String Theory: From Gauge Interactions to Cosmology

NATO Science Series A Series presenting the results of scientific meetings supported under the NATO Science Programme. The Series is published by IOS Press, Amsterdam, and Springer (formerly Kluwer Academic Publishers) in conjunction with the NATO Public Diplomacy Division

Sub-Series I. II. III. IV.

Life and Behavioural Sciences Mathematics, Physics and Chemistry Computer and Systems Science Earth and Environmental Sciences

IOS Press Springer (formerly Kluwer Academic Publishers) IOS Press Springer (formerly Kluwer Academic Publishers)

The NATO Science Series continues the series of books published formerly as the NATO ASI Series. The NATO Science Programme offers support for collaboration in civil science between scientists of countries of the Euro-Atlantic Partnership Council. The types of scientific meeting generally supported are “Advanced Study Institutes” and “Advanced Research Workshops”, and the NATO Science Series collects together the results of these meetings. The meetings are co-organized by scientists from NATO countries and scientists from NATO’s Partner countries –countries of the CIS and Central and Eastern Europe. Advanced Study Institutes are high-level tutorial courses offering in-depth study of latest advances in a field. Advanced Research Workshops are expert meetings aimed at critical assessment of a field, and identification of directions for future action. As a consequence of the restructuring of the NATO Science Programme in 1999, the NATO Science Series was re-organized to the four sub-series noted above. Please consult the following web sites for information on previous volumes published in the Series. http://www.nato.int/science http://www.springer.com http://www.iospress.nl

Series II: Mathematics, Physics and Chemistry – Vol. 208

String Theory: From Gauge Interactions to Cosmology edited by

Laurent Baulieu Université Pierre et Marie Curie, Paris, France

Jan de Boer University of Amsterdam, The Netherlands

Boris Pioline Université Pierre et Marie Curie, Paris, France and

Eliezer Rabinovici Hebrew University, Jerusalem, Israel

Published in cooperation with NATO Public Diplomacy Division

Proceedings of the NATO Advanced Study Institute on String Theory: From Gauge Interactions to Cosmology Cargèse, France 7-19 June 2004 A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN-10 ISBN-13 ISBN-10 ISBN-13 ISBN-10 ISBN-13

1-4020-3732-5 (PB) 978-1-4020-3732-0 (PB) 1-4020-3731-7 (HB) 978-1-4020-3731-3 (HB) 1-4020-3733-3 (e-book) 978-1-4020-3733-7 (e-book)

Published by Springer, P.O. Box 17, 3300 AA Dordrecht, The Netherlands.

www.springer.com

Printed on acid-free paper

All Rights Reserved © 2006 Springer No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Printed in the Netherlands.

Table of Contents

Preface

xi

I

1

Lectures

STRINGS IN A LANDSCAPE by TOM BANKS 1 Introduction . . . . . . . . . . . 2 Cosmology . . . . . . . . . . . 3 Ghost condensation . . . . . . . 4 Strings in a landscape . . . . . 5 Phenomenology of the landscape

3 . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

QUANTIZATION OF HOLOMORPHIC FORMS AND N = 1 SUPERSYMMETRY ON SPECIAL MANIFOLDS by LAURENT BAULIEU 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 N = 1 supersymmetry and the holomorphic BF theory . . . . 3 Four dimensions: K¨ahler manifold . . . . . . . . . . . . . . . 4 Six dimensions: Calabi–Yau three–fold . . . . . . . . . . . . 5 Eight dimensions . . . . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

3 4 4 5 13

19 . . . . .

19 20 22 27 31

LOEWNER CHAINS by MICHEL BAUER AND DENIS BERNARD 1 Critical interfaces and stochastic processes . . . . . . . . . . . . . . . . . 2 Loewner chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

41

THEORY OF COSMOLOGICAL PERTURBATIONS AND APPLICATIONS TO SUPERSTRING COSMOLOGY by ROBERT H. BRANDENBERGER 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 Overview of Inﬂationary Cosmology . . . . . . . . . . . 3 Newtonian Theory of Cosmological Perturbations . . . . 4 Relativistic Theory of Cosmological Fluctuations . . . . 5 Quantum Theory of Cosmological Fluctuations . . . . . 6 Conceptual Problems of Inﬂationary Cosmology . . . . . 7 The Trans-Planckian Window for Superstring Cosmology

v

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

44 61

79 . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. 79 . 81 . 86 . 92 . 99 . 104 . 106

vi 8

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

TOWARDS A TOPOLOGICAL G2 STRING by JAN DE BOER, ASAD NAQVI AND ASSAF SHOMER 1 Introduction . . . . . . . . . . . . . . . . . . . . . 2 G2 sigma models . . . . . . . . . . . . . . . . . . 3 Tri-Critical Ising Model . . . . . . . . . . . . . . . 4 Topological Twist . . . . . . . . . . . . . . . . . . 5 The BRST Operator . . . . . . . . . . . . . . . . . 6 Topological G2 Strings . . . . . . . . . . . . . . . 7 Conclusions . . . . . . . . . . . . . . . . . . . . .

117 . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

CHALLENGES OF MATRIX MODELS by ALEXEI MOROZOV 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Classical period: introduction of models . . . . . . . . . . . . . . . . . 3 First stringy period: generalizations and hunt for structures . . . . . . . 4 Transcendental period: absolutization of structures . . . . . . . . . . . 5 Towards exhaustive theory of 1-matrix model: the Dijkgraaf-Vafa theory 6 Neoclassical period: back to concrete results and back to physics? . . . PHENOMENOLOGY OF NEUTRINO OSCILLATIONS by LEV B. OKUN 1 A short history of neutrinos [1-16] . . . . . . . . . . . 2 Simple formulas for oscillations . . . . . . . . . . . . 3 “Same energy” vs “same momentum” [12-14], [17-25] 4 Solar neutrinos [26] . . . . . . . . . . . . . . . . . . . 5 Reactor antineutrinos [27-30] . . . . . . . . . . . . . . 6 Atmospheric neutrinos [26] . . . . . . . . . . . . . . . 7 Accelerator neutrinos [27] . . . . . . . . . . . . . . . 8 Patterns of mixing angles and masses [25,31] . . . . . 9 Double beta decay [32] . . . . . . . . . . . . . . . . . 10 Cosmological and stellar inputs [33] . . . . . . . . . . 11 Challenges and Prospects [31,34] . . . . . . . . . . . . CLOSED STRINGS IN MISNER SPACE: A TOY MODEL FOR A BIG BOUNCE ? by BRUNO DURIN AND BORIS PIOLINE 1 Semi-classics of closed strings in Misner space 2 One-loop vacuum amplitude . . . . . . . . . . 3 Tree-level scattering amplitudes . . . . . . . . 4 Discussion . . . . . . . . . . . . . . . . . . . .

. . . . . . .

117 118 119 121 122 124 125 129

. . . . . .

129 130 131 135 143 155 163

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

163 164 166 167 169 170 171 171 172 172 173

177 . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

178 187 191 197

vii PHENOMENOLOGICAL GUIDE TO PHYSICS BEYOND THE STANDARD MODEL by STEFAN POKORSKI 1 Introduction . . . . . . . . . . . . . . . . . . . . . 2 The Standard Model . . . . . . . . . . . . . . . . 3 Hints from the Standard Model for its extensions . 4 Supersymmetric extensions of the Standard Model 5 Summary . . . . . . . . . . . . . . . . . . . . . . INTRODUCTION TO COSMIC F- AND D-STRINGS by JOSEPH POLCHINSKI 1 Introduction . . . . . . . . . . . . . . . . . . . . 2 Cosmic string review . . . . . . . . . . . . . . . 3 Cosmic strings from string theory . . . . . . . . 4 Production of cosmic F- and D-strings . . . . . . 5 Stability of strings . . . . . . . . . . . . . . . . . 6 Seeing Cosmic Strings . . . . . . . . . . . . . . 7 Distinguishing Superstrings . . . . . . . . . . . . 8 Conclusions . . . . . . . . . . . . . . . . . . . .

201 . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

201 203 211 221 227 229

. . . . . . . .

. . . . . . . .

LONG TIME SCALES AND ETERNAL BLACK HOLES ´ AND ELIEZER RABINOVICI by JOSE L. F. BARBON 1 Introduction . . . . . . . . . . . . . . . . . . . . . . 2 Long-time details of thermal quasi-equilibrium . . . 3 Topological diversity and unitarity . . . . . . . . . . 4 Conclusions . . . . . . . . . . . . . . . . . . . . . . SEMICLASSICAL STRINGS AND ADS/CFT by ARKADY A. TSEYTLIN 1 Introduction . . . . . . . . . . . . . . . . . . . . . . 2 Effective actions for coherent states . . . . . . . . . 3 General fast motion in S 5 and SO(6) scalar operators 4 Concluding remarks . . . . . . . . . . . . . . . . . .

II

. . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

229 230 236 239 241 245 248 250 255

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

255 257 260 262 265

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

SEMINARS

265 272 280 284

291

BARYOGENESIS WITH LARGE EXTRA DIMENSIONS by KARIM BENAKLI

293

BRANEWORLD COSMOLOGY ALMOST WITHOUT BRANES by GIANLUCA CALCAGNI 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Slow-roll parameters . . . . . . . . . . . . . . . . . . . . . . 4 Cosmological perturbations: theory and observations . . . . . 5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . .

297 . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

297 297 299 300 302

viii CHAOTIC CASCADES FOR D-BRANES ON SINGULARITIES ´ FRANCO ET AL. by SEBASTIAN 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Cascades in coupling space . . . . . . . . . . . . . . . . . . 3 Duality Walls and Fractals . . . . . . . . . . . . . . . . . . 4 Supergravity Duals . . . . . . . . . . . . . . . . . . . . . . 5 Recent developments . . . . . . . . . . . . . . . . . . . . . N = 1 GAUGE THEORY EFFECTIVE SUPERPOTENTIALS by BEN M. GRIPAIOS 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2 The Gauge Theory/Matrix Model Correspondence . . . . . 3 Pure Gauge Terms . . . . . . . . . . . . . . . . . . . . . . 4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .

305 . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

305 305 306 307 308 311

. . . .

GROMOV-WITTEN THEORY AND AUTOMORPHIC FORMS ¨ by DANIEL GRUNBERG 1 Recap TFT . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Correlation Functions . . . . . . . . . . . . . . . . . . . . . 3 BPS Invariants . . . . . . . . . . . . . . . . . . . . . . . . 4 GW Potentials . . . . . . . . . . . . . . . . . . . . . . . . . 5 KdV Hierarchy . . . . . . . . . . . . . . . . . . . . . . . . 6 Automorphic Products . . . . . . . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

311 312 314 316 319

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

319 320 321 322 322 323

BLACK HOLE INSTABILITIES AND PHASE TRANSITIONS 325 by SEAN HARTNOLL 1 Generalised AdS black holes . . . . . . . . . . . . . . . . . . . . . . . . 325 2 Dual phase transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326 PROBING SINGULARITIES by VERONIKA E. HUBENY 1 Introduction . . . . . . . 2 Black hole singularity . . 3 Cosmological singularity 4 Discussion . . . . . . . .

329 . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

329 330 332 333

INFLATION UNLOADED 337 by MATTHEW KLEBAN 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337 2 Quantiﬁcation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338 3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340 D-BRANE EFFECTIVE ACTIONS by PAUL KOERBER 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Derivative corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

341 341 341 342

ix ADS2 × S 2 AS AN EXACT HETEROTIC STRING BACKGROUND by DOMENICO ORLANDO 1 Intro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 SU (2) asymmetric deformation . . . . . . . . . . . . . . . . . 3 SL (2, R) deformation . . . . . . . . . . . . . . . . . . . . . . 4 AdS2 × S 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . STRINGS AND D-BRANES IN HOLOGRAPHIC BACKGROUNDS ¨ AND ARI PAKMAN by DAN ISRAEL 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 SL(2,R)/U(1) and N=2 Liouville . . . . . . . . . . . . . . . . . 3 D-branes from boundary AdS3 . . . . . . . . . . . . . . . . . . 4 strings near NS5 branes . . . . . . . . . . . . . . . . . . . . . . COSMIC CENSORSHIP IN ADS/CFT by MUKUND RANGAMANI 1 Introduction . . . . . . . . . . . 2 The Framework . . . . . . . . . 3 The Size of the singularity . . . 4 Discussion . . . . . . . . . . . .

345 . . . .

. . . .

. . . .

. . . .

. . . .

345 345 347 348 351

. . . .

. . . .

. . . .

. . . .

. . . .

351 352 353 353 355

. . . .

. . . .

. . . .

TOPICS IN BLACK HOLE PRODUCTION by VYACHESLAV S. RYCHKOV 1 Introduction . . . . . . . . . . . . . . 2 Wavepackets and Semiclassics . . . . 3 Multiple Black Hole Production . . . 4 Conclusions . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

355 356 358 360 363

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

TOWARDS THE EXACT DILATATION OPERATOR OF N SUPER YANG-MILLS THEORY by ANTON V. RYZHOV 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 2 Constraints on D(1) . . . . . . . . . . . . . . . . . . . 3 Summing D(1) to all orders . . . . . . . . . . . . . . . 4 Concluding remarks . . . . . . . . . . . . . . . . . . . MODULI TRAPPING AND STRING GAS COSMOLOGY by SCOTT WATSON 1 Initial Conditions . . . . . . . . . . . . . . . . . . . . 2 Strings at Finite Temperature . . . . . . . . . . . . . . 3 Moduli Trapping . . . . . . . . . . . . . . . . . . . . 4 Conclusions . . . . . . . . . . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

363 363 365 366

=4 371 . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

371 372 374 376 379

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

379 380 381 382

STRINGY GEOMETRY: GENERALIZED COMPLEX STRUCTURE 385 by MAXIM ZABZINE 1 Mathematical preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . 385

x 2 3

Topological model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388 Topological model with WZ term . . . . . . . . . . . . . . . . . . . . . . 390

INTEGRABILITY OF SUPERCONFORMAL FIELD THEORY AND SUSY N=1 KDV 393 by ANTON M. ZEITLIN 1 RTT-relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393 2 The Q-operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394 3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395

APPENDIX

397

LIST OF SPEAKERS

399

LIST OF PARTICIPANTS

401

Preface The present volume is a collection of the lecture and seminar notes delivered at the Summer School “ String Theory : from Gauge Interactions to Cosmology”, which was held in Carg`ese, France, from June 7 to 19, 2004. The main focus of the school was on the inter-relations between the ﬁelds of superstring theory, cosmology and particle physics, which have been the subject of very active research recently. Much of the recent progress on topical problems in the ﬁeld was covered, including the duality between gravity and gauge theories, strings and cosmology, critical phenomena in statistical mechanics, topological string theory, physics beyond the standard model, and the landscape of vacua of string theory. The School was an excellent opportunity for the youngest researchers to establish close relationship with the lecturers and among each other, be it during the vibrant Gong Show session, under the shade of the Wisdom Tree or gazing at the beautiful Corsican coastline. We hope that these proceedings will further serve in ﬁxing the acquired knowledge, and hopefully become a valuable reference for anyone working in this fascinating domain of physics. It is a pleasure to extend our warm thanks to the NATO Division for Scientiﬁc Affairs, under the Advanced Study Institute grant PST.ASI 97 85 17, and to the the Human Factor-Mobility & Marie Curie Actions Division of the European Commission, under the High-Level Scientiﬁc Conference grant HPCFCT-2001-00298, whose generous funding have allowed this meeting to take place. NATO’s announced discontinuation of its support will be sorely felt in forthcoming events. We also wish to thank all the people who have contributed to the organization of this Conference, especially the staff of the Institut d’Etudes Scientiﬁques de Carg`ese its Director, Elisabeth Dubois-Violette. We are also much indebted to Josette Durin, for her invaluable help in preparing this volume. Finally, we are very grateful to all the participants of the School for a creating a wonderful stimulating atmosphere, and to the contributors of this volume for their dedication in preparing their notes.

Laurent Baulieu Jan de Boer Boris Pioline Eliezer Rabinovici

xi

Part I

Lectures

STRINGS IN A LANDSCAPE

TOM BANKS Santa Cruz Institute for Particle Physics, Santa Cruz CA 95064 NHETC, Rutgers U., Piscataway, NJ 08854

Abstract. This is a summary of some of the topics covered at the Cargese Summer School on String Theory in June of 2004. Most space is devoted to a discussion of ideas about the String Landscape, which formed the background for some of the talks and much of the discussion at the School.

1. Introduction This paper is a synthesis of my summary remarks at the Cargese School, and my contribution to the wisdom tree discussion with Arkani-Hamed and Polchinski. The topics of the school were varied. There was a discussion of the interface between cosmology and Planck/String scale physics, in the talks of Brandenberger, Polchinski, Pioline, Elitzur, and Rabinovici. Polchinski and Arkani-Hamed devoted some of their time to discussion of the string landscape and expectations for ﬁne tuning. Arkani-Hamed discussed his beautiful model of spontaneous breakdown of general coordinate invariance (also known as ghost condensation), as well as the split supersymmetry model motivated by the string theory landscape. Polchinski also talked about the emergence of cosmologically stable cosmic strings in one of the regions of moduli space that are being explored in the landscape. Tseytlin’s talks were devoted to the string/gauge theory connection, and in particular to the emergence of integrable structures on both sides of the duality between maximally supersymmetric planar Yang-Mills theory and tree level string theory on AdS5 ×S 5 . Dijkgraaf and Ooguri discussed the emergent non-perturbative formulation of topological string theory. Morozov’s lectures reviewed some of the older connections between exactly soluble string theories, and matrix models. Okun and Pokorski gave us detailed updates on neutrino and collider phenomenology. Finally, D. Bernard reminded us that conformal ﬁeld theory still has interesting things to say about real problems in condensed matter physics. I will conﬁne these paragraphs to brief summaries of cosmology, ghost condensation, and a somewhat more detailed discussion of the debate over the landscape. I begin with cosmology. 3 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 3–17. © 2005 Springer. Printed in the Netherlands.

4

TOM BANKS

2. Cosmology Robert Brandberger gave us a review of the calculation of ﬂuctuations in inﬂationary cosmologies, with emphasis on the possibility of seeing Transplanckian effects in the cosmic microwave background. I will give you my understanding of this issue, which may not coincide with Robert’s. The usual argument for the universality and independence of microscopic physics of the inﬂationary ﬂuctuation spectrum relies on the adiabatic theorem. Given the assumption that the adiabatic theorem applies, effective ﬁeld theory arguments show that the corrections to the leading order predictions for ﬂuctuations are of order (H/MP )2 , which usually means they are unobservable. Attempts to produce effects of order H/MP rely on non-adiabatic initial conditions. Brandenberger showed us models of initial states which do produce such effects, as well as models which don’t. String theory does not yet have the tools to express results of stringy cosmologies in terms of the language of ﬁeld theory in which these results are presented. It is not clear which class of models string theory prefers. What is clear is that models with many e-folds of inﬂation, which incorporate violations of the adiabatic theorem (and avoid making the UV more singular than ﬁeld theory), must be very ﬁne tuned in order to make an observational effect. Such models make large modiﬁcations to conventional predictions, only over a limited range of scales. It requires ﬁne tuning to make this range of scales coincide with the range we observe in the CMB. Elitzur, Pioline and Rabinovici talked about models of string cosmology which use the machinery of world-sheet conformal ﬁeld theory to investigate singular cosmologies. These models fall into two types: those where an inﬁnite universe undergoes a Big Crunch, which may or may not be followed by a Big Bang (with string theory providing the tool to remove and glue together the singularities) and those with closed universes which start with a Big Bang and end with a Big Crunch. Both types of models are generalized orbifold CFTs, i.e. they are gauged σ models, where the gauge group may be continuous or discrete. In the Crunch-Bang models the correlation functions of string vertex operators have the conventional interpretation of a scattering theory. It has been shown pretty deﬁnitively that conventional string perturbation theory breaks down near the singularity. Hopes for progress depend on the notion that the production of string winding modes, which are concentrated near the singularity will smooth out the transition without recourse to Planck scale physics. It is hoped that some modiﬁed perturbation expansion will give a clear picture of what is going on near the singularity. In Bang-Crunch models there are no asymptotic regions in which to deﬁne scattering amplitudes. Mathematical computations rely on inﬁnite space-like regions with closed time-like curves, which have been dubbed “whiskers” (actually the Crunch-Bang orbifolds have similar whiskers). I do not understand the physical interpretation of the world sheet correlators in this context, or how they are related to physics in the actual cosmological part of the space-time. 3. Ghost condensation Arkani-Hamed discussed the theory[1] of ghost condensation, which is amusing and beautiful. He and his collaborators have shown that theories which do not have a stable Lorentz 3 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 3–17. © 2005 Springer. Printed in the Netherlands.

STRINGS IN A LANDSCAPE

5

invariant solution can nonetheless have stable solutions which have the symmetries of hom*ogeneous isotropic cosmology. Moreover, because these Lagrangians describe a partial Higgsing of gravity, their low energy behavior is almost universal. It is described by a single new massive parameter f , analogous to the pion decay constant. The theory has a new ﬁeld φ, whose low energy excitations have a non-Lorentz invariant dispersion relation. They can modify known theories in ways which are not yet ruled out, but are potentially observable. In particular, they can produce interesting non-Gaussian ﬂuctuations in the Cosmic Microwave Background. The theory can also account for the acceleration of the universe, without a cosmological constant, although it does not solve the cosmological constant problem. In order to explain the acceleration, f must be taken very small. The new ﬁeld in the theory couples only to gravity, and for small f its effects on ordinary physics are almost negligible. There are two problems with this model. The ﬁrst is that we cannot derive it from string theory in any known way. The second is that the assumption that the new physics couples to the old only through gravity is artiﬁcial. Standard effective ﬁeld theory reasoning leads us to expect non-renormalizable couplings of φ to standard model ﬁelds, scaled by f . For small values of f this is inconsistent with experiment. It appears to be technically natural to omit the direct couplings of the Goldstone ﬁeld to matter. If we postulate that it couples only to gravity, then the small value of f /MP ensures that radiative corrections do not induce large direct couplings of the Goldstone ﬁeld to the standard model. Nonetheless, one would like to have a principle that explained the absence of direct couplings to the standard model, rather than a simple postulate, however self-consistent. Neither of these criticisms is grounds for abandoning this very interesting new model. In particular, string theory as we know it is built to handle asymptotically ﬂat and AdS backgrounds, and has notorious problems with any kind of real cosmology. Ghost condensation is an intrinsically cosmological model. It does not have any stable, Lorentz invariant background solutions. So, the inability to ﬁnd it among solutions to the effective ﬁeld equations of a known string theory, may be more a problem of the limitations of the present formulation of string theory, than an indication that ghost condensation is an incorrect idea. The wonderful thing about ghost condensation is that it makes predictions for novel phenomena, which are just beyond the current reach of experiment. It should be studied and tested to the best of our ability. 4. Strings in a landscape Although there were no talks devoted explicitly to the string landscape at Cargese, it pervaded many of the discussions. Arkani-Hamed’s split supersymmetry model was motivated by the landscape, as was Polchinski’s discussion of cosmic strings in string theory. And our discussion under the wisdom tree was a debate about the landscape. Part of the confusion in the debate about the landscape is the conﬂation of two notions of effective action which occur in quantum ﬁeld theory. The ﬁrst is the 1PI effective action, which is an exact summary of the entire content of a ﬁeld theory. Knowledge of it enables us to construct all of the correlation functions of the theory, in any of its vacuum states. In perturbation theory, we can compute the 1PI action around any vacuum state,

6

TOM BANKS

and get the same result. It is important to realize that there is NO known analog of the 1PI action in string theory, which applies to all solutions of the theory. For example, both AdS5 × S 5 and ten dimensional Minkowski space, are solutions of the tree level effective action of Type IIB SUGRA, but higher order corrections (not to mention non-perturbative corrections) to the action which generates the S-matrix in Minkowski space will not be the same as those which generate the correlation functions on the boundary of AdS space. There is no single quantum effective action which we can use to get both the S-matrix and the AdS correlation functions by ”shifting the ﬁeld”. The other concept of effective action in ﬁeld theory is the low energy or Wilsonian action. This is deﬁned, either in a single vacuum, or in a set of quasi-degenerate vacua whose energy density differences (as well as the heights of the barriers between them) are small compared to some cutoff scale. The Wilsonian action only contains degrees of freedom whose ﬂuctuations are signiﬁcant at these low energy scales. It is important to avoid using it when the conditions of its validity are violated. For example, in minimally SUSic QCD with NF ≤ NC − 1, the low energy degrees of freedom consist of a meson superﬁeld M and one can compute the exact low energy superpotential of M . The low energy Kahler potential is canonical. If one uses this low energy Lagrangian to compute the energy density of states with a given expectation value of M , this Lagrangian gives a divergent answer at the symmetric point of vanishing eigenvalue. This is not the correct physical answer. The true Kahler potential (in the 1PI sense) of M is modiﬁed at the origin of moduli space, and the energy density there is really of order the QCD scale. We should stop paying attention to the predictions of the Wilsonian action when they are outside its range of validity. In string theory, the effective actions we compute are analogous to Wilsonian actions, but their range of validity is even more constrained. In particular, the stringy derivation of the effective action views it as a tool for calculating boundary correlation functions in a ﬁxed asymptotic space-time background. We tend to forget this because, particularly in situations with a lot of SUSY, the leading low energy term in the effective action is independent of the background. This fosters the illusion that different backgrounds can be viewed as vacuum states of the same theory. In fact, as emphasized in [2], the italicized phrase is borrowed from quantum ﬁeld theory, and refers to concepts which depend entirely on the separation between IR and UV physics of that formalism. In string theory/quantum gravity, UV and IR physics are much more intimately entangled, and the concept of different vacuum states of the same underlying string theory Hamiltonian is much more circ*mscribed. When we have a continuous moduli space of super-Poincare invariant S-matrices, we can do experiments at one value of the moduli, which are sensitive to the S-matrix at other values of the moduli1 . Note however, that the only Hamiltonian form we have for such models is in light cone frame, where different values of the moduli correspond to different Hamiltonians. Similarly, a moduli space of correlation functions on the boundary of Anti-deSitter space, corresponds to a one parameter set of different Hamiltonians, rather than different superselection sectors of the same Hamiltonian. The recognition that changes in background asymptotics correspond to changes in the Hamiltonian, rather than changes of superselection sectors for a given Hamiltonian 1 These experiments are much more difﬁcult than they would be in a SUSic quantum ﬁeld theory, without gravity.

STRINGS IN A LANDSCAPE

7

goes back to [3], and has become commonplace with the advent of AdS/CFT. Changes in non-normalizable modes of bulk ﬁelds2 add relevant terms to the Hamiltonian. Changes in the (negative) cosmological constant correspond to changes in the ﬁxed point which deﬁnes the boundary CFT, and thus to a completely different set of high energy degrees of freedom. These facts lead us3 to be suspicious of attempts to ﬁnd new string theory models by patching together an effective potential using the degrees of freedom of e.g. Type II string theory in ﬂat space-time. Indeed, once we include gravitational effects, the low energy action itself gives us reason to be suspicious of meta-stable de Sitter vacua constructed in this way. Under the wisdom tree, Arkani-Hamed gave us an example of a situation where we feel pretty conﬁdent that we can reliably compute a Wilsonian effective potential with minima that have positive vacuum energy. Simply take the supersymmetric standard model, with cubic (A-term) soft supersymmetry breaking terms. We get a host of minima with energy density differences of order the SUSY breaking scale. There is no doubt that a TeV scale experimenter could verify the existence of these minima by exciting localized coherent excitations of the squark and slepton ﬁelds. Let us however assume that we are living in a Poincare invariant state4 , and see what this potential implies about the geometry of space-time when gravitational effects are taken into account. In order to exhibit a meta-stable dS space, we have to excite a region of order the putative Hubble radius into the meta-stable minimum. Old results of Guth and Farhi[4] show that the external observer can never verify the existence of the inﬂating region. A black hole forms around it. Any observer in the asymptotically ﬂat region, who tries to jump into the black hole to ﬁnd inﬂation, ﬁrst encounters a singularity. So effective ﬁeld theory tells us that we can ﬁnd and explore meta-stable positive energy density minima of an effective potential, but not the meta-stable dS spaces that these minima have been thought to imply. That is, the Guth-Farhi results suggest that the stable asymptotically ﬂat vacuum state does not have excitations which correspond to meta-stable dS vacua, even when it has an effective potential with positive energy meta-stable minima. Rather, it has excitations in which ﬁelds are excited into meta-stable minima only over regions small compared to the Hubble radius at those minima. The attempt to create larger regions succeeds only in creating black holes. We see that in gravitational theories, the criterion for the validity of non-gravitational effective ﬁeld theory reasoning depends on more than just the value of the energy density in Planck units. When the Schwarzschild radius of a region exceeds its physical size in the approximation in which gravity is neglected, a black hole forms. Effective ﬁeld theory remains valid outside the black hole horizon (if it is large enough), but not inside. In the above example, no external observer can probe the putative dS region, without ﬁrst encountering a singularity. We also see that the solutions of the same effective equations of motion may not reside in the same quantum theory. 2

Note that if we want to make ﬁnite rather than inﬁnitesimal changes we must restrict our attention to Breitenlohner-Freedman allowed tachyon ﬁelds. 3 well, at least they lead me 4 In other words, ﬁne tune the c.c. to be zero in one preferred vacuum state.

8

TOM BANKS

The results of Coleman and De Lucia[5], on vacuum tunneling in the presence of gravity, give us a sort of converse to this result. Given the same effective Lagrangian we used in the previous paragraph, we can assume the existence of the meta-stable dS space, and ask what it decays into. Here there is a surprise, particularly for those who constantly repeat the mantra “dS space decays into ﬂat space”. In fact, the analytic continuation of the CDL instanton is a negatively curved Friedmann Robertson Walker cosmology, which (if the potential has a stable zero cosmological constant minimum), is asymptotically matter dominated. Although it locally resembles ﬂat space on slices of large cosmological time, its global structure is completely different. In particular, an attempt to set up an asymptotically Minkowskian coordinate system, starting from the local Minkowski frame of some late time observer, inevitably penetrates into regions where the energy density is high (the energy density is constant on slices of constant negative spatial curvature, and is of order the energy density of the false vacuum at early FRW times) . Thus, both analysis of creation and decay of meta-stable dS states, suggests that if a potential has a stable minimum with vanishing cosmological constant, and another with positive energy density, the Minkowski solution and the meta-stable dS solution are simply not part of the same theory. There remains a possibility of the existence of a theory with a stable, matter dominated, FRW cosmological solution with a meta-stable dS excitation. The problem with this is that it is very unlikely that we can make a reliable exploration of this scenario within the realm of low energy effective ﬁeld theory. The CDL instanton solution is non-singular. The a = 0 point of the FRW coordinate system is just a coordinate singularity marking the boundary between the FRW region and a region of the space-time which continues to inﬂate. However, arbitrary hom*ogeneous perturbations of the CDL solution have curvature singularities. Further, there is a large class of localized perturbations which evolve to Big Crunch singularities, rather than passing smoothly through the a = 0 point. For example, consider a localized perturbation on some hyperbolic time slice a ﬁnite proper time prior to a = 0. Let it be hom*ogeneous in a large enough region that signals from its inhom*ogeneous tail cannot propagate to the a = 0 point. Then we will have a singularity. I will mention a more generic example below. If we re-examine the Guth-Farhi argument in the FRW context, we see that it continues to hold until we go back in time to a point where the cosmic energy density is of order the barrier to the meta-stable minimum. At high enough energy density, there are FRW solutions in which the ﬁeld classically evolves into the meta-stable minimum. It will then decay by tunneling, into the FRW continuation of the CDL instanton. Generic FRW solutions (including arbitrary hom*ogeneous perturbations of the CDL instanton) have curvature singularities at a ﬁnite cosmic time in the past. To establish their existence as genuine theories of quantum gravity one must go beyond effective ﬁeld theory, and probably beyond perturbative string theory. Freivogel and Susskind[6] have suggested a scattering theory in which asymptotic states are associated with incoming and outgoing wave perturbations of the nonsingular, time symmetric Lorentzian continuations of the, CDL instantons for the various metastable vacua of string theory. They claim that in this framework, the breakdown of effective ﬁeld theory is avoided, as long as the effective potential is everywhere smaller than Planck scale. I ﬁnd this suggestion interesting, but it is not based on reliable calculations. If one considers black holes with radius larger than the dS radius, formed in the remote past of the FRW part of the time symmetric Lorentzian CDL geometry, it is hard

STRINGS IN A LANDSCAPE

9

to see how these solutions asymptote to the future CDL geometry without encountering a singularity. Near the a = 0 point where the FRW coordinates on the CDL solution become singular, most of the space-time is isometric to dS space at its minimal radius. If we have formed a black hole in the past, with radius much larger than this, then the entire space-time must end in a singularity. Thus, I would claim that unlike asymptotically ﬂat or AdS space-times, we have no reliable effective ﬁeld theory argument that there are an inﬁnite number of states in spacetimes that asymptote to the time symmetric Lorentzian CDL instanton. An inﬁnite number of states is a minimal requirement for the existence of a quantum theory that can make precise mathematical predictions, which can be self consistently measured in the theory. We must understand the nature of the singularities in these perturbations of the CDL instanton geometry before we can conclude that the framework makes sense. I think it is more likely, that meta-stable dS vacua exist only in the context of a Big Bang cosmology. If we consider the problem of accessing a meta-stable dS minimum of an effective potential, at times when the cosmic energy density is of order the barrier height of the potential, then the Guth-Farhi problem does not appear to exist. Starting from a Big Bang singularity, one can ﬁnd hom*ogeneous solutions where a scalar ﬁeld wanders over its potential surface at a time when the energy is higher than the barriers between minima, and then settles in to a meta-stable minimum with positive cosmological constant. One must understand the Big Bang to make a reliable theory of such a situation, but apart from that the solution is non-singular. In particular, the problem of large black holes in the initial state, is not present for this situation. The difﬁculties are all associated with understanding the Big Bang singularity. To summarize, it is clear from semi-classical calculations alone, that the concept of a vacuum state associated with a point in scalar ﬁeld space is not a valid one in theories of quantum gravity. A given low energy effective ﬁeld theory may have different solutions which do not have anything to do with each other in the quantum theory. One solution may be a classical approximation to a well deﬁned quantum theory, while the other is not. It seems likely that the context in which we will have to investigate the existence of non-existence of the landscape is Big Bang cosmology. This is the only situation in which we can reliably construct a universe which gets stuck in a meta-stable dS minimum. Thus, I claim that if string theory really has a multitude of meta-stable dS states, then exact theory into which they ﬁt is a theory of a Big Bang universe which temporarily gets stuck, with some probability, in each of these states. This is ultimately followed by decay to negatively curved FRW universes. These FRW universes have four inﬁnite dimensions and 6 or 7 large compact dimensions, which are expanding to inﬁnity. It is clear that the probability for ﬁnding a particular dS vacuum is partly determined by the density matrix at the Big Bang and not just by counting arguments. In the next subsection I will describe existing proposals for the cosmological distribution of vacua. My main point here is that the nature of the Big Bang will have to be addressed before we can hope to understand the correct statistics of stringy vacua. 4.1. ETERNAL INFLATION The string landscape seems to ﬁt in well with older ideas which go under the name of eternal inﬂation, the self reproducing inﬂationary universe, etc.. The simplest model which

10

TOM BANKS

exhibits this sort of behavior is one with a single scalar ﬁeld with two minima, one with positive vacuum energy and the other with vanishing energy. One considers the expanding branch of the meta-stable dS universe with the ﬁeld in the false minimum. In the quantum ﬁeld theory approximation this seems to produce an ever-expanding region of space and one allows the dS space to decay by CDL bubble formation independently in each horizon volume. In the eternal inﬂation picture, one tries to interpret the result as a single classical space-time. One obtains a Penrose diagram with a future space-like boundary. The future space-like boundary is fractal, with regions corresponding to singularities5 as well as FRW asymptotics, interspersed in a causally disconnected way. In the landscape there will be many different singular regions of the boundary, as well as many different FRW regions. Advocates of the landscape/eternal inﬂation picture then make the analogy to maps of the observable universe, with different causally disconnected regions being the analogs of different planets. Physics it is said, depends on “where you live” and organisms like ourselves can only live in certain regions of the map. A key difference between different minima of the effective potential in eternal inﬂation, and different planets is that we cannot, even in principle, communicate with causally disconnected regions of the universe. How is one to interpret such a picture in terms of conventional quantum mechanics? I believe that the fundamental clue comes from the principle of Black Hole Complementarity[7]. Black holes also present us with two regions of space-time which are causally disconnected. Hawking showed long ago that this was an artifact of the semi-classical approximation, and that black holes return their energy to the external space-time in which they are embedded. If we assume that the region behind the horizon has independent degrees of freedom, commuting with those in the external space, then we are confronted by the information loss paradox. String theorists have believed for some years now, that this is not the case. The principle of Black Hole Complementarity is the statement that the observables behind the horizon do not commute with those in the external space-time. For a large black hole, (and for a long but ﬁnite time as measured by the infalling observer), these two sets of observables are both individually well described by semiclassical approximations, but the two descriptions are not compatible with each other. Fischler and I tried to relate this principle to the Problem of Time[8]. In the semiclassical quantization of gravity one attempts to solve the Wheeler-DeWitt equation HΨ = 0 with an ansatz Ψ = eiS χ(t, φ), where S is the action of some classical space-time background solution, and χ is the wave functional of a quantum ﬁeld theory in this space-time i∂t χ = H(t)χ. 5 The singular regions correspond to decays to parts of the potential where the vacuum energy is negative. They would exist in the landscape context, but not in the simple model we are discussing.

STRINGS IN A LANDSCAPE

11

The (generally time dependent) Hamiltonian H(t) depends both on the choice of classical background, and on the particular time slicing chosen for that background. For example, for the Schwarzschild background we could choose HSch , the Schwarzschild Hamiltonian, or some time dependent H(t) where t is the proper time of a family of in-falling observers. Even ignoring subtle questions of whether these two Hamiltonian evolutions act on the same Hilbert space, it is clear that they are different and that [H(t), HSchw ] = 0 at any time t. It is therefore not surprising that the semi-classical observables of different observers do not commute with each other. Given this description of black holes, it is natural to conjecture that a similar phenomenon occurs for any space-time with horizons. In [8] this was called Cosmological Complementarity for asymptotically dS spaces. E. Verlinde has suggested the name Observer Complementarity for the general case. If we apply this logic to Eternal Inﬂation, we obtain a picture quite different from the original description of these space-times. We simply associate a single Hilbert space and many different (generally time dependent) Hamiltonians to the fractal Penrose diagram. Each Hamiltonian is associated with the causal patch of a given observer. Mathematically, the situation can be equally well described by saying we have a collection of different theories of the universe. There is a philosophical cachet, associated with the phrase, “physics depends on where you live in the multiverse”, which is absent from this alternative way of describing the physics. In the formalism described in [6] this is almost precisely what is conjectured. For each meta-stable dS point L in the landscape, which can decay into the Dine-Seiberg region of moduli space, and for each (typically 10 or 11 dimensional) Super-Poincare invariant solution V of string theory into which L can decay there is a different unitary S-matrix SL,V 6 . It is claimed that each of these S-matrices contains all the physics of the landscape, because there is a canonical way to compute the unitary equivalence U in the † formula SL,V = UL,V,L ,V SL V UL,V,L ,V . The statement that there is a theory of eternal inﬂation in which all of the points L are meta-stable states is really the statement that the theory contains a canonical algorithm for computing the “gauge transformations” U 7 . The authors of [6] claim that all of the meta-stable dS states will show up as resonances in every S-matrix, SV,L . This claim is plausible if the S-matrices are indeed related by unitary conjugation. The spectrum of the S-matrix is then gauge invariant. In ordinary scattering theory, time delays, which are related to resonance lifetimes, are related to the spectrum of the S-matrix. In the eternal inﬂation context, there is no universal notion of time for the different asymptotic states, so more work is necessary to understand these concepts. If indeed the information about each meta-stable dS vacuum can be extracted from the spectral density of a given S-matrix SV,L , and if we can ﬁnd a reliable framework, for deﬁning and calculating these S-matrices, then the landscape will have a mathematical 6

We can also consider initial and ﬁnal states corresponding to different CDL instantons. In[6] these are claimed to be different gauge copies of the same information in the S matrices. I will mention a different interpretation below. 7 For the moment, no approximate statement of what this algorithm is has been proposed. I would conjecture that, if the formalism makes sense, the transition amplitudes between two different FRW spacetimes, mentioned in the previous footnote, provide the algorithm for calculating the U mappings.

12

TOM BANKS

deﬁnition. From a practical point of view however, we just say that string theory gives us an algorithm for constructing models of the world (in the landscape context this means choosing particular stable or meta-stable minima) which is not unique and that we are trying to use data to constrain which model we choose. In the next section, I will describe the situation in the language of the approximate Hamiltonian of a given meta-stable dS observer, rather than that of eternal inﬂation. I have emphasized the problems with this S-matrix point of view and suggested the alternative notion that the landscape could only make sense in the context of Big Bang Cosmology. One must thus understand how to describe the initial states. There are two possibilities, either there is some principle which picks out a ﬁxed initial state8 at the Big Bang, or there is a generalized S-matrix in which we relate a particular state at the Big Bang to a particular linear combination of ﬁnal scattering states in one of the FRW backgrounds deﬁned by a decaying meta-stable dS space. In a manner analogous to Freivogel and Susskind, one would conjecture that the descriptions in terms of different future FRW backgrounds would be unitarily equivalent to each other, by unitary transformations which do not respect locality. Since we are unlikely to have much control over the initial conditions at the Big Bang, one should choose the in-state at the Big Bang to be a high entropy density matrix. Thus, the practical difference between the two proposals is the entropy of the initial state. I know of two proposals for the initial density matrix at the Big Bang, which might lead to a set of cosmological selection rules for meta-stable points in the landscape. The ﬁrst, modular cosmology[9] postulates an early era in which the universe can be described semi-classically, but the potential on moduli space is smaller than the total energy density. The metric on moduli space then provides a ﬁnite volume measure. Furthermore, the motion of the moduli is chaotic. These facts suggest that the probability of ﬁnding the universe in a given meta-stable minimum of the potential is the volume of the basin of attraction of that minimum, divided by the volume of moduli space. Holographic cosmology[10] gives an alternative view of the initial state of the universe, as a “dense black hole ﬂuid” where standard notions of local ﬁeld theory do not apply. The model contains two phenomenological parameters, which govern the transition between this phase of the universe and a normal phase in which the ﬁeld theory description is valid. There are indications that the transition occurs at an energy density well below the uniﬁcation scale of standard model couplings. We might then expect a transition directly into a state with most of the moduli frozen. In order for the model to provide an adequate account of the ﬂuctuations in the CMB, one must have at least one “active” modulus at these low energies, which can provide for a modest number of e-folds of inﬂation. In such a model, minima of the potential on moduli space with energy higher than the scale at which a ﬁeld theoretic description of the universe is possible, cannot make any sense. At best a small class of low energy minima could be compatible with holographic cosmology9 . If holographic cosmology is compatible with landscape ideas, the probability of accessing a particular minimum will be determined by quantum grav8

e.g. the Hartle Hawking Wave Function of the Universe. The potentials calculated in the landscape have no indication of a cut-off at energy scales far below the uniﬁcation scale. This suggests that the two theoretical frameworks are not compatible, but I am trying to avoid jumping to conclusions. 9

STRINGS IN A LANDSCAPE

13

itational considerations, far removed from effective ﬁeld theory. In this framework, the dense black hole ﬂuid is stable and is the most probable state of the universe. A normal universe like our own is determined by a somewhat improbable initial condition, but one expects the maximum entropy initial state that does not collapse to a dense black hole ﬂuid. The survival probability of a given normal state depends on both the properties of the black hole ﬂuid and the low energy physics of the normal state, so the determination of the most probable meta-stable minimum would be a complicated quantum gravitational calculation. In both of these classes of models, simple enumeration of meta-stable minimum is not a good account of the physical probability distributions. 5. Phenomenology of the landscape For practical purposes , the landscape gives us a large set of alternative effective Lagrangians for describing the physics we have observed or will observe in our universe. These are parametrized by a collection of numbers, which include the dimension of spacetime, the name, rank and representation content of the low energy gauge theory, the value of the cosmological constant, and the values of all the coupling constants and masses of ﬁelds in the Lagrangian10 . These numbers can be collected together and viewed as a multidimensional probability space. In the supergravity approximation, we have a way of calculating an a priori distribution for these numbers. Proponents of the landscape would claim that this is an approximation to some more exact distribution, though no-one has suggested a procedure for calculating the corrections. In the previous section I have suggested that early universe cosmology may make important modiﬁcations of the distribution of metastable minima. If it turned out that the distribution predicted the Lagrangian we observe with high probability, it would be a great triumph for string theory. The value of the cosmological constant tells us that this is not the case. Weinberg’s bound[11],which constrains cosmological parameters by insisting on the existence of galaxies, has the form (in Planck units). Λ ≤ Kρ0 Q3 , where ρ0 is the dark matter density at the beginning of the matter dominated era, Q is the amplitude of primordial density perturbations at horizon crossing, and K is a pure number of order 1.. For any reasonable values of the other parameters, this means Λ is much smaller than the typical value found in the landscape. It is clear then that we must supply additional data from experiment in order to ﬁx our description of the world. The landscape framework supplies some theoretical guidance 2 - it tells us that there are a ﬁnite number of possibilities, of order 1010 (to order of magnitude accuracy in the logarithm). Various authors [12] have begun to investigate the a priori distribution of properties like the gauge group and number of generations, the scale of supersymmetry breaking, the existence of large warp factors which give rise to large hierarchies of energy scales etc., assuming a uniform distribution on the space of 10 In principle we could also have non-trivial conformal ﬁeld theories in the low energy world, at least in some approximation.

14

TOM BANKS

minima. The hope is that correlations will become evident which will tell us that a small number of inputs is enough to extract the Lagrangian of the world we live in from the ensemble of Lagrangians the landscape presents us with. The anthropic principle has also been invoked as an input datum to impose on the ensemble of Lagrangians. For its proponents, the attraction of this principle is that the answer to a single yes/no question, “Is there carbon based life?” puts strong constraints on a collection of parameters in the Lagrangian (assuming all others ﬁxed at their real world values). This attraction may be an illusion. In a probability space, the characteristic function of any subset of data points is a single yes/no question. So physicists must ask if there is any special merit to the particular characteristic function chosen by the anthropic principle. This is a hard question to answer, because we do not have much theoretical understanding of life and intelligence, and we have no experimental evidence about other forms of life in the universe we inhabit. If the typical life form resembles the great red spot on Jupiter rather than us, then this life form would think that the criterion that is most appropriate to apply in our universe is the Redspotthropic principle. To put this in a more positive manner: if considerations of carbon based life lead to explanations of the values of the fundamental parameters, then we are making a prediction about the typical form of life that our descendants will ﬁnd when they explore the universe. It should look just like us, or at least be sufﬁciently similar that the criteria for its existence are close to the criteria for ours. If instead, our descendants’ explorations show that the typical life form could tolerate much larger variations in the fundamental constants, then our so-called explanations would really be a ﬁne-tuning puzzle. The Red Spot People could calculate and understand that we wouldn’t be there if the up quark mass were a little bigger, but they might reasonably ask “Who ordered them?”. Of all soi disant anthropic arguments, Weinberg’s bound on cosmological parameters is the least susceptible to this kind of criticism. If there are no galaxies, there are no planets, no Red Spots, no Black Clouds, perhaps no conceivable form of life. Polchinski and others at the school cited the numerical success of this bound as evidence that anthropic reasoning may have relevance to the real world. It is important to realize that this numerical success depends on keeping all other parameters ﬁxed. Arkani-Hamed reported on unpublished work with Dimopoulos, which showed that if both Λ and MP (really the ratio of these parameters to particle physics scales, which are held ﬁxed) are allowed to vary subject only to anthropic constraints, then the preferred value of Λ is larger than experimental bounds by many orders of magnitude. Similarly the authors of [31] following [14] argued that if both Λ and Q are allowed to vary then the probability of ﬁnding a universe like our own is of order 10−4 . A contrary result was reported in [15], but only by assuming an a priori probability distribution that favored small values of Q. In inﬂationary models of primordial ﬂuctuations, the value of Q depends on details of the inﬂaton potential at the end of slow roll. We would certainly expect this parameter to vary as we jump around the landscape. Similar remarks apply to ρ0 . Allowing ρ0 to vary would further reduce the probability that the anthropic distribution favors the real world. Thus, at least with our current knowledge of the landscape, it seems likely that the numerical success of Weinberg’s bound in the landscape context is not terribly impressive11 . 11

I cannot resist remarking that in the context of Cosmological Supersymmetry Breaking[16],

STRINGS IN A LANDSCAPE

15

The greatest challenge to all methods of dealing with the landscape is the large number of parameters in the standard model which have to be ﬁnely tuned to satisfy experimental constraints. These include the strength of baryon, lepton and ﬂavor violating couplings, θQCD and the values of many quark and lepton masses. Anthropic reasoning helps with some of these parameters, but not all, and is insufﬁcient to explain the lifetime of the proton, the value of θQCD and many parameters involving the second and third generation quarks and leptons. From the landscape point of view, the best way to deal with this (in my opinion) is to ﬁnd classes of vacua in which all of these ﬁne tuning problems are solved, perhaps by symmetries. One can then ask whether there are enough vacua left to solve the cosmological constant problem. This might be a relatively easy task. One could then go on to see whether other features of this class of vacua are in concordance with the real world. It is clear that at a certain point in this process, if we don’t falsify the landscape easily12 , we will run into the problem that current technology does not allow one to calculate the low energy parameters with any degree of precision. Indeed, the error estimates are only guesses because we don’t even know in principle how to calculate the next term in the expansion in large ﬂuxes. A more fundamental framework for the discussion of the landscape is a practical necessity as well as a question of principle. I have suggested that if a rigorous framework for the landscape exists, it is probably to be found in the context of a theory of a Big Bang universe with Eternal Inﬂation, and future FRW asymptotics for any given observer. It is likely that we will be unable to deﬁne more precise calculations of the properties of the landscape without ﬁnding a rigorous mathematical deﬁnition of such a space-time. The fundamental object in such a space-time would be a scattering matrix[8] relating a complete set of states at the Big Bang to states in Lorentzian CDL bubble space-times corresponding to decays of meta-stable dS landscape states into the Dine-Seiberg region of moduli space. The ﬁnal states, in addition to particle labels, would carry indices ,(V, L) describing a particular dS minimum L and a particular “asymptotic vacuum”, V , into which it decays. Thus, we would have matrix elements S(I|V, L, pi ), where I labels an initial state at the Big Bang and pi a set of “particle” labels for localized scattering states in a given CDL bubble. An important unanswered question is whether the S-matrix is unitary for each (L, V )13 or only when all (L, V ) sectors are taken into account. The ﬁrst alternative is analogous to the proposal of [6]. The following argument has a bearing on this question. I have stated above that there was no problem with an inﬁnite number of ﬁnal states for ﬁxed L. This is not necessarily the case. If I extrapolate scattering data on I+ backwards, using the classical equations of motion, and assuming a minimal ﬁnite energy for each particle, then all but a ﬁnite number of states will encounter a space-like singularity before transition to the metastable only Λ varies, and Weinberg’s bound retains its original numerical status. 12 e.g. by showing that the number of vacua left after all the other ﬁne tuning problems are solved is too small to solve the cosmological constant problem. 13 One would then invoke the existence of unitary mappings taking the different unitary Smatrices into each other.

16

TOM BANKS

dS regime. This is the time reverse of the argument about black holes in the initial state of the time symmetric CDL bubble. This suggests that for ﬁxed L, the matrix S(I|V.L, pi ) has ﬁnite rank: only a ﬁnite subspace of the space of all out states on the CDL geometry labeled by L, V would be allowed. This leads to a modiﬁcation of the proposal of [6] in which only the S-matrix for ﬁxed V , keeping all possible values of L, is unitary. However, there is no clear reason now to assume that V should be ﬁxed, so perhaps only the full S-matrix is unitary. A more disturbing conclusion is reached if one combines the claim of [17] that the number of L sectors is ﬁnite, with the above argument. One then concludes that the whole S-matrix has ﬁnite rank, and that the entire landscape ﬁts into a Hilbert space with a ﬁnite number of states. One of the supposed virtues of the landscape picture of metastable dS was that, unlike a stable dS space, the landscape was part of a system with an inﬁnite number of states, which could make inﬁnitely precise quantum measurements on itself. If both Douglas’ claim, and that of the last paragraph are true, this is no longer obvious. All but a ﬁnite number of the ﬁnal states in a given CDL instanton geometry, would not connect to a tunneling process from a meta-stable dS vacuum, but instead would evolve directly from a Big Bang. The part of the scattering matrix that involved meta-stable dS resonances would be of ﬁnite rank. The whole issue of a rigorous framework for the landscape remains as murky as ever. Acknowledgements I would like to thank L. Baulieu for inviting me to the Cargese workshop, and Nima Arkani-Hamed for conversations which contributed to my understanding of the issues discussed here. This research was supported in part by DOE grant number DE-FG03-92ER40689. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

N. Arkani-Hamed, H-C. Cheng, M.A. Luty, Ghost Condensation and a Consistent Infrared Modiﬁcation of Gravity, JHEP 0405:074,2004, hep-th/0312099. T. Banks, A Critique of Pure String Theory: Heterodox Opinions of Diverse Dimensions, hepth/0306074. N. Seiberg, S. Shenker, A Note on Background (In)dependence, Phys. Rev. D45, 4581, (1992), hep-th/9201017. E. Farhi, A. Guth, An Obstacle to Creating a Universe in the Laboratory, Phys. Lett. B183, 149, (1987). S. Coleman, F. DeLuccia, Gravitational Effects On and Of Vacuum Decay, Phys. Rev. D21, 3305, (1980). B. Freivogel, L. Susskind, A Framework for the Landscape, hep-th/0408133. C.R. Stephens, G. ’t Hooft, B.F. Whiting, Black Hole Evaporation Without Information Loss, Class. Quant. Grav. 11:621, (1994); L. Susskind, L. Thorlacius, J. Uglum, The Stretched Horizon and Black Hole Complementarity, Phys. Rev. D48, 3743, (1993). T. Banks, W. Fischler, M-theory Observables for Cosmological Space-times, hep-th/0102077. G. Moore, J. Horne, Chaotic Coupling Constants, Nucl. Phys. B432, 109, (1994);T. Banks, M. Berkooz, G. Moore, S. Shenker, P.J. Steinhardt, Modular Cosmology, Phys. Rev. D52, 3548, (1995). T. Banks, W. Fischler, An Holographic Cosmology, hep-th/0111142; Holographic Cosmology 3.0, hep-th/0310288; T. Banks, W. Fischler, L. Mannelli, Microscopic Quantum Mechanics of

STRINGS IN A LANDSCAPE

11. 12.

13. 14. 15. 16. 17.

17

the p = ρUniverse, hep-th/0408076. S. Weinberg, Anthropic Bound on the Cosmological Constant, Phys. Rev. Lett. 59, 2607, (1987). M.R. Douglas, Statistics of String Vacua, hep-ph/0401004; F. Denef, M.R. Douglas, Distributions of Flux Vacua, JHEP 0405:072, 2004, hep-th/0404116; F. Denef, M.R. Douglas, B. Florea, Building a Better Racetrack, JHEP 0406:034, 2004, hep-th/0404257; M.R. Douglas Statistical Analysis of the Supersymmetry Breaking Scale, hep-th/0405279; Basic Results in Vacuum Statistics, hep-th/0409207; A. Giryvaets, S. Kachru, P.K. Tripathy, On the Taxonomy of Flux Vacua, JHEP 0408:002,2004, hep-th/0404243; O. DeWolfe, A. Giryvaets, S. Kachru, W. Taylor, Enumerating Flux Vacua With Enhanced Symmetries, hep-th/0411061 M.L. Graesser, S.D.H. Hsu, A. Jenkins,M. Wise, Anthropic Distribution for Cosmological Constant and Primordial Density Perturbations, Phys. Lett. B600, 15, (2004). M. Rees, M. Tegmark, Why is the CMB Fluctuation Level 10−5 , Ap. J. 499, 526, (1998). L. Pogosian, M. Tegmark, A. Vilenkin, Anthropic Predictions for Vacuum Energy and Neutrino Masses, JCAP 0407:005, 2004, astro-ph/0404497. T. Banks, Cosmological Breaking of Supersymmetry?, hep-th/0007146; and Int. J. Mod. Phys. A16, 910, (2001). M.R. Douglas, The Statistics of String/M-theory Vacua, JHEP 0305:046,2003, hepth/0303194.

QUANTIZATION OF HOLOMORPHIC FORMS AND N = 1 SUPERSYMMETRY ON SPECIAL MANIFOLDS

LAURENT BAULIEU Laboratoire de Physique Th´eorique et Hautes Energies, Universit´e de Paris VI and Paris VII, France

Abstract: We study the quantization of a holomorphic two–form coupled to a Yang– Mills ﬁeld on special manifolds in various dimensions, and we show that it yields twisted supersymmetric theories. For K¨ahler manifolds in four dimensions, our topological model is related to N = 1 Super Yang–Mills theory. Extended supersymmetries are recovered by considering the coupling with chiral multiplets. 1. Introduction The idea that Poincar´e supersymmetry is a ”phase” of a more fundamental symmetry is appealing. For instance, it was shown that Poincar´e supersymmetry and topological symmetry are deeply related, and that the ﬁeld spectrum of dimensionally reduced N = 1 D = 11 supergravity can be determined in the context of an 8-dimensional gravitational Topological Quantum Field Theory (TQFT) [1]. One can foresee that many models, which are dimensional reductions and truncations of maximal supergravity, might be possibly related by twist to topological models [2]. The BRST operator that characterizes a topological symmetry is a scalar operator which can be deﬁned in any given curved space, while Poincar´e supersymmetry is a delicate concept in curved space. Therefore, topological symmetry could be a more fundamental concept than Poincar´e supersymmetry. On the other hand, in order to perform the twist operation that relates Poincar´e supersymmetry and topological symmetry, one often needs to use manifolds with special holonomy. We will discuss the quantization of (holomorphic) two–forms coupled to a Yang-Mills ﬁeld on special manifolds in various dimensions. These theories are basically ATQFT’s (Almost Topological Quantum Field Theories), in the sense that they are deﬁned in terms of a classical action and a set of observables which are invariant under changes of coordinates belonging to restricted classes, for instance, reparametrizations that respect a complex structure. This is to be compared to genuine TQFT that contain observables invariant under all possible changes of metrics. Interesting cases that we will analyse in detail are K¨ahler manifolds in four dimensions and special manifolds in higher (6,7 and 19 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 19–39. © 2005 Springer. Printed in the Netherlands.

20

LAURENT BAULIEU

8) dimensions. In particular in seven dimensions we will analyse G2 manifolds of the kind recently studied by Hitchin [15]. One of the original motivations for this work was to try to understand how (twisted) N = 1 supersymmetric theories can be directly constructed as TQFT. As we will see in the next section, this immediately leads to the introduction in the classical action of a “charged” 2-form B, valued in the adjoint representation of a Lie algebra. In these models one can also consider the coupling to chiral multiplets. If these transforms in the adjoint representation, one recover in this way also the extended supersymmetry in a twisted form. In Sect.2 we introduce the holomorphic BF model and discuss its relationship with N = 1 (twisted) supersymmetry. Notice that the quantization of this model requires the use of the Batalin–Vilkoviski formalism. In Sect.3 the four dimensional case is considered, including a detailed discussion on the coupling with a chiral multiplet. In Sect.4 we discuss the six–dimensional case on a Calabi–Yau three–fold and show how two different quantizations yields respectively to theories related to the B model and A model of topological string. In Sect.5 we discuss the eight–dimensional theory on a Calabi–Yau four–fold and its dimensional reduction to CY3 × S 1 . The eight–dimensional model is discussed also for manifolds with SU (4) structure. This theory can be regarded as a generalization of the four–dimensional self–dual Yang–Mills model [22]. 2. N = 1 supersymmetry and the holomorphic BF theory The standard construction of a TQFT leads to models with N = 2 supersymmetry. To see this, let us consider the ”prototype” case of Topological Yang–Mills theory in four and eight dimensions. The relevant BRST transformations read δAµ δc

= Ψµ + Dµ c = −Φ − 12 [c, c]

δΨµ δΦ

= Dµ Φ − [c, Ψµ ] = −[c, Φ]

(2.1)

These equations stand for the geometrical identity (δ + d)(A + c) + 12 [A + c, A + c] = F + Ψ + Φ [23]. There are as many components in the topological ghosts as in the gauge ﬁelds, and to gauge ﬁx the topological freedom, one must also introduce as many antighosts as topological ghosts. The antighosts are an anticommuting antiself dual 2-form κµν and an anticommuting scalar η. For each one of the antighosts, there is an associated Lagrange multiplier ﬁeld, and their BRST equations are : δκµν ¯ δΦ

= bµν − [c, κµν ] δbµν ¯ = η − [c, Φ] δη

= −[c, bµν ] = [c, η]

(2.2)

The twist operation is a mapping from these ghost and antighost fermionic degrees of freedom on a pair of spinors, which leads one to reconstruct the spinor spectrum of N = 2 supersymmetry, both in 4 and 8 dimensions. The scalar BRST operator δ can then be identiﬁed as a Lorentz scalar combination of the N = 2 Poincar´e supersymmetry generators. However, this ”twist” operation has different geometrical interpretation in 4 and 8 dimensions. In the former case, it is a redeﬁnition of the Euclidean Lorentz group contained in the global SUL (2) × SUR (2) × SU (2) invariance of the supersymmetric theory. In the latter case, it uses the triality of 8-dimensional space. In the previous works

QUANTIZATION OF HOLOMORPHIC FORMS

21

[4, 5, 6], a constant covariant spinor has been used, which implies that one uses Spin(7) invariant manifolds; one can also use a manifold with SU (4) holonomy. Using self-duality equations as gauge functions, one can build a δ-exact action that provides twisted supersymmetric theories with a δ-exact energy momentum tensor. The cohom*ology of the δ symmetry determines therefore a ring of topological observables, which is a subsector of the familiar set of observables for the gauge particles. The latter is selected from the cohom*ology of the ordinary gauge invariance. From (2.1) and (2.2) one concludes that in these TQFT one has twice as many fermionic degrees of freedom than bosonic ones. This makes seemingly impossible to determine N = 1 models as a twist of TQFTs. We may, however, look for models with a “milder” topological symmetry, such as BF models or Chern–Simons type model, characterized by metric independent classical actions. Such actions are not boundary terms, and thus their topological symmetry cannot be as large as that displayed in (2.1) and (2.2). This might lead us to models that are twisted N = 1 supersymmetric theories. In this paper, we will consider the following holomorphic BF action Tr(Bn,n−2 ∧ F0,2 ) (2.3) In−BF = M2n

which is deﬁned on any complex manifold M of complex dimension n. Some aspects of the classical action (2.3) were studied in [24]. It would be interesting to study its quantization and possible relation with supersymmetry. For the moment we only consider some particular models which can be obtained from (2.3) by choosing a particular form for the ﬁeld Bn,n−2 . Notice that the equations of motion for this ﬁeld coming from (2.3) implies that Bn,n−2 is a holomorphic (n, n − 2) form. One might try to deﬁne a theory that is classically invariant under the following (almost) topological symmetry, which is localised in the holomorphic sector 1 : QAm = Ψm + Dm c Qc = − 12 [c, c]

QΨm = −[c, Ψm ] QAm ¯ = Dm ¯c

(2.4)

This “heterotic” symmetry was already used [26, ?, ?, ?] in four dimensions and in [4, ?] in higher dimensions. Here we recover it as a symmetry associated to the classical action (2.3). If we count the ghost degrees of freedom, we have 2 components for Ψm and one for c. Notice that Ψm cannot have a ghost of ghost symmetry with a ghost of ghost Φ, since QΨm = Dm Φ − [c, Ψm ] and Qc = Φ − 12 [c, c] would imply that Q2 Am ¯ = 0. Modulo gauge transformations, only one degree of freedom for the ﬁeld A is left free by the symmetry in (2.4). Moreover, if we succeed in writing a BRST gauge–ﬁxed action for the classical symmetry in (2.4), this will depend on the ghosts Ψm and as many antighost components as there are in Ψm (four components). Then the number of fermion degrees of freedom will ﬁt with those of a single Majorana spinor and we have a chance to eventually reach N = 1 supersymmetry, as we will explain in detail in the next section. Notice that in these models one can also recover the coupling to a chiral multiplet in the adjoint representation and the corresponding extended supersymmetry in the twisted form. 1 We use the standard notation where the complex indices are denoted with latin letters m, n and ¯ . m, ¯ n ¯ , and the complex coordinates are given by z m and z¯m

22

LAURENT BAULIEU

3. Four dimensions: K¨ahler manifold ¨ 3.1. THE CLASSICAL ACTION FOR A BF SYSTEM ON A KAHLER MANIFOLD On a K¨ahler manifold one can deﬁne a complex structure ¯n J m¯ =0,

J mn = 0 , J m¯n = ig m¯n

(3.1)

¯ and 1 ≤ m, m ¯ ≤ N in which allows one to introduce complex coordinates z m and z m 2N dimensions by ¯ n ¯ Jn¯m z ¯ = −iz m . (3.2) Jnm z n = iz m ,

In four dimensions, the action (2.3) reads Icl (A, B) = TrB2,0 ∧ F0,2 = M4

M4

√ ¯n d4 x g Tr mnm¯ Bmn Fm¯ ¯n

(3.3)

where F = dA + A ∧ A is the curvature of the Yang–Mills ﬁeld A. The equations of motion are Fm¯ ¯n = 0

¯n

mnm¯ Dn¯ Bmn = 0

(3.4)

Classically, Am is undetermined, Am ¯ is a pure gauge and Bmn is holomorphic. Notice that B2,0 has no vector gauge invariance. It counts for one propagating degree of freedom. Altogether, there are two gauge invariant degrees of freedom that are not speciﬁed classically. Modulo gauge invariance, there is a mixed propagation between A and B. The symmetries of the action (3.3) are QAm = Ψm + Dm c Qc = − 12 [c, c] QBmn = −[c, Bmn ]

QΨm = −[c, Ψm ] QAm ¯ = Dm ¯c

(3.5)

In the ﬁrst two lines of the above equation we can recognize the symmetry (2.4). The geometrical interpretation of complete charged 2-forms is a non-trivial issue. However, in the language where form degree and positive and negative ghost number are uniﬁed within a bigrading, the charged 2-form can be understood as a sort of Hodge dual to the Yang–Mills ﬁeld [30]. Here we only consider the (2, 0) component of such an object, and we hopefully avoid the ambiguities for deﬁning its theory. We now explain the BRST quantization of the action (3.3), for the sake of inserting it in a path integral. ¨ 3.2. QUANTIZATION OF THE BF SYSTEM ON A KAHLER MANIFOLD In order to deﬁne a quantum theory, that is, a path integral, we need to gauge ﬁx the (almost) topological symmetry of the BF system, in a way that respect the BRST symmetry associated to this symmetry. As it is well-known, the anti–self–duality condition in 4 dimensions can be expressed in complex coordinates as: Fmn = 0 , Jm¯n F m¯n = 0

Fm¯ ¯n = 0 (3.6)

QUANTIZATION OF HOLOMORPHIC FORMS

23

and one has the identity 1 m¯ ¯n Tr Fm¯ + 12 |Jm¯n F m¯n |2 = Tr(Fµν F µν + Fµν F˜ µν ) (3.7) ¯ nF 4 Modulo ordinary gauge invariance, we have two topological freedoms, corresponding to the two components in Ψm . In order to perform a suitable gauge ﬁxing for the two–form B2,0 and for A1,0 , which is the part of the gauge connection absent from the classical action (3.3), we introduce two anticommuting antighosts κmn and κ, and two Lagrange multipliers bmn and b: Qκmn Qκ

= bmn = b

Qbmn Qb

= =

0 0

(3.8)

Since Q2 = 0 from the beginning, we have a ﬁrst order Batalin-Vilkoviski (BV) system. However, since the treatment of the chiral multiplet in the next Section will produce a non-trivial second order BV system, we ﬁnd convenient to introduce right now BV antiﬁelds for A and B and their ghosts, antighosts and Lagrangian multipliers. The upper left notation ∗ labels antiﬁelds2 . Let us recall that the antiﬁeld ∗ φ of a ﬁeld with ghost number g has ghost number −g − 1 and opposite statistics. For a Q-invariant BV action S one has lS . The property Q2 = 0 is equivalent to the master equation Qφ = ∂∂∗l Sφ and Q∗ φ = − ∂∂φ ∂r S ∂ l S =0 ∂φ ∂ ∗ φ

(3.9)

where (∂r , ∂l ) indicate respectively the derivatives from the left and from the right. The following BV action encodes at once the classical action (3.3) and the deﬁnition of its BRST symmetry: 1 √ ¯n d4 x g Tr mnm¯ Bmn Fm¯ S = ¯n 4 M4 ¯ +∗ Am (Ψm + Dm c) +∗ Am (Dm ¯ c)

−∗ B mn [c, Bmn ] −∗ Ψm [c, Ψm ] − +∗ κmn bmn +∗ κb

1∗ 2 c[c, c]

(3.10)

The BV master equation (3.9) is satisﬁed, which implies the gauge invariance of the classical action as well as the nilpotency Q2 = 0 on all the ﬁelds. It is actually important to note that the invariance of the action (3.10) implies that ¯n ∗ mn Q∗ B mn = mnm¯ Fm¯ ]. ¯ n − [c, B

(3.11)

This equation will shortly play a key role for deﬁning the coupling to scalar ﬁelds. The topological gauge ﬁxing corresponds to the elimination of antiﬁelds by a suitable choice of a gauge function Z. The antiﬁelds are to be replaced in the path integral by the BV formula: ∗

2

φ=

δZ δφ

(3.12)

As stressed in [30], the antiﬁelds of A appear in the ghost expansion of B and vice-versa.

24

LAURENT BAULIEU

The Q-invariant observables are formally independent on the choice of Z. In particular, their mean values are expected to be independent on small changes of the metric that one must introduce to deﬁne Z. In order to concentrate path integral around the anti-self-duality condition (3.6) we choose: m¯ ¯n ) + κ( 12 b + iJm¯n F m¯n ) Z = κmn (Bmn − mnm¯ ¯ nF

(3.13)

The BV equation (3.12) implies that B2,0 is eliminated in the path integral with Bmn = m¯ ¯n . After Gaussian integration on b, the gauge-ﬁxed action reads

mnm¯ ¯ nF √ m¯ ¯n S g.f. = d4 x g Tr Fm¯ + 12 |Jm¯n F m¯n |2 ¯ nF M4 ¯ npq ml ¯ l (3.14) −2 m¯ κm¯ Dm ¯ n Dp Ψq + iκJ ¯Ψ Here and in the following discussions we omit the c–dependent terms in the action. In fact these terms express the covariance of the gauge–ﬁxing conditions (3.6) with respect to the gauge symmetry, and vanish when these conditions are enforced. Moreover, we leave aside the standard gauge ﬁxing of the ordinary gauge invariance ∂ µ Aµ = 0. The action (3.14) can be compared with that of N = 1 SYM on a K¨ahler manifold. It is known [31] that on a complex spin manifold the complex spinors can be identiﬁed with even forms S± ⊗ C ∼ Ω0,odd , so that we can identify our topological ghost Ψm as a left– handed Weyl spinor λα and the topological anti–ghosts (κm¯ ¯ n , κ) as a right–handed Weyl ¯ α˙ . More explicitly, the holonomy group of a four–dimensional K¨ahler manifold is spinor λ locally given by U (2) ∼ SU (2)L ×U (1)R ⊂ SU (2)L ×SU (2)R , so that one can naturally identify the forms σµ α1˙ dxµ and σµ α2˙ dxµ as (1, 0) and (0, 1) forms respectively [26, 27]. Then the twist reads3 Ψm κm¯ ¯n κ

= λα σµ α1˙ eµm , α ˙ ¯ α˙ σ = λ ¯µν eµ¯ eνn¯ , 2˙ m ¯ α˙ = δ2˙α˙ λ

(3.15)

On a Hyperk¨ahler manifold, the twist formula can be reinterpreted by making explicit a constant spinor dependence in (3.15). With this change of variables, it is immediate to recognize that the action (3.14) is the N = 1, D = 4 Yang–Mills action √ 1 ¯ µ Dµ λ d4 x g Tr Fµν F µν + Fµν F˜ µν + λγ (3.16) SSY M = 4 M4 As compared to [26], we started from a classical BF system, which, eventually, gives the N = 1, D = 4 Yang–Mills theory as a microscopic theory in a twisted form. Let us notice that it should be possible to cast the topological BRST symmetry in the form of conditions on curvatures yielding descent equations with asymmetric holomorphic decompositions and eventually solve the cocycle equations for Q, similarly to what has been done in [23] for the Topological Yang–Mills theory. 3

We deﬁne the euclidean σ matrices as σµ = (iτ c , 1), τ c , c = 1, 2, 3 being the Pauli matrices.

QUANTIZATION OF HOLOMORPHIC FORMS

25

3.3. COUPLING OF THE BF TO A CHIRAL MULTIPLET Since the N = 2 SYM theory is an ordinary TQFT, and since its Poincar´e supersymmetric version can be obtained by coupling the N = 1 Yang–Mills multiplet to a chiral multiplet in the adjoint representation of the gauge group, one expects to have an expression of the N = 1 scalar theory as an (almost)TQFT on a K¨ahler manifold. As we shall see, this is slightly more complicated than the N = 1 Super Yang-Mills theory, since it will involve the vector gauge symmetry of a (0,2)-charged form, and thus a 2nd rank BV system arises. In order to introduce the chiral multiplet, we extend the set of classical ﬁelds of the previous section as B2,0 → (B2,0 , B0,2 ), However, we keep the same classical action as in (3.3): ¯n Tr(B2,0 ∧ F0,2 ) = Tr( mnm¯ Bmn Fm¯ Icl (A, B0,2 , B0,2 ) = ¯ n) M4

(3.17)

(3.18)

M4

Having a classical action that is independent of B0,2 is equivalent of having the following symmetry for B0,2 QBm¯ ¯n QΨm ¯

1 ∗ mn = D[m , Φ] ¯ n ] − m¯ ¯ nmn [ B ¯ Ψn ¯ ] − [c, Bm¯ 4 = Dm ¯ Φ − [c, Ψm ¯]

(3.19)

QΦ = −[c, Φ]. In fact, the unique degree of freedom carried by Bm¯ ¯ n is canceled by the two degrees of freedom of the topological ghost Ψn¯ , deﬁned modulo the ghost of ghost symmetry generated by Φ. The presence of the antiﬁeld ∗ B mn in (3.19) is necessary in order that Q2 = 0, as one can verify by using (3.19), and (3.11), together with the usual BRST variation of the ghost c, Qc = −c2 . We thus have a second order BV system, since the BRST variations of the ﬁelds depend linearly on the antiﬁeld. This non trivial property justiﬁes, a posteriori, that the classical action depend on the charged (2, 0)-form B2,0 in an ”almost topological way”, as in (3.3), with the ordinary gauge symmetry QAm ¯ = Dm ¯c and QB2,0 = −[c, B2,0 ]. This determines the relevant Q-transformation of the antiﬁeld of the 2-form, which is eventually necessary to obtain a closed symmetry. The fate of the (0, 2)-form B0,2 is to be gauge-ﬁxed and eliminated from the action, as it was the case for B2,0 , but wit a different gauge function. For this purpose we choose ¯n ¯n , with bosonic Lagrange multiplier bm¯ . a topological antighosts that is a (0,2)-form κm¯ The ghost of ghost symmetry of Ψm ¯ must be gauge ﬁxed, and we introduce a bosonic ¯n ¯ with fermionic Lagrange multiplier η¯. κm¯ and η¯ will be eventually untwisted antighost Φ and provide half of a Majorana spinor for N = 1 supersymmetry One has in fact an ordinary pyramidal structure for a 2-form gauge ﬁeld 4 , which shows that Bm¯ ¯ n truly carries zero degrees of freedom, and can be consistently gauge4 In this table, we indicate explicitly the ghost number of the ﬁelds by a superscript. The BRST symmetry acts on the South–West direction, as indicated by the arrows.

26

LAURENT BAULIEU

ﬁxed to zero

(3.20) (1)

The vector ghost symmetry of the charged (0, 2)-form with ghost Ψm ¯ plays the role of a ¯ will be identiﬁed as the complex topological symmetry. In the untwisted theory, Φ and Φ scalar ﬁeld for the N = 1 chiral multiplet in 4 dimensions. The BV action for the ﬁelds in the Table (3.20) and their antiﬁelds is ∗ 1 √ ¯n ∗ mn Smatter = d4 x g Tr B m¯ (D[m , Φ]) ¯ n ] − m¯ ¯ nmn [ B ¯ Ψn ¯ ] − [c, Bm¯ 4 M4 ¯ ∗ +∗ Ψm (Dm ¯ Φ − [c, Ψm ¯ ]) − Φ[c, Φ] m¯ ¯n ∗ ¯ +∗ κm¯ η + Φ¯ ¯ nb

(3.21)

For consistency, we have to add to the above action the action (3.10) in order to properly deﬁne the variation of ∗ B mn as in (3.11). In order to gauge–ﬁx Smatter , we choose the following BV gauge function: ¯n ¯ ¯ m Bm¯ Z = κm¯ ¯ n + ΦD Ψm ¯

(3.22)

¯n Using the BV equation (3.12) and integrating on the Lagrangian multiplier bm¯ one gets = 0, and ﬁnds Bm¯ ¯n √ g.f. ¯n m ¯ m ¯ ¯ (3.23) Smatter = d4 x g Tr κm¯ D[m Ψ + ΦD D Φ + η ¯ D Ψ m ¯ m ¯ ¯ n ¯] M4

As for the Yang–Mills supermultiplet, we can perform a mapping of the topological ghost m¯ ¯n Ψm , η¯) on left and right handed spinors (ψ α , ψ¯α˙ ) ¯ and of the topological anti–ghosts (κ respectively. With this change of variables, we now recognize that the action (3.23) is the N = 1, D = 4 chiral multiplet action √ ¯ µ Dµ ψ ¯ µ Dµ Φ + ψγ d4 x g Tr ΦD (3.24) SSY M = M4

The sum of both actions (3.10) and (3.21), when the suitable gauge ﬁxing conditions (3.13) and (3.22) are enforced, corresponds to the twisted N = 2 Super Yang–Mills action, with notations that are adapted to a K¨ahler manifold 5 . 5

¯ Φ]2 typical of In order to recover the Yukawa couplings and the quartic term in the potential [Φ, the N = 2 SYM one should slightly modify the gauge–ﬁxing fermion Z in (3.22), but this doesn’t change the results on the topological observables.

QUANTIZATION OF HOLOMORPHIC FORMS

27

However, the BRST algebra we discussed so far is mapped only to a N = 1 subsector of the N = 2 supersymmetry. The complete N = 2 superalgebra on a K¨ahler manifold has been discussed in [25, ?, ?]. Let us brieﬂy display how these results can be recovered in our model. In our construction, we can interchange the role of holomorphic and ¯ antiholomorphic coordinates. We can thus consider another operator ∆ ∆Am ∆Am ¯ ∆Ψm ∆Ψm ¯ ∆Φ

¯ m ∆A ¯ ∆Am ¯ ¯ m ∆Ψ ¯ m ∆Ψ ¯ ¯ ∆Φ

= Ψm = 0 = 0 = Dm ¯Φ = 0

= 0 = Ψm ¯ = Dm Φ = 0 = 0

(3.25)

¯ we indicate the equivariant BRST operators, with the ghost ﬁeld c assowhere with ∆, ∆ ciated to the gauge symmetry set to zero. One has: ¯ m (∆ + ∆)A ¯ (∆ + ∆)Am ¯ ¯ m (∆ + ∆)Ψ ¯ m (∆ + ∆)Ψ ¯ ¯ (∆ + ∆)Φ

= Ψm = Ψm ¯ = Dm Φ = Dm ¯Φ = 0

(3.26)

¯ 2 = 0, and (∆ + ∆) ¯ 2 = {∆, ∆} ¯ = δΦ , where δΦ is a gauge transformation Thus, ∆2 = ∆ ¯ with parameter Φ. The operator (∆ + ∆) is the topological BRST symmetry operator (for c = 0), corresponding to the twisted N = 2 supersymmetry. The classical action which is invariant under the symmetry (3.26) is the (real) BF action plus a “cosmological term” ¯ Tr(B ∧B), with B the complete real two–form. This ﬁeld transform as (∆+ ∆)B = DΨ. In a space where one cannot consistently separate holomorphic and antiholomorphic com¯ which is Lorentz invariant. ponents of forms, the only admissible operation is (∆ + ∆), 2 ¯ Then, to close the BRST symmetry, and get δ = (Q + Q)2 = 0, one must redeﬁne the BRST transformation of the Faddeev–Popov ghost c, as follows: Qc = − 12 [c, c] → Qc = −Φ − 12 [c, c]

(3.27)

In this way one recover the complete symmetry of the Topological Yang–Mills action (2.1). 4. Six dimensions: Calabi–Yau three–fold On a Calabi–Yau three–fold CY3 we can use the holomorphic closed (3, 0)–form Ω3,0 and deﬁne B3,1 = Ω3,0 ∧ B0,1 . The classical action (2.3) then becomes Icl (A, B0,1 ) =

M6

Ω3,0 ∧ Tr(B0,1 ∧ F )

(4.1)

28

LAURENT BAULIEU

The BRST symmetry corresponding to the action (4.1) is QAm QAm ¯

= Ψm + Dm c = Dm ¯c,

QBm = Dm ¯ ¯ χ − [c, Bm ¯], Qχ = −[c, χ] , Qc = − 12 [c, c] , (4.2) The invariance of the action (4.1) under the transformation of the B ﬁeld is guaranteed by part of the Bianchi identity D[m ¯ Fn ¯¯ l] = 0

(4.3)

and the fact that Ω is closed. Notice also that the action (4.1) is invariant under the complexiﬁed gauge group GL(N, C). The BRST symmetry (4.2) follows from the following Batalin–Vilkoviski action S

= M6

√ 6 ¯ p¯ ¯q ∗ m ¯ g d x Tr m Bm ¯ Fp¯ ¯q + B (Dm ¯ χ − [c, Bm ¯ ])

∗ m ¯ ∗ m +∗ Am (Ψm + Dm ¯ c) + A Dm ¯ c − Ψ [c, Ψm ] +∗ c(− 12 [c, c]) ,

(4.4)

¯ is the volume form. where we have normalized the (3, 0)–form Ω such that Ω ∧ Ω Let us now proceed to the quantization of the model: this can be performed in different ways, which lead us to the study of different sets of observables. If one chooses to quantize the theory around the perturbative vacuum corresponding to holomorphic ﬂat connections, the corresponding observables will depend on the complex structure Ω of the manifold, as usually happens in type B topological string theories. In fact, in this case the holomorphic BF model has a deep relationship with the holomorphic Chern–Simons theory, which can be regarded as an effective action for D5 branes in type B topological string [32]. This relationship should be a generalization of that between BF and Chern Simons theories in three (real) dimensions [33], and it deserves further investigations. If instead one quantize the theory around a non-perturbative vacuum corresponding to a stable holomorphic vector bundle, one can show that the BF model correspond to the twisted version of a supersymmetric Yang–Mills theory! In this case the observables are dependent on the K¨ahler data of the manifold, as happens in type A topological string. In fact a direct relationship between a twisted U (1) maximally supersymmetric action and the topological vertex has been shown in [7]. The study of the holomorphic BF model in the abelian case can be then useful to clarify the issue of S–duality in topological strings pointed out in [8, 9, 10], and the relationship between Gromov–Witten and Donaldson–Thomas invariants discussed in [18]. Let us now show the details of the two quantizations.

QUANTIZATION OF HOLOMORPHIC FORMS

29

4.1. PERTURBATIVE QUANTIZATION AND B MODEL In this case, the symmetries are treated as ordinary gauge symmetries and ﬁxed with transversality conditions on the Am ¯ and Bm ¯ ﬁelds: ¯ Am Dm ¯ m ¯

D Bm ¯

=

=

(4.5)

The BV fermion corresponding to these conditions is ¯ ¯ Z = χD ¯ m Bm ¯Dm Am ¯ +c ¯

(4.6)

Once the gauge ﬁxing conditions (4.5) are enforced, one has a well deﬁned mixed propagator between the A and B ﬁeld, and can use it to evaluate the path integral in a perturbative expansion. The partition function of this model in the semiclassical limit should be related to the Ray–Singer holomorphic torsion [34] similarly to what happens for the holomorphic Chern–Simons theory analysed in [35, 36]. The higher order terms in the perturbative expansion should be related to other manifold invariants. It would be interesting to study these invariants along the lines of the perturbative analysis of three–dimensional Chern–Simons theory [37]. 4.2. NON–PERTURBATIVE QUANTIZATION AND A MODEL The shift symmetry on the (0, 1) part of the connection gives rise to three degrees of freedom, while the symmetry on the B ﬁeld to one. These are collected into the ghost ﬁelds (Ψm , χ) respectively. In the non–perturbative case, the gauge ﬁxing conditions are chosen as follows Fmn ¯ J mn Fmn ¯

4 = − mnp B p 3 = 0

(4.7)

and amount to three conditions for the ﬁrst line and one for the second. The reason for the particular choice of the coefﬁcient in the ﬁrst equation of (4.7) will be evident shortly. Notice that the second equation in (4.7) reduces the complex gauge group GL(N, C) to the unitary group U (N ), and as such can be considered as a partial gauge–ﬁxing for the complex gauge symmetry of the classical action (4.1). This has to be completed with a further gauge–ﬁxing for the unitary group, as for example the ordinary Landau gauge ∂ µ Aµ = 0. We will discuss this issue in more detail in Sect.5.1. The BV fermion corresponding to (4.7) is given by 4 ¯n p ¯ (Fm¯ ¯(2iJ mn Fmn Z=χ ¯m¯ ¯ n + mnp B ) + η ¯ − h) , 3

(4.8)

¯n , η¯) are the antighosts associated to the gauge–ﬁxing conditions (4.7), whose where (χ ¯m¯ BV action is given by √ 6 m¯ ¯n ∗ Saux = ¯m¯ + η¯h +∗ c¯b) (4.9) g d x Tr(∗ χ ¯ nh M6

30

LAURENT BAULIEU

Eliminating the anti–ﬁelds by means of (3.12) and implementing the gauge–ﬁxing conditions (4.7) by integration on the Lagrangian multipliers, we get from (4.4) and (4.9) S

g.f.

=

3 √ ¯n m¯ n d6 x g Tr − F m¯ Fm¯ Fm¯n |2 ) ¯ n + |J 2 M6 4 ¯n ¯ ¯ n p¯ +χ ¯m¯ D[m η J mn Dm ¯m¯ D χ . ¯ Ψn + m¯ ¯ np¯χ ¯ Ψn ¯ ] + 2i¯ 3

(4.10)

By using the identity [38] 1 − Tr(F ∧ ∗F ) + J ∧ Tr(F ∧ F ) = 4 3 ¯n mn ¯ 2 Tr − F m¯ Fm¯ Fmn ¯ n + |J ¯ | 2

(4.11)

we can recognize in the ﬁrst line of (4.10) the bosonic part of the N=1 D=6 SYM action, modulo the topological density J ∧ Tr(F ∧ F ), where J is the K¨ahler two–form. Concerning the fermionic part, we can make use of the mapping between chiral fermions odd and complex forms S± ⊗ C ∼ Ωeven to map the topological ghosts (Ψm , χ) into the ¯ and the topological antighosts into the left–handed spinor λ. More right–handed spinor λ explicitly, we can use the covariantly constant spinor ζ of the Calabi–Yau three–fold to perform the mapping Ψm χ m¯ ¯n χ ¯

¯ mζ → λΓ ¯ → λζ ¯n → ζΓm¯ λ

η¯ →

m ¯ n ¯ p¯

m¯ ¯ np¯ζΓ Γ Γ λ

(4.12)

In this way, one can recognize in (4.10) the twisted version of the N = 1 D = 6 Super Yang–Mills action. In order to reproduce the U (1) twisted maximally supersymmetric action discussed in [7], one has to add to the classical action (4.1) the higher Chern class F ∧ F ∧ F and couple this theory to an hypermultiplet, with a procedure similar to that discussed in the four dimensional case. As in Sect.3.3, one has to consider the quantization of a (0, 2)–form B0,2 . The corresponding BRST complex is the same as in Table (3.20), but now with six–dimensional ﬁelds (m, ¯ n ¯ = ¯1, ¯2, ¯3). It is straightforward to realize that the ﬁelds appearing in the Table (3.20) together with the multiplet discussed in this subsection give rise exactly to the spectrum of the twisted maximally supersymmetric Yang–Mills discussed in [7]. An alternative and more economical way would be to proceed from the dimensional reduction of the eight-dimensional model that we are going to discuss in the next section. Notice that, as discussed in Sect.3.3, the coupling to the hypermultiplet does not change the classical action (4.1). Moreover, the higher Chern class (F )3 is only a boundary term which does not affect the propagation of the A and B ﬁelds. Thus the relationship with the perturbatively quantized model of the previous subsection still holds.

QUANTIZATION OF HOLOMORPHIC FORMS

31

5. Eight dimensions 5.1. CALABI–YAU FOUR–FOLD On a Calabi–Yau four–fold we can write the following generalization of the action (3.3) Icl (A, B0,2 ) = Ω4,0 ∧ Tr(B0,2 ∧ F0,2 ) (5.1) M8

Here Ω4,0 is the holomorphic covariantly closed (4,0)-form. This, together with part of the Bianchi identity, ensures the invariance of the classical action (5.1) analogously to the CY3 case of the previous section. Also, as in the previous section, we normalize Ω such ¯ is the volume element on M8 . The action (5.1) displays the symmetry that Ω ∧ Ω QAM

=

ΨM + DM c

QAM¯

= DM¯ c ,

1 ¯¯ = D[M¯ χN¯ ] − [c, BM¯ N¯ ] − M¯ N¯ P¯ Q¯ [∗ B P Q , φ] , 4 QχN¯ = DN¯ φ − [c, χN¯ ] , Qc = − 12 [c, c] ,

QBM¯ N¯

Qφ = −[c, φ]

(5.2)

Notice that c is the complexiﬁed Faddeev–Popov ghost. The BV action corresponding to (5.1) is given by √ 8 ¯ ¯¯¯ S = g d x Tr M N P Q BM¯ N¯ FP¯ Q¯ M8

1 ¯ ¯ ¯¯ +∗ B M N (D[M¯ χN¯ ] − [c, BM¯ N¯ ] − M¯ N¯ P¯ Q¯ [∗ B P Q , φ]) 4 ¯ ¯ +∗ AM (ΨM + DM¯ c) +∗ AM DM¯ c +∗ χN (DN¯ φ − [c, χN¯ ]) ¯ ¯ ¯η +∗ χ ¯M¯ N¯ hM N +∗ χh ¯ +∗ c¯b +∗ φ¯ +∗ c(− 12 [c, c]) −∗ φ[c, φ]

(5.3)

The action (5.1) only deﬁne the propagation of part of the gauge ﬁeld, as in the case studied in section 2. It can be gauge–ﬁxed in by imposing six complex conditions for BM¯ N¯

and a gauge–ﬁxing for χM¯

+ BM ¯N ¯

=

− BM ¯N ¯

− = FM ¯N ¯ ,

0,

¯

DM χM¯ = 0. ± B0,2

(5.4)

(5.5)

The projection on self-dual or anti-self-dual part of the (0, 2)–forms can be done by using the anti-holomorphic (0, 4) form. The conditions (5.4) can be enforced by using ¯ ¯ ¯ ¯ the BRST doublets of complex antighosts and Lagrangian multipliers (χ ¯M N , hM N ) and

32

LAURENT BAULIEU

¯ χ) (φ, ¯ respectively. Then, as a generalization of [4], we complete the above six complex conditions for BM¯ N¯ by the following complex condition : DM c AM¯ = 0 ¯

(5.6)

The real part of (5.6) is the ordinary Landau gauge condition. The imaginary part gives instead a condition analogous to the second line of (4.7): ¯

ImDM AM¯ ¯ M

ReD AM¯

¯

=

0 ⇒ J M N FM N¯ = 0

=

0 ⇒ ∂ Aµ = 0

(5.7)

µ

The gauge–ﬁxing fermion corresponding to the gauge conditions (5.4), (5.6) and (5.7) is Z

¯ ¯+ ¯ ¯− ¯ M¯ χM¯ ¯M N (BM¯ N¯ − 2FM¯ N¯ ) + φD = χ ¯M N BM¯ N¯ + χ ¯

+χ(iJ ¯ M N FM N¯ + 12 h) + c¯(∂ µ Aµ + 12 b)

(5.8)

By using the BV equation (3.12) and enforcing the gauge conditions (5.4), (5.6) and (5.7) by integration on the Lagrangian multipliers, we get the wanted action, as a twisted form of the D = 8 supersymmetric Yang–Mills action. Its gauge invariant part is : √ ¯ ¯ ¯ − 1 MN ¯ M¯ DM¯ φ d8 x g Tr 2F M N − FM FM N¯ |2 ) + φD S g.f. = ¯N ¯ + 2 |J M8 ¯ ¯ ¯ ¯ ¯ ¯M N + D[M¯ ΨN¯ ] + χD ¯ M ΨM + η¯DM χM¯ (5.9) +χ ¯M N − D[M¯ χN¯ ] + χ As for the previous cases, we do not display in (5.9) the gauge dependent part of the action. By using the identity 1 √ ¯ ¯N ¯− − 8 √ 2 M 1 MN d x g Tr(2F FM¯ N¯ + 2 |J FM N¯ | ) + S0 = d8 x g Tr(F µν Fµν ) 4 M8 M8 (5.10) where S0 = M8 Ω ∧ Tr(F0,2 ∧ F0,2 ) is a surface term [4], we can recognize the ﬁrst line of the action (5.9) as the bosonic part of the D = 8 SYM action. Concerning the even fermionic part, one can use the identiﬁcation of fermions with forms S± ∼ Ωodd ,0 , which for D = 8 reads S−

∼ Ω1,0 ⊕ Ω3,0 ,

S+

∼ Ω

0,0

⊕Ω

2,0

(5.11)

⊕Ω

4,0

(5.12)

to identify the topological ghosts (ΨM , χM¯ ) with the right–handed projection of the Ma¯ ¯ ¯ a˙ , a˙ = 1, . . . , 8, and the topological anti–ghosts (χ, ¯ χ ¯M N , η¯) with the jorana spinor λ left–handed projection λa . Notice in fact that on a Calabi–Yau four–fold one can use the holomorphic four–form Ω to identify the (3, 0)–form appearing on the r.h.s. of (5.11) with the the ﬁeld χM¯ [31]. Analogously, one can identify the scalar η¯ with the (4, 0)–form appearing in the r.h.s. of (2.15). More explicitly, one can use the two left–handed covariantly constant spinors (ζ1 , ζ2 ) of the Calabi–Yau four–fold to identify χM χ ¯− MN χ ¯

¯ = M N P Q ζ1 ΓN ΓP ΓQ λ − = ζ1 ΓM N λ = ζ1 λ

(5.13)

QUANTIZATION OF HOLOMORPHIC FORMS

33

and ΨM

¯ = ζ2 ΓM λ

χ ¯+ MN

= ζ2 Γ+ MN λ MP NQ η¯ =

ζ2 ΓM ΓN ΓP ΓQ λ

(5.14)

moduli space and Spin(7) theory: the moduli space probed by the above TQFT is a ¯ 02 = 0 and holomorphic (0, 2)–form DB + FM ¯N ¯

D

¯c M

AM¯

=

(5.15)

=

(5.16)

Notice that the classical action (5.1) is invariant under the group of complex gauge transformations GL(N, C). The moduli space described by (5.15) and (5.16) with complex gauge group GL(N, C) should be equivalent to that described by (5.15) and (5.7) with the unitary group U (N ). This last moduli space is directly related with that explored by a Spin(7) invariant topological action [35]. In fact, by using (5.7) one realizes that the imaginary part of (5.16) together with (5.15) amount to seven real conditions, which ﬁt in the 7 part of the Spin(7) decomposition of the (real) two–forms 28 = 7 ⊕ 21. The real part of (5.16) is the ordinary transversality condition for the unitary gauge group. Then the theory deﬁned by the action (5.1) should be equivalent to that deﬁned by the Spin(7)–invariant action Ψ ∧ Tr(B ∧ F ) (5.17) IΨ−BF = M8

where Ψ is the real Spin(7)–invariant Cayley four–form and B, F are real two–forms. The mapping can be done by identifying the fundamental representation of the SU (4) group with the real spinor representation of Spin(7). 5.2. SEVEN DIMENSIONS: FROM CY4 TO CY3 × S 1 In [4], the case of writing a BRSTQFT for a G2 manifold was directly done by starting from the topological action d7 xcijk Di ϕFjk (5.18) M7

where cijk stand for the G2 -invariant tensor made of octonionic structure coefﬁcients, and ϕ is a Higgs ﬁeld. The BRST quantization of this topological term yields a twisted version of the dimensional reduction to seven dimensions of the D=8 super Yang–Mills action on a manifold with Spin(7) or SU (4) holonomy. Basically, the topological gauge functions are the generalization of Bogolmony equations in 7 dimensions, as shown in [4]. G2 manifolds of the kind Σ6 × S1 , where Σ6 is a Calabi–Yau 3-fold, are of special interest both for mathematical [15] and physical [12] applications. A topological theory for such manifolds can be obtained by considering the dimensional reduction of the model

34

LAURENT BAULIEU

discussed in the previous section, for Calabi–Yau four–fold, although we will shortly give the classical 7-dimensional action that one can directly quantize on such manifolds. Starting from the ATQFT for CY4 , one can set the fourth component of the gauge connection to A¯4 = A7 −iL, where 7 is the direction along the circle S1 and L a real scalar ﬁeld. (L will shortly have a special interpretation in seven dimensions.) Moreover, the ¯ = ¯1, . . . , ¯4. dimensional reduction impose to set i(∂4 − ∂¯4 )AM¯ = ∂8 AM¯ = 0 for any M Then one gets from the classical action (5.1) the following action

√

S= Σ6 ×S 1

¯ np¯ g d7 x Tr 2 m¯ (Bm¯ + B F ) ¯ n Fp¯¯ m ¯ n ¯ p ¯ 4

(5.19)

with Fp¯¯4 = Fp7 ¯ − iDp¯L and Bm ¯ = Bm ¯¯ 4 . In this section we deﬁne the complex three– form Ω0,3 on Σ6 starting from the (normalized) complex four–form in eight dimensions as m¯ ¯ np¯ = m¯ ¯ np¯¯ 4 . We thus consider the following classical action: Icl (A, B0,2 ) =

M7

Ω3,0 ∧ Tr(B ∧ F )

(5.20)

¯ Here the coordinates on the manifold are z m , z m for Σ6 and the periodic real coordinate 7 x for the circle. The only components of the 2-form that have a relevant propagation are Bm¯ ¯ n , Bm7 ¯ . The covariant quantization of Bm¯ ¯ n , Bm ¯ (= Bm7 ¯ ) and A requires seven topological ¯n ¯ , κm , κ). Modulo the ordinary gauge symmetry, we have indeed seven antighosts (κm¯ freedoms for the classical action (5.19) associated to the topological ghosts (χm ¯ , χ) for the ﬁelds (Bm¯ ¯ n , Bm ¯ ) and the ghost Ψm for the ﬁeld Am . The relevant invariance is Bm¯ ¯ n ∼ Bm¯ ¯ n + D[m ¯ ∼ Bm ¯ + Dm χ Am ∼ Am + Ψm , Am ¯ ∼ Am ¯ , A7 ∼ A7 , ¯ χn] , Bm modulo ordinary gauge transformations. Moreover, in the BRST complex for the ﬁeld B ¯ for the ”gauge symit will appear also three commuting scalar ghost of ghosts (Φ, L, Φ), metries” of the topological ghosts and antighosts of B2 . All these ﬁelds are conveniently displayed as elements of the following pyramidal diagram:

From the point of view of the dimensional reduction of the CY4 theory, it is interesting to observe that the medium ghost of ghost L of the B ﬁeld can be identiﬁed with the component A8 of the eight dimensional gauge ﬁeld. The BRST transformations of the

QUANTIZATION OF HOLOMORPHIC FORMS

35

ﬁelds can be read from the following Batalin–Vilkoviski action : √ 7 ¯ np¯ g d x Tr 2 m¯ (Bm¯ S = ¯ n Fp¯¯ ¯ Fn ¯ p¯) 4 + Bm Σ6 ×S 1 ∗ m¯ ¯n

+ B

∗ p¯ 1 (D[m ¯ n ] + 2 m¯ ¯ np¯[ B , Φ]) ¯ χn ¯ ] − [c, Bm¯

¯ ∗ n ¯ p¯ 1 (2D[m +∗ B m ¯ ] + 2 m¯ ¯ np¯[ B , Φ]) ¯ χ7] − 2[c, Bm

¯ ∗ +∗ χm (Dm ¯ Φ − [c, χm ¯ ]) + χ(D¯ 4 Φ − [c, χ]) ∗ m ∗ m ¯ ∗ 7 ∗ m + A (Ψm + Dm c) + A D c ¯ m ¯ + A D7 c − Ψ [c, Ψm ] ¯ η +∗ Lη +∗ c(− 12 [c, c]) +∗ Φ¯ m¯ ¯n m ¯ ∗ + 2∗ κm +∗ κm¯ ¯ nb ¯ b + κb

(5.21)

This action is actually the dimensional reduction of action (5.3). However, the interpretation of the ghost of ghost system is quite different. Notice that the vectorial part of the BRST transformation for the ﬁeld Bm ¯ is δBm ¯ = Dm ¯ χ − D¯ ¯ = Dm ¯ χ − D7 χm ¯ 4 χm

(5.22)

The variation of the action due to the ﬁrst term in (5.22) simply vanishes after integration by parts because of the identity (4.3). In the context of the dimensional reduc¯ = ¯ tion, this identity can be read as the M 4 component of the eight–dimensional one ¯N ¯ P¯ Q ¯ M DM¯ FN¯ Q¯ = 0. The variation associated to the last term in (5.22), together with

that coming from the variation of the ﬁeld Bm¯ ¯ n , QBm¯ ¯ n = D[m ¯ χn ¯ ] , gives after integration ¯ =¯ by parts the other three components M 1, ¯ 2, ¯ 3 of the above eight–dimensional identity. The topological freedom of the classical action (5.19) can be ﬁxed by choosing seven independent gauge functions. The ﬁrst six can be directly obtained from the dimensional reduction of (5.4) Bm¯ ¯n Bm ¯

= = =

p¯¯ 4 1 ¯ n − m¯ ¯ np¯F ) 2 (Fm¯ p¯ ¯q 1 1 ¯ p¯ ¯q F ) ¯¯ 4 − 2 n 2 (Fn p¯ ¯q 1 1 ¯ 7 − iDn ¯ L − 2 n ¯ p¯ ¯q F ) 2 (Fn

(5.23)

while the seventh one corresponds to the imaginary part of the complex gauge–ﬁxing condition (5.6) n 1 m¯ Fm¯n + D7 L = 0 (5.24) 2J Notice in fact that the reduction on a manifold of the kind Σ6 × S 1 breaks the complex group of gauge invariance of the eight–dimensional action (5.1) to the unitary group. The residual gauge invariance under this group can be ﬁxed by the ordinary transversality ¯ m 7 Am condition ∂ m ¯ + ∂ Am + ∂ A7 = 0. The gauge–ﬁxing fermion corresponding to the conditions (5.23) and (5.24) is

¯n p7 ¯ 1 Bm¯ + iDp¯L) Z = κm¯ ¯ n − 2 Fm¯ ¯ n − m¯ ¯ np¯(F 3 ¯ p¯ ¯q 1 (F − − iD L −

F ) Bm +κm ¯ n ¯7 n ¯ ¯ p¯ ¯q 2 n 2 ¯ +κ 2(J mn Fmn ¯ + D7 L) − b ¯ 7 ¯ m χm +Φ(D ¯ + D χ)

(5.25)

36

LAURENT BAULIEU

The bosonic part of the gauge–ﬁxed action, which can be obtained as usual by using the BV equation (3.12) and enforcing the gauge conditions (5.23) and (5.24) by integration on the Lagrangian multipliers, reads 3 √ ¯n m¯ n ¯ d7 x g Tr − F m¯ Fm¯ Fm¯n |2 − 12 F m7 Fm7 S g.f. = ¯ n + |J ¯ 2 M7 ¯ 7 7 m ¯ 7 7 ¯ m +Φ(D Dm (5.26) ¯ + D D )Φ + L(D Dm ¯ + D D )L and can be identiﬁed with the bosonic part of the seven dimensional Super Yang–Mills on M7 = Σ6 × S 1 . Concerning the fermionic sector, we have 8 topological ghosts, Ψm ¯ , χm ¯ , χ and η, and ¯n ¯ , κm , κ and η¯. The mapping with spinors can be obtained 8 topological antighosts κm¯ from the dimensional reduction of the eight–dimensional mapping (5.13) and (5.14). After the dimensional reduction to Σ6 × S 1 , the two covariantly–constant eight–dimensional chiral spinors (ζ1 , ζ2 ) are identiﬁed with the unique covariantly constant Majorana spinor ξ of the Calabi–Yau three–fold Σ6 . On the other side, the eight dimensional spinor λ yields two seven dimensional Majorana spinors (λ1 , λ2 ). This results in the mapping Ψm → ξΓm λ1 , η → ξΓ7 λ1 ¯ χm → mnp ξΓnp λ1 , χ → mnp ξΓmnp λ1

(5.27)

for the topological ghosts and κmn → ξΓmn λ2 , κm → ξΓm7 λ2 κ → ξλ2 , η¯ → mnp Γmnp7 λ2

(5.28)

for the topological antighosts. By this mapping we can identify the topological action (5.19) gauge ﬁxed with the conditions (5.23) and (5.24) as the twisted version of N = 2 seven dimensional Super Yang–Mills6 . The observables of the topological model can be identiﬁed with the dimensional reduction of the eight-dimensional cocycles. 5.3. MANIFOLDS WITH SU (4) GROUP STRUCTURE AND SELF–DUAL YANG–MILLS IN EIGHT–DIMENSIONS + , On a K¨ahler manifold with a SU (4) group structure we can choose B4,2 = J ∧ J ∧ B2,0 + where J is as usual the K¨ahler (1, 1) form. Here the 2-form B2,0 is self dual in the indices [mn]. Self duality is deﬁned from a SU (4) invariant 4-form Ω4,0 , which is globally well deﬁned, but not necessarily closed. It counts for 3 degrees of freedom, according to the SU (4) independent decomposition of a 2-form in 8 dimensions:

28 = 6 ⊕ ¯ 6 ⊕ 15 ⊕ 1

(5.29)

and a further decomposition 6 as 6 = 3 ⊕ 3, using the mnpq tensor. 6

As discussed for the four dimensional case, in order to recover the Yukawa couplings and the ¯ Φ]2 typical of the N = 2 SYM one should slightly modify the quartic term in the potential [Φ, gauge–ﬁxing fermion (5.25).

QUANTIZATION OF HOLOMORPHIC FORMS

37

The novelty of this case is that neither K¨ahler (1, 1)-form nor Ω4,0 are necessarily closed. Both the forms can also be rewritten in terms of spinors, which correspondingly are not covariantly constant with respect to the usual spin connection, but only with respect to a modiﬁed connection including torsion terms. The corresponding classical action is a generalisation of that in (3.3): + + + J ∧ J ∧ Tr(B2,0 ∧ F0,2 ) (5.30) Ik−BF (A, B2,0 , B0,2 ) = M8

The symmetries of this action are QAm = Ψm + QBmn

QΨm

QAm ¯ = Dm ¯c

+ −[c, Bmn ]

= =0,

QΦ = −[c, Φ] ,

,

QΨm ¯

+ + ∗ mn + QBm¯ − [c, Bm¯ , Φ] ¯ n + = D[m ¯ Ψn ¯] ¯ n] + [ B = Dn¯ Φ − [c, Ψn¯ ]

Qc = − 12 [c, c]

(5.31)

The quantization of this action can be worked out using the BV formalism, and is very similar to that already discussed in [4]. + + , B0,2 , A1,0 , A0,1 ). Indeed, Ψm , We have 8 freedoms for gauge-ﬁxing the system (B2,0 Ψm ¯ and c have respectively 4,4 and 1 components, but Ψm ¯ has a ghost of ghost symmetry with ghost of ghost Φ, so it only counts for 3=4-1 freedoms. We can choose the following 7 gauge–ﬁxing conditions in the gauge covariant sector: + Bmn Bm¯ ¯ n+

J m¯n Fm¯n

¯

m¯ ¯ n+ ll pp¯ = mnlpm¯ J J ¯ n¯ lp¯F = 0

=

(5.32)

¯ plus the ordinary transversality condition for the gauge ﬁeld, (∂ m Am + ∂ m Am ¯ = 0). The + transformation law of Bmn implies that a gauge ﬁxing for Ψm ¯ must be also done, with a gauge function : ¯ Ψm (5.33) Dm ¯ =0

The Q-invariant gauge ﬁxing of the action with these functions is standard, and reproduces the action as in [4], that is the twisted form of the eight–dimensional Super Yang–Mills theory. The classical action (5.30) can be considered as an eight–dimensional generalization of the self–dual Yang–Mills in four dimensions, in particular of its realisation studied in [22]. We should notice that we are at the extreme point of the deﬁnition of an ATQFT. The action is the sum of a d-closed term and a Q-exact term. Thus, there is a ring of observables deﬁned from the cohom*ology of the BRST operator. However, the classical action is actually completely dependent of the metrics of the manifold, since it depends on both the K¨ahler form and the complex form at the same time.

Acknowledgements This work was done in collaboration with A. Tanzini. L.B. is grateful to I.M. Singer for interesting discussions.

38

LAURENT BAULIEU

References 1.

L. Baulieu, Gravitational Topological Quantum Field Theory Versus N = 2, D = 8 Supergravity and its lift to N = 1 D = 11 Supergravity, JHEP 0404 (2004) 044, hep-th/0304221 ; L. Baulieu, M. Bellon and A. Tanzini, “Supergravity and the knitting of the Kalb-Ramond two-form in eight-dimensional topological gravity,”, Phys. Lett. B 565 (2003) 211 [arXiv:hepth/0303165]; “Eight-dimensional topological gravity and its correspondence with supergravity,” Phys. Lett. B 543 (2002) 291 [arXiv:hep-th/0207020]. 2. D. Anselmi and P. Fre, “Twisted N=2 supergravity as topological gravity in four-dimensions,” Nucl. Phys. B 392 (1993) 401 [arXiv:hep-th/9208029]; L. Baulieu and A. Tanzini, Topological gravity versus supergravity on manifolds with special holonomy, JHEP 0203 (2002) 015, [hepth/0201109]; P. de Medeiros and B. Spence, “Four-dimensional topological Einstein-Maxwell gravity,” Class. Quant. Grav. 20 (2003) 2075 [arXiv:hep-th/0209115]. 3. L. Baulieu, “Going down from a 3-form in 16 dimensions”, Phys.Lett. B544 (2002) 367-373, hep-th/0207184. 4. L. Baulieu, H. Kanno and I.M. Singer, Special Quantum Field Theories In Eight And Other Dimensions, Commun.Math.Phys.194 (1998) 149, [hep-th/9704167] and [hep-th/9705127]; L. Baulieu and C. Laroche, On Generalized Self-Duality Equations Toward Supersymmetric Quantum Field Theories of Forms, Mod.Phys Lett. A13 (1998) 1115, [hep-th/9801014]. 5. B. S. Acharya, J. M. Figueroa-O’Farrill, B. Spence and M. O’Loughlin, “Euclidean Dbranes and higher-dimensional gauge theory,” Nucl. Phys. B 514 (1998) 583 [arXiv:hepth/9707118]; “Higher-dimensional analogues of Donaldson-Witten theory,” Nucl. Phys. B 503 (1997) 657 [arXiv:hep-th/9705138]. 6. M. Blau and G. Thompson, “Euclidean SYM theories by time reduction and special holonomy manifolds,” Phys. Lett. B 415 (1997) 242 [arXiv:hep-th/9706225]. 7. A. Iqbal, N. Nekrasov, A. Okounkov and C. Vafa, “Quantum foam and topological strings,”, hep-th/0312022. 8. A. Neitzke and C. Vafa, “N = 2 strings and the twistorial Calabi-Yau,” hep-th/0402128. 9. N. Nekrasov, H. Ooguri and C. Vafa, “S-duality and topological strings,” JHEP 0410 (2004) 009, hep-th/0403167. 10. A. Kapustin, “Gauge theory, topological strings, and S-duality,” JHEP 0409 (2004) 034 hepth/0404041. 11. H. Ooguri, A. Strominger and C. Vafa, “Black hole attractors and the topological string,” hep-th/0405146. 12. R. Dijkgraaf, S. Gukov, A. Neitzke and C. Vafa, Topological M-theory as Uniﬁcation of Form Theories of Gravity , hep-th/0411073. 13. A. Gerasimov and S. L. Shatashvili, Towards Integrability of Topological Strings I: Threeforms on Calabi-Yau manifolds, hep-th/0409238. 14. P. A. Grassi and P. Vanhove, Topological M theory from pure spinor formalism, hepth/0411167. 15. N. Hitchin, Stable forms and special metrics, “Proceedings of the Congress in memory of Alfred Gray”, (eds M. Fernandez and J. Wolf), AMS Contemporary Mathematics Series, math.DG/0107101; The geometry of three-forms in six and seven dimensions, math.DG/0010054. 16. A. S. Cattaneo, P. Cotta-Ramusino, F. Fucito, M. Martellini, M. Rinaldi, A. Tanzini and M. Zeni, “Four-dimensional Yang-Mills theory as a deformation of topological BF theory,” Commun. Math. Phys. 197 (1998) 571 [arXiv:hep-th/9705123]. 17. L. Baulieu and S. L. Shatashvili, “Duality from topological symmetry,” JHEP 9903 (1999) 011 [arXiv:hep-th/9811198]. 18. D. Maulik, N.Nekrasov, A. Okounkov and R. Pandharipande, math.AG/0312059, math.AG/0406092. 19. M. J. Plebanski, “On the separation of Einsteinian substructures,” J. Math. Phys. 18 (1977), 2511; G. T. Horowitz, “Exactly Soluble Diffeomorphism Invariant Theories,” Commun. Math. Phys. 125 (1989) 417; C. G. Torre, “Perturbations Of Gravitational Instantons,” Phys. Rev. D 41 (1990) 3620; M. Abe, A. Nakamichi and T. Ueno, “Gravitational instantons and moduli

QUANTIZATION OF HOLOMORPHIC FORMS

20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30.

31. 32. 33. 34. 35. 36. 37. 38. 39. 40.

39

spaces in topological two form gravity,” Phys. Rev. D 50, 7323 (1994), hep-th/9408178. work in progress. E. Witten, “(2+1)-Dimensional Gravity As An Exactly Soluble System,” Nucl. Phys. B 311 (1988) 46. W. Siegel, “N=2, N=4 string theory is selfdual N=4 Yang-Mills theory,” Phys. Rev. D 46 (1992) 3235. L. Baulieu, I. M. Singer, Nucl.Phys.Proc.Suppl. B5 (1988) 12. A. D. Popov, “Holomorphic analogs of topological gauge theories,” Phys. Lett. B 473 (2000) 65 [arXiv:hep-th/9909135]; T. A. Ivanova and A. D. Popov, “Dressing symmetries of holomorphic BF theories,” J. Math. Phys. 41 (2000) 2604 [arXiv:hep-th/0002120]. J. S. Park, N=2 topological Yang-Mills theory on compact Kahler surfaces, Commun. Math. Phys. 163 (1994) 113 [arXiv:hep-th/9304060]; J. S. Park, Holomorphic Yang-Mills theory on compact Kahler manifolds, Nucl. Phys. B 423 (1994) 559 [arXiv:hep-th/9305095]. A. Johansen, Twisting of N=1 SUSY Gauge Theories and Heterotic Topological Theories , Int.J.Mod.Phys. A10 (1995) 4325-4358, hep-th/9403017; E. Witten, Supersymmetric Yang-Mills Theory On A Four-Manifold, J.Math.Phys. 35 (1994) 5101-5135, hep-th/9403195. R. Dijkgraaf, J. S. Park and B. J. Schroers, N = 4 supersymmetric Yang-Mills theory on a Kaehler surface, arXiv:hep-th/9801066. C. Hofman and J. S. Park, “Cohom*ological Yang-Mills theories on Kaehler 3-folds,” Nucl. Phys. B 600 (2001) 133 [arXiv:hep-th/0010103]. L. Baulieu, Algebraic Construction Of Gauge Invariant Theories, Particles and Fields, Cargese, Published in Cargese Summer Inst. 1983, 1; On Forms With Non-Abelian Charges And TheirDualities, Phys.Lett. B441 (1998) 250 , hep-th/9808055; L. Baulieu, E. Rabinovici, Selfduality And New TQFTs For Forms, JHEP 9806 (1998) 0068. H.B. Lawson and M.-L. Michelsohn, Spin geometry, Princeton University Press, Princeton, New Jersey, 1989. E. Witten, “Chern-Simons gauge theory as a string theory,” Prog. Math. 133 (1995) 637 [arXiv:hep-th/9207094]. A. S. Cattaneo, P. Cotta-Ramusino, J. Frohlich and M. Martellini, “Topological BF theories in three-dimensions and four-dimensions,” J. Math. Phys. 36 (1995) 6137 [arXiv:hepth/9505027]. D. Ray and A. Singer, “Analytic torsion for complex manifdlds,” Ann. Math. 98 (1973) 154– 177. S. K. Donaldson and R. P. Thomas, “Gauge theory in higher dimensions,”, Prepared for Conference on Geometric Issues in Foundations of Science in honor of Sir Roger Penrose’s 65th Birthday, Oxford, England, 25-29 Jun 1996; R. P. Thomas, PhD thesis, Oxford, 1997. I. Frenkel, B. Khesin and A. Todorov, Complex counterpart of Chern–Simons–Witten theory and holomorphic linking, preprint. S. Axelrod and I. M. Singer, “Chern-Simons perturbation theory,”, hep-th/9110056; J. Diff. Geom. 39 (1994) 173 [arXiv:hep-th/9304087]. P. de Bartolomeis and G. Tian, J. Diff. Geom. 43 (1996) 231. V.G. Turaev and O. Yu. Viro, “Sum Invariants of 3–Manifolds and Quantum 6j–symbols”, Topology 31 (1992) 865. N. Reshetikhin and V.G. Turaev, “Invariants of 3–Manifolds via Link Polynomials and Quantum Groups”, Invent. Math. 103 (1991) 547.

LOEWNER CHAINS

MICHEL BAUER AND DENIS BERNARD Service de Physique Th´eorique de Saclay CEA-Saclay, 91191 Gif-sur-Yvette, France

Abstract. These lecture notes on 2D growth processes are divided in two parts. The ﬁrst part is a non-technical introduction to stochastic Loewner evolutions (SLEs). Their relationship with 2D critical interfaces is illustrated using numerical simulations. Schramm’s argument mapping conformally invariant interfaces to SLEs is explained. The second part is a more detailed introduction to the mathematically challenging problems of 2D growth processes such as Laplacian growth, diffusion limited aggregation (DLA), etc. Their description in terms of dynamical conformal maps, with discrete or continuous time evolution, is recalled. We end with a conjecture based on possible dendritic anomalies which, if true, would imply that the Hele-Shaw problem and DLA are in different universality classes.

Growth phenomena are ubiquitous in the physical world at many scales, from crystals to plants to dunes and larger. They can be studied in many frameworks, deterministic of probabilistic, in discrete or continuous space and time. Understanding growth is usually a very difﬁcult task. This is true even in two dimensions, the case we concentrate on in these notes. Yet two dimensions is a highly favorable situation because it allows to make use of the power of complex analysis in one variable. In many interesting cases, the growing object in two dimensions can be seen as a domain, i.e. a contractible open subset of the Riemann sphere (the complex plane with a point at inﬁnity added). A deep theorem of Riemann asserts that such a domain, whatever complicated and fancy, is conformally equivalent to a simple reference domain, which is usually taken as the upper-half plane or the unit disk. This conformal equivalence is unique once an appropriate normalization, which may depend on the growth problem at hand, has been chosen. Cauchy’s theorem allows to write down an integral representation for the conformal map as an integral along the boundary of the reference domain, involving a density. This density is time dependent. Then the time derivative of the conformal map has an analogous representation and a nice way to specify the growth rule is often directly on this density. This leads to the concept of Loewner chains, which is the central theme of these notes. We shall illustrate Loewner chains in several situations. 41 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 41–77. © 2005 Springer. Printed in the Netherlands.

42

MICHEL BAUER AND DENIS BERNARD

Our aim is to give a pedagogical introduction to a beautiful subject. We wanted to show that it leads to many basic mathematical structures whose appearance in the growth context is not so easy to foresee, like Brownian motion, integrable systems and anomalies to mention just a few. We have also tried to stress that some growth processes have rules which are easy to simulate on the computer. A few minutes of CPU are enough to get an idea of the shape of the growing patterns, to be convinced that something interesting and non trivial is going on, and even sometimes to get an idea of fractal dimensions. This is of course not to be compared with serious large scale simulations, but it is a good illustration of the big contrast between simple rules, complex patterns and involved mathematical structures. However, other growth models, and among those some have been conjectured to be equivalent to simple ones, have resisted until recently to precise numerical calculations due to instabilities. To avoid any confusion, let us stress that being able to describe a growth process using tools from complex analysis and conformal geometry does not mean that the growth process itself is conformally invariant at all. Conformal invariance of the growth process itself puts rather drastic conditions on the density that appears in the Loewner chain. This is illustrated by the ﬁrst part of these notes, which deals with conformally invariant interfaces and their relation to stochastic Loewner evolutions. This part is an elaboration of the the main points developed during the lectures. The second part is an introduction to a larger class of processes describing the growth of possibly random fractal planar domains and a review of some of their basic properties. Due to lack of time, this part was not presented during the lectures. The study of the continuum limit of non-intersecting curves on the lattice has been a subject of lasting interest both in mathematics and in physics. A famous example is given by self-avoiding random walks. The motivation comes from combinatorics, but also from statistical mechanics. Two dimensions is most interesting because a non-intersecting curve is the boundary between two domains and can very often be interpreted as an interface separating two coexisting phases. At a critical point and for short range interactions, such interfaces are expected to be conformally invariant. The argument for that was given two decades ago in the seminal paper on conformal ﬁeld theory [1]. The rough idea is the following. At a critical point, a system becomes scale invariant. If the interactions on the lattice are short range, the model is described in the continuum limit by a local ﬁeld theory and scale invariance implies that the stress tensor is traceless. In two dimensions this is enough to ensure that the theory transforms simply –no dynamics is involved, only pure kinematics– when the domain where it is deﬁned is changed by a conformal transformation. The local ﬁelds are classiﬁed by representations of the inﬁnite dimensional Virasoro algebra and this dictates the way correlation functions transform. This has led to a tremendous accumulation of exact results using conformal ﬁeld theory (CFT). A situation that is well under control is that of unitary minimal models. The Hilbert space of the system splits as a ﬁnite sum of representations of the Virasoro algebra, each associated to a (local) primary ﬁeld, and the corresponding correlation functions can be described rather explicitly. The study of non-local objects like interfaces at criticality has not seen such a systematic development and only isolated though highly artful results [2, 3, 12] have been

LOEWNER CHAINS

43

discovered using conformal ﬁeld theory techniques. Non local objects like interfaces are not classiﬁed by representations of the Virasoro algebra but the reasoning that led O. Schramm to the crucial breakthrough [4], i.e. the deﬁnition of stochastic Loewner evolutions (SLEs), rests on a fairly obvious but cleverly exploited statement of what conformal invariance means for an interface. Surprisingly it allows to turn this problem into growth problem, something which looks natural only a posteriori. For some years, probabilistic techniques have been applied to interfaces, leading to a systematic understanding that was lacking on the CFT side. A sample, surely biased by our ignorance, can be found in refs.[5, 6, 7, 8, 9]. There is now a satisfactory understanding of interfaces in the continuum limit. However, from a mathematical viewpoint, giving proofs that a discrete interface on the lattice has a conformally invariant limit remains a hard challenge and only a handful of cases has been settled up to now. There is now a good explanation of the –initially mysterious– relationship between SLE and CFT. This was one of the main topics of the lectures. Though this aspect is mostly due to the work of the present authors ([10], but see also [11]), we have decided to be very sketchy. The interested reader is referred to the literature. Stochastic Loewner evolution is a simple but particularly interesting example of growth process for which the growth is local and continuous so that the resulting set is a curve without branching. Of course other examples have been studied in connection with 2d physical systems. The motivations are sometimes very practical. For instance, is it efﬁcient to put a pump in the center of oil ﬁlm at the surface of the ocean to ﬁght against pollution? The answer has to do with the Laplacian growth or Hele-Shaw problem [14]. The names diffusion limited aggregation [15], see Figure (1), and dielectric breakdown [16] speak for themselves. Various models have been invented, sometimes with less physical motivation, but in order to ﬁnd more manageable growth processes. These include various models of iterated conformal maps [17], etc. See refs.[18] for recent reviews. As mentioned before, in most cases the shape of the growing domains is encoded in a uniformizing conformal map whose evolution describes the evolution of the domain. The dynamics can be either discrete or continuous in time, it can be either deterministic or stochastic. But the growth process is always described by a simple generalization of the Loewner equation called a Loewner chain. The general understanding of these Loewner chains is still embryonic, in strong contrast with the case of Loewner evolutions, which deal with local growth. For many problems the most basic question are unanswered. For instance, as we shall explain below, Laplacian growth is at the same time a completely integrable system and an ill-posed problem because it develops singularities in ﬁnite time. Hydrodynamics gives a natural regularization via the introduction of surface tension. It has been argued for some time that in the limit of vanishing surface tension one retrieves a model which is in the same universality class as diffusion limited aggregation. But the experimental and numerical evidence is inconclusive and there is no consensus. In fact, at the end of these notes we shall give an argument suggesting that they belong to different universality classes. This is very conjectural, but as should be clear from this introduction, non-local growth processes are certainly a source of interesting and challenging problems.

44

MICHEL BAUER AND DENIS BERNARD

Figure 1. An example of DLA cluster obtained by iterating conformal maps.

1. Critical interfaces and stochastic processes The following pedagogical introduction to conformally invariant random curves breaks in two part. The ﬁrst is a list of examples of critical interfaces on the lattice and some of their properties. The second is a presentation of O. Schramm’s derivation of SLE. In this part the upper-half plane is used as a reference geometry. 1.1. THREE EXAMPLES AND A GENERALITY Let us start with three examples. Our aim is to explain their deﬁnitions, to show a few samples and, in the ﬁrst two cases, to give a numerical estimate of the fractal dimension. The ﬁrst, loop-erased random walks, belongs to the realm of “pure combinatorics”. The second, percolation, is at the frontier between pure combinatorics and statistical mechanics because it arises very naturally as the domain boundary of a statistical mechanics model, but the Boltzmann weights are trivial. It is however a limiting case of a family of models with non trivial Boltzmann weights and, from that point of view, quantities relevant for percolation are obtained by exploring an inﬁnitesimal neighborhood around the trivial Boltzmann weights. The third example, the Ising model, is deeply rooted in statistical mechanics. We shall end this section by abstracting a crucial property of interfaces which allows to make an efﬁcient use of conformal invariance and derive the Markov property for SLEs. 1.1.1. Loop-erased random walks This example is purely of combinatorial nature. A loop-erased random walk is a random walk with loops erased along as they appear. More formally, if X0 , X1 , · · · , Xn is a ﬁnite sequence of abstract objects, we deﬁne the associated loop-erased sequence by the following recursive algorithm. Until all terms in the sequence are distinct,

LOEWNER CHAINS

45

Step 1 Find the couple (l, m) with 0 ≤ l < m such that the terms with indexes from 0 to m − 1 are all distinct but the terms with indexes m and l coincide. Step 2 Remove the terms with indexes from l + 1 to m, and shift the indexes larger than m by l − m to get a new sequence. Let us look at two examples. For the “month” sequence j, f, m, a, m, j, j, a, s, o, n, d, the ﬁrst loop is m, a, m, whose removal leads to j, f, m, j, j, a, s, o, n, d, then j, f, m, j, leading to j, j, a, s, o, n, d, then j, j leading to j, a, s, o, n, d where all terms are distinct. For the “reverse month” sequence d, n, o, s, a, j, j, m, a, m, f, j, the ﬁrst loop is j, j, leading to d, n, o, s, a, j, m, a, m, f, j, then a, j, m, a leading to d, n, o, s, a, m, f, j. This shows that the procedure is not “time-reversal” invariant. Moreover, terms that are within a loop can survive: in the second example m, f , which stands in the j, m, a, m, f, j loop, survives because the ﬁrst j is inside the loop a, j, m, a which is removed ﬁrst. A loop-erased random walk is when this procedure is applied to a (two dimensional for our main interest) random walk. This is very easy to simulate. Fig.2 represents a looperased walk of 200 steps. The thin lines build the shadow of the random walk (where shadow means that we do not keep track of the order and multiplicity of the visits) and the thick line is the corresponding loop-erased walk. The time asymmetry is clearly visible and allows to assert with little uncertainty that the walk starts on the left.

Figure 2. A loop-erased random walk with its shadow.

To ﬁt with the general SLE framework, let us restrict to loop-erased random walks in the upper-half plane. There are a few options for the choice of boundary conditions. A ﬁrst choice is to consider reﬂecting boundary conditions on the real axis for the random walk. Another choice is annihilating boundary conditions: if the random walk hits the real axis, one forgets everything and starts anew at the origin. Why this is a natural boundary condition has to wait until section 1.1.4. Due to the fact that on a two-dimensional lattice a random walk is recurrent (with probability one it visits any site inﬁnitely many times), massive rearrangement occur with

46

MICHEL BAUER AND DENIS BERNARD

probability one. This means that if one looks at the loop-erased random walk associated to a given random walk, it does not have a limit in any sense when the size of the random walk goes to inﬁnity. Let us illustrate this point. The samples in ﬁg.3 were obtained with reﬂecting boundary conditions. It takes 12697 random walk steps to build a loop-erased walk of length 633, but step 12698 of the random walk closes a long loop, and then the ﬁrst occurrence of a loop-erased walk of length 634 is after 34066 random walk steps. Observe that in the mean time most of the initial steps of the loop-erased walk have been reorganized. 200

100

150

100 50

50

-100

-50

-100

100

Figure 3. On the left: a large loop is about to be created. On the right: the massive rearrangement to go from 633 to 634 steps.

However, simulations are possible because when the length of the random walk tends to inﬁnity, so does the maximal length of the corresponding loop-erased walk with probability one. Though annihilating boundary conditions lead to remove even more parts of the random walk than the reﬂecting ones, the corresponding process can be arranged (conditioned in probabilistic jargon) to solve the problem of convergence as follows. Instead of stopping the process when the loop-erased walk has reached a given length, one can stop it when it reaches a certain altitude, say n, along the y-axis. Whatever the corresponding random walk has been, the only thing that matters is the last part of it, connecting the origin to altitude n without returning to altitude 0. Moreover, the ﬁrst time the loop-erased walk reaches altitude n is exactly the ﬁrst time the random walk reaches altitude n. Now a small miracle happens: if a 1d symmetric random walk is conditioned to reach altitude n before it hits the origin again, the resulting walk still has the Markov property. It is a discrete equivalent to the 3d Bessel process (a Bessel process describes the norm of a Brownian motion, however no knowledge of Bessel processes is needed here, we just borrow the name). When at site m, 0 < m < n, the probability to go to m ± 1 is (1 ± 1/m)/2, independently of all previous steps. Observe that there is no n dependence so that we can forget about n, i.e. let it go to inﬁnity. The discrete 3d Bessel process is not recurrent and tends to inﬁnity with probability one: for any altitude l there is with probability one a time after which the discrete 3d Bessel process remains above l

LOEWNER CHAINS

47

for ever. Henceforth, we choose to simulate a symmetric simple random walk along the x axis and the discrete 3d Bessel process along the y-axis and we erase the loops of this new process. This leads to the convergence of the loop-erased walk and numerically to a more economical simulation. Fig.4 is a simulation of about 105 steps, both for reﬂecting and annihilating boundary conditions. At ﬁrst glance, one observes in both cases similar simple (no multiple points) but irregular curves with possibly fractal behavior. 10000

8000

6000

4000

2000

0 -4000

-2000

2000

4000

6000

8000

Figure 4. A sample of the loop-erased random walk for the two boundary conditions.

To estimate the Hausdorff dimensions in both cases, we have generated samples of random walks, erased the loops and made the statistics of the number of steps S of the resulting walks compared to a typical length L (end-to-end distance for reﬂecting boundary conditions, maximal altitude for annihilating boundary conditions). In both cases, one observes that S ∝ Lδ and a modest numerical effort (a few hours of CPU) leads to δ = 1.25 ± .01. This is an indication that the boundary conditions do not change the universality class. To get an idea of how small the ﬁnite size corrections are, observe ﬁg.5. The altitude was sampled from 24 to 213 . The best ﬁt gives a slope 1.2496 and the ﬁrst two points already give 1.2403. As recalled in the introduction, it is believed on the basis of intuitive arguments that in two dimensions scale invariance implies conformal invariance, providing there are no long range interactions. What does this absence of long range interactions mean for loop-erased random walks? Clearly along the loop-erased walk there are long range correlations, if only because a loop-erased random walk cannot cross itself. However, the relevant feature for the intuitive argument is that, in the underlying 2d physical space, interactions are

48

MICHEL BAUER AND DENIS BERNARD

10

5

2

4

6

8

10

Figure 5. The logarithm of the average length of the loop-erased random walk versus logarithm of the maximum altitude. The numerical results are the circles, the straight line is the linear regression, the error bars are shown.

indeed short range. At each time step, the increment of the underlying random walk is independent of the rest of the walk, and the formation of a loop to be removed is known from data at the present position of the random walk. From the analytical viewpoint, the loop-erased random walk is one of the few systems that has been proved to have a conformally invariant distribution in the continuum limit, the fractal dimension being exactly 5/4. A naive idea to get directly a continuum limit representation of loop-erased walks would be to remove the loops from a Brownian motion. This turns out to be impossible due to the proliferation of overlapping loops of small scale. However, the SLE2 process, to be deﬁned later, gives a direct deﬁnition. In fact, it is the consideration of loop-erased random walks that led Schramm [4] to propose SLE as a description of interfaces. 1.1.2. Percolation To deﬁne a random interface for percolation, one possibility is to pave the upper-half plane with regular hexagons. Each one is, independently of the others, occupied or empty with probability 1/2, with the exception of the ones along the real axis, which are all empty on the positive real axis and occupied on the negative real axis, see ﬁg.6. Then a continuous interface, starting at the origin and separating occupied and unoccupied hexagons, is uniquely deﬁned. As before, this deﬁnes a simple curve on the lattice. The percolation interface has an obvious but very remarkable property that singles it out: locality. Locality means that the percolation interface does not depend on the distribution of occupied and empty sites away from itself. Equivalently, if D is a domain (i.e. in this discrete setting a connected and simply connected family of hexagons in the upper half plane) containing the origin, the law of the percolation interface in D before it hits the boundary of D for the ﬁrst time is independent of the distribution of occupied/empty sites outside D. It is the same law as for what one would deﬁne as the percolation interface in D without ever mentioning the world outside D. This observation makes the percolation interface easy to simulate. Indeed, the con-

LOEWNER CHAINS

49

O

Figure 6. The deﬁnition of the percolation interface.

struction of a percolation interface proceeds inductively, as shown on the ﬁg.7.

Figure 7. The percolation interface as a growth process.

Fig.8 shows a few samples of increasing size. The contrast with the previous example is rather striking. Even for small samples, the percolation interface makes many twists and turns. With the resolution of the ﬁgure, the percolation interface for large samples does not look like a simple curve at all! The intuitive explanation why this does not occur for loop-erased walks is that if it comes close to itself, then with large probability a few more steps of the random walk will close a loop. Getting an estimate of the fractal dimension of the percolation interface proceeds along lines similar to the ﬁrst example. If the number of steps S of the percolation interface is compared to a typical length L one observes that S ∝ Lδ and a modest numerical effort (a few hours of CPU) leads to δ = 1.75 ± .01.

50

MICHEL BAUER AND DENIS BERNARD 100

100 50

-50

50

400

-100

100

1000

200

0 -200

500

-200

-200

200

400

600

Figure 8. Samples of the percolation interface for increasing sizes.

The percolation interface is also build by applying rules involving only a few nearby sites, and again it is expected on general non rigorous arguments that its scale invariance should imply its conformal invariance in the continuum limit. The percolation process is another one among the few systems that has been proved to have a conformally invariant distribution in the continuum limit, the fractal dimension being exactly 7/4. As suggested by numerical simulations, the continuum limit does not describe simple curves but curves with a dense set of double points, and in fact the –to be deﬁned later– SLE6 process describes not only the percolation interface but also the percolation hull [2], which is the complement of the set of points that can be joined to inﬁnity by a continuous path that does not intersect the percolation interface. 1.1.3. The Ising model Our next example makes also use of the same pictorial representation: it is the Ising model on the triangular lattice in the low temperature expansion. The spins are ﬁxed to be up on the left and down on the right of the origin along the real axis. The energy of a conﬁguration is proportional to the length of the curves separating up and down islands. The proportionality constant has to be adjusted carefully to lead to a critical system with long range correlations. This time, making accurate simulations is much more demanding. On the square lattice, the deﬁnition of the interface suffers from ambiguities, but these become less relevant for larger sample sizes. Although there is no question that the fractal dimension of the Ising interface with the above boundary conditions is 11/8 and is described by –to be deﬁned later– SLE3 ,

LOEWNER CHAINS

51

Figure 9. A sample for the critical Ising model. The bottom line, where the spins are frozen –black on the right, white on the left– is not represented. Courtesy of J. Houdayer.

a mathematical proof that a continuum limit distribution for the interface exists and is conformally invariant is still out of reach. We conclude this list of examples with two remarks. First, the pictorial representation by a gas of non-intersecting curves on the hexagonal lattice that we used in the second and the third example applies to another family of models, the O(n) models, to which similar interface considerations would apply. Second, non branching interfaces, described on the lattice by simple curves, are not the generic situation. For instance the Q = 3 states Potts model, three different phases coexist and the physical interfaces have branch points. 1.1.4. A generality Up to now, we have discussed interfaces in –lattice approximations of– the upper half plane. Let us note that they make sense for more general domains. Start with the case of the percolation or of the Ising model. In the plane, take any connected and simply connected collection of hexagons, split it into an interior and a connected boundary, and split the boundary in two connected pieces in such a way that exactly two pairs of adjacent boundary hexagons, marked a and b, carry different colors, see ﬁg.10. Observe that we allow a cut which would correspond to the beginning of the interface. Hexagons separated by the cut are not to be counted as neighbors. Then any coloring of the hexagons in the interior will lead to a well-deﬁned interface. It will also lead to a well deﬁned energy, and then the question of the distribution of the interface in such a domain is meaningful. For percolation the energy is independent of the conﬁguration. For the Ising model, it is proportional to the length of the curves separating up and down sites. Observe that –as the possible cut separates ﬁxed spins– counting interactions across the cut or not in the energy just adds a constant to it, so that it has no inﬂuence on probabilities. For the loop-erased random case, the idea is similar. One takes a simply connected piece of the square lattice with two points a and b on the boundary, again allowing some cuts, see ﬁg.11. Consider all walks from a to b that do not touch the boundary except at a before the ﬁrst step and at b after the last step and give each such walk of length

52

MICHEL BAUER AND DENIS BERNARD b

a

Figure 10. A domain on which percolation and the Ising model can be deﬁned.

b

a

Figure 11. A domain on which loop-erased random walks can be deﬁned.

l a weight 4−l . Observe that this choice is exactly the annihilating boundary condition: in the half plane geometry, a was the origin and b the point at inﬁnity and, due to the inﬁnite extension of the boundary, we had to go through a limiting procedure. Then erase the loops to get a probability distribution for loop-erased random walks from a to b in the domain. The probability for the simple symmetric random walk to hit the boundary for the ﬁrst time at b starting from a can be interpreted as the partition function for loop-erased walks. We can now go to the point we want to make, valid for all the above examples. For any of these, we use P(D,a,b) to denote the probability distribution for the interface γ[ab] from a to b in D. Suppose that we ﬁx the beginning γ[ac] of a possible interface in domain D, up to a certain point c. Then 1) we can consider the conditional distribution for the rest of the interface and 2) we can remove the beginning of the interface from the domain to create a new domain and consider the distribution of the interface in this new domain. This is illustrated on ﬁg.12. We claim that the distributions deﬁned in 1) and 2) coincide. For reasons to be explained in a moment, we call this property “locality at the interface”. In equations P(D,a,b) ( . |γ[ac] ) = P(D\γ[ac[ ,c,b) ( . ). It is obvious that these two probabilities are supported on the same set, namely simple

LOEWNER CHAINS

53

b

b c

a

c a

Figure 12. An illustration of situations 1) and 2) for the case of loop-erased walks. What is the distribution of the dotted curve in both situations ?

curves along the edges of the lattice, going from c to b in D\γ[ac[ . Let us however note that for loop-erased random walks, annihilating boundary conditions are crucial. Reﬂecting boundary conditions clearly do not work, if only because the supports do not coincide in that case. For the case of loop-erased random walks, the argument for the equivalence of 1) and 2) goes as follows. Take any random walk (possibly with loops) W0 = a, W1 , · · · , Wl = b that contributes to an interface γ[ab] which is γ[ac] followed by some γ[cb] . Let m be the largest index for which the walk visits c. Because the interface has to start with γ[ac] , the walk Wm = c, · · · , Wl = b cannot cross γ[ac[ again, so it is in fact a walk in D\γ[ac[ from c to b leading to the interface γ[cb] . The weight for the walk W0 = a, W1 , · · · , Wl = b is 4−l , i.e. simply the product of weights for the walks W0 = a, W1 , · · · , Wm = c and Wm = c = a, · · · , Wl = b. Then a simple manipulation of weights leads directly to the announced result. For the case of percolation and the Ising model, in fact more is true: we can view P(D,a,b) not only as a probability distribution for the interface, but as the full probability distribution for the colors of the hexagons and still check the identity of 1) and 2). Again, the supports are the same for 1) and 2), namely any conﬁguration of the colors, except that the colors on both sides of γ[ac] are ﬁxed. For the case of percolation, the colors are independent of each other so the identity of 1) and 2) is clear. For the Ising model, the difference is that the conditional probabilities in 1) take into account the interactions between the colors along the interface, whereas the probability in 2) does not take into account the interactions between the spins along the cut left by the removal of the interface. However, as already mentioned above, the corresponding colors are ﬁxed anyway, so the Boltzmann weights for the conﬁgurations that are in the support of 1) or 2) differ by a multiplicative constant, which disappears when probabilities are computed. This argument extends immediately to systems with only nearest neighbor interactions. They can be deﬁned on any graph. If any subset of edges is chosen and the conﬁguration at both end of each edge is frozen, it makes no difference for probabilities to consider the model on a new graph in which the frozen edges have been deleted. Instead of looking for further generalizations, we argue more heuristically that the continuum limit for a system with short range interactions should satisfy locality at the interface. The use of this locality property –which, as should be amply evident, has noth-

54

MICHEL BAUER AND DENIS BERNARD

ing to do with conformal invariance– together with the conformal invariance assumption is at the heart of O. Schramm’s derivation of stochastic Loewner evolutions. 1.2. O. SCHRAMM’S ARGUMENT We break the argument in several pieces. The heart of the probabilistic derivation establishes two properties of conformally invariant interfaces: − the Markov property, − the stationarity of increments, see sections 1.2.3, 1.2.2 and 1.2.5. The crucial results needed for the probabilistic part, in particular the basics of Loewner evolutions, are collected in 1.2.1 and 1.2.4. The last two subsections are not directly related to the argument. They collect some basic facts on the fractal properties of SLE interfaces and on the connection with CFT. 1.2.1. Riemann’s theorem and hulls In this section we collect a few indispensable results used in the rest of the course. A domain is a non empty connected and simply connected open set strictly included in the complex plane C. Simple connectedness is a notion of purely topological nature which in two dimensions asserts essentially that a domain has no holes and is contractible. But it is a deep theorem of Riemann that two domains are always conformally equivalent, i.e. there is an invertible holomorphic map between them. For instance, the upper-half plane H is a domain. It is well known that it has a three dimensional Lie group of conformal automorphisms, P SL2 (R), that also acts on the boundary of H. There is a unique automorphism, possible followed by a transposition, that maps any triple of boundary points to any other triple of boundary points. By Riemann’s theorem, this is also true for any other domain, at least if the boundary is not too wild. Riemann’s theorem is used repeatedly in the rest of this course and is the starting point of many approaches to growth phenomena in two dimensions. For later use, we note that one can be a bit more explicit when the domain D differs only locally from the upper half plane H, that is if K = H \ D is bounded. Such a set K is called a hull. The real points in the closure of K in C form a compact set which we call KR . Let f : H → D be a conformal bijection. As the boundary of H is smooth, f has a continuous extension to R ≡ R ∪ ∞, and f −1 (R \ KR ) is a non-empty open set in R with compact complement. We call the complement the cut of f . By the Schwarz z ) for m z ≤ 0 gives an analytic extension of symmetry principle, deﬁning f (z) = f (¯ f to the whole Riemann sphere minus the cut. Across the cut, f has a purely imaginary nonnegative discontinuity which we write as a Radon-Nikodym derivative dµf /dx. One can use the P SL2 (R) automorphism group of H to ensure that f is holomorphic at ∞ and f (w) − w = O(1/w) there. This is called the hydrodynamic normalization. It involves three conditions, so there is no further freedom left. We shall denote this special representative by fK , which is uniquely determined by K: any property of fK is an intrinsic property of K. Cauchy’s theorem yields dµfK (x) 1 , (1.1) fK (w) = w + 2π R x − w

LOEWNER CHAINS

55

A quantity that plays an important role in the sequel is 1 dµfK (x), CK ≡ 2π R a positive (unless K = ∅) number called the capacity of K, which is such that fK (w) = w −CK /w +O(1/w2 ) at inﬁnity. The usefulness of capacity stems from its good behavior under compositions: if K and K are two hulls, K ∪ fK (K ) is a hull and CK∪fK (K ) = CK + CK ,

(1.2)

as seen by straightforward expansion at inﬁnity of fK ◦ fK , the map associated to K ∪ fK (K ). In particular capacity is a continuous increasing function on hulls. Anticipating a little bit, let us note immediately that giving a dynamical rule for the evolution of the ﬁnite positive measure dµfK (x) is a good way to deﬁne growth processes. 1.2.2. Conformally invariant interfaces Consider a domain D, with two distinct points on its boundary, which we call a and b. A simple curve, denoted by γ[ab] , from a to b in D is the image of a continuous one-toone map γ from the interval [0, +∞] to D ∪ {a, b} such that γ(0) = a, γ(∞) = b and γ]ab[ ≡ γ(]0, ∞[) ⊂ D. Alternatively, a simple curve from a to b is an equivalence class of such maps under increasing reparametrizations. A point on it has no preferred coordinate but is has a past and a future. If c ∈ D is an interior point, we use a similar deﬁnition for a simple curve γ[ac] from a to c in D. Note that, apart from the fact that on the lattice we could use lattice length as a parameter along the interface, we have just rephrased in the continuum –but with the same notations– what we did before in a discrete setting. Our aim is to study conformally invariant probability measures on the set of simple curves from a to b in D. There is a purely kinematical step, which demands that if h is any conformal map that sends D to another domain h(D), the measure for (h(D), h(a), h(b)) should be the image by h of the measure for (D, a, b): P(D,a,b) (γ[ab] ⊂ U ) = P(h(D),h(a),h(b)) (h(γ[ab] ) ⊂ h(U )), where P(D,a,b) (γ[ab] ⊂ U ) denotes the probability for the curve γ[ab] to remain in a subset U of D. See ﬁg.13. This condition is natural and it is the one that conformal ﬁeld theory suggests immediately. Let us note however that a totally different deﬁnition of conformal invariance is understood in the familiar statement “two dimensional Brownian motion is conformally invariant”. Observe that we could take any measure for (D, a, b) –well, with the invariance under the one parameter group of automorphisms that ﬁxes (D, a, b)– and declare that the measure in h(D) is obtained by deﬁnition by the rule above. To make progress, we need to combine conformal invariance with locality at the interface. 1.2.3. Markov property and stationarity of increments This short section establishes the most crucial properties of conformally invariant interfaces.

56

MICHEL BAUER AND DENIS BERNARD

h(b)

b γ

h

h(γ ) h(U)

U

h(a)

a Figure 13. Conformal invariance for change of domain.

Take c ∈ D and let γ[ac] be a simple curve from a to c in D. Observe that D \ γ]ac] is a domain. To answer the question “if the beginning of the interface is ﬁxed to be γ[ac] , what of the interface ?” we apply locality at the interface to is the distribution of the rest γ[cb] argue that this is exactly the distribution of the interface in D \ γ]ac] . We map this domain conformally to D via a map hγ sending b to b and c to a, see ﬁg.14.

b

b γ’

hγ

−1

h (U) γ c

h (γ ’) γ

U

γ a

a

Figure 14. Conformal invariance for conditional probabilities.

Conformal invariance implies that the image measure is the original measure in P(D,a,b) , ) is an interface that has forgotten γ[ac] . To summarize : where hγ (γ[cb] ) is independent of γ[ac] (the Markov property) and has the same distribhγ (γ[cb] ution as the original interface itself (stationarity of increments).

1.2.4. Local growth and Loewner evolutions The Markov and stationarity of increments property make it plain that to understand the distribution of the full interface, it is enough to understand the distribution of a small, or even inﬁnitesimal, initial segment, and then glue segments via conformal maps. This calls for a description by differential equations. For this purpose, it is very convenient –even if by no means mandatory– to have a natural parametrization of interfaces. Using conformal invariance, we can restrict ourselves

LOEWNER CHAINS

57

to the situation when (D, a, b) = (H, 0, ∞). If γ[0∞] is a simple curve from 0 to ∞ in H, and c a point on it, we know that H \ γ]0c] is a domain, and γ]0c] itself is a hull. We use capacity as a parametrization and deﬁne a time parameter by 2t(c) ≡ Cγ]0c] . The factor of 2 is just historical. The map t is a continuous increasing function of c varying from 0 to ¯ when c moves from 0 to ∞ along γ[0∞] . The inverse map t → c(t) is well-deﬁned ∞ in R and gives a parametrization of the curve. So we introduce an increasing family of hulls Kt , t ∈ [0, ∞[, by Kt = γ]0c(t)] which has capacity 2t. Let ft ≡ fKt be the conformal homeomorphism from H to H\Kt normalized to satisfy ft (w) = w −2t/w +O(1/w2 ) at inﬁnity. Deﬁne gt : H\Kt → H to be the inverse of ft . Then gt (z) = z +2t/z +O(1/z 2 ) at inﬁnity. To study the evolution of the family of hulls Kt , ﬁx ε ≥ 0 and consider the hull Kε,t ≡ gt (Kt+ε \ Kt ), which has capacity 2ε by eq.(1.2). Deﬁne fε,t ≡ fKε,t . Then gt = fε,t ◦gt+ε on H\Kt+ε . Using the representation of fKε,t in terms of its discontinuity eq.(1.1), we obtain gt+ε − fε,t ◦ gt+ε

1 = gt+ε − gt = 2π

R

dµfε,t (x) gt+ε − x

We introduce now the notion of local growth which is crucial for interfaces. When ε is small, Kε,t is a tiny piece of curve and the support of dµfε,t is small and becomes a point when ε goes to 0. Measures supported at a point are δ functions, so there is a point ξt such that, as a measure, dµfε,t /dx ∼ 2εδ(x−ξt ) as ε → 0+ . If Kt is a more general increasing family of hulls of capacity 2t, we say that the condition of local growth is satisﬁed if the above small ε behavior holds. At ﬁrst sight, it might seem that local growth is only true for curves, but this is not true. We shall give an example below. Letting ε → 0+ , from the local growth condition, we infer the existence of a real function ξt such that dgt 2 (z) = . (1.3) dt gt (z) − ξt It is useful to look at this equation from a slightly different point of view, taking the function ξt as the primary data. The solutions of this equation for a given function ξt with initial condition g0 (z) = z is called a Loewner evolution. The image of ξt by gt−1 is the tip of the curve at time t. The cases when the local growth condition is not satisﬁed are called Loewner chains, see below. Had we used another parametrization of the curve, the 2 in the numerator would be replaced by a positive function of the parameter along the curve. Informally, if Kt is a growing curve, we expect that gt+ε (z) − gt (z) describes an inﬁnitesimal cut. This is conﬁrmed by the explicit solution of eq.(1.3) for the trivial case ξt ≡ 0, which yields gt (z)2 = z 2 + 4t, the branch to be chosen being such that at large z, gt (z) ∼ z. This describes a growing segment along the imaginary axis. So intuitively, the simple pole in eq.(1.3) accounts for the existence of a cut and different functions ξt account for the different shapes of curves. One can also solve the case when Kt is an arc of circle starting from the origin. It can be obtained from the trivial solution above by applying appropriate time dependent P SL2 (R) transformations both to gt and z and then by a time change to recover the capacity parametrization. This is an illuminating exercise that we leave to the reader.

58

MICHEL BAUER AND DENIS BERNARD

Take an arc going from 0 to 2R along a circle of radius R. It turns out that when the arc approaches the real axis to close a half disk, the function ξt has a square root singularity √ ξt ∝ R2 − 2t. The capacity remains ﬁnite and goes to R2 and the map itself has a limit gR2 /2 (z) = z +R2 /(z +R) which has swallowed the half disk without violating the local growth condition. One can start the growth process again. Making strings of such maps with various values of the radii is a simple way to construct growing families of hulls that are not curves and that nevertheless grow locally. Note that a square root singularity for ξt is the marginal behavior: if ξt is H¨older of exponent > 1/2, Loewner evolution yields a simple curve. 1.2.5. Stochastic (or Schramm) Loewner evolutions If we sample locally growing hulls with a certain distribution, we get an associated random process ξt . In the case of a conformally invariant distribution, we have established two crucial properties: Markov property and stationarity of increments. To ﬁnish Schramm’s argument leading to SLE, what remains is to see the implications of these properties on the distribution of ξt . The argument and expressions for the meaning of Markov property and stationarity of increments involved a map h that mapped the tip of the piece of interface to the initial marked point a and the ﬁnal marked point b to itself. The map ht (z) = gt (z) − ξt has the required property when the domain is the upper-half plane with 0 and ∞ as marked points. It behaves like ht (z) = z −ξt +2t/z +O(1/z 2 ) at inﬁnity. We infer that for s > t, ht (Ks \ Kt ) is independent of Kt , t ≤ t (Markov property) and is distributed like a hull of capacity s − t = Cht (Ks \Kt ) (stationarity of increments). The hull determines the corresponding map h, so this can be rephrased as: hs ◦ h−1 t (which uniformizes ht (Ks \ Kt )) is independent of ht , t ≤ t, and distributed like an 2 hs−t . As hs ◦ h−1 t = z − (ξs − ξt ) + (s − t)/z + O(1/z ) at inﬁnity, the driving parameter −1 for the process hs ◦ ht is ξs − ξt . To summarize: the Markov property and stationarity of increments for the interface lead to the familiar statement for the process ξt : for s > t, ξs −ξt is independent of ξt , t ≤ t, (Markov property) and distributed like a ξs−t (stationarity of increments). To conclude, a last physical input is needed: one demands that the interface does not branch, which means that at two nearby times the growth is at nearby points. This implies that ξt is a continuous process, in the sense that it has continuous trajectories. One is now in position to apply a mathematical theorem: a 1d Markov process with continuous trajectories and stationary increments is proportional to√ a Brownian motion. We conclude that there is a real positive number κ such that ξt = κBt for some normalized Brownian motion Bt with covariance E[Bs Bt ] = min(s, t). The same argument without imposing that the time parametrization is given by the capacity of the hull would lead to the conclusion that the driving parameter is a continuous martingale, which is nothing but a Brownian motion after a possibly random time change. A solution of 2 dgt √ (z) = (1.4) dt gt (z) − κBt is called a chordal Schramm-Loewner evolution of parameter κ, in short a chordal SLEκ . The connection of this equation with interfaces relies mainly on conformal invariance.

LOEWNER CHAINS

59

But local growth, absence of branches and to a lower level locality at the interface also play a crucial role. 1.2.6. Miscellanea on SLE Up to now we have only discussed the situation when the interface goes from 0 to ∞ in the upper half plane, or more generally from one point on the boundary of a domain to another one. This is called chordal SLE. The argument can be extended easily to two other situations. The ﬁrst is for an interface starting at a ﬁxed point on the boundary of a domain and ending at a ﬁxed point in the bulk. This is called radial SLE. For the second situation, three points are chosen on the boundary and the interface starts from the ﬁrst point and ends at a random point between the two others. This is called dipolar SLE. Just as the upper-half plane is a convenient geometry for chordal SLE, a semi inﬁnite cylinder is nice for radial SLE and an inﬁnite strip is nice for dipolar SLE. For radial SLE on a cylinder of circumference π, the equation reads 2 dgt √ (z) = dt tan(gt (z) − κBt )

(1.5)

and for dipolar SLE on a strip of width π/2, one has 2 dgt √ (z) = dt tanh(gt (z) − κBt )

(1.6)

The generalization of SLE to Riemann surfaces with moduli is easy for the annulus, but raises non-trivial problems in general, currently subject of active research. The set of exact results obtained for SLE forms an impressive body of knowledge. We shall not mention the ones dealing with the explicit computation of certain crossing probabilities but we list just a few “pictorial” properties with some comments. They (the properties and the comments) should be understood with the standard proviso “almost surely” or “with probability 1”. We start with a surprisingly difﬁcult result[4, 5, 6, 7]. − Whatever the value of κ, the pre-image of the driving parameter limw→√κBt gt−1 (w) is a continuous curve γt , called the SLE trace. The trace never crosses itself. This property is crucial if the trace is to be interpreted as a curve separating two phases. − For κ ∈ [0, 4] the SLE trace is a simple curve. For κ ∈]4, 8[, it has double points. For κ ∈ [8, ∞[, it is space ﬁlling. − The fractal dimension dκ of the trace is 1 + κ/8 for κ ≤ 8 and 2 for κ ≥ 8. Using the formula for the dimension of the trace and confronting with the numerical simulations, it is plausible (actually, these are among the few cases for which a mathematical proof exists) that loop-erased random walks correspond to κ = 2, d = 5/4 and percolation to κ = 6, d = 7/4. This is also compatible with the general shape of the numerical samples, which indicate that loop-erased random walks indeed lead to simple curves and that percolation doesn’t. The hull Kt is by deﬁnition H \ gt−1 (H). It has the following properties − The hull Kt is the complement of the connected component of ∞ in H \ γ]0,t] . − For κ ∈ [0, 4], the SLE hull is a simple curve coinciding with the trace. For κ ∈ ]4, ∞[, the SLE hull has an nonempty connected and relatively dense interior.

60

MICHEL BAUER AND DENIS BERNARD

√ This may seem surprising at ﬁrst sight. It is the sign that for κ > 4, the drift κBt goes fast enough for the swallowing procedure to take place, as described in the closing arc example, but on all scales. This is summarized by ﬁg.15.

γ (t )

0

0. With t parameterizing time, Loewner equation reads:

∂ du w + u ft (w) = wft (w) ρt (u) (2.1) ∂t 2iπu w − u The integration is over the unit circle {u ∈ C, |u| = 1}. The Loewner density ρt (u) codes for the time evolution. It may depends on the map ft in which case the growth process in non-linear. For the inverse maps gt ≡ ft−1 : Ot → D, Loewner equation reads:

∂ du gt (z) + u gt (z) = −gt (z) ρt (u) (2.2) ∂t 2iπu gt (z) − u

62

MICHEL BAUER AND DENIS BERNARD

The behavior of ft at inﬁnity ﬁxes a scale since at inﬁnity, ft (w) Rt w + O(1) where Rt > 0, with the dimension of a [length], is called the conformal radius of Kt viewed from inﬁnity. Rt may be used to analyze scaling behaviors, since Kobe 1/4theorem (see e.g. [13]) ensures that Rt scales as the size of the domain. In particular, the (fractal) dimension D of the domains Kt may be estimated by comparing their area At with their linear size measured by Rt : At RtD for large t – the proportionality factor contains a cutoff dependence which restores naive dimensional analysis.

Ot

Γt

D

gt

Kt ft z − plane

w − plane

Figure 16. Uniformizing maps intertwining the ‘physical’ z-plane and the ‘mathematical’ w-plane.

2.1.2. The boundary curves The boundary curve Γt ≡ ∂Ot is the image of the unit circle by ft . We may parameterize the boundary points by γt;α = ft (u) with u = eiα . The Loewner equation codes for the evolution of the shape of Ot and thus for the normal velocity of the boundary points. Only the normal velocity is relevant as the tangent velocity is parameterization dependent. The tangent to the curve is τ = iuft (u)/|ft (u)| and the outward normal is n = −iτ so that n ∂t ft (u)], or the normal velocity at γt is vn = e[¯ vn = |ft (u)| e[∂t ft (u)/uft (u)]. The r.h.s. is determined by the Loewner equation (2.1) because this equation may be viewed as providing the solution of a boundary value problem. Indeed, recall that for du w+u ˆ ˆ h(u) a real function on the unit circle, h(w) = 2iπu w−u h(u) is the unique function ˆ ˆ analytic outside the unit disk whose real part on the unit circle is h, i.e. e[ h(u) ] = h(u). Thus, since ∂t ft (w)/wft (w) is analytic in D, the Loewner equation (2.1) is equivalent to: vn = |ft (u)| ρt (u) or more explicitly 1 : (∂α γt;α ) (∂t γt;α ) − (∂α γt;α ) (∂t γt;α ) = 2i |ft (u)|2 ρt (u) 1

For SLE this equation has to be modiﬁed according to Itˆo calculus

(2.3)

LOEWNER CHAINS

63

Hence, the evolution of the domain may be encoded either in the evolution law for its uniformizing conformal map as in eq.(2.1) or in the boundary normal velocity as in eq.(2.3). The two equations are equivalent. 2.2. EXAMPLES 2.2.1. Stochastic Loewner evolution (SLE) From this point of view, SLE is singular as it corresponds to a Dirac measure δUt (u) centered at the position of a Brownian motion on the unit circle. The locality of this measure reﬂects the fact that SLE describes growing curves. Continuity of the Brownian motion reﬂects the absence of branching in the SLE growths. To make contact with the previous sections, the reader should check that eq.(1.5) becomes of the form (2.2) after the cylinder has been mapped to the outside of the unit disk. 2.2.2. Laplacian growth (LG) This is a process in which the growth of the domain is governed by the solution of Laplace equation, i.e. by an harmonic function, in the exterior of the domain with appropriate boundary conditions. It originates from the hydrodynamical Hele-Shaw problem to be described below. To be precise, let P be the real solution of Laplace equation, ∇2 P = 0, in Ot with the boundary behavior P = − log |z| + · · · at inﬁnity and P = 0 on the boundary curve Γt = ∂Ot . The time evolution of the domain is then deﬁned by demanding that the normal velocity of points on the boundary curve be equal to minus the gradient of P : vn = −(∇P )n . This problem may be written as a Loewner chain since, as is well known, Laplace equation is solved via complex analysis by writing P as the real part of an analytic function. One ﬁrst solves Laplace equation in the complement of the unit disk with the appropriate boundary conditions and then transports it back to the physical domain Ot using the map ft . This gives: P = −e Φt

with

Φt (z) = log gt (z)

The evolution equation for the map ft is derived using that the boundary normal velocity is vn = −(∇P )n . The above expression for P gives: vn = −(∇P )n = |ft (u)|−1 at point γt = ft (u) on the boundary curve. As explained in the previous section, this is enough to determine ∂t ft (w) for any |w| > 1 since this data speciﬁes the real part on the unit circle of the analytic function ∂t ft (w)/wft (w) on the complement of the unit disk. The result is:

w + u du ∂t ft (w) = wft (w) (2.4) 2 w−u |u|=1 2iπu |ft (u)| It is a Loewner chain with ρt (u) = |ft (u)|−2 . As we shall see below, Laplacian growth is an integrable system, which may be solved exactly, but it is ill-posed as the domain develops singularities (cusps y 2 x3 ) in ﬁnite time. It thus needs to be regularized. There exist different ways of regularizing it.

64

MICHEL BAUER AND DENIS BERNARD

One may also formulate Laplacian growth using a language borrowed from electrostatics by imagining that the inner domain is a perfect conductor. Then V = e Φt is the electric potential which vanishes on the conductor but with a charge at inﬁnity. The = ∇V is E ¯ ≡ Ex − iEy = ∂z Φt . Its normal component En = |f (u)|−1 electric ﬁeld E t is proportional to the surface charge density. 2.2.3. The Hele-Shaw problem (HS) This provides a hydrodynamic regularization of Laplacian growth. The differences with Laplacian growth are in the boundary conditions which now involve a term proportional to the surface tension. It may be formulated as follows [14]. One imagines that the domain Kt is ﬁlled with a non viscous ﬂuid, say air, and the domain Ot with a viscous one, say oil. Air is supposed to be injected at the origin and there is an oil drain at inﬁnity. The pressure in the air domain Kt is constant and set to zero by convention. In Ot the pressure satisﬁes the Laplace equation ∇2 P = 0 with boundary behavior P = −φ∞ log |z| + · · · at inﬁnity reﬂecting the presence of the oil drain. The boundary conditions on the boundary curve are now P = −σκt with σ the surface tension and κt the curvature of the boundary curve 2 . The ﬂuid velocity in the oil domain Ot . Laplace equation for P is just a consequence of incompressibility. The is v = −∇P evolution of the shape of the domain is speciﬁed by imposing that this relation holds on the boundary so that the boundary normal velocity is vn = −(∇P )n as in Laplacian growth. Compared to Laplacian growth, the only modiﬁcation is the boundary condition on the boundary curve. This term prevents the formation of cusps with inﬁnite curvature singularities. The parameter φ∞ sets the scale of the velocity at inﬁnity. In the following we set φ∞ = 1. By dimensional analysis this implies that [time] scales as [length2 ] and the surface tension σ has dimension of a [length]. It plays the role of an ultraviolet cut-off. A standard procedure [14] to solve the equations for the Hele-Shaw problem is by ﬁrst determining the pressure using complex analysis and then computing the boundary normal velocity. By Laplace equation, the pressure is the real part of an analytic function, P = −eΦt . The complex velocity v = vx + ivy is v¯ = ∂z Φt . At inﬁnity Φt (z) log z + · · · and v¯ 1/z + · · · . The boundary conditions on P demand that (Φt ◦ ft )(w) = log w + σϑt (w) where ϑt (w) is analytic in D with boundary value e[ϑt (u)] = κt (ft (u)) with κt the curvature. Explicitly

du w + u κt (ft (u)) ϑt (w) = 2iπu w − u The evolution of ft is then found by evaluating the boundary normal velocity vn = e(∇Φ)n at point γt = ft (u): vn = e[ n∂z Φt ] = |ft (u)|−1 e[1 + σu∂u θt (u)] 2

The curvature is deﬁned by κ ≡ −n.∂s τ /τ 2 = m[¯ τ ∂s τ /|τ |3 ] with τ the tangent and n the uf (u) −1 normal vectors. An alternative formula is: κ = |ft (u)| e[1 + f t(u) ]. For a disk of radius R, t the curvature is +1/R.

LOEWNER CHAINS

65

As above, this determines uniquely ∂t ft (u) and it leads to a Loewner chain (2.1) with density: , t (u) = e[u∂u ϑt (u)] (2.5) ρt (u) = |ft (u)|−2 1 + σ t (u) The difference with Laplacian growth is in the extra term proportional to σ. It is highly non-linear and non-local. This problem is believed to be well deﬁned at all times for σ positive. 2.2.4. Other regularized Laplacian growth (rLG) These regularizations amount to introduce an UV cutoff δ in the physical space by evaluating |ft | at a ﬁnite distance away from ∂Ot . A possible choice [20] is ρt (u)1/2 = δ −1 inf{ε : dist[ft (u + εu); ∂Ot ] = δ}. An estimation gives ρt (u) |ft (u + εˆu u)|−2 where εˆu goes to 0 with δ, so that it naively approaches |ft (u)|−2 as δ → 0. Another possible, but less physical, regularization consists in introducing an UV cutoff ν in the mathematical space so that ρt (u) = |ft (u + νu)|−2 . 2.2.5. Dielectric breakdown and generalizations A larger class of problems generalizing Laplacian growth have been introduced. Their Loewner measures are as in Laplacian growth but with a different exponent: ρt (u) = |ft (u)|−α

,

0 ≤ α ≤ 2.

Using an electrostatic interpretation of the harmonic potential, one usually refers to the case α = 1 as a model of dielectric breakdown because the measure is then proportional to the local electric ﬁeld En = |ft (u)|−1 . This is a phenomenological description. Just as the naive Laplacian growth these models are certainly ill-posed. They also require ultraviolet regularizations, one of which is described below using iterated conformal maps. 2.3. CUSP SINGULARITIES IN LG The naive LG problem, without regularization, corresponds to the Loewner density ρt (u) = |ft (u)|−2 . The occurrence of singularities may be grasped by looking for the evolution of domains with a Zn symmetry uniformized by the maps ft (w) = Rt w(1 +

βt w−n ) n−1

for some n > 2 and with |βt | ≤ 1. This form of conformal maps is preserved by the dynamics. The conformal radius Rt and the coefﬁcients βt evolve with time according to ∂t Rt2 = 2/(1 − βt2 ) and βt = (Rt /Rc )n−2 with Rc some integration constant. The singularity appears when βt touches the unit circle which arises at a ﬁnite time tc . At that time the conformal radius is Rc . At tc the boundary curve Γtc has cusp singularities of the generic local form c (δy)2 (δx)3 with c a characteristic local length scale. In the present simple case √ c Rc . At time t tc , the dynamics is regular in the dimensionless parameter −1 tc − t. The maximum c

66

MICHEL BAUER AND DENIS BERNARD

curvature of the √ boundary curve scales as κmax c /(tc − t) near tc and it is localized at a distance tc − t away from the would be cusp tip. See Figure (17).

κ = lc /(tc −t)

t < tc

3

x = lc y

2

tc

Figure 17. Cups formation in Laplacian growth.

This behavior is quite generic. As we shall see, conformal maps ft (w) such that their derivatives are polynomials in w−1 are stable by the Laplacian growth dynamics. By construction, their zeroes are localized inside the unit disk. A singularity in the boundary curve occurs if one of these zeroes converges to the unit circle. The singularity is then a cusp c y 2 x3 as can be seen by expanding locally the conformal map around its singular point. Once regularized with an explicit ultraviolet cut-off, the processes are believed to be well deﬁned for all time. The effect of the regularization procedure on the domain properties is presently unclear. The domain structures may a priori depend on how the problem has been regularized. In the hydrodynamic regularization –the Hele-Shaw problem– the cusp production is expected to be replaced by unlimited ramiﬁcations leading to dendritic growth. In the regularized model, the curvature of Γt is expected to remain ﬁnite at all time. Using scaling theory, a crude estimate of its maximum around the would be singularities may be obtained by interchanging the short distance scale (tc − t) near the singularity in the unregularized theory with the UV cutoff of the regularized theory. In the hydrodynamic regularization this gives κmax c /σ 2 as σ → 0. 2.4. DISCRETE LOEWNER CHAINS 2.4.1. DLA DLA stands for diffusion limited aggregation [15]. It refers to processes in which the domains grow by aggregating diffusing particles. Namely, one imagines building up a domain by clustering particles one by one. These particles are released from the point at inﬁnity, or uniformly from a large circle around inﬁnity, and diffuse as random walkers. They will eventually hit the domain and the ﬁrst time this happens they stick to it. By convention, time is incremented by unity each time a particle is added to the domain. Thus at each time step the area of the domain is increased by the physical size of the particle. The position at which the particle is added depends on the probability for a random walker to visit the boundary for the ﬁrst time at this position. In a discrete approach one may imagine that the particles are tiny squares whose centers move on a square lattice whose edge lengths equal that of the particles, so that

LOEWNER CHAINS

67

particles ﬁll the lattice when they are glued together. The center of a particle moves as a random walker on the square lattice. The probability Q(x) that a particle visits a given site x of the lattice satisﬁes the lattice version of the Laplace equation ∇2 Q = 0. It vanishes on the boundary of the domain, i.e. Q = 0 on the boundary, because the probability for a particle to visit a point of the lattice already occupied, i.e. a point of the growing cluster, is zero. The local speed at which the domain is growing is proportional to the probability for a site next to the interface but on the outer domain to be visited. This probability is proportional to the discrete normal gradient of Q, since the visiting probability vanishes on the interface. So the local speed is vn = (∇Q)n . It is not so easy to make an unbiased simulation of DLA on the lattice. One of the reasons is that on the lattice there is no such simple boundary as a circle, for which the hitting distribution from inﬁnity is uniform. The hitting distribution on the boundary of a square is not such a simple function. Another reason is that despite the fact that the symmetric random walk is recurrent is 2d, each walk takes many steps to glue to the growing domain. The typical time to generate a single sample of reasonable size with an acceptable bias is comparable to the time it takes to make enough statistics on loop-erased random walks or percolation to get the scaling exponent with two signiﬁcant digits. Still this is a modest time, but it is enough to reveal the intricacy of the patterns that are formed. Fig.18 is such a sample. The similarity with the sample in ﬁg.1, obtained by iteration of conformal maps, is striking. But a quantitative comparison of the two models is well out of analytic control and belongs to the realm of extensive simulations.

Figure 18. A DLA sample.

DLA provides a discrete analogue of Laplacian growth. The particle size plays the role of an ultraviolet cutoff. Since DLA only assumes that the growth is governed by the

68

MICHEL BAUER AND DENIS BERNARD

diffusion of elementary patterns, DLA has been applied to a large variety of aggregation or deposition phenomena, see eg.[18]. During this process the clustering domain gets ramiﬁed and develops branches and fjords of various scales. The probability for a particle to stick on the cluster is much higher on the tip of the branches than deep inside the fjords. This property, relevant at all scales, is responsible for the fractal structure of the DLA clusters. Since its original presentation [15], DLA has been studied numerically quite extensively. There is now a consensus that the fractal dimension of 2d DLA clusters is Ddla 1.71. There is actually a debate on whether this dimension is geometry dependent but a recent study [19] seems to indicate that DLA clusters in a radial geometry and a channel geometry have identical fractal dimension. To add a new particle to the growing domain, a random walk has to wander around and the position at which it ﬁnally sticks is inﬂuenced by the whole domain. To rephrase this, for each new particle one has to solve the outer Laplace equation, a non-local problem, to know the sticking probability distribution. This is a typical example when scale invariance is not expected to imply conformal invariance. 2.4.2. Iterated conformal maps As proposed in [17], an alternative way to mimic the gluing of elementary particles consists in composing elementary conformal maps, each of which corresponds to adding an elementary particle to the domain. One starts with an elementary map corresponding to the gluing of a tiny bump, of linear size λ, to the unit disk. A large variety of choices is possible, whose inﬂuence on the ﬁnal structure of the domain is unclear. An example is given by the following formulæ (gλ is the inverse map of fλ ): z cos λ − 1 z − cos λ (2 cos λ)−1 w + 1 + w2 − 2w cos 2λ + 1

gλ (z) = z fλ (w) =

=

where fλ correspond to the deformation of the unit disk obtained by gluing a semi-disk centered at point 1 and whose two intersecting points with the unit circle deﬁne a cone of angle 2λ. For λ 1, the area of the added bump is of order λ2 . But other choices are possible and have been used. Gluing a bump around point eiθ on the unit circle is obtained by rotating these maps. The uniformizing maps are then fλ;θ (w) = eiθ fλ (we−iθ ) The growth of the domain is obtained by successively iterating the maps fλn ;θn with various values for the size λn and the position θn of the bumps. See Figure (19). Namely, if after n iterations the complement of the unit disk is uniformized into the complement of the domain by the map F(n) (w), then at the next (n + 1)th iteration the uniformizing map is given by: F(n+1) (w) = F(n) ( fλn+1 ;θn+1 (w) )

(2.6)

LOEWNER CHAINS

Fn

69

f λ; θ

F n+1 Figure 19. Iteration of conformal maps.

For the inverse maps, this becomes G(n+1) = gλn+1 ;θn+1 ◦ G(n) . To fully deﬁne the model one has to specify the choice of the parameter λn and θn at each iteration. Since λn codes for the linear size of the added bump and since locally conformal maps act as dilatations, the usual choice is to rescale λn+1 by a power of (eiθn )| as: |F(n) λn+1 = λ0 |F(n) (eiθn )|−α/2 ,

0≤α≤2

The case α = 2 corresponds to DLA as the physical area of the added bump are approximatively constant and equal to λ0 at each iterations. In the other case, the area of the (eiθn )|2−α . added bump scales as |F(n) The positions of the added bump are usually taken uniformly distributed on the mathematical unit circle with a measure dθ/2π. It is clear that this discrete model with α = 2 provides a regularization of Laplacian growth with λ0 playing the role of an ultraviolet cutoff. This may also be seen by looking at the naive limit of a small cutoff. Indeed, a naive expansion as λn 1 gives that F(n+1) = F(n) + δF(n) with w + eiθn δF(n) (w) λn wF(n) (w) w − eiθn where we used the expression of fλ for λ 1. Using the recursive expression for λn and averaging over θ with a uniform distribution yields:

dθ iθ −α w + eiθ |F (e )| δF (w) = λ0 wF (w) 2π w − eiθ For α = 2 this reproduces the Loewner chain for Laplacian growth. But this computation is too naive as the small cutoff limit is not smooth, a fact which is at the origin of the non trivial fractal dimensions of the growing domains. There are only very few mathematical results on these discrete models. The most recent one [27] deals with the simplest (yet interesting but not very physical) model with α = 0. It proves the convergence of the iteration to well-deﬁned random maps uniformizing domains of Hausdorff dimension 1. However, these models have been studied numerically extensively. There exists a huge literature on this subject but see ref.[24] for instance. These studies conﬁrm that the fractal dimension of DLA clusters with α = 2 is

70

MICHEL BAUER AND DENIS BERNARD

Ddla 1.71 but they also provide further informations on the harmonic measure multifractal spectrum. Results on the α dependence of the fractal dimension may be found eg. in ref.[25]. Various generalizations have been introduced. For instance, in ref.[26] a model of iterated conformal maps has been deﬁned in which particles are not added one by one but by layers. These models have one control parameter coding for the degree of coverage of the layer at each iterative step. By varying this parameter the model interpolates between discrete DLA and a discrete version of the Hele-Shaw problem. The fractal dimension of the resulting clusters varies with this parameter and seems to indicate that the fractal dimension of this discrete analogue of the Hele-Shaw problem is 2, a point that we shall discuss further in the last section. 2.5. INTEGRABILITY OF LAPLACIAN GROWTH Laplacian growth is an integrable system, at least up to the cusp formation. Let us recall that it corresponds to a Loewner chain with a density ρt (u) = |ft (u)|−2 , or equivalently to the quadratic equation (∂α γt;α ) (∂t γt;α ) − (∂α γt;α ) (∂t γt;α ) = 2i

(2.7)

for the dynamics of the boundary points γt;α = ft (u), u = eiα . What makes the model integrable is the fact that the r.h.s of eq.(2.7) is constant. Eq.(2.7) is then similar to a quadratic Hirota equation. Hints on the integrable structure were found in [21] and much further developed in [22, 23]. 2.5.1. Conserved quantities We now deﬁne an inﬁnite set of quantities which are conserved in the naive unregularized LG problem. They reﬂect its integrability. We follow ref.[22, 23]. These quantities may be deﬁned via a Riemann-Hilbert problem on Γt speciﬁed by, S+ (γ) − S− (γ) = γ¯

,

γ ∈ Γt

(2.8)

for functions S− and S+ respectively analytic in the outer domain Ot and in the inner domain Kt . We ﬁx normalization by demanding S− (∞) = 0. We assume Γt regular enough for this Riemann-Hilbert problem to be well deﬁned. As usual, S± may be presented as contour integrals:

dγ γ¯ . S± (z) = − 2iπ z−γ Γt The conserved quantities are going to be expressed in terms of S± . We thus need their time evolution. Differentiation of eq.(2.8) with respect to time and use of the evolution equation (2.7) gives: ∂t S+ (γ) − ∂t S− (γ) = 2gt (γ)/gt (γ) Notice now that gt (γ)/gt (γ) is the boundary value of (log gt ) which by construction is analytic in Ot . We may thus rewrite this equation as a trivial Riemann-Hilbert problem, ∂t S+ (γ) − (∂t S− + 2(log gt ) )(γ) = 0, so that both terms vanish: ∂t S+ (z) = 0 and

(∂t S− + 2(log gt ) )(z) = 0

(2.9)

LOEWNER CHAINS

71

Since S+ is analytic around the origin, we may expand it in power of z. Equation ∂t S+ (z) = 0 then tells us that S+ (z) is a generating function of conserved quantities: S+ (z) = k≥0 z k Ik with

Ik = Γt

dγ γ¯ γ −k−1 2iπ

,

∂t Ik = 0.

(2.10)

This provides an inﬁnite set of conserved quantities. Since S− is analytic around inﬁnity, it may be expanded in power of 1/z: S− (z) = γ the area of Kt . The second equation (∂t S− + −At /πz + · · · with At = − 2i Γt dγ¯ 2(log gt ) )(z) = 0 with gt (z) = Rt−1 z + O(1) then implies ∂t At = 2π. The area of the domain grows linearly with time, up to the time at which the ﬁrst cusp singularity appears. 2.5.2. Simple solutions A particularly simple class of conformal maps, solutions of the Laplacian growth equation, are those such that their derivatives are polynomials in w−1 . They may be expanded as: ft (w) =

N

fn w1−n ,

f0 = Rt > 0

(2.11)

n=0

with N ﬁnite but arbitrary. The dynamical variables are the N + 1 coefﬁcient f0 , · · · , fN . They are all complex except f0 which is real. It will be convenient to deﬁne the function f¯t by f¯t (w) = ft (w). The fact that this class is stable under the dynamics follows from the Loewner equation (2.4). The trick consists in using the fact that the integration contour is on the unit circle so that |ft (u)|2 = ft (u)f¯t (1/u). The contour integral then involves a meromorphic function of u so that it can be evaluated by deforming the contour to pick the residues. This is enough to prove that ∂t ft (w) possesses the same structure as ft (w) itself so that the class of functions (2.11) is stable under the dynamics. Alternatively one may expand the quadratic equation (2.7) to get a hierarchy of equations: (1 − n)[fn f¯˙j+n + f¯n f˙−j+n ] = 2δj;0 n≥0

For j = 0, this equation tells us again that the area of the domain grows linearly with time. Besides this relation there are only N independent complex equations for j = 1, · · · , N which actually code for the conserved quantities. To really have an integrable system we need to have as many independent integrals of motion as dynamical variables. Thus we need to have N conserved quantities. These are given by the Ik ’s deﬁned above which may rewritten as

du ft (u)f¯t (1/u) Ik = ft (u)k+1 |u|=1 2iπ Only the ﬁrst N quantities, I0 , · · · , IN −1 = R1−N f¯N are non-vanishing. They are independent. They can be used to express algebraically all fn ’s, n ≥ 0, in terms of the real

72

MICHEL BAUER AND DENIS BERNARD

parameter f0 = Rt . The area law, At = π[Rt2 +

(1 − n)|fn |2 ] = 2πt,

n≥1

with the fn ’s expressed in terms of Rt , then reintroduces the time variable by giving its relation with the conformal radius. 2.5.3. Algebraic curves As was pointed out in [23], solutions of Laplacian growth and their cusp formations have an elegant geometrical interpretation involving Riemann surfaces. Recall that given a sufﬁciently smooth real curve Γt drawn on the complex plane one may deﬁne a function S(z), called the Schwarz function, analytic in a ribbon enveloping the curve such that S(γ) = γ, γ ∈ Γt By construction, the Schwarz function may be expressed in terms of uniformizing maps of the domain bounded by the curves as S(z) = f¯t (1/gt (z)). The Riemann-Hilbert problem (2.8) deﬁning the conserved charges then possesses a very simple interpretation: S± are the polar part of the Schwarz function respectively analytic inside or outside Γt , i.e. S(z) = S+ (z) − S− (z). Thus the polar part S+ , analytic in the inner domain, is conserved. The polar part S− , analytic in the outer domain, evolves according to eqs.(2.9). Since log gt (z) is analytic in the outer domain, these equations are equivalent to the single equation: ∂t S(z) = −2(log gt (z))

(2.12)

Now the physical curve Γt may be viewed as a real slice of a complex curve, alias a Riemann surface. The latter is constructed using the Schwarz function as follows. Recall that s = S(z) is implicitly deﬁned by the relations z = ft (w), s = f¯t (1/w). In the case of polynomial uniformizing maps we get the pair of equations = f0 w + f1 + f2 w−1 + · · · + fN w1−N s = f¯0 w−1 + f¯1 + f¯2 w + · · · + f¯N wN −1

z

Eliminating w yields an algebraic equation for z and s only: R:

R(z, s) = 0

(2.13)

with R a polynomial of degree N in both variables, R(z, s) = f¯N z N + fN sN + · · · . Eq.(2.13) deﬁnes an algebraic curve R. It is of genus zero since by construction it is uniformized by points w of the complex sphere. It has many singularities which have to be resolved to recover a smooth complex manifold. The Riemann surface R may be viewed as a N -sheeted covering of the complex z plane: each sheet corresponds to a determination of s above point z. At inﬁnity, the physical sheet corresponds to z f0 w with w → ∞ so that s (z/f0 )N −1 f¯N , the other N − 1 sheets are ramiﬁed and correspond to z fN /wN −1 and s f¯0 /w with w → 0 so that z (s/f¯0 )N −1 fN . Hence inﬁnity is a branch point of order N − 1.

LOEWNER CHAINS

73

By the Riemann-Hurwitz formula the genus g is 2g − 2 = −2N + ν with ν the branching index of the covering. Since the point at inﬁnity counts for ν∞ = N − 2, there should be N other branch points generically of order two. By deﬁnition they are determined by solving the equations R(z, s) = 0 and ∂s R(z, s) = 0. Since the curve is uniformized by w ∈ C, these two equations imply that z (w)∂z R(z(w), s(w)) = 0. Hence either z (w) = 0, ∂z R = 0, and the point is a branch point, or z (w) = 0, ∂z R = 0 = ∂s R, and the point is actually a singular point which needs to be desingularized. So the N branch points at ﬁnite distance are the critical points of the uniformizing map z = ft (w). The curve R possesses an involution (z, s) → (¯ s, z¯) since R(¯ s, z¯) = R(z, s) by construction. The set of points ﬁxed by this involution has two components: (i) a continuous one parametrized by points w = u, |u| = 1 –this is the real curve Γt that we started with– and (ii) a set of N isolated points which are actually singular points. The cusp singularity of the real curve Γt arises when a isolated real point merges with the continuous real slice Γt . Locally the behavior is as for the curve u2 = ε v 2 + v 3 with ε → 0. The simplest example is for N = 3 with Z3 symmetry so that ft (w) = w + b/w2 and w2 z = w3 + b ,

w s = 1 + bw3

We set f0 = 1 and f3 = b. Without lost of generality we assume b real. The algebraic curve is then R(z, s) ≡ bz 3 + bs3 − b2 s2 z 2 + (b2 − 1)(2b2 + 1)sz − (b2 − 1)3 = 0 Inﬁnity is a branch point of order two. The three other branch points are at z = 3ω (b/4)1/3 , s = ω 2 (2b2 + 1)(2b)−1/3 corresponding to w = ω(2b)1/3 with ω a third root of unity. 2 They are critical points of z(w). There are three √ singular points at z = ω (1 − b )/b, 2 2 2 s = ω (1 − b )/b corresponding to w = ω(1 ± 1 − 4b )/2b. The physical regime is for b < 1/2 in which case the real slice Γt = {z(u), |u| = 1} is a simple curve. The singular points are then in the outer domain and the branch points in the inner domain. The cusp singularities arise for b = 1/2. For b > 1/2 there are no isolated singular points, they are all localized on the real slice so that Γt possesses double points. See Figure (20). 2.6. DENDRITIC ANOMALIES We now would like to discuss a few points concerning the regularized models and their small cut-off limits –which are actually what we are interested in. We shall point out that the hydrodynamic regularization used in the Hele-Shaw problem possesses essential differences with that used in other regularized models, say DLA. This opens the possibility for the Hele-Shaw problem and DLA not to be in the same universality class. This observation is based on the conjectural existence of anomalies in the Hele-Shaw problem –a word which refers to quantities which, although naively vanishing in the small cut-off limit, are actually non zero but ﬁnite in this limit due to compensating effects. An analogy with Burgers turbulence may be useful. The 1d Burgers equation is ∂t u + u∂x u − ν∂x2 u = 0 for a velocity ﬁeld u(t, x). The viscosity ν plays the role of ultraviolet cut-off. At ν = 0 this equation is simply Euler equation which may be easily solved.

74

MICHEL BAUER AND DENIS BERNARD

Figure 20. Subcritical and critical algebraic curves. Black circles are singular points. Crosses are branch points.

At ν = 0, any smooth initial data produces shocks in ﬁnite time with a discontinuity of u. In the regularized equation with ν = 0, there is no shock production but the ﬁeld

ˆ ≡ ν(∂x u)2 , which naively vanishes when ν → 0, has actually a non trivial limit. It does not vanish because there is a compensation between small viscosity ν and large velocity gradient (∂x u). The ﬁeld ˆ codes for the amount of energy dissipated inside the shocks. The analogy that we would like to put forward is that the cusp formation in Laplacian growth is analogous to shock formation in Burgers and that the surface tension σ in Laplacian growth plays the role of the viscosity in Burgers. There are potential compensating effects between small surface tension and large curvature in the Hele-Shaw problem. 2.6.1. Almost conserved quantities We now describe the effect of the regularization procedure on the conserved quantities of Laplacian growth. Recall that they were deﬁned starting from the Riemann-Hilbert problem: h+ (γ) − h− (γ) = γ¯

,

γ ∈ Γt

(2.14)

for functions h− and h+ respectively analytic in Ot and Kt . We use the notation h± instead of S± in order to avoid confusion with the unregularized Laplacian We dγgrowth. γ ¯ . Time again ﬁx normalization by demanding h− (∞) = 0 so that h± (z) = − Γt 2iπ z−γ differentiating eq.(2.14) using the evolution equation (2.3) now gives: ∂t h+ (γ) − ∂t h− (γ) = 2

gt (γ) vn (γ) gt (γ) |gt (γ)|

(2.15)

with vn the boundary normal velocity –in LG, we had vn = |gt (γ)|. In the case of the Hele-Shaw (HS) problem, the hydrodynamic regularization of Laplacian growth, the normal velocity possesses a particular form (2.5) involving the derivative of ϑ. Eq.(2.15) may then be rewritten in terms of the potential, Φt (z) = log gt (z) + σϑt (gt (z)), and of the curvature as: ∂t h+ (γ) − ∂t h− (γ) = 2Φt (γ) − 2σκt (γ)

LOEWNER CHAINS

75

Since Φt is analytic in Ot , the solution of this equation is given by that of the RiemannHilbert problem for κt . So, let κ− and κ+ be respectively analytic in Ot and Kt with boundary values κ+ − κ− = κt on Γt . Then: ∂t h+ (z)

= −2σ∂z κ+ (z)

∂t h− (z)

= −2∂z Φt (z) − 2σ∂z κ− (z)

(2.16)

These equations codes for the behavior of h± in the limit of vanishing surface tension. A noticeable point is that the above r.h.s. are total derivatives. In particular, contribution to the 1/z term in ∂t h− only comes from the logarithmic factor in the potential. This implies that at σ = 0 the area of the domains also grows linearly with time, At 2πt at large time 3 , and it is independent of the UV cutoff σ. This is contrast with other regularizations for which the area growth may be cutoff dependent. 2.6.2. A conjecture The supplementary terms proportional to σκ± in the quasi conservation laws (2.16) naively vanish when the UV cutoff is removed. Dendritic anomalies refer to the possibilities for these terms not to vanish as the cutoff goes to zero, say σ → 0 in the Hele-Shaw (HS) problem. Such possibility arises due to a balance between small surface tension and large curvature. Recall the previous HS estimation of the maximum curvature κmax c /σ 2 with c the local scale associated to the cusp. These maxima are localized around points γc which would turn into cusp singularities in absence of surface tension. So, we may expect that in the HS problem the product σκ(γ) is a function picked around its maximum values and is c of the form σc ϕ( γ−γ σ ) when the point γ moves away from γc with ϕ ﬁnite at the origin and decreasing rapidly away from it. For isolated singular points γc , this would converge in the limit σ → 0 to a sum of Dirac point measures. This leads us to conjecture that the anomalous term σκ(γ) goes to a non trivial but ﬁnite distribution as σ → 0. This distribution then contributes to eq.(2.16) to break the ‘classical’ conservation laws valid in the unregularized theory, hence the name dendritic anomaly. The Loewner equation (2.1) –or the evolution equation (2.3)– are then expected to have a ﬁnite limit as σ → 0 with a non trivial right hand side. The solution would be ˆ t such that, for ﬁxed well-deﬁned for all times and would describe a non trivial familly K ˆ t, Kt is the limit of the domains Kt as σ → 0. This is consistent with the observation that the area of Kt is independant of σ and grows linearly with time for any σ, ie. At 2πt, ˆ t is not ruled out. so that the existence of a limiting set K It is implicit in the previous conjecture that the uniformizing maps ft have a ﬁnite limit, say fˆt , as σ → 0 such that the image of the unit circle by fˆt is the boundary of ˆ t , which is the component ˆ t . This leads us to conjecture that the limiting domain O C\K ˆ of C \ Kt containing inﬁnity, is uniformized onto D by gˆt ≡ fˆt−1 . Let Rt be the conformal radius computed at ﬁnite σ using the maps ft . Recall that if Dhs is the fractal dimension we have RtDhs At for large t. By dimensional analysis 3 This result concerning the area is simply a consequence of the ﬂuid incompressibility. But the general method allows to deal with the complete hierarchy of quasi conserved charges.

76

MICHEL BAUER AND DENIS BERNARD

Rt σ(At /σ 2 )1/Dhs since σ has dimension of a [length]. Because in the HS problem the area grows linearly with time without any dependence on σ, this also reads: RtDhs σ Dhs −2 t Since we expect the limiting domains and the limiting conformal maps to exit, Rt should ˆ t computed have a ﬁnite limit as σ → 0 which coincides with the conformal radius R using fˆt . As consequence, we expect: Dhs = 2 That is: the fractal dimension of Laplacian growth clusters, within the hydrodynamic regularization, is 2. If these conjectures are true, they imply that the Hele-Shaw problem is not in the same universality class as DLA. These conjectures are compatible with that of ref.[26] based on numerical studies of generalized iterated conformal maps. They are not expected to apply to arbitrary regularizations because in these cases the area may depend on the UV cutoff and the previous dimensional analysis does not apply. It would then be natural to wonder whether we may deﬁne renormalized uniformizing maps in the limit of a vanishing cutoff.

Acknowledgements It is a pleasure thank Vincent Pasquier and Paul Wiegmann for discussions on this problem and the organizers of the Carg`ese 2004 ASI for providing us the opportunity to write these notes. Work supported in part by EC contract number HPRN-CT-2002-00325 of the EUCLID research training network. References 1. 2. 3.

4. 5.

A. Belavin, A. Polyakov, A. Zamolodchikov, Inﬁnite conformal symmetry in two-dimensional quantum ﬁeld theory, Nucl. Phys. B241, 333-380, (1984). J. Cardy, J. Phys. A25 (1992) L201–206. The rigourous proof is in S. Smirnov, C.R. Acad. Sci. Paris 333 (2001) 239–244. G. Watts, J. Phys A29 (1996) L363–368. The rigourous proof is in J. Dub´edat, Excursion Decompositions for SLE and Watts’ crossing formula, preprint, arXiv: math.PR/0405074 O. Schramm, Israel J. Math., 118, 221–288, (2000); G. Lawler, O. Schramm and W. Werner, (I): Acta Mathematica 187 (2001) 237–273; arXiv:math.PR/9911084 G. Lawler, O. Schramm and W. Werner, (II): Acta Mathematica 187 (2001) 275–308; arXiv:math.PR/0003156 G. Lawler, O. Schramm and W. Werner, (III): Ann. Henri Poincar´e 38 (2002) 109–123. arXiv:math.PR/0005294.

LOEWNER CHAINS 6. 7. 8. 9. 10.

11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27.

77

S. Rohde and O. Schramm, arXiv:math.PR/0106036; and references therein. V. Beffara, The dimension of the SLE curves, arXiv: math.PR/0211322. J. Dub´edat, SLE(κ, ρ) martingales and duality, to appear in Ann. Probab. arXiv:math.PR/0303128 G. Lawler, introductory texts, including the draft of a book, may be found at http://www.math.cornell.edu/∼lawler W. Werner, Lectures notes of the 2002 Saint Flour summer school M. Bauer and D. Bernard, Commun. Math. Phys. 239 (2003) 493–521, arXiv:hep-th/0210015, and Phys. Lett. B543 (2002) 135–138; M. Bauer and D. Bernard, Phys. Lett. B557 (2003) 309–316, arXiv-hep-th/0301064; M. Bauer and D. Bernard, Ann. Henri Poincar´e 5 (2004) 289–326, arXiv:math-ph/0305061. M. Bauer and D. Bernard, SLE, CFT and zig-zag probabilities, arXiv:math-ph/0401019. M. Bauer, D. Bernard and J. Houdayer, Dipolar SLEs, arXiv:math-ph/0411038 R. Friedrich and W. Werner, Conformal restriction, highest weight representations and SLE, arXiv:math-ph/0301018. , B. Duplantier Conformal Fractal Geometry and Boundary Quantum Gravity, arXiv:mathph/0303034. John B. Conway, Functions of One Complex Variable II, Springer-Verlag, New York, New York, 1995 D. Bensimon, L. Kadanoff, S. Liang, B. Shraiman, C. Tang, Rev. Mod. Phys. 58 (1986) 977, and references therein. T. Witten and L. Sander, Phys. Rev. Lett. 47 (1981) 1400. L. Niemeyer, L. Pietronero and H. Wiesmann, Phys. Rev. Lett. 52 (1984) 1033. M. Hasting and S. Levitov, Physica D116 (1998) 244. T. Hasley, Physics Today, 53, Nov. 2000, 36-41; M. Bazant, D. Crowdy, Conformal methods for interfacial dynamics, arXiv:condmat/0409439, and references therein. E. Somfai, R. Ball, J. deVita, L. Sander, ArXiv:cond-mat/0304458. L. Carleson and N. Makarov, Commun. Math. Phys. 216 (2001) 583. B. Shraiman, D. Bensimon, Phys. Rev A 30 (1984) 2840. S. Richardson, Euro. J. of Appl. Math. 12 (2001) 571. I. Krichever, M. Mineev-Weinstein, P. Wiegmann and A. Zabrodin, arXiv:nlin.SI/0311005, and refs. therein. B. Davidovitch, H. Hentschel, Z. Olami, I. Procaccia, L. Sander, E. Somfai, Phys. Rev. E59 (1999) 1368. M. Jensen, A. Levermann, J. Mathiesen, I. Procaccia, Phys. Rev. E65 (2002) 046109. M. Hastings, arXiv:cond-mat/0103312. H. Hentschel, A. Levermann, I. Procaccia, arXiv:cond-mat/0111567. S. Rohde, M. Zinsmeister, Some remarks on Laplacian growth, preprint April 2004.

THEORY OF COSMOLOGICAL PERTURBATIONS AND APPLICATIONS TO SUPERSTRING COSMOLOGY

ROBERT H. BRANDENBERGER Physics Department, McGill University 3600 rue Universit´e Montreal, QC, H3A 2T8, CANADA and Brown University Physics Department Providence, RI 02912, USA

Abstract. The theory of cosmological perturbations is the main tool which connects theories of the early Universe (based on new fundamental physics such as string theory) with cosmological observations. In these lectures, I will provide an introduction to this theory, beginning with an overview of the Newtonian theory of ﬂuctuations, moving on to the analysis of ﬂuctuations in the realm of classical general relativity, and culminating with a discussion of the quantum theory of cosmological perturbations. I will illustrate the formalism with applications to inﬂationary cosmology. I will review the basics of inﬂationary cosmology and discuss why - through the evolution of ﬂuctuations - inﬂation may provide a way of observationally testing Planck-scale physics.

1. Introduction Recent years have provided a wealth of observational data about the cosmos. We have high resolution maps of the anisotropies in the temperature of the cosmic microwave background (CMB) [1], surveys of the large-scale structure (LSS) - the distribution of galaxies in three-dimensional space - are increasing in size and in accuracy (see e.g. [2] and [3]), and new techniques which will allow us to measure the distribution of the dark matter are being pioneered. All of this data involves small deviations of the cosmos from hom*ogeneity and isotropy. The cosmological observations reveal that the Universe has non-random ﬂuctuations on all scales smaller than the present Hubble radius. Parallel to this spectacular progress in observational cosmology, new cosmological scenarios have emerged within which it is possible to explain the origin of non-random inhom*ogeneities by means of causal physics. The scenario which has attracted most attention is inﬂationary cosmology [4, 5], according to which there was a period in the early Universe in which space was expanding at an accelerated rate. However, there are also alternative proposals [7, 8] in which our current stage of cosmological expansion 79 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 79–116. © 2005 Springer. Printed in the Netherlands.

80

ROBERT H. BRANDENBERGER

is preceded by a phase of contraction. These scenarios have in common the fact that for scales of cosmological interest today, although their physical wavelength is larger than the Hubble length during most of the history of the universe, it is smaller than the Hubble radius at very early times, thus in principle allowing for a causal origin of the cosmological ﬂuctuations. In order to connect theories of fundamental physics providing an origin of perturbations with the data on the late time universe, one must be able to evolve cosmological ﬂuctuations from earliest times to today. Since on large scales (scales larger than about 10 Mpc - 1 Mpc being roughly three million light years) the relative density ﬂuctuations are smaller than one today, and since these relative ﬂuctuations grow in time as a consequence of gravitational instability, they were smaller than one throughout their history - at least in a universe which is always expanding. Thus, it is reasonable to expect that a linearized analysis of the ﬂuctuations will give reliable results. In most current models of the very early universe it is assumed that the perturbations originate as quantum vacuum ﬂuctuations. Thus, quantum mechanics is important. On the other hand, since for most of the history of the universe the wavelengths corresponding to scales of cosmological interest today were larger than the Hubble radius, it is crucial to consider the general relativistic theory of ﬂuctuations. Hence, a quantum theory of general relativistic ﬂuctuations is required. At the level of the linearized theory of cosmological perturbations, a uniﬁed quantum theory of the formation and evolution of ﬂuctuations exists - and this will be the main topic of these lectures. Note that all of the conceptual problems of merging quantum mechanics and general relativity have been thrown out by hand by restricting attention to the linearized analysis. The question of “linearization stability” of the system, namely whether the solutions of the linearized equations of motion in fact correspond to the linearization of solutions of the full equations, is a deep one and will not be addressed here (see e.g. [6] for discussions of this issue). Inﬂationary cosmology is at the present time the most successful framework of connecting physics of the very early universe with the present structure (although alternatives such as the Pre-Big-Bang [7] and Ekpyrotic [8] scenarios have been proposed and may turn out to be successful as well). I will thus begin these lectures with a review of inﬂationary cosmology, and of scalar ﬁeld-driven models for inﬂation. Next, I will provide a detailed discussion of the theory of cosmological perturbations, beginning with the Newtonian theory, moving on to the classical general relativistic analysis, and ending with the quantum theory of cosmological perturbations 1 In the ﬁnal sections of these notes, I will return to inﬂationary cosmology, and focus on some important conceptual problems which are not addressed in current realizations of inﬂation in which the accelerated expansion of space is driven by a scalar matter ﬁeld. Addressing these conceptual problems is a challenge and great opportunity for string theory. Since cosmological inﬂation leads to a quasi-exponential increase in the wavelength of inhom*ogeneities, it provides a microscope with which string-scale physics can in principle be probed in current observations. I will conclude these lectures with a discussion of this “window of opportunity” for string theory which inﬂation provides [9].

1

These sections are an updated version of the lecture notes [10].

COSMOLOGICAL PERTURBATIONS

81

2. Overview of Inﬂationary Cosmology To establish our notation and framework, we will be taking the background space-time to be hom*ogeneous and isotropic, with a metric given by ds2 = dt2 − a(t)2 dx2 ,

(2.1)

where t is physical time, dx2 is the Euclidean metric of the spatial hypersurfaces (here taken for simplicity to be spatially ﬂat), and a(t) is the scale factor. The scale factor determines the Hubble expansion rate via H(t) =

a˙ (t) . a

(2.2)

The coordinates x used above are comoving coordinates, coordinates painted onto the expanding spatial hypersurfaces. In standard big bang cosmology, the universe is decelerating, i.e. a ¨ < 0. As a consequence, the Hubble radius lH (t) = H −1 (t) (2.3) is increasing in comoving coordinates. As will be explained later mathematically, the Hubble radius is the maximal distance that microphysics can act coherently over a Hubble expansion time - in particular it is the maximal distance on which any causal process could create ﬂuctuations. If the universe were decelerating forever, then scales of cosmological interest today would have had a wavelength larger than the Hubble radius at all early times. This gives rise to the ﬂuctuation problem for Standard Big Bang (SBB) cosmology, namely the problem that there cannot be any causal process which at early time creates perturbations on scales which are being probed in current LSS and CMB observations 2 . The idea of inﬂationary cosmology is to assume that there was a period in the very early Universe during which the scale factor was accelerating, i.e. a ¨ > 0. This implies that the Hubble radius was shrinking in comoving coordinates, or, equivalently, that ﬁxed comoving scales were “exiting” the Hubble radius. In the simplest models of inﬂation, the scale factor increases nearly exponentially. As illustrated in Figure (1), the basic geometry of inﬂationary cosmology provides a solution of the ﬂuctuation problem. As long as the phase of inﬂation is sufﬁciently long, all length scales within our present Hubble radius today originate at the beginning of inﬂation with a wavelength smaller than the Hubble radius at that time. Thus, it is possible to create perturbations locally using physics obeying the laws of special relativity (in particular causality). As will be discussed later, it is quantum vacuum ﬂuctuations of matter ﬁelds and their associated curvature perturbations which are responsible for the structure we observe today. Postulating a phase of inﬂation in the very early universe also solves the horizon problem of the SBB, namely it explains why the causal horizon at the time trec when photons last scatter is larger than the radius of the past light cone at trec , the part of the last scattering surface which is visible today in CMB experiments. Inﬂation also explains the near ﬂatness of the universe: in a decelerating universe spatial ﬂatness is an unstable ﬁxed 2 The reader is encouraged to ﬁnd the hole in this argument, and is referred to [11, 7, 13] for the answer.

82

ROBERT H. BRANDENBERGER t

t0 H

−1

t (k) f

k t

x

R

t (k) i

t

i

Figure 1. Space-time diagram (sketch) showing the evolution of scales in inﬂationary cosmology. The vertical axis is time, and the period of inﬂation lasts between ti and tR , and is followed by the radiation-dominated phase of standard big bang cosmology. During exponential inﬂation, the Hubble radius H −1 is constant in physical spatial coordinates (the horizontal axis), whereas it increases linearly in time after tR . The physical length corresponding to a ﬁxed comoving length scale labelled by its wavenumber k increases exponentially during inﬂation but increases less fast than the Hubble radius (namely as t1/2 ), after inﬂation.

point of the dynamics, whereas in an accelerating universe it becomes an attractor. Another important aspect of the inﬂationary solution of the ﬂatness problem is that inﬂation exponentially increases the volume of space. Without inﬂation, it is not possible that a Planck scale universe at the Planck time evolves into a sufﬁciently large universe today. Let us now consider how it is possible to obtain a phase of cosmological inﬂation. We will assume that space-time is to be described using the equations of General Relativity 3 . In this case, the dynamics of the scale factor a(t) is determined by the FriedmannRobertson-Walker (FRW) equations a˙ ( )2 = 8πGρ a

(2.4)

a ¨ = −4πG(ρ + 3p) a

(2.5)

and

where for simplicity we have omitted the contributions of spatial curvature (since spatial curvature is diluted during inﬂation) and of the cosmological constant (since any small cosmological constant which might be present today has no effect in the early Universe since the associated energy density does not increase when going into the past). In the above, ρ and p denote the energy density and pressure, respectively. From (2.5) it is clear 3 Note, however, that the ﬁrst model of exponential expansion of space [14] made use of a higher derivative gravitational action.

COSMOLOGICAL PERTURBATIONS

83

that in order to obtain an accelerating universe, matter with sufﬁciently negative pressure 1 p < − ρ 3

(2.6)

is required. Exponential inﬂation is obtained for p = −ρ. Conventional perfect ﬂuids have positive semi-deﬁnite pressure and thus cannot yield inﬂation. In addition, we know that a description of matter in terms of classical perfect ﬂuids must break down at early times. An improved description of matter will be given in terms of quantum ﬁelds. Scalar matter ﬁelds are special in that they allow at the level of a renormalizable action the presence of a potential energy term. The energy density and pressure of a scalar ﬁeld ϕ with canonically normalized action 4 L =

√ 1 −g ∂µ ϕ∂ µ ϕ − V (ϕ) 2

(2.7)

(where Greek indices are space-time indices, V (ϕ) is the potential energy density, and g is the determinant of the metric) are given by ρ

=

p

=

1 (ϕ) ˙ 2+ 2 1 (ϕ) ˙ 2− 2

1 −2 a (∇ϕ)2 + V (ϕ) 2 1 −2 a (∇ϕ)2 − V (ϕ) . 6

(2.8)

Thus, it is possible to obtain an almost exponentially expanding universe provided the scalar ﬁeld conﬁguration 5 satisﬁes 1 (∇p ϕ)2 2 1 (ϕ) ˙ 2 2

V (ϕ) ,

(2.9)

V (ϕ) .

(2.10)

In the above, ∇p ≡ a−1 ∇ is the gradient with respect to physical as opposed to comoving coordinates. Since spatial gradients redshift as the universe expands, the ﬁrst condition will (for single scalar ﬁeld models) always be satisﬁed if it is satisﬁed at the initial time 6 . It is the second condition which is harder to satisfy. In particular, this condition is in general not preserved in time even it is initially satisﬁed. It is sufﬁcient to obtain a period of cosmological inﬂation that the slow-roll conditions for ϕ are satisﬁed. Recall that the equation of motion for a hom*ogeneous scalar ﬁeld in a cosmological space-time is (as follows from (2.7)) is ϕ¨ + 3H ϕ˙ = −V (ϕ) ,

(2.11)

where a prime indicates the derivative with respect to ϕ. In order that the scalar ﬁeld roll slowly, it is necessary that ϕ¨ 3H ϕ˙ (2.12) 4

See [15] for a discussion of ﬁelds with non-canonical kinetic terms. The scalar ﬁeld yielding inﬂation is called the inﬂaton. 6 In fact, careful studies [16] show that since the gradients decrease even in a non-inﬂationary backgrounds, they can become subdominant even if they are not initially subdominant. 5

84

ROBERT H. BRANDENBERGER

such that the ﬁrst term in the scalar ﬁeld equation of motion (2.11) is negligible. In this case, the condition (2.10) becomes (

V 2 ) 48πG V

(2.13)

and (3) becomes

V 24πG . (2.14) V In the initial model of inﬂation using scalar ﬁelds (“old inﬂation” [4]), it was assumed that ϕ was initially in a false vacuum with large potential energy. Hence, the conditions for inﬂation are trivially satisﬁed. To end inﬂation, a quantum tunneling event from the false vacuum to the true vacuum [75] was invoked (see e.g. [18] for a pedagogical review). This model, however, has a graceful exit problem since the tunneling leads to an initially microscopical bubble of the true vacuum which cannot grow to encompass our presently observed universe - the ﬂatness problem of SBB cosmology in a new form. Hence, attention shifted to models in which the scalar ﬁeld ϕ is slowly rolling during inﬂation. There are many models of scalar ﬁeld-driven inﬂation. Many of them can be divided into three groups [19]: small-ﬁeld inﬂation, large-ﬁeld inﬂation and hybrid inﬂation. Small-ﬁeld inﬂationary models are based on ideas from spontaneous symmetry breaking in particle physics. We take the scalar ﬁeld to have a potential of the form V (ϕ) =

1 λ(ϕ2 − σ 2 )2 , 4

(2.15)

where σ can be interpreted as a symmetry breaking scale, and λ is a dimensionless coupling constant. The hope of initial small-ﬁeld models (“new inﬂation” [20]) was that the scalar ﬁeld would begin rolling close to its symmetric point ϕ = 0, where thermal equilibrium initial conditions would localize it in the early universe. At sufﬁciently high temperatures, ϕ = 0 is a stable ground state of the one-loop ﬁnite temperature effective potential VT (ϕ) (see e.g. [18] for a review). Once the temperature drops to a value smaller than the critical temperature Tc , ϕ = 0 turns into an unstable local maximum of VT (ϕ), and ϕ is free to roll towards a ground state of the zero temperature potential (2.15). The direction of the initial rolling is triggered by quantum ﬂuctuations. The reader can easily check that for the potential (2.15) the slow-roll conditions cannot be satisﬁed if σ mpl , where mpl is the Planck mass which is related to G. If the potential is modiﬁed to a ColemanWeinberg [21] form V (ϕ) =

ϕ2 λ 4 λ2 4 25

ϕ + ϕ ln( 2 ) − 2 4 44π M 6

(2.16)

(where M denotes some renormalization scale) then the slow-roll conditions can be satisﬁed. However, this corresponds to a severe ﬁne-tuning of the shape of the potential. A further problem for most small-ﬁeld models of inﬂation (see e.g. [22] for a review) is that the slow-roll trajectory is not an attractor in phase space. In order to end up close to the slow-roll trajectory, the initial ﬁeld velocity must be constrained to be very small. This initial condition problem of small-ﬁeld models of inﬂation effects a number of recently proposed brane inﬂation scenarios, see e.g. [23] for a discussion.

COSMOLOGICAL PERTURBATIONS

85

There is another reason for abandoning small-ﬁeld inﬂation models: in order to obtain a sufﬁciently small amplitude of density ﬂuctuations, the interaction coefﬁcients of ϕ must be very small (this problem is discussed at length towards the end of these lectures). In particular, this makes it inconsistent to assume that ϕ started out in thermal equilibrium. In the absence of thermal equilibrium, the phase space of initial conditions is much larger for large values of ϕ. This brings us to the discussion of large-ﬁeld inﬂation models, initially proposed in [24] under the name “chaotic inﬂation”. The simplest example is provided by a massive scalar ﬁeld with potential 1 (2.17) V (ϕ) = m2 ϕ2 , 2 where m is the mass. It is assumed that the scalar ﬁeld rolls towards the origin from large values of |ϕ|. It is a simple exercise for the reader to verify that the slow-roll conditions (2.13) and (2.14) are satisﬁed provided 1 mpl . |ϕ| > √ 12π

(2.18)

Values of |ϕ| comparable or greater than mpl are also required in other realizations of large-ﬁeld inﬂation. Hence, one may worry whether such a toy model can consistently be embedded in a realistic particle physics model, e.g. supergravity. In many such models V (ϕ) receives supergravity-induced correction terms which destroy the ﬂatness of the potential for |ϕ| > mpl . However, as recently discussed in [25], if the ﬂatness of the potential is protected by some symmetry, then it can survive inclusion of the correction terms. As will be discussed later, a value of m ∼ 1013 GeV is required in order to obtain the observed amplitude of density ﬂuctuations. Hence, the conﬁguration space of ﬁeld values with |ϕ| > mpl but V (ϕ) < m4pl is huge. It can also be veriﬁed that the slow-roll trajectory is a local attractor in ﬁeld initial condition space [16], even including metric ﬂuctuations at the perturbative level [26]. With two scalar ﬁelds it is possible to construct a class of models which combine some of the nice features of large-ﬁeld inﬂation (large phase space of initial conditions yielding inﬂation) and of small-ﬁeld inﬂation (better contact with conventional particle physics). These are models of hybrid inﬂation [27]. To give a prototypical example, consider two scalar ﬁelds ϕ and χ with a potential V (ϕ, χ) =

1 1 1 λχ (χ2 − σ 2 )2 + m2 ϕ2 − g 2 ϕ2 χ2 . 4 2 2

(2.19)

In the absence of thermal equilibrium, it is natural to assume that |ϕ| begins at large values, values for which the effective mass of χ is positive and hence χ begins at χ = 0. The parameters in the potential (2.19) are now chosen such that ϕ is slowly rolling for values of |ϕ| somewhat smaller than mpl , but that the potential energy for these ﬁeld values is dominated by the ﬁrst term on the right-hand side of (2.19). The reader can easily verify that for this model it is no longer required to have values of |ϕ| greater than mpl in order to obtain slow-rolling 7 The ﬁeld ϕ is slowly rolling whereas the potential 7 Note that the slow-roll conditions (2.13) and (2.14) were derived assuming that H is given by the contribution of ϕ to V which is not the case here.

86

ROBERT H. BRANDENBERGER

energy is determined by the contribution from χ. Once |ϕ| drops to the value λχ |ϕc | = σ g

(2.20)

the conﬁguration χ = 0 becomes unstable and decays to its ground state |χ| = σ, yielding a graceful exit from inﬂation. Since in this example the ground state of χ is not unique, there is the possibility of the formation of topological defects at the end of inﬂation (see [11, 7, 13] for reviews of topological defects in cosmology, and the lectures by Polchinski at this school [28] for a discussion of how this scenario arises in brane inﬂation models). After the slow-roll conditions break down, the period of inﬂation ends, and the inﬂaton begins to oscillate around its ground state. Due to couplings of ϕ to other matter ﬁelds, the energy of the universe, which at the end of the period of inﬂation is stored completely in ϕ, gets transferred to the matter ﬁelds of the particle physics Standard Model. Initially, the energy transfer was described perturbatively [29, 30]. Later, it was realized [31, 32, 33, 34] that through a parametric resonance instability, particles are very rapidly produced, leading to a fast energy transfer (“preheating”). The quanta later thermalize, and thereafter the universe evolves as described by SBB cosmology. After this review of inﬂationary cosmology (see e.g. [35] for a more complete recent review), we turn to the discussion of the main success of inﬂationary cosmology, namely the fact that it provides a causal mechanism for generating small inhom*ogeneities. The reader is referred to [5] for a comprehensive analysis of this theory of cosmological perturbations. 3. Newtonian Theory of Cosmological Perturbations 3.1. INTRODUCTION The growth of density ﬂuctuations is a consequence of the purely attractive nature of the gravitational force. Imagine (ﬁrst in a non-expanding background) a density excess δρ localized about some point x in space. This ﬂuctuation produces an attractive force which pulls the surrounding matter towards x. The magnitude of this force is proportional to δρ. Hence, by Newton’s second law ¨ ∼ Gδρ , δρ (3.1) where G is Newton’s gravitational constant. Hence, there is an exponential instability of ﬂat space-time to the development of ﬂuctuations. Obviously, in General Relativity it is inconsistent to consider density ﬂuctuations in a non-expanding background. If we consider density ﬂuctuations in an expanding background, then the expansion of space leads to a friction term in (3.1). Hence, instead of an exponential instability to the development of ﬂuctuations, the growth rate of ﬂuctuations in an expanding Universe will be as a power of time. It is crucial to determine what this power is and how it depends both on the background cosmological expansion rate and on the length scale of the ﬂuctuations. Note that in the following two subsections x will denote the physical coordinates, and q the comoving ones. The materials covered in this section are discussed in several excellent textbooks on cosmology, e.g. in [37, 38, 39, 40].

COSMOLOGICAL PERTURBATIONS

87

3.2. PERTURBATIONS ABOUT MINKOWSKI SPACE-TIME To develop some physical intuition, we ﬁrst consider the evolution of hydrodynamical matter ﬂuctuations in a ﬁxed non-expanding background. In this context, matter is described by a perfect ﬂuid, and gravity by the Newtonian gravitational potential ϕ. The ﬂuid variables are the energy density ρ, the pressure p, the ﬂuid velocity v, and the entropy density S. The basic hydrodynamical equations are ρ˙ + ∇ · (ρv) = 1 v˙ + (v · ∇)v + ∇p + ∇ϕ = ρ ∇2 ϕ = S˙ + (v · ∇)S = p =

0 0 4πGρ

(3.2)

0 p(ρ, S) .

The ﬁrst equation is the continuity equation, the second is the Euler (force) equation, the third is the Poisson equation of Newtonian gravity, the fourth expresses entropy conservation, and the last describes the equation of state of matter. The derivative with respect to time is denoted by an over-dot. The background is given by the background energy density ρo , the background pressure p0 , vanishing velocity, constant gravitational potential ϕ0 and constant entropy density S0 . As mentioned above, it does not satisfy the background Poisson equation. The equations for cosmological perturbations are obtained by perturbing the ﬂuid variables about the background, ρ = ρ0 + δρ v = δv p = p0 + δp ϕ = ϕ0 + δϕ S

(3.3)

= S0 + δS ,

where the ﬂuctuating ﬁelds δρ, δv, δp, δϕ and δS are functions of space and time, by inserting these expressions into the basic hydrodynamical equations (3.2), by linearizing, and by combining the resulting equations which are of ﬁrst order in time. We get the following differential equations for the energy density ﬂuctuation δρ and the entropy perturbation δS ¨ − c2 ∇2 δρ − 4πGρ0 δρ δρ s ˙ δS

= σ∇2 δS = 0,

(3.4)

where the variables c2s and σ describe the equation of state δp = c2s δρ + σδS with c2s =

δp δρ |S

(3.5)

(3.6)

88

ROBERT H. BRANDENBERGER

denoting the square of the speed of sound. Sincethe equations are linear, we can work in Fourier space. Each Fourier component δρk (t) of the ﬂuctuation ﬁeld δρ(x, t) (3.7) δρ(x, t) = eik·x δρk (t) evolves independently. The ﬂuctuations can be classiﬁed as follows: If δS vanishes, we have adiabatic ﬂuc˙ = 0, we speak of an entropy ﬂuctuation. tuations. If the δS is non-vanishing but δρ The ﬁrst conclusions we can draw from the basic perturbation equations (3.4) are that 1) entropy ﬂuctuations do not grow, 2) adiabatic ﬂuctuations are time-dependent, and 3) entropy ﬂuctuations seed an adiabatic mode. Taking a closer look at the equation of motion for δρ, we see that the third term on the left hand side represents the force due to gravity, a purely attractive force yielding an instability of ﬂat space-time to the development of density ﬂuctuations (as discussed earlier, see (3.1)). The second term on the left hand side of (3.4) represents a force due to the ﬂuid pressure which tends to set up pressure waves. In the absence of entropy ﬂuctuations, the evolution of δρ is governed by the combined action of both pressure and gravitational forces. Restricting our attention to adiabatic ﬂuctuations, we see from (3.4) that there is a critical wavelength, the Jeans length, whose wavenumber kJ is given by kJ =

4πGρ0 1/2 . c2s

(3.8)

Fluctuations with wavelength longer than the Jeans length (k kJ ) grow exponentially δρk (t) ∼ eωk t with ωk ∼ 4(πGρ0 )1/2

(3.9)

whereas short wavelength modes (k kJ ) oscillate with frequency ωk ∼ cs k. Note that the value of the Jeans length depends on the equation of state of the background. For a background dominated by relativistic radiation, the Jeans length is large (of the order of the Hubble radius H −1 (t)), whereas for pressure-less matter it goes to zero. 3.3. PERTURBATIONS ABOUT AN EXPANDING BACKGROUND Let us now improve on the previous analysis and study Newtonian cosmological ﬂuctuations about an expanding background. In this case, the background equations are consistent (the non-vanishing average energy density leads to cosmological expansion). However, we are still neglecting general relativistic effects (the ﬂuctuations of the metric). Such effects turn out to be dominant on length scales larger than the Hubble radius H −1 (t), and thus the analysis of this section is applicable only to smaller scales. The background cosmological model is given by the energy density ρ0 (t), the pressure p0 (t), and the recessional velocity v0 = H(t)x, where x is the Euclidean spatial coordinate vector (“physical coordinates”). The space- and time-dependent ﬂuctuating ﬁelds

COSMOLOGICAL PERTURBATIONS

89

are deﬁned as in the previous section: ρ(t, x) v(t, x)

= ρ0 (t) 1 + δ (t, x) = v0 (t, x) + δv(t, x)

p(t, x)

= p0 (t) + δp(t, x) ,

(3.10)

where δ is the fractional energy density perturbation (we are interested in the fractional rather than in the absolute energy density ﬂuctuation!), and the pressure perturbation δp is deﬁned as in (3.5). In addition, there is the possibility of a non-vanishing entropy perturbation deﬁned as in (3.3). We now insert this ansatz into the basic hydrodynamical equations (3.2), linearize in the perturbation variables, and combine the ﬁrst order differential equations for δ and δp into a single second order differential equation for δρ . The result simpliﬁes if we work in “comoving coordinates” q which are the coordinates painted onto the expanding background, i.e. x(t) = a(t)q(t) . (3.11) After some algebra, we obtain the following equation which describes the time evolution of density ﬂuctuations: c2 σ δ¨ + 2H δ˙ − s2 ∇2q δ − 4πGρ0 δ = δS , a ρ0 a2

(3.12)

where the subscript q on the ∇ operator indicates that derivatives with respect to comoving coordinates are used. In addition, we have the equation of entropy conservation ˙ = 0. δS

(3.13)

Comparing with the equations (3.4) obtained in the absence of an expanding background, we see that the only difference is the presence of a Hubble damping term in the equation for δ . This term will moderate the exponential instability of the background to long wavelength density ﬂuctuations. In addition, it will lead to a damping of the oscillating solutions on short wavelengths. More speciﬁcally, for physical wavenumbers kp kJ (where kJ is again given by (3.8)), and in a matter-dominated background cosmology, the general solution of (3.12) in the absence of any entropy ﬂuctuations is given by δk (t) = c1 t2/3 + c2 t−1 ,

(3.14)

where c1 and c2 are two constants determined by the initial conditions, and we have dropped the subscript in expressions involving δ . There are two fundamental solutions, the ﬁrst is a growing mode with δk (t) ∼ a(t), the second a decaying mode with δk (t) ∼ t−1 . On short wavelength, one obtains damped oscillatory motion: −1/2 δk (t) ∼ a (t)exp ±ics k dt a−1 (t ) . (3.15) As a simple application of the Newtonian equations for cosmological perturbations derived above, let us compare the predicted cosmic microwave background (CMB) anisotropies

90

ROBERT H. BRANDENBERGER

in a spatially ﬂat Universe with only baryonic matter - Model A - to the corresponding anisotropies in a ﬂat Universe with mostly cold dark matter (pressure-less non-baryonic dark matter) - Model B. We start with the observationally known amplitude of the relative density ﬂuctuations today (time t0 ), and we use the fact that the amplitude of the CMB anisotropies on the angular scale θ(k) corresponding to the comoving wavenumber k is set by the primordial value of the gravitational potential φ - introduced in the following section - which in turn is related to the primordial value of the density ﬂuctuations at Hubble radius crossing (and not to its value of the time trec ). See e.g. Chapter 17 of [5]). In Model A, the dominant component of the pressure-less matter is coupled to radiation between teq and trec , the time of last scattering. Thus, the Jeans length is comparable to the Hubble radius. Therefore, for comoving galactic scales, k kJ in this time interval, and thus the fractional density contrast decreases as a(t)−1/2 . In contrast, in Model B, the dominant component of pressure-less matter couples only weakly to radiation, and hence the Jeans length is negligibly small. Thus, in Model B, the relative density contrast grows as a(t) between teq and trec . In the time interval trec < t < t0 , the ﬂuctuations scale identically in Models A and B. Summarizing, we conclude, working backwards in time from a ﬁxed amplitude of δk today, that the amplitudes of δk (teq ) in Models A and B (and thus their primordial values) are related by δk (teq )|A

a(trec ) δk (teq )|B . a(teq )

(3.16)

Hence, in Model A (without non-baryonic dark matter) the CMB anisotropies are predicted to be a factor of about 30 larger [76] than in Model B, way in excess of the recent observational results. This is one of the strongest arguments for the existence of nonbaryonic dark matter. Note that the precise value of the enhancement factor depends on the value of the cosmological constant Λ - the above value holds for Λ = 0. 3.4. CHARACTERIZING PERTURBATIONS Let us consider perturbations on a ﬁxed comoving length scale given by a comoving wavenumber k. The corresponding physical length increases as a(t). This is to be compared to the Hubble radius H −1 (t) which scales as t provided a(t) grows as a power of t. In the late time Universe, a(t) ∼ t1/2 in the radiation-dominated phase (i.e. for t < teq , and a(t) ∼ t2/3 in the matter-dominated period (teq < t < t0 ). Thus, we see that at sufﬁciently early times, all comoving scales had a physical length larger than the Hubble radius. If we consider large cosmological scales (e.g. those corresponding to the observed CMB anisotropies or to galaxy clusters), the time tH (k) of “Hubble radius crossing” (when the physical length was equal to the Hubble radius) was in fact later than teq . As we will see in later sections, the time of Hubble radius crossing plays an important role in the evolution of cosmological perturbations. Cosmological ﬂuctuations can be described either in position space or in momentum space. In position space, we compute the root mean square mass ﬂuctuation δM/M (k, t) in a sphere of radius l = 2π/k at time t. A scale-invariant spectrum of ﬂuctuations is deﬁned by the relation δM (k, tH (k)) = const. . (3.17) M

COSMOLOGICAL PERTURBATIONS

91

Such a spectrum was ﬁrst suggested by Harrison [42] and Zeldovich [43] as a reasonable choice for the spectrum of cosmological ﬂuctuations. We can introduce the “spectral index” n of cosmological ﬂuctuations by the relation δM 2 (k, tH (k)) ∼ k n−1 , M

(3.18)

and thus a scale-invariant spectrum corresponds to n = 1. To make the transition to the (more frequently used) momentum space representation, we Fourier decompose the fractional spatial density contrast d3 k δ˜ (k, t)eik·x . (3.19) δ (x, t) = The power spectrum Pδ of density ﬂuctuations is deﬁned by Pδ (k) = k 3 |δ˜ (k)|2 ,

(3.20)

where k is the magnitude of k, and we have assumed for simplicity a Gaussian distribution of ﬂuctuations in which the amplitude of the ﬂuctuations only depends on k. We can also deﬁne the power spectrum of the gravitational potential ϕ: 2 ˜ . Pϕ (k) = k 3 |δϕ(k)|

(3.21)

These two power spectra are related by the Poisson equation (3.2) Pϕ (k) ∼ k −4 Pδ (k) .

(3.22)

In general, the condition of scale-invariance is expressed in momentum space in terms of the power spectrum evaluated at a ﬁxed time. To obtain this condition, we ﬁrst use the time dependence of the fractional density ﬂuctuation from (3.14) to determine the mass ﬂuctuations at a ﬁxed time t > tH (k) > teq (the last inequality is a condition on the scales considered) δM 2 t 4/3 δM 2 (k, t) = (k, tH (k)) . M tH (k) M

(3.23)

The time of Hubble radius crossing is given by

and thus

a(tH (k))k −1 = 2tH (k) ,

(3.24)

tH (k)1/2 ∼ k −1 .

(3.25)

Inserting this result into (3.23) making use of (3.18) we ﬁnd δM 2 (k, t) ∼ k n+3 . M

(3.26)

Since, for reasonable values of the index of the power spectrum, δM/M (k, t) is dominated by the Fourier modes with wavenumber k, we ﬁnd that (3.26) implies |δ˜ |2 ∼ k n ,

(3.27)

92

ROBERT H. BRANDENBERGER

or, equivalently, Pϕ (k) ∼ k n−1 .

(3.28)

4. Relativistic Theory of Cosmological Fluctuations 4.1. INTRODUCTION The Newtonian theory of cosmological ﬂuctuations discussed in the previous section breaks down on scales larger than the Hubble radius because it neglects perturbations of the metric, and because on large scales the metric ﬂuctuations dominate the dynamics. Let us begin with a heuristic argument to show why metric ﬂuctuations are important on scales larger than the Hubble radius. For such inhom*ogeneities, one should be able to approximately describe the evolution of the space-time by applying the ﬁrst FRW equation (2.4) of hom*ogeneous and isotropic cosmology to the local Universe (this approximation is made more rigorous in [44]). Based on this equation, a large-scale ﬂuctuation of the energy density will lead to a ﬂuctuation (“δa”) of the scale factor a which grows in time. This is due to the fact that self gravity ampliﬁes ﬂuctuations even on length scales λ greater than the Hubble radius. This argument is made rigorous in the following analysis of cosmological ﬂuctuations in the context of general relativity, where both metric and matter inhom*ogeneities are taken into account. We will consider ﬂuctuations about a hom*ogeneous and isotropic background cosmology, given by the metric (2.1), which can be written in conformal time η (deﬁned by dt = a(t)dη) as (4.1) ds2 = a(η)2 dη 2 − dx2 . The theory of cosmological perturbations is based on expanding the Einstein equations to linear order about the background metric. The theory was initially developed in pioneering works by Lifsh*tz [45]. Signiﬁcant progress in the understanding of the physics of cosmological ﬂuctuations was achieved by Bardeen [46] who realized the importance of subtracting gauge artifacts (see below) from the analysis (see also [47]). The following discussion is based on Part I of the comprehensive review article [5]. Other reviews - in some cases emphasizing different approaches - are [48, 49, 50, 51]. 4.2. CLASSIFYING FLUCTUATIONS The ﬁrst step in the analysis of metric ﬂuctuations is to classify them according to their transformation properties under spatial rotations. There are scalar, vector and second rank tensor ﬂuctuations. In linear theory, there is no coupling between the different ﬂuctuation modes, and hence they evolve independently (for some subtleties in this classiﬁcation, see [52]). (0) We begin by expanding the metric about the FRW background metric gµν given by (4.1): (0) + δgµν . (4.2) gµν = gµν The background metric depends only on time, whereas the metric ﬂuctuations δgµν depend on both space and time. Since the metric is a symmetric tensor, there are at ﬁrst sight 10 ﬂuctuating degrees of freedom in δgµν .

COSMOLOGICAL PERTURBATIONS

93

There are four degrees of freedom which correspond to scalar metric ﬂuctuations (the only four ways of constructing a metric from scalar functions): 2φ −B,i δgµν = a2 , (4.3) −B,i 2 ψδij − E,ij where the four ﬂuctuating degrees of freedom are denoted (following the notation of [5]) φ, B, E, and ψ, a comma denotes the ordinary partial derivative (if we had included spatial curvature of the background metric, it would have been the covariant derivative with respect to the spatial metric), and δij is the Kronecker symbol. There are also four vector degrees of freedom of metric ﬂuctuations, consisting of the four ways of constructing metric ﬂuctuations from three vectors: 0 −Si 2 , (4.4) δgµν = a −Si Fi,j + Fj,i where Si and Fi are two divergence-less vectors (for a vector with non-vanishing divergence, the divergence contributes to the scalar gravitational ﬂuctuation modes). Finally, there are two tensor modes which correspond to the two polarization states of gravitational waves: 0 0 2 , (4.5) δgµν = −a 0 hij where hij is trace-free and divergence-less hii = hjij = 0 .

(4.6)

Gravitational waves do not couple at linear order to the matter ﬂuctuations. Vector ﬂuctuations decay in an expanding background cosmology and hence are not usually cosmologically important. The most important ﬂuctuations, at least in inﬂationary cosmology, are the scalar metric ﬂuctuations, the ﬂuctuations which couple to matter inhom*ogeneities and which are the relativistic generalization of the Newtonian perturbations considered in the previous section. 4.3. GAUGE TRANSFORMATION The theory of cosmological perturbations is at ﬁrst sight complicated by the issue of gauge invariance (at the ﬁnal stage, however, we will see that we can make use of the gauge freedom to substantially simplify the theory). The coordinates t, x of space-time carry no independent physical meaning. They are just labels to designate points in the space-time manifold. By performing a small-amplitude transformation of the space-time coordinates (called “gauge transformation” in the following), we can easily introduce “ﬁctitious” ﬂuctuations in a hom*ogeneous and isotropic Universe. These modes are “gauge artifacts”. We will in the following take an “active” view of gauge transformation. Let us consider two space-time manifolds, one of them a hom*ogeneous and isotropic Universe M0 , the other a physical Universe M with inhom*ogeneities. A choice of coordinates can be considered to be a mapping D between the manifolds M0 and M. Let us consider a

94

ROBERT H. BRANDENBERGER

˜ which will map the same point (e.g. the origin of a ﬁxed coordinate second mapping D ˜ we system) in M0 into different points in M. Using the inverse of these maps D and D, can assign two different sets of coordinates to points in M. Consider now a physical quantity Q (e.g. the Ricci scalar) on M, and the corresponding physical quantity Q(0) on M0 Then, in the ﬁrst coordinate system given by the mapping D, the perturbation δQ of Q at the point p ∈ M is deﬁned by (4.7) δQ(p) = Q(p) − Q(0) D−1 (p) . ˜ the perturbation is deﬁned by Analogously, in the second coordinate system given by D, ˜ ˜ −1 (p) . δQ(p) = Q(p) − Q(0) D (4.8) The difference ˜ ∆Q(p) = δQ(p) − δQ(p)

(4.9)

is obviously a gauge artifact and carries no physical signiﬁcance. Some of the metric perturbation degrees of freedom introduced in the ﬁrst subsection are gauge artifacts. To isolate these, we must study how coordinate transformations act on the metric. There are four independent gauge degrees of freedom corresponding to the coordinate transformation ˜µ = xµ + ξ µ . (4.10) xµ → x The zero (time) component ξ 0 of ξ µ leads to a scalar metric ﬂuctuation. The spatial three vector ξ i can be decomposed as i + γ ij ξ,j ξ i = ξtr

(4.11)

i which has two (where γ ij is the spatial background metric) into a transverse piece ξtr degrees of freedom which yield vector perturbations, and the second term (given by the gradient of a scalar ξ) which gives a scalar ﬂuctuation. To summarize this paragraph, there are two scalar gauge modes given by ξ 0 and ξ, and two vector modes given by i . Thus, there remain two physical scalar and two vector the transverse three vector ξtr ﬂuctuation modes. The gravitational waves are gauge-invariant. Let us now focus on how the scalar gauge transformations (i.e. the transformations given by ξ 0 and ξ) act on the scalar metric ﬂuctuation variables φ, B, E, and ψ. An immediate calculation yields:

φ˜

=

˜ B ˜ E

=

ψ˜

=

=

a 0 ξ − (ξ 0 ) a B + ξ0 − ξ

φ−

(4.12)

E−ξ a ψ + ξ0 , a

where a prime indicates the derivative with respect to conformal time η. There are two approaches to deal with the gauge ambiguities. The ﬁrst is to ﬁx a gauge, i.e. to pick conditions on the coordinates which completely eliminate the gauge freedom, the second is to work with a basis of gauge-invariant variables.

COSMOLOGICAL PERTURBATIONS

95

If one wants to adopt the gauge-ﬁxed approach, there are many different gauge choices. Note that the often used synchronous gauge determined by δg 0µ = 0 does not totally ﬁx the gauge. A convenient system which completely ﬁxes the coordinates is the so-called longitudinal or conformal Newtonian gauge deﬁned by B = E = 0. If one prefers a gauge-invariant approach, there are many choices of gauge-invariant variables. A convenient basis ﬁrst introduced by [46] is the basis Φ, Ψ given by Φ

=

Ψ

=

1 (B − E )a a a ψ − (B − E ) . a φ+

(4.13) (4.14)

It is obvious from the above equations that the gauge-invariant variables Φ and Ψ coincide with the corresponding diagonal metric perturbations φ and ψ in longitudinal gauge. Note that the variables deﬁned above are gauge-invariant only under linear space-time coordinate transformations. Beyond linear order, the structure of perturbation theory becomes much more involved. In fact, one can show [53] that the only ﬂuctuation variables which are invariant under all coordinate transformations are perturbations of variables which are constant in the background space-time. 4.4. EQUATIONS OF MOTION We begin with the Einstein equations Gµν = 8πGTµν ,

(4.15)

where Gµν is the Einstein tensor associated with the space-time metric gµν , and Tµν is the energy-momentum tensor of matter, insert the ansatz for metric and matter perturbed (0) about a FRW background gµν (η), ϕ(0) (η) : gµν (x, η) ϕ(x, η)

(0) = gµν (η) + δgµν (x, η) = ϕ0 (η) + δϕ(x, η) ,

(4.16) (4.17)

(where we have for simplicity replaced general matter by a scalar matter ﬁeld ϕ) and expand to linear order in the ﬂuctuating ﬁelds, obtaining the following equations: δGµν = 8πGδTµν .

(4.18)

In the above, δgµν is the perturbation in the metric and δϕ is the ﬂuctuation of the matter ﬁeld ϕ. Note that the components δGµν and δTνµ are not gauge invariant. If we want to use the gauge-invariant approach, we note [5] that it is possible to construct a gauge-invariant (gi) µ via tensor δGν (gi) 0

≡

(gi) 0

≡

(gi) i

≡

δG0 δGi

δGj

δG00 + ((0) G00 )(B − E ) 1 δG0i + ((0) G0i − (0) Gkk )(B − E ),i 3 δGij + ((0) Gj i )(B − E ) ,

(4.19)

96

ROBERT H. BRANDENBERGER

where ( 0)Gµν denote the background values of the Einstein tensor. Analogously, a gauge(gi) µ invariant linearized stress-energy tensor δTν can be deﬁned. In terms of these tensors, the gauge-invariant form of the equations of motion for linear ﬂuctuations reads (gi) (gi) = 8πGδTµν . δGµν

(4.20)

If we insert into this equation the ansatz for the general metric and matter ﬂuctuations (which depend on the gauge), only gauge-invariant combinations of the ﬂuctuation variables will appear. In a gauge-ﬁxed approach, one can start with the metric in longitudinal gauge

(4.21) ds2 = a2 (1 + 2φ)dη 2 − (1 − 2ψ)γij dxi dxj and insert this ansatz into the general perturbation equations (4.18). The shortcut of inserting a restricted ansatz for the metric into the action and deriving the full set of variational equations is justiﬁed in this case. Both approaches yield the following set of equations of motion: −3H HΦ + Ψ + ∇2 Ψ HΦ + Ψ ,i

2H + H2 Φ + HΦ + Ψ + 2HΨ δji 1 1 + ∇2 Dδji − γ ik D,kj 2 2

(gi) 0

=

4πGa2 δT0

=

4πGa2 δTi

(gi) 0

(4.22)

(gi) i

= −4πGa2 δTj

,

where D ≡ Φ − Ψ and H = a /a. If we work in longitudinal gauge, then δTj = δTji , Φ = φ and Ψ = ψ. The ﬁrst conclusion we can draw is that if no anisotropic stress is present in the matter at linear order in ﬂuctuating ﬁelds, i.e. δTji = 0 for i = j, then the two metric ﬂuctuation variables coincide: Φ = Ψ. (4.23) (gi) i

This will be the case in most simple cosmological models, e.g. in theories with matter described by a set of scalar ﬁelds with canonical form of the action, and in the case of a perfect ﬂuid with no anisotropic stress. Let us now restrict our attention to the case of matter described in terms of a single scalar ﬁeld ϕ which can be expanded as ϕ(x, η) = ϕ0 (η) + δϕ(x, η)

(4.24)

in terms of background matter ϕ0 and matter ﬂuctuation δϕ(x, η). Then, in longitudinal gauge, (4.22) reduce to the following set of equations of motion (making use of (4.23)) ∇2 φ − 3Hφ − H + 2H2 φ

φ + Hφ φ + 3Hφ + H + 2H2 φ

= = =

4πG ϕ0 δϕ + V a2 δϕ

4πGϕ0 δϕ 4πG ϕ0 δϕ − V a2 δϕ ,

(4.25)

COSMOLOGICAL PERTURBATIONS

97

where V denotes the derivative of V with respect to ϕ. These equations can be combined to give the following second order differential equation for the relativistic potential φ: ϕ0 ϕ0 2 (4.26) φ + 2 H − φ − ∇ φ + 2 H − H φ = 0. ϕ0 ϕ0 This is the ﬁnal result for the classical evolution of cosmological ﬂuctuations. First of all, we note the similarities with the equation (3.12) obtained in the Newtonian theory. The ﬁnal term in (4.26) is the force due to gravity leading to the instability, the second to last term is the pressure force leading to oscillations (relativistic since we are considering matter to be a relativistic ﬁeld), and the second term is the Hubble friction term. For each wavenumber there are two fundamental solutions. On small scales (k > H), the solutions correspond to damped oscillations, on large scales (k < H) the oscillations freeze out and the dynamics is governed by the gravitational force competing with the Hubble friction term. Note, in particular, how the Hubble radius naturally emerges as the scale where the nature of the ﬂuctuating modes changes from oscillatory to frozen. Considering the equation in a bit more detail, observe that if the equation of state of the background is independent of time (which will be the case if H = ϕ0 = 0), then in an expanding background, the dominant mode of (4.26) is constant, and the subdominant mode decays. If the equation of state is not constant, then the dominant mode is not constant in time. Speciﬁcally, at the end of inﬂation H < 0, and this leads to a growth of φ (see the following subsection). To study the quantitative implications of the equation of motion (4.26), it is convenient to introduce [54, 55] the variable ζ (which, up to correction term of the order ∇2 φ which is unimportant for large-scale ﬂuctuations, is equal to the curvature perturbation R in comoving gauge [56]) by 2 H −1 φ˙ + φ ζ ≡ φ+ , (4.27) 3 1+w where p (4.28) w= ρ characterizes the equation of state of matter. In terms of ζ, the equation of motion (4.26) takes on the form 3˙ ζH(1 + w) = O(∇2 φ) . (4.29) 2 On large scales, the right hand side of the equation is negligible, which leads to the conclusion that large-scale cosmological ﬂuctuations satisfy ˙ + w) = 0. ζ(1

(4.30)

This implies that ζ is constant except possibly if 1 + w = 0 at some point in time during the cosmological evolution (which occurs during reheating in inﬂationary cosmology if the inﬂaton ﬁeld undergoes oscillations - see [57] and [74, 59] for discussions of the consequences in single and double ﬁeld inﬂationary models, respectively). In single matter ﬁeld models it is indeed possible to show that ζ˙ = 0 on super-Hubble scales independent of assumptions on the equation of state [66, 67]. This “conservation law” makes it

98

ROBERT H. BRANDENBERGER

easy to relate initial ﬂuctuations to ﬁnal ﬂuctuations in inﬂationary cosmology, as will be illustrated in the following subsection. 4.5. APPLICATION TO INFLATIONARY COSMOLOGY Let us now return to the space-time sketch of the evolution of ﬂuctuations in inﬂationary cosmology - see Figure (1) - and use the conservation law (4.30) - in the form ζ = const on large scales - to relate the amplitude of φ at initial Hubble radius crossing during the inﬂationary phase (at t = ti (k)) with the amplitude at ﬁnal Hubble radius crossing at late times (at t = tf (k)). Since both at early times and at late times φ˙ = 0 on super-Hubble scales as the equation of state is not changing, (4.30) and (4.27) lead to φ(tf (k))

(1 + w)(tf (k)) φ(ti (k)) . (1 + w)(ti (k))

(4.31)

This equation will allow us to evaluate the amplitude of the cosmological perturbations when they re-enter the Hubble radius at time tf (k), under the assumption (discussed in detail in the following section) that the origin of the primordial ﬂuctuations is quantum vacuum oscillations. The time-time perturbed Einstein equation (the ﬁrst equation of (4.22)) relates the value of φ at initial Hubble radius crossing to the amplitude of the fractional energy density ﬂuctuations. This, together with the fact that the amplitude of the scalar matter ﬁeld quantum vacuum ﬂuctuations is of the order H, yields

V φ(ti (k)) ∼ H (ti (k)) . V

(4.32)

In the late time radiation dominated phase, w = 1/3, whereas during slow-roll inﬂation 1 + w(ti (k))

ϕ˙ 20 (ti (k)) . V

(4.33)

Making, in addition, use of the slow roll conditions satisﬁed during the inﬂationary period H ϕ˙ 0

H2

−V 8πG V, 3

(4.34)

we arrive at the ﬁnal result φ(tf (k)) ∼

V 3/2 (ti (k)) , V m3pl

(4.35)

which gives the position space amplitude of cosmological ﬂuctuations on a scale labelled by the comoving wavenumber k at the time when the scale re-enters the Hubble radius at late times, a result ﬁrst obtained in the case of the Starobinsky model [14] of inﬂation in [68], and later in the context of scalar ﬁeld-driven inﬂation in [69, 70, 71, 54]. In the case of slow roll inﬂation, the right hand side of (4.35) is, to a ﬁrst approximation, independent of k, and hence the resulting spectrum of ﬂuctuations is nearly scaleinvariant.

COSMOLOGICAL PERTURBATIONS

99

5. Quantum Theory of Cosmological Fluctuations 5.1. OVERVIEW As already mentioned in the previous section, in many models of the very early Universe, in particular in inﬂationary cosmology, but also in the Pre-Big-Bang and in the Ekpyrotic scenarios, primordial inhom*ogeneities emerge from quantum vacuum ﬂuctuations on microscopic scales (wavelengths smaller than the Hubble radius). The wavelength is then stretched relative to the Hubble radius, becomes larger than the Hubble radius at some time and then propagates on super-Hubble scales until re-entering at late cosmological times. In the context of a Universe with a de Sitter phase, the quantum origin of cosmological ﬂuctuations was ﬁrst discussed in [68] - see [72] for a more general discussion of the quantum origin of ﬂuctuations in cosmology, and also [73, 74] for earlier ideas. In particular, Mukhanov [68] and Press [73] realized that in an exponentially expanding background, the curvature ﬂuctuations would be scale-invariant, and Mukhanov provided a quantitative calculation which also yielded the logarithmic deviation from exact scaleinvariance. To understand the role of the Hubble radius, consider the equation of a free scalar matter ﬁeld ϕ on an unperturbed expanding background: ϕ¨ + 3H ϕ˙ −

∇2 ϕ = 0. a2

(5.1)

The second term on the left hand side of this equation leads to damping of ϕ with a characteristic decay rate given by H. As a consequence, in the absence of the spatial gradient term, ϕ˙ would be of the order of magnitude Hϕ. Thus, comparing the second and the third term on the left hand side, we immediately see that the microscopic (spatial gradient) term dominates on length scales smaller than the Hubble radius, leading to oscillatory motion, whereas this term is negligible on scales larger than the Hubble radius, and the evolution of ϕ is determined primarily by gravity. Note that in general cosmological models the Hubble radius is much smaller than the horizon (the forward light cone calculated from the initial time). In an inﬂationary universe, the horizon is larger by a factor of at least exp(N ), where N is the number of e-foldings of inﬂation, and the lower bound is taken on if the Hubble radius and horizon coincide until inﬂation begins. It is very important to realize this difference, a difference which is obscured in most articles on cosmology in which the term “horizon” is used when “Hubble radius” is meant. Note, in particular, that the hom*ogeneous inﬂaton ﬁeld contains causal information on superHubble but sub-horizon scales. Hence, it is completely consistent with causality [57] to have a microphysical process related to the background scalar matter ﬁeld lead to exponential ampliﬁcation of the amplitude of ﬂuctuations during reheating on such scales, as it does in models in which entropy perturbations are present and not suppressed during inﬂation [74, 59]. There are, however, general relativistic conservation laws [60] which imply that adiabatic ﬂuctuations produced locally must be Poisson-statistic suppressed on scales larger than the Hubble radius. For example, ﬂuctuations produced by the formation of topological defects at a phase transition in the early universe are initially isocurvature (entropy) in nature (see e.g. [61] for a discussion). Via the source term in the equation of motion

100

ROBERT H. BRANDENBERGER

(3.4), a growing adiabatic mode is induced, but at any ﬁxed time the spectrum of the curvature ﬂuctuation on scales larger than the Hubble radius has index n = 4 (Poisson). A similar conclusion applies to recently discussed models [62, 63] of modulated reheating as a new source of density perturbations (see [64] for a nice discussion), and to models in which moduli ﬁelds obtain masses after some symmetry breaking, their quantum ﬂuctuations then inducing cosmological ﬂuctuations. A prototypical example is given by axion ﬂuctuations in an inﬂationary universe (see e.g. [65] and references therein). To understand the generation and evolution of ﬂuctuations in current models of the very early Universe, we need both Quantum Mechanics and General Relativity, i.e. quantum gravity. At ﬁrst sight, we are thus faced with an intractable problem, since the theory of quantum gravity is not yet established. We are saved by the fact that today on large cosmological scales the fractional amplitude of the ﬂuctuations is smaller than 1. Since gravity is a purely attractive force, the ﬂuctuations had to have been - at least in the context of an eternally expanding background cosmology - very small in the early Universe. Thus, a linearized analysis of the ﬂuctuations (about a classical cosmological background) is self-consistent. From the classical theory of cosmological perturbations discussed in the previous section, we know that the analysis of scalar metric inhom*ogeneities can be reduced - after extracting gauge artifacts - to the study of the evolution of a single ﬂuctuating variable. Thus, we conclude that the quantum theory of cosmological perturbations must be reducible to the quantum theory of a single free scalar ﬁeld which we will denote by v. Since the background in which this scalar ﬁeld evolves is time-dependent, the mass of v will be time-dependent. The time-dependence of the mass will lead to quantum particle production over time if we start the evolution in the vacuum state for v. As we will see, this quantum particle production corresponds to the development and growth of the cosmological ﬂuctuations. Thus, the quantum theory of cosmological ﬂuctuations provides a consistent framework to study both the generation and the evolution of metric perturbations. The following analysis is based on Part II of [5]. 5.2. OUTLINE OF THE ANALYSIS In order to obtain the action for linearized cosmological perturbations, we expand the action to quadratic order in the ﬂuctuating degrees of freedom. The linear terms cancel because the background is taken to satisfy the background equations of motion. We begin with the Einstein-Hilbert action for gravity and the action of a scalar matter ﬁeld (for the more complicated case of general hydrodynamical ﬂuctuations the reader is referred to [5])

√ 1 1 R + ∂µ ϕ∂ µ ϕ − V (ϕ) , (5.2) S = d4 x −g − 16πG 2 where R is the Ricci curvature scalar. The simplest way to proceed is to work in longitudinal gauge, in which the metric and matter take the form

(5.3) ds2 = a2 (η) (1 + 2φ(η, x))dη 2 − (1 − 2ψ(t, x))dx2 ϕ(η, x)

=

ϕ0 (η) + δϕ(η, x) .

COSMOLOGICAL PERTURBATIONS

101

The next step is to reduce the number of degrees of freedom. First, as already mentioned in the previous section, the off-diagonal spatial Einstein equations force ψ = φ since δTji = 0 for scalar ﬁeld matter (no anisotropic stresses to linear order). The two remaining ﬂuctuating variables φ and ϕ must be linked by the Einstein constraint equations since there cannot be matter ﬂuctuations without induced metric ﬂuctuations. The two nontrivial tasks of the lengthy [5] computation of the quadratic piece of the action is to ﬁnd out what combination of ϕ and φ gives the variable v in terms of which the action has canonical kinetic term, and what the form of the time-dependent mass is. This calculation involves inserting the ansatz (4.7) into the action (2.1), expanding the result to second order in the ﬂuctuating ﬁelds, making use of the background and of the constraint equations, and dropping total derivative terms from the action. In the context of scalar ﬁeld matter, the quantum theory of cosmological ﬂuctuations was developed by Mukhanov [75, 76] (see also [77]). The result is the following contribution S (2) to the action quadratic in the perturbations: 1 z

(5.4) S (2) = d4 x v 2 − v,i v,i + v 2 , 2 z where the canonical variable v (the “Mukhanov variable” introduced in [76] - see also [72]) is given by ϕ

v = a δϕ + 0 φ , (5.5) H with H = a /a, and where aϕ0 . (5.6) z = H

In both the cases of power law inﬂation and slow roll inﬂation, H and ϕ0 are proportional and hence (as long as the equation of state does not change over time) z(η) ∼ a(η) .

(5.7)

Note that the variable v is related to the curvature perturbation R in comoving coordinates introduced in [56] and closely related to the variable ζ used in [54, 55]: v = zR .

(5.8)

The equation of motion which follows from the action (5.4) is

v − ∇2 v − or, in momentum space,

z v = 0, z

(5.9)

z vk = 0 , (5.10) z where vk is the k’th Fourier mode of v. As a consequence of (5.7), the mass term in the above equation is given by the Hubble scale

vk + k 2 vk −

2 kH

z H2 . ≡ z

(5.11)

102

ROBERT H. BRANDENBERGER

Thus, it immediately follows from (5.10) that on small length scales, i.e. for k > kH , the solutions for vk are constant amplitude oscillations . These oscillations freeze out at Hubble radius crossing, i.e. when k = kH . On longer scales (k kH ), the solutions for vk increase as z: (5.12) vk ∼ z , k kH . Given the action (5.4), the quantization of the cosmological perturbations can be performed by canonical quantization (in the same way that a scalar matter ﬁeld on a ﬁxed cosmological background is quantized [1]). The ﬁnal step in the quantum theory of cosmological perturbations is to specify an initial state. Since in inﬂationary cosmology all pre-existing classical ﬂuctuations are redshifted by the accelerated expansion of space, one usually assumes (we will return to a criticism of this point when discussing the trans-Planckian problem of inﬂationary cosmology) that the ﬁeld v starts out at the initial time ti mode by mode in its vacuum state. Two questions immediately emerge: what is the initial time ti , and which of the many possible vacuum states should be chosen. It is usually assumed that since the ﬂuctuations only oscillate on sub-Hubble scales, the choice of the initial time is not important, as long as it is earlier than the time when scales of cosmological interest today cross the Hubble radius during the inﬂationary phase. The state is usually taken to be the Bunch-Davies vacuum (see e.g. [1]), since this state is empty of particles at ti in the coordinate frame determined by the FRW coordinates (see e.g. [79] for a discussion of this point), and since the Bunch-Davies state is a local attractor in the space of initial states in an expanding background (see e.g. [80]). Thus, we choose the initial conditions 1 vk (ηi ) = √ 2ωk √ ωk vk (ηi ) = √ 2

(5.13)

where here ωk = k, and ηi is the conformal time corresponding to the physical time ti . Let us brieﬂy summarize the quantum theory of cosmological perturbations. In the linearized theory, ﬂuctuations are set up at some initial time ti mode by mode in their vacuum state. While the wavelength is smaller than the Hubble radius, the state undergoes quantum vacuum ﬂuctuations. The accelerated expansion of the background redshifts the length scale beyond the Hubble radius. The ﬂuctuations freeze out when the length scale is equal to the Hubble radius. On larger scales, the amplitude of vk increases as the scale factor. This corresponds to the squeezing of the quantum state present at Hubble radius crossing (in terms of classical general relativity, it is self-gravity which leads to this growth of ﬂuctuations). As discussed e.g. in [81], the squeezing of the quantum vacuum state sets up the classical correlations in the wave function of the ﬂuctuations which are an essential ingredient in the classicalization of the perturbations. 5.3. APPLICATION TO INFLATIONARY COSMOLOGY We will now use the quantum theory of cosmological perturbations developed above to calculate the spectrum of curvature ﬂuctuations in inﬂationary cosmology.

COSMOLOGICAL PERTURBATIONS

103

We need to compute the power spectrum PR (k) of the curvature ﬂuctuation R deﬁned in (5.8), namely H (5.14) R = z −1 v = φ + δϕ ϕ0 The idea in calculating the power spectrum at a late time t is to ﬁrst relate the power spectrum via the growth rate (5.12) of v on super-Hubble scales to the power spectrum at the time tH (k) of Hubble radius crossing, and to then use the constancy of the amplitude of v on sub-Hubble scales to relate it to the initial conditions (5.13). Thus PR (k, t) ≡ k 3 R2k (t)

= = = ∼

k 3 z −2 (t)|vk (t)|2 z(t) 2 |vk (tH (k))|2 k 3 z −2 (t) z(tH (k))

(5.15)

k 3 z −2 (tH (k))|vk (tH (k))|2 k 3 a−2 (tH (k))|vk (ti )|2 ,

where in the ﬁnal step we have used (5.7) and the constancy of the amplitude of v on sub-Hubble scales. Making use of the condition a−1 (tH (k))k = H

(5.16)

for Hubble radius crossing, and of the initial conditions (5.13), we immediately see that PR (k, t) ∼ k 3 k −2 k −1 H 2 ,

(5.17)

and that thus a scale invariant power spectrum with amplitude proportional to H 2 results, in agreement with what was argued on heuristic grounds in Section (4.5). 5.4. QUANTUM THEORY OF GRAVITATIONAL WAVES The quantization of gravitational waves parallels the quantization of scalar metric ﬂuctuations, but is more simple because there are no gauge ambiguities. Note that at the level of linear ﬂuctuations, scalar metric ﬂuctuations and gravitational waves are independent. Both can be quantized on the same cosmological background determined by the background scale factor and the background matter. However, in contrast to the case of scalar metric ﬂuctuations, the tensor modes are also present in pure gravity (i.e. in the absence of matter). Starting point is the action (2.1). Into this action we insert the metric which corresponds to a classical cosmological background plus tensor metric ﬂuctuations:

(5.18) ds2 = a2 (η) dη 2 − (δij + hij )dxi dxj , where the second rank tensor hij (η, x) represents the gravitational waves, and in turn can be decomposed as x (5.19) hij (η, x) = h+ (η, x)e+ ij + hx (η, x)eij x into the two polarization states. Here, e+ ij and eij are two ﬁxed polarization tensors, and h+ and hx are the two coefﬁcient functions.

104

ROBERT H. BRANDENBERGER

To quadratic order in the ﬂuctuating ﬁelds, the action consists of separate terms involving h+ and hx . Each term is of the form

a2 2 (5.20) h − (∇h)2 , d4 x S (2) = 2 leading to the equation of motion

hk + 2

a h + k 2 hk = 0 . a k

(5.21)

The variable in terms of which the action (5.20) has canonical kinetic term is µk ≡ ahk ,

(5.22)

a µk = 0 . µk + k 2 − a

(5.23)

and its equation of motion is

This equation is very similar to the corresponding equation (5.10) for scalar gravitational inhom*ogeneities, except that in the mass term the scale factor a(η) replaces z(η), which leads to a very different evolution of scalar and tensor modes during the reheating phase in inﬂationary cosmology during which the equation of state of the background matter changes dramatically. Based on the above discussion we have the following theory for the generation and evolution of gravitational waves in an accelerating Universe (ﬁrst developed by Grishchuk [82]): waves exist as quantum vacuum ﬂuctuations at the initial time on all scales. They oscillate until the length scale crosses the Hubble radius. At that point, the oscillations freeze out and the quantum state of gravitational waves begins to be squeezed in the sense that (5.24) µk (η) ∼ a(η) , which, from (5.22) corresponds to constant amplitude of hk . The squeezing of the vacuum state leads to the emergence of classical properties of this state, as in the case of scalar metric ﬂuctuations. 6. Conceptual Problems of Inﬂationary Cosmology After this detailed survey of the theory of cosmological perturbations applied to inﬂationary cosmology we can now turn to some conceptual problems of cosmological inﬂation, and ways in which string theory may help address these issues. The ﬁrst two problems relate to the cosmological perturbations we have just discussed. The ﬁrst problem concerns the amplitude of the cosmological ﬂuctuations. Considering the simplest large-ﬁeld potential V (ϕ) =

1 2 2 m ϕ , 2

(6.1)

the result (4.35) for the amplitude of the gravitational potential φ at late times and large scales (which modulo a factor of order unity gives the amplitude of the CMB ﬂuctuations

COSMOLOGICAL PERTURBATIONS

105

on large angular scales and hence should be of the order 10−5 ) yields (making use of the fact that the result from (4.35) must be evaluated for ﬁeld values ϕ ∼ mpl when the relevant scales exit the Hubble radius) φ(tf (k)) ∼

m . mpl

(6.2)

Hence, the value of m must be chosen to be about 1013 GeV, introducing a new hierarchy problem into particle physics model building. In a model with quartic potential V (ϕ) =

1 4 λϕ 4

(6.3)

we obtain a severe constraint on the value of the self coupling constant (λ 10−10 modulo factors of 2π), a constraint which implies that the inﬂaton cannot be in thermal equilibrium before inﬂation. To reach this conclusion we have used “naturalness” considerations on coupling constants which state that the lower bound on the self coupling constant λ implies lower bounds on the coupling constants describing interactions of ϕ with other matter ﬁelds, since such interactions generate at higher loop order contributions to the renormalized value of λ. As shown in [83], this hierarchy problem is quite general. As has recently been shown [59], this problem is worse in many inﬂationary models with entropy ﬂuctuations. If the entropy perturbations are not suppressed during inﬂation, they can be parametrically ampliﬁed during reheating [84, 57, 74, 59]. This results in ﬂuctuations which are nonlinear after inﬂation independent of the values of the coupling constants, a result derived including the back-reaction of these ﬂuctuations in the Hartree approximation [85]. Such models are thus phenomenologically ruled out. The second problem is more important and will be discussed at length in the next section. Basically, since in most scalar ﬁeld-driven inﬂationary models the period of inﬂation lasts much longer than the minimal number of e-foldings required in order that scales of current cosmological interest start out inside the Hubble radius at the beginning of inﬂation, in such models these scales thus originate with a wavelength much smaller than the Planck length, and hence the justiﬁcation for using the formalism of the previous section to compute the evolution of ﬂuctuations is doubtful. This is the trans-Planckian problem for inﬂationary cosmology [9] which becomes the trans-Planckian window of opportunity for string theory. Scalar ﬁeld-driven inﬂationary models have been shown to be geodesically incomplete in the past [86]. Hence, we know that this model cannot describe the very early Universe. A major challenge for string cosmology is to provide this description. Most importantly, scalar ﬁeld-driven inﬂation uses the time-independent part of V (ϕ) to generate inﬂation. However, it is observationally known that (but theoretically not understood why) the time independent quantum vacuum contribution to the energy density of any quantum ﬁeld does not gravitate. This is the famous cosmological constant problem. It may turn out that the solution of the cosmological constant problem will remove not only quantum vacuum energy but also the part of V (ϕ) which generates inﬂation. I view this issue as the Achilles heel of scalar ﬁeld-driven inﬂationary cosmology.

106

ROBERT H. BRANDENBERGER

Finally, standard model particle physics does not provide a candidate for the inﬂaton. Models of particle physics beyond the standard model open the window for providing realizations of inﬂation. Why can string theory help? First of all, string theory contains many ﬁelds which are massless in the early Universe, namely the moduli ﬁelds. Thus, it provides many candidates for an inﬂaton. In addition, it is quite possible that the hierarchy of scales required to give the right magnitude of density ﬂuctuations emerges from a hierarchy of symmetry breaking scales in string theory. Secondly, string theory is supposed to describe the physics on all scales. Thus, in string cosmology the equations for evolving the ﬂuctuations will be unambiguous (even if they are not known today), and the trans-Planckian problem will be solved. One of the main goals of string theory is to provide a nonsingular cosmology (and a concrete realization of this goal was provided in [87]). Thus, string theory should be able to provide a consistent theory of the very early Universe. This theory might connect with late-time cosmology through a period of inﬂation, or through scenarios more similar to [7, 8]. 7. The Trans-Planckian Window for Superstring Cosmology The same background dynamics which yields the causal generation mechanism for cosmological ﬂuctuations, the most spectacular success of inﬂationary cosmology, bears in it the nucleus of the “trans-Planckian problem”. This can be seen from Fig. (2). If inﬂation lasts only slightly longer than the minimal time it needs to last in order to solve the horizon problem and to provide a causal generation mechanism for CMB ﬂuctuations, then the corresponding physical wavelength of these ﬂuctuations is smaller than the Planck length at the beginning of the period of inﬂation. The theory of cosmological perturbations is based on classical general relativity coupled to a weakly coupled scalar ﬁeld description of matter. Both the theories of gravity and of matter will break down on trans-Planckian scales, and this immediately leads to the trans-Planckian problem: are the predictions of standard inﬂationary cosmology robust against effects of trans-Planckian physics [9]? This question has recently been addressed using a variety of techniques. The simplest method is to replace the usual dispersion relation for cosmological perturbations by an adhoc modiﬁed relation, as was done in [88, 89] in the context of studying the dependence of the thermal spectrum of black hole radiation on trans-Planckian physics. We will discuss the application of this method to cosmology [90, 91, 92] below. Other methods include considerations of modiﬁcations of the evolution of cosmological ﬂuctuations coming from string-motivated space-space uncertainty relations [93, 94, 95, 96, 97], string-motivated space-time uncertainty relations [98] (reviewed below), from minimal length considerations [99], from an effective action analysis [100], from a minimal trans-Planckian physics viewpoint (starting each mode out in the vacuum state of the usual action for cosmological perturbations at the time when the physical wavelength is equal to the new fundamental length) [101, 102], and from the point of view of boundary renormalization group analysis [103]. The simplest way of modeling the possible effects of trans-Planckian physics [90, 91, 92] on the evolution of cosmological perturbations, while keeping the mathematical analysis simple, is to replace the linear dispersion relation ωphys = kphys of the usual equation for cosmological perturbations by a non standard dispersion relation ωphys =

COSMOLOGICAL PERTURBATIONS

107

t

b

tR

xp

a

Phase III

t H (k)

Phase II c t i (k) d Phase I ti l

pl

Figure 2. Space-time diagram (physical distance vs. time) showing the origin of the trans-Planckian problem of inﬂationary cosmology: at very early times, the wavelength is smaller than the Planck scale Pl (Phase I), at intermediate times it is larger than Pl but smaller than the Hubble radius H −1 (Phase II), and at late times during inﬂation it is larger than the Hubble radius (Phase III). The line labeled a) is the physical wavelength associated with a ﬁxed comoving scale k. The line b) is the Hubble radius or horizon in SBB cosmology. Curve c) shows the Hubble radius during inﬂation. The horizon in inﬂationary cosmology is shown in curve d).

ωphys (kphys ) which differs from the standard one only for physical wavenumbers larger 2 (n, η) than the Planck scale. This amounts to replacing k 2 appearing in (5.10) with keﬀ deﬁned by k 2 2 2 2 k → keﬀ (k, η) ≡ a (η)ωphys . (7.1) a(η) For a ﬁxed comoving mode, this implies that the dispersion relation becomes time-dependent. Therefore, the equation of motion of the quantity vk (η) takes the form (with z(η) ∝ a(η)) a 2 vk + keﬀ (k, η) − (7.2) vk = 0 . a A more rigorous derivation of this equation, based on a variational principle, has been provided [104] (see also Ref. [105]). The evolution of modes thus must be considered separately in three phases, see Fig. (2). In Phase I the wavelength is smaller than the Planck scale, and trans-Planckian physics can play an important role. In Phase II, the wavelength is larger than the Planck scale but smaller than the Hubble radius. In this phase, trans-Planckian physics will have a negligible effect (this statement can be quantiﬁed [106]). Hence, by the analysis of the

108

ROBERT H. BRANDENBERGER

previous section, the wave function of ﬂuctuations is oscillating in this phase, vk = B1 exp(−ikη) + B2 exp(ikη)

(7.3)

with constant coefﬁcients B1 and B2 . In the standard approach, the initial conditions √ are ﬁxed in this region and the usual choice of the vacuum state leads to B1 = 1/ 2k, B2 = 0. Phase III starts at the time tH (k) when the mode crosses the Hubble radius. During this phase, the wave function is squeezed. One source of trans-Planckian effects [90, 91] on observations is the possible nonadiabatic evolution of the wave function during Phase I. If this occurs, then it is possible that the wave function of the ﬂuctuation mode is not in its vacuum state when it enters Phase II and, as a consequence, the coefﬁcients B1 and B2 are no longer given by the standard expressions above. In this case, the wave function will not be in its vacuum state when it crosses the Hubble radius, and the ﬁnal spectrum will be different. In particular, since short wavelength modes spend more time in the trans-Planckian phase, it is possible that a deviation of the spectrum from scale-invariance could be induced in a background which is expanding almost exponentially. This would be an order one effect on the spectrum of cosmological perturbations. In general, B1 and B2 are determined by the matching conditions between Phase I and II. If the dynamics is adiabatic throughout (in particular if the a /a term is negligible), the WKB approximation holds and the solution is always given by η 1 exp −i keﬀ dτ , (7.4) vk (η) 2keﬀ (k, η) ηi where ηi is some initial time. Therefore, if we start with a positive frequency solution only and use this solution, we ﬁnd that no negative frequency solution appears. Deep in Region II where keﬀ k the solution becomes 1 vk (η) √ exp(−iφ − ikη), 2k

(7.5)

i.e. the standard vacuum solution times a phase which will disappear when we calculate the modulus. To obtain a modiﬁcation of the inﬂationary spectrum, it is necessary to ﬁnd a dispersion relation such that the WKB approximation breaks down in Phase I. A concrete class of dispersion relations for which the WKB approximation breaks down is [89] pl 2m 2 2 2 keﬀ (k, η) = k − k |bm | , (7.6) λ(η) where λ(η) = 2πa(η)/k is the wavelength of a mode. If we follow the evolution of the modes in Phases I, II and III, matching the mode functions and their derivatives at the junction times, the calculation [90, 91, 107] demonstrates that the ﬁnal spectral index is modiﬁed and that superimposed oscillations appear. It has recently been shown [108] that in the case of this class of dispersion relations, the spectrum of black hole Hawking radiation is also affected. However, the above example suffers from several problems. First, in inﬂationary models with a long period of inﬂationary expansion, the dispersion relation (7.6) leads to complex frequencies at the beginning of inﬂation for scales which are of current interest in

COSMOLOGICAL PERTURBATIONS

109

Figure 3. Sketch of the dispersion relation of [104]. The adiabaticity condition on the evolution of ﬂuctuations is broken when the physical frequency (vertical axis) is smaller than the Hubble expansion rate. During the phase of inﬂation, this is the case when the physical wavenumber (horizontal axis) is in the interval between Λ1 and Λ2 . Ultraviolet scales are k > k0 , and the value k = k1 is the value when the dispersion relation turns over.

cosmology. Furthermore, the initial conditions for the Fourier modes of the ﬂuctuation ﬁeld have to be set in a region where the evolution is non-adiabatic and the use of the usual vacuum prescription can be questioned. These problems can be avoided in a toy model in which we follow the evolution of ﬂuctuations in a bouncing cosmological background which is asymptotically ﬂat in the past and in the future. The analysis of [109] shows that even in this case the ﬁnal spectrum of ﬂuctuations depends on the speciﬁc dispersion relation used. An example (see Fig. (3)) of a dispersion relation which breaks the WKB approximation in the trans-Planckian regime but does not lead to the problems mentioned in the previous paragraph was investigated in [104]. It is a dispersion relation which is linear for both small and large wavenumbers, but has an intermediate interval during which the frequency decreases as the wavenumber increases, much like what happens in (7.6). The violation of the WKB condition occurs for wavenumbers near the local minimum of the ω(k) curve. In this model, modes can be set up in the far ultraviolet in their Bunch-Davies vacuum. During the time interval when the physical wavenumber kp passes through the interval Λ1 < kp < Λ2 when the physical frequency is smaller than the value of the Hubble constant Hinf in the inﬂationary phase, the mode is squeezed and thus is no longer in the local vacuum state when kp < Λ1 . A justiﬁed criticism against the method summarized in the previous analysis is that the non-standard dispersion relations used are completely ad hoc, without a clear basis in trans-Planckian physics. Other recently explored approaches are motivated by string theory. For example, there has been a lot of recent work [93, 94, 95, 96, 97] on the implication of space-space uncertainty relations [110, 111] for the evolution of ﬂuctuations. The application of the uncertainty relations to the ﬂuctuations lead to two effects: Firstly, the equation of motion of the ﬂuctuations is modiﬁed. Secondly, for ﬁxed comoving length

110

ROBERT H. BRANDENBERGER

scale k, the uncertainty relation is saturated at critical time ti (k). Thus, in addition to a modiﬁcation of the evolution, trans-Planckian physics leads to a modiﬁcation of the boundary conditions for the ﬂuctuation modes. The upshot of this work is that the spectrum of ﬂuctuations is modiﬁed. The magnitude of the deviations can be of the order H/mpl , and is thus in principle measurable. In [98], the implications of the stringy space-time uncertainty relation [112, 113] ∆xphys ∆t ≥ ls2

(7.7)

on the spectrum of cosmological ﬂuctuations was studied. Again, application of this uncertainty relation to the ﬂuctuations leads to two effects. Firstly, the coupling between the background and the ﬂuctuations is nonlocal in time, thus leading to a modiﬁed dynamical equation of motion (a similar modiﬁcation also results [114] from quantum deformations, another example of a consequence of non-commutative basic physics). Secondly, the uncertainty relation is saturated at the time ti (k) when the physical wavelength equals the string scale ls . Before that time it does not make sense to talk about ﬂuctuations on that scale. By continuity, it makes sense to assume that ﬂuctuations on scale k are created at time ti (k) in the local vacuum state (the instantaneous WKB vacuum state). Let us for the moment neglect the nonlocal coupling between background and ﬂuctuation, and thus consider the usual equation of motion for ﬂuctuations in an accelerating background cosmology. We assume that a(t) scales as a power of time, i.e. we consider power-law inﬂation, and we distinguish two ranges of scales. Ultraviolet modes are generated at late times when the Hubble radius is larger than ls . On these scales, the spectrum of ﬂuctuations does not differ from what is predicted by the standard theory, since at the time of Hubble radius crossing the ﬂuctuation modes will be in their vacuum state. However, the evolution of the infrared modes which are created when the Hubble radius is smaller than ls is different. The ﬂuctuations undergo less squeezing than they do in the absence of the uncertainty relation, and hence the ﬁnal amplitude of ﬂuctuations is lower. From the equation (5.15) for the power spectrum of ﬂuctuations, and making use of the condition a(ti (k)) = kls (7.8) for the time ti (k) when the mode is generated, it follows immediately that the power spectrum is scale-invariant PR (k) ∼ k 0 . (7.9) In the standard scenario of power-law inﬂation the spectrum is red (PR (k) ∼ k n−1 with n < 1). Taking into account the effects of the nonlocal coupling between background and ﬂuctuation mode leads [98] to a modiﬁcation of this result: the spectrum of ﬂuctuations in a power-law inﬂationary background is in fact blue (n > 1). Note that, if we neglect the nonlocal coupling between background and ﬂuctuation mode, the result of (7.9) also holds in a cosmological background which is NOT accelerating. Thus, we have a method of obtaining a scale-invariant spectrum of ﬂuctuations without inﬂation. This result was also subsequently obtained in [115], however without a micro-physical basis for the prescription for the initial conditions. An approach to the trans-Planckian issue pioneered by Danielsson [101] which has recently received a lot of attention is to avoid the issue of the unknown trans-Planckian

COSMOLOGICAL PERTURBATIONS

111

physics and to start the evolution of the ﬂuctuation modes at the mode-dependent time when the wavelength equals the limiting scale. Obviously, the resulting spectrum will depend sensitively on which state is taken to be the initial state. The vacuum state is not unambiguous, and the choice of a state minimizing the energy density depends on the space-time splitting [102]. The signatures of this prescription are typically oscillations superimposed on the usual spectrum. The amplitude of this effect depends sensitively on the prescription of the initial state, and for a ﬁxed prescription also on the background cosmology. For a discussion of these issues and a list of references on this approach the reader is referred to [116]. Note, in particular, that for a ﬁxed background cosmology and for a ﬁxed initial condition prescription, the amplitude of the correction terms in the spectrum may be different for scalar cosmological perturbations on one hand and gravitational waves or test scalar matter ﬁelds on the other hand. If the ultraviolet modes are not in their vacuum state for wavelengths between the Planck or string scale and the Hubble radius, then the question of their back-reaction on the background geometry [117, 118, 119] arises. The naive expectation is that such ultraviolet modes have the equation of state of radiation, and that the mode occupation number hence must be very small (of order (Hinf /mpl )2 ) in order for the total energy density carried by these modes not to prevent inﬂation. This issue can be analyzed in the model of [104] in a setting in which there are no ultraviolet problems (in the naive approach one needs to assume that modes are continuously generated). In this approach, all modes start in their vacuum state, they are squeezed when Λ1 < kp < Λ2 , and they oscillate at an excited level thereafter. In an exponentially expanding background it is clear by time-translation invariance that the energy density in the ultraviolet modes is constant. The stress-energy density of a test scalar ﬁeld with this dispersion relation has recently been calculated [120], yielding the result that up to correction terms suppressed by (Hinf /mpl )2 , the equation of state is that of a cosmological constant. Thus, the back-reaction of the excited states does not prevent inﬂation but rather leads to a renormalization of the cosmological constant. The back-reaction does effect the speed of rolling of the scalar ﬁeld, and this yields constraints on the mode occupation numbers in the ultraviolet, constraints which, however, are much weaker than the ones conjectured in [117, 118, 119]. Another constraint on trans-Planckian physics arises from the observational limits on the ﬂux of ultra-high-energy cosmic rays. Such cosmic rays would be produced [121] in the present Universe if Trans-Planckian effects would lead to non-adiabatic mode evolution in the ultraviolet today. In the model of [104] this does not happen if the present Hubble rate H0 is smaller than the local minimum of the dispersion relation, as in the situation depicted in Fig. (3). In summary, due to the exponential red-shifting of wavelengths, present cosmological scales originate at wavelengths smaller than the Planck length early on during the period of inﬂation. Thus, Planck-scale physics may well encode information in these modes which can now be observed in the spectrum of microwave anisotropies. Two examples have been shown to demonstrate the existence of this “window of opportunity” to probe trans-Planckian physics in cosmological observations. The ﬁrst method makes use of modiﬁed dispersion relations to probe the robustness of the predictions of inﬂationary cosmology, the second applies the stringy space-time uncertainty relation on the ﬂuctuation modes. Both methods yield the result that trans-Planckian physics may lead to

112

ROBERT H. BRANDENBERGER

measurable effects in cosmological observables. 8. Conclusions These lectures have focused on the quantum theory of the generation and evolution of cosmological ﬂuctuations. This is the theory which connects the fundamental physics of the very early Universe with current cosmological observations. The general theory was illustrated in the context of the current paradigm of early universe cosmology, namely the inﬂationary universe scenario. In the context of the current approach to early universe cosmology which involves the coupling of matter described by quantum ﬁeld theory to classical general relativity, inﬂationary cosmology suffers from important conceptual problems. Resolving these problems is a major goal for superstring cosmology. Thus, I would argue that inﬂationary cosmology requires string theory. Whether string theory requires inﬂation is a different issue. String theory provides many candidates for an inﬂaton, and thus it is possible (maybe even likely) that inﬂation can be implemented in string theory. The study of this issue has recently attracted a lot of attention. However, one should keep in mind that current observations do not prove the correctness of inﬂation. They show that the spectrum of ﬂuctuations is nearly scaleinvariant and nearly adiabatic. However, there are other ways to obtain such a spectrum, although the existing alternatives such as the ad hoc model mentioned in the previous section of these lecture notes, or the Pre-Big-Bang [7] and Ekpyrotic [8] scenarios are either not based on fundamental physics or not (yet) as developed as inﬂationary cosmology. Thus, it may be possible that string theory connects with observations using a cosmological scenario different than inﬂation. Even in this case, however, the theory of cosmological perturbations developed here is applicable. A stringy early Universe may involve the dynamics of higher dimensions in an important way. This is the case for the Ekpyrotic scenario (see e.g. [122, 123] for some initial work). In this case, the formalism of cosmological perturbations needs to be extended (see e.g. [124] for a formalism which is a direct generalization of what was discussed here). There may be many interesting effects of bulk ﬂuctuation modes which cannot be seen in a four-dimensional effective ﬁeld theory approach. This is a rich ﬁeld which merits much more work. Acknowledgements The author wishes to thank the organizers of this Carg`ese school for the invitation to lecture and for their hospitality. He also wishes to thank Pei-Ming Ho and J´erˆome Martin for collaborations on the issue of the trans-Planckian problem of inﬂationary cosmology. He is grateful to the Perimeter Institute for hospitality and ﬁnancial support during the time when these lectures were prepared. He is supported in part by an NSERC Discovery grant (at McGill) and by the US Department of Energy under Contract DE-FG02-91ER40688, TASK A. References 1.

C. L. Bennett et al., Astrophys. J. Suppl. 148, 1 (2003) [arXiv:astro-ph/0302207].

COSMOLOGICAL PERTURBATIONS 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. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36.

113

M. Colless et al., arXiv:astro-ph/0306581. K. Abazajian et al. [SDSS Collaboration], Astron. J. 128, 502 (2004) [arXiv:astroph/0403325]. A. H. Guth, Phys. Rev. D 23, 347 (1981). A. Linde: Particle Physics and Inﬂationary Cosmology, (Harwood, Chur, 1990). V. Moncrief, J. Math. Phys. 17, 1893 (1976); V. Moncrief, Prepared for Directions in General Relativity: An International Symposium in Honor of the 60th Birthdays of Dieter Brill and Charles Misner, College Park, MD, 27-29 May 1993 M. Gasperini and G. Veneziano, M. Gasperini and G. Veneziano, Astropart. Phys. 1, 317 (1993) [arXiv:hep-th/9211021]. J. Khoury, B. A. Ovrut, P. J. Steinhardt and N. Turok, Phys. Rev. D 64, 123522 (2001) [arXiv:hep-th/0103239]. R. H. Brandenberger: Inﬂationary cosmology: Progress and problems. In: 1st Iranian International School on Cosmology: Large Scale Structure Formation, Kish Island, Iran, 22 Jan 4 Feb 1999 Kluwer, Dordrecht, 2000. (Astrophys. Space Sci. Libr. ; 247), ed. by R. Mansouri and R. Brandenberger (Kluwer, Dordrecht, 2000). [arXiv:hep-ph/9910410]. R. H. Brandenberger, Lect. Notes Phys. 646, 127 (2004) [arXiv:hep-th/0306071]. A. Vilenkin and E.P.S. Shellard; Cosmic Strings and Other Topological Defects, (Cambridge Univ. Press, Cambridge, 1994). M. B. Hindmarsh and T. W. Kibble, Rept. Prog. Phys. 58, 477 (1995) [arXiv:hep-ph/9411342]. R. H. Brandenberger, Int. J. Mod. Phys. A 9, 2117 (1994) [arXiv:astro-ph/9310041]. A. A. Starobinsky, Phys. Lett. B 91, 99 (1980). C. Armendariz-Picon, T. Damour and V. Mukhanov, Phys. Lett. B 458, 209 (1999) [arXiv:hepth/9904075]. R. H. Brandenberger and J. H. Kung, Phys. Rev. D 42, 1008 (1990). S. R. Coleman, Phys. Rev. D 15, 2929 (1977) [Erratum-ibid. D 16, 1248 (1977)]; C. G. . Callan and S. R. Coleman, Phys. Rev. D 16, 1762 (1977). R. H. Brandenberger, Rev. Mod. Phys. 57, 1 (1985). S. Dodelson, W. H. Kinney and E. W. Kolb, Phys. Rev. D 56, 3207 (1997) [arXiv:astroph/9702166]. A. D. Linde, Phys. Lett. B 108, 389 (1982); A. Albrecht and P. J. Steinhardt, Phys. Rev. Lett. 48, 1220 (1982). S. R. Coleman and E. Weinberg, Phys. Rev. D 7, 1888 (1973). D. S. Goldwirth and T. Piran, Phys. Rept. 214, 223 (1992). R. Brandenberger, G. Geshnizjani and S. Watson, Phys. Rev. D 67, 123510 (2003) [arXiv:hepth/0302222]. A. D. Linde, Phys. Lett. B 129, 177 (1983). A. Linde, arXiv:hep-th/0402051. H. A. Feldman and R. H. Brandenberger, Phys. Lett. B 227, 359 (1989); R. H. Brandenberger, H. Feldman and J. Kung, Phys. Scripta T36, 64 (1991). A. D. Linde, Phys. Rev. D 49, 748 (1994) [arXiv:astro-ph/9307002]. J. Polchinski, arXiv:hep-th/0412244. A. D. Dolgov and A. D. Linde, Phys. Lett. B 116, 329 (1982). L. F. Abbott, E. Farhi and M. B. Wise, Phys. Lett. B 117, 29 (1982). J. H. Traschen and R. H. Brandenberger, Phys. Rev. D 42, 2491 (1990). L. Kofman, A. D. Linde and A. A. Starobinsky, Phys. Rev. Lett. 73, 3195 (1994) [arXiv:hepth/9405187]. Y. Shtanov, J. H. Traschen and R. H. Brandenberger, Phys. Rev. D 51, 5438 (1995) [arXiv:hepph/9407247]. L. Kofman, A. D. Linde and A. A. Starobinsky, Phys. Rev. D 56, 3258 (1997) [arXiv:hepph/9704452]. A. Liddle and D. Lyth, Cosmological Inﬂation and Large-Scale Structure, (Cambridge Univ. Press, Cambridge, 2000). V. F. Mukhanov, H. A. Feldman and R. H. Brandenberger, Phys. Rept. 215, 203 (1992).

114 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. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82.

ROBERT H. BRANDENBERGER S. Weinberg: Gravitation and Cosmology, (Wiley, New York, 1972). P.J.E. Peebles: The Large-Scale Structure of the Universe, (Princeton Univ. Press, Princeton, 1980). T. Padmanabhan: Structure Formation in the Universe, (Cambridge Univ. Press, Cambridge, 1993). J. Peaco*ck: Cosmological Physics, (Cambridge Univ. Press, Cambridge, 1999). R. K. Sachs and A. M. Wolfe, Astrophys. J. 147, 73 (1967). E. R. Harrison, Phys. Rev. D 1, 2726 (1970). Y. B. Zeldovich, Mon. Not. Roy. Astron. Soc. 160, 1 (1972). N. Afshordi and R. H. Brandenberger, Phys. Rev. D 63, 123505 (2001) [arXiv:gr-qc/0011075]. E. Lifsh*tz, J. Phys. (USSR) 10, 116 (1946); E. M. Lifsh*tz and I. M. Khalatnikov, Adv. Phys. 12, 185 (1963). J. M. Bardeen, Phys. Rev. D 22, 1882 (1980). W. Press and E. Vishniac, Astrophys. J. 239, 1 (1980). H. Kodama and M. Sasaki, Prog. Theor. Phys. Suppl. 78, 1 (1984). M. Bruni, G. F. Ellis and P. K. Dunsby, Class. Quant. Grav. 9, 921 (1992). J. c. Hwang, Astrophys. J. 415, 486 (1993). R. Durrer, Helv. Phys. Acta 69, 417 (1996). J. Stewart, Class. Quant. Grav. 7, 1169 (1990). J. Stewart and M. Walker, Proc. R. Soc. London A 341, 49 (1974). J. M. Bardeen, P. J. Steinhardt and M. S. Turner, Phys. Rev. D 28, 679 (1983). R. H. Brandenberger and R. Kahn, Phys. Rev. D 29, 2172 (1984). D. H. Lyth, Phys. Rev. D 31, 1792 (1985). F. Finelli and R. H. Brandenberger, Phys. Rev. Lett. 82, 1362 (1999) [arXiv:hep-ph/9809490]. B. A. Bassett and F. Viniegra, Phys. Rev. D 62, 043507 (2000) [arXiv:hep-ph/9909353]. F. Finelli and R. H. Brandenberger, Phys. Rev. D 62, 083502 (2000) [arXiv:hep-ph/0003172]. J. H. Traschen, Phys. Rev. D 29, 1563 (1984). J. H. Traschen, N. Turok and R. H. Brandenberger, Phys. Rev. D 34, 919 (1986). G. Dvali, A. Gruzinov and M. Zaldarriaga, Phys. Rev. D 69, 023505 (2004) [arXiv:astroph/0303591]. L. Kofman, arXiv:astro-ph/0303614. F. Vernizzi, Phys. Rev. D 69, 083526 (2004) [arXiv:astro-ph/0311167]. M. Axenides, R. H. Brandenberger and M. S. Turner, Phys. Lett. B 126, 178 (1983). S. Weinberg, Phys. Rev. D 67, 123504 (2003) [arXiv:astro-ph/0302326]. W. B. Lin, X. H. Meng and X. M. Zhang, Phys. Rev. D 61, 121301 (2000) [arXiv:hepph/9912510]. V. F. Mukhanov and G. V. Chibisov, JETP Lett. 33, 532 (1981) [Pisma Zh. Eksp. Teor. Fiz. 33, 549 (1981)]. A. H. Guth and S. Y. Pi, Phys. Rev. Lett. 49, 1110 (1982). A. A. Starobinsky, Phys. Lett. B 117, 175 (1982). S. W. Hawking, Phys. Lett. B 115, 295 (1982). V. N. Lukash, Pisma Zh. Eksp. Teor. Fiz. 31, 631 (1980); V. N. Lukash, Sov. Phys. JETP 52, 807 (1980) [Zh. Eksp. Teor. Fiz. 79, (1980)]. W. Press, Phys. Scr. 21, 702 (1980). K. Sato, Mon. Not. Roy. Astron. Soc. 195, 467 (1981). V. F. Mukhanov, JETP Lett. 41, 493 (1985) [Pisma Zh. Eksp. Teor. Fiz. 41, 402 (1985)]. V. F. Mukhanov, Sov. Phys. JETP 67, 1297 (1988) [Zh. Eksp. Teor. Fiz. 94N7, 1 (1988 ZETFA,94,1-11.1988)]. M. Sasaki, Prog. Theor. Phys. 76, 1036 (1986). N. Birrell and P.C.W. Davies: Quantum Fields in Curved Space, (Cambridge Univ. Press, Cambridge, 1982). R. H. Brandenberger, Nucl. Phys. B 245, 328 (1984). R. H. Brandenberger and C. T. Hill, Phys. Lett. B 179, 30 (1986). D. Polarski and A. A. Starobinsky, Class. Quant. Grav. 13, 377 (1996) [arXiv:gr-qc/9504030]. L. P. Grishchuk, Sov. Phys. JETP 40, 409 (1975) [Zh. Eksp. Teor. Fiz. 67, 825 (1974)].

COSMOLOGICAL PERTURBATIONS 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. 122.

115

F. C. Adams, K. Freese and A. H. Guth, Phys. Rev. D 43, 965 (1991). B. A. Bassett, D. I. Kaiser and R. Maartens, Phys. Lett. B 455, 84 (1999) [arXiv:hepph/9808404]. J. P. Zibin, R. H. Brandenberger and D. Scott, Phys. Rev. D 63, 043511 (2001) [arXiv:hepph/0007219]. A. Borde and A. Vilenkin, Phys. Rev. Lett. 72, 3305 (1994) [arXiv:gr-qc/9312022]. R. H. Brandenberger and C. Vafa, Nucl. Phys. B 316, 391 (1989). W. G. Unruh, Phys. Rev. D51, 2827 (1995). S. Corley and T. Jacobson, Phys. Rev. D54, 1568 (1996) [arXiv:hep-th/9601073]. J. Martin and R. H. Brandenberger, Phys. Rev. D63, 123501 (2001) [arXiv:hep-th/0005209]. R. H. Brandenberger and J. Martin, Mod. Phys. Lett. A16, 999 (2001) [arXiv:astroph/0005432]. J. C. Niemeyer, Phys. Rev. D63, 123502 (2001) [arXiv:astro-ph/0005533]. C. S. Chu, B. R. Greene and G. Shiu, Mod. Phys. Lett. A 16, 2231 (2001) [arXiv:hepth/0011241]. R. Easther, B. R. Greene, W. H. Kinney and G. Shiu, Phys. Rev. D64, 103502 (2001) [arXiv:hep-th/0104102]. R. Easther, B. R. Greene, W. H. Kinney and G. Shiu, Phys. Rev. D 67, 063508 (2003) [arXiv:hep-th/0110226]. F. Lizzi, G. Mangano, G. Miele and M. Peloso, JHEP 0206, 049 (2002) [arXiv:hepth/0203099]. S. F. Hassan and M. S. Sloth, Nucl. Phys. B 674, 434 (2003) [arXiv:hep-th/0204110]. R. Brandenberger and P. M. Ho, Phys. Rev. D66, 023517 (2002) [arXiv:hep-th/0203119]. A. Kempf and J. C. Niemeyer, Phys. Rev. D64, 103501 (2001) [arXiv:astro-ph/0103225]. C. P. Burgess, J. M. Cline, F. Lemieux and R. Holman, JHEP 0302, 048 (2003) [arXiv:hepth/0210233]; C. P. Burgess, J. M. Cline and R. Holman, JCAP 0310, 004 (2003) [arXiv:hep-th/0306079]. U. H. Danielsson, Phys. Rev. D 66, 023511 (2002) [arXiv:hep-th/0203198]; U. H. Danielsson, JHEP 0207, 040 (2002) [arXiv:hep-th/0205227]. V. Bozza, M. Giovannini and G. Veneziano, JCAP 0305, 001 (2003) [arXiv:hep-th/0302184]. K. Schalm, G. Shiu and J. P. van der Schaar, JHEP 0404, 076 (2004) [arXiv:hep-th/0401164]. M. Lemoine, M. Lubo, J. Martin and J. P. Uzan, Phys. Rev. D65, 023510 (2002) [arXiv:hepth/0109128]. T. Jacobson and D. Mattingly, Phys. Rev. D63, 041502 (2001) [arXiv:hep-th/0009052]. N. Kaloper, M. Kleban, A. E. Lawrence and S. Shenker, Phys. Rev. D 66, 123510 (2002) [arXiv:hep-th/0201158]. J. Martin and R. H. Brandenberger, Phys. Rev. D65, 103514 (2002) [arXiv:hep-th/0201189]. W. G. Unruh and R. Schutzhold, arXiv:gr-qc/0408009. R. H. Brandenberger, S. E. Joras and J. Martin, Phys. Rev. D 66, 083514 (2002) [arXiv:hepth/0112122]. D. Amati, M. Ciafaloni and G. Veneziano, Phys. Lett. B197, 81 (1987). D. J. Gross and P. F. Mende, Nucl. Phys. B303, 407 (1988). T. Yoneya, Mod. Phys. Lett. A4, 1587 (1989). M. Li and T. Yoneya, arXiv:hep-th/9806240. S. Cremonini, Phys. Rev. D 68, 063514 (2003) [arXiv:hep-th/0305244]. S. Hollands and R. M. Wald, Gen. Rel. Grav. 34, 2043 (2002) [arXiv:gr-qc/0205058]. J. Martin and R. Brandenberger, Phys. Rev. D 68, 063513 (2003) [arXiv:hep-th/0305161]. T. Tanaka, arXiv:astro-ph/0012431. A. A. Starobinsky, Pisma Zh. Eksp. Teor. Fiz. 73, 415 (2001) [JETP Lett. 73, 371 (2001)] [arXiv:astro-ph/0104043]. M. Porrati, Phys. Lett. B 596, 306 (2004) [arXiv:hep-th/0402038]. R. H. Brandenberger and J. Martin, arXiv:hep-th/0410223. A. A. Starobinsky and I. I. Tkachev, JETP Lett. 76, 235 (2002) [Pisma Zh. Eksp. Teor. Fiz. 76, 291 (2002)] [arXiv:astro-ph/0207572]. A. J. Tolley, N. Turok and P. J. Steinhardt, Phys. Rev. D 69, 106005 (2004) [arXiv:hep-

116

ROBERT H. BRANDENBERGER

th/0306109]. 123. T. J. Battefeld, S. P. Patil and R. Brandenberger, Phys. Rev. D 70, 066006 (2004) [arXiv:hepth/0401010]. 124. C. van de Bruck, M. Dorca, R. H. Brandenberger and A. Lukas, Phys. Rev. D 62, 123515 (2000) [arXiv:hep-th/0005032].

TOWARDS A TOPOLOGICAL G2 STRING

JAN DE BOER, ASAD NAQVI Instituut voor Theoretische Fysica Valckenierstraat 65, 1018XE Amsterdam, The Netherlands AND ASSAF SHOMER Santa Cruz Institute for Particle Physics 1156 High Street, Santa Cruz, 95064 CA, USA

Abstract. We deﬁne new topological theories related to sigma models whose target space is a 7 dimensional manifold of G2 holonomy. We show how to deﬁne the topological twist and identify the BRST operator and the physical states. Correlation functions at genus zero are computed and related to Hitchin’s topological action for three-forms. We conjecture that one can extend this deﬁnition to all genus and construct a seven-dimensional topological string theory. In contrast to the four-dimensional case, it does not seem to compute terms in the low-energy effective action in three dimensions.

1. Introduction Topological strings on Calabi-Yau manifolds describe certain solvable sectors of superstrings and as such provide simpliﬁed toy models of string theory. There are two inequivalent ways to twist the Calabi-Yau sigma model. This yields topological theories known as the A-model and the B-model, which at ﬁrst sight depend on different degrees of freedom: the A-model apparently only involves the K¨ahler moduli and the B-model only the complex moduli. However, this changes once branes are included, and it has been conjectured that there is a version of S-duality which maps the A-model to the B-model [1]. Subsequently, several authors found evidence for the existence of seven and/or eight dimensional theories that unify and extend the A and B-model [2, 3, 4, 5, 6]. This was one of our original motivations to take a closer look at string theory on seven-dimensional manifolds of G2 holonomy, and to see whether it allows for a topological twist, though we were motivated by other issues as well, such as applications to M-theory compactiﬁcations on G2 -manifolds, and as a possible tool to improve our understanding of the relation between supersymmetric gauge theories in three and four dimensions. The outline of this note is as follows. We will ﬁrst review sigma-models on target spaces of G2 holonomy, and the structure of chiral the algebra of these theories. The 117 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 117–127. © 2005 Springer. Printed in the Netherlands.

118

JAN DE BOER, ASAD NAQVI AND ASSAF SHOMER

latter is a non-linear extension of the c = 21 2 N = 1 superconformal algebra that contains 7 . This describes a minimal model, the tricritical Ising an N = 1 subalgebra with c = 10 model, which plays a crucial role in the twisting. We then go on to describe the twisting, the BRST operator, the physical states, and we end with a discussion of topological G2 strings. Here we brieﬂy summarize our ﬁndings. A more detailed discussion will appear elsewhere [7]. There is extensive literature about string theory and M-theory compactiﬁed on G2 manifolds. The ﬁrst detailed study of the world-sheet formulation of strings on G2 manifolds appeared in [8]. The world-sheet chiral algebra was studied in some detail in [8, 9, 10, 11]. For more about type II strings on G2 manifolds and their mirror symmetry, see e.g. [12, 13, 14, 15, 16, 17, 18, 19, 20, 21]. A review of M-theory on G2 manifolds with many references can be found in [22]. 2. G2 sigma models We start from an N = (1, 1) sigma model describing d chiral superﬁelds X µ = φµ (z) + θψ µ (z) S=

d2 z d2 θ (Gµν + Bµν )Dθ Xµ Dθ¯Xν .

(2.1)

The super stress-energy tensor is given by T (z, θ) = G(z)+θT (z) = − 12 Gµν Dθ X µ ∂z X ν . This N = (1, 1) sigma model can be formulated on an arbitrary target space. However, the target space theory will have some supersymmetry only when the manifold has special holonomy. This condition ensures the existence of covariantly constant spinors which are used to construct supercharges. The existence of a covariantly constant spinor on the manifold also implies the existence of covariantly constant p-forms given by φ(p) = T Γi1 ...ip dxi1 ∧ · · · ∧ dxip .

(2.2)

This formal expression may be identically zero. The details of the target space manifold dictate which p-forms are actually present. If the manifold has special holonomy H ⊂ SO(d), the non-vanishing forms (2.2) are precisely the forms that transform trivially under H. The existence of such covariantly constant p-forms on the target space manifold implies the existence of extra elements in the chiral algebra. For example, given a covariantly constant p form, φ(p) = φi1 ···ip dxi1 ∧· · ·∧dxip satisfying ∇φi1 ···ip = 0, we can construct a holomorphic superﬁeld current given by J(p) (z, θ) = φi1 ···ip Dθ X i1 · · · Dθ X ip , which satisﬁes Dθ¯J(p) = 0 on shell. In components, this implies the existence of a dimension p2 and a dimension p+1 2 current. For example, on a Kahler manifold, the ex¯ ¯ istence of a covariantly constant Kahler two form ω = gi¯j (dφi ∧ dφj − dφj ∧ dφi ) ¯ implies the existence of a dimension 1 current J = gi¯j ψ i ψ j and a dimension 32 current ¯ ¯ j j i i G (z) = gi¯j (ψ ∂z φ − ψ ∂z φ ), which add to the (1, 1) superconformal currents G(z) and T (z) to give a (2, 2) superconformal algebra.

TOWARDS A TOPOLOGICAL G2 STRING

119

A generic seven dimensional Riemannian manifold has SO(7) holonomy. A G2 manifold has holonomy which sits in a G2 subgroup of SO(7). Under this embedding, the eight dimensional spinor representation 8 of SO(7) decomposes into a 7 and a singlet of G2 , and the latter corresponds to the covariantly constant spinor. The p-form (2.2) is non-trivial only when p = 3, 4. In other words, there is a covariantly constant 3-form (3) φ(3) = φijk dxi ∧dxj ∧dxk . The hodge dual 4-form is then also automatically covariantly constant. By the above discussion, the 3-form implies the existence of a superﬁeld current (3) J(3) (z, θ) = φijk Dθ X i Dθ X j Dθ X k ≡ Φ + θK. Explicitly, Φ is a dimension 32 current (3)

(3)

Φ = φijk ψ i ψ j ψ k and K is its dimension 2 superpartner K = φijk ψ i ψ k ∂φk . Similarly, the 4-from implies the existence of a dimension 2 current X and its dimension 52 superpartner M . The chiral algebra of G2 sigma models thus contains 4 extra currents on top of the two G, T that constitute the N = 1 superconformal algebra. These six generators form a closed quantum algebra which appears explicitly e.g. in [9, 8, 10] (see also [11]). An important fact, which will be crucial in almost all the remaining analysis, is that the generators Φ and X form a closed sub-algebra: if we deﬁne the supercurrent GI = √i15 Φ and stress-energy tensor TI = − 15 X we recognize that this is the unique N = 1 super7 known as the Triconformal algebra of the minimal model with central charge c = 10 critical Ising Model. This sub-algebra plays a role similar to the U (1) R-symmetry of the N = 2 algebra in compactiﬁcations on Calabi-Yau manifolds. In fact, with respect to the conformal symmetry, the full Virasoro algebra decomposes in two commuting1 Virasoro algebras: T = TI + Tr with TI (z)Tr (w) = 0. This means we can classify conformal primaries by two quantum numbers, namely its tri-critical Ising model highest weight and its highest weight with respect to Tr : |primary = |hI , hr . Perhaps it is worth emphasizing the logic here: classically, we ﬁnd a conformal algebra with six generators in sigma-models on manifolds of G2 holonomy. In the quantum theory we expect, in the absence of anomalies other than the conformal anomaly, to ﬁnd a quantum version of this classical algebra. In [9] all quantum extensions were analyzed, and a two-parameter family of quantum algebras was found. Requiring that the quantum algebra has the right central charge (necessary to have a critical string theory) and that it contains the tricritical Ising model (necessary for space-time supersymmetry) ﬁxes the two-parameters. This motivates why this is the appropriate deﬁnition for string theory on G2 manifolds. 3. Tri-Critical Ising Model Unitary minimal models are labeled by a positive integer p = 2, 3, . . . and occur only on 6 . The Tri-Critical Ising Model is the “discrete series” at central charges c = 1 − p(p+1) 7 . It is at the same time also a the second member (p = 4) which has central charge c = 10 minimal model for the N = 1 superconformal algebra. 1 This decomposition only works for the Virasoro part of the corresponding N = 1 algebras. The full N = 1 structures do not commute. For example the superpartner of Φ with respect to the full N = 1 algebra is K whereas its superpartner with respect to the N = 1 of the tri-critical Ising model is X.

120

JAN DE BOER, ASAD NAQVI AND ASSAF SHOMER

The conformal primaries of unitary minimal models are labeled by two integers 1 ≤ n ≤ p and 1 ≤ n < p. The weights in this range are arranged into a “Kac table”. The 2 −1 conformal weight of the primary Φn n is hn n = [pn −(p+1)n] . In the Tri-critical Ising 4p(p+1) 1 6 3 7 3 . Below we write the model (p = 4) there are 6 primaries of weights 0, 10 , 10 , 2 , 16 , 80 Kac table for the tricritical Ising model. Beside the Identity operator (h = 0) and the N = 1 supercurrent (h = 32 ) the NS sector (ﬁrst and third columns) contains a primary 1 6 7 3 and its N = 1 superpartner (h = 10 ). The primaries of weight 16 , 80 of weight h = 10 are in the Ramond sector (middle column). n \ n

1

2

3

1

7 16

3 2

2

1 10

3 80

6 10

3

6 10

3 80

1 10

4

3 2

7 16

The Hilbert space of the theory decomposes in a similar way, H = ⊕n,n Hn ,n . A central theme in this work relies on the fact that since the primaries Φn n form a closed algebra under the OPE they can be decomposed into conformal blocks which connect two Hilbert spaces. Conformal blocks are denoted by Φnl ,l,n,m m which describes the restriction of Φn ,n to a map that only acts from Hm ,m to Hl ,l . An illustrative example, which will prove crucial in what follows, is the block structure of the primary Φ2,1 of weight 1/10. General arguments show that the fusion rule of this ﬁeld with any other primary Φn n is φ(2,1) × φ(n ,n) = φ(n −1,n) + φ(n +1,n) . The only non-vanishing conformal blocks in the decomposition of Φ2,1 are those that connect a primary with the primary right above it and the primary right below n −1,n n +1,n in the Kac table, namely2 , φ2,1,n ,n and φ2,1,n ,n . This can be summarized formally by deﬁning the following decomposition3 Φ2,1 = Φ↓2,1 ⊕ Φ↑2,1 .

(3.1)

Similarly, the fusion rule of the Ramond ﬁeld Φ1,2 with any primary is φ(1,2) × φ(n ,n) = φ(n,n−1) + φ(n ,n+1) showing that it is composed of two blocks, which we denote + as follows Φ1,2 = Φ− 1,2 ⊕ Φ1,2 . Conformal blocks transforms under conformal transformations exactly like the primary ﬁeld they reside in but are usually not single-valued functions of z(¯ z ). 2

Note the confusing notation where down the Kac table means larger n’ and vice-versa. We stress that this decomposition is special to the ﬁeld Φ2,1 and does not necessarily hold for other primaries which may contain other blocks. 3

TOWARDS A TOPOLOGICAL G2 STRING

121

3.1. CHIRAL PRIMARY STATES The chiral-algebra associated with manifolds of G2 holonomy4 allows us to draw several conclusions about the possible spectrum of such theories. It is useful to decompose the generators of the chiral algebra in terms of primaries of the tri-critical Ising model and primaries of the remainder. The commutation relations of the G2 algebra imply that the some 14 , K(z) = of the generators of the chiral algebra decompose as [8]: G(z) = Φ2,1 ⊗ψ 10 24 14 Φ3,1 ⊗ ψ 14 and M (z) = aΦ ⊗ χ + b[X , Φ ] ⊗ ψ , with ψ, χ primaries of the 2,1 −1 2,1 10 10 10 indicated weights in the Tr CFT and a, b constants. 7 3 2 Ramond ground states of the full c = 21 2 SCFT are of the form | 16 , 0 and | 80 , 5 . The 7 existence of the | 16 , 0 state living just inside the tricritical Ising model plays a crucial role in the topological twist. Coupling left and right movers, the only possible RR ground 7 7 , 0L ⊗ | 16 , 0R ground state states compatible with the G2 chiral algebra5 are a single | 16 3 2 3 2 , 5 L ⊗ | 80 , 5 R . By studying operator and a certain number of states of the form | 80 product expansions of the RR ground states we get the following “special” NSNS states 1 2 1 2 6 2 6 2 , 5 L ⊗ | 10 , 5 R , | 10 , 5 L ⊗ | 10 , 5 R and | 32 , 0L ⊗ | 32 , 0R |0, 0L ⊗ |0, 0R , | 10 corresponding to the 4 NS primaries Φn ,1 with n = 1, 2, 3, 4 in the tri-critical Ising model. Note that for these four states there is a linear relation between the Kac label n of the tri-critical Ising model part and the total conformal weight htotal = n 2−1 . In fact, it can be shown that, similar to the BPS bound in the N = 2 case, primaries of the G2 chiral algebra satisfy a (non-linear) bound of the form hI + hr ≥

1+

√ 1 + 80hI . 8

(3.2)

which is precisely saturated for the four NS states listed above. We will therefore refer 7 ﬁeld to those states as “chiral primary” states. Just like in the case of Calabi-Yau, the 16 maps Ramond ground states to NS chiral primaries and is thus an analogue of the “spectral ﬂow” operators in Calabi-Yau. 4. Topological Twist To construct a topologically twisted CFT we usually proceed in two steps. First we deﬁne a new stress-energy tensor, which changes the quantum numbers of the ﬁelds and operators of the theory under Lorentz transformations. Secondly, we identify a nilpotent scalar operator, usually constructed out of the supersymmetry generators of the original theory, which we declare to be the BRST operator. Often this BRST operator can be obtained in the usual way by gauge ﬁxing a suitable symmetry. If the new stress tensor is exact with respect to the BRST operator, observables (which are elements of the BRST cohom*ology) are metric independent and the theory is called topological. In particular, the twisted stress tensor should have a vanishing central charge. 4 We loosely refer to it as “the G2 algebra” but it should not be confused with the Lie algebra of the group G2 . 5 Otherwise the spectrum will contain a 1-form which will enhance the chiral algebra. Geometrically this is equivalent to demanding that b1 = 0.

122

JAN DE BOER, ASAD NAQVI AND ASSAF SHOMER

In practice [23, 24], for the N = 2 theories, an n-point correlator on the sphere in the twisted theory can conveniently be deﬁned6 as a correlator in the untwisted theory of the same n operators plus two insertions of a spin-ﬁeld, related to the space-time supersymmetry charge, that serves to trivialize the spin bundle. For a Calabi-Yau 3-fold target space there are two SU (3) invariant spin-ﬁelds which are the two spectral ﬂow operators U± 12 . This discrete choice in the left and the right moving sectors is the choice between the +(−) twists [25] which results in the difference between the topological A/B models. In [8] a similar expression was written down for sigma models on G2 manifolds, this time involving the single G2 invariant spin ﬁeld which is the unique primary Φ1,2 7 of weight 16 . It was proposed that this expression could be a suitable deﬁnition of the correlation functions of a putative topologically twisted G2 theory. In other words the twisted amplitudes are deﬁned as7 V1 (z1 ) . . . Vn (zn )twist ≡ Σ(∞)V1 (z1 ) . . . Vn (zn )Σ(0)untwist .

(4.1)

In [8] further arguments were given, using the Coulomb gas representation of the minimal model, that there exists a twisted stress tensor with vanishing central charge. This argument is however problematic, since the twisted stress tensor proposed there does not commute with Felder’s BRST operators [26] and therefore it does deﬁne a bona ﬁde operator in the minimal model. In addition, a precise deﬁnition of a BRST operator was lacking. We will proceed somewhat differently. We will ﬁrst propose a BRST operator, study its cohom*ology, and then use a version of (4.1) to compute correlation functions of BRST invariant observables. We will then comment on the extension to higher genus and on the existence of a topologically twisted G2 string. 5. The BRST Operator Our basic idea is that the topological theory for G2 sigma models should be formulated not in terms of local operators of the untwisted theory but in terms of its (non-local)8 conformal blocks. By using the decomposition (3.1) into conformal blocks, we can split any ﬁeld whose tri-critical Ising model part contains just the conformal family Φ2,1 into its up and down parts. In particular, the N = 1 supercurrent G(z) can be split as G(z) = G↓ (z) + G↑ (z).

(5.1)

We claim that G↓ is the BRST current and G↑ carries many features of an anti-ghost. The basic N = 1 relation 2c/3 2T (0) G(z)G(0) = G↓ (z) + G↑ (z) G↓ (0) + G↑ (0) ∼ 3 + z z 6

(5.2)

Up to proper normalization. Up to a coordinate dependent factor that we omit here for brevity and can be found in [7]. 8 It should be stressed that this splitting into conformal blocks is non-local in the simple sense that conformal blocks may be multi-valued functions of z(¯ z ). 7

TOWARDS A TOPOLOGICAL G2 STRING

123

proves the nilpotency of this BRST current (and of the candidate anti-ghost) because the RHS contains descendants of the identity operator only and has trivial fusion rules with the primary ﬁelds of the tri-critical Ising model and so (G↓ )2 = (G↑ )2 = 0. More formally, denote by Pn the projection operator on the sub-space Hn of states whose tri-critical Ising model part lies within the conformal family of one of the four NS primaries Φn ,1 . The 4 projectors add to the identity P1 + P2 + P3 + P4 = 1,

(5.3)

because this exhaust the list of possible highest weights in the NS sector of the tri-critical Ising model9 . We can now deﬁne our BRST operator in the NS sector more rigorously Pn +1 G− 12 Pn . (5.4) Q = G↓− 1 ≡ 2

n

The nilpotency Q2 = 0 is easily proved Pn +2 G2− 1 Pn = Pn +2 L−1 Pn = 0, Q2 = 2

n

(5.5)

n

where we could replace the intermediate Pn +1 by the identity because of property 5.1 and the last equality follows since L−1 maps each Hn to itself. Q does not commute with the local operator O∆(1),0 , O∆(2), 25 , O∆(3), 25 and O∆(4),0 1 2 6 2 corresponding to the chiral states |0, 0, | 10 , 5 , | 10 , 5 and | 32 , 0 (for brevity we will denote those 4 local operators juts by their minimal model Kac label Oi , i = 1, 2, 3, 4). However, it can be checked that the following blocks, An =

Pn +m−1 On Pm ,

(5.6)

m

which pick out the maximal “down component” of the corresponding local operator, do commute with Q and are thus in its operator cohom*ology. Thus the chiral operators of the twisted model are represented in terms of the blocks 5.6 of the local operators corresponding to the chiral states. Furthermore, it can be shown easily that those chiral operator form a ring under the OPE. By doing a calculation at large volume, we see that the BRST cohom*ology has one operator of type O1 ⊗ O1 , b2 + b3 of type O2 ⊗ O2 , b4 + b5 of type O3 ⊗ O3 , and one of type O4 ⊗ O4 . The total BRST cohom*ology is thus precisely given by H ∗ (M ). Also, one ﬁnds that the b3 operators of type O2 ⊗ O2 are precisely the geometric moduli of the G2 target space10 . In the topological G2 theory, genus zero correlation functions of chiral primaries between BRST closed states are position independent. Indeed, the generator of translations on the plane, namely L−1 , is BRST exact 9 For simplicity, we will set Pn = 0 for n ≤ 0 and n ≥ 5, so that we can simply write n Pn = 1 instead of (5.3). 10 In this work we ignore the b2 moduli corresponding to the B-ﬁeld.

124

JAN DE BOER, ASAD NAQVI AND ASSAF SHOMER

{G↓− 1 , G↑− 1 } = 2

2

n

Pn G− 12 (Pn −1 + Pn +1 ) G− 12 Pn =

Pn L−1 Pn = L−1 .

n

(5.7) This is a crucial ingredient of topological theories. ˜ − 1 | 1 , 2 L ⊗ G− 1 | 1 , 2 R Moreover, it can be shown [8] that the upper components G 2 10 5 2 10 5 correspond to exactly marginal deformations of the CFT preserving the G2 chiral algebra, completely in agreement with the identiﬁcation of them as the geometric moduli of the theory. Focusing momentarily on the left movers, we can show that [Q, {G− 12 , O2 }] = ∂A2 11 so that the very same deformation is physical, namely Q exact, also in the topological theory. Note that the deformation is given by a conventional operator that does not involve any projectors. Combining with the right-movers, we ﬁnd that the deformations in the action of the topological string are exactly the same as the deformations of the nontopological string as is expected because both should exist on an arbitrary manifold of G2 holonomy. Most correlation functions at genus zero vanish. The most interesting one is the threepoint function of three operators Y = O2 ⊗ O2 . These correspond to geometric moduli. If we introduce coordinates ti on the moduli space of G2 metrics, then we obtain from a large volume calculation ∂g aa ∂g bb ∂g cc √ d7 x gφabc φa b c . (5.8) Yi Yj Yk = ∂ti ∂tj ∂tk M One might expect, based on general arguments, that this is the third derivative of some prepotential if suitable ‘ﬂat’ coordinates are used. We do not know the precise deﬁnition of ﬂat coordinates for the moduli space of G2 metrics, but if we take for example M = T 7 and take coordinates such that φ is linear in them, then we can verify ∂3 1 φ ∧ ∗φ. (5.9) Yi Yj Yk = − 21 ∂ti ∂tj ∂tk The prepotential appearing on the right hand side is exactly the same as the action functional introduced by Hitchin in [27, 28]. A similar action was also used as a starting point for topological M-theory in [3] (see also [2]). This strongly suggests that our topological G2 ﬁeld theory is somehow related to topological M-theory. 6. Topological G2 Strings In the case of N = 2 theories, the computation of correlation functions at genus zero outlined above can be generalized to higher genera [23, 24]. An n-point correlator on a genus-g Riemann surface in the twisted theory can be deﬁned as a correlator in the untwisted theory of the same n operators plus (2 − 2g) insertions of the spin-ﬁeld that is related to the space-time supersymmetry charge. For a Calabi-Yau 3-fold target space on a Riemann surface with g > 1, the meaning of the above prescription is to insert 2g − 2 of the conjugate spectral ﬂow operator. 11

Recall that A2 was deﬁned in (5.6).

TOWARDS A TOPOLOGICAL G2 STRING

125

To generalize this to the G2 situation, we would like to have something similar. However, there is only one G2 invariant spin-ﬁeld. This is where the decomposition in conformal blocks given in section 3 is useful: the spin-ﬁeld Φ2,1 could be decomposed12 in − + a block Φ+ 2,1 and in a block Φ2,1 . At genus zero we needed two insertions of Φ2,1 , so the − natural guess is that at genus g we need 2g − 2 insertions of Φ2,1 . However, this is not the full story. We also need to insert 3g − 3 copies of the anti-ghost and integrate over the moduli space of Riemann surfaces to properly deﬁne a topological string theory. The anti-ghost is very close to G↑ , and the fusion rules of the tri-critical Ising model tell us that there is indeed a non-vanishing contribution to correlation functions of 2g − 2 Φ− 2,1 ’s and 3g − 3 G↑ . This prescription would therefore work very nicely if we would have found the right anti-ghost. The candidates we tried so far all seem to fail in one way or another. One possible conclusion might be that a twisted stress tensor does not exist and that there is only a sensible notion of topological G2 sigma models but not of topological G2 strings. The fact that so far so many properties of the N = 2 topological theories appeared to hold also in our G2 model leads us to believe that a sensible extension to higher genera indeed exists. Identifying the correct twisted stress tensor remains an open problem. Barring this important omission the coupling to topological gravity goes pretty much along the same lines as for the N = 2 topological string (details can be found in [7]). 7. Conclusions An important application of topological strings stems from the realization [23, 24, 25] that its amplitudes agree with certain amplitudes of the physical superstring. Just like N = 2 topological strings compute certain F-terms in four dimensional N = 2 gauge theories, one might wonder whether the G2 topological string similarly computes F-terms in three dimensional N = 2 gauge theories. Since G2 manifolds are Ricci-ﬂat we can consider compactifying the type II superstring on R1,2 × N7 where N7 is a 7 dimensional manifold of G2 holonomy. This reduces the supersymmetry down to two real supercharges in 3 dimension from each worldsheet chirality so we end up with a low energy ﬁeld theory in 3 dimensions with N = 2 supergravity. By studying amplitudes in some detail we observe that, except perhaps at genus zero, the amplitudes do involve a sum over conformal blocks, and the topological G2 string therefore seems to compute only one of many contributions to an amplitude. This is in contrast to the four dimensional case where the possibility to look only at the (anti)-self dual gravitons allowed to isolate the topological contributions. Nevertheless, we believe that topological G2 strings are worthwhile to study. They may possibly provide a good deﬁnition of topological M-theory, and a further study may teach us many things about non-topological G2 compactiﬁcations as well. We leave this, as well as the generalization to spin(7) compactiﬁcations [29] and the study of branes in these theories to future work.

12 In terms of the Coulomb gas representation, one of these can be represented as an ordinary vertex operator, the other one involves a screening charge.

126

JAN DE BOER, ASAD NAQVI AND ASSAF SHOMER

Acknowledgements It is a pleasure to thank Nathan Berkovits, Volker Braun, Lorenzo Cornalba, Robbert Dijkgraaf, Pietro Antonio Grassi, Wolfgang Lerche, Hiroshi Ooguri, Volker Shomerus, Annamaria Sinkovics, Kostas Skenderis and Cumrun Vafa for useful discussions. We would also like to thank the other organizers of the Carg`ese 2004 summer school. This work was partly supported by the stichting FOM. References 1.

A. Neitzke and C. Vafa, “N = 2 strings and the twistorial Calabi-Yau,” arXiv:hep-th/0402128; N. Nekrasov, H. Ooguri and C. Vafa, “S-duality and topological strings,” JHEP 0410, 009 (2004) [arXiv:hep-th/0403167]. 2. A. A. Gerasimov and S. L. Shatashvili, “Towards integrability of topological strings. I: Threeforms on Calabi-Yau manifolds,” JHEP 0411, 074 (2004) [arXiv:hep-th/0409238]. 3. R. Dijkgraaf, S. Gukov, A. Neitzke and C. Vafa, “Topological M-theory as uniﬁcation of form theories of gravity,” arXiv:hep-th/0411073. 4. N. Nekrasov, “A la recherche de la m-theorie perdue. Z theory: Chasing m/f theory,” arXiv:hep-th/0412021. 5. P. A. Grassi and P. Vanhove, “Topological M theory from pure spinor formalism,” arXiv:hepth/0411167. 6. L. Anguelova, P. de Medeiros and A. Sinkovics, “On topological F-theory,” arXiv:hepth/0412120. 7. J. de Boer, A. Naqvi and A. Shomer, “The Topological G2 String”, to appear. 8. S. L. Shatashvili and C. Vafa, “Superstrings and manifold of exceptional holonomy,” arXiv:hep-th/9407025. 9. R. Blumenhagen, “Covariant construction of N=1 superW algebras,” Nucl. Phys. B 381, 641 (1992). 10. J. M. Figueroa-O’Farrill, “A note on the extended superconformal algebras associated with manifolds of exceptional holonomy,” Phys. Lett. B 392, 77 (1997) [arXiv:hep-th/9609113]. 11. D. Gepner and B. Noyvert, “Unitary representations of SW(3/2,2) superconformal algebra,” Nucl. Phys. B 610, 545 (2001) [arXiv:hep-th/0101116]; B. Noyvert, “Unitary minimal models of SW(3/2,3/2,2) superconformal algebra and manifolds of G(2) holonomy,” JHEP 0203, 030 (2002) [arXiv:hep-th/0201198]; 12. B. S. Acharya, “N=1 M-theory-Heterotic Duality in Three Dimensions and Joyce Manifolds,” arXiv:hep-th/9604133. 13. B. S. Acharya, “On mirror symmetry for manifolds of exceptional holonomy,” Nucl. Phys. B 524, 269 (1998) [arXiv:hep-th/9707186]. 14. T. Eguchi and Y. Sugawara, “CFT description of string theory compactiﬁed on non-compact manifolds with G(2) holonomy,” Phys. Lett. B 519, 149 (2001) [arXiv:hep-th/0108091]. 15. R. Roiban and J. Walcher, “Rational conformal ﬁeld theories with G(2) holonomy,” JHEP 0112, 008 (2001) [arXiv:hep-th/0110302]. 16. T. Eguchi and Y. Sugawara, “String theory on G(2) manifolds based on Gepner construction,” Nucl. Phys. B 630, 132 (2002) [arXiv:hep-th/0111012]. 17. R. Blumenhagen and V. Braun, “Superconformal ﬁeld theories for compact G(2) manifolds,” JHEP 0112, 006 (2001) [arXiv:hep-th/0110232]; “Superconformal ﬁeld theories for compact manifolds with Spin(7) holonomy,” JHEP 0112, 013 (2001) [arXiv:hep-th/0111048]. 18. R. Roiban, C. Romelsberger and J. Walcher, “Discrete torsion in singular G(2)-manifolds and real LG,” Adv. Theor. Math. Phys. 6, 207 (2003) [arXiv:hep-th/0203272]. 19. K. Sugiyama and S. Yamaguchi, “Coset construction of noncompact Spin(7) and G(2) CFTs,” Phys. Lett. B 538, 173 (2002) [arXiv:hep-th/0204213]. 20. T. Eguchi, Y. Sugawara and S. Yamaguchi, “Supercoset CFT’s for string theories on noncompact special holonomy manifolds,” Nucl. Phys. B 657, 3 (2003) [arXiv:hep-th/0301164].

TOWARDS A TOPOLOGICAL G2 STRING 21. 22. 23. 24. 25. 26. 27. 28. 29.

127

M. R. Gaberdiel and P. Kaste, “Generalised discrete torsion and mirror symmetry for G(2) manifolds,” JHEP 0408, 001 (2004) [arXiv:hep-th/0401125]. B. S. Acharya and S. Gukov, “M theory and Singularities of Exceptional Holonomy Manifolds,” Phys. Rept. 392, 121 (2004) [arXiv:hep-th/0409191]. M. Bershadsky, S. Cecotti, H. Ooguri and C. Vafa, “Kodaira-Spencer theory of gravity and exact results for quantum string amplitudes,” Commun. Math. Phys. 165, 311 (1994) [arXiv:hepth/9309140]. I. Antoniadis, E. Gava, K. S. Narain and T. R. Taylor, “Topological amplitudes in string theory,” Nucl. Phys. B 413, 162 (1994) [arXiv:hep-th/9307158]. E. Witten, “Mirror manifolds and topological ﬁeld theory,” arXiv:hep-th/9112056. G. Felder, “Brst Approach To Minimal Methods,” Nucl. Phys. B 317, 215 (1989) [Erratumibid. B 324, 548 (1989)]. N. Hitchin, “The geometry of three-forms in six and seven dimensions,” arXiv:math. dg/0010054. N. Hitchin, “Stable forms and special metrics,” arXiv:math.dg/0107101. J. de Boer, A. Naqvi and A. Shomer, “Topological Strings on Exceptional Holonomy Manifolds”, to appear.

CHALLENGES OF MATRIX MODELS

ALEXEI MOROZOV ITEP, Moscow, Russia

1. Introduction Matrix models appear again and again at the front line of theoretical physics, and every new generation of scientists discovers something new in this seemingly simple subject. It is getting more and more obvious that matrix models capture the very essence of general quantum ﬁeld theory and provide the crucial representative example for the string theory [1]: matrix models may play the same role for string theory as the harmonic oscillator for quantum mechanics. An essential difference is, however, that we have a nearly exhaustive understanding of harmonic oscillator, while nothing like a complete description of matrix models is yet available. Probably, the time is coming to begin a systematic analysis with the goal of building up a theory of matrix model partition functions as the ﬁrst special functions of string theory [2]. Various applications of matrix model techniques should use these functions as building blocks for the formulation of their results, thus separating the physical content of different applications from a common mathematical formalism. The goal can be to build up an analogue of the powerful free-ﬁeld formalism, developed in 1980’s, which allowed to reduce many problems in perturbative string theory [3] (and related issues in representation theory of Kac-Moody algebras, 2d conformal ﬁeld theory and ﬁnite-zone solutions in integrable-systems theory) to an almost classical set of special functions: Riemann’s theta-functions [4, 5], associated with Riemann surfaces. The problem is that matrix model partition functions are non-trivial generalization of Riemann’s theta-functions, which were not fully investigated in XIX-th century, therefore the theory should involve essential new ideas, both in physics and in mathematics. This is the reason why the progress in this ﬁeld – actually known under the nick-name of nonperturbative string theory – has been considerably slower. More and more people begin to realize that generic problems of non-perturbative string theory, i.e. the general theory of phase transitions, have a lot in common, and these most interesting common properties are clearly represented at the simplest level of matrix models. Complications introduced by the sophisticated context of particular applications can and should be separated from the important content, captured at the matrix model level. In the development of matrix model theory one can distinguish several stages. In fact, these stages are typical for every chapter of the string theory, and they characterize not 129 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 129–162. © 2005 Springer. Printed in the Netherlands.

130

ALEXEI MOROZOV

so much the history, but the kind of questions that are posed and the level of abstraction analyzing these questions. Whenever a new problem arises (or is re-addressed again), it is quite useful to understand its place in this hierarchy. 2. Classical period: introduction of models During the classical period the main task was to study concrete phenomena, involving matrix models. Since in these notes we are not going to discuss applications, we just mention a few crucial theoretical methods, developed at this stage. The most important idea was to reduce the N 2 -fold matrix integrals like 1 k Z(t|N ) = dΦ exp tk Tr Φ (2.1) Vol(U (N )) N ×N k

to N -fold integrals over eigenvalues φi of matrices Φ = U † diag(φi )U . This idea, closely related in general context to the problem of gauge invariance, is technically based on the possibility to explicitly integrate over ”angular variables” in the simplest situations, of which (2.1) is a representative example. The single known exactly solvable generalization, which goes slightly beyond this example, involves the Itzykson-Zuber integral [7] [dU ] exp(TrAU † BU ), (2.2) N ×N

which is used in construction of two important classes of matrix theories: generalized Kontsevich model [8, 9], dM exp [tr (W (M ) + LM )] (2.3) ZGKM (L|W ) ∼ m×m

and even further generalized Kazakov-Migdal-Kontsevich model [10, 11], dM (α) [dU (αβ) ][dU (αµ) ] · ZG(KM )2 ({Lµ }|{Wα }) ∼ µ,α

m×m

† † Mβ Uαβ + Lµ Uαµ Mα Uαµ , · exp tr Wα (Mα ) + Mα Uαβ

(2.4)

deﬁned for any graph with Hermitian ﬁelds Mα and background ﬁelds Lµ standing in the vertices and unitary ﬁelds Uαβ and Uαµ on the links. In more general situations (sometime quite important for applications, e.g. involving non-trivial multi-brane backgrounds), angular integration is highly non-trivial and remains an open problem. The simplest of such difﬁculties are echoed in the Gribov copies problem in Yang-Mills theory, while treatment of more sophisticated unitary-matrix integrals face problems, typical for adequate treatment of quantum gravity. The class of theories where gauge (angular) variables can be integrated out in effective way is referred to as eigenvalue models [6]. Most of further development concerns eigenvalue models, which until recently remained implicitly synonymous to matrix models.

MATRIX MODELS

131

The next idea of the classical period used the fact, that angular integration usually provides non-trivial, but very special measures on the space of eigenvalues, which is often made from Van-der-Monde determinants [12]. For example, (2.1) can be transformed into N N k Z(t|N ) = t k φi (φi − φj )2 , (2.5) dφi exp i=1

k

i m2 is called normal hierarchy. Normal hierarchy is analogous to that of quarks and charged leptons (their masses are given in GeV): Q = 2/3 t 175 c 1.3 u 5 · 10−3

Q = −1/3 b 5 s 0.15 d 10 · 10−3

Q = −1 τ 1.9 µ 0.1 e 0.5 · 10−3

Q=0 ν3 ∆m2atm = 25 · 10−4 eV2 ν2 ν1 ∆m2 = 0.8 · 10−4 eV2

Inverted hierarchy is more promising for solving the Majorana–Dirac alternative in neutrinoless double β-decay: the heavier m2 , the larger < m >, see section 9. There is also a possibility of quasidegenerate masses: m1 m2 m3 0.2 − 0.3 eV.

172

LEV B. OKUN

For the absolute value of mνe only the upper limit is known from two experiments on the electron spectrum in tritium β-decay: mνe < 2.2 eV (Troitsk, 2001), Mainz (2003). An experiment KATRIN (Karlsruhe Tritium Neutrino Experiment) is designed to measure the mass of electron antineutrino directly with sensitivity of 0.2 eV. As for mixing angles, let us note, that for neutrinos two of them are much larger than for quarks. 9. Double beta decay [32] While in case of Dirac neutrino the emission of electron proceeds either with absorption of neutrino, or with emission of antineutrino, the Majorana neutrino can be both emitted and absorbed. This refers not only to real neutrinos, but to virtual as well. Therefore in the second order of weak interaction two neutrons in a nucleus can decay into two protons and two electrons by exchanging a virtual Majorana neutrino. The amplitude of such process is proportional to a product of two factors: A(ββ)0ν ∼< m > ·M . The factor < m > represents the neutrino propagator: 2 2 iα21 2 iα31 < m >= m1 Ue1 + m2 Ue2 e + m3 Ue3 e .

The coefﬁcients Ueα can be read off from the ﬁrst line of the matrix U at the end of Section 2. From neutrino oscillations we know that |Ue2 | |Ue3 |. That is why the probability of 2β0ν decay is larger for larger value of m2 (inverted hierarchy). The factor M represents the nuclear matrix element estimated on the basis of measurements of the half-life times of the decays 2β2ν. The most advanced seems to be at present the search of 2β0ν decay of 76 Ge by two collaborations: IGEX and Heidelberg – Moscow. Both give upper limits | < m > | < (0.35 ÷ 1.05) eV at 90% CL , where factor 3 uncertainty reﬂects the uncertainty of M . (A part of the latter collaboration claims to observe the 2β0ν decay and gives a lower limit | < m > | > 0.11 eV.) Collaboration NEMO3 (100 Mo; 82 Se) and CUORICINO (130 Te) expect to lower the upper limit for | < m > | to 0.2 eV. In a more distant future a number of collaborations promise an order of magnitude better sensitivity: CUORE (130 Te), GENIUS (76 Ge), EXO (136 Xe), MAJORANA (76 Ge), MOON (100 Mo). 10. Cosmological and stellar inputs [33] Though neutrinos play crucial role in astrophysics, up to now the observations of the stars have not contributed signiﬁcantly to our knowledge of neutrino masses and mixings. Even the most spectacular supernova SN 1987 A provided only a rather modest upper limit on neutrino masses.

PHENOMENOLOGY OF NEUTRINO OSCILLATIONS

173

From the theory of white dwarfs an upper limit on neutrino magnetic moment can be deduced µνe ≤ 10−10 µB , where µB is Bohr magneton. The early universe turns out to be more informative than stars. The precision measurements of cosmic microwave background (CMB) by Wilkinson Microwave Anisotropy Probe (WMAP) in conjuction with data on large scale structure, 2dFGRS, Lyman Alpha forest and a host of other astronomical measurements allow to ﬁnd the best ﬁt to a set of cosmological parameters and in particular to the upper limit on Ων – the energy density in neutrinos. In case of 3 degenerate neutrino species Σ = 94Ων h2 , where Σ = m1 + m2 + m3 , Ων is in units of critical density, and Hubble constant h = H/H0 , where H0 = 100km · s−1 · Mpc−1 . According to the 2003 ﬁt of WMAP, Σ < 0.7 eV. 11. Challenges and Prospects [31,34] Challenge #1. Absolute values of neutrino masses (from oscillations we know only the values of |∆m212 | and |∆m223 |). The problem could be solved by KATRIN. At present even the sign of ∆m223 is not known. Challenge #2. Majorana vs Dirac. The problem could be solved by observing the neutrinoless double beta decay. Challenge #3. CP-violation in leptonic sector. This is especially difﬁcult because we know that θ13 1 and ∆m221 /∆m232 1. Note that for θ13 = 0 and/or ∆m221 /∆m232 = 0 the CP-violating effects disappear (recall the mechanism by Kobayashi and Maskawa). Of special interest would be the measurement of α21 and α31 if the neutrinoless double beta decay is discovered. In most of the cases only disappearance effects have been observed (disappearance of νµ in atmospheric neutrinos, disappearance of νe and ν¯e in solar and reactor neutrinos). What is needed is the appearance effects, such as the appearance of ντ in the CERN Gran Sasso project (detectors Opera, Icarus). It is necessary to reach much higher statistics, high intensity, high energy, clean beams, such as beta beams from beta decay of relativistic nuclei, or from neutrino factories where neutrinos appear in the decays of stored relativistic muons. In the nearest perspective of great interest is the 50 GeV JHF PS. It is necessary to reach much higher accuracy in measurements of the key parameters: δ(∆m232 ) = 10−4 eV2 , δ(sin2 2θ13 ) = 5 · 10−3 , δ(sin2 2θ32 ) = 10−2 . Solving the puzzles of neutrinos is an absolute must, because these particles are at the focal point of particle physics, astrophysics and cosmology.

174

LEV B. OKUN

Acknowledgements Many thanks to Laurent Baulieu and Boris Pioline for their warm hospitality. I am grateful to A.D. Dolgov, O.V. Lychkovsky, A.A. Mamonov, M.V. Rotaev, M.G. Schep kin, V.L. Telegdi, and M.I. Vysotsky for many valuable comments. This work was partly supported by RFBR grant No.04-02-16538 and by A.von Humboldt Award. 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. 27. 28. 29. 30. 31.

H. Becquerel, Comp. Rend. 12 (1896) 501, 509. Discovery of radioactivity of uranium salts (in fact, of thorium β-rays). E. Rutherford, Phil. Mag., ser.6, 21 (1911) 669. W. Pauli, Letter of 4 December 1930 to a meeting of “radioactive ladies and gentlemen” at T¨ubingen (reproduced in L. Brown Phys. Today 31, No.9 (1978) 23. E. Fermi, Zs. f. Phys. 88 (1934) 161. Theory of β-decay. E. Majorana, Nuovo Cimento 14 (1937) 171. Majorana neutrino. B. Pontecorvo, Report PD-205, Chalk River, 1946, National Research Council of Canada. F. Reines, C.L. Cowan, Phys. Rev. 92 (1953) 830. G. Danby, J.-M. Gaillard, K. Goulianos, L.M. Lederman, N. Mistry, M. Schwartz, J. Steinberger, Phys. Rev. Lett. 9 (1962) 36. Experimental discovery of two kinds of neutrinos. B. Pontecorvo, ZhETF 33 (1957) 549. The idea of νe ↔ ν¯e oscillations. B. Pontecorvo, ZhETF 37 (1959) 1751. Proposal to search for νµ . Z. Maki, M. Nakagawa and S. Sakata, Progr. Theor. Phys. 28 (1962) 870. The idea of νe ↔ νµ mixing. V. Gribov, B. Pontecorvo, Phys. Lett. 28B (1969) 493. Description of νe ↔ νµ oscillations. S. Bilenky, B. Pontecorvo, Lett Nuovo Cim. 17 (1976) 569. H. Fritzsch, P. Minkowski, Phys. Lett. B62 (1976) 72. Formula for νe ↔ νµ oscillations without “factor of 2” misprint. L. Wolfenstein, Phys. Rev. D17 (1978) 2369. Neutrino oscillations in matter. S. Mikheev, A. Smirnov, Yad. Fiz. 42 (1985) 1441. Resonance neutrino transitions in matter. Sov. J. Nucl. Phys. 42 (1986) 913. I. Kobzarev, B. Martemyanov, L. Okun, M. Schepkin, Sov. J. Nucl. Phys. 35 (1982) 708. Example of equal energy scenario. L. Stodolsky, Phys. Rev. D58 (1998) 036006. Arguments in favor of equal energy scenario. H. Lipkin, hep-ph/0212093. On the role of static detector in neutrino oscillations. M.I. Vysotsky, CKM matrix and CP violation in B-mesons, Surveys in High Energy Physics, 2003, vol. 18 (1-4), pp. 19-54; hep-ph/0307218, p.18. A. Dolgov, Phys. Rep. 370 (2002) 333. M. Beuthe, Phys. Rep. 375 (2003) 105, hep-ph/0109119. On the role of wave packets. M. Nauenberg, Phys. Lett. B447 (1999) 23, Err. Phys. Lett. B452 (1999) 439, hepph/9812441. Oscillations of neutrinos from pion decay. B. Kayser, Phys. Lett. B592, Issues 1-4 (July 2004) 145. Neutrino mass, mixing and ﬂavor change. R.N. Mohapatra, APS Neutrino Study: Theory of neutrinos, hep-ph/0412099. H. Back et al., APS Neutrino Study: Report of the Solar and Atmospheric Neutrino Experiments Working Group, hep-ex/0412016. M. Goodman, Long-Baseline Neutrino News. Monthly selection of news from . Neutrino oscillation industry. A website on neutrino experiments. E. Abouzaid et al., APS Neutrino Study: Report of the Reactor working group, www.aps.org/neutrino; http://apsreactor.uchicago.edu/report.html. M. Apolonio et al., CHOOZ, hep-ex/0301017. S.J. Freedman, B. Kayser, APS Neutrino Study: The neutrino matrix, physics/0411216.

PHENOMENOLOGY OF NEUTRINO OSCILLATIONS 32. 33. 34.

175

C. Aalseth et al., APS Neutrino Study: Neutrinoless double beta decay and direct searches for neutrino mass, hep-ph/0412300. S.W. Barwick et al., APS Neutrino Study: Report of the Neutrino Astrophysics and Cosmology Working Group, astro-ph/0412544. C. Albright et al., APS Neutrino Study: Neutrino Factory and Beta Beam Experiments and Development, www.aps.org/neutrino.

CLOSED STRINGS IN MISNER SPACE: A TOY MODEL FOR A BIG BOUNCE ?

BRUNO DURIN LPTHE, Universit´e Paris 6, 4 place Jussieu, 75252 Paris cedex 05, France AND BORIS PIOLINE LPTHE, Universit´e Paris 6, 4 place Jussieu, 75252 Paris cedex 05, France and LPTENS, D´epartement de Physique de l’Ecole Normale Sup´erieure 24 rue Lhom*ond, 75231 Paris cedex 05, France

Abstract. Misner space, also known as the Lorentzian orbifold R1,1 /boost, is one of the simplest examples of a cosmological singularity in string theory. In this lecture, we review the semi-classical propagation of closed strings in this background, with a particular emphasis on the twisted sectors of the orbifold. Tree-level scattering amplitudes and the one-loop vacuum amplitude are also discussed.

Thus I was moving along the sloping curve of the time loop towards that place in which the Friday me before the beating would change into the Friday me already beaten. I. Tichy, [1]

Despite their remarkable success in explaining a growing body of high precision cosmological data, inﬂationary models, just as the Hot Big Bang Model, predict an Initial Singularity where effective ﬁeld theory ceases to be valid [2]. As a quantum theory of gravity, String Theory ought to make sense even in this strongly curved regime, possibly by providing an initial quantum state if the Initial Singularity is truely an Origin of Time, or by escaping it altogether if stringy matter turns out to be less prone to gravitational collapse than conventional ﬁeld-theoretic matter. Unfortunately, describing cosmological singularities and, less ambitiously, time dependence in string theory has been a naggingly difﬁcult task, partly because of the absence of a tractable closed string ﬁeld theory framework. Unless stringy (α ) corrections in the two-dimensional sigma model are sufﬁcient to eliminate the singularity, quantum (gs ) corrections are expected to be important due to 177 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 177–200. © 2005 Springer. Printed in the Netherlands.

178

BRUNO DURIN AND BORIS PIOLINE

the large blue-shift experienced by particles or strings as they approach the singularity, invalidating a perturbative approach. Nevertheless, one may expect cosmological production of particles, strings and other extended states near the singularity to qualitatively alter the dynamics, and it is not unplausible, though still speculative, that their contribution to the vacuum energy be sufﬁcient to lead to a Big Bounce rather than an Big Bang. In order to make progress on this issue, it is useful to study toy models where at least α corrections are under control, and study string production to leading order in gs . Orbifolds, being locally ﬂat, are immune to α corrections, and thus a good testing playground. One of the simplest examples of time-dependent orbifolds1 is the Lorentzian orbifold R1,1 /boost [7, 8, 9], formerly known as Misner space [3] in the gravity literature. Introduced as a local model for the cosmological singularity and chronological horizon of Lorentzian Taub-NUT space, Misner space was shown long ago to exhibit divergences from quantum vacuum ﬂuctuations in ﬁeld theory, at least for generic choices of vacua [22]. Not surprisingly, this is also true in string theory, although less apparent since the local value of the energy-momentum tensor is not an on-shell observable [5]. Similarly, just as in ﬁeld theory, tree-level scattering amplitudes of ﬁeld-theoretical (untwisted) states have been found to diverge, as a result of large graviton exchange near the singularity [32]. While these facts ominously indicate that quantum back-reaction may drastically change the character of the singularity, experience from Euclidean orbifolds suggests that twisted states may alleviate the singularities of the effective ﬁeld theory description, and that it may be worthwhile to investigate their classical behaviour, overpassing the probable inconsistency of perturbation theory. Indeed, in the context of Misner space, twisted states are just strings that wind around the collapsing spatial direction, and become the lightest degrees of freedom near the singularity. In these notes, we review classical aspects of the propagation of closed strings in Misner space, with particular emphasis on twisted states, based on the recent works [4, 5, 6]. The outline is as follows. In Section 2, we describe the semi-classical dynamics of charged particles and winding strings, and compute their cosmological production rate, at tree level in the singular Misner geometry – although our approach is applicable to more general cases. In Section 3, we analyze the imaginary part of the one-loop amplitude, which carries the same information in principle. In Section 4, we review recent results on scattering amplitudes of untwisted and twisted states, and their relation to the problem of classical back-reaction from a “condensate” of twisted states. Section 5 contains our closing remarks. 1. Semi-classics of closed strings in Misner space 1.1. MISNER SPACE AS A LORENTZIAN ORBIFOLD Misner space was ﬁrst introduced in the gravity literature as a local model [3] for the singularities of the Taub-NUT space-time [23] . It can be formally deﬁned as the quo1

The orbifold of R1,1 under time reversal may be even simpler, but raises further puzzles related to time unorientability [10]. Discussions of other exact cosmological backgrounds in string theory include [11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21].

CLOSED STRINGS IN MISNER SPACE

179

tient of two-dimensional2 Minkowski space R1,1 by the ﬁnite boost transformation B : (x+ , x− ) → (e2πβ x+ , e−2πβ x− ), where x± are the light-cone coordinates. As such, it is a locally ﬂat space, with curvature localized at the ﬁxed locus under the identiﬁcation, i.e. on the light-cone x+ x− = 0. The geometry of the quotient can be pictured as four Lorentzian cones touching at their apex (See Figure 1), corresponding to the four quadrants of the covering space R1,1 . Choosing coordinates adapted to the boost B, √ x± = T e±βθ / 2 , x+ x− > 0 (Milne regions) (1.1) √ ± ±βη + − x = ±re / 2, x x < 0 (Rindler regions) (1.2) where, due to the boost identiﬁcation, the coordinates θ and η are compact with period 2π, the metric of the quotient can be written as " # −dT 2 + β 2 T 2 dθ2 2 + − (1.3) ds = −2 dx dx = dr2 − β 2 r2 dη 2 The two regions x+ < 0, x− < 0 (P) and x+ > 0, x− > 0 (F), describe contracting and expanding cosmologies where the radius of the spatial circle parameterized by θ changes linearly in time, and are often called (compactiﬁed) Milne regions. The space-like cones x+ > 0, x− < 0 (R) and x+ < 0, x− > 0 (L), often termed “whiskers”, are instead timeindependent Rindler geometries with compact time η 3 . The Milne and Rindler regions, tensored with a sphere of ﬁnite size, describe the Taub and NUT regions, respectively, of the Taub-NUT space-time in the vicinity of one of its inﬁnite sequence of cosmological singularities. It also captures the local geometry in a variety of other cosmological string backgrounds [12, 15, 17, 18]. It is also interesting to note that, combining the boost B with a translation on a spectator direction, one obtains the Gott space-time, i.e. the geometry around cosmic strings in four dimensions [25]. Due to the compactness of the time coordinate η, both Misner and Taub-NUT spacetimes contain closed timelike curves (CTC) which are usually considered as a severe pathology. In addition to logical paradoxes and exciting prospects [1] raised by timeloops, the energy-momentum tensor generated by a scalar ﬁeld at one-loop is typically divergent, indicating a large quantum back-reaction. According to the Chronology Protection Conjecture, this back-reaction may prevent the formation of CTC altogether [26]. String theorists need not be intimidated by such considerations, and boldly go and investigate whether the magics of string theory alleviate some of these problems. String theory on a quotient of ﬂat space4 is in principle amenable to standard orbifold conformal ﬁeld theory techniques, although the latter are usually formulated for Euclidean orbifolds. While backgrounds with Lorentzian signature can often be dealt with by (often subtle) Wick rotation from Euclidean backgrounds, the real complication stems from the fact that the orbifold group is inﬁnite, and its action non proper 5 . This however need not 2

Higher dimensional analogues have also been considered [24]. This should not be confused with thermal Rindler space, which is periodic in imaginary time. 4 String theory on Taub-NUT space, which is not ﬂat, has been studied recently using heterotic coset models [20]. 5 Deﬁning X + = Z , X − = −Z¯ in the Rindler region, one obtains an orbifold of R2 by a rotation with an irrational angle. A related model has been studied recently in [27]. 3

180

BRUNO DURIN AND BORIS PIOLINE

Figure 1. Free particles or untwisted strings propagate from the past Milne region to the future Milne region, with a temporary excursion in the whiskers.

be a problem at a classical level: as we shall see, free strings propagate in a perfectly well-deﬁned fashion on this singular geometry. 1.2. PARTICLES IN MISNER SPACE As in standard orbifolds, part of the closed string spectrum consists of conﬁgurations on the covering space, which are invariant under the orbifold action. Such “untwisted” states behave much like point particles of arbitrary mass and spin. Their trajectory, aside from small-range string oscillations, consist of straight lines on the covering space: ± X0± = x± 0 +p τ ,

(1.4)

where m2 = 2p+ p− includes the contribution from momentum in the transverse directions to Misner space as well as string oscillators. The momentum along the compact − + − direction is the “boost momentum” j = x+ 0 p + x0 p , and is quantized in units of 1/β in the quantum theory. A massive particle with positive energy (p+ , p− > 0) thus comes in from the inﬁnite past in the Milne region at τ = −∞ and exits in the future Milne region at τ = +∞, after wandering in the Rindler regions for a ﬁnite proper time. As the particle approaches the light-cone from the past region, its angular velocity dθ/dT ∼ 1/T along the Milne circle increases to inﬁnity by the familiar “spinning skater” effect. It is therefore expected to emit abundant gravitation radiation, and possibly lead to large backreaction. From the point of view of an observer in one of the Rindler regions, an inﬁnite number of particles of Rindler energy j are periodically emitted from the horizon at r = 0 and travel up to a ﬁnite radius r = |j|/M before being reabsorbed into the singularity – and so on around the time loop. Quantum mechanically, the center of mass of a (spinless) untwisted string is described by a wave function, solution of the Klein-Gordon equation in the Misner geometry. Diagonalizing the boost momentum j, the radial motion is governed by a Schr¨odinger equation √ " " 2 −∂x − m2 e2y − j 2 = 0 T = ±√ 2x+ x− = ex where (1.5) −∂y2 + m2 e2y − j 2 = 0 r = ± −2x+ x− = ey

CLOSED STRINGS IN MISNER SPACE

181

The particle therefore bounces against an exponentially rising, Liouville-type wall in the Rindler regions, while it is accelerated in a Liouville-type well in the Misner regions. Notice that, in both cases, the origin lies at inﬁnite distance in the canonically normalized coordinate x (y, resp.). Nevertheless, in and out type of wave functions can be deﬁned in each region, and extended to globally deﬁned wave functions by analytic continuation across the horizons at x+ x− = 0. Equivalently, the wave function for an untwisted string in Misner space may be obtained by superposing a Minkowski plane wave with its images under the iterated boosts B n , n ∈ Z. Performing a Poisson resummation over the integer n, one obtains wave functions with a well deﬁned value of the boost momentum j, as a continuous superposition of plane waves ∞ + − fj,m2 ,s (x , x ) = dv exp ip+ X − e−2πβv + ip− X + e2πβv + ivj + vs (1.6) −∞

where s denotes the SO(1, 1) spin in R1,1 [9, 5]. This expression deﬁnes global wave functions in all regions, provided the v-integration contour is deformed to (−∞−i , +∞+ i ). In particular, there is no overall particle production between the (adiabatic) in vacuum at T = −∞ and the out vacuum at T = +∞, however there is particle production between the (adiabatic) in vacuum at T = −∞ and the (conformal) out vacuum at T = 0− . This is expected, due to the “spinning skater” inﬁnite acceleration near the singularity, as mentioned above. 1.3. WINDING STRINGS IN MISNER SPACE In addition to the particle-like untwisted states, the orbifold spectrum contains string conﬁgurations which close on the covering space, up to the action of an iterated boost B w : X ± (σ + 2π, τ ) = e±2πwβ X ± (σ, τ )

(1.7)

In the Milne regions, they correspond to strings winding w times around the compact space-like dimension Sθ1 , which become massless at the cosmological singularity. They are therefore expected to play a prominent rˆole in its resolution, if at all. In the Rindler regions, they instead correspond to strings winding around the compact time-like dimension Sη1 . Given that a time-loop exist, there is nothing a priori wrong about a string winding around time: it is just a superposition of w static (or, more generally, periodic in time) strings, stretched (in the case of a cylinder topology) over an inﬁnite radial distance. In order to understand the semi-classical aspects of twisted strings [5], let us again truncate to the modes with lowest worldsheet energy, satisfying (1.7): X0± (σ, τ ) =

1 ∓νσ ± ±ντ ±α0 e e ∓α ˜ 0± e∓ντ . ν

(1.8)

where ν = −wβ. As usual, the Virasoro (physical state) conditions determine the mass and momentum of the state in terms of the oscillators, M 2 = 2α0+ α0− ,

˜ 2 = 2˜ M α0+ α ˜ 0−

(1.9)

182

BRUNO DURIN AND BORIS PIOLINE

˜ 2 = m2 − jν) are the contributions of the left-moving (resp. where (M 2 = m2 + jν, M right-moving) oscillators. Restricting√to j = 0 for simplicity, one may thus choose α± and α ˜ ± to be all equal in modulus to m/ 2, up to choices of sign leading to two qualitatively different kinds of twisted strings: ˜ − > 0, one obtains short string conﬁgurations − For α+ α m X0± (σ, τ ) = √ sinh(ντ ) e±νσ ν 2

(1.10)

winding around the Milne space-like circle, and propagate from inﬁnite past to inﬁnite future (for α+ > 0). When j = 0, they also extend in the Rindler regions to 2 ˜ )2 /(4ν 2 ), after experiencing a signature ﬂip on the = (M − M a ﬁnite distance r− worldsheet. ˜ − < 0, one obtains long string conﬁgurations, − For α+ α m X0± (σ, τ ) = √ cosh(ντ ) e±νσ ν 2

(1.11)

propagating in the Rindler regions only, and winding around the time-like circle. They correspond to static conﬁgurations which extend from spatial inﬁnity in L or R 2 ˜ )2 /(4ν 2 ), and folding back to inﬁnity again. = (M + M to a ﬁnite distance r+ Notice how, in contrast to Euclidean orbifolds, twisted strings are in no sense localized near the singularity ! ˜ 0± become hermitian operators Quantum mechanically, the (quasi) zero-modes α0± , α with commutation rules [9, 4] [α0+ , α0− ] = −iν ,

[˜ α0+ , α ˜ 0− ] = iν

(1.12)

Representing α0+ as a creation operator in a Fock space whose vacuum is annihilated by α0− , introduces an imaginary ordering constant iν/2 in (1.9) after normal ordering, which cannot be cancelled by any of the higher modes in the spectrum6 . Thus, in this scheme, there are no physical states in the twisted sector [9]. However, this quantization does not maintain the hermiticity of the zero-mode operators. The analogy of (1.12) to the problem of a charged particle in an electric ﬁeld will take us to the appropriate quantization scheme in the next section. 1.4. WINDING STRINGS VS. CHARGED PARTICLES Returning to (1.11), one notices that the complete worldsheet of a twisted closed string can be obtained by smearing the trajectory of the left-movers (i.e. a point with τ + σ = cste) ˜ 0± and under the action of continuous boosts (See Figure 2). In particular, setting a± 0 = α ± ± α0 /ν, the trajectory of the left-movers becomes x0 = ∓˜ X ± (τ ) = x± 0 ±

a± 0 ±ντ e . ν

(1.13)

6 Higher excited modes have energy n±iν, and can be quantized in the usual Fock space scheme, despite the Lorentzian signature of the light-cone directions [4].

CLOSED STRINGS IN MISNER SPACE

183

Figure 2. Closed string worldsheets in Misner space are obtained by smearing the trajectory of a charged particle in Minkowski space with a constant electric ﬁeld. Short (resp. long) strings correspond to charged particles which do (resp. do not) cross the horizon.

which is nothing but the worldline of a particle of charge w in a constant electric ﬁeld E = β ! Indeed, it is easily veriﬁed that the short (long, resp.) string worldsheet can be obtained by smearing the worldline of a charged particle which crosses (does not, resp.) the horizon at x+ x− = 0. Quantum mechanically, it is easy to see that this analogy continues to hold [4, 5] : the usual commutation relations for a particle in an electric ﬁeld − [a+ 0 , a0 ] = −iν ,

− [x+ 0 , x0 ] = −

i , ν

(1.14)

reproduce the closed string relations (1.12) under the identiﬁcation above. The mass of the charged particle M 2 = α+ α− +α− α+ reproduces the left-moving Virasoro generator m2 + νj as well. It is therefore clear that the closed string zero-modes, just as their charge particle counterpart, can be represented as covariant derivatives acting on complex wave functions φ(x+ , x− ): α0± = i∇∓ = i∂∓ ±

ν ± x , 2

˜ ∓ = i∂∓ ∓ α ˜ 0± = i∇

ν ± x 2

(1.15)

in such a way that the physical state conditions are simply the Klein-Gordon operators for a particle with charge ±ν in a constant electric ﬁeld, M 2 = ∇+ ∇− + ∇− ∇+ ,

˜− + ∇ ˜+ ˜2 = ∇ ˜ +∇ ˜ −∇ M

(1.16)

corresponding to the location of Coordinates x± are the (Heisenberg picture) operators √ the closed string at σ = 0. The radial coordinate ±2x+ x− associated to the coordinate representation (1.15), should be thought of as the radial position of the closed string in the Milne or Rindler regions. From this point of view, it is also clear while the quantization scheme based on a Fock space has failed: the Klein-Gordon equation of a charged particle in an electric ﬁeld is equivalent, for ﬁxed energy pt , to a Schr¨odinger equation with an inverted harmonic potential, (1.17) −∂x2 + m2 − (pt − Ex)2 ≡ 0 In contrast to the magnetic case which leads to a positive harmonic potential with discrete Landau levels, the spectrum consists of a continuum of delta-normalization scattering

184

BRUNO DURIN AND BORIS PIOLINE

states which bounce off (and tunnel through) the potential barrier. These scattering states are the quantum wave functions corresponding to electrons and positrons being reﬂected by the electric ﬁeld, and their mixing under tunneling is a reﬂection of Schwinger production of charged pairs from the vacuum. In order to apply this picture to twisted closed strings however, we need to project on boost invariant states, and therefore understand the charged particle problem from the point of view of an accelerated observer in Minkowski space, i.e. a static observer in Rindler space. 1.5. CHARGED PARTICLES IN MISNER SPACE Charged particles in Rindler space have been discussed in [28]. Classical trajectories are, of course, the ordinary hyperbolae from Minkowski space, translated into the Rindler coordinates (y = er , η). For a ﬁxed value j of the energy conjugate to the Rindler time η, the radial motion is governed by the potential 2 2 ˜2 ˜2 ν 2 1 2 M 2M M2 + M 2 2 − r = − (1.18) V (y) = M r − j + νr 2 ν2 2ν 2 where, in the last equality, we have translated the charged particle data into closed string data. In contrast with the neutral case (ν = 0), the potential is now unbounded from below at r = ∞. For j < M 2 /(2ν) (which is automatically obeyed in the closed string ˜ 2 > 0 for non-tachyonic states), the r = 0 and r = ∞ asymptotic regions case, where M are separated by a potential barrier (See Figure 3). Particles on the right (r → ∞) of the barrier correspond to electrons coming from and returning to Rindler inﬁnity, while, for j > 0 (resp. j < 0), particles on the left (r → 0) correspond to positrons (resp. electrons) emitted from and reabsorbed by the Rindler horizon. Quantum tunneling therefore describes both Schwinger pair production in the electric ﬁeld (when j > 0), and Hawking emission of charged particles from the horizon (when j < 0). Similarly, the trajectories of charged particles in Milne space correspond to other branches of the same hyperbolae, and their motion along the cosmological time T , for a ﬁxed value of the momentum j conjugate to the compact spatial direction θ, is governed by the potential 2 2 ˜2 ˜2 ν 2 1 2 M 2M M2 + M 2 2 + T = − (1.19) V (T ) = −M T − j + νT 2 ν2 2ν 2 The potential is maximal and negative at T = 0, although this is at inﬁnite distance in the canonically normalized coordinate x. The classical motion therefore covers the complete time axis T ∈ R. Quantum mechanically, the Klein-Gordon equation in the Rindler region is equivalent to a Schr¨odinger equation in the potential (1.18) or (1.19) at zero-energy, and can be solved in terms of Whittaker functions [28]. Bases of in and out modes can be deﬁned in each quadrant and analytically continued accross the horizons, e.g. in the right Rindler region j = e−ijη r−1 M−i( j − M 2 ),− ij (iνr2 /2) (1.20) Vin,R 2

2ν

2

CLOSED STRINGS IN MISNER SPACE

185

j0 V e

e

y

0j1 2 V e

e

y

j1 2 V e

e

y

j^2

Figure 3.

Left: Classical trajectories of a charged particle in Rindler/Milne space. j labels the Rindler energy or Milne momentum, and is measured in units of M 2 /ν. Right: Potential governing the radial motion in the right Rindler region, as a function of the canonical coordinate y = er . corresponds to incoming modes from Rindler inﬁnity r = ∞, while j Uin,R = e−ijη r−1 Wi( j − M 2 ), ij (−iνr2 /2) 2

2ν

2

(1.21)

corresponds to incoming modes from the Rindler horizon r = 0. As usual in time-

186

BRUNO DURIN AND BORIS PIOLINE

dependent backgroungs, the in and out vacua are related by a non-trivial Bogolubov transformation, which implies that production of correlated pairs has taken place. The Bogolubov coefﬁcients have been computed in [28, 4], and yield the pair creation rates in the Rindler and Milne regions, respectively: QR = e−πM

2

| sinh πj| /2ν , ˜ 2 /2ν cosh π M

QM = e−πM

2

/2ν

˜ 2 /2ν cosh π M | sinh πj|

,

(1.22)

˜ 2 ν, these indeed agree with the tunnelling (or scattering In the classical limit M 2 , M over the barrier, in the Milne regions) rate computed from (1.19). 1.6. SCHWINGER PAIR PRODUCTION OF WINDING STRINGS Having understood the quantum mechanics of charged particles in Minkowski space from the point of view of an accelerating observer, we now return to the dynamics of twisted strings in Misner space. The wave function of the quasi-zero-modes α0± , α0± is governed by the same Klein-Gordon equation as in the charged particle case, although only the dependence on the radial coordinate r is of interest. Its interpretation is however rather different: e.g, a particle on the right of the potential in the right Rindler region corresponds to an inﬁnitely long string stretching from inﬁnity in the right whisker to a ﬁnite radius r+ and folded back onto itself, while a particle on the left of the potential is a short string stretching from the singularity to a ﬁnite radius r− . Quantum tunneling relates the two type of states by evolution in imaginary radius r, and can be viewed semi-classically as an Euclidean strip stretched between r+ and r− . The wave functions in the Milne region are less exotic, corresponding to incoming or outgoing short strings at inﬁnite past, future or near the singularity. In order to compute pair production, one should in principle deﬁne second quantized vacua, i.e. choose a basis of positive and negative energy states. While it is clear how to do so for short strings in the Milne regions, second quantizing long strings is less evident, as they carry an inﬁnite Rindler energy7 , and depend on the boundary conditions at r = ∞. However, they are likely to give the most natural formulation, as any global wave function in Misner space can be written as a state in the tensor product of the left and right Rindler regions: the entire cosmological dynamics may thus be described as a state in a time-independent geometry, albeit with time loops ! Fortunately, even without a proper understanding of these issues, one may still use the formulae (1.22) to relate incoming and outcoming components of the closed string wave functions, and compute pair production for given boundary conditions at Rindler inﬁnity. In particular, it should be noted that the production rate in the Milne regions QM is inﬁnite for vanishing boost momentum j = 0, as a consequence of the singular geometry. Moreover, although our analysis has borrowed a lot of intuition from the analogy to the charged particle problem, we are now in a position to describe pair production of 7

[5].

The latter can be computed by quantizing the long string worldsheet using σ as the time variable

CLOSED STRINGS IN MISNER SPACE

187

winding strings in any geometry of the form ds2 = −dT 2 + a2 (T )dθ2

or

ds2 = dr2 − b2 (r)dη 2

(1.23)

(despite the fact that these geometries are not exact solutions of the tree level string equations of motion, they may be a useful mean ﬁeld description of the back-reacted geometry). Neglecting the contributions of excited modes (which no longer decouple since the metric is not ﬂat), the wave equation for the center of motion of strings winding around the compact direction θ or η, is obtained by adding to the two-dimensional Laplace operator describing the free motion of a neutral particle, the contribution of the tensive energy carried by the winding string: $

1 a(T ) ∂T a(T ) 1 b(r) ∂r b(r)

∂T + ∂r +

j2 1 2 2 2 a2 (T ) + 4 w a (T ) − m = 2 j 1 2 2 2 b2 (r) + 4 w b (r) − m = 0

(1.24)

Choosing a(T ) = βT or a(r) = βr and multiplying out by (a2 (T ), b2 (r)), these equations indeed reduce to (1.19) and (1.18)8 . In particular, for a smooth geometry, the production rate of pairs of winding strings is ﬁnite. 2. One-loop vacuum amplitude In the previous section, we have obtain the production rate of winding strings in Misner space, from the Bogolubov coefﬁcients of the tree-level wave functions. In principle, the same information could be extracted from the imaginary part of the one-loop amplitude. In this section, we start by reviewing the vacuum amplitude and stress-energy in ﬁeld theory, and go on to study the one-loop amplitude in string theory, both in the twisted and untwisted sectors. 2.1. FIELD THEORY The one-loop energy-momentum tensor generated by the quantum ﬂuctuations of a free ﬁeld φ with (two-dimensional) mass M 2 and spin s can be derived from the Wightman functions at coinciding points (and derivatives thereof). These depend on the choice of vacuum: in the simplest “Minkowski” vacum inherited from the covering space, any Green function is given by a sum over images of the corresponding one on the covering space. Using a (Lorentzian) Schwinger time representation and integrating over momenta, we obtain µ

µ

G(x ; x ) =

∞ l=−∞

∞

dρ (iρ)−D/2 exp iρM 2 − 2πsl

(2.1)

i + 2πβl + − −2πβl − x )(x − e x ) exp − (x − e 4ρ

8 Notice that, in disagreement to a claim in the literature [29], the wave equation for j = 0 is not regular at the origin.

188

BRUNO DURIN AND BORIS PIOLINE

where s is the total spin carried by the ﬁeld bilinear. Taking two derivatives and setting x = x , one ﬁnds a divergent stress–energy tensor [22] Tµν dxµ dxν =

1 K −dT 2 − 3T 2 dη 2 2 4 12π T

(2.2)

where the constant K is given by ∞

K=

cosh[2πnsβ]

n=1

2 + cosh 2πnβ (cosh 2πnβ − 1)2

(2.3)

This divergence is expected due to the large blue-shift of quantum ﬂuctuations near the singularity. Notice that for spin |s| > 1, the constant K itself becomes inﬁnite, a reﬂection of the non-normalizability of the wave functions for ﬁelds with spin. In string theory, the local expectation value 0|Tab (x)|0ren is not an on-shell quantity, hence not directly observable. In contrast, the integrated free energy, given by a torus amplitude, is a valid observable9 . In ﬁeld theory, the free energy may be obtained by integrating the propagator at coinciding points (2.1) once with respect to M 2 , as well as over all positions, leading to F=

∞

+

dx dx

−

∞

l=−∞

dρ D (iρ)1+ 2

exp −8i sinh2 (πβl)x+ x− + iM 2 ρ

(2.4)

In contrast to the ﬂat space case, the integral over the zero-modes x± , x does not reduce to a volume factor, but gives a Gaussian integral, centered on the light cone x+ x− = 0. Dropping as usual the divergent l = 0 ﬂat-space contribution and rotating to imaginary Schwinger time, one obtains a ﬁnite result F=

+∞ l=−∞,l =0

∞

2

dρ e−M ρ−2πβsl D 2 ρ1+ 2 sinh (πβl)

(2.5)

Consistently with the existence of globally deﬁned positive energy modes for (untwisted) particles in Misner space, F does not have any imaginary part, implying the absence of net particle production between past and future inﬁnity. 2.2. STRING AMPLITUDE IN THE UNTWISTED SECTOR We may now compare the ﬁeld theory result (2.5) to the one-loop vacuum amplitude in string theory with Euclidean world-sheet and Minkowskian target space, as computed in [9, 16]: 2 2 2 R2 ∞ dρd¯ ρ e−2πβ w ρ2 − 4πρ2 |l+wτ | (2.6) Abos = 2 2 13 |η 21 (ρ) θ1 (iβ(l + wρ); ρ)| F l,w=−∞ (2π ρ2 ) 9 Of course, the spatial dependence of the one-loop energy may be probed by scattering e.g. gravitons at one-loop.

CLOSED STRINGS IN MISNER SPACE

189

where θ1 is the Jacobi theta function, θ1 (v; ρ) = 2q 1/8 sin πv

∞

(1 − e2πiv q n )(1 − q n )(1 − e−2πiv q n ) ,

q = e2πiρ (2.7)

n=1

In this section, we restrict to the untwisted sector w = 0. Expanding in powers of q, it is apparent that the string theory vacuum amplitude can be viewed as the ﬁeld theory result (2.5) summed over the spectrum of (single particle) excited states, satisfying the matching condition enforced by the integration over ρ1 . As usual, ﬁeld-theoretical UV divergences at ρ → 0 are cut-off by restricting the integral to the fundamental domain F of the upper half plane. In contrast to the ﬁeld theory result, where the integrated free energy is ﬁnite for each particle separately, the free energy here has poles in the domain of integration, at ρ=

βl m +i n 2πn

(2.8)

Those poles arise only after summing over inﬁnitely many string theory states. Indeed, each pole originates from the (1 − e±2πiv q n ) factor in (2.7), hence re-sums the contributions of a complete Regge trajectory of ﬁelds with mass M 2 = kn and spin s = k (k ∈ Z). In other words, the usual exponential suppression of the partition function by the increasing masses along the Regge trajectory is overcome by the spin dependence of that partition function. Regge trajectories are a universal feature of perturbative string theory, and these divergences are expected generically in the presence of space-like singularities. Since the pole (2.8) occurs both on the left- and right-moving part, the integral is not expected to give any imaginary contribution, in contrast to the charged open string case considered in [30]. 2.3. STRING AMPLITUDE IN THE TWISTED SECTORS We now turn to the interpretation of the string one-loop amplitude in the twisted sectors (w = 0), following the analysis in [5]. As in the rest of this lecture, it is useful to truncate the twisted string to its quasi-zero-modes, lumping together the excited mode contribu˜ 2 . Equivalently, we truncate tions into a left and right-moving mass squared M 2 and M the path integral to the “mini-superspace” of lowest energy conﬁgurations on the torus of modulus ρ = ρ1 + iρ2 , satisfying the twisted boundary conditions, X± = ±

1 ± ∓(νσ−iAτ ) 1 ± ∓(νσ+iAτ ˜ ) α e α ˜ e ∓ 2ν 2ν

(2.9)

k˜ l + ρ1 w A˜ = + iβ ρ2 ρ2

(2.10)

where A=

l + ρ1 w k − iβ , ρ2 ρ2

where k, k˜ are a pair of integers labelling the periodic trajectory, for ﬁxed twist numbers (l, w). Notice that (2.9) is not a solution of the equations of motion, unless ρ coincides with one of the poles. In order to satisfy the reality condition on X ± , one should restrict ˜ α± = −(˜ to conﬁgurations with k = k, α± )∗ . Nevertheless, for the sake of generality we

190

BRUNO DURIN AND BORIS PIOLINE

shall not impose these conditions at this stage, but only exclude the case of a degenerate ˜ worldsheet k = −k. We can now evaluate the Polyakov action for such a classical conﬁguration, after rotating τ → iτ : π 2 2 2 ν R2 ρ − [k − i(βl + νρ )] 1 2 2ν 2 ρ2 2 π 2 2 ˜ ˜2 R − 2 ν ρ2 − k + i(βl + νρ1 ) 2ν ρ2

S =−

(2.11)

− 2πijνρ1 + 2πµ2 ρ2 ˜ 2 , summarizes the contributions of excited where the last line, equal to √ −iπρM 2 + iπ ρ˜M √ ± ±η ± ±˜ ˜ η / 2. ˜ = ±Re modes, and α = ±Re / 2, α The path integral is thus truncated to an integral over the quasi-zero-modes α± , α ˜±. 2 2 ˜ Since the action (2.11) depends only on the boost-invariant products R and R , a ﬁrst divergence arises from the integration over η − η˜, giving an inﬁnite factor, independent of the moduli, while the integral over η + η˜ is regulated to the ﬁnite value β after dividing out by the (inﬁnite) order of the orbifold group. ˜ whenever In addition there are divergences coming from the integration over R and R ρ1 = −

ν βl ˜ , − i (k − k) ν 2

ρ2 =

˜ |k + k| 2ν

(2.12)

˜ are precisely the double poles (2.8). These poles are interpreted as which, for k = k, ˜ coming from infrared divergences due to existence of modes with arbitrary size (R, R). ˜ the double poles are now in the complex ρ1 plane, and may contribute for For k = k, speciﬁc choices of integration contours, or second-quantized vacua. In either case, these divergences may be regulated by enforcing a cut off |ρ − ρ0 | > on the moduli space, or an infrared cut-off on R. It would be interesting to understand the deformation of Misner space corresponding to this cut off, analogous to the Liouville wall in AdS3 [31]. ˜ ﬁrst, which is ill-deﬁned at ρ satisfying (2.12), Rather than integrating over R, R we may choose to integrate over the modulus ρ ﬁrst. The integral with respect to ρ1 is Gaussian, dominated by a saddle point at ρ1 = −

˜ 2 − kR2 k˜R βl jνρ2 +i − 2i 2 2 2 ˜ ˜2 ν ν(R + R ) R +R

(2.13)

It is important to note that this saddle point is a local extremum of the Euclidean action, unstable under perturbations of ρ1 . The resulting Bessel-type action has again a stable saddle point in ρ2 , at ˜ ˜ + k| RR|k ρ2 = ! ˜ 2 )(4µ2 − R2 − R ˜ 2 ) − 4j 2 ν 2 ν (R2 + R

(2.14)

CLOSED STRINGS IN MISNER SPACE

191

Integrating over ρ2 in the saddle point approximation, we ﬁnally obtain the action as a ˜ function of the radii R, R: ! ˜ R ˜ 2 )(R2 + R ˜ 2 − 4µ2 ) + 4j 2 ν 2 ˜ (R2 + R |k + k|R S = ˜2) ν(R2 + R ±2πj

˜ 2 − kR2 k˜R ± 2πiβjl ˜2) ν(R2 + R

(2.15)

˜ S admits an extremum at the on-shell where the sign of the second term is that of k + k. values ˜ πM M ˜ ˜ 2 = µ2 + jν R2 = µ2 − jν, |k + k| (2.16) R with action Sk,k˜ = ν Notice that these values are consistent with the reality condition, since the boost momentum j is imaginary in Euclidean proper time. Evaluating (ρ1 , ρ2 ) for the values (2.16), we reproduce (2.12), which implies that the integral is indeed dominated by the region ˜ directions around the saddle point around the double pole. Fluctuations in (ρ1 , ρ2 , R, R) ˜ 2 up have signature (+, +, −, −), hence a positive ﬂuctuation determinant, equal to M 2 M to a positive numerical constant. This implies that the one-loop amplitude in the twisted sectors does not have any imaginary part, in accordance with the naive expectation based on the double pole singularities. It is also in agreement with the answer in the untwisted sectors, where the globally deﬁned in and out vacua where shown to be identical, despite the occurence of pair production at intermediate times. Nevertheless, the instability of the Euclidean action under ﬂuctuations of ρ1 and ρ2 indicates that spontaneous pair production takes place, by condensation of the two unstable modes. Thus, we ﬁnd that winding string production takes place in Misner space, at least for vacua such that the integration contour picks up contributions from these states. This is consistent with our discussion of the tree-level twisted wave functions, where tunneling in the Rindler regions implies induced pair production of short and long strings. The periodic trajectories (2.9) describe the propagation across the potential barrier in imaginary proper time, and correspond to an Euclidean world-sheet interpolating between the Lorentzian world-sheets of the long and short strings. 3. Tree-level scattering amplitudes After this brief incursion into one-loop physics, we now return to the classical realm, and discuss some features of tree-level scattering amplitudes. We start by reviewing the scattering of untwisted modes, then turn to amplitudes involving two twisted modes, which can still be analyzed by Hamiltonian methods. We conclude with a computation of scattering amplitudes for more than 2 twisted modes, which can be obtained by a rather different approach. Our presentation follows [32, 6]. 3.1. UNTWISTED AMPLITUDES Tree-level scattering amplitudes for untwisted states in the Lorentzian orbifold are easily deduced from tree-level scattering amplitudes on the covering space, by the “inheritance

192

BRUNO DURIN AND BORIS PIOLINE

principle”: expressing the wave functions of the incoming or outgoing states in Misner space as superpositions of Minkowski plane waves with well-deﬁned boost momentum j via Eq. (1.6) (with spin s = 0), the tree-level scattering amplitude is obtained by averaging the standard Virasoro-Shapiro amplitude Γ − α4 s Γ − α4 t Γ − α4 u AM ink = δ pi (3.1) Γ 1 + α4 s Γ 1 + α4 t Γ 1 + α4 u i ± ±βvi ± pi , with weight eiji vi , under the actions of continuous boosts p± i → pi (v) = e on all (but one) external momenta. Possible divergences come from the boundary of the parameter space spanned by the vi , where some of the momenta p± i (v) become large. In a general (Gross Mende, (s, t, u → ∞ with s/t, s/u ﬁxed) high energy regime, the Virasoro-Shapiro amplitude is exponentially suppressed [33] and the integral over the vi converges. However, there are also boundary conﬁgurations with s, u → ∞ and ﬁxed t where the Virasoro-Shapiro amplitude has Regge behavior st , in agreement with the √ fact that the size of the string at high energy grows like log s. In this regime, using the Stirling approximation to the Gamma functions in (3.1), it is easy to see that the averaged amplitude behaves as ∞ 1 i i 2 dv exp v i(j2 − j4 ) − α (p1 − p3 ) + 1 (3.2) AM isner ∼ 2

hence diverges for small momentum transfer (pi1 − pi3 )2 ≤ 2/α in the directions transverse to Misner space. There are similar collinear divergences in the other channels as well, both in the bosonic or superstring case. The situation is slightly improved in the case of Grant space (analogous to the “null brane” considered in [34]), i.e. when the boost identiﬁcation is combined with a translation of length R on a direction x2 transverse to the light-cone: in this case, the boost momentum is no longer quantized (although the sum Rp2 + βj still is), and one can construct wave packets which are regular on the horizon, by superposition of states of different boost momentum [34, 16]. Collinear divergences remain, albeit in a reduced range of momentum transfer [6], √ ( 1 + 2α E 2 − 1)2 β 2 (3.3) , E= ( p1 + p3 ) ≤ 2 (α E) R As R → 0, this reduces to Misner space case as expected. As a matter of fact, these divergences may be traced to large tree-level graviton exchange near the singularity, or, in the Grant space case, near the chronological horizon [32]. Collinear divergences can in principle be treated in the eikonal approximation, i.e. by resumming an inﬁnite series of ladder diagrams. While a naive application of the ﬂat space result [35] suggests that this resummation may lead to ﬁnite scattering amplitudes of untwisted states in Misner space [19], a consistent treatment ensuring that only boostinvariant gravitons are exchanged has not been proposed yet, and prevents us from drawing a deﬁnitive conclusion. More generally, it would be extremely interesting to develop eikonal techniques in the presence of space-like singularities, and re-evaluate the claim in [36] that a single particle in Misner space will ineluctably cause the space to collapse.

CLOSED STRINGS IN MISNER SPACE

193

3.2. TWO-TWIST AMPLITUDES As we reviewed in Section 1.4, the zero-mode wave functions in the twisted sectors form a continuum of delta-normalizable states with arbitrarily negative worldsheet energy. In contrast to the standard case of twist ﬁelds of ﬁnite order in Euclidean rotation orbifolds, twisted states in Misner space should thus be described by a continuum of vertex operators with arbitrarily negative conformal dimension. While the conformal ﬁeld theory of such operators remains ill-understood, amplitudes with two twisted ﬁelds only can be computed by ordinary operator methods on the cylinder, in the twisted vacua at τ = ±∞ [6]. Stringy fuzziness. Vertex operators for untwisted states are just a boost-invariant superposition of the ordinary ﬂat space vertex operators. In order to compute their scattering amplitude against a twisted string, it is convenient to write them as a normal ordered expression in the twisted Hilbert space. Since the twisted oscillators have an energy n ± iν with n ∈ Z, normal ordering gives a different contribution than in the untwisted state, − + − + , X≺0 ] − [X>0 , X,< ± X ,≺ ) denote the positive and negative frequency parts (excluding the (quasi) zero-mode contributions) of the embedding coordinates X ± (τ, σ), as deﬁned by the untwisted (resp. twisted) mode expansion. As a result, the vertex operator for an untwisted tachyon becomes

+ − − + + − − + : ei(k X +k X ) :(un.) = exp −k + k − ∆(ν) : ei(k X +k X ) :(ν−tw.) (3.5)

Such a factor is in fact present for all untwisted states, although the normal ordering prescription is slightly more cumbersome for excited states. Since this normal ordering constant depends on the winding number w = −ν/β, it cannot be reabsorbed by a ﬁeld redeﬁnition of the untwisted state, nor of the twisted string. Instead, it can be interpreted as the form factor acquired by untwisted states in the background of a twisted string, due to the zero-point quantum ﬂuctuations of the winding string. The latter polarizes untwisted string states into a cloud of r.m.s. size ∆(ν). which, while proportional to ν at small ν, grows logarithmically with the winding number, ∆(ν) = 2ζ(3)ν 2 + O(ν 4 ) = 2 log ν −

23 + O(ν −2 ) 20

(3.6)

Notice that this logarithmic growth winding can be viewed as the T-dual of the Regge growth with energy. It is also interesting to observe the analogy of the form factor in (3.5) with similar factors appearing in non-commutative gauge theories with matter in the fundamental representation – in line with the general relation between twisted strings and charged particles outlined in Section 1.4. Zero-mode overlaps. In general, the S-matrix element factorizes into a product of an excited mode contribution, which can be evaluated, just as in ﬂat space, by normal ordering and commutation, and a

194

BRUNO DURIN AND BORIS PIOLINE

(quasi)-zero-mode contribution. In the real space representation (1.15) for the quasi-zeromodes, the latter reduces to an overlap of twisted and untwisted wave functions, e.g. in the three tachyon case, − + + − (3.7) dx+ dx− f1∗ (x+ , x− ) ei(p2 x +p2 x ) f3 (x+ , x− ) where f1 and f3 denote eigenmodes of the charged Klein-Gordon equation, and f2 is an eigenmode of the neutral Klein-Gordon equation, each of which with ﬁxed angular momentum ji . Considering higher excited modes such as the graviton would introduce extra factors of covariant derivatives α± in (3.7). In order to evaluate these overlaps, it is convenient to use a different representation and diagonalize half of the covariant derivative operators, e.g. ˜ + = iν∂α˜ − α− = iν∂α+ , α

(3.8)

acting on functions of the variables α+ , α ˜ − taking values in the quadrant R × R˜. Onshell wave functions are now powers of their arguments, ˜2 M2 − 1 − M −1 ˜ 2iν 2 ˜ − ) = Nin α+ 2iν 2 ˜ α f (α+ , α

(3.9)

The notation Nin for the normalization factor anticipates the fact that this representation ˜+ is appropriate to describe an in state. The choice of the signs and ˜ of α− and α distinguishes between short strings ( ˜

= 1) and long strings ( ˜

= −1). Of course, the oscillator representation (3.8) can be related to the real-space representation via the intertwiner + − + − + α+ dα− Φin ˜−) (3.10) f (x , x ) = d˜ ν,α ˜ + ,α− (x , x )f (α , α where the kernel is given by + − Φin ν,α+ ,α ˜ − (x , x ) = exp

iνx+ x− i − iα+ x− − i˜ α− x+ + α+ α ˜− 2 ν

(3.11)

This kernel may be viewed as the wave function for an off-shell winding state with “mo˜ − . Equivalently, one may diagonalize the complementary set of operamenta” α+ and α tors, ˜ − = −iν∂α˜ + (3.12) α+ = −iν∂α− , α leading to on-shell wave functions ˜2 1 − M 2 − 1 + − M 2iν − 2 ˜ f (α− , α ˜ + ) = Nout α− 2iν 2 ˜ α

Those are related to the real-space representation by the kernel iνx+ x− i + − out + − + − − + − i˜ α x − iα x − α ˜ α Φν,α˜ + ,α− (x , x ) = exp − 2 ν

(3.13)

(3.14)

Replacing f1∗ (x+ , x− ) and f3 (x+ , x− ) by their expression in terms of the out and in wave functions (3.9), (3.13) respectively, renders the x± Gaussian (albeit with a nonpositive deﬁnite quadratic form). The remaining α± , α ˜ ± integrals can now be computed

CLOSED STRINGS IN MISNER SPACE

195

in terms of hypergeometric functions. Including the form factor from the excited modes, we obtain, for the 3-point amplitude, + − − + gs + −˜ ∆(ν) δ ji δ p⊥ p 1| : ei(p2 X +p2 X ) : |3 = exp −p i 2 2 2ν + − + µ−1 − µ˜−1 p p p + p− ˜ µ −p2 −p2 U λ, µ, i 2 2 U λ, ˜, i 2 2 (3.15) ν ν ˜ where the non-locality parameter ∆(ν) includes the contribution of the quasi-zero-mode, ˜ ∆(ν) = ψ(iν) + ψ(1 − iν) − 2ψ(1)

(3.16)

The parameters of the Tricomi conﬂuent hypergeometric functions U appearing in (3.15) are given by ˜2 ˜ = 1 + M3 λ 2 2iν ˜2 −M ˜2 M 1 µ ˜ =1+i 3 . 2iν

1 M32 + 2 2iν M 2 − M12 µ=1+ 3 2iν λ=

(3.17)

The amplitude is ﬁnite, and it is proportional to the overlap of the zero-mode wavefunctions, up to the smearing due to the form factor of the untwisted string in the background of the twisted string. Similar expressions can be obtained for 3-point functions in superstring theory involving an untwisted massless state. Four-point amplitudes. The same techniques allow to compute 4-point amplitudes, which now include an integral over the location of the 4-th vertex, as well as on the relative boost parameter v between the two untwisted vertices. The complete expression can be found in [6] and is somewhat abstruse, however it is useful to consider the factorization limit z → 0 where T (3), T (4) (resp. T (1), T (2)) come together: 1|T (2)T (3)|4 →

∞

gs2

i(j3 −j1 )v

dv e −∞

δ−j1 +j2 +j3 +j4 δ − p1 +

3

pi

i=1

− + − ˜ dzd¯ z |z|2p3 ·p4 +p3 ·p3 −2 exp − p+ 2 p2 + p3 p3 ∆(ν) 1

2

1

˜2

−λ −λ µ−λ−1 µ ˜ −λ−1 − 2 M1 − 2 − 2 M1 − 2 (−1)µ+˜µ (p+ (p− (p+ (p− z z¯ 2) 2) 3) 3) ˜

˜

iν

iν

(3.18)

The amplitude diverges whenever j3 = j1 due to the propagation of winding strings with vanishing boost momentum in the intermediate channel. This result closely parallels the discussion in Ref. [32], where tree-level scattering amplitudes of four untwisted states where found to diverge, due to large graviton exchange near the singularity. 3.3. MORE THAN TWO TWISTED STRINGS Scattering amplitudes involving three or more twisted states can be obtained by mapping to an analogous problem which is now very well understood: the Wess-Zumino-Witten

196

BRUNO DURIN AND BORIS PIOLINE

model of a four-dimensional Neveu-Schwarz plane wave [37, 38, 39], with metric 1 ˜ 2 ds2 = −2dudv + dζdζ˜ − ζ ζdu , 4

H = dudxd¯ x

(3.19)

where ζ = x1 + ix2 is the complex coordinate in the plane. In the light-cone gauge u = pτ , it is well known that the transverse coordinate X has the mode expansion of a complex scalar ﬁeld twisted by a real, non rational angle proportional to the light-cone momentum p [39]. In fact, there exists a free-ﬁeld representation where the vertex operator of a physical state with non-zero p is just the product of a plane wave along the (u, v) light cone coordinates, times a twist ﬁeld10 creating a cut z p on the world-sheet. Correlation functions of physical states have been computed using standard WZW techniques [41, 42], and, by removing the plane wave contribution, it is then possible to extract the correlator of twist ﬁelds with arbitrary angle. Referring the reader to [6] for more details, we simply quote the result for the three twist amplitude: in real-space representation (1.15), the amplitude (hence, the OPE coefﬁcient of 3 twist ﬁelds) is given by the overlap

± + − − exp (x+ dx± 1 dx2 1 − x2 )(x1 − x2 )Ξ(ν1 , ν2 ) ± ν1 x± ± ± ∗ ± 1 + ν2 x2 (3.20) [f1 (x1 )f2 (x2 )] f3 x3 − ν1 + ν2 where the characteristic size of the kernel is given by the ratio Ξ(ν1 , ν2 ) = −i

1− 1+

γ(iν3 ) iν3 ν1 ν2 γ(iν1 )γ(iν2 ) γ(iν3 ) iν3 ν1 ν2 γ(iν1 )γ(iν2 )

(3.21)

with γ(p) ≡ Γ(p)/Γ(1 − p). As νi → 0, Ξ(ν1 , ν2 ) ∼ 1/(2ζ(3) ν32 ) so that the interaction becomes local, as√expected for ﬂat space vertex operators. For larger ν however, the nonlocality scale 1/ Ξ diverges when ν1 ν2 γ(iν1 )γ(iν2 ) = iν3 γ(iν3 ). The origin of this divergence is not well understood at present. 3.4. TOWARD CLASSICAL BACK-REACTION While computing the back-reaction from the quantum production of particles and strings remains untractable with the present techniques, the results above give us a handle on a related problem, namely the linear response of closed string ﬁelds to a classical (coherent) condensate of winding strings. Indeed, consider deforming Misner space away from the orbifold point, by adding to the free worldsheet action a condensate of marginal twist operators: ¯ − + λ−w V+w + λ+w V−w (3.22) Sλ = d2 σ ∂X + ∂X While this deformation is marginal at leading order, it implies a one-point function for untwisted ﬁelds (3.23) eikX λ ∼ λw λ−w w|eikX | − w , 10

For integer p, new “spectrally ﬂowed” states appear describing long strings [40].

CLOSED STRINGS IN MISNER SPACE

197

which needs to be cancelled by deformating S at order λ2 by an untwisted ﬁeld: this is the untwisted ﬁeld classically sourced by the winding string with vertex operator V±w . In addition, the same winding string also sources twisted states whose winding number is a multiple of w: (3.24) V−2w λ ∼ λw λw w|V−2w |w , The 3-point functions in (3.23), (3.24) are precisely the amplitudes which have been computed the two previous sections. It is thus possible to extract the corrections to the metric and other string ﬁelds to leading order in the deformation parameter λw . In the Euclidean orbifold case, such a procedure allows to resolve a conical ALE singularity into a smooth Eguchi Hanson gravitational instanton. Whether the same procedure allows to resolve the divergences of the Lorentzian orbifold remains an intriguing open question. 4. Discussion In this lecture, we have taken a tour of the classical aspects of the propagation of closed strings in a toy model of a cosmological singularity: Misner space, a.k.a. the Lorentzian orbifold R1,1 /Z. Our emphasis has been particularly on twisted sectors, which play such an important rˆole in resolving the conical singularities of Euclidean orbifolds. In particular, we have obtained a semi-classical understanding of the pair production of winding strings, as a tunneling effect in the Rindler regions, in close analogy to Schwinger pair creation in an electric ﬁeld. Despite the fact that the one-loop amplitude remains real, indicating no overall particle production between inﬁnite past and inﬁnite future, it is clear that abundant production of particles and strings takes place near the singularity. While tree-level scattering amplitudes exhibit severe divergences due to the inﬁnite blue-shift near the singularity, it is quite conceivable that the back-reaction from the cosmological production of particles and winding strings may lead to a smooth cosmology, interpolating between the collapsing and expanding phases. Indeed, winding strings behave much like a two-dimensional positive cosmological constant, and may thus lead to a transient inﬂation preventing the singularity to occur. Unfortunately, incorporating back-reaction from quantum production lies outside the scope of current perturbative string technology at present. A second quantized deﬁnition of string theory would seem to be a prerequisite to even formulate this question, however, unlike the open string case, a ﬁeld theory of off-shell closed strings has remained elusive, and may even be excluded on general grounds. A generalization of the usual ﬁrst quantized approach allowing for non-local deformations of the worldsheet [43] may in principle incorporate emission of correlated pairs of particles, however do not seem very tractable at present. Instead, the most practical approach seems to consider classical deformations by twisted ﬁelds away from the orbifold point. In contrast to the problem of quantum backreaction, this may be treated in conformal perturbation theory, and we have taken some steps in this direction. It remains to see whether Misner space is a good approximation to the resulting space. More importantly, Misner space appears to be a very ﬁnely tuned example of the space-like singularities which are generically expected to occur in classical Einstein gravity: as shown long ago by Belinsky, Khalatnikov and Lifsh*tz, and independently by Mis-

198

BRUNO DURIN AND BORIS PIOLINE

ner himself (see e.g. [44] for a recent review), the generic approach to a cosmological singularity consists of a chaotic sequence of “Kasner” epochs (of which Milne/Misner space is a special example with zero curvature) and curvature-induced bounces, occuring heterogeneously through space. An outstanding question is therefore to understand string theory in Misner (Mixmaster) space. Acknowledgements Both authors are grateful to M. Berkooz, D. Reichman and M. Rozali for a very enjoyable collaboration, and to B. Craps, G. d’Appolonio, E. Kiritsis, and G. Moore for useful discussions. B. D. would like to thank the organizers of the Carg`ese 2004 ASI for giving him the opportunity to present part of this work in the Gong Show. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

11.

12. 13. 14. 15. 16.

S. Lem, “The Seventh Voyage”, in The Star Diaries, Varsaw 1971, english translation New York, 1976. A. Borde and A. Vilenkin, “Eternal inﬂation and the initial singularity,” Phys. Rev. Lett. 72 (1994) 3305 [arXiv:gr-qc/9312022]. C. W. Misner, in Relativity Theory and Astrophysics I: Relativity and Cosmology, edited by J. Ehlers, Lectures in Applied Mathematics, Vol. 8 (American Mathematical Society, Providence, 1967), p. 160. M. Berkooz, and B. Pioline, “Strings in an electric ﬁeld, and the Milne universe,” JCAP 0311 (2003) 007 [arXiv:hep-th/0307280]. M. Berkooz, B. Pioline and M. Rozali, “Closed strings in Misner space: Cosmological production of winding strings,” JCAP 07 (2004) 003 [arXiv:hep-th/0405126]. M. Berkooz, B. Durin, B. Pioline and D. Reichmann, “Closed strings in Misner space: Stringy fuzziness with a twist,” arXiv:hep-th/0407216. G. T. Horowitz and A. R. Steif, “Singular String Solutions With Nonsingular Initial Data,” Phys. Lett. B 258, 91 (1991). J. Khoury, B. A. Ovrut, N. Seiberg, P. J. Steinhardt and N. Turok, “From big crunch to big bang,” Phys. Rev. D 65 (2002) 086007 [arXiv:hep-th/0108187]. N. A. Nekrasov, “Milne universe, tachyons, and quantum group,” arXiv:hep-th/0203112. V. Balasubramanian, S. F. Hassan, E. Keski-Vakkuri and A. Naqvi, “A space-time orbifold: A toy model for a cosmological singularity,” Phys. Rev. D 67 (2003) 026003 [arXiv:hepth/0202187]; R. Biswas, E. Keski-Vakkuri, R. G. Leigh, S. Nowling and E. Sharpe, “The taming of closed time-like curves,” JHEP 0401 (2004) 064 [arXiv:hep-th/0304241]. I. Antoniadis, C. Bachas, J. R. Ellis and D. V. Nanopoulos, “Cosmological String Theories And Discrete Inﬂation,” Phys. Lett. B 211 (1988) 393; I. Antoniadis, C. Bachas, J. R. Ellis and D. V. Nanopoulos, “An Expanding Universe In String Theory,” Nucl. Phys. B 328 (1989) 117; I. Antoniadis, C. Bachas, J. R. Ellis and D. V. Nanopoulos, “Comments On Cosmological String Solutions,” Phys. Lett. B 257 (1991) 278. C. R. Nappi and E. Witten, “A Closed, expanding universe in string theory,” Phys. Lett. B 293, 309 (1992) [arXiv:hep-th/9206078]. C. Kounnas and D. Lust, “Cosmological string backgrounds from gauged WZW models,” Phys. Lett. B 289 (1992) 56 [arXiv:hep-th/9205046]. E. Kiritsis and C. Kounnas, “Dynamical topology change in string theory,” Phys. Lett. B 331 (1994) 51 [arXiv:hep-th/9404092]. S. Elitzur, A. Giveon, D. Kutasov and E. Rabinovici, “From big bang to big crunch and beyond,” JHEP 0206, 017 (2002) [arXiv:hep-th/0204189]; S. Elitzur, A. Giveon and E. Rabinovici, “Removing singularities,” JHEP 0301, 017 (2003) [arXiv:hep-th/0212242]. L. Cornalba and M. S. Costa, “A New Cosmological Scenario in String Theory,” Phys. Rev.

CLOSED STRINGS IN MISNER SPACE

17.

18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39.

199

D 66, 066001 (2002) [arXiv:hep-th/0203031]; L. Cornalba, M. S. Costa and C. Kounnas, “A resolution of the cosmological singularity with orientifolds,” Nucl. Phys. B 637, 378 (2002) [arXiv:hep-th/0204261]; L. Cornalba and M. S. Costa, “On the classical stability of orientifold cosmologies,” Class. Quant. Grav. 20 (2003) 3969 [arXiv:hep-th/0302137]; B. Craps, D. Kutasov and G. Rajesh, “String propagation in the presence of cosmological singularities,” JHEP 0206, 053 (2002) [arXiv:hep-th/0205101]; B. Craps and B. A. Ovrut, “Global ﬂuctuation spectra in big crunch / big bang string vacua,” Phys. Rev. D 69 (2004) 066001 [arXiv:hep-th/0308057]. E. Dudas, J. Mourad and C. Timirgaziu, “Time and space dependent backgrounds from nonsupersymmetric strings,” Nucl. Phys. B 660, 3 (2003) [arXiv:hep-th/0209176]. L. Cornalba and M. S. Costa, “Time-dependent orbifolds and string cosmology,” Fortsch. Phys. 52, 145 (2004) [arXiv:hep-th/0310099]. C. V. Johnson and H. G. Svendsen, “An exact string theory model of closed time-like curves and cosmological singularities,” arXiv:hep-th/0405141. N. Toumbas and J. Troost, “A time-dependent brane in a cosmological background,” JHEP 0411 (2004) 032 [arXiv:hep-th/0410007]. W. A. Hisco*ck and D. A. Konkowski, “Quantum Vacuum Energy In Taub - Nut (NewmanUnti-Tamburino) Type Cosmologies,” Phys. Rev. D 26 (1982) 1225. A. H. Taub, “Empty Space-Times Admitting A Three Parameter Group Of Motions,” Annals Math. 53, 472 (1951); E. Newman, L. Tamburino and T. Unti, “Empty Space Generalization Of The Schwarzschild Metric,” J. Math. Phys. 4 (1963) 915. J. G. Russo, “Cosmological string models from Milne spaces and SL(2,Z) orbifold,” arXiv:hep-th/0305032. J. R. I. Gott, “Closed Timelike Curves Produced By Pairs Of Moving Cosmic Strings: Exact Solutions,” Phys. Rev. Lett. 66, 1126 (1991); J. D. Grant, “Cosmic strings and chronology protection,” Phys. Rev. D 47 (1993) 2388 [arXiv:hep-th/9209102]. S. W. Hawking, “The Chronology protection conjecture,” Phys. Rev. D 46, 603 (1992). D. Kutasov, J. Marklof and G. W. Moore, “Melvin Models and Diophantine Approximation,” arXiv:hep-th/0407150. C. Gabriel and P. Spindel, “Quantum charged ﬁelds in Rindler space,” Annals Phys. 284 (2000) 263 [arXiv:gr-qc/9912016]. N. Turok, M. Perry and P. J. Steinhardt, “M theory model of a big crunch / big bang transition,” Phys. Rev. D 70 (2004) 106004 [arXiv:hep-th/0408083]. C. Bachas and M. Porrati, “Pair Creation Of Open Strings In An Electric Field,” Phys. Lett. B 296, 77 (1992) [arXiv:hep-th/9209032]. J. M. Maldacena, H. Ooguri and J. Son, “Strings in AdS(3) and the SL(2,R) WZW model. II: Euclidean black hole,” J. Math. Phys. 42, 2961 (2001) [arXiv:hep-th/0005183]. M. Berkooz, B. Craps, D. Kutasov and G. Rajesh, “Comments on cosmological singularities in string theory,” arXiv:hep-th/0212215. D. J. Gross and P. F. Mende, “The High-Energy Behavior Of String Scattering Amplitudes,” Phys. Lett. B 197, 129 (1987). H. Liu, G. Moore and N. Seiberg, “Strings in a time-dependent orbifold,” JHEP 0206, 045 (2002) [arXiv:hep-th/0204168]; H. Liu, G. Moore and N. Seiberg, “Strings in time-dependent orbifolds,” JHEP 0210, 031 (2002) [arXiv:hep-th/0206182]. D. Amati, M. Ciafaloni and G. Veneziano, “Classical And Quantum Gravity Effects From Planckian Energy Superstring Collisions,” Int. J. Mod. Phys. A 3 (1988) 1615. G. T. Horowitz and J. Polchinski, “Instability of spacelike and null orbifold singularities,” Phys. Rev. D 66, 103512 (2002) [arXiv:hep-th/0206228]. C. R. Nappi and E. Witten, “A WZW model based on a nonsemisimple group,” Phys. Rev. Lett. 71, 3751 (1993) [arXiv:hep-th/9310112]; D. I. Olive, E. Rabinovici and A. Schwimmer, “A Class of string backgrounds as a semiclassical limit of WZW models,” Phys. Lett. B 321 (1994) 361 [arXiv:hep-th/9311081]. E. Kiritsis and C. Kounnas, “String Propagation In Gravitational Wave Backgrounds,” Phys. Lett. B 320 (1994) 264 [Addendum-ibid. B 325 (1994) 536] [arXiv:hep-th/9310202]; E. Kiritsis, C. Kounnas and D. Lust, “Superstring gravitational wave backgrounds with space-time

200

40. 41. 42. 43.

44.

BRUNO DURIN AND BORIS PIOLINE supersymmetry,” Phys. Lett. B 331, 321 (1994) [arXiv:hep-th/9404114]. E. Kiritsis and B. Pioline, “Strings in hom*ogeneous gravitational waves and null holography,” JHEP 0208, 048 (2002) [arXiv:hep-th/0204004]. G. D’Appollonio and E. Kiritsis, “String interactions in gravitational wave backgrounds,” arXiv:hep-th/0305081. Y. K. Cheung, L. Freidel and K. Savvidy, “Strings in gravimagnetic ﬁelds,” JHEP 0402 (2004) 054 [arXiv:hep-th/0309005]. O. Aharony, M. Berkooz and E. Silverstein, “Multiple-trace operators and non-local string theories,” JHEP 0108 (2001) 006 [arXiv:hep-th/0105309]; M. Berkooz, A. Sever and A. Shomer, “Double-trace deformations, boundary conditions and spacetime singularities,” JHEP 0205 (2002) 034 [arXiv:hep-th/0112264]; E. Witten, “Multi-trace operators, boundary conditions, and AdS/CFT correspondence,” arXiv:hep-th/0112258. T. Damour, M. Henneaux and H. Nicolai, “Cosmological billiards,” Class. Quant. Grav. 20, R145 (2003) [arXiv:hep-th/0212256].

PHENOMENOLOGICAL GUIDE TO PHYSICS BEYOND THE STANDARD MODEL

STEFAN POKORSKI Institute for Theoretical Physics, University of Warsaw and Theory Division, CERN, Geneve

Abstract. Various aspects of physics beyond the Standard Model are discussed from the perspective of the fantastic phenomenological success of the Standard Model, its simplicity and predictive power.

1. Introduction The Standard Model is a successful theory of interactions of quarks and leptons at energies up to about hundred GeV. Despite that success it is widely expected that there is physics beyond the Standard Model, with new characteristic mass scale(s), perhaps up to, ultimately, a string scale. The expectation is motivated by several fundamental questions that remain unanswered by the Standard Model. The most pressing one is better understanding of the mechanism of the electroweak symmetry breaking. The origin of ﬂavour and of the pattern of fermion masses and of CP violation also remain beyond its scope. Moreover, we know now that the physics of the Standard Model cannot explain the baryon asymmetry in the Universe. And on the top of all that come two recent strong experimental hints for physics beyond the Standard Model, that is very small neutrino masses and the presence of dark matter in the Universe. The list can be continued by including dark energy and inﬂation. The Standard Model does not explain the scale of the electroweak symmetry breaking. It is a free parameter of the theory, taken from experiment. Moreover, once we accept the point of view that the Standard Model is only an effective “low energy” theory which is somehow cut-off at a mass scale M , and if M MW,Z , the electroweak symmetry breaking mechanism based on use of an elementary Higgs ﬁeld is unstable against quantum corrections (this is the so-called hierarchy problem). Many different extensions of the Standard Model have been proposed to avoid the hierarchy problem and, more ambitiously, to calculate the scale of the electroweak breaking in terms of, hopefully, more fundamental parameters. Some extensions give the complete 201 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 201–228. © 2005 Springer. Printed in the Netherlands.

202

STEFAN POKORSKI

Standard Model, with one Higgs doublet, as their low energy approximation in the sense of the Appelquist - Carazzone decoupling and in some others the mechanism of the electroweak symmetry breaking cannot be decoupled from the bigger theory. The general idea is that the bigger theory has some characteristic mass scale M only order of magnitude bigger than MW,Z , which plays the role of a cut-off to the electroweak sector. All those extensions of the Standard Model have distinct experimental signatures. The experiments at the LHC will, hopefully, shed more light on the mechanism of the electroweak symmetry breaking and will support one of those (or still another one ?) directions. One approach is based on low energy supersymmetry. The scale M is identiﬁed with the mass scale of supersymmetric partners of the Standard Model particles. Supersymmetry is distinct in several very important points from all other proposed solutions to the hierarchy problem. First of all, it provides a general theoretical framework which allows to address many physical questions. Supersymmetric models, like the Minimal Supersymmetric Standard Model or its simple extensions satisfy a very important criterion of “perturbative calculability”. In particular, they are easily consistent with the precision electroweak data. The Standard Model is their low energy approximation in the sense of the Appelquist - Carazzone decoupling, so most of the successful structure of the Standard Model is built into supersymmetric models. Unfortunately, there are also some troublesome exceptions: there are new potential sources of Flavour Changing Neutral Current (FCNC) transitions and of CP violation, and baryon and lepton numbers are not automatically conserved by the renormalizable couplings. But even those problems can at least be discussed in a concrete way. The quadratically divergent quantum corrections to the Higgs mass parameter (the origin of the hierarchy problem in the Standard Model) are absent in any order of perturbation theory. Therefore, the cut-off to a supersymmetric theory can be as high as the Planck scale and “small” scale of the electroweak breaking is still natural. But the hierarchy problem of the electroweak scale is solved at the price of a new hierarchy problem of the soft supersymmetry braking scale versus the Planck (string) scale. Spontaneous supersymmetry breaking and its transmission to the visible sector is a difﬁcult problem and a fully satisfactory mechanism has not yet been found. Again on the positive side, supersymmetry is not only consistent with Grand Uniﬁcation of elementary forces but, in fact, makes it very successful. And, ﬁnally, supersymmetry is needed for string theory. All other extensions of the electroweak theory proposed as solutions to the hierarchy problem rely on an onset of some kind of strong dynamics at energy scales not much higher than the electroweak scale. In some of them, like Higgless models with dynamical electroweak symmetry breaking or strong gravity in large extra dimensions, the strong dynamics is simply a cut-off directly to the electroweak sector and appears already at O(1 TeV). In models with the Higgs boson as a pseudo-Goldstone boson (e.g. Little Higgs models) and models with gauge ﬁelds present in extra dimensions the cut-off scale M is identiﬁed with the characteristic scale of new perturbative physics, e.g. with the scale of breaking of some global symmetry or with the radii of extra dimensions. However, since those models are non-renormalizable and, moreover, in the bigger theory the quadratic divergences to the scalar mass parameter are absent typically only at one loop level, new physics itself has to be cut-off by some unknown strong dynamics at a scale one or two orders of magnitude higher than the M . Generally speaking, there is no Appelquist Carazzone decoupling of new physics and the precision tests of such a version of the

PHENOMENOLOGICAL GUIDE TO PHYSICS

203

electroweak theory are not possible at the same level of accuracy as in the renormalizable Standard Model. It is clear that models with early onset of strong dynamics cannot be easily, if at all, reconciled with Grand Uniﬁcation. Also, they are very strongly constrained by precision electroweak data. There have been constructed models that work but simple models are usually ruled out. Moreover, various aspects of ﬂavour physics are often very obscure. Spontaneous symmetry breaking in the condensed matter physics and in QCD is due to some collective effects. In supersymmetric models, such effects are presumably responsible for spontaneous breaking of supersymmetry and, in consequence, for the generation of soft mass terms. However, the electroweak symmetry breaking is driven by perturbative quantum corrections, generated by the large top quark Yukawa coupling, to the scalar potential. In the Little Higgs models, the Higgs boson is a Goldstone boson of a bigger spontaneously broken global symmetry group. The Higgs potential needed for the electroweak symmetry breaking is also given by quantum corrections, with important contribution from the top quark Yukawa coupling. Thus, one thing many models have in common is that the electroweak symmetry is broken by perturbative quantum effects and linked to the large mass of the top quark. At present, all extensions of the Standard Model remain speculative and none is fully satisfactory. Remembering the simplicity, economy and success of the Standard Model, one may wonder if in our search for its extensions shouldn’t the Hipocrates principle Primum non nocere play more important role than it does. Indeed, various new ideas offer surprisingly low ratios of beneﬁts to losses. It is, therefore, appropriate to begin by reviewing the basic structure of the Standard Model that underlies its success. It is likely that it gives us important hints for the physics beyond. 2. The Standard Model 2.1. BASIC STRUCTURE The underlying principles of the electroweak theory are: 1. local SU (2)L × U (1)Y gauge symmetry and electroweak uniﬁcation 2. spontaneous breaking of SU (2)L × U (1)Y gauge symmetry to U (1)EM , by the Higgs mechanism with one Higgs doublet 3. matter content (chiral fermions) 4. renormalizability Massless chiral fermions are the fundamental objects of matter: left-handed, with helicity λ = −1/2, and right-handed, with helicity λ = 1/2. It is so because parity and charge conjugation are not the symmetries of our world. The left-handed fermions carry different weak charges from the right-handed fermions. Chiral fermion ﬁelds are twocomponent (Weyl) spinors (see e.g. [1]): SU (2)L doublets q1 ≡

u d

q2 ≡

c s

q3 ≡

t b

(2.1)

204

STEFAN POKORSKI

l1 ≡

νe e

l2 ≡

νµ µ

l3 ≡

ντ τ

(2.2)

with the electric charge and the hypercharge (Y = Q − T3 ) assigned as below u

d

ν

e

Q

2/3

-1/3

-1

Y

1/6

1/6

-1/2

-1/2

These are left-handed chiral ﬁelds in the representation (0, 1/2) of the SL(2, C), each describing two massless degrees of freedom: a particle with the helicity λ = −1/2 and its antiparticle with λ = +1/2. (The chiral ﬁelds can also be written as four-component spinors (see e.g. [1]) but in the following we shall be using the Weyl notation). Right-handed ﬁelds [(1/2, 0) of SL(2, C)] in the same representations of SU (2)L × U (1)Y as the left-handed ﬁelds (2.1) and (2.2) do not exist in Nature. Instead, we have SU (2)L singlets uR , cR , tR dR , sR , bR eR , µR , τR in (1, +2/3), (1, −1/3) and (1, −1) of SU (2)L × U (1)Y , respectively. These are righthanded chiral ﬁelds in the (1/2, 0) representation of the group SL(2, C). For constructing a Lorentz invariant Lagrangian, it is more convenient to take as fundamental ﬁelds only the left-handed chiral ﬁelds. Thus, we introduce left-handed chiral ﬁelds, e.g uc , cc , tc

(2.3)

in (1, −2/3) of SU (2)L × U (1)Y , such that u¯c ≡ CP uc (CP )−1 = uR

(2.4)

Indeed, CP transformation results in the simultaneous change of chirality and charges (representation R → R∗ for internal symmetries). Moreover, we see that the electric charge Q = T3 + Y satisﬁes, e.g. Quc = −2/3 = −Qu

(2.5)

c

and the two left-handed ﬁelds u and u become charge conjugate to each other when UEM (1) remains the only unbroken symmetry: Cuc C −1 = u

(2.6)

We note that the matter chiral ﬁelds of the SM do not include a right-handed neutrino ﬁeld νR in (1, 1) of SU (2)L × U (1)Y (such a charge assignment preserves the relation Q = T3 + Y ) or equivalently, a left-handed ﬁeld ν c such that νR = CP ν c (CP )−1

(2.7)

PHENOMENOLOGICAL GUIDE TO PHYSICS

205

but we can supplement the Standard Model with such a particle, if useful. The breaking of+the electroweak symmetry is generated by the potential of the Higgs H doublet H = with the hypercharge Y = +1/2: H0 V = m2 H † H +

λ † 2 (H H) 2

(2.8)

When m2 < 0 is chosen, the Higgs doublet acquires the vacuum expectation value. Indeed, the minimum of the potential is for H † H = −

v2 m2 ≡ λ 2

(2.9)

By SUL (2) rotation we can always redeﬁne the vacuum so that only the VEV of the lower component of the Higgs doublet is non-zero. The SUL (2) × UY (1) symmetry is 3 then brokendown to U (1) which is identiﬁed with UEM (1) with Q = T + Y because 0 (T 3 + Y ) = 0. The parameters m and λ are free parameters of the Standard v Model. Equivalently, the scale of the electroweak symmetry breaking is not predicted by the theory and must be taken from experiment.

2.2. FERMION MASSES Higgs doublets (and only doublets) have SU (2)L × U (1)Y invariant renormalizable couplings to the chiral fermions of the Standard Model. For the charged fermions, we can write down the following Yukawa couplings: LY ukawa = −YlBA Hi∗ liA ecB − YdBA Hi∗ qiA dcB − YuBA ij Hi qiA ucB + hc

(2.10)

where i is the SU (2)L index and A, B are generation indices. We use the fact that the two-dim representation of SU (2) is real and iτ2 H transforms as H ∗ , i.e. as 2∗ (≡ 2) of SU (2). Therefore, (iτ2 Hq) = ij Hi qj is also an invariant of SU (2). After spontaneous breaking of SU (2)L × U (1)Y to U (1)EM by the Higgs boson vacuum expectation value v we obtain the Dirac masses Lmass = −v(YlBA eA ecB + YdBA dA dcB + YuBA uA ucB ) + hc

(2.11)

However, at the level of the full, SU (2)L × U (1)Y invariant theory, there is no renormalizable term that would give neutrino mass. It is so because ν c is absent from the spectrum of the SM. Thus, in the SM, neutrinos are massless. The interactions (2.10,2.11) are written in some ”electroweak” basis deﬁned by eigenvectors of the SU (2)L ×U (1)Y symmetry group. In such a basis, both the fermion masses and the Yukawa couplings are in general non-diagonal in the ﬂavour indices (A, B). However, we can introduce another set of ﬁelds (say, primed ﬁelds) describing physical particles (mass eigenstates). The ﬂavour of the primed ﬁelds is deﬁned in the mass eigenstate

206

STEFAN POKORSKI

basis. The two sets of ﬁelds are related to each other by unitary transformations: u = UL u

d = DL d

uc = uc UR†

† dc = dC DR

e = EL e

(2.12)

† ec = ec ER

which, of course, do not commute with the SU (2)L × U (1)Y gauge transformations and can be performed only after the spontaneous breakdown of the gauge symmetry. In eq.(2.12), the ﬁelds u, d , e denote three-dimensional vectors in the ﬂavour space. The transformations (2.12) diagonalize the mass terms and the Yukawa couplings deﬁned by (2.10). After diagonalization we can combine the chiral ﬁelds into Dirac ﬁelds. The weak currents can be expressed in terms of the physical (mass eigenstates) ﬁelds: u ¯A σ ¯µ (VCKM )AB dB Jµ− = A,B

+

ν¯A σ ¯µ eA

(2.13)

A

where the Cabibbo-Kobayashi-Maskawa matrix VCKM = UL† DL . Note that the lepton current is diagonal in ﬂavour (deﬁned in the charged lepton mass eigenstate basis) because L νB where the massless neutrino ﬁeld can be redeﬁned by the transformation νA = EAB L EAB is the transformation diagonalizing the charged lepton mass matrix (see (2.12)). Thus, for the lepton current, VCKM = E L† E L = 1. It is important to remember that in the SM (with only one Higgs doublet) the Yukawa couplings to the physical Higgs boson (and, in fact, also the couplings to the Z 0 ) and the mass terms are diagonalized by the same unitary rotations. So they are ﬂavour diagonal. The only source of ﬂavour non-conservation resides in VCKM . In particular, not only the global lepton number but also each ﬂavour lepton number is separately conserved. 2.3. APPROXIMATE CUSTODIAL SYMMETRY OF THE STANDARD MODEL AND THE PRECISION ELECTROWEAK DATA The Higgs sector of the SM is invariant under global SO(4) symmetry acting on four real components of the complex doublet. The group SO(4) SU (2)L × SU (2)R and the Higgs doublet can be written as a 2 × 2 matrix Φ + H 0∗ H , (2.14) Φ= H 0 −H − which transforms as (2, 2) of the latter group (whose ﬁrst factor is just the gauged weak isospin group): Φ −→ Φ = UL ΦUR .

(2.15)

PHENOMENOLOGICAL GUIDE TO PHYSICS

207

The vacuum expectation value of the Higgs ﬁeld breaks SU (2)L ×SU (2)R to its diagonal subgroup called “custodial” SU (2) acting on the three would-be Goldstone bosons Ga : H = H+

1 iGa τ a H 0 −→ √ e v 0 v + h0 2

(2.16)

In the rest of the electroweak Lagrangian the SU (2)R subgroup and therefore also the custodial SU (2) symmetry is broken by the Yukawa interactions and by the U (1)Y coupling. However, the custodial symmetry is still seen at the tree level since by the Higgs mechanism it ensures that the gauge bosons W + , W − and W 0 of SU (2)L transform as a custodial triplet. In consequence, the ratio of the strength of charged and neutral current interaction at the tree level is equal to one ρ≡

2 MW =1, θW MZ2

cos2

(2.17)

where cos2 θW = g22 /(g22 + g12 ). This relation is consistent to a very good approximation with experimental data but this is not the end of the success of the SM. Since the rest of the Lagrangian violates the custodial symmetry, there are quantum corrections to the relation ρ = 1. Since the global SU (2)L × SU (2)R symmetry of the Higgs sector ﬁxes the structure of counterterms in the scalar potential, quantum corrections to the relation ρ = 1 must be ﬁnite! In one-loop approximation one gets ∆ρ ≡ ρ − 1 =

M2 3g22 m2t − m2b g 2 11 ln 2h . − 12 2 2 64π MW 64π 3 MW

(2.18)

The ﬁrst term is of the order of 1% and, is in perfect agrrement with the precision electroweak data. Thus, ﬁts to the data give Mh ∼ O(MW ) although logarithmic dependence of ∆ρ on Mh does not allow for precise determination of this mass. We shall discuss later on the importance of the Higgs particle mass for various extensions of the SM. We conclude that the approximate custodial symmetry of the SM is in fantastic agreement with experimental data. Any alternative mechanism of the electroweak symmetry breaking or any extension of the SM must not violate the custodial symmetry of the electroweak interactions. Furthermore, we see that the renormalizable SM with one Higgs doublet, has very strong predictive power which allows for its precision tests at the level of one per mille. 2.4. GIM MECHANISM AND THE SUPPRESSION OF FCNC AND CP VIOLATING TRANSITIONS s)It is well established experimentally that the amplitudes for processes such as e.g. K 0 (d¯ ¯ mixing caused by electroweak interactions, K 0 |Hweak |K ¯ 0 (ds) ¯ 0 , are strongly supK pressed in comparison with the amplitudes for the charged current transitions like n → pe− ν¯e or µ− → e− ν¯e νµ . A good measure of the mixing is the mass difference between ¯ 0 ) and |K 0 = √1 (|K 0 + |K ¯ 0 ) neutral kaon mass eigenstates |KL0 = √12 (|K 0 − |K S 2 −12 (we neglect here even smaller CP violation): ∆MK = 3 × 10 MeV and is suppressed

208

STEFAN POKORSKI

by factor 106 compared to what one could expect for a generic electroweak transition. This fact ﬁnds a very elegant explanation in the Standard Model. The charged current transitions shown in Fig. 1 are present at the tree level whereas it follows from the structure of the theory that the diagrams shown in Fig. 2 are absent. In the diagrams the quark ﬁelds are of course mass eigenstate ﬁelds and the couplings are obtained by rotating from an electroweak basis, in which the theory is formulated, to the mass eigenstate basis, in which the quark ﬂavour is deﬁned. It is obvious that the Z 0 couplings are ﬂavour diu

e−

d W−

νe W−

Figure 1. Charged current transitions in the Standard Model.

s

µ−

d Z0

e−

s d Z0

h0

Figure 2. Neutral current transitions absent at the tree level in the SM.

agonal. More interesting is the absence in the SM of the scalar ﬂavour changing neutral currents. This result follows from the fact that in the SM there is only one Higgs doublet. Because of that, diagonalizing the fermion mass matrices one simultaneously obtains also diagonal Yukawa couplings to the physical Higgs boson. In models with more Higgs doublets, additional discrete symmetry has to be imposed to ensure that only one Higgs doublet couples to the quarks of the same charge. In the minimal supersymmetric model two doublets are needed for supersymmetric theory but (by the holomorphicity of the superpotential) only one Higgs doublet can couple to the same charge quarks. The absence of ﬂavour non-diagonal neutral currents at the tree level is not sufﬁcient to account for the observed suppression of processes like kaon mixing or b → sγ. For example the 1-loop ¯ transitions and the coefﬁcient C ¯ 0 (ds) s)-K diagrams shown in Fig. 3 generate the K 0 (d¯ in the effective Lagrangian sL γµ dL )(¯ sL γµ dL ) Leﬀ = C (¯

(2.19)

describing their contribution (in the limit of external quark momenta small compared to 2 MW ) is generically of order C ∼ Mα2 ∼ α GF . However, in the SM the sum of all W

such contributions is suppressed by a factor ∼ 10−4 due to the so-called (generalized to 3 generations of quarks) Glashow-Illiopoulos-Maiani mechanism. The coefﬁcient C

PHENOMENOLOGICAL GUIDE TO PHYSICS s

d

W±

W± W±

u, c, t

s

s

d

u, c, t

209

u, c, t s

d

W±

u, c, t d

¯ 0 mixing. Figure 3. Leading SM contribution to K 0 -K

generated by diagrams of Fig. 3 (and the diagrams in which one or both W ± are replaced by the unphysical would-be Goldstone bosons G± ) is ﬁnite and has dimension mass−2 . The whole effective Lagrangian can be written as Leﬀ

1 = 2 ×

g √2 2

4

Vis Vid Vjs Vjd

(2.20)

i,j=u,c,t

d4 q [Ψs γµ PL ( q + mqi )γν PL Ψd ][Ψs γ ν PL ( q + mqj )γ µ PL Ψd ] 2 )2 (2π)4 (q 2 − m2qi )(q 2 − m2qj )(q 2 − MW

The top quark contribution is suppressed by the smallness of the product Vts∗ Vtd . The rest contributes to the coefﬁcient C in (2.19) 4 m2qi m2qj g2 1 √ Vis Vid Vjs Vjd 1 + O Auc ∼ 2 2 , M2 MW MW 2 W i,j=u,c ⎧ ⎞⎫ ⎛ ⎨ m2q ⎬ 2 ∼ α GF (Vts Vtd ) + O ⎝ (2.21) Vis Vid Vjs Vjd 2 ⎠ ⎩ MW ⎭ i,j=u,c

Vud + Vcs Vcd = where in the last line, unitarity of the CKM matrix has been used: Vus −Vts Vtd . From this example it is clear that for the empirical pattern of quark masses and mixing angles there is strong suppression of FCNC in the SM. It is much stronger than “naturally” expected. The predictions of the SM for the FCNC transitions are in very good agreement with experimental data. This is also true for CP violation. The only source of CP violation in the SM is the phase of the CKM matrix. As a result, the effects of CP violation in the kaon system, in which they were ﬁrst observed, are proportional to the masses of the light quarks and small CKM mixing angles and hence very small (this is not so for the B-meson systems in which CP violation is probed by present experiments). The strong suppression of the FCNC and CP violating transitions, so nicely consistent with the SM is a big challenge for any of its extension. This is easy to understand on a qualitative basis. Let us suppose that new physics contributes to such transitions at 1-loop

210

STEFAN POKORSKI

level (any contribution at the tree level would be a total disaster!) with the couplings of order of the strong coupling constant αs ≈ 0.12 and with the scale M of the particle masses in the loop. Then 2 α2 α2 αs 2 MW ∆C ∼ s2 = 2 (2.22) M MW α M Thus, for such contributions to be comparable or smaller than the SM one, the new physics scale M has to be higher than 103 MW ∼ 100 TeV. If the new couplings are of the order > 10 TeV. If the scale of new physics extension of the SM is below of α then we get M ∼ these limits, the new physics must somehow control the ﬂavour effects! 2.5. BARYON AND LEPTON NUMBER CONSERVATION The principles (1) - (4) imply global U (1) symmetries of the theory: baryon and lepton number conservation ∆B = ∆L = 0. In fact, for leptons the implication is even stronger, namely Ue (1) × Uµ (1) × Uτ (1) is a global symmetry of the electroweak Lagrangian and the lepton ﬂavour numbers are separately conserved: ∆Le = ∆Lµ = ∆Lτ =0. For quarks, quark mixing explicitly breaks quark ﬂavour U (1)’s and only the total baryon number is conserved. The conservation of the baryon and lepton numbers by the renormalizable couplings of the Standard Model is beautifully consistent with experimental limits on the life time > 1033 years, and on the branching ratios for the lepton ﬂavour viof the proton , τp ∼ < 10−11 . Proton decay and lepton ﬂavour violating olating decays, e.g. BR(µ → eγ) ∼ decays occur, if at all, many orders of magnitude less frequently than generic electroweak processes. Actually, in the Standard Model those conservation laws are violated by chia ’s are SU (2)L gauge currents and ral anomaly. The diagrams shown in Fig. 4 where jµL jµL,B is the baryon or lepton current of the U (1) global symmetries gives (insisting on the conservation of the gauge currents [1]) ∂ µ jµB

∝ (TrB)

3

a ˜ aµν W Wµν

a=1

∂ µ jµL ∝ (TrL)

3

a ˜ aµν W Wµν

(2.23)

a=1 a Only SUL (2) doublets contribute to the traces, so they do not vanish, and Wµν is the SU (2)L ﬁeld strength. Non-perturbative effects give, in general a non-zero condensate a ˜ µν W (topological baryon and lepton number non-conservation) but the effect is toWµν tally negligible at zero temperature. At non-zero temperature, the topological baryon and lepton number non-conservation is enhanced and can play important physical role because the (B − L) current is anomaly free: ∂ µ jµB−L ∝ Tr(B − L) = 0 and the quantum number B −L is conserved. Thus in the presence of some hypothetical perturbative lepton number and CP violation, topological effects may convert leptogenesis into baryogenesis. Incidentally, with the right-handed neutrino included in the spectrum, the diagrams in Fig. 5 do not give any anomaly, neither, and the (B − L) symmetry can, therefore, be gauged.

PHENOMENOLOGICAL GUIDE TO PHYSICS

211

a jνL

jµL,B

a jνL

jµL,B

fL

fL

b jκL

b jκL

Figure 4. Anomalies in the B and L currents in the SM.

B−L jνL

jµB−L

B−L jνL

jµB−L

f B−L jκL

f B−L jκL

Figure 5. [U (1)B−L ]3 anomaly.

3. Hints from the Standard Model for its extensions 3.1. IS THE EFFECTIVE LOW ENERGY ELECTROWEAK THEORY INDEED THE RENORMALIZABLE STANDARD MODEL? In the construction of the SM Lagrangian we have been, so far, guided by its renormalizability. Accepting the fact that the SM is only an effective theory one may wonder, however, how important is its renormalizability. Unitarity and symmetries are certainly more fundamental requirements and indeed e.g. the physics of pions is described by a nonrenormalizable effective low energy theory (non-linear σ-model). It is, therefore, useful to recall the main differences between the two classes of quantum ﬁeld theories. In a renormalizable theory its cut-off can be taken to inﬁnity and the whole UV sensitivity is hidden in a ﬁnite number of free parameters, the same at any order of perturbation expansion. Calculations with arbitrary precision are, therefore, possible with a ﬁxed number of parameters whose values can be determined from the experimental data. If some theory gives as its low energy approximation a renormalizable theory, then according to the Appelquist-Carazzone decoupling theorem, the effects of heavy degrees of freedom characterized by a mass scale M show up only as corrections in the form of higher dimensional operators allowed by the symmetries of the renormalizable theory: 1 (3.1) O4+n Leﬀ = Lrenormalizable + O Mn This is a window to new physics (if such corrections are needed by experiment) even if we do not know the theory at the scale M .

212

STEFAN POKORSKI

It is worth putting the view at the Standard Model as an effective low energy theory into the better known perspective. We know now that Quantum Electrodynamics (QED) is a renormalizable theory and at the same time it is the low energy approximation to the electroweak theory. Its renormalizability means calculability with arbitrary precision. But it is only an effective theory so we know that its predictions disagree with experiment at the level ∼ O(E/Mw ), where the energy E is the characteristic energy for a given process. For example, let us have a look at the lepton magnetic moment. It gets contributions from the diagrams depicted in Fig. 6. Thus, for the nonrelativistic effective

γ l l

γ

γ

l

l

W

l γ

l

l

W ν

l

Figure 6. One loop contributions to the anomalous lepton magnetic moment in the SM.

interaction with the magnetic ﬁeld we get Heﬀ =

m2 e α + O α 2l σ · B(1 + + . . .) 2ml 2π MW

(3.2)

where the role of the energy scale is played by the lepton mass ml . The ”weak” correction is calculable in the full electroweak theory, but at the level of QED as an effective theory it has to be added as a new non-renormalizable (but U (1)EM invariant) interaction

Figure 7. Effective photon-lepton vertex.

Leﬀ =

ml ¯ ψσµν ψF µν M2

(dim 5)

(3.3)

This would have been a way to discover weak interactions (and to measure the weak scale) in purely electromagnetic processes: we extend QED to a non-renormalizable theory by adding higher dimension operators and look for their experimental manifestation in purely electromagnetic processes once the experimental precision is high enough. Luckily enough for us, effective QED may also contain other than (3.3) non-renormalizable

PHENOMENOLOGICAL GUIDE TO PHYSICS

213

corrections, U (1)EM invariant but violating the conservation of quantum numbers that are accidentally conserved in QED, for instance ﬂavour. Such corrections manifest themselves as different type of interactions - weak interactions - and were easy to discover experimentally. Similarly, among many possible non-renormalizable corrections to the SM which respect the SU (2)L × U (1)Y gauge symmetry there are such that violate e.g. the lepton and/or baryon number conservation or give Majorana masses to neutrinos. We shall discuss them in the following. In a non-renormalizable theory one either has to keep explicit logarithmic cut-off de1 Λ pendence ( 16π 2 ln µ where µ is some low energy scale) or the number of counterterms (i.e. the number of free parameters of the theory) must increase at each order of perturbation expansion. The value of the cut-off Λ is dictated by the consistency between one-loop calculations and the contribution of the higher dimensional operators. Typically, the theory becomes strongly interacting above the cut-off scale. A physically important non-renormalizable effective theory is the theory of pions [2, 1]. The pions are pseudo-Goldstone bosons of the (approximate) global chiral symmetry SU (2)L ×SU (2)R of strong interactions which is spontaneously broken down to SU (2)V of isospin. The physics of the light degrees of freedom (pions) is described by a nonlinear σ-model. The chiral symmetry SU (2)L × SU (2)R , non-linearly realized on the pion ﬁelds, requires non-renormalizable interaction. The lowest dimension one is Lpions = fπ2 Tr(∂µ U † ∂ µ U )

(3.4)

where the ﬁelds U = eiπ

a

τ a /fπ

(3.5)

transform under SU (2)L × SU (2)R linearly: U → VL U VR† . The constanf fπ is the pion decay constant. The chiral symmetry cannot be reconciled with an effective renormalizable theory of pions and the Appelquist-Carazzone decoupling does not work. The important question is: is the true low energy approximation to the more fundamental theory which explains the mechanism of electroweak symmetry breaking the renormalizable SM (like in supersymmetric extensions of the SM) or non-renormalizable electroweak theory (like in higgsless models with dynamical electroweak symmetry breaking and in models with the Higgs boson as a pseudo-Goldstone boson of some spontaneously broken bigger global symmetry)? The predictive power and the phenomenological success of the SM suggests the ﬁrst case. On the other hand, one may argue that the second option would more resemble spontaneous symmetry breaking in the condensed matter physics and in strong interactions. 3.2. MATTER CONTENT AND DEEPER UNIFICATION? There are two striking aspects of the matter spectrum in the Standard Model. One is the chiral anomaly cancellation [2, 1], which is necessary for a unitary (and renormalizable) theory, and occurs thanks to certain conspiracy between quarks and leptons suggesting a deeper link between them. The potential source of chiral anomalies in the Standard Model are the triangle diagrams like the ones shown in Figs. 4 in which now the external lines correspond to all possible triplets of currents coupled to the three types of gauge ﬁelds

214

STEFAN POKORSKI

in the electroweak theory: U (1)Y gauge ﬁeld Bµ , SU (2)L gauge ﬁelds Wµa (a = 1, 2, 3) and/or SU (3)C gauge ﬁelds Aaµ (a = 1, . . . , 8) and internal fermion lines correspond to all chiral fermions in the theory. Most of the anomaly coefﬁcients vanish due to the group structure. The most interesting ones are the anomalies with one U (1)Y current and two SU (2)L currents and the one with three U (1)Y currents. They vanish (a necessary condition for consistency of the electroweak theory) provided (3.6) Qi = 0 , (where Qi ’s are the electric charges of the ﬁelds) separately for doublets and singlets of SU (2)L . Incidently, the same condition is sufﬁcient for vanishing of the mixed U (1)Y gravitational anomaly given by the diagrams like those shown in Figs. 4 but now with two currents corresponding to the energy-momentum tensors and the third one to the U (1)Y current. The condition (3.6) is satisﬁed in the SM because quark and lepton contributions cancel each other. The second striking feature of the matter spectrum in the Standard Model is that it ﬁts into simple representations of the SU (5) and SO(10) groups [3]. Indeed, we have, for SU (5) νe ∗ , dcL 5 = − e L u 10 = , ucL , ecL dL 1 = νLc (if the right − handed neutrino is added to the spectrum) and, for SO(10), 16 = 5∗ + 10 + 1 The assignment of fermions to the SU (5) representations ﬁxes the normalization of the U (1)Y generator: + 5 12 11 L , (3.7) Q = T3 + Y = L + 3 where Lij are the SU (5) generators satisfying the normalization condition [Lij , Lkl ] = 1 ik jl 2δ δ . Both facts, the anomaly cancellation and the pattern of fermion spectrum, strongly suggest some kind of quark and lepton uniﬁcation, at least at some very deep level, with some big group and some mechanism of its breaking. In addition, in line with the above conclusion is a well known fact that, with normalization given by eq (3.7), the running gauge couplings of the Standard Model approach each other at high scale of order 1013 GeV. Although uniﬁcation of the gauge couplings in the Standard Model is only very approximate, it is nevertheless a remarkable fact that the strength of strong and electroweak interactions become comparable at certain energy scale. 3.3. NEUTRINO MASSES: EVIDENCE FOR NEW VERY HIGH MASS SCALE? There is at present strong experimental evidence for neutrino oscillations whose most obvious and most natural explanation is that neutrinos are massive and their mass eigenstates

PHENOMENOLOGICAL GUIDE TO PHYSICS

215

are different from the weak interaction eigenstates. This is the ﬁrst experimental evidence for physics beyond the Standard Model. The smallness of the neutrino masses can be easily understood as due to the presence of a new very high mass scale M . If it makes sense to rely on the Appelquist-Carazzone decoupling theorem then the mass scale M would manifest itself via higher dimension operators. Neutrino mass terms may appear as dimension ﬁve operator 1 (HlA )λAB (HlB ) M

(3.8)

where we use the following notation: (Hl) ≡ ij H i lj denotes SU (2)L contraction and ll ≡ αβ lα lβ denote Lorentz contraction. After spontaneous SM gauge symmetry breaking by the Higgs boson VEV the operator (3.8) gives indeed a Majorana mass matrix for neutrinos: Lν mass = −mAB νA νB + H.c. ,

mAB =

v2 λAB M

(3.9)

Small neutrino masses are obtained for big value of M , with the constants λAB ∼ O(1). This is called a see-saw mechanism. A possible and, in fact quite elegant, origin of the mass scale M would be the existence of another left-handed particle ν c , a singlet of SU (2)L × U (1)Y , i.e. a ﬁeld such that CP ν c (CP )−1 ≡ νR

(3.10)

AB c c LMajorana = MMaj νA νB B + H.c.

(3.11)

with a Majorana mass term

It can be interpreted as a right-handed neutrino ﬁeld. Moreover, we can construct Yukawa interactions c YνBA ljA + H.c. (3.12)

ij Hi νB with a new set of (neutrino) Yukawa couplings YνBA . Both terms are SU (2)L × U (1)Y invariant and even renormalizable. We can consider then the diagram shown in Fig. 8. At H

H

νc l

MMaj ×

νc l

Figure 8. Diagram generating the dimension 5 operator.

the electroweak scale v, if MM aj v, we obtain the effective interaction shown in Fig. 9 described by the operator

216

STEFAN POKORSKI

H

H

l

l

Figure 9. Effective dimension 5 operator. −1 DC CB

ij Hi ljA YνDA (MM Y

ij Hi ljB aj )

(3.13)

We recognize the previously introduced operator (3.8) with λAB −1 = (YνT MM AJ Yν )AB M

(3.14)

The see-saw mechanism with a new mass scale M is the most compelling explanation of the smallness of the neutrino masses [4, 5]. Indeed i) the smallness of mν is then related to its zero electric charge ii) the smallness of mν is also related to lepton number violation at the scale M v2 , v = 240 GeV and for Y ∼ O(1) we get mν ∼ (0.01 ÷ 0.1) eV iii) with mν ∼ Y 2 M 15 for M ∼ (10 ÷ 1013 ) GeV So, the scale M is close to the GUT scale. iv) ν c completes the spinor representation of S0(10) v) heavy ν c can play important role in baryogenesis via leptogenesis. 3.4. HIERARCHY PROBLEM IN THE SM: HINT FOR A NEW LOW MASS SCALE? Quantum corrections to the Higgs potential mass parameter m2 in eq. (2.8) in the SM are quadratically divergent. If the SM is an effective low energy theory and has a cut-off at some mass scale M of new physics, ΛSM ∼ M , it means then that quantum corrections to m2 are quadratically dependent on the new mass scale present in the underlying more fundamental theory. When M MZ this is very unnatural even if m2 , that is MZ , remains a free parameter of this underlying theory, and particularly difﬁcult to accept if in the underlying theory m2 is supposed to be ﬁxed or indeed is ﬁxed by some more fundamental considerations (as e.g. in supersymmetric and Little Higgs models, respectively). The latter is necessary if the underlying theory is to predict the scale of the electroweak symmetry breaking in terms of “more fundamental” parameters and, generically, in terms of its own cut-off ΛNew . Thus, for naturalness of the Higgs mechanism in the SM there > M , say only order of magnitude higher than M and should exist a new mass scale M ∼ Z Z better understanding the mechanism of the electroweak symmetry breaking is, hopefully, a bridge to new physics that will be explored at the LHC. For a more quantitative discussion of this so-called hierarchy problem we recall that in general in a ﬁeld theory with a cut-off Λ and some scalar ﬁeld(s) φ that can acquire

PHENOMENOLOGICAL GUIDE TO PHYSICS VEV(s) the 1-loop effective potential is

1 Λ d4 k ∆V1−loop (φ) = STr ln k 2 − M2 (φ) = cΛ4 + c Λ4 ln Λ2 2 (2π)4 1 1 M2 (φ) 2 2 4 + Λ STrM (φ) + STrM (φ) ln + ... 32π 2 64π 2 Λ2

217

(3.15)

where StrM2 (φ) =Tr[(−)F M 2 (φ)] with F - the fermion number operator and M2 (φ) is the full φ-dependent mass matrix for all ﬁelds of the theory. The ﬁrst terms in the expansion are the φ-independent contribution to the vacuum energy. We are interested in quantum corrections δm2 to the mass parameter m2 of the φ ﬁeld potential. They are obtained by expanding STrM2 (φ) = c2 φ2 + . . . ,

STrM4 (φ) = c4 φ2 + . . . ,

(3.16)

The corrections proportional to c2 (to c4 ) are in general quadratically (logarithmically) dependent on the cut-of scale Λ. In the SM with a cut-of ΛSM we ﬁnd 2 ∂ ∆V1−loop (φ) 3 2 2 3g2 + g12 + λ − 8yt2 Λ2SM + . . . , (3.17) = δm = 2 2 ∂φ 64π min If the SM was the correct theory up to the mass scale suggested by the see-saw mechanism, ΛSM ∼ MGU T 2 |δm2 | ∼ 1030 GeV2 ∼ 1026 MW !

Clearly, this excludes the possibility of understanding the magnitude of Fermi constant −2 2 2 2 in any sensible way. We also see that for |δm2 | ∼ MW , 10MW , 100MW GF ∼ MW < 0.5 TeV, < 1 TeV, < 6 TeV, respectively. If the scale of new physics is one needs ΛSM ∼ ∼ ∼ in the above range it should be discovered at the LHC. However, for a solution of the hierarchy problem it is not enough to have a low physical cut-off scale of the SM. The deeper theory has its own cut-off scale ΛNew and the dependence on it of δm2 calculated in this deeper theory should be mild enough, in order not to reintroduce the hierarchy problem for the electroweak scale.1 Many theoretical ideas have been proposed for solving the hierarchy problem of the electroweak scale. In supersymmetric extensions of the SM the dependence on their own cut-off scales ΛNew is only logarithmic because the quadratic divergences cancel out at any order of the perturbation expansion. Since the effective potential V1−loop (φ) depends only on ln ΛNew (to any order of the perturbation expansion) the scale ΛNew can be as high as the Planck scale. The quadratic dependence on the SM cut-off scale ΛSM , that is on the mass scale MSUSY of the superparticles, shows up in 2 ) φ2 + . . . STrM4 (φ) = f (MSUSY

(3.18)

2

and more explicitly, at the 1-loop as δm2 = 1

2 1 2 Λ2New 2 2 3g M + g − 12y ln , 2 1 t SUSY 2 16π 2 MSUSY

(3.19)

The hierarchy of some other (new) scales is nevertheless usually present. > The formula (3.19) applies in fact to m2H2 which for tan β ∼ 5 is the most important for electroweak symmetry breaking. 2

218

STEFAN POKORSKI

where we have replaced all soft supersymmetry breaking mass terms including the Higgs boson mass by MSUSY . Eq. (3.19) shows that in supersymmetric models the electroweak scale is calculable in terms of the known coupling constants and the (unknown) scales MSUSY and ΛNew . For a natural solution to the hierarchy problem of the electroweak scale MSUSY has to < O(10)M . However, a new very difﬁcult question appears about be low, say MSUSY ∼ W the hierarchy ΛNew /MSUSY . This is the question about the mechanism of supersymmetry breaking. In gravity mediation scenarios ΛNew ∼ MPl . In gauge mediation scenarios ΛNew is low but it is a new, introduced by hand, scale. Other ideas for solving the hierarchy problem of the electroweak scale are more “pragmatic”. Focusing on the scenarios with some predictive power, their general structure is the following: the low energy electroweak theory (but not necessarily the renormalizable SM) is embedded in a bigger one with a characteristic mass scale ΛSM ∼ M ∼ O(1 TeV). > O(10 TeV), high The new physics is under perturbative control up to its cut-off ΛNew ∼ enough to avoid any conﬂict with precision electroweak data (to be discussed later). For > O(10 TeV) to be useful for solving the electroweak hierarsuch scenarios with ΛNew ∼ chy problem the dependence of δm2 on ΛNew calculated in the extended theory has to be weak enough. This is obtained by ensuring that at least 1-loop contribution to the effective potential (3.15) have no quadratic dependence on ΛNew : (δm2 )1−loop = 0 · Λ2New + O(ln ΛNew ) + const.

(3.20) 2

E.g. in the Little Higgs models the vanishing of the c2 in STrM (φ) in eq. (3.16) is ensured by cancellation between contributions from particles of the same statistics. Such models predict the existence of new quark-like fermions and gauge bosons with masses ∼ M . In these models, the quadratic dependence of δm2 on ΛNew is present in higher order of the perturbation expansion but it is suppressed by loop factors. The tree level Higgs mass parameter m2 usually vanishes m2 (M )tree = 0 as e.g. the Higgs boson is a Goldstone boson of some bigger (approximate) symmetry spontaneously broken at the scale M , i.e. M is identiﬁed with the “decay constant” and ΛNew ∼ 4πM . The electroweak symmetry is broken by quantum corrections. The electroweak scale is then > predicted e.g. MW = MW (couplings,M, ΛNew ) with a mild dependence on ΛNew ∼ O(10 TeV) at which new unknown strong interactions set on. The crucial role is played by the new physics parameter M . In judging the plausibility of such ideas it is worth remembering our remarks in 3.1. 3.5. NEW LOW MASS SCALE AND PRECISION ELECTROWEAK DATA The presence of new physics at low energy scale, M ∼ 1 TeV raises the question on its contribution to the electroweak observables. We can address this question in a model independent way if again we assume the Appelquist-Carazzone decoupling scenario, i.e. renormalizable SM and corrections to it from new physics as higher dimension operators: ci ˆ n+4 . O (3.21) Leﬀ = LSM + Mn i n+4 ˆ O i

This time we are interested in operators which contribute to the electroweak observables. Such operators are necessarily of dimension n ≥ 6. One can classify various contributions

PHENOMENOLOGICAL GUIDE TO PHYSICS

219

from new physics according to the value of the coefﬁcients ci in the Lagrangian (3.21): ci ∼ O(1) for new tree-level contributions or contributions from new strong interacting sector; ci ∼ O(1/16π 2 ) for contributions from perturbative new physics at 1-loop. Fitting the electroweak observables one obtains limits on Mci2 . Strictly speaking, the limits are applicable to new physics which gives renormalizable SM as its low energy limit but the results are also indicative for new mass scales in models like e.g. Little Higgs, in which the Appelquist-Carazzone decoupling does not work. In any case, the constraints on such models from the electroweak observables can be discussed model by model. The task of using electroweak data to put limits on the new scale M is greatly facilitated by the expectation that the dominant corrections from new physics will show up as corrections to the gauge boson self-energies (the so-called oblique corrections). This expectation is based on the experience gained in the SM and with its various hypothetical extensions. Several authors have obtained limits on the scale M under plausible assumption that the main corrections to the SM ﬁts appear in gauge boson self-energies A comment is in order here. Gauge boson self-energies are not gauge invariant objects. The vertex and box contributions contain also pieces which are independent of the external legs. They cancel the gauge dependence of the full gauge boson self-energies and restore Ward-identity. Therefore, strictly speaking, there is an ambiguity in extracting the gauge boson self energies from ﬁts to experimental data. Fortunately, the vertex and box contributions are in the SM in the commonly used gauges much smaller than the gauge boson self energies and the neglection of gauge dependence of the latters in obtaining experimental information about their magnitude seems to be a reasonable approach. In the SM, with unbroken electromagnetic U (1) gauge symmetry there are four independent gauge boson self-energies Πij (q 2 ):

d4 x e−iq·x Jiµ (x)Jjν (0) = ig µν Πij (q 2 ) + q µ q ν term

(3.22)

For instance, we can take Πγγ , Π3γ , Π33 and Π11 as independent quantities (i = 1, 3 are SU (2)L indices; the QED Ward identity implies Π11 (q 2 ) = Π22 (q 2 ). In the limit of q 2 M 2 we can expand Πγγ (q 2 ) ≈ q 2 Πγγ (0) Π3γ (q 2 ) ≈ q 2 Π3γ (0) Π33 (q 2 ) ≈ Π33 (0) + q 2 Π33 (0)

(3.23)

Π11 (q 2 ) ≈ Π11 (0) + q 2 Π11 (0) (Πγγ (0) = 0 by QED Ward identity; the only non-zero contribution to Π3γ (0) comes from the W ± -charged Goldston boson loop). Thus, oblique corrections to the electroweak observables are to a good approximation parametrized by six constants. Three of them (or three linear combinations) are ﬁxed in terms of α, MZ and GF by the renormalization procedure. In the remaining three combinations the UV divergences must cancel.3 One 3 The ﬁnitness of the gauge sector contribution to S, T and U requires the inclusion of the terms with Π3γ (0).

220

STEFAN POKORSKI

usually deﬁnes [6, 7] ΠW W (0) e2 [Π11 (0) − Π33 (0)] = 2 2 s c2 MZ2 MW

≡ 4e2 Π33 (0) − Π3γ (0) ∝ Π3Y (0) ≡ 4e2 [Π11 (0) − Π33 (0)]

αT ≡ αS αU

−

ΠZZ (0) MZ2 (3.24)

It is clear from their deﬁnition that the parameters S, T and U have important symmetry properties: T and U vanish in the limit of unbroken custodial SU (2)V symmetry. The parameter S vanishes when SU (2)L is unbroken; unbroken SU (2)V is not sufﬁcient for vanishing of S because S ∝ Π3Y (0) = Π3L,3R (0) + Π3L,B−L (the decomposition is labelled by the SU (2)L × SU (2)R × U (1)B−L quantum numbers) and 3L × 3R = 1 + 5 under SU (2)V . It turns out that in the SM the quantum corrections to the ρ parameter deﬁned in section 2 as a ratio of physical (and measured) observables are to a very good approximation given by αT : ∆ρ ≡ ρ − 1 =

MW − 1 ≈ αT MZ cos θ

(3.25)

and, according to the eq. (2.18), depend quadratically on the top quark mass and logarithmically on the SM Higgs boson mass. Eq. (3.25) is a good approximation because other corrections to ρ (vertex corrections) are in the SM negligibly small. Similarly, in the SM e2 Mh2 m2t αS = −2 ln 2 + ln 2 . (3.26) 48π 2 MZ MZ It was discussed in section 2.3 that the SM quantum corrections agree excellently with electroweak data. The only free parameter in the ﬁts of the SM to these data is the Higgs boson mass. The main part of this dependence enters through the ρ parameter and the data favour negligible contribution to ρ from ln(Mh /MW ) (see eq. (2.18)). The ﬁts give Mh ≈ O(100 GeV), with a big error since the dependence of the ﬁts on the Higgs boson mass is only logarithmic. Such ﬁts determine the values of parameters SSM , TSM and USM for the best ﬁtted value of Mh . We can discuss the room for new physics contribution to the electroweak ﬁts by writing more generally: T = TSM (Mh ) + ∆T , S = SSM (Mh ) + ∆S ,

(3.27)

U = USM (Mh ) + ∆U , where TSM (Mh ) etc. is the SM contribution for some ﬁxed value of the Higgs boson mass. A ﬁt to the data gives now some values for ∆T , ∆S and ∆U as a function of the assumed value of Mh and shows how much room we have for new physics for different values of Mh . It is clear that for, say, Mh = 115 GeV such ﬁts give ∆T , ∆S and ∆U consistent with zero and the only room for new physics is in the errors of the ﬁtted values

PHENOMENOLOGICAL GUIDE TO PHYSICS

221

of these quantities. For larger values of Mh we have more room for new physics contributions. As we can see from eqs. (2.18) and (3.26) it must be positive to T and negative to S to balance the contribution of the larger Higgs boson mass. The ﬁtted values of ∆T , ∆S and ∆U for different values of Mh give limits on the coefﬁcients ci /M of the dimension six operators that contribute to ∆T , ∆S and ∆U . Several interesting conclusions have been reached in such studies [8]. First of all, independently of the assumed value of Mh , for ci ∼ O(1) the ﬁts to the electroweak > O(4 TeV) This limit is reached only for very correlated data give a lower limit M ∼ < signs of the coefﬁcients ci . Thus, qualitatively speaking, any new physics with M ∼ O(10 TeV) must be perturbative and cannot contribute at the tree level to be consistent with electroweak data. This strongly suggests a perturbative solution to the hierarchy problem. Secondly, if new physics is indeed perturbative and shows up only at loop level, then < 240 GeV independently of the value of the scale M . the ﬁts give Mh ∼ > 300 GeV then new physics with the scale M ∼ 10÷30 TeV and with Finally, if Mh ∼ ci ∼ O(1) (i.e. strongly interacting) is actually needed! Thus, experimental discovery of the Higgs boson and the determination of its mass will be a strong hint about the kind of new physics one may expect. 4. Supersymmetric extensions of the Standard Model In the rest of these lectures we discuss supersymmetric models as at present the most complete theoretical framework going beyond the Standard Model [9]. It has quite a few attractive features and also a number of difﬁculties, so that the full success of the SM is not automaticaaly recovered. We have already mentioned some of them in several places in these lectures but here we collect them together and extend our discussion. We shall not discuss one fundamental unresolved issue for supersymmetry which is the mechanism of spontaneous supersymmetry breaking and its transmission to the the SM sector. On the phenomenological side, the related problem is that of new and potentially dangerous sources in the soft supersymmetry breaking parameters of the FCNC and CP violating transitions. For a review of all these aspects of supersymmetric and supergravity models see e.g. [10, 11, 12]. 4.1. PRECISION ELECTROWEAK DATA As discussed in section 3.4 supersymmetric models like the Minimal Supersymmetric Standard Model (MSSM) or its simple extensions satisfy a very important criterion of calculability. Most of the structure of the Standard Model is built into them, so the renormalizable Standard Model is their low energy approximation. Supersymmetric models are easily consistent [14] with the electroweak data since the supersymmetric quantum corrections to the Standard Model ﬁts are suppressed by powers of the mass scale MSU SY of supersymmetric particles and for MSU SY > O(500) GeV are well below experimental errors (in particular, the custodial symmetry breaking by the sfermion masses is sufﬁciently suppressed). Thus, the predictive power of the Standard Model remains intact and its success is not accidental.

222

STEFAN POKORSKI

4.2. THE ELECTROWEAK SYMMETRY BREAKING Supersymmetric models solve the hierarchy problem of the electroweak scale. In the limit of unbroken supersymmetry the quadratically divergent quantum corrections to the Higgs mass parameter are absent in any order of perturbation theory. When supersymmetry is softly broken by a mass scale M , the superpartners get their masses from the electroweak breaking and from the supersymmetry breaking mass terms ∼ M . They decouple at energies smaller than M and the quadratically divergent Standard Model contribution to the Higgs mass parameter is cut-off by M and, therefore, depends quadratically on M. Thus, < O(1) TeV. The cut-off the hierarchy problem of the electroweak scale disappears if M ∼ to a supersymmetric theory can be as high as the Planck scale and “small” scale of the electroweak symmetry breaking is still natural. The electroweak symmetry breaking may be triggered by radiative corrections to the Higgs potential: Λ (4.1) (δm2H2 )1−loop ∼ −O(0.1) M 2 ln M This formula follows from eq. (3.19). If we assume that m20 ∼ M 2 , i.e. that the tree level Higgs mass parameter is approximately equal to the soft supersymmetry breaking scale the radiative electroweak symmetry breaking ((m2H2 )1−loop < 0) is triggered by the large top quark Yukawa coupling, hidden in the numerical factors of eq. (4.1). With the Higgs boson self-interactions ﬁxed by the gauge couplings of the Standard Model λφ4 → g 2 φ4

(4.2)

one obtains the correct prediction for the electroweak scale for Λ ∼ MGU T, P l . This nicely ﬁts with uniﬁcation of the gauge couplings. 4.3. THE MASS OF THE LIGHTEST HIGGS BOSON Supersymmetric models typically restrict the couplings in the Higgs potential and give strong upper bounds on the mass of the lightest Higgs particle [13]. In the minimal model the Higgs boson self-coupling comes from the D-terms and its self-coupling is the gauge coupling, eq.(4.2). Therefore, at the tree level MHiggs < MZ ≈ 91 GeV

(4.3)

There are large quantum corrections to this result. They depend quadratically on the top quark mass and logarithmically on the stop mass scale Mt˜ ∼ MSUSY : 2 MHiggs = λv 2

(4.4)

where λ is given by λ=

1 2 (g + g12 ) cos2 2β + ∆λ , 8 2

with

∆λ =

Mt˜2 3g22 m4t ln . 2 8π 2 v 2 MW m2t

(4.5)

> 500 GeV and for The present experimental limit MHiggs > 114 GeV requires Mt˜ ∼ Mt˜ < 1 TeV, MHiggs < 130 GeV. The closer the Higgs mass would be to the present

PHENOMENOLOGICAL GUIDE TO PHYSICS

223

experimental limit, the better it would be for the “naturalness” of the electroweak scale. Clearly, in the MSSM, the tunning in the Higgs potential depends exponentially on the Higgs mass and one may eventually have some tension here (see eqs. (4.5), (3.19)). One can depart from the minimal model and relax the bound on the Higgs mass. For instance, with an additional chiral superﬁeld which is a Standard Model singlet, one may couple the singlet to the Higgs doublets and get additional contributions to the Higgs self< 1 TeV, the coupling. Explicit calculations show that in such and other models, with M ∼ bound on the Higgs mass cannot be raised above ∼ 150 GeV if one wants to preserve perturbative gauge coupling uniﬁcation. 4.4. GAUGE COUPLING UNIFICATION It is well known [15] that in the framework of the MSSM with degenerate sparticle spectrum characterized by MSU SY ≈ 1 TeV the three experimentally measured gauge couplings unify with high precision at the scale MGU T ∼ 1016 GeV. This gives support to perturbative new physics at O(1 TeV). Supersymmetry and the idea of grand uniﬁcation (see section 3.2) mutually strengthen their attractiveness. A closer look at the uniﬁcation is interesting. One may ask how precise is the uniﬁcation when the superpartner masses are not degenerate and different from 1 TeV. It has been understood that even for nondegenerate superpartner spectrum the superpartner mass dependence of the RG evolution of the gauge couplings can be described to a very good approximation by a single effective parameter T . The superparticle threshold effects are correctly included in the supersymmetric 1-loop RGE whose running starts at T , with the SM RG equations used below the scale T . For consistency, 2-loop running should also be included. T depends strongly on the higgsino (µ) and gaugino (Mi ) mass parameters and much weaker on the sfermion masses. Exact uniﬁcation of the measured gauge couplings requires T ≈ 1 TeV, i.e. the higgsino and the gaugino physical masses ∼ 1 TeV if degenerate. However, a more plausible assumption that the parameters µ and Mi are approximately degenerate and ∼ 1 TeV at the GUT scale gives T ≈ 100 GeV because of strong renormalization effects. Thus, a realistic spectrum does not give exact uniﬁcation and one may wonder about the accuracy of uniﬁcation in the MSSM. In order to deﬁne what we understand by ’successful uniﬁcation’ let us ﬁrst recall the one-loop renormalization group equations in the SM and MSSM. At one-loop the gauge couplings α ˜ i of the three group factors of GSM run according to the equations: (i)

1 1 b = − 0 ln α ˜ i (Q) α ˜ i (MZ ) 2π

Q MZ

+ δi

(4.6)

Here, 1/˜ αi (MZ ) = (58.98 ± 0.04, 29.57 ± 0.03, 8.40 ± 0.14) are the experimental (i) values of the gauge couplings at the Z 0 -pole and b0 are the one-loop coefﬁcients of the 1 4 4 4 relevant beta-functions. They read b0 = ( 10 + 3 Ng , − 43 6 + 3 Ng , −11 + 3 Ng ) in the SM 3 and b0 = ( 5 + 2Ng , −5 + 2Ng , −9 + 2Ng ) in the MSSM, where Ng is the number of generations. Threshold corrections (e.g. from heavy GUT gauge bosons) are represented by the parameters δi . As explained earlier using the MSSM RG equations directly from

224

STEFAN POKORSKI

the electroweak scale for T ≈ MZ means that the supersymmetric threshold corrections corresponding to a realistic mass spectrum are properly included. In the bottom-up approach one can speak about the gauge coupling uniﬁcation if in some range of scales Q the couplings deﬁned by eq. (4.6) with, in general, Q-dependent δi (Q) can take a common value αi (Q) = αGUT for reasonably small values of δi (Q) −1 ).4 The condition for the uniﬁcation can be succintly written as (compared to αGUT 1 (j) (k) + δi (b0 − b0 ) = 0

ijk (4.7) α ˜ i (MZ ) Putting in the experimental values for αi (MZ ) and the beta-function coefﬁcients we get: −41.1 + 3.8δ1 − 11.1δ2 + 7.3δ3 = 0 −0.9 + 4δ1 − 9.6δ2 + 5.6δ3 = 0

(SM) (MSSM)

(4.8)

We see, that to achieve the gauge coupling uniﬁcation at the one-loop level we need −1 in the SM, while in the MSSM we the threshold corrections δi to be of order 10% αGUT −1 need only δi ∼1% αGUT . In the MSSM once the two loop effects are inluded one needs δi ’s by factor 2 larger. The uniﬁcation of the gauge cuplings in the MSSM is indeed very precise: it admits (and requires) only 2% threshold corrections from the GUT physics. These 2% corrections give 10% effect on αs at MZ scale, but the precision of uniﬁcation in the MSSM should be judged by the necessary for exact uniﬁcation threshold corrections at the GUT scale. Uniﬁcation of the gauge couplings does not necessarily imply the standard GUT theories with all their problems, like spontaneous breaking of the GUT gauge group by VEVs of some Higgs ﬁelds, the doublet-triplet splitting problem, etc. Many different solutions have been proposed. With the threshold corrections of the right order of magnitude, the uniﬁcation scale can be estimated from the equation: MGUT 1 1 (1) 1 (2) − − (b − b0 ) ln (4.9) + (δ1 − δ2 ) = 0 α1 (MZ ) α2 (MZ ) 2π 0 MZ For the sake of concreteness, we assume here that δ1 = δ2 = 0 and that all threshold corrections are accounted for by δ3 (thus, the uniﬁcation point is assumed to be where α1 and α2 intersect). Putting in the experimental numbers and the beta-function coefﬁcients we get MGUT ≈ 1 × 1013 GeV in the SM and MGUT ≈ 2 × 1016 GeV in the MSSM. The scale of uniﬁcation in the MSSM is determined very precisely to be in the range (2 ÷ 4) × 1016 GeV. This is interesting because it is very close to the reduced Planck scale MPl = 2 × 1018 GeV and could be considered as evidence for uniﬁcation including gravity.5 But one to two orders of magnitude difference between the two scales needs some explanation. Of course, new particles in incomplete SU (5) representations would 4

Whether there exists a uniﬁed model able to provide such values of δi (Q)’s is a different question. 5 In string theories without the stage of Grand Uniﬁcation below the compactiﬁcation scale MS the couplings unify at MS which e.g. in weakly coupled heterotic string is about factor 5 below MPl .

PHENOMENOLOGICAL GUIDE TO PHYSICS

225

alter the running and could push the uniﬁcation scale closer to the Planck scale. However, one must not destroy the precision of the uniﬁcation by new threshold corrections, so this possibility looks very ﬁne-tuned and ad hoc. An interesting possibility would be to unify the three gauge interactions with gravity at MGU T ∼ 1016 GeV by changing the energy dependence of the gravity coupling. This is possible if gravitational interactions (and only they) live in more than 3 spatial dimentions. The effective four-dimensional Planck constant is then 2+n 2 n = MPl MPl (4+n) R ,

(4.10)

2+n where n is the number of extra dimensions, R is their compactiﬁcation radius and MPl (4+n) is the Planck scale in 4 + n dimensions which we would like to take equal to MGU T ∼ 1016 GeV. For n = 1 (like in the M -theory of Horava and Witten) we get 1/R ∼ 1014 GeV.

4.5. PROTON DECAY In the SM the baryon number is (perturbatively) conserved since there are no renormalizable couplings violating this symmetry (see section. 2.5). Experimental search for proton decay, e.g p → e+ π 0 , p → K + ν is one of the most fundamental tasks for particle physics. The present limit is τp > 1033 years. If the SM is only an effective low energy theory the remnants of new physics should show up as non-renormalizable corrections to the SM. The lowest dimension operators for the proton decay (with the particle spectrum of the SM) is the set of dimension 6 operators of the form (6)

ˆ (6) = ci qqql O i 2 M(6)

(4.11)

For such operators for the proton lifetime we get (6) 1 [c ]2 m5p = i 4 τp 16π M(6)

!

(6)

The limit τp > 10 years gives then M(6) > ci × 1016 GeV (this is only a very rough estimate which neglects strong interaction effects). Any new physics with lower (6) mass scale that could lead to proton decay should be coupled with ci 1. For instance, (6) > ×1015 GeV which is still too high for the SM for ci ∼ αGUT ≈ 1/25 we get M(6) ∼ uniﬁcation (MGUT ≈ 1013 GeV). In supersymmetric extensions of the SM, with softly broken low energy supersymmetry there are low mass scalars in the spectrum with masses M ∼ O(1 TeV), which may have renormalizable couplings to quarks and leptons. Indeed, the most general renormalizable superpotential in the minimal supersymmetric model is 33

ˆH ˆu + D ˆH ˆd + E ˆH ˆd + H ˆu ˆ cQ ˆcL ˆ dH ˆ cQ w=U c c c c c ˆL ˆ+E ˆ L ˆL ˆ+U ˆ D ˆ D ˆ +L ˆ Q ˆH ˆu , +D

(4.12)

226

STEFAN POKORSKI

(the coupling constants and the ﬂavour indices are suppressed). The second line is also consistent with the SM gauge symmetry but these interactions do not conserve baryon and lepton numbers and give renormalizble couplings of scalars to fermions. After integrating the scalars out one gets dimension 6 operators as in (4.11) with M(6) ∼ MSUSY from diagrams shown in Fig. 10 and to be consistent with the limit on the proton lifetime we q

uc d˜c λ1

−→ λ2

c(6) c ¯c ¯ d ql 2 u M(6)

dc

l

Figure 10. Diagram generating the dimension 6 operator.

need c(6) = λ1 λ2 < 10−26 . One can forbid the terms in the second line of eq. (4.12) by imposing a discrete symmetry, the so-called matter parity Rp = (−1)3(B−L) . Such a symmetry could for instance, be a discrete remnant of the gauged U (1)B−L in the SO(10) theory [17]. Matter parity is equivalent to R-parity R = (−1)2S+3(B−L) acting on the component ﬁelds, where S is their spin, since Lorentz-invariant interactions preserve (−1)2S . We get then a stable LSP - candidate for dark matter in the Universe. In supersymmetric GUT models, even with R−parity imposed, there is still another source of dangerous contributions to the proton decay amplitudes. These are the dimension 5 operators (5)

ˆ (5) = ci qq q˜˜l O i 2 M(5)

(4.13)

which when inserted into one loop diagrams with gaugino exchanges give rise via diagrams shown in Fig. 11 to dimension 6 operators. In the effective dimension 6 oper-

q

˜l ˜ × W

q

q˜

Figure 11. Loop diagram generating the dimension 6 operator form the dimension 5 operator.

PHENOMENOLOGICAL GUIDE TO PHYSICS

227

2 ator of the form (4.11) one then gets c(6) = αGUT c(5) , M(6) = M(5) MSUSY . From 33 τp > 10 years and for MSUSY ∼ O(1 TeV), M(5) ∼ MGUT ∼ 1016 GeV, one gets < 10−7 . Thus we need small couplings in the amplitudes generating the dimenc(5) ∼ sion 5 operators. In SUSY GUT’s dimension 5 operators originate from the exchange of the colour triplet scalars present in the Higgs boson GUT multiplets as shown in Fig. 12, so c(5) ∼ Y 2 (Yukawa couplings for the quarks of the ﬁrst two generations)

˜l

q˜

Hc q

MHc ×

Hc q

Figure 12. Diagram generating the dimension 5 operator for proton decay.

and M(5) ∼ MHc ∼ MGUT . Given various uncertainties, e.g. in the unknown squark, gaugino and heavy Higgs boson mass spectrum, such contributions in supersymmetric GUT models are marginally consistent with the experimental limits on the proton lifetime, particularly in models more complicated than the minimal supersymmetric SU (5) model. Concrete classical GUT models are, however, not very attractive and plagued with various problems like e.g. doublet-triplet splitting problem for the Higgs boson multiplets. There are several interesting other ideas like uniﬁcation in (small) extra dimensions or in string theory [18], which offer the possibility of avoiding those difﬁculties and simultaneously preserving the attractive features of GUT’s. In some of such models proton is stable or its lifetime makes its decays unobservable experimentally. An interesting question is: what if proton after all decays but slow enough to rule out classical GUT models? 5. Summary Thinking about new physics from the perspective of the extremely succesful Standard Model is very challenging. At present there is no approach that fully and convincingly incorporates this succes into its structure. Focusing on the electroweak symmetry breaking alone, one may wonder if the high predictive power of the renormalizable SM for the electroweak observables and its perfect agreement with experimental data is signiﬁcant or partly accidental. If signiﬁcant - it supports supersymmetry; if partly accidental - we have more room for various speculations about the mechanism of electroweak symmetry breaking. Experiments are needed to put us on the right track, and hopefully, experiments at the LHC will give us the necessary insight.

228

STEFAN POKORSKI

Acknowledgements I thank Emilian Dudas, Gordon Kane, Jan Louis, Hans Peter Nilles, Jacek Pawe¯lczyk, Fernando Quevedo, Zurab Tavartkiladze and especially my close colaborator Piotr Chankowski for many useful conversations about physics beyond the Standard Model. My special thanks go to Jihad Mourad who patiently reviewed with me the content of these lectures during my visit to the Astroparticle and Cosmology Insititute (Paris VII) in June 2004, from the perspective of a string theorist. I thank also the Organizers of the Cargese School for a very enjoyable time. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

See for instance, S. Pokorski, Gauge Field Theories, Cambridge Monographs on Mathematical Physics, Cambridge University Press, Second Edition 2000 See for instance, S. Weinberg, The Quantum Theory of Fields vols. I-III, Cambridge University Press, Cambridge 1996-2000. See for instance, G.G. Ross, Grand Uniﬁed Theories, Addison-Wesley, Redwood City CA, 1985. For a review on quantum corrections to neutrino masses and mixing see P.H. Chankowski and S. Pokorski, Int. J. Mod. Phys. A17 (2002) 575. For a review of the works on the neutrino mass theory see G. Altarelli and F. Feruglio, Theoretical Models of Neutrino Masses and Mixing, hep-ph/0206077. G. Altarelli and R. Barbieri, Phys. Lett. B253 (1991) 161. M. E. Peskin and T. Takeuchi, Phys. Rev. D46 (1992) 381. See for instance, L. J. Hall and C. F. Kolda, Phys. Lett. B459 (1999) 213; C. F. Kolda and H. Murayama, JHEP 0007 (2000) 035. Perspectives in supersymmetry, G.L. Kane ed., World Scientiﬁc, Singapore 1998. H.-P. Nilles, Phys. Rept. C110 (1984) 1. F. Gabbiani, E. Gabrielli, A. Masiero and L. Silvestrini, Nucl. Phys. B477 (1996) 321. D. J. H. Chung, L. L. Everett, G. L. Kane, S. F. King, J. Lykken and L. T. Wang, hepph/0312378. Perspectives on Higgs Physics II, G.L. Kane ed., World Scientiﬁc, Singapore 1998. See for instance, S. Pokorski, in Proceedings of the International Workshop on Supersymmetry and Uniﬁcation of Fundamental Interactions (SUSY 95), Palaiseau, France, (hepph/9510224). S. Raby, in Review of Particle Physics, Phys. Lett. B592 1. Heavy ﬂavour Physics, Adv. Ser. Direct. High Energy Phys. 15 A.J. Buras and M. Lindner eds., World Scientiﬁc, Singapore 1998. S.P. Martin, Phys. Rev. D42 (1992) R2769. See for instance, K. R. Dienes, Phys. Rept. 287 (1997) 447.

INTRODUCTION TO COSMIC F- AND D-STRINGS

JOSEPH POLCHINSKI KITP, University of California, Santa Barbara, CA USA

Abstract. In these lectures I discuss the possibility that superstrings of cosmic length might exist and be observable. I ﬁrst review the original idea of cosmic strings arising as gauge theory solitons, and discuss in particular their network properties and the observational bounds that rule out cosmic strings as the principal origin of structure in our universe. I then consider cosmic superstrings, including the ‘fundamental’ F-strings and also D-strings and strings arising from wrapped branes. I discuss the conditions under which these will exist and be observable, and ways in which different kinds of string might be distinguished. We will see that each of these issues is model-dependent, but that some of the simplest models of inﬂation in string theory do lead to cosmic superstrings. Moreover, these could be the ﬁrst objects seen in gravitational wave astronomy, and might have distinctive network properties.

1. Introduction Seeing superstrings of cosmic size would be a spectacular way to verify string theory. Witten considered this possibility in the context of perturbative string theory, and found that it was excluded for several reasons [1]. Perturbative strings have a tension close to the Planck scale, and so would produce inhom*ogeneities in the cosmic microwave background far larger than observed. The scale of this tension also exceeds the upper bound on the energy scale of the inﬂationary vacuum, and so it would have been difﬁcult to produce these strings after inﬂation, while any strings produced earlier would have been diluted beyond observation. Ref. [1] also identiﬁed instabilities that would prevent long strings from surviving on cosmic time scales. In recent years we have understood that there are much more general possibilities for the geometry of the compact dimensions of string theory, including localized branes, and this allows the string tension to be much lower, anything between the Planck scale and the weak scale. Also, we have found new kinds of extended object in string theory. Thus the question of cosmic superstrings (and branes) must be revisited, and this has been done beginning in [2, 3]. A necessary set of conditions is: 1. The strings must be produced after inﬂation. 229 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 229–253. © 2005 Springer. Printed in the Netherlands.

230

JOSEPH POLCHINSKI

2. They must be stable on cosmological timescales. 3. They must be observable in some way, but not already excluded. Ref. [1] thus showed that perturbative strings fail on all three counts. If we do ﬁnd models that satisfy these three conditions then there is one more condition that we would hope to satisfy: 4. Cosmic superstrings should be distinguishable from other kinds of cosmic string, in particular gauge theory solitons. We will see that each of these four issues is separately model-dependent but that some fortuitous things do happen, and there are simple models in which all these conditions are met, including some of the most fully-developed models of inﬂation in string theory. These lectures give a pedagogical introduction to this subject. Section 2 reviews the story of cosmic strings that arise as gauge theory solitons. I discuss the network properties of these strings, and their observational signatures and bounds. Section 3 discusses the cosmic string candidates in various compactiﬁcations of superstring theory. Sections 4 through 7 go through each of the conditions on the list above, explaining the issues and their dependence on the details of the compactiﬁcation. Section 8 presents brief conclusions. The outline of these lectures (and some of the prose!) follows [4], but the treatment is more detailed and pedagogical. 2. Cosmic string review 2.1. STRING SOLITONS Cosmic strings might also arise as solitons in a grand uniﬁed gauge theory, and for some time these were a candidate for the source of the inhom*ogeneities that led to the structure in our universe. I start with a review of this subject, since almost all of the ideas and results carry over to superstrings. A complete treatment of this subject would require at least a full volume, and the interested reader is directed to the book by Vilenkin and Shellard [5] and the review by Hindmarsh and Kibble [6]; for a recent look at the subject see [7]. I will have space only to give an overview of the most important ideas. In any ﬁeld theory with a broken U (1) symmetry, there will be classical solutions that are extended in one dimension [8, 9].1 Consider the standard symmetry-breaking potential for a complex scalar ﬁeld, V (Φ) = λ(|Φ|2 − v 2 )2 .

(2.1)

This has a ring of degenerate minima, Φ = veiψ for any phase ψ. Suppose we have a conﬁguration in two space dimensions such that the potential energy is nonzero only in a localized region, falling off rapidly at inﬁnity. At sufﬁciently large radius R the ﬁeld must then be approximately at a minimum of the potential, but possibly a different minimum in different directions: Φ(R, θ) ≈ veiψ(θ) . 1

I will make a more precise statement, in terms of hom*otopy groups, in section 4.

(2.2)

COSMIC F- AND D-STRINGS

231

As we circle the origin at long distance, the phase ψ(θ) must make some integer number n of circuits of the ring of minima. This winding number is conserved in time because it is determined by the topology of the ﬁeld at arbitrarily long distance. Thus there is a conserved topological quantum number, and the minimum energy conﬁguration in each sector is a stable topological soliton. By continuity, whenever the winding number n is nonzero the ﬁeld must somewhere pass through zero, so a nonzero winding number implies that there must be a nontrivial ‘lump’ of potential energy. The √ size of this core region is set by the Higgs mass mH ∼ a λ. Quantum corrections shift the minimum energy in each sector, but do not destabilize the soliton unless they actually restore the U (1) symmetry. Assuming rotational symmetry gives the form Φ(r, θ) = f (r)einθ ,

f (0) = 0 ,

f (∞) → v .

(2.3)

We can add in the third dimension simply by extending the polar coordinates to cylindrical coordinates (r, θ, z) with the ﬁeld independent of z, so that the lump becomes a long straight string. The broken U (1) can be either a global or a gauge symmetry. In the global case the Hamiltonian density at long distance is 2 2 2 + V (Φ) → v n , H = |Π|2 + |∂Φ| r2

r→∞,

(2.4)

because the gradient of Φ falls as 1/r. The string tension, dr 2πrH, is logarithmically divergent at long distance. Nevertheless these solutions are of interest. For example, for two oppositely oriented straight strings separated by a distance L, the divergence is cut off at L. This conﬁguration is topologically trivial and unstable to annihilation of the long strings, but the force between them falls as 1/R so in fact the strings can be quite long-lived. − iAΦ = ∂Φ When the U (1) is gauged, the gradient in the covariant derivative DΦ can be cancelled by an appropriate gauge ﬁeld, ˆ θ) = θg(r) A(r, ,

g(0) = 0 ,

g(∞) →

n . r

The energy is then ﬁnite. The core of the gauge string contains a magnetic ﬂux

F = A = 2πn ,

(2.5)

(2.6)

C

where C is a large circle around the origin. Most of the literature on cosmic strings deals with gauge strings. Most of the superstrings that we will encounter will indeed be analogous to gauge strings, and so this is what we assume in the discussion of network properties and signatures below, but we will see that there is also a possibility of global strings. 2.2. FORMATION OF STRINGS These classical string solutions exist whenever there is a broken U (1) symmetry, and whenever a U (1) symmetry becomes broken during the evolution of the universe a network of strings must actually form. Such a symmetry-breaking transition can occur as a

232

JOSEPH POLCHINSKI

result either of coupling to a time-dependent scalar ﬁeld or of thermal effects. In the ﬁrst case the potential for Φ might be of the form V (Φ) = λ(|Φ|2 − gχ)2

(2.7)

with χ a scalar ﬁeld that is slowly rolling from negative values, for which the potential is minimized at Φ = 0, to positive values, where the symmetry is broken. In the thermal case, the vacuum is found by minimizing the free energy E − T S. The entropy is usually larger in the unbroken phase, because it has a greater number of massless degrees of freedom. Thus the unbroken phase might be favored at high temperature and then the symmetry break as the universe cools and we approach the zero-temperature potential (2.1). The Higgs ﬁeld Φ thus starts at zero, and then when the symmetry breaks it rolls down to one of the vacua. However, it will not role in the same direction everywhere: by causality, it cannot be correlated on distances greater than the horizon scale. In practice, the correlation length is usually less than this, determined by the intrinsic length scales of the ﬁeld theory. Thus it chooses random directions in different places, and inevitably there will be some trapped winding, so that strings are left over at the end. This is the Kibble argument [10], and indeed it is what simulations show. A fraction of the string is in the form of ﬁnite sized loops, and a fraction is in the form of inﬁnite strings; the latter enter and leave the boundaries of the simulation volume no matter how large this is taken to be. These populations are cleanly separated because the distribution of lengths of ﬁnite loops falls rapidly for long loops, giving a convergent integral for the total amount of string in loops. Presumably the existence of the inﬁnite strings is implied by the causality argument. They are important because they begin to stretch with the expansion of the universe, while the small loops quickly decay away. 2.3. EVOLUTION OF STRING NETWORKS To get some understanding of the evolution of the string networks, let us ﬁrst make the crude assumption that the string network just expands along with the growth of the universe. If we look at a given comoving region, the length of string within, and therefore the total energy, grows as the scale factor a(t). The comoving volume scales as a(t)3 , so the energy density ρs scales as a(t)−2 . This would dominate over matter, ρm ∝ a(t)−3 , and radiation, ρr ∝ a(t)−4 . However, there are several processes that reduce the string energy density relative to this estimate:2 1. The long strings, which form as random walks, expand uniformly only on scales greater than the horizon length t. On shorter scales they tend to straighten over time. 2. When two strings collide they can either pass through each other or they can reconnect (intercommute), as in Fig. 1. For gauge theory solitons, the value of P is essentially 1. In an adiabatic collision, gauge theory strings always reconnect, P = 1, 2

For reasons of space I will focus on strings whose only long-distance interactions are gravitational. Global strings would also have a long-ranged Goldstone boson ﬁeld, and could decay via Goldstone boson emission. There are also superconducting strings [11], which have strong couplings to gauge ﬁelds. These are perhaps less likely to arise in the superstring case, for reasons that I will explain.

COSMIC F- AND D-STRINGS

prob. P

233

prob. 1−P

Figure 1. When two strings of the same type collide, they either reconnect, with probability P , or pass through each other, with probability 1 − P . For classical solitons the process is deterministic, and P = 1 for the velocities relevant to the string network. When the strings reconnect, a sharp kink is left on each new string, and each of these separates into a right-moving and a left-moving kink: the result is four kinks, as shown.

because reconnection allows the ﬂux in the string core to take an energetically favorable shortcut. Simulations show that this remains true at the moderately relativistic velocities that are present in string networks [12, 13, 14].3 (a) When a long string intersects itself, a loop breaks off. Loops on scales shorter than the horizon do not expand and so behave like massive matter. (b) When two different long strings reconnect they produce two long kinked strings. The kinks tend to straighten in time, as in point 1, which reduces the total string length. (This effect is not enough to produce a scaling solution without also closed loop formation from self-intersection). 3. The closed loops eventually decay through gravitational radiation. Simulations show that these processes act with the maximum efﬁciency allowed by causality, so that the appearance of the string network at any time looks the same when viewed at the horizon scale t, with a few dozen strings spanning the horizon volume and a gas of loops of various sizes [15, 16, 17], as in Fig. 2. This is known as the scaling solution. The scaling solution is an attractor: if we start with too much string, the higher collision rate will reduce it, while if we start with too little then there will be few collisions until the amount of string per horizon volume (which is increasing relative to the comoving volume) approaches the scaling value. Thus the details of the initial distribution do not matter, as long as there are some inﬁnite strings. The total length of string within a horizon is then a numerical constant times t, and the horizon volume is t3 , so the energy density ρs is a constant times µt−2 , where µ is the string tension. The scale factor is proportional to t1/2 in the radiation-dominated era and t2/3 in the matter dominated era, so it works out that the ratio of the string energy density to the dominant energy density in each era is a constant. The simulations show that ρs /ρm ∼ 60Gµ during the matter-dominated era and ρs /ρr ∼ 400Gµ during the radiation dominated era. This is our ﬁrst encounter with the dimensionless combination Gµ, about which we will have much more to say, but for now we just assert that we are interested in values of Gµ less than 10−6 . The strings would thus be a small but ﬁxed fraction of the total density. In particular, it would take some very exotic network scenario for strings to be the dark matter. 3 Cosmic strings move at moderately relativistic speeds. In ﬂat spacetime the virial theorem implies that v 2 has a mean value of 12 ; this is somewhat reduced by the expansion of the universe.

234

JOSEPH POLCHINSKI

Figure 2. Cosmic string simulation. A side of the cube is around a third of the horizon length. Strings that appear to end are just leaving the simulation volume. From Allen and Shellard [17].

The horizon distance, approximately t, sets the overall scale of the string network, but there are important features on smaller length scales. In particular, there is a lower cutoff on the size of loops, coming from their decay via gravitational radiation. Dimensionally, the gravitational wave power emitted by a loop of size l is Gµ2 . Equating the loop mass µl to the power times t determines the typical lifetime for a loop of size l; equivalently it determines the size l, as a function of the time t, below which the decay becomes rapid. Inserting the constants from a more careful treatment, the result is l ∼ 50Gµt .

(2.8)

Taking Gµ = 10−9 , which will turn out to be a typical value, gives l ∼ 500 lightyears today. Thus there is large hierarchy of scales in the network. Another important property of the network is the existence of a kinked structure at short distance [16]. We have noted that reconnection of loops produces kinks. As t increases the number of kinks per horizon length necessarily increases, and this effect is stronger than the tendency of the kinks to straighten as the string stretches. The study of this structure in simulations is limited by the resolution of the simulations, but it has generally been assumed that it is cut off by gravitational radiation at the same scale (2.8) (see however [18]). The short distance structure is important because the simulations indicate that it determines the typical size of the loops that break off the inﬁnite strings. In summary, we should note that the behavior of string networks is challenging to understand in detail: the large ratio of scales makes numerical simulation difﬁcult, and while there exist various analytic models, the problem is very nonlinear and existing models

COSMIC F- AND D-STRINGS

235

only capture part of the physics. If cosmic strings are ever detected, it will become important to work toward a much more precise understanding. Finally, it is interesting to contrast the properties of one-dimensional defects with those of two-dimensional and zero-dimensional defects. If a network of domain walls forms and subsequently annihilates to the maximum extent allowed by causality, then the typical spacing between walls will be of order t. The energy density in domain walls will then scale as 1/t, as compared to the 1/t2 of strings, and this would come to dominate over the matter and radiation densities. Such a network is therefore excluded. For point particles, annihilation is not able to reduce the density to the scaling value: points have a harder time ﬁnding each other than strings. The spacing then scales as a(t)−1 and the density as a(t)−3 , just like other massive matter; such defects would therefore come to dominate in the radiation-dominated era and are excluded. Thus, whereas strings represent a cosmic opportunity, domain walls and point defects are forbidden. In section 4 we will discuss potential instabilities that could eliminate cosmic strings, and the analogous decays would also provide solutions to the domain wall and point defect problems. However, there is one kind of point defect that provides a serious challenge, namely the magnetic monopole [19, 20]. These exist in all theories that unify electromagnetism with the other interactions (for a recent overview see [21]). They cannot decay (magnetic charge conservation) and they are not conﬁned (at least today), so they present a vexing cosmological problem. However, if the phase transition in which they are produced took place before inﬂation, they would have been diluted the point that a monopole would probably never be seen; this was one of the original motivations for inﬂation [22]. By the same token, cosmic strings will only be of interest if they are produced at the end of inﬂation, or later. 2.4. STRING SIGNATURES AND BOUNDS Since we are considering strings whose only interactions are gravitational, all of their effects are controlled by the dimensionless product of Newton’s constant and the string tension, Gµ. One can think of this in two ways: it is the string tension in Planck units, and it sets the size of the typical metric perturbation produced by a string. For example, the geometry around a long straight string is conic, with a deﬁcit angle 8πGµ. The string network produces inhom*ogeneities proportional to Gµ, and because of the scaling property of the string network these are scale invariant. A value Gµ ∼ 10−5.5 would give scale-invariant perturbations of the right magnitude to produce the galaxies and CMB ﬂuctuations.4 This was a viable theory for some time [23, 24], but it is now excluded. For example, the CMB power spectrum is wrong: the actual spectrum shows a pattern of peaks and dips, whereas the spectrum from strings would be smooth. There is a simple reason for this. The ﬂuctuations produced in inﬂation have a deﬁnite phase. This phase is maintained from the end of inﬂation until the perturbations go nonlinear, and is imprinted as oscillations of the power spectrum. Strings, on the other 4 I am going to quote values of Gµ to the nearest half order of magnitude. This is less precise than most numbers given in the literature, but it is all I will need and it roughly reﬂects the uncertainties in the understanding of the string network. To give more precise numbers would require a more detailed discussion, and so the reader should consult the references.

236

JOSEPH POLCHINSKI

hand, each keep their own time, there is no common phase. Fitting cosmological constant plus cold dark matter plus strings to the CMB power spectrum gives an upper limit Gµ < 10−6 [25, 26]. Beyond the power spectrum, strings will produce nongaussianities in the CMB. Recent limits are somewhat stronger than those from the power spectrum, around Gµ < 10−6.5 [27, 28].5 Incidentally, strings affect the CMB in more than one way. Strings at early times will induce the matter inhom*ogeneities that are reﬂected in the CMB, but even if the CMB were uniform, strings at the present time would produce apparent inhom*ogeneities. We have noted that a static string bends light. If a string moves with velocity v transverse to the observer there will be a differential redshift of order 8πvGµ between the two sides. Another limit on Gµ comes from pulsar timing. Because the energy in the strings eventually goes into gravitational waves, strings produce a large stochastic gravitational wave background. The classic reference on this subject [29] quotes a limit on the energy density in stochastic gravitational waves, per logarithmic frequency range, as ΩGW < 1.2 × 10−7 , using h2 = 0.5 for the Hubble parameter. In the frequency range of interest one can estimate the stochastic background from a network of strings as ΩGW = 0.04Gµ [5], meaning that Gµ < 10−5.5 . Different analyses of the same data, with different statistical methods, give a (controversial) bound a factor of 6 stronger [30] and another a factor of 1.6 weaker [31]. The limits from pulsars should increase rapidly with greater observation time, and recent work [32] quotes a bound 30 times stronger than [29], meaning that Gµ < 10−7 . However, this paper again obtains much weaker limits using other methods and so this should not be regarded as a bound until there is agreement on the analysis.6 The bounds here are from gravitational waves with wavelengths comparable to the size of the emitting string loop, and do not include a potentially substantial enhancement due to high-frequency cusps; see section 6. Thus far we have quoted upper bounds, but there are possible detections of strings via gravitational lensing. A long string will produce a pair of images symmetric about an axis, very different from lensing by a point mass. Such an event has been reported recently [33, 34]. The separation of around two arc-seconds corresponds to Gµ equal to 4 × 10−7 times a geometric factor of order 1. I have heard varying opinions on how seriously to take such an observation, as similar pairs in the past have turned out to coincidental. There is further discussion in the review [7], which also discusses a possible time-dependent lens, as from an oscillating loop, with Gµ ∼ 10−7.5 [35]. 3. Cosmic strings from string theory Now let us consider the candidates for cosmic superstrings that might arise in various vacua of string theory, starting with the perturbative heterotic string. The gravitational and gauge couplings in this case have a common origin from the closed string interaction, and so by calculating the trilinear amplitudes one ﬁnds the relation (e.g. Eq. 18.2.4 of [36]) 4κ2 = α g 2 . 5

(3.1)

This limit is for a network of strings. For a single stray cosmic string, the limit is an order of magnitude weaker. 6 I would like to thank E. Flanagan and H. Tye for communications on this point.

COSMIC F- AND D-STRINGS

237

The gauge and gravitational couplings are equal up to a factor of α required by the dimensions, and a factor of 4 which arises from various conventions. This holds for the ten-dimensional couplings, but because the gauge and gravitational ﬁelds both live in the 2 /V where V is the compactbulk they reduce in the same way, κ24 = κ210 /V and g42 = g10 iﬁcation volume. Expressing the relation in terms of G = κ24 /8π, µ = 1/2πα , and the heterotic gauge coupling αh = g42 /4π gives Gµ =

αh . 16π

(3.2)

The heterotic gauge coupling should be at or above the minimal GUT value 1/20, so Gµ is at least of order 10−3 . The existence of such cosmic strings is therefore excluded [1]. We cannot turn this around and use it as evidence against perturbative heterotic string theory, because in fact a cosmic string network of this type could not form. The CMB limit on tensor perturbations implies an upper limit on the vacuum energy during inﬂation, < (δT /T )2 ∼ 10−10 . In most cases this bounds the tension of any strings that G2 Vinf ∼ < (G2 V )1/2 < 10−5 . Thus, perturbative heterotic strings might subsequently form, Gµ ∼ inf ∼ could only have been produced in a phase transition before inﬂation, and the many efoldings of inﬂation then make it unlikely that any would be found within our horizon. There is a second reason that heterotic strings would not reach cosmic size, as we see in section 5. For the perturbative type I string the dimensional analysis is a bit different but the conclusions are the same. The strongly coupled E8 × E8 heterotic string can have a lower tension [37]. This is because the gauge ﬁelds live on the nine-dimensional boundary while gravity lives in the ten-dimensional bulk. More generally, if the gauge ﬁelds are conﬁned to a brane while gravity propagates in the bulk, the string tension is suppressed by some power of R/LP where R is the size of the dimensions transverse to the brane [38, 39]. Even without large compact dimensions, the string tension can be reduced by a gravitational redshift (warp factor) [40] ds2 = e2A(y) ηµν dxµ dxν + . . .

(3.3)

where y are the compact coordinates. The string is localized at some point y0 in the compact space, so the string worldsheet action is proportional to µ=

e2A(y0 ) . 2πα

(3.4)

In interesting compactiﬁcations there are often throats where the warp factor e2A is much less than 1. The string will fall to the bottom of such a throat, so the four-dimensional tension µ measured in four dimensions can be reduced by a large factor relative to the tension 1/2πα seen by a local ten-dimensional observer. This is the RS idea, that different four-dimensional scales arise from a single underlying scale through gravitational redshifting. Thus the tension is essentially a free parameter, which can lie anywhere from the Planck scale down to the experimental limit near the weak scale. For example, in the warped models it is determined by the depth of the throat, which comes out as the exponential of a function of ﬂux integers that can take a wide range of values [41].

238

JOSEPH POLCHINSKI

The range becomes much smaller when we focus on brane-antibrane inﬂation models [42, 43, 44, 45]. This is a nice geometric idea for obtaining a slow-roll inﬂationary potential, and the basis for most current attempts to describe inﬂation in string theory. That is, the early universe could have contained an extra brane-antibrane pair, separated in the compact directions. The potential energy of these branes would drive inﬂation. The inﬂaton ﬁeld is then the separation between the branes: this has a potential which is rather ﬂat when the branes are separated and steepens as they approach, until at some point a ﬁeld becomes tachyonic and the brane and antibrane annihilate rapidly. To understand inﬂationary cosmology in detail one needs to know the scalar potential. This potential has long been problematic, especially for states of positive vacuum energy, because of potential instabilities in the moduli directions. Recently the tools have been developed to identify a large class of stable solutions [46, 47]. Thus we will focus on the resulting KLMT model [48], since it is the most detailed model of inﬂation in string theory. This is based on a strongly warped compactiﬁcation of the IIB string theory, with inﬂation arising from a D3/D3 pair at the bottom of a throat. In models of brane inﬂation, the value of Gµ can be deduced from the observed magnitude of the CMB ﬂuctuations δT /T . That is, one assumes that δT /T arises from the quantum ﬂuctuations of the inﬂaton; this is natural given the ﬂat form of the inﬂaton potential. For any given brane geometry the inﬂaton potential has a deﬁnite functional form. For example, in the KLMT model the D3/D3 potential is V ∼ V0 [1 − O(φ−4 )]. One then ﬁnds (Eq. C.12 of [48]) G2 V0 ∼ 0.05(δT /T )3 Ne−5/2 ∼ 1.5 × 10−20 ;

(3.5)

here Ne ∼ 60 is the number of e-foldings at the wavelengths responsible for structure formation. The vacuum energy from a D3/D3 in the warped space is 2T3 e4A0 where A0 refers to the bottom of the inﬂationary throat. The F-strings produced after inﬂation (the mechanism for which is the subject of the next section) will sit in the same throat and so µ = e2A0 /2πα . Since T3 = TF2 /2πgs in ten dimensions [36] we have for the tension of the IIB string in the inﬂationary throat √ (3.6) µF ∼ 2 × 10−10 gs , where gs is the string coupling. One might take gs to have a representative value of 0.1. There is one very speciﬁc class of models in which it is related to observed quantities. If the matter ﬁelds live on D3’s or D3’s, then gs turns out to be exactly αYM . However, gs is really eΦ at the position of the brane, and the dilaton Φ in most solutions varies in the compact space and so will not have the same value at the position of the strings as it has at the Standard Model brane. In the special case that it is constant, we can identify gs in the brane tension with the uniﬁed coupling αGUT . Uniﬁcation in brane models is generally nonstandard, but a likely value for αGUT is still around 0.1. Note that these warped manifolds can have many throats and therefore strings of many different tensions. One possibility is that the Standard Model ﬁelds live in a throat whose depth corresponds to the weak scale, whereas the depth (3.5) of the inﬂationary throat gives something closer to the GUT scale. Thus there would be the possibility to see TeVscale strings and extra dimensions at accelerators, and at the same time GUT-scale strings in the sky.

COSMIC F- AND D-STRINGS

239

In addition to the fundamental F-strings, string theory has many other extended objects. For example the IIB theory has a D-string whose tension in ten dimensions is TD = TF /gs , and so in the KLM MT model √ µD ∼ 2 × 10−10 / gs ;

(3.7)

the geometric mean (µF µD )1/2 is independent even of gs . There are also bound states of p F-strings and q D-strings with a distinctive tension formula (3.8) µ(p,q) = µF p2 + q 2 /gs2 . There are also a variety of higher dimensional D-, NS- and M-branes in the various regimes of string theory. Any of these will look like a string if all but one of its spatial directions is wrapped on a compact cycle. In the KLM MT model, any brane not contained fully in a throat will have a Planck-scale tension because a portion of it passes through the region where the warp factor is close to 1, so we will get additional interesting strings only if there are nontrivial cycles in the throat. The most generic throat, the Klebanov-Strassler throat [2] based on the conifold singularity [50], has no cycles that give additional strings, but there are many less generic singularities with a wide variety of nontrivial cycles; these have been extensively studied in the context of gauge/string duality. For models based on large compact dimensions, no fully stabilized examples are known. One can ﬁx the bulk moduli by hand and study inﬂation based on the brane moduli. This is done in [3, 51], which consider a variety of wrapped branes in type II compactiﬁcations, and ﬁnd values in the range < Gµ < 10−6 . 10−12 ∼ ∼

(3.9)

We should emphasize again that it is the assumption that the inhom*ogeneities arise from the quantum ﬂuctuation of the inﬂaton, rather than some other mechanism, that allow us to tie the tension to observed quantities in brane inﬂation. If another mechanism is responsible, the inﬂation scale and the tension could be much lower. It may still turn out to be the case that our vacuum is well-described by weakly coupled heterotic string theory. If so, inﬂation and cosmic strings might simply arise in the effective low energy ﬁeld theory (this would be true for other compactiﬁcations as well). The possible strings include both the magnetic ﬂux tubes discussed in section 2, and also electric ﬂux tubes that exist in strongly coupled conﬁning theories [1]. The heterotic string, whether weakly or strongly coupled, has the advantage that it more readily makes contact with standard grand uniﬁcation. A recent paper [52] argues that cosmic strings should be present in a wide class of grand uniﬁed inﬂationary models. 4. Production of cosmic F- and D-strings If these branes are D-branes, then there is a U (1) gauge symmetry on each of the brane and antibrane, and this U (1) × U (1) disappears when the branes annihilate. One linear combination of the U (1)’s is Higged. The Kibble argument then applies, so that a network of strings must be left over when the branes annihilate [2, 3]. These are D-strings, as one can see by studying the conserved charges [53, 54]. More precisely, if the branes that

240

JOSEPH POLCHINSKI

annihilate are D(3 + k)-branes, extended in the three large dimensions and wrapped on k small dimensions, then the result is D(1 + k)-branes that extend in one large dimension and are wrapped on the same small dimensions. The simplest case is k = 0, where D3/D3 annihilation produces D1-branes. The second linear combination of U (1)’s is conﬁned. We can think of conﬁnement as dual Higgsing, by a magnetically charged ﬁeld, and so we would expect that again the Kibble argument implies production of strings [55, 56]. These are simply the F-strings, the ‘fundamental’ superstrings whose quantization deﬁnes the theory, at least perturbatively. It is notable that this process produces only strings, and not the cosmologically dangerous monopoles or domain walls [2, 3]. The breaking of a U (1) produces defects of codimension two, and the Kibble argument requires that the codimension be in the large directions: the small directions are in causal contact.7 It is striking that what appears to be the most natural implementation of inﬂation in string theory produces strings and not dangerous defects, but we should now ask how generic this is. Even these models are not, in their current form, completely natural: like all models of inﬂation they require tuning at the per cent level to give a sufﬁciently ﬂat potential [48, 58]. There might well be other ﬂat regions in the large potential energy landscape of string theory. An optimistic sign is that there are arguments entirely independent of string theory to indicate that inﬂation terminates with a symmetry-breaking transition: this is known as hybrid inﬂation [59], and leads to efﬁcient reheating as well as production of strings [60, 61, 62]. However, not every symmetry-breaking pattern produces strings. For example, in strongly coupled heterotic string theory [37], there are M5-branes and M2-branes. It would seem natural to use M5-branes with two wrapped directions to implement brane inﬂation. Since the M2-branes have codimension three relative to the M5-branes it will not be so easy to produce strings; this is currently under investigation.8 In [69] it was proposed that cosmic fundamental strings could have been produced in a Hagedorn transition in the early universe. The Hagedorn transition corresponds to the formation of strings of inﬁnite length, and so some percolating strings would survive as the universe cooled below the Hagedorn temperature. This idea has a simple realization in the warped models.9 We have noted that the effective string tension, and so the Hagedorn temperature, is different in different throats. It is possible that after inﬂation a deeper throat reheats above its Hagedorn temperature, leading to string formation as the universe cools (a black hole horizon would then form at the bottom of the throat, corresponding to the Hawking-Page transition [70, 71]). In fact, this is essentially equivalent to the Kibble mechanism. The throat degrees of freedom have a dual gauge description, in terms of which the Hagedorn transition corresponds to deconﬁnement. The transition to a conﬁning phase as the universe cools is the electric-magnetic dual of a symmetry-breaking transition. The strings produced in this way would necessarily have a lower tension than 7 Ref. [55] identiﬁes processes that might suppress production of D-strings. Ref. [57] argues that in some cases there can be disorder in the compact dimensions and so monopoles and domain walls would be produced. We believe that, at least for the KLMT model, neither of these should be relevant. 8 There has also been recent discussion of a more exotic symmetry-breaking pattern in D-brane inﬂation [63, 64, 65]. For other discussions of string production in brane inﬂation see [66, 67, 68] 9 Similar ideas are being considered by A. Frey and R. Myers.

COSMIC F- AND D-STRINGS

a)

241

b)

Figure 3. Instabilities of macroscopic strings: a) Conﬁnement by a domain wall. b) Breakage.

those produced directly in brane annihilation, because of the inefﬁciency of the extra thermal step. 5. Stability of strings 5.1. FIELD THEORY STRINGS Ref. [1] identiﬁed two instabilities that would prevent superstrings from growing to cosmic size. Actually, these same two instabilities exist for ﬁeld theory soliton strings [72] — one for global strings and the other for gauge strings — so let us ﬁrst discuss them in this context. In the case of global strings, we have noted that the long-ranged Goldstone boson has gradient energy. It does not have potential energy at long distance as long as the broken U (1) symmetry is exact: the broken vacua are then exactly degenerate. However, there are general arguments in string theory that there are no exact global symmetries [73, 36]. More generally, the no-hair theorems imply that black holes can destroy global charges, so in any theory of gravity these can not be exactly conserved. Thus the degeneracy of the vacua will not be exact; the trough in the potential is tilted and so there is a potential energy cost even at long distance for a ﬁeld conﬁguration that circles the trough. This cost is minimized by have the scalar ﬁeld make its excursion in a kink of ﬁnite width, rather than uniformly in angle as in the exact case (2.3). Thus there is a domain wall, with energy proportional to its area, bounded by the string. The wall exerts a transverse force on the string and forces it to collapse, as in Fig. 3. This is clear for a loop of string bounding a domain wall, but is less intuitive for a network with inﬁnite strings. One can picture the network, with the domain walls, as a complicated shape formed from strips whose width is the typical transverse separation between strings. The timescale for the collapse of the strips, and the disappearance of the strings, is set by the width of the strips, not their (potentially unbounded) length [74]. For gauge strings, the U (1) symmetry is exact because it is a gauge symmetry, and all energies fall exponentially with distance from the string. A magnetic ﬂux runs along the core of the string, and it is the conservation of this ﬂux that prevents the string from breaking. However, in any uniﬁed theory one expects that there will be electric and magnetic sources for every ﬂux [21], so that the string can break by creation of a monopole/antimonopole pair. If this is possible, it will happen not only once but everywhere along the length of the string, and so the string breaks up into short segments rather thans growing to cosmic length.

242

JOSEPH POLCHINSKI

There can be absolutely stable strings. Consider the Abrikosov ﬂux tubes in an ordinary superconductor. The Higgs ﬁeld there is an electron pair and has charge 2e, so the tube has total ﬂux 2π/2e. However, because there do exist singly charged electrons, Dirac quantization requires the minimum monopole charge to be twice this, 2π/e. Thus a ﬂux tube cannot end on a monopole, though two can; equivalently one can think of the monopole as a bead on a string, at which the ﬂux reverses. So the Abrikosov ﬂux tubes (in an inﬁnite system) are absolutely stable. One can think of this in terms of an unbroken discrete gauge symmetry (−1)Q/e , which acts as −1 on the electron and +1 on the BCS condensate. As one circles the string, ﬁelds come back to themselves only up to this transformation. Thus the string can be detected by an Aharonov-Bohm experiment at arbitrarily long distance, and so it can never just disappear. We will refer to these absolutely stable strings as ‘discrete’ strings. Gauge strings without a discrete gauge charge are truly invisible at long distance, and so there is no obstacle to their breaking. As an aside, we explain this in terms of the hom*otopy classiﬁcation of defects (see e.g. Ref. [75] for an introduction). For gauge group G broken to H there is an exact sequence Π2 (G/H)

→

Π1 (H) → Π1 (G) → Π1 (G/H)

→

Π0 (H) → Π0 (G) → Π0 (G/H) .

(5.1)

In the discussion of strings we started with G = U (1), for which Π1 (G) = Z, and broke to H = I for which Π1 (H) = Π0 (H) = I. Thus the strings are charged under Π1 (G/H) = Z. Now however let us embed the U (1) into a semisimple group; we are always free to assume this to be simply connected by going to the covering group, so Π1 (G) = I and the stable strings are classiﬁed by Π1 (G/H) = Π0 (H)G — that is, by unbroken discrete symmetries that lie in the connected component of the identity in G. It can also be that strings that are not absolutely stable have decays that are slow on cosmic timescales. For monopole pair creation, for example, the classic Schwinger cal2 culation gives a pair production rate suppressed by e−πM /µ , where M is the monopole mass. If the monopole mass is an order of magnitude larger than the scale of the string tension, then the decay will be slow on cosmological time scales. One can think of a succession of symmetry breakings such as SU (2) → U (1) → 1. There are no stable strings for this pattern, but if the scale of the ﬁrst symmetry breaking is higher than that of the second then there will be string solutions in the effective ﬁeld theory describing the second breaking, and these can be long-lived. Similarly the tension of the domain walls might be suppressed by a hierarchy of scales. 5.2. STRING THEORY STRINGS 5.2.1. Perturbative strings Now let us turn to string theory. Consider ﬁrst perturbative strings [1], compactiﬁed for example on a torus or a Calabi-Yau manifold. The type I string couples to no form ﬁeld, and of course it can break. The heterotic and type II strings are effectively global strings, because they couple to the long-distance form ﬁeld Bµν which in four dimensions is dual to a scalar. A conﬁning force is produced by instantons, which are magnetic sources for Bµν (thus the completeness of the magnetic sources [21] again enters). It is simplest to analyze this in terms of the dual scalar Θ, deﬁned by dΘ = ∗dB modulo Chern-Simons

COSMIC F- AND D-STRINGS

243

terms; Θ is the phase of the earlier Φ. The instantons become electric sources, meaning that they give rise to explicitly Θ-dependent terms in the effective action: the instanton and anti-instanton amplitudes are weighted by e±iΘ . It is an interesting exercise to work out the decay rates for the various strings. The time scale for the breaking of Type I strings is the string scale unless one tunes the string coupling essentially to zero. For the heterotic string, the one deﬁnite contribution comes from QCD instantons. The ﬁeld strength for Bµν is of the form H = dB + ω(A) ,

(5.2)

where dω(A) = Tr F 2 and the trace runs over all gauge ﬁelds. Thus the magnetic source dH contains the QCD instanton number. The scale of the potential is mu md ms ΛQCD ∼ 10−7 GeV4 ≡ Vinst . The potential energy (5.3) MP2 (∂Θ)2 + Vinst cos Θ is of order√MP2 τ −1 + Vinst τ for a domain wall of thickness τ . This is minimized for τ ∼ MP / Vinst (the inverse axion mass), giving tension Tdw ∼ MP Vinst . (5.4) The √ acceleration of a string due to the wall is then a = Tdw /µ which is roughly Vinst /MP . The string network will collapse when this is comparable to√the Hubble expansion rate, which is just when the energy density in matter is of order Vinst , near the QCD scale. This is too early to leave observable traces. Further, the estimate (5.4) is just a lower bound, because it is likely that there are other gauge groups becoming strong at higher energy. Indeed, it is a challenge to ﬁnd an scalar (axion) in string theory whose potential is dominated by QCD instantons and so would solve the strong-CP problem. NS5-brane instantons, wrapped on the whole compact space, are also a magnetic source for Bµν [76]. The NS5-brane instantons have action T5 V6 =

4π 2 α 2π V6 = . 2 2κ10 αh

(5.5)

This is the same as the action for a gauge theory instanton at the string scale. In fact the heterotic NS5-branes are just the string-scale limit of instantons, whereas above we considered QCD-scale instantons. If we take the standard GUT value αh ∼ 0.05 then the contribution to Vinst is of order 10−50 MP4 . This is larger than the QCD contribution; it might be suppressed by fermion masses (from zero modes), or enhanced if the GUT coupling is increased. The same estimates apply to perturbative (non-brane) compactiﬁcations of the Type II theory, with the gauge ﬁelds arising from world-sheet current algebras, although it is known that the Standard Model cannot be obtained from these [77]. 5.2.2. Brane models In the KLM MT model [48] we have noted that the candidate cosmic strings are the F- and D1-strings. In ten dimensions these strings couple to form ﬁelds Bµν and Cµν , but these

244

JOSEPH POLCHINSKI

form ﬁelds have no four-dimensional massless modes in this model. They transform nontrivially under the orientifold projection or F theory monodromy that the model requires, and this removes their zero modes. Thus the strings should be unstable to breakage, and indeed this can occur in several ways. First, the projection that removes the form ﬁeld produces an oppositely oriented image string on the covering space of the compactiﬁcation, and breakage occurs through a segment of the string annihilating with its image. If the image is coincident with the string, breakage will be rapid. If the image is not coincident, then the string must ﬂuctuate to ﬁnd its image. Due to the warp factor the strings feel a potential µ = e2A(y) µ10 .

(5.6)

The strings are normally conﬁned to the throat where inﬂation occurs, which as discussed in section 3 has a warp factor eA0 of order 10−4 . When the string annihilates with its image, the role of the monopole is played by a bit of string that stretches out of the inﬂationary throat over to the image throat. Since this passes through bulk region where −2A0 the warp factor is of order one, the breakage rate is proportional to e−πe and so totally negligible even on cosmological timescales [56]. Thus the strings will be stable to selfannihilation if there is no orientifold ﬁxed point (or F-theory equivalent) in the inﬂationary throat. There is no particular reason for the throat to be coincident with any ﬁxed point. Their relative positions are ﬁxed by the complex structure moduli, which depend on ﬂux integers, and these are expected to take rather generic values [78, 79, 80]. The strings can also break on a brane. The model must include branes on which the Standard Model (SM) ﬁelds live. If these are D3-branes in the inﬂationary throat then the strings will break; if they are D7-branes that pass through the inﬂationary throat then all but the D1-string will break [56]. If they are outside the throat then the strings are stable for the same reason as above. In the simplest implementation of the RS idea the SM branes must be in a different throat: the depth of the inﬂationary throat is something of order the GUT scale, while the depth of the SM throat should be of order the weak scale. One must ask whether other branes might still be in the inﬂationary throat, but this is not possible: these would have low energy degrees of freedom which would receive most of the energy during reheating, rather than the SM ﬁelds. So this gives a simple scenario in which the strings are stable, but much more work is needed to see whether it is viable, and whether it is generic. An important question is whether reheating through the bulk can transfer energy efﬁciently to the SM throat; a recent paper gives an afﬁrmative answer [81]. More complicated geometries with additional cycles in the inﬂationary throat have additional strings from wrapped branes. These strings have not been studied in full detail, but in examples one ﬁnds that they can be gauge, global, or discrete, depending on the details of the topology (both in the throat and globally). Similarly their production and stability depends on details. The stabilization of moduli is not been as fully developed in models based on large compact dimensions, but as a prototype we can consider the T 6 /Z2 orientifold which is T -dual to the Type I string. In the Type I string there are BPS D1- and D5-branes [82], which couple electrically and magnetically to the sole Type I RR form Cµν . There are also non-BPS D3- and D7-branes [54]. All of these give rise to strings when all but one direction are wrapped on the T 6 . After T -duality these become BPS D7- and D3-branes,

COSMIC F- AND D-STRINGS

245

and non-BPS D5- and D1-branes, all of which are strings in four dimensions. The BPS strings couple to RR form ﬁelds and so are global. The relevant instantons come from a Euclidean D-branes. The calculation is similar to that for the NS5-brane instanton above, with the brane volume replacing V6 , but now the compact dimensions can be much larger and so we get essentially stable strings with a long-range RR axion ﬁeld. The non-BPS strings separate into two images as in the earlier discussion. Now, it is important to note that the T 6 /Z2 example is nongeneric in an important sense: because of its high degree of symmetry the strings have moduli and move freely in the compact dimensions. Even if there is not a deep warp factor, supersymmetry breaking will lead to a potential at some scale that will localize the strings. Just like a large warping, a large separation leads to a large action for the decay and so these strings can again be stable if separated from their images. Adding the SM and other branes to the large dimension models, the stability of the strings depends on the exact geometry. In some cases it might appear that a string has both instabilities, in that it couples to a form ﬁeld but can break on a brane [56, 83]. The point is that breakage on a brane implies the existence of a gauge ﬁeld on the brane to ensure continuity of the form source, and this Higgses the form ﬁeld so that the result is a gauge string, subject to breakage but not conﬁnement [56]. We expect that in string theory as in ﬁeld theory these strings willl be exactly stable only if stabilized by a discrete gauge symmetry visible in the Aharanov-Bohm effect: anything that can happen will [21]. In summary, it is encouraging to see that strings can be stabilized as a side effect of certain generic properties such as warping and/or large dimensions, which are needed to lower the inﬂationary scale below the Planck scale in these models. 6. Seeing Cosmic Strings We will consider below strings that have only gravitational interactions. However, the possibility, noted above, of global strings with a long-ranged scalar ﬁeld should not be overlooked. Another possibility is a strong coupling to SM or other light ﬁelds, and in particular superconducting strings [11] which carry massless degrees of freedom charged under an unbroken gauge symmetry. Generically the strings that one ﬁnds are rather boring. In ten dimensions they have massless collective coordinates in the transverse directions, and massless fermions, but these are all gapped by symmetry-breaking effects leaving only the minimal collective coordinates. Light four-dimensional ﬁelds generally live on branes, and we have seen that stability requires cosmic strings to be separated from most other branes, implying interactions of gravitational strength. An example of an exception would be a D7-brane coincident with a D1 string; the SM ﬁelds might live on or couple strongly to the D7-brane. However, a superconducting string requires that the D1 and D7 be exactly coincident; to ensure this would require a discrete symmetry (which might be orbifolded). More interesting strings are thus possible in special cases. The CMB and pulsar bounds on Gµ quoted in section 2 are at the upper end of the brane inﬂation range, ruling out the highest-tension models. Both bounds will improve in the coming decade by at least one or two orders of magnitude. An additional exciting prospect comes from LIGO [84, 85]. Under most circ*mstances LIGO is at a disadvantage looking for cosmological backgrounds because these fall with increasing frequency: LIGO is looking at frequencies that are 1010 times those of the pulsar measurements

246

JOSEPH POLCHINSKI

(100 Hz versus years−1 ). For example, the stochastic background coming from the low harmonics of cosmic strings has a constant energy density per logarithmic scale.10 This 2 ˜ of the strain (fractional change ˜ ∝ ω −1 for the Fourier transform h translates into ω 2 h(ω) in the metric). For a stochastic background (different frequencies uncorrelated) the strain ˜ which is of order ω −1 here. Pulsar from approximate frequency ω is of order ω 1/2 h(ω) −14 −15 or 10 while LIGO is sensitive to much timing is sensitive to strains around 10 smaller strains 10−22 or 10−23 , but the frequency penalty of 1010 more than offsets this. However, something unexpectedly nice happens. When a loop of string in three space dimensions oscillates as governed by the Nambu action, it typically forms a cusp several times per oscillation [86]. To see this go to conformal gauge, where the solution is just a sum or right- and left-movers X µ (σ, τ ) = f µ (σ − τ ) + g µ (σ + τ ) ,

(6.1)

and the gauge conditions imply that f and g are both null. Classically there is enough residual gauge freedom to set X 0 = 2τ so that f 0 = τ − σ, g 0 = τ + σ. Then f and −g both have unit norm, and as functions of their arguments they trace out closed curves on the unit sphere as σ goes around the strings; also, each averages to zero by periodicity. It follows that these curves will typically intersect (an even number of times).11 Each intersection represents an event that occurs once per period, and when it occurs the σparameterization becomes singular. A representative example for the leading behavior at the singularity (where for simplicity we put the intersection at σ = τ = 0) is fx gx

= (σ − τ ) − c2 (σ − τ )3 /2 , f y = c(σ − τ )2 , = −(σ + τ ) + c2 (σ + τ )3 /2 , g y = c (σ + τ )2 ,

(6.2)

with c and c constants of order the inverse size of the loop. It follows (for the generic case that c = c ) that at τ = 0 the string forms a cusp, y ∝ |x|2/3 . The instantaneous velocity (f − g )/2 approaches the speed of light at the cusp. Like the crack of a whip, a great deal of energy is concentrated in the tip, but this whip is perhaps hundreds of light-years long and with a tension not so far below the Planck scale. It thus emits an intense beam of gravitational waves in the direction of its motion [84]. The Fourier transform of a cusp singularity is much larger at high frequency than for a smooth function, so that even when the suppression for being off-axis is included this comes within reach of LIGO. This is shown in Fig. 4, reproduced from [85]. Under optimistic assumptions (but not, I think, too optimistic), even LIGO I is close to discovery sensitivity of one event per year over much of the range of interesting tensions, including the narrow range of 10

In making this statement we must consider the relative redshifting of modes at different frequencies. At both the pulsar and the LIGO frequencies the current bounds come from waves emitted during the radiation dominated era, and in this case the effective red shifts are the same due to the scaling property. (One can deduce the time of emission for a current freqency ω by solving ω = l−1 (t)/(1 + z(t)) with l(t) the loop size (2.8) and z(t) the redshift.) Lower frequency waves emitted during the matter-dominated era will experience less redshifting. 11 A kink implies a discontinuity in f or g and so a gap in one of the curves. Since cosmic strings have a short distance structure with many kinks, there will be many gaps and this may reduce the number of intersections.

COSMIC F- AND D-STRINGS

247

-21 -21.5 -22 -22.5 -23 -23.5 -24 -24.5 -25

-12

-10

-8

-6

-4

Figure 4. Gravitational wave cusp signals, taken from Damour and Vilenkin [85]. The horizontal < < Gµ ∼ 10−6 becomes axis is log10 α where α = 50Gµ. Thus the brane inﬂation range 10−12 ∼ −10.3 < log10 α < −4.3. The vertical axis is log10 h where h is the gravitational strain in the LIGO frequency band. The upper and lower dashed horizontals are the sensitivities of LIGO I and Advanced LIGO at one event per year. The upper two curves are the cusp signal under optimistic and pessimistic network assumptions; the pessimism is that a large number of kinks may suppress the cusps. The lowest solid curve is the signal from kinks, which form whenever strings reconnect. The dashed curve is the stochastic signal.

the KLM MT model. This is remarkable: cosmic superstrings might be the brightest objects in gravitational wave astronomy, and the ﬁrst discovered! LIGO I to date has around 0.1 design-year of data, but it is supposed to begin a new science run in January 2005 at close to design sensitivity and with a good duty cycle. Advanced LIGO is sensitive over almost the whole range, and with a higher event rate; it has a target date of 2009. LISA, a few years later, is even more sensitive. In magnitude of strain it is comparable to LIGO I, but it is looking at a frequency 10,000 times lower and so the typical cusp strains are 1000 times greater [85]. The cusp events might be seen in a search for unmodeled bursts. The shape is not as complex as for stellar and black hole inspirals, but modeling the speciﬁc frequency dependence will increase the signal-to-noise ratio. The power-law frequency dependence of the cusp is distinctive. See also [87] for a discussion of the form of the cusp signal. The dependence of the sensitivity on string tension in Fig. 4 is interesting because it is not monotonic. This comes about as follows [85]. As the string tension decreases, the coupling to gravity becomes weaker and so does the gravitational wave burst from a given cusp. However, since gravitational radiation is the only decay channel for string loops they will live longer.12 Thus as we decrease Gµ there are more but smaller cusps. With smaller intrinsic events one cannot see as far, but there are more events to see; the competition between these effects depends on the geometry of the universe. Thus, the three regimes 12 There is another effect as well, the dependence of the short-distance network structure on the gravitational radiation, but we can overlook this for simplicity.

248

JOSEPH POLCHINSKI

that are evident in Fig. 4 correspond at the smallest tensions to cusps that took place at redshifts less than one, at the intermediate tensions to cusps that are at redshifts greater than one but in the matter-dominated era, and at the largest tensions to cusps that occurred during the radiation-dominated era. The rise in the event rate with decreasing tension in the middle range comes about because the signals from smaller, later, cusps suffer less from redshifting during the relatively rapid matter-dominated expansion. The pulsar bounds are also strengthed by taking into account the cusps. This is analyzed in detail in [88]; this quotes a current bound on Gµ of around 10−7 , but the curve is very ﬂat as a function of Gµ so that a small uncertainty in the pulsar analysis or the network properties produces a much larger uncertainty in Gµ. Correspondingly, improved pulsar data and improved understanding of the networks would make it possible to reach much smaller values of Gµ. One may wonder whether the enormous energy stored in cosmic strings can manifest itself in other spectacular effects. The cusps appear to be the best source potential source of high energy particles and radiation, but [89] argues that this is still not observable. Note that short distance physics affects the string only very near the cusp, and even LIGO frequency range is looking on a much longer scale. 7. Distinguishing Superstrings If gravitational wave cusps or some other signature of cosmic strings are seen, this will be just the beginning of the story. Detailed observations will be able to determine some of the microscopic properties of the strings. For example, it is possible to cleanly distinguish weakly coupled fundamental strings from cosmic strings in perturbative GUTs. 7.1. RECONNECTION The microscopic structure of the string core does not affect the evolution of strings that are light-years in length, except when two strings cross and their cores interact. We have noted that gauge theory solitons will always reconnect. For F-strings, reconnection is a quantum process, and takes place with a probability P of order gs2 . The numerical factors are worked out in [90]. To be precise, P is a function of the relative angle and velocity in the collision, but it is simplest to the value averaged over collision parameters. An important issue is the motion of the string in the compact dimensions. For many supersymmetric compactiﬁcations, strings can wander over the whole compact space. Thus they can miss each other, leading to a substantial suppression of P [51, 55]. However, we have noted that in realistic compactiﬁcations strings will always be conﬁned by a potential in the compact dimensions. Even if the scale of the potential is low, the ﬂuctuations of the strings are only logarithmic in the ratio of scales — this is characteristic of one-dimensional objects [56]. Thus there is no suppression by powers of the size of the compact dimensions, but the logarithm can be numerically important — it tends to offset powers of π that appear in the numerator. The value of gs , and the scale of the conﬁning < P < 1. For D-D colpotential, are not known, but in a variety of models [90] ﬁnds 10−3 ∼ ∼ < P < 1. lisions the situation is more complicated, and in the same models one ﬁnds 10−1 ∼ ∼ For F-D collisions, P can vary from 0 to 1.

COSMIC F- AND D-STRINGS

249

To determine the observational effect of changing P one must feed the given value into the network evolution. A simple argument suggests that the signatures scale as 1/P : the amount of string in the network must be increased by this factor in order for an increased number of collisions per unit length of string to offset the reduced P in each collision ([88] and references therein). This is a bit oversimpliﬁed, because there are issues connected with the sub-horizon scale structure in the string network [16, 91] that can work in either direction, but is supported by simulations as well ([92] and reference therein).13 If we take this 1/P as a model, we see that for the smaller values of P discussed above there can be a substantial increase in the signal even above the encouraging values found in the last section, so that LIGO might soon begin to see many cusps. Of course, the existing bounds become stronger. To ﬁrst approximation there are two relevant parameters, µ and P . Each individual cusp event has only a single parameter to measure, its overall strength h: because it is a power law there is no characteristic frequency scale. (There is a high frequency cutoff, determined by the alignment of the cusp with the detector [85], but this gives no information about the cusp itself.) After O(10) cusps are seen one can begin to plot a spectrum, dN ∼ Ah−B dh, and from the two parameters A and B constrain µ and P . There are degeneracies — B depends primarily on the epoch in which the cusp took place — but with a more detailed spectrum, and ultimately with data from kink events and pulsars, this degeneracy will be resolved. Thus µ and P will be overdetermined, and nonstandard network behavior (such as we are about to discuss) will be detectable. If P is only slightly less than one, say 0.5, then it will require precision simulation of the networks and good statistics on the signatures to distinguish this from 1.0. It should be noted that even with given values of µ and P there are still substantial uncertainties in the current understanding of the behavior of string networks. This has recently been discussed in [88], which concludes that the sensitivities given in Fig. 3 are only weakly dependent on the unknowns. The discussion of cosmic strings from grand uniﬁed theories has focussed on perturbative uniﬁcation, as suggested by the successful prediction of the Weinberg angle. In this case the only strings are the classical solitons, and P = 1 is a fairly robust signature. Thus these can be differentiated from fundamental strings unless we are very unlucky and gs is very close to 1, or we have a very unusual ﬁeld theory.14 Strongly coupled conﬁning gauge theories can have electric ﬂux tube strings. For 2 , so ﬁnding a small value for P these the reconnection probability is of order 1/Ncolor would not rule out a ﬁeld-theoretic origin entirely, just a perturbative one. We should note that perturbative string theory gives a prediction for the functional dependence on the collision parameters, P (v, θ) [95, 90]. It would seem very difﬁcult to determine this from observations. If it were possible to map out the string network in detail, through lensing, then perhaps this function might be studied. 13

Refs. [93, 94] discuss other network evolution issues. We should note that the second of these is concerned with higher-dimensional excitations which, according to our discussion, should be massive and so decoupled except at early times. 14 One example: a soliton string with a massless internal scalar mode (besides its normal collective coordinates) would be like a string moving in a higher dimensional space and so two such strings could ‘miss’ each other in ﬁeld space. However, such modes will always get a mass from symmetry breaking, and I expect that this effect will be signiﬁcant only in rather contrived models.

250

JOSEPH POLCHINSKI

7.2. NETWORKS The second potentially distinguishing feature of the superstring networks is the existence of both F- and D-strings [55, 56, 90], and moreover bound states of p F-strings and q D-strings with the distinctive tension formula (3.8). In this case, when strings of different types collide, rather than reconnecting they form more complicated networks with trilinear vertices. It is then possible that the network does not scale, but gets into a frozen phase where it just stretches with the expansion of the universe [10, 96]. If so, its density would come to dominate at the tensions that we are considering. The F-D networks have not yet been simulated, but simulations of comparable networks suggest that they scale, possibly with an enhanced density of strings [97, 98, 99]. From the discussion above, it follows that one will not directly read off the spectrum (3.8) from the observations, but there should eventually be enough information to distinguish F-D string networks from other types. Warped models with more complicated throats, as well as models with large compact dimensions, can give rise to a richer spectrum of strings [3, 51]. Still, the (p, q) spectrum (3.8) is worth particular attention: it is an attractive possibility that inﬂation takes place in a warped throat, and this is the spectrum that arises in the most generic throat. Thus, while the landscape of string theory may be vast, this particular local geometric feature may be common to a large swath of it. Networks with multiple types of string can also arise in ﬁeld theories, though I do not know any perturbative ﬁeld theory that gives the particular spectrum (3.8). However, because of duality there will be gauge theory strings that are very hard to differentiate from the F- and D-strings that we are discussing. In particular, in the KLM MT model the strings exist in a Klebanov-Strassler throat [2], which has a dual description as a cascading gauge theory, in which there are (p, q) strings.15 Indeed, the existence of dualities between string theories and ﬁeld theories raises the issue, what really is string theory? This is beyond our current scope, but I note that in the present case there is a quantitative question. The Klebanov-Strassler throat has a parameter gs M ; when this is large the string description is the valid one, and when it is small the gauge description is the valid one. If we can get enough information about the string network, perhaps combined with information from the CMB, then there might well be enough information to test the hypothesis that we are seeing a Klebanov-Strassler theory, and to measure gs M . 8. Conclusions As we have seen, each of the four conditions that we discussed at the beginning is independently model-dependent. However, quite a number of things have worked out surprisingly well: the production of strings in brane inﬂation, the possible stabilization of the strings as a side effect of other properties of the models (in particular, of the stabilization of the vacuum itself), the possibility to see strings over many interesting orders of magnitude of tension, and the existence of properties that distinguish different kinds of string so that after the strings are discovered we can do a lot of science with them. In any case, searching 15 For discussions of relations between ﬁeld theory strings and F- and D-strings see [100, 101, 102, 103, 104, 105, 106, 107, 108]; see [109, 110] for some related recent bounds.

COSMIC F- AND D-STRINGS

251

for cosmic strings is a tiny marginal cost on top of experiments that will already be done, and it is great that string theorists will have a stake in these experiments over the coming decade and more. Acknowledgements I would like to thank E. Copeland, M. Jackson, N. Jones, and R. Myers for collaborations, and N. Arkani-Hamed, L. Bildsten, G. Dvali, A. Filippenko, T. Kibble, A. Linde, A. Lo, A. Lommen, J. Preskill, S. Trivedi, H. Tye, T. Vachaspati, and A. Vilenkin for discussions and communications. This work was supported by National Science Foundation grants PHY99-07949 and PHY00-98395. 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. 27. 28. 29. 30. 31. 32. 33. 34.

E. Witten, Phys. Lett. B153, 243 (1985). N. Jones, H. Stoica, and S. H. H. Tye, JHEP 07, 051 (2002), hep-th/0203163. S. Sarangi and S. H. H. Tye, Phys. Lett. B536, 185 (2002), hep-th/0204074. J. Polchinski, arXiv:hep-th/0410082. A. Vilenkin and E. Shellard, Cosmic strings and other topological defects (Cambridge Univ. Press 1994). M. B. Hindmarsh and T. W. B. Kibble, Rept. Prog. Phys. 58, 477 (1995), hep-ph/9411342. T. W. B. Kibble, arXiv:astro-ph/0410073. A. A. Abrikosov, Sov. Phys. JETP 5, 1174 (1957). H. B. Nielsen and P. Olesen, Nucl. Phys. B61, 45 (1973). T. W. B. Kibble, J. Phys. A9, 1387 (1976). E. Witten, Nucl. Phys. B249, 557 (1985). E. P. S. Shellard, Nucl. Phys. B283, 624 (1987). R. A. Matzner, Computers in Physics 2, 51 (1988). K. J. M. Moriarty, E. Myers, and C. Rebbi, Phys. Lett. B207, 411 (1988). A. Albrecht and N. Turok, Phys. Rev. D 40, 973 (1989). D. P. Bennett and F. R. Bouchet, Phys. Rev. D 41, 2408 (1990); PUPT-90-1162, Based on Invited talks given at Cosmic String Workshop, Cambridge, England, Jul 2-7, 1989. B. Allen and E. P. S. Shellard, Phys. Rev. Lett. 64, 119 (1990). X. Siemens, K. D. Olum and A. Vilenkin, Phys. Rev. D 66, 043501 (2002) [arXiv:grqc/0203006]. Y. B. Zeldovich and M. Y. Khlopov, Phys. Lett. B 79, 239 (1978). J. P. Preskill, Phys. Rev. Lett. 43, 1365 (1979). J. Polchinski, Int. J. Mod. Phys. A19S1, 145 (2004), hep-th/0304042. A. H. Guth, Phys. Rev. D 23, 347 (1981). Y. B. Zeldovich, Mon. Not. Roy. Astron. Soc. 192, 663 (1980). A. Vilenkin, Phys. Rev. Lett. 46, 1169 (1981). L. Pogosian, S. H. H. Tye, I. Wasserman, and M. Wyman, Phys. Rev. D68, 023506 (2003), hep-th/0304188. L. Pogosian, M. C. Wyman, and I. Wasserman, astro-ph/0403268. E. Jeong and G. F. Smoot, astro-ph/0406432. A. Lo and E. Wright, in preparation (unpublished). V. M. Kaspi, J. H. Taylor and M. F. Ryba, Astrophys. J. 428, 713 (1994). S. E. Thorsett and R. J. Dewey, Phys. Rev. D 53, 3468 (1996). M. P. McHugh, G. Zalamansky, F. Vernotte and E. Lantz, Phys. Rev. D 54, 5993 (1996). A. N. Lommen, astro-ph/0208572. M. Sazhin et al., Mon. Not. Roy. Astron. Soc. 343, 353 (2003), astro-ph/0302547. M. V. Sazhin et al., (2004), astro-ph/0406516.

252 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. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81.

JOSEPH POLCHINSKI R. E. Schild, I. S. Masnyak, B. I. Hnatyk, and V. I. Zhdanov, (2004), astro-ph/0406434. J. Polchinski, String theory. Vol. 2: Superstring theory and beyond (Cambridge, UK: Univ. Pr., ADDRESS, 1998), p. 531. E. Witten, Nucl. Phys. B471, 135 (1996), hep-th/9602070. N. Arkani-Hamed, S. Dimopoulos, and G. R. Dvali, Phys. Lett. B429, 263 (1998), hepph/9803315. I. Antoniadis, N. Arkani-Hamed, S. Dimopoulos, and G. R. Dvali, Phys. Lett. B436, 257 (1998), hep-ph/9804398. L. Randall and R. Sundrum, Phys. Rev. Lett. 83, 3370 (1999), hep-ph/9905221. S. B. Giddings, S. Kachru, and J. Polchinski, Phys. Rev. D66, 106006 (2002), hep-th/0105097. G. R. Dvali and S. H. H. Tye, Phys. Lett. B450, 72 (1999), hep-ph/9812483. S. H. S. Alexander, Phys. Rev. D65, 023507 (2002), hep-th/0105032. C. P. Burgess et al., JHEP 07, 047 (2001), hep-th/0105204. G. R. Dvali, Q. Shaﬁ, and S. Solganik, hep-th/0105203. S. Kachru, R. Kallosh, A. Linde and S. P. Trivedi, Phys. Rev. D 68, 046005 (2003) [arXiv:hepth/0301240]. E. Silverstein, arXiv:hep-th/0405068. S. Kachru et al., JCAP 0310, 013 (2003), hep-th/0308055. I. R. Klebanov and M. J. Strassler, JHEP 08, 052 (2000), hep-th/0007191. P. Candelas and X. C. de la Ossa, Nucl. Phys. B 342, 246 (1990). N. T. Jones, H. Stoica, and S. H. H. Tye, Phys. Lett. B563, 6 (2003), hep-th/0303269. R. Jeannerot, J. Rocher and M. Sakellariadou, Phys. Rev. D 68, 103514 (2003) [arXiv:hepph/0308134]. A. Sen, JHEP 09, 023 (1998), hep-th/9808141. E. Witten, JHEP 12, 019 (1998), hep-th/9810188. G. Dvali and A. Vilenkin, JCAP 0403, 010 (2004), hep-th/0312007. E. J. Copeland, R. C. Myers, and J. Polchinski, JHEP 06, 013 (2004), hep-th/0312067. N. Barnaby, A. Berndsen, J. M. Cline and H. Stoica, arXiv:hep-th/0412095. N. Iizuka and S. P. Trivedi, Phys. Rev. D 70, 043519 (2004) [arXiv:hep-th/0403203]. A. D. Linde, Phys. Rev. D49, 748 (1994), astro-ph/9307002. J. Yokoyama, Phys. Rev. Lett. 63, 712 (1989). L. Kofman, A. D. Linde, and A. A. Starobinsky, Phys. Rev. Lett. 76, 1011 (1996), hepth/9510119. I. Tkachev, S. Khlebnikov, L. Kofman, and A. D. Linde, Phys. Lett. B440, 262 (1998), hepph/9805209. J. Urrestilla, A. Achucarro, and A. C. Davis, Phys. Rev. Lett. 92, 251302 (2004), hepth/0402032. T. Watari and T. Yanagida, Phys. Lett. B589, 71 (2004), hep-ph/0402125. K. Dasgupta et al., JHEP 08, 030 (2004), hep-th/0405247. E. Halyo, hep-th/0402155. T. Matsuda, Phys. Rev. D70, 023502 (2004), hep-ph/0403092. T. Matsuda, hep-ph/0406064. F. Englert, J. Orloff and T. Piran, Phys. Lett. B 212, 423 (1988). S. W. Hawking and D. N. Page, Commun. Math. Phys. 87, 577 (1983). E. Witten, Adv. Theor. Math. Phys. 2, 505 (1998) [arXiv:hep-th/9803131]. J. Preskill and A. Vilenkin, Phys. Rev. D47, 2324 (1993), hep-ph/9209210. T. Banks and L. J. Dixon, Nucl. Phys. B307, 93 (1988). A. Vilenkin and A. E. Everett, Phys. Rev. Lett. 48, 1867 (1982). S. Coleman, Aspects of Symmetry (Cambridge Univ. Press 1988). C. G. . Callan, J. A. Harvey and A. Strominger, Nucl. Phys. B 359, 611 (1991). L. J. Dixon, V. Kaplunovsky and C. Vafa, Nucl. Phys. B 294, 43 (1987). R. Bousso and J. Polchinski, JHEP 06, 006 (2000), hep-th/0004134. M. R. Douglas, JHEP 05, 046 (2003), hep-th/0303194. S. Ashok and M. R. Douglas, JHEP 01, 060 (2004), hep-th/0307049. N. Barnaby, C. P. Burgess and J. M. Cline, arXiv:hep-th/0412040.

COSMIC F- AND D-STRINGS 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.

253

J. Polchinski and E. Witten, Nucl. Phys. B 460, 525 (1996) [arXiv:hep-th/9510169]. L. Leblond and S. H. H. Tye, JHEP 03, 055 (2004), hep-th/0402072. T. Damour and A. Vilenkin, Phys. Rev. Lett. 85, 3761 (2000), gr-qc/0004075. T. Damour and A. Vilenkin, Phys. Rev. D64, 064008 (2001), gr-qc/0104026. N. Turok, Nucl. Phys. B242, 520 (1984). X. Siemens and K. D. Olum, Phys. Rev. D68, 085017 (2003), gr-qc/0307113. T. Damour and A. Vilenkin, arXiv:hep-th/0410222. J. J. Blanco-Pillado and K. D. Olum, Phys. Rev. D 59, 063508 (1999) [arXiv:gr-qc/9810005]. M. G. Jackson, N. T. Jones, and J. Polchinski, (2004), hep-th/0405229. D. Austin, E. J. Copeland, and T. W. B. Kibble, Phys. Rev. D48, 5594 (1993), hep-ph/9307325. M. Sakellariadou, arXiv:hep-th/0410234. C. J. A. Martins, Phys. Rev. D 70, 107302 (2004) [arXiv:hep-ph/0410326]. A. Avgoustidis and E. P. S. Shellard, arXiv:hep-ph/0410349. J. Polchinski, Phys. Lett. B209, 252 (1988). A. Vilenkin, Phys. Rev. Lett. 53, 1016 (1984). T. Vachaspati and A. Vilenkin, Phys. Rev. D35, 1131 (1987). D. Spergel and U.-L. Pen, Astrophys. J. 491, L67 (1997), astro-ph/9611198. P. McGraw, Phys. Rev. D57, 3317 (1998), astro-ph/9706182. K. Becker, M. Becker and A. Strominger, Phys. Rev. D 51, 6603 (1995) [arXiv:hepth/9502107]. J. D. Edelstein, C. Nunez and F. A. Schaposnik, Nucl. Phys. B 458, 165 (1996) [arXiv:hepth/9506147]. G. Dvali, R. Kallosh, and A. Van Proeyen, JHEP 01, 035 (2004), hep-th/0312005. E. Halyo, JHEP 03, 047 (2004), hep-th/0312268. P. Binetruy, G. Dvali, R. Kallosh, and A. Van Proeyen, Class. Quant. Grav. 21, 3137 (2004), hep-th/0402046. S. S. Gubser, C. P. Herzog, and I. R. Klebanov, (2004), hep-th/0405282. A. Achucarro and J. Urrestilla, JHEP 08, 050 (2004), hep-th/0407193. S. S. Gubser, C. P. Herzog, and I. R. Klebanov, (2004), hep-th/0409186. A. Lawrence and J. McGreevy, (2004), hep-th/0409284. J. Rocher and M. Sakellariadou, arXiv:hep-ph/0405133. J. Rocher and M. Sakellariadou, arXiv:hep-ph/0412143.

LONG TIME SCALES AND ETERNAL BLACK HOLES

´ JOSE L. F. BARBON Instituto de F´ısica Te´orica IFT UAM/CSIC Facultad de Ciencias C-XVI C.U. Cantoblanco, E-28049 Madrid, Spain AND ELIEZER RABINOVICI Racah Institute of Physics, The Hebrew University Jerusalem 91904, Israel

Abstract. We discuss the various scales determining the temporal behaviour of correlation functions in the presence of eternal black holes. We point out the origins of the failure of the semiclassical gravity approximation to respect a unitarity-based bound suggested by Maldacena. We ﬁnd that the presence of a subleading (in the large-N approximation involved) master ﬁeld does restore the compliance with one bound but additional conﬁgurations are needed to explain the more detailed expected time dependence of the Poincar´e recurrences and their magnitude.

1. Introduction Hawking’s semiclassical analysis of black hole evaporation suggests that most of the information contained in initial scattering states is shielded behind the event horizon, never to return back to the asymptotic region far from the evaporating black hole [1]. In this picture, the singularity is capable of absorbing all the infalling information, which is then destroyed or transmitted to other geometrical realms, depending on one’s hypotheses about the microphysics of the singularity. From the point of view of measurements on the Hawking radiation, the evaporation is not described by a unitary S-matrix. Rather, quantum coherence is violated and the linear evolution in Hilbert space takes pure states into mixed states. Still, probability is conserved, since density matrices ρ remain Hermitian, ρ† = ρ, positive, ρ > 0 and normalized, Trρ = 1 under time evolution. The AdS/CFT correspondence [2] is not consistent with this picture. In this construction, quantum gravity in a (d + 1)-dimensional asymptotically Anti-de Sitter spacetime (AdS) of curvature radius R is deﬁned in terms of a conformal ﬁeld theory (CFT) on a spatial sphere Sd−1 of radius R. The effective expansion parameter in the gravity side 1/N 2 ∼ GN /Rd−1 , maps to an appropriate large N limit of the CFT. For example, for 255 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 255–263. © 2005 Springer. Printed in the Netherlands.

256

´ AND ELIEZER RABINOVICI JOSE L. F. BARBON

Figure 1. The energy spectrum of a CFT representing AdSd+1 quantum gravity. The spectrum is discrete on a sphere of radius R, with gap of order 1/R. The asymptotic energy band of very dense “black hole” states sets in beyond energies of order N 2 /R. The corresponding density of states is that of a conformal ﬁxed point in d spacetime dimensions.

two-dimensional CFT’s N 2 is the central charge. When the CFT is a gauge theory, the AdS side is a string theory, N is the rank of the gauge group, and the string perturbative expansion in powers of gs ∼ 1/N is identiﬁed with ’t Hooft’s 1/N expansion in the gauge theory side. According to this deﬁnition, the formation and evaporation of small black holes with Schwarschild radius RS R, should be described by a unitary process in terms of the CFT Hamiltonian. Thus, there is no room for violations of coherence as a matter of principle. Unfortunately, the CFT states corresponding to small black holes are hard to describe, and it remains a challenge to put the ﬁnger on the precise error in Hawking’s semiclassical analysis in that case. For large AdS black holes with Schwarschild radius RS R one may attempt to rise to the challenge, since they are thermodynamically stable and can exist in equilibrium at ﬁxed (high) temperatures 1/β 1/R. Indeed, the corresponding Bekenstein–Hawking entropy scales like that of N 2 conformal degrees of freedom at high energy, √ d−1 (1.1) S ∼ N (E R) d ∼ N 2 (R/β)d−1 . Therefore, large AdS black holes with inverse Hawking temperature β R describe the leading approximation to the thermodynamical functions of the canonical CFT state ρβ =

e−βH , Z(β)

Z(β) = Tr exp(−βH) .

(1.2)

This suggests that we can test the semiclassical unitarity argument by careful analysis of slight departures from thermal equilibrium, rather than studying a complete evaporation instability in the vacuum. Ref. [3] proposes to look at the very long time structure of correlators of the form G(t) = Tr [ ρ A(t) A(0) ] , (1.3)

LONG TIME SCALES AND ETERNAL BLACK HOLES

257

for appropriate Hermitian operators A. In the semiclassical approximation, one expects these correlators to decay as exp(−Γ t) with Γ ∼ β −1 . However, because the CFT lives in ﬁnite volume, the spectrum is actually discrete (c.f. Fig 1), and the correlator must show nontrivial long time structure in the form of Poincar´e recurrences (see [4, ?]). This result, which is straightforward from the boundary theory point of view, has far reaching consequences as far as the bulk physics is concerned. Hence, the failure of G(t) to vanish as t → ∞ can be used as a criterion for unitarity preservation beyond the semiclassical approximation. This argument can be made more explicit by checking the effect of coherence loss on the long-time behaviour of G(t). Using the results of [6] one can simulate the required decoherence by coupling an ordinary quantum mechanical system to a random classical noise. It is then shown in [7] that this random noise forces G(t) to decay exponentially for large t, despite having a discrete energy spectrum. This shows that the long-time behaviour of correlators probes the strict quantum coherence of the bounded system. At the same time, one would like to identify what kind of systematic corrections to the leading semiclassical approximation are capable of restoring unitarity. A proposal was made in [3] in terms of topology-changing ﬂuctuations of the AdS background. Our purpose here is to investigate these questions and offer an explicit estimate of the instanton effects suggested in [3] (see also [8]). Ultimately, this analysis should provide information about the nature of the black hole singularity. 2. Long-time details of thermal quasi-equilibrium Poincar´e recurrences occur in general bounded systems. Classically they follow from the compactness of available phase space, plus the preservation of the phase-space volume in time (Liouville’s theorem). Quantum mechanically, they follow from discreteness of the energy spectrum (characteristic of spatially bounded systems) and unitarity, since Gβ (t) =

1 −βEi e |Aij |2 ei(Ei −Ej )t Z(β) i,j

(2.1)

deﬁnes a quasiperiodic function of time (we have chosen the canonical density matrix for the initial state). After initial dissipation on a time scale Γ−1 , where Γ measures the approximate width of matrix elements of A in the energy basis, the correlator will show O(1) “resurgences” when most of the relevant phases complete a period (c.f. Fig 2). The associated time scale is tH ≡ 1/ω, with ω = Ei − Ej an average frequency in (2.1). We can estimate ω as Γ/∆nΓ , where ∆nΓ is the number of energy levels in the relevant band of width Γ. Introducing the microcanonical entropy in terms of the level-number function as n(E) ≡ exp S(E), we have E+Γ/2 E+Γ/2 dn = ∆nΓ ≈ dE dE β(E) eS(E) ≈ Γ β eS(β) . (2.2) dE E−Γ/2 E−Γ/2 where we have introduced the microcanonical inverse temperature as β(E) ≡ dS/dE. From this analysis we obtain an estimate tH ∼ β eS(β) .

(2.3)

´ AND ELIEZER RABINOVICI JOSE L. F. BARBON

258 L(t) 1

exp (− Γ t )

Γ

−1

t

2 t

H

H

t

Figure 2. Schematic representation of the very long time behaviour of the normalized time correlator L(t) in bounded systems. The initial decay with lifetime of order Γ−1 is followed by O(1) “resurgences” after the Heisenberg time tH ∼ β exp(S) has elapsed. Poincar´e recurrence times can be deﬁned by demanding the resurgences to approach unity with a given a priori accuracy, and scale like a double exponential of the entropy.

Following [9] we call this the Heisenberg time scale. Poincar´e times can be deﬁned in terms of quasiperiodic returns of Gβ (t) with a given a priori accuracy. In a sense, the Heisenberg time is the smallest possible Poincar´e time. A more quantitative criterion can be used by deﬁning a normalized positive correlator, L(t), satisfying L(0) = 1, and its inﬁnite time average, G(t) 2 1 T , L ≡ lim dt L(t) . (2.4) L(t) ≡ T →∞ T 0 G(0) The proﬁle of L(t) is sketched in Fig 2. The time average can be estimated by noticing that the graph of L(t) features positive “bumps” of height ∆L and width Γ, separated a time tH , so that ∆L L∼ . (2.5) Γ tH For the case at hand ∆L ∼ 1, tH ∼ β eS , and we obtain (c.f. [5, 7]) L∼

e−S(β) . βΓ

(2.6)

Since both β and Γ scale as N 0 in the large-N limit of the dual CFT, the “recurrence index” L ∼ exp(−N 2 ) scales as a nonperturbative effect in the semiclassical approximation. Indeed, one ﬁnds L = 0 in gravity perturbation theory in the AdS black hole background. The reason is that the relevant eigenfrequencies ω (the so-called normal modes of the black hole) form a continuous spectrum to all orders in the 1/N expansion. For a static metric of the form ds2 = −g(r) dt2 +

dr2 + r2 dΩ2d−2 , g(r)

(2.7)

LONG TIME SCALES AND ETERNAL BLACK HOLES V

eff

259

V eff

exp (4 π r / β ) * r

−oo

r

r

*

r

*

r

* = π R/2

*

= 0

r

*

*

= π R /2

Figure 3. The effective potential determining the semiclassical normal frequency modes in a large AdS black hole background (left). In Regge–Wheeler coordinates the horizon is at r∗ = −∞, whereas the boundary of AdS is at r∗ = πR/2 (only the region exterior to the horizon appears). There is a universal exponential behaviour in the near-horizon (Rindler) region. The effective one-dimensional Schr¨odinger problem represents a semi-inﬁnite barrier with a continuous energy spectrum. This contrasts with the analogous effective potential in vacuum AdS with global coordinates (right). The domain of r∗ is compact and the spectrum of normal modes is discrete with gap of order 1/R.

the normal frequency spectrum follows from the diagonalization of a radial Schr¨odinger operator ω2 = − with Veﬀ

d−2 g(r) = 2

d2 + Veﬀ (r∗ ) , dr∗2

g (r) d − 4 ∇2Ω 2 + g(r) + g(r) − 2 + m r 2r2 r

(2.8)

(2.9)

for a scalar ﬁeld of mass m (analogous potentials can be deduced for higher spin ﬁelds). Here we have deﬁned the Regge–Wheeler or “tortoise” coordinate dr∗ = dr/g(r). We have shown in Fig. 3 the form of the resulting effective potentials for large AdS black holes, compared with the case of the vacuum AdS manifold. The vacuum AdS manifold, corresponding to the choice g(r) = 1 + r2 /R2 in (2.7), behaves like a ﬁnite cavity, as expected. The distinguishing feature of the black-hole horizon is a a non-degenerate zero, g(r0 ) = 0, which induces the universal scaling Veﬀ (r∗ ) ∝ exp(4πr∗ /β)

as r∗ → −∞ ,

(2.10)

with 1/β = g (r0 )/4π the Hawking temperature and the horizon r = r0 appearing at r∗ = −∞. The spectrum is discrete in pure AdS, and continuous in the AdS black hole. Physically, this just reﬂects the fact that the horizon is an inﬁnite redshift surface, so that we can store an arbitrary number of modes with ﬁnite total energy, provided they are sufﬁciently red-shifted by approaching the horizon [10]. Since the thermal entropy of perturbative gravity excitations in the vacuum AdS spacetime scales as S(β)AdS ∼ N 0 , we see that the perturbative Heisenberg time of the AdS spacetime is of O(1) in the large-N limit, leading to LAdS = O(1). On the other hand, we have Lbh = 0 in this approximation.

260

´ AND ELIEZER RABINOVICI JOSE L. F. BARBON

Figure 4. Summing over large-scale ﬂuctuations of the thermal ensemble in which a black hole spontaneously turns into radiation (and viceversa) is represented in the Euclidean formalism as the coherent sum of thermal saddle points of different topology. The “cigar-like” geometry X represents the black-hole master ﬁeld (in the CFT language) and the cylindrical topology Y represents the thermal gas of particles.

3. Topological diversity and unitarity It is instructive to understand these perturbative results in the Euclidean formalism, obtained by t = −iτ in (2.7), followed by an identiﬁcation τ ≡ τ + β. The resulting metric for the vacuum AdS spacetime has a non-contractible S1 given by the τ compact direction. We call Y this Euclidean manifold. On the other hand, the black hole spacetime with g(r0 ) = 0 has different topology, since the thermal S1 shrinks to zero size at r = r0 . The choice 1/β = g (r0 )/4π ensures smoothness at r = r0 . We call this Euclidean black hole manifold X. The real-time correlation functions in the black hole background, G(t)X , follow by analytic continuation from their Euclidean counterparts. Since X is a completely smooth manifold in the 1/N expansion, so is the Euclidean correlator G(it)X for t = 0. The continuous spectrum arising in the spectral decomposition of G(t)X is a consequence of the contractible topology of X, since the Hamiltonian folliation by τ = constant surfaces is singular at r = r0 . Therefore, it seems that improving on the semiclassical prediction for L requires some sort of topology-change process. The proposal of [3] is precisely that: instead of evaluating the semiclassical correlators on X, one should sum coherently the contribution of X and Y . Normally one neglects the contribution of Y on a quantitative basis (at high temperatures R β). However, here the contribution of X to L vanishes and one is forced to consider the ﬁrst correction. Since Y has a discrete spectrum in perturbation theory, the net result for L should be non-vanishing in

LONG TIME SCALES AND ETERNAL BLACK HOLES

261

L(t) 1

2

exp (− N

)

tc

t + t (Y) c H

t

Figure 5. The instanton approximation to the correlator L(t)inst features the expected exponential decay exp(−Γ t) induced by the contribution of the X-manifold, whereas the resurgences are entirely due to the interference with the Y -manifold, leading to small bumps of order exp(−2∆I) ∼ exp(−N 2 ), separated a time tH (Y ) ∼ N 0 . These bumps are noticeable against the background of the X-manifold after a time tc ∼ ∆I/Γ.

this approximation. Physically, this superposition of Euclidean saddle points (or master ﬁelds, in the language of the CFT) corresponds to large-scale ﬂuctuations in which the AdS black hole is converted into a graviton gas at the same temperature and viceversa. A more detailed estimate of this “instanton” approximation to L yields (c.f. [7]) Linst ≈ C e−2∆I ,

(3.1)

where C = O(N 0 ), ∆I = IY − IX and I = − log Z(β), calculated in the classical gravity approximation. Since IY ∼ −N 0 and IX ∼ −N 2 , the exponential suppression factor is of order exp(−2|IX |) ∼ exp(−N 2 ), reproducing the expected scaling (2.6), at least in order of magnitude (however, in general SX = −2|IX |, even at high temperature). However, the apparent success of (3.1) turns out to be somewhat coincidental. If we consider the full time proﬁle of L(t) rather than the inﬁnite time average, we ﬁnd L(t)inst ≈ L(t)X + C e−2∆I L(t)Y .

(3.2)

The resulting structure is shown in Figs. 5 and 6. The instanton approximation to the normalized correlator features the normal dissipation with lifetime Γ−1 ∼ β coming from the contribution of X. However, the resurgences are controlled by L(t)Y , damped by a factor exp(−2∆I) ∼ exp(−N 2 ), and separated a time tH (Y ) ∼ N 0 . Hence, the very long time behaviour as shown in Fig. 6 is very different from the expected one, although the inﬁnite time average comes out right in order of magnitude: 2

1 ∆L e−N ∼ L∼ ∼ . Γ tH Γ·β Γ · β eN 2

(3.3)

We can also ﬁnd the time scale tc for which the large-scale instantons considered here are quantitatively important on the graph of L(t). This is shown in Fig. 5 and yields tc ∼ ∆I/Γ ∼ N 2 .

´ AND ELIEZER RABINOVICI JOSE L. F. BARBON

262 L(t) 1

2 e −N

t

H

(Y)

t

t H

Figure 6. Schematic representation of the very long time behaviour of L(t)inst (dark line) compared to the expected pattern for the exact quantity L(t). The resurgences of L(t)inst occur with periods of order tH (Y ) = O(N 0 ) and have amplitude of order exp(−N 2 ) 1. The expectations for the exact CFT, in the dashed line, are O(1) resurgences with a much larger period tH ∼ exp(N 2 ) tH (Y ), corresponding to tiny energy spacings of order exp(−N 2 ). Despite 2 the gross difference of both proﬁles, the inﬁnite time average is O(e−N ) for both of them.

A more complicated set of Euclidean saddle points can be analyzed for the threedimensional case of BTZ black holes. The authors of [11] conclude that resummation of an inﬁnite family of SL(2, Z) saddle points is unlikely to alter the conclusions presented here on the basis of the leading instanton approximation. They also point out that the semiclassical approximation breaks down for times longer than tc ∼ c, where c is the central charge of the dual two-dimensional CFT. 4. Conclusions The study of very long time features of correlators in black hole backgrounds is a potentially important approach towards unraveling the mysteries of black hole evaporation and the associated physics at the spacelike singularity. We have seen that large scale topologychanging ﬂuctuations proposed in [3] begin to restore some of the ﬁne structure required by unitarity, but fall short at the quantitative level. Presumably the appropriate instantons occur on microscopic scales and involve stringy dynamics. While semiclassical black holes do faithfully reproduce “coarse grained” inclusive properties of the system such as the entropy (c.f. [12]), additional dynamical features of the horizon may be necessary to resolve ﬁner details of the information loss problem. Roughly, one needs a systematic set of corrections that could generate a “stretched horizon” of Planckian thickness [13]. The crudest model of such stretched horizon is the brick-wall model of ’t Hooft [10]. In this phenomenological description one replaces the horizon by a reﬂecting boundary condition at Planck distance ∼ P from the horizon. This deﬁnes a “mutilated” X manifold, of cylindrical topology, leading to a discrete spectrum of the right spacing in order of magnitude. We have also seen that the characteristic time for large topological ﬂuctuations to be important is tc ∼ O(N 2 ) in the semiclassical approximation. In [14] it was argued

LONG TIME SCALES AND ETERNAL BLACK HOLES

263

that semiclassical two-point functions probe the black hole singularity on much shorter characteristic times, thereby justifying the analysis on the single standard black hole manifold. However, we have seen that detailed unitarity is only restored on time scales of order tH ∼ exp(N 2 ). Thus tc tH and we conclude that such semiclassical analysis of the singularity is bound to be incomplete, as it misses whatever microphysics is responsible for the detailed unitarity restoration in the quantum mechanical time evolution. Acknowledgements E. R. would like to thank the KITP at Santa Barbara for hospitality during the completion of this work, under grant of the National Science Foundation No. PHY99-07949. The work of J.L.F.B. was partially supported by MCyT and FEDER under grant BFM200203881 and the European RTN network HPRN-CT-2002-00325. The work of E.R. is supported in part by the Miller Foundation, the BSF-American Israeli Bi-National Science Foundation, The Israel Science Foundation-Centers of Excellence Program, The GermanIsrael Bi-National Science Foundation and the European RTN network HPRN-CT-200000122. References 1. 2. 3. 4.

5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

S.W. Hawking, Phys. Rev. D14 (1976) 2460. J. Maldacena, Adv. Theor. Math. Phys. 2 (1998) 231 hep-th/9711200. S.S. Gubser, I.R. Klebanov and A.M. Polyakov, Phys. Lett. B428 (1998) 105 hep-th/9802109. E. Witten, Adv. Theor. Math. Phys. 2 (1998) 253 hep-th/9802150. J. Maldacena, J. High Energy Phys. 0304 (2003) 021 hep-th/0106112. L Susskind, Contribution to Stephen Hawking’s 60th birthday celebration. hep-th/0204027. L. Dyson, J. Lindesay and L. Susskind, J. High Energy Phys. 0208 (2002) 045 hep-th/0202163. N. Goheer, M. Kleban and L. Susskind, J. High Energy Phys. 0307 (2003) 056 hep-th/0212209. L. Dyson, M. Kleban and L. Susskind, J. High Energy Phys. 0210 (2002) 011 hep-th/0208013. T. Banks, L. Susskind and M. E. Peskin, Nucl. Phys. B244 (1984) 125. J.L.F. Barb´on and E. Rabinovici, J. High Energy Phys. 0311 (2003) 047 hep-th/0308063. D. Birmingham, I. Sachs, S. N. Solodukhin, Phys. Rev. D67 (2003) 104026 hep-th/0212308. I. Sachs, hep-th/0312287. M. Srednicki, cond-mat/9809360. G. ’t Hooft, Nucl. Phys. B256 (1985) 727. M. Kleban, M. Porrati and R. Rabadan, J. High Energy Phys. 0410 (2004) 030 hep-th/0407192. G.W. Gibbons and S.W. Hawking, Phys. Rev. D15 (1977) 2752. K.S. Thorne, R.H Price and D.A. McDonald, “Black Holes: The Membrane Paradigm”. Yale University Press 1986. L. Susskind, L. Thorlacius and J. Uglum, Phys. Rev. D48 (1993) 3743 hep-th/9306069. P. Kraus, H. Ooguri and S. Shenker, Phys. Rev. D67 (2003) 124022 hep-th/0212277. L. Fidkowski, V. Hubeny, M. Kleban and S. Shenker, hep-th/0306170.

SEMICLASSICAL STRINGS AND ADS/CFT

ARKADY A. TSEYTLIN Physics Department, The Ohio State University, Columbus OH43210-1106, USA and at Imperial College, London and Lebedev Physics Institute, Moscow

Abstract. We discuss AdS/CFT duality in the sector of “semiclassical” string states with large quantum numbers. We review the coherent-state effective action approach, in which similar 2d sigma model actions appear from the AdS5 × S 5 string action and from the integrable spin chain Hamiltonian representing the N=4 super Yang-Mills dilatation operator. We consider mostly the leading-order terms in the energies/anomalous dimensions which match but comment also on higher-order corrections.

1. Introduction The N = 4 SYM theory with SU (N ) gauge group is a family of conformal theories parametrized by the two numbers – N and gYM . Four-dimensional conformal theories have apparently much less symmetry than their two-dimensional cousins and thus should be much harder to solve (i.e. to determine their spectrum of dimensions of conformal primary operators and their correlation functions). There are strong indications that this 2 N =ﬁxed) limit. In the large N problem may simplify in the planar (N → ∞, λ ≡ gYM limit there are no formal objections to integrability of a 4-d quantum ﬁeld theory, and, indeed, the AdS/CFT duality [1, 2, 3] implies the existence of a hidden integrable 2-d structure corresponding to that of AdS5 × S 5 string sigma model (on a sphere or a cylinder). A major ﬁrst step towards the solution of the SYM theory would be to determine the spectrum of anomalous dimensions ∆(λ) of the primary operators built out of products of local gauge-covariant ﬁelds. The AdS/CFT duality implies the equality between the AdS5 energies E√ of quantum R2 λ closed string states (as functions of the effective string tension T = 2πα = 2π and quan5 tum numbers like S angular momenta Ji ) and the dimensions ∆ of the corresponding local SYM operators (see, e.g., [4]). To check the duality one would like also to understand how strings “emerge” from the ﬁeld theory, in particular, which (local, single-trace) gauge theory operators [5] correspond to which “excited” string states and how one may verify the matching of their dimensions/energies beyond the well-understood BPS/supergravity sector. One would then like to use the duality as a guide to deeper level of the structure 265 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 265–290. © 2005 Springer. Printed in the Netherlands.

266

ARKADY A. TSEYTLIN

of quantum SYM theory. For example, the results motivated by comparison to string theory may allow one to “guess” the general structure of the SYM anomalous dimension matrix and may also suggest new methods of computing anomalous dimensions in less supersymmetric gauge theories. Below we shall review some recent progress in checking AdS/CFT correspondence in a subsector of non-BPS string/SYM states with large quantum numbers. 1.1. GENERALITIES Let us start with brief remarks on the SYM and the string sides of the duality. The SYM theory contains a gauge ﬁeld, 6 scalars φm and 4 Weyl fermions, all in adjoint representation of SU (N ). It has global conformal and R-symmetry, i.e. is invariant under SO(2, 4) × SO(6). To determine (in the planar limit) scaling the dimensions of local gauge-invariant operators one, in general, needs to ﬁnd the anomalous dimension matrix to all orders in λ and then to diagonalize it. The special case is that of chiral primary or BPS operators (and their descendants) tr(φ{m1 ...φmk } ) whose dimensions are protected, i.e. do not depend on λ. The problem of ﬁnding dimensions appears to simplify also in the case of “long” operators containing large number of ﬁelds under the trace. One example is provided by “near-BPS” operators [6] like tr(ΦJ1 Φn2 ...) + ..., where J n, and Φk = φk + iφk+3 , k = 1, 2, 3. Below we shall consider “far-from-BPS” operators like tr(ΦJ1 1 ΦJ2 2 ...) + ..., with J1 ∼ J2 1. The type IIB string action in AdS5 × S 5 space has the following structure 1 I=− T 2

2π

dσ (∂ p Y µ ∂p Y ν ηµν + ∂ p X m ∂p X n δmn + ...) ,

dτ

(1.1)

0 √

where Y µ Y ν ηµν = −1, X m X n δmn = 1 , ηµν = (− + + + +−), T = 2πλ and dots stand for the fermionic terms [7] that ensure that this model deﬁnes a 2-d conformal ﬁeld theory. The closed string states can be classiﬁed by the values of the Cartan charges of the obvious symmetry group SO(2, 4) × SO(6) – E, S1 , S2 ; J1 , J2 , J3 , i.e. by the AdS5 energy, two spins in AdS5 and 3 spins in S 5 . The mass shell (conformal gauge constraint) condition then gives a relation E = E(Q, T ). Here T is the string tension and Q = (S1 , S2 , J1 , J2 , J3 ; nk ) where nk stand for higher conserved charges (analogs of oscillation numbers in ﬂat space). The BPS (chiral primary) string states are point-like (supergravity modes), near-BPS (BMN) states are nearly pointlike, while generic semiclassical far-from-BPS states are represented by extended closed string conﬁgurations. According to the AdS/CFT duality quantum closed string states in AdS5 × S 5 should be dual to quantum SYM states at the boundary, i.e. in R × S 3 or, via radial quantization, to local single-trace operators at the origin of R4 . Such operators have the following structure S1 S2 D3+i4 ΦJ1 1 ΦJ2 2 ΦJ3 2 ...) + ... (where scalars and covariant derivatives may be also tr(D1+i2 replaced by gauge ﬁeld strength factors and fermions). The energy of a string state should then be equal to the dimension of the corresponding SYM operator, E(Q, T ) = ∆(Q, λ), where on the SYM side the charges Q should characterise the eigen-operator of the anomalous dimension matrix. By analogy with the ﬂat space case and ignoring α corrections (i.e. assuming R → ∞ or α → 0) the excited string states are expected to have energies

SEMICLASSICAL STRINGS AND ADS/CFT

267

E ∼ m ∼ √1α ∼ λ1/4 [2]. This represents a non-trivial prediction for strong-coupling asymptotics of SYM dimensions. The asymptotics may, however, be different in the limits where the charges Q are also large, e.g., of order λ1/2 as in the semiclassical limit [8]. In general, the natural perturbative expansion on the string side will ∞ (inverse-tension) n , while on the SYM side the usual planar perturbation be given by E = n=−1 (√cλ) n ∞ theory will give the eigenvalues of the anomalous dimension matrix as ∆ = n=0 an λn . The AdS/CFT duality implies that the two expansions should be the strong-coupling and the weak-coupling asymptotics of the same function. To check the relation E = ∆ is then a non-trivial problem, except in the case of 1/2 BPS (single-trace chiral primary) operators which are dual to the supergravity states when the energies/dimensions are protected from corrections [4] and thus can be matched on the symmetry grounds. 1.2. SEMICLASSICAL STRING STATES: BMN AND BEYOND For generic non-BPS states the situation with checking the duality looked hopeless until the remarkable suggestion of [6]1 and then of [8] that a progress can be made by (i) concentrating on a subsector of states with large or “semiclassical” values of quantum √ numbers, Q ∼ T ∼ λ and (ii) considering a new limit Q→∞, On the string side

Q √ λ

˜ ≡ λ = ﬁxed . λ Q2

(1.2)

= √1 plays the role of a semiclassical parameter (like rotation ˜ λ

frequency) which can then be taken to be large. The energy of such states is E = Q + f (Q, λ), where f → 0 in the λ → 0 limit. The duality implies that such semiclassical string states (as well as states represented by small ﬂuctuations near them) should be dual to “long” SYM operators with a large canonical dimension, i.e. containing large number of ﬁelds or derivatives under the trace. In this case the duality map becomes more explicit. 1 The history of the BMN limit is somewhat non-trivial. It started with the important observation that the Penrose limit of the AdS5 × S 5 space [9] leads to a maximally supersymmetric plane wave geometry supported by the Ramond-Ramond 5-form ﬂux. Remarkably, the AdS5 × S 5 string theory action [7] in this limit (i.e. Green-Schwarz string action in plane-wave background) becomes essentially quadratic and thus its spectrum can be found explicitly [10]. Motivated by that, ref. [6] gave a dual SYM interpretation of the corresponding string states, suggested that string energies can be compared to perturbative SYM dimensions computed in the same limit, and indeed directly checked this to the ﬁrst non-trivial order. Ref. [8] then explained that, on the string-theory side, the BMN limit is nothing but a semiclassical expansion near a particular (point-like) string solution. This interpretation was further clariﬁed and extended in [11, 12], suggesting, in particular, how one can in principle compute corrections to the BMN limit (which was later done in full detail in [13] and especially in [14]). This made it clear that the plane-wave connection is not fundamental but rather is a special feature of semiclassical expansion near a certain string conﬁguration (represented by massless AdS5 × S 5 geodesic wrapping S 5 ). Expanding near other string solutions leads to other special “Penrose-type” limits of string geometry (geometry “seen” by a fundamental string probe) which are described by the corresponding quadratic ﬂuctuation actions. Expansions near a class of “fast” string conﬁgurations (discussed below) for which the world sheet becomes null in the effective zero-tension limit [15, 16, 17] may be interpreted as a stringy generalization of the Penrose limit (keeping in mind, of course, that this analogy applies only in the strict zero-tension limit).

268

ARKADY A. TSEYTLIN

The simplest possibility is to start with a BPS state that carries a large quantum number and consider small ﬂuctuations near it, i.e. a set of near-BPS states characterised by a large parameter [6]. The only non-trivial example of such a BPS state is represented by a pointlike string moving along a geodesic in S 5 with a large angular momentum Q = J. Then E = J and the dual operator is trΦJ , Φ = φ1 +iφ2 . The small closed strings representing near-by ﬂuctuations are ultrarelativistic, i.e. their kinetic energy is much larger than their mass. They are dual to SYM operators of the form tr(ΦJ ...) where dots stand for a small number of other ﬁelds and/or covariant derivatives (one needs to sum over different orders of the factors to ﬁnd an eigenstate of the anomalous dimension matrix). The energies of the small ﬂuctuations happen to have the structure [10, 6] E=J+

! ˜ K + O( 1 ) = J + K + k1 λ ˜ + k2 λ ˜ 2 + ... . 1 + n2 λ J

(1.3)

One can argue in general [11, 12] and check explicitly [13, 14] that higher-order quantum string sigma model corrections are indeed suppressed in the limit (1.2), i.e. in the large J, ˜ ≡ λ2 = λ limit. A remarkable feature of this expression is that E is analytic in ﬁxed λ J ˜ suggesting direct comparison with perturbative SYM expansion in λ. λ, ˜ 3 terms in the expansion of the ˜ λ ˜ 2 and λ Indeed, it can be shown that the ﬁrst three λ, square root agree precisely with the one [6], two [19] and three [20, 21, 23] loop terms in the anomalous dimensions of the corresponding operators. There is also (for a 2-impurity ˜ expression may appear K = 2 case) an argument [24] suggesting how the full 1 + n2 λ on the perturbative SYM side (for reviews of various aspects of the BMN limit see also [25]). However, the general proof of the consistency of the BMN limit in the SYM theory ˜ and 1 ) (i.e. that the usual perturbative expansion can be rewritten as an expansion in λ J remains to be given. Also, to explain why the string and the SYM expressions match one ˜ = λ2 → 0) and the SYM limit (ﬁrst should show that the string limit (ﬁrst J → ∞, then λ J λ → 0, then J → ∞) produce exactly the same expressions for the energies/dimensions, even though, in general, the two limits may not commute, cf. [26, 27, 28]. If one moves away from the near-BPS limit and considers, e.g., a non-supersymmetric state with a large angular momentum Q = S in AdS5 [8], a similar direct quantitative check of the duality is no longer possible: here the classical energy is not analytic in λ and quantum corrections are no longer suppressed by powers of S1 (but as usual are suppressed by powers of √1λ ). However, it is still possible to demonstrate a remarkable qualitative agreement between S-dependence of the string energy and of the SYM anomalous dimension. The energy of a folded closed string rotating at the center of AdS5 which is dual [8] to twist 2 operators on the SYM side (tr(Φ∗k DS Φk ), D = D1 + iD2 , and similar operators with spinors and gauge bosons that mix at higher loops [32, 33]) has the following form when expanded at large S :

On the string side

E = S + f (λ) ln S + O(S 0 )

(1.4)

√ c2 f (λ)λ1 = c0 λ + c1 + √ + ... , λ

(1.5)

SEMICLASSICAL STRINGS AND ADS/CFT

269

where c0 = π1 is the classical [8] and c1 = − π3 ln 2 is the 1-loop [11] coefﬁcient. On the gauge theory side one ﬁnds the same S-dependence of the anomalous dimension with the perturbative expansion of the ln S coefﬁcient being f (λ)λ1 = a1 λ + a2 λ2 + a3 λ3 + ...

(1.6)

1 11 where a1 = 2π1 2 [31], a2 = − 96π 2 [32], and a3 = 360×64π 2 [33]. Like in the case of the SYM entropy [34], here one expects the existence of a smooth interpolating function f (λ) that connects the two perturbative expansions. In fact, observing that the factor π1 in (1.5) and factor π12 in (1.6) seem to factorize, one can suggest a simple square root type interpolating formula for f (λ) that seem indeed to give a good ﬁt [32, 33] (cf. also the discussion end of section 4).

1.3. MULTISPIN STRING STATES One may wonder still if examples of quantitative agreement between string energies and SYM dimensions found for the near-BPS (BMN) states exist also for more general nonBPS string states. Indeed, it was noticed already in [11] that a string state that carries large spin in AdS5 as well as large spin J in S 5 has, in contrast to the above J = 0 ˜ = λ2 , just as in the BMN case with a large case, an analytic expansion of its energy in λ J oscillation number K ∼ S. It was observed in [35] that semiclassical string states carrying several large spins (with at least one of them being in S 5 ) have a regular expansion of their ˜ and it was then suggested, by analogy with the near-BPS case, energy E in powers of λ ˜ may be possible to match with the that the expansion of E in small effective tension or λ perturbative expansion of the SYM dimensions. √ 5 √ For a classical rotating closed string solution in S one has E = λE(wi ), Ji = λwi so that E = E(Ji , λ). The required key √ property is that (in contrast to the case of a single spin in AdS5 ) there should be no λ factors in the expansion of the classical energy E in small λ λ λ2 ˜ + c2 λ ˜ 2 + ... . (1.7) E = J + c1 + c2 3 + ... = J 1 + c1 λ J J 3 ˜ ≡ λ2 and cn = cn ( Ji ) are functions of ratios of the spins which Here J = i=1 Ji , λ J J ˜ =ﬁxed. are ﬁnite in the limit Ji 1, λ The simplest example of such solution is provided by a circular string rotating in two orthogonal planes in S 3 part of S 5 with the two angular momenta being equal, J1 = J2 [35]: X1 ≡ X1 + iX2 = cos(nσ) eiwτ ,

X2 ≡ X3 + iX4 = sin(nσ) eiwτ ,

and with the global AdS5 time being t = κτ . The conformal gauge constraint implies κ2 = w2 + n2 and thus 1 ˜ 1 4 ˜2 E = J 2 + n2 λ = J(1 + n2 λ (1.8) − n λ + ...) , 2 8 regular expansion in string where J = J1 + J2 = 2J1 . For ﬁxed J the energy thus has a! tension (in contrast to what happens in ﬂat space where E =

2 α J).

270

ARKADY A. TSEYTLIN

Similar expressions (1.7) are found also for more general rigid multispin closed strings [35, 36, 37, 38, 39, 40]. In particular, for a folded string rotating in one plane of S 5 and with its center of mass orbiting along big circle in another plane the coefﬁcients cn are transcendental functions (expressed in terms of elliptic integrals) [37]. More generally, the 3-spin solutions are described by an integrable Neumann model [38, 39] and the coefﬁcients cn in the energy are expressed in terms of genus two hyperelliptic functions. The reason why choosing a particular string ansatz one gets an integrable effective 1-d model lies in the integrability of the original S 5 = SO(6)/SO(5) classical sigma model [30] (see also [44]). To be able to hope to compare the classical energy to the SYM dimension one should ˜ =ﬁxed. be sure that higher string α corrections are suppressed in the limit J → ∞, λ 1 1 √ √ ; what is more important, the Formally, this is of course the case since α ∼ λ ∼ J

˜ λ

˜ [36], i.e., as in the BMN case, the expansion in large corrections are again analytic in λ ˜ J and small λ is well-deﬁned on the string side, d1 d2 2 ˜ ˜ E = J 1 + λ(c1 + + ...) + λ (c2 + + ...) + ... , (1.9) J J 1 J

with the classical energy (1.7) being the J → ∞ limit of the exact expression. The reason for this particular form of the energy (1.9) can be explained as follows [11, 12, 40]. We are computing the AdS5 × S 5 superstring sigma model loop corrections to the mass of a stationary solitonic solution on a 2-d cylinder (no IR divergences). This theory is conformal (due to the crucial presence of the fermionic ﬂuctuations) and thus it does not depend on a UV cutoff. The relevant 2d ﬂuctuations are massive and their masses scale as w ∼ √1 . As a result, the inverse mass expansion is well-deﬁned and the ˜ λ

˜ This was explicitly quantum corrections should be proportional to positive powers of λ. demonstrated by a 1-loop computation in [36, 66]. Similar expressions are found for the energies of small ﬂuctuations near a given classical solution: as in the BMN case, the ﬂuctuation energies are suppressed by extra factor of J, i.e. ˜ 1 + m1 + ...) + λ ˜ 2 (k2 + m2 + ...) + ... . δE = λ(k (1.10) J J 1.4. ADS/CFT DUALITY: NON-BPS STATES Assuming that the same large J limit is well-deﬁned also on the SYM side, one should then be able to compare the coefﬁcients in (1.9) to the coefﬁcients in the anomalous dimensions of the corresponding SYM operators tr(ΦJ1 1 ΦJ2 2 ΦJ3 3 ) + ... (and also do similar matching for near-by ﬂuctuation modes) [35]. In practice, what is known (at least in principle) is how to compute the dimensions in the different limit: by ﬁrst expanding in λ and then expanding in J1 . One may expect that this expansion of anomalous dimensions may take the form equivalent to (1.9), i.e. ∆ = J + λ(

a1 b1 a2 b2 + 2 + ...) + λ2 ( 3 + 4 + ...) + ... , J J J J

(1.11)

SEMICLASSICAL STRINGS AND ADS/CFT

271

and that the respective coefﬁcients in E and ∆ may agree with each other. The subsequent work [41, 42, 43, 44, 45, 2, 27, 47, 3, 49] did verify this structure of ∆ and, moreover, established the general agreement between the two leading coefﬁcients c1 , c2 in E (1.9) and the one-loop and two-loop coefﬁcients a1 , a2 in ∆ (1.11) (as usual, by “n-loop” term in ∆ we mean the term multiplied by λn ). To compute ∆ one is to diagonalize the anomalous dimension matrix deﬁned on a set of “long” scalar operators, and this is obviously a non-trivial problem. The important step to this goal was made in [41] where it was observed that the one-loop planar dilatation operator in the scalar sector can be interpreted as a Hamiltonian of an integrable SO(6) spin chain and thus can be diagonalized even for large length L = J by the Bethe ansatz method.2 In the simplest case of the “SU (2)” sector of operators tr(ΦJ1 1 ΦJ2 2 ) + ... built out of two chiral scalars, the dilatation operator can be interpreted as “spin up” and “spin down” states of periodic XXX1/2 spin chain with length L = J = J1 + J2 . Then the 1-loop dilatation operator becomes equivalent to the Hamiltonian of the ferromagnetic Heisenberg model J λ (I − σl · σl+1 ) . (1.12) D1 = (4π)2 l=1

By considering the thermodynamic limit (J → ∞) of the corresponding Bethe ansatz ˜ equations, the proposal of [35] was conﬁrmed at the leading order of expansion in λ a1 [42, 43]: for eigen-operators with J1 ∼ J2 1 it was shown (i) that ∆ − J = λ J + ..., and (ii) a remarkable agreement was found between a1 = a1 ( JJ12 ) and the coefﬁcient c1 in the energies (1.9) of various 2-spin string solutions. It was also possible also to match (as in the BMN case) the energies of ﬂuctuations near the circular J1 = J2 solution with the corresponding eigenvalues of (1.12) [42]. Similar leading-order agreement between string energies and SYM dimensions was observed also in other sectors of states with large quantum numbers: (1) in the SU (3) sector: for speciﬁc solutions [35, 38, 39] with 3 spins in S 5 which are dual to the operators tr(ΦJ1 1 ΦJ2 2 ΦJ3 3 ) + ... [45, 50]; (2) in the SL(2) [51] sector: for a folded string state [11] with one spin in AdS5 and one spin in S 5 (with E = J + S + Jλ c1 ( SJ ) + ... [11, 35]) which is dual to the operators tr(DS ΦJ ) + ... [43]; (3) in a “subsector” of SO(6) states containing pulsating (and rotating) solutions which again have regular [52] expansion of the energy in the limit of large oscillation λ + ... [42, 45]. number L, i.e. E = L + c1 L This agreement between the leading-order terms in the expansion of energies of certain semiclassical string states and dimensions of the corresponding “long” SYM operators leaves, however, many questions, in particular: (i) How to understand this agreement beyond speciﬁc examples, i.e. in a more universal way? (ii) Which is the precise relation between proﬁles of string solutions and the structure of the dual SYM operators? 2

Relations between 1-loop anomalous dimension matrix for a certain class of composite operators and integrable spin chain Hamiltonians were observed previously in the large N QCD context [63] (for a review and connections to AdS/CFT see [64]).

272

ARKADY A. TSEYTLIN

(iii) How to characterise the set of semiclassical string states and dual SYM operators to which this direct relation should apply? (iv) Why the agreement holds at all, i.e. why the two limits (ﬁrst J → ∞, and then ˜ → 0, or vice versa) taken on the string and the SYM sides give equivalent results to the λ ˜ Why/when it does not work to all orders in expansion ﬁrst two orders in expansion in λ? 1 ˜ in λ (and J )? The questions (i),(ii) were addressed in [2, 3, 55, 56, 57] using the low-energy effective action approach for coherent states; an alternative approach based on matching the general solution (and the integrable structure) of the string sigma model with that of the thermodynamic limit of the Bethe ansatz in the SU (2) sector was developed in [47]. The question (iii) was addressed in [15, 16, 17, 57], and the question (iv) – in [27, 28, 29]. Still, our understanding of why there is a direct agreement with gauge theory at the ﬁrst ˜ 3 order is ˜ and λ ˜ 2 orders of expansion and why it does not [43, 28] continue to the λ two λ 3 still rather rudimentory. Below we shall review the effective action approach as developed in [2, 3, 56, 57], ˜ concentrating mostly on the leading (“1-loop”) order in expansion in λ. 2. Effective actions for coherent states The suggestion of how to understand the agreement between leading-order terms in the multispin string energies and the corresponding one-loop anomalous dimensions in a universal way was made in [2] and was clariﬁed and elaborated further in [3, 57]. The key idea was that instead of comparing particular solutions one should try to match effective sigma models which appear on the string side and the SYM side. Another related idea of [2, 3, 57] was that since the “semiclassical” string states carrying large quantum numbers are represented in the quantum theory by coherent states, one should be comparing coherent string states to coherent SYM states (and thus to coherent states of the spin chain). In view of the ferromagnetic nature of the dilatation operator (1.12), in the thermodynamic limit J = J1 + J2 → ∞ with ﬁxed large number of impurities (i.e. with ﬁxed JJ12 ) it is favorable to form large clusters of spins. Then a “low-energy” approximation and continuum limit should apply, leading to an effective “non-relativistic” sigma model for a coherent-state expectation value of the spin operator. At the same time, on the string side, taking the “large space-time energy” (or large J) limit directly in the classical string action produces a reduced “non-relativistic” sigma ˜ corrections to energies of model that describes in a universal way the leading-order O(λ) all string solutions in the two-spin sector. The resulting sigma model action turns out to agree exactly [2] with the semiclassical coherent state action found from the SU (2) sector ˜ =ﬁxed limit. This demonstrates how a string action of the spin chain in the J → ∞, λ can directly “emerge” from a gauge theory in the large-N limit and provides a direct map between the “coherent” SYM states (or the corresponding operators built out of two holomorphic scalars) and all two-spin classical string states. Furthermore, the correspondence established at the level of the action implies also (i) the matching of the integrable structures and (ii) the matching of the ﬂuctuations around particular solutions and thus it goes beyond the special examples of rigidly rotating strings. 3

Similar (dis)agreements were found for the 1/J corrections to the BMN states [14].

SEMICLASSICAL STRINGS AND ADS/CFT

273

2.1. COHERENT STATES Let us brieﬂy review the deﬁnition of coherent states (see, e.g., [53]). For a harmonic oscillator ([a, a† ] = 1) one can deﬁne the coherent state |u as a|u = u|u, where u † ∗ is a complex number. Equivalently, |u = R(u)|0, where R = eua −u a so that acting † on the vacuum |0 the operator R is simply proportional to eua . Note that |u can be written as a superposition of the eigenstates |n of the harmonic oscillator Hamiltonian, n ∞ |u ∼ n=0 √un! |n. An alternative deﬁnition of a coherent state is that it is a state with minimal uncertainty for both the coordinate qˆ = √12 (a + a† ) and the momentum pˆ =

− √i2 (a − a† ) operators, ∆hatp2 = ∆ 12 tq 2 = 12 , ∆ˆ p2 ≡ u|ˆ p2 |u − (u|ˆ p|u)2 . For that reason it is the “best” approximation to a classical state. If one deﬁnes a time-dependent q |u = √12 (u+u∗ ), state |u(t) = e−iHt |u then the expectation values of qˆ and pˆ, i.e. u|ˆ

u|ˆ p|u = − √i2 (u − u∗ ) follow the classical trajectories. Starting instead of the Heisenberg algebra with the SU (2) algebra [S3 , S± ] = ±S± , = 1 σ one can [S+ , S− ] = 2S3 and considering the s = 1/2 representation where S 2 deﬁne a spin coherent state as a linear superposition of spin up and spin down states: ∗ |u = R(u)|0. Here R = euS+ −u S− , |0 = | 12 , 12 and u is a complex number. An equivalent way to label the coherent state is by a unit 3-vector n deﬁning a point of S 2 . Then |n = R(n)|0 where |0 corresponds to a 3-vector (0, 0, 1) along the 3rd axis. One can write n = U †σ U, U = (u1 , u2 ), and then R(n) is an SO(3) rotation from a north pole to a generic point of S 2 deﬁned by n. The key property of the coherent state is that n determines the coherent state expectation value of the spin operator: n = 1 n . (2.1) n|S| 2 Similar deﬁnition of coherent states can be given in the case when SU (2) is replaced by a generic group G. Given a semisimple group G with the Cartan basis of its algebra (Hi , Eα , E−α ) ([Hi , Hj ] = 0, [Hi , Eα ] = αi Eα , [Eα , E−α ] = αi Hi , [Eα , Eβ ] = Nαβ Eα+β ) whose interpretation will be a symmetry group of a quantum Hamiltonian (acting in a unitary irreducible representation Λ on the Hilbert space VΛ ) one may deﬁne a set of coherent states by choosing a particular state |0 (with 0||0 = 1) in VΛ and acting on it by the elements of G. A subroup H of G that leaves |0 invariant up to a phase (Λ(h)|0 = eiφ(h) |0) is called maximum stability subgroup. One may then deﬁne the coset space G/H, the elements of which (g = ωh, h ∈ H, ω ∈ G/H, Λ(g) = Λ(ω)Λ(h)) will parametrize the coherent states, |ω, Λ = Λ(ω)|0. This deﬁnition depends on a choice of group G, its representation Λ and the vector |0. It is natural to assume also that |0 is an eigenstate of the Hamiltonian H, e.g., a ground state. For a unitary representation Λ we may choose H†i = Hi , E†α = E−α and select |0 to be the highest-weight vector of the representation Λ, i.e. demand that it is annihilated by “raising” generators and is an eigen-state of the Cartan generators: (i) Eα |0 = 0 for all positive roots α; (ii) Hi |0 = hi |0. In addition, we may demand that |0 is annihilated also by some “lowering” generators, i.e. (iii) E−β |0 = 0 for some negative roots β; the remaining negative roots will be denoted by γ. Then the coherent states are given by

(wγ E−γ − wγ∗ Eγ ) |0 , (2.2) |ω, Λ = exp γ

274

ARKADY A. TSEYTLIN

where γ are the negative roots for which Eγ |0 = 0. wγ may be interpreted as coordinates on G/H where H is generated by (Hi , Eα , E−β ). For example, in the case of G = SU (3) with the Cartan basis (H1 , H2 , Eα , Eβ , Eα+β , E−α , E−β , E−α−β ) and with |0 being the highest-weight of the fundamental representation, i.e. E−β |0 = 0, E−α |0 = 0, E−α−β |0 = 0, the subgroup H is generated by (H1 , H2 , Eβ , E−β ), i.e. is SU (2)×U (1) and G/H = SU (3)/(SU (2)× U (1)) = CP 2 (see also [56]). We shall apply this general deﬁnition of coherent states in section 3. 2.2. LANDAU-LIFsh*tZ MODEL FROM SPIN CHAIN In general, one can rewrite the usual phase space path integral as an integral over the overcomplete set of coherent states (for the harmonic oscillator this is simply a change of variables from q, p to u = √12 (q + ip)): Z=

[du] eiS[u] ,

S=

d dt u|i |u − u|H|u . dt

(2.3)

d The ﬁrst (“Wess-Zumino” or “Berry phase”) term in the action ∼ iu∗ dt u is the analog of the usual pq˙ term in the phase-space action. Applying this to the case of the Heisenberg spin chain Hamiltonian (1.12) one ends up with with the following action for the coherent state variables nl (t) at sites l = 1, ..., J (see, e.g., [54]):

S=

dt

J

l ) · nl − C(n

l=1

λ 2 (nl+1 − nl ) . 2(4π)2

(2.4)

is a monopole potential on S 2 ). In local coordinates Here dC = ijk ni dnj ∧ dnk (i.e. C · dn = 1 cos θdφ. In the (at each site l) one has n = (sin θ cos φ, sin θ sin φ, cos θ), C 2 ˜ = λ2 (which we are interested in) we can take a continuum limit J → ∞ with ﬁxed λ J limit by introducing the 2-d ﬁeld n(t, σ) = {n(t, 2π J l)}, l = 1, ..., J. Then S=J

dt 0

2π

dσ 1˜ 2 C · ∂tn − λ(∂ n ) + ... , σ 2π 8

(2.5)

where dots stand for higher derivative terms suppressed by J1 . Since J appears as a factor in front of the action, in the limit J → ∞ all quantum corrections should be also suppressed by J1 , and thus the above action can be treated classically. The corresponding equations of motion 1˜ 2 (2.6) ∂t ni = λ

ijk nj ∂σ nk 2 are the Landau-Lifsh*tz equations for a classical ferromagnet. An alternative derivation of them is based on ﬁrst writing down the Heisenberg equation for the time evolution of the spin operator directly from the spin chain Hamiltonian, then considering the coherent state expectation value and ﬁnally taking the continuum limit.

SEMICLASSICAL STRINGS AND ADS/CFT

275

2.3. LANDAU-LIFsh*tZ MODEL FROM STRING ACTION The action (2.5) should be describing the coherent states of the Heisenberg spin chain in the above thermodynamic limit. One may wonder how a similar “non-relativistic” action may appear on the string side where one starts with the usual “relativistic” sigma model (1.1). To obtain such an effective action one is to perform the following sequence of steps [2, 3, 57]: (i) isolate a “fast” coordinate α whose momentum pα is large for a class of string conﬁgurations we consider; ˜ ∼ σ where α ˜ is “T-dual” to α); (ii) gauge-ﬁx t ∼ τ and pα ∼ J (or α (iii) expand the action in derivatives of “slow” or “transverse” coordinates (to be identiﬁed with n). To illustrate this procedure let us consider the SU (2) sector of string states carrying two large spins in S 5 , with string motions restricted to S 3 part of S 5 . The relevant part of the AdS5 × S 5 metric is then ds2 = −dt2 + dXi dX∗i , with Xi X∗i = 1. Let us set X1 = X1 + iX2 = u1 eiα ,

X2 = X3 + iX4 = u2 eiα ,

ui u∗i = 1 .

Here α will be a coordinate associated to the total spin in the two planes (which in general will be the sum of orbital and internal spin). ui (deﬁned modulo U (1) gauge transformations) will be the “slow” coordinates determining the “transverse” string proﬁle. Then dXi dX∗i = (dα + C)2 + Dui Du∗i ,

C = −iu∗i dui ,

Dui = dui − iCui ,

where the second |Dui |2 term represent the metric of CP 1 (this parametrisation corresponds to the Hopf ﬁbration S 3 ∼ S 1 × S 2 ). Introducing n = U †σ U, U = (u1 , u2 ) we get 1 dXi dX∗i = (Dα)2 + (dn)2 , Dα = dα + C(n) . (2.7) 4 Writing the resulting sigma model action in phase space form and imposing the (nonconformal) gauge t = τ, pα =const= J, one gets [3] the same action (2.5) with the WZ · ∂tn originating from the pα Dα term in the phase-space Lagrangian (cf. its origin term C on the spin chain side as an analog of the pq˙ in the coherent state path integral action). An equivalent approach [57] leading to the same action (2.5) is based on ﬁrst applying a 2-d duality (or “T-duality”) transformation α → α ˜ and then choosing the “static” gauge t = τ, α ˜ = √1 σ with √1 = √Jλ . Indeed, starting with ˜ λ

˜ λ

L=−

1√ −g g pq − ∂p t∂q t + Dp αDq α + Dp u∗i Dq ui 2

(2.8)

and applying the 2-d duality in α we get L=−

1√ −gg pq − ∂p t∂q t + ∂p α ˜ ∂q α ˜ + Dp u∗i Dq ui + pq Cp ∂q α ˜. 2

(2.9)

Thus the “T-dual” background has no off-diagonal metric component but has a non-trivial NS-NS 2-form coupling in the (˜ α, ui ) sector. It is useful not to use conformal gauge here. Eliminating the 2-d metric g pq we then get the Nambu-type action √ L = pq Cp ∂q α ˜ − h, (2.10)

276

ARKADY A. TSEYTLIN

where h = | det hpq | and hpq = −∂p t∂q t + ∂p α ˜ ∂q α ˜ + D(p u∗i Dq) ui . If we now ﬁx the 2π 1 static gauge, t = τ, α ˜ = √ σ, we ﬁnish with the action I = J dt 0 dσ 2π L, where ˜ λ

+ L = Ct −

˜ σ ui |2 )(1 − |Dt ui |2 ) + 1 λ(D ˜ t u∗ Dσ ui + c.c.)2 . (1 + λ|D i 4

(2.11)

Making the key assumption that the evolution of ui in t is slow, i.e. the time derivatives ˜ and expanding in powers are suppressed (which can be implemented by rescaling t by λ ˜ ˜ of λ), we ﬁnd, to the leading order in λ, 1˜ 2 L = −iu∗i ∂t ui − λ|D σ ui | . 2

(2.12)

This is the same as the CP 1 Landau-Lifsh*tz action (2.5) when written in terms of n. Thus the string-theory counterpart of the WZ term in the spin-chain coherent state effective action originates from the 2-d NS-NS WZ term in the action for the “T-dual” coordinate α ˜ upon the static gauge ﬁxing of the latter [57]. To summarize: (i) (t, α) ˜ are the “longitudinal” coordinates that are gauge-ﬁxed (with α ˜ playing the role of the string direction or the spin chain direction on the SYM side); (ii) U = (u1 , u2 ) or n = U †σ U are the “transverse” coordinates that determine the semiclassical string proﬁle and also the structure of the coherent operator on the SYM side, tr(Πσ ui (σ)Φi ) (see [2, 57] and below). The agreement between the low-energy effective actions on the spin chain and the string side explains not only the matching of energies of coherent states representing conﬁgurations with two large spins (and also the matching of near-by ﬂuctuations) but also the equivalence of the integrable structures (which was observed on speciﬁc examples in [44, 45]). 2.4. HIGHER ORDERS IN λ The above leading-order agreement in SU (2) sector has several generalizations. ˜ and First, we may include higher-order terms on the string side. Expanding (2.11) in λ eliminating higher powers of time derivatives by ﬁeld redeﬁnitions (which can be done since the leading-order equation of motion is ﬁrst order in time derivative) we end up with [3] (n ≡ ∂σ n) ˜ ˜2 · n˙ − λ n2 + λ (n2 − 3 n4 ) L=C 8 32 4 3 ˜ λ 7 25 13

˜4) . − n2 − n2n2 − (nn )2 + n6 + O(λ (2.13) 64 4 2 16 ˜ 2 term is obtained [3] in the coherent state action on the spin chain side by The same λ starting with the sum of the 1-loop dilatation operator (1.12) and the 2-loop term found in [20] D2 =

J 2λ2 (Ql,l+2 − 4Ql,l+1 ) , (4π)4 l=1

Qk,l ≡

1 (I − σk · σl ) . 2

(2.14)

SEMICLASSICAL STRINGS AND ADS/CFT

277

The n4 term originates from a non-trivial quantum correction on the spin chain side. This ˜ observed on explains the matching of energies and dimensions to the ﬁrst two orders in λ, speciﬁc examples (using the generalized Bethe ansatz on the spin chain side) in [27]. The equivalent general conclusion about 2-loop matching was obtained in the integrabilitybased approach in [47]. ˜ 3 term. A natural explanation The order-by-order agreement seems to break down at λ ˜ [27, 28] is that the string limit (ﬁrst J → ∞, then λ → 0) and the SYM limit (ﬁrst λ → 0, then J → ∞) need not produce the same result when applied to the 2-parameter functions E = ∆(λ, J). A proposal of how to “complete” the perturbative gauge-theory expression to make the agreement with string theory manifest appeared in [28]; ref. [29] also suggested a possible form of generalized Bethe ansatz on the string side that would naturally interpolate between the string and gauge theory results (see also [62]). These ˜ 3 disagreement [14] between the string proposals also suggest a resolution of the order λ 1 theory and gauge theory predictions for J corrections to the BMN spectrum. ˜ and λ ˜ 2 terms is that the A possible explanation of why we found the agreement of λ structure of the dilatation operator at one and two loop orders is, in a sense, ﬁxed by the BMN limit, which thus essentially determines the low energy effective action in a unique way. This is no longer so starting with the 3-loop order, where the dilatation operator already contains [20] the 4-spin QQ interactions (cf. (2.14)) which do not contribute to anomalous dimensions in the strict BMN limit. By analogy with non-renormalization (due to underlaying supersymmetry) of few leading terms in low-energy effective actions, one may suggest that here the 3-loop problem may be related to the appearance of non-trivial interpolating functions as coefﬁcients of the Qn (n ≥ 2) terms in the dilatation operator. This would, in particular, explain why different structures of the QQ terms are found in the “gauge theory” [20] and “string theory” [62] limits. As was suggested in [27, 28], the disagreement between the string and gauge theory results at 3-loop order in λ and the leading order in J can be repaired by adding “wrapping” contributions to the dilatation operator (and thus to the Bethe ansatz relations) on the gauge theory side. To illustrate this possibility, let us consider the circular solution λJ case [35], and use the function (1+λ) J [28], which is equal to 1 in the string theory limit λ ˜ (J → ∞ with ﬁxed 2 ≡ λ) but zero in the perturbative gauge theory limit, in order to J

interpolate between the different ∆=J+

λ3 J5

coefﬁcients as follows:

λ2 λJ−3 λ λ3 − 3+ + ... . 5 2J 8J 16J (1 + λ)J−3

(2.15)

√ This expression agrees with both the string result (E = J 2 + λ in (1.8) with n = 1) λ λ2 3 and the perturbative gauge theory result (∆pert = J + 2J − 8J 3 + 0 × λ + ... [27]). Let us add few details about the coherent-state expectation value of higher-loop SYM dilatation operator. This expectation value appears in the action in the coherent state path integral (2.3) of the quantum spin chain. Written in terms of independent permutations or operator is Ql,k ≡ I − Pl,k = 12 (1 − σl · σk ) the “r-loop” term in the planar dilatation expected [20, 61] to contain Q in maximal power [ r+1 ], i.e. D , D ∼ Q, D3 , D4 ∼ 1 2 2

278

Q+

ARKADY A. TSEYTLIN

QQ, etc. Explicitly, D=

∞ r=0

λr Dr , (4π)2r

Dr =

J

Dr (l) ,

(2.16)

l=1

where as in (1.12) [41] and (2.14) [20] D0 = I ,

D1 = 2Ql,l+1 ,

D2 = −2(4Ql,l+1 − Ql,l+2 ) .

(2.17)

The 3-loop term [20, 61] is a special case of a 2-parameter family [21, 62] D3 (α, c) = 4(15Ql,l+1 − 6Ql,l+2 + Ql,l+3 ) + b1 Ql,l+2 Ql+1,l+3 + b2 Ql,l+3 Ql+1,l+2 + b3 Ql,l+1 Ql+2,l+3 ,

(2.18)

b1 = 4 + 2c − α ,

(2.19)

b2 = −4 − 4c + α ,

b3 = −2c + 5α .

The choice of c = 0, α = 0 corresponds to the 3-loop gauge theory operator of [20] whose form is ﬁxed [61] by the superconformal algebra and the structure of Feynman diagrams (and the BMN scaling). This choice is also consistent with integrability of the spin chain. To preserve integrability [21] one should set α = 0 (this is the parameter α2 of [21]), while to have consistency with the gauge-theory perturbative expansion [62] one should set c = 0 (c ≡ c4 in [62]). The case of α = 2, c = 0 corresponds to the operator mentioned in [21] which seemed to agree with some string-theory results. The case of α = 0, c = 1 is the “string” operator [62] which should correspond to the “string” modiﬁcation of the Bethe ansatz equations in [29]. Starting with an SU (2) coherent state satisfying (2.1) for which n|Ql,k |n =

1 1 (1 − nl · nk ) = (nl − nk )2 , 2 4

(2.20)

computing n|D|n and then taking the continuum limit (by introducing a spatial coordi2π nate 0 < σ ≤ 2π, and n(σl ) = n( 2πl J ), so that nl+1 − nl = J ∂σ n + . . ., etc.) we ﬁnd, using Taylor expansion and dropping a total derivative over σ, see [3], ˜ λ λ 1 4 2 n|D |n → + O( ∂ n ) , n 1 (4π)2 8 J2

˜= λ , λ J2

˜2 λ2 λ 1 6 2 n|D2 |n → − n + O( 2 ∂ n) , (4π)4 32 J ˜3 λ3 J 2 4 λ n|D |n → n k 3 1 (4π)6 64 (2π)2 1 + n2 + k2n2n2 + k3 (nn )2 + O( 2 ∂ 8n) . J Here k1 =

1 1 (16b1 + 9b2 + b3 ) = (14 − 3c − α) 16 8

(2.21)

(2.22)

(2.23)

SEMICLASSICAL STRINGS AND ADS/CFT 1 1 (64b1 + 45b2 + b3 ) = − (38 − 27c − 7α) , 96 48 5 5 k3 = − (32b1 + 9b2 + 5b3 ) = − (46 + 9c + α) . 48 24

279

k2 = −

(2.24)

1 Note a relation: k1 + k2 + 10 k3 = 0. The problematic scaling-violating term J 2 (∂12 n)4 in (2.23) does not cancel automatically; it should cancel after one takes into account quan˜ approximation tum corrections (which survive the continuum limit beyond the order λ [3]). Quantum corrections are expected also to be important in order to demonstrate the equivalence of the spin-chain result (for the “string” choice of c = 1, α = 0) with the string-theory result in (2.13). Verifying this equivalence beyond the quadratic n2 -terms (which obviously agree) remains an open problem. Let us also mention that one can sum up all terms in the string effective Hamiltonian that are of second order in n but to arbitrary order in σ-derivatives [3] ! 1 2 ˙ ˜ L = C · n − n 1 − λ ∂ − 1 n + O(n3 ) . (2.25) 4

This expression is in agreement with the leading-order results (2.13) and with the exact BMN spectrum (see also [1]). The coherent analogs of BMN states correspond to small ﬂuctuations near the vacuum state n0 = (0, 0, 1). On the spin chain side these correspond (in discrete version) to the microscopic spin wave excitations or magnons. 2.5. OTHER SECTORS It is possible to extend the approach of [2, 3] to other sectors of rotating string states [55, 56, 57]. First, one is to identify subsectors of operators of the SYM theory which are closed under renormalization at least to one-loop. Such bosonic subsectors are: (i) the three-spin “SU (3)” sector of string conﬁgurations with all three S 5 angular momenta (J1 , J2 , J3 ) being non-zero. They are dual to chiral operators tr(ΦJ1 1 ΦJ2 2 ΦJ3 3 )+ .... These form a set closed only under one-loop renormalization [20], but in the limit when L = J1 + J2 + J3 is large one can treat this sector as closed even beyond one loop (mixings with fermionic operators are suppressed by 1/L [59]). (ii) the “SO(6)” sector of generic (pulsating and rotating) string motions in S 5 which are dual to operators built out of 6 real scalars. Examples of pulsating string states were considered in [52, 45] and more generally in [57]; this sector will be discussed in the next section. As was pointed out in [59], one can also consider SO(3) and SO(4) subsectors of the SO(6) sector which are again closed modulo 1/L corrections. (iii) the two-spin “SL(2)” sector of string conﬁgurations with one AdS5 spin S = S1 S and one S 5 angular momentum J = J3 , which are dual to operators tr(D1+i2 ΦJ ) + ... (forming a set closed under renormalization to all orders [51, 64]). (iv) the three-spin “SU (1, 2)” sector of string conﬁgurations with two AdS5 angular S1 S2 D3+i4 ΦJ ) momenta S1 , S2 and one S 5 spin J = J3 , dual to operators tr(D1+i2 + ... which form a set closed under one-loop renormalization. Operators carrying more general combinations of non-zero spins from the list (S1 , S2 , J1 , J2 , J3 ) mix with fermionic and ﬁeld-strenth operators already at one loop and would require to consider the full superspin chains [51, 65]; on the string side one would then

280

ARKADY A. TSEYTLIN

need to include fermions of the GS action [18]. It may happen that one can again isolate some more general subsectors closed modulo 1/L corrections, but it appears that in order to apply a derivation of the reduced sigma model action on the string side one would need to impose additional constraints on the form of string conﬁgurations [56]. It is indeed straightforward to generalize [55, 56] the leading-order agreement observed in the SU (2) sector to the SU (3) sector of states with three large S 5 spins Ji , i = 1, 2, 3, ﬁnding the CP 2 analog of the CP 1 Landau-Lifsh*tz Lagrangian in (2.5),(2.12) ˜ 1 ui |2 on both the string and the spin chain sides. [56] L = −iu∗i ∂0 ui − 12 λ|D Similar conclusion is reached [56] (see also [69]) in the SL(2) sector of (S, J) states dual to operators like tr(DS ΦJ ) + .... Like in the SU (2) sector here the one-loop dilatation operator D1 may be interpreted as the XXX−1/2 Heisenberg spin chain Hamiltonian [51]. The corresponding coherent states (related to the SU (2) ones by an “analytic continuation” from the 2-sphere to the 2-hyperboloid, n → , η ij i j = −21 +22 +23 = −1) are deﬁned so that for the SL(2) generators Si one has |Si | = − 12 i and then [56] J 2λ 1 ln[ (1 − ηij il jl+1 )] (4π)2 2

|(D1 )SL(2) | =

l=1

=

J

2λ 1 ln 1 + ηij (il − il+1 )(jl − jl+1 ) . 2 (4π) 4

(2.26)

l=1

Note that this is the direct (− + +) signature analog on the classically integrable lattice Hamiltonian for the Heisenberg magnetic [58]. Since we are interested in comparing to the semiclassical string case, S as well as J should be large, and, in view of the ferromagnetic nature of the spin chain, this effectively amounts to a low-energy semiclassical limit of ˜ = λ2 we get, as in the SU (2) case, the chain. Considering the limit J → ∞ with ﬁxed λ J 2πl 2π 1 2 (σl ) = ( J ) ≡ l and l+1 − l = J ∂σ + O( J 2 ∂σ ). To the one loop order, i.e. with only one power of λ in (2.26), in expanding the logarithm we need to keep only the order 1 J 2 term, i.e. the term quadratic in ﬁrst derivatives. This leads to |(D1 )SL(2) | → J 0

2π

dσ 2π

˜ 1 λ ηij i j + O( 2 ∂ 4 ) . 8 J

(2.27)

As in the SU (2) case, the same term in the action (containing also the WZ term) is found on the string theory side. This implies the general agreement between the string and SYM ˜ in the SL(2) sector theory predictions for the energies/dimensions at leading order in λ and thus generalizes the previous results [43] found for particular solutions. 3. General fast motion in S 5 and SO(6) scalar operators One would like to try to understand the general conditions on string states and SYM operators for which the above correspondence works, and incorporate also states with large oscillation numbers. Here we will follow [56, 57] (a closely related approach was developed in [16, 17, 18]). For strings moving in S 5 with large oscillation number the λ ˜ = λ2 → 0 [52]. energy is E = L+c1 L +..., i.e. it is again regular in the limit L → ∞, λ L

SEMICLASSICAL STRINGS AND ADS/CFT

281

The leading-order duality relation between string energies and anomalous dimensions in this case was checked in [42, 45, 59]. The general condition on string solutions for which ˜ has a regular expansion in λ ˜ appears to be that the world sheet metric should E/L = f (λ) ˜ degenerate [16] in the λ → 0 limit, i.e. the string motion should be ultra-relativistic in the small effective string tension limit (in the strict tensionless limit the corresponding states become BPS [15]). For example, in the conformal gauge the 2-d induced metric in general scales as g00 = −κ2 + ... (assuming t = κτ , etc.), or, after a rescaling of the 2-d time coordinate, g00 = ˜ + ..., where we used that κ = √1 . For the fast-moving strings, the leading −1 + O(λ) ˜ λ

˜→0 O(1) term in the metric gets cancelled out, and thus the metric degenerates in the λ limit. ˜ → 0 limit each string piece is following a geodesic (big In the strict tensionless λ circle) in S 5 , while switching on tension leads to a slight deviation from geodesic ﬂow, i.e. to a nearly-null world surface [16]. The dual coherent SYM operators are then “locally BPS”, i.e. each string bit corresponds to a BPS linear combination of 6 scalars (see below). In general, the scalar operators can be written as O = Cm1 ...mL tr(φm1 ...φmL ) .

(3.1)

The planar 1-loop dilatation operator D1 acting on Cm1 ...mL was found in [41] to be equivalent to an integrable SO(6) spin chain Hamiltonian L Hnm11...n ···mL =

L λ nl nl+1 nl nl+1 δ . δ + 4δ δ ml ml+1 [ml ml+1 ] (4π)2

(3.2)

l=1

To ﬁnd the analog of the coherent-state action (2.12) we choose a natural set of coherent states Πl |vl , where at each site |v = R(v)|0. Here R is an G = SO(6) rotation and |0 is the BPS vacuum state corresponding to tr(φ1 + iφ2 )L or v(0) = (1, i, 0, 0, 0, 0), which is invariant under the subgroup H = SO(2) × SO(4). Then the rotation R(v) and thus the coherent state is parametrized by a point in G/H = SO(6)/[SO(4) × SO(2)] , i.e. v belongs to the Grassmanian G2,6 [56]. G2,6 is thus the coherent state target space for the spin chain sigma model since it parametrizes the orbits of the half-BPS operator φ1 + iφ2 under the SO(6) rotations. This is the space of 2-planes passing through zero in R6 , or the space of big circles in S 5 , i.e. the moduli space of geodesics in S 5 [17]. It can be represented also as an 8-dimensional quadric in CP 5 : a complex 6-vector vm ∗ should be subject, in addition to vm vm = 1 (and gauging away the common phase) also ˜ = λ2 and the continuum to vm vm = 0 condition. Taking the limit L → ∞ with ﬁxed λ L limit vlm (t) → vm (t, σ) we then get the G2,6 analog of the CP 1 action (2.5),(2.12) S=L

dt 0

∗ vm v m =1,

2π

dσ 2π

vm vm = 0 ,

1˜ ∗ 2 ∂t vm − λ|D −ivm σ vm | 2

,

Dσ vm = ∂σ vm − (v ∗ ∂σ v)vm .

(3.3)

282

ARKADY A. TSEYTLIN

This is also a generalization of the CP 2 action found in the SU (3) sector [56]. One may wonder how this 8-dimensional sigma model may be related to the string sigma model on R × S 5 where the coordinate space of transverse motions is only 4dimensional. The crucial point is that the coherent state action is deﬁned on the phase space (cf. the harmonic oscillator case in sect. 2.1), and 8 = (1 + 5) × 2 − 2 × 2 is indeed the phase space dimension of a string moving in Rt × S 5 . On the string side, the need to use the phase space description is related to the fact that to isolate a “fast” coordinate α for a generic string motion we need to specify both the position and velocity of each string piece. Given L = −(∂t)2 + (∂Xm )2 in conformal ˙ = 0, X˙ 2 + X 2 = κ2 , X 2 = 1) we ﬁnd that the point-like trajectories gauge (XX m (geodesics) are described by Xm = am cos α + bm sin α ,

α = κτ, a2m = 1, b2m = 1, am bm = 0 .

Equivalently, 1 ∗ ), Xm = √ (eiα vm + e−iα vm 2

1 vm = √ (am − ibm ), |v|2 = 1, v 2 = 0 , 2

where the constant vector vm thus belongs to G2,6 . In general, for near-relativistic string motions vm should change slowly with τ and σ. Then starting with the phase space Lagrangian for (Xm , pm ) 1 1 Xm , L = pm X˙ m − pm pm − Xm 2 2

(3.4)

we may change the variables according to [57] (cf. again the harmonic oscillator case) 1 ∗ Xm = √ (eiα vm + e−iα vm ), 2

i ∗ pm = √ pα (eiα vm − e−iα vm ), 2

(3.5)

where α and vm now depend on τ and σ and vm again belongs to G2,6 . There is an obvious U (1) gauge invariance, α → α − β, vm → eiβ vm . Gauge-ﬁxing the 2-d repara˜ and setting metrizations by t ∼ τ, pα ∼ L (or, doing an approximate T-duality α → α α ˜ ∼ σ as in sect. 2.3) one ﬁnds, after an additional rescaling of the time coordinate, that the phase-space Lagrangian becomes [57]:

1˜ 1 ˜ 2iα 2 (Dσ v)2 + c.c. . L = pα Dt α − λ|D σ v| − λ e 2 4

(3.6)

∗ ∂t vm and the last one averages to zero The ﬁrst term here produces the WZ term −ivm −1 ˜ → ∞. since α ≈ κτ + ..., where κ = ( λ) Equivalently, the α-dependent terms in the action (that were absent in the SU (2) or SU (3) sectors) can be eliminated by canonical transformations [57]. We then end up with the following 8-dimensional phase-space Lagrangian for the “transverse” string motions, ∗ ˜ σ vm |2 , which is the same as found on the spin chain side (3.3). ∂t vm − 12 λ|D L = −ivm The 3-spin SU (3) case is the special case when vm = (u1 , iu1 , u2 , iu2 , u3 , iu3 ), where ui belongs to the CP 2 subspace of G2,6 . The agreement between the spin chain and the string sides in this general G2,6 = SO(6)/[SO(4) × SO(2)] case explains not only the matching for pulsating solutions [52, 45] but also for near-by ﬂuctuations.

SEMICLASSICAL STRINGS AND ADS/CFT

283

Let us now discuss the reason for the restriction v 2 = 0 on the spin chain side and also clarify the structure of the coherent operators corresponding to semiclassical string states. Given the scalar operator O = Cm1 ...mL tr (φm1 . . . φmL ) one may obtain the Schr¨odinger equation for the wave function Cm1 ...mL (t) from4 ∗ ∗ n1 ...nL S = − dt iCm1 ...mL ∂t Cm1 ...mL + Cm1 ...mL Hm1 ···mL Cn1 ...nL . (3.7) In the limit L → ∞ we may consider the coherent state description and assume the factorized ansatz [57] (3.8) Cm1 ...mL = vm1 ...vmL , where each vl = {vml } (l = 1, ..., L) is a complex unit-norm 6-vector. The BPS case corresponding to the totally symmetric traceless Cm1 ...mL is represented by vl = v(0) , 2 = 0. Using (3.2) and substituting the ansatz for Cm1 ...mL into the above action one v(0) ﬁnds L S = − dt iv ∗ l ∂t vl l=1

λ ∗ ∗ ∗ ∗ + v )(v v ) + 2 − 2(v v )(v v ) , (v l l+1 l l+1 l l+1 l l+1 (4π)2

(3.9)

where we suppressed the 6-vector indices in the scalar products. As expected [41], the coherent state expectation value of the Hamiltonian (i.e. order λ term in (3.9)) vanishes for the BPS case when vl does not depend on l and v 2 = 0. More generally, if we assume that vl is changing slowly with l (i.e. vl vl+1 ), then we ﬁnd that (3.9) contains a potential term (v ∗ l v ∗ l )(vl vl ) coming from the ﬁrst “trace” structure in (3.2). This term will lead to large (order λL [41]) shifts of anomalous dimensions, invalidating low-energy expansion, ˜ = λ2 = ﬁxed, and thus i.e. prohibiting one from taking the continuum limit L → ∞, λ L from establishing direct correspondence with string theory along the lines of [2, 3, 56].5 To get solutions with small variations of vl from site to site we are thus to impose vl2 = 0 ,

l = 1, . . . , L

(3.10)

which minimizes the potential energy coming from the ﬁrst term in (3.2). This condition implies that the operator at each site is invariant under half of supersymmetries: if v 2 = 0 the matrix vm Γm appearing in the variation of the operator vm φm , i.e. δ (vm φm ) = 2i ¯(vm Γm )ψ, satisﬁes (vm Γm )2 = 0. This means that vm φm is invariant under the variations associated with the null eigenvalues [57]. One may thus call v 2 = 0 a “locally BPS” condition since the preserved combinations of supercharges in general are different for each vl , i.e. the operator corresponding to C = v1 ...vL is not BPS. Here 4

For the coherent states we consider the corresponding equation of motion may be interpreted as a (non-trivial) RG equation for the coupling constant associated to the operator O. 5 Even SU (2) sector has in general other higher-energy states with dimensions ∼ λL (in addition λ to magnons with energies ∼ Lλ2 and macroscopic spin waves with ∆ ∼ L there are also spinons with ∆ ∼ λL) but these do not correspond to fast strings – they are not captured by the low-energy continuum limit of the coherent state path integral.

284

ARKADY A. TSEYTLIN

“local” should be understood in the sense of the spin chain, or, equivalently, the spatial world-sheet direction.6 In the case when vl are slowly changing we can take the continuum limit as in [2, 3, 56] by introducing the 2-d ﬁeld vm (t, σ) with vml (t) = vm (t, 2πl L ). Then (3.9) reduces to (3.3) (all higher derivative terms are suppressed by powers of L1 and the potential term is absent due to the condition v 2 = 0). Then (3.9) becomes equivalent to the G2,6 LandauLifsh*tz sigma model (3.3) which was derived from the phase space action on the string side. To summarize, considering ultra-relativistic strings in S 5 one can isolate a fast variable α (a “polar angle” in the string phase space) whose momentum pα is large. One may ˜ ∼ Lσ, so that σ or the “operator direction” on the gauge-ﬁx pα to be constant ∼ L or set α SYM side gets interpretation of “T-dual to fast coordinate” direction. As a result, one ﬁnds a local phase-space action with 8-dimensional target space (where one can not eliminate 4 momenta without spoiling the locality). This action is equivalent to the Grassmanian G2,6 Landau-Lifsh*tz sigma model action appearing on the spin chain side. As a by-product, we thus get a precise mapping between string solutions and SYM operators representing coherent spin chain states [2, 57]. Explicit examples corresponding to pulsating and rotating solutions were given in [57]. In the continuum limit we may write the operator corresponding to the solution v(t, σ) as O = tr[ σ v(t, σ)], v ≡ vm (t, σ)φm . This locally BPS coherent operator is the SYM operator naturally associated to a ultra-relativistic string solution. The t-dependence of the string solution thus translates into the RG scale dependence of O, while the σ-dependence describes the ordering of scalar ﬁeld factors under the trace. In general, semiclassical string states represented by classical string solutions should be dual to coherent spin chain states or coherent operators, which are different from the exact eigenstates of the dilatation operator but which should lead to the same energy or anomalous dimension expressions. At the same time, the Bethe ansatz approach [41, 42, 43, 47] is determining the exact eigenvalues of the dilatation operator. The reason why the two approaches happen to be in agreement is that in the limit we consider the problem is essentially semiclassical, and because of the integrability of the spin chain, its exact eigenvalues are not just well-approximated by the classical solutions but are actually exactly reproduced by them for L → ∞, i.e. (just as in the harmonic oscillator or ﬂat space string theory case, cf. also [67]) the semiclassical coherent state sigma model approach happens to be exact. 4. Concluding remarks As discussed above, there exists a remarkable generalization of the near-BPS (BMN) limit to non-BPS but “locally-BPS” sector of string/SYM states (for related reviews see [12, 40, 22]). It remains to understand better when and why this direct relation works or fails, but the hope is to use it as a guide towards ﬁnding the string/SYM spectrum exactly in λ, at least in a subsector of states. The relation between phase-space action 6

This generalizes the argument implicit in [2]; an equivalent proposal was made in [17]. This is related to but different from the “nearly BPS” operators discussed in [15] (which, by deﬁnition, were those which become BPS in the limit λ → 0).

SEMICLASSICAL STRINGS AND ADS/CFT

285

for “slow” variables on the string side and the coherent-state action on the SYM (spin chain) side gives a very explicit picture of how string action “emerges” on the conformal gauge theory side (with the central role played by the dilatation operator). This implies not only an equivalence between string energies and SYM dimensions (established to ﬁrst ˜ but also a direct relation between the two orders in expansion in the effective coupling λ) string proﬁles and the structure of coherent SYM operators. One may try also to use the duality as a tool to uncover the structure of planar SYM theory to all orders in λ by assuming the exact correspondence between particular SYM and string states. For example, demanding the consistency with the BMN scaling limit (along with superconformal algebra) determines the structure of the full 3-loop SYM dilatation operator in the SU (2) sector [20, 21]. One can also use the BMN limit to ﬁx only part of the dilatation operator but to all orders in λ [1]. Generalizing (1.12),(2.14) and the 3- and 4-loop expressions in [20, 21] one can organize [27, 3, 1] the dilatation operator as an expansion in powers of Qk,l = 12 (I − σk · σl ) which reﬂect interactions between spin chain sites, D= Q+ QQ + QQQ + ... . Here the products Q...Q are “irreducible”, i.e. index of each site appears only once. The Q2 -terms ﬁrst appear at 3 loops, Q3 -terms – at 5 loops, etc. [20, 21]. Concentrating on the order-Q part D(1) of D one can write: D

(1)

∞ L λr = Dr (l) , (4π)r r=0

Dr (l) = 2

l=1

r

ar,k Ql,l+k ,

(4.1)

k=1

L L−1 or D(1) = l=1 k=1 hk (L, λ) Ql,l+k . Demanding the agreement with the BMN limit one can then determine the coefﬁcients ar,k and thus the function hk explicitly to all orders in λ [1]. In particular, for large L, i.e. when D acts on “long” operators, one ﬁnds D(1) = 2

L ∞

fk (λ) Ql,l+k ,

fk (λ) =

l=1 k=1

∞ r=k

λr ar,l , (4π)2r

(4.2)

where the coefﬁcients fk (λ) can be summed up in terms of the standard Gauss hypergeometric function [1] k Γ(k − 12 ) λ 1 λ 1 √ (4.3) fk (λ) = 2 F1 (k − , k + ; 2k + 1; − 2 ) , 2 4π 2 2 π 4 π Γ(k + 1) or, equivalently, 1 fk (λ) = 4π(2k − 1)

λ π2

k 0

1

u(1 − u) du 1 + πλ2 u

k−1/2

.

(4.4)

fk goes rapidly to zero at large k, so we get a spin chain with short-range interactions. One may hope that imposing additional constraints coming from correspondence with other string solutions (and using recent insights in [28, 29]) may help to determine the structure of the dilatation operator further.

286

ARKADY A. TSEYTLIN

The function fk (λ) in (4.3) smoothly interpolates between the usual perturbative ex√ pansion at small λ and fk (λ) ∼ λ behaviour at large λ. The latter is the expected behaviour of anomalous dimensions of “long” operators dual to “semiclassical” string states. Similar interpolating functions should appear also in anomalous dimensions of other SYM operators, though for “short” operators the strong-coupling asymptotics of the dimensions should be λ1/4 . Let us consider, for example, the following dimension 4 supersymmetry descendant of the Konishi operator, tr([Φ1 , Φ2 ]2 ) (which belongs to the SU (2) sector and should have the same anomalous dimension as the Konishi operator). The ﬁrst few terms in the perturbative λ-expansion of its anomalous dimension are known to be [68, 20, 33, 61] ¯ − 3λ ¯ 2 + 21 λ ¯5) , ¯ 3 − 705 λ ¯ 4 + O(λ ∆λ1 = 4 + 3λ 4 64

(4.5)

¯≡ λ . λ 4π 2 If one would to ignore all non-linear in Q terms in the dilatation operator (4.1), then the resummed expression for the anomalous dimension would be [1] ∆(1) = 4 +

3 ¯−1 , 1 + 4λ 2

which does not, however, have the expected large λ asymptotics, √ ¯ 1/4 + ... . ∆λ1 = 2λ1/4 + ... = 2 2π(λ)

(4.6)

(4.7)

Note that one cannot reproduce such asymptotics with an interpolating expression for ∆ 2 ¯ built out of rational functions of the square ! of the effective string λ = T : while the ¯ + d) at small λ would have the same expansion of a rational function (like a + bλ form as (4.5), the factors of π would not match√in the strong coupling limit (4.7). The reason for the above extra factor of π in ∆λ1 in (4.7) expressed in terms ¯ can be understood following ref. [2]. The Konishi operator should correspond to of λ the lowest-level scalar string mode. The masses of the AdS5 × S 5 string modes are, in general, non-trivial functions of the string tension (the corresponding wave equations receive α -corrections), but, in the large-tension limit, they should simply be the same as in ﬂat space, i.e. (in units where R = 1) √ √ λ ¯ 4n 2 m = = 4n λ = 8nπT , T = = λ. (4.8) α 2π Then the standard AdS5 formula ∆(∆ − 4) = m2 for the dimension of a scalar ﬁeld with mass m (again, expected to be valid in the large tension limit) predicts that for n = 1 ! 2 ¯ + ... . (4.9) ∆ − 2 = 4 + m → m + ... = 2 2π λ ¯ the dimension ∆ of the It thus appears that instead of being a rational function of λ, Konishi operator should be a transcendental function. In fact, the hypergeometric function

SEMICLASSICAL STRINGS AND ADS/CFT

287

¯ seems to be a natural candidate: one like the one appearing in (4.3), i.e. 2 F1 (a, b; c; −k λ), can choose its arguments (and an overall coefﬁcient) so that to match the powers of λ and π in both the strong and the weak coupling limits (one example with required strongcoupling asymptotics has a = −1/4, b = 3/4, c = 3/2, k = 4). With ∆ ∼ 2 F1 (or ∆ being a rational function of 2 F1 ) it is possible also to satisfy the string-theory requirement that the strong-coupling expansion should be organized as an expansion in powers of the 3 + ...) (cf. (4.9)). inverse string tension, i.e. ∆ = λ1/4 (c1 + √c2λ + (√cλ) 2 At the same time, for “long” operators with large canonical dimensions (like BMN operators or non-BPS operators discussed in the previous sections) the interpolating functions appearing in ∆ may be simple rational functions like square roots: here both the weak (gauge theory) and the strong (string theory) expansions are organized in terms of ¯ = λ2 with the leading coefﬁcients which do not contain extra powers of π (see also λ 4π the discussion below eq. (1.6)). Acknowledgements I would like to thank the organizers of the Carg`ese 2004 Summer School for the invitation to speak. I am grateful to G. Arutyunov, N. Beisert, S. Frolov, M. Kruczenski, J. Russo, A. Ryzhov and B. Stefanski for useful discussions and collaboration on the work described above. This work was supported in part by the DOE grant DE-FG02-91ER40690, the INTAS contract 03-51-6346 and the RS Wolfson award. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

11.

J. M. Maldacena, “The large N limit of superconformal ﬁeld theories and supergravity,” Adv. Theor. Math. Phys. 2, 231 (1998) [Int. J. Theor. Phys. 38, 1113 (1999)] [hep-th/9711200]. S. S. Gubser, I. R. Klebanov and A. M. Polyakov, “Gauge theory correlators from non-critical string theory,” Phys. Lett. B 428, 105 (1998) [hep-th/9802109]. E. Witten, “Anti-de Sitter space and holography,” Adv. Theor. Math. Phys. 2, 253 (1998) [hepth/9802150]. O. Aharony, S. S. Gubser, J. M. Maldacena, H. Ooguri and Y. Oz, “Large N ﬁeld theories, string theory and gravity,” Phys. Rept. 323, 183 (2000) [hep-th/9905111]. I. R. Klebanov, “TASI lectures: Introduction to the AdS/CFT correspondence,” hep-th/0009139. A. M. Polyakov, “Gauge ﬁelds and space-time,” Int. J. Mod. Phys. A 17S1, 119 (2002) [hepth/0110196]. D. Berenstein, J. M. Maldacena and H. Nastase, “Strings in ﬂat space and pp waves from N =4 super Yang Mills,” JHEP 0204, 013 (2002) [hep-th/0202021]. R. R. Metsaev and A. A. Tseytlin, “Type IIB superstring action in AdS5 × S 5 background,” Nucl. Phys. B 533, 109 (1998) [hep-th/9805028]. S. S. Gubser, I. R. Klebanov and A. M. Polyakov, “A semi-classical limit of the gauge/string correspondence,” Nucl. Phys. B 636, 99 (2002) [hep-th/0204051]. M. Blau, J. Figueroa-O’Farrill, C. Hull and G. Papadopoulos, “A new maximally supersymmetric background of IIB superstring theory,” JHEP 0201, 047 (2002) [hep-th/0110242]. R. R. Metsaev, “Type IIB Green-Schwarz superstring in plane wave Ramond-Ramond background,” Nucl. Phys. B 625, 70 (2002) [hep-th/0112044]. R. R. Metsaev and A. A. Tseytlin, “Exactly solvable model of superstring in plane wave Ramond-Ramond background,” Phys. Rev. D 65, 126004 (2002) [hep-th/0202109]. S. Frolov and A. A. Tseytlin, “Semiclassical quantization of rotating superstring in AdS5 ×S 5 ,” JHEP 0206, 007 (2002) [hep-th/0204226].

288 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.

ARKADY A. TSEYTLIN A. A. Tseytlin, “Semiclassical quantization of superstrings: AdS5 × S 5 and beyond,” Int. J. Mod. Phys. A 18, 981 (2003) [hep-th/0209116]. A. Parnachev and A. V. Ryzhov, “Strings in the near plane wave background and AdS/CFT,” JHEP 0210, 066 (2002) [hep-th/0208010]. C. G. Callan, H. K. Lee, T. McLoughlin, J. H. Schwarz, I. Swanson and X. Wu, “Quantizing string theory in AdS5 × S 5 : Beyond the pp-wave,” Nucl. Phys. B 673, 3 (2003) [hepth/0307032]. C. G. Callan, T. McLoughlin and I. Swanson, “Holography beyond the Penrose limit,” hep-th/0404007; “Higher impurity AdS/CFT correspondence in the near-BMN limit,” hep-th/0405153. D. Mateos, T. Mateos and P. K. Townsend, “Supersymmetry of tensionless rotating strings in AdS5 × S 5 , and nearly-BPS operators,” hep-th/0309114. “More on supersymmetric tensionless rotating strings in AdS5 × S 5 ,” hep-th/0401058. A. Mikhailov, “Speeding strings,” JHEP 0312, 058 (2003) [hep-th/0311019]. A. Mikhailov, “Slow evolution of nearly-degenerate extremal surfaces,” hep-th/0402067. “Supersymmetric null-surfaces,” hep-th/0404173. A. Mikhailov, “Supersymmetric null-surfaces,” hep-th/0404173. D. J. Gross, A. Mikhailov and R. Roiban, “Operators with large R charge in N = 4 Yang-Mills theory,” Annals Phys. 301, 31 (2002) [hep-th/0205066]. N. Beisert, C. Kristjansen and M. Staudacher, “The dilatation operator of N = 4 super YangMills theory,” Nucl. Phys. B 664, 131 (2003) [hep-th/0303060]. N. Beisert, “The su(2|3) dynamic spin chain,” hep-th/0310252. N. Beisert, “The dilatation operator of N = 4 super Yang-Mills theory and integrability,” hepth/0407277. “Higher-Loop Integrability in N=4 Gauge Theory,” hep-th/0409147. B. Eden, C. Jarczak and E. Sokatchev, “A three-loop test of the dilatation operator in N = 4 SYM,” hep-th/0409009. A. Santambrogio and D. Zanon, “Exact anomalous dimensions of N = 4 Yang-Mills operators with large R charge,” Phys. Lett. B 545, 425 (2002) [hep-th/0206079]. J. C. Plefka, “Lectures on the plane-wave string / gauge theory duality,” Fortsch. Phys. 52, 264 (2004) [hep-th/0307101]. A. Pankiewicz, “Strings in plane wave backgrounds,” Fortsch. Phys. 51, 1139 (2003) [hep-th/0307027]. D. Sadri and M. M. Sheikh-Jabbari, “The planewave / super Yang-Mills duality,” hep-th/0310119. I. R. Klebanov, M. Spradlin and A. Volovich, “New effects in gauge theory from pp-wave superstrings,” Phys. Lett. B 548, 111 (2002) [hep-th/0206221]. D. Serban and M. Staudacher, “Planar N = 4 gauge theory and the Inozemtsev long range spin chain,” JHEP 0406, 001 (2004) [hep-th/0401057]. N. Beisert, V. Dippel and M. Staudacher, “A novel long range spin chain and planar N = 4 super Yang-Mills,” hep-th/0405001. G. Arutyunov, S. Frolov and M. Staudacher, “Bethe ansatz for quantum strings,” hepth/0406256. K. Pohlmeyer, “Integrable Hamiltonian Systems And Interactions Through Quadratic Constraints,” Commun. Math. Phys. 46, 207 (1976). D. J. Gross and F. Wilczek, “Asymptotically Free Gauge Theories. I,” Phys. Rev. D 8, 3633 (1973). A. V. Kotikov, L. N. Lipatov and V. N. Velizhanin, “Anomalous dimensions of Wilson operators in N = 4 SYM theory,” Phys. Lett. B 557, 114 (2003) [hep-ph/0301021]. A. V. Kotikov, L. N. Lipatov, A. I. Onishchenko and V. N. Velizhanin, “Three-loop universal anomalous dimension of the Wilson operators in N = 4 SUSY Yang-Mills model,” hepth/0404092. S. S. Gubser, I. R. Klebanov and A. A. Tseytlin, “Coupling constant dependence in the thermodynamics of N = 4 supersymmetric Yang-Mills theory,” Nucl. Phys. B 534, 202 (1998) [hep-th/9805156]. S. Frolov and A. A. Tseytlin, “Multi-spin string solutions in AdS5 × S 5 ,” Nucl. Phys. B 668, 77 (2003) [hep-th/0304255]. S. Frolov and A. A. Tseytlin, “Quantizing three-spin string solution in AdS5 × S 5 ,” JHEP 0307, 016 (2003) [hep-th/0306130].

SEMICLASSICAL STRINGS AND ADS/CFT 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.

289

S. Frolov and A. A. Tseytlin, “Rotating string solutions: AdS/CFT duality in nonsupersymmetric sectors,” Phys. Lett. B 570, 96 (2003) [hep-th/0306143]. G. Arutyunov, S. Frolov, J. Russo and A. A. Tseytlin, “Spinning strings in AdS5 × S 5 and integrable systems,” Nucl. Phys. B 671, 3 (2003) [hep-th/0307191]. G. Arutyunov, J. Russo and A. A. Tseytlin, “Spinning strings in AdS5 × S 5 : New integrable system relations,” Phys. Rev. D 69, 086009 (2004) [hep-th/0311004]. A. A. Tseytlin, “Spinning strings and AdS/CFT duality,” hep-th/0311139. J. A. Minahan and K. Zarembo, “The Bethe-ansatz for N = 4 super Yang-Mills,” JHEP 0303, 013 (2003) [hep-th/0212208]. N. Beisert, J. A. Minahan, M. Staudacher and K. Zarembo, “Stringing spins and spinning strings,” JHEP 0309, 010 (2003) [hep-th/0306139]. N. Beisert, S. Frolov, M. Staudacher and A. A. Tseytlin, “Precision spectroscopy of AdS/CFT,” JHEP 0310, 037 (2003) [hep-th/0308117]. G. Arutyunov and M. Staudacher, “Matching higher conserved charges for strings and spins,” JHEP 0403, 004 (2004) [hep-th/0310182]. “Two-loop commuting charges and the string / gauge duality,” hep-th/0403077. J. Engquist, J. A. Minahan and K. Zarembo, “Yang-Mills duals for semiclassical strings on AdS5 × S 5 ,” JHEP 0311, 063 (2003) [hep-th/0310188]. M. Kruczenski, “Spin chains and string theory,” hep-th/0311203. V. A. Kazakov, A. Marshakov, J. A. Minahan and K. Zarembo, “Classical/quantum integrability in AdS/CFT,” hep-th/0402207. M. Kruczenski, A. V. Ryzhov and A. A. Tseytlin, “Large spin limit of AdS5 × S 5 string theory and low energy expansion of ferromagnetic spin chains,” hep-th/0403120. M. Lubcke and K. Zarembo, “Finite-size corrections to anomalous dimensions in N = 4 SYM theory,” JHEP 0405, 049 (2004) [hep-th/0405055]. C. Kristjansen, “Three-spin strings on AdS5 × S 5 from N = 4 SYM,” Phys. Lett. B 586, 106 (2004) [hep-th/0402033]. L. Freyhult, “Bethe ansatz and ﬂuctuations in SU(3) Yang-Mills operators,” hep-th/0405167. C. Kristjansen and T. Mansson, “The Circular, Elliptic Three Spin String from the SU(3) Spin Chain,” hep-th/0406176. N. Beisert and M. Staudacher, “The N=4 SYM integrable super spin chain”, Nucl. Phys. B 670, 439 (2003) [hep-th/0307042]. J. A. Minahan, “Circular semiclassical string solutions on AdS5 × S 5 ,” Nucl. Phys. B 648, 203 (2003) [hep-th/0209047]. A. Perelomov, ”Generalized Coherent States and Their Applications”, Berlin, Germany: Springer (1986) 320 p. W. M. Zhang, D. H. Feng and R. Gilmore, “Coherent States: Theory And Some Applications,” Rev. Mod. Phys. 62, 867 (1990). E. H. Fradkin, “Field Theories Of Condensed Matter Systems,” Redwood City, USA: Addison-Wesley (1991) 350 p. (Frontiers in physics, 82). S. Sachdev, “Quantum phase transitions”, Cambridge U. Press (1999) 352 p. R. Hernandez and E. Lopez, “The SU(3) spin chain sigma model and string theory,” JHEP 0404, 052 (2004) [hep-th/0403139]. B. J. Stefanski, Jr. and A. A. Tseytlin, “Large spin limits of AdS/CFT and generalized LandauLifsh*tz equations,” JHEP 0405, 042 (2004) [hep-th/0404133]. M. Kruczenski and A. A. Tseytlin, “Semiclassical relativistic strings in S5 and long coherent operators in N = 4 SYM theory,” hep-th/0406189. L. D. Faddeev and L. A. Takhtajan, “Hamiltonian Methods In The Theory Of Solitons,” Springer, Berlin (1987) 592 p. L. D. Faddeev, “How Algebraic Bethe Ansatz works for integrable model,” hep-th/9605187. J. A. Minahan, “Higher loops beyond the SU(2) sector,” hep-th/0405243. A. V. Ryzhov and A. A. Tseytlin, “Towards the exact dilatation operator of N = 4 super YangMills theory,” hep-th/0404215. N. Beisert, “Higher loops, integrability and the near BMN limit,” JHEP 0309, 062 (2003) [hep-th/0308074]. N. Beisert, “Spin chain for quantum strings,” hep-th/0409054. A. V. Belitsky, V. M. Braun, A. S. Gorsky and G. P. Korchemsky, “Integrability in QCD and

290

64. 65. 66. 67. 68. 69.

ARKADY A. TSEYTLIN beyond,” hep-th/0407232. A. V. Belitsky, A. S. Gorsky and G. P. Korchemsky, “Gauge / string duality for QCD conformal operators,” Nucl. Phys. B 667, 3 (2003) [hep-th/0304028]. A. V. Belitsky, S. E. Derkachov, G. P. Korchemsky and A. N. Manashov, “Superconformal operators in N = 4 super-Yang-Mills theory,” hep-th/0311104. “Quantum integrability in (super) Yang-Mills theory on the light-cone,” hep-th/0403085. S. A. Frolov, I. Y. Park and A. A. Tseytlin, “On one-loop correction to energy of spinning strings in S 5 ,” hep-th/0408187. A. Jevicki and N. Papanicolaou, “Semiclassical Spectrum Of The Continuous Heisenberg Spin Chain,” Annals Phys. 120, 107 (1979). D. Anselmi, “The N = 4 quantum conformal algebra,” Nucl. Phys. B 541, 369 (1999) [hepth/9809192]. S. Bellucci, P. Y. Casteill, J. F. Morales and C. Sochichiu, “sl(2) spin chain and spinning strings on AdS5 × S 5 ,” hep-th/0409086.

Part II

Seminars

BARYOGENESIS WITH LARGE EXTRA DIMENSIONS

KARIM BENAKLI LPTHE, Universit´e Paris 6, 4 pl Jussieu, 75252 Paris cedex, France

Abstract. Baryogenesis is a challenge, particularly for low string scale models because there is an upper bound on the reheat temperature of the Universe, and because certain baryon number violating operators must be suppressed. We review the model using horizontal family symmetries presented in [1] to emphasis its prediction for the weak angle 0.239 very close to experimental value.

In these proceedings I wish to return to the model proposed in [1]: it shows that it is in principle possible to reconcile explanation of the observed baryonic asymmetry of the universe (BAU),with the possibility of large extra dimensions (references will be found in [1]). Baryogenesis is a challenge in the class of models, mainly because the reheat temperature Treh is constrained to be so low that most existing scenarios are ruled out. For example, the reheat temperature must be less than ∼ 100 GeV for a large range of values of the string scale Ms , so the electroweak B + L violating processes are not available for baryogenesis. The low Treh creates a generic difﬁculty. One of the Sakharov conditions for baryogensis is that one needs some out-of-equilibrium dynamics. This can be found at phase transitions, or when some interaction is not fast enough to keep up with the expansion of the Universe. However when the temperature (or energy density) of the Universe is low, the expansion rate is too (H ∼ 10−18 T at T ∼GeV), so interactions have no difﬁculty keeping up with the expansion. Getting the out-of-equilibrium anywhere but a phase transition is hard. If the reheat temperature is less than ∼ 0.1 GeV, then the only phase transition available appears to be the one out of inﬂation. Another difﬁculty for baryogenesis models is the bounds on baryon number violation. For instance, to avoid fast proton decay through |∆B| = |∆L| = 1 operators, and neutron-anti-neutron oscillations through ∆B = 2 operators, one may assume that B is conserved mod 3 . This is problematic for scenarios where the BAU is generated in the out-of-equilibrium decay of a particle X. X must have at least two decay modes with different baryon number in the ﬁnal state, and approximately the same branching ratios. Otherwise the baryon asymmetry generated will be small as a consequence of CPT. If B is conserved mod 3, then X must decay to ﬁnal states with B = 1 and with B = 2 (or 293 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 293–296. © 2005 Springer. Printed in the Netherlands.

294

KARIM BENAKLI generation

q

uc

dc

ec

1 2 3

5 4 1

5 2 0

2 1 1

1 0 0

5 4 1

TABLE 1. Possible charges for the fermions and the Higgs under the horizontal U (1), for three generations. The ﬁrst generation is u, d, e, and so on. These charges generate approximately the right Yukawa couplings.

B = 0 and B = 3), so that X exchange generates a vertex that conserves B mod 3. But B = 2 operators are of higher dimension that B = 1 operators, so the branching ratio of X to the B = 2 ﬁnal state will be very small. Suppose that Ms ≤ 105 GeV, so symmetries are required to forbid the n − n ¯ operator udsuds, and the fast proton decay vertices. Then, one can consider that the BAU is generated in an out-of-equilibrium decay of a scalar particle X. It should decay out of equilibrium to ﬁnal states with different baryon numbers, with enough CP violation in the decay rates to generate a baryon to photon ratio η ∼ 3 × 10−10 . X can be the inﬂaton and it would decay before nucleosynthesis, at about the right time to reheat the Universe. Or,X is a particle generated in the reheating process, with a number density nX = δnγ . The condition such that X annihilations will be out of equilibrium at the reheat temperature and thereafter, so all the Xs will decay is that it is very weakly coupled. Large suppression factor in the couplings could originate from the volume of internal dimensions. We assume that the SM yukawa couplings are generated by some horizontal U (1) gauge symmetry, which is spontaneously broken below Ms . The quarks (q, uc , dc ) and leptons (, ec ) carry positive charges under this symmetry, and the charges are higher for the lighter fermions. The Higgs that breaks the horizontal U (1) with vev θ carries negative charge. By choosing the horizontal charges of the fermions QH f with care, one can generate approximately the right structure for the Yukawa matrices, because the interacH H tion uc uH ∼ mu u ¯u appears multiplied by (θ/Λ)Quc +Qu and tc tH ∼ mt t¯t appears H H multiplied by (θ/Λ)Qtc +Qt . Such a mechanism is probably required in models with a low Ms to avoid FCNC. It will also suppress the problematic uc dc sc uc dc sc operator: at Ms where θ is zero, it is forbidden by the horizontal symmetry (if all the fermions are positively charged), and once the horizontal symmetry is broken, uc dc sc uc dc sc can apH H H pear suppressed by (θ/Λ)2(Quc +Qdc +Qsc ) . For θ/Λ ≡ ∼ .2 and the charges in table 2, the operator uc dc sc uc dc sc will be multiplied by 16 , which is compatible with the experimental limit for Λ ≤ few TeV. The proton is stable enough provided that L is conserved mod 2. Suppose that X is a light (∼ 10 GeV) gauge singlet scalar with L = 1. It can decay to SM particles via the dimension 7 operators Xqqq and Xuc uc dc ec . These violate B respectively by 1 and -1 units, so a baryon asymmetry could be generated. We suppose that the fermions have the charges under the horizontal U (1) that are listed in table 2. In

BARYOGENESIS WITH LARGE EXTRA DIMENSIONS

295

this case the principle decay rates will be Γp¯ ∼ 18

m7X Λ6

¯B ¯ p¯ τ + ) X → cc uc bc τ c (D

(0.1)

Γp ∼ 18

m7X Λ6

X → c s b ντ (D B K p ντ )

(0.2)

m7X X → c d b ντ (D B p ντ ) (0.3) Λ6 where = θ/Λ ∼ .2. We neglect kinematics, factors of 4π, and so on, so these are very approximate estimates. However, for Λ ∼ 3 TeV, and Treh ∼ 3 MeV, we get mX ∼ 25 GeV. This is heavy enough to decay to B– and D–mesons, but light enough to (possibly) be produced in the reheating process, or to be the inﬂaton. For larger Λ, we would need a larger mX . We now need to consider whether a sufﬁcient baryon asymmetry can be generated in the decays. We assume that Γp Γp2 so we neglect Γp2 and all the other smaller decay modes. The net number of baryons produced per X particle will be Γp2 ∼ 20

¯ p¯ − Γ ¯p Γp − Γp¯ + Γ nb ≡ θCP nX Γp + Γp¯

(0.4)

¯ is the CP conjugate decay. The baryon-to-photon ratio nb /nγ ≡ η 3 × 10−10 where Γ will be nX η θCP . (0.5) nγ If X is the inﬂaton, then nX /nγ ∼ Treh /mX ∼ 10−3 . If X is produced in the reheating process, then nX /nγ = δ is a model dependent parameter. One would not expect to make more than one or two Xs in the decay of each inﬂaton, so in this case δ ≤ 10−3 . This means that we need θCP ≥ 10−7 . If we assume that the CP violation arises through loop corrections involving new particles at the scale Ms , then θCP ∼ (mX /Λ)2 ∼ 10−6 , which is approximately right. The family symmetry presented here obviously suffers from anomalies. These might be cancelled in two different ways. The ﬁrst is to assume that massive particles in a hidden sector are charged under this U (1), standard model symmetries and some hidden gauge group. The hidden symmetry might suppress any undesirable non-renormalisable operator. A more interesting possibility is to appeal to a Green-Schwarz mechanism to cancel the anomalies. If the standard model gauge couplings are all given by the vacuum expectation value of a single modulus (dilaton) then anomaly cancellation implies particular tree level relations between the couplings. For the model at hand, the strong, weak and hypercharge U (1) couplings are in the ratio 1 : 1 : 105/33 at Ms ∼ TeV instead of the usual relation 1 : 1 : 5/3 at 1016 GeV. To compare the tree level prediction with experimental measurements we see that the value obtained for sin 0.239 is extremely close to the measured one.! Here I would like to stress this point which we did not in [1]. In conclusion, if the large extra dimension models deal quite easily with particle experiments, explaining the cosmological observation, as to generate in a natural way the

296

KARIM BENAKLI

measured small or big ratio in the universe, remains a challenge. If one accepts a certain amount of ﬁne-tuning, as it is fashionable today, then models can be easily build as shown above. References 1.

K. Benakli, S. Davidson, Phys.Rev. D60 (1999) 025004, hep-ph/9810280.

BRANEWORLD COSMOLOGY ALMOST WITHOUT BRANES

GIANLUCA CALCAGNI Dipartimento di Fisica, Universit`a di Parma INFN – Gruppo collegato di Parma Parco Area delle Scienze 7/A, I-43100 Parma, Italy

Abstract. We review some general aspects of braneworld cosmologies in which an inﬂationary period driven by a scalar ﬁeld conﬁned on the brane is described by a nonstandard effective Friedmann equation. The perturbation spectra, consistency equations and observational consequences of these models are considered.

1. Introduction Motivated by recent developments in string, superstring and M theory, several models for a multidimensional target spacetime have been proposed. Among them, particular attention has been devoted to brane-world scenarios, according to which the visible universe is a (3+1)-dimensional variety (a 3-brane) embedded in a bulk with some either non-compact or compactiﬁed extra dimensions. Typically, the background metric on the brane is assumed to be the Friedmann-Robertson-Walker (FRW) metric and the Einstein equations are modiﬁed in accordance with the gravity model describing spacetime. Their projection on the brane results in the basic FRW equations for the cosmological evolution. For an introduction to the subject and some lists of references, see [1]. In this paper we review how to look for cosmic signatures of high-energy, higherderivative gravity models. In particular, the construction of a nontrivial set of consistency equations permits us to compare theoretical predictions with the perturbation spectra of the cosmic microwave background (CMB). It turns out that CMB experiments of this and next generation might be able to discriminate between the standard four-dimensional lore and the braneworld paradigm. In Secs. 2 and 3 we introduce the basic ingredients of the patch and slow-roll formalisms, taking as examples the ﬁve-dimensional RandallSundrum (RS) scenarios and their Gauss-Bonnet (GB) generalization. In Sec. 4 we outline some results on cosmological perturbations in the presence of an extra dimension and ﬁnd their observational consequences. Conclusions are in Sec. 5 297 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 297–303. © 2005 Springer. Printed in the Netherlands.

298

GIANLUCA CALCAGNI

2. Setup One of the ﬁrst problems one has to deal with when constructing braneworld models is how to stabilize the extra dimension. This can be achieved in a number of ways; in the RS example, Goldberger and Wise have provided a mechanism according to which a 5D massive scalar is put into the bulk with a potential of the same order of the brane tension λ [2]. If the energy density ρ on the brane is smaller than the characteristic energy of the scalar potential, ρ/V ∼ ρ/λ 1, then the radion is stabilized and one gets the standard Friedmann equation H 2 ∝ ρ on the brane. On the contrary, if the brane energy density is comparable with the stabilization potential, ρ/λ 1, the bulk backreacts because it feels the presence of the brane matter, the minimum of the potential is shifted and the well-known quadratic corrections to the Friedmann equation arise: H2 =

κ24 ρ(2λ + ρ) , 6λ

(2.1)

where H is the effective Hubble rate experienced by an observer on the brane and κ4 is the 4D gravitational coupling. The RS model can be viewed as a particular energy limit of a Gauss-Bonnet braneworld, −2/3 and a Friedmann equation characterized by the 5D Planck mass M5 ≡ κ5 H2 =

c+ + c− − 2 , 8α

(2.2)

where α = 1/(8gs2 ) is the Gauss-Bonnet coupling (gs is the string coupling) and, deﬁning α/2 κ25 ≡ σ0−1 , " c± =

3/2

(1 + 4αΛ5 /3)

2

1/2

+ (σ/σ0 )

#2/3 ± σ/σ0

.

(2.3)

Λ5 < 0 is the bulk cosmological constant and σ is the matter energy density which is decomposed into a matter contribution plus the brane tension λ: σ = ρ + λ. When σ/σ0 1, one gets the “pure Gauss-Bonnet” high-energy regime, H2 =

κ25 16α

2/3 ρ2/3 .

(2.4)

When the energy density is far below the 5D scale but ρ λ, one recovers the Friedmann equation of the RS scenario with vanishing 4D cosmological constant, provided that some relations among the parameters of the action are satisﬁed. Here we shall consider nonstandard cosmological evolutions on the brane and extend the RS and GB discussion to arbitrary scenarios we dubbed “patch cosmologies” [3], with H 2 = βq2 ρq .

(2.5)

βq is a constant and the exponent q is equal to 1 in the pure 4D (radion-stabilized) regime, q = 2 in the high-energy limit of the RS braneworld and q = 2/3 in the high-energy limit of the GB scenario. In order to simplify the framework, we make the following assumptions:

BRANEWORLD COSMOLOGY

299

1. There is a conﬁnement mechanism such that matter lives on the brane only, while gravitons are free to propagate in the bulk. This is guaranteed as long as ρ < M54 . 2. The contribution of the Weyl tensor is neglected. 3. The contribution of the anisotropic stress is neglected. 4. We concentrate on the large-scale limit of the cosmological perturbations. Assumption 2 closes the system of equations on the brane and sets aside the nonlocal contributions from the bulk, while assumptions 3 and 4 reduce the number of degrees of freedom of gauge invariant scalar perturbations. This list might seem too restrictive and to spoil almost all the interesting features of the model. However, assumptions 4 and 2 nicely ﬁt in the inﬂationary regime, since the long wavelength region of the spectrum, corresponding to the Sachs-Wolfe plateau, encodes the main physics of the inﬂationary era. Moreover, the dark radiation term, which is the simplest contribution of the Weyl tensor, scales as a−4 and is exponentially damped during the accelerated expansion. Finally, bulk physics mainly affects the small-scale/latetime cosmological structure and can be consistently neglected during inﬂation. This is a highly nontrivial result which has been conﬁrmed with several methods both analytically and numerically [4]. Imposing a perfect ﬂuid on the brane with equation of state p = wρ, the continuity equation governing the cosmological dynamics is the same as in four dimensions, thanks to assumption 2: ρ˙ + 3H(ρ + p) = 0 . (2.6) There are two candidates for the role of inﬂation. The ﬁrst one is an ordinary scalar ﬁeld φ with energy density and pressure ρ = φ˙ 2 /2 + V (φ) = p + 2V (φ) .

(2.7)

The second one is a Dirac-Born-Infeld (DBI) tachyon T such that ρ = V (T )/cS = −V (T )2 /p , cS ≡ 1 − T˙ 2 .

(2.8) (2.9)

From a string-theoretical point of view, the evolution of the tachyon proceeds up to the condensation T˙ → 1 into the closed string vacuum, where no signal of open excitations propagates (Carrollian limit). Also, near the minimum the strong coupling regime emerges, gs = O(1), and the perturbative description implicit in the DBI action may fail down. However, from a cosmological perspective Eq. (2.8) is a toy model and, like in standard inﬂation, an additional reheating mechanism around the condensation is required for gracefully exiting the inﬂationary period. Here we will not consider this and other (indeed solvable) problems concerning the tachyon and just implement the DBI action in the cosmological dynamics as an alternative model of inﬂation. 3. Slow-roll parameters Let ψ denote the inﬂaton ﬁeld irrespectively of its action. Expressions involving ψ will be valid for both the ordinary scalar and the tachyon. The ﬁrst-order slow-roll (SR) parame-

300

GIANLUCA CALCAGNI

ters are deﬁned as

≡−

H˙ , H2

η≡−

ψ¨ , H ψ˙

(3.1)

together with their evolution equations with respect to synchronous time

˙ = H 2 − θ, − 2η , η˙ = H η − ξ 2 ,

(3.2)

¨ ψ) ˙ · /H 2 is a second-order parameter, in the sense that it appears only where ξ 2 ≡ (ψ/ in expressions which are O( 2 , η 2 , η). Here θ, = 2 for the tachyon ﬁeld and θ, = θ ≡ 2(1 − q −1 ) for the ordinary scalar ﬁeld (4D: θ = 0; RS: θ = 1; GB: θ = −1). Note that each time derivative of the SR parameters increases the order of the SR expressions by one. One can construct inﬁnite towers of SR parameters encoding the full dynamics of the inﬂationary model. For instance, Eq. (3.1) provides the ﬁrst entries of the “Hubble” SR tower; another, sometimes more convenient tower is the potential tower deﬁned as

φV,0

≡

φV,n

≡

q V 2 , 6βq2 V 1+q 1/n 1 V (n+1) (V )n−1 , 3βq2 V nq

(3.3) n ≥ 1,

(3.4)

in the case of the normal scalar ﬁeld. The potential tower can be related to the Hubble tower by approximated relations [3]. The ﬁrst SR parameter is actually the time derivative of the Hubble radius RH ≡ H −1 . Because of its purely geometrical content, it cannot be implemented in these SR towers recursively. By deﬁnition, there is inﬂation when < 1: a ¨ = H 2 (1 − ) . a

(3.5)

Under the slow-roll approximation, if the potential term dominates over the kinetic term, then the inﬂaton slowly rolls down its potential, , η 1, and the perfect ﬂuid mimics that of a cosmological constant, p ≈ −ρ. Deviations from the de Sitter behaviour generate large-scale perturbations which explain the anisotropies in the CMB. 4. Cosmological perturbations: theory and observations Quantum ﬂuctuations of the scalar ﬁeld governing the accelerated era are inﬂated to cosmological scales because of the superluminal expansion. They constitute the seeds of both the small anisotropies observed in the microwave sky and the large-scale nonlinear structures around which gravitating matter organizes itself. For an introduction of the subject in the general relativistic case, see [5]. The standard procedure to adopt in order to compute the perturbation spetrum is: (a) Write the linearly perturbed metric in terms of gaugeinvariant scalar quantities. (b) Compute the effective action of the scalar ﬁeld ﬂuctuation and the associated equation of motion. (c) Write the perturbation amplitude as a function

BRANEWORLD COSMOLOGY

301

of an exact solution of the equation of motion with constant SR parameters. (d) Perturb this solution with small variations of the parameters. In scenarios with an extra dimension the full computation is very nontrivial due to either the extra degrees of freedom in the 5D metric and the complicated geometrical background on which to solve the Einstein equations coupled with the junction conditions on the brane. However, as explained above things become simpler when going to the largescale limit. In this case, several arguments show that the resulting spectra are, to lowest SR order, A = z(φ) = z(T ) = z(h) = Fq2

≡

k , 5πz aφ˙ , H aT˙ 1/q

cS βq H θ/2 √ 2a , κ 4 Fq

(4.1) (4.2) ,

3qβq2−θ H θ , ζq κ24

(4.3) (4.4) (4.5)

where A(h) = At is the tensor spectrum of the gravitational sector and ζq is a numerical constant which depends on the concrete gravity model one is considering: it is ζ1 = 1 = ζ2/3 for the 4D and GB cases and ζ2 = 2/3 for RS [6]. To lowest order, the scalar and tensor spectral indices are ﬁrst order in the SR parameter, nt ≡ d ln A2t /d ln k ∼ O( ) ∼ ns − 1 ≡ d ln A2s /d ln k, while their running αs,t ≡ dns,t /d ln k is second order. Here k is the comoving wave number of the perturbation and the subscripts s and t refer to scalar and tensor perturbations, respectively. In the case of exact scale invariance, ns = 1 and nt = 0. The tensor-to-scalar ratio is r ≡ A2t /A2s = /ζq + O( 2 ) .

(4.6)

Combining the SR expressions of the observables, one gets the consistency equations = −(2 + θ)ζq r + O( 2 ) , = (2 + θ)ζq r[(2 + θ)ζq r + (ns − 1)] ,

(4.7) (4.8)

αs (φ) ≈ ζq r[4(3 + θ)ζq r + 5(ns − 1)] , αs (T ) ≈ (3 + θ)ζq r[(2 + θ)ζq r + (ns − 1)] .

(4.9) (4.10)

nt αt

The key point is that the set of consistency relations is not degenerate when considering different patches θ and θ . The only known (accidental) degeneracy is for Eq. (4.7) in the RS and 4D case, where nt = −2r at ﬁrst SR order. However, the second-order version of this equation together with the expressions for the runnings deﬁnitely break the degeneracy. This implies that, at least in principle, braneworld scenarios can be discriminated between each other. To quantify the effect of the extra dimension, we can use the recent CMB data coming from WMAP [7]. With the upper bound r < 0.06 for the tensor-to-scalar ratio and the

302

GIANLUCA CALCAGNI

best-ﬁt value ns ≈ 0.95 for the scalar spectral index, the relative scalar running in two different patches is αs(θ,ψ) − αs(θ ,ψ ) ∼ O(10−2 ) , (4.11) which is close to the WMAP estimate of the experimental error. This estimate will be highly improved by either the updated WMAP data set and near-future experiments, including the European Planck satellite, for which the forecast precision should be ameliorated by one order of magnitude, ∆αs ∼ O(10−3 ). 5. Conclusions In this paper we have summarized some results on braneworld inﬂation and their observable consequences. We have not presented a full 5D calculation but we expect that bulk physics would not dramatically improve large-scale results [4]. The study of the microwave background could give the ﬁrst clues of a wider spacetime in the next years or even months. In addition to the brane conjecture, one may insert other exotic ingredients, borrowed from string and M theory, that may give rise to characteristic predictions, although at the price of increasing the number and complexity of concurring models. For instance, the introduction of a noncommutative scale can generate a blue-tilted spectrum and explain, at least partially, the low-multipole suppression of the CMB spectrum detected by WMAP (e.g. [8]). It would be interesting to ﬁnd new cosmological scenarios with θ = 0, ±1 and exploit the compact formalism provided by the patch formulation of the cosmological dynamics. Certainly there could be a lot of work for M/string theorists in this direction. A ﬁnal important question is in order: Will the CMB be the smoking gun of extra dimensions? In the context of the patch formalism the answer, unfortunately, is no. Some general relativistic models may predict a set of values for the observables {n t, n s, r, α s, . . .} close to that of a braneworld within the experimental sensitivity. Even noncommutativity may not escape this “cosmic degeneracy” since, for example, a blue-tilted spectrum can be achieved by the 4D hybrid inﬂation. So we can talk about clues but not proofs about high-energy cosmologies when examining the experimental data. The subject has to be further explored in a more precise way than that provided here in order to ﬁnd out more compelling and sophisticated predictions, extending the discussion also to the small-scale region of the spectrum. Acknowledgements It is a pleasure to thank the organizers for their kind hospitality at Carg`ese Summer School 2004. References 1. 2.

V.A. Rubakov, Phys. Usp. 44 (2001) 871, hep-ph/0104152; R. Maartens, Living Rev. Relativity 7 (2004) 1, gr-qc/0312059; P. Brax, C. van de Bruck, and A.-C. Davis, hep-th/0404011. W.D. Goldberger and M.B. Wise, Phys. Rev. Lett. 83 (1999) 4922, hep-ph/9907447; W.D. Goldberger and M.B. Wise, Phys. Lett. B 475 (2000) 275, hep-ph/9911457.

BRANEWORLD COSMOLOGY 3. 4.

5. 6. 7.

8.

303

G. Calcagni, Phys. Rev. D 69 (2004) 103508, hep-ph/0402126. D.S. Gorbunov, V.A. Rubakov, and S.M. Sibiryakov, J. High Energy Phys. 10 (2001) 015, hepth/0108017; C. Gordon and R. Maartens, Phys. Rev. D 63 (2001) 044022, hep-th/0009010; K. Ichiki, M. Yahiro, T. Kajino, M. Orito, and G.J. Mathews, Phys. Rev. D 66 (2002) 043521, astro-ph/0203272; B. Leong, A. Challinor, R. Maartens, and A. Lasenby, Phys. Rev. D 66 (2002) 104010, astro-ph/0208015; K. Koyama, Phys. Rev. Lett. 91 (2003) 221301, astroph/0303108; K. Koyama, D. Langlois, R. Maartens, and D. Wands, hep-th/0408222. V.F. Mukhanov, H.A. Feldman, and R.H. Brandenberger, Phys. Rep. 215 (1992) 203. D. Langlois, R. Maartens, D. Wands, Phys. Lett. B 489 (2000) 259, hep-th/0006007; J.F. Dufaux, J.E. Lidsey, R. Maartens, M. Sami, hep-th/0404161; G. Calcagni, hep-ph/0406057. C.L. Bennett et al., Astrophys. J., Suppl. Ser. 148 (2003) 1, astro-ph/0302207; D.N. Spergel et al., Astrophys. J., Suppl. Ser. 148 (2003) 175, astro-ph/0302209; H.V. Peiris et al., Astrophys. J., Suppl. Ser. 148 (2003) 213, astro-ph/0302225; S.L. Bridle, A.M. Lewis, J. Weller, and G. Efstathiou, Mon. Not. R. Astron. Soc. 342 (2003) L72, astro-ph/0302306. G. Calcagni, Phys. Rev. D 70 (2004) 103525, hep-th/0406006; G. Calcagni and S. Tsujikawa, Phys. Rev. D 70 (2004) 103514, astro-ph/0407543.

CHAOTIC CASCADES FOR D-BRANES ON SINGULARITIES

´ FRANCO SEBASTIAN Center for Theoretical Physics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA YANG-HUI HE Dept. of Physics and Math/Physics RG, Univ. of Pennsylvania, Philadelphia, PA 19104, USA CHRISTOPHER HERZOG Kavli Institute for Theoretical Physics, University of California, Santa Barbara, CA 93106, USA AND JOHANNES WALCHER Institute for Advanced Study, Princeton, NJ 08540, USA

Abstract. We brieﬂy review our work on the cascading renormalization group ﬂows for gauge theories on D-branes probing Calabi-Yau singularities. Such RG ﬂows are sometimes chaotic and exhibit duality walls. We construct supergravity solutions dual to logarithmic ﬂows for these theories. We make new observations about a surface of conformal theories and more complicated supergravity solutions.

1. Introduction Extending the revolutionary AdS/CFT correspondence [1] beyond the original relation between N = 4 SYM on N D3-branes and Type IIB supergravity (sugra) in AdS5 × S 5 with N units of RR 5-form ﬂux on the S 5 is important to understanding realistic strongly coupled ﬁeld theories such as QCD. Two standard extensions have been (1) reducing the SUSY to N = 1 by placing the D3-branes transverse to a Calabi-Yau singularity (the dual sugra background becomes AdS5 ×X 5 , where X 5 is some non-spherical horizon); and (2) breaking conformal invariance and inducing an RG ﬂow, by introducing fractional branes, i.e., D5-branes wrapped over collapsing 2-cycles of the singularity (in the sugra dual, 3-form ﬂuxes are turned on). A fascinating type of RG ﬂow is the duality cascade: Seiberg duality is used to switch to an alternative description whenever inﬁnite coupling is reached. This idea was introduced in [2] for the gauge theory on D-branes probing the conifold. 305 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 305–309. © 2005 Springer. Printed in the Netherlands.

´ FRANCO ET AL. SEBASTIAN

306

0.125

0.125

0.375

0.25

0.5

0.625

0.75

0.5

0.625

0.75

0.7

x2 0.6

x1

(a)

0.4

0.22 0.514

23(1)

23(2)

0.5

x3

0.25

0.375

0.225

0.23

0.235

0.513

0.512

(b)

231(2)

231(3)

23(2) 23(1)

0.51

232(3) 232(1)

0.511

(A)

0.509

2

0.22

0.225

0.23

0.235

1.5 xi

0.22325

1

0.22335

0.22355

0.22345

0.2295

0.22955

0.2296

0.22965

0.2297

0.5135

0.51116 0.51349

0.51114

0.5

0.51348

0.51346

0.22325

0.22335

0.22355

0.2295

0.22955

0.2296

2313(1)

2313(2)

231(3)

0.51344 0.51343

2312(3)

0.51345

231(2)

2321(2)

232(1)

0.22345

2321(3)

0.51106 0.51104

2323(2)

(B)

232(3)

0.1 0.2 0.3 0.4 0.5 0.6 0.7 t

0.51347

0.5111

2323(1)

0.51112

0.51108

2312(1)

(c) 0

0.22965

0.2297

(C)

Figure 1.

2. Cascades in coupling space There is an interesting way to look at cascading RG ﬂows. In a gauge theory described by a quiver with k gauge groups, the inverse squared couplings xi ≡ 1/gi2 are positive and deﬁne a k-dimensional cone (R+ )k . Inside this cone, the RG ﬂow generates a trajectory k dictated by the beta functions and satisfying i ri /gi2 = constant. Each step between dualizations then corresponds to a straight line in the simplex deﬁned by the intersection between this hyperplane and the (R+ )k cone. We show such a trajectory in Figure 1(A). Now, each wall of the cone corresponds to one of the gauge couplings going to inﬁnity. Therefore, whenever one of them is reached, we switch to a Seiberg dual theory at weak coupling. There will be then a different simplex associated to the dual theory. The entire cascade corresponds to a ﬂow in the space of glued simplices. From this perspective, which resembles a billiard bouncing in coupling space, one foresees that cascading RG ﬂows will exhibit chaotic behavior. 3. Duality Walls and Fractals After introducing the notion of a duality cascade, it is natural to wonder whether some supersymmetric extension of the Standard Model, such as the MSSM, can sit at the IR endpoint of a cascade. This question was posed by Matthew Strassler [3]. Generically, while trying to reconstruct such a RG ﬂow, one encounters a UV accumulation point beyond which Seiberg duality cannot proceed. This phenomenon is dubbed a duality wall and has been constructed for gauge theories engineered with D-branes on singularities [4]. Figure 1(B) shows the behavior of couplings for a cascade with a duality wall for the

CHAOTIC CASCADES FOR D-BRANES ON SINGULARITIES

307

theory on D-branes over a complex cone over the Zeroth-Hirzebruch surface F0 1 . Postponing the question of a possible UV completion of duality walls, we can study the dependence of its position on initial couplings. Illustrating with F0 , the result is remarkable and is presented in Figure 1(C). The curve is a fractal, with concave and convex cusps. Whenever we zoom in on a convex cusp, an inﬁnite, self-similar structure of more cusps emerges. One subtle point which was not emphasized in [5] involves the existence in coupling space of a codimension two surface of conformal theories for F0 and the other del Pezzo quiver gauge theories. If the number of gauge couplings is n + 2, then a naive counting of the linearly independent β-functions constrains only two combinations of gauge couplings when the theory is conformal, leaving an n-dimensional surface of conformal theories. This n-dimensional surface is parametrized on the gravity side by the dilaton and the integral of the NSNS B2 form through n − 1 independent 2-cycles. The existence of this codimension two surface may well affect the existence and behavior of the duality wall for F0 . In [4], it was assumed that a generic choice of initial couplings would lie on the conformal surface. However, if the initial conditions do not lie on the conformal surface, one expects large coupling constant corrections to the anomalous dimensions, which will in turn affect the strengths of the β-functions. 4. Supergravity Duals The main support for the idea of a cascading RG ﬂow in the original case of the conifold comes from a supergravity dual construction. This dual reproduces the logarithmic decrease in the effective number of colors towards the IR and also matches the beta functions for the gauge couplings. In [5], analog supergravity solutions were constructed describing logarithmic cascades for the gauge theories on D-branes probing complex cones over del Pezzo surfaces. The fact that this was possible is remarkable, since they were obtained without knowing the explicit metric. These supergravity solutions are of the general type studied by Gra˜na and Polchinski [6]. The general form of the metric is a warped product of ﬂat four-dimensional Minkowski space and a Calabi-Yau X ds2 = Z −1/2 ηµν dxµ dxν + Z 1/2 ds2X ,

(4.1)

The solution also carries 3-form ﬂux G3 = F3 − gis H3 . In order to preserve N = 1 supersymmetry, G3 must be supported only on X, imaginary self-dual, a (2, 1) form and harmonic. Indeed, it is possible to construct a G3 satisfying all these condition. It has the form n dr aI (η + i ) ∧ φI (4.2) G3 = r I=1

where the φI , I = 1 . .. n, are a basis of (1, 1) forms orthogonal to the K¨ahler class of the del Pezzo and η = 13 dψ + σ . The one-form σ satisﬁes dσ = 2ω, with ω the K¨ahler form on dPn , and 0 ≤ ψ < 2π is the angular coordinate on the circle bundle over dPn . 1

We refer the reader to [4, 5] for a detailed description of the associated quiver theory.

´ FRANCO ET AL. SEBASTIAN

308

The intersection product between the φI is dPn φI ∧ φJ = −AIJ , where AIJ is the Cartan matrix for the exceptional Lie algebra En . There is a different type of fractional brane associated to each φI , given by D5-branes wrapping the 2-cycle in the del Pezzo Poincar´e dual to φI . Let us now study the number of D5-branes and D3-branes associated to these solutions D5-Branes: The number of D5-branes is given by the Dirac quantization of the RR 3form F3 : aJ = 6πα M J . Hence, this family of solutions are dual to cascades in which the number of fractional branes of each type remains constant. D3-Branes: Similarly, the effective number of D3-branes is computed from F5 = F5 + ∗F5 where F5 = d4 x ∧ d(Z −1 ) and Z is the warp factor in (3.1). The factor Z satisﬁes the equation 1 ∇2X Z = − |H3 |2 . (4.3) 6 In [5], |F3 |2 was assumed to be a function only of the radius, in which case Z(r) =

2 · 34 2 2 α gs 9−n

ln(r/r0 ) 1 + 4 4 r 4r

M I AIJ M J

(4.4)

i,j

and from Diracquantization, the number of D3-branes will grow logarithmically: N = 3 I J I,J M AIJ M . However, generically, Z may depend on other coordi2π gs ln(r/r0 ) nates on the Calabi-Yau cone X.2 The function Z averaged over the other coordinates may still be logarithmic in r [7]. 5. Recent developments Recently, there has been further progress in the study of quiver theories and their sugra duals. In [8], a-maximization [9] was used to compute the volume of the 5d horizon of the dual of the dP1 gauge theory, yielding an irrational value. This result corrected previous computations in the literature and was obtained by carefully taking into account the global symmetries that are actually preserved by the superpotential. In [5], the duality cascade for dP1 was analyzed using naive R-charges that did not take into account these global symmetries. A stable elliptical region in coupling space was found with a selfsimilar logarithmic cascade. Redoing the analysis with the new R-charges, we ﬁnd the same elliptical region albeit with a slightly different shape and center. The 5d horizon for the complex cone over dP1 is called Y 2,1 and is a member of an inﬁnite family of Sasaki-Einstein geometries denoted Y p,q . They have S 2 × S 3 topology and their metrics have been explicitly constructed [10]. Furthermore, the gauge theory duals to the entire Y p,q family have been constructed [11]. These developments change profoundly the status of the AdS/CFT, providing an inﬁnite number of ﬁeld theories with explicit sugra duals. 2

We would like to thank Q. J. Ejaz for telling us about this possibility.

CHAOTIC CASCADES FOR D-BRANES ON SINGULARITIES

309

Acknowledgements We would like to thank Qudsia Jabeen Ejaz, Ami Hanany, Pavlos Kazakopoulos Igor Klebanov and Joe Polchinski for useful discussions. This research is funded in part by the CTP and the LNS of MIT and by the department of Physics at UPenn. The research of S. F. was supported in part by U.S. DOE Grant #DE-FC02-94ER40818. The research of Y.-H. H. was supported in part by U.S. DOE Grant #DE-FG02-95ER40893 as well as an NSF Focused Research Grant DMS0139799 for “The Geometry of Superstrings”. C. H. was supported in part by the NSF under Grant No. PHY99-07949. The research of J. W. was supported by U.S. DOE Grant #DE-FG02-90ER40542. S. F. would like to thank the organizers of Cargese Summer School, where this material was presented. References 1. 2. 3.

4. 5. 6. 7. 8. 9. 10. 11.

J. M. Maldacena, “The large N limit of superconformal ﬁeld theories and supergravity,” Adv. Theor. Math. Phys. 2, 231 (1998) [Int. J. Theor. Phys. 38, 1113 (1999)] [arXiv:hepth/9711200]. I. R. Klebanov and M. J. Strassler, “Supergravity and a conﬁning gauge theory: Duality cascades and chiSB-resolution of naked singularities,” JHEP 0008, 052 (2000) [arXiv:hepth/0007191]. M. J. Strassler, “Duality in Supersymmetric Field Theory and an Application to Real Particle Physics,” Talk given at International Workshop on Perspectives of Strong Coupling Gauge Theories (SCGT 96), Nagoya, Japan. Available at http://www.eken.phys.nagoyau.ac.jp/Scgt/proc/ S. Franco, A. Hanany, Y. H. He and P. Kazakopoulos, “Duality walls, duality trees and fractional branes,” arXiv:hep-th/0306092. S. Franco, Y. H. He, C. Herzog and J. Walcher, Phys. Rev. D 70, 046006 (2004) [arXiv:hepth/0402120]. M. Grana and J. Polchinski, “Supersymmetric three-form ﬂux perturbations on AdS(5),” Phys. Rev. D 63, 026001 (2001) [arXiv:hep-th/0009211]. C. Herzog, Q. J. Ejaz,and I. Klebanov, “Cascading RG Flows for New Sasaki-Einstein Manifolds,” arXiv:hep-th/0412193. M. Bertolini, F. Bigazzi and A. L. Cotrone, “New checks and subtleties for AdS/CFT and a-maximization,” arXiv:hep-th/0411249. K. Intriligator and B. Wecht, “The exact superconformal R-symmetry maximizes a,” Nucl. Phys. B 667, 183 (2003) [arXiv:hep-th/0304128]. D. Martelli and J. Sparks, “Toric geometry, Sasaki-Einstein manifolds and a new inﬁnite class of AdS/CFT duals,” arXiv:hep-th/0411238, and references therein. S. Benvenuti, S. Franco, A. Hanany, D. Martelli and J. Sparks, “An inﬁnite family of superconformal quiver gauge theories with Sasaki-Einstein duals,” arXiv:hep-th/0411264.

N = 1 GAUGE THEORY EFFECTIVE SUPERPOTENTIALS

BEN M. GRIPAIOS Rudolf Peierls Centre for Theoretical Physics, University of Oxford, 1, Keble Road, Oxford, OX1 3NP, United Kingdom

Abstract. Recent advances in computing effective superpotentials of four dimensional N = 1 SUSY gauge theories coupled to matter are described. The correspondence with matrix models and two proofs of this are discussed. By considering the case with matter transforming in the fundamental representation, a Wilsonian derivation of the VenezianoYankielowicz effective superpotential for a pure gauge theory is given using the generalized Konishi anomaly.

1. Introduction This work provides a brief introduction to recent advances in the study of four-dimensional gauge theories with the minimal amount (that is N = 1) of supersymmetry. We consider general theories of this type, with arbitrary gauge group coupled to matter in an arbitrary representation, and with an arbitrary superpotential for the matter. The aim is to compute the effective superpotential for the light gauge superﬁelds obtained (`a la Wilson) by integrating out the matter1 . Minimising this effective superpotential then yields the quantum vacua of the theory. In doing so, one discovers two remarkable things [1]. The ﬁrst remarkable thing is that the ‘integrating out’ part, which involves evaluating Feynman diagrams in superspace, reduces to the computation of planar diagrams in a matrix model. That is, the entire dependence on (super)space ‘disappears’!2 The second remarkable thing is that, having done the perturbative computation in this way to n loops, one ﬁnds (upon minimising the effective superpotential) that one has calculated the n-instanton correction to the vacuum structure. Thus, a perturbative computation leads to non-perturbative information about the physics! We stress that although the methods apply to general minimally supersymmetric fourdimensional gauge theories, they do not tell us everything about such theories. A complete speciﬁcation of the theory is given by the effective action; here one is only able to calculate 1

The matter must be massive, so that we can sensibly integrate it out. The disappearance is not a straightforward dimensional reduction however. If it were, the arguments presented here would hold in arbitrary dimensions; in fact they are speciﬁcally fourdimensional. 2

311 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 311–317. © 2005 Springer. Printed in the Netherlands.

312

BEN M. GRIPAIOS

the so-called F -terms (the effective superpotential) in the effective action. The K¨ahler, or D-terms (which lack the strong constraint of holomorphy) are not determined. In the next section, we sketch two proofs [2, 3] of the gauge theory/matrix model correspondence (conjectured in [1]) and show how the superpotentials are calculated in each case. We then show that there are matter-independent (i.e. pure gauge) contributions to the effective superpotential which are undetermined. These contributions turn out to be non-perturbative and provide the bridge between the perturbative computation and the non-perturbative physics mentioned above. In section 3 we show that these contributions can in fact be determined in this framework [4], and we do this. These pure gauge theory contributions correspond to the Veneziano-Yankielowicz superpotential [5, 6]. In the framework presented here, the effective superpotential is derived by integrating out massive matter and the Veneziano-Yankielowicz superpotential is thus a bona-ﬁde Wilsonian effective superpotential in this context. 2. The Gauge Theory/Matrix Model Correspondence We employ N = 1 superspace notations (see e.g. [7]). The gauge ﬁelds (and their superpartners) are written as components of a real vector superﬁeld V ; by acting with superspace covariant derivatives, one can form the analogue of the gauge ﬁeld strength, ¯ 2 e−V Dα eV and the gauge-invariant glueball chiral superﬁeld S ∼ trW α Wα . Wα ∼ D The matter is represented by chiral superﬁelds Φ with a tree-level matter superpotential in the action which is polynomial in the matter superﬁelds3 (2.1) d4 xd2 θWtree = d4 xd2 θgk Φk . Here, the coefﬁcients gk are called the tree-level matter couplings. We consider integrating out the matter Φ in some background glueball ﬁeld S to obtain an effective superpotential Weﬀ , which depends on S, the gk , and the gauge coupling (which we write in terms of the dimensionally-transmuted scale Λ). It was claimed in the introduction that the perturbative computation of Weﬀ reduces to the evaluation of planar diagrams in a bosonic matrix model, and one might well ask how this can be demonstrated. Two proofs have appeared. The ﬁrst [2] simply considers the contributing Feynman diagrams in superspace and shows that the momentum dependence of the bosons and their fermion superpartners cancels in all such diagrams. The only things left to consider are insertions of S, factors of gk coming from the vertices, and numerical symmetry factors. One can show that these can be obtained from planar diagrams of the matrix model F (S) Wtree (φ) exp 2 = dφ exp , (2.2) gs gs where φ are N × N bosonic matrices and S = gs N . The restriction to planar diagrams is enforced by taking the ’t Hooft limit: N 1 and gs 1, with S ﬁxed. The action 3 Only polynomials of degree three or less are renormalizable. However, since we claim that the computation of the effective superpotential reduces to a matrix model, the results must be independent of the momenta and any momentum cutoff. The results are thus independent of the UV completion of the theory and one is free to consider ‘non-renormalizable’ tree-level superpotentials.

N = 1 GAUGE THEORY EFFECTIVE SUPERPOTENTIALS

313

of the matrix model is given by the tree-level matter superpotential Wtree with the matter superﬁelds Φ replaced by bosonic matrices φ. To compute the perturbative computation to Weﬀ obtained by integrating out the matter (e.g. for gauge group SU (N )), one evaluates N

∂F (S) , ∂S

(2.3)

where F (S) is the perturbative free energy of the matrix model in the planar limit. The second proof [3] is rather different. One considers the effect of general chiral changes of variables δΦ = f (Φ, Wα ) in the path integral. These lead to anomalous Ward identities generalizing the Konishi anomaly [8, 9]. For example, the variation δΦ =

Φ (Φ) yields . ∂Φ ∂Wtree −S Φ = 0. (2.4) ∂Φ ∂Φ From the general chiral change of variables speciﬁed by f , one obtains a complete set of anomalous Ward identities for the chiral matter ﬁelds, and one can show that these are in one-to-one correspondence with the complete set of Ward identities in the matrix model (which, since the matrix model partition function (2.2) is just an integral, correspond to integration by parts identities). This establishes the correspondence between the SUSY gauge theory and the bosonic matrix model. Having established the correspondence, one can go on and calculate the effective superpotential for any given theory. To do this, one needs to solve the complete set of Ward identities to obtain the expectation values Φk appearing in the tree-level matter superpotential in terms of the background glueball superﬁeld S and the couplings gk . The effective superpotential can then be determined from the partial differential equations ∂Weﬀ = Φk , ∂gk

(2.5)

which follow from standard supersymmetry and holomorphy arguments. We note that these partial differential equations only specify the effective superpotential up to a term which is independent of the matter couplings gk , but which may depend on both S and the gauge coupling scale Λ. This term must contain any contribution to the effective superpotential coming from the pure gauge theory without matter. So let us ask the question: is there a pure gauge theory contribution? It turns out that there is, as was shown many years ago by Veneziano and Yankielowicz [5, 6] using the U (1)R symmetry of the pure gauge theory. For the gauge group SU (N ), for example, the pure gauge theory superpotential is S (2.6) Weﬀ (S, Λ) = N −S log 3 + S . Λ Such terms are non-perturbative. One way to see this is to minimise Weﬀ with respect to S. This reproduces the vacuum condensate S N = Λ3N , due to (non-perturbative) instantons [10]. In the next section, we show how such terms can be derived using the

314

BEN M. GRIPAIOS

generalized Konishi anomaly in the presence of matter discussed above [4]. This then renders the above approach self-contained, as well as providing an independent derivation of the Veneziano-Yankielowicz terms. Before doing so, one might ask where the missing terms are hidden in the perturbative approach using Feynman diagrams. One might assume that they correspond to diagrams with gauge superﬁelds in the loops; this is not really correct, since the missing terms are non-perturbative and thus cannot show up in any diagram. Intriguingly, these terms can be generated by the measure of the matrix model (i.e. the volume of the gauge group) [11, 1], though it is not at all clear why. 3. Pure Gauge Terms In order to derive the pure gauge theory contributions, we determine the effective superpotential in the case where the matter sector consists of F ﬂavours of ‘quarks’ transforming in the fundamental representation of the gauge group, which we take to be SU (N ) (though the argument can be applied to any classical Lie group). Furthermore, we choose a tree-level superpotential in which the quarks can have either zero or non-zero classical expectation values at the minima. If a quark has a non-zero vev, then since the quarks transform non-trivially under the gauge group, the gauge group must be spontaneously broken via the Higgs mechanism. By putting each of the F quarks at zero or non-zero minima, we can engineer the gauge symmetry breaking such that the unbroken gauge group is anything from SU (N ) down to SU (N − F ). We then solve the Konishi anomaly Ward identities (2.4) and the resulting partial differential equations (2.5), determining the effective superpotential in each vacuum, up to a constant term (by ‘constant’ we mean ‘independent of the tree-level matter couplings’). The tree-level matter couplings are free parameters in the theory. We vary them such that both the quark masses and the Higgs vevs (which determine the masses of the massive gauge bosons) become large. In that limit, the massive matter decouples from the unbroken low energy gauge group, and the effective superpotential contains a sum of contributions from the decoupled matter and the low energy gauge group. Once we have identiﬁed the contribution of the massive matter and discarded it, we are left with the superpotential of the low energy gauge group. This includes the constant term. Now any two distinct vacua have different unbroken gauge groups, but the same constant term. If we subtract the two superpotentials (with the massive matter discarded), the constant cancels and we are left with a difference equation for the pure gauge theory superpotential. The solution to this difference equation yields precisely the VenezianoYankielowicz terms (2.6). To determine the constant term in any theory, one then demands that Weﬀ (S, gk , Λ) reproduces the correct decoupled contributions of the unbroken gauge group and massive matter in any vacuum in the massive limit [12]. Incidentally, the fact that the matching in one vacuum correctly reproduces the superpotential in all vacua justiﬁes a posteriori the assumption that the constant term is the same for each vacuum branch. Having explained the argument, let us now carry it out. Since quarks are Dirac fermions and chiral supermultiplets contain Weyl fermions, we represent F ﬂavours of quarks by F chiral superﬁelds Qi transforming in the fundamental representation of SU (N ) ˜ j transforming in the anti-fundamental representation. and a further F chiral superﬁelds Q

N = 1 GAUGE THEORY EFFECTIVE SUPERPOTENTIALS

315

The tree-level matter superpotential is written in terms of the gauge invariant mesons ˜ j as Mij = Qi Q Wtree = mtrM − λtrM 2 .

(3.1)

mMij − 2λMik Mkj = 0,

(3.2)

The classical vacua are then

with F− eigenvalues at Mii = 0 and F+ = F − F− eigenvalues at Mii = m/2λ. If Mii ˜ i , and the gauge symmetry is broken. The low has a non-zero vev, then so have Qi and Q energy gauge group is thus broken down to SU (N − F+ ). The quantum theory has the Konishi anomaly and the classical vacua are modiﬁed to (2.4) mMij − 2λMik Mkj = δij S, with F± eigenvalues at Mii

m = 4λ

1±

+

8λS 1− 2 m

(3.3)

.

(3.4)

The partial differential equations following from holomorphy and supersymmetry are [3] ∂Weﬀ = trM , ∂m ∂Weﬀ = −trM 2 . ∂λ

(3.5)

We shall not write here the expression for the effective superpotential Weﬀ which is obtained by integrating these equations (it is rather cumbersome). Taking the limit of Weﬀ in which the quark mass m and Higgs vev m/2λ become large and subtracting the superpotentials for the vacua in which the number of Higgsed quarks is F1,2 , one obtains m2 S + (F1 − F2 ) S log 2 −S . (3.6) Weﬀ,1 − Weﬀ,2 → (F1 − F2 ) 4λ m /2λ The ﬁrst term represents the decoupled matter: it is given (according to the non-renormalization theorem) by the classical expectation value of Wtree . The second term must therefore represent the contribution of the low energy pure gauge group SU (N − F1,2 ).4 It seems peculiar that what we have identiﬁed as the superpotential of the low energy gauge group contains the matter couplings m and λ. However, these are precisely the factors needed to convert the SU (N ) gauge coupling scale Λ to the SU (N − F1,2 ) scales Λ1,2 via the scale-matching relation 2 F1 2 F2 m m 3(N −F1 ) 3(N −F2 ) 3N −F F =Λ m = Λ2 . (3.7) Λ1 2λ 2λ 4 There is a subtlety here: the glueball superﬁeld S includes the massive gauge bosons, which should be integrated out by replacing them with their vevs. However, in the decoupled limit, the massive gauge bosons have zero vevs, so the ﬁeld S is equivalent to the glueball superﬁeld of the low energy gauge group once the massive gauge bosons have been integrated out.

316

BEN M. GRIPAIOS

This relation comes from requiring that the coupling constants of the high energy theory (with dynamic matter) and the low energy theory (with matter integrated out) match at the Higgs and quark mass scales (see e.g. [13]). Replacing the matter couplings by the appropriate gauge coupling scales in this way (and discarding the massive matter) leads to the difference equation Weﬀ,1 − Weﬀ,2 =

S (N − F1 ) −S log 3 + S Λ1 with solution

Weﬀ (S, Λ) = N

−S log

S − (N − F2 ) −S log 3 + S , (3.8) Λ2 S +S Λ3

+ f (S).

(3.9)

Here, f (S) is an arbitrary function of S alone; it is independent of all other parameters. On dimensional grounds, f (S) ∝ S and one sees that the ambiguity in f (which can be re-written as a pure number multiplying Λ3N ) corresponds to the freedom to choose a renormalisation group scheme [13]. 4. Discussion The methods summarised above provide a very powerful framework in which to study gauge theories with N = 1 SUSY, and it is certainly of interest to go on and study the vacuum structure and phases of speciﬁc models. More general extensions to this work include the question of whether similar results hold in dimensions other than four [14], the extension to supergravity (rather than supergauge) backgrounds [1, 15, 16, 17, 18] and whether dynamical breaking of supersymmetry may be studied in this framework. References 1. 2. 3. 4. 5. 6. 7. 8. 9.

R. Dijkgraaf and C. Vafa, A perturbative window into non-perturbative physics, hep-th/0208048. R. Dijkgraaf, M. T. Grisaru, C. S. Lam, C. Vafa, and D. Zanon, Perturbative computation of glueball superpotentials, hep-th/0211017. F. Cachazo, M. R. Douglas, N. Seiberg, and E. Witten, Chiral rings and anomalies in supersymmetric gauge theory, JHEP 12 (2002) 071, [hep-th/0211170]. B. M. Gripaios and J. F. Wheater, Veneziano-Yankielowicz superpotential terms in N = 1 SUSY gauge theories, Phys. Lett. B587 (2004) 150. G. Veneziano and S. Yankielowicz, An effective Lagrangian for the pure N=1 supersymmetric Yang-Mills theory, Phys. Lett. B113 (1982) 231. T. R. Taylor, G. Veneziano, and S. Yankielowicz, Supersymmetric QCD and its massless limit: An effective Lagrangian analysis, Nucl. Phys. B218 (1983) 493. S. J. Gates, M. T. Grisaru, M. Rocek, and W. Siegel, Superspace, or one thousand and one lessons in supersymmetry, Front. Phys. 58 (1983) 1–548, [hep-th/0108200]. K. Konishi, Anomalous supersymmetry transformation of some composite operators in SQCD, Phys. Lett. B135 (1984) 439. K.-I. Konishi and K.-I. Shizuya, Functional integral approach to chiral anomalies in supersymmetric gauge theories, Nuovo Cim. A90 (1985) 111.

N = 1 GAUGE THEORY EFFECTIVE SUPERPOTENTIALS 10.

317

V. A. Novikov, M. A. Shifman, A. I. Vainshtein, and V. I. Zakharov, Instanton effects in supersymmetric theories, Nucl. Phys. B229 (1983) 407. 11. H. Ooguri and C. Vafa, Worldsheet derivation of a large N duality, Nucl. Phys. B641 (2002) 3–34, [hep-th/0205297]. 12. A. Brandhuber, H. Ita, H. Nieder, Y. Oz, and C. Romelsberger, Chiral rings, superpotentials and the vacuum structure of N = 1 supersymmetric gauge theories, hep-th/0303001. 13. J. Terning, Tasi-2002 lectures: Non-perturbative supersymmetry, hep-th/0306119. 14. R. Dijkgraaf and C. Vafa, N = 1 supersymmetry, deconstruction, and bosonic gauge theories, hep-th/0302011. 15. A. Klemm, M. Marino, and S. Theisen, Gravitational corrections in supersymmetric gauge theory and matrix models, JHEP 03 (2003) 051, [hep-th/0211216]. 16. R. Dijkgraaf, A. Sinkovics, and M. Temurhan, Matrix models and gravitational corrections, hep-th/0211241. 17. J. R. David, E. Gava, and K. S. Narain, Konishi anomaly approach to gravitational F-terms, hep-th/0304227. 18. B. M. Gripaios, Superpotentials for glueball and conformal supergravity backgrounds, Commun. Math. Phys. in press, hep-th/0311025.

GROMOV-WITTEN THEORY AND AUTOMORPHIC FORMS

¨ DANIEL GRUNBERG visiting Racah Institute, Jerusalem

Abstract. Here’s a quick introduction to Gromov-Witten invariants, their origin in topological strings, their relation to BPS states counting, and the automorphic properties of their generating function (GW potential). The plan is to interpolate between Morozov’s and Dijkgraaf’s lectures, i.e. between KdV hierarchies (or their solutions as τ -functions) and topological strings (or correlators of operators on a worldsheet).

1. Recap TFT Topological Field Theory studies maps f from a Riemann surface Σg of genus g (string worldsheet) to a target space X. Examples are obtained by twisting non-linear sigma models with CY threefold as targets. So what is twisting ? For a NLσM, bosons φ are simply coordinates on the target space (like the map above), while fermions squat in bundles: ψ+ ∈ K 1/2 ⊗ f ∗ (T X)

¯ 1/2 ⊗ f ∗ (T X) ψ− ∈ K

K is the canonical bundle1 on Σg . Since X is a complex manifold (even K¨ahler), we can ¯i i split ψ± into holomorphic and anti-holomorphic components, ψ± and ψ± . ±1/2 Twisting consists in multiplying the bundles by K so that now i ψ+ ∈ f ∗ (T X)

¯i ψ+ ∈ K ⊗ f ∗ (T¯X).

Idem for ψ− . There are two kinds of twisting, leading to the A- and B-models, according to whether you twist the holomorphic or anti-holomorphic components of ψ+ and ψ− in the same way or opposite. Twisting ensures that the Lagrangian can now be rewritten in very compact form (here for the A-model): f ∗ (γ) +it {Q, V } L=t /

Σg

01

2

Σg

=:d (degree of f ) 1 ie the highest power of the cotangent bundle (here: power 1, hence line bundle Ω of differential forms, c1 (K) = 2g − 2).

319 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 319–324. © 2005 Springer. Printed in the Netherlands.

¨ DANIEL GRUNBERG

320

Σg

f X

Figure 1. An instanton, ie a map from the worldsheet to the target space.

where t plays the role of coupling constant (t = 1/gs ), Q is the BRST operator and ¯ We shall incorporate t into the K¨ahler form γ on X, so that it becomes a V ∼ gi¯j ψ+ ∂φ. complex K¨ahler parameter, and Σg f ∗ (γ) = f∗ [Σg ] γ = d γ = d · t. Here, d ∈ H2 (X) is called the degree of the map f . Varying the complex structure of X will vary the metric gi¯j and hence V , which will merely generate vanishing terms of the form {Q, . . . }; the action will remain unchanged. Thus our model does not depend on the complex structure of X nor on that of the worldsheet ! This is the reason why we call it topological. 2 For the same reason, e−it{Q, V } is actually independent of t, and so we’ll go to the limit Re t → ∞ to evaluate the path integral for the action. What sort of maps (or ‘instantons’) contribute the most to the path integral for e−L = e−dt e−it{Q, V } ? Since t is large, the phase will wobble (and cancel itself out) for any value of V , except if V ¯ = 0. Since the boson φ is just the map f (I should have used the same vanishes, ie if ∂φ letter), this means only holomorphic maps contribute, ie holomorphic instantons. Warning: there are four different wordings in the literature for one and the same limit Re t → ∞: weak coupling limit (since t ∼ 1/gs ), large volume limit (since t became a K¨ahler parameter of X), holomorphic limit and topological limit (reasons above). 2. Correlation Functions What operators can we insert on the worldsheet Σg to build correlators ? Each operator will stand in correspondence with a complex subvariety of X. Choose n subvarieties Z1 , . . . , Zn ; their hom*ology classes have duals γ1 , . . . , γn ∈ H 2∗ (X). For each choice of Z (or γ), we build the operator ik i1 Oγ (p) := γi1 ...ik (f (p)) ψ+ . . . ψ+

of ghost number k, and supported on Z. The point p is a marking on Σg such that f (p) ∈ Z, representing the insertion of (the pullback of) Oγ . 3 We are now in a position do deﬁne the correlators: DφDψ e−L Oγ1 . . . Oγn , Oγ1 . . . Oγn g,d := Mg,n (X,d)

2

For the B-model, there is no dependence on the K¨ahler structure of X and on the complex structure of the worldsheet. 3 As an aside, I can disclose to you that {Q, Oγ } = −Odγ , that is, the BRST cohom*ology of the A-model TFT is isomorphic to the de Rham cohom*ology of X. In other words, the chiral ring is isomorphic to the (quantum) cohom*ology ring.

GROMOV-WITTEN THEORY AND AUTOMORPHIC FORMS

321

where Mg,n (X, d) is the moduli space of n-pointed holomorphic maps from Σg to X of degree d. We go again to the large t limit to make connection with the correlators en vogue among algebraic geometers: ev∗ (γ1 ∧ · · · ∧ γn ), γ1 . . . γn g,d := Mg,n (X,d)

where the evaluation map at each point pi is simply evi : Mg,n (X, d) → X : {mapf } → f (pi ). Again, this is to vanish unless f (pi ) ∈ Zi ∀i ! Note that from the mathematical perspective, the operators have dropped off of the picture. The enumerative meaning associated with γ1 . . . γn g,d is the ‘number of maps’ of degree d such that f (pi ) ∈ Zi . These are the Gopakumar-Vafa invariants (invariants of the complex structure of X). 3. BPS Invariants One has to be careful in your choice of differential forms γi (how many and of what degrees); after all, their total degree has to match the dimension of the moduli space to be integrated over, which is known to be dimC Mg,n (X, d)

= d · c1 (X) + (dimX − 3)(1 − g) + n = n for CY threefolds = 3d − 1 + n for rational curves (ie. g = 0) in CP2 ! deg γi = 12

Thus, when studying plane rational curves (embeddings of the line CP1 ), we could choose 3d − 1 four-forms γi [pt.] so that lhs and rhs equal 6d − 2. The latter are dual to points Zi in the complex plane, and so the geometric meaning of f (pi ) ∈ Zi is: you need to specify 3d − 1 points to end up with a ﬁnite number of rational curves (of degree d) passing through them (eg. degree 1: through 2 points passes 1 line). That in itself is quite cool to know, but you can do better: compute that number, [pt.]3d−1 0,d , via the so-called WDVV equation (asserts associativity of the quantum cohom*ology ring) and end-up with a recursion formula (by Kontsevich): 1,1,12, 620,... for d = 1, 2, 3, 4, .... Similarly, for CY threefolds, we could very well choose n = 0, ie no insertion of operators (hence no γi ), and both the integrand and integration space are 0-dimensional. In that case, the output is deﬁned to be the degree of the 0-cycle [Mg,n (X, d)] and denoted by 1g,d or anew by Ng,d . There’s no constraint f (pi ) ∈ Zi needed, ie embedded curves are already isolated, they don’t show up in inﬁnite families. Here too, the N0,d were computed (this time using Mirror Symmetry for the quintic, by Candelas et al), and yielded 2875, 4876875/8,... for d = 1, 2, .... But they are not integers! What went wrong ? In fact, N0,d are not quite the numbers of rational curves in a CY threefold, but rather the ‘numbers of maps’. If nd is the number of rational curves of degree d, then it will also contribute to N0,2d via a double cover; etc. The two quantities are linearly related: N0,d = k|d k −3 nd/k , so that now we do obtain integers: nd = 2875, 609250,... for d = 1, 2, .... For higher genus, the linear relation still holds: Ng,d = k|d,g ≥g . . . ng ,d/k where the dots are universal coefﬁcients (rational numbers, known for CY threefolds).

¨ DANIEL GRUNBERG

322

These new instanton invariants ng,d (believed to be integers by the Gopakumar-Vafa conjecture) count BPS states (or D-branes) wrapped around cycles of X. There are attempts to deﬁne them via the moduli space M of such objects: ng,d = (−1)dimM χ(M) (can be negative!). Their mathematical meaning is still shadowy. 4. GW Potentials The genus-g topological string amplitude is obtained by summing over all degrees of the instanton: 1 Oγ1 . . . Oγn g,d Fg := n! d,n≥0

The algebro-geometric version of this is the GW potential at genus g: Fg (ti ) :=

1 ti . . . tin γ1 . . . γn g,d n! i ...i 1

d,n≥0

1

n

The complex parameters ti are gathered together into one class γ = ti γi (γi form a basis of H 2∗ (X)), and the last sum is simply γ n g,d . For CY threefolds, the correlators γ n g,d vanish except when γ ∈ H 2 (X), ie a K¨ahler class, in which case they give (−d · t)n 1g,d (by the divisor axiom). So the GW potential can be more suggestively written as (−d · t)n −d·t Fg (ti ) = = 1g,d Ng,d e = exp f ∗ (γ) n! Σg d≥0

n≥0

d≥0

holo maps f

which is the starting point of topological string theory: summing over all holomorphic maps, whereby Ng,d is the size of the moduli space of maps of a given degree. Let us collect all amplitudes into a single generating function, the full GW potential or ‘free energy’: Fg λ2g−2 F (ti , λ) = g≥0

where λ is the string coupling. The partition function is then Z = eF , which can always be rewritten as an inﬁnite product – at least formally: Z = eF =

(1 − y l q d )cl,d ,

y := eiλ , q := e−t

(4.1)

l,d

Compare this with the vacuum τ -function in Morozov’s lecture in this volume: log τ = 2g−2 , which is related to the solution u of the KdV hierarchy by ∂x2 log τ = u. g≥0 Fg λ 5. KdV Hierarchy Indeed, the Virasoro conjecture claims that u := ∂t20 F is a solution of the ‘KdV hierarchy’, a single equation which can be unwound into an inﬁnite set (or hierarchy) – denoting

GROMOV-WITTEN THEORY AND AUTOMORPHIC FORMS

323

t0 by x: k=1: k=2: etc,

λ−2 ∂t1 u λ−4 ∂t2 u

= 12 uux + 14 uxxx = 12 u2 ux + 16 ux uxx +

1 12 uxxx

+

1 5 240 ∂x u

of which the ﬁrst line is the traditional KdV equation (hence the name). This KdV hierarchy for ∂t20 F is equivalent to another hierarchy for Z, namely Lm Z = 0 for m ≥ −1, where the Lm form the Virasoro algebra [Lk , Lm ] = (k−m)Lk+m (hence the name for the conjecture). Explicitly: L−1 = ∂t0 + t20 + i≥0 ti+1 ∂ti L0 = ∂t1 + i≥0 ti ∂ti L1 = ... (see again Morozov’s lectures.) If we rather work with the simpler moduli space of curves, Mg,n , instead of Mg,n (X, d), yet keep all the formalism above, then the KdV hierarchy is known to hold (proof by Kontsevich, conjecture by Witten). In this case, the correlators are sim ply γi1 . . . γin g,d := Mg,n c1 (L1 )i1 . . . c1 (Ln )i1 for the tautological line bundles4 Li (no mention of X nor of any map f anymore). 6. Automorphic Products For an example of (4.1), the product form is explicitly known for the contribution from constant maps (d = 0) in case of CY threefolds with Euler characteristic χ (see ‘melting crystals’): −χ/2 (1 − y l )l . Zconst = The reason we would like to know the explicit product form of the full partition function Z (ie know the powers cl,d ) is that such products can have enticing automorphic properties, ie the function transforms covariantly under some group (eg: modular forms have automorphism group SL(2, Z)). Borcherds has proved that the product (4.1) is automorphic under the Lorentz group O(2, s + 2; Z) if the powers cl,d are the coefﬁcients in the Taylor expansion of a modular form. This can be generalised to the lifting of Jacobi forms. Eg: Set j(τ )−744 =: c(n)q n, a modular function of weight 0. Then p−1 m>0,n∈Z (1 − pm q n )c(mn) is an automorphic form of weight 0 for the group O(2, 2; Z). Since this group is isomorphic to SL(2, Z) × SL(2, Z) · Z2 , we expect a function modular in each variable p, q and symmetric under swap of p and q. Indeed, it’s easy to show the product boils down to j(p) − j(q). This is the case s = 0 from Borcherds’ theorem. 2 Eg: More easily, since 12n θ24= 12(1 + 2q + 2q + . . . ) is a modular form of weight 1/2, the product q (1 − q ) is an automorphic form for SL(2, Z). Indeed, this is just the discriminant ∆. Similarly, you could write all Eisenstein series in product form 4 The ﬁbre of Li at each point of Mg,n (ie above each n-pointed genus-g curve Σg,n ) is simply the cotangent line at the ith marked point.

¨ DANIEL GRUNBERG

324

with powers given in a closed formula: E4 = (1 − q)240 (1 − q 2 )... . . . . Also: j(τ ) = q −1 (1 − q)1728 . . . . Now what can we learn from this ? Lots. Automorphic forms – let alone modular forms – tie the knot with virtually all corners of maths. For instance, in number theory, their coefﬁcients (in Taylor expansion of q) often have interesting Ramanu properties. c(n)q n , and discovered jan studied the generating function for partitions, q 1/24 /η =: smashing congruences: c(7n + 5) ≡ 0 mod 7, or c(11n + 6) ≡ 0 mod 11, etc. If GW invariants Ng,d or BPS invariants ng,d are contained in powers of GW potentials that have automorphic properties, its natural to expect that they have similar congruences. Wild conjecture! Tiny evidence yet! Only in few examples of heterotic string compactiﬁcations on certain CY threefold are there known results for the prepotential F0 (genus-0) with a suitable product formula for exp(F0 ). If true (or even half-true), these speculations turn would shed light on the enumerative meaning of these invariants, and on the geometry of the curves or of the moduli space they pertain to. Acknowledgements I am particularly grateful for the support from the Racah Institute that enabled my visit to Israel’s vibrant string theory community. Ditto for the organisers of the Carg`ese 2004 school. References 1.

D. Gr¨unberg, Gromov-Witten Theory from a Stringy Perspective, unofﬁcial thesis, unpublished, http://www.physik.hu-berlin.de/˜grunberg/thesis.ps (or .pdf)

BLACK HOLE INSTABILITIES AND PHASE TRANSITIONS

SEAN HARTNOLL DAMTP, Centre for Mathematical Sciences, Cambridge University Wilberforce Road, Cambridge CB3 OWA, UK

Abstract. Generalised black holes have a horizon given by an arbitrary Einstein manifold. I review a criterion for the classical stability of these black holes. Roughly, spherical horizons are stable but lemon-shaped horizons can be unstable. In Anti-de Sitter space these black holes are dual to gauge theory on a curved background given by the same Einstein manifold. The dual thermal ﬁeld theory effect is a novel phase transition induced by inhom*ogeneous Casimir pressures and characterised by a “condensation of pressure”.

1. Generalised AdS black holes The AdS/CFT correspondence relates the physics of black holes in AdSd+2 to a thermal ﬁeld theory living on the ‘boundary’ S 1 × S d . Here we discuss an instability and dual phase transition that arises when the black hole has boundary S 1 × M, with M a general Einstein manifold. Let us review the stability of generalised black holes [1, 2] with a negative cosmological constant. Generalised black holes take the form ds2 = −f (r)dt2 +

dr2 + r2 ds2d , f (r)

(1.1)

2 where f (r) = 1−(1+r+ /L2 )(r+ /r)d−1 +r2 /L2 . The event horizon is at r = r+ . If ds2d were the round metric on S d , then the spacetime would be the usual Schwarzschild black hole in AdS space. However, the d + 2 dimensional vacuum Einstein equations allow the horizon metric ds2d to be any d dimensional Einstein metric on a Riemannian manifold M Rαβ = (d − 1)gαβ , (1.2)

where α, β are indices for M. The resulting spacetime is called a generalised black hole. The asymptotic geometry of the background has now changed and the dual thermal ﬁeld theory lives on S 1 × M rather than S 1 × S d . The linearised stability of generalised black holes under perturbations gab → gab +hab has recently been investigated [1, 2, 3]. For large, r+ /L 1, generalised AdS black holes 325 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 325–328. © 2005 Springer. Printed in the Netherlands.

326

SEAN HARTNOLL

one ﬁnds

2 r+ ⇔ instability , (1.3) L2 where λL is the minimum eigenvalue of the Lichnerowicz operator on M and A2 is a positive O(1) number that may be determined numerically [2]. Recall that the Lichnerowicz operator acts on symmetric rank two tensors as

λL < −A2

(∆L h)ab = 2Rc abd hd c + Rca hc b + Rcb hc a − ∇c ∇c hab .

(1.4)

Note from (1.3) that the minimum Lichnerowicz eigenvalue needs to be very negative in order for instability to occur. Examples of Einstein manifolds with large negative Lichnerowicz eigenvalues are the B¨ohm metrics on S 5 . . . S 9 and on products of spheres [3].

Riem 2 >> R 2

Riem 2 ~ R 2

Figure 1. The B¨ohm metrics have regions of large curvature near θ = 0 and θ = θf . The Ricci scalar is constant.

The B¨ohm metrics on S 5 have the form , 22 , ds25 = dθ2 + a(θ)2 dΩ22 + b(θ)2 dΩ

(1.5)

, 2 are round metrics on spheres S 2 and S,2 . The coordinate θ has range where dΩ22 and dΩ 2 0 to θf . The Einstein equations imply some nonlinear differential equations for a(θ) and b(θ). The boundary conditions for topology S 5 are that a(0) = 0, b(0) = b0 and a(θf ) = a0 , b(θf ) = 0. In fact, the solutions to the equations on S 5 have a(θ) = b(θf − θ). There is a discrete inﬁnity of solutions, at speciﬁc values of b0 . Solutions exist for arbitrarily small b0 . As b0 → 0 the metric develops conical singularities at θ = 0, θ = θf . Thus b0 gives the resolution of the singularity. Figure 1.1 illustrates a B¨ohm metric. Also as b0 → 0 one ﬁnds that λL → −∞ [3]. Therefore if we use B¨ohm metrics as the horizon of a generalised black hole then the black hole will be unstable if the curvature at the tips of the manifold is sufﬁciently large. The connection between large curvatures and the Lichnerowicz spectrum which appears for the B¨ohm metrics is more general. On Einstein manifolds with positive curvature the minimum Lichnerowicz eigenvalue is related to the Weyl curvature. Roughly, one expects [3] (1.6) λ2L ∼ Cαβγδ C αβγδ |max. , where Cαβγδ C αβγδ |max. is the maximum value taken by Cαβγδ C αβγδ on the manifold M. One ﬁnds that the unstable mode tends to relax the curvatures of the spacetime near the horizon.

BLACK HOLE INSTABILITIES AND PHASE TRANSITIONS

327

2. Dual phase transition The criterion for instability (1.3) may be translated into a dual ﬁeld theory language using the standard AdS/CFT dictionary to give [2] T 2 < −λL × O(1)

⇔

instability ,

(2.1)

where T is the temperature of the ﬁeld theory. Thus the duality predicts a critical temperature in the dual ﬁeld theory on S 1 × M: TC2 ∼ |λL | .

(2.2)

The unstable mode of the generalised black hole is a tensor mode on M. That is, the mode has no legs in the t or r directions and satisﬁes ∇α hαβ = 0 ,

hα α = 0 .

(2.3)

Furthermore the unstable mode has ﬁnite energy and is thus the “normalisable mode” in the AdS/CFT language. Therefore the growth in hαβ due to the instability is mapped via the AdS/CFT correspondence to a growth in the expectation value of the spatial components of the energy momentum tensor hαβ ∼ eωt

⇒

< Tαβ > ∼ eωt .

(2.4)

Combining the two facts (2.2) and (2.4) suggests that the dual thermal ﬁeld theory undergoes a phase transition when cooled below the temperature TC . The phase transition is characterised by a “condensation of pressure”, that is, by the spatial components of the energy momentum tensor gaining a vacuum expectation value. The appearance of the pressure suggests a thermodynamic characterisation of the instability. Recent work on black brane instabilities has supported the Gubser-Mitra conjecture which states that translationally invariant horizons are classically stable if and only if they are locally thermodynamically stable. Local thermodynamic stability requires the Helmholtz free energy of the system to be a concave function of temperature and a convex function of volume. The known instabilities of black branes are due to a negative heat capacity ∂ 2 F < 0. (2.5) CV = −T ∂T 2 V

The large generalised AdS black holes which have been considered here always have positive heat capacity. The alternative thermodynamic instability is to have a negative isothermal compressibility [4] ∂ 2 F ∂P = −V < 0. (2.6) KT−1 = V ∂V 2 T ∂V T If this were indeed the thermodynamic instability of the dual ﬁeld theory that triggers the phase transition, then generalised black holes would provide an example of a generalisation of the Gubser-Mitra correspondence to horizons without translational invariance.

328

SEAN HARTNOLL

A negative compressibility would most likely arise in the regions of the manifold with large curvature, as this is where the Casimir pressures are greatest. The compressibility would want to relax the curvatures to reduce the Casimir effect. This mimics the behaviour of the bulk unstable mode mentioned above. The ﬁeld theory background is not dynamical, so the condensation of pressure provides an alternative means of cancelling the wrong-sign Casimir pressure. Acknowledgements It is my please to thank Gary Gibbons and Chris Pope for the initial collaborations which stimulated my interest in generalised black holes. I would like to thank also the organisers of the Carg`ese 2004 ASI. References 1. 2. 3. 4.

G. Gibbons and S. A. Hartnoll, Phys. Rev. D 66 (2002) 064024. S. A. Hartnoll, JHEP 0308 (2003) 019. G. W. Gibbons, S. A. Hartnoll and C. N. Pope, Phys. Rev. D 67 (2003) 084024. S. A. Hartnoll, Phys. Rev. D 70 (2004) 044015.

PROBING SINGULARITIES

VERONIKA E. HUBENY Department of Physics & LBNL University of California, Berkeley, CA 94720, USA

Abstract. We review the method of holographically extracting information about the singularity of a Schwarzschild-AdS black hole and contrast this with probing cosmological singularities.

1. Introduction Curvature singularities are a rather generic feature in general relativity. As demonstrated by the singularity theorems [1], even relatively mild initial data can evolve to a singularity. Indeed, in nature, we expect singularities inside black holes as well as cosmological ones. Of course, the laws of general relativity break down at a singularity, thereby precluding us from evolving the classical spacetime into its causal future. But the real world is not classical, and it should not suffer from singularities; in fact we expect quantum effects to become important long before the actual singularity is reached. Unfortunately, to answer the intriguing and fundamental question of what actually happens requires understanding of quantum gravity, which is beyond the scope of our present knowledge. Nevertheless, probing singularities within the quantum framework which is available— namely string theory—should provide us with further clues to the full theory of quantum gravity. String theory has already proved very successful in resolving many timelike singularities. In simple examples like orbifold singularities, perturbative string dynamics sufﬁces to explain the resolution, while more complicated examples such as the conifold singularity require stringy non-perturbative effects. Spacelike singularities are even more subtle, and we expect that a fully non-perturbative formulation of string theory will be essential for their resolution. While there already exist several remarkable nonperturbative formulations of string theory, the one best suited for our purposes is the well-known AdS/CFT correspondence [2]. In particular, we will use the duality between 10-dimensional IIB string theory on asymptotically AdS5 × S 5 spacetimes and 4-dimensional N = 4 Super Yang Mills gauge theory, living on the boundary of AdS. Since the bulk has gravity, it can contain black holes, as well as cosmological regions. Once we discover how these are represented by 329 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 329–335. © 2005 Springer. Printed in the Netherlands.

330

VERONIKA E. HUBENY

the dual (non-gravitational) gauge theory, we will be able to learn how the respective singularities are resolved. In the following section, we review the method proposed by FHKS [3] of probing the spacelike singularity inside a Schwarzschild-AdS black hole. Although one can extract speciﬁc information about the singularity from certain correlators in the dual CFT, this extraction is rather subtle and in particular requires analytic continuation through a branch cut. This reﬂects the fact that the singularity is hidden behind an event horizon, and therefore is not directly “visible” from the boundary where the CFT resides. On the other hand, if the singularity could be directly visible from the boundary, one might expect the CFT to encode it more directly and hence provide a more explicit mechanism for singularity resolution. HHM [3] had presented scenarios which seemed to guarantee the requisite naked singularities; however after more detailed study [5], we believe it unlikely that one will be able to ﬁnd a clean example with naked singularities (see M. Rangamani’s talk in this volume). The set-up of [3], however, included a cosmological region in the bulk, with Big Bang and Big Crunch singularities, cast within the AdS/CFT framework. One may therefore ask whether one can revert to the methods of FHKS to probe these cosmological singularities. In section 3, we show that the answer is no; the black hole and cosmological singularities in fact behave very differently from each other. 2. Black hole singularity To study a singularity and an event horizon in the bulk gravity theory, we will consider the Schwarzschild-AdS geometry, describing a large black hole in AdS, whose gauge dual corresponds to an approximately thermal state. By analyzing the properties of this state, such as the expectation values of various operators, we can hope to extract information about the black hole. Indeed, due to the powerful nature of the AdS/CFT duality, we expect to be able to decode the full quantum near-singularity bulk physics from the gauge dual. However, since such decoding requires that the gauge theory encodes behind-thehorizon physics in the ﬁrst place, we should ﬁrst ask, how much of the bulk physics does the gauge dual encode? Naive considerations might lead one to expect that the CFT encodes only the region of the bulk which is causally connected to the boundary, i.e. outside the event horizon. On the other hand, event horizon is a global object (deﬁned as the boundary of the past of future inﬁnity), which means that we cannot determine the presence or position of the event horizon without knowing the entire future evolution of the spacetime. Hence if the horizon were to bound what the CFT can encode, then the AdS/CFT correspondence would have to be very nonlocal in time. We illustrate this point by a simple gedanken-experiment [6]: The holographic nature of the AdS/CFT correspondence suggests that as a boundary observer, one can obtain instantaneous information about the event in the bulk, for instance by measuring appropriately decorated Wilson loop [7]. Once this information is obtained, i.e. entirely to the future of this measurement, the boundary observer may collapse in a shell which forms a black hole whose horizon will have encompassed the measured event. While the above argument indicates that the CFT should encode at least some physics inside a horizon, it does not lend itself to detailed computational analysis. Instead, it is

PROBING SINGULARITIES

331

more fruitful to turn to a simpler set-up, namely that of the eternal black hole, which is static outside the horizon and has both future and past singularities, as well as two boundaries. This geometry was analyzed previously in 3 dimensions by a number of authors [8, 9, 10]; here we study the higher-dimensional analog [3]. The real-time or thermoﬁeld formalism for thermal ﬁeld theory [11] is especially well-suited for this analysis. One copy of the CFT resides on each of the two asymptotic regions; the CFTs are noninteracting but entangled through the Hartle-Hawking state [9]. In this approach, a single boundary thermal description is recovered by tracing over the Hilbert space of the other boundary CFT. Moreover, as demonstrated previously for 3 dimensions [10], one can probe physics behind the horizon by studying the correlator of two operators, one on each asymptotic boundary, each creating a large mass bulk particle. As the mass m → ∞, the correlator can be evaluated in the semi-classical geodesic approximation and is given by exp(−mL), where L is the (regularized) proper length of the spacelike geodesic joining the boundary points. Because the geodesic passes through spacetime regions inside the horizon, this boundary correlator reveals information about the geometry behind the horizon. Unfortunately, the three dimensional case, which is simple enough to study analytically, is rather special. The geometry is locally that of pure AdS3 , and the black hole singularity is merely a result of the orbifold nature of this geometry [12]. Consequently, the geodesics are not sensitive to the position of the singularity, and correspondingly the correlation function is relatively structureless. On the other hand, this situation changes drastically in higher dimensions. In particular, the singularity in d > 3 dimensions approaches that of the d-dimensional Schwarzschild black hole singularity, which has a dramatic effect on the spacelike geodesics: pictorially, spacelike radial geodesics “bounce off” from the singularity. For simplicity, we focus on d = 5 where the boundary CFT is four dimensional N = 4 SYM; however, our results are qualitatively similar to those in other higher dimensions as well. The ﬁrst surprise, underlining the marked difference from the previously-studied 3dimensional case, is that the Penrose diagram of the Schw-AdS5 spacetime is different. While the Schw-AdS3 geometry has a square-shaped Penrose diagram, this can no longer hold in higher dimensions [3]. One reason, apparent from Fig.1.1, is that outgoing radial null geodesics starting at the past singularity do not reach the boundary at a timesymmetric point. Apart from the differences in the Penrose diagrams, there are two important features in the symmetric radial spacelike geodesics in higher dimensions which are absent in the 3-D case: There exists a particular time tc beyond which there are no geodesics connecting the two boundaries, and the geodesics cross each other. This has an important consequence for our present goal of ﬁnding a signature of the singularity in the dual ﬁeld theory. Consider the boundary to boundary correlator φφ(t) between two high-dimension operators inserted in a symmetric fashion on the two boundaries at time t. Since this correlation function is related to the proper length of a spacelike geodesic connecting the two points, we expect that it reﬂects the special behaviour of the geodesics as t → tc . In fact, direct evaluation suggests that we should see a pole in the correlation function: φφ(t) ∼ e−mL(t) ∼

1 (t − tc )2m

as

t → tc .

(2.1)

332

VERONIKA E. HUBENY

t=0

tc

(a)

(b)

Figure 1. Penrose diagrams of Schw-AdS in (a) 3 and (b) 5 dimensions. The top and bottom curves are the future and past singularities, the vertical lines are the boundaries, the dashed diagonal lines are the event horizons, and the remaining curves are spacelike radial geodesics.

This corresponds to a light-cone singularity in the ﬁeld theory, since the geodesics are becoming almost null. Had this been the full story, we would have a striking signature of the black hole singularity in the gauge theory dual. However, general considerations of the boundary ﬁeld theory rule this out; one can easily show that |φφ(t)| ≤ |φφ(0)| < ∞. What went wrong? In evaluating Eq.(2.1) we assumed that the correlator is dominated by a single geodesic, namely the real, “bounce” geodesic shown in Fig.1.1b. But, in fact, this geodesic does not dominate the correlator, for there are in general multiple geodesics that connect the two boundary points. One indicator of this fact is the intersection of nearby geodesics around t = 0; in a cut-off ﬁeld theory there would be three geodesics connecting the two points at (cut-off) boundaries. Alternately, in a Euclidean picture, one real and two complex geodesics contribute. At t = 0 their proper distances coincide, creating a branch point in the correlator which behaves as L ∼ t4/3 for small t. (The 3 in the denominator of the exponent, implying a 3-sheeted Riemann surface, corresponds to the 3 geodesics.) By studying various resolutions of this branch point, one can show that as t increases from 0, the correlator deﬁned by the boundary CFT is given by a symmetric sum of the two complex branches of this expression, each attributed to a complex geodesic in the complexiﬁed spacetime. But the correlator is an analytic function of t and can be continued onto the real sheet. This is the essential feature of the CFT which enables us to extract information which is not directly measurable. On the real sheet, the “light cone singularity” does appear. So the boundary correlator does contain information about the singularity, albeit in a subtle way. 3. Cosmological singularity In the preceding section we have seen that despite the singularity being spacelike, we can study it in the CFT by analyzing analytically continued correlators which correspond to

PROBING SINGULARITIES

333

spacelike geodesics which bounce off the singularity. Let us now contrast this story with a different type of spacelike singularity, namely a cosmological (Big Crunch) singularity. Since one can embed cosmological singularities into an AdS/CFT set-up [3], one might expect that CFT correlators, related to bulk geodesics as above, would similarly yield information about the singularity. Surprisingly, this is not the case, due to a classical GR distinction between these two scenarios. Namely, spacelike geodesics can bounce off the black hole singularity, but they cannot bounce off the cosmological one. To see this, let us for simplicity consider spherically symmetric spacetimes and focus on the radial geodesics. Writing the radial part of the metric suggestively as ds2 = −dt2 + a2 (t) dr2 ,

(3.1)

we obtain the t-component of the geodesic equation: t¨ + a(t) a (t) r˙ 2 = 0

(3.2)

d d where ˙ ≡ dλ (λ being the afﬁne parameter along the geodesic), and ≡ dt . In order for the geodesic to bounce off from a (future) spacelike singularity, there must be a point along the geodesic at which t˙ = 0 while t¨ < 0. Evidently, the ﬁrst condition can be satisﬁed for spacelike geodesics; whether the second condition can be satisﬁed as well depends on the sign of a (t) (since a(t) > 0), as dictated by Eq.(3.2). Consider now a hom*ogeneous FRW cosmology with a Big Crunch. Then a(t) is simply the scale factor. Near the future singularity the space is imploding, so that a (t) < 0, which implies that t¨ > 0 everywhere along the geodesic. Hence, spacelike geodesics cannot bounce off the singularity in these cosmologies.1 Since spacelike geodesics cannot bounce, they will hit the singularity, which makes these FRW cosmologies spacelike geodesically incomplete. As we have seen, this is very different from the behaviour in the black hole geometry.2 As emphasized in the previous section, the classical feature that spacelike geodesics bounce off from the black hole singularity was crucial for extracting information about the Schwarzschild-AdS black hole singularity via the CFT [3]; in fact, studying the CFT correlators may provide further insight into the singularity resolution in quantum gravity. On the other hand, this feature of the singularity being absent in the cosmological case, the corresponding singularity resolution should be very different from that of the black hole3 .

1 Note that to bounce off from a past Big Bang singularity, we would need t¨ > 0 when t˙ = 0; but this cannot be satisﬁed either, because in that region a (t) > 0. 2 In order to write the metric in the form of Eq.(3.1), we need to perform a change of coordinates from the usual Schwarzschild form; near the singularity, this yields a(t) ∼ −1/t, so that a (t) > 0, which allows t¨ < 0 when t˙ = 0. In other words, the Schwarzschild “time” direction, which is spacelike inside the horizon, is actually expanding as one approaches the future singularity. 3 This argument assumes that the difference in the behaviour of certain speciﬁc probes of the singularity are manifestations of the difference in the actual singularity itself; this is by no means guaranteed, merely suggestive.

334

VERONIKA E. HUBENY

4. Discussion In summary, we have seen that the CFT encodes physics behind the horizon, including the near-singularity region. Unlike the previously-studied three dimensional case, in higher dimensions there is a genuine curvature singularity which is bowed in on the Penrose diagram, leading to a special behaviour of the geodesics as a critical time tc is approached. Despite subtleties related to the nontrivial analytic structure of L(t), the CFT correlators reveal distinct signals of the black hole singularity. In fact, given4 the CFT data φφ, the properties of the singularity are computationally accessible [3]. Since the “tc ” singu1 and gs , as well as for small ls , we can also extract the larity persists to all orders in m stringy and quantum behaviour near the black hole singularity. Thus, using analyticity, we have demonstrated that a signiﬁcant amount of information from behind the horizon, and in particular from near the singularity, is encoded in boundary theory correlators. Since analyticity was essential for our arguments, it would be worthwhile to understand its implications in the context of AdS/CFT at a much deeper level. We have given an example of how we can learn about (and presumably resolve) spacelike singularities in the bulk, using AdS/CFT. Hitherto, it has been commonly assumed that understanding the black hole singularity will help us understand the cosmological Big Bang or Big Crunch singularities as well. We have seen that this may not be a wellfounded expectation; the black hole singularities should be encoded very differently from the cosmological ones. The fact that, even at the classical level, the near-singularity geodesic properties differ may be viewed as a hint to the singularity resolutions being different. As already evidenced in our experience with timelike singularities, there are many different channels of resolution depending on the details of the singularity in question; it is not surprising that this situation has an analog in the context of spacelike singularities. Acknowledgements I wish to thank my collaborators, L. Fidkowski, M. Kleban, and especially M. Rangamani and S. Shenker. I would also like to thank also the organizers of the Carg`ese 2004 ASI, and the Kavli Institute for Theoretical Physics for hospitality. This work is supported in part by the funds from the Berkeley Center for Theoretical Physics, and also by the DOE grant DE-AC03-76SF00098 and the NSF grants PHY-0098840 and PHY99-0794. References 1. 2. 3. 4. 5.

S. W. Hawking and G. F. R. Ellis, The Large Scale Structure of Space-Time, Cambridge University Press, (1973). O. Aharony, S.S. Gubser, J. Maldacena, H. Ooguri, Y. Oz, Phys. Rept. 323 (2000) 183, hepth/9905111. L. Fidkowski, V. Hubeny, M. Kleban and S. Shenker, JHEP 0402, 014 (2004) hep-th/0306170. T. Hertog, G. T. Horowitz and K. Maeda, Phys. Rev. D 69, 105001 (2004) hep-th/0310054. V. E. Hubeny, X. Liu, M. Rangamani and S. Shenker, hep-th/0403198.

4 Of course, it is not presently feasible to obtain the CFT “data” φφ in the relevant regime by a direct computation: the CFT is strongly coupled, and m → ∞ implies inﬁnitely large dimension operators. Nevertheless, it is still of use to ask the matter-of-principle question: what does a given CFT information tell us about the bulk?

PROBING SINGULARITIES 6. 7. 8. 9. 10. 11. 12.

335

V. E. Hubeny, Int. J. Mod. Phys. D 12, 1693 (2003) hep-th/0208047. L. Susskind and N. Toumbas, Phys. Rev. D 61, 044001 (2000) hep-th/9909013. B. Freivogel, S. B. Giddings and M. Lippert, Phys. Rev. D 66, 106002 (2002) hep-th/0207083. V. Balasubramanian and S. F. Ross, Phys. Rev. D 61, 044007 (2000) hep-th/9906226. J. Louko, D. Marolf and S. F. Ross, Phys. Rev. D 62, 044041 (2000) hep-th/0002111. J. M. Maldacena, JHEP 0304, 021 (2003) hep-th/0106112. P. Kraus, H. Ooguri and S. Shenker, Phys. Rev. D 67, 124022 (2003) hep-th/0212277. M. Le Bellac, Thermal Field Theory, Cambridge University Press (1966). M. Banados, C. Teitelboim and J. Zanelli, Phys. Rev. Lett. 69, 1849 (1992) hep-th/9204099.

INFLATION UNLOADED

MATTHEW KLEBAN Institute for Advanced Study, Einstein Drive Princeton, NJ 08540, USA

Abstract. I present a brief review of astro-ph/0406099, which argues that there is a limit on the number of efolds of inﬂation which are observable in a universe which undergoes an eternally accelerated expansion in the future. Such an acceleration can arise from an equation of state p = wρ, with w < −1/3, and it implies the existence of event horizons. In some respects the future acceleration acts as a second period of inﬂation, and “initial perturbations” (including signatures of the ﬁrst inﬂationary period) are inﬂated away or thermalize with the ambient Hawking radiation. Thus the current CMB data may be looking as far back in the history of the universe as will ever be possible even in principle, making our era a most opportune time to study cosmology.

1. Introduction Quantum ﬂuctuations produced during inﬂation are imprinted on the curvature, and are subsequently stretched by inﬂation to super (Hubble) horizon scales1 . Once there, they “freeze out”, i.e. their amplitude approaches a constant set by the horizon crossing condition, and their wavelength scales with the particle horizon, λ(t) = λ0 a(t)/a0 . What happens next depends on the subsequent evolution of a(t). When inﬂation ends, the Hubble horizon begins to grow linearly in time, but the wavelength stretches more slowly, as λ(t) ∼ a(t). If the vacuum energy is zero, this situation will persist indeﬁnitely, and the Hubble horizon eventually catches up with the perturbation (see the left panel of Fig. 1), after which it can collapse and form structure. A patient observer in such a universe can see arbitrarily far back in time by measuring these perturbations: the longer she waits, the farther back during inﬂation the ﬂuctuations she sees were generated. However, if at some time the post-inﬂationary universe begins to accelerate (and continues to do so forever), there will be event horizons [2]. In this case a part of the global spacetime is permanently inaccessible to any given observer, and the evolution of perturbations is very different. Depending on when they were produced, inﬂationary ﬂuctuations may or may not re-enter the Hubble horizon during matter domination (see the right panel 1 Because space is tightly constrained in these proceedings, I have included only the most immediately relevant references. Please see [1] for more.

337 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 337–340. © 2005 Springer. Printed in the Netherlands.

338

MATTHEW KLEBAN

Figure 1.

Evolution of the wav elengths of some typical inﬂationary perturbations in a uni v erse without (left panel) and with (right panel) event horizons. In the left panel, all ﬂuctuations ev entually reenter the Hubble horizon. In the right panel, in the case a), a ﬂuctuation is stretched outside of the Hubble horizon during inﬂation, remains there f or a time, then reenters during a matter dominated era after inﬂation, and ev entually gets expelled out of the horizon once more during the ﬁnal stage of acceleration. In b), the ﬂuctuation w ould ha v e reentered about no w , but the late acceleration just pre v ents that. In c), the late acceleration pre v ents the ﬂuctuation from ev er reentering the Hubble horizon. AH = apparent horizon, RS = reheating surf ace, PT = photon trajectory , F(P) H = future(past) horizon, t 0 = n ow .

of Fig. 1). If they do not re-enter by the time the universe begins to accelerate, they will never do so and hence will never be directly observable. The photons which comprise the CMB originate on the slice (i.e. a sphere) of the last scattering surface which is separated from the observer by null geodesics (labelled PT in Fig. 1). In a decelerating universe the radius of this last scattering sphere grows without bound, and new information about inﬂation continues to become available over time. In a universe which accelerates, the last-scattering sphere asymptotes to the size of the event horizon at the time of last scattering, which is ﬁnite. Therefore, the pattern of temperature anisotropies in the CMB “freezes” after the transition to future acceleration. Eventually even this remnant will be permanently erased. Spacetimes with event horizons contain Hawking particles, and as the cosmological expansion continues, the CMB redshifts until it is colder than the Hawking radiation. After this time, any remaining information in the CMB will be masked by quantum effects. 2. Quantiﬁcation The condition that an initial Hubble-scale perturbation (generated at some time ti during inﬂation with scale Hi ) has expanded to ﬁll the observable universe today (subscript 0 refers to now) is: (2.1) a(ti )H(ti ) = a0 H0 . Using a(t) = ae exp(Hi (t − te )) for times during inﬂation yields ae Hi N ≡ Hi (te − tb ) = ln , a0 H0

(2.2)

INFLATION UNLOADED

339

Figure 2. On the left, the evolution of the comoving Hubble scale a(t)H (t) for a universe which inﬂates, followed by radiation and matter domination; on the right, the same quantity for a universe that enters a late-time accelerating phase.

for some time tb during inﬂation. After inﬂation, the universe grew by a factor of about a0 /ae ∼ Te /T0 , where Te is the reheating temperature and T0 ∼ 10−3 eV the current CMB temperature. Taking this ratio to be about 1026 − 1028 and the scale of inﬂation to be Hi ∼ 1014 GeV , one ﬁnds N ∼ 60. Hence, to use inﬂation to solve the horizon and ﬂatness problems requires at least 60 efolds (this is somewhat model dependent). If the universe accelerates in the future, the comoving Hubble scale a(t)H(t) grows at late times. At a time tf when the comoving Hubble scale equals its value at reheating, the last perturbation generated during inﬂation will be larger than the horizon, and afterwards no new structure will form from inﬂationary perturbations. The equality a(tf )H(tf ) = a(te )Hi ,

(2.3)

(where te is the time at reheating) deﬁnes tf ; it is calculated below). Spacetimes with event horizons contain (approximately) thermal Hawking particles, with a characteristic temperature TH = H/2π. Being a quantum effect, this spectrum does not redshift in the usual way. If the universe is accelerating, the CMB temperature TCM B will eventually decrease to a point where it is equal to TH ∼ H(t). This occurs at a time tT when (2.4) a(tT )H(tT ) = a(te )Te . Taking the ratio Hi /Te ∼ 1, tf ∼ tT . Using (2.3), eqs. (2.1) and (2.2), and the scaling a(t)H(t) ∼ a0 H0 (t/t0 )−(1+3w)/[3(1+w)] when −1 < w < −1/3: tT ∼ tf ∼ 1078(1+w)/|1+3w| t0 .

(2.5)

In the limit w → −1/3 the time diverges. The limit w → −1 yields tT ∼

60 . H0

(2.6)

Therefore if the cosmic acceleration never ends, only those inﬂationary ﬂuctuations produced in the interval between the end of inﬂation and 60 efolds before the end will ever be observable. Further, the information which is accessible now will be lost after the

340

MATTHEW KLEBAN

time tT . It is interesting that the future acceleration appears in this sense to shield us from the past incompleteness of inﬂation, e.g. the big bang singularity. 3. Summary Eternal dark energy with w < −1/3 prevents us from ever detecting inﬂationary perturbations which originated before the ones currently observable. Further, it slowly degrades the information stored in the currently observable perturbations. This allows us to reformulate the “Why now?” problem in a novel and interesting way: why are we living in the best time to do cosmology; the time at which we can see back the farthest? References 1. 2.

N. Kaloper, M. Kleban and L. Sorbo, Phys. Lett. B 600, 7 (2004). S. Hellerman, N. Kaloper and L. Susskind, JHEP 0106 (2001) 003; W. Fischler, A. KashaniPoor, R. McNees and S. Paban, JHEP 0107 (2001) 003.

D-BRANE EFFECTIVE ACTIONS

PAUL KOERBER UBC, Department of Physics & Astronomy 6224 Agricultural Rd., Vancouver, B.C., V6T 1Z1, Canada

Abstract. We present an alternative way to construct the D-brane effective action in the case where the background gauge ﬁeld strengths are not slowly varying and/or in the case of multiple coinciding D-branes.

1. Introduction It has been known for quite some time that the effective action for a single D-brane in the presence of slowly varying background gauge ﬁeld strengths is the Dirac-Born-Infeld action [1]. A class of derivative corrections was calculated to all orders in α in [2] using boundary conformal ﬁeld theory methods. When multiple D-branes coincide the gauge group becomes non-abelian and much less is known about the effective action beyond the leading orders. As became clear from an analysis of the spectrum of strings stretching between intersecting branes, the answer is not just a symmetrized trace version of the abelian Dirac-Born-Infeld action [3]. Furthermore, the corrections containing commutators of ﬁeld strengths turn out to be intimately related to the derivative corrections. We present an alternative way of constructing these D-brane effective actions based on requiring the existence of a type of BPS equations which are an α -deformation of the equations introduced in [4]. For more details and further references, I refer to my PhD thesis [5]. 2. Derivative corrections In [4] the (anti-)self-duality equations of euclidian Yang-Mills theory in 4 dimensions were generalized to dimensions greater than 4. One can show that solutions to these generalized equations satisfy a Bogomolny bound and preserve part of the supersymmetry, so we will call these equations BPS equations. Their symmetry is only a subgroup G of the rotation group SO(d). We will focus on the case were G = U (k) ⊂ SO(2k). Introducing complex coordinates the BPS equations read

Fαβ

= Fα¯ β¯ = 0,

(2.1)

Fαα¯

=

(2.2)

0.

α

341 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 341–343. © 2005 Springer. Printed in the Netherlands.

342

PAUL KOERBER

The ﬁrst equation is the condition for a holomorphic vector bundle while the second one is called the DUY condition [6]. It turns out that the abelian Born-Infeld action is (without derivative corrections) the unique deformation of the Maxwell action allowing for an α deformation of these BPS equations to automatically solve the equations of motion. The holomorphicity condition stays uncorrected while the DUY condition becomes arctanh 2πα Fαα¯ = 0 . (2.3) α

It is still possible to establish a Bogomolny bound and show that part of the supersymmetry is unbroken by solutions to this deformed equation. Interestingly, if we study the derivative corrections introduced in [2], which can be written as (2.4) SWyl = −τp dp+1 x (1 + P ) − det (η + 2πα F ) , with P S ρ1 ρ2 µ1 µ2

δ δ 1 ρ1 S ρ2 µ1 µ2 S ρ2 ρ1 µ3 µ4 , 48 δ(2πα Fµ1 µ2 ) δ(2πα Fµ3 µ4 ) 2 = hρ1 ρ3 (2πα ) ∂ρ1 ∂ρ3 Fµ1 µ2 +2 (2πα )3 hν1 ν2 ∂ρ1 F[µ1 |ν1 ∂ρ3 | Fµ2 ]ν2 . =

one can still ﬁnd a deformation of the DUY condition which is now given by (1 + P ) [arctanh 2πα F ]αα¯ = 0 .

(2.5)

(2.6)

α

Assuming that these BPS equations should still exist in the non-abelian case, we were successful in using them as a constraint to construct the non-abelian D-brane effective action up to order α4 [7]. Our results passed the spectrum test which the symmetrized trace Born-Infeld action failed [8]. 3. Outlook Another interesting aspect is that the transverse scalars describing the positions of the Dbranes become matrices if the D-branes coincide [9]. One can study general coordinate invariance in this context [10]. The most interesting transformations seem to be ones that mix the transversal and the longitudinal coordinates. Invariance under these transformations seems again to impose strong constraints on the form of the non-abelian D-brane effective action [11]. Acknowledgements It is a pleasure to thank A. Sevrin for the stimulating collaboration and encouragement during the work on my PhD thesis. I would also like to thank the organizers of the Carg`ese 2004 summer school for the chance to present this work at the Gong Show.

D-BRANE EFFECTIVE ACTIONS

343

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

E. Fradkin and A. Tseytlin, Phys.Lett. B163 (1985) 123; R. Leigh, Mod.Phys.Lett. A4 (1989) 2767. N. Wyllard, Nucl.Phys. B598 (2001) 247-275, hep-th/0008125; N. Wyllard, JHEP 0108 (2001) 027, hep-th/0107185. A. Hashimoto and W. Taylor, Nucl.Phys. B503 (1997) 193, hep-th/9703217. E. Corrigan, C. Devchand, D. Fairlie and J. Nuyts, Nucl.Phys. B214 (1983) 452. P. Koerber, Fortsch.Phys. 52 (2004) 871, hep-th/0405227. K. Uhlenbeck and S.T. Yau, Comm. Pure Appl. Math. 39 (1986) 257 and Comm. Pure Appl. Math. 42 (1989) 703; S.K. Donaldson Duke Math. J. 54 (1987) 231. P. Koerber and A. Sevrin, JHEP 0210 (2002) 046, hep-th/0208044. A. Sevrin and A. Wijns, JHEP 0308 (2003) 059, hep-th/0306260. E. Witten, Nucl.Phys. B460 (1996) 335, hep-th/9510135. J. de Boer and K. Schalm, JHEP 0302 (2003) 041, hep-th/0108161; D. Brecher, K. Furuuchi, H. Ling and M. Van Raamsdonk, JHEP 0406 (2004) 020, hep-th/0403289. D. Brecher, H. Ling, P. Koerber and M. Van Raamsdonk, work in progress.

ADS2 × S 2 AS AN EXACT HETEROTIC STRING BACKGROUND

DOMENICO ORLANDO ´ CPHT - Ecole Polytechnique 91128 Palaiseau, France

Abstract. An exact heterotic string theory on an AdS2 × S 2 background is found as deformation of an SL (2, R) × SU (2) WZW model. Based on [1].

1. Intro Anti de Sitter in three dimensions and S 3 are among the most simple and yet interesting string backgrounds. They are exact solutions to the string equations beyond the supergravity approximation and, at the same time, are simple to deal with although non-trivial thanks to the presence of non-vanishing curvatures. For this reason they constitute an unique setting in which to analyze AdS/CFT correspondence, black-hole physics, littlestring theory. String propagation in these backgrounds is described in terms of WZW models for the SL (2, R) and SU (2) groups, hence marginal deformations of such models allow to study moduli space of the string vacua. In particular well-known class of marginal deformations for wzw models are those driven by left–right current bilinears [2, 3]. On the other hand S 3 and AdS3 are embedded in larger structures so one can consider marginal deformations where just one of the currents belongs to the SU (2) or AdS3 algebra, the other belonging to some other U (1) corresponding to an internal magnetic or electric ﬁeld. This kind of deformation generates a continuous line of exact CFT’s. In this note we will show how with an appropriate choice for the deforming current we obtain a boundary in moduli space and that this boundary can be given a simple geometric interpretation in terms of the AdS2 × S 2 near-horizon geometry of the Bertotti-Robinson black hole [4, 5]. 2. SU (2) asymmetric deformation In the SU (2) case, there exists just one possible choice for the deforming current the two other being related by inner automorphisms, since the group has rank one, is compact and 345 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 345–349. © 2005 Springer. Printed in the Netherlands.

346

DOMENICO ORLANDO

its Lie algebra simple. Take the WZW model for SU (2): $

3 3 k ¯ + ∂β ∂β ¯ + ∂γ ∂γ ¯ + 2 cos β ∂α∂γ ¯ + ¯ a ∂α∂α ψ a ∂ψ SSU (2)k d2 z 4 a=1 (2.1) where ψ a are the left-moving free fermions, superpartners of the bosonic SU (2)k currents, and (α, β, γ) are the usual Euler angles parameterizing the SU (2) group manifold. The left-moving fermions transform in the adjoint of SU (2); there are no right-moving superpartners but a right-moving current algebra of total charge c = 16 can be realized in terms of right-moving free fermions. This means that we can build a N = (1, 0) worldsheet supersymmetry-compatible deformation given by: 1 = 2π

δSmagnetic

√ kkG H = d2 z J 3 + ıψ 1 ψ 2 J¯G ; 2π

(2.2)

where J 3 belongs to the SU (2) algebra and J¯g is the current of the algebra at level kg realized by the right-moving free fermions. An exact CFT is obtained for any value of the deformation parameter H. 2.1. GEOMETRY These new backgrounds all present a constant dilaton, a magnetic ﬁeld, a NS - NS ﬁeld proportional to the unperturbed one and a metric retaining a residual SU (2) × U (1) isometry [6]. The most remarkable property is that the deformation line in moduli space has a boundary corresponding to a critical value of the deformation parameter H 2 = 1/2. At this point the U (1) subgroup decompactiﬁes and the resulting geometry is the left coset SU (2) /U (1) ∼ S 2 which is thus found to be an exact CFT background only supported by a magnetic ﬁeld (the dilaton remains constant and NS ﬁeld vanishes). A geometrical interpretation for this process can be given as follows: the initial S 3 sphere is a Hopf ﬁbration of an S 1 ﬁber generated by the J 3 current over an S 2 base; the deformation only acts on the ﬁber, changing its radius up to the point where this seems to vanish, actually marking the trivialization of the ﬁbration: S 3 −−2−−−− → R × S2, 2 H →Hmax

(2.3)

If we turn our attention to the gauge ﬁeld one can show that a quantization of the magnetic charge is only compatible with levels of the afﬁne algebras such that kkG = p2 , p ∈ Z. We will ﬁnd the same condition in terms of the partition function for the boundary deformation. Although this construction has been implicitly carried on for ﬁrst order in α background ﬁelds, it is important to stress that the resulting metric is nevertheless exact at all orders since the renormalization boils down to the redeﬁnition of the level k that is simply shifted by the dual Coxeter number (just as in the WZW case).

ADS2 × S 2 AS AN EXACT HETEROTIC STRING BACKGROUND

347

2.2. PARTITION FUNCTION Consider the case of kg = 2 (one right-moving C fermion). The relevant components of the initial partition funciton are given by a SU (2)k−2 -modular-invariance-compatible combination of SU (2)k−2 supersymmetric characters and fermions from the gauge sector. For our pourposes it is useful to further decompose the supersymmetric SU (2)k characters in terms of those of the N = 2 minimal models: a 2ν j j a χ (τ ) ϑ (τ, ν) = Cm Θm,k τ, − . (2.4) b b k m∈Z2k

The deformation acts as a boost on the left-lattice contribution of the Cartan current of the supersymmetric SU (2)k and on the right current from the gauge sector: √ 2 1 h 1 h 2 h 2kn+ √m 2k Θm,k ϑ¯ e−ıπg(n¯ + 2 ) q 2 q¯2 (n¯ + 2 ) = g n,¯ n −→

1

e−ıπg(n¯ + 2 ) q 2 h

√ 2 cosh x+(n 2kn+ √m ¯+ h 2 ) sinh x 2k

n,¯ n

1 n ¯ + h cosh x+ × q¯2 ( 2 )

√

2kn+ √m

2k

2 sinh x

, (2.5)

1 where the boost parameter x is given by cosh x = 1−2H 2. Although an exact CFT is obtained for any value of the deformation parameter H we will concentrate, as before, on the boundary value H 2 = 1/2. In this case the √ boost 2k¯ n+ parameter diverges thus giving the following constraints: 4(k + 2)n + 2m + 2 √ 2kh = 0. Therefore, the limit is well-deﬁned only if the level of the supersymmetric SU (2)k satisﬁes the quantization condition k = 2p2 , p ∈ Z i.e. the charge quantization for the ﬂux of the gauge ﬁeld. Under these constraints the U (1) corresponding to the combination of charges orthogonal our condition decouples and can be removed. In this way we end up with the expression for the S 2 partition function: h a; h a j¯ j ZS 2 M j¯j eıπg(N + 2 ) Cp(2N (2.6) = ¯ −h) b χ b; g j,¯ j N ∈Z2p

in agreement with the result found in [8] by using the coset construction. The remaining charge N labels the magnetic charge of the state under consideration. 3. SL (2, R) deformation The same construction as above can be repeated for the SL (2, R) WZW model. In this case the moduli space is somewhat richer for it is possible to realize three different asymmetric deformations using the three generators of the group. These are not equivalent (SL (2, R) is not compact) and in fact they lead to three physically different backgrounds. The elliptic deformation line, in example, contains the G¨odel universe [7], the parabolic deformation gives the superposition of AdS3 and a gravitational plane wave. Two of these deformation lines present the same boundary effect as the SU (2) deformation. In particular the

348

DOMENICO ORLANDO

elliptic deformation leads to the hyperbolic space H2 = SL (2, R) /U (1) supported by an immaginary magnetic ﬁeld, ie an exact but non-unitary CFT. The hyperbolic deformation, on the other hand, leads to AdS2 = SL (2, R) /U (1) supported by an electric ﬁeld. No charge quantization is present in this case, because of the non-compact nature of the background. In this latter case it is not yet possible to give the same construction for the partition function as for the SU (2) case since this would require the decomposition of the initial partition function in a basis of hyperbolic characters which is not a simple exercize. Nevertheless by following the same procedure as before it is possible to evaluate the effect of the deformation on the spectrum of primaries and hence give the resulting AdS2 background spectrum. 4. AdS2 × S 2 The S 2 and AdS2 backgrounds can be combined so to give an exact CFT corresponding to the AdS2 × S 2 near-horizon geometry of the BR black-hole. Let us now consider the complete heterotic string background which consists of the AdS2 × S 2 space–time times an N = 2 internal conformal ﬁeld theory M, that we will assume to be of central charge cˆ = 6 and with integral R-charges. The levels k of SU (2) and kˆ of SL(2, R) are such that the string background is critical: cˆ =

2(k − 2) 2(kˆ + 2) ˆ + = 4 =⇒ k = k. k kˆ

(4.1)

This translates into the equality of the radii of the corresponding S 2 and AdS2 factors, which is in turn necessary for supersymmetry. Furthermore, the charge quantization condition for the two-sphere restricts further the level to k = 2p2 , p ∈ N. The combined AdS2 × S 2 background can give new insights about the physics of the BR black hole in particular by analizing the Schwinger-pair production in such background, or the study of the stability and propagation of D-branes. Acknowledgements Research partially supported by the EEC under the contracts HPRN-CT-2000-00131, HPRN-CT-2000-00148, MEXT-CT-2003-509661, MRTN-CT-2004-005104 and MRTNCT-2004-503369. References 1. 2. 3. 4. 5.

D. Israel, C. Kounnas, D. Orlando, and P. M. Petropoulos, Electric / magnetic deformations of s∗∗ 3 and ads(3), and geometric cosets, [hep-th/0405213]. S. Chaudhuri and J. A. Schwartz, A criterion for integrably marginal operators, Phys. Lett. B219 (1989) 291. S. Forste and D. Roggenkamp, Current current deformations of conformal ﬁeld theories, and wzw models, JHEP 05 (2003) 071, [hep-th/0304234]. B. Bertotti, Uniform Electromagnetic Field In The Theory Of General Relativity, Phys. Rev. 116 (1959) 1331. I. Robinson, Bull. Acad. Polon. Sci. 7 (1959) 351.

ADS2 × S 2 AS AN EXACT HETEROTIC STRING BACKGROUND 6. 7. 8.

349

E. Kiritsis and C. Kounnas, Infrared behavior of closed superstrings in strong magnetic and gravitational ﬁelds, Nucl. Phys. B456 (1995) 699–731, [hep-th/9508078]. D. Isra¨el, Quantization of heterotic strings in a G¨odel/anti de Sitter spacetime and chronology protection, JHEP 01 (2004) 042, [hep-th/0310158]. P. Berglund, C. V. Johnson, S. Kachru and P. Zaugg, Heterotic Coset Models and (0,2) String Vacua, Nucl. Phys. B 460 (1996) 252, [hep-th/9509170].

STRINGS AND D-BRANES IN HOLOGRAPHIC BACKGROUNDS

¨ AND ARI PAKMAN DAN ISRAEL The Racah Institute of Physics, The Hebrew University of Jerusalem, 91904 Jerusalem, Israel

Abstract. We review recent progress in the study of non-rational (boundary) CFTs and their applications to holographic backgrounds in superstring theory. We focus on the supersymmetric coset SL(2, R)/U (1) and its dual N=2 Liouville. We discuss the modular properties of their characters, their partition function and boundary states for their Dbranes. Then these results are used to study the CFT of the NS5-brane background, with applications to Little String Theories.

1. Introduction Conformal ﬁeld theories (CFT) are the natural building blocks for the exact perturbative description of superstring theory in non-trivial backgrounds without RR ﬂuxes. When the curved space is non-compact, the CFTs needed are non-rational (NRCFTs). They are more difﬁcult to handle that the rational ones, and the simplest examples have been fairly understood only in the last years. This progress has been largely motivated by their relevance to describe holographic backgrounds. Among the latter, the most famous are AdS(3) and strings in d ≤ 2. The holographic nature of ﬁve-branes solutions is less clear, mainly due to the non-triviality of the dual non-gravitational theory, called Little String Theory. These constructions can typically be recast as non-rational analogues of Gepner models, with a non-compact factor given by the supersymmetric coset SL(2, R)/U (1) . It is this last theory that we shall study in detail in this note. The non-rational CFTs are deﬁned in opposition to rational ones. The techniques used to solve the latter do not work for the former, or need to be adapted. The complications arise because the Hilbert space has a continuous spectrum, due to the non-compactness of the target space. In general the spectrum of these theories splits into two categories of states. On the one hand the spectrum contains continuous representations, corresponding to asymptotic states propagating in the non-compact “radial” direction(s). On the other hand the states of the discrete representations correspond to (a ﬁnite set of) localized bound states. As we may infer from general considerations of scattering theory, these two kinds of states mix. For example, in the coset SL(2, R)/U (1) , the reﬂection amplitude 351 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 351–354. © 2005 Springer. Printed in the Netherlands.

352

¨ AND ARI PAKMAN DAN ISRAEL

for continuous representations has poles whenever it is analytically continued to a discrete representation. The modular properties of the characters exhibit a similar pattern. While the continuous representations modular transform into themselves, the discrete representations yield both discrete and continuous ones. These results hold both in the bosonic and susy cosets [1], and are shown using a lemma proven in [1]. These new features complicate the construction of partition functions and boundary states by a large amount. A last important aspect is that for speciﬁc values of the level, these theories acquire some kind of rational behavior. Remember that a free a boson on a circle at rational radius, has an extended chiral symmetry, and summing over the orbits of this symmetry one gets a ﬁnite set of extended characters. The same holds for the (super) coset√SL(2, R)/U (1) at rational level, since the target space asymptotes a cylinder of radius 2k. In particular, in the supersymmetric case one obtain a ﬁnite set of rational N=2 R-charges, which is desirable to construct space-time supersymmetric vacua. 2. SL(2,R)/U(1) and N=2 Liouville The theory is obtained applying the rules of an elliptic coset construction to the supersymmetric WZW model SL(2, R) at level k. For the axial coset the sigma model is well deﬁned because the action of the gauge ﬁeld has no ﬁxed point, and corresponds to an Euclidean two-dimensional black hole. The spectrum of primaries is obtained by descent from AdS3 . The worldsheet-supersymmetric partition function has been computed using the powerful techniques of marginal deformations of WZW models[3]. This amplitude splits naturally into (non-minimal) characters of the N=2 superconformal algebra. An exact decomposition of the partition function has been carried out in[2]. We have found ﬁrst a contribution of discrete representations, with the correct multiplicities for all the descendants. For the continuous representations, their contribution is divergent, due to the inﬁnite volume available for them. An infrared regularization of the partition function is possible, leading to a ﬁnite non-trivial density of continuous representations, compatible with N=2 supersymmetry. However this regulator breaks (super)conformal symmetry, and there is a price to pay: the partition function contains an extra non-universal contribution which is not related to the N=2 algebra. The super-coset SL(2, R)/U (1) has been conjectured to be dual to the N=2 Liouville theory. Evidence for this equivalence comes from a sigma-model mirror symmetry and from the agreement between perturbative computations of correlators. However, we argue that these results come from a more fundamental structure of these theories. Indeed, both theories possess the same chiral algebra, which is the (non-minimal) N=2 SCA. This algebra can be decomposed into the bosonic coset SL(2, R)/U (1) and a free boson[1]. Since conformal bootstrap results for the Euclidean AdS3 are obtained using only the chiral symmetries of the model (without an explicit reference to a speciﬁc action), they can be applied to the supersymmetric SL(2, R)/U (1) coset and to the N=2 Liouville theory as well, because they both lift to the same current algebra[4]. Therefore the theories can differ only by the way the left and right representations of the algebra are glued in the closed string spectrum.

STRINGS AND D-BRANES IN HOLOGRAPHIC BACKGROUNDS

353

The vector coset SL(2, R)/U (1) has a singular sigma-model, therefore it receives substantial corrections; its single cover is a Zk orbifold of the cigar[3]. It is likely that the N=2 Liouville is the corrected form of this vector coset. 3. D-branes from boundary AdS3 The study of D-branes in these exact superstrings backgrounds is essential, in order to understand the non-perturbative dynamics in these non-compact manifolds, and it may give indications about its holographic degrees of freedom. The construction of the exact boundary states, which contain all the information about the couplings of the D-branes to the closed string states, follows the same logic as before. Conformal bootstrap methods were employed successfully to study D-branes in the bosonic Liouville theory in Euclidean AdS3 . The results have been used later to study the bosonic coset.

Figure 1. The consistent D-branes in the cigar: D0-, D1- and D2-branes (left to right)

In[4] we constructed the D-branes in the super-coset SL(2, R)/U (1) using similar methods. An important aspect of this analysis is that the boundary conformal bootstrap uses only chiral symmetries. Therefore the arguments given above about the duality super-coset SL(2, R)/U (1) / N=2 Liouville extend straightforwardly in the presence of a boundary. However, one should be careful about the way left and right representations are glued, in order to construct the basis of Ishibashi states. Then the various boundary states of the theory can be constructed by descent from D-branes of Euclidean AdS3 . There is a 4 R)algebra direct connection between the boundary conditions for the currents of the sl(2, and the gluing conditions of type A or B deﬁned in. The D-branes that we obtained satisfy by construction the factorization constraints, since they descend from consistent D-branes in Euclidean AdS3 . We also checked the Cardy condition, i.e. the open-closed channel consistency of the annulus diagram. The various D-branes of the cigar are depicted in ﬁg. 1. 4. strings near NS5 branes The results given above can be lifted to the background created by NS5-branes distributed on a topologically trivial circle, known to be T-dual to an exact CFT. When all the NS5branes are separated from each other, the background is perturbative and one can take a double scaling limit where gravity decouples and this perturbative nature holds. We have shown in[2] that the complicated solution for the ring of NS5 in this limit can be obtained as a null gauging of the super-WZW model SL(2, R) × SU (2). This CFT has worldsheet N=4 SCA, so there are no perturbative corrections metric. However, instantons corrections show up, and they are indeed captured by the supergravity solution. The null coset can be recast as a Zk orbifold of SL(2, R)/U (1) ×SU (2)/U (1). The Zk orbifold can be thought as coming from the GSO projection generalizing Gepner models. However, this orbifold

¨ AND ARI PAKMAN DAN ISRAEL

354

changes deeply the background of the effective theory, because in the semi-classical limit its twisted sectors become very light; hence the correct geometry is not given by the sum of the two coset factors. Once this identiﬁcation is understood it is possible to write the one-loop amplitude for this NS5-brane backgroud[2]. Various BPS D-branes in this background are now under study. In particular there are new non-factorizable D-branes that can be constructed out of the coset D-branes[5]. In type IIA we have also D4-branes stretched between the NS5branes on which a D=4, N=2 SYM theory lives. Quite remarkably, the one-point function for these D4-branes can be related to the beta-function of the gauge theory. Acknowledgements It is a pleasure to thank C. Kounnas and J. Troost for the collaborations leading to the present results. We also thank C. Bachas, S. Elitzur, A. Giveon, E. Kiritsis, D. Kutasov, B. Pioline, E. Rabinovici, S. Ribault and V. Schomerus for stimulating discussions. References 1. 2. 3. 4. 5.

D. Israel, A. Pakman and J. Troost, JHEP 0404, 045 (2004) D. Israel, C. Kounnas, A. Pakman and J. Troost, JHEP 0406, 033 (2004) D. Israel, C. Kounnas and M. P. Petropoulos, JHEP 0310, 028 (2003) D. Israel, A. Pakman and J. Troost, arXiv: hep-th/0405259. D. Israel, A. Pakman and J. Troost, to appear

COSMIC CENSORSHIP IN ADS/CFT

MUKUND RANGAMANI Department of Physics & LBNL, Univesity of California Berkeley, CA 94720, USA

Abstract. We review recent attempts at violations of cosmic censorship in asymptotically Anti-deSitter spacetimes. The essential logic behind the potential counter-examples was that there exist smooth initial data that are guaranteed to evolve in the future to a singularity, which nonetheless are not going to be cloaked by a black hole horizon, owing to the conﬁguration possessing insufﬁcient energy to produce the desired black hole. We demonstrate that a reﬁned analysis indicates that the kinematics always allows for black hole formation, thereby upholding cosmic censorship in this context.

1. Introduction Singularities are a ubiquitous feature in classical general relativity. As attested to by the singularity theorems, mild initial data can often evolve to singular spacetimes. Within the classical framework, formation of singularities implies lack of predictability. Moreover, it implies that formation of singularities in some local region would lead to drastic effects arbitrarily far away. To avert this consequence it was proposed by Penrose [1] that the is a universal cosmic censor that prevents these singularities from being causally accessible to asymptotic observers. In a more colloquial phrasing the cosmic censorship conjecture may be stated as follows: whenever initial data under evolution form singularities, the singularities are hidden behind event horizons of black holes, thereby preventing them from inﬂuencing the physics at large distances. While the spirit of the cosmic censorship conjecture is appropriate within the limited sense of classical general relativity, it is not clear that this is a desirable state of affairs in the quantum gravity context. As such one would expect that accessing the large curvature effects should provide insight into the workings of quantum gravity (see V. Hubeny’s talk in this volume for a discussion of these issues). To this end it is useful and instructive to analyze the domain of validity of the cosmic censorship conjecture. At the same time given its recalcitance to being proven, one wonders whether there might be some important piece of physics that we are missing. Ideally one would like to talk about cosmic censorship violation in contexts where we have some level of conﬁdence about what we expect in the full quantum gravity context. 355 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 355–362. © 2005 Springer. Printed in the Netherlands.

356

MUKUND RANGAMANI

-0.5

Φ

a

V

0.5

1

1.5

2

Φ

b

Φ0

-5 -10 -15 r -20

R0

Figure 1. The scalar potential and the ﬁeld proﬁle.

As of date the framework wherein this criterion is met is the AdS/CFT correspondence. Here we know that there is an in principle formulation of quantum gravity in asymptotically Anti-deSitter spacetimes in terms of a quantum ﬁeld theory. This makes it ideal to explore consequences of cosmic censorship in asymptotically AdS spacetimes. The ﬁrst steps in the direction of potential cosmic censorship violation were taken in [2] and more recently this was extended to a context with greater control from the AdS/CFT point of view in [3]. The essence of their argument is to produce a simple conﬁguration that reliably evolves into a singularity, and then show that there is insufﬁcient energy available to make a black hole whose horizon is big enough to cloak the singularity. A careful study of this scenario with the purpose of identifying a region in parameter space where we can reliably argue that naked singularities will form and moreover the dual CFT will be well-behaved was carried out in [5]. It was argued that there aren’t any initial conditions that satisfy the aforementioned criteria. In particular, we carry out a reﬁned estimate of the proper size of the singularity and the energitics in the system to show that there is always a potential for black hole formation. This is in agreement with the numerical results of [4], [6], and [7], wherein full evolution of the initial data was carried out. We begin by reviewing the framework in which we will consider cosmic censorship violation and then proceed to proceed to detail how we can make reliable estimates of the size of the singularity and the energetics involved and end with a short discussion. 2. The Framework The simplest context in which we can discuss the formation of naked singularities in asymptotically AdS is to consider a scalar gravity system with the scalar potential being picked arbitrarily, with the only constraint that it admits AdS vacua. In [2], such a set-up was chosen, where it was argued that with a suitable choice of scalar potential one can ensure that it is possible to ﬁnd conﬁgurations with very little energy, which nonetheless make black holes. Choosing an arbitrary scalar potential is not viable within the AdS/CFT context. Nonetheless, [3] argue that it is possible to create naked singularities in this case as well. The idea is to use the well known fact that in AdS, scalar ﬁelds can have negative mass squared, but nevertheless be stable. In fact such ﬁelds are available in the context of N = 8 gauged supergravity on AdS5 . Consider then the following action, which is a

COSMIC CENSORSHIP IN ADS/CFT

357

T π/2Η

Ts

ε

rh1

Ttrust

T=0

rh0

Rs0

Rtrust

Ro

R

Figure 2. A sketch of the causal diagram for the evolution of the scalar ﬁeld proﬁle presented in Fig 1.1b.

consistent truncation of the supergravity theory: S=

√

1 1 2 −g R − (∇φ) − V (φ) , 2 2

(2.1)

where the potential is plotted in Fig. 1.1.a, and is given by V (φ) = −2 e2φ/

√

3

− 4 e−φ/

√

3

.

(2.2)

The initial data is prescribed in the form of the ﬁeld proﬁle φ(r) as indicated in Fig. 1.1b. We are assuming that we have time symmetric initial data and will be parametrising the metric on the inital data surface as ds2Σ

=

m(r) 1− + r2 3 π 2 r2

−1 dr2 + r2 dΩ23 ,

(2.3)

so that r denotes the proper size of the S3 on the initial data surface and m(r) is related to the ADM mass of the conﬁguration and is determined in terms of φ(r) by virtue of the constraint equations. As is clear from Fig. 1.1.b, there are two parameters at our disposal: R0 which is size of the hom*ogeneous region of the scalar ﬁeld and φ0 the value of the scalar ﬁeld in the hom*ogeneous region. The choice of this hom*ogeneous proﬁle has been made to ensure that the future evolution of the initial data results in a singularity. Thinking in terms of FRW cosmologies, a hom*ogenous scalar ﬁeld tends to roll down the potential and collapse to a singularity in open universes with a negative cosmological constant. The asymptotic fall-off of the scalar ﬁeld is dictated by the requirement that the geometry be asymptotically AdS. In Fig. 1.2 we sketch the conformal diagram for the evolution of the scalar ﬁeld in the domain of dependence of the region r ≤ R0 . For the sake of concreteness there are two distinct scalar proﬁles which we can consider:

358

MUKUND RANGAMANI

− Normalizable fall-off φ(r) = φ0 ≡

A R02

r ≤ R0 ,

φ(r) =

A r2

r > R0 .

(2.4)

− Non-normalizable fall-off with a rigid cut-off at r = R1 φ(r) = φ0 ,

r ≤ R0

φ(r) =

φ1 R12 ln r , ln R1 r2

R0 < r ≤ R 1 .

(2.5)

With the explicit expressions for the ﬁeld proﬁle, the ADM mass can be calculated from the constraint equations. We refer the reader to [5] for the details. In what follows we assume that we denote Mconﬁg (φ0 , R0 ) as the energy contained in the initial data. Given the kinematical information about how much energy is contained in the initial data, the strategy is to ask whether we have enough energy to form a black hole which can enclose the singularity which is guaranteed to form upon evolution. To this end we need to be aware of the size of the singularity, in order to ascertain the size of the event horizon of the black hole. This we turn to next. 3. The Size of the singularity The ﬁrst step is to model the evolution of a hom*ogeneous scalar ﬁeld in hyperbolic slices of AdS. This is an accurate description of the dynamics in the domain of dependence of the hom*ogeneous region r ≤ R0 for the conﬁgurations considered in the body of the paper. The metric in d dimensions can be written as dR2 2 2 2 2 2 + R dΩd−2 . ds = −d T + a(T ) (3.1) 1 + R2 where the scale factor a(T ) and the scalar ﬁeld φ(T ) are to be determined from the Einstein’s equations, which read: 1 ˙2 2 a2 a˙ 2 − φ +V =1 (d − 1)(d − 2) 2 a ¨ 2 (d − 2) ˙ 2 − V − φ =0 a (d − 1)(d − 2) 2 a ˙ (3.2) φ¨ + (d − 1) φ˙ + V,φ = 0 . a Using (3.2) it can be shown that we get a curvature singularity as a → 0, unless a(T ) ∼ cos T as for pure AdS, where a = 0 corresponds merely to a coordinate singularity. Evolving these equations is difﬁcult to do analytically, but it is relatively straightforward numerically. There are two cases where we can make analytic statements, The ﬁrst is where we assume that the ﬁeld is small and study only the Gaussian ﬂuctuations in the potential, and the other is the near-singularity regime, where the potential is negligible. Armed with this information it is possible to show that a curvature singularity will occur in T < π2 .

COSMIC CENSORSHIP IN ADS/CFT

359

T π/2Η

Ts Region of trapped surfaces r h2

Tmax

T=0

R s0

r h3

Rs1

Ro

R

Figure 3. Reﬁned causal digram for the FRW region where the hom*ogeneous scalar ﬁeld collapses.

Let us now turn to the next step, namely estimating the mass of the black hole required to cloak the singularity. This involves a bit more detailed analysis, since it depends on the behaviour of the null geodesics in the dynamical background. What we can do is to estimate when the pure AdS metric cease to approximate the actual geometry. We will then use that as the cutoff for the ingoing radial null geodesic starting at T = 0 at radius R0 ; namely, we approximate the black hole size rh0 by the proper radius reached by this null geodesic at the cut-off time Ttrust , cf., Fig. 1.2. Carrying out this exercise we ﬁnd: 2/3

rh0 = 0.27 R0 φ0

.

(3.3)

We must emphasise here that the above serves as an estimate for the lower bound on the size of the black hole that needs to form. To get a more reliable estimate on the size of the black hole necessary to cloak the singularity, we solve FRW equations (3.2) numerically. With an explicit numerical solution to the scale factor a(T ), we can study the property of null geodesics in the resulting big crunch geometry. Note that this is still a much easier problem than having to solve for the full evolution. In particular we can obtain the behavior of rh0 (φ0 ). As indicated in Fig. 1.3, we start from a point on the initial data slice where the spheres have proper size R0 and consider a radially ingoing null geodesic. At some point very close to the singularity we cut-off the geometry and from the end-point of the geodesic from R0 onto this cut-off surface we take a radially ingoing null geodesic. This would intersect the initial data slice at Rs0 . However, the proper size of the spheres will peak at some intermediate time Tmax . Call this radius rh2 ; this will be the lower bound on the size of the black hole. We use the numerical solution for a(T ) to estimate rh2 . In Fig. 1.4 we show the behavior of rh2 (φ0 )/R0 . We see that for large φ0 , rh2 is independent of φ0 . This follows from the fact that the potential is exponential in φ. Moreover, we ﬁnd that for small φ0 , 2/3

rh2 (φ0 ) = 0.37 R0 φ0

(3.4)

360

MUKUND RANGAMANI rh2R0 , rh3R0

rh3R0

0.5

0.15

0.1

0.3

0.05 0.1

1

3

5

Φ0

0.02

0.04

0.06

0.08

0.1

Φ0

Figure 4. Accurate numerical estimate for the size of the singularity with the domain of dependence.

The numerical evaluation of rh2 differs from our weak ﬁeld estimate rh0 only in the numerical factor, which is attributable to a more accurate determination of the point where the geometry deviates signiﬁcantly from pure AdS. The estimate for the black hole size (3.4), does not constitute the tightest lower bound. In deriving it we assumed that the singularity stretched out to the boundary of the domain of dependence at time Ts . However, we can argue that it necessarily has to be bigger and so obtain more stringent bounds. By following an outgoing congruence of null geodesics from the initial data surface, we can show that the best estimate of the size of the singularity is given by rh3 in Fig. 1.3. The basic idea is to exploit the fact that we are able to move all the way to the boundary of the domain of dependence (see [5] for details). In fact, with a bit of numerical work we can show that for small φ0 1/2

rh3 (φ0 ) = 0.58 R0 φ0

,

for φ0 1 ,

(3.5)

is the best estimate that can be obtained within the domain of dependence. Moreover we can show that for rh3 → σs R0 = 0.577 R0 ,

for φ0 1 .

(3.6)

Plots of rh3 as function of φ0 are shown in Fig. 1.4. With the numerical information about the singularity in hand, we can ask what is the energy required to form the black hole of the requisite size. For the normalizable scalar mode, the black hole in question is the Schwarzschild AdS black hole, while for the nonnormalizable conﬁguration it is the scalar hair black hole. Referring the interested reader to [5] for details; we ﬁnd at the end of the day: Mconﬁg − MBH > 0

(3.7)

indicating that there is always enough energy in the system to form the black hole of the appropriate size. One might worry that since we only estimate the lower bound on the black size by the ray tracing arguments it might be possible that the singularity will be a bit bigger and there would be a potential for a naked singularity. However, numerical evolution of the data [7] that our estimate for the size of the singularity is quite robust and hence there is no violation of cosmic censorship.

COSMIC CENSORSHIP IN ADS/CFT

361

4. Discussion In this talk we have discussed the potential violation of cosmic censorship within the AdS/CFT context. At the face value one is tempted to think that naked singularity formation should be easy in AdS spacetimes, since it is easy to form singularities (as attested to by the FRW collapse scenario) and moreover black holes are quite heavy in AdS spacetimes. Nonetheless in the context of AdS/CFT correspondence we see that there is an interesting obtacle to violation of cosmic censorship. In fact, the arguments can be extended to the original context of cosmic censorship violation in AdS spacetimes as suggested in [2]; see [5] and [8]. The essential reason for the upholding of the cosmic censorship conjecture is that the singularity that forms from the initial data is rather small, and there is always sufﬁcient energy in the system to cloak it with a black hole of the desired size. At the same time, one begins to wonder whether there is an interesting lesson to be learnt from the CFT. Although we have not explicitly discussed the issue, the estimates of the energy in the initial conﬁguration are most reliably carried out from the CFT side owing to the fact that the bulk ADM energy has ﬁnite boundary contributions. In any event it is the CFT which lives on the boundary that has a well deﬁned Hamiltonian evolution and it is the eigenvalues of this conserved Hamiltonian that are of interest to us. There are other physical effects from the string theoretic viewpoint that we have not discussed here. It is shown in [5] that in the situations where we turn on the nonnormalizable modes we have to content with a D-brane instability of the system. In these contexts one can imagine that there are stringy effects at work which serve to cloak the singularity. At this stage one might speculate whether there is an underlying mechanism in the ﬁeld theory that serves to prevent cosmic censorship violation. After all one might argue that smooth initial data in the bulk maps to a well deﬁned state in the CFT and since the Hamiltonian evolution is well behaved, there is no room for singularities. Fleshing out this statement in a quantitative fashion could serve to prove the cosmic censorship conjecture. Acknowledgements I wish to thank my collaborators, Veronika Hubeny, Xiao Liu and Stephen Shenker for a wonderful collaboration. I would also like to thank also the organizers of the Carg`ese 2004 ASI for an excellent summer school and the Kavli Institute for Theoretical Physics for hospitality. This work was supported by the funds from the Berkeley Center for Theoretical Physics, and also by the DOE grant DE-AC03-76SF00098 and the NSF grant PHY-0098840. References 1. 2. 3. 4. 5.

R. Penrose, Riv. Nuovo Cim. 1, 252 (1969) [Gen. Rel. Grav. 34, 1141 (2002)]. T. Hertog, G. T. Horowitz and K. Maeda, Phys. Rev. Lett. 92, 131101 (2004) [arXiv:grqc/0307102]. T. Hertog, G. T. Horowitz and K. Maeda, Phys. Rev. D 69, 105001 (2004) [arXiv:hepth/0310054]. M. Gutperle and P. Kraus, JHEP 0404, 024 (2004) [arXiv:hep-th/0402109]. V. E. Hubeny, X. Liu, M. Rangamani and S. Shenker, arXiv:hep-th/0403198.

362 6. 7. 8.

MUKUND RANGAMANI D. Garﬁnkle, Phys. Rev. D 69, 124017 (2004) [arXiv:gr-qc/0403078]. A. V. Frolov, arXiv:hep-th/0409117. T. Hertog, G. T. Horowitz and K. Maeda, arXiv:gr-qc/0405050.

TOPICS IN BLACK HOLE PRODUCTION

VYACHESLAV S. RYCHKOV Institute for Theoretical Physics, University of Amsterdam, 1018XE Amsterdam, The Netherlands

Abstract. We revisit Voloshin’s model of multiple black hole production in transPlanckian elementary particle collisions in D = 4. Our revised computation shows that the cross section to produce n additional black holes is suppressed by s−1 , rather than being enhanced as was originally found. We also review the semiclassical gravity picture of black hole production from hep-th/0409131, making additional comments about the meaning of wavepacket subdivision.

1. Introduction Black hole (BH) production in trans-Planckian elementary particle collisions (E EP lanck ) has long been considered a theoretical possibility. If TeV-scale gravity scenarios based on large extra dimensions or warped compactiﬁcations are realized in nature, this possibility may be realized in practice at future accelerators (see [1] for a recent review). The key question is the cross section of this process, which is usually assumed to be set by the horizon radius of the produced BH (the so called “geometric cross section” σ ∼ πrh2 ). In this note I would like to discuss two aspects of the BH production problem relevant for justifying the geometric cross section estimate. First I will explain how to derive this estimate from a controlled semiclassical gravity approximation to the BH production process, adding some comments to the original discussion of [2] (also recently reviewed in [3]). Then I will estimate the cross section of multiple BH production due to collisions between virtual gravitons emitted by the primary particles, in a model ﬁrst proposed by Voloshin [4]. The conclusion (contrary to [4]) is that multiple BH production gives a subdominant contribution to the total cross section. The size of produced BHs in large extra dimension scenarios is typically much smaller than the size of extra dimensions, and their production may be considered as happening in ﬂat D dimensional spacetime. Since we are focussing on theoretical issues, to keep the discussion clear we will work in D = 4. We will use Planck units, setting EP lanck = 1. 363 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 363–369. © 2005 Springer. Printed in the Netherlands.

364

VYACHESLAV S. RYCHKOV

2. Wavepackets and Semiclassics In this note we will adhere to the standard believe that in quantum theory of gravity classical BHs with mass M 1 will be realized as long-lived resonance states, decaying via Hawking radiation, with lifetime ∼ M 3 . It is then energetically allowed to produce such BHs in trans-Planckian (E 1) elementary particle collisions. To estimate the cross section of this process, Eardley and Giddings [5] looked at the grazing collision of two ultrarelativistic point particles in classical general relativity, using formation of a closed trapped surface (CTS) as a sufﬁcient condition for BH formation. In this totally classical description, lower bound for the cross section is given by πb2max , where bmax is the maximal impact parameter for which we are able to ﬁnd a CTS in the spacetime formed by two colliding Aichelburg-Sexl shock waves. [5] found bmax ∼ E ∼ rh , and this implies σ ∼ πrh2 . Recourse to such an indirect method is necessary, because explicit solutions of Einstein’s equations exhibiting the ﬁnal BH state are out of reach. How do we justify this approach from quantum gravity point of view? A point of immediate concern is that particles in a collider experiment are described by wide wavepacket states of macroscopic size (set essentially by the beam radius, which in turn is determined by the focussing ability of accelerator magnets). These wavepackets are vastly larger than BHs whose production we are trying to describe. If we use the energy momentum tensor of these wavepacket states |ψ in the RHS of the semiclassical Einstein’s equations Rµν − 12 gµν R = ψ|Tµν |ψ,

(2.1)

we won’t see any BH production whatsoever, since the energy is spread out over a huge volume, and the energy density is insufﬁcient to cause collapse. Informally speaking, particles “do not ﬁt” inside a BH. However, this does not mean that BHs do not form. The correct interpretation is that the part of the gravitational ﬁeld wavefunction corresponding to BH production was erased—averaged away—by eq. (2.1). Some averaging is always inevitable when using the semiclassical ﬁeld equations, since we substitute Tµν by its expectation value. Unfortunately, in this case it destroys precisely the part of the wavefunction we are interested in. To see BH production, one should instead proceed as follows. First of all, we have to subdivide the initial wavepackets into much smaller wavepackets of size w rh : |ψ = N −1/2

N i=1

|ψi .

(2.2)

This subdivision is carried out so that the small wavepackets |ψi in the RHS are almost orthogonal. This orthogonality is quite obvious in the position representation (see Fig. 1). Because of the orthogonality, collisions between different pairs of small wavepackets are mutually excluded possibilities, and probabilities of BH production in each such elementary collision should be added. Now, it is the collisions of the small wavepackets that we are going to analyze using eq. (2.1). Condition w rh ensures that the small wavepackets produce a collision spacetime which is a small perturbation of the one corresponding to point particles of the same energy. Thus the Eardley-Giddings analysis applies, and adding probabilities results in the geometric cross section.

BLACK HOLE PRODUCTION

365

Using ﬁnite-size wavepackets instead of point particles has an additional bonus in that it puts the conditions of applicability of the semiclassical approximation under control. For example, curvature blows up when the shock fronts of Aichelburg-Sexl waves corresponding to point particles collide [6]. However, taking wavepacket size into account regulates the curvature and brings it below the Planck value, so that we can trust the semiclassical gravity approximation [2]. 3. Multiple Black Hole Production According to the above discussion, the geometric cross section formula provides a lower bound for a single large BH production cross section in a trans-Planckian collision. However, as the energy of the particles grows, multiple BH production also becomes energetically allowed. It is important to understand which process is dominant at asymptotically high energies. If multiple BHs dominate, it will be much harder to observe dipole patterns of emitted particles expected in the Hawking evaporation of a single large BH. We are going to discuss a model proposed by Voloshin [4], in which multiple BHs are produced due to collisions between virtual gravitons emitted by the trans-Planckian projectiles. Such virtual graviton emission is a quantum effect: classically, any radiation happens after the particles collide. The process is studied diagrammatically, with a typical diagram shown in Fig. 2. Only the case of small peripheral BHs (1 mi E, i = 1 . . . n) is considered, so that the gravitons are “soft”. The production amplitude is computed from the diagrams using the standard QFT propagators and particular vertices for soft graviton emission and for BH production (Fig. 3). The amplitude to emit a positive helicity graviton of energy ω E and small transverse momentum k = (k2 , k3 ), |k| ω is given by (see Appendix A) A ∝ (E/ω)2 (k2 + ik3 )2 .

(3.1)

(We will not pay attention to constant numerical factors. Thus our ﬁnal result (3.3) is valid up to a factor cn .) Although we are unable to compute the elementary BH production vertex f (q 2 ), the geometric cross section allows to ﬁx the combination |f (q 2 )|2 ρ(q 2 ) ∼ (q 2 )2 ,

(3.2)

where ρ(q 2 ) is the density of BH states at mass q 2 . A crucial ﬁnal element of the model is a condition which ensures that the emitted gravitons do not subsequently fall into a common large BH. Such an infall may happen

Figure 1. Subdivision of the right-moving particle wavepacket. Similar subdivision has to be done for the left-moving particle.

366

VYACHESLAV S. RYCHKOV

due to graviton rescattering diagrams, which we are not going to compute. Thus, without a “fall safe” condition we would be in danger of greatly overestimating the multiple BH amplitude. Voloshin’s “fall safe” condition limits the transverse momenta of emitted gravitons by |k| 1/E. To derive this condition, we note that a typical emitted graviton will be off-shell by ∆E ∼ k2 /ω. It will exist for a time interval ∆t ∼ 1/∆E, during which it will reach transverse separation ∆z ∼ (k/ω)∆t from the projectile. Voloshin’s condition arises if we require that this transverse separation is larger than the horizon radius of the big BH formed in the collision of the primary particles: ∆z E. In the described model, our computation (see Appendix B) gives the following amplitude to produce n additional BHs with 4-momenta qi : n (qi2 )−2 f (qi2 ) . (3.3) f (n) (s, qi2 ) ∼ f (s) i=1

The original computation of [4], Eq. (5), gave an amplitude larger than (3.3) by a factor of (s qi2 )n . We believe that our result is correct; see Appendix B for an explanation. Using (3.2), we can compute from (3.3) the contribution of the diagrams from Fig. 2 into the total cross section σn to produce one large (m2 ∼ s) and n small BHs. This contribution will behave like s1−n , the suppression being due to the phase space restriction |k| 1/E satisﬁed by the small BHs as a consequence of the “fall safe” condition. However, for n ≥ 2 there are diagrams which give a larger contribution, so that σn ∼ const is likely for any n ≥ 1. Consider, e.g., Fig. 4, where the primary particles emit “fall safe” gravitons of energy E1 , 1 E1 E, and it is these gravitons that form n smaller BHs according to the previous model. The allowed phase space for this diagram will be much bigger, since the individual small BHs can now have much larger transverse momenta |k| 1/E1 , only their sum being 1/E. Choosing E1 above the threshold of n BH production, we will get an s-independent contribution to σn .1 In any case, we see that σn is suppressed compared to the geometric cross section value σ0 ∼ s. 4. Conclusions In this note we discussed two aspects of the BH production problem. In Section 2 we discussed how wavepacket arguments can be used to justify the use of semiclassical gravity in this problem. In Section 3, we revisited Voloshin’s model of multiple BH production, and presented a revised computation which shows that this process is suppressed compared to the production of a single large BH. Our main conclusion is that the geometric 1 It is also easy to see that the inclusive cross section of multiple BH production cannot decay with s. This is because the particles may ﬁrst reduce their energy by emitting one or more “fall safe” gravitons (which costs no s-dependent factor, see [4], Eq. (3)), and then collide to form BHs. I am grateful to M. Voloshin for this remark.

Figure 2. A typical diagram for multiple BH production in Voloshin’s model

BLACK HOLE PRODUCTION

367

cross section estimate seems to be in rather good health, surviving all checks and resisting any disproving attempts. Acknowledgements I would like to thank Steve Giddings for the opportunity to collaborate on [2]. A part of Section 2 arose as an answer to Marco Cavagli`a’s interesting questions. I am especially grateful to Misha Voloshin for the e-mail correspondence which helped me to understand his model, and for the critical remarks about the early versions of the computation. It is also a pleasure to thank the organizers of the Carg`ese 2004 String Theory school. This work was supported by Stichting FOM. Appendix A. Soft graviton emission vertex Consider emission of a positive helicity collinear graviton with energy ω and small transverse momentum k = (k2 , 0), k2 ω. Its polarization tensor, satisfying the constraints hij kj = 0, hii = 0, is given by ⎞ ⎛ 2 2 −k2 /ω −ik2 /ω k2 /ω ⎠ . 1 i (4.1) hij ≈ ⎝ −k2 /ω 2 2 i −1 − k2 /ω −ik2 /ω The emission amplitude is A ∝ p − k|Tij |phij , where the energy momentum tensor matrix element is 1 P |Tµν |p = P |φ,µ φ,ν − ηµν (∂φ)2 |p = pµ Pν + pν Pµ − ηµν (pP ) . (4.2) 2 From this we ﬁnd A ∝ E 2 k22 /ω 2 . B. Multiple BH production amplitude We will compute the n = 2 amplitude in the case when both small BHs are produced at rest in the c.m. frame: qi = (mi , 0, 0, 0) in Fig. 5. The calculation naturally splits into 3 steps: 1) compute the loop integral over the longitudinal momenta; 2) multiply by the emission vertices and integrate over the transverse momenta; 3) multiply by the BH production vertices. It is convenient to write the graviton momenta as ki = (

xi − yi mi xi + yi mi + , − , ki ), 2 2 2 2

ki = qi − ki

Figure 3. Graviton emission and BH production vertices

(4.3)

368

VYACHESLAV S. RYCHKOV

Figure 4. A diagram giving an s-independent contribution to σn

Figure 5. A diagram for one large and two small BHs in the ﬁnal state

(i = 1, 2). The longitudinal loop integral now separates into the integrals over xi and yi . The part depending on xi is 1 (4.4) I = dx1 dx2 2 2 k1 k2 (p − k1 )2 (p − k1 − k2 )2 dx1 dx2 . ≈ 2 2 (m1 x1 − k1 )(m2 x2 − k2 )(−Ex1 − k21 )(−E(x1 + x2 ) − k212 ) The +i is implicit in each denominator. This integral is easy to compute by closing the contour in the lower half-plane. Omitting corrections of the order m/E 1, we have I ∝ [E 2 k21 (k21 + k22 )]−1 .

(4.5)

Since we used complex analysis, it is important to check that the integral is dominated by soft, almost real gravitons. This is indeed true, the important region being |xi | k2i /mi . The tail corresponding to |xi | k2i /mi can be estimated directly as: 1 dx1 dx2 I. (4.6) m1 m2 E 2 x21 x22 This check also justiﬁes post factum neglecting the dependence of the graviton emission vertices on xi , as well as the omission of x2 terms in the denominators of (4.4). Before proceeding to the next step, we have to add diagrams differing by the order of graviton emission (Fig. 6). This summation has the effect I → [E 2 k21 k22 ]−1 . Finally, we

Figure 6. One of the 3 permuted diagrams

BLACK HOLE PRODUCTION

369

multiply by the same factor arising from the y-integration, and get: [E 2 k21 k22 ]−2

(longitudinal part) .

(4.7)

We will assume that the small BHs are produced in the spin 0 state, so that the colliding gravitons have opposite helicities. The product of the corresponding emission vertices is ∝ (E/m)4 (k2 + ik3 )2 (k2 − ik3 )2 = (E/m)4 (k2 )2 .

(4.8)

Multiplying (4.7) by such factors for both BHs and integrating over |ki | 1/E (the “fall safe” condition), we get a number ∼ (m1 m2 )−4 . Finally, we multiply by the BH production vertices and arrive at f (2) ∼ f (s) f (m21 ) f (m22 ) (m1 m2 )−4 .

(4.9)

This formula agrees with the general result (3.3) in the considered case. Extension of the above computation to the general case is quite straightforward. A point worth mentioning is the use of the standard identity [a1 (a1 + a2 ) · · · (a1 + . . . + an )]−1 = [a1 · · · an ]−1 (4.10) perm

when summing over the order of graviton emission in the n > 2 case. In [4], the estimate d4 ki /(ki2 ki 2 ) ∼ O(1) was used in computing the amplitude. However, our analysis shows that the correct estimate is: 1 dx dy d2 k d4 k ∼ 2 2 . ∝ (4.11) k 2 k 2 (mx − k2 + i )(my − k2 + i ) m E This extra factor, for each of n BHs, explains the difference between (3.3) and the result of [4]. References 1. 2. 3. 4. 5. 6.

P. Kanti, hep-ph/0402168. S. B. Giddings and V. S. Rychkov, Phys. Rev. D 70, 104026, hep-th/0409131. V. S. Rychkov, hep-th/0410041. M. B. Voloshin, Phys. Lett. B 524, 376 (2002), hep-ph/0111099. D. M. Eardley and S. B. Giddings, Phys. Rev. D 66, 044011 (2002), gr-qc/0201034. V. S. Rychkov, Phys. Rev. D 70, 044003 (2004), hep-ph/0401116; hep-ph/0405104.

TOWARDS THE EXACT DILATATION OPERATOR OF N = 4 SUPER YANG-MILLS THEORY

ANTON V. RYZHOV Department of Physics, Brandeis University Waltham, MA 02454, USA

Abstract. I present a summary of hep-th/0404215, which suggested a novel way of organizing the dilatation operator D of planar N = 4 SYM in the SU (2) sector. Instead of the usual perturbative expansion in powers of λ, we split D into parts D(n) according to the number n of independent pairwise interactions between spins at different sites. The BMN limit ﬁxes D(1) completely, and it has regular expansions at both small and large values of λ. Anomalous dimensions of “long” operators in the two-scalar sector then generically √ scale as λ at large λ, i.e. in the same way as energies of semiclassical states in the dual AdS5 × S 5 string theory.

1. Introduction The N = 4 supersymmetric SU (N ) Yang-Mills theory is a family of CFT’s parametrized 2 N . The problem of computing dimensions of by N and the ‘t Hooft coupling λ = gYM local gauge invariant operators should simplify in the planar limit of N → ∞, λ ﬁxed. In this limit the AdS/CFT duality conjecture suggests that conformal dimensions should be smooth functions of λ, and have regular expansions at both large and small λ. The main obstacle to verifying the correspondence explicitly isn our lack of tools for obtaining exact all-order results on either gauge theory ( cn λ ) or string theory n ( (√bλ) n ) side. One potentially fruitful idea of how to go beyond the ﬁrst few orders in SYM perturbation theory is to try to determine the exact structure of the dilatation operator D by imposing additional conditions (like superconformal symmetry, BMN limit, integrability, etc.), implied by the expected correspondence with AdS5 × S 5 string theory. One may then be able to see if they admit a regular expansion not only at small but also at large λ. This is the approach explored in [1]. Using regularity of a BMN-type scaling limit [2, ?] as an input, we concentrate on the planar SU (2) sector of single trace SYM operators built out of chiral combinations X and Z of two the 6 SYM adjoint scalars, i.e. tr(X...XZ...ZX...) with canonical dimension L. This sector is closed under renormalization. The eigen-operators of D with J1 Z’s and J2 X’s (so L = J1 + J2 ) should be 371 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 371–377. © 2005 Springer. Printed in the Netherlands.

372

ANTON V. RYZHOV

dual to string states with two components of the SO(6) spin. On general grounds, the SYM dilatation operator computed in the planar limit should be a series in λ D=

∞ r=0

λr D2r . (4π)2r

(1.1)

Restricting D to planar graphs suggests that D2r should be given by local sums over sites a = 1, ..., L with Z and X interpreted as a spin “up” and spin “down” state of a periodic (a + L ≡ a) spin chain for which D is the Hamiltonian. The one-loop term D2 turns out to be equivalent to the Hamiltonian of the ferromagnetic XXX 12 Heisenberg spin chain, and r-loop contributions involve interactions of r + 1 nearest neighbors. Generic r-loop contributions to 1.1 contain terms linear in the projectors Qa,b = 1 − Pa,b (Pa,b permutes spins at sites a and b), a term quadratic in Qa,b , and so on: D2r

(1)

(2)

= D2r + D2r + ..., r (1) ar,c Qa,a+c , D2r = 2

(1.2) (n)

D2r ∼

Qn .

(1.3)

c=1 (n)

Qn in D2r stands for products of independent projectors, i.e. with all indices corresponding to different sites. Then D = D0 + D(1) + D(2) + ...,

D(1) = 2

∞ r=1

L r λr ar,c Qa,a+c . (4π)2r a=1 c=1

(1.4)

Using the periodicity of the chain (Qa,b+L = Qa,b , etc.) D(1) can be rewritten as D(1) =

L L−1

hc (L, λ)Qa,a+c

(1.5)

a=1 c=1

In [1] we determined the general expression for the coefﬁcients ar,c and thus the functions hc (L, λ), i.e. found the spin-spin (linear in Q) part of the exact dilatation operator D. The key point was to demand that the BMN-type scaling limit L → ∞,

˜ ≡ λ = ﬁxed λ L2

(1.6)

of the coherent-state expectation value of D(1) is well deﬁned [2, 3]. This turns out to be (nearly) equivalent to the consistency with the BMN expression for the anomalous dimensions of the 2-impurity operators. Imposing the condition of agreement with the BMN square root formula ﬁxes one remaining free coefﬁcient at each order in r. 2. Constraints on D(1) Following [2, 3], we consider the coherent state path integral representation for the quantum mechanics of D as a generalized spin chain Hamiltonian. The corresponding action

TOWARDS THE EXACT DILATATION OPERATOR OF N = 4 SYM

373

for a collection of unit 3-vectors na (t) at each of L sites of the chain (n| σai |n = nia , (nia )2 = 1, i = 1, 2, 3) is then given by S=

dt

L

LWZ (na ) − D ,

D = n| D |n ,

(2.1)

a=1

where LWZ (na ) = Ci (na )n˙ ia ensures the proper SU (2) commutation relations if one considers na back as spin operators. The spin chain in question is ferromagnetic,1 and in the long spin chain limit L → ∞ one expects that the low energy excitations of the spin chain will be captured by the semiclassical dynamics of 2.1. The correspondence with string theory then suggests that the low energy effective action for the system governed by (2.1) should have a well deﬁned continuum limit. To take the continuum limit one may introduce a ﬁeld n(σ, t), 0 < σ ≤ 2π, with na (t) = n( 2πa L , t), so that 2.1 becomes S→L

dσ ˜ . Ci (n)n˙ ia − H(∂n, ∂ 2n, ...; λ) 2π

2π

dt 0

(2.2)

˜ and H which originated from D should be a regular function of the effective coupling λ 1 ˜ σ-derivatives of n(t, σ) in the limit L → ∞, λ ﬁxed (with subleading Ln terms omitted). Quantum corrections are then suppressed because of the large prefactor L in front of the action. Since the Qa,b in (1.4) satisfy n|Qa,b |n =

1 1 (1 − na · nb ) = (na − nb )2 , 2 4

(2.3)

n| D(1) |n contains terms quadratic in n (but all orders in derivatives); n| D(2) |n, terms quartic in n, etc. The approximation that distinguishes D(1) from all higher D(k) in 1.2 is the one in which one keeps only small ﬂuctuations of n(t, σ) near its (“all-spins where |δn| 1, so that up”) ground-state value n0 = (0, 0, 1). Then n = n0 + δn, higher powers of the ﬂuctuating ﬁeld δn are suppressed, regardless the number of spatial derivatives acting on them. Such conﬁgurations correspond to semiclassical spinning string states with J1 J2 , and are close to a single-spin BPS state. They should indeed represent semiclassical or coherent-state analogs of few-impurity BMN states, having the same BMN energy-spin relation which is indeed reproduced in the limit J1 J2 by the classical two-spin string solutions. In this BMN-type approximation, demanding that the continuum version of n| D(1) |n have a regular scaling limit 1.6 implies (after integrating by parts) λr (1) D (4π)2r 2r

r

= →

1

λr ar,c (na − na+c )2 2(4π)2r c=1 2r+2 ˜ r (∂ r n)2 + O ∂ dr λ , L2

Then the state with all spins “up” (represented by the operator trZ L ) is a true vacuum.

(2.4)

374

ANTON V. RYZHOV

with the coefﬁcients ar,c satisfying ar,c =

(−1)r−c (2r)! ar,r , (r − c)!(r + c)!

c = 1, ..., r − 1 .

(2.5)

r (−1)r−1 2r (−1)r−1 c a = ar,r , r,c 22r (2r)! c=1 22r+1

(2.6)

The ﬁrst non-vanishing coefﬁcient becomes dr =

and the r-loop contribution to the expectation value of D(1) takes the form (cf. 2.4) 2π dσ r 2 1 λr (1) r ˜ (∂ ) (2.7) D → L d n ) + O( λ r (4π)2r 2r 2π L 0 What remains is to ﬁnd the values of ar,r , which can be done by analyzing the spectrum of two-impurity BMN operators OBMN n

J πn(2p + 1) 1 √ cos tr(XZ p XZ J−p ) . J +1 J + 1 p=0

=

(2.8)

Their anomalous dimensions can be computed in both string theory and gauge theory in the large J, ﬁxed Jλ2 limit, and one ﬁnds + ∆BMN = J + 2 1 +

λ 2 1 n + O( ) . J2 J

(2.9)

To reproduce 2.9 by acting with the dilatation operator (1.1) on OBMN , we should have n ar,c =

(−1)r−c Γ(2r + 1) Γ(2r − 1) . Γ(r − c + 1) Γ(r + c + 1) Γ(r) Γ(r + 1)

(2.10)

3. Summing D(1) to all orders The expression for D(1) can actually be summed up explicitly. First, summing 2.7 over r with dr in 2.6 gives a very simple formula for the quadratic in n (“small ﬂuctuation” or “BMN”) part of the coherent state effective Hamiltonian in 2.2: D(1)

= →

∞

λr (1) D2r 2r (4π) r=1 ! 2π dσ 1 ˜ ∂ 2 − 1 n + O( 1 ) . n 1− λ L 2π 4 L 0

(3.1)

Remarkably, this is the same expression that follows from the classical AdS5 × S 5 string sigma model action expanded in the limit 1.6. Next, let us substitute the values 2.10 for

TOWARDS THE EXACT DILATATION OPERATOR OF N = 4 SYM

375

the coefﬁcients ar,c we have found above into D(1) in 1.4 and try to formally perform the summation over r ﬁrst, independently for each Qa,a+c term. We get D(1)

=

2

∞ L

fc (λ) Qa,a+c ,

fc (λ) =

∞ r=c

a=1 c=1

λr ar,c . (4π)2r

(3.2)

Remarkably, the series representation for the coefﬁcients fc (λ) can then be summed up in terms of the standard hypergeometric functions, fc (λ) =

λ 4π 2

c

Γ(c − 12 ) √ 4 π Γ(c + 1)

2 F1 (c

1 λ 1 − , c + ; 2c + 1; − 2 ) 2 2 π

(3.3)

c

2λ The coefﬁcient in front of 2 F1 is equal to (4π) 2c ac,c . The fc go to 0 rapidly at large c, so we effectively have a spin chain with short range interactions. The resulting coefﬁcients fc (λ) are smooth positive functions of λ having regular expansion at both small λ and large λ, in which case we ﬁnd

fc (λ)λ→∞

=

√ 1 λ 1 + O( ) , π 2 4c2 − 1 λ

(3.4)

√ The square root fc → λ asymptotics of (3.4) is related to the cut structure of 2 F1 . So far, we have treated all Qa,a+c terms as independent but for ﬁnite L the terms with c and c + mL are the same because of the periodicity of the chain (implied by cyclicity of the trace in the operators). Also, under the sum over a one has Qa,a+c = Qa,a+L−c . Therefore, for ﬁnite L the sum over c should, in fact, be restricted to run from c = 1 to c = L − 1, D(1)

=

2

L ∞ a=1 c=1

fc (λ) Qa,a+c =

L L−1

hc (L, λ) Qa,a+c .

(3.5)

a=1 c=1

The new coefﬁcients hc depend on both the the length of the chain and ’t Hooft coupling hc (L, λ)

=

c+mL ∞ Γ(c + mL − 12 ) λ √ 2 4π 4 π Γ(c + mL + 1) m=0 1 λ 1 × 2 F1 (c + mL − , c + mL + ; 2c + 2mL + 1; − 2 ) 2 2 π + (c → L − c) .

(3.6)

For ﬁnite L, we may expand the hypergeometric functions in 3.6 at large λ as in (3.4) and then do the sum over m. Ignoring the issue of convergence of the resulting strongcoupling expansion, that leads to the following simple result for the leading-order term hc (L, λ)λ→∞

√ π sin L λ −1 = + O(λ ) π 2πL cos L − cos 2πc L

(3.7)

376

ANTON V. RYZHOV

We can apply the above relations to the ﬁrst non-trivial case of small length. The nonBPS operator with L = 4 is the level four descendant K of the Konishi scalar operator K = tr[X, Z]2 = 2 tr(XZXZ − XXZZ) .

(3.8)

The action of D(1) 3.5 on K is determined by noting that Π1 K = Π3 K = 6K,

Π2 K = 0 ,

D(1) K = γ (1) K ,

and one ﬁnds the following surprisingly simple result + 3 λ (1) 1+ 2 −1 . γ = 2 π

(3.9)

(3.10)

4. Concluding remarks The resulting D(1) may be interpreted as a Hamiltonian of a periodic spin chain with long-range interactions. One could conjecture that this spin chain may be integrable; and, furthermore, the higher-order terms D(2) , D(3) , ... in 1.4 may be effectively determined by D(1) , e.g., expressed in terms of higher conserved charges of the chain. This would then determine the full D. Inspired by recent work in spin chains, in [1] we suggested to organize the dilatation operator as an expansion 1.4, 1.2 in powers of independent projection operators Qa,b [3] at L sites of spin chain D = D0 +

∞

D(n) with D0 = L,

n=1

D(1) =

L L−1

hc (L, λ) Qa,a+c , ...

(4.1)

a=1 c=1

where D(n) are given by sums of products of n Q’s at independent sites of the spin chain. We determined the coefﬁcients in D(1) by demanding that its BMN-type scaling limit [2, 3] be regular, and found that it admits a very simple representation 3.2, applicable at least in the large L limit. This representation includes all √ orders in λ and suggests that the corresponding anomalous dimensions should grow as λ for large λ. A natural extension of this work would be to try to ﬁnd the next term in the expansion (4.1), namely D(2) =

L

L−1

hc1 ;c2 ,c3 (L, λ) Qa,a+c1 Qa+c2 ,a+c2 +c3

(4.2)

a=1 c1 ,c2 ,c3 =1

The prime on the sum here means that certain terms should be omitted: since Qa,b Qb,c + Qb,c Qa,b = Qa,b + Qb,c − Qa,c the terms with c2 = c1 and c2 = c3 − c1 have already been included in D(1) . The contributions of higher D(n) terms should be crucial at ﬁnite √ λ asymptotL, resolving, in particular, the above-mentioned contradiction between the √ ics of the coefﬁcients in D(1) and the expected 4 λ scaling of dimensions of operators corresponding to string modes.

TOWARDS THE EXACT DILATATION OPERATOR OF N = 4 SYM

377

The AdS/CFT duality suggests that D should correspond to an integrable spin chain. The simplest possibility could be that, by analogy with the Inozemtsev chain , the operator D(1) (with interaction coefﬁcients given by 3.6) represents a Hamiltonian of an integrable spin 1/2 chain, while all higher order terms D(n) are effectively determined by D(1) through integrability. Satisfying the requirements of integrability, BMN scaling and consistency with gauge theory should be essentially unique. References 1. 2. 3.

A. V. Ryzhov and A. A. Tseytlin, “Towards the exact dilatation operator of N=4 super YangMills theory,” hep-th/0404215. M. Kruczenski, “Spin chains and string theory,” hep-th/0311203. M. Kruczenski, A. V. Ryzhov and A. A. Tseytlin, “Large spin limit of AdS5 ×S 5 string theory and low energy expansion of ferromagnetic spin chains,” hep-th/0403120.

MODULI TRAPPING AND STRING GAS COSMOLOGY

SCOTT WATSON Department of Physics, Brown University, Box 1843, Providence, RI 02912, USA

Abstract. In this talk I will discuss the role of ﬁnite temperature quantum corrections in string cosmology and show that they can lead to a stabilization mechanism for the volume moduli. I will show that from the higher dimensional perspective this results from the effect of states of enhanced symmetry on the one-loop free energy. These states lead not only to stabilization, but also suggest an alternative model for ΛCDM. At late times, when the low energy effective ﬁeld theory gives the appropriate description of the dynamics, the moduli will begin to slow-roll and stabilization will generically fail. However, stabilization can be recovered by considering cosmological particle production near the points of enhanced symmetry leading to the process known as moduli trapping.

1. Initial Conditions One problem in string cosmology is the issue of initial conditions. Not only must models of string cosmology address the standard initial condition problems in cosmology, but string theory also predicts the existence of extra dimensions. The usual prescription for dealing with the extra dimensions is to take them small, stable, and unobservable. However a complete model of string cosmology should explain how this came about and why the explicit breaking of Lorentz invariance should be allowed. A step in this direction was ﬁrst proposed by Brandenberger and Vafa[1]. They argued that, by considering the dynamics of a string gas in nine, compact spatial dimensions initially taken at the string scale, one could explain why three dimensions grow large while six stay compact. The crux of their argument is based on the fact that in addition to the usual Kaluza Klein modes of a particle on a compact space, strings also possess winding modes. These extra degrees of freedom will generically halt cosmological expansion; however, if these modes could annihilate with their anti-partners through the process, W + + W − unwound string,

(1.1)

this would allow the dimension they occupy to expand. Then, the fact that strings generically intersect (interact) in at most three spatial dimensions means that in the remaining six dimensions thermal equilibrium cannot be maintained. Thus, the winding modes will 379 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 379–383. © 2005 Springer. Printed in the Netherlands.

380

SCOTT WATSON

drop out of equilibrium and the six spatial dimensions will be frozen near the string scale. Furthermore, once all the winding modes in the three large dimensions annihilate, the universe emerges ﬁlled with a gas of momentum modes, which evolves as a radiation dominated universe. 2. Strings at Finite Temperature The usual starting point of string cosmology is the action, 1 2 1 d+1 √ −2ϕ 2 R + 4(∇ϕ) − H + O(α ) + O(gs ). d x −ge S= 2 2κ 12

(2.1)

For simplicity we will ignore the Ramond sector and set H = 0. Here we have in mind the heterotic string on a toroidal background. Motivated by the Brandenberger-Vafa scenario we will take the background to be R4 × T 6 , where we assume that the three spatial dimensions have grown large enough to be approximated by an FRW universe and the six small dimensions are toroidal and near the string scale. To include time dependence we make use of the adiabatic approximation, which implies that we can replace static quantities by slow varying functions of time1 . The action (2.1) represents a double expansion in both the string coupling gs ∼ eϕ 1 and the string tension, T = 4πα . We now want to include terms coming from gs corrections at ﬁnite temperature[2, 3]. Let us consider the 1-loop free energy ¯ F = T Z1 ∼ β −1 (2.2) e−βM (n,ω,N,N ) , ¯ ) is the string mass, and β = where Z1 is the one loop partition function, M (n, ω, N, N 1/T is the inverse temperature. In the √ early universe we are interested in temperatures near or below the string scale (β α ) where the major contribution to the one-loop free energy can be seen to come from the massless modes of the string. In the case of the √heterotic string there are additional massless states that occur at the special radius R = α . These extra states include winding and momentum modes of the string. To understand the dynamics that result by including these states we can ﬁnd the energy density and pressure, which follow from the free energy as 1 ∂ (βF ) V ∂β ∂F 1 pi = − , 9V ∂(lnRi ) ρ=

(2.3)

where V is the spatial volume and Ri is the scale factor in the ith direction. As discussed above, we take initial conditions where three dimensions have grown large and six remain near the string scale. For such initial data, the above pressure in three dimensions corresponds to the equation of state of a radiation dominated universe p3 = 13 ρ, whereas the 1

This is known to be a good approximation in the early universe, and is equivalent to saying that the time scale for string interactions is greater than the rate of expansion. However, one should keep in mind that such an approximation will mostly likely break down near a curvature singularity.

MODULI TRAPPING AND STRING GAS COSMOLOGY

381

Figure 1. The evolution of the volume modulus of the extra dimensions for different initial values near the self-dual radius. We see that the moduli are eventually stabilized at the self-dual radius where the pressure is found to vanish.

pressure in the small dimensions gives the behavior, √ √ p6 < 0 for R6 > α and p6 > 0 for R6 < α .

(2.4)

In dilaton gravity negative pressure implies a contracting universe, whereas positive pressure leads to expansion. Thus, as can be seen in Figure 2, pressure leads to a stabilizing effect for the scale factor of √ the extra dimensions driving the radius toward the enhanced symmetry point R(6) = α where the pressure vanishes. At this location the gauge symmetry of the heterotic string is enhanced, E8 × E8 + U (1)6 −→ E8 × E8 + SU (2)6 . In addition to stabilizing the volume moduli, it has been shown that the remnant string modes, if taken in the dark sector, can also lead to an interesting cold dark matter candidate[4, 5] (see also[6]). 3. Moduli Trapping If one attempts to extend the arguments above to the 4D effective ﬁeld theory, one ﬁnds that the stabilization mechanism no longer holds. This is not surprising since the pressure in the extra dimensions has no analog from the 4D perspective. However, one thing that should remain is the idea of enhanced symmetry.

382

SCOTT WATSON

√ Recall that it was the contribution of the enhanced symmetry states near R = α that led to the pressure terms in (2.4) stabilizing the extra dimensions. We can account for these enhanced states from the effective ﬁeld theory (EFT) perspective √ by considering the effects of particle production near the enhanced symmetry point, R = α . To understand how this mechanism works let us consider the simplest case of heterotic strings on the background R4 × S 1 . The low energy effective action comes from the compactiﬁcation of the action (2.1). The dynamics are then given by dilaton gravity coupled to a chiral U (1) gauge theory, Lm = (∂σ)2 −

1 1 (Fµν )2 − 2 (F¯µν )2 , 2 4g 4g

(3.1)

¯ is the left (right) gauge theory resulting from the compactiﬁcawhere F = dA (F¯ = dA) tion of the higher dimensional metric and ﬂux and g is the gauge coupling. The scalar σ gives the radius of the compactiﬁcation√and can be scaled to measure the departure from the self-dual radius, i.e. σ = 0 at R = α . We see that σ has only a kinetic term and the lack of a potential implies the radius is free to take any value. However, as the modulus passes near the self-dual radius we have noted that there are additional massless degrees of freedom. If our theory is to be complete these extra degrees of freedom must be included in the low energy effective action. This is accomplished by lifting the effective lagrangian in (3.1) to a non-abelian gauge theory, in this case chiral SU (2). We introduce the covariant derivative, Dµ σ = ∂µ σ + gAµ σ.

(3.2)

This leads to a time dependent mass for the new vector Aµ 1 2 g σ(t)2 A2µ . 2

(3.3)

This time dependent mass implies particle production in our cosmological space-time. The stabilization of the radius σ can now be realized as follows; initially σ is dominated strictly by the kinetic term, however once it passes near the enhanced symmetry point σ = 0, Aµ particles will be produced. Then, as σ continues its trajectory the mass of the Aµ ’s will increase and this leads to backreaction on σ. This force, along with friction from the cosmological expansion, will eventually stabilize σ at the enhanced symmetry point σ = 0. This is a simple example of moduli trapping[7, 8, 9]. Although we have considered here a simple toy model of a string on a circle, points of enhanced symmetry are present in nearly all string and M-theory compactiﬁcations. Moreover, it is worth mentioning that this mechanism need not apply only to volume moduli. In fact, in Kofman, et. al.[7] the modulus of interest was the distance between two branes which is of course related by T-duality to the case we have considered here. 4. Conclusions We have seen that from the 10D perspective it is possible to stabilize the volume modulus of a heterotic string compactiﬁcation on a T 6 . The stabilization was found to be the

MODULI TRAPPING AND STRING GAS COSMOLOGY

383

result of the pressure exerted on the compact space due to the presence of enhanced symmetry states contributing to the one-loop free energy of the strings at ﬁnite temperature. In the 4D effective ﬁeld theory such effects can be understood by considering particle production near the enhanced symmetry point. Near this point additional massless states are allowed which can be particle produced in the cosmological space-time. These states then get masses via the string Higgs Effect and backreact on the modulus stabilizing the extra dimensions at the self-dual radius. Acknowledgements I would like to thank Robert Brandenberger, Sera Cremonini, Steve Gubser, Laur Jarv, Liam McAllister, and Thomas Mohaupt for useful comments and discussions during this work. I would also like to thank the organizers of the Carg`ese 2004 ASI. This work was supported by NASA GSRP. References 1. 2. 3. 4. 5. 6. 7. 8. 9.

R. H. Brandenberger and C. Vafa, “Superstrings In The Early Universe,” Nucl. Phys. B 316, 391 (1989). J. Kripfganz and H. Perlt, “Cosmological Impact Of Winding Strings,” Class. Quant. Grav. 5, 453 (1988). A. A. Tseytlin and C. Vafa, “Elements of string cosmology,” Nucl. Phys. B 372, 443 (1992) [arXiv:hep-th/9109048]. S. S. Gubser and P. J. E. Peebles, “Cosmology with a dynamically screened scalar interaction in the dark sector,” arXiv:hep-th/0407097. S. S. Gubser and P. J. E. Peebles, “Structure formation in a string-inspired modiﬁcation of the cold dark matter model,” arXiv:hep-th/0402225. T. Battefeld and S. Watson, “Effective ﬁeld theory approach to string gas cosmology,” JCAP 0406, 001 (2004) [arXiv:hep-th/0403075]. L. Kofman, A. Linde, X. Liu, A. Maloney, L. McAllister and E. Silverstein, “Beauty is attractive: Moduli trapping at enhanced symmetry points,” JHEP 0405, 030 (2004) [arXiv:hepth/0403001]. S. Watson, “Moduli stabilization with the string Higgs effect,” arXiv:hep-th/0404177. L. Jarv, T. Mohaupt and F. Saueressig, “Singular compactiﬁcations and cosmology,” arXiv:hep-th/0311016.

STRINGY GEOMETRY: GENERALIZED COMPLEX STRUCTURE

MAXIM ZABZINE LPTHE, Universit´e Paris 6, 4 pl Jussieu, 75252 Paris cedex, France

Abstract. We consider a worldsheet realization of generalized complex geometry, a notion introduced recently by Hitchin which interpolates between complex and symplectic manifolds.

The recently developed notion of generalized complex geometry naturally extends and uniﬁes complex and symplectic geometries, in general interpolating between the two [1, 2]. There have been many hints that this geometry should be relevant to string theory. In this note, we present toy 2D models which provide the “physical” derivation of the generalized complex geometry. This note is based on the joint with U. Lindstr¨om, R. Minasian and A. Tomasiello [3]. 1. Mathematical preliminaries In this Section we review some basic notions and ﬁx notations. Namely we collect general facts concerning the generalized complex structure, see [1] and [2] for further details. Also we work out the coordinate form of the integrability conditions for the generalized complex structure. Let us start by recalling the deﬁnition of the standard complex structure on a manifold M (dim M = d). An almost complex structure is deﬁned as a linear map on the tangent bundle1 J : T → T such that J 2 = −1d . This allows the deﬁnition of projectors on T , π± =

1 (1d ± iJ). 2

(1.1)

An almost complex structure is called integrable if the projectors π± deﬁne integrable distributions on T , namely if (1.2) π∓ [π± X, π± Y ] = 0 for any X, Y ∈ T where [ , ] is a standard Lie bracket on T . 1

In this note we consider the complexiﬁed tangent and cotangent bundles.

385 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 385–391. © 2005 Springer. Printed in the Netherlands.

386

MAXIM ZABZINE

A generalization of the notion of complex structure has been proposed by Hitchin [1]. In Hitchin’s construction T is replaced by T ⊕ T ∗ and the Lie bracket is replaced by the appropriate bracket on T ⊕T ∗ , the so called Courant bracket. Thus a generalized complex structure is an almost complex structure J on T ⊕T ∗ whose ±i-eigenbundles are Courant involutive. In other words, a generalized complex structure is splitting of T ⊕ T ∗ into a sum of two complementary Dirac subbundles. A detailed study of generalized complex geometry can be found in Gualtieri’s thesis [2]. Now let us give detailed deﬁnitions. On T ⊕ T ∗ there is a natural indeﬁnite metric deﬁned by (X + ξ, X + ξ) = −iX ξ. In the coordinate basis (∂µ , dxµ ) we can write this metric as follows 0 1d I= . (1.3) 1d 0 A generalized almost complex structure is a map J : T ⊕ T ∗ → T ⊕ T ∗ such that J 2 = −12d and that I is hermitian with respect to J , J t IJ = I. On T ⊕ T ∗ there is a Courant bracket which is deﬁned as follows 1 [X + ξ, Y + η]c = [X, Y ] + LX η − LY ξ − d(iX η − iY ξ). 2

(1.4)

This bracket is skew-symmetric but in general does not satisfy the Jacobi identity. However if there is a subbundle L ⊂ T ⊕ T ∗ which is involutive (closed under the Courant bracket) and isotropic with respect to I then the Courant bracket on the sections of L does satisfy the Jacobi identity. This is a reason for imposing hermiticity of I with respect to J . One important feature of the Courant bracket is that, unlike the Lie bracket, this bracket has a nontrivial automorphism deﬁned by a closed two-form b, eb (X + ξ) = X + ξ + iX b.

(1.5)

[eb (X + ξ), eb (Y + η)]c = eb [X + ξ, Y + η]c .

(1.6)

such that We can construct the projectors on T ⊕ T Π± =

∗

1 (I ± iJ ) ; 2

(1.7)

the almost generalized complex structure J is integrable if Π∓ [Π± (X + ξ), Π± (Y + η)]c = 0,

(1.8)

for any (X + ξ), (Y + η) ∈ T ⊕ T ∗ . This is equivalent to the single statement [X + ξ, Y + η]c − [J (X + ξ), J (Y + η)]c + J [J (X + ξ), Y + η]c + +J [X + ξ, J (Y + η)]c = 0

(1.9)

which resembles the deﬁnition of the Nijenhuis tensor. To relate this construction to the physical models we have to reexpress the above deﬁnitions in coordinate form. The map J can be written in the form J P J = , (1.10) L K

STRINGY GEOMETRY: GENERALIZED COMPLEX STRUCTURE

387

where J : T M → T M, P : T ∗ M → T M, L : T M → T ∗ M and K : T ∗ M → T ∗ M and hence they correspond to the tensor ﬁelds, J µν , Lµν , P µν and Kµν . Then the condition J 2 = −12d becomes J µν J νλ + P µν Lνλ = −δ µλ ,

(1.11)

J µν P νλ + P µν Kν λ = 0, Kµν Kν λ + Lµν P νλ = −δ µλ , Kµν Lνλ + Lµν J νλ = 0.

(1.12) (1.13) (1.14)

The hermiticity of I with respect to J translates into the following conditions J µν + Kµν = 0,

P µν = −P νµ ,

Lµν = −Lνµ .

(1.15)

In local coordinates the integrability condition (1.9) is equivalent to the following four conditions

J µν,ρ P ρλ J λν L[λρ,γ]

+

J ν[λ J µρ],ν + J µν J ν[λ,ρ] + P µν L[λρ,ν] = 0

(1.16)

P [µ|ν P |λρ] ,ν = 0

(1.17)

+ P ρλ,ν J µρ − J λρ,ν P µρ + J λν,ρ P µρ − P µλ,ρ J ρν Lνλ J λ[γ,ρ] + J λ[ρ Lγ]ν,λ + Lλρ J λγ,ν + J λρ Lλγ,ν

=0

(1.18)

=0

(1.19)

To summarize, the generalized complex structure J is deﬁned by three tensor ﬁelds J µν , Lµν and P µν which satisfy the algebraic conditions (1.11)-(1.15) and the differential conditions (1.16)-(1.19). The usual complex structure J is embedded in the notion of generalized complex structure J 0 . (1.20) J = 0 −J t One can check that all properties (1.11)-(1.19) are satisﬁed provided that J is a complex structure. Also, a symplectic structure is an example of a generalized complex structure 0 −ω −1 J = (1.21) ω 0 where ω is an ordinary symplectic structure (dω = 0). More exotic examples exist and are given by manifolds, that do not admit any known complex or symplectic structure, but do admit a generalized complex structure [2]. Consider a generalized complex structure J ; a new generalized complex structure can be generated by 1 0 1 0 J (1.22) Jb = b 1 −b 1 if b ∈ Ω2closed (M). The structure Jb is integrable due to the fact that the transformation (1.5) is an automorphism of the Courant bracket. The transformation (1.5) is called a b-transform.

388

MAXIM ZABZINE

The key feature of a complex manifold is that is locally equivalent to ICk via a diffeomorphism. For symplectic manifolds the Darboux theorem states that a symplectic structure is locally equivalent, via diffeomorphism, to the standard symplectic structure (R I 2k , ω), where (1.23) ω = dx1 ∧ dx2 + ... + dx2k−1 ∧ dx2k . For generalized complex manifolds there exists a generalized Darboux theorem [2], which states that in a neighborhood of a regular point2 a generalized complex structure on a manifold M is locally equivalent via a diffeomorphism and a b-transform (see (1.22)), to the product of an open set in ICk and an open set in the standard symplectic space (R I d−2k , ω). The Courant bracket on T ⊕ T ∗ can be twisted by a closed three form H. Namely given a closed three form H one can deﬁne another bracket on T ⊕ T ∗ by [X + ξ, Y + η]H = [X + ξ, Y + η]c + iX iY H.

(1.24)

This bracket has similar properties to the Courant bracket. Again if a subbundle L ⊂ T ⊕ T ∗ is closed under the twisted Courant bracket and isotropic with respect to I, then the Courant bracket on the sections of L does satisfy the Jacobi identity. Thus in the integrability condition (1.9) the Courant bracket [ , ]c can be replaced by the new twisted Courant bracket [ , ]H . In local coordinates the new integrability condition is equivalent to four expressions J ν[λ J µρ],ν + J µν J ν[λ,ρ] + P µν (L[λρ,ν] + J σ[λ Hρ]σν ) = 0 P J µν,ρ P ρλ

+

P ρλ,ν J µρ

+

J λ[ν,ρ] P µρ

−

P µλ,ρ J ρν

−P

λσ

[µ|ν

P

µρ

P |λρ] ,ν

(1.25)

=0

(1.26)

Hσρν = 0

(1.27)

J λν L[λρ,γ] + Lνλ J λ[γ,ρ] + J λρ Lγν,λ + J λγ Lνρ,λ + Lλρ J λγ,ν + +J λρ Lλγ,ν + Hργν − J λ[ρ J σγ Hν]λσ = 0

(1.28)

2. Topological model In this Section we consider a toy topological model which will provide a “physical” derivation of generalized complex geometry. The model has the following action Stop = d2 σ dθ S+µ ∂= Φµ (2.1) written in N = (1, 0) superﬁelds where (++, =) are worldsheet indices and (+, −) are two-dimensional spinor indices. This is a topological system which describes the holomorphic maps Φ : Σ → M. The model is manifestly N = (1, 0) supersymmetric and can be deﬁned over any differential manifold M. We would like to ﬁnd the restrictions on M arising from the requirement that the model admits (2, 0) supersymmetry. 2 P is a Poisson structure and it will deﬁne a symplectic foliation. The point is called regular if P has constant rank in a neighborhood.

STRINGY GEOMETRY: GENERALIZED COMPLEX STRUCTURE

389

We have to look for additional (non manifest) supersymmetry transformations. The general transformations of S+ and Φ are given by the following expressions δ( )Φµ = + D+ Φν J µν − + S+ν P µν ,

(2.2)

δ( )S+µ = i + ∂++ Φν Lµν − + D+ S+ν Kµν + + S+ν S+ρ Nµ νρ + + + D+ Φν D+ Φρ Mµνρ + + D+ Φρ S+ν Qµρν

(2.3)

Classically the Ansatz (2.2) and (2.3) is unique on dimensional grounds and by Lorentz covariance [4]. This Ansatz involves seven different tensors on M. We have to require the standard N = (2, 0) supersymmetry algebra, i.e. the manifest and non-manifest supersymmetry transformations commute and the nonmanifest supersymmetry transformations satisfy the following conditions + µ [δ( 2 ), δ( 1 )]Φµ = 2i + 1 2 ∂++ Φ ,

+ [δ( 2 ), δ( 1 )]S+µ = 2i + 1 2 ∂++ S+µ .

(2.4)

Since the nonmanifest transformations are written in (1, 0) superﬁeld then the ﬁrst requirement is automatically satisﬁed. Next we have to calculate the commutator of two nonmanifest supersymmetry transformations. Imposing the condition (2.4) implies four algebraic and eleven differential conditions on the seven tensors introduced in (2.2) and (2.3). Before analyzing the algebra in detail it is useful to look at the invariance of the action. The action (2.1) is invariant under (2.2) and (2.3) if the following algebraic conditions J µν + Kν µ = 0,

Lµν = −Lνµ ,

P µν = −P νµ

(2.5)

1 L[µν,ρ] = Mρνµ 2

(2.6)

as well as the differential conditions 1 µν P ,ρ = −Nρ µν , 2

J µ[ν,ρ] = Qνρµ ,

are satisﬁed. The differential conditions (2.6) allow us to express all three index tensors in terms of appropriate derivatives of two index tensors J, P , L and K. These two index tensors can be combined as a single object J P J = , (2.7) L K where J : T ⊕ T ∗ → T ⊕ T ∗ . It is easy to see that the algebraic part of the supersymmetry algebra (2.4) can be written as a single equation, namely that J 2 = −12d . Passing then to the action, the algebraic condition (2.5) is equivalent to a hermiticity of I with respect to J (i.e., the natural pairing on T ⊕ T ∗ , see previous Section). Therefore J is an almost generalized complex structure. Finally we have to analyze the eleven differential conditions coming from the algebra using (2.6). Using the results from the previous Section, we can show that all differential conditions arising from (2.4) are the same as the conditions (1.16)-(1.19). Therefore we have proved that the differential conditions that come from the supersymmetry algebra are equivalent to integrability of J with respect to the Courant bracket. To summarize the topological model (2.1) admits (2, 0) supersymmetry if and only if manifold M is generalized complex manifold.

390

MAXIM ZABZINE

As we brieﬂy mentioned in the previous Section, a generalized complex manifold is equivalent locally, via diffeomorphism and b-transform, to a product of a symplectic and a complex manifolds. If we choose the Darboux coordinates (label n) along the symplectic leaf and the standard complex coordinates (label i, ¯i) transverse to the leaf the supersymmetry transformations (2.2) and (2.3) is simpliﬁed drastically and have the following form δΦi = i + D+ Φi , δS+i = i + D+ S+i , δΦn = − + S+(n+1) , n+1

δΦ

+

¯

¯

δΦi = −i + D+ Φi δS+¯i = −i + D+ S+¯i δS+(n+1) = −i + ∂++ Φn

= S+n ,

+

n+1

δS+n = i ∂++ Φ

(2.8) (2.9) (2.10) (2.11)

3. Topological model with WZ term In the previous Section we presented the topological model for which the extended supersymmetry is related to the generalized complex structure with integrability deﬁned with the respect to the Courant bracket. The natural question is now the following: if in the integrability condition the Courant bracket is replaced by the twisted Courant bracket, can we then construct a model which incorporates twisted integrability? This is in fact possible and the solution is related to the WZ term. We consider the topological model with an additional term 1 (3.1) d2 σ dθ D+ Φµ ∂= Φν Bµν Stop = d2 σ dθ S+µ ∂= Φµ − 2 The last term is a WZ term and it depends only on a closed three-form H Hµνλ =

1 (Bµν,λ + Bλµ,ν + Bνλ,µ ) 2

(3.2)

if the world-sheet does not have a boundary. The model (3.1) has N = (1, 0) supersymmetry and can be deﬁned over any differential manifold M equipped with a closed three-form H. The Ansatz for the nonmanifest supersymmetry transformations is given by the same expressions as before, (2.2) and (2.3). The off-shell supersymmetry algebra is exactly the same, (2.4). The main difference comes from the action. Namely invariance of the new action (3.1) under the transformations (2.2) and (2.3) leads to new relations between the three and two index tensors in the supersymmetry transformations. The action (3.1) is invariant under (2.2) and (2.3) if the following algebraic conditions are satisﬁed J µν + Kν µ = 0,

Lµν = −Lνµ ,

P µν = −P νµ

(3.3)

as well as the differential conditions 1 µν P ,ρ = −Nρ µν , 2

J µ[ν,ρ] + P µλ Hλνρ = Qνρµ ,

1 1 L[µν,ρ] + J λ[µ Hν]λρ = Mρνµ . 2 2

(3.4)

STRINGY GEOMETRY: GENERALIZED COMPLEX STRUCTURE

391

The algebraic part of all conditions remains the same as in the previous Section and therefore the two-index tensors can be combined in a single object J which is an almost generalized complex structure. However the differential conditions will change. Using (3.4) we have to study the supersymmetry algebra (2.4). We can show that the three differential conditions arising from (2.4) are the same as conditions (1.25)-(1.28). Therefore we have proved that the differential conditions coming from the supersymmetry algebra are equivalent to integrability of J with respect to the twisted Courant bracket. Finally let us remark that both actions (2.1) and (3.1) are invariant under following transformation with b ∈ Ω2closed (M) Φ → Φ,

1 S+µ → S+µ + bµν D+ Φν 2

(3.5)

which correspond to a b-transform in T ⊕ T ∗ (since ∂= Φ → ∂= Φ). Acknowledgements It is a pleasure to thank Ulf Lindstr¨om, Ruben Minasian and Alessandro Tomasiello for the stimulating collaboration. I would like to thank also the organizers of the Carg`ese 2004 ASI. References 1. 2. 3. 4.

N. Hitchin, “Generalized Calabi-Yau manifolds,” Q. J. Math. 54 (2003), no. 3, 281–308, arXiv:math.DG/0209099. M. Gualtieri, “Generalized complex geometry,” Oxford University DPhil thesis, arXiv:math.DG/0401221. U. Lindstr¨om, R. Minasian, A. Tomasiello and M. Zabzine, “Generalized complex manifolds and supersymmetry,” arXiv:hep-th/0405085, to appear in Comm.Math.Phys. U. Lindstr¨om, “Generalized N = (2,2) supersymmetric non-linear sigma models,” Phys. Lett. B 587 (2004) 216 [arXiv:hep-th/0401100].

INTEGRABILITY OF SUPERCONFORMAL FIELD THEORY AND SUSY N=1 KDV

ANTON M. ZEITLIN Department of High Energy Physics, Physics Faculty, St. Petersburg State University, Ul’yanovskaja 1, Petrodvoretz, St.Petersburg, 198904, Russia

Abstract. The quantum SUSY N=1 hierarchy based on sl(2|1)(2) twisted afﬁne superalgebra is considered. The construction of the corresponding Baxter’s Q-operators and fusion relations is outlined. The relation with the superconformal ﬁeld theory is discussed.

One of the most famous integrable systems (IS) is the Korteweg-de Vries hierarchy. It is related with the superconformal ﬁeld theory because its Poisson brackets give the Virasoro algebra and the involutive family of integrals of motion (IM) providing the integrability of the conformal ﬁeld theory (CFT). Since the late 1980-s the supersymmetric and fermionic extensions of the KdV system have been known (see e.g. [1], [2], [3] and references therein), which in turn are related with superconformal ﬁeld theory (SCFT). During the following years it was extensively studied both on the classical and the quantum level. However, up to the present nobody has applied the most successful method in the theory of integrable systems, the so-called quantum inverse scattering method (QISM) to these IS. In this short paper we demonstrate some algebraic tools giving possibility to study SUSY N=1 KdV via QISM. 1. RTT-relation The SUSY N=1 KdV model is related to the following L-operator: LF = Du,θ − Du,θ Φhα − (eδ−α + eα ), where hα , eδ−α ≡ eα0 , eα are the Chevalley generators of twisted afﬁne Lie superalgebra sl(2|1)(2) ∼ = osp(2|2)(2) ∼ = C(2)(2) , Du,θ = ∂θ + θ∂u is a superderivative, the variable u lies on a cylinder of circumference 2π, θ is a Grassmann variable, Φ(u, θ) = φ(u) − √i θξ(u) is a bosonic superﬁeld with the following Poisson brackets: {Du,θ Φ(u, θ), 2 Du ,θ Φ(u , θ )} = Du,θ (δ(u − u )(θ − θ )). Making a gauge transformation of the L3 Φ(u, θ) operator we obtain a new superﬁeld U(u, θ) ≡ Du,θ Φ(u, θ)∂u Φ(u, θ) − Du,θ √ =−θU (u) − iα(u)/ 2, where U and α generate the superconformal algebra under the 393 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 393–396. © 2005 Springer. Printed in the Netherlands.

394

ANTON M. ZEITLIN

Poisson brackets: {U (u), U (v)}

= δ (u − v) + 2U (u)δ(u − v) + 4U (u)δ (u − v),

{U (u), α(v)} {α(u), α(v)}

= =

3α(u)δ (u − v) + α (u)δ(u − v), 2δ (u − v) + 2U (u)δ(u − v).

The SUSY N=1 KdV system has an inﬁnite number of conservation laws and the ﬁrst nontrivial one gives the SUSY N=1 KdV equation: Ut = −Uuuu + 3(UDu,θ U)u . The integrals of motion are generated by the logarithm of the supertrace of the corresponding monodromy matrix, which has the following form: M

(cl)

2πiphα1

=e

P exp 0

2π

i du √ ξ(u)e−φ(u) eα1 2

i − √ ξ(u)eφ(u) eα0 − e2α1 e−2φ(u) − e2α0 e2φ(u) − [eα1 , eα0 ] . 2 Its quantum generalization can be represented in the quantum P-exponential form (for the explanation of this notion see below and [3] for details): M

(q)

2πiP hα1

=e

(q)

P exp

2π

du(W− (u)eα1 + W+ (u)eα0 ). 0

Vertex operators W± are deﬁned in the following way W± (u) = dθ : e±Φ(u,θ) := ∓ √i2 ξ(u) : e±φ(u) :. The universal R-matrix with the lower Borel subalgebra represented 2π by (q −1 − q)−1 0 duW± (u) is equal to L = e−πiP hα1 M(q) . Due to this fact L satisﬁes the RTT-relation: Rss Ls ⊗ I I ⊗ Ls = (I ⊗ Ls Ls ⊗ I Rss , where s, s mean that the corresponding object is considered in some representation of Cq (2)(2) . Thus the supertraces of the monodromy matrix (“transfer matrices”) ts = strMs commute, providing the quantum integrability. It is very useful to consider the evaluation representations of Cq (2)(2) , ρs (λ), where now the symbol s means integer and half-integer numbers. Denoting ρs (λ)(M) as Ms (λ) we ﬁnd that ts (λ) = strMs (λ) commute: [ts (λ), ts (µ)] = 0. The expansion of log(t 12 (λ)) in λ (the transfer matrix in the fundamental 3-dimensional representation) is believed to give us as coefﬁcients the local IM, the quantum counterparts of the mentioned IM of SUSY N = 1 KdV. 2. The Q-operator Using the super q-oscillator representations of the upper Borel subalgebra of the quantum afﬁne superalgebra Cq (2)(2) we deﬁne the Q± operators (see [4] for details). The transfer-matrices in different evaluation representations can be expressed in such a way: 1

1

1

1

2cos(πP )ts (λ) = Q+ (q s+ 4 λ)Q− (q −s− 4 λ) + Q+ (q −s− 4 λ)Q− (q s+ 4 λ),

INTEGRABILITY OF SCFT AND SUSY N=1 KDV

395

where s runs over integer and half-integer nonnegative numbers. Q± operators satisfy quantum super-Wronskian relation: 1

1

1

1

2cos(πP ) = Q+ (q 4 λ)Q− (q − 4 λ) + Q+ (q − 4 λ)Q− (q 4 λ). One should note, that we use only 4s + 1-dimensional “osp(1|2)-induced” representations (sometimes called atypical) of C(2)(2) . It allows, however, to construct the fusion relations, see below. To construct the relations like Baxter’s ones we introduce additional “quarter”-operators, constructed “by hands” from the Q-operators: 1

1

1

1

2cos(πP )t k (λ) = Q+ (q 4 + 4 λ)Q− (q − 4 − 4 λ) − Q+ (q − 4 − 4 λ)Q− (q 4 + 4 λ) k

k

k

k

4

for odd integer k. The Baxter’s relations are: 1

1

t 14 (λ)Q± (λ) = ±Q± (q 2 λ) ∓ Q± (q − 2 λ), 1

1

1

t 12 (q 4 λ)Q± (λ) = t 14 (q 2 λ)Q± (q − 2 λ) + Q± (qλ). (1)

The fusion relations have the following form very similar to the A1 case: 1

1

tj (q 4 λ)tj (q − 4 λ) = tj+ 14 (λ)tj− 14 (λ) + (−1)4j . But they are only “fusion-like” because the “quarter”-operators do not seem to correspond to any representation of C(2)(2) . The truncation of these relations for different values of q, being the root of unity: q N = ±1, N ∈ Z, N > 0 has the following form: t N (λ) + t N − 1 (λ) = 2cos(πN P ). 2

2

2

In the case when p = l+1 N , where l ≥ 0, l ∈ Z there exists an additional number of truncations: t N − 1 (λ) = 0, t N (λ) = t N − 1 (λ) = (−1)l+1 , 2

4

2

2

2

N 2

t N − 1 −s (λq ) = (−1)4s ts (λ)(−1)l+1 . 2

2

These relations allow us to rewrite the fusion relation system in the Thermodynamic Bethe Ansatz Equations of D2N type. 3. Conclusions In this paper we studied algebraic relations arising from the integrable structure of CFT provided by the SUSY N =1 KdV hierarchy. The construction of the Q-operator as a “transfer”-matrix corresponding to the inﬁnite-dimensional q-oscillator representation could be also applied to the lattice models. The relations like Baxter’s and fusion ones will be also valid in the lattice case because they depend only on the decomposition properties of the representations. In the following we also plan to study the quantization of N > 1 SUSY KdV hierarchies, related with super-W conformal/topological integrable ﬁeld theories.

396

ANTON M. ZEITLIN

Acknowledgements I am very grateful to my supervisor Prof. P. Kulish. It is a pleasure to thank Prof. M. sem*nov-Tian-Shansky and F. Smirnov for useful discussions and Prof. L. Baulieu, B. Pioline and LPTHE, Univ. Paris 6 for support and hospitality. This work was supported by Dynasty Foundation. References 1. 2. 3. 4.

P.P. Kulish, A.M. Zeitlin, Zapiski Nauchn. Seminarov POMI, 291 (2002) 185 (in Russian), English translation in: Journal of Mathematical Sciences, Kluwer Academic Publishers, hepth/0312158. P.P. Kulish, A.M. Zeitlin, Phys. Lett. B 581 (2004) 125, hep-th/0312159; P.P. Kulish, A.M. Zeitlin, Theor. Math. Phys. 142 (2005), in press. P.P. Kulish, A.M. Zeitlin, Phys. Lett. B 597 (2004) 229, hep-th/0407154. P.P. Kulish, A. M. Zeitlin, Nucl. Phys. B, 2005, in press.

Appendix

LIST OF SPEAKERS

Arkani-Hamed Nima, [emailprotected], Dept. of Physics, Harvard University, Cambridge, MA 02138, USA Baulieu Laurent, [emailprotected], L.P.T.H.E., Universit´es Paris VI et Paris VII, Boˆıte 126, 4 Place Jussieu, 75252 Paris Cedex 05, France Bernard Denis, [emailprotected], SPhT-Saclay, CEA/Saclay, 91191 Gif-sur-Yvette Cedex, France Brandenberger Robert, rhb@het. brown. edu, Box 1843, Brown University, Providence, RI 02912-1843, USA De Boer Jan, [emailprotected], Institute for Theoretical Physics Valckenierstraat 65, 1018 XE Amsterdam, The Netherlands Dijkgraaf Robbert, [emailprotected], Institute for Theoretical Physics Valckenierstraat 65, 1018 XE Amsterdam, The Netherlands Elitzur Shmuel, [emailprotected], Racah Institute of Physics, Hebrew University, Jerusalem, 91904 Israel Morozov Alexei, [emailprotected], ITEP, Bol. Cheremushkinskaya, 25 117218, Moscow, Russia Okun Lev, [emailprotected], ITEP, Bol. Cheremushkinskaya, 25 117218, Moscow, Russia Ooguri Hirosi, [emailprotected], Caltech, Pasadena, CA 91125, USA Pioline Boris, [emailprotected], L.P.T.H.E., Universit´es Paris VI et Paris VII, Boˆıte 126, 4 Place Jussieu, 75252 Paris Cedex 05, France Pokorski Stefan, [emailprotected], IFT UW, Hoza 69, PL-00-681 Warszawa, Poland Polchinski Joseph, [emailprotected], KITP, University of California, Santa Barbara, CA 93106, USA Rabinovici Eliezer, [emailprotected], Racah Institute of Physics, Hebrew University, Jerusalem, 91904 Israel Tseytlin Arkady, [emailprotected], Physics Department, The Ohio State University, Columbus OH43210-1106, USA

399

LIST OF PARTICIPANTS

Alkalaev Konstantin, [emailprotected], P.N.Lebedev Physical Institute, 53 Leninsky prospect, Moscow, 119991 Russia Angelantonj Carlo, [emailprotected], Humboldt Univ., Unter den Linden 6, 10099 Berlin, Germany Anisimov Alexey, [emailprotected], Max-Planck-Institute, F¨ohringer Ring 6 80805 M¨unchen, Germany Arcioni Giovanni, [emailprotected], Racah Institute of Physics, Hebrew University, Jerusalem, 91904 Israel Arvidsson Par, [emailprotected], Institute of Theoretical Physics, Chalmers Univerity of Technology S-412 96 Gothenburg, Sweden Ashok Sujay, [emailprotected], Department of Physics k Astronomy, Rutgers University 136 Frelinghuysen Road, Piscataway, New Jersey 08854, USA Banks Thomas, [emailprotected], University of California, Santa Cruz, 1156 High Street Santa Cruz, CA 95064, USA / Department of Physics & Astronomy, Rutgers University 136 Frelinghuysen Road, Piscataway, New Jersey 08854, USA Belov Dmitriy, [emailprotected], Department of Physics & Astronomy, Rutgers University 136 Frelinghuysen Road, Piscataway, New Jersey 08854, USA Benakli Karim, [emailprotected], L.P.T.H.E., Universit´es Paris VI et Paris VII, Boˆıte 126, 4 Place Jussieu, 75252 Paris Cedex 05, France Boels Rutger, [emailprotected], University of Amsterdam, Postbus 19268, 1000 GG Amsterdam, The Netherlands Calcagni Gianluca, [emailprotected], Parma University, Via Universit, 12 - 1-43100 Parma, Italy Ceresole Anna, [emailprotected], Department of Theoretical Physics, University of Turin Via P. Giuria 1, 10125 Torino, Italy and INFN Dell’aquila Eleonora, [emailprotected], Department of Physics and Astronomy, Rutgers University, 136 Frelinghuysen Road, NJ 08854-8019 Piscataway, USA

401 L. Baulied et al. (eds.), String Theory: From Gauge Interactions to Cosmology, 401–404. © 2005 Springer. Printed in the Netherlands.

402

ANTON M. ZEITLIN

Durin Bruno, [emailprotected], LPTHE, 4 place Jussieu, tour 24, 5eme e´ tage, boˆıte 126, 75252 Paris cedex 05, France Escoda Cristina, [emailprotected], DAMTP, Centre for Mathematical Sciences, Wilberforce Road, CB3 0WA Cambridge, United Kingdom Evslin Jarah, [emailprotected], INFN Sezione di Pisa, via Buonarroti, 2, 56127 Pisa, Italy Falkowski Adam, [emailprotected], Institute of Theoretical Physics, Warsaw University, ul. Hoza 69, 00-681 Warsaw, Poland Flink Erik, [emailprotected], Department of Theoretical Physics, Chalmers University of Technology, SE-41296 G¨oteborg, Sweden Flournoy Alex, [emailprotected], Department of Physics, Technion, 32000 Haifa, Israel Franco Sebastian, [emailprotected], Center for Theoretical Physics, Massachusetts Institute of Technology, 77 Massachusetts Avenue, MA 02139 Cambridge, USA Giryavets Alexander, [emailprotected], Department of Physics, Stanford University, Stanford CA 94305, USA Grana Mariana, [emailprotected], LPTENS, D´epartement de Physique de l’ENS, 24, Rue Lhom*ond, 75005 Paris, France Gripaios Ben, [emailprotected], Theoretical Physics, University of Oxford, 1 Keble Road, OX1 3NP Oxford, United Kingdom Grunberg Daniel, [emailprotected], KdV Institute, Plantage Muidergr 24, 1018 TV Amsterdam, The Netherlands Gursoy Umut, [emailprotected], Massachusetts Institute of Technology, 77 Massachussetts Ave, Cambridge MA 02139, USA Halmagyi Nick, [emailprotected], Department of Physics and Astronomy, University of Southern California, Los Angeles CA 90089, USA Hartnoll Sean, [emailprotected], DAMTP, Centre for Mathematical Sciences, Wilberforce road, CB3 0WA Cambridge, United Kingdom Heise Rainer, [emailprotected], Institute for Theoretical Physics, Chalmers University of Technology and Goteborg University, 41296 G¨oteborg, Sweden Herbst Manfred, [emailprotected], CERN, Theory Division, 1211 Geneva 23, Switzerland Hubeny Veronika, [emailprotected], Department of Physics,382 Via Pueblo Mall,Stanford CA 94305, USA

INTEGRABILITY OF SCFT AND SUSY N=1 KDV

403

Jarv Laur, [emailprotected], Friedrich-Schiller-Universit¨at Jena, D - 07740 Jena, Germany Kirsch Ingo, [emailprotected], Humboldt-Univ. Berlin, Unter den Linden 6, 10099 Berlin, Germany Kleban Matthew, [emailprotected], Institute for Advanced Study, Einstein Drive, Princeton, NJ 08540, USA Koerber Paul, [emailprotected], Vrije Univ. Brussel, Plainlaan 2, 1050 Brussel, Belgium Kohlprath Emmanuel, [emailprotected], ´ CPHT, Ecole Polytechnique, F91128 Palaiseau Cedex, France Kuperstein Stanislav, [emailprotected], Tel Aviv University, P.O. Box 39040, Tel Aviv 69978, Israel Lee Sangmin, [emailprotected], CERN, Theory Division, 1211 Geneva 23, Switzerland Lindstrom Ulf, [emailprotected], Department of Theoretical Physics, Box 803, SE-751 08 Uppsala, Sweden Maillard Tristan, [emailprotected], Eidgenssische Technische Hochschule Z¨urich, ETH Zentrum, R¨amistrasse 101, CH-8092 Z¨urich, Switzerland Maoz Liat, [emailprotected], Institute for Theoretical Physics, Amsterdam Valckenierstraat 65, 1018XE Amsterdam, the Netherlands Naqvi Asad, [emailprotected], Institute for Theoretical Physics, Amsterdam Valckenierstraat 65, 1018XE Amsterdam, the Netherlands Olsson Martin, [emailprotected], Department of Theoretical Physics, Box 803, SE-751 08 Uppsala, Sweden Orlando Domenico, [emailprotected], ´ CPHT, Ecole Polytechnique, F91128 Palaiseau Cedex, France Ozer Aybike, [emailprotected], Trinity College Dublin, College Green, Dublin 2, Ireland Pakman Ari, [emailprotected], Racah Institute of Physics, Hebrew University, Jerusalem, 91904 Israel Papadodimas Kyriakos, [emailprotected], Faculty of Arts and Sciences, Harvard University, Cambridge, MA 02138, USA Park Jongwon, [emailprotected], Particle Theory Group, Enrico Fermi Institute, University of Chicago, 5640 South Ellis Avenue, Chicago, Illinois 60637, USA Patir Assaf, [emailprotected], Department of Particle Physics, Weizmann Institute of Science, Rehovot, Israel Paulot Louis, [emailprotected], Universit´e Libre de Bruxelles, Campus Plaine, boulevard du Triomphe, Physique Th´eorique et Math´ematique C.P. 231 1050 Bruxelles, Belgium

404

ANTON M. ZEITLIN

Prezas Nikolaos, [emailprotected], Humboldt Univ., Unter den Linden 6, 10099 Berlin, Germany Rangamani Mukund, [emailprotected], Physics Dept University of California, Berkeley, CA 94720 and Physics Dept., Harvard University, 17 Oxford St. Cambridge, MA 02138, USA Roest Diederik, [emailprotected], Institute for Theoretical Physics, University of Gr¨oningen, Nijenborgh 4 9747 AG, Gr¨oningen, The Netherlands Rotaev Mikhail, [emailprotected], Institute of Theoretical and Experimental Physics (ITEP), B. Cheremushkinskaya, 25, Moscow, 117259, Russia Rulik Ksenya, [emailprotected], Department of Theoretical Physics, University of Turin Via P. Giuria 1, 10125 Torino, Italy Rychkov Vyacheslav, [emailprotected], University of Amsterdam, Postbus 19268, 1000 GG Amsterdam, The Netherlands Ryzhov Anton, [emailprotected], Brandeis University, 415 South St. Waltham, MA 02454-9110, USA Sever Amit, [emailprotected], Racah Institute of physics, Hebrew University, Jerusalem, 91904 Israel Shaynkman Oleg, [emailprotected], P.N. Lebedev Physical Institute, 53 Leninsky prospect, Moscow, 119991 Russia Sulkowski Piotr, [emailprotected], Institute of Theoretical Physics, Warsaw University, ul. Hoa 69, PL-00-681 Warsaw, Poland Tanzini Alessandro, [emailprotected], L.P.T.H.E., Universit´es Paris VI et Paris VII, Boˆıte 126, 4 Place Jussieu, 75252 Paris Cedex 05, France Vikman Alexander, [emailprotected], Ludwig-Maximilians-Universitt, Theresienstraﬂe 37, 80333 M¨unchen, Germany Watson Scott, [emailprotected], Physics Department, Brown University, Providence RI 02912, USA Williams Brook, [emailprotected], University of California, Santa Barbara, CA 93106, USA Zabzine Maxim, [emailprotected], L.P.T.H.E., Universit´es Paris VI et Paris VII, Boˆıte 126, 4 Place Jussieu, 75252 Paris Cedex 05, France Zeitlin Anton, [emailprotected], Institute of Problems of Mechanical Engineering, Russian Academy of Sciences, V.O., Bolshoj pr., 61 St. Petersburg, 199178 Russia