shear deﬁned by 1 (ψ,11 − ψ,22 ), 2

(22)

γ2 = |γ| sin(2φ) = ψ,ij .

(23)

γ1 = |γ| cos(2φ) =

In general the quantity a is called a spin-s quantity when a changes as a = eisφ a,

(24)

under the rotation of coordinates by an angle φ. The convergence and gravitational shear are then the spin-0 and spin-2 quantities, respectively. Using this property, it is convenient to deﬁne the tangential and radial components of the shear relative to some reference point θ 0 . Noticing that the tangential component of the shear is obtained from γ1 by rotating the angle π/2 − φ (where φ is the polar angle of θ − θ 0 ), we have the following: γt (θ; θ 0 ) = −Re(γe−2iφ ) = −γ1 cos 2φ + γ2 sin 2φ,

(25)

γ× (θ; θ0 ) = −Im(γe−2iφ ) = γ1 sin 2φ − γ2 cos 2φ,

(26)

where Re(z) and Im(z) denote the real and imaginary part of a complex number z, respectively. γt and γ× are referred to as tangential and cross shears, respectively (Fig. 4). Under the parity transformation, these change as γt → γt , γ× → −γ× . Tangential shear plays an important role in cluster mass reconstruction and cosmic shear, as shown later. The convergence and the shear are related in the following way:

1 (27) γ(θ) = d2 θ D(θ − θ )κ(θ ) π with D(θ) =

2 1 ∂ ∂2 ∂2 θ2 − θ12 − 2iθ1 θ2 e2iφ − + i ln |θ| = 2 =− 2 , 2 2 4 2 ∂θ1 ∂θ2 ∂θ1 ∂θ2 |θ| θ

Fig. 4.

The relationship between γ1 , γ2 and γt , γ× .

(28)

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-182

where we have introduced polar coordinates (θ, φ) such that θ = (θ cos φ, θ sin φ). Note that there exists a global transformation that leaves the shear γ(θ) invariant such that κ(θ) → κ(θ) + κ0 ,

(29)

where κ0 is an arbitrary constant. This degeneracy leads to an ambiguity in the mass reconstruction based only on the image distortion in weak lensing analysis. In practical calculations, it is more convenient to work in Fourier space because one can make use of the Fast Fourier transform. The kernel is expressed as follows in Fourier space: k 2 − k22 + 2ik1 k2 ˆ = πe2iφk , D(k) =π 1 k2

(30)

where k = (k, φk ) is the vector of the polar coordinates in Fourier space. Since D(k)D∗ (k) = π 2 , the inverse of (26) is straightforward: 1 ∗ D (k)ˆ γ (k), (31) π where κ ˆ and γˆ are the Fourier transforms of the corresponding quantities. The geometrical meaning of convergence and shear is easily seen by considering the deformation of a circular source. Let us consider a circle with radius dβ. Consider a small circle deﬁned by the equation κ ˆ (k) =

dβ T dβ = 1,

(32)

where dβ is a position vector from the origin and dβ T is its transverse. Then the image mapped by the lensing satisﬁes the following equation: 1 = dθT AT Adθ.

(33)

The matrix A has the following eigenvalues: λ± = 1 − κ ± |γ| = (1 − κ)(1 ± |g|),

(34)

where g = γ/(1 − κ) is called the reduced shear. The associated eigenvectors are π cos φ − cos φ cos φ 2 V+ = = (35) π , V− = sin φ , − sin φ sin φ − 2 where φ = 12 tan−1 (g2 /g1 ) is the position angle of the eigenvector V− . Using the orthogonal matrix O which diagonalizes the matrix A, one ﬁnds that the unit circle is mapped to the following ellipse (see Fig. 5) dΘ1 2 dΘ2 2 1 + 1 = 1, (36) λ− λ+ where dΘ = Odθ. The dΘ1 is the coordinate along the eigenvector V− .

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-183

Fig. 5. The unit circle is mapped to an ellipse with the length of the major axis 1/|λ− | and the minor axis 1/|λ+ | for kappa < 1 and its inclination from dθ axis is φ.

Since 1/|λ− | > 1/|λ+ | for κ < 1, this is the equation of an ellipse elongated in the direction of φ (tangential direction) with the length of the major axis 1/|λ− | and minor axis |1/λ+ |. For κ > 1 the ellipse is elongated in the direction φ − π/2 (radial direction) with the length of the major axis 1/|λ+ | and the minor axis 1/|λ− |. It is interesting to note that the distortion disappears along the curve deﬁned by κ(θ) = 1, which always lies in the odd-parity region. Moreover in the region deﬁned by κ 1 and γ 0 the mapped image is highly magniﬁed without deformation. These are referred to as GRAavitationally lensed yet Morphologically Regular images, or GRAMORs. Such an image was predicted in 1998 by Williams and Lewis47 and Futamase, Hattori and Hamana48 and was discovered in the galaxy cluster MACS J1149.5+2223 (Broadhurst et al.49 ). Thus the convergence expresses an isotropic magniﬁcation (κ > 0) or demagniﬁcation (κ < 0). The (reduced) shear g expresses a deformation without changing the area. The component g1 expresses the deformation along the θ1 (θ2 ) axis for g1 > 0 (g1 < 0). The component g2 expresses the deformation along the axis rotated 45◦ from the θ1 axis. 2.3. Caustic and critical curves According to Liouville’s theorem, surface brightness is conserved by gravitational lensing. This may be shown as follows. The energy in the interval dν radiated in the solid angle dΩ through the area ds in the time dt may be written as IcdtdsdνdΩ (I is the surface brightness). This may also be written using the photon distribution

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-184

function f (x, p) as IcdtdsdνdΩdν = hνf d3 xd3 p = 4πh3 ν 3 f cdtdsdνdΩ.

(37)

Thus I ∝ ν 3 f . Liouville’s theorem tells us that f = constant and ν = constant. Because there is no energy exchange in gravitational lensing, the surface brightness is conserved. Thus the magniﬁcation of the ﬂux due to lensing is given simply by ratio of the solid angle between image and source. This is given by the inverse of the determinant of the Jacobian matrix A: µ=

1 1 1 = = . 2 2 det A (1 − κ) (1 − |g| ) (1 − κ)2 − |γ|2

(38)

Images with det A > 0 (λ− > 0 or λ+ < 0) have the same parity as the source, while images with det A < 0 (λ+ > 0 and λ− < 0) have parity opposite to the source. The closed curves in the image plane on which the point image is magniﬁed indeﬁnitely are called “critical curves” which is deﬁned by the condition λ± = 0. The corresponding curves in the source plane are called “caustics”. The critical curves separate the image plane into even- and odd-parity regions. In practice, realistic objects like galaxies have a ﬁnite size and det A = 0 applies to every part of the object to be magniﬁed indeﬁnitely. Thus the extended source is not magniﬁed indeﬁnitely on the critical curve, but is highly magniﬁed. Such images are called luminous giant arcs. If the lensing object is suﬃciently massive and satisﬁes κ > 1 in some region, then det A(θ) = 0 is satisﬁed somewhere in the image plane and a critical curve exists. When convergence is a decreasing function of the distance from the center of the lensing object, the curve deﬁned by λ+ = 0 is called the inner critical curve and the curve deﬁned by λ− = 0 is called the outer critical curve. Since κ > 1 on the inner critical curve, the image is elongated in the radial direction (radial arc), and thus the inner critical curve is sometimes called the radial critical curve. On the other hand κ < 1 on the outer critical curve, so the image on the outer critical curve is elongated in the tangential direction (tangential arc) and the outer critical curve is called the tangential critical curve. The conﬁguration of caustics determines the number of images. This is explained by the lens model examples below. The point determined by the condition κ(θ) = 1, namely Σ(θ) = Σcr is the critical point in the absence of shear. Although the shear is non-zero in general, the region where κ(θ) ≥ 1 is referred to as the strong lensing region associated with multiple images, while the region where κ(θ) 1 is referred to as the weak lensing region. 2.3.1. Circular lenses Let us now consider simple examples of a lens model. First, we consider a lens with a spherically symmetric mass distribution. In this case, the bending angle at a point θ is determined by the projected mass M (θ) enclosed by a circle of radius θ = |θ|

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-185

around the center of mass, and is given by α(θ) =

DLS 4GM (θ) θ=κ ¯(θ)θ, D L D S c2 θ 2

(39)

where κ ¯ is the average of κ inside the radius θ: κ ¯ (θ) =

1 M (θ) DLS 4GM (θ) = . 2 Σcr π(DL θ) D L D S c2 θ 2

(40)

The lens equation becomes the scalar equation: β(θ) = θ − κ ¯ (θ)θ.

(41)

This shows immediately that the image with θ ≥ 0 satisﬁes θ ≥ β and an image with θ < 0 (opposite side of the source projected on the image plane) lies in the region with κ ¯ (θ) > 1. The Jacobian matrix is

2 2 1 0 2θ1 θ2 θ d¯ κ κ (θ1 ) − (θ2 ) 1 d¯ A(θ) = 1 − κ ¯− − . (42) 2 dθ 2θ dθ 0 1 2θ1 θ2 −(θ1 )2 + (θ2 )2 Remembering that the mean convergence is a decreasing function of the distance from the center in general and γ = −θ/2d¯ κ/dθ, the eigenvalues are then calculated to be d dβ [θ¯ κ(θ)] = λ+ = 1 − κ + γ = 1 − , (43) dθ dθ β λ− = 1 − κ − γ = 1 − κ ¯(θ) = . (44) θ As explained above, the eigenvalue λ− describes the tangential deformation of the image due to a geometrical eﬀect, and λ+ describes the radial deformation of the image due to the gravitational tidal eﬀect which depends on the radial density proﬁle of the lensing object. In the spherical lens case, the lens mapping may be understood by specifying a curve in the (β, θ) plane (see Fig. 6, lens mapping). The curve is characterized by a mass distribution. Figure 6 shows the mapping given by a point mass where there are only two images. If the lensing mass distribution is non-singular at the origin, the mapping is shown in Fig. 7. As seen in this ﬁgure, the caustic corresponding to the outer critical curve is the origin in the source plane (actually the line in this case). When a source of a ﬁnite size approaches the caustic from the inside, two images with opposite parity approach perpendicularly to the inner critical line and these two images disappear when the source passes across the caustics. The opposite phenomenon occurs when the source approaches the caustic from the outside. Thus, the source must be inside the caustics to make multiple images. This may be more easily seen in Fig. 8, where the critical lines and caustics are shown in the image plane and source plane, respectively. This is a general property for caustics which are locally regarded as a straight line. Such caustics are called “fold caustics”.

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-186

Fig. 6.

Fig. 7.

Lens mapping given by a point mass.

Lens mapping given by a non-singular lens model.

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

Fig. 8.

II-187

Caustics and critical curves for a spherical lens. (For color version, see page II-CP12.) Source: http://gravitationallensing.pbworks.com/w/page/15553257/Strong%20Lensing.

2.3.2. The Einstein radius and radial arcs From symmetry, the tangential caustics are given by β = 0. The image is a ring on the tangential critical curve, provided that the lens conﬁguration is super-critical. Such a ring is known as an Einstein ring and the angular radius is called the Einstein radius. The Einstein radius is obtained by solving the lensing equation by taking β = 0: 1/2 1/2

1/2 4GM (θE ) DLS M (θE ) DLS = 16.5 θE = . (45) c2 DL DS 1014 h−1 M DL DS In actual lensing events, a source such as a galaxy is not perfectly located at β = 0. In that case, we observe a pair of highly magniﬁed, stretched images (called arcs) along the tangential critical curve, with a shorter arc just inside and a longer arc just outside of the critical curve. Thus, the position θarc of the tangential arcs is used to roughly estimate the mass of a lensing object inside the circle traced by the arc as follows:

2 DL DS θE 14 −1 M (θarc ) M (θE ) = 3.31 × 10 h M . (46) DLS 30 We now consider radial arcs which are sometimes observed in galaxy clusters. As shown above, the condition needed to have a radial arc is λ+ = 0. Radial arcs are very useful to study the radial density proﬁle. For example, let suppose that the convergence has a simple power law form as

n θ0 (47) κ(θ) = κ0 θ with 0 < n < 2. The mean convergence is then given by κ ¯ (θ) =

κ(θ) n. 1− 2

(48)

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-188

The condition λ+ = 0, i.e. d[θ¯ κ]/dθ = 0 becomes

θr =

1 κ0 (1 − n)

1/n θ0 .

(49)

Thus the condition n < 1 is required for a lens to have a radial critical curve. As typical examples of a spherical lens, we consider two models. The ﬁrst is the singular isothermal sphere (SIS) model. The 3-dimensional density is given by ρ(r) =

σv2 . 2πGr2

(50)

The convergence and the averaged convergence are given by σ 2 D 1 v LS κ(r) = 2πG , c DS θ σ 2 D 1 v LS κ ¯ (θ) = 4πG . c DS θ

(51) (52)

The next example is the so-called NFW proﬁle, which is a numerically predicted universal proﬁle for CDM halos given by Navarro, Frenk and White (NFW).175 The density is given by ρ(r) =

rcrit δc

2 , r r 1+ rs rs

(53)

where ρcrit , ρc and rc represent the critical density of the universe, the characteristic overdensity of the CDM halo, and the scale radius, respectively. The convergence and the averaged convergence are given by Bartelmann50 and Wright et al.51

θ κ(r) = κs f , (54) θs

−2 θ θ κ ¯ (θ) = 2κs g , (55) θs θs where κs = 2δc ρcrit rs Σ−1 crit and θs = rs /DL . The function f (x) given by 1 2 1−x −1 + arctanh 2 1+x (1 − x2 ) 1 − x f (x) = 1 3 1 2 x − 1 arctan x2 − 1 1 − 2 x+1 (x − 1)

and g(x) are then

x1

(56)

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-189

2 1−x arctanh 2 1+x (1 − x ) x g(x) = ln + 1 2 2 x−1 arctan 2 x +1 (x − 1)

x1

2.3.3. Non-circular lenses There are no exactly circular lenses in nature except a star, which we do not treat here. Any lensing object is somehow non-circular. Usually rather than giving the explicit distribution for the surface mass density or 3-dimensional mass density, the gravitational potential is given. Several interesting models exist for non-circular potentials. As far as an elliptical potential is concerned, the general model contains at least ﬁve parameters (an indicator of the strength of the lens such as the Einstein angle in the circular case, core radius, ellipticity, the power law index of density falloﬀ and the angle of the major axis from a ﬁxed direction). For example, the Tilted Plummer Elliptical potential is given by α2 ψ(x, y) = E η

ωp2 + rep α2E

η/2 ,

(58)

2 = x2 (1 − p ) + y 2 (1 + p ), ωp is related to the core radius, p is the ellipwhere rep ticity and η is the power law index. η = 1 corresponds to the isothermal case. The reader interested in this model should consult the paper by Grogin and Narayan.52 Although one needs a speciﬁc potential for model ﬁtting, general properties of elliptical models are similar for any known model. Figure 9 shows small circular sources in the source plane and the corresponding images in the image plane. In the elliptical case, there exist inner caustic and inner critical curves. The inner caustics have points where the derivative along the curve is not well deﬁned. Such a point is called a cusp. Five images exist when the source is inside the inner caustic. Three images exist when the source is between the inner and outer caustics. As seen in Fig. 9, when a suﬃciently small ﬁnite source approaches the cusp from the inside, three images (two images parallel along and outside the caustics, one perpendicular to and inside the caustics) approach a corresponding point on the critical curve and vanish when the source passes through the cusp. When the source approaches a cusp from the outside, the opposite phenomenon occurs and three images seem to appear from the corresponding point on the inner critical curve. See Blandford and Kochanek53 for a more systematic understanding of caustic and critical curves.

May 4, 2017 9:54 b2167-ch17

II-190

T. Futamase

Fig. 9. The caustics and critical curves of an elliptical lens with suﬃciently small core. The upper and bottom panels are the source plane and image plane, respectively. The sources are circular with same size. The corresponding (same color with source) image positions and shapes are shown. The left part of ﬁgures show the case when the source is close to inner fold caustics. On the other hand, the right part of ﬁgures show the case when the source is close to inner cusp caustics (Hattori et al.54 ). (For color version, see page II-CP12.)

3. Strong Lensing 3.1. Methods of solving the lens equation: LTM and non-LTM In strong gravitational lensing, there are a number of observables such as the relative positions of images, relative ﬂuxes of images and time delays of luminosity changes between lensed images. These observables depend not only on the mass distribution of the source but also on the cosmological parameters. There are essentially two methods for constructing mass models of the source. The ﬁrst is called the parametric method, which assumes a simple but physically reasonable model of potential such as an isothermal sphere, NFW model, elliptical pseudo-isothermal model and so on with parameters that have clear physical meanings. Physically reasonable means that one infers the shape of the lensing potential from the image position as well as the lens position and shape assuming the existence of a strong correlation between light and mass. Thus, this method is now called the light traces mass (LTM) method. In the case of a galaxy lens, the relative position of the images and the existence of arcs gives us enough information to determine the form of the potential.

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-191

The other method is called the non-parametric method, which is also referred to as the “grid-based” method or non-LTM method because one does not assume the potential shape from the beginning. This method has become both possible and popular since deep observations using 8–10 m telescopes and the HST reveal detailed shape information of images for galaxy lenses, and a multitude of arcs and images for cluster lenses. Such a wealth of information allows us to establish the grid by grid correspondence between the image plane and the source plane. The basic principle of this method is the conservation of the surface brightness. Namely, the surface brightness of two grids related by the lens equation is the same. We will not go into further details of these two methods, partly because there are many strong lens model software codes using both LTM and non-LTM methods, which are publicly available. Instead we shall focus on the basic properties of strong gravitational lensing. For more detailed treatment of strong lensing, see an extensive review by Kochanek.55 3.2. Image magnification In strong lensing with multiple images, the ratio of the luminosity between images is an important observable and is used to constrain the mass distribution on the local scale inside the lensing galaxy. To derive the ﬂux ratio between images, it should be noted that gravitational lensing does not change the surface brightness. Therefore, the magniﬁcation of the image compared with the source is just the ratio of the size between the image and source, and thus the magniﬁcation of the image is the inverse of the determinant of the Jacobian. We also deﬁne the magniﬁcation matrix as the inverse matrix of the Jacobian M = A−1 .

(59)

The magniﬁcation of an image is not known because we do not know the luminosity of the source. However since there are multiple images, the ratio of magniﬁcation is observable and is an important constraint of the mass model of the lensing object. Suppose we have two images A and B, then using the magniﬁcation matrix, we have δθB = M (θB )ij δβ = M (θ B )M −1 (θA )δθ A ≡ MBA δθ A .

(60)

The ﬂux ratio is very useful to constrain the lens model. Although a simple smooth lens potential is enough to predict image positions consistent with observation, this is not the case for the ﬂux ratio between images. This may be regarded as an indication of the existence of dark matter substructure in the halo of the lensing galaxy.56,57 3.3. Time delays Another possible observable in strong lensing is time diﬀerence (or time delay) between arrival of light rays in multiple images. (Refsdal9 ). Since the photons from

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-192

the source travel along diﬀerent paths to make the diﬀerent images, the time taken by the photon from the source to the observer diﬀers from image to image, resulting in an observable time delay. Time delays have been observed for a number of systems, and were used as a method to measure the global Hubble parameter before WMAP measured the Hubble parameter to within several percent accuracy. The method is free from any empirical assumptions associated with the method of the distance ladder. However this method was actually not very powerful in the past because of poor time delay measurements, resulting from simpliﬁed and inaccurate assumptions about the lens mass distribution. Nonetheless, this method is potentially very useful if we have enough information to determine the lens mass distribution accurately enough. It has been shown that a single gravitational lens with well-measured time delays and a detailed lens model can be used to measure time delay distances deﬁned as D∆ ≡ (1 + zL )DL DS /DLS (see below) to 6–7% total uncertainty (random and systematic Suyu et al.58 ). In fact, Suyu et al.59 have shown that in quadruple lens systems the lens model inaccuracy can be less important than the inaccuracy in the time delay measurements. Their result is based on detailed modelling of the gravitational lens RXJ1131-1231 with the spatially extended Einstein ring observed by the HST and well-measured time delays between its multiple images. On the order of 100 new quadruple lensed quasars are expected to be discovered in the near future with imaging surveys such as the Subaru Hyper SupremeCam Survey, Pan-STARRS-1,60 the Dark Energy Survey (DES).61 When their time delay data are available, it is expected that together with supernovae and cosmic microwave background information, we can improve the dark energy ﬁgure of merit by almost a factor of 5, and determine the matter density parameter Ωm,0 to 0.004, the Hubble parameter H0 to 0.7%, and the dark energy equation of state time variation parameter wa to ±0.26, systematics permitting (Linder62 ). Here we explain the basic theory of time delays. If a source produces a short burst of light, then the photons of the burst will arrive at a time 1 + zL dL dS 1 2 τ= (θ − β) − ψ(θ) , (61) H0 dLS 2 where dX is the angular diameter distance in unit of c/H0 . The ﬁrst term is just the result of the increase in the path length and the second term is the result of a decrease in the speed of light near the gravitating object. The time delay between the two images A and B is given by ∆AB = τA − τB =

1 + zL dL dS H0 dLS

1 (θA − β)2 − (θB − β)2 − (ψ(θA ) − ψ(θB )) . 2

(62)

For practical application of time delays, we have to take into account the fact that the lens galaxy is sometimes a member of a small group of galaxies or is

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-193

under the inﬂuence of nearby galaxies. In that case, we consider these eﬀects by the external potential in the following form: 1 ψc (θ) ψc (0) + ∇ψc (0) · θ + κc (0) θx2 + θy2 2 1 + γ1c (0) θx2 − θy2 + γ2c (0)θx θy , 2

(63)

where we take the origin at the position of the lens galaxy and assume that the length scale of the external ﬁeld is large compared with the separation of two images and thus neglect the higher order terms. The coeﬃcients are given by κc =

1 (ψc,xx + ψc,yy ), 2

γ1c =

1 (ψc,xx − ψc,yy ), 2

γ2c = ψc,xy .

(64)

It is easy to show that the ﬁrst two terms of the above expansion does not change any observables in the strong lensing system. The third term expresses the contribution from the uniform sheet. It does have any eﬀect on the time delay. We now show that a uniform sheet will bring an ambiguity into the Hubble constant determination. We consider the transformation of the potential c ψ → (1 − c)ψ + (θx2 + θy2 ) 2

(65)

with a constant c. This changes the bending angle as follows: α → (1 − c)α + cθ.

(66)

The lensing equation tells us that θ does not change by changing the source position β → (1 − c)β. This will change the magniﬁcation matrix as M → M/(1 − c), but the relative magniﬁcation matrix MBA = MB MA−1 does not change. However the time delay equation (62) tells that the combination H0 ∆τBA changes as above H0 ∆τBA → (1 − c)H0 ∆τBA .

(67)

This means that the observed time delay does not change if we can change the distance scale at the same time as H0 → (1 − c)H0 .

(68)

Thus even if we determine the lens model by image positions, as well as the ratio of relative magniﬁcation between images and time delay, there remains an ambiguity in the measurement of Hubble parameter H0 = 100h(1 − κc (0)) [km/s/Mpc].

(69)

The ambiguity κc (0) is the projected surface mass density near the lens position. Weak lensing analysis in the region around the lens galaxy is a direct method to determine κc . In fact there have been attempts to measure κc (0) in QSO0957+561 A, B by weak lensing.63,64

May 4, 2017 9:54 b2167-ch17

II-194

T. Futamase

3.4. Comparison of lens model software Analysis of strong gravitational lensing relies on a software analysis of observational data. There have been many lens model software codes available and thus it is important to have a systematic comparison between them. A number of software codes used in research studies are not publicly available. We will review here several of the publicly available codes that have been used in previous studies. A comprehensive review is available in Lefor, Futamase and Akhlaghi.65 3.4.1. Non-light traces mass software These codes were formerly referred to as non-parametric, and include: • PixeLens: This program uses a non-parametric model with Bayesian statistics. Inputs to the program consist of model constants (redshifts, pixel size, etc.) and image data. The radius of the mass map in pixels and the redshifts for the lens and source must be given. The code reconstructs a pixellated mass map by generating large ensembles of models with a Metropolis algorithm (Saha and Williams66 ). • Lensview: Lensview is based on the LensMEM algorithm, which ﬁnds the best ﬁtting lens model and source brightness using a maximum entropy constraint. Lensview is used to study lens inversion, obtaining a model of the source based on lensed images. Due to its comprehensive nature, it is possible to specify very complicated lens models based on one or more components (Wayth and Webster67 ). • LensPerfect: LensPerfect uses a parametric model but is also model-free as described by its developers, who further characterize it as non-LTM. LensPerfect solutions are given as sums of basis functions. While most parametric models are based on a physical object, the basis functions used by LensPerfect have no physical interpretation. Input to the program is via a text ﬁle and graphical output is shown immediately on the display (Coe et al.68 ). • GRALE: GRALE can be used to simulate gravitational lenses and to invert lensing systems. The GRALE algorithm uses a non-parametric technique to infer the mass distribution of a gravitational lens system with multiple strong lensed systems (Liesenborgs et al.69 ). 3.4.2. Light traces mass software These codes were formerly referred to as parametric codes, and include: • Lenstool: Lenstool was ﬁrst described in 1993 and has been continually upgraded. The current version uses a Markov Chain Monte Carlo sampler to avoid local minima in likelihood functions. The software uses a text ﬁle for data input and describes the model using a wide array of parameters. Multiple lens potentials are supported, all at the same redshift. Lenstool is a comprehensive program for gravitational lens modeling (Jullo et al.25 ).

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-195

• Glaﬁc: Glaﬁc allows both point and extended sources in the analysis of lensed images as well as a large catalog of lens potentials. A text ﬁle is used to describe the model and image data as well as priors. Glaﬁc can accurately recover lens model parameters of known lensing systems (Oguri70 ). • Gravlens/Lensmodel: This software has a large catalog of mass models and uses a text ﬁle for commands and data input. The software uses a tiling algorithm to determine the number and position of lensed images associated with a given source, and uses a polar grid centered on the main galaxy (Keeton71 ). This list of software is by no means comprehensive but provides a list of some of the commonly used software for strong gravitational lens models available for public download. All the lens model software described here is available directly from, or using links provided by, the Astrophysical Source Code Library (http://www.ascl.net). 3.5. Lens statistics The statistics of strong lensing events have interesting applications to cosmology. There have been many studies of statistics using quasars, giant luminous arcs and recently GRAMORs. It has attracted much attention because the statistics are sensitive to the cosmological constant via the cosmological volume element (e.g. Turner72 ; Fukugita, Futamase and Kasai30). As shown below, a detailed and accurate knowledge of the properties of the source and lensing object is necessary to use lens statistics as a useful method to measure the cosmological parameters and the redshift evolution of the lensing galaxies (e.g. Kochanek73 ; Ofek et al.74 ; Chae and Mao75 ; Matsumoto and Futamase76 ; Chae77 ; Cao and Zhu82 ). Furthermore, it has been shown that sample incompleteness can bias the results signiﬁcantly (Capelo and Natarajan78 ). Recent large scale galaxy surveys such as SDSS overcame these diﬃculties and made the statistics a useful tool in observational cosmology (see Oguri et al.79 ). Here we explain the basic formulation of quasar statistics as an example of lens statistics. Consider the probability that a quasar with redshift zs is lensed by a lensing object with redshift range zL ∼ zL + dzL . Writing the cross-section as σL , the diﬀerential probability may be written as follows. We have cdt dzL , dτ (zL ) = n(zL )σL (70) dzL where n(zL ) is the number density of lensing objects. The cross-section is given by solving the lens equation once we choose an appropriate lens potential, and depends on the redshifts of the source and the lens as well as the relative position of the source and lens.

L

Φ µ d2 β , (71) σL = µΦ(L) S

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-196

where the integration region S in the source plane is over the position where multiple images are produced. Here we include the magniﬁcation bias. This is necessary because some observed lensed quasars are magniﬁed and become brighter than a limited magnitude. Thus we need a magniﬁcation factor µ for each source position and the quasar luminosity function Φ(L). The cosmological dependence is in the quantity cdt/dzL , which is calculated in FRW geometry to be cdt c 1 = , dzL H0 (1 + zL ) Ωm,0 (1 + zL )3 + (1 − Ωm,0 − ΩΛ,0 )(1 + z)2 + ΩΛ,0

(72)

where Ωm,0 is the matter density parameter and ΩΛ,0 = Λ/3H02 is the normalized cosmological constant. In practical applications, the number density of the lens object is calculated using the luminosity function Φ(L) = dn(L)/L or the velocity function ψ(σv ) = dn(σv )/dσv . For example, using the velocity function the diﬀerential probability of lensing per unit redshift with an image separation between θ and θ + d∆θ is given by d2 τ = d∆θdzL

0

zS

dσv φ(σv )(1 + zL )3

cdt dσL . dzL d∆θ

(73)

A recent study based on the lens sample from the Sloan Digital Sky Survey Quasar Lens Search (SQLS) (Oguri et al.79 ) used 19 lensed quasars selected from 50,836 source quasars where the velocity dispersion of the lensing galaxies suggested by SDSS data (Bernardi et al.80 ) and the quasar luminosity function for 0.4 < z < 2.3 measured by the combined analysis of SDSS and 2dF data (Croom et al.81 ) are used. It is shown that the SQLS sample constrains the cosmologi+0.06 +0.06 cal constant to ΩΛ,0 = 0.790.07 (stat.)0.06 (syst.) for a ﬂat universe. It is also shown that the dark energy equation of state parameter is consistent with w = 1 when the SQLS data is combined with constraints from baryon acoustic oscillation (BAO) measurements or results from the Wilkinson Microwave Anisotropy Probe (WMAP). Furthermore, no redshift evolution of the galaxy velocity function at z ≤ 1 is found. 4. Weak Lensing Although strong lensing is an important and spectacular phenomenon, it is a rare event and most light rays from distant sources propagate through relatively weak gravitational ﬁelds such as present in the outskirts of clusters and regions sourced by the large scale structure. In this case the intervening gravitational ﬁelds only slightly distort the shape of background sources, resulting in a systematic distortion pattern of background source images known as weak gravitational lensing. In the past decade, weak lensing has become a powerful, reliable method to measure the distribution of matter in clusters, dominated by invisible dark matter (DM), without

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-197

assumptions concerning the physical and dynamical state of the system.83–91 Not only clusters, but also weak lensing due to large scale structure, attracts much attention because it carries information regarding the growth of structure which is aﬀected by the nature of dark energy. For a general review of weak gravitational lensing, see Bartelmann and Schneider92 and Umetsu.93,94

4.1. Basic method We shall explain one method of mass reconstruction using weak lensing mass observations in some detail. For clarity of the argument we treat the simplest situation in this subsection. The complexities are mentioned later. First we select a sample of background galaxies for each lensing object considered. Since the lensing signal depends on the redshifts of the sources, it is necessary to have redshift information for accurate determination of the mass distribution of the lensing object. For example, in a weak lensing analysis of galaxy clusters, the contamination by faint member galaxies dilutes the lensing signal, in particular in cluster central regions.95 In the case of cluster lensing, the background galaxies are selected by color information. Namely, their colors are redder than the colormagnitude sequence of cluster member galaxies due to large k-corrections (for more detail, see Medezinski et al.,96 Okabe et al.34 However, more detailed redshift information regarding the background galaxies is necessary for accurate determination of cosmic shear (see below). Once we deﬁne the sample of background objects, the next step is to measure the shape of each background object. Several shape measurement schemes have been developed. We concentrate here on the so-called moment method, which uses multipole moments of brightness distribution to characterize the shape of the background objects (KSB33 and Okura and Futamase97 ). Although this is not the optimal method as seen in Fig. 10 below, it is still the most commonly used method. In the moment method we measure the quadrupole moments of surface brightness for each background object

(obs) (74) Qij = d2 θ∆θi ∆θj I(θ), where I(θ) is the surface brightness distribution and ∆θi = θi − θ0,i with θ0,i the center of the image. Below we deﬁne the center as the point at which the dipole moment vanishes. Then we deﬁne a purely spin-2 quantity from the measured quadrupole moments (obs) (obs) (obs) Q11 − Q22 2Q12 (obs) = , (obs) , (75) obs) (obs) (obs) Q11 + Q22 Q11 + Q22 which is called the ellipticity. For another deﬁnition of ellipticity, see Bartelmann (s) and Schneider.92 Similarly we deﬁne the quadrupole moments Qij and ellipticity

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-198

(s) for each of the corresponding sources which are not observable. Using the lens equation we ﬁnd the relationship between the ellipticity of the observed image and of the intrinsic source (s) =

(obs) − 2g + g 2 (obs) . 1 + |g|2 − 2Re(g(obs) ∗ )

(76)

(s) (obs) − 2g.

(77)

In the weak lensing limit

Since we can expect that the intrinsic ellipticities of nearby galaxies on the sky do not correlate with each other, we can average over a certain number of background galaxies to obtain

σe obs 2g + O √ , (78) N where σe 0.3 − 0.4 is the dispersion of the intrinsic ellipticity of galaxies, and N is the number of background galaxies averaged over. The more galaxies we average over, the more accurate the estimation of the gravitational shear we have. In the best seeing condition at the summit of Mauna Kea we may have N ≥ 50 per arcmin2 . In a realistic situation, the number is less than 30 or so. In the case of cluster lensing, the lensing signal is on the order of 0.1 and thus we can safely detect the lensing signal by clusters. Once we have the averaged shear ﬁeld we can convert it to a convergence map which is the normalized surface mass density using the relationship between the convergence and the shear. Although this is the essence of weak lensing mass reconstruction, things are much more complicated for real observations.

Fig. 10. Stellar ellipticity distributions before and after PSF anisotropy correction. The left panel shows the observed ellipticity components (e∗1 , e∗2 ) before PSF correction, and the right panel shows the residual ellipticity components (δe∗1 , δe∗2 ) after PSF anisotropy correction (Umetsu and Broadhurst101 ).

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-199

4.1.1. Shape measurements In practical observations of weak lensing, the observed image is not the lensed image but rather the image smeared by various sources of noise such as the distortion by atmospheric turbulence, imperfect optics, pixelization by the CCD chip, and photon noise. Thus we have to correct these eﬀects before applying a mass reconstruction method. It is supposed that these eﬀects are expressed by the point spread function (PSF) P (θ) as follows:

(obs) (θ) = d2 ψI(ψ)P (θ − ψ). (79) I In order to improve the accuracy of weak lensing analysis, we need to reduce various sources of noise and to correct the PSF eﬀects without introducing other errors. In this review we focus our attention on PSF correction in some detail because this has the most important eﬀect in weak lensing analysis. There are several methods for PSF correction available today. Here we employ the moment-based approach (see Kaiser, Squires and Broadhurst33; hereafter KSB). In realistic observations, since the observed shapes of background galaxies are rather noisy, particularly for small faint galaxies, we need to introduce a weight function W to reduce the random noise in the outer part of the image, so that

θ (obs) , (80) Qij = d2 θθi θj I (obs) (θ)(θ)W σ2 where σ is a typical scale for the galaxy. The simple choice is the spherical Gaussian shape:

|θ|2 θ (81) W ∝ e− σ 2 . σ It is known that this choice introduces a systematic error in measuring the shape that overestimates the shear for images with large ellipticities. The error is overcome by introducing an elliptical window function (see Okura and Futamase98–100 ). The PSF correction proceeds in three steps. First we decompose the PSF function into isotropic and anisotropic part as follows:

(82) P (θ) = d2 θ q(θ )P (iso) (θ − θ ), where q, P (iso) are normalized and have vanishing 1-order moments.

dθq(θ) = dθP (iso) (θ) = 1,

(83)

dθθi q(θ) =

dθθi P (iso) (θ) = 0.

(84)

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-200

(i) Anisotropic PSF correction We deﬁne

d2 θ I(θ )P (iso) (θ − θ ).

(85)

Then the observed intensity may be written as follows:

(obs) (θ) = dθ q(θ − θ )I (iso) (θ ). I

(86)

I (iso) (θ) =

Assuming a nearly spherical shape of the anisotropic PSF function, we only consider the quadrupole moment of the anisotropic PSF function

(87) qij = d2 θq(θ)qi qj . Then the quadrupole moment Q(iso) calculated from I (iso) is related to the observed quadrupole moment as follows: 1 Q(iso) = Q(obs) − Zijk qk , 2 where q1 = q11 − q22 , q2 = 2q12 and

Zijk = d2 θI (obs) (θ)

2 |θ| ∂2 θi θj W . ∂θi ∂θj σ2

(88)

(89)

Using the deﬁnition of ellipticity and the above expression, we ﬁnd the ellipticity corrected by the anisotropic PSF eﬀect is (iso) = e(obs) − Pijsm qj ,

(90)

where P sm is called the smear polarizability which expresses the response of the ellipticity against the anisotropic PSF eﬀect (a detailed expression may be found in the original paper by KSB and the review by Bartelmann and Schneider). From dimensional arguments P sm ∝ rg−2 (where rg is the Gaussian scale of the object), and thus the PSF anisotropic eﬀect becomes important for objects with small S/N . The above correction requires a knowledge of qi which is given by applying the equation to stars. Since stars do not suﬀer lensing and their images have no ellipticity after isotropic smearing, e(∗,iso) = 0 and thus we have ∗,(obs)

qi∗ = (P ∗,sm )−1 ij ej

.

(91)

The quantities with an asterisk denote quantities associated with stars. To obtain the smooth map of qi used in Eq. (97), the co-added mosaic image is divided into small chunks with scales determined by the typical coherent scale of PSF anisotropy patterns. By doing this, PSF anisotropy in individual chunks is well described by fairly low-order polynomials. Typically the rms value of stellar ellipticities is reduced from a few % to (4 − 8) × 10−3 after anisotropic PSF correction.

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-201

Figure 10 shows the distribution of stellar ellipticity before and after the PSF anisotropy correction using the Subaru data for the cluster Abell 1689 (Umetsu and Broadhurst101). (ii) Isotropic PSF correction Next, we have to correct for magniﬁcation due to the PSF eﬀect. For this purpose ˆ we consider the intensity I(β) in the source plane corresponding to I (iso) (θ). If we ˜ then we can write write the lensing equation as β = Aθ, ˆ Aθ). ˜ I (iso) (θ) = I(

(92)

The inverse relationship may be written by introducing the PSF in the source plane and may be written as

ˆ (93) I(β) = d2 β I s (β )Pˆ (β − β ), where we have introduced the PSF in the source plane corresponding to the isotropic part of the PSF in the image plane, so that Pˆ (β) =

1 P (iso) (A˜−1 θ). det A˜

(94)

This function is normalized and has a zero mean. Note that Pˆ does have an anisotropic part from the lensing eﬀect. Thus we decompose into isotropic and anisotropic parts, so that

q (β − β ). (95) Pˆ (β) = d2 β Pˆ (iso) (β )ˆ Then the same argument in the image plane applies here to give eˆ0i = eˆi − Pijsm qˆj , where eˆ0 is the ellipticity of the intensity Iˆ0 deﬁned by

ˆ I(β) = d2 β Iˆ0 qˆ(β − β ).

(96)

(97)

We have used I (obs) instead of Iˆ in P sm as before. The ﬁnal step is to relate the artiﬁcial ellipticity eˆ0 to the isotropically smeared ellipticity. This is simply obtained by using the relationship between I (iso) and ˆ which is deﬁned above. Using this relationship, the quadrupole moment of the I, artiﬁcial source Iˆ is calculated to be

2

|β| 2 ˆ ˆ Qij = d ββi βj I(β)W (98) σ ˆ2

2

|θ| − δm ηm (θ) 2 (iso) ˜ ˜ ˜ (θ)W = (det A)Aik Aj d θθk θ I , (99) σ2

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-202

where σ ˆ 2 = (1 − κ)2 (1 + |g|2 )σ 2 and δi = 2gi /(1 + |g|2 ). Under the assumption that qˆ is small, we can replace I (iso) in the observed image I (obs) , and then by expanding the window function in the ﬁrst order of δi , we obtain (iso)

ei

= eˆi + Pijsh gj ,

(100)

where P sh is called shear polarizability and expresses the response of the ellipticity against the shear. The detailed expression may be found in the original paper which describes KSB. Combining these two relationships, we have (iso)

ei

= eˆ0i + Pijsh gj + Pijsm qˆj .

(101)

By using star images, we have ∗,sh qˆi = −(P ∗,sm )−1 ik Pkj gj .

e

(102)

Finally, by using the relationship between e(iso) and the observed ellipticity , we ﬁnally have

(obs)

eˆ0 = e(obs) − Pijsm qj − Pijg gj ,

(103)

∗,sh sm (P ∗,sm )−1 Pijg = Pijsh − Pik k P j .

(104)

where we have deﬁned

Since the intrinsic ellipticity of background galaxies do not have a correlation, averaging gives us gi = (P g )−1 ij (ej

(obs)

− Pijsm qj ) ,

(105)

∗,(obs)

with qi = (P ∗,sm )−1 . ij ej As seen from the above argument, one has to make some assumptions such as a small anisotropic part q and the use of the observed image I (obs) in the evaluation of smear polarizability and shear polarizability. These assumptions caused systematic errors in the original KSB weak lensing analysis. There have been some proposals to improve the accuracy of the analysis. Erben et al. have proposed using the trace part in P g to reduce the noise which creates non-diagonal components in P g . g Ps,ij =

1 T r(P g )δij ≡ Psg δij . 2

(106)

(Hudson et al.102 ; Hoekstra et al.103 ; Erben et al.104 ; Hetterscheidt et al.105 ). The diagonal parts are measured for individual galaxies, but they are very noisy for small and faint galaxies. To avoid the noisy part, an averaging in object-parameter space was introduced by Van Waerbeke et al.35 ; see also Erben et al.104 ; Hamana et al.106

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-203

Although in the KSB method the result in principle should not depend on the choice of the scale in the weight function, this is not the case for real observations. Hoekstra et al.103 have shown that the scale used in the weight function of galaxies gives a better result. In practice, the averaged shear over the observed region is estimated by introducing the weight function ug,I such that ug,I =

1 2 + α2 , σg,I

(107)

where I labels the object in the region, the constant α is the softening parameter and σg is the rms error of the complex distortion measurement. Usually α σg2 0.4 is used, which is a typical value of the mean rms σ ¯g over the background sample. Then the averaged shear is given by ωg (θ − θI )ug,I gi,I gi (θ) = I , (108) I ωg (θ − θ I )uI where gi,I is the reduced shear estimated of the√Ith galaxy at an angular position θI , and ωg (θ) ∝ exp(−θ2 /θg2 ) with θg = FWHM/ 4 ln 2 which pixelizes the distortion data into a regular grid of pixels. The error variance for the above smoothed shear is given as 2 2 ωg,I u2g,I σi,I 2 (θ) = I σg

2 ,

(109)

ωg,I uI

I 2 δij δIJ has been where ωg,I = ωg (θ − θ I ) and the relationship gi,I gj,J = (1/2)σg,I used. The smoothing scale is chosen to optimize the weak lensing detection of target mass structures, depending on both the size of the structure and the strength of noise power (∝ σ ¯g2 /ng ). Once the smoothed shear ﬁeld is thus obtained, it can be converted to the convergence ﬁeld using the relation κ(k) = πD∗ (k)γ(k). Actually this inversion method has to be applied in an inﬁnite space. For a ﬁnite space one need to use the ﬁnite-ﬁeld solution of the inversion problem for the reconstruction kernel. Such method has been developed by Seitz and Schneider.107 For a suﬃciently wide ﬁeld where the data ﬁeld is not dominated by a positively or negatively biased density ﬁeld, both methods give the almost same convergence ﬁeld. This is not the case for a nearby cluster for which we have to use a ﬁnite ﬁeld method.

4.2. E/B decomposition Observations and data analysis suﬀer from errors and noise. In particular, weak lensing analysis needs many corrections such as for the PSF distortion to isolate

May 4, 2017 9:54 b2167-ch17

II-204

T. Futamase

a physical signal of a weak tidal gravitational ﬁeld. If the corrections are not perfect, they may systematically change the signal. Thus it is critically important to characterize the eﬀect of systematic errors on the ﬁnal results. Here we consider a method for the evaluation of systematic errors called E/B decomposition. This method is based on the fact that gravitational shear is generated by a Newtonian potential in the lowest order, and thus the gravitational ﬁeld is rotation-free. Let us ﬁrst deﬁne a new 2D vector ﬁeld u from the reconstructed convergence κ by γ1,1 + γ2,2 . (110) u ≡ ∇κ = γ2,1 − γ1,2 Using the vector ﬁeld, we deﬁne the E-mode κ and B-mode κ as follows: ∇2 κE ≡ ∇ · u,

(111)

∇2 κB ≡ ∇ × u = u2,1 − u1,2 ,

(112)

which gives both E-mode and B-mode potentials ∇2 φE,B = 2κE,B .

(113)

Note that the gravitational eﬀect is given by the Newtonian potential Ψ in the lowest order so that φE = Ψ and φB = 0, but in practical observations, φE = Ψ and φB = 0. The complex shear may be expressed by the complex potential φ = φE +iφB as follows: 1 E 1 B B E B γ= + + i φ − φ − φ φ,11 − φE (114) φ ,22 ,12 ,12 ,22 . 2 2 ,11 This expression shows that the transformation γ (θ) = iγ(θ) is equivalent to an interchange of the E- and B-modes of the original maps. By noticing that the shear transforms as γ = γe2iφ under the rotation of an angle φ, this operation is a rotation of each ellipticity by π/4 with each potential being ﬁxed. Figure 11 shows the E and B mode shear. There are several sources of B-mode shear. One artiﬁcial source is PSF residuals. Since the PSF correction so far proposed adopts some sort of approximation, it gives rise to B-mode shear. In addition to artiﬁcial B-modes, there are also physical B-modes. Intrinsic ellipticities of background galaxies contribute to the shear estimate. By assuming randomness of orientation of the intrinsic ellipticities, such uncorrelated ellipticities yield statistically identical contributions to the E- and B-modes. By increasing the number density of background galaxies, these modes may be reduced. Another physical B-mode is due to the existence of intrinsic alignments,

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-205

E mode

B mode

Fig. 11.

Shear patterns of E- and B-modes (Van Waerbeke and Mellier108 ).

which may be generated by tidal forces between halos of galaxies. There are two kinds of intrinsic alignments. One is called II correlation. Galaxies around a large halo may have some correlation of the orientation of ellipticity by a tidal gravitational force. This eﬀect creates both E- and B-modes. It is reasonable to assume that the eﬀect only applies to nearby galaxies, and thus the eﬀect is eliminated by excluding nearby galaxy pairs from the background sample.109 Another eﬀect is called GI correlation, which is the correlation between the intrinsic ellipticities and large scale structure.110 This is because the ellipticites are inﬂuenced by a local density ﬁeld and the same density ﬁeld causes weak lensing for high-z galaxies (cosmic shear). Thus the B-mode signals serve as a useful null check for systematic eﬀects. Finally we highlight an ambiguity in weak lensing analyses. As pointed out above, the relationship between the convergence and the shear is not unique. Any constant mass sheet does not change the shear ﬁeld which is obtained by the weak lensing limit. This is called the mass-sheet degeneracy. Actually as shown above, the observable is not shear itself but the reduced shear γ(θ) . (115) g(θ) = 1 − κ(θ) This ﬁeld is invariant under the following global transformation: κ(θ) → λκ(θ) + 1 − λ,

γ(θ) → λγ(θ),

(116)

where λ = 0 is an arbitrary constant. There is a method to break this degeneracy once one realizes that the above transformation is equivalent to the following scaling

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-206

for the Jacobian matrix: A(θ) → λA(θ).

(117)

One can exploit magniﬁcation eﬀects since the inverse of the determinant of the Jacobian is the magniﬁcation. 4.3. Magnification bias Recent advances in weak lensing observations made the measurement of the magniﬁcation bias possible by measuring the depletion of the surface density of background objects. There are two eﬀects as a result of lensing magniﬁcation. One is the expansion of the area of sky. The other is the ampliﬁcation of the apparent magnitude which is given by mobs = m + 2.5 log µ,

(118)

where m is the apparent magnitude without lensing. These eﬀects change the magnitude-limited number density n(< mlim ) of the background objects, so that n(< mlim ) =

1 n0 (< mlim + 2.5 log µ), µ

(119)

where mlim is the limiting magnitude and n0 is the unlensed number density count per solid angler. Let us suppose that n0 near the limited magnitude obeys a power law, α=

d log n0 (< m) , dm

(120)

where α is the power law index around m = mrmlim . Then we have n(< mlim ) = n0 (< mlim )µ2.5α−1 .

(121)

Since µ > 1, n(< mlim ) < n0 (< mlim ) if α < 0.4. The ratio is b=

n < mlim ) = µ2.5α−1 1 + (5α − 2), n0 (< mlim )

(122)

where we have used the weak lensing limit in the ﬁnal step. For red background galaxies at a median redshift ∼ 1, the intrinsic power law index at faint magnitudes is observed to be relatively ﬂat α 0.1 and thus causes depletion of the number count. This depletion has been observed in several massive clusters (Broadhurst et al.95 ; Umetsu and Broadhurst101; Umetsu et al.111 ). 4.3.1. Simulation test Although many improvements in weak lensing analysis have taken place, systematic errors in shape measurement must be reduced by a factor of 5–10 by the time

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-207

the next generation large scale galaxy surveys are conducted. It is important to test the accuracy of shear measurement methods. This is done by using simulated data with realistic complicated galaxies and PSF eﬀects such as the GREAT08 and GREAT10 challenges (Bridle et al.112,113 ; Kitching et al.114,116,119 ) and STEP (Heymans et al.117 ; Massey et al.118 ). The accuracy of these methods is measured by the two parameters m and c, deﬁned by γ obs − γ true = mγ true + c,

(123)

where m is called the calibration bias and c is called PSF residual. Figure 12 shows the results using various methods of weak lensing analysis.119 4.3.2. Higher-order weak lensing-flexion and HOLICs In the usual treatment of weak lensing, the quadrupole moments of the shape are measured to evaluate the ellipticity which are spin-2 quantities. However, the quadrupole moments are not enough to characterize the shape of galaxies in general. In fact, there are many background galaxies, in particular small faint galaxies with shapes not well described by simple ellipses. It would be very hard to describe general complicated shapes mathematically, but if the shape deviates slightly from an ellipse, higher-order moments such as the octopole moments may be useful. Remembering that the spin-2 combination is constructed by a combination of the quadrupole moments, the quantities with deﬁnite values of spin should be constructed by appropriate combinations of higher order moments (Okura, Umetsu and Futamase121 ). Such quantities are called Higher Order Lensing Image’s Characteristics (HOLICs). Here we concentrate on the octoHOLICs constructed from octopole moments.

Fig. 12. Left: The multiplicative m and additive c biases of various measurement methods are shown. The symbols indicate the method (shown in the right panel). The central ﬁgure expands the x- and y-axes to show the best performing methods. The ﬁgures are cited from Kitching et al.119 The detailed description and references of the methods are also found there.

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-208

The following quantities are the octo-HOLICs: ζ≡

(Q111 + Q122 ) + i(Q112 + Q222 ) , ξ

(124)

δ≡

(Q111 − 3Q122 ) + i(3Q112 − Q222 ) , ξ

(125)

where Qijk and Qijk are the octopole and 16th pole moments, respectively and ξ is the normalization factor deﬁned as ξ ≡ Q1111 + 2Q1122 + Q2222 .

(126)

The ﬁrst HOLICs ζ has spin-1, and the second HOLICs δ has spin-3. These HOLICs have dimensions of [angle]−1 . As the spin-2 ellipticity is related to gravitational shear which is the second derivative of the lens potential, the above HOLICs are related to the third derivative of the lens potential which is called ﬂexion (Goldberg and Bacon, 2005; Bacon et al., 2006). Higher-order shape change is calculated as 1 dβi = Aij dθj + Dijk dθj dθk + O(θ3 ), 2

(127)

where Dijk = Aij,k = −ψ,ijk (Bacon, Rowe & Taylo, 2005). Using this equation, we can show a relationship between the intrinsic HOLICs and the observed HOLICs. By linearizing the relationship we have ζ (s) ζ −

9 F , 41−κ

(128)

δ (s) δ −

3 G , 41−κ

(129)

F = ∂κ,

G = ∂γ,

(130)

where F and G are deﬁned as

where ∂ = ∂1 + i∂2 is the complex gradient operator. Thus F has spin-1 and is called the ﬁrst ﬂexion. G has spin-3 and is called the second ﬂexion. By averaging over the background galaxies, we know that the HOLICs are an estimator of the ﬂexion. So far, the spin-1 HOLICs ζ has been measured in Abell 1689 and used for mass reconstruction (Leonard et al.122 ; Okura, Umetsu and Futamase121 ; Leonard et al.123 ; Cain, Schechter and Bautz124 ). For recent development and other approach for Flexion analysis, see, for example, Viola, Melchior and Bartelmann,125 Rowe et al.126 and Levinson.127 4.4. Cluster mass reconstruction Since weak lensing analysis cannot resolve mass structure smaller than a ﬁnite scale on the order of 1 arcmin, it is applied mainly to clusters of galaxies and to the large

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-209

scale structure of the universe. Clusters of galaxy are the largest self-gravitating objects in the universe, and thus their statistical properties contain indispensable informations on the initial density ﬂuctuations and their nonlinear growth. The most important such information will be their mass distribution. On the other hand, a standard structure formation scenario based on CDM and a cosmological constant predicts that larger structures are the result of assembly of smaller structures and that the spherically averaged CDM halo is well described by the so-called NFW proﬁle.175 Weak lensing thus plays an important role to study cluster properties. In this subsection, we explain frequently used techniques for measuring the cluster mass proﬁle and then give two examples of weak lensing observations. For a more comprehensive treatment of cluster weak lensing, see the reviews by Umetsu93 and by Kneib and Natarajan.176 It is convenient to write the critical mass density in the form Σcrit =

c2 β −1 , 4πGD

β(z , zs ) ≡

D s (zs ) , Ds (zs )

(131)

where β(zs ) is a function of the source redshift zs with ﬁxed z and represents the geometrical strength of lensing for a source with redshift zs , and thus is called the lensing depth. If the source redshift distribution is given by N (z), the mean lensing depth is

β ≡

∞

0

dzN (z)β(z) ∞

.

(132)

dzN (z)

0

It is convenient to use the normalized mean lensing depth deﬁned by177 ω=

β . β(zs → ∞)

(133)

The convergence and shear may then be written in terms of the corresponding quantities for a ﬁducial source at inﬁnite redshift as κ(θ) = ωκ∞ (θ) and γ(θ) = ωγ∞ (θ). In this way, the redshift information of background galaxies is taken into account. Actually the observable is the reduced shear g = γ/(1 − κ) which is nonlinear in κ. The averaging by source redshift is complicated operation. The detailed treatment of the averaging is given by Seitz and Schneider177 and by Hoekstra et al.178 As explained above, the tangential shear g+ is a coordinate independent quantity for a given reference point. For cluster lensing, the reference point is naturally the center of the cluster, which can be determined from the symmetry of the strong lensing pattern, the X-ray centroid position, or the position of brightest cD galaxy. It is also convenient that the cross-shear g× may be used as an estimator of the systematic error since g× is the divergence-free, curl-type distortion pattern

May 4, 2017 9:54 b2167-ch17

II-210

T. Futamase

of background images. The reason to use tangential shear for measuring the mass proﬁle is that the azimuthal average of the tangential shear allows us to directly compare with the theoretical model by using the relationship ¯ (< θ, θ0 ) − κ (θ, θ 0 ), γt (θ; θ 0 ) = κ

(134)

where ·· is the azimuthal average and κ ¯ (< θ, θ0 ) is the average of the convergence within the radius θ. In the following, we understand that θ is the radius from the origin θ 0 and will not write the origin. The above relationship is obtained by the deﬁnition of κ ¯ in the following form:

θ

θ ∂φ ¯ (< θ) = 2 dθ θ κ(θ) = θ2 κ (135) dψ , 2π ∂θ 0 where we have used the fact that 2κ = ∆φ, and the 2D version of Gauss’s theorem. By taking the derivative with respect to θ and using the relationship ∂2φ = κ − γt , ∂θ2

(136)

dθ2 κ ¯ (< θ) = θ¯ κ(< θ) + θ[ κ (θ) − γt (θ)]. dθ

(137)

we have

This is the required relationship by noting d(θ2 κ ¯ /)dθ = 2 κ . It is also useful to deﬁne the following quantity, over the annulus bounded by θ and θout located outside the mass to be measured: ¯ (< θ) − κ ¯ (θ < θ < θout ), ζ(θ, θout ) ≡ κ

(138)

where θ and θout are the inner and outer radii of the annulus centered on some position θ0 usually taken at the position of the brightest galaxy in the cluster. This is referred to as ζ-statistics (Fahlman et al.128 ). Using the above equations, we can derive

θout 2 ζ(θ, θout ) = d ln θ γt (θ) . (139) θ2 θ 1− 2 θout Thus ζ is directly obtained by observing the image distortion of background galaxies in the weak lensing regime where |κ|, |γ| 1. Since κ ¯ (θ < θ < θout ) is positive deﬁnite, κ ¯ (< θ) ≥ ζ(θ, θout ). Thus we can deﬁne the lowest projected mass contained within the radius θ as Mζ (< θ) = π(Dd θ)2 Σcr ζ(θ, θout ).

(140)

By ﬁxing the outer radius θout suﬃciently large (but not so large to contain neighboring clusters), ζ(θ, θout ) as a function of θ is called the radial proﬁle. This proﬁle does not depend upon the invariance transformation. We can ﬁt the observed radial proﬁle with some theoretical models.

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-211

Using the technique explained above one can observe the mass distribution of galaxy clusters. The structure formation scenario based on CDM makes particular predictions for the averaged density proﬁle of the main halo and statistical properties of dark matter subhalos. Thus the measurement of the mass distribution by weak lensing will give us a direct test of the CDM structure formation scenario. There are several on-going projects aiming at an accurate measurement of the cluster mass proﬁle.129–133 Here we give two examples of weak lensing observations of clusters performed using the Subaru telescope. If many lensed images with redshift information are observed in the central region of a cluster, a very accurate determination of the mass distribution in the central region is possible. Also the magniﬁcation bias is observed in some massive clusters. One can then combine all information to acquire an accurate model independent determination of mass distribution over a wide range of cluster radii. The CRASH survey179 is designed for this purpose, combining high-resolution 16-band HST imaging with wide-ﬁeld Subaru observations for a sample of 20 X-ray selected relaxed clusters (Zitrin et al.180 ; Coe et al.181 ; Umetsu et al. 2012). The Bayesian approach to combine this information in a joint mass-proﬁle analysis has been proposed by Umetsu et al.182,183 For another approach to combine the strong lensing and weak lensing, see Merten et al.184 4.4.1. Density profile It has been known that the density proﬁle of a collapsed object is well described by the NFW proﬁle regardless of scale, as the result of numerical simulations.175 Here we show the result of our measurement of the mean density proﬁle by stacking the weak lensing signals of approximately 50 mass-selected clusters in the redshift range 0.15 < z < 0.3 (Okabe et al.34 ). These are selected from the ROSAT All Sky Survey catalogs (Ebeling et al.134,135 ; Boehringer et al.136 ) that satisfy LX [0.1 − 2.4 keV]/E(z)2.7 ≥ 4.2 × 1044 ergs−1 , 0.15 ≤ z ≤ 0.30, nH < 7 × 1020 cm−2 , and −25◦ < δ < +65◦, where E(z) = H(z)/H0 is the normalized Hubble parameter. As mentioned above, the selection of background galaxies is important for an accurate mass reconstruction. The sample of background galaxies basically consists of red galaxies satisfying the condition ∆C ≡ (V − i ) − (V − i )ES0 > 0 for each cluster. We made a positive color cut ∆C > 0.475 to eliminate contamination by faint red cluster members due to statistical errors and possible intrinsic scatter in galaxy color (for details, see Okabe et al.34 ). This results in just 1% contamination of the sample by foreground and cluster members. Each individual cluster is detected at a typical peak signal-to-noise ratio (S/N ) 4 using the 2D Kaiser–Squires mass reconstructions. The signal is stacked by the shear catalogs in physical length units centered on the respective brightest cluster galaxies (BCGs) and the average cluster mass distribution for the full sample, with a peak S/N of 28. The stacked lensing mass measurement is less sensitive to cluster internal structures and 3-dimensional halo orientation.

May 4, 2017 9:54 b2167-ch17

II-212

T. Futamase

γ Fig. 13. Stacked tangential shear proﬁle (∆Σ+ Σcr 1−κ ) of 50 clusters in units of projected mass density, where diﬀerent cluster and background redshift galaxies are weighted by the lensing kernel. The projected radius is computed from the weighted mean cluster redshift. The solid, dashed, dotted and dashed-dotted curves are the best-ﬁt NFW, SIS, gNFW, and Einasto proﬁles, respectively. The lower portion shows the results of the 45◦ test for systematic errors (Okabe et al.34 ). (For color version, see page II-CP12.)

Thus it enables us to measure an averaged cluster mass proﬁle for the sample. When the signals from 50 clusters are combined, the number density of background galaxies becomes 266.3 arcmin−2 . The B-mode signal is smaller by at least an order of magnitude than the E-mode signal. Figure 13 shows the stacked shear proﬁle over the radial range 100 h−1 kpc < r < 2.8 h−1 Mpac. This shows that the proﬁle is well described by the predicted NFW proﬁle with parameters +0.53 +0.49 × 1014 h−1 M and cvir = 5.41−0.45 with less than 10% statistical Mvir = 7.19−0.50 error. The concentration parameter obtained from stacked lensing proﬁle is broadly in line with theoretical predictions (e.g. Bhattacharya et al.137 ). No evidence of departures from the NFW proﬁle is found. 4.4.2. Dark matter subhalos in the coma cluster According to the CDM structure formation scenario, less massive halos are accreted into more massive halos, which are then subsequently eroded by eﬀects combined with tidal stripping and dynamical friction of the host halo, eventually becoming a smooth component. In this process the central regions of subhalos have survived under the overdensity ﬁeld, and constitute their population. Numerical simulations of, and analytic approaches to CDM based structure formation scenario predict that subhalo mass functions at the intermediate and low mass scales follow a power

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-213

−α law, dn/dlnMsub ∝ Msub with α ∼ 0.9−1.0 (e.g. Taylor and Babul138–140 ; Oguri 141 and Lee ; van den Bosch et al.142 ; Diemand et al.143 ; De Lucia et al.144 ; Gao et al.;145 Shaw et al.146 ; Angulo et al.147 ; Giocoli et al.;148 Klypin et al.149 ; Gao et al.150 ; Wu et al.151 ). Thus the statistical properties of cluster subhalos such as mass function and spatial distribution, provide us with useful information about the mass assembly history. Their measurement is the most stringent test of CDM predictions on scales of less than several Mpc. A characteristic feature of the subhalo mass function is also critically important to constrain the nature of dark matter, because it depends on the particle mass of dark matter. Furthermore, measurement of the correlation between member galaxies and subhalo masses sheds important insight on the physics of galaxy evolution associated with dark matter. Thus it is important to measure the mass function directly from observations without assuming a relationship between dark matter and luminous matter and the dynamical state of the system. Weak lensing observations of very nearby clusters (z ≤ 0.05) are ideal for this purpose because their large apparent size enables us to easily resolve less massive subhalos inside the clusters and provides a correspondingly large number of background galaxies, which leads to low statistical errors and compensates for the low lensing eﬃciency to achieve a high S/N. We show the results of a 4.1 deg2 weak gravitational lensing survey of subhalos in the very nearby Coma cluster at redshift z = 0.0236 using the Subaru/Suprime-Cam (Okabe et al., 2012). The observed area is about 80% within r200 inside of which the mean density is 200 times the critical density at the cluster redshift. Note that one arcmin corresponds to 20 h−1 kpc for the cosmology with Ωm,0 = 0.27, ΩΛ,0 = 0.73 and H0 = 100 hkms−1Mpc−1 . Figure 14 shows the projected mass distribution with a smoothing scale of FWHM= 4 . and units of signiﬁcance of ν = κ/σκ . As shown in this ﬁgure, 32 subhalos were detected and their masses are measured in a model-independent manner, down to the order of 103 of the virial mass of the cluster. All of them are associated with a small group of cluster members. The mean distortion proﬁles stacked over subhalos show a sharply truncated feature well-ﬁtted by a Navarro– Frenk–White (NFW) mass model with the truncation radius, as expected due to tidal destruction by the main cluster. It is found that subhalo masses, truncation radii, and mass-to-light ratios decrease toward the cluster center. Figure 15 shows the subhalo mass function dn/dlnMsub constructed by weak lensing mass measurements. It covers a range of two orders of magnitude in mass and is well described by a single power law

dn −α ∝ Msub , d ln Msub +0.42 . with α = 1.09−0.32

(141)

May 4, 2017 9:54 b2167-ch17

II-214

T. Futamase

Fig. 14. Projected mass distribution with a smoothing scale of FWHM= 4 and units of significance of ν = κ/σκ . The shear is used without taking into account the LSS lensing eﬀect. The contours of signiﬁcance start at 1σ with a step value of 1σ (Okabe et al.158 ). (For color version, see page II-CP13.)

This agrees with the predicted value of ∼ 0.9–1.0 from an N-body simulation based on the CDM scenario. 4.5. Cosmic shear The most challenging observation in weak lensing is cosmic shear, which is weak lensing due to the large scale structure of the universe. It attracted much attention because its signal depends on the expansion rate of the universe and the growth rate of structure. It therefore carries useful information about dark energy and the theory of gravity. Cosmic shear has already been detected by several groups.35–38 Subsequent detections have put useful constraints on the matter density parameter Ωm,0 and on the matter density spectrum normalization σ8 (Bacon et al.152 ; Hoekstra, Yee and Gladders153 ; Bacon et al.154 ; Hamana et al.106 ). More recent results are found, for example, in Fu et al.155 for the CFHTLS third data release; Schrabback et al.156

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-215

Fig. 15. Subhalo mass function spanning two orders of magnitude of subhalo masses. The red solid and blue dotted lines are the best-ﬁt power-law model and Schechter function, respectively. The best-ﬁt powers are in remarkable agreement with CDM predictions. Green dashed lines are the mass function for spurious peaks. The thick and thin dashed lines are the best-ﬁt and the 68% C.L. uncertainty, respectively (Okabe et al.158 ). (For color version, see page II-CP13.)

-5

-1.0 10

Fig.E-mode, are 16. Theand variance black open of thepoints aperture are mass B-mode. measured The signal by the in CFHTLS full angular Wide. scaleRed is shown ﬁlled points in the enlargement in the panel. It shows a very small B-mode on very large scales (120 –230 ).155

May 4, 2017 9:54 b2167-ch17

II-216

T. Futamase

for the Hubble Space Telescope COSMOS survey; Benjamin et al.157 and Kilbinger et al.159 for CFHTLens; and Jee et al.160 for the Deep Lens Survey. In particular, the Canada–France–Hawaii Telescope Lensing Survey (CFHTLenS) covers 154 deg2 in ﬁve optical bands where the shapes of 4.2 million galaxies with 0.2 < z < 1.3 (photometric redshift) are measured. Combining with WMAP7, BOSS and an HST distance-ladder prior on the Hubble parameter, it is found that Ωm,0 = 0.283±0.010 and σ8 = 0.813 ± 0.04 for a ﬂat ΛCDM model. On-going and planned galaxy surveys aiming to observe cosmic shear including the Panoramic Survey Telescope & Rapid Response System (Pan-STARRS),60 the Subaru Hyper Suprime Cam (HSC) Survey,161 the Dark Energy Survey (DES),162 the Large Synoptic Survey Telescope (LSST),163 the KIlo-Degree Survey (KIDS)164 and Euclid.165 However the signal is extremely small and contains many systematic errors such as inaccurate shape measurement, and inaccurate determination of redshifts. Also, we do not yet have a complete theoretical understanding of small scale clustering. Thus it is essential to reduce these systematic errors down to a required level in order to place a useful constraint on the cosmological information (Huterer et al.166 ; Cropper et al.167 ; Massey et al.168 ). This is an ongoing and actively developing research ﬁeld. Here we give an introduction to understand the physics of cosmic shear. Cosmic shear observations measure the coherent distortion of background galaxies. This is related to the convergence of the matter density ﬁeld and thus the matter power spectrum. In fact, there is a relationship between the shear and convergence: ˆ γ (k)ˆ γ (k) = ˆ κ(k)ˆ κ(k ) = (2π)2 C(k)δD (k − k ),

(142)

where the quantities with a “hat” are Fourier transforms of the corresponding quantities, and C(k) is the lensing power spectrum of the shear (convergence). As easily imagined, the power spectrum is related to the matter power spectrum, and thus the observation of the cosmic shear provides a method to measure the evolution of the matter power spectrum. The details are as follows. First, the cosmological lens equation may be written as the equation for the convergence ﬁeld by the matter distribution at the comoving distance λ(z) as follows:

3Ωm H02 λH (λ − λ )λ dχ δ(x , λ), (143) κ(θ, λ) = 2 a(λ )λ 0 where δ is the matter density contrast and the Limber approximation has been used. We have neglected the contribution from integration along the line of sight. As seen from the above equation, the dependence of the cosmological parameters appears in two parts, in the comoving distance and in the evolution of the density contrast. The latter depends not only on the global geometry of the universe but also on the local gravitational force, and thus cosmic shear is in principle able to distinguish between dark energy and an alternative theory of gravity as the origin of the accelerated expansion.

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-217

The power spectrum for shear (convergence) may be written in terms of the matter power spectrum Pδ as follows:

9H04 Ωm λH g 2 (λ) k C(k) = dλ 2 Pδ ;χ . (144) 4 a (λ) λ 0 This is the desired relationship between the power spectrum of shear and the matter power spectrum. Here g(χ) is deﬁned as

λH λ − λ dλ p(λ) g(λ) = , (145) λ 0 with p(χ) being the observed probability distribution of galaxies for the survey. A precise knowledge of p(χ) is only obtained by spectroscopic observation, which is very time consuming for a large survey. Assuming a close correlation between the dark matter density contrast δm (z) and the number density of galaxies ng (z), the sum of a huge number of galaxies makes the convergence observable. Although large and deep redshift surveys are planned in the future, a photometric determination of the redshift (photo-z method) is used for existing surveys and those planned in the near future. One of the errors in the observation of cosmic shear arises from the error in the photo-z method. 4.5.1. How to measure the cosmic density field In practice, the following two point correlation functions are observed: ξ± (θ) = γt γt (θ) ± γ× γ× (θ),

ξ× (θ) = γt γ× (θ).

(146)

Since γt → γt nad γ× → −γ× under the parity transformation, ξ× should vanish. These correlation functions may be expressed in terms of the power spectrum. For this purpose, it is convenient to choose the coordinate as θ = (θ, 0) and then γ1 = −γt , γ2 = −γ× . Thus

d ξ+ = γ(0)γ(θ) = J0 (θ)C(). (147) 2π Similarly

ξ− =

d J4 (θ)C(), 2π

(148)

where Jn (x) is the nth Bessel function. However, the power spectrum is expressed in terms of these two point functions

(149) C() = 2π dθθξ+ (θ)J0 (θ) = 2π dθθξ− (θ)J4 (θ), where we use the orthogonality of the Bessel functions

∞ 1 dxxJn (ux)Jn (vx) = δD (u − v). u 0

(150)

May 4, 2017 9:54 b2167-ch17

T. Futamase

II-218

In practice, the averaged tangential shear is used. For this purpose, ﬁrst we deﬁne the window function U , which satisﬁes the property

(151) dθθU (θ; θr ) = 0, where θr characterizes the radius of the averaging. The ﬁltered convergence is then deﬁned as follows:

(152) F (θr , ; θ0 ) ≡ d2 θ U (|θ − θ 0 |; θr )κ(θ ), where θ 0 is the center of the region averaged over. One can easily conﬁrm that the above may be written as the integral of the tangential shear

(153) F (θr ; θ 0 ) ≡ d2 θ Q(|θ − θ0 |; θr )γt (θ ), where

θ 2 dθ θ U (θ ; θr ) − U (θ; θr ). θ2 0 Consequently, the ensemble average is given by

2

2 dθθU (θ, θr )J0 (θ) . F (θr ; θ 0 ) = 2π dC()

Q(θ; θr ) =

(154)

(155)

One can obtain information regarding the power spectrum (which depends on the cosmological parameters) by observing the dispersion of the aperture mass which is constructed by measuring the tangential shear. There are several choices for the window function. Here we use one proposed by Schneider et al. (2002):

θ2 1 θ2 9 1 − − θ < θr 2 θr2 3 θr2 U (θ) = πθr , (156) 0 θ > θr

2 2 6 θ 1− θ θ < θr 2 2 θr2 Q(θ) = πθr θr . (157) 0 θ > θr Consequently, the ensemble average is called the aperture mass and is written as follows:

2 (158) Map (θr ) = 2π dC()I(θr ) with

I(θr ) =

2 12 J (θ ) . 4 r π(θr )2

(159)

The functional form of the kernel I(θr ) for the aperture mass shows that it plays a role of picking up modes θr ∼ 5 in the power spectrum, and is considered a good statistical quantity to evaluate the power spectrum.

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

II-219

Since the power spectrum may be expressed by the correlation functions ξ±

1 2θr θdθ θ θ 2 Map (θr ) = T+ ξ+ (θ) + T− (160) ξ− (θ) , 2 0 θr2 θr θr where

∞

T+ (x) = 576 0

∞

T− (x) = 576 0

we can also deﬁne

dt J0 (xt) [J4 (t)]2 , t3

(161)

dt 2 J4 (xt) [J4 (t)] , t3

(162)

M× =

d2 θQ(θ)γ× (θ).

(163)

2 This quantity vanishes identically as long as the convergence is real. Thus M×

may be used for the evaluation of noise:

1 ∞ θdθ θ θ 2 M× (θr ) = T+ ξ+ (θ) − T− (164) ξ− (θ) . 2 0 θr2 θr θr 2 In an ideal situation, M×

should vanish identically because ξ× = 0. This does not happen in practical observations due to the existence of noise. Actually it may 2

is the be shown by using the E/B decomposition of the convergence that M× B-mode part of the aperture mass. However the observed region is ﬁnite and the number density of background galaxies is ﬁnite in practical observations. Thus we cannot take a correlation √ between two galaxies separated more than some θmax and less than θmin ∼ 1/ ng . Thus E- and B-mode decomposition is not perfect and there exists leakage from the E-mode to the B-mode on small scales if θr ≤ 2θmax . To avoid such leakage it is possible to choose a window function to deﬁne a ﬁltered convergence that it can be calculated using only the two point correlation function with a ﬁnite separation.170 In practical measurements of the two-point correlation function, we need to construct an unbiased estimator and a covariance matrix in order to compare observables with theory. We ask the reader to consult relevant references such as Schneider et al.171 and Semboloni et al.172 for detail.

5. Conclusion and Future At 100 years after its birth, general relativity has become an indispensable tool for understanding our universe. Gravitational lensing is not the only example, but is certainly one of the most actively investigated examples, which was brought about by rapid advances in observational apparatus and techniques since the 1980s. This tendency will continue in the future with ongoing and planned weak lensing surveys such as the Subaru HSC survey, Pan-STARR, DES, KIDS and Euclid. The role of gravitational lensing will become increasingly important in observational cosmology.

May 4, 2017 9:54 b2167-ch17

II-220

T. Futamase

In this review, we have provided a pedagogical introduction to the basics of gravitational lensing and highlighted some of the recent achievements. The research area where gravitational lensing plays a role is very wide. In fact, any cosmological observation suﬀers from the eﬀect of gravitational lensing to some extent. For example, the magnitude–redshift (m–z) relationship in Type Ia Supernovae (SNIa) was used to conﬁrm the acceleration of the universe173,174 and it is expected that future supernovae surveys will determine the nature of dark energy. Since the m–z relation is inﬂuenced by the lensing by large scale structure, this program is possible only with an accurate estimate of the lensing eﬀect. Lensing eﬀects by large scale structure in cosmological observables have been regarded as noise from an observational point of view, but their measurement may turn out to yield important information on structure formation and cosmology if we have an accurate theoretical understanding of its origin. Acknowledgments I would like to thank my collaborators, Alan Kawarai Lefor, Nobuhiro Okabe, Yuki Okura, Masahiro Takada, and Keiichi Umetsu. Parts of this review are based on the work done with them. Unfortunately I am only able to cover a part of this broad research ﬁeld in this review, and there are many other interesting applications, such as CMB lensing. We ask the reader to consult other reviews on gravitational lensing for a more comprehensive treatment. I thank Masayo Morioka for preparing the ﬁgures in this review and Alan Kawarai Lefor for improving the English usage. The author would like to thank Prof. Wei-Tou Ni for the opportunity to contribute to the Centennial General Relativity book. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

J.-P. Kneib, Ph.D. Thesis (1993). D. Valls-Gabaud, Albert Einstein Century Int. Conf. 861 (2006) 1163. J. Soldner, Astronomishes Jahrbuch fur 1804 (1801) 161. O. Chwolson, Astronomische Nachrichten 221 (1924) 329. A. Einstein, Science 84 (1936) 506. F. Link, Comptes Rendus Acad. Sci. 202 (1936) 917. F. Link, Bull. Astron. 10 (1937) 73. F. Zwicky, PR 51 (1937) 290. S. Refsdal, Mon. Not. R. Astron. Sco. 128 (1964) 307. S. Liebes, Phys. Rev. 133 (1964) 835. D. Walsh, R. F. Carswell and R. J. Weymann, Nature 279 (1979) 381. T. Kundi´c et al., Astrophys. J. 482 (1997) 75. R. Lynds and V. Petrosian, Bull. Am. Astron. Soc. 18 (1986) 1014. G. Soucail, B. Fort, Y. Mellier and J. P. Picat, AA 172 (1987) L14. J. N. Hewitt, E. L. Turner, D. P. Schneider, B. F. Burke and G. I. Langston, Nature 333 (1988) 537. 16. J.-P. Kneib, R. S. Ellis, I. Smail, W. J. Couch and R. M. Sharples, Astrophys. J. 471 (1996) 643.

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61.

II-221

E. Jullo et al., Science 329 (2010) 924. R. S. Ellis, arXiv:0109294. J.-P. Kneib, R. S. Ellis, M. R. Santos and J. Richard, Astrophys. J. 607 (2004) 697. J. Richard et al., Astrophys. J. 685 (2008) 705. R. J. Bouwens et al., Astrophys. J. 690 (2009) 1764. D. Coe et al., Astrophys. J. 762 (2013) 32. M. Bradaˇc et al., Astrophys. J. 652 (2006) 937. M. Bradaˇc et al., Astrophys. J. 706 (2009) 1201. E. Jullo et al., New J. Phys. 9 (2007) 447. E. Jullo and J.-P. Kneib, Mon. Not. R. Astron. Soc. 395 (2009) 1319. J. Richard et al., arXiv:1405.3303. E. L. Turner, J. P. Ostriker and J. R. Gott, Astrophys. J. 284 (1984) 1. E. Turner, Astrophys. J. 365 (1990) L43. M. Fukugita, T. Futamase and M. Kasai, Mon. Not. R. Astron. Soc. 246 (1990) 24P. J. A. Tyson, R. A. Wenk and F. Valdes, ApJL 349 (1990) L1. N. Kaiser and G. Squires, Astrophys. J. 404 (1993) 441. N. Kaiser, G. Squires and T. Broadhurst, Astrophys. J. 449 (1995) 460. N. Okabe, G. P. Smith, K. Umetsu, M. Takada and T. Futamase, ApJL 769 (2013) L35. L. van Waerbeke et al., A&A 358 (2000) 30. D. J. Bacon, A. R. Refregier and R. S. Ellis, Mon. Not. R. Astron. Soc. 318 (2000) 625. N. Kaiser, G. Wilson and G. A. Luppino, arXiv:astro-ph/0003338. R. Maoli et al., A&A 368 (2001) 766. L. Van Waerbeke et al., A&A 393 (2002) 369. D. Wittman et al., Nature 405 (2000) 143. B. Paczynski et al., Astrophys. J. 435 (1994) L113. P. Schneider, J. Ehlers and E. E. Falco, Gravitational Lenses, Vol. XIV (SpringerVerlag, Berlin, 1992). R. D. Blandford and R. Narayan, 30 (1992) 311. S. Refsdal and J. Surdej, Rep. Prog. Phys. 57 (1994) 117. R. Narayan and M. Bartelmann, arXiv:astro-ph/9606001. T. Futamase, Mon. Not. R. Astron. Soc. 237 (1989) 187. L. L. R. Williams and G. F. Lewis, Mon. Not. R. Astron. Soc. 294 (1998) 299. T. Futamase, M. Hattori and T. Hamana, ApJL 508 (1998) L47. A. Zitrin and T. Broadhurst, ApJL 703 (2009) L132. M. Bartelmann, AA 303 (1995) 643. C. O. Wright and T. G. Brainerd, arXiv:9908213. N. A. Grogin and R. Narayan, Astrophys. J. 464 (1996) 92. R. D. Blandford and C. S. Kochanek, Dark Matter in the Universe, J. Bahcall, T. Piran and S. Weinberg, (eds.) (World Scientific, Springer), p. 133. M. Hattori, J. Kneib and N. Makino, Progress Theor. Phys. Suppl. 133 (1999) 1. C. S. Kochanek, in Proc. 33rd Saas-Fee Advanced Course, G. Meylan, P. Jetzer and North P. (eds.) (Springer-Verlag, Berlin). S. Mao and P. Schneider, Mon. Not. R. Astron. Soc. 295 (1998) 587. M. Chiba, Astrophys. J. 565 (2002) 17. S. H. Suyu et al., Astrophys. J. 766 (2013) 70. S. H. Suyu et al., ApJL 788 (2014) L35. N. Kaiser, W. Burgett and K. Chambers et al., Proc. SPIE 7733 (2010) 77330. M. Oguri and P. J. Marshall, MNRAS 405 (2010) 2579.

May 4, 2017 9:54 b2167-ch17

II-222

T. Futamase

62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88.

E. V. Linder, Phys. Rev. D 84 (2011) 123529. H. Dahle, S. J. Maddox and P. B. Lilje, Astrophys. J. 435 (1994) L79. P. Fisher, G. Bernstein, G. Rhee and J. A. Tyson, AJ 113 (1997) 521. A. T. Lefor, T. Futamase and M. Akhlaghi, NewAR 57 (2013) 1. P. Saha and L. L. R. Williams, AJ 127 (2004) 2604. R. B. Wayth and R. L. Webster, Mon. Not. R. Astron. Soc. 372 (2006) 1187. D. Coe et al., Astrophys. J. 681 (2008) 814. J. Liesenborgs et al., Mon. Not. R. Astron. Soc. 386 (2008) 307. M. Oguri, PASJ 62 (2010) 1017. C. R. Keeton, arXiv:astro-ph/0102340. E. L. Turner, ApJL 365 (1990) L43. C. S. Kochanek, Astrophys. J. 466 (1996) 638. E. O. Ofek, H.-W. Rix and D. Maoz, Mon. Not. R. Astron. Soc. 343 (2003) 639. K.-H. Chae and S. Mao, ApJL 599 (2003) L61. A. Matsumoto and T. Futamase, Mon. Not. R. Astron. Soc. 384 (2008) 843. K.-H. Chae, Mon. Not. R. Astron. Soc. 402 (2010) 2031. P. R. Capelo and P. Natarajan, New J. Phys. 9 (2007) 445. M. Oguri et al., AJ 143 (2012) 120. M. Bernardi et al., Mon. Not. R. Astron. Soc. 404 (2010) 2087. S. M. Croom et al., Mon. Not. R. Astron. Soc. 399 (2009) 1755. S. Cao and Z.-H. Zhu, A&A 538 (2012) A43. J. Tyson, R. A. Wenk and F. Valdes, Astrophys. J. 349 (1990) L1. N. Kaiser and G. Squires, Astrophys. J. 404 (1993) 441. P. Schneider and C. Seitz, AA 294 (1995) 411. T. Brainerd, R. D. Blandford and I. Smail, Astrophys. J. 466 (1996) 623. M. J. Huudson, S. D. Gwyn, H. Dahle and N. Kaiser, Astrophys. J. 503 (1998) 531. D. Clowe, G. A. Luppino, N. Kaiser, J. P. Henry and I. M. Gioia, Astrophys. J. 498 (1998) L61. H. Hoekstra, M. Franx, K. Kuijken and P. G. van Dokkum, Mon. Not. R. Astron. Soc. 333 (2002) 911. H. Dahle et al., ApJS 139 (2002) 313. E. S. Cypriano, L. Sordre, J.-P. Kneib and L. E. Campusano, Astrophys. J. 613 (2004) 95. M. Bartelmann and P. Schneider, Phys. Rep. 340 (2001) 291. K. Umetsu, arXiv:1002.3952. K. Umetsu, M. Tada and T. Futamase, PTP Suppl. 133 (1999) 53. T. Broadhurst et al., Astrophys. J. 619 (2005) L143. E. Medezinski et al., Astrophys. J. 663 (2007) 717. Y. Okura and T. Futamase, Astrophys. J. 699 (2009) 143. Y. Okura and T. Futamase, Astrophys. J. 730 (2011) 9. Y. Okura and T. Futamase, Astrophys. J. 748 (2012) 112. Y. Okura and T. Futamase, Astrophys. J. 771 (2012) 37. K. Umetsu and T. Broadhurst, Astrophys. J. 684 (2008) 177. M. J. Hudson, S. D. J. Gwyn, H. Dahle and N. Kaiser, Astrophys. J. 503 (1998) 531. H. Hoekstra, M. Franx, K. Kuijken and G. Squires, Astrophys. J. 504 (1998) 636. T. Erben, L. van Waerbeke, E. Bertin, Y. Mellier and P. Schneider, A&A 366 (2001) 717. M. Hetterscheidt et al., A&A 468 (2007) 859. T. Hamana et al., Astrophys. J. 597 (2003) 98.

89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106.

May 4, 2017 9:54 b2167-ch17

Gravitational lensing in cosmology

107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135. 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149.

II-223

S. Seitz and P. Schneider, AA 305 (1998) 383. L. van Waerbeke and Y. Mellier, arXiv:astro-ph/0305089. M. Takada and M. White, Astrophys. J. 601 (2004) L1. C. M. Hirata and U. Seljak, PRD 70 (2004) 063526. K. Umetsu et al., arXiv:1404.1375U. S. Bridle et al., Ann. Appl. Stat. 3 (2009) 6. S. Bridle et al., Mon. Not. R. Astron. Soc. 405 (2010) 2044. T. D. Kitching, A. F. Heavens and L. Miller, Mon. Not. R. Astron. Soc. 413 (2011) 2923. T. D. Kitching et al., Mon. Not. R. Astron. Soc. 423 (2012) 3163. T. D. Kitching et al., ApJS 205 (2013) 12. C. Heymans et al., Mon. Not. R. Astron. Soc. 368 (2006) 1323. R. Massey et al., Mon. Not. R. Astron. Soc. 376 (2007) 13. T. D. Kitching et al., Mon. Not. R. Astron. Soc. 423 (2012) 3163. Y. Okura, K. Umetsu and T. Futamase, Astrophys. J. 660 (2007) 995. Y. Okura, K. Umetsu and T. Futamase, Astrophys. J. 680 (2008) 1. A. Leonard, D. M. Goldberg, J. L. Haaga and R. Massey, Astrophys. J. 666 (2007) 51. A. Leonard, L. J. King and D. M. Goldberg, Mon. Not. R. Astron. Soc. 413 (2011) 789. B. Cain, P. L. Schechter and M. W. Bautz, Astrophys. J. 736 (2011) 43. M. Viola, P. Melchior and M. Bartelmann, arXiv:1107.3920. B. Rowe et al., arXiv:1211.0966. R. S. Levinson, Publ. Astron. Soc. Jpn. 125 (2013) 1474. G. Fahlman, N. Kaiser, G. Squires and D. Woods, Astrophys. J. 437 (1994) 56. D. E. Applegate et al., Mon. Not. R. Astron. Soc. 439 (2014) 48. H. Hoekstra et al., Mon. Not. R. Astron. Soc. 427 (2012) 1298. M. Oguri et al., Mon. Not. R. Astron. Soc. 420 (2012) 3214. N. Okabe and K. Umetsu, Publ. Astron. Soc. Jpn. 60 (2008) 345. A. B. Newman et al., Astrophys. J. 765 (2013) 24. H. Ebeling et al., Mon. Not. R. Astron. Soc. 301 (1998) 881. H. Ebeling et al., VizieR Online Data Catalog 730 (2000) 10881. H. B¨ ohringer et al., Astron. Astrophys. 425 (2004) 367. A. Bhattacharya et al., Astrophys. J. 766 (2013) 32. J. E. Taylor and A. Babul, Mon. Not. R. Astron. Soc. 348 (2004) 811. J. E. Taylor and A. Babul, Mon. Not. R. Astron. Soc. 364 (2005) 515. J. E. Taylor and A. Babul, Mon. Not. R. Astron. Soc. 364 (2005) 535. M. Oguri and J. Lee, Mon. Not. R. Astron. Soc. 355 (2004) 120. F. C. van den Bosch, X. Yang and H. J. Mo, Mon. Not. R. Astron. Soc. 340 (2003) 771. J. Diemand, B. Moore and J. Stadel, Mon. Not. R. Astron. Soc. 352 (2004) 535. G. De Lucia et al., Mon. Not. R. Astron. Soc. 348 (2004) 333. L. Gao, S. D. M. White, A. Jenkins, F. Stoehr and V. Springel, Mon. Not. R. Astron. Soc. 355 (2004) 819. L. D. Shaw, J. Weller, J. P. Ostriker and P. Bode, Astrophys. J. 646 (2006) 815. R. E. Angulo, C. G. Lacey, C. M. Baugh and C. S. Frenk, Mon. Not. R. Astron. Soc. 399 (2009) 983. C. Giocoli, G. Tormen, R. K. Sheth and F. C. van den Bosch, Mon. Not. R. Astron. Soc. 404 (2010) 502. A. A. Klypin, S. Trujillo-Gomez and J. Primack, Astrophys. J. 740 (2011) 102.

May 4, 2017 9:54 b2167-ch17

II-224

T. Futamase

150. L. Gao et al., Mon. Not. R. Astron. Soc. 425 (2012) 2169. 151. H.-Y. Wu, O. Hahn, R. H. Wechsler, P. S. Behroozi and Y.-Y. Mao, Astrophys. J. 767 (2013) 23. 152. D. J. Bacon, A. Refregier, D. Clowe and R. S. Ellis, Mon. Not. R. Astron. Soc. 325 (2001) 1065. 153. H. Hoekstra, H. K. C. Yee and M. D. Gladders, Astrophys. J. 577 (2002) 595. 154. D. J. Bacon, R. J. Massey, A. R. Refregier and R. S. Ellis, Mon. Not. R. Astron. Soc. 344 (2003) 673. 155. L. Fu et al., Astron. Astrophys. 479 (2008) 9. 156. T. Schrabback et al., Dark Matter Astropart. Part. Phys., Proceedings of the 6th International Heidelberg Conference, ed. H. V. Klapdor-Kleingrothaus of Geraint F. Lewis (World Scientific Publishing Co) (2008), pp. 260–273. 157. J. Benjamin et al., Mon. Not. R. Astron. Soc. 431 (2013) 1547. 158. N. Okabe, T. Futamase, M. Kajisawa and R. Kuroshima, Astrophys. J. 784 (2014) 90. 159. M. Kilbinger et al., Mon. Not. R. Astron. Soc. 430 (2013) 2200. 160. M. J. Jee et al., Astrophys. J. 765 (2013) 74. 161. The HSC collab., http://member.ip,u.jp/masahiro.takada/proposal rv.pdf. 162. D. L. DePoy et al., SPIE 7014 (2008) 70140E. 163. J. A. Tyson, D. M. Wittman, J. F. Hennawi and D. N. Spergel, in Proc. 5th Int. UCLA Symp. (2002). 164. G. Verdoes Kleijn et al., arXiv:1112.0886. 165. Euclid Science Study, arXiv:0912.0914. 166. D. Huterer, M. Takada, G. Bernstein and B. Jain, Mon. Not. R. Astron. Soc. 366 (2006) 101. 167. M. Cropper et al., Mon. Not. R. Astron. Soc. 431 (2013) 3103. 168. R. Massey et al., Mon. Not. R. Astron. Soc. 429 (2013) 661. 169. J. A. Peacock and S. J. Dodds, Mon. Not. R. Astron. Soc. 280 (1996) L19. 170. P. Schneider and M. Kilbinger, Astron. Astrophys. 462 (2007) 841. 171. P. Schneider, L. van Waerbeke, M. Kilbinger and Y. Mellier, Acta Astronom. 396 (2002) 1. 172. E. Semboloni et al., Mon. Not. R. Astron. Soc. 375 (2007) L6. 173. M. Schmidt et al., Astrophys. J. 507 (1998) 46. 174. S. Perlmutter et al., Nature 391 (1998) 51. 175. J. F. Navarro, C. S. Frenk and S. D. White, Astrophys. J. 490 (1997) 493. 176. J.-P. Kneib and P. Natarajan, Astron. Astrophys. Rev. 19 (2011) 47. 177. C. Seitz and P. Schneider, Astron. Astrophys. 318 (1997) 687. 178. H. Hoekstra, M. Franx and K. Kuijken, Astrophys. J. 532 (2000) 88. 179. M. Postman et al., ApJCS 199 (2012) 25. 180. A. Zitrin et al., Astrophys. J. 742 (2011) 117. 181. D. Coe et al., Astrophys. J. 757 (2012) 22. 182. K. Umetsu, Astrophys. J. 795 (2014) 163. 183. K. Umetsu, T. Broadhurst, A. Zitrin, E. Medezinski and L. Hsu, Astrophys. J. 729 (2011) 127. 184. J. Merten, M. Cacciato, M. Meneghetti, C. Mignone and M. Bastelmann, Astron. Astrophys. 500 (2009) 681.

May 4, 2017 9:54 b2167-ch18

Chapter 18 Inflationary cosmology: First 30+ years

Katsuhiko Sato National Institutes of Natural Sciences, 2F Hulic Kamiyacho Building, 4-3-13 Toranomon, Minato-ku, Tokyo 105-0001, Japan Department of Physics, School of Science and Engineering, Meisei University, 2-1-1 Hodokubo, Hino-shi, Tokyo 191-8506, Japan Kavli Institute for the Physics and Mathematics of the Universe (Kavli IPMU), UTIAS, WPI, The University of Tokyo, Kashiwanoha, Kashiwa, Chiba 277-8568, Japan [email protected] Jun’ichi Yokoyama Research Center for the Early Universe (RESCEU), Department of Physics, School of Science, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan Kavli Institute for the Physics and Mathematics of the Universe (Kavli IPMU), UTIAS, WPI, The University of Tokyo, Kashiwanoha, Kashiwa, Chiba 277-8568, Japan [email protected]

Starting with an account of historical developments in Japan and Russia, we review inflationary cosmology and its basic predictions in a pedagogical manner. We also introduce the generalized G-inflation model, in terms of which all the known single-field inflation models may be described. This formalism allows us to analyze and compare the many inflationary models that have been proposed simultaneously and within a common framework. Finally, current observational constraints on inflation are reviewed, with particular emphasis on the sensitivity of the inferred constraints to the choice of datasets used. Keywords: Cosmology; inflation; cosmic microwave background.

1. Introduction The classical Big Bang Cosmology1 was the ﬁrst theory of the universe starting from physical principles that succeeded in interpreting in a uniﬁed manner the three fundamental cosmological observations of the time: The homogeneous cosmic expansion, originally discovered by Hubble2 and Lemaˆıtre3; the cosmic microwave II-225

May 4, 2017 9:54 b2167-ch18

II-226

K. Sato and J. Yokoyama

background radiation (CMB) discovered by Penzias and Wilson4 ; and the abundance of light elements.5,6 The classical Big Bang cosmology, however, suﬀered from several fundamental diﬃculties such as the horizon problem, the ﬂatness problem,7 and the initial singularity problem. All these problems relate to the initial conditions for the universe. Because reliable theories to describe the very early universe were lacking, these problems were not studied seriously until quite long after the conﬁrmation of the classical Big Bang cosmology by the discovery of CMB. Such a situation was somewhat changed by the introduction of grand uniﬁed theories (GUTs) by Georgi and Glashow.8 Since the uniﬁcation scale of GUTs, MGUT = 1015−16 GeV is on one hand too big to be experimentally accessible but on the other hand is just below the Planckian scale, MPl = (c5 /G)1/2 = 1.2 × 1019 GeV, where quantum gravitational eﬀects become important, there emerged a trend to use the high temperature and density regime in the early universe as an arena for its veriﬁcation, stimulating serious study of the birth of the universe. At the same time, however, GUT theories challenged the classical Big Bang cosmology by predicting a phase transition9 which overproduces10 magnetic monopoles11,12 that would dominate the matter content of the universe, leaving no room for ordinary baryonic matter, with Ωmonopole /Ωb 1015 . Fortunately, however, the path toward resolving this serious problem was already present in the GUT itself, as observed independently by Sato13–15 and Guth.16 Inﬂationary cosmology thus came into being. These pioneers observed that in the symmetric state of the Higgs ﬁeld, which is realized thermally shortly after the Big Bang, the ﬁeld has a large potential energy 4 , and that if this state is metastable with a long enough density of order of MGUT lifetime, the universe would be dominated by the false vacuum energy density which would not decrease as long as the Higgs ﬁeld remains in that state no matter how much the universe expands. As a result, an exponential cosmic expansion is realized which was referred to as cosmic inﬂation. Since both the particle horizon and the curvature radius are exponentially stretched during inﬂation, it can in principle provide solutions to the horizon and the ﬂatness problems. Apart from the GUT-based approach mentioned above, there was a longer tradition in the Russian school to pursue theories of the very early universe. Although many were speculative, a number of important ideas were proposed. An important such idea was Starobinsky’s curvature square theory which realized exponential cosmic expansion followed by Friedmann regime.17 Thus nowadays these three authors are regarded as the pioneers of inﬂationary cosmology18 before the slow-roll scenario was proposed.19,20 Before starting to review standard inﬂation updating Ref. 21, we ﬁrst describe early history of inﬂation. Since much has already been written about the developments in America (see e.g. Refs. 22 and 23), we emphasize those in Japan and Russia, which followed remarkably diﬀerent paths and have not been discussed in much detail in the English language literature. We take units c = = 1.

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-227

1.1. Developments in Japan In Japan theories of the early universe were developed in close connection with uniﬁed theories of elementary interactions. Compact astrophysical objects were also used as their testing ground. Uniﬁcation of elementary interactions in nature has been a long-standing issue of theoretical physics which is also called the Einstein’s dream. As the ﬁrst step, the theory to unify the weak interaction with the electromagnetic force was proposed by Weinberg24 and Salam25 in 1967, but it was not immediately widely accepted in the community. Only after 1973 when the Gargamelle experiment at CERN discovered the neutral current did this theory become popular. The importance of the neutral current interaction in the neutrino diﬀusion in supernovae explosions was shown in Ref. 26, which was conﬁrmed by the Nobel-prize winning discovery of the neutrino bursts from SN1987A.27 The mass of the Higgs boson in the electroweak theory was constrained by CMB28 and stellar evolution.29 Symmetry restoration and further breakdown in high density matter was also pointed out in Refs. 30 and 31 in the electroweak theory. Further progress toward a uniﬁed theory was made by Georgi and Glashow, who proposed the grand uniﬁed theories (GUTs).8 They argued that the three fundamental interactions unify at an extremely high energy scale around MGUT = 1015−16 GeV. At this high energy scale baryon nonconserving interactions characteristic of GUTs are expected to have occurred without suppression. Yoshimura32 proposed a model of baryogenesis in the early universe making use of these interactions. Applying studies of high-temperature behavior of uniﬁed gauge theories33–36 to the hot big bang universe, Sato and Sato proposed that all the four elementary interactions were uniﬁed above the Planckian scale when gravity bifurcated, and at the GUT phase transition, electroweak and strong interactions were separated from each other, followed by the Weinberg–Salam phase transition. Then the cosmic history of the elementary interactions can be expressed by a tree diagram as shown in Fig. 1. Sato also realized that, if the symmetry-breaking phase transition of the GUT Higgs ﬁeld in the early universe is ﬁrst-order, the universe would expand exponentially in a supercooled stage when the false vacuum energy dominates the universe, and that a number of interesting cosmological consequences would be obtained. In the ﬁrst paper of the series, Sato showed that the particle horizon would be exponentially stretched, so that even if CP symmetry is softly broken in baryogenesis, one may realize a large enough domain with positive baryon number instead of a mixture of small domains with positive and negative baryon numbers.14 He then showed that ﬂuctuations associated with the phase transition are exponentially stretched and may account for the origin of seed ﬂuctuations for large-scale structures.13 Although the current paradigm of cosmological structure formation attributes the origin of seed ﬂuctuations to quantum ﬂuctuations of a nearly massless scalar ﬁeld which induces inﬂation, the essential features of

May 4, 2017 9:54 b2167-ch18

II-228

K. Sato and J. Yokoyama

Fig. 1. Evolutionism of interactions. The original picture was first published in an article by Humitaka Sato and Katsuhiko Sato in June 1976 in a Japanese science magazine “Shizen” (meaning Nature) published by Chuo-Koron-sha.

the generation of superhorizon ﬂuctuations and its reentry to the Hubble radius in the subsequent Friedmann stage were already there.13 Finally, the monopole problem was also shown to be solved as the exponential expansion can push it away from the observable domain.15 Monopoles are diluted by subsequent entropy production. Unfortunately, however, this type of scenario making use of a ﬁrst-order phase transition, now referred to as old inﬂation, was not successful because inﬂation could not be terminated to realize a hot Friedmann universe over the comoving scale large enough to correspond to the current Hubble horizon. On the other hand, Sato and collaborators seriously considered geometrical outcomes after the phase transition and reached the revolutionary view that our universe is not unique but there can be multiple universes that are causally disconnected from each other.37,38 Nowadays the multiverse is very popular in uniﬁed theories such as string theories and even the probability that a universe like ours is created is seriously considered. One should remember that these papers were ﬁrst to discuss multiverse as an outcome of modern uniﬁed theories. 1.2. Developments in Russia In former Soviet Union, the possibility of exponential cosmic expansion was ﬁrst suggested by Gliner39 in 1965, who classiﬁed the possible algebraic structures of

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-229

the energy–momentum tensor. Among them he found a form of matter called the µ-vacuum, which has an equation of state (EOS) P = −µ with the energy density ρ = µ. This state has the same invariance as the vacuum and with µ > 0 the metric takes the de Sitter form. This proposal, however, cannot be considered as the birth of inﬂationary cosmology since a discussion of the transition to the Friedmann regime is absent. In fact, the observation that the matter with EOS P = −ρ induces exponential expansion was already recognized by Lemaˆıtre.40 This idea was pushed forward by Gurevich41 to solve the problem of causal connection of galaxies. Although he made some thermodynamic arguments of a “Λ” ﬁeld (he used the character Λ instead of µ), its transition to normal matter remained unclariﬁed. Almost at the same time Gliner and Dymnikova42 considered a cosmology with EOS P = −ρ only in the very early universe to avoid the initial singularity. In this scenario the EOS was supposed to change to that of normal radiation or matter in the course of expansion, but again its physical mechanism was left unexplored. A paper that was very interesting from modern point of view but not recognized as such at that the time was the paper43 by Starobinsky in 1978 on a nonsingular isotropic cosmological model whose material content is a massive scalar ﬁeld. Although his objective was to ﬁnd a viable bouncing solution to avoid the initial singularity, all the currently well-known ﬁeld equations and their slow-roll solutions that are used in chaotic inﬂation44 had already been given there. As noted by Starobinsky himself in his talk at the Tomalla Prize ceremony in 2009, the decisive step for Linde to propose chaotic inﬂation was to throw away the contraction phase (and replace it with large quantum ﬂuctuations in the Planckian regime to establish natural initial conditions). Thus in 1970s much work was done in Russia about the birth and the very early stage of the universe, but many of the papers were speculative and not based on concrete models, and moreover did not make contact with observation. One should note that gems of truly novel ideas often emerge from such speculation. One such novel idea is the calculation of the primordial tensor perturbations in the early de Sitter stage which was ﬁrst elaborated by Starobinsky45 in 1979, even before the concept of inﬂationary cosmology was introduced. His original motivation was physically sound with the desire to probe the initial state of the universe. He had quantum creation of the universe in mind and postulated that the initial state of the universe should be highly symmetric. Therefore, he considered de Sitter spacetime at the outset. He then considered the quantum mechanical behavior of gravitons to probe the initial state of the universe before the big bang, when other particles with stronger interactions were likely thermalized thus losing the memory of the initial state. In this paper again, however, the physical discussion on the transition to the Friedmann regime was absent. The sound physical mechanism to realize de Sitter expansion as well as subsequent transition to the Friedmann regime was ﬁnally given by Starobinsky by incorporating higher-order curvature terms in the action

May 4, 2017 9:54 b2167-ch18

II-230

K. Sato and J. Yokoyama

based on quantum corrections.17 Mukhanov and Chibisov46 calculated quantum ﬂuctuations generated in this model and pointed out the possibility that they may account for the origin of seed ﬂuctuations for galaxy formation. Meanwhile, apart from these mostly geometrical considerations, Kirzhnits and Linde33,34 investigated the high temperature behavior of uniﬁed gauge theories and phase transitions which may have occurred in the hot or dense early universe. Their study formed a basis of the inﬂationary cosmology. Indeed soon after the old inﬂation model was proposed and its problems were elucidated, Linde19 proposed the new inﬂation model based on a Coleman–Weinberg type potential for the GUT Higgs ﬁeld, in which inﬂation occurs as the scalar ﬁeld induces a slow-roll over phase transition toward its zero-temperature minimum after thermal correction of the potential has disappeared due to cosmic expansion.19,20 This was the ﬁrst slow-roll inﬂation model and the ﬁrst small-ﬁeld inﬂation model using a concave potential which is sometimes called hill-top inﬂation.47 Unfortunately, however, the original new inﬂation predicted a too short epoch of inﬂation with too large density ﬂuctuations,48 and thus could not reproduce our universe. Furthermore, as realized by Linde himself, the universe would have been far from a thermal state at the GUT era, hence one cannot expect thermal symmetry restoration to set the appropriate initial condition for new inﬂation. These consideration led to chaotic inﬂation.44 1.3. Inflation paradigm Now the standard paradigm of inﬂation is that the accelerated expansion is realized when a scalar ﬁeld, dubbed as the inﬂaton, slowly rolls down its potential toward a global minimum in a time scale longer than the Hubble time.19,20,44 The remarkable feature of the slow-roll inﬂation is that it cannot only explain the global properties of the observed universe, but also provide seeds of density and curvature ﬂuctuations that evolve into large-scale structures.46,48–50 The temperature (and polarization) anisotropies generated on CMB at the same time provide indirect cosmological tests of the inﬂation paradigm (see e.g. Refs. 51–53 among others). In fact, the slow-roll inﬂation driven by a potential is not the only possibility. There are presently two alternatives. One class of models realizes inﬂation without introducing any inﬂaton ﬁeld but modiﬁes gravity from the Einstein’s general relativity. As mentioned in Sec. 1.2 Starobinsky was the ﬁrst to show that quasiexponential expansion and subsequent transition to power-law expansion could be realized by incorporating higher-order curvature terms in the action based on quantum corrections.17 Nowadays its simpler version including square scalar curvature term besides the Einstein action is referred to as the Starobinsky model. The other alternative is comprised of models which make use of a scalar ﬁeld with a higherorder kinetic function. If we can realize a state with a constant canonical kinetic function its energy density can have the same equation of state as the cosmological constant to drive exponential inﬂation. Such models are referred to as k-inﬂation.54

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-231

All these models can be regarded as a subclass of the generalized G-inﬂation model,55,56 which is the most general theory of single-ﬁeld inﬂation with its ﬁeld equations given by second-order diﬀerential equations as with most of the other theories of fundamental physics. 2. Resolution of Fundamental Problems Although the original inﬂation models of Sato and Guth realized exactly exponential cosmic expansion, it is not a necessary condition to resolve horizon and ﬂatness problems. All we need is to realize a suﬃciently long period of accelerated expansion, when the particle horizon increases more rapidly than t and the Hubble horizon, so the horizon problem can be solved. From the Einstein equations for the cosmic scale factor a(t) 2 K 8πG a˙ + 2 = ρ, (1) a a 3 a ¨ 4πG =− (ρ + 3P ), (2) a 3 we see that such an expansion law is possible if cosmic energy density ρ and pressure P satisfy ρ + 3P < 0. Then the energy density decreases less rapidly than a−2 (t) so that the curvature term in the Friedmann equation (1) K/a2 (t) becomes unimportant. As a result, the ﬂatness problem is also solved simultaneously. Of course, such a stage dominated by a new energy component should not last forever but the universe must be converted to a state dominated by radiation to realize the initial state of the hot Big Bang cosmology, since we do not wish to demolish its success. The accelerated or inﬂationary expansion must be followed by creation of radiation and entropy. In fact it is this epoch rather than the inﬂationary expansion itself that the monopole and other unwanted relic problems such as domain walls, gravitinos, etc. are solved. Since inﬂation exponentially dilutes both monopole density and entropy density in the same way, the current universe with a huge entropy with a negligible amount of monopoles would never be realized without entropy production at the reheating. Let us consider quantitatively how much inﬂationary expansion is required to solve the horizon and ﬂatness problems in terms of entropy consideration. First the entropy contained in the current Hubble radius H0−1 = 4.2 × 103 Mpc is given by that of the CMB photon and neutrino background with the eﬀective temperature Tν = 1.95K as S0 = 2.6×1088. We consider the condition that the comoving Hubble −3 at t = ti , contains more entropy than volume at the beginning of inﬂation, 4π 3 H S0 after the reheating, so that currently observable region is well inside the initial Hubble radius at the onset of inﬂation. For simplicity, suppose that the energy density takes a constant value ρinf during inﬂation and it continues from t = ti to tf to stretch the scale factor by af /ai ≡ eN , and that after inﬂation the universe is dominated by a component with equationof-state P = wρ until the universe becomes radiation dominant with the reheating

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-232

temperature TR . At this time the initial Hubble radius has been stretched to 1 2 − 3(1+w) π g∗ TR4 −1 N H e ≡ rH . 30ρinf

(3)

The corresponding volume contains the entropy 1 − 1+3w −1+3w 1+w 1+w 4π 2 g∗ 3 4π 3 16π 3 e3N 45g∗w 1+w H TR S= TR × rH = , 90 3 270 4π 3 MPl MPl (4) where g∗ is number of relativistic degrees of freedom at reheating. Requiring this to be larger than S0 , we ﬁnd g r 12.5 − 10.8w w 1 + 3w ∗ N > 67.7 − − ln + ln 1+w 3 + 3w 106.75 6(1 + w) 0.01 2 TR 1 − 3w H + ln ≡ Nmin , r ≡ 0.01 , (5) 3 + 3w 108 GeV 2.4 × 1013 GeV where r is the tensor-to-scalar ratio to be deﬁned later. As mentioned above, this is the condition that the initial Hubble radius at the onset of inﬂation is stretched to be larger than the current Hubble radius. However, we must note that the observed density perturbation on this scale is observed to be 10−5 . We expect that the initial amplitude of ﬂuctuation on this scale can be close to unity just to initiate inﬂation. Thus it is not suﬃcient to have N = Nmin but 1 we need more inﬂation by a factor of (10−5 )− 2 ∼ 500 to suppress the amplitude of ﬂuctuations to an acceptable level.57 Thus the minimal condition to solve the horizon problem is actually N > Nmin + ln 500 = Nmin + 6.2. Taking this extra number of e-folds into account we ﬁnd 1 r 1 TR N > 55 + ln + ln (6) 6 0.01 3 108 GeV for w = 0, and

1 g∗ 1 r 1 TR N > 67 − ln + ln − ln 6 106.75 3 0.01 3 108 GeV

(7)

for w = 1. Next let us consider the evolution of the total density parameter Ωtot using K = H 2 (Ωtot − 1). a2 We ﬁnd Ωtot (t0 ) − 1 = Ωtot (ti ) − 1

a(ti )Hinf a 0 H0

2

= e−2(N −Nmin ) < 500−2 = 4 × 10−6 .

(8)

(9)

Thus, if the horizon problem is solved by inﬂation, the ﬂatness problem is automatically solved and such inﬂation models predict the total density parameter today is equal to unity with four to ﬁve digit accuracy.

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-233

3. Realization of Inflation 3.1. Three mechanisms As mentioned in the introduction one can classify realization mechanisms of inﬂation into three broad categories. The standard one is the slow-roll inﬂation models. Consider a scalar ﬁeld φ, the inﬂaton, with a canonical kinetic term X ≡ − 21 g µν ∂µ φ∂ν φ and a potential V [φ]. Taking the variation of the action √ √ S = L −gd4 x = (X − V [φ]) −gd4 x (10) with respect to the metric tensor, one ﬁnds the energy–momentum tensor 2 δS Tµν = − √ = ∂µ φ∂ν φ + Lgµν . −g δg µν

(11)

For homogeneous ﬁeld conﬁguration, the energy density, ρ, and the pressure, P , are given by ρ=

1 ˙2 φ + V [φ], 2

P =

1 ˙2 φ − V [φ], 2

(12)

respectively. Hence if the potential is so ﬂat that V [φ] > φ˙ 2 is satisﬁed for a long enough period, the equation-of-state parameter w ≡ P/ρ is smaller than −1/3 and an accelerated expansion can be realized. Next, in order to see how a scalar ﬁeld with a noncanonical kinetic term may aﬀect the dynamics of the universe, let us rewrite the Lagrangian in more general form as L = K(X, φ). Then calculating the energy–momentum tensor in the same way as (11), we ﬁnd ρ = 2XK X − K,

P =K

(13)

in this case. Thus an accelerated expansion is possible even without any potential, if XK X < −K is satisﬁed. In particular, in the case KX = ∂K/∂X = 0 holds we ﬁnd de Sitter expansion. This is what is called k-inﬂation. If we expand the Lagrangian as a power series of X like K(X, φ) = K1 (φ)X + K2 (φ)X 2 . . .

(14)

we immediately ﬁnd that K1 (φ) and K2 (φ) should have an opposite sign in order to realize inﬂation. So far we have implicitly assumed the gravity sector is expressed by the Einstein– Hilbert action √ 1 (15) Sg = 2 R −gd4 x 2κ

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-234

and considered a single scalar ﬁeld as a matter ingredient. We may realize inﬂation even without any scalar ﬁeld matter by modifying gravity as √ 1 Sg = 2 f (R) −gd4 x, (16) 2κ where f (R) is a function of scalar curvature R. The trace of ﬁeld equation is given by 3f (R) + f (R)R − 2f (R) = 0.

(17)

One can therefore ﬁnd a de Sitter solution, R = 12H =const., if f (R) satisﬁes f (R) = 2f (R), namely, f (R) ∝ R2 . A pure R2 model cannot terminate inﬂation nor has the proper Einstein limit, so we adopt 2

R2 . (18) 6M 2 This theory is a nonperturbative extension of the Einstein gravity, which contains additional scalar degree of freedom called scalaron. These mechanisms may not work in a mutually isolated manner but the actual inﬂation may have been induced by their combinations58 with possible higher-order derivative terms.56 f (R) = R +

3.2. Inflation scenario Let us describe cosmic evolution using a slow-roll inﬂation model as a prototype. Once inﬂationary expansion sets in, the universe rapidly becomes homogeneous, isotropic, and spatially ﬂat, and the energy density other than the inﬂaton is soon diluted away. Hence we may consider a spatially ﬂat FLRW universe from the beginning except when we discuss the feasibility of inﬂation in a generic inhomogeneous and anisotropic spacetime. Thus the scalar ﬁeld equation and the Einstein equation read (19) φ¨ + 3H φ˙ + V [φ] = 0, 2 8πρφ a˙ ρφ 1 = H2 = = , ρφ = φ˙ 2 + V [φ], (20) 2 2 a 3MPl 3MG 2 √ √ 1 respectively, where MG = MPl / 8π = 1/ 8πG = (c5 /8πG) 2 = 2.4 × 1018 GeV is the reduced Planck mass. To realize successful inﬂation the potential V [φ] must remain the dominant contribution to ρφ for a suﬃciently long time, that is, the scalar ﬁeld must remain practically constant over the cosmic expansion time scale. For this purpose φ¨ must contribute negligibly in the equation of motion. Then ﬁeld equations read (21) 3H φ˙ + V [φ] = 0, 2 a˙ 8πV [φ] V [φ] = H2 = = 2 2, a 3MPl 3MG which are called the slow-roll equations of motion.

(22)

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-235

For these approximate equations to hold, the potential must satisfy 2 2 MG V [φ] 2 V [φ] V ≡ 1, ηV ≡ MG , |ηV | 1. 2 V [φ] V [φ]

(23)

Here V and ηV are called the (potential) slow-roll parameters. Conversely, if these slow-roll conditions (23) are satisﬁed, inﬂationary expansion is likely to start as soon as the potential energy dominates the cosmic energy density. Inﬂation or accelerated ˙ becomes expansion is terminated when φ˙ 2 becomes larger than V [φ] or when |H| 2 larger than H . Under the slow-roll approximation, (21) and (22), the former occurs at V = 3/2 and the latter at V = 1. The discrepancy is due to the invalidity of the slow-roll approximation in such a regime. Under the slow-roll approximation, the number of e-folds of inﬂation after φ has crossed a value φN is given by φf dφ 1 φf dφ √ H N = Hdt = = (24) , M G φN 2V φ˙ φN with φf being the value at the end of inﬂation. Figure 2 shows typical shape of the inﬂation potential, which has two possible ﬁeld ranges to realize inﬂation. Near the local maximum at φ = 0 the potential is so ﬂat that inﬂation may be possible. Models of such class are called small-ﬁeld models. On the other hand, if the potential increases at most with a power-law, one can see that the slow-roll conditions (23) are satisﬁed for super-Planckian ﬁeld values, so that inﬂation is also realized there. Such a model is called a large-ﬁeld model. In both cases the slow-roll conditions no longer holds once the ﬁeld approaches the global minimum at φ = v, and it starts to oscillate around it. Thus the potential energy density is transferred to ﬁeld oscillation energy, which will eventually decay V[φ] Large field model Small field model

v Fig. 2.

φ

Scalar field potential for single-field slow-roll inflation.

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-236

into particles interacting with φ. The universe will then be heated up to realize the initial condition of the big bang universe physically in the modern context. In fact, Starobinsky’s R2 inﬂation can be regarded as a variant of a largeﬁeld model because the modiﬁed gravity action can be converted to the Einstein gravity plus a scalar ﬁeld with a potential. Speciﬁcally, in terms of conformal transformation, R (25) g¯µν ≡ 1 + gµν , 3M 2 we ﬁnd the action is equivalent with √ √ 1 1 µν 4 ¯ 4 Sg = −¯ gd xR + −¯ gd x g¯ ∂µ φ∂ν φ − V [φ] , 2κ 2 √ 3 R 3 2 2 − 23 κφ 2 κφ ≡ ln 1 + ) . , V [φ] ≡ M M (1 − e G 2 3M 2 4

(26)

(27)

Thus the system is equivalent with a scalar ﬁeld matter in the Einstein gravity. 2 For φ MG = κ−1 , the potential has a plateau with a height 3MG M 2 /4 where inﬂation can occur. Inﬂation is followed by scalar ﬁeld oscillation around the origin and reheating proceeds through gravitational decay of the inﬂaton φ. Finally, k-inﬂation of the type (14) is terminated when both K1 (φ) and K2 (φ) become positive. Then X starts to decrease rapidly and only the ﬁrst term becomes relevant. Now the universe is dominated by the kinetic energy of a free scalar ﬁeld whose energy density dissipates in proportion to a−6 (t) with the equationof-state parameter w = 1. This abrupt change from de Sitter to a power-law a(t) ∝ t1/3 induces gravitational particle production to reheat the universe in this model. 4. Slow-Roll Inflation Models Let us focus on speciﬁc models of slow-roll inﬂation with a canonical kinetic term. 4.1. Large-field models This model was originally proposed by Linde under the name “chaotic inﬂation”,44 since it makes use of the chaotic initial condition of the universe at the Planck time tPl = (G/c5 )1/2 = 5.4 × 10−44 sec, when we expect large quantum ﬂuctuations to dominate. Let us consider the following simple Lagrangian as an example assuming that Einstein gravity starts to hold at that epoch. 1 Lφ = − (∂φ)2 − V [φ], 2

V [φ] =

1 2 2 m φ . 2

(28)

By virtue of the uncertainty principle, at the Planck time we expect that both gradient energy and potential energy densities were ﬂuctuating with the Planckian

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-237

magnitude, so that 1 1 2 2 4 4 − (∂φ)2 MPl , m φ MPl , (29) 2 2 in each coherent domain. Taking m MPl we ﬁnd φ can take a large value up 2 /m MPl . If the magnitude of φ is saturated, from the gradient to φ ∼ MPl energy constraint we ﬁnd the coherent length of the ﬁeld, L, satisﬁes 12 (φ/L)2 ∼ −1 4 4 /(Lm)2 MPl , or L m−1 MPl . That is, φ is homogeneous over the MPl Compton wavelength well above the horizon scale at the Planckian regime. Then over the horizon scale one may regard the ﬁeld as homogeneous and apply the homogeneous ﬁeld equations derived in the previous section. Solving the ﬁeld equation (19) and the Einstein equation (20), one ﬁnds mMPl (30) φ(t) = φi − √ (t − ti ), 2 3π 4π m m2 a(t) = ai exp φi (t − ti ) − (t − ti )2 3 MPl 6

1 φ2 (t) = af exp − 2π 2 . (31) 2 MPl √ Inﬂationary expansion is terminated around φ MPl / 4π when V becomes ˙ unity and |φ/φ| becomes as large as H. Therefore, in order to solve the horizon problem one needs an initial amplitude φi 3MPl . The most stringent constraint on this model comes from the amplitude of density and curvature perturbation generated during inﬂation, which sets the mass m ∼ = 1013 GeV and constrains the self-coupling of the form λφ4 /4 to λ 10−12 (see Sec. 13.1).59 4.2. Small-field model As mentioned in Sec. 1.2, the ﬁrst small-ﬁeld model is Linde’s new inﬂation model19 which actually has a problem in realizing the required initial condition. There is another class of small-ﬁeld inﬂation models called topological inﬂation60,61 in which this problem is solved for topological reason without resorting to thermal symmetry restoration. Consider a simple Lagrangian 1 λ L = − (∂φ)2 − V [φ], where V [φ] = (φ2 − v 2 )2 , (32) 2 4 which has a local maximum at φ = 0. Since φ is a real scalar ﬁeld, this model admits a domain wall solution connecting two vacuum states φ = ±v. Neglecting gravitational eﬀect for the moment, we ﬁnd a solution

λ vz (33) φ(x) = v tanh 2

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-238

that describes a domain wall on the xy-plane. Its thickness d0 can be estimated by −1 2 equating (∇φ)2 ∼ (v/d0 ) and V [0] ≡ Vc as d0 ≈ vVc 2 . On the other hand, the Hubble horizon scale corresponding to the vacuum energy density Vc is given by − 12 12 8πG 3 −1 Hc = = MPl Vc , (34) 3 8πVc which is smaller than the thickness of the wall if v MPl . In this case one can ﬁnd a region with a large potential energy density V ∼ Vc whose dimension is larger than the Hubble scale near the center of the domain wall. Such a ﬁeld conﬁguration provides a suﬃcient condition to initiate inﬂation there without being aﬀected by the conﬁguration outside the Hubble horizon. In this model, an initially random ﬁeld conﬁguration in the global spacetime naturally creates domain walls inside of which inﬂation sets in near the central core. In this sense this model provides a natural mechanism of small-ﬁeld inﬂation provided that the universe continues to expand until the energy density of the domain wall is locally dominant. We can also show that inﬂation ends after a ﬁnite time except for the locus with φ = 0, and the reheating process proceeds just as in the chaotic inﬂation model. 4.3. Hybrid inflation This is a model to induce inﬂation by a false vacuum energy density through a nonthermal symmetry restoration by virtue of an extra scalar ﬁeld.62 The simplest model of hybrid inﬂation consists of a real scalar ﬁeld (φ) and a complex scalar ﬁeld (χ) with a Lagrangian 1 L = −(∂χ)† (∂χ) − (∂φ)2 − V [χ, φ], 2 where V [χ, φ] =

λ 1 (|χ|2 − v 2 )2 + g 2 φ2 |χ|2 + m2 φ2 . 2 2

(35)

From ∂V = λ(|χ|2 − v 2 )χ + g 2 φ2 χ, ∂χ† ∂2V ≡ Mχ2 = λ(2|χ|2 − v 2 ) + g 2 φ2 , ∂χ∂χ†

(36)

we ﬁnd that χ = 0 is a potential minimum if the inequality g 2 φ2 > λv 2 is satisﬁed initially. Then the potential reads λ 4 1 2 2 v + m φ . (37) 2 2 If the ﬁrst term dominates √ the energy density of the universe, inﬂation takes place. It is terminated at φ < gλv when χ induces a phase transition, which occurs within V [χ = 0, φ] =

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-239

one Hubble time typically unless ﬁne-tuning of parameters is applied. This model is attractive from particle physics viewpoint because enough inﬂation is possible even if φ is much smaller than the Planck scale63 and neither λ nor g requires ﬁne-tuning. Instead the initial ﬁelds conﬁguration of the two scalar ﬁelds do need tuning to realize inﬂation.64,65 In this model, since χ is a complex scalar ﬁeld, a network of strings will be formed after the phase transition. Where χ a real scalar ﬁeld, a network of domain walls would be formed and dominate the energy density. Other mechanisms of formation for the topological defects have been proposed in Refs. 66–70 near or at the end of inﬂation. Hence there may well remain some relic defects even if their energy scale is higher than the maximum temperature after inﬂation so that symmetry restoration by thermal eﬀects is impossible. 5. Reheating The entropy production process required after inﬂation to realize the initial condition of hot Big Bang cosmology is called reheating, although in modern inﬂationary cosmology the universe had never previous been dominated by radiation. As is clear from the discussion in the previous section, the universe is dominated by coherent scalar ﬁeld oscillation after potential-driven slow-roll inﬂation. Let us take the large-ﬁeld√massive scalar model as an example. As the inﬂaton starts to satisfy |φ| MPl / 6π the scalar ﬁeld starts to oscillate around the origin with a period 2π/m. Such ﬁeld oscillations are equivalent to the condensation of the homogeneous zero-mode of the scalar ﬁeld which decreases its amplitude through cosmic expansion, to decay into radiation ﬁnally. In the initial oscillatory regime when its amplitude is large, some nonperturbative particle production also takes place eﬃciently to some extent. This is called preheating.71–73 But the ﬁnal reheating stage is always dominated by the perturbative decay of the scalar ﬁeld, which is accounted for by incorporating a decay term Γφ φ˙ in its equation of motion.74–76 Here Γφ is the decay rate of a φ particle which is much smaller than m since the inﬂation must be weakly coupled to other ﬁelds to suppress quantum ﬂuctuations to an acceptable level as discussed in the following sections. Multiplying φ˙ to (19) after introducing the above mentioned dissipation term, we ﬁnd d 1 ˙2 1 2 2 φ + m φ = −(3H + Γφ )φ˙ 2 . (38) dt 2 2 Since φ is rapidly oscillating over the cosmic expansion time scale, we can replace 2 φ˙ 2 in the right-hand side by an average over the oscillation period, φ˙ , which is identical to the total energy density of φ, ρφ , thanks to the virial theorem.77 Thus, we ﬁnd a Boltzmann equation dρφ = −(3H + Γφ )ρφ , dt

(39)

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-240

which is associated with that for radiation energy density ρr dρr = −4Hρr + Γφ ρφ . dt These two equations can be solved as

−3 a(t) ρφ (t) = ρφ (tf ) exp[−Γφ (t − tf )], a(tf ) −4 t

a(t) ρr (t) = Γφ ρφ (τ )dτ. tf a(τ )

(40)

(41)

(42)

Thus the universe starts to be dominated by radiation at t Γ−1 φ , when the radiation temperature TR is given by

12 1 8π 8π π 2 g∗ 4 2 ∼ 1 ∼ 1 = ρ T = = Γφ , R 2 r 2 30 3MPl 2t 3MPl 2 1 1 12 200 4 200 4 Γφ 11 ∼ 0.1 TR ∼ M Γ 10 GeV. = Pl φ = g∗ g∗ 105 GeV

H=

(43)

(44)

Such is the case inﬂation is followed by scalar ﬁeld oscillation. If inﬂation is induced purely by kinetic terms as in the case of k-inﬂation54 and G-inﬂation,55 the entropy after inﬂation is attributed to gravitational particle production78 associated with the change of the cosmic expansion law. Here we describe this process following Ref. 79. Let us consider the case where the cosmic expansion law changes from de Sitter with the Hubble parameter Hinf to a power law with the index t1/3 corresponding to a universe dominated by the kinetic energy of a free massless homogeneous scalar ﬁeld which is called the kination regime. From the standard calculation of particle creation,78,80,81 we consider the creation of massless minimally coupled scalar particles. The relevant Bogolubov coefﬁcient βω is given by ∞ i βω = e−2iωη V (η)dη, (45) 2ω −∞ with V (η) = 16 a2 (η)R(η), which yields expressions for the number density and energy density ∞ 1 nr = |βω |2 ω 2 dω, (46) 2π 2 a3 0 ∞ 1 ρr = |βω |2 ω 3 dω. (47) 2π 2 a4 0 If we approximate the transition from de Sitter to the power law expansion occurred abruptly we ﬁnd the resultant radiation energy density diverges. We must adopt

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-241

a model where the scale factor evolves smoothly up to the second time derivative. We consider the following model where the scale factor a2 (η) ≡ f (Hinf η) evolves as 1 (x < −1), x2 f (x) = (48) a0 + a1 x + a2 x2 + a3 x3 (−1 < x < x0 − 1), b0 (x + b1 ) (x0 − 1 < x) requiring f (x), f (x), f (x) to be continuous at x = −1 and x = x0 − 1, so that the scalar curvature is continuous. Since the transition occurs within the Hubble time x0 is smaller than unity. The coeﬃcients are determined as (up to O((x0 )0 )), a0 = 6 −

1 , x0

a1 = 8 −

a3 = −

1 , x0

3 , x0

b0 = 2,

a2 = 3 − b1 =

3 , x0

3 . 2

(49) (50)

Then the energy density of created particles is given by ρr = where

I =−

x

−∞

dx1

x

−∞

4 Hinf I, 128π 2 a4

dx2 ln(|x1 − x2 |)V˜ (x1 )V˜ (x2 ),

1 f f − (f )2 2 V˜ (x) = . f2

(51)

(52)

(53)

The upper limit of integration x = Hη is the time when V˜ falls well below unity so that the notion of the particle is well-deﬁned. We ﬁnd I −36 ln x0 , so that the energy density in the beginning of the kination regime is given by 4 1 9Hinf ln ρr = , (54) 32π 2 a4 Hinf ∆t where ∆t is the time elapsed during the transition from de Sitter to kination. We will ignore the logarithmic dependence on ∆t and take ln(1/(H∆t)) ∼ 1. If there are N modes of species created this way, (54) should be multiplied by N . Since the energy density of the scalar ﬁeld in the kination regime behaves as 2 2 Hinf a−6 , ρφ = 3MG

(55)

the scale factor when the universe becomes radiation dominant is given by a2R =

2 32π 2 MG , 3NH 2inf

(56)

May 4, 2017 9:54 b2167-ch18

II-242

K. Sato and J. Yokoyama

so the energy density of the relativistic particles is ρr |R =

8 81N 3 Hinf 4. (32π 2 )3 MG

(57)

Thus the reheating temperature is given in terms of the tensor-to-scalar ratio r by 1 2 3 g − 14 r 3N 4 30 4 Hinf 3 ∗ 6 4 TR = GeV. (58) 3.9 × 10 N 3 2g 2 106.75 0.01 M π 4 (32π ) G ∗ 6. Generation of Quantum Fluctuations that Eventually Behave Classically Thanks to the smallness of the slow-roll parameters, quantum ﬁeld theoretic properties of the inﬂaton during inﬂation is quite similar to that of a massless minimally coupled scalar ﬁeld ϕ(x, t) in de Sitter space ds2 = −dt2 + e2Ht dx2 .

(59)

Such a scalar ﬁeld is known to exhibit anomalous growth of the square vacuum expectation value according to82–84 2 H 2

ϕ(x, t) = Ht. (60) 2π Decomposing the scalar ﬁeld as d3 k ϕ(x, t) = ˆ†k ϕ∗k (t)e−ik ·x ) (ˆ ak ϕk (t)eik ·x + a (2π)3/2 d3 k ≡ ϕˆk (t)eik ·x , (2π)3/2

(61)

˙ t), hence we ﬁnd the momentum conjugate to ϕ(x, t) reads π(x, t) = a3 (t)ϕ(x, the canonical commutation relation [ϕ(x, t), π(x , t)] = iδ(x − x ) is equivalent to ˆ†k ] = δ (3) (k − k ) if we impose the normalization condition [ˆ ak , a ϕk (t)ϕ˙ ∗k (t) − ϕ˙ k (t)ϕ∗k (t) =

i a3 (t)

.

(62)

Thus a ˆ†k and a ˆk satisfy the usual commutation relations and act as creation and annihilation operators of a k-mode, respectively. The mode function ϕk (t) satisﬁes

2 d k2 d + 2Ht ϕk (t) = 0, + 3H (63) dt2 dt e as derived from the Klein–Gordon equation in de Sitter space, which is solved as 3 iH π (1) ϕk (t) = H(−η) 2 H 3 (−kη) = √ (1 + ikη)e−ikη , (64) 2 4 2k 3

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-243

under the condition (62). Here η is the conformal time deﬁned by

t

η≡

dt = a(t)

t

dt 1 =− . eHt HeHt

(65)

Note that since (63) is a second-order diﬀerential equation there are two inde(2) (1) pendent solutions proportional to H 3 (−kη) and H 3 (−kη). We have adopted only 2 2 (64) which coincides with the positive-frequency mode in the Minkowski space at large wavenumber limit where the eﬀect of cosmic expansion is negligible. k Note also that −kη = Ha(t) is nothing but the ratio of physical wavenumber to the Hubble parameter. In the superhorizon limit k a(t)H, we ﬁnd iH iH ϕk = √ (1 + ikη)e−ikη = √ 3 2k 2k 3 2 iH k → √ 1+O , Ha 2k 3

1−

ik Ha

k

ei Ha (66)

so that we can put ϕ∗k (t) = −ϕk (t) and regard ϕˆk (t) = ϕk (t)(ˆ ak − a ˆ†−k ) holds when the wavelength corresponding to k is much larger than the Hubble radius. Then the ˆk (t) = a(t)3 ϕ˙ k (t)(ˆ ak − a ˆ†−k ), which means momentum conjugate to ϕˆk is given by π ˆk have the same operator dependence in the superhorizon regime. that both ϕˆk and π Thus in this regime, where the decaying mode is negligible, the quantum operˆk commute with each other, so that quantum ﬂuctuations behave as ators ϕˆk and π classical statistical ﬂuctuations.85 Furthermore, the absolute square of mode function behaves as |ϕk (t)|2 =

H2 H2 1 + (kη)2 → 3 , 3 2k 2k

when

k → 0, Ha(t)

(67)

that is, it takes a constant value proportional to k −3 . Multiplying the phase space 4πk2 density over a unit logarithmic frequency interval, (2π) 3 dk with dk = kd ln k = k, we ﬁnd the dispersion takes a constant over each logarithmic frequency interval. One can immediately see that (60) can be reproduced from this mode function by summing up only superhorizon modes only as

HeHt

ϕ (x, t) 2

d3 k |ϕk (t)| = (2π)3 2

H

H 2π

2

Ht,

(68)

where the infrared cutoﬀ is identiﬁed with the mode that left the Hubble radius at the beginning of inﬂation set to t = 0 here.86 The behavior that the square expectation value increases in proportion to time is the same as that of Brownian motion with a step ±H/(2π) at each time interval

May 4, 2017 9:54 b2167-ch18

II-244

K. Sato and J. Yokoyama

H −1 . Thus one can describe the quantum ﬂuctuations generated during inﬂation as follows. H For each Hubble time H −1 , a quantum ﬂuctuation with amplitude 2π , which is the Hawking temperature of de Sitter space, is continuously generated on the scale of Hubble radius, and it is stretched by subsequent cosmic expansion to form the long wavelength ﬂuctuations.

Finally as a preparation for the calculations of curvature and tensor perturbations, let us write down the action of a massive scalar ﬁeld ϕ using conformal time with the line element ds2 = a2 (η)(−dη 2 + dx2 ).

√ 1 µν 1 2 2 4 S= −gd x − g ∂µ φ∂ν φ − m φ 2 2 1 = dηd3 x a2 φ2 − (∇φ)2 − a4 m2 φ2 , (69) 2 where a prime denotes diﬀerentiation with respect to conformal time. Introducing a rescaled variable χ ≡ aφ, we ﬁnd

a 1 3 2 2 2 2 S= dηd x χ − (∇χ) − a m − (70) χ2 2 a which is equivalent to the action of a scalar ﬁeld with a time-dependent mass term in Minkowski space. In de Sitter space with a(η) = −1/(Hη), the mode function satisﬁes χk + k 2 χk −

2 χk = 0, (−η)2

(71)

where we have taken m = 0. Its solution may be expressed as πη 12 ϕk (t) (1) χk (η) = − H 3 (−kη) = 2 4 a(t)

(72)

in agreement with (64) with the normalization condition χ χ∗ − χχ∗ = i equivalent to (62). 7. Cosmological Perturbation We now superimpose linearized metric perturbations onto the spatially ﬂat homogeneous and isotropic background spacetime ds2 = −dt2 + a(t)2 dx2 ,

(73)

obtaining the perturbed spacetime ds2 = −(1 + 2A)dt2 − 2aBj dtdxj + a2 (δij + 2HL δij + 2HT ij )dxi dxj ,

(74)

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-245

where all new variables are functions of time and position87–89 with Tr HT ij = 0. As is well known, a spatial vector Bj can be decomposed to rotation-free component and divergence-free component as j , Bj = ∂j B + B

j = 0. where ∂j B

(75)

Similarly, the spatial tensor HT ij can be decomposed as δij 2 T j + ∂j H T i + HT T ij , HT ij = ∂i ∂j − ∇ HT + ∂i H 3 T j = 0, where ∂j H

∂j HT T k j = 0,

HT T j j = 0.

(76)

Here HT T ij is the transverse-traceless component corresponding to gravitational waves. Thus perturbation variables are classiﬁed into scalar, vector, and tensor variables according to the spatial transformation law. Variables of diﬀerent types are not mixed in linear perturbation theory and we can treat them independently. Since the vector perturbation has only a decaying mode under normal circumstances in the linear perturbation theory,87,88 we do not consider it hereafter and focus on the scalar and tensor perturbations. First let us consider scalar perturbations which are related with the density and curvature ﬂuctuations, keeping only A, B, HL and HT . In fact, not all of these variables are geometrical quantities but gauge modes are included as well. They arise from the arbitrariness of the choice of the background spacetime with respect to which we deﬁne the perturbation variables. One should remember that the real physical entity is an inhomogeneous and anisotropic spacetime, and that it may be regarded as homogeneous and isotropic only after taking some average whose deﬁnition is not unique in general. To characterize the gauge dependence, let us consider two background spacetimes with coordinates xµ and xµ , and assume that these two coordinates are mutually related by the coordinate transformation x0 = x0 + δx0 ≡ x0 + T,

xi = xi + δxi ≡ xi + ∂ iL,

(77)

where δxµ are small quantities whose relative magnitudes are of the same order as the perturbation variables. Here T and L are functions of the space and time coordinates. Under this coordinate transformation, we ﬁnd that the scalar perturbation variables at the same coordinate values in two diﬀerent backgrounds are related according to A = A − T˙ , 1 B = B + aL˙ + T, a H L = HL −

∇2 L − HT, 3

H T = HT + L,

(78)

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-246

respectively. Here T and L represent gauge degrees of freedom on scalar perturbation. In order to avoid their appearance, one may either ﬁnd combinations of perturbation variables that remain unchanged after gauge transformation, or ﬁx the gauge completely. For example, if we set H T = 0, L is ﬁxed. If we further set B = 0, T is also ﬁxed and there remains no gauge mode. In this case the perturbed metric is given by ds2 = −(1 + 2Ψ)dt2 + a2 (1 + 2Φ)dx2 ,

(79)

where we have rewritten A = Ψ and HL = Φ to stress their gauge invariance. This is called the longitudinal gauge. In order to ﬁx the gauge, one may combine transformation properties of matter variables such as a scalar ﬁeld, which transforms as ˙ φ(t, x) = φ(t − T, xj − ∂ j L) = φ(t − T ) + δφ = φ(t) − φ(t)T + δφ,

(80)

˙ δφ = δφ − φT.

(81)

that is,

8. Generation of Curvature Fluctuations in Inflationary Cosmology Here, we consider the generation of curvature perturbations in single-ﬁeld inﬂation models that include not only potential-driven slow-roll models but also k-inﬂation. Further generalization to the case of the most general single-ﬁeld inﬂation model with second-order ﬁeld equations will be carried out in Sec. 10 based on Ref. 56. We start with the action

2 MG 4 √ R + K(X, φ) . (82) S = d x −g 2 In the spatially ﬂat homogeneous and isotropic spacetime (73) the background equations read 2 3MG H 2 = ρ = 2XKX − K,

2 ˙ 2 H + 3MG 2MG H 2 = −P = −K,

(83)

together with the scalar ﬁeld equation of motion KXφ 2 ˙ 2 Kφ 2 φ¨ + 3Hc2s φ˙ + c φ − c = 0, KX s KX s

c2s ≡

PX KX = . ρX KX + 2XKXX

(84)

In order to derive the second-order action with respect to the curvature perturbation in the comoving gauge, R, it is convenient to use (3 + 1) formalism ds2 = −N 2 dt2 + hij (dxi + N i dt)(dxj + N j dt),

hij ≡ a2 (t)e2R δij ,

(85)

in which the constraint equations for the perturbation variables are easily obtained.90 This coordinate choice still has a gauge degree of freedom with respect to time translation by which the scalar ﬁeld ﬂuctuation is transformed as (81). Setting δφ = 0 one can ﬁx this gauge freedom.

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-247

Then writing the action (82) in this gauge, we ﬁnd √ √ 1 M2 2 (3) S= d4 x hN (MG R + 2K) + G d4 x hN −1 (Eij E ij − E 2 ) + · · · , 2 2 (86) where 1 ˙ (hij − Ni|j − Nj|i ), E ≡ Tr E, (87) 2 where terms replaced by · · · are all total derivative terms, so they do not aﬀect the dynamics of the system. Here R(3) is scalar curvature of 3-space and | denotes covariant derivative with respect to hij . Since we are concerned with scalar-type ﬂuctuations, we set N = 1 + α and Ni = ∂i ψ, where α and ψ are perturbation variables of the same order as R. Diﬀerentiating the action (86) with respect to N , we ﬁnd the Hamiltonian constraint Eij ≡

R(3) + 2

K KX 1 ij 2 2 − 4X M 2 − N 2 (Eij E − E ) = 0, MG G

(88)

which yields H 2 1 2 ∇ ψ = − 2 ∇2 R + Σα, MG Σ ≡ XKX + 2X 2 KXX . (89) 2 a a The momentum constraint obtained by diﬀerentiation with respect to Ni reads

1 j j (E − Eδi ) = 2Hα,i − 2R˙ ,i = 0, (90) N i |j ˙ which gives α = R/H. Inserting these relations into the action, we obtain an action for R only as

Σ ˙2 H˙ (∇R)2 2 3 3 R S2 = MG dtd xa − ≡ − , . (91) H H H2 a2 H2 1

1

Introducing new variables z ≡ a(2Σ) 2 /H = a(2H ) 2 /cs , v ≡ MG zR and using conformal time, we ﬁnd

1 z 2 3 2 2 2 dηd x v − cs (∇v) + v , S2 = (92) 2 z which like (70) is an action of a free scalar ﬁeld with a time-dependent mass term

z ηH ηH s˙ η˙ H 2 2 − H − s + = (aH) 1−s+ − + z 2 2 H 2H (93) ≡ (aH)2 (2 + q), ˙H c˙s ηH ≡ , s≡ , HH Hcs with the important diﬀerence that the “sound velocity” cs may not be identical to unity. Still, one can quantize v in the same way as a free scalar ﬁeld if the time variation of cs is negligible. q is a small number consisting of slow-roll parameters.

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-248

Using the de Sitter scale factor a = −1/(Hη) one can obtain the mode function in the same way as in Eq. (72), so that πη 12 1 i (1) ∼ vk = − Hν (−kcs η) = √ 1− e−ikcs η , 4 kcs η 2kcs (94) 1 3 4 2 ∼ 3 ν= 1+ q = , 2 9 2 which approaches to a constant in the long wave limit. Thus the power spectrum of curvature perturbations multiplied by the phase space density reads 2 4πk 3 4πk 3 vk H2 2 PR (k) ≡ |R | = = (95) k 2c . 3 3 2 (2π) (2π) MG z 8π MG s H Since each quantity in the right-hand side has a weak time dependence, it is evaluated at the time k-mode left the sound horizon when |kcs η| = 1 was ﬁrst satisﬁed. The validity of this approximation relies on the fact that the mode amplitude remains constant after the mode has exited the horizon. The spectral index of curvature perturbation is deﬁned and given by d ln PR (k) (96) = −2H − ηH − s. d ln k In case the scalar ﬁeld has a canonical kinetic term as in the standard theory of 2 H 2 ). Thus we ﬁnd particle physics, we ﬁnd cs = 1 and H = φ˙ 2 /(2MG 2 2 2 H δϕ PR (k) = = H . (97) 2π φ˙ φ˙ ns − 1 ≡

Here δϕ = H/(2π) is the amplitude of scalar ﬁeld ﬂuctuation. This equality is intuitively understandable since it represents relative ﬂuctuation of local scale factor δa δϕ = δN = Hδt = H , a φ˙

(98)

where N ≡ ln a. Since this δN gives appropriate expression for long wave curvature ﬂuctuations if we take an appropriate gauge, this δN method91,92 is often used in the literature. For slow-roll inﬂation models with a canonical kinetic term, from (21) and (22) we ﬁnd H = V and ηH = −2ηV + 4V , so the spectral index is given by ns − 1 = −6V + 2ηV .

(99)

Furthermore the scale dependence of the spectral index, which is called the running spectral index, is given by dns = 16V ηV − 242V − 2ξV , d ln k

4 ξV ≡ MG

V (φ)V (φ) . V 2 (φ)

(100)

These quantities are observationally measured using CMB, as we shall see in Sec. 12.

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-249

9. Tensor Perturbation Next we consider tensor perturbations generated during inﬂation as quantum gravitational waves.45,93,94 In order to derive ﬁeld equation for the graviton, ﬁrst we superimpose a small metric perturbation hµν onto Minkowski space background as gµν = ηµν + hµν . The indices of hµν is raised by the Minkowski tensor. We again need to calculate the action up to second-order in the perturbed variables, so we should use g µν = η µν − hµν + hµα hνα . Taking the longitudinal and traceless gauge, ,j i we ﬁnd h00 = h0i = 0, hα α = hi = 0, hij = 0, so that the Ricci scalar is given by 3 ij,µ 1 R = hij h,µ hij,µ − hij,l hjl,i . ij,µ + h 4 2

(101)

Here the Greek indices runs from 0 to 3 while the Latin indices from 1 to 3. Since what we need is a perturbed action in cosmological background with the metric (102) ds2 = a2 (η) −dη 2 + (δij + hij )dxi dxj ≡ g˜µν dxµ dxν . We make use of a conformal transformation g˜µν = Ω2 gµν to calculate the desired quantity. The Ricci tensor transforms as ˜ µν = Rµν − 2(ln Ω);µν − gµν g στ (ln Ω);τ σ R + 2(ln Ω);µ (ln Ω);ν − gµν g στ (ln Ω);σ (ln Ω);τ . Taking Ω = a(η), we ﬁnd the Ricci scalar in the metric (102) is given by ˜ = a−2 R + 6 a − 3 a hij hij . R a a

(103)

(104)

Since we wish to calculate properties of quantum gravitational waves generated during inﬂation, let us assume inﬂation is driven by constant vacuum energy density, which is equivalent to a cosmological constant Λ, and calculate the second-order action for hij , which reads 2 MG ˜ − 2Λ) −˜ S2T = gd4 x (R 2 second-order 2 M = G dηd3 xa2 (hij hji − hij,l hj,l (105) i ), 8 where we have used Λa4 = 2aa − a2 . This action has the same form as a free scalar ﬁeld again, so we can quantize it in the same way as before. Indeed taking new variables as zT ≡ a/2, uij ≡ MG zT hij , we ﬁnd

1 a 2 3 2 2 u , S2T = (106) dηd x uij − (∇uij ) + 2 a ij which is indeed of the same form as (70).

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-250

For exact de Sitter space with H = 0 we have a = −1/(Hη), whereas in case H = 0, one can express the scale factor around an arbitrary time η∗ using the Hubble parameter at that time H∗ as 1 −ν 1 1 31− 3 −η 2 T a=− , νT = . (107) H∗ η∗ 1 − −η∗ 2 1− Therefore, we ﬁnd the normalized mode function πη 12 uA = − Hν(1) (−kη)eA ij ij (k), T 4

A = +, ×,

(108)

A ∗ijB where eA (k) = δ AB . ij (k) is a polarization tensor that satisﬁes eij (k)e Thus, the power spectrum of long-wave tensor perturbation is given by

PT (k) ≡

∗ijA 4πk 3 4πk 3 uA 2H 2 ij u ∗ij h h = = ij 2 2 2 (2π)3 (2π)3 MG zT π 2 MG

(109)

with the spectral index d ln PT (k) = −2H . d ln k Finally the ratio of this power spectrum to the curvature perturbation, nt ≡

r≡

PT (k) = 16cs H = −8cs nt PR (k)

(110)

(111)

is called the tensor-to-scalar ratio. It is one of the most important quantity subject to ongoing and future observational tests of inﬂation. In Sec. 11, we present the relation between r and nt in more generic theories. 10. The Most General Single-Field Inflation So far we have considered a rather limited class of inﬂationary universe models including only potential-driven slow-roll inﬂation and k-inﬂation. But theories consisting of tensor and a single scalar degrees of freedom have much more functional degrees of freedom under the conditions that they are ghost free and stable. In fact, the most general single-ﬁeld action with gravity which yields second-order ﬁeld equations was formulated by Horndeski95 back in 1974. A theory constructed with the same objective but totally diﬀerent appearance was also proposed by covariantizing theories with Galilean invariance96 ∂µ φ → ∂µ φ + bµ and referred to as generalized Galileon.97 The equivalence of two theories was shown in Ref. 56. Here, making use of this theory, we provide a comprehensive and thorough study of the most general noncanonical and nonminimally coupled single-ﬁeld inﬂation models yielding second-order ﬁeld equations based on Ref. 56, which is called generalized G-inﬂation since it is a natural and comprehensive extension of G-inﬂation model,55 which was the ﬁrst inﬂation model using Galileon-like higher derivative theory.

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-251

We use the modern notation of Ref. 97, which contains four arbitrary functions of φ and the canonical kinetic function X ≡ −∂µ φ∂ µ φ/2 as L2 = K(φ, X),

(112)

L3 = −G3 (φ, X)φ,

(113)

L4 = G4 (φ, X)R + G4X (φ, X)[(φ)2 − (∇µ ∇ν φ)2 ],

(114)

L5 = G5 (φ, X)Gµν ∇ ∇ φ µ

ν

1 − G5X (φ, X)[(φ)3 − 3(φ)(∇µ ∇ν φ)2 + 2(∇µ ∇ν φ)3 ], 6

(115)

where Gµν is the Einstein tensor, (∇µ ∇ν φ)2 = (∇µ ∇ν φ)(∇µ ∇ν φ), (∇µ ∇ν φ)3 = (∇µ ∇ν φ)(∇ν ∇λ φ)(∇λ ∇µ φ), and GiX = ∂Gi /∂X. The Lagrangian of the Einstein gravity may be automatically included by taking 2 /2, and the extension to a nonminimally coupled scalar ﬁeld is also G4 = MG immediate. Thus, the total action we consider is given by 5 √ d4 x −gLi , S= (116) i=2

which is the most general single scalar theory resulting in equations of motion containing derivatives up to second-order. Then the standard slow-roll inﬂation can be achieved by K = X − V [φ] with all the other terms except for the abovementioned Einstein term equal to zero. It can also realize other inﬂation models easily; k-inﬂation54 is realized by K(φ, X), G-inﬂation55 by K and G3 terms, and even New Higgs inﬂation,98 which couples the Einstein tensor with the scalar kinetic term, can be realized taking an appropriate form of G5 . We also note that the curvature-square inﬂation17 or more general f (R) inﬂation can also be recast in the present form by deﬁning a new ﬁeld as φ = df/dR. Below we describe ﬁeld equations for the homogeneous background and present two examples based on Refs. 56 and 99. 10.1. Homogeneous background equations The background equations of motion can be derived from (116) taking φ = φ(t) and the spacetime metric ds2 = −N 2 (t)dt2 + a2 (t)dx2 . Variation with respect to N (t) gives the constraint equation corresponding to the Friedmann equation, which can be written as 5

Ei = 0,

(117)

i=2

where E2 = 2XK X − K,

(118)

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-252

˙ E3 = 6X φHG 3X − 2XG3φ ,

(119)

˙ 4φX − 6H φG ˙ 4φ , E4 = −6H 2 G4 + 24H 2 X(G4X + XG4XX ) − 12HX φG

(120)

2 ˙ E5 = 2H 3 X φ(5G 5X + 2XG5XX ) − 6H X(3G5φ + 2XG5φX ).

(121)

Being the constraint equation, the above terms contain derivatives of the metric and the scalar ﬁeld only up to ﬁrst-order. The evolution equation is obtained by diﬀerentiation with respect to a(t), which reads 5 Pi = 0, (122) i=2

where P2 = K,

(123)

¨ 3X ), P3 = −2X(G3φ + φG

(124)

˙ 4 − 12H 2 XG4X − 4H XG ˙ 4X − 8HXG ˙ ˙ P4 = 2(3H 2 + 2H)G 4X − 8HX XG4XX ˙ 4φ + 4XG4φφ + 4X(φ¨ − 2H φ)G ˙ 4φX , + 2(φ¨ + 2H φ)G

(125)

¨ 5X − 4H 2 X 2 φG ¨ 5XX P5 = −2X(2H 3 φ˙ + 2H H˙ φ˙ + 3H 2 φ)G ˙ 5φφ . + 4HX (X˙ − HX )G5φX + 2[2(HX ) + 3H 2 X]G5φ + 4HX φG

(126)

Although the background quantities Ei and Pi are deﬁned in an analogous way in which the energy density and the isotropic pressure of a usual scalar ﬁeld are deﬁned, the partition between the gravitational and scalar-ﬁeld parts of the Lagrangian is ambiguous, because the gravitational contributions are treated with the matter contributions on an equal footing in the above expressions. Indeed, one can see 2 /2, E4 and P4 reduce to the Einstein tensor multiplied by −1: that, for G4 = MG 2 2 2 ˙ (3H 2 + 2H). E4 = −3MG H and P4 = MG Finally, variation with respect to φ(t) gives the scalar-ﬁeld equation of motion, 1 d 3 (a J) = Pφ , a3 dt

(127)

where ˙ X + 6HX G3X − 2φG ˙ 3φ + 6H 2 φ(G ˙ 4X + 2XG4XX ) − 12HX G4φX J = φK ˙ 5φ + XG5φX ) + 2H 3 X(3G5X + 2XG5XX ) − 6H 2 φ(G

(128)

and ¨ 3φX ) + 6(2H 2 + H)G ˙ 4φ + 6H(X˙ + 2HX )G4φX Pφ = Kφ − 2X(G3φφ + φG ˙ 5φX . − 6H 2 XG5φφ + 2H 3 X φG

(129)

The solutions of the above equations admitting accelerated cosmic expansion constitute generalized G-inﬂation. We present two examples below.

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-253

10.2. Kinetically driven G-inflation If we impose shift symmetry, φ → φ + c, the scalar ﬁeld has no potential. In this case, the ﬁeld equations are 5

˙ − K − 6H 2 (G4 − 2XG4X ) + 4H 3 X φG ˙ 5X = 0, Ei = φJ

(130)

i=2 5

˙ − 2X φG ¨ 3X + 2 d [2H(G4 − 2XG4X ) − H 2 X φG ˙ 5X ] = 0, (Ei + Pi ) = φJ dt i=2 (131) d 3 (a J) = 0. dt

(132)

From Eq. (132), one immediately ﬁnds that J ∝ a−3 → 0. Shift-symmetric models thus have an attractor, J = 0, along which H = const., φ˙ = const., satisfying ˙ X + 6HXG3X + 6H 2 φ˙ (G4X + 2XG4XX ) φK + 2H 3 X (3G5X + 2XG5XX ) = 0, ˙ 5X = 0. K + 6H 2 (G4 − 2XG4X ) − 4H 3 X φG

(133) (134)

Provided that Eqs. (133) and (134) have a nontrivial root, H = 0, φ˙ = 0, we obtain inﬂation driven by φ’s kinetic energy. As a simple example, let us take K = −X +

X2 , 2M 3 µ

G3 =

X , M3

(135)

with M and µ being constants with unit mass dimension. Writing the ﬁeld equations ˙ for this Lagrangian, we can easily ﬁnd a de Sitter solution with a constant φ, X = M 3 µx, H2 =

M 3 (1 − x)2 , 18µ x

(136)

√ where x (0 < x < 1) is a constant satisfying (1 − x)/x 1 − x/2 = 6µ/MG . For √ µ MG , it can be seen that x 1 − 3µ/MG and hence the Hubble rate during 2 ). We can show that inﬂation is given in terms of M and µ as H 2 M 3 µ/(6MG this inﬂationary solution is an attractor of the theory. Inﬂation can end if the shift symmetry is broken at some φ where the ﬁrst term of K changes sign. Then the universe will soon become dominated by the kinetic energy of the scalar ﬁeld, which decreases in proportion to a−6 after the higherorder term in K becomes negligible compared with the ﬁrst term. Since the shift symmetry of the original Lagrangian prevents direct interaction between φ and standard-model sector, reheating proceeds only through gravitational particle production.78,79 The reheating temperature in such a case is given by (58).

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-254

10.3. Potential-driven slow-roll G-inflation Next we consider the opposite extreme. Suppose that the functions in the Lagrangian can be expanded in terms of X as K(φ, X) = −V (φ) + K(φ)X + · · · ,

(137)

Gi (φ, X) = gi (φ) + hi (φ)X + · · · ,

(138)

and consider the case in which the inﬂaton ﬁeld value φ(t) changes very slowly. In this case, the potential term manifestly breaks the shift symmetry, so that the model is capable of a graceful exit from inﬂation.100 Neglecting all the terms multiplied by φ˙ in the gravitational ﬁeld equations, we obtain 5

Pi −

i=2

5

Ei −V (φ) + 6g4 (φ)H 2 ,

(139)

i=2

where we have assumed ˙ H2 |H|

¨ |H φ|. ˙ and |φ|

(140)

We may thus have slow-roll inﬂation with H2

V . 6g4

(141)

During slow-roll, we approximate ˙ |HJ|, |J|

|g˙ i | |Hgi |,

|h˙ i | |Hhi |.

(142)

Under the above approximation, we have the slow-roll equation of motion for φ, 3HJ −Vφ + 12H 2 g4φ ,

(143)

J Kφ˙ − 2g3φ φ˙ + 6(Hh3 X + H 2 h4 φ˙ − H 2 g5φ φ˙ + H 3 h5 X).

(144)

with

Which term is dominant in Eq. (144) depends on the magnitude of the coeﬃcients hi (φ) of X. Note here that we can set g3 = 0 and g5 = 0 without loss of generality, because g3φ can be absorbed into a redeﬁnition of K and g5φ into h4 , that is, K − 2g3φ → K, h4 − g5φ → h4 . 2 /2 with all the other coeﬃcients equal to zero, If we take K = 1 and g4 = MG we recover the standard slow-roll inﬂation in the Einstein gravity driven by the potential V (φ). Identifying φ with the real neutral component of the standard Higgs ﬁeld, we can reproduce various types of Higgs inﬂation by taking each of the remaining functions as follows: K = 1 + κφ2n ,

running kinetic inﬂation,101

(145)

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

h3 (φ) =

φ , M4

2 MG ξ − φ2 , 2 2 1 h4 (φ) = , 2µ2

g4 (φ) =

II-255

Higgs G-inﬂation,102

(146)

original Higgs inﬂation,103–105

(147)

new Higgs inﬂation,98

(148)

φ , running Einstein inﬂation.99 (149) M6 Thus all types of Higgs inﬂation models can be expressed in the framework of generalized G-inﬂation. Furthermore, combinations of these models can also be analyzed coherently in this context.99 h5 (φ) =

11. Power Spectrum of Perturbations in Generalized G-inflation Next we incorporate scalar and tensor perturbations into the homogeneous solution again based on Ref. 56. To do this we consider the perturbed metric ds2 = −N 2 dt2 + γij (dxi + N i dt)(dxj + N j dt), in the unitary gauge φ = φ(t), where N = 1 + α,

Ni = ∂i β,

γij = a (t)e 2

2ζ

1 δij + hij + hik hkj . 2

(150)

(151)

Here α, β, and ζ are scalar perturbations and hij is a tensor perturbation satisfying √ hii = 0 = hij,j . With the above deﬁnition of the perturbed metric, −g does not contain hij up to second-order, and the coeﬃcients of ζ 2 and αζ vanish, thanks to the background equations. 11.1. Tensor perturbations The quadratic action for the tensor perturbations is found to be

1 FT (2) ST = dtd3 x a3 GT h˙ 2ij − 2 (∇hij )2 , 8 a

(152)

where ¨ 5X + G5φ )], FT = 2[G4 − X(φG

(153)

˙ 5X − G5φ )]. GT = 2[G4 − 2XG4X − X(H φG

(154)

One may notice that GT can also be expressed as 1 ∂Pi . 2 i=2 ∂ H˙

(155)

FT . GT

(156)

5

GT = The squared sound speed is given by

c2T =

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-256

One sees from the action (152) that ghost and gradient instabilities are avoided provided that FT > 0,

GT > 0.

(157)

Note that c2T is not necessarily unity in the general case, contrary to the standard inﬂation models. To canonically normalize the tensor perturbation, we deﬁne cT a dyT = dt, zT = (FT GT )1/4 , vij = zT hij , (158) a 2 and then the quadratic action is written as

1 zT 2 (2) 3 2 2 ST = dyT d x (vij ) − (∇vij ) + v , 2 zT ij

(159)

where a prime denotes diﬀerentiation with respect to yT . In terms of the Fourier wavenumber k, sound horizon crossing occurs when k 2 = zT /zT ∼ 1/yT2 for each mode. On superhorizon scales, the two independent solutions to the perturbation equation that follows from the action (159) are dyT . (160) vij ∝ zT and zT zT2 In terms of the original variables, the two independent solutions on superhorizon scales are given by t dt hij = const. and . (161) a3 GT The second solution corresponds to a decaying mode. 2 ˙ To evaluate the primordial power spectrum, let us assume that = −H/H const., fT =

F˙ T G˙T const. and gT = const. HFT HGT

(162)

We also deﬁne the variation parameter of the sound velocity of tensor perturbations as sT =

c˙T 1 = (fT − gT ). HcT 2

(163)

Clearly only two of the three parameters are independent. We additionally impose conditions gT fT + > 0, 2 2

(164)

3 − + gT > 0.

(165)

1−−

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-257

The former (equivalent to 1 > + sT ) guarantees that the time coordinate yT runs from −∞ to 0 as the universe expands. The latter implies that the second solution in (161) indeed decays. We see that zT can be written as

1/2−νT (−yT ) 3/4 F ∗ 1 (−yT ∗ ) zT = T1/4 , fT gT H (−y ) T∗ 2GT ∗ ∗ 1−− + 2 2

(166)

where the quantities with ∗ are those evaluated at some reference time yT = yT ∗ . The normalized mode solution to the perturbation equation is given in terms of the Hankel function: √ π√ vij = −yHν(1) (−kyT ) eij , (167) T 2 where νT =

3 − + gT , 2 − 2 − fT + gT

(168)

and eij is a polarization tensor. Notice that the conditions (164) and (165) guarantee the positivity of νT . On superhorizon scales, −kyT 1, we obtain Γ(νT ) (−yT )1/2−νT 3/2−νT k 3/2 hij ≈ 2νT −2 k eij . 3 zT Γ 2 Thus, we ﬁnd the power spectrum of the primordial tensor perturbation: 1/2 GT H 2 PT = 8γT 3/2 2 , FT 4π −kyT =1

(169)

(170)

where γT = 22νT −3 |Γ(νT )/Γ(3/2)|2 (1 − − fT /2 + gT /2). The tensor spectral tilt is given by nT = 3 − 2νT .

(171)

Contrary to the predictions of the conventional inﬂation models, a blue spectrum nT > 0 can be obtained if the following condition is satisﬁed: 4 + 3fT − gT < 0.

(172)

This condition is easily compatible with conditions (164) and (165). Thus, positive and large gT compared with and fT can lead to a blue spectrum of tensor perturbations. In deriving the above formulas, we only assumed that , fT , and gT are constant. These parameters may not necessarily be very small as long as the inequalities (164) and (165) are satisﬁed under a sensible background solution.

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-258

11.2. Scalar perturbations We now focus on scalar ﬂuctuations setting hij = 0. Plugging the perturbed metric into the action and expanding it to second-order, we obtain

FT (2) 3 3 SS = dtd xa −3GT ζ˙2 + 2 (∇ζ)2 + Σα2 a ∇2 ∇2 ∇2 ˙ ˙ − 2Θα 2 β + 2GT ζ 2 β + 6Θαζ − 2GT α 2 ζ , (173) a a a where ˙ Σ = XKX + 2X 2 KXX + 12H φXG 3X ˙ 2 G3XX − 2XG3φ − 2X 2 G3φX − 6H 2 G4 + 6H φX + 6[H 2 (7XG4X + 16X 2 G4XX + 4X 3 G4XXX ) ˙ 4φ + 5XG4φX + 2X 2 G4φXX )] − H φ(G 3˙ 2 ˙ + 30H 3 φXG 5X + 26H φX G5XX

˙ 3 G5XXX − 6H 2 X(6G5φ + 9XG5φX + 2X 2 G5φXX ), + 4H 3 φX

(174)

2 ˙ ˙ ˙ Θ = −φXG 3X + 2HG4 − 8HX G4X − 8HX G4XX + φG4φ + 2X φG4φX 2 ˙ − H 2 φ(5XG 5X + 2X G5XX ) + 2HX (3G5φ + 2XG5φX ).

(175)

It is interesting to see that even in the most generic case, some of the coeﬃcients are given by FT and GT , i.e. the functions characterizing the tensor perturbation, and only two new functions show up in the scalar quadratic action. Note that the following relations hold: Σ=X

5 ∂Ei i=2

1 ∂Ei + H , ∂X 2 i=2 ∂H 5

1 ∂Ei , 6 i=2 ∂H

(176)

5

Θ=−

(177)

which compactify the above lengthy expressions. Varying the action (173) with respect to α and β, we obtain the constraint equations Σα − Θ

2 ∇2 ˙ − GT ∇ ζ = 0, β + 3Θ ζ a2 a2 Θα − GT ζ˙ = 0.

(178) (179)

Using the constraint equations, we eliminate α and β from the action (173) and ﬁnally arrive at

FS (2) 3 3 2 2 ˙ SS = dtd x a GS ζ − 2 (∇ζ) , (180) a

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-259

where 1 d a 2 G − FT , a dt Θ T Σ GS = 2 GT2 + 3GT . Θ

FS =

(181) (182)

The analysis of the curvature perturbation hereafter is completely parallel to that of the tensor perturbation. The squared sound speed is given by c2S = FS /GS , and ghost and gradient instabilities are avoided as long as FS > 0,

GS > 0.

(183)

In the case of standard potential driven inﬂation and k-inﬂation, where G3 = 0 = 2 2 /2, we ﬁnd FS = MG . This means that positive H˙ is prohibited G5 and G4 = MG 106 In the generalized G-inﬂation, however, we may by the stability requirement. realize positive H˙ without violating the stability condition. This point was already clear in G-inﬂation and kinetic gravity braiding for which G3 = 0.55,107 Stable cosmological solutions with H˙ > 0 oﬀer a radical and very interesting scenario of the earliest universe.108–111 Using the new variables dyS =

cS dt, a

zS =

√ 1/4 2a (FS GS ) ,

u = zS ζ,

(184)

the curvature perturbation is canonically normalized and the action is now given by

1 z (2) SS = dyS d3 x (u )2 − (∇u)2 + S u2 , (185) 2 zS where a prime denotes diﬀerentiation with respect to yS . Each perturbation mode exits the sound horizon when k 2 = zS /zS ∼ 1/yS2 , where k is the Fourier wavenumber. The two independent solutions on superhorizon scales are ζ = const. and

t

dt . a3 GS

(186)

During inﬂation, it may be assumed that GS is slowly varying. In this case, the second solution decays rapidly. Note however that the nontrivial dynamics of the scalar ﬁeld can induce a temporal rapid evolution of GS , which would aﬀect the superhorizon behavior of the curvature perturbation through the contamination of the second mode in the same way as in Refs. 112–115. Given the speciﬁc background dynamics, one can evaluate such an eﬀect using our general formulas. Closely following the procedure above for the tensor perturbations, we now evaluate the power spectrum of the primordial curvature perturbations. We assume

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-260

that const.,a fS =

F˙ S const., HFS

gS =

G˙S const. HGS

(187)

and then deﬁne νS =

3 − + gS . 2 − 2 − fS + gS

(188)

The power spectrum is given by

1/2 γS GS H 2 Pζ = , 2 F 3/2 4π 2 S −kyS =1

(189)

where γS = 22νS −3 |Γ(νS )/Γ(3/2)|2 (1 − − fS /2 + gS /2). The spectral index is ns − 1 = 3 − 2νS .

(190)

An exactly scale-invariant spectrum is obtained if 3 1 + fS − gS = 0. (191) 4 4 Here again, , fS , and gS are not necessarily very small (as long as ns − 1 0). Taking now the limit , fT , gT , fS , gS 1, the tensor-to-scalar ratio is given by 3/2 −1/2 FS GS F S cS r = 16 = 16 . (192) FT GT F T cT In the case of potential-driven slow-roll inﬂation considered in Sec. 10.3, we need have to leading order in slow-roll ˙ X 4φX (K + 6H 2 h4 ) + (h3 + H 2 h5 ), 2 H H ˙ X 6φX GS 2 (K + 6H 2 h4 ) + (h3 + H 2 h5 ) H H

FS

(193) (194)

2 ˙ and FT GT 2g4 , where we used the slow-roll equation 2g4 + g˙ 4 /H φJ/2H . 2 In this case, we have cT 1 and nT −(2+gT ) with fT gT g˙ 4 /(Hg4 ). If the K 2 ˙ ) g4 (2+gT ), or h4 term dominates in J, we have c2S 1 and FS GS J φ/(2H which yields the standard consistency relation:

r −8nT .

(195)

It is often argued that a negative nT is a generic prediction of inﬂationary cosmology, and that if observationally a positive nT was conﬁrmed, it would falsify inﬂation. It is clear from the above discussion that such a statement is incorrect. The signature of tensor spectral index is not a discriminant of inﬂation and exotic cosmology but simply that of standard inﬂation and generalized G-inﬂation. a By

defining the variation parameter of the sound velocity of scalar perturbations as sS = c˙ S /(HcS ) = (fS − gS )/2, the formulae with fS and/or gS can be rewritten in terms of cS .

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-261

12. Inflationary Cosmology and Observations Various predictions of inﬂationary cosmology are now confronting observational tests using CMB anisotropy and large-scale structures. Since the ﬁrst-year WMAP results were disclosed,52,116,117 the cosmological parameters of homogeneous and isotropic universe as well as properties of ﬂuctuations have been determined quite accurately with well chaaracterized error bars. Indeed, until the WMAP papers were published, although we had the concordance cosmology, we did not have any reliable error estimates because the concordance values of the cosmological parameters had been obtained by combining a number of diﬀerent observational results and we did not have means to calculate error bars to make a sensible weighted average. Current precision cosmology makes use of the Markov Chain Monte Carlo (MCMC) method to constrain values of cosmological parameters as well as the amplitude and spectrum of primordial ﬂuctuations using temperature and E-mode polarization data of CMB.116 Note, however, that the ﬁnal values of these parameters as well as the magnitude of errors obtained by these procedures change considerably depending on which observational datasets one uses as well as which parameters are included in the model. Recent cosmological data as a whole show good agreement with spatially ﬂat universe with a nonvanishing cosmological constant (Λ) and cold dark matter (CDM) based on simple single-ﬁeld inﬂation paradigm that predicts almost scaleinvariant adiabatic ﬂuctuations. Thus it makes sense to estimate various parameters of ΛCDM model using MCMC techniques. Below we quote results from the seven- and nine-year WMAP118,119 denoted respectively by W7 and W9 as well as Planck 201353,120 (P13) and 2015121,122 (P15) results to give an idea about what constraints are likely to survive in future independent of the subtleties of each experiment and accumulation of the data. Abbreviations in parentheses after each quoted number refers to the dataset used, whose meaning is self-explanatory. First as the most important test of the global spacetime, the spatial curvature is constrained as −0.0133 < ΩK0 ≡ −

K a20 H02

−0.079 < ΩK0 < 0.007 −0.0065 < ΩK0 < 0.0012 −0.09 < ΩK0 < 0.001 −0.0075 < ΩK0 < 0.0052 −0.019 < ΩK0 < 0.011 −0.0039 < ΩK0 < 0.0048

< 0.0084 (W7HSTBAO 95%CL), (W9 68%CL), (W9eCMBBAOH0 68%CL), (P13WPhighL 95%CL), (P13lensWPhighLBAO 95%CL), (P15TTTEEElens 95%CL), (P15TTTEEElensExt 95%CL),

which is consistent with (9). Hence from now on we ﬁx ΩK0 = 0.

(196)

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-262

The amplitude and the spectral index of curvature perturbations measured by WMAP read PR (k0 ) = (2.43 ± 0.11) × 10−9 ,

ns = 0.967 ± 0.014 (W7 68%CL),

PR (k0 ) = (2.430 ± 0.091) × 10−9 , ns = 0.968 ± 0.012 (W7HSTBAO 68%CL), (197) PR (k0 ) = (2.41 ± 0.10) × 10−9 , ns = 0.972 ± 0.013 (W9 68%CL), +0.078 ) × 10−9 , ns = 0.971 ± 0.010 (W9BAOH0 68%CL), PR (k0 ) = (2.427−0.079

(198) where the pivot scale is taken as k0 = 0.002 Mpc−1 . Planck, on the other hand, takes the pivot scale at k0 = 0.05 Mpc−1 and ﬁnds +0.053 PR (k0 ) = (2.195−0.059 ) × 10−9 ,

ns = 0.9603 ± 0.0073 (P13WP 68%CL),

+0.056 ) × 10−9 , PR (k0 ) = (2.200−0.054

ns = 0.9608 ± 0.0054 (P13WPhighLBAO 68%CL), (199)

PR (k0 )

=

+0.076 (2.207−0.073 )

× 10

−9

,

ns = 0.9645 ± 0.0049 (P15TTTEEElowP 68%CL),

+0.076 ) × 10−9 , PR (k0 ) = (2.204−0.074

ns = 0.9673 ± 0.0045 (P15TTlowPBAO 68%CL). (200)

If we allow scale-dependence of the spectral index, the values change considerably as +0.050 , ns = 1.027−0.051

ns = 1.008 ± 0.042, ns = 1.009 ± 0.049, ns = 1.020 ± 0.029,

dns = −0.034 ± 0.026 (W7 68%CL), d ln k dns = −0.022 ± 0.020 (W7HSTBAO 68%CL), d ln k

(201)

dns = −0.019 ± 0.025 (W9 68%CL), d ln k dns = −0.023 ± 0.011 (W9eCMBBAOH0 68%CL). d ln k (202)

These numbers refer to the values at k0 = 0.002 Mpc−1 . Planck ﬁnds dns = −0.013 ± 0.009 (P13WP 68%CL), d ln k dns = −0.011 ± 0.008 (P13lensWPhighL 68%CL), d ln k

(203)

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

dns = −0.0057 ± 0.0071 (P15TTTEEElowP 68%CL), d ln k dns = −0.0033 ± 0.0074 (P15lowPlens 68%CL), d ln k

II-263

(204)

on their pivot scale k0 . Thus there is no need to consider inﬂation models with appreciable running any more.123–125 The amplitude of tensor perturbations, which tells us the energy scale of inﬂation, is also an important quantity to distinguish inﬂation models. If we incorporate tensor perturbations in power-law ΛCDM model the constraints are r < 0.36 (W7 95%; no running), r < 0.24 (W7BAOH0 95%; no running), r < 0.38 (W9 95%; no running), r < 0.13 (W9eCMBBAOH0 95%; no running),

(205)

r < 0.11 (P13WPhighL 95%; no running), r < 0.11 (P15TTlowPlens 95%; no running), r < 0.10 (P15TTTEEElowP 95%; no running), using the pivot scale k0 = 0.002 Mpc−1 , whereas if we allow a running spectral index the above constraint relax substantially to r < 0.49 (W7 95%; with running), r < 0.49 (W7BAOH0 95%; with running), r < 0.50 (W9 95%; with running), r < 0.47 (W9eCMBBAOH0 95%; with running),

(206)

r < 0.26 (P13WPhighL 95%; with running), r < 0.18 (P15TTlowP 95%; with running), r < 0.15 (P15TTTEEElowP 95%; with running). Thus the above constraints change signiﬁcantly if we allow running spectral index of curvature perturbation or not, since they are indirectly obtained through the temperature anisotropy and the E-mode polarization. Because of this sensitivity to the nobel assumptions, it is important to constrain r directly through B-mode polarization of CMB, which is not generated by scalar perturbations so that the result does not rely on assumptions concerning the spectrum of curvature perturbations. In March 2014, BICEP2 collaboration announced possible detection of primordial tensor perturbations through their observation of B-mode polarization of CMB at the South Pole,126 favoring a fairly large value of

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-264

r ∼ 0.2, namely high-scale inﬂation. It turned out, however, that their result was signiﬁcantly contaminated by the galactic dust and large r is no longer favored. The joint analysis of BICEP2/Keck Array data and Planck 2015 data was performed obtaining the upper bound, r0.05 < 0.12 at 95% CL.127 Thus the current constraint from B-mode polarization is at the same level as those obtained indirectly from temperature anisotropy and E-mode polarization with the power-law model. These constraints are often graphically displayed in a (ns , r) plane128 with oneand two-sigma contours and compared with various inﬂation models. We do not do the same here for two reasons. First constraints vary appreciably depending on what dataset one uses and what underlying model one adopts as seen above. Second occupying center of the likelihood contour at the current level of observations does not necessarily mean that such a model is the right one, and we should be open minded for further development of the observations. Below we instead calculate these observables in each class of models discussed in Sec. 4.

12.1. Large-field models Consider a massive scalar model as an example of large-ﬁeld models. The number of e-folds of inﬂation after φ crossed φN is given from (31) as 1 N∼ = 4

φN MG

2 .

(207)

Hence the slow-roll parameters are V = ηV = 2

MG φN

2 =

1 , 2N

ξV = 0

(208)

which gives the amplitude of curvature perturbation, (95), PR (k) =

1 6π 2

mN MG

2 .

(209)

Setting N = 55 at the pivot scale tentatively, we ﬁnd m = 1.6 × 1013 GeV and the quartic coupling is constrained as λ 8 × 10−13 . The spectral index and its running are given by ns = 1 −

2 = 0.964, N

dns 2 = − 2 = −6.6 × 10−4 , d ln k N

(210)

in agreement with the observations. The amplitude of tensor perturbation, r = 16 = 0.15, is so large that it is now strongly disfavored by observations.127 It is remarkable that the simplest and most attractive model of inﬂation is now about to be ruled out by the observation.

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-265

Let us also consider R2 inﬂation in the Einstein frame as a kind of large-ﬁeld models. From (23), (24) and (27), we ﬁnd 3 √2 4 1 3 ∼ √2 N∼ , = e 3 κφN , V = = κφ 4 3 (e 3 N − 1)2 4N 2 √2 (211) 4 e 3 κφN − 2 ∼ 1 √ ηV = − =− , 3 (e 23 κφN − 1)2 N which yield ns = 1 −

2 = 0.964, N

r=

12 = 4.0 × 10−3 , N2

(212)

for N = 55. 12.2. Small-field model Next we consider a small ﬁeld model (32). From V [φ] = parameters are given by V =

2 2 φ 8MG , 2 (φ − v 2 )2

ηV =

2 (3φ2 − v 2 ) 4MG , 2 (φ − v 2 )2

λ 2 4 (φ

ξV =

− v 2 )2 the slow-roll

4 2 φ 96MG . 2 (φ − v 2 )3

(213)

Since inﬂation ends when H = V = 1, the ﬁeld value, φf , at that time is given by √ 2 2 + 8M 2 v 2 ∼ (β 2 − 2 2β)M 2 , φ2f = v 2 + 4MG − 16MG (214) = G G where we have set v ≡ βMG and the last approximate equality holds when β 10. The number for e-folds from φ = φN to φf reads φf dφ β2 φf 1 β2 1 φf 2 2 N= H = ln − − ln . (215) 2 (φf − φN ) 4 4 φN 8MG φN 2 φ˙ φN The last approximation is valid when β 10 − 20. The amplitude of curvature perturbation is given by 2 λ(φ2N − v 2 )4 MG λβ 8 8N λβ 8 √ PR (k0 ) = exp + 1 , 6 φ2 768π 2 MG 768π 2 φN β2 768π 2 (β − 2)2 N (216) where the second and third approximations are based on φN v and φf (β − √ 2)MG , respectively. The spectral index and its running are given by ns − 1 = −

2 8(3φ2N + v 2 )MG 8 − 2, 2 2 2 (φN − v ) β

4 dns (320v 2 φ2N + 192φ4N )MG 320 =− − 6 2 2 4 d ln k (φN − v ) β

(217) φN MG

2 .

(218)

May 4, 2017 9:54 b2167-ch18

II-266

K. Sato and J. Yokoyama

For β = 15 and N = 55, for example, we ﬁnd ns = 0.964 with negligibly small running, and the self-coupling is given by λ 7 × 10−14 . 12.3. Hybrid inflation model Here, we consider the case where vacuum energy drives inﬂation, which is distinct from the above two classes. The potential in this regime is approximated as V [φ] = V0 +

m2 2 φ 2

and so the slow-roll parameters are given by 2 1 m 4 φ m2 V = , ηV = , 18 H MG 3H 2

(219)

ξV = 0.

(220)

In this case one cannot determine the energy scale of inﬂation from the amplitude of curvature ﬂuctuations alone. The spectral index is usually determined by ηV as ns − 1 2ηV =

2m2 > 1, 3H 2

(221)

which is inconsistent with current observations. The running spectral index takes a small value 3 2 2 dns 2m2 φ φ 3 16V ηV = (ns − 1) (222) d ln k 3H 2 MG MG so does the tensor-to-scalar ratio, 2 2 2 2m2 φ φ 2 r=2 (n − 1) . s 3H 2 MG MG

(223)

One should note that if the second term dominates over the vacuum energy in (219), the prediction becomes closer to a large-ﬁeld model and the spectral index can be smaller than unity. 12.4. Noncanonical models and multi-field models So far we have focused on the amplitude and spectral index of primordial curvature and tensor perturbations. Indeed in single-ﬁeld slow-roll inﬂation there is only one ﬂuctuating degree of freedom, so only adiabatic ﬂuctuations are generated with no isocurvature counterparts. Furthermore, the curvature perturbation generated in slow-roll inﬂation behaves as a practically free massless scalar ﬁeld during inﬂation, and so its vacuum ﬂuctuation is Gaussian distributed. This is why the power spectrum fully quantiﬁes the properties of ﬂuctuations. In some classes of noncanonical inﬂation models including DBI inﬂation,129,130 ghost inﬂation,131 k-inﬂation54 and G-inﬂation,55 the inﬂaton may change rapidly and its interaction may not be negligible. In such circumstances deviation from

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-267

Gaussian statistics may be signiﬁcant.90,132–135 Furthermore, in models with multiple ﬂuctuating ﬁelds besides the inﬂaton, one may also expect large deviation from Gaussianity.136–138 Much work was done in recent years to quantify the skewness or the bispectrum of the temperature anisotropy, and a number of contrived models were proposed that predicted an observationally falsiﬁable magnitude of the deviation from Gaussianity by Planck.139 The constraints on non-Gaussianity provided by analyzing the Planck data140 falsiﬁed most of these models that had been proposed predicting large bispectral non-Gaussianity; therefore, we do not discuss this issue in more detail. Another possible signature of multi-ﬁeld models would be the presence of isocurvature modes. But again CMB observations do not need any isocurvature components at least on relevant scales.119 13. Conclusion The fact that we live in a big and old universe with a large amount of entropy can be regarded as a primary consequence of inﬂation in the early universe. Any attempt to avoid inﬂation by providing alternative explanations to the other predictions of inﬂationary cosmology, namely, generation of adiabatic density perturbation with nearly scale-invariant spectrum as well as tensor perturbation should also have sound explanation to the aforementioned global properties of the spacetime. Some people, however, try to refute this viewpoint claiming that the global properties of the universe, namely the homogeneity, isotropy, and longevity had been known long before the ﬁrst inﬂation models13,16,17 were proposed so that they cannot be regarded as the predictions of inﬂationary cosmology. While this statement is correct, one should also recognize that theories in physics have a double role. The ﬁrst is to make new prediction which may be conﬁrmed or falsiﬁed by experiment and observation. The second is to provide sensible explanations to known mysteries of the nature. Inﬂationary cosmology certainly provides a simple mechanism to explain fundamental properties of the global structure of the space, playing the latter role of theoretical physics. Thus the above mentioned criticism is unjustiﬁed. Furthermore, the simplest class of slow-roll inﬂation models predicted nearly scale-invariant, adiabatic, and Gaussian curvature and density ﬂuctuations, all of which are now being conﬁrmed by dedicated CMB experiments. In particular, the observed angular correlation between temperature anisotropy and E-mode polarization clearly indicates that the observed ﬂuctuations were generated as super-Hubble ﬂuctuations as predicted by inﬂation. We are tempted to interpret the fact that Planck did not conﬁrm any sizable non-Gaussianity140 despite the recent eﬀorts of many theorists to make models of non-Gaussian ﬂuctuations as indicating that nature preferred the simplest inﬂation models. This situation, on the other hand, highlights the diﬃculty of singling out the right model responsible for our universe by exploiting the small numbers of

May 4, 2017 9:54 b2167-ch18

K. Sato and J. Yokoyama

II-268

observational clues available. For example, among the models occupying central regions of the likelihood contours in (ns , r) plane now, it is very diﬃcult to distinguish between R2 model17,141,142 and a Higgs inﬂation model with a large and negative nonminimal coupling103–105 using observations of perturbation variables alone. We should combine with the analysis of the reheating processes of these models for more detailed comparison, which may be observationally probed by future space-based laser interferometers such as DECIGO.143,144 On the other hand, the latest observational upper limits on B-mode polarization of CMB is ruling out the simplest inﬂation model, namely, chaotic inﬂation with a massive scalar ﬁeld.127 This is somehow a shocking result since we also have sensible particle physics models for this inﬂation mechanism. This may indicate that our understanding of the gravity sector at such high energy scale may be wrong, as there is no reason believe that the Planck scale remains intact up to the Planck scale. Acknowledgments Part of the present chapter is an update of the review paper written by JY21 reproduced here through the Creative Commons license. This paper is also partially based on a Japanese text book written by JY and edited by KS.145 JY is grateful to Alexei A. Starobinsky for recounting the history of developments in Russia and to Tsutomu Kobayashi and Masahide Yamaguchi for fruitful collaboration on generalized G-inﬂation. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20.

G. Gamow, Phys. Rev. 70 (1946) 572. E. Hubble, Proc. Natl. Acad. Sci. 15 (1929) 168. G. Lemaˆıtre, Ann. Societe Sci. Brux. 47 (1927) 49. A. A. Penzias and R. W. Wilson, Astrophys. J. 142 (1965) 419. R. A. Alpher, H. Bethe and G. Gamow, Phys. Rev. 73 (1948) 803. C. Hayashi, Prog. Theor. Phys. 5 (1950) 224. R. H. Dicke and P. J. E. Peebles, The big bang cosmology: Enigmas and nostrums, in General Relativity, An Einstein Centenary Survey, Chap. 9 (Cambridge, 1979). H. Georgi and S. L. Glashow, Phys. Rev. Lett. 32 (1974) 438. T. W. B. Kibble, J. Phys. A 9 (1976) 1387. J. Preskill, Phys. Rev. Lett. 43 (1365) 1979. G. ’t Hooft, Nucl. Phys. B 79 (1974) 276. A. M. Polyakov, JETP Lett. 20 (1974) 194. K. Sato, Mon. Not. R. Astron. Soc. 195 (1981) 467. K. Sato, Phys. Lett. B 99 (1981) 66. M. B. Einhorn and K. Sato, Nucl. Phys. B180 (1981) 385. A. H. Guth, Phys. Rev. D 23 (1981) 347. A. A. Starobinsky, Phys. Lett. B 91 (1980) 99. Class for Physics of the Royal Swedish Academy of Sciences, AAPPS Bull. 11 (2011). A. D. Linde, Phys. Lett. B 108 (1982) 389. A. Albrecht and P. J. Steinhardt, Phys. Rev. Lett. 48 (1982) 1220.

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69.

II-269

J. Yokoyama, Inﬂation: 1980-201X, PTEP 2014(6) (2014) 06B103. S. K. Blau and A. H. Guth, Inﬂationary Cosmology (1986). A. H. Guth, Inﬂation and cosmological perturbations (2003), 725. S. Weinberg, Phys. Rev. Lett. 19 (1967) 1264. A. Salam, Conf. Proc. C 680519 (1968) 367. K. Sato, Prog. Theor. Phys. 54 (1975) 1325. K. Hirata et al., Phys. Rev. Lett. 58 (1987) 1490. K. Sato and H. Sato, Prog. Theor. Phys. 54 (1975) 912. K. Sato and H. Sato, Prog. Theor. Phys. 54 (1975) 1564. K. Sato and T. Nakamura, Prog. Theor. Phys. 55 (1976) 1133. K. Sato and T. Nakamura, Prog. Theor. Phys. 55 (1976) 978. M. Yoshimura, Phys. Rev. Lett. 41 (1978) 281. D. A. Kirzhnits and A. D. Linde, Phys. Lett. B 42 (1972) 471. D. A. Kirzhnits and A. D. Linde, Sov. Phys. JETP 40 (628) 1975. L. Dolan and R. Jackiw, Phys. Rev. D 9 (1974) 3320. S. Weinberg, Phys. Rev. D 9 (1974) 3357. K. Sato, M. Sasaki, H. Kodama and K. Maeda, Prog. Theor. Phys. 65 (1443) 1981. K. Sato, H. Kodama, M. Sasaki and K. Maeda, Phys. Lett. B 108 (1982) 103. ´ B. Gliner, Sov. J. Exp. Theor. Phys. 22 (1966) 378. E. G. Lemaˆıtre, Proc. Natl. Acad. Sci. 20 (1934) 12. L. E. Gurevich, Astrophys. Space Sci. 38 (1975) 67. ´ B. Gliner and I. G. Dymnikova, Soviet Astron. Lett. 1 (93) 1975. E. A. A. Starobinskii, Sov. Astron. Lett. 4 (1978) 82. A. D. Linde, Phys. Lett. B 129 (1983) 177. A. A. Starobinsky, JETP Lett. 30 (1979) 682 (in Russian). V. F. Mukhanov and G. V. Chibisov, JETP Lett. 33 (1981) 532 (in Russian). L. Boubekeur and D. H. Lyth, J. Cosmol. Astropart. Phys. 0507 (010) 2005. A. A. Starobinsky, Phys. Lett. B 117 (1982) 175. S. W. Hawking, Phys. Lett. B 115 (1982) 295. A. H. Guth and S. Y. Pi, Phys. Rev. Lett. 49 (1982) 1110. G. F. Smoot et al., Astrophys. J. 396 (1992) L1. H. V. Peiris et al., Astrophys. J. Suppl. 148 (213) 2003. P. A. R. Ade et al., Planck 2013 results. XXII. Constraints on inﬂation (2013). C. Armendariz-Picon, T. Damour and V. F. Mukhanov, Phys. Lett. B 458 (1999) 209. T. Kobayashi, M. Yamaguchi and J. Yokoyama, Phys. Rev. Lett. 105 (2010) 231302. T. Kobayashi, M. Yamaguchi and J. Yokoyama, Prog. Theor. Phys. 126 (2011) 511. A. Kashlinsky, I. I. Tkachev and J. Frieman, Phys. Rev. Lett. 73 (1994) 1582. L. A. Kofman, A. D. Linde and A. A. Starobinsky, Phys. Lett. B 157 (1985) 361. D. S. Salopek, Phys. Rev. Lett. 69 (1992) 3602. A. Vilenkin, Phys. Rev. Lett. 72 (1994) 3137. A. D. Linde, Phys. Lett. B 327 (1994) 208. A. D. Linde, Phys. Rev. D 49 (1994) 748. D. H. Lyth, Phys. Rev. Lett. 78 (1997) 1861. N. Tetradis, Phys. Rev. D 57 (1998) 5997. L. E. Mendes and A. R. Liddle, Phys. Rev. D 62 (2000) 103511. J. Yokoyama, Phys. Lett. B 212 (1988) 273. J. Yokoyama, Phys. Lett. B 231 (1989) 49. J. Yokoyama, Phys. Rev. Lett. 63 (1989) 712. K. Kamada, K. Nakayama and J. Yokoyama, Phys. Rev. D 85 (2012) 043503.

May 4, 2017 9:54 b2167-ch18

II-270

K. Sato and J. Yokoyama

70. K. Kamada, Y. Miyamoto and J. Yokoyama, J. Cosmol. Astropart. Phys. 1210 (2012) 023. 71. J. H. Traschen and R. H. Brandenberger, Phys. Rev. D 42 (1990) 2491. 72. L. Kofman, A. D. Linde and A. A. Starobinsky, Phys. Rev. Lett. 73 (1994) 3195. 73. L. Kofman, A. D. Linde and A. A. Starobinsky, Phys. Rev. D 56 (1997) 3258. 74. L. F. Abbott, E. Farhi and M. B. Wise, Phys. Lett. B 117 (1982) 29. 75. A. Albrecht, P. J. Steinhardt, M. S. Turner and F. Wilczek, Phys. Rev. Lett. 48 (1437) 1982. 76. J. Yokoyama, Phys. Rev. D 70 (2004) 103511. 77. M. S. Turner, Phys. Rev. D 28 (1983) 1243. 78. L. H. Ford, Phys. Rev. D 35 (1987) 2955. 79. T. Kunimitsu and J. Yokoyama, Phys. Rev. D 86 (2012) 083541. 80. L. Parker, Phys. Rev. 183 (1969) 1057. 81. Ya. B. Zeldovich and A. A. Starobinsky, Sov. Phys. JETP 34 (1972) 1159. 82. T. S. Bunch and P. C. W. Davies, Proc. R. Soc. Lond. A 360 (1978) 117. 83. A. Vilenkin and L. H. Ford, Phys. Rev. D 26 (1982) 1231. 84. A. D. Linde, Phys. Lett. B 116 (1982) 335. 85. D. Polarski and A. A. Starobinsky, Class. Quantum Grav. 13 (1996) 377. 86. A. Vilenkin, Nucl. Phys. B 226 (1983) 527. 87. J. M. Bardeen, Phys. Rev. D 22 (1980) 1882. 88. H. Kodama and M. Sasaki, Prog. Theor. Phys. Suppl. 78 (1984) 1. 89. V. F. Mukhanov, H. A. Feldman and R. H. Brandenberger, Phys. Rep. 215 (1992) 203. 90. D. Seery and J. E. Lidsey, J. Cosmol. Astropart. Phys. 0506 (2005) 003. 91. A. A. Starobinsky, JETP Lett. 42 (1985) 152. 92. M. Sasaki and E. D. Stewart, Prog. Theor. Phys. 95 (1996) 71. 93. V. A. Rubakov, M. V. Sazhin and A. V. Veryaskin, Phys. Lett. B 115 (1982) 189. 94. L. F. Abbott and M. B. Wise, Nucl. Phys. B 244 (1984) 541. 95. G. W. Horndeski, Int. J. Theor. Phys. 10 (1974) 363. 96. A. Nicolis, R. Rattazzi and E. Trincherini, Phys. Rev. D 79 (2009) 064036. 97. C. Deﬀayet, X. Gao, D. A. Steer and G. Zahariade, Phys. Rev. D 84 (2011) 064039. 98. C. Germani and A. Kehagias, Phys. Rev. Lett. 105 (2010) 011302. 99. K. Kamada, T. Kobayashi, T. Takahashi, M. Yamaguchi and J. Yokoyama, Phys. Rev. D 86 (2012) 023504. 100. M. Kawasaki, M. Yamaguchi and T. Yanagida, Phys. Rev. Lett. 85 (2000) 3572. 101. K. Nakayama and F. Takahashi, J. Cosmol. Astropart. Phys. 1011 (2010) 009. 102. K. Kamada, T. Kobayashi, M. Yamaguchi and J. Yokoyama, Phys. Rev. D 83 (2011) 083515. 103. J. L. Cervantes-Cota and H. Dehnen, Nucl. Phys. B 442 (1995) 391. 104. F. L. Bezrukov and M. Shaposhnikov, Phys. Lett. B 659 (2008) 703. 105. A. O. Barvinsky, A. Yu. Kamenshchik and A. A. Starobinsky, J. Cosmol. Astropart. Phys. 0811 (2008) 021. 106. J. Garriga and V. F. Mukhanov, Phys. Lett. B 458 (1999) 219. 107. C. Deﬀayet, O. Pujolas, I. Sawicki and A. Vikman, J. Cosmol. Astropart. Phys. 1010 (2010) 026. 108. P. Creminelli, M. A. Luty, A. Nicolis and L. Senatore, J. High Energy Phys. 0612 (080) 2006. 109. P. Creminelli, A. Nicolis and E. Trincherini, J. Cosmol. Astropart. Phys. 1011 (2010) 021. 110. D. Pirtskhalava and L. Santoni, J. High Energy Phys. 1412 (2014) 151.

May 4, 2017 9:54 b2167-ch18

Inflationary cosmology: first 30+ years

II-271

111. T. Kobayashi, M. Yamaguchi and J. Yokoyama, Galilean Creation of the Inflationary Universe (2015). 112. O. Seto, J. Yokoyama and H. Kodama, Phys. Rev. D 61 (2000) 103504. 113. S. M. Leach and A. R. Liddle, Phys. Rev. D 63 (2001) 043508. 114. S. Inoue and J. Yokoyama, Phys. Lett. B 524 (2002) 15. 115. Y. Takamizu and J. Yokoyama, Phys. Rev. D 83 (2011) 043504. 116. C. L. Bennett et al., Astrophys. J. Suppl. 148 (2003) 1. 117. D. N. Spergel et al., Astrophys. J. Suppl. 148 (2003) 175. 118. E. Komatsu et al., Astrophys. J. Suppl. 192 (2011) 18. 119. G. Hinshaw et al., Astrophys. J. Suppl. 208 (2013) 19. 120. P. A. R. Ade et al., Planck 2013 results. XVI. Cosmological parameters, A and AP (2013). 121. P. A. R. Ade et al., Planck 2015 results. XX. Constraints on inﬂation, A and AP (2015). 122. P. A. R. Ade et al., Planck 2015 results. XIII. Cosmological parameters. (2015). 123. M. Kawasaki, M. Yamaguchi and J. Yokoyama, Phys. Rev. D 68 (2003) 023508. 124. M. Yamaguchi and J. Yokoyama, Phys. Rev. D 68 (2003) 123520. 125. M. Yamaguchi and J. Yokoyama, Phys. Rev. D 70 (2004) 023513. 126. P. A. R. Ade et al., BICEP2 I: Detection of B-mode Polarization at degree angular scales (2014). 127. P. A. R. Ade et al., Phys. Rev. Lett. 114(10) (2015) 101301. 128. S. Dodelson, W. H. Kinney and E. W. Kolb, Phys. Rev. D 56 (1997) 3207. 129. M. Alishahiha, E. Silverstein and D. Tong, Phys. Rev. D 70 (2004) 123505. 130. E. Silverstein and D. Tong, Phys. Rev. D 70 (2004) 103505. 131. N. Arkani-Hamed, P. Creminelli, S. Mukohyama and M. Zaldarriaga, J. Cosmol. Astropart. Phys. 0404 (2004) 001. 132. D. Langlois, S. Renaux-Petel, D. A. Steer and T. Tanaka, Phys. Rev. Lett. 101 (2008) 061301. 133. D. Babich, P. Creminelli and M. Zaldarriaga, J. Cosmol. Astropart. Phys. 0408 (2004) 009. 134. T. Kobayashi, M. Yamaguchi and J. Yokoyama, Phys. Rev. D 83 (2011) 103524. 135. A. De Felice and S. Tsujikawa, Phys. Rev. D 84 (2011) 083504. 136. N. Bartolo, S. Matarrese and A. Riotto, Phys. Rev. D 69 (2004) 043503. 137. T. Suyama and M. Yamaguchi, Phys. Rev. D 77 (2008) 023505. 138. D. H. Lyth and Y. Rodriguez, Phys. Rev. Lett. 95 (2005) 121302. 139. N. Bartolo, E. Komatsu, S. Matarrese and A. Riotto, Phys. Rep. 402 (2004) 103. 140. P. A. R. Ade et al., Planck 2013 Results. XXIV. Constraints on primordial nonGaussianity (2013). 141. A. Vilenkin, Phys. Rev. D 32 (1985) 2511. 142. M. B. Mijic, M. S. Morris and W.-M. Suen, Phys. Rev. D 34 (1986) 2934. 143. N. Seto, S. Kawamura and T. Nakamura, Phys. Rev. Lett. 87 (2001) 221103. 144. S. Kawamura et al., Class. Quantum Grav. 23 (2006) S125. 145. J. Yokoyama, Utyuron I, 2nd edn. Modern Astronomy Series, Vol. 2 Nihon Hyoron-sha (2012).

May 4, 2017 9:54 b2167-ch19

Chapter 19 Inflation, string theory and cosmic strings

David F. Chernoﬀ Department of Astronomy, Cornell University, Ithaca, New York, USA chernoﬀ@astro.cornell.edu S.-H. Henry Tye Jockey Club Institute for Advanced Study and Department of Physics, Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong Department of Physics, Cornell University, Ithaca, New York, USA [email protected]

At its very beginning, the universe is believed to have grown exponentially in size via the mechanism of inﬂation. The almost scale-invariant density perturbation spectrum predicted by inﬂation is strongly supported by cosmological observations, in particular the cosmic microwave background (MB) radiation. However, the universe’s precise inﬂationary scenario remains a profound problem for cosmology and for fundamental physics. String theory, the most-studied theory as the ﬁnal physical theory of nature, should provide an answer to this question. Some of the proposals on how inﬂation is realized in string theory are reviewed. Since everything is made of strings, some string loops of cosmological sizes are likely to survive in the hot big bang that followed inﬂation. They appear as cosmic strings, which can have intricate properties. Because of the warped geometry in ﬂux compactiﬁcation of the extra spatial dimensions in string theory, some of the cosmic strings may have tensions substantially below the Planck or string scale. Such strings cluster in a manner similar to dark matter leading to hugely enhanced densities. As a result, numerous fossil remnants of the low tension cosmic strings may exist within the galaxy. They can be revealed through the optical lensing of background stars in the near future and studied in detail through gravitational wave emission. We anticipate that these cosmic strings will permit us to address central questions about the properties of string theory as well as the birth of our universe. Keywords: Inﬂation; string theory and cosmology; cosmic strings; early universe; gravitational microlensing. PACS Number(s): 98.80.Bp, 98.80.Cq, 98.80.Es, 11.25.−w, 11.27.+d

1. Introduction By 1930, quantum mechanics, general relativity and the Hubble expansion of our universe were generally accepted by the scientiﬁc community as three triumphs of modern physical science. Particles and ﬁelds obey quantum mechanical rules and II-273

May 4, 2017 9:54 b2167-ch19

II-274

D. F. Chernoﬀ and S.-H. H. Tye

spacetime bends and warps according to Einstein’s classical description of gravity. The dynamical arena for particles, ﬁelds and spacetime is the universe and modern cosmology is born as the application of these physical laws to the universe itself. The path forward seemed clear but in the 1950s and 1960s physicists ﬁrst began to appreciate that trying to quantize the classical gravitational ﬁeld led to severe inconsistencies. This nonrenormalizability problem was further sharpened once the quantization of gauge (vector) ﬁelds was well understood in early 1970s. At about the same time, string theory — a quantum mechanical theory of onedimensional objects — was found to contain a graviton (a massless spin-2 particle). The promise of a self-consistent theory of quantum gravity has attracted huge attention since the 1980s. Tremendous progress has been made in the past 30 years in understanding many aspects of string theory; yet today we are no closer to knowing which of many theoretical possibilities might describe our universe and we still lack experimental evidence that the theory describes nature. By now, string theory is such a large topic with so many research directions that we have to choose a speciﬁc area to discuss. The centenary of general relativity prompts us to reﬂect on two very amazing manifestations of general relativity: namely black holes and cosmology. Given the tremendous progress in cosmology in the past decades, both in observation and in theory, we shall focus here on the intersection of string theory and cosmology. Our fundamental understanding of low energy physics is grounded in quantum ﬁeld theory and gravity but if we hope to work at the highest energies then we cannot escape the need for an ultraviolet completion to these theories. String the√ α is the ory provides a suitable framework. The string scale, MS = 1/ α where √ Regge slope, is expected to be below the reduced Planck scale MPl = 1/ 8πG 2.4 × 1018 GeV (where G is the Newton’s constant) but not too far from the grand uniﬁed scale MGUT 1016 GeV (where the proton mass is about 1 GeV, in units where c = = 1). Particle physics properties at scales much lower than MS are hard to calculate within our present understanding of string theory. High energy experiments can reach only multi-TeV scales, orders of magnitude below the expected MS . The diﬃculty of ﬁnding a common arena to compare theory and observation is hardly a new dilemma: although standard quantum mechanics works well at and above eV energy scales, it is woefully inadequate to describe properties of proteins which typically involve milli-eV scales and lower. Where should we look? One place to look for stringy eﬀects is in the early universe. If the energy scale of the early universe reaches MS , an abundance of physical eﬀects will provide a more direct path to test our ideas than currently feasible in high energy experiments. In fact, attempts to interpret observations will tell us whether our present understanding of physics and the universe is advanced enough to permit us to ask and answer sensible questions about these remote physical regimes. We shall focus on the inﬂationary universe scenario,1–3 which reaches energy scales suﬃciently high to create ample remnants. The hot big bang’s nucleosynthetic epoch has direct observational support at and below MeV scales whereas the energies of interest to

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-275

us are ∼ MS . We are most fortunate that precision cosmology has begun to let us address questions relevant to these extreme conditions. Even in this rather specialized topic of inﬂationary cosmology and string theory, a comprehensive review by Baumann and McAllister4 appeared recently. So we shall take the opportunity here to present a brief introduction and share within this framework some of our own views on a particularly interesting topic, namely, cosmic strings. The subject of cosmic strings has been extensively studied too,5 so here we shall focus on low tension cosmic strings, which may appear naturally in string theory. Since current gravitational observations are insensitive to quantum eﬀects, one may wonder why an ultraviolet completion of the gravity theory is needed if inﬂation, in fact, occurs several orders of magnitude below the Planck scale. To partly answer this, we like to recall that in 1920s, there was no obvious necessity for a quantum theory consistent with special relativity given that electrons in atoms and materials move at nonrelativistic speed only. However, the Dirac theory predicted anti-matter, the value of the electron’s gyromagnetic moment and a host of interesting properties (e.g. the Lamb shift) that are well within reach of the nonrelativistic study of electrons and atoms. The standard electroweak model, the ultraviolet completion of the Fermi’s weak interaction model, is another example. Among other successes, the standard electroweak model predicted the presence of the neutral current as well as explains the π 0 decay to two photons via anomaly and the basis of the quark–lepton family via the anomaly free constraints. All these phenomena are manifest at energy scales far below the electroweak scale. In the current context, we need to search for stringy eﬀects that are unexpected in classical general relativity and might be observed in the near future at energy scales far below the Planck scale. There are many speciﬁc ways to realize inﬂation within string theory. These may be roughly grouped into models with small and large ﬁeld ranges. They have rather diﬀerent properties and predictions. Recall that 6 of the 9 spatial dimensions must be compactiﬁed into a Calabi–Yau like manifold with volume V6 , where 2 MPl ∼ MS6 V6 1. (1) MS During inﬂation the inﬂaton ﬁeld φ moves from an initial to ﬁnal position within the manifold, covering a path or ﬁeld range ∆φ. Now, the distance scale inside the compactiﬁed manifold is bounded by MPl . The case ∆φ < MPl is known as small ﬁeld inﬂation. The opposite limit when the ﬁeld range far exceeds the typical scale of the manifold, ∆φ MPl , is large ﬁeld inﬂation. In the latter case the inﬂaton wanders inside the compactiﬁed manifold for a while during inﬂation. This scenario may occur if the inﬂaton is an axion-like ﬁeld and executes a helical-like motion. Flux compactiﬁcation of the six spatial dimensions in string theory typically introduces many axions and naturally sets the stage for this possibility. The recent B-mode polarization measurement in the cosmic microwave background (CMB) radiation6 by BICEP2 is most interesting. Primordial B-mode

May 4, 2017 9:54 b2167-ch19

II-276

D. F. Chernoﬀ and S.-H. H. Tye

polarization is sourced by tensor perturbations during inﬂation and measurable primordial B-mode polarization implies that ∆φ MPl . This is a prototypical example of how it is now possible to probe high energies and early moments of the universe’s history by means of cosmological observations. However, since dust can contaminate today’s observed signal,7,8 it remains to be seen whether the BICEP2 observations deﬁnitively imply that large ﬁeld inﬂation takes place. We believe this issue should be settled soon. In the meantime, we shall entertain both possibilities here. In both small and large ﬁeld ranges macroscopic, one-dimensional objects, hereafter cosmic strings, can appear rather naturally. In the former case, the cosmic strings can be the fundamental superstrings themselves, while in the latter case, they can be fundamental strings and/or vortices resulting from the Higgs like mechanism. Cosmic strings with tension above the inﬂation scale will not be produced after inﬂation and we should expect only relatively low tension strings to be produced. Because of the warped geometry and the presence of throats in the compactiﬁcation in string theory, some cosmic strings can acquire very low tensions. In ﬂux compactiﬁcation such strings tend to have nontrivial tension spectrum (and maybe even with beads). If the cosmic string tension is close to the present observational bound (about Gµ < 10−7 ), strings might contribute to the B-mode polarization at large multipoles independent of whether or not BICEP2 has detected primordial B mode signals. The string contribution to the CMB power spectrum of temperature ﬂuctuations is limited to be no more than ∼ 10% and this constrains any string B-mode contribution. Because the primordial tensor perturbation from inﬂation decreases quickly for large multipoles one very informative possibility is that cosmic strings are detected at large and other astrophysical contributions are sub-dominant. On the other hand, if tension is low (Gµ 10−7 ) then the cosmic string eﬀect for all is negligible compared to known sources and foreground contributions. Other signatures must be sought. For small Gµ string loops are long lived and the oldest (and smallest) tend to cluster in our galaxy, resulting in an enhancement of ∼ 105 in the local string density. This enhancement opens up particularly promising avenues for detecting strings by microlensing of stars within the galaxy and by gravitational wave emission in the tension range 10−14 < Gµ 10−7 . As a cosmic string passes in front of a star, its brightness typically doubles, as the two images cannot be resolved. As a string loop oscillates in front of a star, it generates a unique signature of repeated achromatic brightness doubling. Here, we brieﬂy review the various observational bounds on cosmic strings and estimate the low tension cosmic string density within our galaxy as well as the likelihood of their detection in the upcoming observational searches. It is encouraging that searches of extrasolar planets and variables stars also oﬀer a chance to detect the micro-lensing of stars by cosmic strings. Once a location is identiﬁed by such a detection, a search for gravitational wave signals should follow.

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-277

Detection of cosmic strings followed by the measurement of their possible different tensions will go a long way in probing superstring theory. Although a single string tension can easily originate from standard ﬁeld theory, a string tension spectrum should be considered as a distinct signature of string theory. It is even possible that some cosmic strings will move in the compactiﬁed dimensions with warped geometry, which can show up observationally as strings with varying tension, both along their lengths as well as in time. In summary, cosmic strings probably oﬀer the best chance of ﬁnding distinct observational support for the string theory.

2. The Inflationary Universe So far, observational data agrees well with the simplest version of the slow-roll inﬂationary universe scenario, i.e. a single, almost homogeneous and isotropic, scalar inﬂaton ﬁeld φ = φ(t) subject to potential V (φ). For a Friedmann–Lemaˆıtre– Robertson–Walker metric, general relativity yields schematic simple equations for the cosmic scale factor a(t) and for φ, 2 a˙ 1 φ˙ 2 ρc (0) ρm (0) ρr (0) = V (φ) + + 2 + H = + + ···, 2 a 3MPl 2 a a3 a4 2

dV φ¨ + 3H φ˙ = − = −V (φ), dφ

(2)

(3)

where the Hubble parameter H = a/a ˙ (a(t) ˙ = da/dt) measures the rate of expansion of the universe, and ρc (0), ρm (0) and ρr (0) are the curvature, the nonrelativistic matter and the radiation densities at time t = 0.a In an expanding universe, a(t) grows and the curvature, the matter and the radiation terms diminish. The energy and pressure densities for the inﬂaton are ρφ =

φ˙ 2 + V (φ), 2

pφ =

φ˙ 2 − V (φ), 2

(4)

where a canonical kinetic term for φ is assumed. If V (φ) is suﬃciently ﬂat and if φ˙ is initially small then Eq. (3) implies that φ moves slowly in the sense that its kinetic energy remains small compared to the potential energy, φ˙ 2 V (φ). In the limit that V (φ) is exactly constant and dominates in Eq. (2) then H is constant and a(t) ∝ eHt . More precisely one may deﬁne the inﬂationary epoch as that period when the expansion of the universe is accelerating, i.e. a ¨ > 0, or, in a spatially ﬂat energy density has been divided into ρm and ρr . The ultra-relativistic massive particles are lumped into ρr and the marginally relativistic massive particles included in ρm . This division is epoch-dependent and schematic. a The

May 4, 2017 9:54 b2167-ch19

II-278

D. F. Chernoﬀ and S.-H. H. Tye

universe, 2 ˙ H = −(ρ + p), 2MPl

=−

H˙ 3 p = 1 + H2 2 ρ

(5)

< 1. Slow-roll inﬂation means small , H nearly constant and a(t) ∼ a(0)eHt . In a typical slow-roll inﬂationary model inﬂation ends at t = tend when the inﬂaton encounters a steeper part of the potential and > 1. The number of e-folds of inﬂation is Ne Htend , a key parameter. The energy released from the potential V (φ) heats the universe at the end of inﬂation and starts the hot big bang. The period of slow-roll must last at least Ne > 50 e-folds to explain three important observations about our universe that are otherwise unaccounted for in the normal big bang cosmology: ﬂatness, lack of defects and homogeneity. For any reasonable initial curvature density ρc (0), the ﬁnal curvature density ρc (tend ) < ρc (0)e−150 will be totally negligible, thus yielding a ﬂat universe. This is how inﬂation solves the ﬂatness problem. Any defect density (probably included in ρm (0)) present in the universe before the inﬂationary epoch will also be inﬂated away, thus solving the so-called “monopole” or defect problem. Since the cosmic scale factor grows by a huge factor, the universe we inhabit today came from a tiny patch of the universe before inﬂation. Any original inhomogeneity will be inﬂated away. Inﬂation explains the high degree of homogeneity of the universe. In summary, if Ne is suﬃciently large then inﬂation accounts for the universe’s observed ﬂatness, defect density and homogeneity. What is amazing is that inﬂation also automatically provides a mechanism to create primordial inhomogeneities that ultimately lead to structure formation in our universe. As the inﬂaton slowly rolls down the potential in the classical sense, quantum ﬂuctuations yield slightly diﬀerent ending times tend and so slightly diﬀerent densities in diﬀerent regions. The scalar and the metric ﬂuctuations may be treated perturbatively, and one obtains the dimensionless power spectra in terms of H, ∆2S (k) =

1 H2 2 || , 8π 2 MPl

∆2T (k) =

2 H2 2 2 = r∆S (k), π 2 MPl

(6)

r = 16, where ∆2S and ∆2T are the scalar and the tensor modes, respectively. A scalar scaleinvariant power spectrum corresponds to constant ∆2S , which occurs when a space expands in a nearly de-Sitter fashion for a ﬁnite length of time. The scalar mode is related to the temperature ﬂuctuations ﬁrst measured by COBE.9 Measurements and modeling have been reﬁned over the past two decades. Since φ rolls in the

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-279

nonﬂat potential in the inﬂationary scenario, ∆2S will have a slight k dependence. This is usually parametrized with respect to the pivot wave number kp in the form ∆2S (k) = ∆2S (kp ) ∆2T (k)

=

∆2T (kp )

k kp k kp

ns −1+(dns /d ln k) ln(k/kp )/2+···, (7) nt +(dnt /d ln k) ln(k/kp )/2+···, (8)

where the PLANCK and WMAP best ﬁt values for ΛCDM quoted in Ref. 10 for kp = 0.05 Mpc−1 are ns = 0.9603 ± 0.0073, +0.58 × 10−9 . ∆2S (kp ) = 2.19−0.53

(9) (10)

These do not change appreciably when a possible tensor component is included and PLANCK constraints on r are given at pivot 0.002 Mpc−1 . In the slow-roll approximation, where φ¨ is negligible in Eq. (3), the deviation from scale-invariance is quantiﬁed by the spectral tilt ns − 1 =

d ln ∆2S (k) = −6 + 2η, d ln k

dns = −16η + 242 + 2ξ 2 , d ln k

(11)

where the parameters that measure the deviation from ﬂatness of the potential are 2 M2 V = Pl , (12) 2 V η=

2 V MPl , V

ξ2 =

4 V V MPl . V2

(13)

where and η are the known as the slow-roll parameters which take small values during the inﬂationary epoch. The tensor mode comes from the quantum ﬂuctuation of the gravitational wave. Its corresponding tilt is nt = −2.

(14)

For a small ﬁeld range ∆φ, the constraint on Ne dictates a rather ﬂat potential and small , which in turn implies a very small r. This relation can be quantiﬁed by the Lyth bound,11 ∆φ r ≥ Ne . (15) MPl 8 If 60 ≥ Ne ≥ 40, we ﬁnd that typical values of r satisﬁes r < 0.005 for ∆φ < MPl . This is much smaller than r 0.2 reported by BICEP2.6

May 4, 2017 9:54 b2167-ch19

II-280

D. F. Chernoﬀ and S.-H. H. Tye

While ∆2S is a combination of H 2 and 1/, ∆2T provides a direct measurement of H 2 and hence the magnitude of the inﬂaton potential during inﬂation: r V (2.2 × 1016 GeV)4 . (16) 0.2 The imprint of tensor ﬂuctuations are present in the CMB but ∆2T ∆2S so it is not possible to measure ∆2T directly from total temperature ﬂuctuations at small .b Fortunately, the CMB is linearly polarized and can be separated into E-mode and B-mode polarizations. It happens that ∆2T contributes to both modes while ∆2S contributes only to the E-mode. So a measurement of the primordial B-mode polarized CMB radiation is a direct measurement of ∆2T . Searching for the B-mode CMB is very important. Besides the intrinsic smallness of the B-mode signal, which makes detection a major challenge, the primordial ﬂuctuation may be masked by the interstellar dust. It is likely that the uncertainty from dust would have been resolved by the time this paper appears. So we may consider this as a snapshot after the announcement of the BICEP2 data and before a full understanding of the impact of dust on the reported detection. Due to this uncertainty, we shall consider both a negligibly small r (say r < 0.002) and an observable primordial r ≤ 0.2. 3. String Theory and Inflation By now, string theory is a huge research subject. So far, we have not yet identiﬁed the corner where a speciﬁc string theory solution ﬁts what happens in nature. It is controversial to state whether we are close to ﬁnding that solution or we are still way oﬀ. By emphasizing the cosmological epoch, we hope to avoid the details but try to ﬁnd generic stringy features that may show up in cosmological observations. Here we give a lightning pictorial summary of some of the key features of Type IIB superstring theory so readers can have at least a sketchy picture of how string theory may be tested. Because of compactiﬁcation, we expect modes such as Kaluza–Klein modes to appear in the eﬀective four-dimensional theory. Their presence will alter the relation between the inﬂaton potential scale and r (16) to: 1 r (2.2 × 1016 GeV)4 , (17) V ˆ 2 0.2 N tensor and scalar perturbations have diﬀerent dependencies experimentalists deduce the scalar spectrum from high measurements where scalar power will dominate any tensor contribution. Then they observe the total temperature ﬂuctations at low where tensors should make their largest impact. If the theoretical relation between low and high ﬂuctuations is known (“no running” being the simplest possibility) then the total observed low power limits whatever extra contribution might arise from the tensor modes. The WMAP12 and Planck13 collaborations (incorporating data from the South Pole Telescope (SPT) and Atacama Cosmology Telescope (ACT) microwave background experiments and baryon acoustic oscillation observations) placed limits on r, r < 0.13 and r < 0.11, respectively. These upper limits on r are somewhat less than the r values reported by BICEP2 and under investigation.14 b Since

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-281

ˆ eﬀectively counts the number of universally coupled (at one-loop) degrees where N of freedom below the energy scale of interest here.15 Present experimental bound ˆ is very loose. Surprisingly, a value as big as N ˆ ∼ 1025 is not ruled out. on N 3.1. String theory and flux compactification Here is a brief description on how all moduli are dynamically stabilized in ﬂux compactiﬁcation. Recall that 2-form electric and magnetic ﬁeld strengths follow from the 1-form ﬁeld Aµ in the electromagnetic theory, under which point-like particles (such as electrons) are charged. In analogy to this, fundamental strings in string theory are charged under a 2-form ﬁeld Bµν which yields 3-form ﬁeld strengths. In general, other dimensional objects are also present in string theory, which are known as branes. A p-brane spans p spatial dimensions, so a membrane is a 2-brane while a 1-brane is string like; that is a 1-brane is really a string. In the Type IIB superstring theory, there are a special type of Dp-branes where p is an odd integer. Among other properties of Dp-branes, we like to mention two particularly relevant ones here : (1) each end of an open string must end on a Dp-brane, and (2) with the presence of D1-string (or -brane), there exists another 2-form ﬁeld Cµν under which a D1-string is charged. Self-consistency of Type IIB superstring theory requires it to have nine spatial dimensions. Since only three of them describe our observable world, the other six must be compactiﬁed. Consider a stack of D3-branes of cosmological size. Such a stack appear as a point in the six compactiﬁed dimensions. In the brane world scenario, all standard model particles (electrons, quarks, gluons, photons, etc.) are light open string modes whose ends can move freely inside the D3-branes, but cannot move outside the branes, while the graviton, being a closed string mode, can move freely in the branes as well as outside, i.e. the bulk region of the six compactiﬁed dimensions. In this sense, the D3-branes span our observable universe. We can easily replace the D3-branes by D7-branes, with 4 of their dimensions wrapping a 4-cycle inside the compactiﬁed six-dimensional space. Dark matter may come from unknown particles inside the same stack, or from open string modes sitting in another stack sitting somewhere else in the compactiﬁed space. As a self-consistent theory, the six extra dimensions must be dynamically compactiﬁed (and stabilized). This is a highly nontrivial problem. Fortunately, the 3-form ﬁeld strengths of both the NS–NS ﬁeld Bµν and the R–R ﬁeld Cµν are quantized in string theory. Wrapping 3-cycles in the compactiﬁed dimensions, these ﬂuxes contribute to the eﬀective potential V in the low energy approximation. Their presence can render the shape and the size of the compactiﬁed six-dimensional space to be dynamically stabilized. This is known as ﬂux compactiﬁcation. Here the matter content and the forces of nature are dictated by the speciﬁc ﬂux compactiﬁcation.16,17 Warped internal space appears naturally. This warped geometry will come to play an important role in cosmology.

May 4, 2017 9:54 b2167-ch19

II-282

D. F. Chernoﬀ and S.-H. H. Tye

To describe nature, a Calabi–Yau like manifold is expected, with branes and orientifold planes. At low energy, a particular manifold may be described by a set of dynamically stabilized scalar ﬁelds, or moduli. One or more K¨ ahler moduli parameterize the volume while the shape is described by the complex structure moduli, whose number may reach hundreds. A typical ﬂux compactiﬁcation involves many moduli and three-form ﬁeld strengths with quantized ﬂuxes (see Ref. 18). With such a large set of dynamical ingredients, we expect many possible vacuum solutions; collectively, this is the string theory landscape, or the so-called cosmic landscape. Here, a modulus is a complex scalar ﬁeld. Written in polar coordinate, we shall refer to the phase degree of freedom as an axion. For a given Calabi–Yau like manifold, we can, at least in principle, determine the four-dimensional low energy supergravity eﬀective potential V for the vacua. To be speciﬁc, let us consider only 3-form ﬁeld strengths F3i wrapping the threecycles inside the manifold. (Note that these are dual to the four-form ﬁeld strengths in four-dimensional spacetime.) We have V (F3i , φj ) → V (ni , φj ), (i = 1, 2, . . . , N , j = 1, 2, . . . , K) where the ﬂux quantization property of the 3-form ﬁeld strengths F3i allow us to rewrite V as a function of the quantized values ni of the ﬂuxes present and φj are the complex moduli describing the size and shape of the compactiﬁed manifold as well as the couplings. There are barriers between diﬀerent sets of ﬂux values. For example, there is a (ﬁnite height) barrier between n1 and n1 − 1, where tunneling between V (n1 , n2 , . . . , nN , φj ) and V (n1 − 1, n2 , . . . , nN , φj ) may be achieved by brane-ﬂux annihilation.19 For a given set of ni , we can locate the meta-stable (classically stable) vacuum solutions V (ni , φj ) by varying φj . We sift through these local minima which satisfy the following criteria: they have vanishingly small vacuum energies because that is what is observed in today’s universe, and long decay lifetimes to lower energy states because our universe is long-lived. These criteria restrict the manifolds, ﬂux values and minima of interest; nonetheless, within the rather crude approximation we are studying, there still remain many solutions; and statistically, it seems that a very small cosmological constant is preferred.20 3.2. Inflation in string theory When the universe was ﬁrst created (say, a bubble created via tunneling from nothing), φj are typically not sitting at φj,min , so they tend to roll toward their stabilized values. Heavier moduli with steeper gradients probably reached their respective minima relatively quickly. The ones with less steep directions took longer. The last ones to reach their stabilized values typically would move along relatively ﬂat directions, and they can play the role of inﬂatons. So the vacuum energy that drives inﬂation is roughly given by the potential when all moduli except the inﬂaton have already reached bottom. Since the ﬂux compactiﬁcation in string theory introduces dozens or hundreds of moduli (each is a complex scalar ﬁeld in the low energy eﬀective theory approximation), one anticipates that there are many

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-283

candidates for inﬂation. Even if we assume that just one ﬁeld is responsible for the inﬂationary epoch in our own universe it seems likely that string theory can choose among many possibilities to realize single ﬁeld inﬂation in many diﬀerent ways. In fact, the picture from string theory may oﬀer additional interesting possibilities beyond ﬁeld theory. For example, suppose there was a pair of brane–anti-brane present in the early universe whose tensions drive the inﬂation. As we shall see, the inﬂaton ﬁeld happens to be the distance between them. After their annihilation that ends inﬂation, the inﬂaton ﬁeld no longer exists as a degree of freedom in the low energy eﬀective ﬁeld theory. Some inﬂationary scenarios in string theory generate unobservably small primordial r while others give observably large primordial r. Small ﬁeld range implies small r, but large ﬁeld range does not necessary imply large r, which happens when the Lyth bound (15) is not saturated. To obtain enough e-folds, we need a ﬂat enough potential, so it is natural to consider an axion as the inﬂaton, as ﬁrst proposed in natural inﬂation.21,22 The axion has a natural shift symmetry φ → φ + any constant. This continuous symmetry can easily be broken, via nonperturbative eﬀects, to a discrete symmetry, resulting in a periodic potential that can drive inﬂation. The typical ﬁeld range is sub-Planckian implying a small r. We shall see shortly that there are many string theory models built out of the axion that permit large ﬁeld ranges and some of these can yield large r. Here, we shall describe the various string theory realizations of inﬂation within the framework of Type IIB string theory. The pictorially simplest models occur in the brane world scenario. We shall present a few sample models to give a picture of the type of models that have been put forward. Readers can ﬁnd a more complete list in Ref. 4. If r turns out to be large, it will be interesting to see whether some of the small r models can be adapted or modiﬁed to have a large r. For example, one can try to modify a brane inﬂationary scenario into a warm inﬂationary scenario with a relatively large r. 4. Small r Scenarios Besides fundamental superstrings, Type IIB string theory has Dp-branes where the number of spatial dimensions p is odd. Furthermore, supersymmetry can be maintained if there are only D3- and D7-branes; so, unless speciﬁed otherwise, we shall restrict ourselves to this case. String theory has nine spatial dimensions. Since our observable universe has only three spatial dimensions, the other six spatial dimensions have to be compacted in a manifold with volume V6 . The resulting (i.e. dynamically derived) MPl is related to the string scale MS via Eq. (1) and the typical ﬁeld range ∆φ is likely to be limited by MPl > MS . For such a small ﬁeld range, the potential has to be ﬂat enough to allow 50 or more e-folds and the Lyth bound (15) for 60 ≥ Ne ≥ 40 implies r < 0.005. We shall refer to these models as small r scenarios.

May 4, 2017 9:54 b2167-ch19

II-284

D. F. Chernoﬀ and S.-H. H. Tye

4.1. Brane inflation The discovery of branes in string theory demonstrated that the theory encompasses a multiplicity of higher-dimensional, extended objects and not just strings. In the brane world scenario, our visible universe lies inside a stack of D3-branes, or a stack of D7-branes. Here, 6 of the 9 spatial dimensions are dynamically compactiﬁed while the three spatial dimensions of the D3-branes (or 3 of the D7-branes) are cosmologically large. The six small dimensions are stabilized via ﬂux compactiﬁcation16,17 ; the region outside the branes is referred to as the bulk region. The presence of RR and NS–NS ﬂuxes introduces intrinsic torsion and warped geometry, so there are regions in the bulk with warped throats (Fig. 1). Since each end of an open string must end on a brane, only closed strings are present in the bulk away from branes. There are numerous such solutions in string theory, some with a small positive vacuum energy (cosmological constant). Presumably the standard model particles are open string modes; they can live either on D7-branes wrapping a 4-cycle in the bulk or (anti-)D3-branes at the bottom of a warped throat (Fig. 1). The (relative) position of a brane is an open string mode. In brane inﬂation,23 one of these modes is identiﬁed as the inﬂaton φ, while the inﬂaton potential is generated by the classical exchange of closed string modes including the graviton. From the open string perspective, this exchange of a closed string mode can be viewed as a quantum loop eﬀect of the open string modes.

Fig. 1. A pictorial sketch of a generic ﬂux compactiﬁed six-dimensional bulk, with a number of warped throats (4 of which are shown here). Besides the warped throats, there are D7-branes ¯ wrapping 4-cycles. The blue dots stand for mobile D3-branes while the red dots are D3-branes ¯ sitting at the bottoms of throats. In the D3-D3-brane inﬂationary scenario, the tension of the brane pairs provide the vacuum energy that drives inﬂation as the D3-brane moves down A¯ throat. Inﬂation ends as the D3-brane annihilates with the D3-brane in A-throat. The standard model branes may live in A-throat or S-throat. As an alternative, inﬂation may take place while branes are moving out of B-throat. (For color version, see page II-CP14.)

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-285

¯ 4.1.1. D3-D3-brane inﬂation In the early universe, besides all the branes that are present today, there is an extra 24,25 ¯ ¯ pair of D3-D3-branes. Due to the attractive forces present, the D3-brane is expected to sit at the bottom of a throat. Here again, inﬂation takes place as the D3¯ brane moves down the throat toward the D3-brane, with their separation distance as the inﬂaton, and inﬂation ends when they collide and annihilate each other, allowing the universe to settle down to the string vacuum state that describes our universe today. Although the original version encounters some ﬁne-tuning problems, the scenario becomes substantially better as we make it more realistic with the introduction of warped geometry.26 Because of the warped geometry, a consequence of ﬂux compactiﬁcation, a mass M in the bulk becomes hA M at the bottom of a warped throat, where hA 1 is the warped factor (Fig. 1). This warped geometry tends to ﬂatten, by orders of ¯ magnitude, the inﬂaton potential V (φ), so the attractive D3-D3-brane potential is rendered exponentially weak in the warped throat. The attractive gravitational (plus RR) potential together with the brane tensions takes the form 1 φ4A , (18) V (φ) = 2T3 h4A 1 − NA φ4 where T3 is the D3-brane tension and the eﬀective tension is warped to a very small value T3 h4A . The warp factor hA depends on the details of the throat. Crudely, h(φ) ∼ φ/φedge , where φ = φedge when the D3-brane is at the edge of the throat, so h(φedge ) 1. At the bottom of the throat, where φ = φA φedge , hA = h(φA ) = φA /φedge . The potential is further warped because NA 1 is the D3 charge of the ¯ throat. The D3-D3-brane pair annihilates at φ = φA . In terms of the potential (18), a tachyon appears as φ → φA so inﬂation ends as in hybrid inﬂation. The energy released by the brane pair annihilation heats up the universe to start the hot big bang. To ﬁt data, hA ∼ 10−2 . If the last 60 e-folds of inﬂation takes place inside the throat, then φedge ≥ φ ≥ φA during this period of inﬂation. This simple model yields ns 0.97, r < 10−5 and vanishing non-Gaussianity. The above model is very simple and well motivated. There are a number of interesting variations that one may consider. Within the inﬂaton potential, we can add new features to it. In general, we may expect an additional term in V (φ) of the form βH 2 φ2 , 2 where H is the Hubble parameter so this interaction term behaves like a conformal coupling. Such a term can emerge in a number of diﬀerent ways: • Contributions from the K¨ ahler potential and various interactions in the superpo26 tential as well as possible D-terms,27 so β may probe the structure of the ﬂux compactiﬁcation.28,29 • It can come from the ﬁnite temperature eﬀect. Recall that ﬁnite temperature T induces a term of the form T 2 φ2 in ﬁnite temperature ﬁeld theory. In a de-Sitter

May 4, 2017 9:54 b2167-ch19

II-286

D. F. Chernoﬀ and S.-H. H. Tye

universe, there is a Hawking–Gibbons temperature of order H thus inducing a term of the form H 2 φ2 . ¯ • The D3-brane is attracted to the D3-brane because of the RR charge and the gravitational force. Since the compactiﬁed manifold has no boundary, the total RR charge inside must be zero, and so is the gravitational “charge.” A term of the above form appears if we introduce a smooth background “charge” to cancel the gravitational “charges” of the branes.30 On the other hand, negative tension is introduced via orientifold planes in the brane world scenario. If the throat is far enough away from the orientifold planes, it is reasonable to ignore this eﬀect. Overall, β is a free parameter and so is naively expected to be of order unity, β ∼ 1. However, the above potential yields enough inﬂation only if β is small enough, β < 1/5.31 The present PLANCK data implies that β is essentially zero or even slightly negative. Another variation is to notice that the six-dimensional throat can be quite nontrivial. In particular, if one treats the throat geometry as a Klebanov–Strassler deformed conifold, gauge-gravity duality leads to the expectation that φ will encounter steps as the D3-brane moves down the throat.32 Such steps, though small, may be observed in the CMB power spectrum. 4.1.2. Inﬂection point inﬂation Since the six-dimensional throat has one radial and ﬁve angular modes, V (φ) gets corrections that can have angular dependencies. As a result, its motion toward the bottom of the throat may follow a nontrivial path. One can easily imagine a situation where it will pass over an inﬂection point. Around the inﬂection point, V (φ) may take a generic simple form V (φ) V0 + Aφ +

Bφ2 Cφ3 + , 2 3

where η = 0 at the inﬂection point B + 2Cφ = 0. As shown in Refs. 29 and 33, √ given that V (φ) is ﬂat only around the inﬂection point, Ne ∼ 1/ , so must be very small, resulting in a very small r. Here ns − 1 = 2η so we expect ns to be very close to unity, with a slight red or even blue tilt. Although motivated in brane inﬂation, an inﬂection point may be encountered in other scenarios of the inﬂationary universe, so it should not be considered as a stringy feature. 4.1.3. DBI model Instead of modifying the potential, string theory suggests that the kinetic term for a D-brane should take the Dirac–Born–Infeld form34,35 1 µ 1 1 ∂ φ∂µ φ → − , (19) 1 − f (φ)∂ µ φ∂µ φ + 2 f (φ) f (φ)

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-287

where f (φ) ∼ h(φ)−4 ∼ A/φ4 is the warp factor of a throat. This DBI property is intrinsically a stringy feature. Here, the warp factor plays the role of a brake that slows down the motion of the D-brane as it moves down the throat, f (φ)(∂t φ)2 < 1. This braking mechanism is insensitive to the form of V (φ), so many e-folds are assured as φ → 0. It produces a negligibly small r but a large non-Gaussianity in the equilateral bi-spectrum as the sound speed becomes very small. The CMB data has an upper bound on the non-Gaussianity that clearly rules out the dominance of this stringy DBI eﬀect. Nevertheless, it is a clear example that stringy features of inﬂationary scenarios may be tested directly by cosmological observation. Instead of moving down a throat, one may also consider inﬂation while a brane is moving out of a throat.36,37 In this case, the predictions are not too diﬀerent from that of the inﬂection point case. 4.1.4. D3-D7-brane inﬂation Here, the inﬂaton is the position of a D3-brane moving relative to the position of a higher-dimensional D7-brane.38 Since the presence of both D3 and D7-branes preserves supersymmetry (as opposed to the presence of D5-branes), inﬂation can be driven by a D-term and a potential of the form like V (φ) = V0 + a ln φ − bφ2 + cφ4 + · · · may be generated. Such a potential yields ns 0.98, which may be a bit too big. One may lower the value a little by considering variations of the scenario. In any case, one ends up with a very small r. 4.2. K¨ ahler moduli inflation In ﬂux compactiﬁcation, the K¨ ahler moduli are typically lighter than the complex structure moduli. Intuitively, this implies their eﬀective potentials are ﬂatter than those for the complex structure moduli. For a Swiss-cheese like compactiﬁcation, besides the modulus for the overall volume, we can also have moduli describing the sizes of the holes inside the manifold. So it is reasonable to ﬁnd situations where a K¨ ahler modulus plays the role of the inﬂaton. A potential is typically generated by nonperturbative eﬀects, so examples may take the form39 V V0 (1 − Ae−kφ ),

(20) 2 −kφ

< where both A and k are positive and of order unity or bigger. Here, η −Ak e 0 and η 2 /2k 2 , so the potential is very ﬂat for large enough φ. If the inﬂaton measures the volume of a blow-up mode corresponding to the size of a 4-cycle √ in a Swiss-Cheese compactiﬁcation in the large volume scenario, we have k ∼ V ln V, where the compactiﬁcation volume is of order V ∼ 106 in string units, so k is huge and r 2(ns − 1)2 /k 2 ∼ 10−10 . Models of this type are not close to saturating

May 4, 2017 9:54 b2167-ch19

II-288

D. F. Chernoﬀ and S.-H. H. Tye

the Lyth bound (15). Other scenarios40 of this type again have small r. It will be interesting whether this type of models can be modiﬁed to have a larger r. 5. Large r Scenarios The Lyth bound (15) implies that large r requires large inﬂaton ﬁeld range ∆φ MPl . A phenomenological model with relatively ﬂat potential and large ﬁeld range is not too hard to write down, e.g. chaotic inﬂation. However, the range of a typical modulus in string theory is limited by the size of compactiﬁcation, ∆φ < MPl . Even if we could extend the ﬁeld range (e.g. by considering an irregular shaped manifold), the corrections to a generic potential may grow large as φ explores a large range. Essentially, we lose control of the approximate description of the potential. Axions allow one to maintain control of the approximation used while exploring large ranges. Here we shall brieﬂy review two ideas, namely the Kim–Nilles–Peloso Mechanism41 and the axion monodromy.42,43 In both cases, an axion moves in a helicallike path. The ﬂatness of the potential and the large ﬁeld range appear naturally. 5.1. The Kim–Nilles–Peloso mechanism Let us start with a simple model and then build up to a model that is relatively satisfactory. 5.1.1. Natural inﬂation It was noticed long ago that axion ﬁelds may be ideal inﬂaton candidates because an axion ﬁeld φ, a pseudo-scalar mode, has a shift symmetry, φ → φ + const. This symmetry is broken to a discrete symmetry by some nonperturbative eﬀect so that a periodic potential is generated,21,22 φ m2 2 A 2 V (φ) = A 1 − cos φ + · · · ∼ φ , → (21) f 2f 2 2 where we have set the minima of the potential at φ = 0 to zero vacuum energy. With suitable choice of A and f the resulting axion potential can be relatively small and ﬂat, an important property for inﬂation. As f becomes large, this model approaches the (quadratic form) chaotic inﬂation,44 which is well studied. To have enough e-folds, we may need a large ﬁeld range, say, ∆φ > 14 MPl, which is possible here only if the decay constant f > ∆φ MPl . This requires a certain degree of ﬁne tuning since a typical f is expected to satisfy f < MPl (see, for example, Fig. 2 in Ref. 45). To ﬁt the scalar mode perturbations of COBE m 7 × 10−6 MPl . The range of predictions45 in the r versus ns plot is shown in Fig. 2. 5.1.2. N-ﬂation There are ways to get around this to generate enough e-folds. One example is to extend the model to include N diﬀerent axions (similar to the idea of using many

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-289

Fig. 2. The tensor-to-scalar ratio r versus the primordial tilt ns plot for pivot scale 0.002 Mpc−1 . Pairs of solid and dashed semi-circular lines encompass 68% and 95% conﬁdence limits for r and ns as given in Ref. 10. The regions are based on PLANCK data combined with: WMAP [large scale polarization] (gray), WMAP plus BAO [baryon acoustic oscillations in galaxy surveys] (blue) and WMAP plus higher CMB data [ACT and SPT] (red) as described in detail in Ref. 10. All remaining lines, points and arrows describe theoretical models. Convex potentials lie above the straight, solid black line; concave potentials below it. Power law inﬂation generates results along the straight, dashed line. String-related models include the short yellow segment for a linear inﬂaton potential V ∼ φ, the short red segment for V ∼ φ2/3 and the short black segment (chaotic inﬂation) for V ∼ φ2 . Each small (large) dot stands for 50 (60) e-folds. Two purple lines moving to the left show the full range of predictions for natural inﬂation. All small ﬁeld models and ˆ some large ﬁeld ones suppress r (these are not plotted). The black arrow is the vertical shift ∆, illustrating how the results for cosine potential diﬀer from those of the φ2 potential. (For color version, see page II-CP14.)

scalar ﬁelds46,47 ) each with a term of the form in Eq. (21), with a diﬀerent decay constant fi < MPl . This N-ﬂation model48 is a string theory inspired scenario since ﬂux compactiﬁcation results in the presence of many axions. In the approximation that the axions are independent of each other (that is, their couplings with each other are negligible), the analysis is quite straightforward and interesting.49 The predictions are similar to that of a large ﬁeld model, that is, 0.93 < ns < 0.95 but r ≤ 10−3 . In general, the axions do couple to each other and the situation can be quite complicated. For a particularly simple case, inspired by string theory and supergravity, a statistical analysis has been carried out and clear predictions can be made.49 For a large number of axions, one has φi φi φj αi cos + V (φi ) = V0 + βij cos − . (22) fi fi fj i i,j The statistical distribution of results yields small values for r quite similar to that of the N-ﬂation model.

May 4, 2017 9:54 b2167-ch19

D. F. Chernoﬀ and S.-H. H. Tye

II-290

5.1.3. Helical inﬂation Let us consider a particularly interesting case of a 2-axion model of the form41

φ1 φ1 φ2 V (φ1 , φ2 ) = V1 + V2 = V0 1 − cos + A 1 − cos − , (23) f1 f1 f2 where we take f1 f1 MPl . Note that the second term in V (23), namely V2 , vanishes at φ1 φ2 − = 0. f1 f2

(24)

This is the minimum of V2 and the bottom of the trough of the potential V . For f1 large enough A, inﬂaton will follow the trough as it rolls along its path. Now 2πf 1 measures the number of cycles that φ2 can travel for 0 < φ1 < f1 . Although φ2 has a shift symmetry φ2 → φ2 + 2πf2 , the path does not return to the same conﬁguration after φ2 has traveled for one period because φ1 has also moved. Thus, instead of a shift symmetry, the system has a helical symmetry. Moving along the path of the trough (24), we see that the second term in the potential (23) vanishes and so the potential V reduces to that with only the ﬁrst term and the range of the inﬂaton ﬁeld φ2 can easily be super-Planckian. To properly normalize the ﬁelds, one deﬁnes two normalized orthogonal directions, f1 φ1 + f2 φ2 X= , f12 + f22

f2 φ1 − f φ2 Y = 2 1 2 , f1 + f2

(25)

the inﬂaton will roll along the X-direction while Y -direction is a heavy mode which can be integrated out. Note that the system is insensitive to the magnitude of A as long as it is greater than O(1) such that Y -direction is heavy enough. For instance, the slow roll parameters and η are not aﬀected which means that the observables ns and r are insensitive to A. Since the X path is already at the minimum of the second term, at Y = 0, the eﬀective potential along X is determined by V1 ,

X cos θ 1 V0 X 2 V (X) = V0 1 − cos ≈ cos2 θ, (26) f1 2 f12 cos θ ≡

f1 f12 + f22

.

(27)

So this helical model is reduced to the single cosine model (21) where f1 f1 f2 f= f12 + f22 f1 f1 can be bigger than MPl even if all the individual fi < MPl . So it is not diﬃcult to come up with stringy models that can ﬁt the existing data. In view of the large B-mode reported by BICEP2,6 this model was revisited by a number of groups.50–57 In fact, one can consider a more general form for V1 (φ1 ),58 which can lead to a large ﬁeld range model with somewhat diﬀerent predictions.

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-291

Since the shift symmetry of an axion is typically broken down to some discrete symmetry, the above cosine model is quite natural. The generalization to more than two axions is straightforward, and one can pile additional helical motions on top of this one, increasing the value of the eﬀective decay constant f by additional big factors. All models in this class reduce to a model with a single cosine potential, which in turn resembles the quadratic version of chaotic inﬂation. This limiting behavior is quite natural in a large class of axionic models in the supergravity framework.59 Because of the periodic nature of an axionic potential, this cosine model can have a smaller value of r than chaotic inﬂation.44 The deviations from chaotic inﬂation occur in a well-deﬁned fashion. Let 2 ˆ = 16∆ = r + 4(ns − 1) = −4MPl , ∆ (28) f2 ˆ = 0 while the cosine model has ∆ ˆ < 0. All other where φ2 chaotic inﬂation has ∆ quantities such as runnings of spectral indices have very simple dependencies on ˆ As shown in Table 1, this deviation is quite distinctive of the periodic nature of ∆. the inﬂaton potential. For ns = 0.96, r = 0.16 in the φ2 model, while r can be as ˆ = −0.12) in the cosine model. As data improves, a negative small as r = 0.04 (or ∆ ˆ can provide a distinctive signature for a periodic axionic potential for value of ∆ inﬂation. 5.2. Axion monodromy The above idea of extending the eﬀective axion range can be carried out with a single axion, where the axion itself executes a helical motion. This is the axion Table 1. Comparison between φ2 chaotic inﬂation and the cosine model for the various physically measurable quantities51 where ∆ (28) measures the diﬀerence. Here MPl = 1.

V (φ) =

1 (V /V )2 2

η = V /V

1 2 2 m φ 2

)] V0 [1 − cos( φ f

1 r 16

1 r 16

1 r 16

1 r 16

+ 2∆

ξ 2 = V V /V 2

0

ω 3 = V 2 V /V 3

0

ns − 1 = 2η − 6

− 41 r

− 14 r + 4∆

nt = −2

− 81 r

− 18 r

dns d ln k

= −16η + 242 + 2ξ 2

1 2 r 32

dnt d ln k

= −4η + 82

1 2 r 64

d2 ns d ln k2

= −1923 + 1922 η − 32η2

−24ξ 2 + 2ηξ 2 + 2ω 3

1 3 − 128 r

1 r∆ 2

1 2 r ∆ 32

1 2 r 32

1 2 r 64

+ r∆2

− r∆

− 12 r∆

1 3 − 128 r + 38 r 2 ∆ − 4r∆2

May 4, 2017 9:54 b2167-ch19

II-292

D. F. Chernoﬀ and S.-H. H. Tye

monodromy model.42,43 In this scenario, inﬂation can persist through many periods around the conﬁguration space, thus generating an eﬀectively large ﬁeld range with an observable r. Recall that a gauge ﬁeld, or one-form ﬁeld (i.e. with one spacetime index), is sourced by charged point-like ﬁelds, while a 2-form (anti-symmetric tensor) ﬁeld is sourced by strings. So we see that there will be at least one 2-form ﬁeld in string theory. Consider a 5-brane that ﬁlls our four-dimensional spacetime and wraps a 2-cycle inside the compactiﬁed manifold. The axion is the integral of a 2form ﬁeld over the 2-cycle. Integrating over this 2-cycle, the six-dimensional brane action produces a potential for the axion ﬁeld in the resulting four-dimensional eﬀective theory. Here, the presence of the brane breaks the axion shift symmetry and generates a monodromy for the axion. For NS5-branes, a typical form of the axion potential (coming from the DBI action) is (29) V (φ) = A b2 + φ2 . For small parameter b, V (φ) Aφ. The prediction of such a linear potential is shown in Fig. 2. One can consider D5-branes instead. D7-branes wrapping 4-cycles in the compactiﬁed manifold is another possibility. For large values of φ, a good axion monodromy model requires that there is no uncontrollable higher-order stringy or quantum corrections that would spoil the above interesting properties.60–62 Variants of this picture may allow a more general form of the potential, say V (φ) ∼ φp where p can take values such as p = 2/3, 4/3, 2, 3, thus generating r 0.04, 0.09, 0.13 and 0.2, respectively. The p = 2 case predicts a ns value closest to r ∼ 0.16 reported by BICEP2 after accounting for dust. The observational bounds are currently under scrutiny. Readers are referred to Ref. 4 for more details. 5.3. Discussions If BICEP2’s detection of r is conﬁrmed, it does not necessarily invalidate completely all the small r models discussed above. It may be possible to modify some of them ¯ to generate a large enough r. As an example, if one is willing to embed the D3-D3brane inﬂation into a warm inﬂationary model, a r 0.2 may be viable.63 It will be interesting to re-examine all the small r string theory models and see whether and how any of them may be modiﬁed to produce a large r. As string theory has numerous solutions, it is not surprising that there are multiple ways to realize the inﬂationary universe scenario. With cosmological data available today, theoretical predictions and contact with observations are so far quite limited. As a consequence, it is rather diﬃcult to distinguish many string theory inspired predictions from those coming from ordinary ﬁeld theory (or even supergravity models). There are exceptions, as pointed out earlier. For example, the DBI inﬂation prediction of an equilateral bi-spectrum in the non-Gaussianity, or the determined spacing of steps in the power spectrum itself, may be considered to be distinct enough that if either one is observed, some of us may be convinced that

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-293

it is a smoking gun of string theory. Although searches for these phenomena should and would continue, so far, we have not been lucky enough to see any hint of them. Here we like to emphasize that there is another plausible signature to search for. Since all fundamental objects are made of superstrings (we include D1-strings here), and the universe is reheated to produce a hot big bang after inﬂation, it is likely that, besides strings in their lowest modes which appear as ordinary particles, some relatively long strings will also be produced, either via the Kibble mechanism or some other mechanism. They will appear as cosmic strings. When there are many axions, or axion-like ﬁelds, with a variety of plausible potentials, the possibilities may be quite numerous and so predictions may be somewhat imprecise. For any axion or would-be-Goldstone bosons, with a continuous U (1) symmetry, we expect a string-like defect which can end up as cosmic strings if generated in early universe. In ﬁeld theory, a vortex simply follows from the Higgs mechanism where the axion a appears as the phase of a complex scalar ﬁeld, Φ = ρeia/f . In string theory, we note that such an axion a is dual to a 2-form tensor ﬁeld Cλκ , i.e. ∂ µ a = µνλκ ∂ν Cλκ . As pointed out earlier, such a 2-form ﬁeld is sourced by a string. So the presence of axions would easily lead to cosmic strings (i.e. vortices, fundamental strings and D1-strings) and these may provide signatures of string theory scenarios for the inﬂationary universe. This is especially relevant if cosmic strings come in a variety of types with diﬀerent tensions and maybe even with junctions. Following Fig. 1, we see that besides the axions responsible for inﬂation, there may be other axions. Some of them may have mass scales warped to very small values. If a potential of the form (21) is generated, a closed string loop becomes the boundary of a domain wall, or membrane. It will be interesting to study the eﬀect of the membrane on the evolution of a cosmic string loop. The tension of the membrane and the axion mass are of order √ A σ ∼ Af, m2 = 2 . f It is interesting to entertain the possibility that this axion can contribute substantially to the dark matter of the universe. If so, its contribution to the energy density is roughly given by A while its mass is estimated to be m 10−22 eV.64 This yields √ −118 4 3 A 10 MPl , f A/m 10−10 MPl and σ 10−69 MPl 10−14 GeV3 . For such a small membrane tension, the evolution of the corresponding cosmic string is probably not much changed. 6. Relics: Low Tension Cosmic Strings Witten’s65 original consideration of macroscopic cosmic strings was highly inﬂuential. He argued that the fundamental strings in heterotic string theory had tensions too large to be consistent with the isotropy of the COBE observations. Had they been produced they would be inﬂated away. It is not even clear how inﬂation might be realized within the heterotic string theory.

May 4, 2017 9:54 b2167-ch19

II-294

D. F. Chernoﬀ and S.-H. H. Tye

However, with the discovery of D-branes,66 the introduction of warped geometries67 and the development of speciﬁc, string-based inﬂationary scenarios, the picture has changed substantially. Open fundamental strings must end on branes; so both open and closed strings are present in the brane world. Closed string loops inside a brane will break up into pieces of open strings, so only vortices (which may be only metastable) can survive inside branes. Here, D1-strings (i.e. D1-branes) may be treated as vortices inside branes and survive long enough to be cosmologically interesting.68 The warped geometry will gravitationally redshift the string tensions to low values and, consequently, strings can be produced after inﬂation. A string with tension T in the bulk will be warped to µ = h2 T with h being the warp factor, which can be very small when the string is sitting at the bottom of a throat. (It is h = hA in Eq. (18) in throat A.) That is important because relics produced during inﬂation are rapidly diluted by expansion. Only those generated after (or very near the end of) inﬂation are potentially found within the visible universe. Since the Type IIB model has neither D0-branes nor D2-branes, the well-justiﬁed conclusion that our universe is dominated neither by monopoles nor by domain walls follows automatically. Scenarios that incorporate string-like relics may prove to be consistent with all observations. Such relics appear to be natural outcomes of today’s best understood string theory scenarios. The physical details of the strings in Type IIB model can be quite nontrivial, including the types of diﬀerent species present and the range of string tensions. Away from the branes, p fundamental F1-strings and q D1-strings can form a (p, q) bound state. Junctions of strings will be present automatically. If they live at the bottom of a throat, there can be beads at the junctions as well. To be speciﬁc, let us consider the case of the Klebanov–Strassler warped throat,69 whose properties are relatively well understood. On the gravity side, this is a warped deformed conifold. Inside the throat, the geometry is a shrinking S 2 ﬁbered over a S 3 . The tensions of the bound state of p F1-strings and that of q D1-strings were individually computed.70 The tension formula for the (p, q) bound states is given by Ref. 71 in terms of the warp factor h, the strings scale MS and the string coupling gs , 2 h2 MS2 q 2 bM 2 πp Tp,q + sin , (30) 2π gs2 π M where b = 0.93 numerically and M is the number of fractional D3-branes (that is, the units of 3-form RR ﬂux F3 through the S 3 ). Interestingly, the F 1-strings are charged in ZM and are non-BPS. The D-string on the other hand is charged in Z and is BPS with respect to each other. Because p is ZM -charged with nonzero binding energy, binding can take place even if (p, q) are not coprime. Since it is a convex function, i.e. Tp+p < Tp + Tp , the p-string will not decay into strings with

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-295

smaller p. M fundamental strings can terminate to a point-like bead with mass70 3/2 gs bM h Mbead = MS 4π π 3 irrespective of the number of D-strings around. Inside a D-brane, F1-strings break into pieces of open strings so we are left with D1-strings only, in which case they resemble the usual vortices in ﬁeld theory. For M → ∞ and b = h = 1, the tension (30) reduces to that for ﬂat internal space.72 Cosmological properties of the beads have also been studied.73 Following from gauge-gravity duality, the interpretation of these strings in the gauge theory dual is known. The F 1-string is dual to a conﬁning string while the D1string is dual to an axionic string. Here the gauge theory is strongly interacting and the bead plays the role of a “baryon”. It is likely that diﬀerent throats have diﬀerent tension spectra similar to that for the Klebanov–Strassler throat. Finally, there is some evidence that strings can move in both internal and external dimensions and are not necessarily conﬁned to the tips of the throat.74 This behavior can show up as a cosmic string with variable tension. Following gauge-gravity duality, one may argue that one can also obtain these types of strings within a strongly interacting gauge theory; however, so far we are unable to see how inﬂation can emerge from such a nonperturbative gauge theory description. As we will describe in more detail, all indications suggest that, once produced after inﬂation, a scaling cosmic string network will emerge for the stable strings. Before string theory’s application to cosmology, the typical cosmic string tension was presumed to be set by the grand uniﬁed theory’s (GUTs) energy scale. Such strings have been ruled out by observations. We will review the current limits shortly. Warped geometry typically allows strings with tensions that are substantially smaller, avoiding the observational constraints on the one hand and frustrating easy detection on the other. The universe’s expansion inevitably generates sub-horizon string loops and if the tension is small enough the loops will survive so long that their peculiar motions are damped and they will cluster in the manner of cold dark matter. This results in a huge enhancement of cosmic string loops within our galaxy (about 105 times larger at the Sun’s position than the mean throughout the universe) and makes detection of the local population a realistic experimental goal in the near future. We will focus on the path to detection by means of microlensing. Elsewhere, we will discuss blind, gravitational wave searches. A microlensing detection will be very distinctive. The nature of a microlensing loop can be further conﬁrmed and studied through its unique gravitational wave signature involving emission of multiple harmonics of the fundamental loop period from the precise microlensing direction. Since these loops are essentially the same type of string that makes up all forms of microscopic matter in the universe, their detection will be of fundamental importance in our understanding of nature.

May 4, 2017 9:54 b2167-ch19

II-296

D. F. Chernoﬀ and S.-H. H. Tye

6.1. Strings in brane world cosmology String-like defects or fundamental strings are expected whenever reheating produces some closed strings toward the end of an inﬂationary epoch. Once inﬂation ends and the radiation dominated epoch begins, ever larger sections of this cosmic string network re-enter the horizon. The strings move at relativistic speeds and long lengths collide and break oﬀ sub-horizon loops. Loops shrink and evaporate by emitting gravitational waves in a characteristic time τ = l/(ΓGµ) where l is the invariant loop size, µ is the string tension while Γ is numerically determined and Γ ∼ 50 for strings coupled only to gravity.5 To ease discussion, we shall adopt this value for Γ. String tension is the primary parameter that controls the cosmic string network evolution, ﬁrst explored in the context of phase transitions in grand uniﬁed ﬁeld theories (GUTs), which may be tied to the string scale. Assuming that the inﬂation scale is comparable to the GUT scale, inﬂation generated horizoncrossing defects whose tension is set by the characteristic grand uniﬁcation energy.5 These GUT strings with Gµ ∼ 10−6 would have seeded the density ﬂuctuations for galaxies and clusters but have long been ruled out by observations of the CMB.9,75,76 Here, string theory comes to the rescue. Six of the string theory’s 9 spatial dimensions are stably compactiﬁed. The ﬂux compactiﬁcation involves manifolds possessing warped throat-like structures which redshift all characteristic energy scales compared to those in the bulk space. In this context, cosmic strings produced after inﬂation living in or near the bottoms of the throats can have diﬀerent small tensions.77–81 The quantum theory of one-dimensional objects includes a host of eﬀectively one-dimensional objects collectively referred to here as superstrings. For example, a single D3-brane has a U (1) symmetry that is expected to be broken, thus generating a string-like defect. (For a stack of n branes, the U (n) ⊃ U (1) symmetry is generic.) Note that fundamental superstring loops can exist only away from branes. Any superstring we observe will have tension µ exponentially diminished from that of the Planck scale by virtue of its location at the bottom of the throat. Values like Gµ < 10−14 (i.e. energies < 1012 GeV) are entirely possible. A typical manifold will have many throats and we expect a distribution of µ, presumably with some Gµ > 10−14 . In addition to their reduced tensions, superstrings should diﬀer from standard ﬁeld theory strings (i.e. vortices) in other important ways: long-lived excited states with junctions and beads may exist, multiple noninteracting species of superstrings may coexist and, ﬁnally, the probability for breaking and rejoining colliding segments (intercommutation) can be much smaller than unity.82 Furthermore a closed string loop may move inside the compacted volume as well. Because of the warped geometry there, such motion may be observed as a variable string tension both along the string length and in time.

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-297

6.2. Current bounds on string tension Gµ and probability of intercommutation p Empirical upper bounds on Gµ have been derived from null results for experiments involving lensing,89–91 gravitational wave background and bursts,92–105 pulsar timing84,106–111 and CMB radiation.9,10,75,76,112–124 We will brieﬂy review some recent results but see Refs. 125 and 126 for more comprehensive treatments. All bounds on string tension depend upon uncertain aspects of string physics and of network modeling. The most important factors include: • The range of loop sizes generated by network evolution. “Large” means comparable to the Hubble scale, “small” can be as small as the core width of the string. Large loops take longer to evaporate by emission of gravitational radiation. • The probability of intercommutation p is the probability that two crossing strings break and reconnect to form new continuous segments. Field theory strings have p ∼ 1 but superstrings may have p as small as 10−3 , depending on the crossing angle and the relative speed. The eﬀect of lowering p is to increase the network density of strings to maintain scaling. • The physical structure of the strings. F1 strings are one-dimensional, obeying Nambu–Goto equations of motion. D1 strings and ﬁeld theory strings are vortices with ﬁnite cores. • The mathematical description of string dynamics used in simulations and calculations. The Abelian Higgs model is the simplest vortex description but the core size in calculations is not set to realistic physical values. Abelian Higgs and Nambu–Goto descriptions yield diﬀerent string dynamics on small scales. • The number of stable string species. Superstrings have more possibilities than simple ﬁeld theory strings. These include bound states of F1 and D1 strings with beads at the junctions and noninteracting strings from diﬀerent warped throats. • The charges and/or ﬂuxes carried by the string. Many eﬀectively one-dimensional objects in string theory can experience nongravitational interactions. • The character of the discontinuities on a typical loop. The number of cusps and/or the number of kinks governs the emitted gravitational wave spectrum. While there has been tremendous progress, all of these areas are under active study. When the model-related theoretical factors are ﬁxed each astrophysical experiment probes a subset of the string content of spacetime. For example, CMB power spectrum ﬁts rely on well-established gross properties of large-scale string networks which are relatively secure but do not probe small sized loops which may dominate the total energy density and which would show up only at large . Analysis of combined PLANCK, WMAP, SPT and ACT data124 implies Gµ < 1.3 × 10−7 for Nambu–Goto strings and Gµ < 3.0 × 10−7 for ﬁeld theory strings. Limits from optical lensing in ﬁelds of background galaxies rely on the theoretically wellunderstood deﬁcit angle geometry of a string in spacetime but require a precise

May 4, 2017 9:54 b2167-ch19

II-298

D. F. Chernoﬀ and S.-H. H. Tye

accounting for observational selection eﬀects. Analysis of the GOODS and COSMOS optical surveys90,91 yields Gµ < 3 × 10−7 . Taken together these observations −7 imply Gµ < ∼ 1–3 × 10 . There is a well-established bound on the gravitational energy density at the time of Big Bang Nucleosynthesis because an altered expansion rate impacts light element yields (e.g. see Ref. 127). The gravitational radiation generated by any string network cannot exceed the bounds. If a network forms large Nambu–Goto loops of one type of string with intercommutation probability p an estimate of the −7 2 BBN constraint is Gµ < ∼ 5 × 10 p (Ref. 128) (this depends implicitly on the loop formation size and a still-emerging understanding of how network densities vary with p). LIGO’s experimental bound on the stochastic background gravitational radiation104 from a similar network implies a limit on Gµ of the same general form as the BBN limit but weaker. Advanced LIGO is projected to reach Gµ ∼ 10−12 for p = 1.128 The same LIGO results can be used to place a reliable, conservative bound of Gµ < 2.6 × 10−4 over a much wider range of possible models, almost independent of loop size and of frequency scaling of the emission.126 More stringent bounds rely on additional assumptions. Consider a speciﬁc set of choices for the secondary parameters of strings: one string species, Nambu–Goto

Fig. 3. Observational constraints limit string tension and intercommutation probability if the network forms large loops. Each line is a particular observational constraint. The region below and to the right of a line is disfavored. Smaller p increases the number density of loops and larger Gµ increases gravitational signal amplitude. Conversely, the region to the left and above a line remains consistent with observational limits. The illustrated constraints are: Big Bang Nucleosynthesis (orange line) and CMB (yellow line) from the gravitational wave spectra in Ref. 128, pulsar time of arrival for the Parkes Pulsar Timing Array (blue line, based on the analytic form in Ref. 109) and for NANOGrav (red dotted line) in Ref. 111. All lines depend upon theoretical modeling of the network, especially the fraction of large loops formed. For example, the initials BP (green)132 are an example in which this fraction has been inferred from simulations and the bound on Gµ at p = 1 indicated (the line has not been calculated). This result is approximately ∼ 30 times less restrictive on Gµ than an equivalent analysis in which all loops are born at a single large size as has been assumed in all the other analyses. (For color version, see page II-CP14.)

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-299

dynamics, only gravitational interactions, large loops (α = 0.1), each with a cusp. The time of arrival of pulses emitted by a pulsar vary on account of the gravitational wave background. When the sequence is observed to be regular the perturbing −9 amplitude of the waves is limited. The current limit is Gµ < ∼ 10 for p = 1 and −12 −3 111 < for p = 10 . Figure 3 is a graphical summary that illustrates some of ∼ 10 the bounds discussed. The BBN (orange line) and CMB (yellow line) constraints are shown as a function of string tension Gµ and intercommutation probability p. The Parkes Pulsar Timing Array limit (blue line)109 and the NANOGrav limit (red dotted line)111 are based on radiating cusp models. Each constraint rules out the area below and to the right of a line. Most analyses do not account for the fact that only a small fraction of horizon-crossing string actually form large loops which are ultimately responsible for variation in arrival times. In this sense the lines may be over-optimistic (see ﬁgure caption). Superconducting strings have also been proposed.129 The bound on supercon−10 130 Interestingly, it was pointed out ducting cosmic strings is about Gµ < ∼ 10 . that the recently observed fast radio bursts can be consistent with being produced by superconducting cosmic strings.131 In short, cosmic superstrings are generically produced toward the end of inﬂation and observations imply tensions substantially less than the original GUT-inspired strings. Multiple, overlapping approaches are needed to minimize physical uncertainties and model-dependent aspects. There is no known theoretical impediment to the magnitude of Gµ being either comparable to or much lower than the current observational upper limits. 7. Scaling, Slowing, Clustering and Evaporating (see Ref. 133) Simulated cosmological string networks converge to self-similar scaling solutions.5,82,114 Consequently Ωlong (Ωloop ), the fraction of the critical density contributed by horizon-crossing strings (loops), is independent of time while the characteristic size of a loop formed at time t scales with the size of the horizon: l = αt for some ﬁxed α. To achieve scaling, long strings that enter the horizon must be chopped into loops suﬃciently rapidly — if not, the density of long strings increases and over-closes the universe. In addition, loops must be removed so that Ωloop stabilizes — if not, loops would come to dominate the contribution of normal matter. Scaling of the string network is an attractor solution in many well-studied models and the universe escapes the jaws of both Scylla and Charybdis. The intercommutation probability determines the eﬃciency of chopping and µ determines the rate of loop evaporation. Studies of GUT strings5 took Gµ ∼ 10−6 (large enough to generate perturbations of interest at matter-radiation equality) and α small (set by early estimates of gravitational wave damping on the long strings) with the consequence Hτ ∼ α/ΓGµ 1 where H is the Hubble constant. Newly formed GUT loops decay quickly. The superstrings of interest here have smaller Gµ so that loops of

May 4, 2017 9:54 b2167-ch19

II-300

D. F. Chernoﬀ and S.-H. H. Tye

a given size live longer. In addition, recent simulations134–138 produce a range of < large loops: 10−4 < ∼ α ∼ 0.25. The best current understanding is that ∼ 10–20% of the long string length that is cut up goes into loops comparable to the scale of the horizon (α ∼ 0.1) while the remaining ∼ 80–90% fragments to much smaller size scales.139,140 The newly formed, large loops are the most important contribution for determining today’s loop population. Cosmic expansion strongly damps the initial relativistic center of mass motions of the loops and promotes clustering of the loops as matter perturbations grow.141,142 Clustering was irrelevant for the GUT-inspired loops. They moved rapidly at birth, damped brieﬂy by cosmic expansion and were re-accelerated to mildly relativistic velocities by the momentum recoil of anisotropic gravitational wave emission (the rocket eﬀect) before fully evaporating.93,143,144 GUT loops were homogeneously distributed throughout space. By contrast, below a critical tension Gµ ∼ 10−9 all superstring loops accrete along with the cold dark matter.142 Loops of size l = lg ≡ ΓGµt0 are just now evaporating where t0 is the age of the universe. The mean number density of such loops is dominated by network fragmentation when the universe was most dense, i.e. at early times. When they were born they came from the large end of the size spectrum, i.e. a substantial fraction of the scale of the horizon. The epoch of birth is ti = lg /α = ΓGµt0 /α. For Gµ < 7 × 10−9 (α/0.1)(50/Γ) loops are born before equipartition in ΛCDM, i.e. ti < teq . The smallest loops today have lg ≈ 40pc(Gµ/2×10−10) with characteristic mass scale Mg = 1.7 × 105 M (Gµ/2 × 10−10 )2 for Γ = 50. Loop number and energy densities today are dominated by the scale of the gravitational cutoﬀ, the smallest loops that have not yet evaporated. The characteristic number density dn/d log l ∝ (ΓGµ)−3/2 (αteq /t0 )1/2 /t30 and the energy density dρloop /d log l ∼ ΓGµt0 dn/d log l. The latter implies Ωloops ∝ αGµ/Γ whereas Ωlong ∝ ΓGµ. In scenarios with small µ it’s the loops that dominate long, horizoncrossing strings in various observable contexts. The probability and rate of local lensing are proportional to the energy density. Loops are accreted as the galaxy forms. Figure 4 illustrates schematically the constraints for a loop with typical initial peculiar velocity to be captured during galaxy formation and to remain bound today. The loop must lie within the inner triangular region which delimits small enough tension and early enough time of formation. Above the horizontal line capture is impossible; below the diagonal line detachment by the rocket eﬀect has already occurred. The critical tension for loop clustering in the galaxy is set by the right-hand corner of the allowed region. To brieﬂy summarize: F , the enhancement of the galactic loop number density over the homogeneous mean, simply traces E, the enhancement of cold dark matter over ΩDM ρc , for critical density ρc . This encapsulates conclusions of a study of the growth of a galactic matter perturbation and the simultaneous capture and escape of network-generated loops.142 At a typical galactocentric distance of 11 kpc log10 F¯ = log10 E¯+f (y) where log10 E¯ = 5.5 and f (y) = −0.337−0.064y 2 −g(y) and

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-301

Fig. 4. Bounds on formation time and string tension for a loop with initial velocity vi = 0.1 to be captured at physical radius 30 kpc by galaxy formation. (1) Upper bounds on the formation time ti /t0 are given by the horizontal lines. The condition that cosmic drag lower the velocity to less than the circular rotation velocity today is given by the red line. The more stringent condition that capture occur at 30 kpc is given by the turquoise line. (2) Upper bounds on the string tension Gµ are given by the diagonal lines. The condition that the loop be younger than its gravitational wave decay timescale is given by the red line. The more stringent condition that the loop not be accelerated out of the galaxy by today is given by the green line. (3) The shaded region encompasses string tensions and formation times giving bound loops at 30 kpc for vi = 0.1 and α = 0.1. The critical value of Gµ below which clustering is possible is determined by the upper right hand corner of the green and turquoise lines. Lowering vi raises the limit on ti /t0 (horizontal lines moves upward); lowering α shifts the bound to smaller Gµ (diagonal lines moves leftward). Shifting the loop orbital scale to smaller values (say the solar position) requires earlier formation times (horizontal lines shifts down) and allows larger Gµ (green line shifts to the right but is limited by the red line which is ﬁxed). (4) The geometric symbols illustrate the sensitivity of the rocket eﬀect. These are numerical experiments examining the outcome today (t = t0 ) for groups of 10–20 loops captured at 30 kpc with slightly diﬀerent string tensions: stars = all loops bound, boxes = all loops ejected, triangles = some bound and some ejected (for clarity the points are slightly oﬀset in the vertical but not the horizontal direction). (For color version, see page II-CP15.)

g(y) = 5(1 + tanh(y − 7)). The tension-dependent deviation from loops as passive tracers of cold dark matter is f (y), y = log10 (Gµ/10−15 ), ﬁt for 0 ≤ y ≤ 5 from the capture study. Clustering saturates for small tensions: f (y) = f (0) for y < 0. The over-bar indicates quantities averaged over the spherical volume and, in the case of the loops, a weighting by loop length in the capture study. The extra piece g(y) plays a role for Gµ > 10−10 where it describes the suppression in clustering as one approaches the upper right corner of the triangle in Fig. 4. Numerical simulations have not yet accurately determined it. A single f (y) ﬁts a range of radii as long as E¯ 1 (typically, galactic distances less than ∼ 100 kpc). So we can easily have an enhancement F ∼ 105 for Gµ ∼ 10−14 at the Sun’s position. By comparison, GUT loops would have F = 1 for all positions and tensions.

May 4, 2017 9:54 b2167-ch19

II-302

D. F. Chernoﬀ and S.-H. H. Tye

The model accounts for the local population of strings and we have used it to estimate microlensing rates, and LISA-like, LIGO-like and NANOGrav-like burst rates. 7.1. Large-scale string distribution We will start with a “baseline” description (loops from a network of a single, gravitationally interacting, Nambu–Goto string species with reconnection probability p = 1). The detailed description142 was motivated by analytic arguments139,140 that roughly 80% of the network invariant length was chopped into strings with very small loop size (α ∼ Gµ) and by the numerical result134–138 that the remaining 20% formed large, long-lived loops (α = 0.1). At a given epoch loops are created with a range of sizes but only the “large” ones are of interest for the local population. The baseline description is supposed to be directly comparable to numerical simulations which generally take p = 1. The most recent simulations132 are qualitatively consistent. Next, we parameterize the actual “homogeneous” distribution in the universe when string theory introduces a multiplicity of string species and the reconnection probability p may be less than 1. And ﬁnally we will form the “local” distribution which accounts for the clustering of the homogeneous distribution. In a physical volume V with a network of long, horizon-crossing strings of tension µ with persistence length L there are V /L3 segments of length L. The physical energy density is ρ∞ = µL/V = µ/L2 . The persistence length evolves as the universe expands. A scaling solution demands L ∝ t during power law phases. There are also loops within the horizon; their energy is not included in ρ∞ and the aim of the model is to infer the number density of loops of a given invariant size. Kibble145 developed a model for the network evolution for the long strings and loops in cosmology. It accounted for the stretching of strings and collisional intercommutation (long string segments that break oﬀ and form loops; loops that reconnect to long string segments). A variety of models of diﬀering degrees of realism have been studied since then, guided by ever more realistic numerical simulations of the network. As a simple approximate description we focus on the Velocity One Scale model146 in a recently elaborated form.147,148 The reattachment of loops to the network turns out to be a rather small eﬀect and is ignored. We extended existing treatments by numerically evaluating the total loop creation rate in ﬂat ΛCDM cosmology. The loop energy in a comoving volume varies like E˙ l = Cρ∞ pva3 /L where C is the chopping eﬃciency, p is the intercommutation probability and v is the string velocity. All quantities on the right-hand side except p vary in time; C is a parameterized ﬁt in matter and radiation eras. By integrating the model from large redshift to the current epoch one evaluates the fraction of the network that is lost to loop formation. The rate at which the loop energy changes is E˙ l = Aµa3 /(p2 t3 ) where A is a slowly varying function of redshift z and p shown in Fig. 5. The plotted variation of A with redshift indicates

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings p2 t3 µ

II-303

a 3 dE loop dt

8

6

4

2

2

0

2

4

6

Lo g 10 z

Fig. 5. The redshift-dependent rate of energy loss to loops experienced by the string network in ΛCDM cosmology for a given intercommutation probability p. This is an important input for predicting properties of today’s loops. The ordinate is the dimensionless measure of the rate of energy loss to loops, A = E˙ l /(µa3 /(p2 t3 )). If the scale factor a(t) were a power law in time then A would be constant with respect to z. This is not the case in the ΛCDM cosmology with transitions from radiation-to-matter-to-Λ dominated regimes. Diﬀerent lines show results for p = 1 (top) to p = 10−3 (bottom) in powers of 10. If network density varied as 1/p2 then all these lines would overlap. This is approximately true for small p but not for p ≥ 0.1.

departures from a pure scaling solution (consequences of the radiation to matter transition for a(t) and the implicit variation of chopping eﬃciency C). Knowing A as a function of redshift and p is the input needed to evaluate of the loop formation rate. Loops form with a range of sizes at each epoch. Let us call the size at the time of formation αt. A very important feature predicted by theoretical analyses is that a large fraction of the invariant length goes into loops with α ∼ Gµ or smaller. The limiting behavior of numerical simulations as larger and larger spacetime volumes are modeled suggests ∼ 80% of the chopped up long string bears that fate. Such loops evaporate rapidly without contributing to the long-lived local loop population. The remaining 20% goes into large loops with α ∼ 0.1. Let us call the fraction of large loops f = 0.2. The birth rate density for loops born with size αtb is dn fA δ(l − αtb ). (31) = dtdl αp2 t4b A loop formed at tb with length lb shrinks by gravitational wave emission. Its size is l = lb − ΓGµ(t − tb )

(32)

at time t (Γ ∼ 50). The number density of loops of size l at time t is the integral of the birth rate density over loops of all length created in the past. For l < αt or, equivalently,

May 4, 2017 9:54 b2167-ch19

D. F. Chernoﬀ and S.-H. H. Tye

II-304

tb < t we have dn (l, t) = dl tb =

f Aα2 p2

a(tb ) a

3

Φ3 , (l + ΓGµt)4

(33)

l + ΓGµt , αΦ

(34)

ΓGµ . α

(35)

Φ = 1+

The scaling dn/dl ∝ (l + ΓGµt)−4 was already noted by Kibble.145 The loop number distribution peaks at zero length but the quantity of interest in lensing is number weighted by loop length, ldn/dl. The characteristic scale at time t is lg = ΓGµt, i.e. roughly the size of a newly born loop that would evaporate in total time t. The distribution ldn/dl peaks at l = (2/3)lg . For Gµ < 7 × 10−9 (α/0.1)(50/Γ) the loops near lg today were born before equipartition, teq . We use a ∝ t1/2 to simplify the expression to give l

dn x = 5/2 dl (1 + x)

fA p2 t30

(ΓGµ)−3/2

αteq t0

1/2 ,

(36)

where t0 is today and x ≡ l/lg . The numerical results are A ∼ 7.68 for p = 1, f = 0.2, α = 0.1 and Γ = 50 (and from ΛCDM teq = 4.7 × 104 yr and t0 = 4.25 × 1017 s). These give dn x −3/2 l = 1.15 × 10−6 µ kpc−3 , (37) dl baseline (1 + x)5/2 −13 lg = 0.0206 µ−13 pc,

(38)

Mg = 0.043 µ2−13 M .

(39)

Here, µ−13 ≡ Gµ/c2 /10−13 is an abbreviation for the dimensionless string tension in units of 10−13 . The baseline distribution lies below147,148 on account of loss of a signiﬁcant fraction of the network invariant length to small strings and of adoption the ΛCDM cosmology. Next, string theory modiﬁcations to the baseline are lumped into a common factor G dn dn =G (40) dl homog dl baseline to give the description of the actual homogeneous loop distribution. Prominent among expected modiﬁcations is the intercommutation factor p. Large scale string simulations have not reached a full understanding of the impact of p < 1 though there is no question that G increases as a result. The numerical treatment of the Velocity One Scale model implies that A is a weak function of p for small p and,

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-305

in that limit, dn/dl ∝ 1/p2 . Ultimately, this matter will be fully settled via network simulations with p < 1. String theory calculations of the intercommutation probability suggests p = 10−1 – 10−3 implying G = 102 – 106. The number of populated, noninteracting throats that contain other types of superstrings is a known unknown and is unexplored. In our opinion, there could easily be 100’s of such throats for the complicated bulk spaces of interest. In the string theory scenarios G = 1 is a lower limit. For the purposes of numerical estimates in this paper we adopt G = 102 (with a given tension) as the most reasonable lower limit. Much larger G are not improbable while lower G are unlikely. We should emphasize that strings in diﬀerent throats would have diﬀerent tensions, so adopting a single tension here yields only a crude estimate. 7.2. Local string distribution If a loop is formed at time t with length l = αt then its evaporation time τ = l/ΓGµ. For Hubble constant H at t the dimensionless combination Hτ = α/(ΓGµ) is a measure of lifetime in terms of the universe’s age. Superstring loops with α = 0.1 and small µ live many characteristic Hubble times. New loops are born with relativistic velocity. The peculiar center of mass motion is damped by the universe’s expansion. A detailed study of the competing eﬀects (formation time, damping, evaporation, eﬃcacy of anisotropic emission of gravitational radiation) in the context of a simple formation model for the galaxy shows that loops accrete when µ is small. The degree of loop clustering relative to dark matter clustering is a function of µ and approximately independent of l. Smaller µ means older, more slowly moving loops and hence more clustering. Below we give a simple ﬁt to the numerical simulations to quantify this eﬀect. The spatially dependent dark matter enhancement in the galaxy is E(r) =

ρDM (r) , ΩDM ρc

(41)

where ρDM (r) is the local galactic dark matter density and ΩDM ρc is average dark matter density in the universe. Low tension string loops track the dark matter with a certain eﬃciency as the dark matter forms gravitationally bound structures.142 We ﬁt the numerical results by writing the spatially dependent string enhancement to the homogeneous distribution as equal to the dark matter enhancement times an eﬃciency factor F (r) = E(r)10f (y) , −0.337 − 0.064y 2 − 5(1 + tanh(y − 7)) f (y) = −0.337 y = 2 + log10 µ−13

(42) for 0 ≤ y, for y < 0,

(43) (44)

−13

and µ−13 is the dimensionless tension in units of 10 . The clustering is never 100% eﬀective because the string loops eventually evaporate. The eﬃciency saturates at

May 4, 2017 9:54 b2167-ch19

D. F. Chernoﬀ and S.-H. H. Tye

II-306

y = 0 or Gµ/c2 = 10−15 . The local string population is enhanced by the factor F with respect to the homogeneous distribution dn dn dn (r) = F (r) = F (r)G . (45) dl local dl homog dl baseline To complete the description of the local loop population we adopt empirical ﬁts to the galaxy’s dark matter halo.149 Model I is a “cored galaxy” center in which the central, limiting dark matter density is zero. Since winds from stars and ejection/heating by supernovae lift baryons out of star-forming regions, the center, they reduce the gravitational potential and tend to lower the dark matter density. Model II is a “cusped galaxy” center in which the density formally diverges. Many N-body simulations show that collisionless structure formation in ΛCDM yields such proﬁles. The two models are thought to bracket the range of physical possibilities in the central regions and are in general agreement on large scales. Let us deﬁne the eﬀective number density of loops as the energy density in the distribution of loops of all size, ρloop , divided by energy of a loop of size equal to the characteristic gravitational cutoﬀ ρloop µlg dn l = dl . dl lg

n ¯=

G c2 10 Log10 n kpc

15

10

(46) (47)

8

3

5

R kpc 5

10

15

20

25

5

10

Fig. 6. The eﬀective number density of string loops in two galactic models for a range of string tensions. The dotted red lines show results for a galactic model with a cusp at the center, the solid blue lines for a model with a core. The tension is Gµ/c2 = 10−15 for the uppermost curve and increases by powers of 10 until reaching 10−8 for the lowest curve. The eﬀective number density depends upon the clustering of loops, the dark matter distribution within the galaxy, the network scaling and loop distribution and G the enhancement of string theory strings compared to ﬁeld theory strings. (For color version, see page II-CP15.)

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-307

Figure 6 plots the eﬀective number density of string loops for two descriptions of the galaxy’s dark matter distribution (core and cusp) as a function of galactocentric radius. The diﬀerent lines show a range of string tensions, all for G = 102 . The quantity n ¯ depends upon the eﬃcacy of loop clustering, the dark matter distribution within our galaxy and the network scaling model. There are two curves plotted for each tension. These agree at large radii but diﬀer near the galactic center where the detailed form of the dark matter distribution is uncertain. For many experiments the rate of detection scales as n ¯ weighted by powers of lg . The microlensing rate, for example, is proportional to the product of two factors: the number of string loops along a given line of sight ∝ n ¯ and the length of an individual ¯ lg ∝ n ¯ Gµ. At the Sun’s position, the string loop ∝ lg . The event rate scales like n eﬀective number density increase by ∼ 15 orders of magnitude as Gµ/c2 drops from 10−8 to 10−15 . The microlensing event rate increases by ∼ 8 orders of magnitude. As we will discuss shortly, for the same variation in tension the timescale for an individual microlensing event decreases by ∼ 8 orders of magnitude. This creates a huge range of timescales of interest for the duration and rate of microlensing events. 8. Detection 8.1. Detection via Microlensing Consider a straight segment of string oriented perpendicular to the observer’s line of sight with respect to a background source as shown in Fig. 7. Let two photons from the source travel toward the string. The photons do not suﬀer any relative

Fig. 7. (Left) The conical spacetime geometry near a short segment of string (red line) permits photons to travel two paths (black arrows) from source (yellow ball) to observer when source, string and observer are nearly aligned. The ideal observer perceives two images of the source split by a small angle (black arc) proportional in size to string tension. For low string tension, the observer cannot resolve the separate images; however, the ﬂux may be easily measured to be twice that of a single image (in the point source limit) at each wavelength (negligible Kaiser–Stebbins eﬀect). (Right) The microlensing ampliﬁcation of ﬂux for a point source as a function of time for a string (green) is compared to a Newtonian point masses (blue; the ratio of impact parameter to Einstein radius is 0.31). No other known astrophysical sources produce this digital microlensing. (For color version, see page II-CP15.)

May 4, 2017 9:54 b2167-ch19

II-308

D. F. Chernoﬀ and S.-H. H. Tye

deﬂection during the ﬂy-by as long as they pass around the string in the same sense. However, there is a small angular region about the string with two paths from the source to the observer. Background sources within angle ΘE = 8πGµ = 1.04 × 10−3(Gµ/2 × 10−10) arcsec form two images. Unlike the case of a point mass, shear and distortion are absent. The angular size of a sun-like star at distance R is Θ /ΘE = 4.5 × 10−4 (2 × 10−10 /Gµ)(10 kpc/R) so galactic stellar sources generally −13 appear point-like for Gµ > ∼ 10 . Compact halo objects lens background sources150,151 and unresolved, lensed sources will appear to ﬂuctuate achromatically in brightness. This is microlensing. Experimental eﬀorts to detect microlensing phenomena have borne considerable fruit.152 Likewise, loops of superstring microlens background sources but have a special property: the source brightness varies by a factor of 2 as the angular region associated with the string passes across the observer-source line of sight. The amplitude variation, schematically compared in Fig. 7, is quite distinctive for point sources. The internal motions of a loop are relativistic and generally dominate the motion of the source and the observer. Numerical calculations have established that microlensing occurs when light passes near a relativistic, oscillating string loop153 ; the eﬀect is not limited to a stationary string. of loops dn/dl is proThe total rate of lensing RL implied for a distribution √ 2 portional to the solid angle swept out per time cl/ 3r for loop l at distance r, or √ √ RL = dldrr2 dn/dl(cl/ 3r2 ). Small tensions give large lensing rates, RL ∝ 1/ Gµ because the integral over dn/dl ∝ l−2.5 is dominated by l ∼ lg , the gravitational cutoﬀ. Figure 8 illustrates the hierarchy of relevant timescales. The characteristic duration of the event is δt = RΘE /c ∼ 630 s(R/10 kpc)(Gµ/2 × 10−10 ). Loops bound to the galaxy have center-of-mass motions vh ∼ 220 km s−1 ; microlensing of a given source will repeat ∼ c/vh times; new sources are lensed at rate ∼ (vh /c)RL . The characteristic loop oscillation timescale which governs the intervals in repetitive microlensing of a single source is tosc ∼ lg /c = 135 yr (Gµ/2 × 10−10 ). The unique ﬁngerprint of loop lensing is repeated achromatic ﬂux doubling (digital microlensing). Detection eﬃciency for digital lensing depends upon the string tension, the magnitude and angular size of the background stars and the time sampling of the observations. Estimates can be made for any experiment which repeatedly measures the ﬂux of stellar sources. The ﬁrst string microlensing search154 was recently completed using photometric data from space-based missions CoRoT155,c and RXTE.156,d The methodology was potentially capable of detecting strings with tensions 10−16 < Gµ < 10−11 though the expected number of detections was limited by the available lines of sight studied c The

2006 mission was developed and operated by the CNES, with the contribution of Austria, Belgium, Brazil, ESA (RSSD and Science Program), Germany and Spain. d The 1995 mission was developed and operated by NASA.

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-309

Flux

Time

Fig. 8. A Schematic Pattern. Digital microlensing doubles the ﬂux over the time period that source, loop and observer are aligned within the deﬁcit angle created by the string (red line). The repetition interval for lensing a particular source is the loop oscillation timescale (green line) and ∼ 103 repetitions in total occur (yellow line). The timescale for a new source to be lensed by the original loop (or the original source by a new loop) is much longer (black line). (For color version, see page II-CP16.)

in the course of the missions. In principle, any photometry experiment that makes repeated ﬂux measurements of an intrinsically stable astronomical source has the power to limit a combination of the number density of loops and string tension. One interesting possibility is the satellite GAIA launched by the European Space Agency in December 2013 designed for astrometry. It is monitoring each of about 1 billion stars about 70 times over a period of 5 years. Another is the Large Synoptic Survey Telescope (LSST) to be sited in Chile within the decade to photograph the entire observable sky every few days. Previous estimates for the rate of detections for GAIA141 and LSST157 were encouraging. Now, detailed calculations for WFIRST are also available. 8.2. WFIRST microlensing rates (see Ref. 158) The Wide-Field Infrared Survey Telescopee (WFIRST) is a NASA space observatory designed to perform wide-ﬁeld imaging and slitless spectroscopic surveys of the near infrared sky. WFIRST will carry out a microlensing survey program for exo-planet detection in the direction of the galactic bulge by observing a total of 2.8 square degrees for 1.2 years primarily in a broad long wavelength ﬁlter (Wband 0.927–2.0 µm). Repeated, short exposures (∼ 1 minute) of the same ﬁelds are the key to searching for and to monitoring the ampliﬁcation of bulge sources by star–planet systems along the line of sight. Fortuitously, the WFIRST experiment e http://wﬁrst.gsfc.nasa.gov/.

May 4, 2017 9:54 b2167-ch19

D. F. Chernoﬀ and S.-H. H. Tye

II-310

is also sensitive to cosmic strings. We have evaluated the expected microlensing rate by cosmic strings for a realistic distribution of stars (stellar types, distances, velocities, etc.), dust obscuration and survey parameters (ﬂux sensitivity, time of exposure and angular scale of stars). The lensing rate is split into digital events (the ﬂux doubles), analog events (all potentially measurable ﬂux enhancements given the signal-to-noise of the observations) and total events (all geometric conﬁgurations that can lens, whether detectable or not) evaluated for two galactic dark matter models (with cusp and with core at galactic center). Figure 9 shows the string lensing rate per square degree as a function of string tension Gµ/c2 for G = 102 . The lensing rate for digital and analog events exceeds 10 per square degree per year for 10−14 < Gµ/c2 < 10−10 for the cusp model and for 10−13 < Gµ/c2 < 10−11 for the core model. The results for the cusp and core models should be regarded as establishing the range of astrophysical possibilities for the loop density distribution within the experiment. The range of string theory possibilities (the variation of G) remains substantial. Digital and spatially resolved analog events can be reliably detected. Blending of sources in analog events needs to be simulated to fully characterize the fraction of such events that will be detectable. In any case, the decrease from blending is limited: sources brighter than magnitude 23 in the J-band (1.131– 1.454 µm) contribute at least 10% of the total analog rate, are well-separated and produce light curves detectable at high signal-to-noise. These results show that WFIRST microlensing searches can probe hitherto unexplored ranges of string tension. Future surveys of wider areas of the bulge and/or surveys lasting for longer periods of time can potentially scale up the total Log10 4

2

0

14

12

10

8

Log10

2

4

Fig. 9. The intrinsic string lensing rate (events per sq degree per year) for WFIRST’s study of the bulge for G = 102 . The three solid lines show digital (red), analog (green) and total (blue) event rates for the dark matter model with a cusp. The three dashed lines (same meaning for the colors) give the event rate for the model with a core. The analog rate does not account for blending which may diminish the detectable rate by no more than a factor of 10. (For color version, see page II-CP16.)

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-311

expected rate of detection by as much as 100. Surveys with shorter exposures times are able to discern larger numbers of digital events. 8.3. Gravitational waves Besides uncovering a relic from the earliest moments of the universe’s formation, the observation of a superstring microlensing event provides some immediate information. If the event is resolved in time, the characteristic string tension is inferred. If repetitions are seen, the characteristic loop size is constrained. And any detection provides a precise sky location for follow up. Some classes of strings create electromagnetic signatures but all generate gravitational radiation possibly observable by LISA-like (hereafter, LISA) and LIGO-like instruments. The scenario that LISA detects the distinctive signature of local loop emission has been contemplated for a blind, all-sky search.159 They relied on the enhancement from local clustering141 and concluded that strings with 10−19 < Gµ < 10−11 were potentially detectable via fundamental and low-order harmonics. They also estimated the background from the galaxy and from the universe as a whole. Therefore, it is no surprise that a loop microlensing event oﬀers some exciting possibilities. The key considerations are: (1) The precise location aﬀorded by microlensing dramatically reduces the signal search space on the sky. (2) A generic loop radiates a full range of harmonics with frequency fn = nf1 = 2cn/l because the equations of motion are intrinsically nonlinear. (3) Many string harmonics fall near LISA’s peak sensitivity n∗ = f∗ /f1 1 (hLISA ∼ 10−23.8 at f∗ ∼ 10−2.3 Hz for 1 year periodic variation160 ; tension µcrit = 2.9 × 10−12 corresponds to a 1 year period; n∗ = 1.6 × 105(µ/µcrit )). (4) Galactic binary interference becomes problematic only at f < fWD where fWD ∼ 10−2.6 Hz.161 LISA’s veriﬁcation binaries have known positions on the sky, orbital frequencies, masses, distance limits etc. determined by optical measurements and other means.160 A long-lived superstring loop detected by microlensing shares similar observational advantages: it has known position on the sky, distance upper limit estimated based on the microlensed star and emits a distinctive signal. This creates the ideal observational situation in which much of the data analysis can be conducted with a single fast Fourier transform. Diﬀerent types of loops generate diﬀerent gravitational wave signatures. The solution for a Nambu–Goto loop’s motion is the sum of left- and right-moving onedimensional waves subject to nonlinear constraints. When both modes are smooth (no derivative discontinuities) the dynamical solution can form a cusp. At that instant a bit of the string reaches the speed of light. When one mode is smooth but the other has a discontinuity the loop’s dynamical solution includes a kink, a discontinuity in the string’s tangent vector that moves along the string at the speed of light in one direction. When both modes have discontinuities we call the solution a k-kink. The presence of cusps, kinks and k-kinks may be inferred from the paths of the tangent vectors of the left and right moving modes which are constrained to

May 4, 2017 9:54 b2167-ch19

D. F. Chernoﬀ and S.-H. H. Tye

II-312

lie on the surface of a sphere. These distinctions are relevant for gravitational wave emission: the cusp generates bursts of beamed radiation, the kink has a pulsar-like beam that generically sweeps across a substantial angle in the sky and the k-kink emits into a substantial fraction of the sky. The tendency for string loops to explore part of the internal space of a typical compactiﬁcation during oscillations74 alters the cusp and kink beaming patterns with potentially important repercussions for rates of detection in gravitational wave experiments.162–164 We model the beaming anisotropy, the harmonic power and phase of the gravitational wave signal (qualitatively and quantitatively extending previous analyses159 ). Standard signal detection theory characterizes the strength of detection and the precision with which measurements can be made by comparing the putative signal to noise sources. Here, we consider only detector noise described as an additive, stationary, random Gaussian process with one-sided spectral noise at frequency f is S(f ). The strength of signal h(t) is ρSNR = 2h, h where the symmetric ∞ ˜ ∗ (f ) )h inner product is g, h = −∞ df g˜(f ˜(f ) is the Fourier transform of g(t) S(f ) and g (the factor of 2 stems from one-sided noise versus two-sided signal). Figure 10 summarizes ρSNR for detecting a single source with beamed cusp, beamed kink and generic, unbeamed emission. Upper and lower line pairs show the eﬀect of ignoring and accounting for the galactic white dwarf interference. Confusion is minimal because the string signal extends to frequencies where the binary density is small and because the signal overlap between string and binary sources is small. Log10 Ρ 8

6

4

2

14

13

12

11

10

9

Log10 G

c2

Fig. 10. The log of LISA’s signal strength (log10 ρSNR ) for 1 year of observation of cusp, kink and k-kink (top to bottom) at a distance of 1 kpc with a known position on the sky as a function of string tension. The orientation of the cusp and kink is for the maximum signal. Upper/lower line pairs show the eﬀect of ignoring/accounting for white dwarf interference. For tensions with Gµ/c2 > 3 × 10−12 the fundamental loop period exceeds the length of the observation so excess power is seen but the frequency resolution is insuﬃcient to measure individual harmonics.

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-313

The time-averaged microlensing rate is roughly proportional to the loop’s maximum projected area divided by its period of oscillation but the gravitational wave emissivity is dominated by speciﬁc loop conﬁgurations having rapidly moving segments when the loop is most contracted. Microlensing and beamed, gravitational wave emission generally will not and need not be simultaneous to be informative. We elaborate on this important point. Generic cusps and kinks have characteristic beaming patterns described above and undergo fully three-dimensional motions that yield time-averaged microlensing rates that are not particularly angle-dependent. Microlensing is a favorable means to locate such loops but the high frequency, directed gravitational wave beams they emit are visible only from special directions. A microlensing detection does not signiﬁcantly inﬂuence the probability of intercepting beamed emission in experiments with duration exceeding the fundamental loop period. Instead, there is a roughly constant, low probability that a cusp that is present will beam in the observer’s direction and modest probability that a kink will (the beam’s angular size is smaller at higher frequencies of emission). The most likely conﬁguration for k-kinks is a ﬂat, degenerate box-like orbit with a well-deﬁned plane. Special observers who lie in that plane will not see microlensing but most others are sensitive to the loop’s projected motion and see the eﬀect if a suitable background source exists. The gravitational wave emission of galactic k-kinks is not strongly beamed but, as the signal estimates show, is still potentially detectable. Unbeamed emission from loops that happen to contain cusps and kinks should also be detectable whether or not the high frequency beam impinges on the observer. To summarize, there are excellent chances to detect some of this emission once the precise location of the source is known. If the fundamental string period is less than the duration of the experiment then the string’s frequency comb can be resolved and ﬁt. Otherwise, there will be excess frequency-dependent power seen in the microlensing direction. The issue of confusion with local and cosmological string loops will be addressed in future studies. While nondetection by LISA would place bounds on tension and loop size the most informative scenario will be measuring the individual harmonics of gravitational wave emission. Cusps have almost smooth spectral amplitude and phase distributions, kinks show some mode-to-mode variation and k-kinks progressively more. The variation with harmonic order provides immediate information regarding derivative discontinuities, e.g. an experimental determination of cusp and kink content. This is directly applicable for understanding the string network in a cosmological context. Speciﬁcally, the discontinuities are created when horizon-crossing strings are chopped into loops but are modiﬁed as the strings are stretched out by the expanding universe. It is important to understand if the loops exclusively contain discontinuities or, also, cusps. This relates to anticipated cosmologically distant string sources that contribute to the gravitational wave background currently sought.

May 4, 2017 9:54 b2167-ch19

II-314

D. F. Chernoﬀ and S.-H. H. Tye

A measurement of the fundamental frequency is equivalent to a precise measurement of the invariant string length. It can be determined in 1 year to relative accuracy at least as good as ∆f1 /f1 ∼ 1/ρSNR . The gravitational decay time for the typical string loop is likely to be of order the age of universe. A study of Fig. 10 suggests that the expected decay in a year ∆f1 /f1 ∼ 10−10 is too small to measure directly unless the loop is unexpectedly close to the end of its life or very nearby. Instead, ﬁrst, one will measure changes in frequency from center of mass motion and acceleration within the galaxy and place upper limits on ∆f1 /f1 . The latter will constrain string couplings to axion-like ﬁelds and astrophysical eﬀects of nearby matter. Since loops are long-lived sources emitting over a wide range of harmonics, eventually, multiple detectors will be brought to bear and the loop’s decay measured in a direct fashion. This program can yield a precise determination of the string tension.

9. Summary String theory contains a consistent quantum gravity sector and possesses a deep mathematical and physical structure. It has become apparent that the theory’s richness enables it to describe a huge multiplicity of solutions and that testing whether string theory is the theory of nature is no easy task. Fortunately and fundamentally, it possesses an unambiguous property, namely, the presence of strings. If the evolution of the early universe is able to produce stable or metastable closed strings that stretch across the sky, then these unusual objects may provide distinctive evidence for the theory. As more and more observational data have been collected and analyzed, most cosmologists now conclude that the universe started with an inﬂationary epoch. This conclusion quite naturally motivates investigations of the following interrelated issues: how inﬂation is realized, how and what type of cosmic strings are produced, how to search for and detect these objects and how to measure their properties. Here, we review how inﬂation may be realized within string theory and whether there are distinctive stringy features in string theory-motivated inﬂationary scenarios. In lieu of presenting a full review we sample some of the scenarios proposed already to give readers a taste of what can be achieved. As there are many ways to realize inﬂation within the string theory framework, we hope better data will eventually indicate how inﬂation happens in string theory. As we have pointed out, some stringy inﬂationary scenarios can have interesting distinct features to be detected. We are also encouraged to note that cosmic superstrings are naturally produced after inﬂation. Next, we review and discuss cosmic superstring features and production. One underlying message is that, because of the intricacies of ﬂux compactiﬁcation, multiple types of cosmic superstrings with a wide range of properties may be produced. Such superstrings are quite diﬀerent from the vortices well studied within ﬁeld theory. The tension may be drawn from a discrete spectrum and diﬀerent string states

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-315

can cooperate to form junctions and beads. A future experimental observation of any of these properties can legitimately be taken as a smoking gun for string theory. We review ongoing searches for strings in cosmology. Although speciﬁc experimental results are model-dependent the diversity of approaches has jointly lowered the upper limit for string tension by about two or more orders of magnitude below GUT scales. This is an important progress. However, well-motivated string models with warped geometries can yield characteristic string tensions that are far smaller than current upper limits, perhaps all the way to the general scales of the Standard Model itself. As Feynman famously said about miniaturization, “there’s plenty of room at the bottom”. Likewise, for string theory, there is a huge range of plausible tensions yet to be explored. We review the new physics and cosmology of these relatively low tension cosmic strings, particularly the tendency for string loops to track structure formation in the matter-dominated eras. As low tension strings decay slow enough, their clustering happens like dark matter, with a 105 order of magnitude enhancement in density within the galaxy. This opens up intriguing possibilities to seek strings having tensions many orders of magnitude less than current upper limits. We review one promising possibility that, for tension 10−14 ≤ Gµ < 10−7 , string loops within the Galaxy microlens stars within the Galaxy. This type of microlensing qualitatively diﬀers from the well-studied microlensing eﬀect of compact objects in the Galactic halo. Ongoing (e.g. GAIA of ESA) as well as planned (e.g. LSST, WFIRST) searches for variable stars and/or exo-planets are sensitive to cosmic superstring microlensing. This gives us hope that much improved upper limits on string tension and/or actual detections will emerge in the coming years. We outline the important role that gravitational wave observations might play if and when a string loop is detected. Acknowledgments We thank Jolyon Bloomﬁeld, Tom Broadhurst, Jim Cordes, John Ellis, Eanna Flanagan, Romain Graziani, Mark Hindmarsh, Renata Kallosh, Andrei Linde, Liam McAllister, Levon Pogosian, Ben Shlaer, Yoske Sumitomo, Alex Vilenkin, Barry Wardell, Ira Wasserman and Sam Wong for discussions. We gratefully acknowledge the support of the John Templeton Foundation (University of Chicago 37426Cornell FP050136-B). DFC is supported by NSF Physics (Astrophysics and Cosmology Theory, No. 1417132); SHHT is supported by the CRF Grants of the Government of the Hong Kong SAR under HUKST4/CRF/13G. References 1. A. H. Guth, Phys. Rev. D 23 (1981) 347, http://adsabs.harvard.edu/abs/ 1981PhRvD..23..347G. 2. A. D. Linde, Phys. Lett. B 108 (1982) 389, http://adsabs.harvard.edu/abs/ 1982PhLB..108..389L.

May 4, 2017 9:54 b2167-ch19

II-316

D. F. Chernoﬀ and S.-H. H. Tye

3. A. Albrecht and P. J. Steinhardt, Phys. Rev. Lett. 48 (1982) 1220, http:// adsabs.harvard.edu/abs/1982PhRvL..48.1220A. 4. D. Baumann and L. McAllister, Inﬂation and string theory, arXiv:1404.2601. 5. A. Vilenkin and E. P. S. Shellard, Cosmic strings and other topological defects (2000), http://adsabs.harvard.edu/abs/2000csot.book.....V. 6. P. Ade et al., Phys. Rev. Lett. 112 (2014) 241101, doi: 10.1103/PhysRevLett.112. 241101. 7. R. Flauger, J. C. Hill and D. N. Spergel, Toward an understanding of foreground emission in the BICEP2 region, arXiv:1405.7351. 8. M. J. Mortonson and U. Seljak, A joint analysis of Planck and BICEP2 B modes including dust polarization uncertainty, arXiv:1405.5857. 9. G. F. Smoot, C. L. Bennett, A. Kogut, E. L. Wright, J. Aymon, N. W. Boggess, E. S. Cheng, G. de Amici, S. Gulkis, M. G. Hauser, G. Hinshaw, P. D. Jackson, M. Janssen, E. Kaita, T. Kelsall, P. Keegstra, C. Lineweaver, K. Loewenstein, P. Lubin, J. Mather, S. S. Meyer, S. H. Moseley, T. Murdock, L. Rokke, R. F. Silverberg, L. Tenorio, R. Weiss and D. T. Wilkinson, Astrophys. J. 396 (1992) L1, http://adsabs.harvard.edu/abs/1992ApJ...396L...1S. 10. Planck Collab. (P. A. R. Ade et al.), Planck 2013 results. XXII. Constraints on inﬂation, arXiv:1303.5082. 11. D. H. Lyth, Phys. Rev. Lett. 78 (1997) 1861, doi: 10.1103/PhysRevLett.78.1861. 12. G. Hinshaw, D. Larson, E. Komatsu, D. N. Spergel, C. L. Bennett, J. Dunkley, M. R. Nolta, M. Halpern, R. S. Hill, N. Odegard, L. Page, K. M. Smith, J. L. Weiland, B. Gold, N. Jarosik, A. Kogut, M. Limon, S. S. Meyer, G. S. Tucker, E. Wollack and E. L. Wright, Astrophys. J. Suppl. Series 208 (2013) 19, doi: 10.1088/00670049/208/2/19. 13. Planck Collab. (P. A. R. Ade et al.), Astron. Astrophys. 571 (2014) A16, doi: 10.1051/0004-6361/201321591. 14. K. M. Smith, C. Dvorkin, L. Boyle, N. Turok, M. Halpern, G. Hinshaw and B. Gold, Phys. Rev. Lett. 113 (2014) 031301, doi: 10.1103/PhysRevLett.113.031301. 15. I. Antoniadis and S. P. Patil, The eﬀective planck mass and the scale of inﬂation, arXiv:1410.8845. 16. S. B. Giddings, S. Kachru and J. Polchinski, Phys. Rev. D 66 (2002) 106006, doi: 10.1103/PhysRevD.66.106006. 17. S. Kachru, R. Kallosh, A. D. Linde and S. P. Trivedi, Phys. Rev. D 68 (2003) 046005, doi: 10.1103/PhysRevD.68.046005. 18. M. R. Douglas and S. Kachru, Rev. Mod. Phys. 79 (2007) 733, doi: 10.1103/RevModPhys.79.733. 19. R. Bousso and J. Polchinski, J. High Energy Phys. 0006 (2000) 006, doi: 10.1088/1126-6708/2000/06/006. 20. Y. Sumitomo and S.-H. H. Tye, J. Cosmol. Astropart. Phys. 2 (2013) 006, doi: 10.1088/1475-7516/2013/02/006. 21. K. Freese, J. A. Frieman and A. V. Olinto, Phys. Rev. Lett. 65 (1990) 3233, doi: 10.1103/PhysRevLett.65.3233. 22. F. C. Adams, J. R. Bond, K. Freese, J. A. Frieman and A. V. Olinto, Phys. Rev. D 47 (1993) 426, doi: 10.1103/PhysRevD.47.426. 23. G. Dvali and S. H. Tye, Phys. Lett. B 450 (1999) 72, http://adsabs.harvard. edu/abs/1999PhLB..450...72D. 24. C. P. Burgess, M. Majumdar, D. Nolte, F. Quevedo, G. Rajesh and R.-J. Zhang, J. High Energy Phys. 7 (2001) 047, doi: 10.1088/1126-6708/2001/07/047.

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-317

25. G. Dvali, Q. Shaﬁ and S. Solganik, D-brane inﬂation, arXiv:hep/th 0105203, http://adsabs.harvard.edu/abs/2001hep.th....5203D. 26. S. Kachru et al., J. Cosmol. Astropart. Phys. 0310 (2003) 013, doi: 10.1088/14757516/2003/10/013. 27. C. Burgess, R. Kallosh and F. Quevedo, J. High Energy Phys. 0310 (2003) 056. 28. M. Berg, M. Haack and B. Kors, On the moduli dependence of nonperturbative superpotentials in brane inﬂation, arXiv:hep-th/0409282. 29. D. Baumann et al., J. High Energy Phys. 0611 (2006) 031, doi: 10.1088/11266708/2006/11/031. 30. S. Buchan, B. Shlaer, H. Stoica and S. H. Tye, J. Cosmol. Astropart. Phys. 02 (2004) 013, http://adsabs.harvard.edu/abs/2004JCAP...02..013B. 31. H. Firouzjahi and S.-H. H. Tye, J. Cosmol. Astropart. Phys. 0503 (2005) 009, doi: 10.1088/1475-7516/2005/03/009. 32. R. Bean, X. Chen, G. Hailu, S.-H. H. Tye and J. Xu, J. Cosmol. Astropart. Phys. 0803 (2008) 026, doi: 10.1088/1475-7516/2008/03/026. 33. A. Krause and E. Pajer, J. Cosmol. Astropart. Phys. 7 (2008) 023, doi: 10.1088/14757516/2008/07/023. 34. E. Silverstein and D. Tong, Phys. Rev. D 70 (2004) 103505, doi: 10.1103/PhysRevD.70.103505. 35. M. Alishahiha, E. Silverstein and D. Tong, Phys. Rev. D 70 (2004) 123505, doi: 10.1103/PhysRevD.70.123505. 36. X. Chen, Phys. Rev. D 71 (2005) 063506, doi: 10.1103/PhysRevD.71.063506. 37. X. Chen, J. High Energy Phys. 8 (2005) 045, doi: 10.1088/1126-6708/2005/08/045. 38. K. Dasgupta, C. Herdeiro, S. Hirano and R. Kallosh, Phys. Rev. D 65 (2002) 126002, doi: 10.1103/PhysRevD.65.126002. 39. J. P. Conlon and F. Quevedo, J. High Energy Phys. 1 (2006) 146, doi: 10.1088/11266708/2006/01/146. 40. C. P. Burgess, M. Cicoli and F. Quevedo, J. Cosmol. Astropart. Phys. 11 (2013) 003, doi: 10.1088/1475-7516/2013/11/003. 41. J. E. Kim, H. P. Nilles and M. Peloso, J. Cosmol. Astropart. Phys. 0501 (2005) 005, doi: 10.1088/1475-7516/2005/01/005. 42. E. Silverstein and A. Westphal, Phys. Rev. D 78 (2008) 106003, doi: 10.1103/PhysRevD.78.106003. 43. L. McAllister, E. Silverstein and A. Westphal, Phys. Rev. D 82 (2010) 046003, doi: 10.1103/PhysRevD.82.046003. 44. A. D. Linde, Phys. Lett. B 129 (1983) 177, doi: 10.1016/0370-2693(83)90837-7. 45. K. Freese and W. H. Kinney, Natural Inﬂation: Consistency with Cosmic Microwave Background Observations of Planck and BICEP2, arXiv:1403.5277. 46. A. R. Liddle, A. Mazumdar and F. E. Schunck, Phys. Rev. D 58 (1998) 061301, doi: 10.1103/PhysRevD.58.061301. 47. A. Mazumdar, S. Panda and A. P´erez-Lorenzana, Nucl. Phys. B 614 (2001) 101, doi: 10.1016/S0550-3213(01)00410-2. 48. S. Dimopoulos, S. Kachru, J. McGreevy and J. G. Wacker, J. Cosmol. Astropart. Phys. 8 (2008) 003, doi: 10.1088/1475-7516/2008/08/003. 49. R. Easther and L. McAllister, J. Cosmol. Astropart. Phys. 5 (2006) 018, doi: 10.1088/1475-7516/2006/05/018. 50. K. Choi, H. Kim and S. Yun, Phys. Rev. 90 (2014) 023545, doi: 10.1103/PhysRevD.90.023545. 51. S.-H. H. Tye and S. S. C. Wong, Helical inﬂation and cosmic strings, arXiv:1404.6988.

May 4, 2017 9:54 b2167-ch19

II-318

D. F. Chernoﬀ and S.-H. H. Tye

52. R. Kappl, S. Krippendorf and H. P. Nilles, Aligned natural inﬂation: Monodromies of two axions, arXiv:1404.7127. 53. I. Ben-Dayan, F. Gil Pedro and A. Westphal, Hierarchical Axion Inﬂation, arXiv:1404.7773. 54. C. Long, L. McAllister and P. McGuirk, Phys. Rev. 90 (2014) 023501, doi: 10.1103/PhysRevD.90.023501. 55. K. Harigaya and M. Ibe, Inﬂaton potential on a Riemann surface, arXiv:1404.3511. 56. T. Higaki and F. Takahashi, Natural and multi-natural inﬂation in axion landscape, arXiv:1404.6923. 57. T. C. Bachlechner, M. Dias, J. Frazer and L. McAllister, A new angle on chaotic inﬂation, arXiv:1404.7496. 58. M. Berg, E. Pajer and S. Sj¨ ors, Phys. Rev. D 81 (2010) 103535, doi: 10.1103/PhysRevD.81.103535. 59. R. Kallosh, A. Linde and B. Vercnocke, Natural inﬂation in supergravity and beyond, arXiv:1404.6244. 60. F. Marchesano, G. Shiu and A. M. Uranga, F-term axion monodromy inﬂation, arXiv:1404.3040. 61. L. McAllister, E. Silverstein, A. Westphal and T. Wrase, The powers of monodromy, arXiv:1405.3652. 62. M. Arends, A. Hebecker, K. Heimpel, S. C. Kraus, D. Lust, C. Mayrhofer, C. Schick and T. Weigand, D7-Brane moduli space in axion monodromy and ﬂuxbrane inﬂation, arXiv:1405.0283. 63. M. R. Setare, A. Sepehri and V. Kamali, Phys. Lett. B 735 (2014) 84, doi: 10.1016/ j.physletb.2014.05.081. 64. H.-Y. Schive, T. Chiueh and T. Broadhurst, Nat. Phys. 10 (2014) 496, doi: 10.1038/ nphys2996. 65. E. Witten, Phys. Lett. B 153 (1985) 243, http://adsabs.harvard.edu/abs/ 1985PhLB..153..243W. 66. J. Polchinski, Phys. Rev. Lett. 75 (1995) 4724, doi: 10.1103/PhysRevLett.75.4724. 67. L. Randall and R. Sundrum, Phys. Rev. Lett. 83 (1999) 3370, doi: 10.1103/PhysRevLett.83.3370. 68. L. Leblond and S.-H. H. Tye, J. High Energy Phys. 3 (2004) 055, doi: 10.1088/11266708/2004/03/055. 69. I. R. Klebanov and M. J. Strassler, J. High Energy Phys. 8 (2000) 052, doi: 10.1088/1126-6708/2000/08/052. 70. S. S. Gubser, C. P. Herzog and I. R. Klebanov, J. High Energy Phys. 9 (2004) 036, doi: 10.1088/1126-6708/2004/09/036. 71. H. Firouzjahi, L. Leblond and S.-H. H. Tye, J. High Energy Phys. 5 (2006) 047, doi: 10.1088/1126-6708/2006/05/047. 72. E. J. Copeland, R. C. Myers and J. Polchinski, J. High Energy Phys. 06 (2004) 013, http://adsabs.harvard.edu/abs/2004JHEP...06..013C. 73. L. Leblond and M. Wyman, Phys. Rev. D 75 (2007) 123522, doi: 10.1103/PhysRevD.75.123522. 74. A. Avgoustidis, S. Chadburn and R. Gregory, Phys. Rev. D 86 (2012) 063516, doi: 10.1103/PhysRevD.86.063516. 75. C. L. Bennett, A. J. Banday, K. M. Gorski, G. Hinshaw, P. Jackson, P. Keegstra, A. Kogut, G. F. Smoot, D. T. Wilkinson and E. L. Wright, Astrophys. J. 464 (1996) L1, http://adsabs.harvard.edu/abs/1996ApJ...464L...1B. 76. D. N. Spergel, R. Bean, O. Dor´e, M. R. Nolta, C. L. Bennett, J. Dunkley, G. Hinshaw, N. Jarosik, E. Komatsu, L. Page, H. V. Peiris, L. Verde, M. Halpern, R. S.

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88.

89.

90. 91. 92. 93. 94. 95.

96. 97.

II-319

Hill, A. Kogut, M. Limon, S. S. Meyer, N. Odegard, G. S. Tucker, J. L. Weiland, E. Wollack and E. L. Wright, Astrophys. J. Suppl. Series 170 (2007) 377, http://adsabs.harvard.edu/abs/2007ApJS..170..377S. N. Jones, H. Stoica and S. H. Tye, J. High Energy Phys. 07 (2002) 051, http:// adsabs.harvard.edu/abs/2002JHEP...07..051J. S. Sarangi and S. H. Tye, Phys. Lett. B 536 (2002) 185, http://adsabs.harvard. edu/abs/2002PhLB..536..185S. N. T. Jones, H. Stoica and S. H. Tye, Phys. Lett. B 563 (2003) 6, http://adsabs. harvard.edu/abs/2003PhLB..563....6J. H. Firouzjahi and S. H. Tye, J. Cosmol. Astropart. Phys. 03 (2005) 009, http:// adsabs.harvard.edu/abs/2005JCAP...03..009F. S. E. Shandera and S. H. Tye, J. Cosmol. Astropart. Phys. 05 (2006) 007, http://adsabs.harvard.edu/abs/2006JCAP...05..007S. M. G. Jackson, N. T. Jones and J. Polchinski, J. High Energy Phys. 10 (2005) 013, http://adsabs.harvard.edu/abs/2005JHEP...10..013J. A. Vilenkin, Phys. Lett. B 133 (1983) 177, http://adsabs.harvard.edu/abs/ 1983PhLB..133..177V. C. Hogan and R. Narayan, Mon. Not. R. Astron. Soc. 211 (1984) 575, http:// adsabs.harvard.edu/abs/1984MNRAS.211..575H. A. Vilenkin, Astrophys. J. 282 (1984) L51, http://adsabs.harvard.edu/abs/ 1984ApJ...282L..51V. A. A. de Laix, Phys. Rev. D 56 (1997) 6193, http://adsabs.harvard.edu/abs/ 1997PhRvD..56.6193D. F. Bernardeau and J. Uzan, Phys. Rev. D 63 (2001) 23005, http://adsabs. harvard.edu/abs/2001PhRvD..63b3005B. M. Sazhin, G. Longo, M. Capaccioli, J. M. Alcal, R. Silvotti, G. Covone, O. Khovanskaya, M. Pavlov, M. Pannella, M. Radovich and V. Testa, Mon. Not. R. Astron. Soc. 343 (2003) 353, http://adsabs.harvard.edu/abs/2003MNRAS. 343..353S. M. V. Sazhin, M. Capaccioli, G. Longo, M. Paolillo, O. S. Khovanskaya, N. A. Grogin, E. J. Schreier and G. Covone, arXiv:astro-ph/0601494, http://adsabs.harvard.edu/abs/2006astro.ph..1494S. J. L. Christiansen, E. Albin, K. A. James, J. Goldman, D. Maruyama and G. F. Smoot, Phys. Rev. D 77 (2008) 123509, doi: 10.1103/PhysRevD.77.123509. J. L. Christiansen, E. Albin, T. Fletcher, J. Goldman, I. P. W. Teng, M. Foley and G. F. Smoot, Phys. Rev. D 83 (2011) 122004, doi: 10.1103/PhysRevD.83.122004. C. Hogan and R. Narayan, Mon. Not. R. Astron. Soc. 211 (1984) 575, http://adsabs.harvard.edu/abs/1984MNRAS.211..575H. T. Vachaspati and A. Vilenkin, Phys. Rev. D 31 (1985) 3052, http://adsabs. harvard.edu/abs/1985PhRvD..31.3052V. A. Economou, D. Harari and M. Sakellariadou, Phys. Rev. D 45 (1992) 433, http://adsabs.harvard.edu/abs/1992PhRvD..45..433E. R. A. Battye, R. R. Caldwell and E. P. S. Shellard, Gravitational waves from cosmic strings, in Topological Defects in Cosmology, Proc of a conference held in Rome, Italy, 10–12 October, 1996. World Scientiﬁc Publishers, eds. M. Signore and F. Melchiorri (1998), p. 11. T. Damour and A. Vilenkin, Phys. Rev. Lett. 85 (2000) 3761, http://adsabs. harvard.edu/abs/2000PhRvL..85.3761D. T. Damour and A. Vilenkin, Phys. Rev. D 64 (2001) 64008, http://adsabs. harvard.edu/abs/2001PhRvD..64f4008D.

May 4, 2017 9:54 b2167-ch19

II-320

D. F. Chernoﬀ and S.-H. H. Tye

98. T. Damour and A. Vilenkin, Phys. Rev. D 71 (2005) 63510, http://adsabs. harvard.edu/abs/2005PhRvD..71f3510D. 99. X. Siemens, J. Creighton, I. Maor, S. R. Majumder, K. Cannon and J. Read, Phys. Rev. D 73 (2006) 105001, http://adsabs.harvard.edu/abs/2006PhRvD.. 73j5001S. 100. C. J. Hogan, Phys. Rev. D 74 (2006) 43526, http://adsabs.harvard.edu/abs/ 2006PhRvD..74d3526H. 101. X. Siemens, V. Mandic and J. Creighton, Phys. Rev. Lett. 98 (2007) 111101, http://adsabs.harvard.edu/abs/2007PhRvL..98k1101S. 102. B. Abbott et al., Phys. Rev. D 76 (2007) 082001, http://adsabs.harvard.edu/ abs/2007PhRvD..76h2001A. 103. B. P. Abbott et al., Phys. Rev. D 80 (2009) 062002, doi: 10.1103/PhysRevD.80. 062002. 104. The LIGO and Virgo Coallab. (B. P. Abbott et al.), Nature 460 (2009) 990, http://adsabs.harvard.edu/abs/2009Natur.460..990A. 105. J. Aasi et al., Phys. Rev. Lett. 112 (2014) 131101, doi: 10.1103/PhysRevLett.112. 131101. 106. F. R. Bouchet and D. P. Bennett, Phys. Rev. D 41 (1990) 720, http://adsabs. harvard.edu/abs/1990PhRvD..41..720B. 107. R. R. Caldwell and B. Allen, Phys. Rev. D 45 (1992) 3447, http://adsabs. harvard.edu/abs/1992PhRvD..45.3447C. 108. V. M. Kaspi, J. H. Taylor and M. F. Ryba, Astrophys. J. 428 (1994) 713, http://adsabs.harvard.edu/abs/1994ApJ...428..713K. 109. F. A. Jenet, G. B. Hobbs, W. van Straten, R. N. Manchester, M. Bailes, J. P. W. Verbiest, R. T. Edwards, A. W. Hotan, J. M. Sarkissian and S. M. Ord, Astrophys. J. 653 (2006) 1571, http://adsabs.harvard.edu/abs/2006ApJ...653.1571J. 110. M. R. Depies and C. J. Hogan, Phys. Rev. D 75 (2007) 125006, http://adsabs.harvard.edu/abs/2007PhRvD..75l5006D. 111. P. B. Demorest, R. D. Ferdman, M. E. Gonzalez, D. Nice, S. Ransom, I. H. Stairs, Z. Arzoumanian, A. Brazier, S. Burke-Spolaor, S. J. Chamberlin, J. M. Cordes, J. Ellis, L. S. Finn, P. Freire, S. Giampanis, F. Jenet, V. M. Kaspi, J. Lazio, A. N. Lommen, M. McLaughlin, N. Palliyaguru, D. Perrodin, R. M. Shannon, X. Siemens, D. Stinebring, J. Swiggum and W. W. Zhu, Astrophys. J. 762 (2013) 94, doi: 10.1088/0004-637X/762/2/94. 112. L. Pogosian, S. H. Tye, I. Wasserman and M. Wyman, Phys. Rev. D 68 (2003) 23506, http://adsabs.harvard.edu/abs/2003PhRvD..68b3506P. 113. L. Pogosian, M. Wyman and I. Wasserman, J. Cosmol. Astropart. Phys. 09 (2004) 008, http://adsabs.harvard.edu/abs/2004JCAP...09..008P. 114. S. H. Tye, I. Wasserman and M. Wyman, Phys. Rev. D 71 (2005) 103508, http://adsabs.harvard.edu/abs/2005PhRvD..71j3508T. 115. M. Wyman, L. Pogosian and I. Wasserman, Phys. Rev. D 72 (2005) 23513, http://adsabs.harvard.edu/abs/2005PhRvD..72b3513W. 116. L. Pogosian, I. Wasserman and M. Wyman, On vector mode contribution to CMB temperature and polarization from local strings (2006), arXiv: astro-ph/0604141. 117. U. Seljak, A. Slosar and P. McDonald, J. Cosmol. Astropart. Phys. 10 (2006) 014, http://adsabs.harvard.edu/abs/2006JCAP...10..014S. 118. N. Bevis, M. Hindmarsh, M. Kunz and J. Urrestilla, Phys. Rev. D 75 (2007) 65015, http://adsabs.harvard.edu/abs/2007PhRvD..75f5015B. 119. A. A. Fraisse, J. Cosmol. Astropart. Phys. 03 (2007) 008, http://adsabs. harvard.edu/abs/2007JCAP...03..008F.

May 4, 2017 9:54 b2167-ch19

Inﬂation, string theory and cosmic strings

II-321

120. N. Bevis and P. M. Saﬃn, Phys. Rev. D 78 (2008) 23503, http://adsabs.harvard. edu/abs/2008PhRvD..78b3503B. 121. L. Pogosian, S. H. Tye, I. Wasserman and M. Wyman, J. Cosmol. Astropart. Phys. 02 (2009) 013, http://adsabs.harvard.edu/abs/2009JCAP...02..013P. 122. R. Battye and A. Moss, Phys. Rev. D 82 (2010) 023521, doi: 10.1103/PhysRevD.82.023521. 123. J. Dunkley et al., Astrophys. J. 739 (2011) 52, doi: 10.1088/0004-637X/739/1/52. 124. Planck Collab. (P. A. R. Ade et al.), Planck 2013 results. XXV. Searches for cosmic strings and other topological defects, arXiv:1303.5085. 125. E. J. Copeland, L. Pogosian and T. Vachaspati, Class. Quantum Grav. 28 (2011) 204009, doi: 10.1088/0264-9381/28/20/204009. 126. S. A. Sanidas, R. A. Battye and B. W. Stappers, Phys. Rev. D 85 (2012) 122003, doi: 10.1103/PhysRevD.85.122003. 127. A. Buonanno, TASI lectures on gravitational waves from the early universe, arXiv:grqc/0303085. ¨ 128. S. Olmez, V. Mandic and X. Siemens, Phys. Rev. D 81 (2010) 104028, http://adsabs.harvard.edu/abs/2010PhRvD..81j4028O. 129. E. Witten, Nucl. Phys. B 249 (1985) 557, doi: 10.1016/0550-3213(85)90022-7. 130. K. Miyamoto and K. Nakayama, J. Cosmol. Astropart. Phys. 7 (2013) 012, doi: 10.1088/1475-7516/2013/07/012. 131. Y.-W. Yu, K.-S. Cheng, G. Shiu and H. Tye, J. Cosmol. Astropart. Phys. 11 (2014) 040, doi: 10.1088/1475-7516/2014/11/040. 132. J. J. Blanco-Pillado, K. D. Olum and B. Shlaer, Phys. Rev. D 89 (2014) 023512, doi: 10.1103/PhysRevD.89.023512. 133. D. F. Chernoﬀ and S. H. Tye, in preparation. 134. V. Vanchurin, K. Olum and A. Vilenkin, Phys. Rev. D 72 (2005) 63514, http:// adsabs.harvard.edu/abs/2005PhRvD..72f3514V. 135. M. Sakellariadou, J. Cosmol. Astropart. Phys. 04 (2005) 003, http://adsabs. harvard.edu/abs/2005JCAP...04..003S. 136. C. J. Martins and E. P. Shellard, Phys. Rev. D 73 (2006) 43515, http://adsabs. harvard.edu/abs/2006PhRvD..73d3515M. 137. A. Avgoustidis and E. P. Shellard, Phys. Rev. D 73 (2006) 41301, http://adsabs. harvard.edu/abs/2006PhRvD..73d1301A. 138. C. Ringeval, M. Sakellariadou and F. R. Bouchet, J. Cosmol. Astropart. Phys. 02 (2007) 023, http://adsabs.harvard.edu/abs/2007JCAP...02..023R. 139. J. Polchinski and J. V. Rocha, Phys. Rev. D 74 (2006) 83504, http://adsabs. harvard.edu/abs/2006PhRvD..74h3504P. 140. J. Polchinski and J. V. Rocha, Phys. Rev. D 75 (2007) 123503, http://adsabs. harvard.edu/abs/2007PhRvD..75l3503P. 141. D. F. Chernoﬀ and S. H. H. Tye, Cosmic string detection via microlensing of stars, arXiv:0709.1139, http://arXiv.org/abs/0709.1139. 142. D. F. Chernoﬀ, Clustering of superstring loops, arXiv:0908.4077, http://arxiv. org/abs/0908.4077. 143. C. J. Hogan, Nature 326 (1987) 853, http://adsabs.harvard.edu/abs/ 1987Natur.326..853H. 144. C. J. Hogan and M. J. Rees, Nature 311 (1984) 109, http://adsabs.harvard.edu/ abs/1984Natur.311..109H. 145. T. W. B. Kibble, Nucl. Phys. B 252 (1985) 227, http://adsabs.harvard.edu/ abs/1985NuPhB.252..227K.

May 4, 2017 9:54 b2167-ch19

II-322

D. F. Chernoﬀ and S.-H. H. Tye

146. C. J. A. P. Martins and E. P. S. Shellard, Phys. Rev. D 53 (1996) 575, doi: 10.1103/PhysRevD.53.R575. 147. S. Kuroyanagi, K. Miyamoto, T. Sekiguchi, K. Takahashi and J. Silk, Phys. Rev. D 86 (2012) 023503, doi: 10.1103/PhysRevD.86.023503. 148. S. Kuroyanagi, K. Miyamoto, T. Sekiguchi, K. Takahashi and J. Silk, Phys. Rev. D 87 (2013) 023522, doi: 10.1103/PhysRevD.87.023522. 149. J. Binney and S. Tremaine, Galactic Dynamics, 2nd edn. (Princeton University Press, 2008). 150. S. Liebes, Phys. Rev. 133 (1964) 835, http://adsabs.harvard.edu/abs/ 1964PhRv..133..835L. 151. B. Paczynski, Astrophys. J. 304 (1986) 1, http://adsabs.harvard.edu/abs/ 1986ApJ...304....1P. 152. S. Mao, Res. Astron. Astrophys. 12 (2012) 947, doi: 10.1088/1674-4527/12/8/005. 153. J. K. Bloomﬁeld and D. F. Chernoﬀ, Phys. Rev. D 89 (2014) 124003, doi: 10.1103/ PhysRevD.89.124003. 154. M. S. Pshirkov and A. V. Tuntsov, Phys. Rev. D 81 (2010) 083519, doi: 10.1103/ PhysRevD.81.083519. 155. COROT: COnvection, ROtation & planetary Transits. http://smsc.cnes.fr/COROT (2014), http://smsc.cnes.fr/COROT. 156. Rossi X-Ray Timing Explorer (RXTE): December 1995 — January 2012. http:// heasarc.gsfc.nasa.gov/docs/xte/XTE.html (Sept. 2014). http://heasarc. gsfc.nasa.gov/docs/xte/XTE.html. 157. D. F. Chernoﬀ, White paper: Microlensing of superstring loops. unpublished document contributed to LSST (2009). 158. D. F. Chernoﬀ, R. Graziani, A. Kwok and T. Morris, to appear. 159. M. R. DePies and C. J. Hogan, Harmonic gravitational wave spectra of cosmic string loops in the galaxy, http://adsabs.harvard.edu/abs/2009arXiv0904.1052D. 160. K. Danzmann and T. A. Prince, Mission documents [LISA International Science Team]. http://list.caltech.edu/mission documents (2011). 161. A. J. Farmer and E. S. Phinney, Mon. Not. R. Astron. Soc. 346 (2003) 1197, http://adsabs.harvard.edu/abs/2003MNRAS.346.1197F. 162. E. O’Callaghan and R. Gregory, J. Cosmol. Astropart. Phys. 3 (2011) 004, doi: 10.1088/1475-7516/2011/03/004. 163. E. O’Callaghan, S. Chadburn, G. Geshnizjani, R. Gregory and I. Zavala, J. Cosmol. Astropart. Phys. 9 (2010) 013, doi: 10.1088/1475-7516/2010/09/013. 164. E. O’Callaghan, S. Chadburn, G. Geshnizjani, R. Gregory and I. Zavala, Phys. Rev. Lett. 105 (2010) 081602, doi: 10.1103/PhysRevLett.105.081602.

May 4, 2017 9:53 b2167-ch20

Chapter 20 Quantum gravity: A brief history of ideas and some prospects

Steven Carlip†,∗∗ , Dah-Wei Chiou‡,§,†† , Wei-Tou Ni¶,‡‡ and Richard Woodard,§§ †Department of Physics, University of California Davis, Davis, California 95616, USA ‡Department of Physics, National Taiwan Normal University, Taipei 11677, Taiwan §Department of Physics and Center for Condensed Matter Sciences, National Taiwan University, Taipei 10617, Taiwan ¶Center for Gravitation and Cosmology, Department of Physics, National Tsing Hua University, Hsinchu 30013, Taiwan Department

of Physics, University of Florida, Gainesville, Florida 32611, USA ∗∗[email protected] ††[email protected] ‡‡[email protected] §§ [email protected]

We present a bird’s-eye survey on the development of fundamental ideas of quantum gravity, placing emphasis on perturbative approaches, string theory, loop quantum gravity (LQG) and black hole thermodynamics. The early ideas at the dawn of quantum gravity as well as the possible observations of quantum gravitational eﬀects in the foreseeable future are also brieﬂy discussed. Keywords: Perturbative quantum gravity; string theory; loop quantum gravity; black hole thermodynamics; quantum gravity phenomenology. PACS Number: 04.60.−m

1. Prelude Quantum gravity is the research that seeks a consistent uniﬁcation of the two foundational pillars of modern physics — quantum theory and Einstein’s theory of general relativity. It is commonly considered as the paramount open problem of theoretical physics, and many fundamental issues — such as the microscopic structure of space and time, the origin of the universe, the resolution of spacetime singularities, etc. — relies on a better understanding of quantum gravity. II-325

May 4, 2017 9:53 b2167-ch20

II-326

S. Carlip et al.

The quest for a satisfactory quantum description of gravity began very early. Einstein after proposing general relativity thought that quantum eﬀects must modify general relativity in his ﬁrst paper on gravitational waves in 19161 (although he switched to a diﬀerent point of view working on the uniﬁcation of electromagnetism and gravitation in the 1930s). Klein argued that the quantum theory must ultimately modify the role of spatiotemporal concepts in fundamental physics2–4 and his ideas were developed by Deser.5 With the interests and developments of Rosenfeld, Pauli, Blokhintsev, Heisenberg, Gal’perin, Bronstein, Frenkel, van Dantzig, Solomon, Fierz and many other researchers, three approaches to quantum gravity after World War II had already been enunciated in 1930s pre-World War II as summarized by Stachel6 : (i) Quantum gravity should be formulated by analogy with quantum electrodynamics (Rosenfeld, Pauli and Fierz). (ii) The unique features of gravitation will require special treatment — the full theory with its nonlinear ﬁeld equations must be quantized and generalized as to be applicable in the absence of a background metric (Bronstein and Solomon). (iii) General relativity is essentially a macroscopic theory, e.g. a sort of thermodynamics limit of a deeper, underlying theory of interactions between particles (Frenkel and van Dantzig). Many ideas of the early time continue to provide valuable insight about the nature of quantum gravity even today. For example, the 1939 work on linearized general relativity as a spin-2 ﬁeld by Fierz and Pauli7 inspired a recent development on massive gravity, bimetric gravity, etc. Modern work on quantum gravity, however, did not really start-oﬀ until the development of a canonical formalism in 1959–1961 by Arnowitt, Deser and Misner (ADM) for the case of asymptotically ﬂat boundary conditions8–18 (for a review, see Ref. 19). This served as the basis for Schoen and Yau’s proof that the classical energy is bounded below20,21 and is the starting point for numerical simulations of classical general relativity. Of course the ADM formalism also provides the Hamiltonian and canonical variables whose quantization deﬁnes quantum general relativity. However, the sacriﬁce of manifest covariance made perturbative computations prohibitively diﬃcult, although many strategies to tackle this diﬃculty have been suggested, e.g. in loop quantum gravity (LQG). Ninety-nine years passed since the very ﬁrst conception of Einstein. Today, quantum gravity has grown into a vast area of research in many diﬀerent (both perturbative and nonperturbative) approaches. Many directions have led to signiﬁcant advances with various appealing and ingenious ideas; these include causal sets, dynamical triangulation, emergent gravity, H-space theory, LQG, noncommutative geometry, string theory, supergravity, thermogravity, twistor theory and much more. (For surveys and references on various approaches of quantum gravity,

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

II-327

see Refs. 22–25; for detailed accounts of the history and development, see Refs. 6 and 26; for a popular science account of quantum gravity, see the book Ref. 27.)

2. Perturbative Quantum Gravity The quantization of general relativity on the perturbative level (with or without matter) was the life work of DeWitt, for which he was awarded the 1987 Dirac Medal and the 2005 Einstein Prize.28–30 His program eventually succeeded — against the expectation of most expert opinion during the 1960s — in generalizing Feynman’s covariant quantization of quantum electrodynamics so that it could be applied as well to theories such as general relativity and Yang–Mills that are based on nonAbelian gauge symmetries. There were three key steps: (i) The introduction of the background field method for representing the eﬀective action as a gauge invariant functional of the ﬁelds.29,31 This made clear the connection between invariant counterterms and all possible ultraviolet divergences of scattering amplitudes at a ﬁxed order in the loop expansion. (ii) The realization that non-Abelian gauge symmetries require the inclusion of opposite-normed ghost fields to compensate the eﬀect of unphysical polarizations in loop corrections.29,32 (iii) The development of invariant regularization techniques, the ﬁrst of which was dimensional regularization.33,34 This permitted eﬃcient computations to be made without the need for noninvariant counterterms. The ﬁrst application of these techniques was made in 1974 by ’t Hooft and Veltman.35 They showed that general relativity without matter has a ﬁnite S-matrix at one-loop order, but that the addition of a scalar ﬁeld leads to the need for higher curvature counterterms that would destabilize the universe. Very shortly thereafter Deser and van Nieuwenhuizen showed that similar unacceptable counterterms are required at one-loop for general relativity plus Maxwell’s electrodynamics,36,37 for general relativity plus a Dirac fermion38 and (with Tsao) for general relativity plus Yang–Mills.39 The fate of pure general relativity was settled in 1985 when Goroﬀ and Sagnotti showed that higher curvature counterterms are required at two-loop order.40–42 These results show that quantum general relativity is not fundamentally consistent as a perturbative quantum ﬁeld theory. Opinion is divided as to whether this means that general relativity must be abandoned as the fundamental theory of gravity or whether quantum general relativity might make sense nonperturbatively. String theory discards general relativity, and does not even employ the metric as a fundamental dynamical variable. LQG and causal dynamical triangulation are approaches that attempt to make sense of quantum general relativity without employing perturbation theory. At this time it would be fair to say that there is not yet any fully successful quantum theory of gravity.

May 4, 2017 9:53 b2167-ch20

II-328

S. Carlip et al.

But being unable to do everything is not at all the same as lacking the ability to do anything. Perturbative quantum general relativity (with or without matter) can still be used in the standard sense of low-energy eﬀective ﬁeld theory,43,44 in the same way that Fermi theory was for years employed to understand the weak interactions, even including loop eﬀects.45,46 The most interesting eﬀects tend to occur in curved spacetime backgrounds, on which one can also consider quantizing only matter, without dynamical gravity.47 When working on curved backgrounds one often ﬁnds the formalism of ﬂat space scattering theory to be inappropriate because it is based on incorrect assumptions about free vacuum at very early times (which actually began with an initial singularity) and at very late times (which may be ﬁlled with ensembles of particles created by the curved geometry). Solving the eﬀective ﬁeld equations under these conditions is especially troublesome because the in–out eﬀective ﬁeld equations at µ a spacetime point xµ contain contributions from points x far in its future, and because even the in–out matrix elements of Hermitian operators such as the metric and the Maxwell ﬁeld strength tensor can develop imaginary parts. For these reasons it is often more appropriate to employ the Schwinger–Keldysh formalism. This is a diagrammatic technique that is almost as simple to use as the standard Feynman rules, which gives true expectation values instead of in–out matrix elements. It was devised for quantum mechanics in 1960 by Schwinger.48 Over the next few years it was generalized to quantum ﬁeld theory by Mahanthappa and Bakshi,49–51 and to statistical ﬁeld theory by Keldysh.52 Although Schwinger–Keldysh eﬀective µ ﬁeld equations are nonlocal, they are causal — in the sense that only points x in the past of xµ contribute — and the solutions for Hermitian ﬁelds are real.53,54 They are the natural way to study quantum eﬀects in cosmology55 and nonequilibrium quantum ﬁeld theory.56 Weinberg recently devised a variant of the formalism which is especially adapted to computing the correlation functions of primordial inﬂation.57,58 For a recent review of perturbative quantum gravity, see Ref. 59. 3. String Theory On the other hand, the search for a completely consistent theory of quantum gravity has never stopped. Although it does not embrace general relativity at the foundational level and it is disputable whether it can be genuinely formulated in a background-independent, nonperturbative fashion, string theory is arguably considered by many as the most promising candidate for a consistent theory of quantum gravity on the grounds that its low-energy limit surprisingly gives rise to (the supersymmetric extension of) the theory of (modiﬁed) gravity plus other force and matter ﬁelds. In the framework of string theory, the point-like particles of particle physics are replaced by (the diﬀerent quantum excitations of) minuscule one-dimensional objects called strings. In addition to being a quantum theory of gravity, string theory also aims to unify all fundamental forces and all forms of matter, striving

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

II-329

for the ultimate goal of being a “theory of everything”. (For textbooks on string theory, see Refs. 60–62; for a detailed account of its history and development, see Ref. 63.) String theory was born in the late 1960s as a never completely successful theory of the strong nuclear force and was later recognized as a suitable framework for a quantum theory of gravity (see Ref. 64 for its “prehistory”). The idea of identifying the string action as the area of the worldsheet of the string traveling in spacetime was introduced independently by Nambu,65 Goto66 and Hara67 in the early 1970s. The modern treatment of string theory based on the worldsheet path integral of the Polyakov action was initiated by Polyakov in 1981 (see Refs. 68 and 69) and has led to an intimate link with the conformal ﬁeld theory (CFT). The worldsheet CFT in the presence of spacetime background ﬁelds was studied by Callan et al.70 ; the conformal invariance demands beta functions of the worldsheet ﬁeld theory to vanish identically and consequently yields the equations of motion of the background ﬁelds that bear remarkable resemblance to the Einstein ﬁeld equation and (non-Abelian) Maxwell’s equations with higher-order corrections. Quantization of the bosonic string requires the number of spacetime dimensions to be 26, of which 22 extra spatial dimensions are thought to be compactiﬁed in the deep microscopic scale and thus undetectable at low energies. The bosonic string theory is unsatisfactory in two aspects: First, it entails existence of negative-normed tachyon ﬁelds and consequently is nonunitary and inconsistent; second, it does not contain fermions and thus cannot account for the quarks and leptons in the standard model. In the 1980s, bosonic string theory was extended into superstring theory, which incorporates both bosonic and fermionic degrees of freedom via supersymmetry and requires six extra spatial dimensions to be compactiﬁed on a Calabi–Yau manifold. Superstring theory not only gets rid of negative-normed ﬁelds but also includes fermions as desired. There are basically two (equivalent) approaches to embody supersymmetry in string theory: (i) The Ramond–Neveu–Schwarz (RNS) formalism with manifest supersymmetry on the string worldsheet71,72 ; (ii) the Green–Schwarz (GS) formalism with manifest supersymmetry on the background spacetime.73–75 The first superstring revolution began in 1984 with the discovery by Green and Schwarz76 that the cancellation of gauge and gravitational anomalies demands a very strong constraint on the gauge symmetry: The gauge group must be either SO(32) or E8 × E8 . Eventually, ﬁve consistent but distinct superstring theories were found: Type I, type IIA, type IIB, SO(32) heterotic and E8 × E8 heterotic. In the late 1980s, it was realized that the two type II theories and the two heterotic theories are related by T-duality, which, simply speaking, maps a string winding around a compactiﬁed dimension of radius R to that of radius 2s /R, where s is the string length scale. (For reviews on T-duality, see Refs. 77 and 78.) A few years later, S-duality was discovered as another kind of duality that maps the string coupling constant gs to 1/gs . The two basic examples are the duality that maps the type I theory to the SO(32) heterotic theory79 and the duality that maps the type

May 4, 2017 9:53 b2167-ch20

II-330

S. Carlip et al.

IIB theory to itself.80 As S-duality relates a strongly coupled theory to a weakly coupled theory, it provides a powerful tool to explore nonperturbative behaviors of a theory with gs > 1, given the knowledge of the dual theory perturbatively obtained with gs < 1. In the mid-1990s, understanding of the nonperturbative physics of superstring theory progressed signiﬁcantly, revealing that superstring theory contains various dynamical objects with p spatial dimensions called p-branes, in addition to the fundamental strings. Particularly, type I and II theories contain a class of p-branes, known as D-branes (or Dp-branes, more speciﬁcally), upon which open strings can end in Dirichlet boundary conditions. In 1995, Polchinski identiﬁed D-branes as solitonic solutions of supergravity that are understood as (generalized) black holes,81 a landmark discovery that heralded the prominence of D-brane dynamics. Furthermore, studies on the S-duality of type II82,83 and E8 × E8 heterotic84,85 theories sprang a big surprise: These two theories grow an eleventh dimension of size gs s at strong coupling. These discoveries together with the aforementioned relations between diﬀerent superstring theories via T- and S-dualities brought about the second superstring revolution that took place around 1994–1997, suggesting the existence of a more fundamental theory in 11 dimensions. The existence of such a theory was ﬁrst conjectured and named as M-theory by Witten at a string theory conference at the University of Southern California in 1995. M-theory is supposed to be the fundamental theory of everything, of which the ﬁve distinct superstring theories along with the eleven-dimensional theory of supergravity are regarded as diﬀerent limits. Although the complete formulation of M-theory remains elusive (and whether “M” should stand for “mother”, “magic”, “mystery”, “membrane”, “matrix”, etc. remains a matter of taste), a great deal about it can be learned by virtue of the fact that the theory should describe two- and ﬁve-dimensional objects known as M-branes (M2- and M5-branes, respectively) and reduce to elevendimensional supergravity theory at the low-energy limit. Investigations of M-theory have inspired a great number of important theoretical results in both physics and mathematics. A collection of coincident D-branes in string theory or M-branes in M-theory produces a warped spacetime with ﬂux of gauge ﬁelds akin to a charged black hole, as the branes are sources of gauge ﬂux and gravitational curvature. The low-energy limit of the gauge theory on the branes’ worldvolume (referred to as “boundary”) is found to describe the same physics of string theory or M-theory in the near-horizon geometry (referred to as “bulk”). In this way, one is led to conjecture a remarkable duality that relates conventional (nongravitational) quantum ﬁeld theory on the boundary to string theories or M-theory on the bulk. This gauge theory/string theory duality is often referred to as anti-de Sitter (Ads)/CFT correspondence, which was ﬁrst spelled out in a seminal paper by Maldacena in 1997.86 (A detailed review on AdS/CFT correspondence was given in Ref. 87.) By considering N coincident D3-branes in the type IIB theory, one obtains the celebrated example: The

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

II-331

type IIB theory in AdS5 × S 5 (the product space of ﬁve-dimensional anti-de Sitter space and ﬁve-dimensional sphere) is equivalent to the SU (N ) super Yang–Mills theory with N = 4 supersymmetry on the four-dimensional boundary (which is a superconformal ﬁeld theory). Many other examples have also been substantiated to various degrees of rigor. In the attempt to construct the complete formulation of M-theory, the AdS/CFT correspondence sets out a new strategy based on the holographic principle, which posits that the physics of a bulk region is completely encoded on its lower-dimensional boundary, as originally propounded by ’t Hooft in 199388 and elaborated by Susskind in 1994.89 In string theory, the shape and size of the compactiﬁed manifold and consequently the fundamental physical constants (lepton masses, coupling constants, cosmological constant, etc.) are dynamically determined by vacuum expectation values of scalar (moduli) ﬁelds. The crucial question arises known as the modulispace problem or moduli-stabilization problem: What mechanism stabilizes the compactiﬁed manifold and uniquely determines the physical constants? Answers to this problem have been proposed in the approach of flux compactifications as a generalization of conventional Calabi–Yau compactiﬁcations, which were ﬁrst introduced by Strominger in 198690 and by de Wit et al. in 198791 and now have become a rapidly developing area of research. The idea is to generate a potential that stabilizes the moduli ﬁelds by compactifying string theory or M-theory on a warped geometry. In a warped geometry, the ﬂuxes associated with certain tensor ﬁelds thread cycles of the compactiﬁed manifold; as the magnetic ﬂux of an n-form ﬁeld strength through an n-cycle depends only on the homology of the n-cycle, the ﬂux stabilizes due to the ﬂux quantization condition and gives rise to a nonvanishing potential. Flux compactiﬁcations were ﬁrst studied in the context of M-theory by Becker and Becker in 1996.92 (A comprehensive review on ﬂux compactiﬁcations was given in Ref. 93.) In 1999, Gukov et al. made it evident that ﬂux compactﬁcations generate nonvanishing potential for moduli ﬁelds, leading to a solution to the moduli-space problem.94 However, since the ﬂuxes can take many diﬀerent discrete values over diﬀerent homology cycles of diﬀerent (generalized) Calabi–Yau manifolds, ﬂux compactiﬁcations typically yield a vast multitude (commonly estimated to be of the order 10500 (see Ref. 95)) of possible vacuum expectation values, referred to as the string theory landscape, and therefore we have to abandon the long-cherished hope that the fundamental physical constants are supposed to be uniquely ﬁxed by string theory or M-theory. In response to this problem, Susskind in 200396 proposed the anthropic argument of the string theory landscape as a concrete implementation of the “anthropic principle”, which suggests that physical constants take their values not because fundamental laws of physics dictate so but rather because such values are at the loci of the landscape that are suitable to the existence of (intelligent) life (in order to measure these constants). The properties of the string theory landscape was shortly analyzed in a statistical approach by Douglas.95 The scientiﬁc relevance of the anthropic landscape has sparked ﬁerce debate and remained highly

May 4, 2017 9:53 b2167-ch20

II-332

S. Carlip et al.

controversial, yet it has gradually gained popularity (especially in the context of cosmology). The application of ﬂux compactiﬁcations as well as the string theory landscape to the study of cosmology has spawned a whole new growing research area known as string cosmology. A notable subﬁeld of it is brane cosmology, which posits that our four-dimensional universe is restricted to a brane inside a higher-dimensional space, as ﬁrst proposed by Randall and Sundrum in 1999.97 The primary goal of string cosmology is to place ﬁeld-theoretic models of cosmological inﬂation on a ﬁrmer logical ground from the perspective of string theory or, more boldly, to open entirely new realms of pre-big bang scenarios that cannot be described by any ﬁeldtheoretic models. In 2003, Kachru et al. outlined the construction of metastable de Sitter vacua98 and shortly investigated the application to inﬂation.99 String cosmology has made some concrete predictions that will soon be confronted with the near-future astrophysical observations. (For reviews on string cosmology, see Refs. 100 and 101 and the textbook Ref. 102.)

4. Loop Quantum Gravity The primary competitor of string theory for a complete quantum theory is LQG. In sharp contrast to the ambitious goal of string theory, LQG does not intend to pursue a theory of everything that uniﬁes all force and matter ﬁelds but deliberately adopts the “minimalist” approach in the sense that it focuses solely on the search for a consistent quantum theory of gravity without entailing any extraordinary ingredients such as extra dimensions, supersymmetry, etc. (although many of these can be incorporated compatibly). The beauty of LQG lies in its faithful attempt to establish a conceptual framework whereby the apparently conﬂicting tenets of quantum theory and Einstein’s theory of general relativity conjoin harmoniously, essentially in a nonperturbative, background-independent fashion. (For textbooks on LQG, see Refs. 103–105; for a detailed account of its history and development, see Ref. 106; for a recent review of LQG, see Ref. 107.) The research of LQG originated from the works by Ashtekar in 1986–1987,108,109 which reformulated Einstein’s general relativity into a new canonical formalism in terms of a self-dual spinorial connection and its conjugate momentum, now known as Ashtekar variables, casting general relativity in a language closer to that of the Yang–Mills gauge theory. The shift from metric to Ashtekar variables provided the possibility of employing nonperturbative techniques in gauge theories and opened a new avenue eventually leading to LQG. Shortly thereafter, Jacobson and Smolin discovered that Wilson loops of the Ashtekar connection are solutions of the Wheeler–DeWitt equation (the formal equation of quantum gravity).110 This remarkable discovery led to the “loop representation of quantum general relativity” introduced by Rovelli and Smolin in 1988–1990.111,112 The early developments of LQG were reviewed in Ref. 113.

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

II-333

In the 1990s, by the works of Br¨ ugmann et al.,114,115 it became clear that intersections of loops are essential for the consistency of the theory and quantum states of gravity should be formulated in terms of intersecting loops, i.e. graphs with links and nodes. Later on, inspired by Penrose’s speculation on the combinatorial structure of space,116 Rovelli and Smolin117 obtained an explicit basis of states of quantum geometry known as spin networks, which are oriented graphs labeled by numbers associated with spin representations of SU (2) on each link and node. The mathematical construction of spin networks was systematized by Baez.118,119 The idea that LQG could predict discrete quantum geometry was ﬁrst suggested in the study of “weave states”.120 In 1995, the area and volume operators were deﬁned upon spin networks and their eigenvalue spectra were found to be discrete,121 showing that quantum geometry is indeed quantized in the Planck scale as had long been speculated. Rigorous and systematic studies of geometry operators and their eigenvalues were further elaborated (e.g. see Refs. 122–124). The physical interpretation of spin networks (or more precisely, diﬀeomorphisminvariant knot classes of spin networks, known as s-knots) is extremely appealing: They represent diﬀerent quantized three-dimensional geometries, which are not quantum excitations in space but of space. This picture manifests the paradigm of background independence of general relativity, as any reference to the localization of spin networks is dismissed. Furthermore, inclusion of matter ﬁelds into LQG does not require a major revamp of the underlying framework. (For a systematic account of inclusion of matter, see Ref. 104.) The quantum states of space plus matter naturally extend the notion of spin networks with additional degrees of freedom. In the presence of matter ﬁelds, the background independence becomes even more prominent, as geometry and matter ﬁelds are genuinely on the equal footing and reside on top of one another via their contiguous relations in the spin network without any reference to a given background. While kinematics of LQG is well understood in terms of spin networks, its dynamics (i.e. evolution of spin networks) and low-energy (semiclassical) physics is much more diﬃcult and unclear. An anomaly-free formulation of the quantum dynamics was ﬁrst obtained by Thiemann in 1996,125 and shortly the formulation was fully developed in his remarkable “QSD” series of papers.126–128 Since then, a great number of variant approaches have been pursued, but the quantum dynamics remains very obscure, largely because implementation of the quantum Hamiltonian constraint is very intricate. The Master constraint program 129 initiated by Thiemannn proposed an elegant solution to the diﬃculties and has evolved into a fully combinatorial theory known as algebraic quantum gravity (AQG).130–133 Meanwhile, various techniques, notably by the idea of holomorphic coherent states,134–139 have been devised to investigate the low-energy physics. Up to now, LQG has achieved considerable progress in understanding the quantum dynamics and providing contact with the low-energy physics.

May 4, 2017 9:53 b2167-ch20

II-334

S. Carlip et al.

Applying principles of LQG to cosmological settings leads to the symmetryreduced theory known as loop quantum cosmology (LQC), which was originally proposed by Bojowald in 1999 and reached a rigorous formulation around 2006. Providing a “bottom–up” approach to the full theory of LQG, LQC has become the most well-developed subﬁeld of LQG and led to many signiﬁcant successes. Most notably, it suggests a new cosmological scenario where the big bang is replaced by the quantum bounce, which bridges the present expanding universe with a preexistent contracting counterpart. Absence of cosmological singularities has been shown to be robust for a great variety of LQC models, therefore aﬃrming the long-held conviction that singularities in classical general relativity should be resolved by the eﬀects of quantum gravity. (For textbooks and reviews on LQC, see Refs. 140–143.) The standard formalism of LQG adopts the canonical (Hamiltonian) approach but is closely related to the covariant (sun-over-histories) approach of quantum gravity known as the spin foam theory. The concept of a spin foam, which can be viewed as a (discrete) “worldsurface” swept out by a spin network traveling and transmuting in time, was introduced in 1993 as inspired by the work of Ponzano– Regge model144 and later developed into a systematic framework of quantum gravity by Perez et al. in the 2000s. A spin foam represents a quantized spacetime in the same sense that a spin network represents a quantized space; the transition amplitude from one spin network to another is given as the discrete sum (with appropriate weights) over all possible spin foams that connect the initial and ﬁnal spin networks. (For reviews on the spin foam theory, see Refs. 145–148.) LQG has grown into a very active research ﬁeld pursued in many directions, both in the traditional canonical approach and in the covariant (i.e. spin foam) approach. Recently, the precise connection between the canonical and covariant approaches has become one of the central topics of LQG. The merger of diﬀerent approaches has yielded profound insight and suggested a new theoretical framework referred to as covariant LQG, which provides new conceptual principles and could pave a royal road to a complete theory of quantum gravity as advocated by Rovelli149–151 (also see Ref. 105 for a comprehensive account). 5. Black Hole Thermodynamics Perhaps the most notable achievement in the study of quantum gravity so far was the discovery that black holes are not really black, but emit thermal Hawking radiation. The surprising fact that black holes behave as thermodynamic objects has radically aﬀected our understanding of general relativity and given valuable hints about the nature of quantum gravity. The study of black hole thermodynamics can be traced back to 1972 when Hawking proved that an area of an event horizon can never decrease152 — a property reminiscent of the second law of thermodynamics. The resemblance between black hole mechanics and thermodynamics was further enhanced by the discovery of analogs of other thermodynamical laws and formally elaborated in the paper by

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

II-335

Bardeen et al.153 in terms of the “four laws of black hole mechanics”, which are in clear parallel with the four usual laws of thermodynamics as the surface gravity (κ) plays the role of temperature and the area of horizon (Ahor ) of entropy. Around the same time, considering a series of thought experiments initiated by his advisor, Wheeler,154 Bekenstein argued that the black hole entropy has to be of the form SBH = ηkB Ahor /(G), where kB is Boltzmann’s constant and η is a proportional constant of order one, and correspondingly obtained the “generalized second law of thermodynamics”, which asserts that the total of ordinary entropy of matter plus the black hole entropy never decreases.155–157 In 1974, using the newly developed techniques of quantum ﬁeld theory in curved spacetime, Hawking158,159 showed that all black holes radiate as black bodies with the Hawking temperature TH = κ/(2π). The ﬁrst law of black hole mechanics then determines the entropy as SBH = 14 kB Ahor /(G), which conﬁrms Bekenstein’s expression with η = 1/4 and is often referred to as the Bekenstein–Hawking formula. As the equivalence principle implies that the gravitational ﬁeld near a black hole horizon is locally equivalent to uniform acceleration in a ﬂat spacetime, it is expected that an accelerated observer should perceive an eﬀect similar to the Hawking radiation. In 1976, Unruh160 demonstrated that an observer moving with a constant proper acceleration a in Minkowski spacetime indeed see a thermal ﬂux of particles with the Unruh temperature TU = a/(2π), which is in almost exact analogy with the Hawking temperature. At almost the same time, Bisognano and Wichmann gave an independent and mathematically rigorous proof of the Unruh eﬀect in quantum ﬁeld theory.161 To resolve the worrying issue whether the ﬂux of particles formally obtained via the standard quantum ﬁeld-theoretical deﬁnitions of vacuum states and particle numbers can be physically interpreted as “particles” as seen by detectors, Unruh160 and DeWitt162 devised simple models of particle detectors and showed that the answer is aﬃrmative for these detectors. The Hawking temperature and the Bekenstein–Hawking entropy are inherently quantum gravitational, in the sense that they depend explicitly on both Planck’s constant and Newton’s constant G. They are also surprisingly universal. The entropy, for example, depends only on the horizon area, and takes the same simple form regardless of the black hole’s charges, angular momentum, horizon topology and even the number of spacetime dimensions. The thermal properties of black holes have not been directly observed, but by now they have been derived in so many diﬀerent ways — from Hawking’s original calculation of quantum ﬁeld theory in curved spacetime and Unruh’s approach of an accelerated observer to Euclidean partition functions,161,163–165 canonical quantization in Hamiltonian formulation,166 quantum tunneling,167,168 anomaly techniques,169–174 the computation of pair production amplitudes175–178 and many others developed since early on to very recently — that their existence seems very nearly certain. The natural question is then whether black hole thermodynamics, like the ordinary thermodynamics of matter, has a microscopic “statistical mechanical” explanation. In 1996, using methods based on D-branes and string duality in string

May 4, 2017 9:53 b2167-ch20

II-336

S. Carlip et al.

theory, Strominger and Vafa suggested a strategy of “turning down” the strength of the gravitational interaction until a black hole becomes a weakly coupled system of strings and D-branes so that at weak coupling one can count the number of states; this strategy led to the same Bekenstein–Hawking entropy for extremal (i.e. maximally charged) supersymmetric (BPS) black holes in ﬁve dimensions.179 This remarkable result was quickly extended to a large number of extremal, nearextremal and some particular nonextremal black holes.180,181 However, it becomes harder and uncertain to obtain the right factor η = 1/4 for far-from-extremal black holes (the Schwarzschild black hole is a typical case). This issue is currently under intense investigation in various approaches but has yet to be fully elucidated. In 2005, Mathur proposed running the analysis backwards: Starting at weak coupling with a particular collection of strings and D-branes, one then turns the gravitational coupling up and see what geometry appears at strong coupling.182 The result is typically not a black hole but instead a “fuzzball” — a conﬁguration with no horizon and no singularity but with a geometry akin to that of a black hole outside a would-be horizon.183 In a few special cases, one can count the number of such classical fuzzball geometries and reproduce the Bekenstein–Hawking entropy. In general, however, many of the relevant states may not have classical geometric descriptions. In the context of string theory, the notable AdS/CFT correspondence also provides a feasible scheme to derive the black hole entropy. For asymptotically Ads black holes, one can in principle compute the entropy by counting states in the (nongravitational) dual CFT. The most straightforward application of this correspondence is for the (2+1)-dimensional BTZ black hole.184 In 1997–1998, Strominger185 and Birmingham et al.186 independently computed the BTZ black hole entropy, which precisely reproduces the Bekenstein–Hawking expression. Since many higherdimensional near-extremal black holes have a near-horizon geometry of the form BTZ × trivial, the BTZ results can be used to obtain the entropy of a large class of string theoretical black holes.187 The second major research program to derive the black hole entropy from a microscopic picture is LQG. The suggestion that counting the quanta of area in LQG could oﬀer the statistical description of black hole thermodynamics was ﬁrst proposed by Krasnov in 1997.188 The idea essentially is that the statistical ensemble is composed of the microstates of the horizon geometry that give rise to a speciﬁed total area. By introducing the concept of “isolated horizons”, Ashtekar et al. rigorously developed the framework of the idea and derived the right Bekenstein– Hawking formula for the Schwarzschild black hole.189–191 These results were soon extended to other black holes with rotation, distortion, etc.192–197 In 2004, the combinatorial problem of counting the microscopic states in LQG was rephrased in a more manageable way by Domagala and Lewandowski198 and by Meissner.199 These works corrected a ﬂawed assumption considered true for several years and enabled one to compute the formula for Ahor G to the subleading

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

II-337

−1 1 A hor order as kB S = γγM A 4G − 2 ln G +· · ·, by which the Barbero–Immirzi parameter γ is ﬁxed as γ = γM ≈ 0.23653. For small areas, the precise number counting was ﬁrst suggested in 2006200,201 and later thoroughly investigated by employing combinatorial methods (see Ref. 202 for a detailed account and Ref. 203 for a review). The key discovery is that, for microscopic black holes, the so-called black hole degeneracy spectrum when plotted as a function of the area exhibits a striking “staircase” structure, which makes contact in a nontrivial way with the evenly-spaced black hole horizon area spectrum predicted by Bekenstein and Mukhanov.204 Recently, an explicit SU (2) formulation for the black hole entropy has been developed by Engle et al. based on covariant Hamiltonian methods,205–207 giving rigorous support for the earlier proposal that the quantum black hole degrees of freedom could be described by an SU (2) Chern–Simons theory. (Also see the review Ref. 203 and references therein for other recent advances.) In addition to string theory and LQG, we now have a number of statistical mechanical approaches to black hole thermodynamics. These include entanglement entropy,208–212 induced gravity213–215 and many others. While they may diﬀer on the subleading corrections,216 they all seem to give the correct Bekenstein–Hawking −1 A hor S=A entropy. (In general, one expects kB 4G − α ln G + · · ·, where the coeﬃcient α depends on the quantum theory.) A new puzzle is why such a diverse set of quantum gravitational approaches should all agree. (There are some indications that even the subleading corrections might be universal, up to diﬀerences on the treatment of angular momentum and conserved charges.216) Even more puzzling is the “information loss problem”: If a black hole is formed by collapse of matter in a pure quantum state, how can the seemingly thermal ﬁnal state be compatible with unitary evolution? This question is currently the subject of intense research, and may yield surprising new insights into the nature of space and time. (For a detailed account of the information paradox, see Ref. 217.) For a recent review of black hole thermodynamics, see Ref. 218.

6. Quantum Gravity Phenomenology Quantum gravity is often denounced as merely a theoretical enterprise that has no direct contact with experimental or observational realms, as quantum gravitational eﬀects are appreciable only at the Planck scale, which is thought to be completely out of current reach. Today, as idea ﬂourishes and technology advances drastically, the situation could be changed soon and quantum gravity might become reachable in the foreseeable future. Many quantum gravity models suggest departures from the equivalence principle, CPT symmetry and/or local Lorentz invariance in the deep Planck regime; among them the speciﬁc examples are string theory,219 braneworld scenarios,220 LQG and spin foam theory,221,222 noncommutative geometry,223–225 emergent gravity,226 etc. Deviations from the symmetries at the Planck scale in turn modify the

May 4, 2017 9:53 b2167-ch20

II-338

S. Carlip et al.

pure metric spacetime structure and the local Lorentzian energy–momentum dispersion relation at lower-energy scale. Modiﬁcations of the pure metric spacetime structure implies deviations from the Einstein Equivalence Principle (EEP). The most stringent test of EEP comes from cosmic observations. The null result of birefringence in cosmic propagation of polarized photons and polarized gamma rays so far has ascertained the light-cone structure and the core metric to ultrahigh precision of 10−38 ; this high precision already probes the second-order in the ratio of W -boson mass or proton mass to the Planck mass.227 To this precision, the only freedoms left over regarding the metric structure are a scalar degree of freedom (dilaton) and a pseudo-scalar degree of freedom (axion). The dilaton alters the amplitude while the axion rotates the linear polarization of the cosmic propagation. Based on the cosmic propagation since the last scattering surface of the cosmic microwave background (CMB), the fractional variation of the dilaton degree of freedom is constrained to 8×10−4 by the precision agreement of the CMB spectrum to the black-body spectrum.228 The cosmic axion degree of freedom is constrained by comic polarization rotation (CPR): Uniform CPR to less than 0.02 (rad);229 CPR ﬂuctuations to 0.02 (rad).229,230 Observations on ultrahigh energy cosmic Rays (UHECR)231 — cosmic rays with kinetic energy greater than 1018 eV, the most energetic particles ever observed — provide a chance to see possible modiﬁcations to the local Lorentzian energy– momentum dispersion relation of ultrarelativistic (v ∼ c) particles. The UHECR spectrum measured at the Pierre Auger Observatory thus far has put very strong constraints on deviations from the Lorentzian dispersion relation.232 Future investigation with improved precision will either impose even stronger constraints or pick up signals of new physics including quantum gravitational eﬀects (see Refs. 233 and 234 for detailed surveys). Similarly, observations on propagation of ultrahigh energy cosmic electromagnetic waves can be used to test any momentum dependence of the speed of photons, as an indication of a breakdown of the Lorentz symmetry. Particularly, the detailed analysis of the relationship between the arrival time, photon momentum and redshift of gamma-ray bursts is approaching very closely to the desired sensitivity to the Planck-scale physics.235 On the other hand, the ultraprecise measurements of “atom-recoil frequency” in the cold-atom experiments allow us to probe the energy–momentum dispersion relation of nonrelativistic (v c) particles with Planck-scale sensitivity, thus providing revealing insight into various types of modiﬁcations to the dispersion relation considered in the quantum gravity literature.236,237 Combining studies of the two complementary regimes of the nonrelativistic and ultrarelativistic dispersion relations would tell us a great deal about the nature of quantum spacetime.237 In the context of astrophysics, many theories — such as attractor theory,238 the DGP model,239 massive gravity,240 etc. — predict the eﬀects of quantum gravity in the solar system dynamics. None of these eﬀects have been observed yet, but

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

II-339

they might be testable in the proposed space missions designed to test relativistic gravity241,242 and/or to detect and measure gravitational waves by using laser interferometry (see Ref. 243 for a detailed survey on space-based gravitationalwave detectors). Sensitive tests of the equivalence principle, local Lorentz invariance, modiﬁcations to Einstein’s ﬁeld equations, etc. might also be possible in the coming space missions.244 In the context of cosmology, quantum eﬀects in the pre-inﬂationary era may give rise to suﬃcient deviations from the standard inﬂationary scenario and leave footprints on the CMB. In string theory, various observable imprints on anisotropies and polarization of the CMB have been suggested by diﬀerent models of string cosmology and by the existence of cosmic strings (see Ref. 102 for a review). In LQG, the standard inﬂationary scenario is extended to the epoch from the big bounce in the Planck era to the onset of slow-roll inﬂation.245 Some research works have attempted to reveal possible observable imprints of loop quantum eﬀects on the CMB.246–251 There are a few experiments currently in operation or in ongoing development, aiming to make accurate measurements of anisotropies and polarization of the CMB, and their results will soon lead to great excitement. Very recently, the joint analysis of BICEP2/Keck Array and Planck data found no statistically signiﬁcant evidence for tensor modes in the B-mode polarization, yielding an upper limit r < 0.12 at 95% conﬁdence on the tensor-to-scalar ratio r.252 This upper limit already ruled out many “large-r scenarios” of pre-inﬂationary models, including those in string cosmology. The conclusive result of B-mode polarization expected to arrive in the near future will further discriminate between diﬀerent quantum gravity theories. There are many other cosmological and astrophysical observations as well as laboratory experiments that could reveal possible quantum gravitational eﬀects. At the turn of the centennial of the birth of Einstein’s general relativity, quantum gravity phenomenology will in time become an important area of relevant research. (For a comprehensive survey on quantum gravity phenomenology, see Ref. 253; also see Ref. 254 for a survey from the perspective of LQG.) Acknowledgements S.C. was partially supported by US Department of Energy grant DE-FG0291ER40674; D.W.C. by the Ministry of Science and Technology of Taiwan under the Grant No. 101-2112-M-003-002-MY3; and R.W. by NSF Grant PHY-1506513 and by the Institute for Fundamental Theory at the University of Florida. References 1. 2. 3. 4.

A. Einstein, Sitz.ber. Preuss. Akad. Wiss. Berl. Math. Phys. 1916 (1916) 688. O. Klein, Z. Phys. 46 (1927) 688. O. Klein, Kosmos (Sweden) 32 (1954) 33. O. Klein, Generalization of Einsteins theory of gravitation considered from the point of view of quantum ﬁeld theory, in F¨ uenfzig Jahre Relativit¨ atstheorie: Proc. of the

May 4, 2017 9:53 b2167-ch20

II-340

5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.

20. 21. 22.

23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46.

S. Carlip et al.

Jubilee of Relativity Theory, Bern, 11–16 July, 1955, eds. A. Mercier and M. Kervaire, Helvetica Physica Acta Supplementum, Vol. 4 (Birkh¨ auser, Basel, 1956). S. Deser, Rev. Mod. Phys. 29 (1957) 417. J. Stachel, Early history of quantum gravity (1916–1940), in Black Holes, Gravitational Radiation and the Universe, eds. B. R. Iyer and B. Bhawal (Kluwer, 1999). M. Fierz and W. Pauli, Proc. Roy. Soc. Lond. A 173 211 (1939). R. Arnowitt and S. Deser, Phys. Rev. 113 745 (1959). R. L. Arnowitt, S. Deser and C. W. Misner, Phys. Rev. 116 (1959) 1322. R. L. Arnowitt, S. Deser and C. W. Misner, Phys. Rev. 117 (1960) 1595. S. Deser, R. Arnowitt and C. W. Misner, Il Nuovo Cimento 15 (1960) 487. R. Arnowitt, S. Deser and C. W. Misner, Phys. Rev. 118 (1960) 1100. S. Deser, R. Arnowitt and C. W. Misner, J. Math. Phys. 1 (1960) 434. R. L. Arnowitt, S. Deser and C. W. Misner, Phys. Rev. 120 (1960) 313. S. Deser, R. Arnowitt and C. W. Misner, Ann. Phys. 11 (1960) 116. S. Deser, R. Arnowitt and C. W. Misner, Il Nuovo Cimento 19 (1961) 668. R. L. Arnowitt, S. Deser and C. W. Misner, Phys. Rev. 121 (1961) 1556. R. L. Arnowitt, S. Deser and C. W. Misner, Phys. Rev. 122 (1961) 997. R. L. Arnowitt, S. Deser and C. W. Misner, Gen. Rel. Gravit. 40 (2008) 1997, arXiv:gr-qc/0405109: Originally appeared as Chap. 7, pp. 227–264, in Gravitation: An Introduction to Current Research, ed. L. Witten (Wiley, New York, 1962). R. Schoen and S. T. Yau, Commun. Math. Phys. 65 (1979) 45. R. Schoen and S. T. Yau, Commun. Math. Phys. 79 (1981) 231. C. Rovelli, Strings, loops and others: A Critical survey of the present approaches to quantum gravity, in Gravitation and Relativity: At the Turn of the Millennium (Inter-University Centre for Astronomy and Astrophysics, Pune, 1998), pp. 281–331. S. Carlip, Rep. Prog. Phys. 64 (2001) 885, arXiv:gr-qc/0108040. R. P. Woodard, Rep. Prog. Phys. 72 (2009) 126002, arXiv:0907.4238 [gr-qc]. C. Kiefer, Quantum Gravity (Oxford University Press, Oxford, 2012). C. Rovelli, arXiv:gr-qc/0006061. L. Smolin, Three Roads to Quantum Gravity (Weidenfeld and Nicolson, London, 2000). B. S. DeWitt, Phys. Rev. 160 (1967) 1113. B. S. DeWitt, Phys. Rev. 162 (1967) 1195. B. S. DeWitt, Phys. Rev. 162 (1967) 1239. L. F. Abbott, Acta Phys. Pol. B 13 (1982) 33. L. D. Faddeev and V. N. Popov, Phys. Lett. B 25 (1967) 29. G. ’t Hooft and M. J. G. Veltman, Nucl. Phys. B 44 (1972) 189. C. G. Bollini and J. J. Giambiagi, Nuovo Cimento B 12 (1972) 20. G. ’t Hooft and M. J. G. Veltman, Ann. Poincar´e Phys. Theor. A 20 (1974) 69. S. Deser and P. van Nieuwenhuizen, Phys. Rev. Lett. 32 (1974) 245. S. Deser and P. van Nieuwenhuizen, Phys. Rev. D 10 (1974) 401. S. Deser and P. van Nieuwenhuizen, Phys. Rev. D 10 (1974) 411. S. Deser, H. S. Tsao and P. van Nieuwenhuizen, Phys. Rev. D 10 (1974) 3337. M. H. Goroﬀ and A. Sagnotti, Phys. Lett. B 160 (1985) 81. M. H. Goroﬀ and A. Sagnotti, Nucl. Phys. B 266 (1986) 709. A. E. M. van de Ven, Nucl. Phys. B 378 (1992) 309. J. F. Donoghue, Phys. Rev. Lett. 72 (1994) 2996, arXiv:gr-qc/9310024. J. F. Donoghue, Phys. Rev. D 50 (1994) 3874, arXiv:gr-qc/9405057. G. Feinberg and J. Sucher, Phys. Rev. 166 (1968) 1638. S. D. H. Hsu and P. Sikivie, Phys. Rev. D 49 (1994) 4951, arXiv:hep-ph/9211301.

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60.

61.

62. 63. 64. 65.

66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81.

II-341

B. S. DeWitt, Phys. Rep. 19 (1975) 295. J. S. Schwinger, J. Math. Phys. 2 (1961) 407. K. T. Mahanthappa, Phys. Rev. 126 (1962) 329. P. M. Bakshi and K. T. Mahanthappa, J. Math. Phys. 4 (1963) 1. P. M. Bakshi and K. T. Mahanthappa, J. Math. Phys. 4 (1963) 12. L. V. Keldysh, Zh. Eksp. Teor. Fiz. 47 (1964) 1515 [Sov. Phys.-JETP 20 (1965) 1018]. K. C. Chou, Z. B. Su, B. L. Hao and L. Yu, Phys. Rep. 118 (1985) 1. R. D. Jordan, Phys. Rev. D 33 (1986) 444. E. Calzetta and B. L. Hu, Phys. Rev. D 35 (1987) 495. E. Calzetta and B. L. Hu, Phys. Rev. D 37 (1988) 2878. S. Weinberg, Phys. Rev. D 72 (2005) 043514, arXiv:hep-th/0506236. S. Weinberg, Phys. Rev. D 74 (2006) 023508, arXiv:hep-th/0605244. R. P. Woodard, Chapter 21 in this book; Int. J. Mod. Phys. D 23 (2014) 1430020, arXiv:1407.4748 [gr-qc]. M. B. Green, J. H. Schwarz and E. Witten, Superstring Theory: Introduction, Vol. 1 (Cambridge University Press, Cambridge, 1987); Superstring Theory: Loop Amplitudes, Anomalies and Phenomenology, Vol. 2 (Cambridge University Press, Cambridge, 1987). J. Polchinski, String Theory: An Introduction to the Bosonic String, Vol. 1 (Cambridge University Press, Cambridge, 1998); String Theory: Superstring Theory and Beyond, Vol. 2 (Cambridge University Press, Cambridge, 1998). K. Becker, M. Becker and J. H. Schwarz, String Theory and M-Theory: A Modern Introduction (Cambridge University Press, Cambridge, 2007). D. Rickles, A Brief History of String Theory: From Dual Models to M-Theory (Springer-Verlag, Berlin, 2014). J. H. Schwarz, arXiv:0708.1917 [hep-th]. Y. Nambu, Duality and Hadrodynamics, Notes prepared for the Copenhagen High Energy Symposium (1970); Reprinted in Broken Symmetry: Selected Papers of Y. Nambu, eds. T. Eguchi and K. Nishijima (World Scientiﬁc, Singapore, 1995). T. Goto, Prog. Theor. Phys. 46 (1971) 1560. O. Hara, Prog. Theor. Phys. 46 (1971) 1549. A. M. Polyakov, Phys. Lett. B 103 (1981) 207. A. M. Polyakov, Phys. Lett. B 103 (1981) 211. C. G. Callan, Jr., E. J. Martinec, M. J. Perry and D. Friedan, Nucl. Phys. B 262 (1985) 593. P. Ramond, Phys. Rev. D 3 (1971) 2415. A. Neveu and J. H. Schwarz, Nucl. Phys. B 31 86 (1971). M. B. Green and J. H. Schwarz, Nucl. Phys. B 181 (1981) 502. M. B. Green and J. H. Schwarz, Nucl. Phys. B 198 (1982) 252. M. B. Green and J. H. Schwarz, Phys. Lett. B 109 (1982) 444. M. B. Green and J. H. Schwarz, Phys. Lett. B 149 (1984) 117. A. Giveon, M. Porrati and E. Rabinovici, Phys. Rep. 244 (1994) 77, arXiv:hep-th/ 9401139. E. Alvarez, L. Alvarez-Gaume and Y. Lozano, Nucl. Phys. Proc. Suppl. 41 (1995) 1, arXiv:hep-th/9410237. J. Polchinski and E. Witten, Nucl. Phys. B 460 (1996) 525, arXiv:hep-th/9510169. C. M. Hull and P. K. Townsend, Nucl. Phys. B 438 (1995) 109, arXiv:hep-th/ 9410167. J. Polchinski, Phys. Rev. Lett. 75 (1995) 4724, arXiv:hep-th/9510017.

May 4, 2017 9:53 b2167-ch20

II-342

S. Carlip et al.

82. 83. 84. 85. 86.

P. K. Townsend, Phys. Lett. B 350 (1995) 184, arXiv:hep-th/9501068. E. Witten, Nucl. Phys. B 443 (1995) 85, arXiv:hep-th/9503124. P. Hoˇrava and E. Witten, Nucl. Phys. B 460 (1996) 506, arXiv:hep-th/9510209. P. Hoˇrava and E. Witten, Nucl. Phys. B 475 (1996) 94, arXiv:hep-th/9603142. J. M. Maldacena, Int. J. Theor. Phys. 38 (1999) 1113; Adv. Theor. Math. Phys. 2 (1998) 231, arXiv:hep-th/9711200. O. Aharony, S. S. Gubser, J. M. Maldacena, H. Ooguri and Y. Oz, Phys. Rep. 323 (2000) 183, arXiv:hep-th/9905111. G. ’t Hooft, Dimensional reduction in quantum gravity, in Salamfestschrift: A Collection of Talks from the Conf. on Highlights of Particle and Condensed Matter Physics, Trieste, Italy, 8–12 March, 1993 (World Scientiﬁc, 1993), arXiv:gr-qc/9310026. L. Susskind, J. Math. Phys. 36 (1995) 6377, arXiv: hep-th/9409089. A. Strominger, Nucl. Phys. B 274 (1986) 253. B. de Wit, D. J. Smit and N. D. H. Dass, Nucl. Phys. B 283 (1987) 165. K. Becker and M. Becker, Nucl. Phys. B 477 (1996) 155, arXiv:hep-th/9605053. M. Gra˜ na, Phys. Rep. 423 (2006) 91, arXiv:hep-th/0509003. S. Gukov, C. Vafa and E. Witten, Nucl. Phys. B 584 (2000) 69, Erratum 608 (2001) 477, arXiv:hep-th/9906070. M. R. Douglas, J. High Energy Phys. 0305 (2003) 046, arXiv:hep-th/0303194. L. Susskind, The Anthropic landscape of string theory, in Universe or Multiverse? ed. C. Bernard (Cambridge University Press, 2009), pp. 247–266. L. Randall and R. Sundrum, Phys. Rev. Lett. 83 (1999) 4690, arXiv:hep-th/9906064. S. Kachru, R. Kallosh, A. D. Linde and S. P. Trivedi, Phys. Rev. D 68 (2003) 046005, arXiv:hep-th/0301240. S. Kachru, R. Kallosh, A. D. Linde, J. M. Maldacena, L. P. McAllister and S. P. Trivedi, J. Cosmol. Astropart. Phys. 0310 (2003) 013, arXiv:hep-th/0308055. F. Quevedo, Class. Quantum Grav. 19 (2002) 5721, arXiv:hep-th/0210292. U. H. Danielsson, Class. Quantum Grav. 22 (2005) S1, arXiv:hep-th/0409274. D. Baumann and L. McAllister, Inflation and String Theory (Cambridge University Press, Cambridge, 2015). C. Rovelli, Quantum Gravity (Cambridge University Press, Cambridge, 2004). T. Thiemann, Modern Canonical Quantum General Relativity (Cambridge University Press, Cambridge, 2007). C. Rovelli and F. Vidotto, Covariant Loop Quantum Gravity: An Elementary Introduction to Quantum Gravity and Spinfoam Theory (Cambridge University Press, Cambridge, 2014). C. Rovelli, Class. Quantum Grav. 28 (2011) 153002, arXiv:1012.4707 [gr-qc]. D. W. Chiou, Chapter 23 in this book; Int. J. Mod. Phys. D 24 (2014) 1530005, arXiv:1412.4362 [gr-qc]. A. Ashtekar, Phys. Rev. Lett. 57 (1986) 2244. A. Ashtekar, Phys. Rev. D 36 (1987) 1587. T. Jacobson and L. Smolin, Nucl. Phys. B 299 (1988) 295. C. Rovelli and L. Smolin, Phys. Rev. Lett. 61 (1988) 1155. C. Rovelli and L. Smolin, Nucl. Phys. B 331 (1990) 80. C. Rovelli, Class. Quantum Grav. 8 (1991) 1613. B. Br¨ ugmann, R. Gambini and J. Pullin, Phys. Rev. Lett. 68 (1992) 431. B. Br¨ ugmann, R. Gambini and J. Pullin, Nucl. Phys. B 385 (1992) 587, arXiv:hepth/9202018. R. Penrose, Angular momentum: An approach to combinatorial spacetime, in Quantum Theory and Beyond, ed. T. Bastin (Cambridge University Press, Cambridge 1971).

87. 88.

89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105.

106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116.

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

II-343

117. C. Rovelli and L. Smolin, Phys. Rev. D 52 (1995) 5743, arXiv:gr-qc/9505006. 118. J. C. Baez, Adv. Math. 117 (1996) 253, arXiv:gr-qc/9411007. 119. J. C. Baez, Spin networks in nonperturbative quantum gravity, in The Interface of Knots and Physics, ed. L. H. Kauﬀman (American Mathematical Soc. San Francisco, 1995), pp. 167–203. 120. A. Ashtekar, C. Rovelli and L. Smolin, Phys. Rev. Lett. 69 (1992) 237, arXiv:hepth/9203079. 121. C. Rovelli and L. Smolin, Nucl. Phys. B 442 (1995) 593, Erratum Nucl. Phys. B 456 (1995) 753, arXiv:gr-qc/9411005. 122. R. De Pietri and C. Rovelli, Phys. Rev. D 54 (1996) 2664, arXiv:gr-qc/9602023. 123. T. Thiemann, J. Math. Phys. 39 (1998) 3347, arXiv:gr-qc/9606091. 124. J. Lewandowski, Class. Quantum Grav. 14 (1997) 71, arXiv:gr-qc/9602035. 125. T. Thiemann, Phys. Lett. B 380 (1996) 257, arXiv:gr-qc/9606088. 126. T. Thiemann, Class. Quantum Grav. 15 (1998) 839, arXiv:gr-qc/9606089. 127. T. Thiemann, Class. Quantum Grav. 15 (1998) 875, arXiv:gr-qc/9606090. 128. T. Thiemann, Class. Quantum Grav. 15 (1998) 1207, arXiv:gr-qc/9705017. 129. T. Thiemann, Class. Quantum Grav. 23 (2006) 2211, arXiv:gr-qc/0305080. 130. K. Giesel and T. Thiemann, Class. Quantum Grav. 24 (2007) 2465, arXiv:gr-qc/ 0607099. 131. K. Giesel and T. Thiemann, Class. Quantum Grav. 24 (2007) 2499, arXiv:gr-qc/ 0607100. 132. K. Giesel and T. Thiemann, Class. Quantum Grav. 24 (2007) 2565, arXiv:gr-qc/ 0607101. 133. K. Giesel and T. Thiemann, Class. Quantum Grav. 27 (2010) 175009, arXiv: 0711.0119 [gr-qc]. 134. A. Ashtekar, J. Lewandowski, D. Marolf, J. Mourao and T. Thiemann, J. Funct. Anal. 135 (1996) 519, arXiv:gr-qc/9412014. 135. T. Thiemann, Class. Quantum Grav. 23 (2006) 2063, arXiv:gr-qc/0206037. 136. B. Bahr and T. Thiemann, Class. Quantum Grav. 26 (2009) 045011, arXiv:0709.4619 [gr-qc]. 137. B. Bahr and T. Thiemann, Class. Quantum Grav. 26 (2009) 045012, arXiv:0709.4636 [gr-qc]. 138. C. Flori, arXiv:0904.1303 [gr-qc]. 139. E. Bianchi, E. Magliaro and C. Perini, Phys. Rev. D 82 (2010) 024012, arXiv:0912.4054 [gr-qc]. 140. M. Bojowald, Living Rev. Rel. 11 (2008) 4, http://www.livingreviews.org/ Articles/lrr-2008-4. 141. A. Ashtekar and P. Singh, Class. Quantum Grav. 28 (2011) 213001, arXiv:1108.0893 [gr-qc]. 142. M. Bojowald, Lect. Notes Phys. 835 (2011) 1. 143. M. Bojowald, Canonical Gravity and Applications: Cosmology, Black Holes, and Quantum Gravity (Cambridge University Press, Cambridge, 2011). 144. N. J. LaFave, arXiv:gr-qc/9310036. 145. A. Perez, Class. Quantum Grav. 20 (2003) R43, arXiv:gr-qc/0301113. 146. D. Oriti, arXiv:gr-qc/0311066. 147. A. Perez, arXiv:gr-qc/0409061. 148. A. Perez, Living Rev. Rel. 16 (2013) 3, http://www.livingreviews.org/Articles/ lrr2013-3, arXiv:1205.2019 [gr-qc]. 149. C. Rovelli, Class. Quantum Grav. 28 (2011) 114005, arXiv:1004.1780 [gr-qc]. 150. C. Rovelli, PoS (QGQGS) 2011 (2011) 003, arXiv:1102.3660 [gr-qc].

May 4, 2017 9:53 b2167-ch20

II-344

151. 152. 153. 154. 155. 156. 157. 158. 159. 160. 161. 162.

163. 164. 165. 166. 167. 168. 169. 170. 171. 172. 173. 174. 175. 176. 177. 178. 179. 180.

181. 182. 183.

S. Carlip et al.

C. Rovelli, Lect. Notes Phys. 863 (2013) 57. S. W. Hawking, Commun. Math. Phys. 25 (1972) 152. J. M. Bardeen, B. Carter and S. W. Hawking, Commun. Math. Phys. 31 (1973) 161. J. A. Wheeler and K. Ford, Geons, Black Holes, and Quantum Foam (W. W. Norton, New York, 1998), Chap. 4. J. D. Bekenstein, Lett. Nuovo Cimento 4 (1972) 737. J. D. Bekenstein, Phys. Rev. D 7 (1973) 2333. J. D. Bekenstein, Phys. Rev. D 9 (1974) 3292. S. W. Hawking, Nature 248 (1974) 30. S. W. Hawking, Commun. Math. Phys. 43 (1975) 199, Erratum 46 (1976) 206. W. G. Unruh, Phys. Rev. D 14 (1976) 870. J. J. Bisognano and E. H. Wichmann, J. Math. Phys. 17 (1976) 303. B. S. DeWitt, Quantum gravity: The new synthesis, in General Relativity: An Einstein Centenary Survey, eds. S. W. Hawking and W. Israel (Cambridge University Press, Cambridge, 1979). G. W. Gibbons and M. J. Perry, Phys. Rev. Lett. 36 (1976) 985. G. W. Gibbons and M. J. Perry, Proc. R. Soc. Lond. A 358 (1978) 467494. G. W. Gibbons and S. W. Hawking, Phys. Rev. D 15 (1977) 2752. S. Carlip and C. Teitelboim, Class. Quantum Grav. 12 (1995) 1699, arXiv:gr-qc/ 9312002. M. K. Parikh and F. Wilczek, Phys. Rev. Lett. 85 (2000) 5042, arXiv:hep-th/9907001. L. Vanzo, G. Acquaviva and R. Di Criscienzo, Class. Quantum Grav. 28 (2011) 183001, arXiv:1106.4153 [gr-qc]. S. P. Robinson and F. Wilczek, Phys. Rev. Lett. 95 (2005) 011303, arXiv:gr-qc/ 0502074. S. Iso, H. Umetsu and F. Wilczek, Phys. Rev. Lett. 96 (2006) 151302, arXiv:hep-th/ 0602146. S. Iso, H. Umetsu and F. Wilczek, Phys. Rev. D 74 (2006) 044017, arXiv:hep-th/ 0606018. S. Iso, T. Morita and H. Umetsu, Phys. Rev. D 77 (2008) 045007, arXiv:0710.0456 [hep-th]. L. Bonora, M. Cvitan, S. Pallua and I. Smoli´c, J. High Energy Phys. 0812 (2008) 021, arXiv:0808.2360 [hep-th]. L. Bonora, M. Cvitan, S. Pallua and I. Smoli´c, Phys. Rev. D 80 (2009) 084034, arXiv:0907.3722 [hep-th]. D. Garﬁnkle, S. B. Giddings and A. Strominger, Phys. Rev. D 49 (1994) 958, arXiv:gr-qc/9306023. J. D. Brown, Phys. Rev. D 51 (1995) 5725, arXiv:gr-qc/9412018. F. Dowker, J. P. Gauntlett, D. A. Kastor and J. H. Traschen, Phys. Rev. D 49 (1994) 2909, arXiv:hep-th/9309075. R. B. Mann and S. F. Ross, Phys. Rev. D 52 (1995) 2254, arXiv:gr-qc/9504015. A. Strominger and C. Vafa, Phys. Lett. B 379 (1996) 99, arXiv:hep-th/9601029. A. W. Peet, TASI lectures on black holes in string theory, in TASI 99 : Strings, Branes and Gravity, eds. J. Harvey, S. Kachru and E. Silverstein (World Scientiﬁc, Singapore, 2001). S. R. Das and S. D. Mathur, Annu. Rev. Nucl. Part. Sci. 50 (2000) 153, arXiv:grqc/0105063. S. D. Mathur, Fortschr. Phys. 53 (2005) 793, arXiv:hep-th/0502050. S. D. Mathur, Class. Quantum Grav. 23 (2006) R115, arXiv:hep-th/0510180.

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

II-345

184. M. Banados, C. Teitelboim and J. Zanelli, Phys. Rev. Lett. 69 (1992) 1849, arXiv:hep-th/9204099. 185. A. Strominger, J. High Energy Phys. 9802 (1998) 009, arXiv:hep-th/9712251. 186. D. Birmingham, I. Sachs and S. Sen, Phys. Lett. B 424 (1998) 275, arXiv:hep-th/ 9801019. 187. K. Skenderis, Lect. Notes Phys. 541 (2000) 325, arXiv:hep-th/9901050. 188. K. V. Krasnov, Phys. Rev. D 55 (1997) 3505, arXiv:gr-qc/9603025. 189. A. Ashtekar, J. Baez, A. Corichi and K. Krasnov, Phys. Rev. Lett. 80 (1998) 904, arXiv:gr-qc/9710007. 190. A. Ashtekar, A. Corichi and K. Krasnov, Adv. Theor. Math. Phys. 3 (1999) 419, arXiv:gr-qc/9905089. 191. A. Ashtekar, J. C. Baez and K. Krasnov, Adv. Theor. Math. Phys. 4 (2000) 1, arXiv:gr-qc/0005126. 192. A. Ashtekar, S. Fairhurst and B. Krishnan, Phys. Rev. D 62 (2000) 104025, arXiv:grqc/0005083. 193. A. Ashtekar, C. Beetle and J. Lewandowski, Phys. Rev. D 64 (2001) 044016, arXiv:grqc/0103026. 194. A. Ashtekar, C. Beetle and J. Lewandowski, Class. Quantum Grav. 19 (2002) 1195, arXiv:gr-qc/0111067. 195. A. Ashtekar, A. Corichi and D. Sudarsky, Class. Quantum Grav. 20 (2003) 3413, arXiv:gr-qc/0305044. 196. A. Ashtekar and A. Corichi, Class. Quantum Grav. 20 (2003) 4473, arXiv:grqc/0305082. 197. A. Ashtekar, J. Engle and C. Van Den Broeck, Class. Quantum Grav. 22 (2005) L27, arXiv:gr-qc/0412003. 198. M. Domagala and J. Lewandowski, Class. Quantum Grav. 21 (2004) 5233, arXiv:grqc/0407051. 199. K. A. Meissner, Class. Quantum Grav. 21 (2004) 5245, arXiv:gr-qc/0407052. 200. A. Corichi, J. Diaz-Polo and E. Fernandez-Borja, Class. Quantum Grav. 24 (2007) 243, arXiv:gr-qc/0605014. 201. A. Corichi, J. Diaz-Polo and E. Fernandez-Borja, Phys. Rev. Lett. 98 (2007) 181301, arXiv:gr-qc/0609122. 202. I. Agullo, J. Fernando Barbero, E. F. Borja, J. Diaz-Polo and E. J. S. Villasenor, Phys. Rev. D 82 (2010) 084029, arXiv:1101.3660 [gr-qc]. 203. J. F. G. Barbero, J. Lewandowski and E. J. S. Villasenor, PoS (QGQGS) 2011 (2011) 023, arXiv:1203.0174 [gr-qc]. 204. J. D. Bekenstein and V. F. Mukhanov, Phys. Lett. B 360 (1995) 7, arXiv:grqc/9505012. 205. J. Engle, A. Perez and K. Noui, Phys. Rev. Lett. 105 (2010) 031302, arXiv:0905.3168 [gr-qc]. 206. J. Engle, K. Noui, A. Perez and D. Pranzetti, Phys. Rev. D 82 (2010) 044050, arXiv:1006.0634 [gr-qc]. 207. J. Engle, K. Noui, A. Perez and D. Pranzetti, J. High Energy Phys. 1105 (2011) 016, arXiv:1103.2723 [gr-qc]. 208. R. D. Sorkin, On the entropy of the vacuum outside a horizon, in Proc. of the Tenth Int. Conf. General Relativity and Gravitation, Padova, 4–9 July, 1983, Vol. II (Consiglio Nazionale della Ricerche, Roma, 1983), pp. 734–736. 209. L. Bombelli, R. K. Koul, J. Lee and R. D. Sorkin, Phys. Rev. D 34 (1986) 373. 210. M. Srednicki, Phys. Rev. Lett. 71 (1993) 666, arXiv:hep-th/9303048. 211. S. Ryu and T. Takayanagi, Phys. Rev. Lett. 96 (2006) 181602, arXiv:hep-th/0603001.

May 4, 2017 9:53 b2167-ch20

II-346

S. Carlip et al.

212. V. E. Hubeny, M. Rangamani and T. Takayanagi, J. High Energy Phys. 0707 (2007) 062, arXiv:0705.0016 [hep-th]. 213. A. D. Sakharov, Sov. Phys.-Dokl. 12 (1968) 1040, Dokl. Akad. Nauk Ser. Fiz. 177 (1967) 70; Sov. Phys.-Usp. 34 (1991) 394; Gen. Relat. Gravit. 32 (2000) 365. 214. V. P. Frolov, D. V. Fursaev and A. I. Zelnikov, Nucl. Phys. B 486 (1997) 339, arXiv:hep-th/9607104. 215. V. P. Frolov and D. V. Fursaev, Phys. Rev. D 56 (1997) 2212, arXiv:hep-th/9703178. 216. S. Carlip, Class. Quantum Grav. 17 (2000) 4175, arXiv:gr-qc/0005017. 217. S. D. Mathur, Lect. Notes Phys. 769 (2009) 3, arXiv:0803.2030 [hep-th]. 218. S. Carlip, Chapter 22 in this book; Int. J. Mod. Phys. D 23 (2014) 1430023, arXiv:1410.1486 [gr-qc]. 219. V. A. Kostelecky and S. Samuel, Phys. Rev. D 39 (1989) 683. 220. C. P. Burgess, J. M. Cline, E. Filotas, J. Matias and G. D. Moore, J. High Energy Phys. 0203 (2002) 043, arXiv:hep-ph/0201082. 221. R. Gambini and J. Pullin, Phys. Rev. D 59 (1999) 124021, arXiv:gr-qc/9809038. 222. C. Rovelli and S. Speziale, Phys. Rev. D 83 (2011) 104029, arXiv:1012.1739 [gr-qc]. 223. J. Lukierski, H. Ruegg and W. J. Zakrzewski, Ann. Phys. 243 (1995) 90, arXiv:hepth/9312153. 224. G. Amelino-Camelia and S. Majid, Int. J. Mod. Phys. A 15 (2000) 4301, arXiv:hepth/9907110. 225. S. M. Carroll, J. A. Harvey, V. A. Kostelecky, C. D. Lane and T. Okamoto, Phys. Rev. Lett. 87 (2001) 141601, arXiv:hep-th/0105082. 226. C. Barcelo, S. Liberati and M. Visser, Living Rev. Rel. 8 (2005) 12; Living Rev. Rel. 14 (2011) 3, http://relativity.livingreviews.org/Articles/lrr-2011-3/, arXiv:grqc/0505065. 227. W. T. Ni, Phys. Lett. A 379 (2015) 1297, arXiv:1411.0460 [gr-qc]. 228. W. T. Ni, Phys. Lett. A 378 (2014) 3413, arXiv:1410.0126 [gr-qc]. 229. S. di Serego Alighieri, Int. J. Mod. Phys. D 24 (2015) 1530016, arXiv:1501.06460 [astro-ph.CO]. 230. H. H. Mei, W. T. Ni, W. P. Pan, L. Xu and S. di Serego Alighieri, Astrophys. J. 805 (2015) 107, arXiv:1412.8569 [astro-ph.CO]. 231. K. Kotera and A. V. Olinto, Annu. Rev. Astron. Astrophys. 49 (2011) 119, arXiv:1101.4256 [astro-ph.HE]. 232. Pierre Auger Collab. (T. Yamamoto), arXiv:0707.2638 [astro-ph]. 233. F. W. Stecker and S. T. Scully, New J. Phys. 11 (2009) 085003, arXiv:0906.1735 [astro-ph.HE]. 234. S. Liberati and L. Maccione, J. Phys., Conf. Ser. 314 (2011) 012007, arXiv:1105.6234 [astro-ph.HE]. 235. G. Amelino-Camelia, A. Marciano, M. Matassa and G. Rosati, arXiv:1006.0007 [astro-ph.HE]. 236. G. Amelino-Camelia, C. L¨ ammerzahl, F. Mercati and G. M. Tino, Phys. Rev. Lett. 103 (2009) 171302, arXiv:0911.1020 [gr-qc]. 237. F. Mercati, D. Mazon, G. Amelino-Camelia, J. M. Carmona, J. L. Cortes, J. Indurain, C. L¨ ammerzahl and G. M. Tino, Class. Quantum Grav. 27 (2010) 215003, arXiv:1004.0847 [gr-qc]. 238. T. Damour and K. Nordtvedt, Phys. Rev. Lett. 70 (1993) 2217. 239. G. R. Dvali, G. Gabadadze and M. Porrati, Phys. Lett. B 485 (2000) 208, arXiv:hepth/0005016. 240. C. de Rham, Living Rev. Rel. 17 (2014) 7, http://relativity.livingreviews.org/ Articles/lrr-2014-7/, arXiv:1401.4173 [hep-th].

May 4, 2017 9:53 b2167-ch20

Quantum gravity: A brief history of ideas and some prospects

II-347

241. W. T. Ni, Class. Quantum Grav. 26 (2009) 075021, arXiv:0812.0887 [astro-ph]. 242. ASTROD Collab. (C. Braxmaier et al.), Exp. Astron. 34 (2012) 181, arXiv:1104.0060 [gr-qc]. 243. J. R. Gair, M. Vallisneri, S. L. Larson and J. G. Baker, Living Rev. Rel. 16 (2013) 7, http://relativity.livingreviews.org/Articles/lrr-2013-7/, arXiv:1212.5575 [gr-qc]. 244. C. L¨ ammerzahl, arXiv:gr-qc/0402122. 245. A. Ashtekar, arXiv:1303.4989 [gr-qc]. 246. J. Grain and A. Barrau, Phys. Rev. Lett. 102 (2009) 081301, arXiv:0902.0145 [gr-qc]. 247. A. Barrau, arXiv:0911.3745 [gr-qc]. 248. J. Mielczarek, T. Cailleteau, J. Grain and A. Barrau, Phys. Rev. D 81 (2010) 104049, arXiv:1003.4660 [gr-qc]. 249. J. Grain, A. Barrau, T. Cailleteau and J. Mielczarek, Phys. Rev. D 82 (2010) 123520, arXiv:1011.1811 [astro-ph.CO]. 250. A. Barrau, PoS (ICHEP) 2010 (2010) 461, arXiv:1011.5516 [gr-qc]. 251. I. Agullo, A. Ashtekar and W. Nelson, Class. Quantum Grav. 30 (2013) 085014, arXiv:1302.0254 [gr-qc]. 252. BICEP2 and Planck Collab. (P. A. R. Ade et al.), Phys. Rev. Lett. 114(10) (2015) 101301, arXiv:1502.00612 [astro-ph.CO]. 253. G. Amelino-Camelia, Living Rev. Rel. 16 (2013) 5, http://relativity.livingreviews. org/Articles/lrr-2013-5/, arXiv:0806.0339 [gr-qc]. 254. L. Smolin, Lect. Notes Phys. 669 (2005) 363.

May 4, 2017 9:53 b2167-ch21

Chapter 21 Perturbative quantum gravity comes of age

R. P. Woodard Department of Physics, University of Florida, Gainesville, FL 32611, USA [email protected]

I argue that cosmological data from the epoch of primordial inﬂation is catalyzing the maturation of quantum gravity from speculation into a hard science. I explain why quantum gravitational eﬀects from primordial inﬂation are observable. I then review what has been done, both theoretically and observationally, and what the future holds. I also discuss what this tells us about quantum gravity. Keywords: Inﬂation; quantum gravity; cosmology. PACS Number(s): 04.60.−m, 04.62.+v, 98.80.−k

1. Introduction Gravity was the ﬁrst of the fundamental forces to impress its existence upon our ancestors, because it is universally attractive and long range. These same features ensure its precedence in cosmology. Gravity also couples to stress-energy, which is why quantum general relativity is not perturbatively renormalizable,1–13 and why identiﬁable eﬀects are unobservably weak at low energies.14 These problems have hindered the study of quantum gravity until recently. This paper is about how interlocking developments in the theory and observation of inﬂationary cosmology have changed that situation, and what the future holds. The experiences of two Harvard graduate students serve to illustrate the situation before inﬂation. The ﬁrst is Leonard Parker who took his degree in 1967, based on his justly famous work quantifying particle production in an expanding universe.15–17 Back then people believed that the expansion of the universe had been constantly slowing down or “decelerating.” Parker’s work was greeted with indiﬀerence on account of the small particle production associated with the current expansion, and on the inability of a decelerating universe to preserve memories of early-times when the expansion rate was much higher. The ruling dogma of the 1960s was S-matrix theory, whose more extreme proponents believed they could II-349

May 4, 2017 9:53 b2167-ch21

II-350

R. P. Woodard

guess the strong interaction S-matrix based on a very few properties such as analyticity and unitarity. Through a curious process this later morphed into string theory. Quantum ﬁeld theory was regarded as a failed formalism whose success for quantum electrodynamics was an accident. Conﬁrmation of the Standard Model had changed opinions about quantum ﬁeld theory by my own time at Harvard (1977–1983). However, the perturbative nonrenormalizability of quantum general relativity led to dismissive statements such as, “only old men should work on quantum gravity.” The formalism of quantum ﬁeld theory had also become completely tied to asymptotic scattering experiments. For example, no one worried about correcting free vacuum because inﬁnite time evolution from “in” states to “out” states was supposed to do this automatically. Little attention was paid to making observations at ﬁnite times because the S-matrix was deemed the only valid observable, the knowledge of which completely deﬁned a quantum ﬁeld theory. My thesis on developing an invariant extension of local Green’s functions for quantum gravity was only accepted because Brandeis Professor Stanley Deser vouched for it. I left it unpublished for eight years.18 The situation was no better during the early stages of my career. As a postdoc, I worked with a very bright graduate student who dismissed the quantum gravity community as “la-la land” and made no secret of his plan to change ﬁelds. And there is no denying that any number of crank ideas were treated with perfect seriousness in those days, which validated our critics. I recall knowledgeable people questioning why anyone bothered trying to quantize gravity in view of the classical theory’s success. That opinion was never viable in view of the fact that the lowest divergences of quantum gravity4–9 derive from the gravitational response to matter theories which are certainly quantum, whether or not gravitons exist.14 The diﬀerence between then and now is that I can point to data — and quite a lot of it — from the same gravitational response to quantum matter. Today cosmological particle production is recognized as the source of the primordial perturbations which seeded structure formation. There is a growing realization that these perturbations are quantum gravitational phenomena,14,19,20 and that they cannot be described by any sort of S-matrix or by the use of in–out quantum ﬁeld theory.21,22 This poses a challenge for fundamental theory and an opportunity for its practitioners, which dismays some physicists and delights others. All of the problems that had to be solved for ﬂat space scattering theory in the mid 20th century are being re-examined, in particular, deﬁning observables which are infrared (IR) ﬁnite, renormalizable (at least in the sense of low energy eﬀective ﬁeld theory) and in rough agreement with the way things are measured.23,24 People are also thinking seriously about how to perturbatively correct the initial state.25 This revolutionary change of attitude did not result from any outbreak of sobriety within the quantum gravity community, or of toleration from our colleagues. The transformation was forced upon us by the overwhelming data in support of inﬂationary cosmology. In the coming sections of this paper I review the theory

May 4, 2017 9:53 b2167-ch21

Perturbative quantum gravity comes of age

II-351

behind that data, in particular: • • • • •

Why quantum gravitational eﬀects from inﬂation are observable; Why the tree order power spectra are quantum gravitational eﬀects; Loop corrections to the primordial power spectra; Other potentially observable eﬀects; and What the future holds.

2. Why Quantum Gravitational Eﬀects from Primordial Inﬂation are Observable Three things are responsible for the remarkable fact that quantum gravitational eﬀects from the epoch of primordial inﬂation can be observed today: • The inﬂationary Hubble parameter is large enough that quantum gravitational eﬀects are small, but not negligible; • Long wavelength gravitons and massless, minimally coupled scalars experience explosive particle production during inﬂation; and • The process of ﬁrst horizon crossing results in long wavelength gravitons and massless, minimally coupled scalars becoming fossilized so that they can survive down to the current epoch. I will make the ﬁrst point at the beginning, in the subsection on the inﬂationary background. Then the subsection on perturbations discusses the second and third points. 2.1. The background geometry On scales larger than about 100 Mpc the observable universe is approximately homogeneous, isotropic and spatially ﬂat. The invariant element for such a geometry can be put in the form, ds2 = −c2 dt2 + a2 (t)dx · dx.

(1)

Two derivatives of the scale factor a(t) have great signiﬁcance, the Hubble parameter H(t) and the ﬁrst slow roll parameter (t), H(t) ≡

a˙ , a

(t) ≡ −

H˙ . H2

(2)

Inﬂation is deﬁned as H(t) > 0 with (t) < 1. One can see that it is possible from the current values of the cosmological parameters (denoted by a subscript zero),26 H0 ≈ 2.2 × 10−18 Hz,

0 ≈ 0.47.

(3)

May 4, 2017 9:53 b2167-ch21

II-352

R. P. Woodard

However, the important phase of inﬂation for my purposes is Primordial Inﬂation, which is conjectured to have occurred during the ﬁrst 10−32 s of existence. If the BICEP2 detection of primordial B-mode polarization is accepted, then we ﬁnally know the values of H(t) and (t) near the end of primordial inﬂation,27 Hi ≈ 1.8 × 10+38 Hz,

i ≈ 0.013.

(4)

I will comment later on the signiﬁcance of Hi . Let us here note that i is very near the de Sitter limit of = 0 at which the Hubble parameter becomes constant. This is a very common background to use when estimating quantum eﬀects during primordial inﬂation. We have direct observational evidence that both the scale factor and its logarithmic time derivative H(t) have changed over many orders of magnitude during cosmic history. In contrast, the deceleration parameter only varies over the small range 0 ≤ (t) ≤ 2. Figure 1 shows what we think we know about (t) as a function of the number of e-foldings since the end of primordial inﬂation at t = te , a(t) N (t) ≡ ln . (5) a(te ) It is a tribute to decades of observational work that only a small portion of this ﬁgure is really unknown, corresponding to the phase of re-heating at the end of inﬂation.

2 as a function of the ˙ Fig. 1. The red curve shows the ﬁrst slow roll parameter (t) ≡ −H/H number of e-foldings N since the end of primordial inﬂation. A question mark stands for the phase of reheating between the epochs of primordial inﬂation and radiation domination, because there are many models for this period. Signiﬁcant events marked on the graph are Big Bang Nucleosynthesis (BBN) when the seven lightest isotopes were produced, matter-radiation equality (eq) when the energy density was composed of equal amounts of relativistic and nonrelativistic matter, and recombination (rec) when neutral Hydrogen formed and the cosmic microwave radiation began free-streaming. Observable cosmological perturbations experience ﬁrst horizon crossing near the lower left hand corner of the graph.

May 4, 2017 9:53 b2167-ch21

Perturbative quantum gravity comes of age

II-353

Primordial inﬂation was advanced in the late 1970s and early 1980s to explain the absence of observed relics (primordial black holes, magnetic monopoles, cosmic strings) and the initial conditions (homogeneous, isotropic and spatially ﬂat) for the long epoch of radiation domination which is visible on Fig. 1. After some notable precursors,28–31 the paper of Guth32 focused attention on the advantages of a early epoch of inﬂation and, incidentally, coined the name. Important additional work concerned ﬁnding an acceptable way to commence inﬂation and to make it end.33,34 The ﬁrst completely successful model was Linde’s “Chaotic Inﬂation.”35 One of the most powerful motivations for primordial inﬂation is that it explains the Horizon Problem of why events far back in our past light-cone seem so uniform. I will review the argument here because the same analysis is useful for the next subsection. From the cosmological geometry (1), we can easily compute the coordinate distance R(t2 , t1 ) traversed by a light ray whose trajectory obeys ds2 = 0, t1 cdt R(t2 , t1 ) ≡ . (6) a(t) t2 Now note the relation,

d 1 1 ˙ = 1− . dt ( − 1)Ha a ( − 1)2 H

(7)

One can see from Fig. 1 that (t) was nearly constant over long periods of cosmic evolution, in particular during the epoch of radiation domination, which would extend back to the beginning if it were not for primordial inﬂation. So we can drop the second term of (7) to conclude, R(t2 , t1 ) ≈

c c − . (1 − 1)H1 a1 (2 − 1)H2 a2

(8)

One additional exact relation brings the horizon problem to focus, d [H(t)a(t)] = −[(t) − 1]H 2 (t)a(t). dt

(9)

Combining Eq. (9) with (8) reveals a crucial distinction between inﬂation ((t) < 1) and deceleration ((t) > 1): During deceleration the radius of the light-cone is dominated by its upper limit, whereas the lower limit dominates during inﬂation. The horizon problem derives from assuming that there was no phase of primordial inﬂation so that the epoch of radiation domination extends back to the beginning of the universe. Suppose that the universe began at t = t2 and we view some early event such as recombination (rec on Fig. 1) or big bang nucleosynthesis (BBN on Fig. 1). At time t = t1 we can see things out to the radius of our past light-cone R(t1 , t0 ) which is vastly larger than the radius of the forward light-cone R(t2 , t1 ) ≈ c/[(1 − 1)H1 a1 ] that anything can have traveled from the beginning of time. For example, the cosmic microwave radiation is uniform to one part in 105 , which is far better thermal equilibrium than the air of the room in which you are sitting. Without a phase of primordial inﬂation we are seeing about 2200 diﬀerent

May 4, 2017 9:53 b2167-ch21

II-354

R. P. Woodard

patches of the sky which have not even had time to exchange a single photon, much less achieve a high degree of thermal equilibrium.14 Of course the problem just gets worse the further back we look. At the time of big bang nucleosynthesis BBN we are seeing about 1015 causally disconnected regions, which are nonetheless in rough thermal equilibrium.14 Without inﬂation the radius R(t2 , t1 ) of the forward light-cone is almost independent of the beginning of time t2 . No matter how early we make t2 , it is not possible to increase R(t2 , t1 ) more than about c/[(1 − 1)H1 a1 ]. Hence, the high degree of uniformity we observe in the early universe would have to be a spectacularly unlikely accident. Primordial inﬂation solves the problem neatly by making the lower limit of the forward light-cone dominate, R(t2 , t1 ) ≈ c/[(1 − 2 )H2 a2 ]. We can make the radius of the forward light-cone much larger than the radius of the past light-cone, so that causal processes would have had plenty of time to achieve the high degree of equilibrium that is observed. Before closing this subsection, I want to return to the numerical values quoted for H0 and Hi in relations (3)–(4). The loop counting parameter of quantum gravity can 2 ≡ G/c5 ≈ 2.9×10−87 s2 . be expressed in terms of the square of the Planck time, TPl Quantum gravitational eﬀects from a process whose characteristic frequency is ω 2 . For inﬂationary particle production the characteristic are typically of order ω 2 TPl frequency is of course the Hubble parameter, so we can easily compare the strengths of quantum gravitational eﬀects during the current phase of inﬂation and from the epoch of primordial inﬂation, GH02 ≈ 1.4 × 10−122 , c5

GHi2 ≈ 9.4 × 10−11 . c5

(10)

The minuscule ﬁrst number is why we will never detect quantum gravitational eﬀects from the current phase of inﬂation. Although the second number is tiny, it is not so small as to preclude detection, if only the signal can persist until the present day. In the next subsection, I will explain how that can happen. The loop counting parameter GH 2 /c5 is the quantum gravitational analog of the quantum electrodynamics ﬁne structure constant α ≡ e2 /4π0 c ≈ 7.3 × 10−3 . Both parameters control the strength of perturbative corrections. Recall that a result in quantum electrodynamics — for example, the invariant amplitude of Compton scattering — typically consists of a lowest, tree order contribution of strength α, then each additional loop brings an extra factor of α. In the same way, the lowest, tree order quantum gravity eﬀects from inﬂationary particle production have strength GH 2 /c5 , and each addition loop brings an extra factor of GH 2 /c5 . Because the quantum gravitational loop counting parameter from primordial inﬂation is so much smaller than its quantum electrodynamics cousin, we expect that quantum gravitational perturbation theory should be wonderfully accurate. In fact, all that can be resolved with current data is the tree order eﬀect, although I will argue in Sec. 4.4 that the one loop correction may eventually be resolved. Beyond that there is no hope.

May 4, 2017 9:53 b2167-ch21

Perturbative quantum gravity comes of age

II-355

2.2. Inﬂationary particle production The phenomenon of polarization in a medium is covered in undergraduate electrodynamics. The medium contains a vast number of bound charges. The application of an electric ﬁeld makes positive charges move with the ﬁeld and the negative changes move opposite. That charge separation polarizes the medium and tends to reduce the electric ﬁeld strength. One of the amazing predictions of quantum ﬁeld theory is that virtual particles are continually emerging from the vacuum, existing for a brief period, and then disappearing. How long these virtual particles can exist is controlled by the energy– time uncertainty principle, which gives the minimum time ∆t needed to resolve and energy diﬀerence ∆E, ∆t∆E .

(11)

If one imagines the emergence of a pair of positive and negatively charged particles of mass m and wave vector ±k then the energy went from zero to E = 1 2[m2 c4 + 2 c2 k 2 ] 2 . To not resolve a violation of energy conservation, the energy– time uncertainty principle requires the pair to disappear after a time ∆t given by,

∆t ∼ √ . m 2 c4 + 2 c2 k 2

(12)

The rest is an exercise in classical (that is, nonquantum) physics. If we ignore the change in the particles’ momentum then their positions obey, d2 2 2 ±2 eE 2 k 2 ∆x ) = ±eE ⇒ ∆x (∆t) = ( m c + (13) ± ± 3 . cdt2 2c[m2 c2 + 2 k 2 ] 2 Hence the polarization induced by wave vector k is, p = +e∆x+ (∆t) − e∆x− (∆t) =

2 e 2 E 3

c[m2 c2 + 2 k 2 ] 2

.

(14)

The full vacuum polarization density comes from integrating d3 k/(2π)3 . The simple analysis I have just sketched gives pretty nearly the prediction from one loop quantum electrodynamics, which is in quantitative agreement with experiment. It allows us to understand two features of vacuum polarization which would be otherwise obscure: • That the largest eﬀect derives from the lightest charged particles because they have the longest persistence times ∆t and therefore induce the greatest polarization; and • That the electrodynamic interaction becomes stronger at short distances because the longest wavelength (hence smallest k) virtual particles could induce more polarization than is allowed by the travel time between two very close sources. Cosmological expansion can strengthen quantum eﬀects because it causes the virtual particles which drive them to persist longer. This is easy to see from the

May 4, 2017 9:53 b2167-ch21

R. P. Woodard

II-356

geometry (1). Because spatial translation invariance is unbroken, particles still have conserved wave numbers k. However, because the physical distance is the coordinate distance scaled by a(t), the physical energy of a particle with mass m and wave number k = 2π/λ becomes time-dependent, 2 c2 k 2 E(t, k) = m2 c4 + 2 . (15) a (t) Hence, the relation for the persistence time ∆t of a virtual pair which emerges at time t changes from (12) to, t+∆t dt E(t , k) ∼ . (16) t

Massless particles persist the longest, just as they do in ﬂat space. However, for inﬂation it is the lower limit of (16) which dominates, so that even taking ∆t to inﬁnity does not cause the integral to grow past a certain point. One can see this from the de Sitter limit, t+∆t ck ck dt = [1 − e−Hi ∆t ]. (17) a(t ) H i a(t) t A particle with ck < H(t)a(t) is said to be super-horizon, and we have just shown that any massless virtual particle which emerges from the vacuum with a superhorizon wave number during inﬂation will persist forever. It turns out that almost all massless particles possess a symmetry known as conformal invariance which suppresses the rate at which they emerge from the vacuum. This keeps the density of virtual particles small, even though any that do emerge can persist forever. One can see the problem by specializing the Lagrangian of a massless, conformally coupled scalar ψ(t, x) to the cosmological geometry (1), R 2√ a3 ψ˙ 2 1 ∂i ψ∂i ψ (H˙ + 2H 2 )ψ 2 µν √ −g − ψ −g → L = − ∂µ ψ∂ν ψg − − . 2 12 2 c2 a2 c2 (18) The equation for a canonically normalized, spatial plane wave of the form ψ(t, x) = v(t, k)eik · x can be solved for a general scale factor a(t), t dt 2 2 exp −ick c k ti a(t ) ˙ + 2H 2 v = 0 ⇒ v(t, k) = v¨ + 3H v˙ + + H . (19) 2 a 2ck a(t) The factor of 1/a(t) in (19) suppresses the emergence rate, even though destructive interference from the phase die oﬀ, just as the energy–time uncertainty principle (17) predicts. The stress-energy contributed by this ﬁeld is, 1 1 1 ρ σ ρσ Tµν = δµ δν − gµν g ∂ρ ψ∂σ ψ + Rµν − gµν R + gµν − Dµ Dν ψ 2 , (20) 2 6 2

May 4, 2017 9:53 b2167-ch21

Perturbative quantum gravity comes of age

II-357

is the covariant d’Alembertian. We where Dµ is the covariant derivative and can get the zero-point energy of a single wave vector k by specializing T00 to the cosmological geometry (1) and multiplying by a factor of a3 (t), 2 2

a3 c k ck 2 2 2 ∗ ∗ E(t, k) = |v| ˙ + + H |v| + H(v v˙ + vv ˙ ) = . (21) 2 a2 2a(t) This is just the usual 12 ω term which is not strengthened but rather weakened by the cosmological expansion. Only gravitons and massless, minimally coupled scalars are both massless and not conformally invariant so that they can engender signiﬁcant quantum eﬀects during inﬂation. Because they obey the same mode equation36,37 it will suﬃce to specialize the scalar Lagrangian to the cosmological geometry (1), 1 1 3 φ˙ 2 1 µν √ L = − ∂µ φ∂ν φg −g → a − 2 ∂i φ∂i φ . (22) 2 2 c2 a The equation for a canonically normalized, spatial plane wave of the form φ(t, x) = u(t, k)eik · x is simpler than that of its conformally coupled cousin (19) but more diﬃcult to solve, so I will specialize the solution to de Sitter, t dt exp −ick c2 k 2 iHi a(t) ti a(t ) u ¨ + 3H u˙ + 2 u = 0 ⇒ u(t, k) = 1+ . a 2ck ck a(t) (23) The minimally coupled mode function u(t, k) has the same phase factor as the conformal mode function (19), and they both fall oﬀ like 1/a(t) in the far subhorizon regime of ck Hi a(t). However, they disagree strongly in the super-horizon regime during which v(t, k) continues to fall oﬀ whereas u(t, k) approaches a phase times Hi /2c3 k 3 . One can see from the equation on the left of (23) that u(t, k) approaches a constant for any inﬂating geometry. The zero-point energy in wave vector k is,

2 c2 k 2 2 Hi a(t) 1 3 ck 1 2 E(t, k) = a |u| ˙ + 2 |u| = + . (24) 2 a a(t) 2 2ck Because each wave vector is an independent harmonic oscillator with mass proportional to a3 (t) and frequency k/a(t) we can read oﬀ the occupation number from expression (24), 2 Hi a(t) N (t, k) = . (25) 2ck As one might expect, this number is small in the sub-horizon regime. It becomes of order one at the time tk of horizon crossing, ck = H(tk )a(tk ), and N (t, k) grows explosively afterward. This is crucial because it means that inﬂationary particle production is an IR eﬀect. That means we can study it reliably using quantum general relativity, even though that theory is not perturbatively renormalizable.

May 4, 2017 9:53 b2167-ch21

II-358

R. P. Woodard

The ﬁnal point I wish to make is that the mode function u(t, k) becomes constant after ﬁrst horizon crossing. For de Sitter this constant is calculable, ick u(t, k)|dS → iHi exp − . (26) 2c3 k 3 Hi a i However, one can see from the mode equation on the left-hand side of (23) that the approach to a constant happens for any inﬂating geometry. Recall from equation (9) that the inverse horizon length c−1 H(t)a(t) grows during inﬂation and decreases during the later phase of deceleration which encompasses so much of the cosmological history depicted in Fig. 1. Hence, we can give the following rough summary of the “life cycle of a mode” of wave number k: • At the onset of primordial inﬂation the mode has ck H(t)a(t) so the mode function oscillates and fall oﬀ like 1/a(t); • If inﬂation lasts long enough the mode will eventually experience ﬁrst horizon crossing ck = H(tk )a(tk ), after which mode function becomes approximately constant; and • During the phase of deceleration which follows primordial inﬂation, modes which experienced ﬁrst horizon crossing near the end of inﬂation re-enter the horizon ck = H(Tk )a(Tk ), after which they begin participating in dynamical processes with amplitude larger by a factor of a(Tk )/a(tk ) than they would have had without ﬁrst horizon crossing. This is how quantum gravitational eﬀects from the epoch of primordial inﬂation become fossilized so that they can be detected now. 3. Tree Order Power Spectra Although the evidence for primordial inﬂation is overwhelming, there is not yet any compelling mechanism for causing it. The simplest class of successful models is based on general relativity plus a scalar inﬂaton ϕ(x) whose potential V (ϕ) is regarded as a free function,35 4 √ c R 1 L= − ϕ,µ ϕ,ν gµν − V (ϕ) −g. (27) 16πG 2 Here gµν (x) is the D-dimensional, spacelike metric with Ricci curvature R. (I will work in D spacetime dimensions to facilitate the use of dimensional regularization, even though the D = 4 limit must eventually be taken for physical results). The purpose of this section is to show how this simple model can not only drive primordial inﬂation but also the quantum gravitational ﬂuctuations whose imprint on the cosmic background radiation has been imaged with stunning accuracy.26,38–40 I ﬁrst demonstrate that the potential V (ϕ) can be chosen to support the cosmological geometry (1) with any scale factor a(t) for which the Hubble parameter is monotonically decreasing. I also comment on the many problems of plausibility which seem to point to the need for a better model. I then decompose perturbations

May 4, 2017 9:53 b2167-ch21

Perturbative quantum gravity comes of age

II-359

about the background (1) into a scalar ζ(x) and a transverse-traceless tensor hij (x). Owing to the particle production mechanism adumbrated in Sec. 2.2, certain modes of ζ and hij become hugely excited during primordial inﬂation, and then freeze-in so that they can survive to much later times. The strength of this eﬀect is quantiﬁed by the primordial scalar and tensor power spectra, which I deﬁne and compute at tree order, along with associated observables. I then discuss the controversy which has arisen concerning an alternate deﬁnition of the tree order power spectra. The section closes with an explanation of why the tree order power spectra are the ﬁrst quantum gravitational eﬀects ever to have been resolved. I will adopt the notation employed in recent studies by Maldacena41 and by Weinberg,21 however, the original work for tensors was done in 1979 by Starobinsky,42 and for scalars in 1981 by Mukhanov and Chibisov.43 Important subsequent work was done over the course of the next several years by Hawking,44 Guth and Pi,45 Starobinsky,46 Bardeen et al.47 and by Mukhanov.48 Some classic review articles on the subject are Refs. 49–51. 3.1. The background for single-scalar inﬂation There is no question that a minimally coupled scalar potential model of the form (27) can support inﬂation because there is a constructive procedure for ﬁnding the potential V (ϕ) given the expansion history a(t).52–55 For the geometry (1) the scalar depends just on time ϕ0 (t) and only two of Einstein’s equations are nontrivial, 1 8πG ϕ˙ 20 (D − 2)(D − 1)H 2 = 2 + V (ϕ ) (28) 0 , 2 c c2 1 8πG ϕ˙ 20 2 ˙ −(D − 2)H − (D − 2)(D − 1)H = 2 − V (ϕ0 ) . (29) 2 c c2 By adding (28) and (29) one obtains the relation, 8πG −(D − 2)H˙ = 4 ϕ˙ 20 . (30) c Hence, one can reconstruct the scalar’s evolution provided the Hubble parameter is monotonically decreasing, and that relation can be inverted (numerically if need be) to solve for time as a function of t ˙ ) −(D − 2)c4 H(t ϕ0 (t) = ϕ0 (t2 ) ± dt ⇔ t = T (ϕ0 ). (31) 8πG t2 One then determines the potential by subtracting (29) from (28) and evaluating the resulting function of time at t = T (ϕ0 ), (D − 2)c2 ˙ [H(T (ϕ0 )) + 3H 2 (T (ϕ0 ))]. (32) 16πG Just because scalar potential models (27) can be adjusted to work does not mean they are particularly plausible. They suﬀer from six sorts of sometimes contradictory V (ϕ0 ) =

May 4, 2017 9:53 b2167-ch21

R. P. Woodard

II-360

ﬁne-tuning problems: (1) Initial Conditions — Inﬂation must begin with the inﬂaton approximately homogeneous, and potential-dominated, over more than a Hubble volume56 ; (2) Duration — The inﬂation potential must be ﬂat enough to make inﬂation last for at least 50 e-foldings33,34 ; (3) Scalar Perturbations — Getting the right magnitude for the scalar power spec2 trum requires G3 /c11 × V 3 /V ∼ 10−11 (Ref. 49); (4) Tensor Perturbations — Getting the right magnitude for the tensor power spectrum requires c4 /G × (V /V )2 ∼ 1 (Ref. 49); (5) Reheating — The inﬂation must couple to ordinary matter (its gravitational couplings do not suﬃce) so that its post-inﬂationary kinetic energy produces a hot, radiation-dominated universe57 ; (6) Cosmological Constant — The minimum of the scalar potential must have the right value G2 Vmin /c7 ≈ 10−123 to contribute the minuscule vacuum energy we observe today.58–61 Note that adding the matter couplings required to produce reheating puts 2–4 at risk because matter loop eﬀects induce Coleman–Weinberg corrections to the inﬂation eﬀective potential. Nor does fundamental theory provide any explanation for why the cosmological constant is so small.62,63 The degree of ﬁne-tuning needed to enforce these six conditions strains credulity, and disturbs even those who devised the early models.64–67 Opinions diﬀer, but I feel it is a mistake to make too much of the defects of single-scalar inﬂation. The evidence for an early phase of accelerated expansion is overwhelming and really incontrovertible, irrespective of what caused it. Further, all that we know about low energy eﬀective ﬁeld theory conﬁrms that the general relativity portion of Lagrangian (27) must be valid, even at the scales of primordial inﬂation. That suﬃces to establish the quantum gravitational character of primordial perturbations, even without a compelling model for what caused inﬂation. So I will go forward with the analysis on the basis of the single-scalar model (27), ﬁrm in the belief that whatever eventually supplants it must exhibit many of the same features. 3.2. Gauge-ﬁxed, constrained action We decompose gµν into lapse, shift and spatial metric according to Arnowitt, Deser and Misner (ADM),68–70 gµν dxµ dxν = −N 2 c2 dt2 + gij (dxi − N i cdt)(dxj − N j cdt).

(33)

ADM long ago showed that the Lagrangian has a very simple dependence upon the lapse,68–70 √ c4 g B L = (Surface Terms) − N ·A+ . (34) 16πG N

May 4, 2017 9:53 b2167-ch21

Perturbative quantum gravity comes of age

II-361

The quantity A is a potential energy, A = −R +

16πG 1 ij V (ϕ) + ϕ ϕ g , ,i ,j c4 2

(35)

where R is the (D − 1)-dimensional Ricci scalar formed from gij . The quantity B in (34) is a sort of kinetic energy.

2 8πG ϕ˙ i 2 ij i B = (E i ) − E Eij − 4 − ϕ,i N , (36) c c where Eij /N is the extrinsic curvature, 1 [Ni;j + Nj;i − c−1 g˙ ij ] (37) 2 and a semicolon denotes spatial covariant diﬀerentiation using the connection compatible with gij . ADM ﬁx the gauge by specifying N (t, x) and N i (t, x), however, Maldacena41 and Weinberg21 instead impose the conditions, Eij ≡

G0 (t, x) ≡ ϕ(t, x) − ϕ0 (t) = 0,

(38)

Gi (t, x) ≡ ∂j hij (t, x) = 0.

(39)

The transverse-traceless graviton ﬁeld hij (t, x) is deﬁned by decomposing the spatial metric into a conformal part and a unimodular part gij , √ gij (t, x) ⇒ g = aD−1 e(D−1)ζ . gij = a2 (t)e2ζ(t,x) (40) The unimodular part is obtained by exponentiating the transverse-traceless graviton ﬁeld hij (t, x), 1 gij ≡ [eh ]ij = δij + hij + hik hjk + O(h3 ), 2

hii = 0.

(41)

The Faddeev–Popov determinant associated with (38), (39) depends only on hij , 2 ˙ . and becomes singular for = 0. Recall that H(t) = a/a ˙ and (t) = −H/H Of course, no gauge can eliminate the physical scalar degree of freedom which is evident in (27). With condition (38) the inﬂation degree of freedom resides in ζ(t, x) and linearized gravitons are carried by hij (t, x). In this gauge the lapse and shift are constrained variables which mediate important interactions between the dynamical ﬁelds but contribute no independent degrees of freedom. Varying (34) with respect to N produces an algebraic equation for N , B B A− 2 =0⇒N = . (42) N A This gives the constrained Lagrangian a “virial” form,71 √ c4 g √ Lconst. = (Surface Terms) − AB. 8πG

(43)

May 4, 2017 9:53 b2167-ch21

R. P. Woodard

II-362

From relations (30–32) one can see that the background values of the potential and kinetic terms are equal, A0 = B0 = (D − 2)c−2 [H˙ + (D − 1)H 2 ]. Hence, the background value of the lapse is unity. There is unfortunately no nonperturbative solution for the shift N i (t, x) in terms of ζ and hij , so its constraint equation must be solved perturbatively. One ﬁrst employs (40) to exhibit how the potential (35) depends on ζ and hij , A = A0 − R ≡ A0 (1 + α).

(44)

Here the spatial Ricci scalar is, R=

e−2ζ 2 ζ − (D − 2)(D − 3)ζ ,k [R − 2(D − 2)∇ g k ζ, ], a2

(45)

2 ≡ ∂i gij ∂j is the = O(h2 ) is the Ricci scalar formed from gij and ∇ where R conformal scalar Laplacian. The full scalar Laplacian is, 1 √ ∇2 ≡ √ ∂i [ gg ij ∂j ]. g

(46)

At this stage one can recognize that the dimensionless three-curvature perturbation R is just ζ, in D = 4 dimensions and to linearized order,50

D−2 1 R(t, x) ≡ − 2 R = (47) ζ(t, x) + O(ζ 2 , ζh, h2 ). 4∇ 2 The kinetic energy (36) can be expressed as, k) − N k k,k ] − E k E B ≡ A0 (1 + β) = A0 + 2(D − 2)c−1 H[(D − 1)(c−1 ζ˙ − ζ,k N

2

˙ ζ˙ ζ k k + (N k N k )2 . + (D − 2)(D − 1) − ζ,k N − 2(D − 2) − ζ,k N ,k ,k c c (48) j and E i ≡ gij N ij ≡ 1 [N i ≡ N i, N i;j + N j;i − c−1 h˙ ij ]. Here we deﬁne N 2 The next step is to expand the volume part of the constrained Lagrangian in powers of α and β, √ c4 g √ c4 aD−1 e(D−1)ζ − AB = − A0 (1 + α)(1 + β) (49) 8πG 8πG c4 aD−1 e(D−1)ζ (α + β) (α − β)2 =− A0 1 + − + ··· . (50) 8πG 2 8 As Weinberg noted, the terms involving no derivatives of ζ or hij sum up to a total derivative,21 aD−1 A0 =

∂ [(D − 2)HaD−1 ]. ∂t

(51)

May 4, 2017 9:53 b2167-ch21

Perturbative quantum gravity comes of age

II-363

i and ζ can be elimiAnother important fact is that quadratic mixing between N 71 nated with the covariant ﬁeld redeﬁnition, −2ζ k + gk ∂ 1 ce 2 ζ − (c−1 ζ˙ − ζ,i N i) . Sk ≡ N ∇ (52) 2 Ha2 ∇ After much work the quadratic Lagrangians emerge,

D−3+ c4 aD−1 LS 2 = ∂ Sk ∂ Sk + ∂ Sk ∂k S , 32πG D−1− (D − 2)c4 aD−1 ζ˙2 ∂k ζ∂k ζ Lζ 2 = − , 16πG c2 a2 Lh2

c4 aD−1 = 64πG

h˙ ij h˙ ij ∂k hij ∂k hij − c2 a2

(53)

(54)

.

(55)

These results suﬃce for the analysis of this section. To consider loop corrections (or non-Gaussiantity) one must solve the constraint equation for Si ,

D−3+ i ∂j ∂j S + (56) ∂i ∂j Sj = O(ζ 2 , ζh, h2 , Sζ). D−1− That is a tedious business which has only been carried out to a few orders. I will review what is known in Sec. 4.1.

3.3. Tree order power spectra As we will see in Sec. 4.3, there is not yet general agreement on how to deﬁne the primordial power spectra when loop corrections are included.23,24 At tree order we can dispense with dimensional regularization, and also forget about the distinction between ζ and the dimensionless three-curvature perturbation (47). The following deﬁnitions suﬃce: k3 (57) d3 x e−ik·x Ω|ζ(t, x)ζ(t, 0)|Ω , ∆2R (k, t) ≡ 2π 2 k3 ∆2h (k, t) ≡ (58) d3 x e−ik·x Ω|hij (t, x)hij (t, 0)|Ω . 2π 2 Although it is useful to retain the time dependence in expressions (57), (58), the actual predictions of primordial inﬂation are obtained by evaluating the timedependent power spectra safely between the ﬁrst and second horizon crossing times tk and Tk described in Sec. 2.2, ∆2R (k) ≡ ∆2R (k, t)|tk tTk ,

∆2h (k) ≡ ∆2h (k, t)|tk tTk .

(59)

May 4, 2017 9:53 b2167-ch21

R. P. Woodard

II-364

The state | Ω in expressions (57), (58) is annihilated by α(k) and β(k, λ) in the free ﬁeld expansions of ζ and hij , d3 k ζ(t, x) = {v(t, k)eik·x α(k) + v ∗ (t, k)e−ik·x α† (k)}, (60) (2π)3 d3 k hij (t, x) = {u(t, k)eik·x ij (k, λ)β(k, λ) + Conjugate}. (61) (2π)3 λ=±

The polarization tensors ij (k, λ) are the same as those of ﬂat space. If we adopt the usual normalizations for the creation and annihilation operators, [α(k), α† (k )] = (2π)3 δ 3 (k − k ),

[β(k, λ), α† (k , λ )] = δλλ (2π)3 δ 3 (k − k ) (62)

then canonical quantization of the free Lagrangians (54), (55) implies that the mode functions obey, ˙ c2 k 2 i4πG v¨ + 3H + v˙ + 2 v = 0, v v˙ ∗ − vv ˙ ∗= 2 3 , (63) a c a u ¨ + 3H u˙ +

c2 k 2 u = 0, a2

uu˙ ∗ − uu ˙ ∗=

i32πG . c2 a 3

(64)

It has long been known that the graviton mode function u(t, k) obeys the same Eq. (23) as that of a massless, minimally coupled scalar.36,37 Only their normalizations diﬀer by the square root of 32πG/c2 . By substituting the free ﬁeld expansions (60), (61) into the deﬁnitions (57), (58) of the power spectra, and then making use of the canonical commutation relations (62), one can express the tree order power spectra in terms of scalar and tensor mode functions, ∆2R (k, t) =

k 3 |v(t, k)| , 2π 2

∆2h (k, t) =

k 3 |u(t, k)|2 k 3 |u(k, t)|2 ij ∗ij = . 2 2π π2

(65) (66)

λ=±

One of the frustrating things about primordial inﬂation is that we do not know what a(t) is, so we need results which are valid for any reasonable expansion history. This means that even tree order expressions such as (65), (66) can only be evaluated approximately because there are no simple expressions for the mode functions for a general scale factor a(t).72–74 One common approximation is setting (t) to a constant, the reliability of which can be gauged by studying the region (at N ≈ −40) of Fig. 1 at which currently observable perturbations freeze-in. (The necessity of nonconstant (t) later is not relevant for the validity of assuming constant (t) to estimate the amplitude at freeze-in.) For constant (t) both mode functions are proportional to a Hankel

May 4, 2017 9:53 b2167-ch21

Perturbative quantum gravity comes of age

II-365

function of the ﬁrst kind, u(t, k) ˙ = 0 ⇒ v(t, k) = √ , 8

8π 2 G ck (1) u(t, k) = H , ν (1 − )c2 H(t)a3 (t) (1 − )H(t)a(t)

(67) ν=

1 2

3− . 1−

(68)

Between ﬁrst and second horizon crossing (tk t Tk ) we can take the small argument limit of the Hankel function, (t) ˙ = 0,

tk t T k

⇒ u(t, k) ≈ =

ν 8π 2 G −iΓ(ν) 2(1 − )H(t)a(t) × (1 − )c2 H(t)a3 (t) π ck

1 16π 2 G −iΓ(ν) 2(1 − )H(t)a (t) 1− × . c5 k 3 π (ck)

(69)

(70)

Constant (t) also implies H(t)a (t) is a constant, which we may as well evaluate at the time of ﬁrst horizon crossing, H(t)a (t) = H 1− (tk )(ck) . With the doubling √ formula (Γ(2x) = 22x−1 / π × Γ(x)Γ(x + 12 )) we at length obtain, 1 2 1 − 1− Γ( 1− ) 16πGH 2 (tk ) (t) ˙ = 0, tk t Tk ⇒ u(t, k) ≈ −i × . 1 2 c5 k 3 Γ( 1− ) (71) The factor multiplying the square root has nearly unit modulus for small — and the BICEP2 result is i = 0.013,27 while previous data sets give the even smaller bound of i < 0.007 at 95% conﬁdence.38–40 Hence it should be reliable to drop this factor, resulting in the approximate forms, ∆2R (k) ≈

GH 2 (tk ) , πc5 (tk )

∆2h (k) ≈

16GH 2 (tk ) . πc5

(72)

The Wentzel–Kramers–Brillouin (WKB) approximation is another common technique for estimating the freeze-in amplitudes of v(t, k) and u(t, k) which appear in expressions (65), (66) for the tree order power spectra. Recall that the method applies to diﬀerential equations of the form f¨ + ω 2 (t)f = 0. From expression (64) one can see that reaching this form for the tensor mode function requires, the rescal3 ing f (t, k) = a 2 (t) × u(t, k). It is then simple to recognize the correctly normalized WKB solution and its associated frequency as, t 16πG exp −i u(t, k) ≈ dt ω(t , k) , (73) c2 a3 (t)|ω(t, k)| t2 9 3 c2 k 2 ω (t, k) ≡ 2 − − (t) H 2 (t). a (t) 4 2 2

(74)

May 4, 2017 9:53 b2167-ch21

II-366

R. P. Woodard

In the sub-horizon regime of ck H(t)a(t) the frequency is real ω(t, k) ≈ ck/a(t) and the solution (73) both oscillates and fall oﬀ like 1/a(t). Freeze-in occurs in the super-horizon regime of ck H(t)a(t) during which the frequency is imaginary ω(t, k) ≈ iH(t)[ 32 − 12 (t)]. We can estimate the freeze-in amplitude by computing the real part of the exponent, 3 t t 1 ˙ ) 2 (t)H 2 (t) 3 a(t ˙ ) 1 H(t a dt ω(t , k) ≈ dt −i + . (75) = ln 1 3 2 a(t ) 2 H(t ) a 2 (tk )H 2 (tk ) tk tk Substituting (75) in (73) and using H(tk )a(tk ) = ck gives the following result for the freeze-in modulus, 16πGH 2 (t ) k . tk t Tk ⇒ |u(t, k)|WKB ≈ (76) 5 3 3 1 c k − (t) 2 2 If we drop the order one factor of | 32 − 12 (t)| the result is (72). From expression (63) we see that reaching the WKB form for the scalar mode 1 3 function requires the rescaling f (t, k) = 2 (t) × a 2 (t) × v(t, k). Its frequency is simpler to express if I ﬁrst introduce the (Hubble form of the) second slow roll parameter η(t), (t) ˙ . (77) 2H(t)(t) The correctly normalized WKB approximation for the scalar mode function and its frequency is, t 2πG v(t, k) ≈ exp −i dt ν(t , k) , (78) c2 (t)a3 (t)|ν(t, k)| t2

3 3 c2 k 2 ˙ − η˙ ν 2 (t, k) ≡ 2 − + (t) − η(t) − η(t) + (79) H 2 (t). a (t) 2 2 H(t) η(t) ≡ (t) −

Freeze-in occurs as one evolves from the sub-horizon regime of ν(t, k) ≈ ck/a(t) to the super-horizon regime of ν(t, k) ≈ iH(t)[ 32 + 12 (t) − η(t)]. The real part of the exponent of (78) is, t t ˙ ) 1 (t 3 a(t ˙ ) 1 H(t ˙ ) dt ν(t , k) ≈ dt −i + + 2 a(t ) 2 H(t ) 2 (t ) tk tk 1 1 3 a 2 (t)H 2 (t) 2 (t) = ln . (80) 1 1 3 a 2 (tk )H 2 (tk ) 2 (tk ) Substituting in (78) gives a freeze-in modulus which is again roughly consistent with (72), 2πGH 2 (tk ) . tk t Tk ⇒ |v(t, k)|WKB ≈ (81) 3 1 5 3 c k (tk ) + (t) − η(t) 2 2

May 4, 2017 9:53 b2167-ch21

Perturbative quantum gravity comes of age

II-367

It should be obvious from the discordant factors of order one in expressions (71), (76) and (81), that the results (72) for the tree order power spectra are only approximate. This is conﬁrmed by numerical integration of explicit models.75,76 In addition to order one factors, depending on the geometry at t = tk there is also a nonlocal “memory factor” which depends on the precise manner in which the mode evolves up to ﬁrst horizon crossing.74 Most of the ambiguity derives from not having a deﬁnitive model for what caused inﬂation. Once the expansion history is known it is possible to derive wonderfully accurate results by numerically integrating the mode functions.77 It is even more eﬃcient to numerically evolve the time-dependent power spectra directly, without the irrelevant phase information.78 The tree order power spectra (59) give the tensor-to-scalar ratio r(k), the scalar spectral index ns (k), and the tensor spectral index nt (k), r(k) ≡

∆2h (k) ≈ 16(tk ), ∆2R (k)

ns (k) ≡ 1 + nt (k) ≡

∂ ln(∆2R (k)) ≈ 1 − 4(tk ) + 2η(tk ), ∂ ln(k)

∂ ln(∆2h (k)) ≈ −2(tk ). ∂ ln(∆2h (k))

(82) (83) (84)

In each case the deﬁnition is exact, and the approximate result derives from expressions (72) with an additional approximation to relate dk to dtk , cdk = [1 − (tk )]H 2 (tk )a(tk )dtk ≈ ck × H(tk )dtk .

(85)

Comparison of (82) and (84) implies an important test on single-scalar inﬂation which is violated in more general models,79–81 r ≈ −8nt .

(86)

Certain general trends are also evident from the approximate results (72): • r < 1 because 1; • nt < 0 because H(t) decreases; and • ns − 1 < nt because (t) tends to increase. To anyone who works in quantum gravity it is breath-taking that we have any data, so it seems petulant to complain that the some of the