a NATO Advanced Study Institute

LES HOUCHES

SESSION LXXVI 30 July – 31 August 2001 Unity from duality: Gravity, gauge theory and strings L’unité de la physique fondamentale : gravité, théorie de jauge et cordes

Edited by C. BACHAS, A. BILAL, M. DOUGLAS, N. NEKRASOV and F. DAVID

Springer

Les Ulis, Paris, Cambridge Berlin, Heidelberg, New York, Hong Kong, London, Milan, Paris, Tokyo

Published in cooperation with the NATO Scientific Affair Division

ISSN 0924-8099 print edition ISSN 1610-3459 online edition ISBN 3-540-00276-6 ISBN 2-86883-625-9 Springer-Verlag Berlin Heidelberg New York EDP Sciences Les Ulis

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms or in other ways, and storage in data banks. Duplication of this publication or parts thereof is only permitted under the provisions of the French and German Copyright laws of March 11, 1957 and September 9, 1965, respectively. Violations fall under the prosecution act of the French and German Copyright Laws. © EDP Sciences; Springer-Verlag 2002 Printed in France

Preface

The 76th session of the Les Houches Summer School in Theoretical Physics was devoted to recent developments in string theory, gauge theories and quantum gravity. As frequently stated, Superstring Theory is the leading candidate for a uniﬁed theory of all fundamental physical forces and elementary particles. This claim, and the wish to reconcile general relativity and quantum mechanics, have provided the main impetus for the development of the theory over the past two decades. More recently the discovery of dualities, and of important new tools such as D-branes, has greatly reinforced this point of view. On the one hand there is now good reason to believe that the underlying theory is unique. On the other hand, we have for the ﬁrst time working (though unrealistic) microscopic models of black hole mechanics. Furthermore, these recent developments have lead to new ideas about compactiﬁcation and the emergence of low-energy physics. While pursuing the goal of uniﬁcation we have also witnessed a dramatic return to the “historic origins” of string theory as a dual model for meson physics. Indeed, the study of stringy black branes has uncovered a surprising relation between string theory and large-N gauge dynamics. This was cristallized in the AdS/CFT correspondence, which has revived the old hope for a string description of the strong interaction. The AdS/CFT correspondence is moreover a prime illustration of the central role of string theory in modern theoretical physics. Much like quantum ﬁeld theory in the past, it provides a fertile springboard for new tools, concepts and insights, which should have ramiﬁcations in wider areas of physics and mathematics. The main lectures of the Les Houches school covered most of the recent developments, in a distilled and pedagogical fashion. Students were expected to have a good knowledge of quantum ﬁeld theory, and of basic string theory at the level, for instance, of the ﬁrst ten chapters of Green, Schwarz and Witten. The emphasis was on acquiring a working knowledge of advanced string theory in its present form, and on critically assessing open problems and future directions. The lectures by Bernard de Wit were a comprehensive introduction to supergravities in diﬀerent dimensions and with various numbers of supersymmetries. Topics covered include the allowed low-energy couplings, duality symmetries, compactiﬁcations and supersymmetry in curved backgrounds.

xxii

Part of this is older material not easily accessible in the literature, and presented here from a modern perspective. Eliezer Rabinovici lectured on supersymmetric gauge theories, reviewing earlier and more recent results for N = 1, 2 and 4 supersymmetries in four dimensions. These results include the structure of the eﬀective lagrangians, non-renormalization theorems, dualities, the celebrated Seiberg-Witten solution and brane engineering of eﬀective gauge theories. M-theory and string dualities were introduced in the lectures by Ashoke Sen. He reviewed the conjectured relations between the ﬁve perturbative string theories, the maximal N = 1 supergravity in eleven dimensions and their compactiﬁcations. He summarized our present-day knowledge of the still elusive fundamental or “M theory”, from which the above theories derive as special limits. More recent topics include non-BPS branes, where duality is of limited (but not zero) use. Philip Candelas gave a pedagogical introduction to the important subject of Calabi Yau compactiﬁcations. He ﬁrst reviewed the older material, and then discussed more recent aspects, including second quantized mirror symmetry, conifold transitions and some intriguing relations to number theory. Unfortunately a written version of his lectures could not be included in this volume. The holographic gauge/string theory correspondence was the subject of the lectures by Juan Maldacena and by Igor Klebanov. Maldacena introduced the conjectured equivalence between string theory in the near-horizon geometries of various black branes and gauge theories in the large Ncolor limit. He focused on the celebrated example of N = 4 four-dimensional super Yang Mills dual to string theory in AdS5 × S5 , and gave a critical review of the existing evidence for this correspondence. He also discussed analogous conjectures in other spacetime dimensions, in particular those relevant to the study of stringy black holes, and of the still elusive little string theory. Igor Klebanov then concentrated on this duality in the phenomenologically more interesting contexts of certain N = 1 and 2 supersymmetric gauge theories in four dimensions. He reviewed the relevant geometries on the supergravity side, which include non-trivial ﬂuxes and fractional branes, and discussed the gravity duals of renormalization group ﬂow, conﬁnement and chiral symmetry breaking. These results have revived and made sharper the old ideas about the “master ﬁeld” of large N gauge theory. The lectures of Michael Green dealt with some ﬁner aspects of string dualities and of the gauge theory/string theory correspondence. He discussed higher derivative couplings in eﬀective supergravity actions, focusing in particular on the contributions of instantons both in string theory and on the

xxiii

Yang Mills side. His lectures also included some introductory material on Dbranes. Unfortunately a written version of his lectures could not be included in this volume. Andrew Strominger gave a detailed introduction to quantum gravity in a de Sitter spacetime. He discussed in particular whether ideas of holography, that have worked well in anti de Sitter, could also be applied in this case. This was one of the more speculative subjects in the school, but a fascinating one not the least because astrophysical observations seem to indicate that we actually live in an accelerating universe. Finally Michael Douglas gave three lectures on D-brane geometry, and in particular on the problem of classifying all N = 1 string-theory vacua, while Alexander Gorsky discussed N = 1 and N = 2 supersymmetric gauge theories and their relation to integrable models. Nikita Nekrasov lectured on open strings and non-commutative gauge theories. Some more advanced and/or topical subjects were covered in the accompanying series of seminars. Seminar speakers included Laurent Baulieu, Mirjam Cvetiˇ, Frank Ferrari, Dan Freedman, Bernard Julia, Peter Mayr, c Soo-Jong Rey, Augusto Sagnotti, Samson Shatashvili, and one of the organizers (C.B.). There was also a lively weekly student seminar and discussion sessions, which contributed greatly to the lively and stimulating atmosphere of the school. Some of the seminar speakers have kindly accepted to contribute to the present volume. In the year that has elapsed since the end of the school there have been further developments in the subject. The pp-waves, which arise as Penrose limits of near-horizon geometries, oﬀer for instance a new line of attack on the important problem of solving string theory in Ramond-Ramond backgrounds. Such developments and others will no doubt make, one day, the present volume obsolete. This is of course no reason for regret – to the contrary we hope that this may happen sooner rather than later, and that the participants of this school will help shape the (non-recognisable?) future form of M theory. Acknowledgements This Les Houches school and the present proceedings were made possible due to the contributions of many individuals and several diﬀerent funding sources. Funding sources included the NATO Scientiﬁc Aﬀairs Division through its Advanced Study Institute program, the European Union through its High-Level Scientiﬁc Conferences program, the Commissariat ` l’Energie a ´ Atomique, the Universit´ Joseph Fourier de Grenoble, and two European e

xxiv

IHP Networks: “Superstring Theory” (HPRN-CT-2000-00122) and “The Quantum Structure of Spacetime” (HPRN-CT-2000-00131). Among the many people contributing to the success of the school, we should mention - the board of the School and in particular Fran¸ois David, who has c worked tirelessly at all diﬀerent stages (funding applications, admissions, running the session, preparation of proceedings) exceeding often the organizers in zeal and energy; - the secretaries Mmes G. D’Henry, I. Leli`vre and B. Rousset (and the e other personnel of the school), who helped solve administrative and everyday problems; and last but not least - the lecturers, for their eﬀorts in presenting hard material in a clear and pedagogical fashion, and also for writing up their lecture notes.

Costas Bachas Adel Bilal Michael Douglas Nikita Nekrasov Fran¸ois David c

CONTENTS

Lecturers Participants Pr´face e Preface Contents

xi xiii xvii xxi xxv

**Lecture 1. Supergravity by B. de Wit
**

1 Introduction 2 Supersymmetry in various dimensions 2.1 The Poincar´ supersymmetry algebra . . . . e 2.2 Massless supermultiplets . . . . . . . . . . . 2.2.1 D = 11 supermultipets . . . . . . . . 2.2.2 D = 10 supermultiplets . . . . . . . 2.2.3 D = 6 supermultiplets . . . . . . . . 2.3 Massive supermultiplets . . . . . . . . . . . 2.4 Central charges and multiplet shortening . . 2.5 On spinors and the R-symmetry group HR

1

3 4 5 7 8 9 13 17 19 22 28 28 32 35 38 43 46 53 55 56 61 66 69

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

3 Supergravity 3.1 Simple supergravity . . . . . . . . . . . . . . . . . . 3.2 Maximal supersymmetry and supergravity . . . . . . 3.3 D = 11 supergravity . . . . . . . . . . . . . . . . . . 3.4 Dimensional reduction and hidden symmetries . . . 3.5 Frames and ﬁeld redeﬁnitions . . . . . . . . . . . . . 3.6 Kaluza–Klein states and BPS-extended supergravity 3.7 Nonmaximal supersymmetry: Q = 16 . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

4 Homogeneous spaces and nonlinear sigma models 4.1 Nonlinearly realized symmetries . . . . . . . . . . . . . . 4.2 Geometrical quantities . . . . . . . . . . . . . . . . . . . 4.3 Nonlinear sigma models with homogeneous target space 4.4 Gauged nonlinear sigma models . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

xxvi

5 Gauged maximal supergravity in 4 and 5 dimensions 5.1 On E7(7) /SU (8) and E6(6) /U Sp(8) cosets . . . . . . . . 5.2 On ungauged maximal supergravity Lagrangians . . . . 5.3 Electric–magnetic duality and E7(7) . . . . . . . . . . . 5.4 Gauging maximal supergravity; the T -tensor . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

70 72 74 79 84

6 Supersymmetry in anti-de Sitter space 89 6.1 Anti-de Sitter supersymmetry and masslike terms . . . . . . . . . . 93 6.2 Unitary representations of the anti-de Sitter algebra . . . . . . . . 97 6.3 The superalgebras OSp(N |4) . . . . . . . . . . . . . . . . . . . . . 110 7 Superconformal symmetry 116 7.1 The superconformal algebra . . . . . . . . . . . . . . . . . . . . . . 119 7.2 Superconformal gauge theory and supergravity . . . . . . . . . . . 122 7.3 Matter ﬁelds and currents . . . . . . . . . . . . . . . . . . . . . . . 126

**Lecture 2. Supersymmetric Gauge Theories by D.S. Berman and E. Rabinovici
**

1 Introduction 2 Supersymmetric quantum mechanics 2.1 Symmetry and symmetry breaking . . . . . . 2.2 A nonrenormalisation theorem . . . . . . . . 2.3 A two variable realization and ﬂat potentials 2.4 Geometric meaning of the Witten index . . . 2.5 Landau levels and SUSY QM . . . . . . . . . 2.6 Conformal quantum mechanics . . . . . . . . 2.7 Superconformal quantum mechanics . . . . . 3 Review of supersymmetric models 3.1 Kinematics . . . . . . . . . . . . . 3.2 Superspace and chiral ﬁelds . . . . 3.3 K¨hler potentials . . . . . . . . . . a 3.4 F-terms . . . . . . . . . . . . . . . 3.5 Global symmetries . . . . . . . . . 3.6 The eﬀective potential . . . . . . . 3.7 Supersymmetry breaking . . . . . . 3.8 Supersymmetric gauge theories . . 4 Phases of gauge theories

137

141 143 151 152 154 158 159 161 164 165 165 167 169 170 170 172 172 173 179

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

xxvii

5 Supersymmetric gauge theories/super QCD 5.1 The classical moduli space . . . . . . . . . . . 5.2 Quantum moduli spaces . . . . . . . . . . . . 5.3 Quantum moduli space for 0 < NF < NC . . 5.4 Integrating in . . . . . . . . . . . . . . . . . . 5.5 Quantum moduli space for NF ≥ NC . . . . . 5.6 NF = NC . . . . . . . . . . . . . . . . . . . . 5.7 NF = NC + 1 . . . . . . . . . . . . . . . . . . 5.8 Higgs and conﬁnement phases . . . . . . . . . 5.9 Infra-red duality . . . . . . . . . . . . . . . . 5.10 Superconformal invariance in d = 4 . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

180 182 185 185 186 190 190 191 192 194 202 207 210 210 211 213 216 226 229 233 235 237

6 Comments on vacuum energies in scale invariant theories 7 Supersymmetric gauge theories and string theory 7.1 Branes in string theory . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Branes in IIA and IIB string theories . . . . . . . . . . . . . . . . 7.3 The eﬀective ﬁeld theory on branes . . . . . . . . . . . . . . . . . 7.4 Eﬀective D = 4 dimensional systems with N = 2 supersymmetry 7.5 An eﬀective D = 4, N = 1, U (NC ) gauge theory with matter . . 7.6 More pieces of information . . . . . . . . . . . . . . . . . . . . . . 7.7 Obtaining the dual ﬁeld theory . . . . . . . . . . . . . . . . . . . 7.8 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . 8 Final remarks

. . . . . . . .

**Lecture 3. An Introduction to Duality Symmetries in String Theory by A. Sen
**

1 Introduction 2 A brief review of perturbative string 2.1 The spectrum . . . . . . . . . . . . . 2.2 Interactions . . . . . . . . . . . . . . 2.3 Compactiﬁcation . . . . . . . . . . . theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

241

243 245 246 251 254

3 Notion of duality symmetries in string theory 255 3.1 Duality symmetries: Deﬁnition and examples . . . . . . . . . . . . 255 3.2 Testing duality conjectures . . . . . . . . . . . . . . . . . . . . . . 258

xxviii

4 Analysis of low energy eﬀective ﬁeld theory 4.1 Type I–SO(32) heterotic duality in D = 10 . 4.2 Self-duality of heterotic string theory on T 6 . 4.3 Duality between heterotic on T 4 and type IIA 4.4 SL(2, Z) self-duality of type IIB in D = 10 . 4.5 Other examples . . . . . . . . . . . . . . . . .

. . . . on . . . .

. . . . K3 . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

260 261 264 268 272 274

5 Precision test of duality: Spectrum of BPS states 5.1 SL(2, Z) S-duality in heterotic on T 6 and multi-monopole spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 SL(2, Z) duality in type IIB on S 1 and D-branes . . . . . 5.3 Massless solitons and tensionless strings . . . . . . . . . . 6 Interrelation between diﬀerent duality conjectures 6.1 Combining non-perturbative and T -dualities . . . . . 6.2 Duality of dualities . . . . . . . . . . . . . . . . . . . 6.3 Fiberwise duality transformation . . . . . . . . . . . 6.4 Recovering higher dimensional dualities from lower dimensional ones . . . . . . . . . . . . . .

276 moduli . . . . . 280 . . . . . 286 . . . . . 296

299 . . . . . . . . 299 . . . . . . . . 299 . . . . . . . . 301 . . . . . . . . 304

7 Duality in theories with less than sixteen supersymmetry generators 305 7.1 Construction of a dual pair of theories with eight supercharges . . 306 7.2 Test of duality conjectures involving theories with eight supercharges . . . . . . . . . . . . . . . . . . . . . . . . 309 8 M-theory 312 8.1 M-theory in eleven dimensions . . . . . . . . . . . . . . . . . . . . 312 8.2 Compactiﬁcation of M-theory . . . . . . . . . . . . . . . . . . . . . 315

**Lecture 4. Les Houches Lectures on Large N Field Theories and Gravity by J. Maldacena
**

1 General introduction

323

325

2 The correspondence 330 2.1 The ﬁeld ↔ operator correspondence . . . . . . . . . . . . . . . . . 336 2.2 Holography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338 3 Tests of the /CFT correspondence 3.1 The spectrum of chiral primary operators . . . . . . 3.1.1 The ﬁeld theory spectrum . . . . . . . . . . . 3.1.2 The string theory spectrum and the matching 3.2 Matching of correlation functions and anomalies . .

Ë

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

341 342 342 348 351

xxix

4 Correlation functions 353 4.1 Two-point functions . . . . . . . . . . . . . . . . . . . . . . . . . . 355 4.2 Three-point functions . . . . . . . . . . . . . . . . . . . . . . . . . 359 5 Wilson loops 362 5.1 Wilson loops and minimum surfaces . . . . . . . . . . . . . . . . . 363 5.2 Other branes ending on the boundary . . . . . . . . . . . . . . . . 368 6 Theories at ﬁnite temperature 369 6.1 Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369 6.2 Thermal phase transition . . . . . . . . . . . . . . . . . . . . . . . 372

**Lecture 5. D-Branes on the Conifold and N = 1 Gauge/Gravity Dualities by I.R. Klebanov, C.P. Herzog and P. Ouyang
**

1 Introduction

383

385

2 D3-branes on the conifold 388 2.1 Dimensions of chiral operators . . . . . . . . . . . . . . . . . . . . 391 2.2 Wrapped D3-branes as “dibaryons” . . . . . . . . . . . . . . . . . . 393 2.3 Other ways of wrapping D-branes over cycles of T 1,1 . . . . . . . . 394 3 The RG cascade 397 3.1 Matching of the β-functions . . . . . . . . . . . . . . . . . . . . . . 400 4 The chiral anomaly 402 4.1 The anomaly as a classical eﬀect in supergravity . . . . . . . . . . 403 4.2 The anomaly as spontaneous symmetry breaking in AdS5 . . . . . 405 5 Deformation of (KS) the conifold 410 5.1 The ﬁrst-order equations and their solution . . . . . . . . . . . . . 412 5.2 SO(4) invariant expressions for the 3-forms . . . . . . . . . . . . . 413 6 Infrared physics 6.1 Dimensional transmutation and conﬁnement . . . . . . . . . . . . . 6.2 Tensions of the q-strings . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Chiral symmetry breaking and gluino condensation . . . . . . . . . 414 414 416 418

**Lecture 6. De Sitter Space by M. Spradlin, A. Strominger and A. Volovich
**

1 Introduction

423

425

xxx

2 Classical geometry of de Sitter space 427 2.1 Coordinate systems and Penrose diagrams . . . . . . . . . . . . . . 428 2.2 Schwarzschild-de Sitter . . . . . . . . . . . . . . . . . . . . . . . . . 435 2.3 Geodesics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436 3 Quantum ﬁeld theory on de Sitter space 437 3.1 Green functions and vacua . . . . . . . . . . . . . . . . . . . . . . . 437 3.2 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440 3.3 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443 4 Quantum gravity in de Sitter space 446 4.1 Asymptotic symmetries . . . . . . . . . . . . . . . . . . . . . . . . 446 4.2 De Sitter boundary conditions and the conformal group . . . . . . 447 A Calculation of the Brown-York stress tensor 451

**Lecture 7. String Compactiﬁcation with N = 1 Supersymmetry by M.R. Douglas
**

1 Introduction and the question of vacuum selection 2 From ten dimensional geometry to four dimensional eﬀective ﬁeld theory 3 D-branes with stringy corrections 4 Quantum corrections 5 Towards the low energy theory of everything

455

457

461 466 470 472

**Lecture 8. Lectures on Open Strings, and Noncommutative Gauge Theories by N.A. Nekrasov
**

1 Introduction 2 Background independence 2.1 Dolan-Nappi solutions . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Intersecting branes . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 T -duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

477

479 480 483 485 485

xxxi

3 BPS algebras 486 3.1 Noncommutative U (1) instantons . . . . . . . . . . . . . . . . . . . 486 3.2 Higher dimensional instantons . . . . . . . . . . . . . . . . . . . . . 489 4 Fermions in the Y background 490 4.1 Fermions in the instanton background . . . . . . . . . . . . . . . . 490 4.2 The Dirac ﬁeld in the monopole background . . . . . . . . . . . . . 491 5 Non-trivial backgrounds 492 5.1 Example: Γ = Z2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493 6 Conclusions and outlook 494

**Lecture 9. Condensates Near the Argyres-Douglas Point in SU(2) Gauge Theory with Broken N = 2 Supersymmetry by A. Gorsky
**

1 Introduction 2 Matter and gaugino condensates

497

499 501

3 Dyon condensates 504 3.1 Monopole condensate . . . . . . . . . . . . . . . . . . . . . . . . . . 505 3.2 Charge and dyon condensates . . . . . . . . . . . . . . . . . . . . . 506 4 The Argyres-Douglas point: How well the theory is deﬁned 5 Conclusions 508 509

**Seminar 1. Quantum Field Theory with Extra Dimensions by L. Baulieu
**

1 Introduction 2 The ﬁelds 3 The action at level

513

515 516

Ò

518

. . . . . . . . . . . . . . .2 The new Spin(7) holonomy metric . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . .4 Non-critical strings . . . . . . . . Ferrari 1 Introduction 2 Many paths to the gauge/string 2. . . . . . . . . . . 3. . . . . . . . . Four Dimensional Non-Critical Strings by F. . . . . . . . .3. . . . . . . . . . .1.2 Old G2 holonomy metrics and their harmonic forms .3 D-branes . . . . . 2. . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . 3. . . 3. . . .1 Resolved cones over S 2 × S 4 and S 2 × CP2 . 547 549 550 550 551 552 554 559 562 564 566 569 571 . . . . . . . .2 Large N . . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . Special Holonomy Spaces and M-Theory by M. . . . . . . .2 Instantons and large N . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . .1 Motivation . . . . 523 525 527 527 528 529 529 530 531 531 532 533 533 534 534 534 535 536 Ú . . . L¨ and C. duality . 2. . . . . . . . . . . . . . . . . . . . . 3. 2. . . . Cvetiˇ. . . . . . . . . .1 Harmonic forms for the Stenzel metric . . . . harmonic forms and resolved branes 3. . . Gibbons. . .3. . . . . .3 Resolved D2-brane . . .1 Stenzel metric . . . .5 Further insights . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . 2. . . . . q) forms . . . . .W. . . . 3. . . . . . . . . .N. . . . . . . . . . 537 541 Seminar 3. G. . . .1 Conﬁnement .1. . . 3 Special holonomy spaces. . . . . . .4 Other examples . . .1 The old metric and harmonic 4-forms .2 Resolved cone over S 3 × S 3 . . . . .1 Four dimensional CFTs as Kazakov critical points 3. . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Applications: Resolved M2-branes and D2-branes . . . . 4 New G2 holonomy metrics 4. .1 Classiﬁcation of G2 holonomy spaces with S 3 × S 3 orbits 5 Conclusions and open avenues 537 . . . . . . . . . . 3. . . . . . . . . .2 Resolved M2-brane . 3 Four dimensional non-critical strings 3. . . . . . . . . . . . . . . . . .2. . . . . . . .3 New Spin(7) holonomy metrics and their harmonic forms . . . . . . .xxxii Seminar 2. . . . . . . . . . . . . . . . 3. . . . . . .2 Harmonic middle-dimension (p. . . . 3. . H. . . . . Pope c u 1 Introduction transgression 2 Resolution 2. . . . . . 3. .3 A toy model example . . . . . . . . . . . . .4 Exact results in 4D string theory . .

. . .5. . . . . . . . . . . . . . 604 5. . . 607 Seminar 6. . . . . 593 3. 605 5. . . . . . . . . . . . . . Open Wilson Lines and UV-IR Duality by Soo-Jong Rey 1 Introduction and conjectures 2 Flying noncommutative dipole 587 589 591 3 Open wilson lines: How and why? 593 3. . . . . . . . . 571 Non-perturbative non-Borel summable partition functions . . . . .2 Full proofs . . . . . . . . . . . . . . . . . . . . . . . .2 Witten’s w -product is Moyal’s m -product . U-Opportunities: Why is Ten Equal to Ten? by B. Julia 1 Classiﬁcations 2 Real forms of Lie algebras 3 Chaos controlled by symmetry 575 577 581 583 Seminar 5. . . . . . .1 3. . . . . . . . . . .2 Interacting OWLs .1 Open strings as miniature dipoles . . . . . . . . .1 Open Wilson lines . . 572 572 4 Open problems Seminar 4. . . . .3 Closed strings as OWLs . . . . . . . . . . . 595 4 Free and interacting OWLs 597 4. . Exact Answers to Approximate Questions – Noncommutative Dipoles. . . . . . . . 600 5 Closed strings out of open strings 604 5. . . . . . .2 Generalized star products . . .xxxiii 3. . . . . . . . . . .1 Free OWLs . . . . . . . . .5. . . . . . . . . Open-String Models with Broken Supersymmetry by A. . . . . Sagnotti 1 Broken supersymmetry and type-0 models 611 613 . . . . . . . . . . . . . 598 4. . . . . . . . . . . . . . . . .

L. Exceptional Magic by S. Tachyon Condensation and Closed Strings by S.xxxiv 2 Scherk-Schwarz deformations and brane supersymmetry 3 Brane supersymmetry breaking 618 620 Seminar 7. Shatashvili 1 G2 2 627 643 648 657 661 ËÔ Ò(7) 3 Topological twist .L. On a Field Theory of Open Strings. Shatashvili Seminar 8.

U. Piscataway. India of Laboratory. N. University. Jefferson Physical Cambridge. Israel A. NJ 08544. MA 02138. Switzerland M.K. Chhatnag Road. Racah Institute 91904 Jerusalem. . NJ 08854-8019. GREEN. 1 rue Breguet. Rutgers.ORGANIZERS C. NEKRASOV. The State University of New Jersey. BILAL. Laboratoire de Physique Théorique. RABINOVICI. Princeton.S. DOUGLAS. Department of Physics & Astronomy. Princeton University.H.A.S.A. Physics. SEN. 91440 Bures-sur-Yvette. DAMPT. Cambridge CB3 OWA. Université de Neuchâtel. 24 rue Lhomond. 91191 Gif-sur-Yvette. CEA Saclay. U. ENS. U. Jhusi.E. J. Joseph Henry Laboratories. Institute. Harish-Chandra Research 211019 Allahabad. Oxford University. 75231 Paris. U. France A. DAVID. SPhT. 35 route de Chartres. Giles.S. 24-29 St.K. France F. Wilberforce Road. M. E. Mathematical Institute. Harvard Hebrew University. MALDACENA. BACHAS. France LECTURERS P. KLEBANOV.A. Institut de Physique.S. I. I. CANDELAS. U. Oxford OX1 3LB . 2000 Neuchâtel.

JULIA. A. BAULIEU. ITEP. CERN Theory Division. U.A. 117259 Moscow. 75231 Paris Cedex 05. Corea Physics. Italy S. 75231 Paris. REY. PA 19104. GORSKY. 1211 Genova 23. B. Harvard University. U. Pennsylvania. Department of Physics and Astronomy. 24 rue Lhomond.S. University D. Institute for Theoretical Physics. The Netherlands SEMINAR SPEAKERS L.A. . Cheremushkinskaya 25.S. France Curie. Seoul National University. MA 02138.S. SHATASHVILI. FREEDMAN. 00133 Roma. LPTHE. Department of Physics. MAYR. Yale University. Tor Vergata. U.A. Laboratory. CVETIC. Via della Ricerca Scientifica 1. U. Spinoza Institute. A. Department of Physics. Philadelphia.A.S. CT 06520. France P. Switzerland S. Université Pierre et Marie 4 place Jussieu. MA 02139. Tour 16. of M.xii A. LPT/ENS. 35P4 CE Utrecht. New Haven. Center for Theoretical 151-747 Seoul. DE WIT. MIT. Center for Theoretical Physics. SAGNOTTI. STROMINGER. University Roma II. Jefferson Physical Cambridge. B. Utrecht University. Russia B. Cambridge.

ABANIN. BERMAN. Tomsk State University. Santa Barbara. Russia A. GSP-1. U. France G. 50125 Firenze. BRAUN. ALEXANDROV. 117259 Moscow. Humboldt University. 117259 Moscow. 1018 XE Amsterdam. Tour 16. NJ 08540.go Fermi 2. D’APPOLLONIO. B. ITEP. D. 2002 Neuchâtel. Jerusalem 91904. Tomsk 63050. Invalidenstr.A. 110. Dipartimento di Fisica. Département de Physique Théorique. A. Italy V. Racah Institute of Physics.PARTICIPANTS D. Hebrew University. ITEP. Physics Faculty. I. DYMARSKY. The Netherlands Instituut voor Theoretische Fysica. Université Pierre et Marie Curie. COUCHOUD. Vorobevy Gory. Switzerland N. University of California.A. BELOV. B. Steklov Mathematical Moscow 117966. U. Cheremushkinskaya 25. Germany L. Sezione INFN. Lenin Av. Princeton University. Princeton. 36. Cheremushkinskaya ul 25. Jadwin Hall. . Russia D. Physics Department. Russia R. CA 93106. 8. CARLEVARO. Université de Neuchâtel. DUIVENVOORDEN. 4 place Jussieu. Gubkin St. Russia Institute. Theoretical Physics department. DOLGUSHEV. Rue Breguet 1. Moscow 119899. Israel V. LPTHE. BENA. FERRARI.S. L. Russia F. 75252 Paris Cedex 05. Institut de Physique. University Valckenierstraat 65. 10115 Berlin. Physik QFT.S. Department of Physics. of Amsterdam. Joseph Henry Laboratories. Moscow State University.

Orme des Merisiers. HENRY. Russia A. CPHT. Mathematical Institute. Department of Physics. Sakyo-ku. 24 rue Lhomond. Physics Department. Mumbai 400005.xiv B. LOW. M. Japan V. Moscow 119899. Stanford University.A. 382 via Pueblo Mall. Martin Luther Universität Wittenberg. Homi Bhabha Road.S. Quantum Statistics and Quantum Field Theory Department. École Polytechnique. U. France L. Department of Physics. 75005 Paris. 41296 Gothenburg. 24-29 St. Tata Institute of Fundamental Research. KONECHNY. Moscow State University. Moscow State University. HUBENY. Cambridge. Sweden S. I. University of Oxford. Chalmers University of Technology. Stanford University.S.A. Kyoto 606-8502. ENS. Vorobevy Gory. McALLISTER. Jefferson Physical Laboratories. HERZOG. S. Case 907. Stanford. Physics Faculty. GURRIERI. Harvard University. D. Theoretical Physics Department. CA 94305-4060. 91128 Palaiseau. Fachbereich Physik. NEMANI. Theoretical Physics Group. Cambridge. MA 02138. U.A. FLOREA. LPT.S. Berkeley. U. MALYSHEV. CEA Saclay. bâtiment 774. France S. U. France C. Mail Stop 50A-501 4BNL. Kyoto University. Vorobevy Gory. S. Department of Theoretical Physics. Göteborg University. MORIYAMA. 382 Via Pueblo Mall. Princeton. Department of Physics. U. U.S. NJ 08544. MA 02138. University of California Berkeley.A. India . Giles.S. 06108 Halle. V. Oxford OX1 3LB. HAACK. GRAN. Colaba. Friedmann Bach Platz 6. CA 94305.S. KASTE. Russia I. Germany P. Department of Theoretical Physics. Princeton University. KitashirakawaOiwakecho. Moscow 119899. Harvard University. KLEVTSOV. U. U. Physics Faculty. MASINA. Department of Physics. Stanford.A.S. CA 94720. France Théorique Luminy. 91191 Gif-sur-Yvette. GUKOV. Centre de Physique 13288 Marseille Cedex 9. P.A. SPhT.K.

Princeton. SHCHERBAKOV. Ukraine M.A. Dipartimento di Fisica. Uppsala University. 00133 Roma. 9747 AG Groningen. 2300 RA Leiden. The Netherlands C. NJ 08854-8019. Departamento De Fisica Teorica C-XI. RABADAN. G. DAMTP. NY 10003. France A. 28049 Madrid.O.S. 117259 Moscow. PARK. New York. Department of Physics & Astronomy.O. France Théorique. via della Ricerca Scientifica 1. University of Cambridge. Centre de Physique 91128 Palaiseau. V. Via Beirut 4. U. K. SPhT. F.xv F. Rutgers. Washington Road. PESTUN. 751 08 Uppsala. U. 13. Russia S. Institute for Theoretical Physics. University of P. ITEP. NOGRADI. U. Box 2208. Mathematics Department. Wilberforce Road. Universita di Roma “Tor Vergata”. Theoretical Physics Department. J. 136 Frelinghuisen Road. The State University of New Jersey. Department of Theoretical Physics.S.A. V.A. RYCHKOV. 4 Washington Place. Cheremushkinskaya 25. Spain S.K. Centre for Mathematical Sciences. Russia R. École Polytechnique. Ann Arbor. Cheremushkinskaya 25. ROEST. The Netherlands L. 117259 Moscow. Princeton University. Department of Physics. U. ROMELSBERGER. 452-48 Cal Tech. SERVANT. U. SCHÄFER-NAMEKI. Italy G. New York University.A. SISSA. Nijenborgh 4. Dnepropetrovsk 49050. RIBAULT. Heraklion. Greece Crete. ITEP. D. Fine Hall. Naukova St. Sweden . PANDO ZAYAS. The University of Michigan.S. Box 9506. Dnepropetrovsk National University. PANOTOPOULOS.S. CA 91125. Randall Laboratory of Physics.A. Institute Lorentz. CEA Saclay. Physics Department. 1200 East California Blvd. NJ 08544. M. B. Pasadena. RICCIONI.S. 91191 Gif-sur-Yvette. Italy D. University of Leiden. U. Universidad Autonoma de Madrid. B. SARAIKIN. SCHNABL. Box 803. Piscataway. California Institute of Technology.. MI 48109-1120. SMEDBÄCK. P. Cambridge CB3 0WA. NITTI. 34014 Trieste. Michigan Center for Theoretical Physics.

B. ULANOV. St. S. Cambridge CB3 OWA. 17A Oxford Street. ZOTOV. South Korea A.xvi A. Institute for Theoretical Physics. ITEP. Department. YEE. Harvard University. U. Cambridge. Dnepropetrovsk 49050. STAHN. SPRADLIN. Naukova St.. VASSILIEV. Institute for Theoretical Physics. Seoul 151-742. 8093 Zurich. Ul B. WALCHER. 117259 Moscow. C. Harvard University. Cambridge. 117259 Moscow. Department of Applied Mathematics and Theoretical Physics. SOLOVYOV.S. Ukraine M. MA 02138.A. Switzerland J. Dnepropetrovsk National University. U. U. Seoul National University. 14 B Metrologicheskaya St. Cheremushkinskaya 25.S.K. Russia A. Department of Physics. Kiev. MIPT.A. VOLOVICH. School of Physics. 13. Ukraine D. ITEP. Zurich. J. Cheremushkinskaya 25.-T. ETH. Theoretical Physics Department. Russia . Physics MA 02138.

The Netherlands .LECTURE 1 SUPERGRAVITY B. Utrecht University. DE WIT Institute for Theoretical Physics & Spinoza Institute.

. . . . . . . . . . . . . e 2. . . . . 110 7 Superconformal symmetry 116 7. . . . . . .2 Maximal supersymmetry and supergravity . . . . . . . . . 3. . . . . . the T -tensor . . . . . . . . . . . . . . . .4 Gauged nonlinear sigma models . . . . . . . . . . . . . . . . . . . . 2. .2 Massless supermultiplets .5 Frames and ﬁeld redeﬁnitions . . . . . . . . . . . . . . . 5. . .1 The superconformal algebra . . . . . . . . space . . . . . . . . . . 119 7. . . . . . . . . . . . . . . .3 Massive supermultiplets . . . . . . . . . .1 On E7(7) /SU (8) and E6(6) /U Sp(8) cosets . . . . . .3 D = 11 supergravity . . . . . . . . . . . . . . .1 The Poincar´ supersymmetry algebra . . 3 4 5 7 17 19 22 28 28 32 35 38 43 46 53 55 56 61 66 69 70 72 74 79 84 3 Supergravity 3. . . 93 6. . 2. . .2 Superconformal gauge theory and supergravity . . . . . . 4 Homogeneous spaces and nonlinear sigma models 4. . . . . . . . 2. . . . . . . . . . . . 4. . . . . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . .6 Kaluza–Klein states and BPS-extended supergravity 3. . . . . 3. . . . . . . . . . . . . . . 4. . . . . . .2 Unitary representations of the anti-de Sitter algebra .3 Electric–magnetic duality and E7(7) . . . . . . .3 Nonlinear sigma models with homogeneous target 4. . . .4 Central charges and multiplet shortening . . 97 6. . . .Contents 1 Introduction 2 Supersymmetry in various dimensions 2. . . . . . . . . . . . .1 Nonlinearly realized symmetries . . . . . . . . . . . . . . .3 The superalgebras OSp(N |4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 7. . . . . . . . . . 126 . . . . . . . . . . . . . .1 Anti-de Sitter supersymmetry and masslike terms . . 3. . . . . . .4 Gauging maximal supergravity.2 On ungauged maximal supergravity Lagrangians 5. 5 Gauged maximal supergravity in 4 and 5 dimensions 5. . 3. . . . .5 On spinors and the R-symmetry group HR . . . . . . . . . 5. . . . . . . 6 Supersymmetry in anti-de Sitter space 89 6. . . . . . . . . . . . . . .4 Dimensional reduction and hidden symmetries . . . . . . . . . . .1 Simple supergravity . .3 Matter ﬁelds and currents . . . . . . . . .7 Nonmaximal supersymmetry: Q = 16 . . . . . . . . . . . . . . . . . . . .2 Geometrical quantities . . . . . . .

which will cover a number of basic aspects of supergravity. but the subject has grown so much and has so many diﬀerent facets that no comprehensive treatment is available as of today. u Hermann Nicolai. somewhat hypothetical. or rather.g. [1]). c EDP Sciences. Springer-Verlag 2002 . During its historical development the perspective of supergravity has changed. Francisco Morales. yet another. called M-theory. Although we know a lot about Mtheory. see. Murat G¨ naydin. e. Mario Trigiante. Jan Louis. many topics will be left untouched.g. String theory is no longer a theory exclusively of strings but includes other extended objects that emerge in the supergravity context as solitonic objects. its underlying principles have only partly been established. theory. Kostas Skenderis. and in recent developments of the conceptual basis of quantum ﬁeld theory and quantum gravity. The advances made have found their place in many reviews and textbooks (see. Also in these lectures.SUPERGRAVITY B. Peter van Nieuwenhuizen. Looking backwards it becomes clear that there are many reasons why neither superstrings nor supergravity could account I am grateful to Sergio Ferrara. The only candidate for such a theory is superstring theory (for some reviews and textbooks. Ergin Sezgin. [2]). Stefan Vandoren and Toine Van Proeyen for many helpful and stimulating discussions. e. Soo-Jong Rey. in our understanding of many central features of superstring theory. String theory and supergravity in their modern incarnations now represent some of the many faces of M-theory. But nowadays supergravity is primarily viewed as an eﬀective ﬁeld theory describing the low-mass degrees of freedom of a more fundamental underlying theory. Henning Samtleben. Originally it was envisaged as an elementary ﬁeld theory which should be free of ultraviolet divergencies and thus bring about the long awaited uniﬁcation of gravity with the other fundamental forces in nature. de Wit 1 Introduction Supergravity plays a prominent role in our ideas about the uniﬁcation of fundamental forces beyond the standard model.

To appreciate the implications of supersymmetry. In Section 4 gauged nonlinear sigma models with homogeneous target spaces are introduced. where the emphasis is on gauged supergravity with 32 supercharges in 4 and 5 spacetime dimensions. This is one of the motivations for considering anti-de Sitter supersymmetry and the representations of the anti-de Sitter group in Section 6. This section is self-contained. gauge theories and string theory from various perspectives. Here we should make a distinction between a supermultiplet of ﬁelds which transform irreducibly under the supersymmetry transformations.to weak-coupling regimes. This construction is explained in Section 5. Section 7 contains a short introduction to superconformal transformations and superconformally invariant theories. and a supermultiplet of states described by a supersymmetric theory. These theories can describe anti-de Sitter ground states which are fully supersymmetric. Gauge Theory and Strings for all the relevant degrees of freedom and we have learned to appreciate that M-theory has many diﬀerent realizations. . It controls the dynamics and. Because supersymmetry is such a powerful symmetry it plays a central role in almost all these developments. paving the way for the construction of gauged supergravity. mostly concentrating on the maximally supersymmetric cases. because of nonrenormalization theorems. precise predictions can be made in many instances. We intend to stay within the supergravity perspective and to try and indicate what the possible implications of supersymmetry and supergravity are for these subjects. but it is of course related to the discussion in Section 6 on anti-de Sitter representations as well as to the adS/CFT correspondence. In this section1 we concentrate on supermultiplets of states. primarily restricting ourselves to ﬂat Minkowski spacetimes of dimension D. often relating strong. The relevant symmetries in this case form an 1 The material presented in this and the following section is an extension of the second section of [6]. Subsequently supergravity theories are introduced in Section 3. see [1. This school oﬀers a large number of lectures dealing with gravity.4 Unity from Duality: Gravity. 3–6]). 2 Supersymmetry in various dimensions An enormous amount of information about supersymmetric theories is contained in the structure of the underlying representations of the supersymmetry algebra (for some references. Section 2 starts with a detailed discussion of supersymmetry and its representations. Our hope is that the material presented below will oﬀer a helpful introduction to and will blend in naturally with the material presented in other lectures.

charge conjugation preserves the chirality of the spinor.1 The Poincar´ supersymmetry algebra e The generators of the super-Poincar´ algebra comprise the supercharges. such as the supersymmetric extensions of the antide Sitter (or the conformal) algebras. Now there are again two possibilities. e Observe that the issue of central charges is diﬀerent when not in ﬂat space. The spinors are then Dirac spinors. the generators of the Lorentz group. · · · . Here Γµ are the gamma matrices that generate the Cliﬀord algebra C(D − 1. (2. . When this dimension is odd. The result is summarized in Table 1. For the remaining two sequences. +. We adhere to this nomenclature.1) where we suppressed the central charges. which consist of translations and e Lorentz transformations. charge conjugation relates positive. are called “central charges”. see [7]. All spinors in this sequence can be restricted to Majorana spinors. The second one corresponds to those odd dimensions where Majorana spinors do not exist. 1) with Minkowskian metric ηµν = diag(−. it is possible in certain cases to have Majorana spinors. For a recent practical introduction to superalgebras. there exist ﬁve diﬀerent sequences of spinors. Qβ } = −2iPµ (Γµ )αβ . However. ¯ {Qα . for example. 2. many of the concepts that we introduce will also play a role in the discussion of other superalgebras. As shown. where the number of dimensions is a multiple of 4. There are other relevant superalgebras. such as the antide Sitter (or conformal) superalgebras. The ﬁrst step is therefore to determine Q for any given number of spacetime dimensions D. e transforming as spinors under the Lorentz group. in the context of the anti-de Sitter superalgebra (discussed in Sect. and possibly additional generators that commute with the supercharges. These cases constitute the ﬁrst sequence. often called central charges2. In even dimension one may distinguish three sequences. +). but not necessarily with all the generators of the Poincar´ algebra. 6). all charges that commute with the supercharges. The most important anti-commutation relation of the super-Poincar´ e algebra is the one of two supercharges. as can be seen. Usually. corresponding to spacetimes of particular dimensions. In the ﬁrst one. The size of a supermultiplet depends exponentially on the number of independent supercharge components Q.B. These will be encountered in due course. depending on whether Majorana spinors 2 The terminology adopted in the literature is not always very precise.with negative-chirality spinors. the energy and momentum operators. For the moment we ignore these additional charges. de Wit: Supergravity 5 extension of the Poincar´ transformations.

one can consider combinations of N+ positiveand N− negative-chirality spinors. The fourth column denotes the type of spinors: Majorana (M). In addition we include the number of . corresponding to D = 6 mod 8. Dirac (D). D 3. In all these cases there exists a group HR of rotations of the spinors which commute with the Lorentz group and leave the supersymmetry algebra invariant. acting on N -fold reducible spinor charges. for instance. comprise the fourth sequence. For the last sequence with D = 2 mod 8. 9. In the second column. This group. For Weyl charges. Qirr speciﬁes the real dimension of an irreducible spinor in a Ddimensional Minkowski spacetime. 8. In the latter two sequences of spinor charges shown in Table 1. often referred to as the “R-symmetry” group. so that HR decomposes into the product of two such groups. 11. 7. Here we list the real dimension of an irreducible spinor charge and the corresponding spacetime dimension. The third column speciﬁes the group HR for N -extended supersymmetry. Weyl (W) and Majorana-Weyl (MW).6 Unity from Duality: Gravity. we allow N± charges of opposite chirality. mod 8 6. mod 8 5. in the context of the Euclidean rest-frame superalgebra for massive representations or for the anti-de Sitter superalgebra. Another way to present some of the results above. It is often realized as a manifest invariance group of a supersymmetric ﬁeld theory. Those will be discussed later in this section. one for each chiral sector. In Table 1 we have listed the corresponding HR groups for N irreducible spinor charges. Here we have assumed that HR is compact so that it preserves a positive-deﬁnite metric. Majorana spinors exist and the charges can be restricted to so-called Majorana-Weyl spinors. is thus deﬁned as the largest subgroup of the automorphism group of the supersymmetry algebra that commutes with the Lorentz group. mod 8 4. There are other versions of the R-symmetry group HR which play a role. The cases where we cannot have Majorana spinors. is shown in Table 2. One can consider extended supersymmetry. but this is by no means necessary. Gauge Theory and Strings Table 1. 10. The supercharges in ﬂat Minkowski spacetimes of dimension D. deﬁned in the text. mod 8 2. mod 8 Qirr 2(D−1)/2 2(D+1)/2 2D/2 2D/2 2D/2−1 HR SO(N ) U Sp(2N ) U (N ) U Sp(2N+ ) × U Sp(2N− ) SO(N+ ) × SO(N− ) type M D M W MW can exist or not. where the spinor charges transform reducibly under the Lorentz group and comprise N irreducible spinors.

1) has therefore zero eigenvalues. The matrix Pµ Γµ on the righthand side of (2.1) as (using ¯ Q = iQ† Γ0 ). which. satisﬁes P 2 = 0. we may consider the states at arbitrary but ﬁxed momentum Pµ .2) β For light-like P µ = (P 0 . The supersymmetry anti-commutator can then be written as ˜ ˜ {Qα . In a positive-deﬁnite Hilbert space some (linear combinations) of the supercharges must therefore vanish. Q† } = 2 (P Γ0 )αβ . Here Γ is the gamma matrix along the spatial momentum P of the states. (2. de Wit: Supergravity 7 Table 2. This is usually (one of) the smallest possible supermultiplet(s). .2 Massless supermultiplets Because the momentum operators Pµ commute with the supercharges. The third column represents the number of bosonic + fermionic massless states for the shortest supermultiplet. we mean the multiplet with the helicities of the states as low as possible. To exhibit this more explicitly. 5 =4 =3 shortest supermultiplet 128 + 128 8+8 4+4 2+2 1+1 states of the shortest3 supermultiplet of massless states. with phase factors such that ˜ ˜ (ΓD )2 = (Γ⊥ )2 = 1 .3) β αβ ˜ Here ΓD consists of the product of all D independent gamma matrices. / (2. let us rewrite (2. Q† } = 2 P 0 1 + ΓD Γ⊥ . 8. perpendicular to P ).4) 3 By the shortest multiplet. ˜ ˜ [ΓD . 9. {Qα . Qirr 32 16 8 4 2 D D D D D D = 11 = 10. 2. 7 = 6. written as a sum of bosonic and fermionic states. for massless representations. We return to a more general discussion of the R-symmetry groups and their consequences in Section 2. and ˜ ⊥ of the product of all D − 2 gamma matrices in the transverse directions Γ (i. P ) the right-hand side is proportional to a projection operator (1 + Γ Γ0 )/2.e. Simple supersymmetry in various dimensions. Γ⊥ ] = 0 .B. We present the dimension of the irreducible spinor charge with 2 ≤ Qirr ≤ 32 and the corresponding spacetime dimensions D.5. (2..

In D = 6 there are three theories with Q = 32 and one with Q = 24. 4. For a more recent discussion. this is so because the representation is complex. when regarding the 16 spinor charges as gamma matrices. for instance [1. When confronting these results with the last column in Table 2. it follows that the representation space constitutes the spinor representation of SO(16). alternatively. the dimensions are shown in Table 2. whereas the other ones generate a Cliﬀord algebra. 10 and 6 spacetime dimensions. 16. For D = 6. For D = 3 we refer to [8]. half the spinors must vanish on physical states. On the other hand. see [10]. as one might naively expect. The two chiral spinor subspaces correspond to the bosonic and fermionic states. 4 the representation is twice as big because it must also accommodate fermion number (or.8 Unity from Duality: Gravity. To determine the helicity content of the bosonic and fermionic states. because it must be CPT self-conjugate). which decomposes into two chiral subspaces. 2. one corresponding to the bosons and the other one to the fermions. 20 and 24. 24. it turns out that the dimension of the shortest supermultiplet is not just equal to 2Qirr /4 . Some of these theories will be discussed later in more detail (in particular supergravity in D = 11 and 10 spacetime dimensions). In odd-dimensional spacetimes irreducible spinors are subject to the eigenvalue ˜ condition ΓD = ±1. 20. So far these supergravities have played no role in string theory. Therefore (2. The derivation for D = 4 is presented in many places (see. the representation space of the charges decomposes into the two chiral spinor representations of SO(Q/2). . Denoting the real dimension of the supercharges by Q. For D = 3. Bigger supermultiplets can be obtained by combining irreducible multiplets by requiring them to transform nontrivially under the Lorentz group. For the massless multiplets. in D = 5 there exists a theory with Q = 24 [9]. Consequently. one considers 4 In D = 4 there exist theories with Q = 12.1 D = 11 supermultipets In 11 dimensions we are dealing with 32 independent real supercharges. many supergravity theories exist. which projects out half of the spinor space. Pure supergravity theories with spacetime dimension 4 ≤ D ≤ 11 can exist with Q = 32. Gauge Theory and Strings This shows that the right-hand side of (2. Depending on the number of spacetime dimensions. 5]).2.3) is proportional to a projection operator.3) simpliﬁes and shows that the 16 nonvanishing spinor charges transform according to a single spinor representation of the helicity group SO(9). respectively. corresponding to D = 11. We shall demonstrate this below in three relevant cases. 8 and 4 supersymmetries4 . 12.

which is a characteristic property of the group SO(8). In this context one uses the notation 10 . is to investigate how they decompose 5 The representations can be characterized according to the four diﬀerent conjugacy classes of the SO(8) weight vectors. For instance.B.2 D = 10 supermultiplets In 10 dimensions the supercharges are both Majorana and Weyl spinors. denoted as the 8v and 8c representations. According to (2. v and c (see. One way to distinguish the inequivalent representations. It then turns out that one of the 128 representations branches into helicity representations according to 128 → 44 + 84. and 350 . to the other two.3). respectively. nonextended) supersymmetry with 16 charges. each transforming according to a chiral spinor representation. The smallest massless supermultiplet has now been constructed with 8 bosonic and 8 fermionic states and corresponds to the vector multiplet of supersymmetric Yang-Mills theory in 10 dimensions [13]. the nonvanishing charges transform in a chiral spinor representation of the SO(8) helicity group.e. we assign the bosonic and fermionic states.. which are inequivalent. 35c . the smallest massive supermultiplet comprises 32 768 + 32 768 states. without central charges. Hence we are dealing with three 8-dimensional representations of SO(8). the antisymmetric tensor states in the 84 and the gravitini states in the 128 representation of SO(9). They are traditionally distinguished by labels s. respectively. Before constructing the supermultiplets that are relevant for D = 10 supergravity. These multiplets will not be considered here.2. v. let us ﬁrst discuss some other properties of SO(8) representations. de Wit: Supergravity 9 the embedding of the SO(9) spinor representation in the SO(16) vector representation. One is the representation to which we assign the supercharges. ˜ The latter means that they are eigenspinors of ΓD . The above states comprise precisely the massless states corresponding to D = 11 supergravity [11]. such as the adjoint and the singlet representation. s and c. 280 . when we have simple (i. whose irreducible representation space corresponds to the bosonic and fermionic states. 2. while the second one transforms irreducibly according to the 128 representation of the helicity group. . which we will denote by 8s . denoted by 0. Bigger supermultiplets consist of multiples of 256 states. 350 . 350 for 35v . the three types of representation are inequivalent. for instance [12])5 . The fact that SO(8) representations appear in a three-fold variety is known as triality. The graviton states transform in the 44. With 8 nonvanishing supercharges we are dealing with an 8-dimensional Cliﬀord algebra. 35s . With the exception of certain representations.

= 8s + 56s . Therefore there is an SO(7) subgroup under which the 8v representation branches into 8v −→ 7 + 1.and 56-dimensional representations are 28 −→ 7 + 21 . and leads to the second supermultiplet shown in Table 3. There are thus three diﬀerent possibilities. Corresponding branching rules for the 28-. with one of the 8-dimensional representations. 8c −→ 8 . 35c. Multiplication . Multiplying 8v with 8v + 8c yields 8v × 8v bosonic and 8v × 8c fermionic states. 8v × 8s 8s × 8c 8c × 8v = 8c + 56c . 8v × 8v 8s × 8s 8c × 8c = = = 1 + 28 + 35v . = 8v + 56v .g. where 8 is the spinor representation of SO(7). (2. Massless N = 1 supermultiplets in D = 10 spacetime dimensions containing 8 + 8 or 64 + 64 bosonic and fermionic degrees of freedom. 56v 56c. 35v −→ 1 + 7 + 27 . 35. Under this SO(7) the other two 8-dimensional representations branch into 8s −→ 8 . 1 + 28 + 35s . 1 + 28 + 35c .10 Unity from Duality: Gravity. Using the multiplication rules for SO(8) representations. Gauge Theory and Strings Table 3.s −→ 21 + 35 . supermultiplet vector multiplet graviton multiplet gravitino multiplet gravitino multiplet bosons 8v 1 + 28 + 35v 1 + 28 + 35c 8v + 56v fermions 8c 8s + 56s 8s + 56s 8c + 56c into representations of an SO(7) subgroup. which can be associated with the states of the graviton in D = 10 dimensions (the ﬁeld-theoretic identiﬁcation of the various states has been clariﬁed in many places.6) it is straightforward to obtain these new multiplets. (2. This supermultiplet contains the representation 35v . Therefore this supermultiplet will be called the graviton multiplet.5) In order to obtain the supersymmetry representations relevant for supergravity we consider tensor products of the smallest supermultiplet consisting of 8v + 8c . see e.s −→ 35 . −→ 8 + 48 . the Appendix in [6]). each leading to a 128-dimensional supermultiplet. Each of the 8-dimensional representations leaves a diﬀerent SO(7) subgroup of SO(8) invariant.

This feature reﬂects itself in the multiplet decomposition. since there are two inequivalent gravitino multiplets. This is the multiplet corresponding to IIB supergravity [14]. Therefore these two supermultiplets are called gravitino multiplets. where the 1. there will also be two inequivalent N = 2 supermultiplets containing the states corresponding to a graviton and two gravitini. In this case the supercharges constitute two Majorana-Weyl spinors of opposite chirality. we interchange the boson and fermion assignments in these products. 28 and 56s representations are degenerate and constitute doublets under this SO(2) group. By combining a graviton and a gravitino multiplet it is possible to construct an N = 2 supermultiplet of 128 + 128 bosonic and fermionic states. A second supermultiplet may be viewed as the tensor product of a (8v +8s ) supermultiplet with a second supermultiplet (8v +8c ). . except that we will associate the 8c and 8s representations with fermionic quantities (note that these are the representations to which the fermion states of the Yang-Mills multiplet and the supersymmetry charges are assigned). However. which can be associated with gravitino states. Such a multiplet follows if one starts from a supersymmetry algebra based on two MajoranaWeyl spinor charges Q with the same chirality. Multiplication with 8c then leads to 8c × 8c bosonic and 8c × 8v fermionic states. graviton and gravitino supermultiplets are shown in Table 3. Hence the automorphism group HR is equal to SO(2). whereas multiplication with 8s gives 8s × 8c bosonic and 8s × 8v fermionic states. Because the supercharges have the same chirality. respectively. de Wit: Supergravity 11 with 8c or 8s goes in the same fashion. The explicit SO(8) decompositions of the vector. We have thus established the existence of two inequivalent gravitino multiplets.7) fermions: 8s + 8s + 56s + 56s . According to the construction presented above. These supermultiplets contain fermions transforming according to the 56s and 56c representations. Consequently. one N = 2 supermultiplet may be be viewed as the tensor product of two identical supermultiplets (namely 8v + 8c ). one can perform rotations between these spinor charges which leave the supersymmetry algebra unaﬀected. 8s . The states of this multiplet decompose as follows: Chiral N = 2 supermultiplet (IIB) bosons: 1 + 1 + 28 + 28 + 35 + 35 v c (8v + 8c ) × (8v + 8c ) =⇒ (2. but no graviton states as those transform in the 35v representation.B.

The formula below summarizes the massless states of IIA supergravity from an 11-dimensional perspective. which in D = 9 dimensions comprises the rest-frame rotations. However. 84 and 128 representation of the helicity group SO(9).8) fermions: 8s + 8c + 56s + 56c . This leads to precisely the same supermultiplet as the reduction of the states of IIA supergravity. the massive states are still characterized in terms of the group SO(8). a 3-rank antisymmetric gauge ﬁeld and the gravitino ﬁeld.9) Clearly. characterized in terms of representations of the helicity group SO(7): −→ 1 1 −→ 1 + 7 8v 44 =⇒ 35v −→ 1 + 7 + 27 84 =⇒ 8s 8c 128 =⇒ 56s 56c 28 56v −→ 7 + 21 −→ 21 + 35 −→ −→ −→ −→ 8 8 8 + 48 8 + 48. in D = 9 we have a degeneracy of states. We also show how the 10-dimensional states can subsequently be branched into 9-dimensional states. transforming in the 27 and 48 representations of the SO(7) helicity group. respectively. the reductions of IIA and IIB supergravity to 9 dimensions. One could also take the states of the IIB supergravity and decompose them into D = 9 massless states. (2. Gauge Theory and Strings Now the supermultiplet decomposes as follows: Nonchiral N = 2 supermultiplet (IIA) bosons: 1 + 8 + 28 + 35 + 56 v v v (8v + 8s ) × (8v + 8c ) =⇒ (2. associated with the group HR = SO(2). yield the same theory [16–18].12 Unity from Duality: Gravity. Indeed. They correspond to the degrees of freedom described by the metric. Therefore the Kaluza-Klein states that one obtains when compactifying the ten-dimensional theory on a circle remain inequivalent for the IIA and IIB theories (see [19] for a discussion of this phenomenon and its . The latter follows from the fact that two D = 10 Majorana-Weyl spinors with opposite chirality can be combined into a single D = 11 Majorana spinor. The massless states of 11-dimensional supergravity transform according to the 44. It can be obtained by a straightforward reduction of D = 11 supergravity. We note the presence of graviton and gravitino states. This is the multiplet corresponding to IIA supergravity [15].

2. The states transform both in the SU+ (2) helicity group and under a U Sp(2N+ ) group. Shortest massless supermultiplets of D = 6 N+ -extended chiral supersymmetry. vector and 6 The content of this table also speciﬁes the shortest massive supermultiplets in four dimensions as well as with the shortest massless multiplets in ﬁve dimensions. Let us now turn to speciﬁc supermultiplets. Tables 10 and 11). gravitini. for even N+ they can be chosen real. Of course. The ﬁeld content of the maximal Q = 32 supergravity theories for dimensions 3 ≤ D ≤ 11 will be presented in two tables (cf. The SU (2) group is then associated with spin or with helicity. 3 or 4 supercharges. for N+ chiral spinors. Maximal supergravity will be introduced in Section 3. In Table 4 we present the decomposition of the various helicity representations of the smallest supermultiplets based on N+ = 1. except for D = 10. but almost all information follows from ﬁrst considering the purely chiral case. Likewise N− negative-chirality spinors transform under U Sp(2N− ). When we have supercharges of only one = chirality. In principle we have N+ positiveand N− negative-chirality charges. respectively. an identical table can be given for negative-chirality spinors.2.B. as is shown in Table 46 . de Wit: Supergravity 13 Table 4. This feature is already incorporated in Table 1. which are not Majorana. Because the charge conjugated spinor has the same chirality. the chiral rotations of the spinors can be extended to the group U Sp(2N+). In D = 6 dimensions the helicity group SO(4) decomposes into the product of two SU (2) groups: SO(4) ∼ (SU + (2) × SU − (2))/Z2 . Let us recall that the helicity assignments of the states describing gravitons. It turns out that the Q = 32 supergravity multiplets are unique in all spacetime dimensions D > 2.3 D = 6 supermultiplets In 6 dimensions we have chiral spinors. SU+ (2) 5 4 3 2 1 N+ = 1 N+ = 2 N+ = 3 1 6 14 14 (32 + 32)C 1 2 (2 + 2)C 1 4 5 (8 + 8)R N+ = 4 1 8 27 48 42 (128 + 128)R consequences). the smallest supermultiplet will only transform under one SU (2) factor of the helicity group. . For odd values of N+ the representations are complex. 2.

there exists also a nonchiral version with 16 supercharges. 3) helicity representation yields the (2. one obtains the states of (1. Here (m. a doublet of chiral spinors and a scalar. consisting of the graviton. as well as the lower-dimensional ones. and a four-plet of chiral fermions. 3) or (3. then corresponds to the (2. 3). 3). 0) tensor multiplet. with a vector state. which contain neither gravitons nor gravitini. 2). (3. 1) helicity representation gives the (1. shown in Table 4. 0) and (0. 0) hypermultiplet. N− ) = (1. see for instance one of the Appendices in [6]. 3) representation. The smallest multiplet. We now elucidate the construction of the supermultiplets listed in Table 5. with a selfdual tensor. 1) supersymmetry. (2. 2). When a supermultiplet contains (3. with the bosonic states decomposing into a selfdual tensor. There are no such multiplets for more than Q = 32 supercharges. Multiplication with the (1. Again. 2) . There are supermultiplets with higher SU (2) helicity representations. and a ﬁve-plet of spinless states. 1). the selfdual tensors of the tensor and of the supergravity supermultiplet are of opposite selfduality phase. where the smallest supermultiplet is the (1. 3). because gravitini without a graviton are not expected to give rise to a consistent interacting ﬁeld theory. a spinless state and a doublet of chiral spinors. 0) tensor multiplet. 1) or (1. 2) or (2. with the vector state and four scalars. The smallest multiplet is now given by the tensor product of the supermultiplets with (1. This yields the vector multiplet. The multiplets of this type are shown in Table 5. Gauge Theory and Strings (anti)selfdual tensor gauge ﬁelds. N− ) = (2. 2). 0). 3) helicity representation. 0) supersymmetry. The tensor product with the (1. (3. (3. 1) or (1. Some of these multiplets are shown in Table 6 and we will discuss them in due course. consisting of a complex doublet of spinless states and a chiral spinor. and (2. and spinor ﬁelds are (3. 1) or (1. 0) supergravity.14 Unity from Duality: Gravity. N− ) = (1. Multiplying the hypermultiplet with the (2. Hence we have only (3. For the derivation of these assignments. (2. the latter transforming with respect to the (2. Of course. 0) vector multiplet. 2) helicity representation yields the (1. 3) representations. Taking the tensor product of the smallest supermultiplet with the (2. 3) representations. n) denotes that the dimensionality of the reducible representations of the two SU (2) factors of the helicity group are of dimension m and n. 0) supergravity multiplet. four chiral gravitini and ﬁve selfdual tensors [20]. Observe that the selfdual tensor ﬁelds in the tensor and supergravity supermultiplet are of opposite selfduality phase. In the following we will ﬁrst restrict ourselves to helicities that correspond to at most the three-dimensional representation of either one of the SU (2) factors. (2. we insist that it will also contain a single (3. 3). Next consider (N+ . namely the one corresponding to (N+ . The simplest case is (N+ . 2).

1. 2) graviton 128 + 128 fermions (2. 1) +(1. 1) (3. 1) (2. 1) (3. 3. 2. 2. .B. 1. 3. 1) (2. m. 2. 1) + (1. 1) + h. 1. 3. 2. 1) (2. 2. 1. 5. multiplet (1. 1. 1. 1. Some relevant D = 6 supermultiplets with (N+ . 5. 2. 1) (2. 5. 4. 1) + (1. 1. 5. corresponding to a graviton. 1) +(3. 1. 4. 0) tensor (1. 2) +(2. 1) (1. 4) + (1. 5) (2. 3. 3. 0) vector (1. 2. 2) (3. 1) graviton # 4+4 4+4 4+4 12 + 12 8+8 24 + 24 8+8 32 + 32 bosons (1. 1. 2. The 32 fermionic states comprise two doublets of chiral gravitini and two chiral spinor doublets. 1) (3. 2. 2. 1) +(1. de Wit: Supergravity 15 Table 5. a tensor. 0) graviton (1. 3. 1. The ˜ ˜ second column lists the number of bosonic + fermionic states for each multiplet. 2. 2. 1) (2. 1) (2. There are two doublets of chiral fermions with opposite chirality. 1. 2. 2. 5. 1. 2) +(1. 2) + (1. 1) +(1. 4) representation of U Sp(2) × U Sp(2). 2. where the latter transform under the (2. 5. 2. 1. 1. 1. 1. 5) + (1. 4. 3. 2) representation of the helicity group yields the states of the (1. 1) + (1. 0) graviton (2. 1. 0) hyper (1. 1. n) are assigned to (m. 1. 1) +(2. 1) graviton 64 + 64 (2. 1. N− ) supersymmetry. 5. 4. 4. 5. transforming as doublets under the appropriate U Sp(2) group. 1) supergravity multiplet. n) representations of U Sp(2N+ ) × U Sp(2N− ). 1) +(2. 2) +(2. 1) (3. 1) + (2. 3. a scalar and four vector states. 3. 1) (2. 1) (3. 1) +(2. 2. Taking the tensor product of the vector multiplet with the (2. 1. 1.c. 3. 1. 1. 1) +(1. n. 1. It consists of 32 bosonic states. 0) tensor (2. 1) + (1. 1. 4. 1. 3. 1. 1. 2) (3. The states (m. 1. 1. 2) +(2. 4) +(2. n) representations of the helicity group ˜ ˜ SU+ (2) × SU − (2) and (m. 1) + (3. 4. 1. 1) vector (1. 1. 2) representation of U Sp(2) × U Sp(2). 1. 3. 1. 5) +(1. 2. 2. 3. 1) +(1. 2) +(2. 1. 2. (3. 1. 3. 3. 1) + (1. each transforming as a doublet under the corresponding U Sp(2) group. 1) + (3. 4. 1. 1) (3. 2. 1) (3. 1. 2. 1.

2) supergravity multiplet. 14) +(3. 2). 1) + (3. one can continue and classify representations for other values of (N+ . The multiplets are identical to those that underly the ﬁve-dimensional N -extended supergravities. . 4) (4. 0) (2. N− ). 14) (4. 1) tensor multiplet. 1) + (1. Restricting ourselves to 32 supercharges and requiring the highest helicity to be a ﬁve-dimensional representation of one of the SU (2) factors. 1) + (3. The states (n. 0) tensor multiplet with the (0. 1) (5. 0) and (0. 0) (3. N− ) = (2. there are the chiral theories which contain neither graviton nor gravitino states. n) are assigned to the n-dimensional representation of SU (2) and the ˜ n-dimensional representation of U Sp(2N ). As is obvious from the construction that we have presented. 1) + (3. 4) + (2. 0) # 8+8 24 + 24 64 + 64 128 + 128 bosons (5. see [10]. Here a supergravity multiplet exists [21] and can be obtained from the product of the states of the (2. which we discard because it contains gravitini but no graviton states. Some of these multiplets will suﬀer from the fact that they have more than one graviton state. 2) supermultiplets. 0) supersymmetry. In principle. The theories based on these multiplets have only rigid supersymmetry. 1) + (1. The smallest supermultiplet is given by the tensor product of the smallest (2. we turn to the case of (N+ . 8) + (2. 48) Subsequently we discuss the case (N+ . 6) +(2. there are just four theories. Gauge Theory and Strings Table 6. 14) (5. 2) (4. a single (5. N− ) = (2. 27) +(1. These states transform according to representations of U Sp(4) × U Sp(4). 42) fermions (4. There is in fact a smaller supermultiplet. supersymmetry (1. The second column lists the number ˜ of bosonic + fermionic states for each multiplet. However. They are all chiral. so that we expect them to be inconsistent at the nonlinear level.16 Unity from Duality: Gravity. 6) + (2. Finally. 1) + (3. summarized in Table 6. 1) +(3. This yields the 128 + 128 states of the (2. so that the helicity group in six dimensions is restricted to SU (2) × 1 and the states are characterized as representations of U Sp(2N ). this will inevitably lead to states transforming in higher-helicity representations. D = 6 supermultiplets without gravitons and gravitini with (N. 0) (4. 1) highest-helicity state and at most 32 supercharges. 1). For a recent discussion of one of these theories. 5) (5.

4. in D = 4 spacetime dimensions. This group will also be denoted by HR . Knowledge of the relevant groups HR is important and convenient in writing down the supermultiplets. However. Q† } = 2M δαβ . as it is required to commute with a smaller group. The analysis of massive supermultiplets takes place in the restframe. The supercharges transform as spinors under this group. Central charges and multiplet shortening are discussed in Section 2. the relativistic R-symmetry group is equal to U (N ). which contains U (N ) as a subgroup according to 2N = N + N. and massive supermultiplets can be decomposed in terms of massless ones. It can also reveal certain relations between supermultiplets. supermultiplets living in higher dimensions can always be decomposed into supermultiplets living in lower dimensions.3 Massive supermultiplets 17 Generically massive supermultiplets are bigger than massless ones because the number of supercharges that generate the multiplet is not reduced. associated with spin. More explicit derivations can be found in [4]. In particular the N = 4 supermultiplet of Table 7 appears in many places and coincides with the massless N = 8 supermultiplet . even between supermultiplets living in spacetimes of diﬀerent dimension. In this section we assume that the central charges are absent. the D = 4 massive multiplets shown in Table 7 coincide with the massless supermultiplets of chirally extended supersymmetry in D = 6 dimensions shown in Table 4. so that one obtains a Euclidean supersymmetry algebra. This only happens for special values of these charges. de Wit: Supergravity 2. β Just as before. The shortened supermultiplets are known as BPS multiplets. The states then organize themselves into representations of the rest-frame rotation group.B. Table 7 shows the smallest massive representations for N ≤ 4 in D = 4 dimensions as an illustration. unlike for massless supermultiplets where one-half of the supercharges vanishes. in the presence of mass parameters the superalgebra may also contain central charges. Obviously. Obviously the nonrelativistic group can be bigger than its relativistic counterpart. For instance.10) {Qα . Clearly the states of given spin can be assigned to representations of the nonrelativistic group HR = U Sp(2N ) and decomposed in terms of irreducible representations of the relativistic R-symmetry group U (N ). it is the nonrelativistic variant of the R-symmetry group that was introduced previously. which could give rise to a shortening of the representation in a way similar to what happens for the massless supermultiplets. For instance. the spinor charges transform under the automorphism group of the supersymmetry algebra that commutes with the spin rotation group. but sometimes there exists a relationship that is less trivial. SO(D − 1). while the nonrelativistic one is the group U Sp(2N ). (2.

precisely in accord with the N = 2 entry in Table 7. Unfortunately. In fact. By employing such ﬁelds one may be able to deﬁne an oﬀ-shell representation.18 Unity from Duality: Gravity. Extending the number of supercharges beyond 16 will increase the minimal spin of a massive multiplet beyond spin-2. which in turn coincides with the spin rotation group for D = 4. because the transformation rules are implicitly dependent on the action. These are the currents that couple to the ﬁelds of N = 4 conformal supergravity. The N = 4 multiplet in Table 7 corresponds to the gravitational supermultiplet of currents [22]. Since the number of eﬀective supercharges is equal in these cases and given by Qeﬀ = 16 (remember that only half of the charges play a role in building up massless supermultiplets). The D = 5 and the chiral D = 6 massless supermultiplets are subject to the same helicity group SU (2). the multiplets must indeed be identical. for instance. as the reader can easily verify. These states can be described by a ﬁeld theory in which the supercharges generate corresponding supersymmetry variations on the ﬁelds. Such representations are called on-shell representations. which is shown in Tables 10 and 11. one may conclude that conformal supergravity does not exist for more than 16 charges. In certain cases one can improve the situation by introducing extra ﬁelds which do not directly correspond to physical ﬁelds. The reasons for this are clear. There is an oﬀ-shell counting argument. Gauge Theory and Strings in D = 5 dimensions. where the transformations close upon (anti)commutation without the need for imposing ﬁeld equations. this multiplet coincides with the multiplet of the currents that couple to an N = 2 supersymmetric gauge theory. according to which the ﬁeld degrees of freedom should comprise a massive supermultiplet (while the states that are described could be massless). This fact makes is much more diﬃcult to construct an extended variety of actions for these theories. The lack of closure has many consequences. the oﬀ-shell description of the N = 2 vector multiplet in D = 4 dimensions can be formulated in terms of a gauge ﬁeld (with three degrees of freedom). also the relevant automorphism groups HR coincide. Here we also want to brieﬂy draw the attention to the relation between oﬀ-shell multiplets and massive representations. Very often the transformations on the ﬁelds do not close according to the supersymmetry algebra unless one imposes the equations of motion for the ﬁelds. Notorious examples are gauge theories and supergravity theories with 16 or more supercharges. So far we discussed supermultiplets consisting of states on which the supercharges act. Not surprisingly. For instance. many theories do not possess (ﬁnite-dimensional) oﬀ-shell representations. Since higher-spin ﬁelds can usually not be coupled. These ﬁelds are known as auxiliary ﬁelds. when determining quantum corrections. a fermion doublet (with eight degrees of freedom) and a triplet of auxiliary scalar ﬁelds (with three degrees of freedom). For that .

However. As we discuss in due course.B. . there may be solitonic or other states that carry charges. In Section 2. The states are listed as U Sp(2N ) representations which are subsequently decomposed into representations of U (N ). 7 There may be additional gauge transformations that are of interest to us. at the nonperturbative level. and may be regarded as central charges. local supersymmetry transformations and the gauge transformations associated with the antisymmetric gauge ﬁelds7 . However. so these central charges simply vanish on physical states. Subsequently we then discuss some further implications of these results. q). Conformal supergravity will be discussed in Section 7. On such states.4 Central charges and multiplet shortening The supersymmetry algebra of the maximal supergravities comprises general coordinate transformations. spin 1 1/2 0 2 3/2 1 1/2 0 N =1 1=1 2=1+¯ 1 N =3 1=1 6=3+¯ 3 14 = 8 + 3 + ¯ 3 14 = 6 + ¯ + 1 + ¯ 6 1 N =2 1=1 4=2+¯ 2 5=3+1+¯ 1 N =4 1=1 8=4+¯ 4 27 = 15 + 6 + ¯ 6 4 48 = 20 + 20 + 4 + ¯ 42 = 20 + 10 + 10 + 1 + ¯ 1 reason there can be no oﬀ-shell formulations for supergravity with more than 16 charges. dyons. some of the central charges may take ﬁnite values.5 we will present a table listing the various groups HR for spinors associated with certain Cliﬀord algebras C(p. An example are magnetic monopoles. At the M-theory level. the corresponding gauge ﬁelds are composite and do not give rise to physical states (at least. Without further knowledge about the kind of states that may emerge at the nonperturbative level. the theory does not contain charged ﬁelds. these charges are associated with certain brane conﬁgurations. In perturbation theory. q) with corresponding rotation groups SO(p. or black holes. it is possible to have (part of) the automorphism group HR realized as a local invariance. de Wit: Supergravity 19 Table 7. Minimal D = 4 massive supermultiplets without central charges for N ≤ 4. not in perturbation theory). These gauge transformations usually appear in the anticommutator of two supercharges. 2.

5) [6] 27 [5] 63 [4] 135 [3] p=2 1 [55] 1 [45] p=3 p=4 1 [210] 1 [120] 1 [84] 1+3 [56] 10 [35] (10. Gauge Theory and Strings Table 8.20 Unity from Duality: Gravity.11) p where Γµ1 ···µp is the antisymmetrized product of p gamma matrices. 1) +(1. The total number of central charges is equal to 528 − D. 1) +(5. by considering a decomposition of the anticommutator. which . C is the charge-conjugation matrix and Zµ1 ···µp is the central charge. 4) 1+2 [126] 3+¯ 3 [35] [15] 36 [10] 36 + 36 [3] we can generally classify the possible central charges. Qβ } ∝ (Γµ1 ···µp C)αβ Zµ1 ···µp . The second row speciﬁes the number of independent components for each p-rank tensor charge. Hence we may write {Qα . 4) 5 4 3 U Sp(8) U (8) SO(16) [1] 1 + 27 [1] 28 + 28 [1] 120 [1] 1 [10] 2 [10] 2 [9] 3 [8] 5 [7] (1. Decomposition of the central extension in the supersymmetry algebra with Q = 32 supercharge components in terms of p-rank Lorentz tensors. 1) +(1. This anticommutator carries at least two spinor indices and two indices associated with the group HR . because we have not listed the D independent momentum operators. (2. D 11 HR 1 p=0 p=1 10A 1 10B SO(2) 9 8 7 6 SO(2) U (2) U Sp(4) U Sp(4) ×U Sp(4) 1 [1] 1+2 [1] 3+¯ 3 [1] 10 [1] (4. 10) [10] p=5 1 [462] 1+1 [126] 1+2 [126] 1 [36] 1+¯ 1 [28] 1+5 [21] (4.

12) The two central charges. The results of the table are in direct correspondence with the eleven-dimensional superalgebra with the most general central charges. which generically implies that the mass of the multiplet is larger than or equal to the central charges. this phenomenon of multiplet shortening is the same as for massless supermultiplets. Hence one speaks of 1/2-BPS. For lower supersymmetry the multiplets can be 8 For related discussions see. the right-hand side of the anticommutator is subject to a positivity condition. Because we have 32 supercharge components. This means that perturbative results based on BPS supermultiplets can often be extrapolated to a nonperturbative regime. in the same way as half of the charges vanish for the massless supermultiplets. Qualitatively. For BPS multiplets some of the supercharges must vanish on the states. The fact that the BPS supermultiplets have a completely diﬀerent ﬁeld content than the generic massive supermultiplets makes that they exhibit a remarkable stability under “adiabatic” deformations. de Wit: Supergravity 21 transforms as an antisymmetric p-rank Lorentz tensor and depends on possible additional HR indices attached to the supercharges. to indicate which fraction of the supercharges vanishes on the states. depending on whether the (Γµ1 ···µp C)αβ is antisymmetric or symmetric in α. so that the product with Zµ1 ···µp is always symmetric in the combined indices of the supercharges. but here the fraction of the charges that vanishes is not necessarily equal to 1/2. where we list all possible charges and their HR representation assignments.B. the bound may can take a complicated form. Especially in higher dimensions. . Qβ } = −2iPM ΓM + ZMN ΓMN + ZMN P QR ΓMN P QR . This vanishing of some of the supercharges leads to a shortening of the multiplet. β. For the maximal supergravities in spacetime dimensions 3 ≤ D ≤ 11 this classiﬁcation is given in Table 8. etcetera. ZMN and ZMN P QR can be associated with the winding numbers of two.and ﬁve-branes. The central charge must be symmetric or antisymmetric in these indices. 1/4-BPS supermultiplets. the sum of the independent momentum operators and the central charges must be equal to (32 × 33)/2 = 528. When the bound is saturated one speaks of BPS states. For higher extended supersymmetry the diﬀerence in size of BPS supermultiplets and massive supermultiplets can be enormous in view of the fact that the number of states depend exponentionally on the number of nonvanishing central charges. In order to realize the supersymmetry algebra in a positive-deﬁnite Hilbert space. αβ αβ αβ (2. for example [23–25] and references therein. For given spacetime dimension all possible central charges can be classiﬁed8 . ¯ {Qα . This positivity bound is known as the Bogomol’nyi bound.

q) based on p + q generators. On the other 2 hand there is another massive vector supermultiplet. The value for r determines the square of the matrix built from forming the product e1 · e2 · · · ep+q of all the Cliﬀord algebra generators. . the four states of two spin. but nevertheless they are quite diﬀerent. where the hypermultiplet provides the scalar degree of freedom that allows the conversion of the massless to massive spin-1 states. This means that p generators square to the identity and q to minus the identity. For instance. We consider the Cliﬀord algebras C(p. 3). Subsequently we shall discuss some correspondences between the various spinors in diﬀerent dimensions. . so that we have again 8 + 8 states. Minkowski and (anti-)de Sitter spaces as well as the supersymmetry algebras based on these spinors. which is BPS. In view of the very large variety of BPS supermultiplets. For r = 0. This multiplet carries no central charge. . The second supermultiplet. where we restrict q ≤ 2. Without central charges. where r is deﬁned by r ≡ p − q mod 4. the eight states of four irreducible spin. denoted by e1 . When decomposing these multiplets into massless N = 2 supermultiplets. such a multiplet comprises 8 + 8 states. From this table we can gain certain insights into the properties of spinors living in Euclidean. In later sections we will discuss speciﬁc BPS supermultiplets as well as other mechanisms of multiplet shortening in anti-de Sitter space. while for r = 2. corresponding to the three states of spin-1.5 On spinors and the R-symmetry group HR In this section we return once more to the spinor representations and the corresponding automorphism group HR . 2. We list the real dimension of the irreducible Cliﬀord algebra representation. the . . appears as a massive charged vector multiplet when breaking a non-Abelian supersymmetric gauge theory to an Abelian subgroup. and the values r (equal to 0.1 representations. denoted by dC . consider N = 2 massive vector supermultiplets in four spacetime dimensions. . . q).1 representations 2 and two states of spin-0. which is BPS and comprises the three states of spin-1. e2 . Gauge Theory and Strings comparable in size. q). for r = 0. we do not continue this general discussion of supermultiplets with central charges. ep+q . and ﬁve states with spin 0. . This realization is known as the Coulomb phase. the ﬁrst multiplet decomposes into a massless vector multiplet and a hypermultiplet. with a nondegenerate metric of signature (p. 1 this square equals the identity. but with a completely diﬀerent spin content. Therefore. . Hence this is the multiplet one has in the Higgs phase. 3 the square equals minus the identity. also known as the R-symmetry group. Let us ﬁrst elucidate the information presented in the table.22 Unity from Duality: Gravity. Table 9 summarizes information for spinors up to (real) dimension 32 associated with the groups SO(p. These states are subject to a nonvanishing central charge which requires that the states are all doubly degenerate.

0). the dimensions of the corresponding spinors diﬀer by a factor two. q) diﬀers by a factor 16. 1) and SO(p − 1. q) (or C(p. q) (or C(p. which. a group HR should be assigned to each of the chiral sectors separately. leaving a positive-deﬁnite metric invariant. q+8)) and C(p. From a . simply coincide. q) generated by products of even numbers of generators is not simple and breaks into two simple ideals.B. the complex numbers (C). q) times the 16 × 16 real matrices. 0) is the helicity group of massless spinor states in ﬂat Minkowski space of dimension D = p + 1. Finally the table lists the branching of the Cliﬀord algebra representation into SO(p. while for r = 1. according to Schur’s lemma. or U Sp(2N ). We now discuss and clarify a number of correspondences between spinors living in diﬀerent dimensions. according to Schur’s lemma the centralizer of SO(p. the group SO(p. according to which there exists an isomorphism between the Cliﬀord algebras C(p+8. We also present the centralizer of the irreducible representations of the Cliﬀord algebra. q) spinor representations.3) thus share the same automorphism group. q) decomposes into two simple ideals. or the quaternions (H). one or three complex structures. respectively. but they are complex so that their real dimension remains unaltered. respectively. In a ﬁeld-theoretic context the reduction of the spinor degrees of freedom is eﬀected by the massless Dirac equation and the automorphism groups HR that commute with the Lorentz transformations and the transverse helicity rotations. q) commutes with the identity and none. When r = 0 the Cliﬀord algebra representation decomposes into two chiral spinors. one or three complex structures. From a physical perspective. the full Cliﬀord algebra C(p. respectively.1) and (2. The last column gives the compact group HR . U (N ). consisting of the linear transformations that commute with the group SO(p. For r = 0. Observe that for r = 2 we can also have chiral spinors. while their respective groups HR always coincide. q +8)) and C(p. q) must form a division algebra for irreducible spinor representations. Correspondingly. which leads to HR = SO(N ). for any p > 1. For r = 2. q) and act on N irreducible spinors. We note that the results of Table 9 are in accord with the results presented earlier in Tables 1 and 2. which generate the corresponding division algebra. the dimension of the representations of C(p+8. This means that the irreducible representation commutes with the identity and none. must form a division algebra and is thus isomorphic to the real numbers (R). Again. Therefore. Table 9 reﬂects also the so-called periodicity theorem [26]. but nevertheless in certain cases the Cliﬀord algebra representation can still decompose into two irreducible spinor representations. this correspondence can be understood from the fact that SO(p − 1. de Wit: Supergravity 23 subalgebra C+ (p. The ﬁrst correspondence is between spinors of SO(p. According to the table. The two algebras (2. 3 there are no chiral spinors.

and the SO(p. 2) C(9. Gauge Theory and Strings Table 9. 2) C(4. 0) C(6. 2) C(5. 1) C(4. 1) C(7. 1) C(7. 1) C(3. 0) C(6. 0) C(0. Representations of the Cliﬀord algebras C(p. q) C(1. 0) C(4. 1) C(1. We also list the dimensions of the Cliﬀord algebra and spinor representation. as well as r = p − q mod 4. q) with q ≤ 2 and their centralizers. 2) C(8.q) 1 1+1 1+1 2 2 2+2 2+2 2+2 4 4 4 4+4 8 8 8 8+8 16 8+8 16 8+8 8+8 16 16 + 16 16 + 16 32 16 + 16 16 + 16 32 32 32 + 32 32 + 32 32 + 32 HR SO(N ) SO(N ) SO(N ) U (N ) SO(N ) U (N ) SO(N ) SO(N ) U Sp(2N ) U (N ) SO(N ) U Sp(2N ) U Sp(2N ) U (N ) U Sp(2N ) U Sp(2N ) U Sp(2N ) U (N ) U Sp(2N ) SO(N ) SO(N ) SO(N ) U Sp(2N ) U (N ) U Sp(2N ) SO(N ) SO(N ) U (N ) SO(N ) U (N ) SO(N ) SO(N ) . 2) C(7. 1) C(10. 2) C(2. 0) C(5. 2) r 1 3 0 2 1 2 3 0 3 2 1 0 3 2 1 0 3 2 1 3 0 1 0 2 3 3 0 2 1 2 3 0 centralizer R C R R R H C R H C R H H C H H C H H C R R H C H C R R R H C R dSO(p. 2) C(6. 0) C(9. 1) C(8. 0) C(2. 0) C(10. 1) C(9. 1) C(5. 2) C(1. 2) C(10. 2) C(3. q) spinors of maximal real dimension 32 and their Rsymmetry group. dC 1 2 2 2 2 4 4 4 4 4 4 8 8 8 8 16 16 16 16 16 16 16 32 32 32 32 32 32 32 64 64 64 C(p. 1) C(2. 0) C(8. 0) C(3. 1) C(0.24 Unity from Duality: Gravity.

this correspondence is related to the isomorphism C(p.. As the table shows. 1). 3. 1) . but one should consult the table for speciﬁc cases. 7. The correspondence relates spinors of SO(p. 4. have already been discussed in Section 2. D = 2. a single Minkowski spinor can not be elevated to anti-de Sitter space. This can be understood from the fact that the Cliﬀord algebra representations are irreducible with respect to SO(p. For D = 5 the R-symmetry reduces to U (N ). indeed. an underlying isomorphism for the corresponding Cliﬀord algebras is lacking.3. which are proportional to Γ[a Γb] . 1) the R-symmetry group is larger and equal to U Sp(2N ). 10. and one must at least start from an even number of ﬂat Minkowski spinors (so that N is even). 0) with p > 1. the Rsymmetry group is not necessarily the same. Therefore theories formulated in ﬂat Minkowski spacetime of dimension D = 3. as it does for p = 3 and 10. 2) by including the gamma matrices Γa . while for D = 3.13). as it is based on the isomorphism (2. 9. 8. The third correspondence relates spinors of SO(p. However.B. In a number of cases (i. 2) . 6 the R-symmetry remains the same. Physically this correspondence is relevant when considering relativistic massive spinors in ﬂat Minkowski spacetime of dimension D = p + 1. For these cases. 4. In the remaining dimensions. 1) and SO(p. to those of SO(p. 1). the nonrelativistic automorphism group has a tendency to increase. 1) is isomorphic with the real 2 × 2 matrices. 1) with p > 1. 6 can in principle be elevated to anti-de Sitter space. 2) and SO(p. Again the situation depends sensitively on the value for p. so that one can always extend the generators of SO(p. Inspired by the ﬁrst correspondence one may investigate a second one between spinors of SO(p. equivalently. which transform in the restframe under p-dimensional spin rotations. For p = 3 (or. p = 2. 2) spinors allow the R-symmetry group U (N ). The fourth correspondence is again more systematic. 6) the spinor dimension is the same for both groups. p = 4.e. Because the number of irreducible spinors is then doubled. the relativistic spinor decomposes into two nonrelativistic spinors. but not for p = 5 and 9. = (2. 6. D = 4) the implications of the fact that the nonrelativistic automorphism group U Sp(2N ) is bigger than the relativistic one. q) ∼ C(p − 1. The results of the table should therefore be applied with care. In that case the dimension of the automorphism group HR can increase. de Wit: Supergravity 25 mathematical viewpoint. while for SO(4. 4.13) where we note that C(1. 7. it is hard to make general statements about the fate of the R-symmetry when moving to anti-de Sitter space and one has to consult Table 9. For p = 4 the SO(4. In the remaining cases. q − 1) ⊗ C(1. In a number of cases the relativistic spinor transforms irreducibly under the nonrelativistic rotation group. this correspondence is less systematic and. 5. 8 (always modulo 8).

but now with opposite three-momentum −p. 4)p . We return to the superconformal invariance and related aspects in Section 7. The anticommutator of two S-supersymmetry charges yields the conformal boosts. Hence we have a ﬁeld theory with Q = 16 supersymmetries. We start with an N = 4 supersymmetric gauge theory in D = 4 spacetime dimensions. Observe that SO(p. and the entries in the parentheses denote the helicity and the SU (4) representation of the states. (2. so that they can be characterized by their spin. whose massless states are characterized as representations of the SO(2) helicity group and the R-symmetry group SU (4)9 .and S-supersymmetry. It is illuminating to exploit some of the previous correspondences and the relations between various supersymmetry representations in the context of the so-called adS/CFT correspondence [28]. 2 2 where p indicates the three-momentum. 9 Because this multiplet is CPT self-conjugate. This feature is well known [27] and the two supersymmetries are called Q. 1)p + (0. 2) can be regarded as the group of conformal symmetries in a Minkowski space of p dimensions. It implies that the extension of the Poincar´ superalgebra in e D = p spacetime dimensions to a superconformal algebra requires a doubling of the number of supercharges. |p| the energy. or as the isometry group of an anti-de Sitter space of dimension p + 1. . 6)p + ( 1 . 4)p + (− 1 . The resulting supermultiplet consists of 128 + 128 degrees of freedom. the composite multiplet is now a full supermultiplet subject to 16 supersymmetries and the rotation (rather than the helicity) group.14) were only subject to the helicity group and 8 supersymmetries.26 Unity from Duality: Gravity. yields a multiplet with zero momentum and with mass M = 2|p|. 1). of which only 8 are realized on the massless supermultiplet.14) (±1. This correspondence extends this statement to the level of spinors. In the anti-de Sitter context. This supermultiplet decomposes as follows. We close this section by exhibiting a chain of relationships between various supermultiplets. Both set of charges transform under the R-symmetry group of the Poincar´ e algebra. the spinor charge is irreducible but has simply twice as many components. Multiplying this multiplet with a similar one. Gauge Theory and Strings and of SO(p − 1. While the states of the original multiplet (2. For all p > 1 the spinor dimension diﬀers by one-half while the R-symmetry group remains the same. which plays a more basic role in the superconformal algebra as its generators appear in the anticommutator of a Q-supersymmetry and an S-supersymmetry charge. the U (1) subgroup of U (4) coincides with the helicity group and plays no independent role here. As it turns out the helicity states can now be assembled into states that transform under the 3-dimensional rotation group.

divergence-free) and traceless. The spin-0 operators are scalar composite operators. which is conserved (i. Because neither the currents nor the conformal supergravity ﬁelds are subject to any ﬁeld equations (unlike the supersymmetric gauge multiplet from which we started. The N = 4 conformal supergravity theory couples consistently to the N = 4 supersymmetric gauge theory. a fully supersymmetric ground state leads to a U (4) R-symmetry group. and 6 selfdual antisymmetric tensors. which couples to the ﬁelds of conformal supergravity. anti-de Sitter space leads to “remarkable representations”.e. which constitutes only an on-shell supermultiplet).3 . One then ﬁnds that the spin-2 operators correspond to the energy-momentum tensor. associated with the currents of SU (4).2 operators. These are the supersymmetry currents. the R-symmetry group coincides with the U Sp(8) R-symmetry of the relativistic 5-dimensional supersymmetry algebra. inspection shows that this composite multiplet is precisely the N = 4 massive multiplet shown in Table 7. Hence this is the same multiplet that describes D = 5 maximal supergravity. the states are annihilated by half the supercharges and are still classiﬁed according to SO(3). The presence of the traceless and conserved energy-momentum tensor and supersymmetry currents. As we will discuss in Section 6. The spin-1 operators decompose into 15 conserved vectors. Section 7 will further explain the general setting of superconformal theories that is relevant in this context. The oﬀ-shell N = 4 conformal supergravity multiplet in 4 dimensions can also be interpreted as an on-shell massless supermultiplet in 5 dimensions with 32 supersymmetries. it forms the basis for a proper oﬀ-shell theory of N = 4 conformal supergravity [22]. According to Table 9. In this form the relevant R-symmetry group is extended to U Sp(8).B. This is precisely the supermultiplet of currents [22]. and of the SU (4) conserved currents. These are the singletons. 5 for a discussion of this) in which case it can possess an anti-de Sitter ground state. 2 1 Finally there are 20 and 4 chiral and antichiral spin. which now acts as the helicity group. is a consequence of the superconformal invariance of the underlying 4-dimensional gauge theory. which are conserved and traceless (with respect to a contraction with gamma matrices) such that each of them correspond precisely to the 4 components appropriate for spin. Because of the masslessness. de Wit: Supergravity 27 Indeed. Furthermore there are 4 chiral and 4 antichiral vector-spinor operators. so that it has precisely the 5 independent components appropriate for spin-2. The latter theory can be gauged (we refer to Sect. This theory has a nonlinearly realized E6(6) invariance whose linearly realized subgroup (which is relevant for the spectrum) equals U Sp(8). It is possible to cast the above product of states into a product of ﬁelds of the 4-dimensional gauge theory. which do .

We stress that the above excursion...28 Unity from Duality: Gravity. 3 Supergravity In this section we discuss ﬁeld theories that are invariant under local supersymmetry. the invariance under local supersymmetry implies the invariance under spacetime diﬀeomorphisms. with and without a cosmological term. Gauge Theory and Strings not have a smooth Poincar´ limit because they are associated with possible e degrees of freedom living on a 4-dimensional boundary. a. µ. Therefore these theories are necessarily theories of gravity. At the end we brieﬂy discuss some of the nonmaximal theories 3. while the spin-connection ﬁeld is associated with (local) Lorentz transformations of these frames. linking the various supermultiplets in diﬀerent dimensions by a series of arguments. . Then we concentrate on maximal supergravity theories in various dimensions. this gauging allows for an anti-de Sitter max= imally supersymmetric ground state [29]. We exhibit the initial steps in the construction of a supergravity theory. Because of the underlying supersymmetry algebra. b. Thus.1 Simple supergravity The ﬁrst steps in the construction of any supergravity theory are usually based on the observation that local supersymmetry implies the invariance under general coordinate transformation. . . At this stage we thus have to content ourselves with the existence of this remarkable chain of correspondences. This set-up requires the gauge group of 5-dimensional supergravity to be chosen such as to preserve the relevant automorphism group. The vielbein deﬁnes a local set of tangent frames of the spacetime manifold. . is purely based on symmetries. It does not capture the dynamical aspects of the adS/CFT correspondence and has no bearing on the nature of the gauge group in 4 dimensions. their symmetries. Many aspects of these correspondences will reappear in later sections. Therefore one must introduce the a ﬁelds needed to describe general relativity. The world indices. ν. Because the 5dimensional anti-de Sitter superalgebra coincides with the 4-dimensional superconformal algebra. both run from 0 to D−1. and dimensional compactiﬁcations on tori. the circle closes. the 4-dimensional boundary theory must be consistent with superconformal invariance. . Indeed. The vielbein ﬁeld is nonsingular and its inverse µ is denoted by ea . Therefore the gauge group must be equal to SO(6) ∼ SU (4)/Z2 . For an introduction . and the tangent space indices. . namely a vielbein ﬁeld eµ and a ab spin-connection ﬁeld ωµ . Hence it does not come as a surprise that these singleton representations coincide with the supermultiplet of 4dimensional N = 4 gauge theory.

Γ[µν] . µ ab R(e. which carry both a world index and a spinor index and which act as the gauge ﬁelds associated with local supersymmetry. so that P and M denote the translation and the Lorentz generators of the Poincar´ algebra.5) 10 For deﬁniteness we consider a generic supergravity theory with one Majorana gravitino with an antisymmetric charge-conjugation matrix C and gamma matrices Γa satisa fying CΓa C −1 = −ΓT . Hence any supergravity Lagrangian is expected to contain the Einstein-Hilbert Lagrangian of general relativity and the Rarita-Schwinger Lagrangian for the gravitino ﬁeld.4) a ab It is suggestive to regard eµ and ωµ as the gauge ﬁelds of the Poincar´ e group.1) (3. a ab D[µ (ω)Rνρ] (P ) + R[µν (ω) eρ] b = 0 . (3.3) We note that these tensors satisfy the Bianchi identities. Observe that the spinor covariant derivative on ψµ contains no aﬃne connection. In that context R(ω) is written as R(M ). upon using the vielbein postulate. (3. a a a Rµν (P ) = Dµ (ω)eν − Dν (ω)eµ . ab D[µ (ω)Rνρ] (ω) = 0 . ¯ κ2 L = − 1 e R(ω) − 1 eψµ Γµνρ Dν (ω)ψρ + · · · . as it should [30]. κ2 is related to Newton’s constant and a e = det(eµ ). Furthermore Γµ = eµ Γa . ω) = ea ebν Rµν (ω) . denoted by ψµ . but this restriction is not essential10 . a .2) and ωµ ab is the spin-connection ﬁeld whose deﬁnition will be discussed in a sequel. (3. ab ab ab ac b ac b Rµν (ω) = ∂µ ων − ∂ν ωµ + ωµ ων c − ων ωµ c .B. namely the curvature ab a associated with the spin connection Rµν (ω) and the torsion tensor Rµν (P ). For simplicity we only consider a single Majorana gravitino ﬁeld. Γb ] are the generators of the Lorentz 2 4 transformations in spinor space. We note the existence of two covariant tensors. We will e use this notation in later sections when discussing the anti-de Sitter and the conformal algebras. which carries this name because it is proportional to the antisymmetric part ρ of the aﬃne connection. ω) = ebν Rµν (ω) . 4 (3. The matrices 1 Γab = 1 [Γa . Here we will just use the notation R(ω) and deﬁne its contractions with the inverse vielbeine (related to the Ricci tensor and Ricci scalar) by a ab Rµ (e. 2 2 where the covariant derivative on a spinor ψ reads ab Dµ (ω)ψ = ∂µ − 1 ωµ Γab ψ . Furthermore one needs one or several gravitino ﬁelds. de Wit: Supergravity 29 to the vielbein formalism we refer to [30].

(3. The action corresponding to the above Lagrangian is locally supersymmetric up to terms cubic in the gravitino ﬁeld. Let us now include a cosmological term into the above Lagrangian as well as a suitably chosen masslike term for the gravitino ﬁeld. ω) − 1 eψµ Γµνρ Dν (ω)ψρ 2 2 ¯ + 1 g(D − 2)e ψµ Γµν ψν + 1 g 2 (D − 1)(D − 2) e + · · · (3. When the spacetime dimension exceeds eleven. up to terms that are cubic in the gravitino ﬁeld. (3. 2 (3. (3.11) .7) for the spin connection. Gauge Theory and Strings The spin connection can be treated as an independent ﬁeld (ﬁrst-order formalism). For pure gravity the ﬁrst. Extending this Lagrangian to a fully supersymmetric one is not always possible. δeµ a = 1 ¯ Γa ψµ .8) From the spin connection one deﬁnes the aﬃne connection by Γµνρ = ρ ea Dµ (ω)eνa . for instance. L ¯ = − 1 e R(e. or it can be ﬁxed from the beginning (second-order formalism). µ ν c c Ωab c = ea eb (∂µ eν − ∂µ eν ) . 2 δψµ = Dµ (ω) . the aﬃne connection ab becomes equal to the Christoﬀel symbols and Rµνρ σ = Rµν (ω) eρ a ebσ coincides with the standard Riemann tensor.30 Unity from Duality: Gravity.10) 4 2 As it turns out the corresponding action is still locally supersymmetric.6) Such a constraint is called “conventional” because it expresses one ﬁeld in terms of other ﬁelds in an algebraic fashion. conventional supergravity no longer exists.9) where the gravitino variation is the extension to curved spacetime of the spinor gauge invariance of a Rarita-Schwinger ﬁeld.7) where the Ωab c are the objects of anholonomity. Usually it requires additional ﬁelds of lower spin. provided that we introduce an extra term to the transformation rules. (3. δeµ a = 1 ¯ Γa ψµ . which ensures the validity of the vielbein postulate.6) can be solved algebraically and leads to. a Rµν (P ) = 0 . whose existence can be inferred from the knowledge of the possible underlying (massless) supermultiplets of states. The constraint (3. as we shall discuss in the next section. 2 δψµ = Dµ (ω) + 1 gΓµ 2 . by imposing the constraint.and the second-order formulism lead to the same result. ab c ωµ (e) = 1 eµ (Ωab c − Ωb c a − Ωc ab ) . which is then solved in terms of its ﬁeld equations. The supersymmetry transformations contain the terms. With the zero-torsion value (3.

simple supergravity in D = 10 dimensions cannot possibly have (anti-)de Sitter ground states.e.B. see [33. This example does not cover all cases. i.10) reads (suppressing the gravitino ﬁeld). one may verify whether the Minkowski spinors have the right dimension to enable them to live in these spaces.. For instance.13) (3. a MajoranaWeyl spinor in D = 10 spacetime dimensions has only half the number of components as compared to a spinor in (anti-)de Sitter space of the same dimension.10) was ﬁrst written down in [31] in four space-time dimensions and the correct interpretation of the masslike term was given in [32]. Assuming that the theory has an anti-de Sitter or de Sitter ground state. as it implies that the cosmological term is of deﬁnite sign. but there are also explicit studies ruling out supersymmetric cosmological terms in 11 dimensions [35]. Rµν − 1 gµν R + 1 g 2 (D − 1)(D − 2) gµν = 0 . 2 . Here the argument may be invoked that no relevant supersymmetric extension of the anti-de Sitter algebra exists beyond D = 7 dimensions [3]. For an early discussion. if the ground state is to preserve supersymmetry). R = g 2 D(D − 1) . The Einstein equation corresponding to (3. Hence supersymmetry does not a priori forbid a cosmological term. We should point out that there are situations where a cosmological term is not consistent with supersymmetry. Such a counting argument does not exclude anti-de Sitter ground states in D = 11 spacetime dimensions. 1 ¯ − 2 eψµ Γµνρ (Dν (ω) + 1 gΓν )ψρ . de Wit: Supergravity 31 The Lagrangian (3. Consistency requires that gΓµ satisﬁes the same Majorana constraint as ψµ and . where ωµ and eµ act as gauge ﬁelds11 . The reality of g has important consequences. (3. Nevertheless the conclusion that the cosmological term must have this particular sign remains. With the conventions that we have adopted this implies that g is real. but it must be of deﬁnite sign (at least. Rµν = g 2 (D − 1) gµν . Therefore. Observe that the variation for ψµ may be regarded as a generalized ab a covariant derivative.10) is consistent with that interpretation as it can be generated from the Rarita-Schwinger Lagrangian by the same change of the covariant derivative. 2 2 which implies. because D = 11 Lorentz spinors can exist in anti-de Sitter space. 34] and references therein. as one does not always have a single Majorana spinor with the speciﬁed charge conjugation properties. unless one accepts “ghosts”: ﬁelds whose kinetic terms are of the wrong sign.12) 11 The masslike term in (3.

32 Unity from Duality: Gravity.15) Spinors satisfying this equation are called Killing spinors.16) from which one derives that the Riemann tensor satisﬁes (3. (3. For instance.17) = 0. Since (3. while minimal supergravity in D = 4 dimensions does not require additional ﬁelds.15) is a ﬁrst-order diﬀerential equation. implies (3.2 Maximal supersymmetry and supergravity In Section 2 we restricted ourselves to supermultiplets based on Q ≤ 32 supercharge components. We stress once more that. Requiring full supersymmetry. The maximally symmetric solution of this equation is an anti-de Sitter space.14) so that the spinor must live in anti-de Sitter space. We will return to this issue later in Section 6. To see this one notes that also (Dµ (ω) + 1 gΓµ )(Dν (ω) + 1 gΓν ) must vanish. This happens for spinors (x) satisfying Dµ (ω) + 1 gΓµ 2 = 0. 3. The need for certain extra ﬁelds can be readily deduced from the underlying massless supermultiplets. (3. in this section. we have restricted ourselves to the graviton-gravitino sector. so that (3. in D = 11 dimensions an additional antisymmetric gauge ﬁeld is necessary.14) This solution leaves all supersymmetries intact just as ﬂat Minkowski space does.13). which were extensively discussed in the previous section. Therefore supersymmetry requires an Einstein space. one expects that it can be solved provided some integrability condition is satisﬁed. Gauge Theory and Strings Hence we are dealing with a D-dimensional Einstein space. where we discuss the (super)multiplet structure in anti-de Sitter space. To construct the full theory usually requires more ﬁelds and important restrictions arise on the dimensionality of spacetime. (3. whose Riemann curvature equals Rµνab = 2g 2 eµ[a eν b] . (3. Antisymmetrizing this 2 2 expression in µ and ν then yields the (algebraic) integrability condition − 1 Rµνab Γab + 1 g 2 Γµν 4 2 Multiplication with Γν yields Rµν − g 2 (D − 1) gµν Γν = 0 . From the general analysis it is clear that increasing the number of supercharges leads to higher and higher helicity . One can verify this directly by considering the supersymmetry variation of the gravitino ﬁeld and by requiring that it vanishes in the bosonic background.15) holds for any spinor . Hence we have seen that supersymmetry can be realized in anti-de Sitter space.

The fact that an inﬁnite number of ﬁelds can cure certain inconsistencies is by itself not new. conventional supergravity theories are not of this kind. it turns out that symmetric gauge ﬁelds cannot consistently couple. However. 24 and 32 [8]. [39]): (i) an inﬁnite tower of higher-spin gauge ﬁelds. but not with other ﬁelds of the same spin [37]. For instance. de Wit: Supergravity 33 representations. (ii) interactions that are inversely proportional to the cosmological constant. By consistent. An exception is the graviton ﬁeld.g. There is also direct evidence in D = 3. of a massless supermultiplet in D = 4 spacetime dimensions is larger than or equal to 1 16 Q. In 2 view of the supersymmetry algebra an interacting supersymmetric theory of this type should contain gravity. but in higher dimensional spacetimes one expects to arrive at the same conclusions. for instance [36]). upon dimensional reduction. 20. (iii) extensions of the super-Poincar´ or the super-de Sitter algebra with additional fermionic e and bosonic charges. when Q > 8 we have |λmax | ≥ 1. so that theories for these multiplets must include vector gauge ﬁelds. which can interact with itself as well as with low-spin matter. While a massive spin-2 ﬁeld cannot be coupled to gravity. see e. There have been many eﬀorts to circumvent this bound of Q = 32 supersymmetries. However. Those are described by gauge ﬁelds that are symmetric Lorentz tensors. so that the theory should contain Rarita-Schwinger ﬁelds. we mean that the respective gauge invariances of the higher-spin ﬁelds (or appropriate deformations thereof) cannot be preserved at the interacting level. where graviton and gravitini ﬁelds do not describe dynamic degrees of freedom.e. neither to themselves nor to other ﬁelds. when coupling matter to this theory in the form of scalars and spinors. . |λmax |. Most of the search for interacting higher-spin ﬁelds was performed in 4 spacetime dimensions [38]. which are topological. as for Q > 32 no interacting ﬁeld theories seem to exist. Hence the conclusion is that there is a restriction on the number Q of independent supersymmetries.B. the maximal helicity. these theories would give rise to theories that are consistent in D = 4. Hence. one can write down supergravity theories based on a graviton ﬁeld and an arbitrary number of gravitino ﬁelds. Indeed explicit theories have been constructed which demonstrate this. When Q > 16 we have |λmax | ≥ 3 . However. so that in this case we must include λ = 2 states for the graviton. see. It seems clear that one needs a combination of the following ingredients in order to do this (for a review. in Table 5) to list supermultiplets with states transforming in higher-helicity representations. Symmetric tensor gauge ﬁelds for arbitrary helicity states can be constructed (in D = 4 dimensions. the theory does not support more than 32 supercharges. Therefore. Beyond Q = 16 there are four unique theories with Q = 18. This is the reason why we avoided (i. because otherwise. Beyond Q = 32 one is dealing with states of helicity λ > 2.

D 11 10A 10B 9 8 7 6 5 4 3 HR 1 1 SO(2) SO(2) U (2) U Sp(4) U Sp(4) ×U Sp(4) U Sp(8) U (8) SO(16) graviton 1 1 1 1 1 1 1 1 1 1 p = −1 0 1 2 2+1 5+1+¯ 1 14 (5. as can be seen in Kaluza-Klein theory. Bosonic ﬁeld content for maximal supergravities. For p > 0 the ﬁelds can be assigned to representations of a bigger group than HR . The representations [1] and [28] (in D = 8. The highest dimension D = 11 is motivated by the fact that spinors have more that 32 components in ﬂat Minkowski space for spacetime dimensions D > 11. The fact that no uniform decompactiﬁcation is possible is closely related to the T -duality between winding and momentum states that one knows from string theory. Observe. there are scenarios based on spacetime dimensions higher than D = 11. however. The p = 3 gauge ﬁeld in D = 10B has a self-dual ﬁeld strength. 5) 42 35 + 35 128 p=0 0 1 0 2+1 3+¯ 3 10 (4. 1) +(1. but it turns out that the interactions often prefer the rank of the gauge ﬁeld to be as small as possible. 5) p=2 1 1 0 1 [1] p=3 0 0 1∗ the coupling of an inﬁnite number of them can be consistent. Therefore. In D = 3 dimensions. the graviton does not describe propagating degrees of freedom. This will be discussed in due course. These theories have Q = 32 supersymmetries and we restrict our discussion to 3 ≤ D ≤ 11. These transformations cannot be represented on the vector potentials. In this section we review the maximal supergravities in various dimensions.41]. respectively) are extended to U (1) and SU (8) representations through duality transformations on the ﬁeld strengths. For the antisymmetric gauge ﬁelds. . 4.34 Unity from Duality: Gravity. Gauge Theory and Strings Table 10. 4) 27 [28] p=1 0 1 2 2 3 5 (5. The bosonic ﬁelds always comprise the metric tensor for the graviton and a number of (p + 1)-rank antisymmetric gauge ﬁelds. where the extra dimensions can not uniformly decompactify so that the no-go theorem is avoided. that this argument assumes D-dimensional Lorentz invariance. As was stressed in [40. it is a priori unclear whether to choose a (p+1)-rank gauge ﬁeld or its dual (D − 3 − p)-rank partner.

D 11 10A 10B 9 8 7 6 5 4 3 HR 1 1 SO(2) SO(2) U (2) U Sp(4) U Sp(4)×U Sp(4) U Sp(8) U (8) SO(16) gravitini 1 1+1 2 2 2+¯ 2 4 (4. as we will explain in a sequel. Actually. 7 the fermion ﬁelds are counted as symplectic Majorana spinors. Obviously. For D = 4. 3. as in D = 11 dimensions.3 D = 11 supergravity A Supergravity in 11 spacetime dimensions is based on an “elfbein” ﬁeld EM . Fermionic ﬁeld content for maximal supergravities. 4) 8 8+¯ 8 16 spinors 0 1+1 2 2+2 2+¯+4+¯ 2 4 16 (4. always consisting of gravitini and simple spinors. For D = 5. Its Lagrangian can be written as . 6. the maximal supergravity theories have symmetry groups that are much larger than HR .B. 5) + (5. it is the only Q ≥ 16 supergravity theory without a scalar ﬁeld. p = 3 and p = 4 are each other’s dual conjugates. we restrict ourselves to p ≤ 3. which transform in conjugate representations of HR . In Table 11 we also present the fermionic ﬁelds. 8 we include both chiral and antichiral spinor components. 4) 48 56 + 56 128 in Table 10. a Majorana gravitino ﬁeld ΨM and a 3-rank antisymmetric gauge ﬁeld CMN P . the problematic higher-spin ﬁelds are avoided. 0) supergravity in 6 dimensions. de Wit: Supergravity 35 Table 11. This table presents all the ﬁeld conﬁgurations for maximal supergravity in various dimensions. All these ﬁelds are classiﬁed as representations of the R-symmetry group HR . because the only symmetric gauge ﬁeld is the one describing the graviton. In D = 3 dimensions the gravitino does not describe propagating degrees of freedom. With chiral (2. 1) + (1. Note that the simplest versions of supergravity (which depend on no other coupling constant than Newton’s constant) are manifestly invariant under HR .

18) are supercovariant. E = det EM and ΩMAB denotes the spin connection. 6 (3.19) ˆ FN P QR .36 Unity from Duality: Gravity. 2 (3. whereas its supersymmetry variation in the action is treated as if it were an independent ﬁeld [42].18) by replacing the spinˆ connection ﬁeld Ω by (Ω + Ω)/2 in the covariant derivative of the gravˆ itino kinetic term and by replacing FMN P Q in the last line by (FMN P Q + FMN P Q )/2.22) N] 4 The left-hand side is the supercovariant torsion tensor. Ω) − 1 E ΨM ΓMN P DN (Ω)ΨP − 48 E(FMN P Q )2 2 2 κ2 11 √ 1 − 3456 2 εMN P QRST UV W X FMN P Q FRST U CV W X (3. . √ = − 1 2 ¯ Γ[MN ΨP ] . FP QRS F P QRS − 1 FMP QR FNP QR .5-order” formalism. ¯ ˆ D[M (Ω) E A − 1 ΨM ΓA ΨN = 0 . Here the covariant derivative is covariant with respect to local Lorentz transformation DM (Ω) = ∂M − 1 ΩM AB ΓAB . We have the following bosonic ﬁeld equations and Bianchi identities. These substitutions ensure that the ﬁeld equations corresponding to (3. RMN ∂M E F MN P Q ∂[M FN P QR] = = = 1 72 gMN 1 1152 √ N P QRST UV W XY 2ε FRST U FV W XY . L11 = 1 1 ¯ − 1 E R(E. (3. The Lagrangian is derived in the context of the so-called “1. The supersymmetry transformations are equal to A δEM = δCMN P δΨM ¯ ΓA Ψ M . Note the presence of a Chern-Simons-like term F ∧ F ∧ C in the Lagrangian. (3. in which the spin connection is deﬁned as a dependent ﬁeld determined by its (algebraic) equation of motion. 8 √ N 1 ˆ = DM (Ω) + 288 2 ΓMN P QR − 8 δM ΓP QR 1 2 (3.21) The supercovariant spin connection is the solution of the following equation.23) 0. We also wish to point out that the quartic-Ψ terms can be included into the Lagrangian (3. Gauge Theory and Strings follows [11].20) 4 ˆ and FMN P Q is the supercovariant ﬁeld strength √ ˆ ¯ FMN P Q = 24 ∂[M CN P Q] + 3 2 Ψ[M ΓN P ΨQ] . so that the action is only invariant under tensor gauge transformations up to surface terms. A where the ellipses denote terms of order Ψ4 .18) √ 1 ¯ ¯ − 192 2E ΨR ΓMN P QRS ΨS + 12 ΨM ΓN P ΨQ FMN P Q + · · · .

κ2 → e−9α κ2 .26) Under this rescaling the Lagrangian changes according to L11 → e−9α L11 . These are just the charges that can appear as central charges in the supersymmetry algebra (2. respectively. (3. (3. CMN P → e−3α CMN P . The spatial volumes are orthogonal to a p = 2 and a p = 5 brane conﬁguration. . HMN P QRST = A A EM → e−α EM .23) and (3.B.12).29) 12 Note that the rescalings also leave the supersymmetry transformation rules unchanged. √ 1 E εMN P QRST UV W X F UV W X − 1 2 F[MN P Q CRST ] . provided the supersymmetry parameter is changed accordingly. respectively. This change can then be absorbed into a redeﬁnition of κ11 12 . de Wit: Supergravity 37 which no longer depend explicitly on the antisymmetric gauge ﬁeld.28) (3.25) One could imagine that the third equation of (3. the constant 1/κ2 in front of the Lagrangian (3. is undetermined and depends on ﬁxing some length scale. The same situation is present in many other supergravity theories. which 11 carries dimension [length]−9 ∼ [mass]9 . 2 7! (3.24) ΨM → e−α/2 ΨM . Finally. D D (3. κ2 → e(2−D)α κ2 . Concentrating on the Einstein-Hilbert action in D spacetime dimensions. In analogy with the Maxwell theory. where HMN P QRST is the dual ﬁeld strength. LD → e(2−D)α LD . An alternative form of the second equation is [43] ∂[M HN P QRST U] = 0 . namely κ11 . the corresponding scaling property is D D gµν → e−2α gµν .27) This simply means that the Lagrangian depends on only one dimensional coupling constant. To see this consider a continuous rescaling of the ﬁelds.and a 5-dimensional spatial volume. and the corresponding charges are 2.18). These charges are associated with the “ﬂux”-integral of HMN P QRST and FMN P Q over the boundary of an 8. 11 11 (3.and 5-rank Lorentz tensors. the integral H may be associated with electric ﬂux and the integral F with magnetic ﬂux.24) receive contributions from charges that would give rise to source terms on the right-hand side of the equations. Solutions of 11-dimensional supergravity that contribute to these charges were considered in [44–46].

. this is usually not the case and the number of independent supersymmetries will be reduced. D − 1. . In the latter case the resulting 4-dimensional supergravity theory acquires a cosmological term. . the resulting supergravity has the same number of supersymmetries as the original one. We denote the compactiﬁed coordinate by x10 which now parameterizes a circle of length L13 . the other one the compactiﬁcation on a sphere S 7 .4 Dimensional reduction and hidden symmetries The maximal supergravities in various dimensions are related by dimensional reduction. It is a key element in ensuring that solutions of the lowerdimensional theory remain solutions of the original higher-dimensional one. 3. . The aim of the present discussion here is to elucidate a number of features related to these symmetries. Fully supersymmetric compactiﬁcations are rare. In the formulation of the compactiﬁed theory. this implies that the physical value of Newton’s constant. This ensures that various complicated mixtures of massless modes with the tower of massive modes will be avoided. In this reduction some of the spatial dimensions are compactiﬁed on a hypertorus and one retains only the ﬁelds that do not depend on the torus coordinates. mainly in the context of the reduction of D = 11 supergravity to D = 10 dimensions. it is important to decompose the higher-dimensional ﬁelds in such a way that they transform covariantly under the lower-dimensional gauge symmetries and under diffeomorphisms of the lower-dimensional spacetime. The massless modes form the basis of the lower-dimensional supergravity theory. does not necessarily coincide with the parameter κ2 in the Lagrangian but it also D depends on the precise value adopted for the (ﬂat) metric in the ground state of the theory.38 Unity from Duality: Gravity. The ﬁelds are thus decomposed in a Fourier series as periodic functions in x10 on the interval 0 ≤ x10 ≤ L. A subset of the gauge symmetries associated with the compactiﬁed dimensions survive as internal symmetries. This corresponds to the theory one obtains when the size of the torus is shrunk to zero. For instance. Because a toroidal background does not break supersymmetry. For compactiﬁcations on less trivial spaces than the hypertorus. One is the compactiﬁcation on a torus T 7 . Gauge Theory and Strings Of course. 1. . This results in a spectrum of massless modes and an inﬁnite tower of massive modes with masses inversely proportional to the circle length L. 11-dimensional supergravity can be compactiﬁed to a 4-dimensional maximally symmetric spacetime in only two ways such that all supersymmetries remain unaﬀected [47]. which is an obvious requirement for having consistent truncations to the 13 Throughout these lectures we enumerate spacetime coordinates by 0.

Because these originate from supermultiplets that are massless in higher dimensions. which deals with isometries in N = 2 supersymmetric Maxwell-Einstein theories in D = 5. Our derivation here was alluded to in [49]. (3.32) Another symmetry of the lower-dimensional theory is more subtle to identify14 . The emergence of new internal symmetries in theories that originate from a higher-dimensional setting. In the compactiﬁed situation we can also involve the compactiﬁcation length into the dimensional scaling. Let us start with the symmetries associated with the metric tensor. The 11-dimensional metric can be decomposed according to ds2 = gµν dxµ dxν + e4φ/3 (dx10 + Vµ dxµ )(dx10 + Vν dxν ) . [9.g. The massless modes correspond to the x10 -independent parts of the 10-dimensional metric gµν . Following the discussion in [49] we distinguish between symmetries that have a direct explanation in terms of the higher dimensional symmetries.31) The massive modes.6. and symmetries whose origin is obscure from a higher-dimensional viewpoint. Implications of these BPS supermultiplets will be discussed in more detail in Section 3. Its existence in 10-dimensional supergravity was noted long ago (see. leading to Vµ (x) → Vµ (x) + ∂µ ξ(x) . couple to this gauge ﬁeld with a charge that is a multiple of eKK = 2π · L (3. e.B. Another point of interest concerns the nature of the massive supermultiplets. they are 1/2-BPS multiplets which are shortened by the presence of central charges corresponding to the momenta in the compactiﬁed dimension. which did not leave the theory invariant but could be used to adjust the coupling constant κ10 . 4 and 3 dimensions. Speciﬁcally. the vector ﬁeld Vµ and the scalar φ. Here the x10 -independent component of Vµ acts as a gauge ﬁeld associated with reparametrizations of the circle coordinate x10 with an arbitrary function ξ(x) of the 10 remaining spacetime coordinates xµ .30) where the indices µ. de Wit: Supergravity 39 massless states. which correspond to the nontrivial Fourier modes in x10 . . ν label the 10-dimensional coordinates and the factor multiplying φ is for convenience later on. The integration over x11 introduces an overall factor L in the action (we do not incorporate any L-dependent normalizations in the 14 There are various discussions of this symmetry in the literature. we have x10 → x10 − ξ(x) and xµ → xµ . is a standard feature of Kaluza-Klein theories [48]. 50]) and an extensive discussion can be found in [18]. In the previous section we noted the existence of certain scale transformations of the D = 11 ﬁelds. (3.

35) to yield an invariance of the Lagrangian. It simply expresses the length of the x10 -periodicity interval.40 Unity from Duality: Gravity.36) The tensor gauge ﬁeld CMN P decomposes into a 3. these combined transformations are given by15 ea → e−α ea . Its purely bosonic terms read L10 = 1 − 1 e e2φ/3 R(e. Stated diﬀerently..35) so that κ10 remains invariant. (3. x10 -independent) modes. which depends on the coordinatization. as long as we change L accordingly. corresponding to a reparametrization of the 11-th coordinate. Therefore. C11µν → e6α C11µν .e. we may rescale L according to (3. Consequently we are then dealing with a symmetry of the Lagrangian. as those depend on x10 /L which is insensitive to the scale transformation. µ µ φ → φ + 12α . which transform according to Cµνρ → e−3α Cµνρ . Gauge Theory and Strings Fourier sums. because the Kaluza-Klein charges (3. For the ﬁelds corresponding to the 11-dimensional metric. (3.33) and is of dimension [mass]8 .6. In the eﬀective 10-dimensional theory.37) The presence of the above scale symmetry is conﬁrmed by the resulting 10dimensional Lagrangian for the massless (i. This does not imply that the Lagrangian remains invariant when retaining the higher Fourier modes. This issue will be relevant in Section 3. κ10 11 (3. ω) − 1 e e2φ (∂µ Vν − ∂ν Vµ )2 2 8 κ2 10 1 − 48 e e2φ/3 (Fµνρσ )2 − 3 e e−2φ/3 (Hµνρ )2 4 √ µ −µ 1 + 1152 2 ε 1 10 C11µ1 µ2 Fµ3 µ4 µ5 µ6 Fµ7 µ8 µ9 µ10 . Vµ → e−9α Vµ .32) depend explicitly on L. the scale transformations (3. we can reparameterize x10 by some diﬀeomorphism.26) are thus suitably combined with the diﬀeomorphism (3.34) L → e−9α L . the coupling constant that emerges in the 10-dimensional theory equals 1 L 2 = κ2 . because of the invariance under diffeomorphisms. (3. In particular. 15 Note that these transformations apply uniformly to all Fourier modes. x10 → e−9α x10 .38) .and a 2-rank tensor in 10 dimensions. L itself has no intrinsic meaning. However. so that the 10-dimensional and the 11-dimensional ﬁelds are directly proportional). (3.

all the scalars that emerge from dimensional reduction of gauge ﬁelds are subject to constant shift . xi → Oi j xj . or one can reduce at once to lower dimensions. The above example exhibits many of the characteristic features of dimensional reduction and of the symmetries that emerge as a result. 11-dimensional supergravity reduced on a hypertorus thus leads to a Lagrangian for the massless sector in lower dimensions (the massive sector is discussed in Sect. which exhibits a number of invariances that ﬁnd their origin in the diﬀeomorphisms and gauge transformations related to the torus coordinates. and which combine with the previous ones to generate the group SO(n. which act transitively on the manifold (i. 2 4 (3. let us brieﬂy discuss an example of the latter based on gravity coupled to an antisymmetric tensor gauge ﬁeld in D + n spacetime dimensions. one must properly account for the periodicity intervals of the torus coordinates xi .e. The scalar ﬁelds originate from the metric and the antisymmetric gauge ﬁeld with both indices taking values in T n . 2 exactly equal to the number of independent ﬁelds gij . see Sect. de Wit: Supergravity 41 where Hµνρ = 6 ∂[µ Cνρ]11 is the ﬁeld strength tensor belonging to the 2-rank tensor gauge ﬁeld. The homogeneous space can then be identiﬁed as the coset space SO(n.37). Furthermore. n).. There are thus 1 2 n(n − 1) independent shift transformations. consecutively reducing the dimension by unit steps. and n2 scalar ﬁelds.39) After compactiﬁcation on a torus T n .6). The group GL(n) contains the rotation group SO(n). one tensor gauge ﬁeld. 2n Abelian vector gauge ﬁelds. The matrices O generate the group GL(n). which can be regarded as a generalization of the scale transformations (3. act on gij and Bij according to g → OT g O and B → OT B O. in total. 2 they leave no point on the manifold invariant). its remaining part depends on 1 n(n + 1) parameters. However. whose origin is 2 not directly related to the higher-dimensional context. As already explained.e. Special tensor gauge transformations with parameters proportional to Λij xj induce a shift of the massless scalars Bij proportional to the constants Λ[ij] . it turns out that there exist 1 n(n − 1) additional isometries. but the action for the massless ﬁelds remains invariant under continuous GL(n) transformations. The diﬀeomorphisms acting on the torus coordinates xi which are linear in xi .36) and (3. Before continuing our general discussion. the ﬁelds that are independent of the torus coordinates remain massless ﬁelds in D dimensions: the graviton. we have now identiﬁed 1 n(3n − 1) isometries.1). When reducing to lower dimension one can follow the same procedure a number of times. According to the above. n)/(SO(n) × SO(n)). so that they are parametrized by a symmetric tensor gij and an antisymmetric tensor Bij . i. so that. L ∝ − 1 E R − 9 E (∂[M BN P ] )2 . Therefore the manifold is homogeneous (for a discussion of such manifolds.B. 3. 4.

In general these extra symmetries are not necessarily symmetries of the full action. In Table 12 we list the isometry group G and the isotropy group HR of these scalar manifolds for maximal supergravity in dimensions 3 ≤ D ≤ 11. 54]. Therefore we expect that the target space for the scalar ﬁelds is an homogeneous space. the rank of the resulting symmetry group in lower dimensions is equal to the rank of GL(n). Here we merely stress a number of characteristic features of the group G. i. another (noncompact) subgroup is the group GL(11 − D). the symmetries may not leave the Lagrangian. and thus to the number of compactiﬁed dimensions. Gauge Theory and Strings transformations. see [51]). n) is the T -duality group. because the maximal supergravity theory that one obtains from compactiﬁcation on a hypertorus has no additional coupling constants (beyond Newton’s constant) which could induce R-symmetry breaking. but only the ﬁeld equations. n). associated with the reduction on an (11 − D)-dimensional torus. which has a manifest SL(2) in D = 10 dimensions. be found in [9. where D is the spacetime dimension to which we reduce. Earlier versions of such tables can. Of course. As we mentioned already. These scalars and the scalars originating from the metric transform transitively under the isometry group. it is easy to see that the scalar manifold (as well as the rest of the theory) possesses additional symmetries beyond the ones that follow from higher dimensions. Yet another subgroup is SL(2) × SO(n. SL(2) is the S-duality group and SO(n. We expect that HR is also realized as a symmetry. to the subalgebra of GL(11 − D) scale transformations and/or to the subalgebra associated with HR . One of them is that HR is always the maximal compact subgroup of G. In even dimensions. n) is associated with . The reason for this is that the isometries may act by means of duality transformations on ﬁeld strengths associated with antisymmetric tensor gauge ﬁelds of rank 1 D − 1 which cannot be implemented on 2 the gauge ﬁelds themselves. A more recent discussion of these isometry groups from the perspectives of string theory and M-theory can be found in. Usually a counting argument (of the type ﬁrst used in [53]) then readily indicates what the structure is of the corresponding homogeneous space that is parametrized by the scalar ﬁelds. these subalgebras will partly overlap. because the latter do not yet incorporate the full R-symmetry group of the underlying supermultiplet. For supergravity. The group SO(n. where n = 10 − D. invariant.e. which emerges for D < 10 can be understood within the string perspective. with an isometry group whose generators belong to a solvable subalgebra associated with the shift transformations. This group. r = 11 − D. 50].42 Unity from Duality: Gravity. for instance. for example [24. It also follows from the toroidal compactiﬁcations of IIB supergravity. for D = 6 we refer to [52]. In 4 dimensions this phenomenon is known as electric-magnetic duality (for a recent review. Since 11-dimensional supergravity has itself no scalar ﬁelds..

while a standard kinetic term for the scalar ﬁeld φ is lacking. In this form the gravitational ﬁeld and the scalar ﬁeld are entangled and one has to deal with the scalar-graviton . but only under the local HR group.38) does not contain the standard Einstein-Hilbert term for gravity.39)). 5) E6(+6) E7(+7) E8(+8) H 1 1 SO(2) SO(2) U (2) U Sp(4) U Sp(4)×U Sp(4) U Sp(8) SU (8) SO(16) dim [G] − dim [H] 0−0=0 1−0=1 3−1=2 4−1=3 11 − 4 = 7 24 − 10 = 14 45 − 20 = 25 78 − 36 = 42 133 − 63 = 70 248 − 120 = 128 the invariance of toroidal compactiﬁcations that involve the metric and an antisymmetric tensor ﬁeld (cf. The scalars transform linearly under both the rigid duality group as well as under the local HR group. The ﬁelds which initially transform only under the local HR group.B. We discuss some of the consequences for the central charges and the BPS states in Section 3. 4 and 5). listed in Table 10. de Wit: Supergravity 43 Table 12. governed by the Cartan-Maurer equations. which transform under the group HR as we have shown in Table 8. This does not pose a serious problem. (3. as this would be in conﬂict with their own gauge invariance. D 11 10A 10B 9 8 7 6 5 4 3 G 1 SO(1. The corresponding connections are then composite connections. The diﬀerence of the dimensions of G and H equals the number of scalar ﬁelds.5 Frames and ﬁeld redeﬁnitions The Lagrangian (3. the G-transformations become realized nonlinearly (we discuss such nonlinear realizations in detail in Sects. This phenomenon is also realized for the central charges. the fermions) do not transform under the group G. In such a formulation most ﬁelds (in particular. the gauge ﬁelds cannot transform under the local group HR . Homogeneous scalar manifolds G/H for maximal supergravities in various dimensions. After ﬁxing a gauge.6. will now transform under the duality group G through ﬁeld-dependent HR transformations. Here we should add that it is generally possible to realize the group HR as a local symmetry of the Lagrangian. 1)/Z2 SL(2) GL(2) E3(+3) ∼ SL(3)×SL(2) E4(+4) ∼ SL(5) E5(+5) ∼ SO(5. The type-IIB theory cannot be obtained from reduction of 11-dimensional supergravity and is included for completeness. 3.

. The corresponding Lagrangian reads16 LEinstein = 10 1 e − 1 R(e. µ µ (3. this rescaling may depend on the extra coordinate x10 . (3. one applies a so-called Weyl rescaling of the metric gµν by an appropriate function of φ.40) We already stressed that the the compactiﬁcation length L is just a parameter length with no intrinsic meaning as a result of the fact that one can always apply general coordinate transformations which involve x10 . Of course.38) one obtains the Lagrangian in the Einstein frame. Gauge Theory and Strings system as a whole. where the isometries of the scalar ﬁelds do not act on the graviton. respectively.41) Supergravity theories are usually formulated in this frame. which in the metric speciﬁed by (3. µ µ ea = e−φ/3 [ea ]string . After applying the ﬁrst rescaling (3. For these lectures two diﬀerent Weyl rescalings are particularly relevant. To separate the scalar and gravitational degrees of freedom. 16 Note that under a local scale transformation ea → eΛ ea . They are deﬁned by ea = e−φ/12 [ea ]Einstein . In the context of Kaluza-Klein theory this factor is known as the “warp factor”. one may also consider the geodesic length.36.37). 1) transformations.30) is equal to L exp[2 φ /3]. which lead to the so-called Einstein and to the string frame. 3.44 Unity from Duality: Gravity. the geodesic length of the 11-th dimension is invariant under the SO(1.40) to the Lagrangian (3. ω) − 1 (∂µ φ)2 − 1 e e3φ/2 (∂µ Vν − ∂ν Vµ )2 2 4 8 κ2 10 3 1 − 4 e e−φ (Hµνρ )2 − 48 e eφ/2 (Fµνρσ )2 √ 1 + 1152 2 εµ1 −µ10 C11µ1 µ2 Fµ3 µ4 µ5 µ6 Fµ7 µ8 µ9 µ10 . Observe that gauge ﬁelds cannot be redeﬁned by these local scale transformations because this would interfere with their own gauge invariance. In the Einstein frame. the Ricci scalar in D µ µ dimensions changes according to R → e−2Λ R + 2(D − 1)D µ ∂µ Λ + (D − 1)(D − 2)g µν ∂µ Λ ∂ν Λ . In the case that we include the massive modes. This frame is characterized by a standard Einstein-Hilbert term and by a graviton ﬁeld that is invariant under the scale transformations (3.

states with R-R charge are solitonic. (3. deﬁned in (3. which reads R10 = L 2 e 2π φ /3 . On the other hand the vector Vµ and the 3-form Cµνρ describe Ramond-Ramond (R-R) states and the speciﬁc form of their vertex operators forbids any tree-level coupling to the dilaton [18.30). as in (3. whereas in the string frame the Ricci scalar is multiplied by a dilaton term exp(−2φD ). 3. For later purposes let us note that the above discussion can be generalized to arbitrary spacetime dimensions.37).40) leads to the Lagrangian in the string frame. The Weyl rescaling which connects the two frames is given by. or. . the antisymmetric tensor Cµν11 and the dilaton φ always arise in the Neveu-Schwarz sector and couple universally to e−2φ . In particular the Kaluza-Klein gauge ﬁeld Vµ corresponds in the string context to the R-R gauge ﬁeld of type-II string theory.B.44) This result applies to the frame speciﬁed by the 11-dimensional theory17 .42).38). which leads to the Lagrangian (3. R-R charges are carried by the D-brane states.42) This frame is characterized by the fact that R and (Hµνρ )2 have the same coupling to the scalar φ. As we stressed already.32). The metric gµν .41) and (3. ω) + 2(∂µ φ)2 − 3 (Hµνρ )2 2 4 κ2 10 1 − 1 e (∂µ Vν − ∂ν Vµ )2 − 48 e (Fµνρσ )2 8 √ 1 + 1152 2 εµ1 −µ10 C11µ1 µ2 Fµ3 µ4 µ5 µ6 Fµ7 µ8 µ9 µ10 . In the context of 10-dimensional supergravity. The inﬁnite tower of massive KaluzaKlein states carry a charge quantized in units of eKK . 17 This is the frame speciﬁed by the metric given in (3. de Wit: Supergravity 45 The second rescaling (3. In string theory. that gµν and C11µν transform with equal weights under the scale transformations (3.55]. [ea ]string = e2φD /(D−2) [ea ]Einstein . The signiﬁcance of the dilaton factors in the Lagrangian above is well known. (3. L itself has no intrinsic meaning and it is better to consider the geodesic radius of the 11-th dimension. µ µ (3. equivalently. The Einstein frame in any dimension is deﬁned by a gravitational action that is just the Einstein-Hilbert action. In string theory φ coincides with the dilaton ﬁeld that couples to the topology of the worldsheet and whose vacuum-expectation value deﬁnes the string coupling constant according to gs = exp( φ ). Lstring = 10 1 e e−2φ − 1 R(e. respectively.43) Let us now return to 11-dimensional supergravity with the 11-th coordinate compactiﬁed to a circle so that 0 ≤ x10 ≤ L.36.

46) where a and µ refer to the 10-dimensional Lorentz and world indices. From the 11-dimensional expressions.45) It shows that a small 11-th dimension corresponds to small values of exp φ which in turn corresponds to a weakly coupled string theory. M E10 ∂M = e−2φ/3 ∂10 .6 Kaluza–Klein states and BPS-extended supergravity In most of this section we restrict ourselves to pure supergravity. On the supergravity side these states often appear as solitons. The most conspicuous change is that the continuous nonlinearly realized symmetry group G . this result becomes (M KK )string = |eKK | e− φ · (3.49) Massive Kaluza-Klein states are always BPS states. is that the Kaluza-Klein features of 11-dimensional supergravity have a precise counterpart in string theory [54–56]. The surprising insight that emerged. (3. However.46 Unity from Duality: Gravity. In the string frame. we infer that.48) M KK = |eKK | e−2 φ /3 . the above result reads (R10 )string = L e 2π φ . Observe that L is ﬁxed in terms of κ10 and κ11 (cf.33)). M µ Ea ∂M = ea (∂µ − Vµ ∂10 ) . The central charge here is just the 10-th component of the momentum. The presence of these BPS states introduces a number of qualitative changes to the theory which we discuss in this section. which is proportional to the Kaluza-Klein charge. There one has nonperturbative (in the string coupling constant) states which carry R-R charges. which couple to the supergravity theory as massive matter supermultiplets. 3.32)) related by (3. Gauge Theory and Strings In the string frame. (3. (3.47) Hence Kaluza-Klein states have a mass and a Kaluza-Klein charge (cf. when compactifying dimensions one also encounters massive Kaluza-Klein states. (3. meaning that they are contained in supermultiplets that are “shorter” than the generic massive supermultiplets because of nontrivial central charges. the KaluzaKlein masses are multiples of M KK = 1 · R10 (3. in the frame speciﬁed by the 11-dimensional theory.

Therefore nonzero central charges must couple to appropriate gauge ﬁelds in the supergravity theory. These gauge ﬁelds transform (with minor exceptions) linearly with constant matrices under the group G. Here one naively assumes that the U -duality group acts on the central charges of the BPS states and it is simply deﬁned as the arithmetic subgroup of G that leaves the central-charge lattice invariant. 4 and 5). transform nonlinearly under G. In most cases. they do not constitute representations of the U -duality group. one may thus envisage a (possibly local ﬁeld) theory of BPS states coupled to supergravity that is U -duality invariant. the vector gauge ﬁelds that emerge from the higher-dimensional metric upon the toroidal compactiﬁcation. because the central charges that they carry are too restricted. From the eleven-dimensional perspective it is easy to see why the central charges associated with the Kaluza-Klein states are too restricted. via ﬁeld-dependent HR transformations). in view of the fact that they appear in the anticommutator of two supercharges and supersymmetry is realized locally. This is the reason why retaining the KaluzaKlein states in the dimensional compactiﬁcation will lead to a breaking of the U -duality group. This U -duality group has been conjectured to be the exact symmetry group of (toroidally compactiﬁed) M-theory [54]. However. central charges are in principle assigned to representations of the group HR and not of the group G (although the central charges will eventually. known as the U -duality group. shows that the gauge ﬁelds usually appear in the G-representation required for gauging the corresponding central charge. because under U -duality the central charges related to the . the pointlike (ﬁeld-dependent) central charges can be assigned to representations of G for spacetime dimensions D ≥ 4. This theory would exhibit many of the features of M-theory and describe many of the relevant degrees of freedom. i. The BPS states (which are contained in M-theory) should therefore be assigned to representations of the U -duality group. de Wit: Supergravity 47 is broken to an arithmetic subgroup. these HR representations can be elevated to representations of G..e. In this way. The Kaluza-Klein states that we encounter in toroidal compactiﬁcations of supergravity are a subset of the 1/2-BPS states in M-theory. However. Another aspect of the coupling of the BPS states to supergravity is that the central charges should be related to local symmetries.B. Provided that the central charges can be assigned to the appropriate representations of the U -duality group and that the appropriate gauge ﬁelds are available. They carry pointlike central charges and they couple to the Kaluza-Klein photon ﬁelds. by multiplying with the representatives of the coset space G/HR (representatives of coset spaces will be discussed in Sects. Similar observations exist for stringlike and membranelike central charges except that in these cases the dimension must be restricted even further [41]. Inspection of the tables that we have presented earlier. upon gauge ﬁxing.

transform according to identical representations of the SO(7) helicity group and constitute equivalent supermultiplets.3 are the real symmetric Pauli matrices. In this way. (3.48 Unity from Duality: Gravity. In 9 spacetime dimensions with N = 2 supersymmetry the Lorentzinvariant central charges are encoded in a two-by-two real symmetric matrix Z ij . Such extended theories are called BPS-extended supergravity [40. the BPS supermultiplets which carry momentum along the circle. although the theory would presumably not be able to decompactify uniformly to a ﬂat spacetime of more than eleven dimensions. However. 1) × SL(2. The aim of this section is to elucidate some of these ideas in the relatively simple context of N = 2 supergravity in D = 9 spacetime dimensions. Nevertheless. conventional dimensional compactiﬁcation does not involve any brane charges. R). We note that the central charge associated with the parameter a transforms as a doublet under the .e. It is well known that the massive supermultiplets of IIA and IIB string theory coincide. the number of spacetime dimensions could exceed eleven. Henceforth the momentum states of the IIA and the IIB theories will be denoted as KKA and KKB states. remain inequivalent as they remain assigned to the inequivalent representations of the group SO(8) which is now associated with the restframe (spin) rotations of the massive states. in certain cases one may still be able to extend the Kaluza-Klein states with other BPS states. string theory and (super)membranes. so that a U -duality invariant theory is obtained.50) Here σ1. the charges carried by the Kaluza-Klein states) implies that the newly introduced states (associated with wrapped branes) may also have an interpretation in terms of extra dimensions. which can be decomposed as Z ij = b δ ij + a (cos θ σ3 + sin θ σ1 )ij . We start by considering the BPS multiplets that are relevant in 9 spacetime dimensions from the perspective of supergravity. 41]. The fact that they constitute inequivalent supermultiplets. has implications for the winding states in order that T -duality remains valid [19]. The fact that some of the central charges are associated with extra spacetime dimensions (i. (2. respectively. IIA and IIB states that are massless in 9 spacetime dimensions. whereas the massless states comprise inequivalent supermultiplets for the simple reason that they transform according to different representations of the SO(8) helicity group. respectively. When compactifying the theory on a circle. However. In 9 dimensions the R-symmetry group and the duality group are equal to HR = SO(2) and G = SO(1. Gauge Theory and Strings momentum operator in the compactiﬁed dimensions combine with the twoand ﬁve-brane charges (cf. The corresponding interacting ﬁeld theory is the unique N = 2 supergravity theory in 9 spacetime dimensions.12)) in order to deﬁne representations of the U -duality group.

and right-moving momenta. One type has central charges b = 0 and a = 0. . 17]. while IIB winding states constitute KKA supermultiplets. so that either MBPS = |pL | or 2 2 |pR | with pL = pR . according to which there exists a IIA and a IIB perspective. (for IIB) 2 2 ij (3. MBPS = |a| + |b| .B. so that they don’t play a role in what follows.51) Here one can distinguish three types of BPS supermultiplets. the result takes a diﬀerent form for the IIA and the IIB theory as the following formula shows. which can be generated by wrapping the membrane over the corresponding T 2 . However. The KKA supermultiplets that comprise Kaluza-Klein states of IIA supergravity are of this type. 2 2 (3. For type-II string theory one obtains these central charges in terms of the left. For the winding states. namely that IIA momentum states constitute KKA supermultiplets. The KKB supermultiplets that comprise the Kaluza-Klein states of IIB supergravity are of this type. where pL = −pR .52) Z = 1 (p − p )δ ij + 1 (p + p )σ ij . de Wit: Supergravity 49 SO(2) R-symmetry group that rotates the two supercharge spinors. Subsequently one shows that BPS states that carry these charges must satisfy the mass formula. ij 1 (pL + pR )δ ij + 1 (pL − pR )σ3 .53) For pL = pR we conﬁrm the original identiﬁcation of the momentum states. (for IIA) L R L R 3 2 2 The corresponding BPS mass formula is thus equal to MBPS = 1 |pL + pR | + 1 |pL − pR | . It is also possible to view the central charges from the perspective of the 11-dimensional (super)membrane [57]. (3. Another type of 1/2-BPS mulitplets has central charges a = 0 and b = 0. while the central charge proportional to the parameter b is SO(2) invariant. The 1/4-BPS multiplets arise for string states that have either right. Assuming that the two-brane charge takes values in the compact coordinates labelled by 9 and 10. All of this is entirely consistent with T -duality [16. with decompactiﬁcation radii are that inversely proportional and with an interchange of winding and momentum states. one obtains the opposite result: IIA winding states constitute KKB supermultiplets. pL . while IIB momentum states constitute KKB supermultiplets. pR .or left-moving oscillator states. Finally there are 1/4-BPS multiplets (annihilated by one fourth of the supercharges) characterized by the fact that neither a nor b vanishes. These are 1/2-BPS muliplets. because they are annihilated by half of the supercharges. that characterize winding and momentum along S 1 . Observe that the 1/4-BPS states will never become massless.

as the ﬁelds never depend on all the 12 coordinates! Whether this kind of BPS-extended supergravity oﬀers a viable scheme in a more general context than the one we discuss here. the Kaluza-Klein states transform as KKA multiplets. the Kaluza-Klein states constitute KKB multiplets and their charge is SO(2) invariant. Therefore there is a rather natural way to describe the IIA and IIB momentum and winding states starting from a (super)membrane in eleven spacetime dimensions.50 Unity from Duality: Gravity. Their charges transform obviously with respect to an SO(2) associated with rotations of the coordinates labelled by 9 and 10. is not known. ij ij Z ij = Z9 10 δ ij − (P9 σ3 − P10 σ1 ) .55) Here q1. This point was ﬁrst emphasized in [58]. Hence we have a “double” tower of these charges with corresponding KKA supermultiplets. In the case at hand we know a lot about these couplings from our knowledge of the T 2 compactiﬁcation of D = 11 supergravity and the S 1 compactiﬁcation of IIB supergravity. Tm denotes the supermembrane tension. From the perspective of 11-dimensional supergravity compactiﬁed on T 2 . Gauge Theory and Strings one readily ﬁnds the expression. the BPS mass formula takes the form MBPS = = 2 P92 + P10 + |Z9 10 | 1 √ |q1 + τ q2 | + Tm A |p| . A τ2 (3. as well as eleven-dimensional supergravity. from the perspective of 9-dimensional supergravity one is led to couple both towers of KKA and KKB supermultiplets simultaneously. In that case one obtains some dichotomic theory [19]. (3.2 denote the momentum numbers on the torus and p is the number of times the membrane is wrapped over the torus. not even in a uniform decompactiﬁcation limit. Clearly the KKA states correspond to the momentum modes on T 2 while the KKB states are associated with the wrapped membranes on the torus.54) When compactifying on a torus with modular parameter τ and area A. although there is no 12-dimensional Lorentz invariance. This suggests to consider N = 2 supergravity in 9 spacetime dimensions and couple it to the simplest BPS supermultiplets corresponding to KKA and KKB states. On the other hand. However. In the case at hand this new theory describes the ten-dimensional IIA and IIB theories in certain decompactiﬁcation limits. which we refer to as BPS-extended supergravity. from the perspective of IIB compactiﬁed on S 1 . But the theory is in some sense truly 12-dimensional with three compact coordinates. As shown in Tables 8 and 10 there are three central charges and 9-dimensional supergravity possesses precisely three gauge ﬁelds that couple to these charges. . Here we have a “single” tower of KKB supermultiplets.

1) 3 weights through the dimensional reduction of IIB and eleven-dimensional supergravity. We expect that the local SO(2) invariance can be incorporated in the full BPS-extended supergravity theory and can be exploited in the construction of the couplings of the various BPS supermultiplets to supergravity. as the corresponding ﬁelds can be uniquely identiﬁed by their scaling weights under SO(1. which satisfy a constraint φα φα = 1 and are subject to a local SO(2) invariance. by G99 = exp(σ). The resulting BPS-extended theory incorporates 11-dimensional supergravity and the two type-II supergravities in special decompactiﬁcation limits. But. we are dealing with a 12-dimensional theory . In 9 dimensions supergravity has two more scalars. so that they describe precisely two scalar degrees of freedom (α = 1. The coset is described by the complex doublet of ﬁelds φα . a symmetry of the massless theory that emerges upon dimensional reduction and is associated with scalings of the internal vielbeine.B. On the IIA side it originates from the IIA tensor ﬁeld. There are two vector ﬁelds Aα . 1). The third gauge ﬁeld. From the IIB side they originate from the tensor ﬁelds. is a singlet under SL(2) and is the KaluzaKlein photon on the IIB side. de Wit: Supergravity 51 The ﬁelds of 9-dimensional N = 2 supergravity are listed in Table 13. the results follow entirely from supersymmetry. the KKA states are the momentum states on T 2 . The scalar ﬁeld σ is related to G99 . From the IIA perspective these correspond to the Kaluza-Klein states on S 1 and the D0 states. so that it couples to the KKB states. while the KKB states correspond to the membranes wrapped around the torus. which are the µ Kaluza-Klein photons from the T 2 reduction of eleven-dimensional supergravity and which couple therefore to the KKA states. denoted by Bµ . These two ﬁelds transform under SL(2). 2). which can be understood from the perspective of the modular transformation on T 2 as well as from the S-duality transformations that rotate the elementary strings with the D1 strings. where we also indicate their relation with the ﬁelds of 11-dimensional and 10-dimensional IIA/B supergravity upon dimensional reduction. which is equal to exp(− 4 σ). as we stressed above. which conﬁrms that they couple to the IIB (elementary and D1) winding states. From the perspective of the supermembrane. we stress that. likewise it is related to the determinant of the 11-dimensional metric in the compactiﬁed dimensions. R)/SO(2). It is not necessary to work out all the nonlinear ﬁeld redeﬁnitions here. from the perspective of 9-dimensional N = 2 supergravity. While it is gratifying to see how all these correspondences work out. We already mentioned the three Abelian vector gauge ﬁelds which couple to the central charges. which are described by a nonlinear sigma model based on SL(2. the IIB metric component in the compactiﬁed dimension. which is consistent with the fact that the IIA winding states constitute KKB supermultiplets. The precise relationship follows from comparing the SO(1.

One then ﬁnds that m2 mKKB ∝ Tm . type-IIA. q2 . although no ﬁeld can depend nontrivially on all of these coordinates. D = 11 ˆ Gµν ˆ Aµ 9 10 ˆ ˆ Gµ 9 .56) where qα and p refer to the integer-valued KKA and KKB charges. we only have a discrete subgroup that leaves the charge lattice invariant. We should discuss the fate of the group G = SO(1. G9 9 . One simply has the freedom to view the theory from a IIA or a IIB perspective and interpret it accordingly. The last column lists the SO(1. . 8. although the latter has become trivial as the theory is not based on a speciﬁc IIA or IIB perspective. respectively. Gµ 10 ˆ ˆ Aµν 9 . KKA (3. after coupling to the BPS states. We return to them in a moment. C9 D=9 gµν Bµ α Aµ α Aµν IIB Gµν Gµ 9 α Aµ 9 α Aµν SO(1. The 11-dimensional and ˆ 10-dimensional indices. Cµν Cµνρ φ. This is the group SL(2. 10) and M = (µ. 1) 0 −4 3 −1 2 0 7 Aµνρ φ exp(σ) α Aµνρσ φ G9 9 α here. The central charges of the BPS states form a discrete lattice. This formula can be compared to the membrane BPS formula (3. and mKKA and mKKB are two independent mass scales. G9 9 . Both S. R) of pure supergravity after coupling the theory to the BPS multiplets. p) = mKKA e3σ/7 |qα φα | + mKKB e−4σ/7 |p| .52 Unity from Duality: Gravity. Aµν 10 ˆ Aµνρ ˆ ˆ ˆ G9 10 . 9). 1) scaling weights of the ﬁelds. Gauge Theory and Strings Table 13. (3. In this way we are able to deduce the following BPS mass formula. are split as M = (µ. The bosonic ﬁelds of the eleven dimensional. which is aﬀected by this group. The KKA and KKB states and their interactions with the massless theory can be understood from the perspective of compactiﬁed 11-dimensional and IIB supergravity. 1. MBPS (q1 . where µ = 0. . 9.57) . Z). Hence. Cµ Cµν 9 .and T -duality are manifest. 1) × SL(2. G10 10 IIA Gµν Cµ 9 Gµ 9 . respectively.55) in the 11-dimensional frame. The theory has obviously two mass scales associated with the KKA and KKB states. ninedimensional N = 2 and type-IIB supergravity theories. .

Table 15 summarizes Q = 16 supergravity for dimensions 3 ≤ D ≤ 10. we now have the option of introducing additional matter ﬁelds. 4 (3. In D = 3 dimensions the vector ﬁeld is dual to a scalar. For Q = 16 the matter will be in the form of vector supermultiplets. D 10 9 8 7 6 5 4 3 HR 1 1 U (1) U Sp(2) U Sp(2)×U Sp(2) U Sp(4) U (4) SO(8) Aµ 1 1 1 1 1 1 1 φ 0 1 1+¯ 1 3 (2.7 Nonmaximal supersymmetry: Q = 16 For completeness we also summarize a number of results on nonmaximal supersymmetric theories with Q = 16 supercharges. 3. In other words. 2) 4 4+¯ 4 8 with a numerical proportionality constant. All supermultiplets contain a gauge ﬁeld Aµ .56) is that there is no limit in which the masses of both KKA and KKB states will tend to zero.B. ω) 2 κ2 10 3 − 4 e e−2φ/3 (Hµνρ )2 − 1 e(∂µ Aν − ∂ν Aµ )2 . there exists no theory with Q = 32 supercharges in ﬂat Minkowski spacetime of dimensions D > 11. This multiplet comprises 8 + 8 physical degrees of freedom. there is no uniform decompactiﬁcation limit. Field content for maximal super-Maxwell theories in various dimensions.38). which are now restricted to dimensions D ≤ 10. scalars φ and spinors χ and comprises 8 + 8 degrees of freedom. 1) + (1. Table 14 shows the ﬁeld representations for the vector multiplet in dimension 3 ≤ D ≤ 10. de Wit: Supergravity 53 Table 14. However. Therefore.58) . possibly associated with some non-Abelian gauge group. L10 = 1 − 1 e e2φ/3 R(e. The Lagrangian can be obtained by truncation of (3. 2) 5 6∗ 8 χ 1 1 1+¯ 1 2 (2. The 6∗ representation of SU (4) is a selfdual rank-2 tensor. We also consider the Q = 16 supergravity theories. in spite of the fact that we have more than 11 dimensions. However. the most important conclusion to draw from (3. In D = 10 dimensions the bosonic terms of the supergravity Lagrangian take the form [59]. unlike in the case of maximal supergravity.

representing an Abelian vector supermultiplet. D 10 9 8 7 6A 6B 5 4 3 HR 1 1 U (1) U Sp(2) U Sp(2)×U Sp(2) U Sp(4) U Sp(4) U (4) SO(8) # 64 56 48 40 32 24 24 16 8k graviton 1 1 1 1 1 1 1 1 1 p = −1 1 1 1 1 1 1 1+¯ 1 8k p=0 1 1+¯ 1 3 (2.54 Unity from Duality: Gravity. Note also that the kinetic term for the Kaluza-Klein vector ﬁeld in (3. but act on the (Abelian) ﬁeld strengths by duality transformations. 1) 5∗ where. for convenience. # denotes the number of bosonic degrees of freedom.37). the SU (4) transformations cannot be implemented on the vector potentials. A feature that deserves to be mentioned. In D = 3 dimensions supergravity is a topological theory and can be coupled to scalars and spinors.36. 1) and (2. depends on φ. The tensor ﬁeld in the 6B theory is selfdual.59) C11µν → e C11µν . we have included a single vector gauge ﬁeld Aµ . with the exception of the 6B and the 4-dimensional theory. Chern-Simons terms play an important role in the anomaly cancellations of this theory. Bosonic ﬁelds of nonmaximal supergravity with Q = 16. 2) 5+1 [6] p=1 1 1 1 1 (1. 3. This reﬂects itself in the extension of the symmetry transformations noted in (3. µ µ 6α φ → φ + 12α . where Aµ transforms diﬀerently from the Kaluza-Klein vector ﬁeld Vµ . In D = 4 dimensions. Aµ → e3α Aµ (3. Gauge Theory and Strings Table 15. ea → e−α ea . is that the ﬁeld strength Hµνρ associated with the 2-rank gauge ﬁeld contains a Chern-Simons term A[µ ∂ν Aρ] . 0) supergravity. Note that. In the third column. where k is an arbitrary integer. all these theories contain precisely one scalar ﬁeld. k)/(SO(8)×SO(k)). The scalars parametrize the coset space SO(8. unlike the kinetic term for the matter vector ﬁeld in the Lagrangian above. In 6 dimensions type-A and type-B correspond to (1.38). .

ω) − 1 e(∂µ φ)2 2 4 κ2 10 3 − 4 e e−φ (Hµνρ )2 − 1 e e−φ/2 (∂µ Aν − ∂ν Aµ )2 . 4 4 (3. 4 Homogeneous spaces and nonlinear sigma models This chapter oﬀers an introduction to coset spaces and nonlinear sigma models based on such target spaces with their possible gaugings. This is related to the fact that the SO(32) heterotic string theory is S-dual to type-I string theory [60]. µ (3. µ = e−φ/3 [ea ]string . This is the low-energy eﬀective Lagrangian relevant for the heterotic string. Eventually the matter gauge ﬁeld has to be part of an non-Abelian gauge theory based on the groups SO(32) or E8 × E8 in order to be anomaly-free. In the Einstein frame.60) It is straightforward to obtain the corresponding Lagrangians. The aim . this is the low-energy eﬀective action of the type-I string. de Wit: Supergravity 55 In this case there are three diﬀerent Weyl rescalings that are relevant. namely ea µ ea µ ea µ = e−φ/12 [ea ]Einstein .63) Now the dilaton seems to appear with the wrong sign. 4 4 (3. Finally. the graviton is again invariant under the isometries of the scalar ﬁeld. The bosonic terms read LEinstein = 10 1 − 1 e R(e.B. Lstring = 10 1 −2φ e − 1 e R(e. As it turns out. ω) + 2e(∂µ φ)2 2 κ2 10 − 3 e (Hµνρ )2 − 1 e(∂µ Aν − ∂ν Aµ )2 . ω) + 2(∂µ φ)2 2 κ2 10 − 3 e (Hµνρ )2 − 1 e eφ (∂µ Aν − ∂ν Aµ )2 . (3.62) which shows a uniform coupling to the dilaton. where the type-I dilaton must be associated with −φ. µ = eφ/6 [ea ]string .61) 4 The second Weyl rescaling leads to the following Lagrangian. the third Weyl rescaling yields Lstring 10 = 1 e e2φ − 1 R(e.

Obviously. We start by introducing the concept of a coset space G/H. q).1 Nonlinearly realized symmetries As an example consider the n-dimensional sphere S n of unit radius which we may embed in an (n+1)-dimensional real vector space Rn+1 . denoted by H. we can reach each point by a suitable rotation. It is possible to elevate the Abelian gauge group associated with the vector gauge ﬁelds to a non-Abelian group. equal to E7(7) and E6(6) . A homogeneous space is a space where every two points can be connected by an isometry transformation. . 5 spacetime dimensions. However. Such invariances are called isometries and G = SO(n + 1) is therefore known as the isometry group. . and present their description in a form that emphasizes a local gauge invariance associated with the group H. the group of (n + 1)-dimensional rotations. respectively. It is convenient to choose a certain point on the sphere (let us call it the north pole) with coordinates in Rn+1 given by (0. Then we discuss the corresponding nonlinear sigma models. 0. for a homogeneous manifold. the rotation connecting these two points is not unique as every point on S n is invariant under an SO(n) subgroup. these theories have a nonlinearly realized symmetry group. concentrating on the maximal supergravities in D = 4. . However. such as [61. 62]. Finally we introduce the so-called gauging of this class of nonlinear sigma models. where we explain the gauging of maximal supergravity. where H is a subgroup of a group G. Most of this material is standard and can be found in textbooks. The latter enable us to include some material on de Sitter and anti-de Sitter spacetimes. consisting of the following orthogonal matrices embedded . which we make use of in later chapters. In this introduction we try to be as general as possible but in the examples we restrict ourselves to pseudo-orthogonal groups: SO(n) or the noncompact versions SO(p.56 Unity from Duality: Gravity. From the north pole. Clearly the sphere is such a homogeneous space as every two points on S n can be related by an SO(n + 1) isometry. 1). which is a subgroup of these exceptional groups. based on homogeneous target spaces. Gauge Theory and Strings of this introduction is to facilitate the discussion in the next chapter. the isotropy groups for two arbitrary points are isomorphic (but not identical as one has to rotate between these points). This group is called the isotropy group (or stability subgroup). 4. . the north pole itself is invariant under the SO(n) isotropy group. The sphere is obviously invariant under SO(n + 1). As discussed earlier. The construction of these gaugings makes an essential use of the concepts and techniques discussed here.

Every such space can be described in terms of appropriate G/H cosets based on an isometry group G and a isotropy subgroup H. Such equivalence classes are called cosets. ∗ ··· ∗ 0 ··· 0 0 .2) satisﬁes the condition (4. h=. we are dealing with a compact space. Using the same decomposition as in (4. is thus equivalent to giving a parametrization of the sphere. Here i. (4. . (4. de Wit: Supergravity into SO(n + 1).. . q) thus satisfy g −1 = η 0 0 1 gT η 0 0 1 . q) we have a diagonal metric with p eigenvalues +1 and q eigenvalues −1 (or vice versa as the overall sign is not relevant). we choose this metric of the form diag (η. .2) g(α) = exp −αj 0 with some element of H.3) so that the metric η is obviously H-invariant.2) is no longer orthogonal and the . 1). 0 1 57 h∈H. where η is again a diagonal metric with p (or q) eigenvalues equal to −1 and q − 1 (or p − 1) eigenvalues equal to +1. (4. ∗ ··· ∗ . (4. which assigns a single SO(n + 1) element to every coset.1) Therefore. The sphere is obviously just one particular example of a homogeneous space. It is not diﬃcult to ﬁnd such a parametrization. When η has negative eigenvalues the matrix (4. n. . so that we can also deal with noncompact spaces. then the transformation g2 = g1 · h will do the same.1). The coset representative (4. provided that αi = ηij αj . . j = 1. if a rotation g1 ∈ G. Elements of SO(p. ∗ .B. . . and therefore the space S n is a coset space G/H with G = SO(n + 1) and H = SO(n). Therefore points on the sphere can be associated with the class of group elements g ∈ G that are equivalent up to multiplication by elements h ∈ H from the right.4) When η equals the unit matrix. .3). . A parametrization of the cosets of SO(n + 1)/SO(n). The noncompact groups leave an indeﬁnite metric invariant. with G = SO(n + 1) maps the north pole onto a certain point on the sphere. For SO(p. One ﬁrst observes that every element of SO(n + 1) can be decomposed as the product of 0 αi . In view of various applications we extend our notation to noncompact versions of the orthogonal group.

because there are disconnected components. which ensures that in an inﬁnitesimal neighbourhood of a point invariant under H. we have the anti-de Sitter space SO(n − 1. 2)/SO(n − 1. the generators k are deﬁned up to additive terms belonging to elements of the algebra associated with H. The elements g(α) ∈ G deﬁne a representative of the G/H coset space. (4. 18 Noncompact . is “mostly plus” for de Sitter. Of course. However. We also note that. k. such as in the coset representatives (4. +. the coset representative is not unique. h. 1).5) The ﬁrst commutation relation states that the h form a subalgebra. 19 We are a little cavalier here with our terminology. We reserve the term hyperbolic for the Riemannian hyperbolic space. and therefore a parametrization of the corresponding space. for η = (−. This aspect will be important later on when comparing the curvature for these spaces.1). Gauge Theory and Strings space will be noncompact18 . as we shall see later (cf. which eventually will play the role of the tangent-space metric. In general. Also the de Sitter and anti-de Sitter spaces are hyperbolic. In this way we can thus treat a variety of spaces at the same time. k] = [k. In our example. We have decomposed the generators of G into generators h of H and generators k belonging to its complement. Obviously the above relations involve a choice of basis. the latter have been used to generate the coset representative. −. note that η. the latter are associated with the last row and column of (4. and for η = (−.58 Unity from Duality: Gravity. 1)/SO(n). h] = [h. When they do not. We will not discuss these issues here. Let us mention some examples. in the noncompact case. +). have a positive or negative deﬁnite metric and are thus Riemannian spaces.2). +) we have the de Sitter space SO(n. · · · . 1)/SO(n − 1. The second one implies that the generators k form a representation of H. In the more general case the third commutator may also yield the generators k. so that they satisfy (schematically). there are other decompositions than (4. coset spaces where H is the maximal compact subgroup of a noncompact group G. the coordinates αi rotate under H according to that representation. 1). Unlike the de Sitter and anti-de Sitter spaces the former spaces are double-sheeded. groups are not fully covered by exponentiation. · · · . and “mostly minus” for the hyperbolic and the antide Sitter space. For η = −1 we have the hyperbolic space19 SO(n.2).10)). (4. k] = h. which oﬀer distinct advantages. but they are pseudo-Riemannian. [h. the homogeneous space is symmetric.

Acting with (4. for instance20 . i δj + αi αj cos α2− 1 α g(α) = − sin α αj α 59 sin α αi α . Obviously the space is compact when η is positive. Observe that the appearance of η in the above formulae is the result of the fact that the generators k are normalized according to tr(ki kj ) = −2 ηij . This corresponds to the sphere S n . In all other cases the parameter space is obviously noncompact and the sine and cosine may change to the hyperbolic sine and cosine in those parts of the space where α2 is negative. One may use the coset representative to sweep out the coset space from one point (i.6) where α2 = ηij αi αj .2). α y i (4. We may also chose diﬀerent coordinates. the “north pole”) in the (n+1)-dimensional embedding space. the upper or the lower hemisphere).8) equals the square root of the matrix δj − y i yj .8) on the point (0. ± 1 − y 2 ) are sometimes called homogeneous coordinates. Y A = (y i . Note that i the n × n submatrix in (4. . cos α (4. ± 1 − y 2 ) .2). such as.B. y i = αi sin α .e. For the sphere the range of the coordinates is restricted by y 2 = Σi (y i )2 ≤ 1. One may choose a diﬀerent parametrization of the cosets by making a diﬀerent decomposition than in (4.8) ± 1 − y2 where. we parametrize diﬀerent parts of the space (for the sphere S n . . Inhomogeneous coordinates are the ratios y i / 1 − y 2 . 0.9) 20 The coordinates (y i .7) so that the coset representative reads ± 1 − y2 − 1 i i δj + y y j y2 g(y) = −yj (4. 1) yields the following coordinates in the embedding space. Diﬀerent parametrizations are generally related through (coordinate-dependent) H transformations acting from the right. . . depending on the sign choice. (4. de Wit: Supergravity Let us now proceed and evaluate (4. because the G-transformations act linearly on these coordinates. . because in that case the parameter space can be restricted to 0 ≤ α < π.

we ﬁnd the relation.14) . q + 2 1). Writing oG ≈ 1 + g ˆ and oH ≈ 1 + h(y).11) Hence the eﬀect is a change of coordinates y → y in a way that satisﬁes the group multiplication laws. one ﬁnds that there are two types of isometries. After this multiplication the result is in general no longer compatible with the coset representative g(y). one has g(y) −→ oG g(y) = g(y ) oH (y) . α (4.13) where the i are n constant parameters. which will induce corresponding transformations in the coset space. oH (y). q) in the spinor representation. (4.3) one then shows that the coset space is embedded in n + 1 dimensions according to ηij Y i Y j + (Y n+1 )2 = 1 . ˆ g ξ i (y) ∂i g(y) = ˆ g(y) − g(y) h(y) . In other words. which changes the coordinates y i by constant rotations.60 Unity from Duality: Gravity. but by applying a suitable y-dependent H transformation. we can again bring g(y) in the desired form. To see this. from the right. we multiply g(y) by a constant element oG ∈ G from the left.12) We return to this result in the next section. The inﬁnitesimal transformation y i → y i = ˆ y i + ξ i (y) deﬁnes the so-called Killing vectors ξ i (y). as one can verify by including extra generators equal to the matrices 1 2 Γi . Gauge Theory and Strings Using (4. The other corresponds to n coordinate dependent shifts. Applying this construction to the case at hand.10) Since the g(y) are contained in G one may examine the eﬀect of G transformations acting on g(y). Coset representatives can be deﬁned in diﬀerent representations of the group G. g(α) = exp[ 1 Γi αi ] = cos(α/2) 1 + i 2 sin(α/2) i α Γi . δy i = i 1 − y2 . (4. q). but in fact it transforms also as a spinor under SO(p. both types of transformations take the form of a constant G transformation on the embedding coordinates Y A which leaves the embedding condition (4. (4. Consequently we can deﬁne a representative of SO(p. The representation transforms (not necessarily irreducibly) under SO(p.10) invariant. As the reader can easily verify. One corresponding to the group H. q + 1)/SO(p. q) generated by the matrices 1 Γij . The most interesting one is the spinor representation. (4. Assume that we have a representation of the Cliﬀord algebra C(p.

The signiﬁcance of this fact will be clear in a sequel. 4. with indices i that label the coordinates and indices a that label the generators k. the one-form g −1 dg takes its values in the Lie algebra associated with G. deﬁned with the help of the representative (4.e. de Wit: Supergravity 61 √ with α deﬁned as before. Γj } = −2ηij 1.15) at the end of the previous subsection. Therefore the one-forms can be decomposed into the generators h and k. speciﬁed at the north pole. As a ﬁrst consequence we note that the spinor ψ(y). (4. similar results can be obtained in other representations of G. introduced earlier. This construction can applied as well to cosets of other (pseudo-)orthogonal groups. Obviously. because they are invariant under left multiplication of g with constant elements of G. satisﬁes the equation. associated with tangent-space rotations that belong to the group H. g −1 dg = ω + e . which deﬁne a tangent frame at each point of the space. we deﬁne ψ(y) = g −1 (y) ψ(0). and {Γi . The resulting y-dependent spinor ψ(y) is a so-called Killing spinor of the coset space. i. The one-forms g −1 dg are called left-invariant. α2 = αi αi ..B. In terms of the coordinates y i the representative reads g(y) = 1 2 1+y+ y i Γi √ 1−y 1+ √ · 1+y+ 1−y (4. It is convenient to use the language of diﬀerential forms. These one-forms e are thus related to the vielbeine of the coset space. The one-forms ω deﬁne the spin connection21 . Hence e deﬁnes a square matrix. The exterior derivative dg(y) describes the change of g induced by an inﬁnitesimal variation of the coset-space coordinates y i .16) where ω is decomposable into the generators h and e into the generators k.15) One can act with this representative on a constant spinor.17) 21 Observe that in supergravity we have deﬁned the spin connection ﬁeld with opposite sign.e.2 Geometrical quantities Geometrical quantities of the homogeneous space are deﬁned from the leftinvariant one-forms g −1 dg.16) is of central importance for the geometry of the coset spaces. d + ω + e ψ(y) = 0 . so that the one-forms take their value in the Lie algebra associated with G. Because the g’s themselves are elements of G. Equation (4. i. .but by no means essential. We shall exhibit this below. where g(y) ∈ G. (4.

In general it is not necessary to specify the coset representative. (4.e. g(y) −→ g(y) h(y) .18) (ω + e) −→ h−1 (ω + e)h + h−1 dh . ej ] − [ωj . It is straightforward to see how ω and e transform under (4. (4. yi ωi ei −→ y i + ξ i . d(ω + e) = −(ω + e) ∧ (ω + e) . Rij (G/H) = ∂i ej − ∂j ei + [ωi .23) (4. (4. which yields ωi ei −→ h−1 ωi h + h−1 ∂i h . ei ] . as diﬀerent representatives are related by y-dependent H transformation acting from the right on g.24) d(g −1 dg) = −(g −1 dg) ∧ (g −1 dg) . ωj ] .22) The values of these curvatures follow from the Cartan-Maurer equations. −→ h−1 ei h .e. i. (4. Furthermore it follows from (4.16). ω acts as a gauge connection for the local H transformations.. we note the relations [Di . Introducing H-covariant derivatives. −→ ei − ∂i ξ j ej − ξ j ∂j ei .21) We can also deﬁne Lie-algebra valued curvatures associated with ωi and ei . (4. Let us now proceed and investigate the properties of the one-forms ω and e. Dj ] = −Rij (H) .62 Unity from Duality: Gravity. or. in terms of ω and e. (3. h(y) ∈ H .27) that ωi and ei transform as covariant vectors under coordinate transformations. Gauge Theory and Strings Upon writing this out in terms of the gamma matrices. Rij (G/H) = Di ej − Dj ei .25) . −→ ωi − ∂i ξ j ωj − ξ j ∂j ωi . Rij (H) = ∂i ωj − ∂j ωi + [ωi .20) Obviously.19) This equation can again be decomposed (using the ﬁrst two relations (4. i.18) This leads to a diﬀerent parametrization of the coset space.15)). one recovers precisely the so-called Killing spinor equation (cf. To derive these equations we take the exterior derivative of the deﬁning relation (4. (4.5) in terms of the generators h and k. (4.

ωi (y) dy i ei (y) dy i .B. (4. b.29) Furthermore.8) one readily obtains ab ωi = = b a y a δi − y b δi a δi + 1∓ 1 1 − y2 y2 . that label the generators k are called tangent-space indices (or local Lorentz indices in the context of general relativity). because the group H was precisely the (pseudo)orthogonal group. we ﬁnd Rij (H) = −[ei . Usually the group H can be embedded into SO(n) (or a noncompact version thereof) and leaves some target-space metric invariant (we will see the importance of this fact shortly).27) g −1 dg = i −ei (y) dy 0 From (4. (4. using the same matrix decomposition as before. b. Rij (G/H) = 0 . which follows from the (pseudo)orthogonality of H. . . we ﬁnd explicit expressions for the vielbein and the spin connection. In the context of diﬀerential geometry the indices i are called world indices. The quantity ωi thus acts as the connection associated with rotations of the tangent frames.5). the ﬁelds ei can be decomposed into the generators k and thus deﬁne a set of vielbeine eia that specify a tangent frame at each point in the coset space. because they refer to the coordinates of a manifold. These aspects are easily recognized in the examples we are discussing. whereas the indices a.28) eia yi y a ± y2 1 − y2 −1 . using the relations (4.26) Note that the vanishing of Rij (G/H) is a consequence of the fact that we assumed that the coset space was symmetric (see the text below (4. the curvatures introduced before.5)). Because the generators k form a representation of H. this group rotates the tangent frames. Hence. ab where. (4. and therefore we call it the spin connection. ej ] . Note that ωi is antisymmetric in a. are readily identiﬁed with the curvature of the spin connection and with the torsion tensor. From the . As we already alluded to earlier. The inverse vielbein reads i i e a = δa + ya y i ± y2 1 − y2 − 1 . . indices are raised and lowered with η. de Wit: Supergravity 63 Decomposing this equation in terms of the Lie algebra generators. as before. (4.

Gauge Theory and Strings Cartan-Maurer equations.33) Because the torsion is zero. [i j] Di eja − Dj eia = 0 . which indicates that we are dealing with a maximally symmetric space. but we stress that this is related to the signature choice that we made for the metric. Because ωi diﬀers in sign as compared to the spin connection used in ab Section 3.35) ξ i (y) (ωi (y) + ei (y)) = g −1 (y)ˆ g(y) − ˆ .12) with (4.16). (4. In the parametrization (4. (4. a “mostly plus” metric requires to include a minus sign in the deﬁnition (4. 2 All coset spaces have isometries corresponding to the group G. c l upon contraction with ηac ek eb .34) is proportional to the metric. ab Rij (ω) = 2 ea eb . g h(y) (4. we obtain. This means that the maximal number of isometries (equal to 1 n(n + 1)) is realized for this space.1. which we introduced earlier.31) for the hyperbolic and anti-de Sitter spaces. 1 − y2 g ij = η ij − y i y j .31) When there are several H-invariant tensors. there is a more extended class of metrics that one may consider.28) one obtains for the metric and its inverse. this leads to Γijk = y k gij . from which we can deﬁne the Riemann curvature. (4. l l Rijk l = −gki δj + gkj δi .32) Given the fact that we have already made a choice for η previously. so that gij = ηab eia ejb . .30) To deﬁne a metric gij one contracts an H-invariant symmetric rank-2 tensor with the vielbeine. Thus the curvature is of deﬁnite sign. Combining (4. This sign is important when comparing to spheres or de Sitter spaces. but in the case at hand the metric is unique up to a proportionality factor. The Riemann curvature (4. the Riemann tensor is equal to minus the curvature Rij (ω). explained above.31). For the examples at hand. (4. The diﬀeomorphisms associated with these isometries are generated by Killing vectors ξ i (y). gij = ηij + yi yj . we thus ﬁnd in components. (4. From the vielbein postulate. we know that the aﬃne connection is equal to Γijk = eak Di eja . as we have discussed above.64 Unity from Duality: Gravity.34) where gij is the metric tensor deﬁned by (4. the connection coincides with Christoﬀel symbol. The obvious invariant tensor is ηab . and we ﬁnd the following result for the Riemann curvature tensor.

h(y)] = ˆ ˆ ∂i ξ j ωj + ξ j ∂j ωi + ∂i h(y) + [ωi . β. For the vielbein and spin connection.41) Observe that both terms vanish separately for a symmetric space. we established (cf. g g (4.12). (4.36) yields Rij (G/H) ξ j + [ei . Since the metric is Hinvariant by construction. are invariant under the group G.39) Combining the ﬁrst equation with the ﬁrst equation (4. g (4. Therefore. ˜(y)]G/H = 0 . (4. This follows from (4. g ˜ −ξ i (y) ωi (y) + h(y) . it thus follows that the metric is invariant under the diﬀeomorphism associated with G.11)) that the G-transformation acting on the left can be decomposed into a diﬀeomorphism combined with a coordinate-dependent H-transformation. Rij (H) ξ j .40) 0. it follows that e and ω are both invariant as well. and introduce structure constants by g [ˆα . which transform under H. (4. The diﬀeomorphisms generated by the Killing vector ﬁelds will give rise to the group G. ˆβ ] = fαβγ ˆγ .42) . 0. Let us label the generators of the group G by indices α. the vielbein e and the spin connection ω are invariant under these combined transformations. from which e and ω were constructed.36) . Therefore. we ﬁnd ˆ ∂i ξ j ej + ξ j ∂j ei + [ei . we ﬁnd ξ i (y) ei (y) = ˆ h(y) = where ˜ h(y) ˜(y) g = = ˆ g −1 (y) g g(y) ˆ g −1 (y) g g(y) H 65 ˜(y) .36). . ˜ ∂i ξ j ej + ξ j Dj ei + [ei . de Wit: Supergravity Decomposing this equation according to the Lie algebra. Now we return to the observation that the left-invariant forms. h(y)] = ˜ Di h(y) = 0.B. h(y)] = ˜ In terms of h(y) these results take a more covariant form. Hence. .37) G/H ˆ The contribution h(y) is only relevant for those quantities that live in the tangent space. . . Moreover.38) where ξi = gij ξ j and ξ i is the so-called Killing vector deﬁned by (4. (4. δgij = Di ξj (y) + Dj ξi (y) = 0 . (4. . (4.

The local H transformations depend on the ﬁelds φ(x) and thus indirectly on the spacetime coordinates. (4. i j ˜ fαβγ hγ + ξα ξβ Rij (H) . (4. To show this we note the identity i i i j ˜ ˜ ˜ ˜ (−ξα Di + hα )(−ξβ Di + hβ ) = ξα ξβ (Di Dj − Rij (H)) + hα hβ j i ˜ j ˜ j +[ξα (Di ξβ ) + hα ξβ + hβ ξα ]Dj . Gauge Theory and Strings The Killing vectors and the corresponding H-transformations then satisfy corresponding group multiplication properties. One starts from scalar ﬁelds which take their values in the homogeneous space. On such ﬁelds the isometries are generated by the operators.45) 4.44) On the basis of the results above one can show that these operators satisfy the commutation relations of the Lie algebra associated with the isometry group G. j i j i ξβ ∂j ξα − ξα ∂j ξβ = i fαβγ ξγ . which now depends on n ﬁelds. one uses the analogue of (4. Pµ (φ) = ei (φ) ∂µ φi . (4. Hence we may follow the same procedure as before and deﬁne a coset representative V(φi (x)) ∈ G. ˜ ˜ [hα . where Qµ is decomposable into the generators h and Pµ into the generators k. so that the ﬁelds φi (x) deﬁne a map from the spacetime to the coset space. Therefore one may elevate these transformations to transformations that depend arbitrarily on xµ . which are functions of the coset space coordinates assigned to a representation of the group H.66 Unity from Duality: Gravity.48) . Obviously one has the relations Qµ (φ) = ωi (φ) ∂µ φi . (4. to deﬁne Lie-algebra valued quantities Qµ and Pµ . hβ ] = (4. (4.43) One can consider ﬁelds on the coset space.46) V −1 ∂µ V = Qµ + Pµ .16). Under such transformations we have V(φ) → V(φ) h(x) . Subsequently.3 Nonlinear sigma models with homogeneous target space It is now rather straightforward to describe a nonlinear sigma model based on a homogeneous target space by making use of the above framework.47) The above expressions show that Qµ and Pµ are just the pull backs of the target space connection and vielbein to the spacetime. i ˜ −ξα Di + hα .

26). In the ﬁrst one regards the gauge ﬁeld Qµ as an independent ﬁeld. 2 (4.52) Here we made use of the commutation relations (4. Pν ] . Obviously the Lagrangian must be invariant under both the rigid G transformations and the local H transformations.49) Hence Qµ acts as a gauge ﬁeld associated with the local H transformations.5) There are several ways to write down the Lagrangian of the corresponding nonlinear sigma model.54) Clearly this Lagrangian is invariant under the group G. L = − 1 tr Pµ P µ . (4. From (4. (4. The result can be written as. as soon as one imposes a gauge and restricts V to a coset representative parametrized by certain ﬁelds φi .and in a second-order form.47) reads V −1 Dµ V = Pµ . (4. Qν ] = −[Pµ .53) L = 1 tr Dµ V −1 Dµ V . Dµ V = ∂µ V − VQµ . 2 One can interpret this result in a ﬁrst. Hence we write (4. Qµ (x) Pµ (x) −→ h−1 (x) Qµ (x) h(x) + h−1 (x) ∂µ h(x) . After imposing the gauge condition on V(x) one obtains the coset representative V(φ(x)). so that (4. −→ h−1 (x) Pµ (x) h(x) . When this is not done.50) (4. we introduced new degrees of freedom into V associated with the group H. After substituting the result one obtains the second-order form. which presupposes (4.48) we derive the following local H-transformations. At this stage one still has the full gauge invariance with respect to local H transformations and one can impose a gauge restricting V to a coset representative.51). It is convenient to introduce a corresponding H-covariant derivative. the theory is invariant under Grigid × Hlocal with both groups acting linearly. Pν] ] = 0 . Furthermore both Pµ and Qµ are invariant under rigid G-transformations. one derives the Cartan-Maurer equations (4. but until that point V will just be an unrestricted spacetime dependent element of the group G. = ∂[µ Pν] + [Q[µ . Eventually we will ﬁx this gauge freedom.B. whose ﬁeld equations are algebraic and are solved by (4. the . de Wit: Supergravity 67 By allowing ourselves to perform such local gauge transformations.51) from the beginning.51) Just as before. However. Fµν (Q) D[µ Pν] = ∂µ Qν − ∂ν Qµ + [Qµ .

gauge ﬁelds cannot couple in this way as their gauge invariance would be in conﬂict with the local invariance under the group H. In order to deal with complicated supergravity theories that involve homogeneous spaces. This combined subgroup still generates a representation of the group G. the matter ﬁelds transform nonlinearly under the group G with transformations that take the form of φ-dependent H-transformations. We emphasize that the presentation that we followed so far was rather general. the maximally symmetric spaces that we considered served only as an example. so that H is the maximally compact subgroup of G. As we intend to demonstrate. The exceptional groups are noncompact and are divided by their maximal compact subgroups. for the spaces of interest). so that one is always dealing with a manifest linearly realized symmetry group Grigid × Hlocal . 2 (4. Matter ﬁelds are assigned to a representation of the local H group. The corresponding spaces have dimension 70 and 42. However. respectively. gauge ﬁelds have to transform under the rigid group G. L = − 1 gij (φ) ∂µ φi ∂ µ φj .or secondorder formalism. but it is now realized in a nonlinear fashion. Usually the ﬁelds will remain invariant under the group G as long as one does not ﬁx the gauge and choose a speciﬁc coset representative. Also here we can proceed in ﬁrst.47) will acquire some extra terms that depend on the matter ﬁelds. this strategy allows for a systematic approach. In this form the local H invariance is absent. Upon choosing a gauge. symmetric target spaces.68 Unity from Duality: Gravity. . whereas the gauge-ﬁxed approach leads to unsurmountable diﬃculties (at least. These target spaces are usually noncompact and Riemannian.54) leads to the standard form of the nonlinear sigma model.31). the strategy is to postpone this gauge choice till the end. so that they couple to the sigma model ﬁelds through the connection Qµ that appears in the covariant derivatives.36). In ﬁrst-order form the equation (4. but the invariance under G is still there and realized as target space isometries generated by corresponding Killing vectors. Therefore. It is straightforward to demonstrate that (4. Gauge Theory and Strings residual subgroup is such that the H transformations are linked to the G transformations and depend on the ﬁelds φi . In this form it is easy to see how to couple matter ﬁelds to the sigma model in a way that the invariance under the isometries remains unaﬀected. Later in this chapter we will be dealing with the E7(7) /SU (8) and E6(6) /U Sp(8) coset spaces. determined by (4. In supergravity we are often dealing with sigma models based on homogeneous.55) where the target space metric is given by (4.

Fµν (Q) = = D[µ Pν] = ∂µ Qν − ∂ν Qµ + [Qµ . The modiﬁcations caused by the new minimal couplings are minor and the eﬀects can be concisely summarized by the Cartan-Maurer equations. Of course. but the derivative is now covariantized and modiﬁed by the terms depending on the new gauge ﬁelds Aµ . They remain expressed by (4. (4. the expressions for Qµ and Pµ will change. as we will discuss in the next section.B. (4. de Wit: Supergravity 4. As explained above. Qν ] . The consistency of this procedure is obvious as (4. [Pµ .58) When imposing a gauge condition. Pν ] − g V −1 Fµν (A)V 1 − 2 g V −1 Fµν (A)V G/H H . where we have introduced a coupling constant g to keep track of the new terms introduced by the gauging. one can gauge some or all of these isometries in the usual way: one elevates the parameters of the isometry group (of a subgroup thereof) to arbitrary functions of the spacetime coordinates and introduces the necessary gauge ﬁelds (with their standard gauge-invariant Lagrangian containing a kinetic term) and corresponding covariant derivatives. for sigma models based on homogeneous target spaces one can proceed in a way in which all transformations remain linearly realized. Because Qµ and Pµ now depend on the gauge connections Aµ .51) is fully covariant. When gauging isometries in a generic nonlinear sigma model .36) (in the gauge where V equals the coset representative). Hence. the last result for Pµ exhibits precisely the Killing vectors (4.51). µ (0) Pµ = Pµ − g[V −1 Aµ V]G/H .57) . To adopt this approach is extremely important for the construction of gauged supergravity theories. the original rigid invariance under G transformations from the left is now broken by the embedding of the new gauge group into G. With this change.4 Gauged nonlinear sigma models 69 Given a nonlinear sigma model with certain isometries. Since these new gauge transformations involve the isometry group they must act on the group element V as a subgroup of G. according to Qµ = Q(0) − g[V −1 Aµ V]H . We will always use the second-order formalism so that the H-connection Qµ will not be an independent ﬁeld. Hence the covariant derivative of V is now changed by the addition of the corresponding (dynamical) gauge ﬁelds Aµ which take their values in the corresponding Lie algebra (which is a subalgebra of the Lie algebra associated with G).56) Dµ V(x) = ∂µ V(x) − V(x) Qµ (x) − g Aµ (x) V(x) . (4.

However. Table 12) and the group G is also realized as a symmetry of the full theory. This gauging has an obvious KaluzaKlein origin. The modiﬁcations in the matter sector arise through the order g contributions to Qµ . Supersymmetry requires the presence of extra terms of order g and g 2 in the Lagrangian. It is an obvious question whether these theories allow an extension in which the Abelian gauge ﬁelds are promoted to non-Abelian ones. Gauge Theory and Strings (cf. 5 Gauged maximal supergravity in 4 and 5 dimensions The maximally extended supergravity theories introduced in Section 3 were obtained by dimensional reduction from 11-dimensional supergravity on a hypertorus. Generically the ﬁelds transform as follows. Apart from the gauge ﬁeld interactions there are fermionic masslike terms of order g and a scalar potential of order g 2 . one replaces the derivatives according to ∂µ φ → ∂µ φi −Aµ ξ i (φ). For example. Note that Pµ and Qµ are invariant under the new gauge group (but transform under local H-transformations. An obvious gauging in D = 4 dimensions is based on the group SO(8). which has an SO(8) isometry group. as before). They contain an extra parameter g. In the next section we will discuss the application of this formulation to gauged supergravity. in 4 dimensions the 28 Abelian vector ﬁelds do not constitute a representation of the group E7(7) . (4. The graviton is invariant.3. where for simplicity we assumed a single isometry. . but at the level of the combined ﬁeld equations and Bianchi identities. In this compactiﬁcation the four-index ﬁeld strength acquires a nonzero values when all its indices are in the four-dimensional spacetime. we concentrate on the maximal gauged supergravities in D = 4 and 5 spacetime dimensions. the (Abelian) gauge ﬁelds transform linearly under G and the fermions transform linearly under the group H. The group emerges as the gauge group of the compactiﬁed theory formulated in 4 dimensions. as the Lagrangian has a manifest SO(8) invariance and there are precisely 28 vector ﬁelds [63]. In this case the group G is realized by electric-magnetic duality and acts on the ﬁeld strengths. and arises when compactifying seven coordinates of D = 11 supergravity on the sphere S 7 . In this theory the E7(7) invariance group is broken to a local SO(8) group so that the resulting theory is invariant under SU (8)local × SO(8)local . which is the non-Abelian coupling constant.70 Unity from Duality: Gravity. rather than on the vector ﬁelds.55)). in some dimensions the G-invariance is not realized at the level of the action. We return to electric-magnetic duality in Section 5. This turns out to be possible and the corresponding theories are known as gauged supergravities. In these theories the scalar ﬁelds parametrize a G/H coset space (cf. The latter may give rise to groundstates with nonzero cosmological constant. To explain the construction of gauged supergravity theories.

In this solution the ﬁve-index (self-dual) ﬁeld strength acquires a nonzero value whenever the ﬁve indices take all values in either S 5 or in the ﬁve-dimensional spacetime. so that the symmetry group of the Lagrangian equals the SL(2) × SO(6) subgroup of E6(6) . This poses an obvious problem. Type-IIB supergravity has a manifest SL(2) invariance and the isometry group of S 5 is SO(6). In D = 5 dimensions they are E6(6) and U Sp(8). We will brieﬂy comment on this issue at the end of the section. where SO(6) is realized as a local gauge group. respectively. 22 Strictly speaking the isotropy groups are SU (8)/Z2 and U Sp(8)/Z2 . Before continuing with supergravity we ﬁrst discuss some basic features of the two coset spaces E7(7) /SU (8) and E6(6) /U Sp(8). Hµν 1 56 1 27 χijk 56 1 48 1 uijIJ . which leaves 12 Abelian gauge ﬁelds which are charged with respect to the same group. as the Abelian gauge transformations of these 12 ﬁelds will be in conﬂict with their transformations under the SO(6) gauge group. Again the Kaluza-Klein scenario can serve as a guide. Representation assignments for the various supergravity ﬁelds with respect to the groups G and H. Note that the tensors Fµν . This implies that 15 of the 27 gauge ﬁelds become associated with the non-Abelian group SO(6). as there is no obvious 27-dimensional gauge group. In D = 4 dimensions these groups are E7(7) and SU (8). Both these exceptional Lie groups can be introduced in terms of 56-dimensional matrices22 . it turns out that many other subgroups of E7(7) can be gauged. a eµ i ψµ Fµν . type-IIB supergravity has a compactiﬁcation on the sphere S 5 . In D = 5 dimensions the possible gaugings are not immediately clear.B. . de Wit: Supergravity 71 Table 16. Gµν . The solutions is to convert these 12 gauge ﬁelds to antisymmetric tensor ﬁelds. Gµν and/or Hµν denote the ﬁeld strengths of the vector ﬁelds and/or (for D = 5) possible tensor ﬁelds. The Lagrangian can thus be written in a form that is invariant under U Sp(8)local × SO(6)local × SL(2) [29]. v ijIJ 28 + 28 56 27 + 27 27 + 27 SU (8) E7(7) U Sp(8) E6(6) 1 1 1 1 8 1 8 1 However. Also in 5 dimensions other gauge groups are possible. While D = 11 supergravity has no obvious compactiﬁcation to ﬁve dimensions.

5) is invariant. (5. This group is the group of electric-magnetic dualities of maximal supergravity in D = 4 dimensions. (5. 8. ΛI I = 0 . IJ IJ (z1 . Gauge Theory and Strings 5. the number of the noncompact generators minus the number of compact ones . ΣIJKL = [K ΛI J = −ΛJ I . L = 1. for which the ΣIJKL is fully antisymmetric and the generators are further restricted according to ΛIJKL = δ[I ΛJ]L] . 1 24 εIJKLMN P Q ΣMN P Q .6) Obviously the matrices ΛIJ generate the group SU (8).1 On E7(7) /SU(8) and E6(6) /USp(8) cosets We discuss the E7(7) and E6(6) on a par for reasons that will become obvious. since ΣIJKL comprise 70 real parameters. where the indices are antisymmetrized index pairs [IJ] and [KL] and I. where ω and Ω are given by ω= 0 1 1 0 . .2) The corresponding group elements constitute the group Sp(56.1) where ΛIJKL and ΣIJKL are subject to the conditions (ΛIJKL )∗ = ΛIJKL = −ΛKLIJ . Consider now inﬁnitesimal transformations of the form. The deﬁning properties of elements E of Sp(56.72 Unity from Duality: Gravity. . J. Hence the (zIJ . In passing we note that the real subgroup (in this pseudoreal representation) is equal to the group GL(28). δzIJ δz IJ = = ΛIJKL zKL + ΣIJKL z KL . R) are E∗ = ω E ω . Because SU (8) is the maximal compact subgroup. . (5. which has dimension 63. z KL ) with z IJ = (zIJ )∗ . z KL ) span a 56-dimensional vector space. (5. ΛIJKL z KL + ΣIJKL zKL (5. (5. . R) in a pseudoreal basis. Let us now consider the E7(7) subgroup. Ω= 1 0 0 −1 . The matrices ΛIJ KL are associated with its maximal compact subgroup. the dimension of E7(7) equals 63+70 = 133. To deﬁne the groups we consider the fundamental representation.4) E −1 = Ω E † Ω . (ΣIJKL )∗ = ΣKLIJ .3) The above properties ensure that the sequilinear form. acting on a pseudoreal vector (zIJ . K. which is equal to U (28). z2 ) = z1 z2IJ − z1IJ z2 .

J ΩIK ΩKJ = −δI .6) with ΛIJ and ΣIJKL real. deﬁnes the group SL(8. For further information the reader is encouraged to read the appendices of [53]. ΩIJ ΣIJKL = 0 . It is possible to choose conventions in which the E7(7) matrices have a diﬀerent block decomposition than (5. It is straightforward to show that these matrices close under commutation and generate the group E7(7) . The group E7(7) has a quartic invariant. To show this one needs a variety of identities for selfdual tensors [64]. which. The other restrictions on the generators concern ΣIJKL . To see this we observe that the following restrictions are preserved by the group. Because of the constraints (5. We note that the subgroup generated by (5. one of them is that the contraction ΣIKLM ΣJKLM is traceless.7) IJKLMN P Q εIJKLMN P Q z IJ KL MN P Q z z z +ε zIJ zKL zMN zP Q . which is not compact. de Wit: Supergravity 73 is equal to 70 − 63 = 7. The maximal compact subgroup U Sp(8) thus has dimension 64 − 28 = 36. This is the group U Sp(8). zIJ = ±ΩIK ΩJL z KL .1) is reducible and decomposes into two singlets and a 27 and a 27 representation. however. This is the group SL(8). rather than to SU (8). .10) The ﬁrst one suppresses the singlet representation and the second one projects out the 27 or the 27 representation. Here one introduces a skew-symmetric tensor ΩIJ .1) and where the diagonal blocks correspond to the group SL(8).8) Now we restrict ourselves to the subgroup of U (8) that leaves ΩIJ invariant. while the diﬀerence between the numbers of noncompact and compact generators equals 42 − 36 = 6. Altogether we thus have 36 + 42 = 78 generators. E7(7) has another maximal 63-dimensional subgroup.9) ΣIJKL = ΩIM ΩJN ΩKP ΩLQ ΣMN P Q . J4 (z) = zIJ z JK zKL z LI − 1 (zIJ z IJ )2 4 1 + 96 (5. plays no role in the following. (5. satisfying ΩIJ = −ΩJI . However. while there are 70 − 28 = 42 generators associated with ΣIJKL . These numbers conﬁrm that we are indeed dealing with E6(6) and its maximal compact subgroup U Sp(8). (5. for which the restrictions are rather similar.B. (ΩIJ )∗ = ΩIJ . R). ΛIJKL = δ[I ΛJ]L] . ΩIJ z IJ = 0 . (5. [K ΛI J = −ΛJ I . Λ[I K ΩJ]K = 0 . Another subgroup is the group E6(6) . Altogether we have the conditions.9) the 56-dimensional representation deﬁned by (5.

For those and other details we refer to the original literature [29.12) where the rank-4 antisymmetric tensor Σ satisﬁes the algebraic restrictions appropriate for the exceptional group. hopefully. Gauge Theory and Strings The group E6(6) has a cubic invariant. A coset representative is obtained by exponentiation of the generators deﬁned in (5. which is a 56 × 56 matrix that belongs to the group E7(7) or E6(6) . 5. deﬁned by J3 (z) = z IJ z KL z MN ΩJK ΩLM ΩN I . where the gauge group is embedded into the group SL(6).11) which plays a role in the E6(6) invariant Chern-Simons term in the supergravity Lagrangian.1). where we have listed these spaces). that will be relevant in the following. (5. etc. V(x) = exp 0 Σ(x) Σ(x) 0 . which is noncompact. We will not be complete here but sketch a number of features of the maximal supergravity theories in D = 4. spinor conventions. without getting entangled in issues that depend on the spacetime dimension. Because the spinless ﬁelds always appear in nonpolynomial form. One starts by introducing a so-called 56-bein V(x). We will be rather cavalier about numerical factors. and which plays a role in many of the known gaugings. As explained in the previous section.2 On ungauged maximal supergravity Lagrangians An important feature of pure extended supergravity theories is that the spinless ﬁelds take their values in a homogeneous target space (cf. In this way we will. be able to bring out the main features of the G/H structure. Table 12. the 56-bein is reducible for E6(6) . which has dimension 35 + 3 = 38. so that SL(2) remains as a rigid invariance group of the Lagrangian.74 Unity from Duality: Gravity. This is the group SL(6) × SL(2). just as in the previous sections. depending on whether we are in D = 4. we assume that the 56-bein transforms under the exceptional group from the left and under the local SU (8) (or . There is another maximal subgroup of E6(6) . (5. Hence. as well as in the gauging. or 5 dimensions. 5 where the coset structure plays an important role. Our notation will be based on a description in terms of right cosets. Schematically. which may diﬀer from the notations used in the original references where one sometimes uses left cosets. but we will use the reducible version in order to discuss the two theories on a par. 63]. it is vital to exploit the coset structure explained in the previous section in the construction of the supersymmetric action and transformation rules.

[kl]). In the above the row indices are ([IJ]. [KL]). 0. i.e.B. conversely. upon interchanging V and V −1 .. Observe also that (5. e.3) the inverse V −1 can be expressed in terms of the complex conjugates of the submatrices of V. (5.16) The corresponding equations for E6(6) are identical.18) with similar identities for the v ijIJ . de Wit: Supergravity 75 U Sp(8)) from the right.14) V −1 (x) = klIJ kl v (x) u KL (x) Consequently we derive the identities. The notation of the submatrices is chosen such as to make contact with [63]. 8 (5. According to (5. 8 IJ IJ δKL → δKL − 1 ΩKL ΩIJ . ij = δkl . The latter are the indices associated with the local SU (8) or U Sp(8). V(x) = (5. where left cosets were chosen. uijIJ uklIJ − v ijIJ vklIJ uijIJ v klIJ − v ijIJ uklIJ or. we have ﬁxed the gauge with respect to local SU (8) or U Sp(8).13) ijKL KL −v (x) ukl (x) with the usual conventions uijIJ = (uijIJ )∗ and vijIJ = (v ijIJ )∗ . uijKL ΩIK ΩJL = Ωik Ωjl uklIJ . (5. for E7(7) . Observe that the indices of the matrix are antisymmetrized index pairs [IJ] and [ij]. The 56-bein can be decomposed in block form according to ij −vklIJ (x) u IJ (x) . IJ uij (x) vijKL (x) . = 0.13) gauge ﬁxing is not assumed.g. uijIJ ΩIJ = 0 . whereas in (5. In this case the (pseudoreal) matrices uij IJ ± ΩIK ΩJL v ijKL and their complex conjugates deﬁne (irreducible) .15) (5. (5.17) Furthermore the matrices u and v vanish when contracted with the invariant tensor Ω and they are pseudoreal. except that the antisymmetrized Kronecker symbols on the right-hand sides are replaced according to ij ij δkl → δkl − 1 Ωkl Ωij . and the column indices are ([ij]. uijIJ uijKL − vijIJ v ijKL uijIJ vijKL − vijIJ uijKL = = IJ δKL .12) is a coset representative.

. Hence. Qµ ijkl ijkl Pµ = = uijIJ ∂µ uklIJ − vijIJ ∂µ v klIJ . We note the identity uij IJ + ΩIK ΩJL v ijKL ukl IJ − ΩIM ΩJN vklMN ij = δkl − 1 Ωkl Ωij . ijkl (5. Gauge Theory and Strings elements of E6(6) corresponding to the 27 and 27 representations.19) In this case we can thus decompose the 56-bein in terms of a 27-bein and a 27-bein. ijkl Pµ = 1 24 εijklmnpq Pµ mnpq . so that Qµ ij equals Qµ ij = 2 3 [k (5.76 Unity from Duality: Gravity.25) Because of the underlying Lie algebra the connections Qµ ij satisfy Qµij = −Qµj i and Qµi i = 0.24) uikIJ ∂µ ujkIJ − vikIJ ∂µ v jkIJ . to the 42-dimensional representation of U Sp(8). V −1 ∂µ V = which leads to the expressions. Therefore it transforms according to the 70-dimensional representation of SU (8). with the reality condition. v ijIJ ∂µ uklIJ − uijIJ ∂µ v klIJ . Qµ ijkl = δ[i Qµ j]l] . with the reality condition.20) (5. (5.21) ijkl The important observation is that Qµijkl and Pµ are subject to the same constraints as the generators of the exceptional group listed in the previijkl ous section. 8 (5. Likewise Qµ transforms as a connection associated with SU (8) or U Sp(8).23) Pµ = Ωim Ωjn Ωkp Ωlq Pµ mnpq . Hence Qµ ijkl must satisfy the decomposition. Subsequently we evaluate the quantities Qµ and Pµ . respectively. (5. (5.22) or. as well as an extra symmetry condition in the case of U Sp(8) (cf. (5. Pµ is fully antisymmetric and subject to a reality constraint. Qµ ijmn klmn Pµ Pµ ijpq Qµklpq .9)).

6) or (5. de Wit: Supergravity Furthermore we can evaluate the Maurer-Cartan equations (4. Dµ Σijkl = ∂µ Σijkl + 2Qµ m[i Σjkl]m . where Σ is an H-covariant expression proportional to the supersymmetry parameter i and the fermion ﬁelds χijk . up to a local H-transformation. (5.28) is consistent with both groups G and H.28) 0 Σ Σ 0 . (5.e. the generators denoted by k in the previous section). we can avoid the nonlinearities completely and carry out the calculations in a transparent way. Under this variation the quantities Qµ and Pµ transform systematically. Because Σ takes the form of an H-covariant tensor.27). (5.9).29) Observe that. as δV = V or. Before ﬁxing a gauge. in terms of submatrices. Hence it must be of the form Σijkl ∝ ¯[i χjkl] . Fµν (Q)i j ijkl D[µ Pν] 77 = ∂µ Qν ij − ∂ν Qµ ij + Q[µ ik Qν]kj = − 4 P[µjklm Pν]iklm . δuijIJ = −Σijkl v klIJ .B. Continuing along similar lines we turn to a number of other features that are of interest for the Lagrangian and transformation rules. (5.e.27). (5. Such decompositions are an important tool for dealing with the spinless ﬁelds in this nonlinear setting. δvijIJ = −Σijkl uklIJ . The supersymmetry variation can be written in the form (5. . As was stressed above. any variation of V can be decomposed into (5.20). Fixing the gauge prematurely and converting to a speciﬁc coset representative for G/H would lead to unsurmountable diﬃculties. The ﬁrst one is the observation that any variation of the 56-bein can be written..52). Furthermore Σ must satisfy the restrictions associated with the exceptional group. the variation (5.27) where Σijkl is the rank-four antisymmetric tensor corresponding to the generators associated with G/H (i. up to a local H-transformation. up to complex conjugation and possible contractions with H-covariant tensors. δQµ ij ijkl δPµ = = 2 3 jklm Σjklm Pµ iklm − Σiklm Pµ . this establishes that the SU (8) tensors Qµ and Pµ can be assigned to the adjoint representation of the group G. 3 ijkl = ∂[µ Pν] + 2Q[µ m[i Pν] jkl]m = 0. In particular this applies to supersymmetry transformations. as is also obvious from the decomposition (5.26) Observe that the use of the Lie algebra decomposition for G/H is crucial in deriving these equations. i.

The last term represents the 1 Noether coupling term for the spin-0/spin. spinors and scalars take the following form.9). For the ﬁrst term ijkl we note that δ1 Σ2 leads to a term proportional to Pµ . (5.78 Unity from Duality: Gravity. this contribution can be written in the form [δ1 .28) applied in diﬀerent orders on the 56-bein. Apart from higher-order spinor terms.20). up to a local H-transformation proportional to Qµ . ω) − 1 ψµ γ µνρ (∂ν − 1 ων γab )δi + 1 Qν ij ψρj 2 2 2 ab l 1 − 12 χijk γ µ (∂µ − 1 ωµ γab )δk + 3 Qµ kl χijl − ¯ 2 2 ijkl 1 − 12 χijk γ ν γ µ ψνl Pµ . Taking into account the various H-covariant combinations in the actual expressions implied by (5. Here we distinguish the Einstein-Hilbert term for gravity. as follows from (5. so that up to proportionality constants we must have a variation. the Dirac Lagrangian and the nonlinear sigma model associated with the G/H target space. ¯ ijkl 1 96 Pµ µ Pijkl (5. δ2 ] V ∝ (¯i γ µ 1 2i [i jkl]m − ¯i γ µ 2 1 i) V 0 Pµ Pµ 0 . of the form (¯1 γ µ 2 m ) Pµ . ijkl δχijk ∝ Pµ γ µ l . (5. Hence up to a number of ﬁelddependent H-transformations. 0 Σ2 Σ2 0 . the Rarita-Schwinger Lagrangian for the gravitini. we have [δ1 . which incorporates the spacetime derivatives of the spinless ﬁelds. Let us now turn to the action.32) This is precisely a spacetime diﬀeomorphism. e−1 L1 = j ab 1 ¯i − 2 R(e.6) or (5. the supersymmetry commutator closes on V into a spacetime diﬀeomorphism (up to terms of higher-order in the spinors that we suppressed). (5.2 system.30) The veriﬁcation of the supersymmetry algebra on V is rather easy.33) This Lagrangian is manifestly invariant with respect to E7(7) or E6(6) . Gauge Theory and Strings The supersymmetry variation of the spinor χijk contains the quantity ijkl Pµ .31) The last term is just an inﬁnitesimal H-transformations. For D = 4 the fermion ﬁelds are chiral spinors and we have to add the contributions from the . the terms in the Lagrangian pertaining to the graviton. combined with two supersymmetry parameters. 1 and 2 . respectively. Under two consecutive (ﬁeld-dependent) variations (5. δ2 ] V = V 0 2 δ[1 Σ2] 2 δ[1 Σ2] 0 +V 0 Σ1 Σ1 0 . gravitini.

can lead to an equivalent set of ﬁeld equations and Bianchi identities. Here Oij denotes a covariant tensor antisymmetric in both spacei time and U Sp(8) indices and quadratic in the fermion ﬁelds. up to terms quartic in the fermion ﬁelds. and which can be used to convert E6(6) to U Sp(8) indices.3 Electric–magnetic duality and E7(7) For D = 4 the Lagrangian is not invariant under E7(7) but under a smaller group. We continue the discussion of the D = 4 theory in the next section. so that they satisfy the reality KL constraint Bµ IJ = ΩIK ΩJL Bµ .35) where we distinguish the kinetic term. for D = 5 we are dealing with so-called symplectic Majorana spinors. µν (5. The result obtained by combining the Lagrangians (5. which acts on the vector ﬁelds (but not necessarily on the 56-bein) according to a 28-dimensional subgroup of GL(28). 5.B. uijIJ +v ijKL ΩIK ΩJL . µν (5. The vector ﬁelds bring in new features. and the Lagrangian is manifestly invariant under the corresponding transformations. It is impossible to construct an invariant action just for the vector ﬁelds and one has to make use of the scalars. Here we disregard such details and concentrate on the symmetry issues. the phenomenon is not restricted to . where the symmetry group of ﬁeld equations and Bianchi identities can be bigger than that of the Lagrangian. de Wit: Supergravity 79 spinors of opposite chirality. This situation is typical for D = 4 theories with Abelian vector ﬁelds. the combined equations of motion and the Bianchi identities are invariant under the group E7(7) . ψµ and χijk . as this requires to ﬁrst introduce the concept of electric-magnetic duality. which can be written in terms of the 27-beine. which are diﬀerent for spaceIJ time dimensions D = 4 and 5. However.35) gives the full supergravity Lagrangian invariant under rigid E6(6) and local U Sp(8) transformations. and where diﬀerent Lagrangians not related by local ﬁeld redeﬁnitions.11) and a moment coupling with the µν fermions. However. a Chern-Simons interaction associated with the E6(6) cubic invariant (5. L2 = 1 − 16 e Fij Fµνkl Ωik Ωjl µν IJ KL MN 1 − 12 εµνρσλ Bµ ∂ν Bρ ∂σ Bλ ΩJK ΩLM ΩN I µν 1 + 4 e Fij Oij .34) The invariant Lagrangian of the vector ﬁelds then reads. Observe that the dependence on the spinless ﬁelds is completely implicit. Any additional dependence would aﬀect the invariance under E6(6) . Hence we deﬁne a U Sp(8) covariant ﬁeld strength for the vector ﬁelds.33) and (5. In D = 5 dimensions the vector ﬁelds Bµ transform as the 27 representation of E6(6) . equal to IJ IJ Fij = (uijIJ − v ijKL ΩIK ΩJL )(∂µ Bν − ∂ν Bµ ) .

[52]). which can only have a solution (for generic Lagrangians) provided that the matrix . such that the new tensors Gµν follow from this Lagrangian as in (5.37) can again follow from a Lagrangian. e.36) ±α Together Fµν and G± α comprise the electric and magnetic ﬁelds and inµν ductions. n × n submatrices and β n denotes the number of independent gauge potentials. in Minkowski space. The Bianchi identities and equations of motion for the Abelian gauge ﬁelds take the form ∂µ F + − F − α µν = ∂ µ G+ − G− µν α = 0. β. Wαβ and Z αβ are constant. .38) where U α . . . ﬁeld strengths. rotated. but for the moment we will remain more general and label the ﬁeld strengths by α. so that the rotation is associated with a 56 × 56 matrix.g. In N = 8 supergravity we have 56 such ﬁeld strengths of each duality. where one can perform (Hodge) duality rotations.g. see. The 4-dimensional version has been known for a long time and is commonly referred to as electric-magnetic duality (for a recent review of this duality in supergravity. Its simplest form arises in Maxwell theory in four-dimensional (ﬂat or curved) Minkowski space. Vαβ . This duality can be generalized to any D = 4 dimensional ﬁeld theory with Abelian vector ﬁelds and no charged ﬁelds. Gauge Theory and Strings 4 dimensions and can occur for antisymmetric tensor gauge ﬁelds in any even number of spacetime dimensions (see. [51]). real. which commute with the Lorentz group and rotate the electric and magnetic ﬁelds and inductions according to E ↔ H and B ↔ D. are related by complex conjugation) and so are the ﬁeld strengths G± α that appear in µν the ﬁeld equations. This condition amounts to an integrability condition. so that the gauge ﬁelds enter the Lagrangian only through their (Abelian) ﬁeld strengths. rotations of the ﬁeld strengths F ± and G± ±α Fµν G± β µν −→ U W Z V ±α Fµν G± β µν . (5. e. which are deﬁned by G± α = ± µν ∂L 4i · e ∂F ± αµν (5.80 Unity from Duality: Gravity. These ﬁeld strengths (in the case at hand we have 28 of them.36). labelled by antisymmetric index pairs [IJ]. constant.) are decomposed into selfdual and ±α anti-selfdual components Fµν (which. More precisely.37) which are obviously invariant under real. (5. The relevant question is whether the rotated equations (5. does there exist a new Lagrangian depending on the new.

R).39) We distinguish two subgroups of Sp(2n. there exist diﬀerent Lagrangians with diﬀerent symmetry groups. n. but in the context of a duality invariance rather than of a reparametrization. R) was ﬁrst derived in [65]. One is the invariance group of the combined ﬁeld equations and Bianchi identities. Of course. a generic theory does not have such an invariance group. (5.13)) in a way that we will discuss below. (5. R)23 . Furthermore the Lagrangian is not uniquely deﬁned (it can always be reparametrized via an electric-magnetic duality transformation) and neither is its invariance group. Z TV = V TZ . we form the linear combinations. which usually requires the other ﬁelds in the Lagrangian to transform as well. as this requires the gauge group to be contained in the invariance group of the Lagrangian. +AB F+ AB = 1 (i G+ AB + Fµν ) . R). Furthermore. de Wit: Supergravity 81 is an element of the group Sp(2n. This implies that the submatrices satisfy the constraint U TV − W TZ = V U T − W Z T = 1 . Observe that these groups do not constitute an invariance of the theory. which are related to the index pairs [IJ] of the 56-bein (cf. we assign diﬀerent indices to the ﬁeld strengths and the underlying gauge groups than to the 56-bein V. but maximal supergravity is known to have an E7(7) ⊂ Sp(56. R). U TW = W TU . Namely. whose Bianchi identities and equations of motion are the same (modulo a linear transformation) and are invariant under the same group (which contains the symmetry groups of the various Lagrangians as subgroups). to remain in the context of the pseudoreal basis used previously. constitute the group Sp(2n. is usually smaller and restricted by Z = 0 and U −1 = V T . one can show that in Minkowski spaces of dimensions D = 4k. the rotations of the ﬁeld equations and Bianchi identities associated with n rank-(k − 1) antisymmetric gauge ﬁelds that are described by a Lagrangian. Given the fact that we can rotate the ﬁeld strengths by electric-magnetic duality transformations. The subgroup that is a symmetry of the Lagrangian. this invariance group is not necessarily realized as a symmetry of the Lagrangian.B. we label the ﬁelds strengths by independent index pairs [AB]. the subgroup associated with the matrices U equals GL(n). The fact that the symplectic redeﬁnitions of the ﬁeld strengths constitute the group Sp(2n. . 1µν µν 2 +AB F+AB = 1 (i G+ AB − Fµν ) . this group equals SO(n. In this respect our presentation is more in the spirit of a later treatment in [66] for N = 2 vector multiplets coupled to supergravity (duality invariances for these theories were introduced in [67]). For rank-k antisymmetric gauge ﬁelds in D = 2k + 2 dimensions.40) 2µν µν 2 23 Without any further assumptions. R) invariance. These issues are extremely important when gauging a subgroup of the invariance group. However. but merely characterize an equivalence class of Lagrangians. (5. More precisely.

which is simply deﬁned by the above condition.42) one can straightforwardly determine the relevant terms in the Lagrangian. For convenience. V −1 E F+ AB 1µν F+AB 2µν = F+ ij µν +kl Oµν . R)/GL(28). the real GL(28) subgroup is induced by corresponding linear transformations of the vector ﬁelds. Without going into the details we mention that the chirality + and duality of Oµν is severely restricted so that the structure of (5.3)). the 56-bein plays a crucial role. R) according to the matrices E speciﬁed in (5. the relation among the ﬁeld strengths must have the form. In the basis (5. On the other hand. 28 × 28 matrices in the oﬀ-diagonal blocks. in elements of E7(7) \Sp(56.3).42) + where Oµν is an SU (8) covariant tensor quadratic in the fermion ﬁelds and independent of the scalar ﬁelds. which are manifestly invariant under diﬀerent subgroups of E7(7) . R) matrices E. E= UIJAB VKLAB VIJCD UKLCD . i. The tensor F+ ij is an SU (8) covariant ﬁeld strength that µν appears in the supersymmetry transformation rules of the spinors. [63]). (5. but such a redeﬁnition cannot be carried out at the level of the Lagrangian. The various Lagrangians are thus encoded in Sp(56. R) matrix E (so that it satisﬁes the conditions (5. Hence the matrix E allows the ﬁeld strengths and the 56-bein to transform under E7(7) in an equivalent but nonidentical way. The relation involves a constant Sp(56. Gauge Theory and Strings Anti-selfdual ﬁeld strengths (F−AB . One could consider absorbing this matrix into the deﬁnition of the ﬁeld strengths (F1 . let us AB remember that the tensors Fµν and Gµν AB are related by (5.42) is unique (cf. To exhibit how one can deal with a continuous variety of Lagrangians. it can be absorbed into the 56-bein V. F2 ). symmetric. which appears in the moment couplings in the Lagrangian.41) On the basis of E7(7) and SU (8) covariance. when E ∈ E7(7) .36) and this relationship must be consistent with E7(7) . From (5. In order to establish this consistency.82 Unity from Duality: Gravity.e. we redeﬁne the 56-bein by absorbing the .3) the generators of GL(28) have a block decomposition with SO(28) generators in both diagonal blocks and identical real. (5. up to multiplication by GL(28) from the right and multiplication by E7(7) from the left. F− AB ) follow from complex conjuga1µν 2µν tion. On this basis the ﬁeld strengths rotate under Sp(56. unless it belongs to a GL(28) subgroup which can act on the gauge ﬁelds themselves.

rather than with I. (5. +AB L3 = − 1 e Fµν F +CD µν [(u + v)−1 ]ABij (uijCD − v ijCD ) 8 +AB − 1 e Fµν [(u + v)−1 ]ABij O+µν ij 2 + h. . .38) and (5. (5. de Wit: Supergravity matrix E. .6). According to the ˆ analysis leading to (5. Note. The gauge ﬁelds transform under the real subgroup (i.. but only after a proper conversion of the I. however. . .47) In order to be a subgroup of E7(7) as well. . .45) AB where the 28 × 28 matrices satisfy [(u + v)−1 ]ABij (uijCD + v ijCD ) = δCD . this subgroup is generated on V by matrices Λ and Σ. Because the matrix E−1 V is an element of Sp(56.44) ν µ and which take the form. one has to make a certain choice for the matrix E. B. the matrix multiplying the two ﬁeld strengths in (5. but with indices A.c.B.1). R). that the E7(7) tensors Qµ and Pµ are not aﬀected by ˆ the matrix E and have identical expressions in terms of V and V. µνij (5. satisfying Im ΣABCD + ΛABCD = 0 . . AB Fµν = ∂µ AAB − ∂ν AAB . the imaginary parts of 24 Such symmetry properties follow from the symmetry under interchanging index pairs in the products (uijAB − vijAB ) (uklAB + vklAB ) and (uijAB + vijAB ) (uijCD + vijCD ).43) (x) ˆ where we have to remember that V is now no longer a group element of E7(7) . B. In order that the Lagrangian be invariant under a certain subgroup of E7(7) . . J. just as in (5. This is not the case for the terms in the Lagrangian that contain the Abelian ﬁeld strengths. (5. . J.. (5.e.. to A. indices. and is equal to +AB +ij Fµν = (uijAB + v ijAB ) F+ − (uijAB + vijAB ) Oµν .39). . . they must also satisfy (5.46) Clearly the Lagrangian depends on the matrix E. . ˆ V(x) = E−1 V(x) = uijAB (x) −v ij CD 83 −v kl AB (x) uklCD (x) . + The SU (8) covariant ﬁeld strength Fµνij will appear in the supersymmetry transformation rules for the fermions.45) is symmetric under the interchange of [AB] ↔ [CD]24 . .

TCD ] = fAB. in D = 5 this obstruction can be avoided. When the latter are charged under the gauge group. we will mainly concentrate on this theory. as one can deduce from the symmetry groups that are realized in maximal supergravity in higher dimensions. so this subtlety does not arise. then there is a potential obstruction to the gauging as the gauge invariance of these gauge ﬁelds cannot coexist with the nonAbelian gauge transformations. Typically this conversion of vector into tensor ﬁelds leads to terms that are inversely proportional to the gauge coupling [68]. This aspect is most nontrivial for the D = 4 theory. The biggest group whose existence can be inferred in this way. one must start from a Lagrangian that is symmetric under the desired gauge group.49) It turns out that the viability for a gauging depends sensitively on the choice of the gauge group and its corresponding embedding.4 Gauging maximal supergravity. The latter can transform under the gauge group. 6−p) involves 15 non-Abelian gauge ﬁelds and 12 antisymmetric tensor ﬁelds.CD TEF . (5. However. Therefore. In D = 5 dimensions. When eﬀecting the gauging the vector ﬁelds may decompose into those associated with the non-Abelian gauge group and a number of remaining gauge ﬁelds. 1). because (charged) vector ﬁelds can alternatively be described as antisymmetric rank-2 tensor ﬁelds. in view of electric-magnetic duality. the Lagrangian is manifestly symmetric under E6(6) . the gauging of SO(p.48) where the gauge group generators TAB are 56 × 56 matrices which span a subalgebra of maximal dimension equal to the number of vector ﬁelds. For instance. the T-tensor The gauging of supergravity is eﬀected by switching on the gauge coupling constant. because they are not realized as tensor gauge ﬁelds. after assigning the various ﬁelds to representations of the gauge group embedded in E7(7) or E6(6) . 5. embedded in the Lie algebra of E7(7) or E6(6) .84 Unity from Duality: Gravity. which requires one to make a suitable choice of the matrix E. A large variety of symmetry groups exists. In D = 4 dimensions. Hence the Abelian ﬁeld strengths are changed to non-Abelian ones and derivatives of the scalars are covariantized according to ∂µ V → ∂µ V − gAAB TAB V . . Only the gauge ﬁelds themselves and the spinless ﬁelds can transform under this gauge group. However. The structure constants of the gauge group are given by EF [TAB . µ (5. is E6(6) × SO(1. which is the group that one obtains from the D = 5 Lagrangian upon reduction to D = 4 dimensions. Gauge Theory and Strings the generators act exclusively on the 56-bein).

because the new terms in the Lagrangian lead to new variations. When establishing supersymmetry of the action one needs the CartanMaurer equations at an early stage to cancel variations from the gravitino kinetic terms and the Noether term (the term in the Lagrangian proportional to χψµ Pν ). provided we change the generators TAB into ˆ TAB = E−1 TAB E . 2 (5.53) v ijIJ (∆AB uklIJ ) − uijIJ (∆AB v klIJ ) where ∆AB u and ∆AB v indicate the change of submatrices in V induced by multiplication with the generator TAB . For convenience we now restrict ourselves to D = 4 dimensions. on which the E7(7) transformations act in the basis that is appropriate for the ﬁeld strengths. The leading variations are induced by the modiﬁcation (5.54) This is done below. Note that we could have expressed ˆ the above quantities in terms of the modiﬁed 56-bein V. in the basis appropriate for V. (5. . implies that the quantities QAB and PAB satisfy the constraints. (5. de Wit: Supergravity 85 to write down a corresponding Lagrangian requires an even number of tensor ﬁelds.B. This modiﬁcation was already noted in (4. The order-g terms in the Maurer-Cartan equation yield ¯ .51) where V −1 TAB V = QAB ijmn klmn PAB These modiﬁcations are the result of the implicit dependence of Qµ and Pµ on the vector potentials AAB . (5.50) PAB ijpq QABklpq .48) of the Cartan-Maurer equations. The fact that the matrices TAB generate a µ subalgebra of the algebra associated with E7(7) . ijkl PAB = = QAB ijkl 1 ijklmnpq PAB mnpq 24 ε [k δ[i QAB j]l] . It is straightforward to write down the explicit expressions for QAB and PAB .52) while QABij is antihermitean and traceless. Introducing the gauging leads directly to a loss of supersymmetry. 3 AB ijkl − 1 g Fµν PAB .57) and takes the form Fµν (Q)i j ijkl D[µ Pν] = = AB − 4 P[µjklm Pν]iklm − g Fµν QAB i j . QAB ij ijkl PAB = = 2 3 uikIJ (∆AB ujkIJ ) − vikIJ (∆AB v jkIJ ) . (5.

58) This shows that the SU (8) covariant T -tensors can be assigned to a representation of E7(7) . which is a special feature of D = 4 dimensions. (5. Along the same lines as before we can show that the SU (8) tensors QAB and PAB transform according to the adjoint representation of E7(7) . We will concentrate on properties which are perhaps less obvious. which follows directly from (5.86 Unity from Duality: Gravity.28). which is known as the T -tensor. which allows one to derive. First we observe that SU (8) covariantized variations of the T -tensor are again proportional to the T -tensor. 3 Σp[ijk Tl] j Σklmn Timn . This includes the SU (8) covariant derivative of the T -tensor. these properties are generic. The ﬁrst variation is proportional to an SU (8) tensor Tijkl . Tijkl = = j 3 4 QAB i 1 2 (uklAB + v klAB ) (5. Apart from the ˆ distinction between V and V.27) and (5. They are linearly proportional to the fermion ﬁelds and read.57) mnAB ˆ vijCD (∆AB uklCD ) ). These variations are induced by (5. δTijkl mn δTijkl = = kl Σjmnp Timnp − kl 1 jmnpqrst Σimnp Tqrst + 24 ε pmn pqrs 4 1 − 24 εijklpqrs Σmntu Ttu . Some of them are rather obvious (such as Tiijk = 0). Gauge Theory and Strings the leading variations of the Lagrangian.58) upon the . Before completing the analysis leading to a consistent gauging we stress that all variations are from now on expressed in terms of the T -tensor. δL = ρ µν i 1 4 g(¯j γ γ ψρ 1 + 288 εijklmnpq − ¯i γ ρ γ µν ψρj ) QAB i j (uklAB + v klAB ) F+ µνkl χijk γ µν l PAB mnpq (ursAB + v rsAB ) F+ ¯ µνrs (5. Another component of the T -tensor appears in the second variation and is equal to mn Tijkl = = 1 2 PAB ijkl 1 2 (umnAB + v mnAB ) − ˆ uijCD (∆AB vklCD ) (umnAB +v (5. ˆ The T -tensor is thus a cubic product of the 56-bein V which depends in a nontrivial way on the embedding of the gauge group into E7(7) .56) ˆ ˆ uimCD (∆AB ujmCD ) − vimCD (∆AB v jmCD ) (uklAB + v klAB ) . It satisﬁes a number of important properties. and follow rather straightforwardly from the deﬁnition. as its variations yield again the same tensor.c. ˆ ˆ ˆ ˆ where ∆AB u and ∆AB v are the submatrices of TAB V.55) + h. This property will play an important role below.

A viable gauging requires that the T -tensor satisﬁes a number of rather nontrivial identities. must be linearly related to the T -tensor. The transformation properties were given in (5. √ ¯i δg ψµ = − 2g Aij ¯j γµ . de Wit: Supergravity 87 substitutions δ → Dµ and Σ → Pµ . First of all.lmn χijk χlmn + h. 56 × 133 = 56 + 912 + 6480 . ¯ 3 (5. . irrespective of the gauge group. A1 is symmetric in (ij). Note that the tensors Aij . P (V) = g 2 jkl 2 1 24 |A2i | − 1 |Aij |2 · 1 3 (5. The three SU (8) covariant tensors. irrespective of the gauge group.62) .B. which depend only on the spinless ﬁelds. A2ijkl and Aijk. to cancel the variations (5.c.59) whose presence necessitates corresponding modiﬁcations of the supersymmetry transformations of the fermion ﬁelds. Let us ﬁrst describe these new terms. (5. A3: 28 + 420 + 1176 + 1512 .59). but the new terms in the Lagrangian and transformation rules have a universal form. let us analyze the SU (8) content of the T -tensor. A2: 28 + 420 . A1 . This implies that these tensors transform under SU (8) according to the representations A1: 36 . To see how this can be the case. √ j ¯ ¯i Lmasslike = g e 1 2 A1ij ψ i γ µν ψν + 1 Ajkl ψµ γ µ χjkl µ 2 6 2i +Aijk. and as such is constitutes a certain tensor that transforms under E7(7) .55) we need masslike terms in the Lagrangian.lmn have certain symmetry prop1 3 erties dictated by the way they appear in the Lagrangian (5. because they were introduced for the purpose of cancelling the variations (5. as we will discuss shortly. A2 is fully antisymmetric in [jkl] and A3 is antisymmetric in [ijk] as well as in [lmn] and symmetric under the interchange [ijk] ↔ [lmn]. 1 δg χijk = −2g A2lijk ¯l .58). where we made use of the fact that the T -tensor consists of a product of the fundamental times the adjoint representation of E7(7) .60) Finally at order g 2 one needs a potential for the spinless ﬁelds. As we mentioned already. A2 and A3 . the T -tensor is cubic in the 56-bein. Hence the T -tensor comprises the representations. (5.61) These last three formulae will always apply.55). To be speciﬁc.

as follows from the deﬁnition (5. we should stress once more that many of the above features are generic and apply in other dimensions. which suppresses the 28 representation of the T -tensor.lmn 3 (5. (5. which gave rise to Tijkl mn Tijkl = = = 3 − 4 A2ijkl + 3 A1 δi . Comparing these representations to the SU (8) representations to which the tensors A1 −A3 (and their complex conjugates) belong. respectively.64) Note that these conditions are necessary.58) the T -tensor must be restricted by suppressing complete representations of E7(7) in order that its variations and derivatives remain consistent. 2 j[k l] 4 − 3 δ[i Tjkl] .65) Observe that a contraction with the ﬁrst upper index is also equal to zero.88 Unity from Duality: Gravity.62). but not suﬃcient as one also needs nontrivial identities quadratic in the T -tensors in order to deal with the variations of the Lagrangian of order g 2 . Indeed this was found by explicit calculation. Hence the T -tensor transforms under E7(7) according to the 912 representation which decomposes into the 36 and 420 representations of SU (8) and their complex conjugates residing in the tensors A1 and A2 . n] [m Aijk. the unrestricted T -tensors in D = 5 and 3 dimensions belong . √ n] 1 − 108 2 εijkpqr[lm Tpqr . One then ﬁnds that there is yet another constraint. so that it must consist of the 28 + 36 + 420 representation of SU (8) (and its complex conjugate). Ajkl = − 4 Ti 2i 3 [jkl] .63) 6480 = 28 + 28 + 420 + 420 + 1280 + 1280 + 1512 + 1512 . with the result 56 = 28 + 28 . (5. This proves that the T tensor cannot contain the entire 6480 representation of E7(7) .56). Ti [jk]i = 0. we note that there is a mismatch between (5.63) and (5. In view of (5. Gauge Theory and Strings The representations on the right-hand side can be decomposed under the action of SU (8).66) Although we concentrated on the D = 4 theory. 912 = 36 + 36 + 420 + 420 . Aij = 1 ikj 4 21 Tk . (5. whereas A3 is not an independent tensor and can be expressed in terms of A2 . This implies that the T -tensor is decomposable into A1 and A2 . For instance.

de Wit: Supergravity to the following representations of E6(6) and E8(8) . respectively. which decompose as follows under the action of U Sp(8) and SO(16).B. We close with a few comments regarding the various gauge groups that have been considered. but those can be included by adding Chern-Simons terms. In D = 3 dimensions there is no guidance from Kaluza-Klein compactiﬁcations and one has to rely on the group-theoretical analysis described above. To give a really exhaustive classiﬁcation remains cumbersome. SO(5) and SO(6). namely ﬂat Minkowski space and anti-de Sitter space with a cosmological 25 The D = 3 theory has initially no vector ﬁelds. for the 5-dimensional theory they were also realized in [29] and in [72]. see [74. . 351 = 36 + 315 . Noncompact gauge groups were initiated in [71] for the 4-dimensional theory.1 we presented the ﬁrst steps in the construction of a generic supergravity theory. These terms lose their topological nature when gauging some of the E8(8) isometries [69]. We established the existence of two supersymmetric gravitational backgrounds. 89 = 3: 248 × 248 = 1 + 248 + 3875 + 27 000 + 30 380 . 6 Supersymmetry in anti-de Sitter space In Section 3. (5. Gaugings can also be constructed via a so-called Scherk-Schwarz reduction from higher dimensions [73].68) These representations correspond to the tensors A1 and A2 . respectively. 75]. As we mentioned at the beginning of this section. The S 7 and the S 4 [70] compactiﬁcations of 11dimensional supergravity and the S 5 compactiﬁcation of IIB supergravity. starting with the Einstein-Hilbert Lagrangian for gravity and the Rarita-Schwinger Lagrangians for the gravitino ﬁelds.67) In these cases a successful gauging requires the T -tensor to be restricted to the 351 and the 1 + 3875 representations. 1 + 3875 = 1 + 135 + 1820 + 1920 . gave rise to the gauge groups SO(8). (5. For explorations based on the group-theoretical analysis explained above. In that case there exists a large variety of gauge groups of rather high dimension [69]. for D = 5 A3 is again dependent while for D = 3 there is an independent tensor A3 associated with the 1820 representation of SO(16). respectively25 D D = 5: 27 × 78 = 27 + 351 + 1728 . the ﬁrst gaugings were to some extent motivated by corresponding KaluzaKlein compactiﬁcations.

σ σ Rµνρσ = g 2 (gµρ δν − gνρ δµ ) . . and HR = U Sp(2N ) for D = 6. Note again that the derivation in Section 3. according to Table 9. whose bosonic generators correspond to the Poincar´ group. they have 1 D(D + 1) independent isometries which comprise 2 the Poincar´ group or the group SO(D − 1. These N supercharges transform under a compact R-symmetry group. According to (4. −.14)). . HR = U (N ) for D = 5. its bosonic subalgebra can no longer be restricted to the sum of the anti-de Sitter algebra and the R-symmetry algebra. so that we have coordinates Y A with A = −. In contrast to this.2) Obviously.10) it is possible to describe anti-de Sitter space as a hypersurface in a (D + 1)-dimensional embedding space. In this section we will mainly be dealing with simple anti-de Sitter supersymmetry and we will always assume that 3 < D ≤ 7.1) Because both ﬂat Minkowski space and anti-de Sitter space are maximally symmetric. each transforming as a spinor under the anti-de Sitter group. 1. 2). 1) and thus belongs to the coset spaces that were discussed extensively in Section 4. In that case the bosonic subalgebra coincides with the anti-de Sitter algebra. Denoting the extra coordinate of the embedding space by Y − . . For D > 7 the superalgebra is no longer simple [3]. +) invariant. +. As we discussed in Section 2. we have HR = SO(N ) for D = 4. . 2. there exists an (N -extended) super-Poincar´ algebra associated with ﬂat e Minkowski space of any dimension. +. . . Anti-de Sitter space is isomorphic to SO(D−1. whose generators will ap¯ pear in the {Q. The two cases are clearly related and ﬂat space is obtained in the limit g → 0. 2).1 was incomplete and in general one will need to introduce additional ﬁelds. but one needs extra bosonic generators that transform as high-rank antisymmetric tensors under the Lorentz group (see also [76]). (6. Q} anticommutator. as can for instance be seen from the expression of the Riemann curvature (cf. 2)/SO(D−1.90 Unity from Duality: Gravity.5. (3. this hypersurface is deﬁned by −(Y − )2 − (Y 0 )2 + Y 2 = ηAB Y A Y B = −g −2 . the hypersurface is invariant under linear transformations that leave the metric ηAB = diag (−. Gauge Theory and Strings constant proportional to g 2 . . There exist N -extended versions where one introduces N supercharges. . These transformations constitute the group SO(D − 1. In D = 3 spacetime dimensions the anti-de Sitter group SO(2. 0. 2) is not simple. 7. possibly augmented with the R-symmetry generators e associated with rotations of the supercharges. The 1 D(D + 1) generators 2 . where g was some real coupling constant proportional to the the inverse anti-de Sitter radius. The algebra of e the combined bosonic and fermionic symmetries is called the anti-de Sitter superalgebra. (6. respectively. D − 1. the R-symmetry group is in general not the same as in Minkowski space.

4) Anti-de Sitter space has the topology of S 1 [time] × RD−1 [space] and has closed timelike curves. which has the topology of RD . +.B = a (6. . 2) group in the spinor representation. For later use we record the (simple) anti-de Sitter superalgebra. because the R-symmetry generators will appear on the ¯ right-hand side of the {Q. +) is the D-dimensional . which will be deﬁned later. Qβ } = ¯ [MAB .B. D−1. There exist no Cauchy surfaces in this space. e On spinors. Our gamma matrices satisfy the Cliﬀord property {Γa . In the limit g → 0. Qα ] = 1 − 2 (ΓAB )αβ M AB . . As we alluded to earlier. which in addition to (6. where η ab = diag (−.4) contains the (anti-)commutation relations ¯ {Qα . MAB = 1 ΓAB = 2 1 2 Γab 1 2 Γa for for A. the anti-de Sitter algebra can be realized by the following combination of gamma matrices Γa in D-dimensional Minkowski space. Q → g −1/2 Q. Here one rescales the generators according to o u M−A → g −1 PA . are the generators of SO(D − 1. (6. There are many ways to coordinatize anti-de Sitter space. de Wit: Supergravity denoted by MAB act on the embedding coordinates by MAB = YA ∂ ∂ − YB . this algebra changes its form when considering N supercharges which rotate under R-symmetry. b . Q} anticommutator. Any attempt to determine the outcome of some evolution or wave equation from a spacelike surface requires fresh information coming from a timelike inﬁnity which takes a ﬁnite amount of time to arrive [77. B = a. ∂Y B ∂Y A 91 (6. A = −. . MCD ] = ηBC MAD − ηAC MBD − ηBD MAC + ηAD MBC . keeping the remaining generators MAB corresponding to the Lorentz subalgebra unchanged. . 1 ¯ 2 (Q ΓAB )α . but we will avoid using explicit coordinates. so that one ﬁnds the universal covering space denoted by CadS. . b = 0. Spatial inﬁnity is a timelike surface which cannot be reached by timelike geodesics. . The relation with the Minkowski case proceeds by means of a so-called Wigner-In¨n¨ contraction. (6. It is now easy to evaluate the commutation relations for the MAB [MAB . the generators PA will form a commuting subalgebra and the full algebra contracts to the super-Poincar´ algebra. These curves can be avoided by unwrapping S 1 .5) Here the matrices ΓAB . Γb } = 2 η ab 1. 1. .3) where we lower and raise indices by contracting with ηAB and its inverse η AB . .6) with a. 78].

2) has more Casimir operators when D > 3.8) vanishes and the proof is elementary (see.g.7) The group SO(D − 1. it is important that we establish the relation between the wave operator for ﬁelds that live in anti-de Sitter space. the second term in (6. (6. [δ1 . the reader is advised to consult Section 2. 2). Here we consider only the gravitational sector of the theory. For scalar ﬁelds. The appropriate formulae were given at the end of Section 4. Let us now brieﬂy return to the supersymmetry algebra as it is realized on the vielbein ﬁeld.44) and (4. This result can be proven for any symmetric. homogeneous. 1). 2 (6. for a complete theory there are additional contributions. e.11) The commutator of two supersymmetry transformations yields an inﬁnitesimal generalcoordinate transformation and a tangent-space Lorentz transformation.2 (cf. for example [79]).45)) and from them one can derive. For example. To make contact between the masslike terms in the wave equations and the properties of the irreducible representations of the anti-de Sitter group. we obtain for the vielbein. but nevertheless the above terms remain and 26 Note that when the gravitino is a Majorana spinor. (4.92 Unity from Duality: Gravity. Of central importance is the quadratic Casimir operator of the isometry group SO(D − 1. 2 2 1 2 ¯2 Γa δ1 ψµ − 1 ¯1 Γa δ2 ψµ 2 (6. and the quadratic Casimir operator C2 . which we will discuss in Section 6.9) The ﬁrst term corresponds to a spacetime diﬀeomorphism and the second one to a tangent space (local Lorentz) transformation. . Concerning the R-symmetry group in anti-de Sitter space. space (see.1. Gauge Theory and Strings Lorentz-invariant metric26 .5.8) Lorentz where C2 is the quadratic Casimir operator for the representation of the Lorentz group to which the ﬁelds have been assigned. We remind the reader that ﬁelds in anti-de Sitter space are multi-component functions of the anti-de Sitter coordinates that rotate irreducibly under the action of the Lorentz group SO(D − 1. δ2 ] eµ a = = Dµ ( 1 ¯2 Γa 1 ) + 1 g (¯2 Γab 1 ) eµb . Using the transformation rules (3. deﬁned by C2 = − 1 M AB MAB . [80]). which involves the appropriately covariantized D’Alembertian 2adS . C2 = 2adS g=1 Lorentz + C2 . but these are of higher order in the generators and will not play a role in the following. the quantities Γ AB should satisfy the same Majorana constraint.

i. As for all Killing vectors. so that masslike terms will not necessarily be the same for diﬀerent ﬁelds belonging to the same multiplet.. the diﬀeomorphism with parameter ξ µ = 1 ¯2 Γµ 1 . higher derivatives can be decomposed into the Killing vector and its ﬁrst derivative. We allow for two constants a and b in the supersymmetry transformations. because the momentum operators commute with the supersymmetry charges. (6. 6. a pseudoscalar ﬁeld B and a Majorana spinor ﬁeld ψ. A scalar chiral supermultiplet in 4 spacetime dimensions consists of a scalar ﬁeld A. In the anti-de Sitter background.e.2) .2 for generic coset spaces.B.15) so that the gravitino ﬁeld remains zero under supersymmetry. 2 because Dµ (g ¯2 Γνρ 1 ) = −g 2 gµ[ρ ξν] is antisymmetric in µ and ν. / (6.10).38)). For supermultiplets in anti-de Sitter space this is no longer the case. following (3. In the notation of this section.e. This phenomenon will be now illustrated below in a speciﬁc example. δA δψ = 1 ¯ψ . We have already discussed the interpretation of masslike terms for the gravitino. which is a Killing spinor in the anti-de Sitter space. is an anti-de Sitter Killing vector (i. The last term in (6. the parameters of the supersymmetry transformations are Killing spinors satisfying (3. The Killing vector can be decomposed into the 1 D(D + 1) 2 Killing vectors of the anti-de Sitter space. (x) must satisfy the Killing spinor equation (3. we ﬁnd Dµ (g ¯2 Γνρ 1 ) = −g 2 gµ[ρ ξν] in the case at hand. so that the combination of symmetries on the right-hand side of (6. de Wit: Supergravity 93 (6. Indeed. Further clariﬁcation from an algebraic viewpoint will be given later in Section 6. δB = 1 i¯γ5 ψ . which we parametrize as follows. were the gravitino ﬁeld vanishes.9) should be realized uniformly on all the ﬁelds. Indeed. this equation reads a ∂µ − 1 ω ab γab + 1 g eµ γa 4 2 = 0.3. In anti-de Sitter space the supersymmetry transformations of the ﬁelds are proportional to a spinor parameter (x). 4 4 = ∂ (A + iγ5 B) − (a A + ib γ5 B) .9) should vanish when 1 and 2 are Killing spinors.1 Anti-de Sitter supersymmetry and masslike terms In ﬂat Minkowski space all ﬁelds belonging to a supermultiplet are subject to ﬁeld equations with the same mass. Therefore both the gravitino and the vielbein are left invariant under supersymmetry. namely a scalar chiral supermultiplet in D = 4 spacetime dimensions.9) is a compensating target space transformation of the type we have been discussing extensively in Section 4.15). so that P 2 is a Casimir operator. it satisﬁes (4.1) where we made the anti-de Sitter vierbein and spin connection explicit..

In order to construct the corresponding ﬁeld equations for A and B. To determine the constants a and b and the ﬁeld equations of the chiral multiplet. However. we consider the variation of the fermionic ﬁeld equation. which is proportional to m. Now we note that the right-hand side should constitute a coordinate transformation and a Lorentz transformation. This yields the following second-order diﬀerential equations [2adS + 2g 2 − m(m − g)] A = 0 . Obviously. can be accounted for by introducing an auxiliary ﬁeld to the supermultiplet. If we now deﬁne m = 1 (a + b). which gives rise to the wave operator 2adS + 1 [/ D]− m2 .9) but the correct Lorentz transformation is only reproduced provided that a − b = 2g. which is proportional to g. e. the supersymmetry transformation of ψ equals δψ = D(A + iγ5 B) − m(A + iγ5 B) − g(A − iγ5 B) .g.4) Consequently. D D 2 D. The coeﬃcient of the ﬁrst term in δψ has been chosen such as to ensure that [δ1 . we consider the closure of the supersymmetry algebra on the spinor ﬁeld. Again we have to take into account that derivatives on the supersymmetry parameter are not equal to zero. 1 [δ1 . one evaluates (/ − m)(/ + m)ψ. so that the last term is just 2 the Dirac equation with mass m.3) We point out that derivatives acting on (x) occur in this calculation at an intermediate stage and should not be suppressed in view of (6. the coordinate transformation coincides with (6. they produce terms proportional to g which turn out to cancel in the above commutator. (6. / The commutator yields the Riemann curvature of the anti-de Sitter space. [2adS + 2g 2 − m(m + g)] B = 0 . δ2 ]ψ = ξ µ Dµ ψ + 16 (a− b) ¯2 γ ab 1 γab ψ − 1 ξ ρ γρ [/ + 1 (a+ b)ψ] . δ2 ] yields the correct coordinate transformation ξ µ Dµ on the ﬁelds A and B. can be understood as a compensating S-supersymmetry transformation associated with auxiliary ﬁelds in the supergravity sector (see.1). (6.5) and the fermionic ﬁeld equation equals (/ + m)ψ = 0. possibly up to a ﬁeld equation. The second term in D (6.7) . The third term. In an anti-de Sitter space of arbitrary dimension D this equation then reads [2adS + 1 D(D − 1)g 2 − m2 ]ψ = 0 . / (6.5). b = m−g. we ﬁnd a = m+g. Namely. 4 (6.6) The last equation follows from the Dirac equation. [81]). After some Fierz reordering we obtain the result. Gauge Theory and Strings In this expression the anti-de Sitter vierbein ﬁeld has been used to contract the gamma matrix with the derivative. [2adS + 3g 2 − m2 ] ψ = 0. Dψ 2 2 (6.94 Unity from Duality: Gravity.

6). [81]). whose ﬁeld equation reads (antisymmetrizing over indices ν1 . spinor. (6.11) (6. Combining this result with (6. e. Indeed.7. in spite of the fact that they belong to the same supermultiplet [82]. 2+ 1D−2 R = 2 + 1 D(D − 2) g 2 . Observe that the Dirac operator D is also conformally / invariant and so is the wave equation associated with the Maxwell ﬁeld.12. vector and 8 tensor ﬁelds respectively.6). (6. The quadratic Casimir operator of the Lorentz group takes the values 0. . νn ) (n + 1) Dµ ∂[µ Cν1 ···νn ] − m2 Cν1 ···νn = 0 .6) all have diﬀerent mass terms.g.9) and 27 When m = 0. 6. de Wit: Supergravity 95 which. the role of mass is quite diﬀerent in anti-de Sitter space as compared to ﬂat Minkowski space.9) This can be generalized to an antisymmetric tensor of rank n. In the same way as before. 1 D(D − 1). so that (6.8) reads D2 Aν −[Dµ . for D = 4 agrees with the last equation of (6.8) Dµ (∂µ Aν − ∂ν Aµ ) − m2 Aν = 0 . for D = 4. Consequently. in anti-de Sitter space [2adS + (D − 1)g 2 − m2 ] Aµ = 0 .10) The g 2 term in the ﬁeld equations for the scalar ﬁelds can be understood from the observation that the scalar D’Alembertian (in an arbitrary gravitational background) can be extended to a conformally invariant operator (see. . 6. Using (6. This leads to27 Dµ Aµ = 0. this leads to [2adS + n(D − n)g 2 − m2 ] Cν1 ···νn = 0 . Dν ]Aµ −m2 Aν = 0 or. we do reproduce the g 2 dependence in the ﬁrst two equations (6. (6. .2.B.12) which seems the obvious candidate for a massless wave operator for scalar ﬁelds. spinor. . otherwise one can impose this equation as a gauge condition. 4 4D−1 (6. This will be elucidated later in Section 6. D − 1 and n(D − n) for scalar. vector and tensor ﬁelds. A striking feature of the above result is that the ﬁeld equations (6.8) we can now determine the values for the quadratic Casimir operator for the representations described by scalar. . For future applications we also evaluate the Proca equation for a massive vector ﬁeld.

for m2 = 0. m2 is not the coeﬃcient in the mass term of the Kleinscalar Gordon equation.13) For spinless ﬁelds. m and m do correspond to the mass terms in the Dirac and Proca equations.8) yields the following values for the quadratic Casimir operators. The fact that the value of C2 does not depend on the rank of the tensor ﬁeld is in accord with the fact that. 1 2 whereas for spin-0 we ﬁnd that m ≥ − 4 . (6. while. which implies that the masslike terms for scalar ﬁelds can have a negative coeﬃcient µ2 subject to the inequality. which ensures the stability of an anti-de Sitter background against small ﬂuctua1 tions of the scalar ﬁelds. as that is given by the value for C2 . vector and tensor ﬁelds upon putting m = 0. This can also be seen from the observation that (massless) antisymmetric tensor gauge ﬁelds of rank n = D − 2 are on-shell equivalent to massless scalar ﬁelds. contrary . for spinor 2 and vector ﬁelds. Hence. (6. with m2 as deﬁned in (6. While we concluded above that these tensor ﬁelds lead to C2 = 0.13). We thus derive speciﬁc values for C2 for massless scalar. a rank-n and a rank-(D − n − 2) tensor gauge ﬁeld are equivalent on shell (also in curved space). In the next section we study unitary representations of the anti-de Sitter algebra and this study will conﬁrm some of the results found above. For spin. 4 = − 1 D(D − 1) + m2 . In a number of cases this decoupling is more extreme and one obtains a so-called singleton representation which does not have a smooth Poincar´ limit. = m2 . In the next section we will derive a rather general lower bound on the value of C2 for the lowest-weight representations of the anti-de Sitter algebra (cf. 1 µ2 ≥ − 4 (D − 1)2 .10. Gauge Theory and Strings (6. which uniquely identiﬁes the massless representations and their values of C2 . 6. respectively.96 Unity from Duality: Gravity.13). massless 1 scalar ﬁelds have C2 = − 4 D(D − 2) according to (6. scalar C2 spinor C2 tensor C2 vector C2 = − 1 D(D − 2) + m2 .2 the bound on C2 implies that m2 ≥ 0.14) This result is known as the the Breitenlohner-Freedman bound [82]. In those cases there is no decoupling of a representation e that could be identiﬁed as massless and therefore there remains a certain ambiguity in the deﬁnition of “massless” representations. spinor. 8 = m2 . we see that the interpretation of the mass parameter is not straightforward in the context of anti-de Sitter space. (6. The diﬀerence may not be entirely surprising in view of the fact that the antisymmetric tensor Lagrangian is not conformally invariant for arbitrary values of D.26)). For massless representations of higher spin there is a decoupling of degrees of freedom.

acting on spinors which are ﬁnite-dimensional objects. de Wit: Supergravity 97 to the scalar ﬁeld Lagrangian. The eigenvalues of this generator. are quantized in integer units in order to have single-valued functions. 6. (6. . ± Ma = −iM0a ± M−a . The remaining generators M−a and M0a are combined into D −1 pairs of mutually conjugate operators. 2). and to [80. = . .2 Unitary representations of the anti-de Sitter algebra In this section we discuss unitary representations of the anti-de Sitter algebra. . The compact subgroup of the anti-de Sitter group is SO(2) × SO(D − 1) corresponding to rotations of the compact anti-de Sitter time and spatial rotations. † MAB = −MAB .17) 28 The case of D = 3 is special because SO(2. In order to underline the general features we will stay as much as possible in general spacetime dimensions D > 328 . spinless states which correspond to diﬀerent values for the anti-de Sitter Casimir operator C2 . .B. For these representations the generators are anti-hermitean. We refer to [83] for some of the original work. b = 1. . 85] where part of this work was reviewed. 2 First. b. It is convenient to decompose the 1 D(D+1) generators as follows. where we discuss unitary representations of the anti-de Sitter algebra.16) Obviously the generators of the spatial rotations are the operators Mab with a. have diﬀerent hermiticity properties. R))/Z2 . . 2) ∼ (SL(2. refer to space indices. (6. Note that we have changed notation: here and henceforth in this section the indices a. The anti-de Sitter isometry group. is noncompact. The connection between the local ﬁeld theory description and the anti-de Sitter representations tends to be subtle. 84. associated with motions along the circle. we have established the existence of two diﬀerent ﬁeld representations that describe massless. (6. R) × SL(2. At the end of the next section. The energy operator H will thus be deﬁned as H = −iM−0 . however. we brieﬂy return to the issue of massless representations. the generator M−0 is related to the energy operator when the radius of the anti-de Sitter space is taken to inﬁnity. At any rate. 2) has the generators 1 Γµν 2 and 1 Γµ . which implies that unitary representations will be inﬁnitely dimensional. D − 1. .15) Here we note that the cover group of SO(D − 1. SO(D − 1. These 2 generators. unless one passes to the covering space CadS.

5)) {Qα . A restricted class of representations will be discussed in a sequel. but in passing. C2 = − 1 M AB MAB 2 + − = H 2 − 1 {Ma . so that irreducible representations are characterized by two spin values. the Ma play the role of raising and lowering operators: when ± applied to an eigenstate of H with eigenvalue E. deﬁned by 1 J 2 = − 2 (Mab )2 . (6. 2 . Summarizing: J2 = s(s + 1) for D = 4 .18) = ± Obviously. 2s+ (s+ + 1) + 2s− (s− + 1) for D = 5 .21) The SO(D−1) representations for D > 5 are speciﬁed by giving the eigenvalues of additional (higher-order) SO(D − 1) Casimir operators. (6. For D = 5. a more general discussion of all possible representations requires a more technical set-up and is outside the scope of these lectures. Mb ] = − + [Ma . In this section we restrict ourselves to the bosonic case. Q† } β = H δαβ − 1 iMab (Γa Γb Γ0 )αβ 2 + − + 1 (Ma Γa (1 + iΓ0 ) + Ma Γa (1 − iΓ0 ))αβ . (6. For D = 4 it is expressed in terms of the “spin” s which is an integer for bosons and a half-integer for fermions and the spin-s representation has dimension 2s + 1 and J 2 = s(s + 1). let us already brieﬂy indicate how some of the other (anti-)commutators of the simple anti-de Sitter superalgebra decompose (cf. Ma } − 1 (Mab )2 2 2 + − = H(H − D + 1) + J 2 − Ma Ma . Their dimension is equal to (2s+ + 1)(2s− + 1) and 2 2 2 J 2 = 2(J+ + J− ) with J± = s± (s± + 1). We also give the Casimir operator in this basis. Ma ] = ± ± [Ma . 0. the corresponding rotation group SO(4) is the product of two SU (2) groups.20) In simple cases. −2(H δab + Mab ) . application of Ma yields a state with eigenvalue E ± 1. s− ). (6. Mb ] ± ±Ma . (s+ . Gauge Theory and Strings + − satisfying (Ma )† = Ma . The anti-de Sitter commutation relations then read ± [H.19) where J 2 is the total spin operator: the quadratic Casimir operator of the rotation group SO(D − 1). (6. its value is well known.98 Unity from Duality: Gravity.

Among other things.24) in the expression (6. when there is a state with some eigen− value E0 that is annihilated by all the Ma . C2 = E0 (E0 − D + 1) + J 2 . However. we expect the representation to cover an inﬁnite range of eigenvalues. 2 99 (6. we derive at once the eigenvalue of the quadratic Casimir operator associated with this representation in terms of E0 and J 2 .19) is positive. Continuous representations cover the whole range of eigenvalues E extending from −∞ to ∞.B. J . Qα ] = ∓ 1 (Γa (1 ∓ iΓ0 ) Q)α . 2 this number will depend on whether the eigenvalues E take integer or halfinteger values. this value holds for any state belonging to the corresponding irreducible representation. This will be discussed later in Section 6. as we shall see below. We now turn to irreducible representations of the anti-de Sitter algebra (6.18). (6. all separated by integers. (6. Qα ] = − 1 i(Γ0 Q)α .22) For the anti-de Sitter superalgebra.23) where the subscript indicates that one must choose the minimal value that J 2 + (E − 1 (D − 1))2 takes in the representation. J = 0 .3.18) applied to the ground state |E0 . 2 ± [Ma . For a unitary representation the + − Ma Ma term in (6. For real values of E0 the Casimir . For unitary lowest-weight representations one can derive various lower bounds. We start with the observation that the energy operator can be diagonalized so that we can label the states according to their eigenvalue ± E.24) The unitarity upper bound (6. which implies that the Casimir operator is bounded by C2 ≤ − 1 (D − 1)2 + J 2 + E − 1 (D − 1) 4 2 2 minimal . J and satisﬁes − Ma |E0 . (6.23) on C2 is primarily useful for continuous representations. Substituting the condition (6.25) Since C2 is a Casimir operator. This is therefore not a continuous representation but a so-called lowest-weight representation. Because application of Ma leads to the states with higher and lower eigenvalues E. then only states with eigenvalues E > E0 will appear in the representation. all the bosonic operators can be expressed as bilinears of the supercharges. The ground state of this representation (which itself transforms as an irreducible representation of the rotation group and may thus be degenerate) is denoted by |E0 . de Wit: Supergravity [H . so that in principle one could restrict oneself to fermionic operators only and employ the projections (1±iΓ0)Q as the basic lowering and raising operators.

but those will be similar and need not to be discussed separately. They generally lead to a lower bound for E0 and thus to a corresponding lower bound for C2 . In what follows we restrict ourselves to lowest-weight representations.j s s s s s s s s s s s 0 1 2 3 4 5 6 Fig. These two values correspond to two diﬀerent solutions of the ﬁeld equations subject to diﬀerent boundary conditions at spatial inﬁnity. because those have a natural interpretation in the limit of large anti-de Sitter radius in terms of Poincar´ representations. Each point corresponds to the spherical harmonics of S D−1 : traceless.100 Unity from Duality: Gravity. permissible values for E0 with the same value for C2 .26) As we already discussed at the end of the previous section. Alternatively we can e construct highest-weight representations. . for scalar ﬁelds (J 2 = 0) this is just the Breitenlohner-Freedman bound [82]. operator is bounded by C2 ≥ J 2 − 1 (D − 1)2 . 1. States of the spinless representation in terms of the energy eigenvalues E and the angular momentum j. 4 (6. symmetric tensors Y a1 ···al of rank l = j. Gauge Theory and Strings E 6 E0 + 5 E0 + 4 E0 + 3 E0 + 2 E0 + 1 E0 s . Unless this bound supersedes (6.26) there can exist a degeneracy in the sense that there are two possible. Additional restrictions based on unitarity will be derived shortly.

In the following we consider a number of representations of SO(D − 1) that exist for any dimension. We list the corresponding eigenvalues of the quadratic Casimir operator J 2 . for general D and for the speciﬁc cases of D = 4. They are symmetric l-rank tensor spinors that vanish upon contraction by a gamma matrix and appear when taking products of spherical harmonics with a simple spinor. For D = 5 there are two such numbers. as we explained in the text. To be precise. spanned by l-rank traceless. de Wit: Supergravity 101 Table 17. spanned by mixed tensors Y b.B. Table 17 lists the value of J 2 for these representations. One is the symmetric traceless tensor representation (corresponding to the spherical harmonics on S D−2 ) denoted by l-rank tensors Y A1 ···Al . 2 That implies that every point with spin j in Figure 1 generates two points with spin j ± 1 . representation D D=4 D=5 Y A1 ···Al l(l + D − 3) s=l s± = 1 l 2 Y B. s± . The result of this is shown in Figure 2. For D = 4 these representations are characterized by an integer spin s. for general D. Multiplying such tensors with the vector representation gives rise to two of these representations with rank l ± 1. These states are all shown in Figure 1. all states of energy E = E0 +n are constructed by an n-fold product of creation + operators Ma In this way one obtains states of higher eigenvalues E with higher spin.1 state. 5.A1 ···Al of rank l + 1 (which is not independent for D = 4). For the bosons we consider the spherical harmonics.A1 ···Al (l + D − 4)(l + 1) s=l s± = s∓ + 1 = 1 (l + 1) 2 The full lowest-weight representation can now be constructed by acting with the raising operators on the ground state |E0 .a1 ···al . J . symmetric tensors Y a1 ···al . 2 . denoted by Y α. the states decompose into the state of the highest spin generated by the traceless symmetric product of E − E0 + operators Ma and a number of lower-spin descendants. Armed with this information it is straightforward to ﬁnd the decompositions of the spinor representation of the anti-de Sitter algebra. with the exception of points associated with j = 0. In a similar Table 18 we list the value of J 2 for the irreducible symmetric tensor-spinors. and the representation spanned by mixed tensors Y B. which 2 will simply move to j = 1 . and a “mixed” representation. The simplest case is the one where the vacuum has no spin (J = 0). For given eigenvalue E.a1 ···al . One simply takes the direct product of the spinless representation with a spin. Two generic SO(D − 1) representations.

where 2 we have only one state for each given spherical harmonic. The eigenvalues of the quadratic SO(D−1) Casimir operator J 2 for the symmetric tensor-spinor representation spanned by tensors Y α. the spinless and the spinor representations that we have constructed so far are not necessarily irreducible. Gauge Theory and Strings Table 18. but with an energy equal to E0 +2 or E0 +1. E0 ≥ 1 (D − 3) .27) This leads to 2E0 + 3 − D = 1 2 − Ma |E0 + 2. not surprisingly. (6.a1 ···al for general dimension D and for the speciﬁc cases D = 4. 5. for the scalar and spinor representations. 2 (6. representation J2 D=4 D=5 Y α. In the context of the oscillator method. . 2 has the same value for C2 . spinless 2 . 29 The singleton representation was ﬁrst found by Dirac [86] in 4-dimensional anti-de Sitter space and was known as a “remarkable representation”. but now with E0 = 1 (D+1).28) so that unitary of the representation requires the inequality. spinless 2 .a1 ···al l(l + D − 2) + 1 (D − 1)(D − 2) 8 s=l+ 1 2 s± = s∓ − 1 = 1 l 2 2 However. which we will refer to later. To see this consider the excited state that has the same spin content as the ground state. and compare their value for the Casimir operator with that of the corresponding ground state.102 Unity from Duality: Gravity.30) The excited state then constitutes the ground state for a separate irreducible spinless representation. which. (6. The Casimir eigenvalue for this representation equals C2 (spinless singleton) = − 1 (D + 1)(D − 3) . In this way we ﬁnd for the scalar − E0 (E0 − D + 1) = (E0 + 2)(E0 − D + 3) + Ma |E0 + 2. respectively. 4 (6. singletons in anti-de Sitter spaces of dimension D = 4 are called “doubletons” [87].29) For E0 = 1 (D − 3) we have the so-called singleton representation29. In these lectures we will only use the name singleton to denote these remarkable representations.

which simply move to j = 1. both singleton representations have the same eigenvalue of the Casimir operator. The existence of the singletons was ﬁrst noted by Dirac [86]. (6. An exception are the spinless points.32) so that one obtains the unitarity bound E0 ≥ 1 (D − 2) . de Wit: Supergravity For the spinor representation one ﬁnds a similar result. spinor 2 103 . e.33) For E0 = 1 (D − 2) we have the spinor singleton representation. the singleton spectrum is diﬀerent as the states have a single energy eigenvalue for any given value of the spin. as the resulting multiplet contains states of spin lower than that of the ground state. These representations are characterized by the fact that they do not exist in the Poincar´ limit. continuous tower of modes. In principle. however. note that Poincar´ representations correspond e e to plane waves which are decomposable into an inﬁnite number of spherical harmonics. associated with two spherical harmonics. which is just what one obtains in the limit of vanishing energy increments for the generic spectrum shown in..g.31) As the value for J 2 are the same for the ground state and the excited state one readily derives − 2E0 − D + 3 = Ma |E0 + 1. E0 (E0 − D + 1) + J 2 = (E0 + 1)(E0 − D + 2) + J 2 − + Ma |E0 + 1. as is obvious in Figure 3. spinor 2 . (6. wave functions that constitute singleton representations do not depend on the radius of the anti-de Sitter spacetime and can be regarded as living on the boundary. 2 (6. which again 2 consists of just one state for every value of the total spin. To obtain the spin-1 representation one can take the direct product of the spinless multiplet with a spin-1 state. (6. That means that. for given spin. The result of taking the product is depicted in Figure 4. irrespective of the size of the spatial momentum (related to the energy eigenvalue). one is dealing with an inﬁnite.B. Figure 1. Consequently. This procedure can be . In contradistinction.34) Note that in D = 4. For the spinor representation the value of the Casimir operator equals 1 C2 (spinor singleton) = − 8 (D + 1)(D − 2) . To see this. Now the situation is more complicated. each point in Figure 1 now generates three points. associated with rank-j ±1 tensors as well as mixed tensors of rank j + 1 (so that l = j).

35) . extended directly to ground states that transform as a spherical harmonic Y a1 ···al . Gauge Theory and Strings E 6 E0 + 5 E0 + 4 E0 + 3 E0 + 2 E0 + 1 E0 b b b s s s s s s . In that case we ﬁnd C2 = = so that E0 − l − D + 3 = 1 2 − Ma |E0 + 1.36) .25). (6. l − 1 2 E0 (E0 − D + 1) + l(l + D − 3) . 2 (6. States of the spinor representation in terms of the energy eigenvalues E and the angular momentum.j b b b s s s s s b b s s s s b b s s s b s s b s 0 1 2 3 4 5 6 Fig. 2. the half-integer values for j = l + 1 denote that we 2 are dealing with a symmetric tensor-spinor of rank l. The small circles denote the original spinless multiplet from which the spinor multiplet has been constructed by a direct product with a spinor.104 Unity from Duality: Gravity. Hence we consider the states with E = E0 + 1 and j = l − 1. We compare the value of the Casimir operator for the ﬁrst excited states with minimal spin to the value for the ground state speciﬁed in (6. Along the same lines as before. assuming that the ground state has l ≥ 1. l − 1 − (E0 + 1)(E0 − D + 2) + (l − 1)(l + D − 4) − Ma |E0 + 1. we investigate whether this representation can become reducible for special values of the ground state energy.

37) When E0 = l + D − 3.1 singleton ground state has an energy which is one 2 2 half unit higher. which denotes the spin-0 ground state energy.j s c s c s c s c s c 105 0 1 2 3 4 5 6 Fig. which decouples from the original multiplet together with its corresponding excited states. Therefore we establish the unitarity bound E0 ≥ l + D − 3 (l ≥ 1). Hence massless representations with spin l ≥ 1 are characterized by E0 = l + D − 3 . The spin-0 and spin. The solid dots indicate 2 the states of the spin-0 singleton.38) . however. This can be interpreted as the result of a gauge symmetry. (6. as is explained in the text. 2 It is obvious that singletons contain much less degrees of freedom than a generic local ﬁeld. l − 1 is itself the ground state of an irreducible multiplet. is equal to E0 = 1 (D − 3). The spin. Because these representations have a smooth Poincar´ limit they are not singletons and can therefore be regarded e as massless representations. the circles the states of the spin.1 singleton. de Wit: Supergravity E 6 E0 + 5 E0 + 4 E0 + 3 E0 + 2 E0 + 1 E0 s . the state |E0 + 1. The value of E0 . 3.B.1 singleton representations. (l ≥ 1) (6.

as we already discussed. The double-occupancy points exhibit the structure of a spin-0 multiplet with ground state energy E0 + 1. but this requires further knowledge of the various representations of the rotation group. For a spin-s ground state in 4 spacetime dimensions we . as is explained in the text. 5 dimensions this information is readily available.j s f s s s f s s f s s f f s s f s s s 0 1 2 3 4 5 6 Fig. For certain other cases. This multiplet becomes reducible and can be dropped when E0 = D − 2. in D = 4. States of the spin-1 representation in terms of the energy eigenvalues E and the angular momentum j. We return to this issue at the end of this section. Gauge Theory and Strings E 6 E0 + 5 E0 + 4 E0 + 3 E0 + 2 E0 + 1 E0 s s . This is outside the scope of these lectures.106 Unity from Duality: Gravity. However. The above arguments can be easily extended to other grounds states. at least for general dimension. The remaining points then constitute a massless spin-1 multiplet. the identiﬁcation of massless representation is somewhat ambiguous.39) We recall that this result is only derived for l ≥ 1. (l ≥ 1) (6. indicated by the circle superimposed on the dots and states transforming as mixed tensors (with l = j) denoted by a . shown in Figure 5. 4. For these particular values the quadratic Casimir operator acquires a minimal value equal to C2 (massless) = 2(l − 1)(l + D − 3) . Observe that there are now points with double occupancy.

Now E0 is no longer arbitrary but it is ﬁxed to E0 = D − 2. (s± ≥ 1 ) 2 (6. In 2 that case we ﬁnd that the ground state energy satisﬁes the unitarity bound. even to s = 0.37). This is a 2 special property for D = 4 dimensions.j s s s s s s s s s s s 107 0 1 2 3 4 5 6 Fig. (6. 2 E0 ≥ s + 1 . We brieﬂy summarize some results.41) C2 = 2(s2 − 1).B. because here the rotation group factorizes into two SU (2) groups. 1 conformal ﬁelds. E0 ≥ s+ + s− + 2 . de Wit: Supergravity E 6 E0 + 5 E0 + 4 E0 + 3 E0 + 2 E0 + 1 E0 s . First let us assume the the groundstate has spin (s+ .39) becomes (6. The case of D = 5 requires extra attention. The equation corresponding to (6. for which we cannot use this derivation. 5. When the bound is saturated we obtain a massless representation. It turns out that this result applies to all spin-s representations. immediately derive the unitarity bound (for s > 1 ). s− ) with s± ≥ 1 .42) .40) by following the same procedure as leading to (6. States of the massless s = 1 representation in terms of the energy eigenvalues E and the angular momentum j.

(6. We close this section with a number of comments regarding “massless” representations and their ﬁeld-theoretic description. Our discussion of the shortening of anti-de Sitter supermultiplets in Section 6. This is generically the case for arbitrary dimensions D = 4 and is thus related to the fact that the rotation group is of higher rank. The same phenomenon of multiplet shortening is well known and relevant in conformal ﬁeld theory in 1 + 1 dimensions. The corresponding value for 2 2 the Casimir operator is equal to C2 = (s+ + s− )2 + 2s+ (s+ + 1) + 2s− (s− + 1) − 4 .108 Unity from Duality: Gravity. Here we are not striving for completeness. with a large variety of spin values. As we demonstrated . Gauge Theory and Strings This bound is saturated for massless states. 2 What remains to be considered are the ground states with spin (0. Mb ] commutator acquires zero or negative eigenvalues for certain values of E0 and J 2 . for a recent elementary introduction to this method we refer to [80].45) The singleton representations for s = 0. s).3 will support this point of view. From the above it is clear that we are dealing with the phenomenon of multiplet shortening for speciﬁc values of the energy and spin of the representation. This phenomenon can be understood from the fact that − + the [Ma . When viewed in this way.44) E0 ≥ 1 + s . Here we ﬁnd (6. There is an extensive literature on this. When the bound is saturated we have again a singleton representation. s− − 1 ) decouples. The purpose of this section was to elucidate the various principles that underlie the anti-de Sitter representations and their relation with the ﬁeld theory description. which is applicable in any number of spacetime dimensions and which can also be used for supersymmetric extensions of the anti-de Sitter algebra. Note 2 that for D = 5 there are thus inﬁnitely many singleton representations. unlike in 4 dimensions. (6. There are in fact alternative and often more systematic techniques for constructing the lowest-weight representations. The corresponding values for the Casimir operator are C2 (singleton) = 3(s2 − 1) .43) For s± = 1 l these values are in agreement with earlier result. is known as the oscillator method [88]. just as in the earlier discussions on BPS multiplets in previous sections. the shortening of the representation is qualitatively similar to the shortening of BPS multiplets based on the anticommutator of the supercharges. A powerful method to construct the unitary irreducible representations of the anti-de Sitter algebra. 1 were already found earlier. for which the E = E0 + 1 states with spin (s+ − 1 .

This anti-de Sitter representation is described by a (complex) tensor ﬁeld. In Table 19 we have collected a number of examples of spin-0 and spin. whose ﬁeld equation takes the form. one with spin (s1 . de Wit: Supergravity 109 above. which are obviously massless in view of (6. decouple into diﬀerent irreducible representations.38).44) and (6. On shell the equation (6. according to which every representation should be regarded as massless that appears in the tensor product of two singleton representations. In that case one has representations that contain fewer degrees of freedom. it also contains the l = 0 representation with E0 = D − 3. when they e do not. for a speciﬁc value of E0 .38) does not apply so that the interpretation as a massless representation is less obvious. certain representations can. the spectrum of states is qualitatively independent of the value for E0 . Therefore the concept of mass remains ambiguous. where n is an arbitrary positive integer. which is on-shell equivalent to a scalar. they are called massless. (6. (6.46) where m = ±g. e−1 εµνρσλ Dρ Bσλ + 2im B µν = 0 .1. One of them is tied to the fact that the corresponding ﬁeld equation is conformally invariant.B. as we discussed at the end of Section 6. This phenomenon takes place when some unitarity bound is saturated. Hence these representations should be regarded as massless. s− ) and E0 = 2 + s+ + s− .1. The tensor product of the singleton representations with spin (0. From this equation one can show that Bµν satisﬁes (6. they are called singletons. it is easy to verify that the product of two spinless singletons leads to an inﬁnite series of higher spin representations that are all massless according to (6. 0) and energy E0 = 2 + s + n. as long as E0 does not saturate the unitarity bound and a singleton representation decouples. s− ) and (s+ . to which (6. In maximal gauged supergravity in 5 dimensions with gauge group SO(6). Another one follows from the fact that we are dealing with a gauge ﬁeld.1 representations with the criteria 2 according to which they can be regarded as massless. one of these representations appears as part of the “massless” supergravity multiplet. However. This interpretation can be tested as follows. We also invoke a criterion introduced by G¨naydin (see [89] and the u discussion in [90]). It is interesting to consider this criterion for masslessness in D = 5 dimensions.10) so that C2 = 1 (cf.42). For instance. Taking the product of two singleton representations. for example. We have already discussed this in Section 6. Here the example is an antisymmetric rank-(D − 2) gauge ﬁeld. For the case of spin-0 or spin. 0) leads to ground states with spin (s.46) projects out the .1 repre2 sentations. When these “shortened” representations have a smooth Poincar´ limit. where we emphasized that the absence of mass terms in the ﬁeld equations is also not a relevant criterion for masslessness. 0) and another one with spin (s2 . 0) leads to a ground state with spin (s+ .13)).

the indices α. .e.1. Some unitary anti-de Sitter representations of spin 0 and 1 which are 2 massless according to various criteria. We substitute these deﬁnitions into (6. qα ] = 1 − 2 qα . These anti-de Sitter supermultiplets were ﬁrst discussed in [82. a = 1. but at the end we turn to more general N . In most of the section we discuss simple supersymmetry (i. 91].22) and obtain [H . 85. . This allows us to make contact with the material discussed in Section 6. For deﬁniteness we discuss the case of 4 spacetime dimensions with a Majorana supercharge Q. and the corresponding values for E0 and C2 .22). spin 0 0 0 0 1 2 1 2 D−3 D−1 1 1 2D − 2 5 D− 2 E0 1 2D − 1 2D 1 − 2) − 2) −2(D − 3) 0 1 − 8 D(D − 1) 1 (D2 − 15D + 32) 8 C2 1 − 4 D(D 1 − 4 D(D type conformal scalar conformal scalar ∈ singleton × singleton (D − 2)-rank gauge ﬁeld conformal spinor ∈ singleton × singleton degrees of freedom corresponding to spin (1. Γa = 0 iσ a −iσ a 0 . 3 . 6. 0) or (0. N = 1). We start from the (anti-)commutation relations already presented in (6. .48) † where q α ≡ qα . are two-component spinor indices and a the σ are the Pauli spin matrices. We choose conventions where the 4 × 4 gamma matrices are given by Γ0 = −i1 0 0 i1 . .18) and (6. From this one derives that E0 = 3 (a second solution with E0 = 1 violates the unitarity bound (6.47) and write the Majorana spinor Q in the form Q= qα εαβ q β .44)). Gauge Theory and Strings Table 19.3 The superalgebras OSp(N |4) In this section we return to the anti-de Sitter superalgebras. (6.110 Unity from Duality: Gravity. (6. 1). 84.. depending on the sign of m. β. 2.

0 with ground-state energies E0 and E0 + 1.1 singleton. s which have both spin s and which have energies E0 and E0 + 1. As we will 2 see below. the pseudoscalar ﬁeld B and the spinor ﬁeld ψ of the scalar chiral supermultiplet. For E0 = 1 the 2 2 multiplet degenerates and decomposes into a super-singleton.B. q α |E0 . respectively. (H 1 + J · σ)α β . (6. that we studied in Section 6. We obtain the ground states 2 |E0 . s . 0 . consisting of a spin-0 and a spin. s . characterized by the fact that it is annihilated by the lowering operators qα . so that E0 > 1 . which decompose into the ground states 2 with spin j = s − 1 and j = s + 1 .2 pair of ground states q α |E0 . However. qα |E0 . These ground states are |E0 . We see that the operators qα and q α are lowering and raising operators. It includes the spinless states |E0 .49) where we have deﬁned the (hermitean) angular momentum operators Ja = 1 − 2 i εabc M bc . q β } = {qα . (6. these states can be described by the scalar ﬁeld A. respectively. 2 For s ≥ 1 we are in the generic situation. we study unitary irreducible representations of the OSp(1|4) superalgebra. s and q [α q β] |E0 . There 1 is one spin. s = 0 . Obviously. q α ] = {qα . There are two more (degenerate) ground states. de Wit: Supergravity [H . and another spinless supermultiplet with 2 E0 = 3 . qβ } = {q α . respectively. + Products of the Ma simply lead to the higher-energy states in the antide Sitter representations of given spin that we considered in Section 6. s . 0 and q [α q β] |E0 . The s = 0 case is special since it contains less anti-de Sitter representations than the generic case.1. with energy E0 + 1 . + Ma (σ 2 σ a )αβ . both with energy E0 + 1 .49). We assume that there exists a lowestweight state |E0 .2. Let us brieﬂy discuss these representations for diﬀerent s. we only have to study the antisymmetrized products of the q α . They change the energy of a state by half a unit. q α |E0 . − Ma (σ a σ 2 )αβ .50) In principle we can now choose a ground state and build the whole representation upon it by applying products of raising operators q α . Observe that the relative sign between H and J · σ in the third (anti)commutator is not arbitrary but ﬁxed by the closure of the algebra. the bounds for E0 that we derived in the previous sections should be respected. because the + symmetric ones just yield products of the operators Ma by virtue of (6. 2 2 . s . q β } = 1 α 2q 111 . By restricting ourselves to the antisymmetrized products of the q α we thus restrict ourselves to the ground states upon which the full anti-de Sitter representations are build. s and q [α q β] |E0 . In analogy to the bosonic case.

the expression on the right-hand side of (6.2 . s · (6. Gauge Theory and Strings As in the purely bosonic case of Section 6.112 Unity from Duality: Gravity. whose applicability is extended to 1 spin. Obviously. To determine its eigenvalues. the right-hand side is manifestly hermitean as well. the pseudo-scalar ﬁeld B and .52) (J · σ)2 + (J · σ) = s(s + 1)1 . Armed with these results we return to the masslike terms of Section 6. s E0 . There is an extended literature on this. where the latter is again regarded as a (4s + 2)-by-(4s + 2) matrix. This shows that J · σ has only two (degenerate) eigenvalues (assuming s = 0. The corresponding multiplets are then again called massless.51) has zero eigenvalues so that there are zero-norm states in the multiplet which decouple.40). there can be situations in which states decouple so that we are dealing with multiplet shortening associated with gauge invariance in the corresponding ﬁeld theory. As we already mentioned one can also use the oscillator method to construct the irreducible representations. The ground state with s = 1 and E0 = 3 leads to the massless 2 2 vector supermultiplet in 4 spacetime dimensions.53) If the bound is saturated. s|(E0 1 + J · σ)α β |E0 . Because we assume that the representation is unitary. s| qα q β |E0 . for instance [87. s E0 . but in order to be positive deﬁnite the eigenvalue E0 of H must be big enough to compensate for possible negative eigenvalues of J · σ. s = = E0 . E0 must satisfy the inequality E0 ≥ s + 1 . as one can verify by inserting a complete set of intermediate states between the operators qα and q β in the matrix element on the left-hand side. In that case we must be dealing with a massless multiplet. for s ≥ 1 · 2 (6. This is the bound (6. The reader may consult.51) to be positive deﬁnite. we consider the matrix elements of the operator qα q β between the (2s + 1)-degenerate ground states |E0 . so that it is (4s + 2)-by-(4s + 2). s|{qα . 92].2. q β }|E0 . if E0 = s + 1. i. this matrix must be positive deﬁnite. Hence in order for (6. as follows by straightforward calculation. Namely. we note that J · σ satisﬁes the following identity (6.1 for the chiral supermultiplet. namely s and −(s + 1).e. The ground-state energy for anti-de Sitter multiplets corresponding to the scalar ﬁeld A. so that the above equation is not trivially satisﬁed). We now discuss this in a general way analogous to the way in which one discusses BPS multiplets in ﬂat space.51) This expression constitutes an hermitean matrix in both the quantum numbers of the degenerate groundstate and in the indices α and β.

Observe that we could have made a slightly diﬀerent identiﬁcation here. = (E0 + 1)(E0 − 2) . an alternative ﬁeld representation for describing this supermultiplet.41). are equal to E0 . For toroidal compactiﬁcations the massive Kaluza-Klein states belong to BPS multiplets whose central charges are the momenta associated with the compactiﬁed dimensions. The spinless representation with 1 E0 = 2 can be described by a scalar ﬁeld. de Wit: Supergravity 113 the Majorana spinor ﬁeld ψ. = E0 (E0 − 1) . Comparison with the ﬁeld equations (6. = (E0 − 1)2 . For nontrivial compactiﬁcations that correspond to supersymmetric . but also the inﬁnite tower of massive Kaluza-Klein states should comprise shortened supermultiplets. respectively.55) The terms on the right-hand side are not present for massless ﬁelds and we should therefore identify them somehow with the common mass parameter m of the supermultiplet. E0 +1 and E0 + 1 . and the second spinless representation with E0 = 3 by a rank-2 tensor ﬁeld. = (E0 + 1 )(E0 − 5 ) + 3 · 2 2 4 (6. the spin. we know that the quadratic Casimir operator is given by (6.B.6) shows (for g = 1) that we obtain the correct contributions provided we make the identiﬁcation E0 = m + 1.54) For massless anti-de Sitter multiplets. The Lagrangian for the tensor supermultiplet is not conformally invariant in 4 dimensions. From Kaluza-Klein compactiﬁcations of supergravity one can deduce that there should also exist shortened massive supermultiplets. it should be regarded as massless. (6. We have not constructed this supermultiplet in anti-de Sitter space. When compactifying to an anti-de Sitter ground state with supersymmetry the massless supermultiplets remain shortened by the same mechanism. i. so we present the value for C2 − 2(s2 − 1) for the three multiplets. When E0 = 2 there exists. C2 (A) + 2 C2 (B) + 2 C2 (ψ) + 3 2 = (E0 − 1)(E0 − 2) . 2 The Casimir operator therefore takes the values C2 (A) C2 (B) C2 (ψ) = E0 (E0 − 3) . the above result remains the same under the interchange of A and B combined with a change of sign in m (the latter is accompanied by a chiral redeﬁnition of ψ).e. and this could account for the unusual ground state energy for the spinor representation. The reason is that the underlying supergravity multiplet in higher dimensions is shortened because it is massless. in principle.2 representation with 5 E0 = 2 by a spinor ﬁeld. in view of the fact that it contains a tensor gauge ﬁeld.

4). .56). . namely the Hamiltonian. q βj } = δi δα H + δi J · σα β + δα T · Σi j .56) where T are the hermitean 1 N (N − 1) generators of SO(N ) which act 2 on the supercharges in the fundamental representation. the massive Kaluza-Klein states must be shortened according to the mechanism exhibited in this section. s. The supercharges now carry an extra SO(N ) index and are denoted by qαi and q αi .51)). As it turns out. We have already determined the possible eigenvalues of J ·σ which are equal to s or −(s+1). Gauge Theory and Strings anti-de Sitter ground states. (6. which decomposes into three terms. with q αi = (qαi )† with i = 1. the analysis is rather similar. .57) The construction of lowest-weight representations proceeds in the same way as before. Denoting the SO(N ) representation by t (which can be expressed in terms of the eigenvalues of the Casimir operators or Dynkin labels). One starts with a ground state of energy E0 which has a certain spin and transforms according to a representation of SO(N ) which is annihilated by the qαi . To exhibit this we generalize the previous analysis to the N -extended superalgebra. q βj } = = − Ma (σ a σ 2 )αβ δij . (6.59) q [α1 i1 · · · q αn in ] |E0 . As before the unitarity limits follow from the separate limits on the anti-de Sitter representations and from the right-hand side of the anticommutator (6. denoted by OSp(N. + Ma (σ 2 σ a )αβ δ ij . .58) qαi |E0 . s. qβj } {q αi . t · Here the antisymmetrization applies to the combined (αi) labels. Hence the generic N -extended representations decompose into ordinary anti-de Sitter representation whose ground states have energy E0 + 1 n and which 2 can be written as (6. with exception of the combination that leads to the opera+ tors Ma which will generate the full anti-de Sitter representations. The most relevant change to the (anti)commutators (6. t = 0 . generated by the hermitean matrices Σ. Excited states are generated by application of the q αi . (6. the rotation generators and the R-symmetry generators. The last two anticommutators are given by {qαi . Therefore it follows that there must exist shortened massive representations of the extended supersymmetric anti-de Sitter algebra.49) is in the third one. In a similar way one can determine the eigenvalues for T · Σ by noting that it satisﬁes a polynomial . The singleton multiplets decouple from the Kaluza-Klein spectrum. N . which are mutually anticommuting. we have (6.114 Unity from Duality: Gravity. taken in the space of ground state conﬁgurations (cf. which reads j β j β {qαi .

52) with coeﬃcients determined by the Casimir operators. Two prominent examples were noted (see. we derive the unitarity bound.56) in the space of degenerate ground state conﬁgurations has the following six eigenvalues: E0 + s − t. 1 or t + 1. namely the singleton representations in D = 5 and 7 anti-de Sitter space. Combining these results we ﬁnd that the right-hand side of (6. de Wit: Supergravity 115 matrix equation such as (6. 0) tensor multiplet in D = 6 dimensions. 92]). The ground states with s = 0 and E0 = t deﬁne massive shortened representations of the type that appear in Kaluza-Klein compactiﬁcations [84]. Here the issue is whether the singletons play only a group-theoretic role or whether they have also a more dynamical signiﬁcance.g. unless t = 0. which correspond to N = 4. 2 2 for s ≥ 1 . precisely because they are related to boundary degrees of freedom. Group-theoretically they are of interest because their products lead to the massless and massive representations that one encountered in the Kaluza-Klein context. E0 − s − t − 1. [87. the combined result takes the following form. Long before the formulation of the AdS/CFT correspondence it was realized that supersingleton representations could be described by conformal supersymmetric ﬁeld theories on a boundary. All these eigenvalues must be positive. E0 ≥ 1 + s + t E0 ≥ 1 + s E0 ≥ t for s ≥ 1 . t = 0 . Incorporating also the possibility that s or t vanishes. There is an extended literature to which we refer the reader for applications and further details. E0 − s or E0 + t + 1. E0 + s + 1. For instance.B. 2 for s = 0 . This equation shows that the eigenvalues of T · Σ take the values −t.60) where T 2 = t(t + 1) 1. E0 + s + t + 1. for N = 3 we derive. E0 ≥ 1 + s+ t. either by changing the number of supersymmetries or by changing the spacetime dimension. so that in the generic case where s and t are nonvanishing. Obviously these techniques can be extended to other cases. Another theme addresses the connection between singletons and higher-spin theories. D = 4 supersymmetric gauge theories and the chiral (2. certain anti-de Sitter representations must decouple. Before closing the section we want to return to the remarkable singleton representations. In [94] the reader may also ﬁnd a . We refrain from speculating about these questions and just refer to some recent papers [93–95]. t ≥ 1 · 2 (6. In the Poincar´ limit these representations become e all massless. t ≥ 1 . −(T · Σ)3 + 2(T · Σ)2 + (t2 + t − 1)(T · Σ) = t(t + 1) 1 . The singletons decouple from the KaluzaKlein spectrum. e. (6.61) Whenever one of these bounds is saturated.

Gauge Theory and Strings summary of some useful results about singletons as well as an extensive list of references. For conformal Killing vectors.2) Solutions to this equation are called conformal Killing vectors. which are based on the same anti-de Sitter algebra. In the next section we will move to a discussion of superconformal symmetries. where ξ = 0 one then proves that (D − 2)Dµ Dν ξ and Dµ Dµ ξ are determined in terms of lower derivatives. A space that has the maximal number of isometries is called maximally symmetric. the Killing vector is thus fully determined by its value at that point and the values of its ﬁrst derivatives (which are antisymmetric in view of (7. We draw the attention of the reader to the fact that in Section 7. For D = 2 there can be inﬁnitely many conformal Killing vectors.1)). For general dimension D > 2 there are at most 1 (D + 1)(D + 2) conformal 2 Killing vectors. D will always denote the spacetime dimension of the superconformal theory. This suﬃces to derive the maximal number of conformal Killing vectors quoted above for D > 2. Dµ ξν + Dν ξµ = 2 gµν Dρ ξ ρ . Note that the above equation is the traceless part of (7. The conformal Killing vectors that are not isometries are thus characterized by a nonvanishing ξ = Dµ ξ µ .116 Unity from Duality: Gravity. Altogether there are thus 1 D(D + 1) initial conditions to be ﬁxed and 2 they parametrize the number of independent Killing vectors. satisfying Dµ ξν + Dν ξµ = 0 . First one shows that Dµ Dν ξρ = Rνρµ σ ξσ − 1 gµν Dρ ξ − gρµ Dν ξ − gρν Dµ ξ . but in spacetime dimension D + 1.1). 7 Superconformal symmetry Invariances of the metric are known as isometries. D (7.3) For Killing vectors (which satisfy ξ = 0) this result implies that the second derivatives of Killing vectors are determined by the vector and its ﬁrst derivatives. The corresponding superalgebra is then the anti-de Sitter superalgebra. Continuous isometries are generated by so-called Killing vectors.1) is. . These result can be derived as follows.1) The maximal number of linearly independent Killing vectors is equal to 1 2 D(D + 1). (7. D (7. A weaker condition than (7. When expanding about a certain point on the manifold. Both ordinary and conformal Killing vectors generate a group.

can be clariﬁed by extending the D-dimensional spacetime parametrized by coordinates xµ with an extra (noncompact) coordinate y. {yyy } = −y −1 .4) ΛD xµ scale transformations (D) (2 xµ xν − x2 η µν )ΛKν conformal boosts (K).2).7) so that the right-hand side of (7. {µνy } = −y −1 δµ . δx = f+ (x + t) + f− (x − t) . y2 (7. Obviously K ξ = D(ΛD +xµ Λµ ). µ Here ξP . SO(2. which decompose as follows. This is the same group as the anti-de Sitter group in D+1 dimensions. for D ≥ 3 the anti-de Sitter and the conformal group coincide for dimensions D + 1 and D. 2). . R))/Z2 . (7.5) However. The corresponding diﬀeomorphisms can be characterized in terms of two independent functions f± and take the form. The above conformal Killing vectors generate the group K SO(D.B.6) The fact that. de Wit: Supergravity 117 In what follows we choose a Minkowski signature for the D-dimensional space. = This follows directly by writing out the inﬁnitesimal transformations (7. ds2 = gµν dxµ dxν + dy 2 . corresponding to two copies of the Virasoro algebra. assuming the line element. The corresponding expressions for the curvature components are Rµνρ σ = D σ Rµνρ σ + 2y −2 gρ[µ δν] . respectively. It is straightforward to derive the nonvanishing Christoﬀel symbols for this extended space ν (7. where {µρν } remains the same for both spaces and all other components vanish. 2) ∼ (SL(2. The case of D = 2 is special because in that case the above transformations generate a semisimple group. δt = f+ (x + t) − f− (x − t) . K K 2 (7. x t δ(x ± t) = (ξP ± ξP ) + (ΛD ∓ xt )(x ± t) + 1 (Λx ∓ Λt )(x ± t)2 . µν = − νµ . which is responsible for the lack of invariance of the line element of the original D-dimensional space. for D = 2 there are inﬁnitely many conformal Killing vectors. R)×SL(2. can be cancelled by a scale transformation of extra coordinate y.8) {µyν } = y −1 gµν . µ spacetime translations (P ) ξP µ xν Lorentz transformations (M ) ν ξµ = (7. ΛD and Λµ are constant parameters.4) for the linear combinations x ± t. a restriction that is mainly relevant when considering supersymmetry. Flat Minkowski spacetime has the maximal number of conformal Killing vectors.

which associated with the SO(1.4). we have C2 = E0 (E0 − D) (observe that we must replace D by D + 1 in (6. Straightforward calculation yields. Nevertheless. [28. In terms of the ground state energy E0 of the anti-de Sitter representation.11) Near the boundary where y is small. 2D+1 = y 2 2D + (y ∂y )2 − D y ∂y . This identiﬁcation of the generators is discussed in more detail in the next section.10) The condition Dµ Dν ξ = 0 holds for the conformal Killing vectors (7. We may compare this to solutions of the Klein-Gordon equation in the anti-de Sitter space. g = 1 in (3. (7.25)). Also the relation between the D’Alembertians of the extended and of the original D-dimensional spacetime is relevant in this context.118 Unity from Duality: Gravity. 2D ξ y (x. σ y −2 δν . 97–101]. for which we know that the D’Alembertian equals the quadratic Casimir operator C2 . and also the lectures presented at this school). . This setting is relevant for the adS/CFT correspondence and there exists an extensive literature on this (see. This was the reason why we adopted a positive signature in the line element (7. and only this group can be extended to isometries of the (D+1)-dimensional space. y) = y ξ(x) .e. Gauge Theory and Strings Rµyρ y Ryνy σ = = y −2 gµρ . ξ µ (x. For D = 2 these vectors generate a ﬁnite subgroup of the inﬁnite-dimensional conformal group. which shows that we have the identiﬁcation ∆ = E0 or ∆ = D − E0 . y) = ξ µ (x) − y2 µ ∂ ξ(x) . near the boundary [77] of the space (y ≈ 0). e.7) for the coordinate y. This identiﬁcation is somewhat remarkable in view of the fact that E0 is the energy eigenvalue associated with the SO(2) generator of the anti-de Sitter algebra and not with the noncompact scale transformation of y.9) With these results one easily veriﬁes that the curvature tensor of the (D+1)dimensional extension of a ﬂat D-dimensional Minkowski space is that of an anti-de Sitter spacetime with unit anti-de Sitter radius (i.14)). D (7. Subsequently one can show that the D-dimensional conformal Killing vectors satisfying Dµ Dν ξ = 0 can be extended to Killing vectors of the (D + 1)-dimensional space. 1) eigenvalue. (7. the conformal Killing vectors generate asymptotic symmetries. the ﬁelds can be approximated by y ∆ φ(x). Such a phenomenon was ﬁrst analyzed in [96].g.

−→ Mab . Pc ] [Mab . 2(Mab + ηab D) . 2) was given in (6. . Kc ] [Pa . According to the discussion in Section 2. D. Pb ] = = = = Ka . 4 η[a[c Md]b] . . which we extend with two extra index values. 0. . (7.14) Q→Q= . (7. Qα and we deﬁne an extended set of gamma matrices ΓA by. D − 1. For the bosonic generators which generate the group SO(D. The algebra associated with SO(D. . (7.5)) we make a diﬀerent decomposition than the one that led to (6. D − 1.1 The superconformal algebra 119 From the relation between the conformal and the anti-de Sitter algebra one can determine the superextension of the conformal algebra generated by the above conformal Killing vectors. Sα ¯ ¯ ¯ ¯ (7. 0.22). . The decomposition of the conjugate spinor is somewhat subtle.18) and (6. −→ −→ 1 2 (Pa 1 2 (Pa − Ka ) . and D generators Ka of the conformal boosts. Mab ] = = = = −Pa . + Ka ). .4) and corresponds to the following commutation relations [D. −2 ηc[a Pb] . Q → Q = (Qα . . 1. Sα ) . Mcd ] [D.13) To obtain the superextension (for D ≤ 6) one must ﬁrst extend the spinor representation associated with the D-dimensional spacetime to incorporate two extra gamma matrices ΓD and Γ− . Table 9) this requires a doubling of the spinor charges. Kb ] = 0 . de Wit: Supergravity 7.12) Here we distinguish the generator D of the dilatations. Γa = Γa 0 0 −Γa ΓD = 0 1 1 0 Γ− = 0 1 −1 0 . Ka ] [Mab . D generators Pa of the translations. 2) we have MD− Mab MDa M−a −→ D . We start from a D-dimensional spacetime of coordinates carrying indices a = 0.B. [Ka . .15) The new charges Sα generate so-called special supersymmetry transformations [27]. [D. Pa ] [Mab . so that A = −. In comparison to the anti-de Sitter algebra and superalgebra (cf. in particular. 1 D(D−1) generators 2 Mab of the Lorentz transformations. (6. 1. .5 (see. −2 ηc[a Kb] . to make contact with the Majorana condition employed for the anti-de Sitter algebra.4) and (6. Pb ] [Ka .

(7. S} . {Sα . Sα ] = −(QΓa )α . Its . S αβ ¯ {Qα . but those follow from the Rsymmetry assignments of the supercharges. From the results quoted in the previous section. S} anticommutator. which are cubic in the fermionic generators. commutators with the R-symmetry generators must be speciﬁed. S 2 ¯ ¯ [Pa . this mismatch will in general remain when including the R-symmetry generators.120 Unity from Duality: Gravity. In addition.16) may have zero eigenvalues. Those multiplets are in one-to-one correspondence with the anti-de Sitter supermultiplets. the bosonic subalgebra will be the sum of the conformal algebra and the R-symmetry algebra. ¯ [D. Gauge Theory and Strings The anticommutation relation for the spinor charges follows from (6. The results given so far suﬃce to discuss the most salient features of the superconformal algebra and henceforth we will be ignoring the contributions of the R-symmetry generators. Qα ] = 1 (QΓab )α . leading to shortened supermultiplets. Qβ } = −Γa Pa . Sβ } = − 1 Γab Mab + ηαβ D. ¯ ¯ [Mab . As before. Q} = = or ¯ {Qα . Qα ] = −(SΓa )α . The R-symmetry can be identiﬁed from Table 9 and the corresponding generators will appear on the righ-hand side of the {Q. up to D = 6. αβ ¯β } = −Γa Ka .18) Here we are assuming the same gamma matrix conventions as in the beginning of Section 3.16) ¯ ¯ {Q. Qα ] = 1 (QΓab )α . the matrix on the right-hand side of (7. Q 2 ¯ ¯ [Ka . the other (anti)commutation relations listed above remain unchanged. requires in general the presence of the R-symmetry charges. The validity of the remaining Jacobi identities. 2 ¯α ] = 1 Sα . Q} ¯ {S. 2 ¯ α ] = − 1 Qa .17) ¯ {S. The above (anti)commutators satisfy the Jacobi identities that are at most quadratic in the fermionic generators. Note also that the numbers of bosonic and fermion generators do not match. we know that. (7. S} 1 − 2 Γab Mab − D −Γa Ka − 1 Γab Mab 2 +D −Γ Pa a The nonvanishing commutators of the spinor charges with the bosonic generators read ¯ ¯ [Mab .5) and can be written as ¯ {Q. ¯ [D. 2 αβ (7. Q} {Q.

Combining (7. x) = exp(−y a Pa ) S(g.4) it then follows that the corresponding stability group of this point is generated by the generators M of the Lorentz group. x) φ(g −1 x) . Generic ﬁelds are thus assigned to representations of this subgroup. ∂ Pa φ(x) = φ(x) .B. Let us assume that the action of these spacetime transformations denoted by g takes the following form on a generic multicomponent ﬁeld φ. In principle. Indeed. D and ˆ a . de Wit: Supergravity 121 eigenvalues are subject to certain positivity requirements in order that the algebra is realized in a positive-deﬁnite Hilbert space. this point equal to xa = 0. we derive that conformal transforK 2 mations act inﬁnitesimally on φ according to δφ(x) = −ξ a ∂a φ(x) + 1 ab 2 (7. K On the other hand.4) in ﬂat spacetime introduced at the beginning of this section. The abstract algebra can be connected to the spacetime transformations (7.20) ∂xa so that φ(x) can be written as exp(xa Pa )φ(0). 0) must form a representation of ˆ ˆ this subgroup. this is an application of the theory of homogeneous spaces discussed in Section 4 and we will demonstrate this for the bosonic transformations [102].19) where S is some matrix acting on the components of φ. it follows that S(g. (7.22) [a 2ΛK xb] + ˆ Mab + ΛD − 2Λa xa K ˆ D+ ˆ Λa K a K φ(x) . To see this we derive how the conformal transformations act on generic ﬁelds.21) Writing this out for inﬁnitesimal transformations with y a = xa + ξ a .20). The procedure applied above is just a simple example of the construction of induced representations on a G/H coset manifold. φ(x)−→φg (x) = S(g. (7. From (7. a supersymmetric extension can be given in superspace. Observe that there exists a subgroup of the conformal group that leaves a point in spacetime invariant and choose. we are describing ﬂat space as a coset manifold. with y = g x . where ξ a denotes the conformal Killing vectors parametrized in (7.4). whose generators are denoted by the matrices Mab . 0) ≈ 1 + 1 ab Mab + ΛD D + Λa K. we want the translation operators to act exclusively on the coordinates xa . so that (the generators have been taken antihermitean). the generator D of the scale transformations and the generators K of the conformal boosts. where the conformal group plays . 0) exp(xb Pb ) .19) with (7. (7. Hence we conclude that the matrices S(g. by a suitable translation. and ˆ ˆ ˆ O(g.

Hence the metric is invariant under the conformal transformations. (3. which we specify below. Explicit evaluation then shows that the invariance of the ﬂat vielbein requires the compensating tangent-space transformations. The second approach starts from the coupling to superconformal matter and the corresponding superconformal currents. We start by associating a gauge ﬁeld to every generator.122 Unity from Duality: Gravity. Note that the special conformal boosts do not act on the tangent space index of the vielbein. The coset representative equals exp(xa Pa ). 7. (7. This conformal group has a priori nothing to do with spacetime transformations and the resulting theory is described in some unspeciﬁed spacetime. This is similar to what we described in Section 3.22). follow directly from the structure constants of the superconformal algebra. In the next two sections we will discuss how one can deviate from ﬂat space in the context of the conformal group. from which it follows (cf.22).23) with parameters speciﬁed by (7. 4 4 . (7. where we imposed a constraint on the torsion tensor (cf.16)) that the vielbein is constant and diagonal and the connections associated with the stability group are zero. δeµa = ab eµ. This approach amounts to imposing the maximal number of conventional constraints. the transformation rules for the gauge ﬁelds. so that the spin connection becomes a dependent ﬁeld and the Riemann tensor becomes proportional to the curvature of the spin connection ﬁeld. There are two approaches here which lead to related results. while the vielbein is invariant after including the compensating transformations represented by the second line of (7.24) Up to normalization factors. generators: gauge ﬁelds: parameters: P a eµ a ξP M ab ωµ ab D bµ ΛD K fµa Λa K Q ψµ S φµ η. as established earlier. δeµa ab δωµ a a = Dµ ξP − ΛD eµ + 1 ¯Γa ψµ . One is to start from a gauge theory of the conformal group.b − ΛD eµa . (4. Then one imposes a constraint on certain curvatures. 2 = Dµ ab [a b] [a ¯ + ΛK eµ − ξP fµb] − 1 ¯Γab φµ + 1 ψµ Γab η .6)).2 Superconformal gauge theory and supergravity In principle it is straightforward to set up a gauge theory associated with the superconformal algebra. Gauge Theory and Strings the role of the isometry group G and the stability group plays the role of the isotropy group H.

i. [µ ν] a 2 a ¯ 2 D[µ fν] − 1 φ[µ Γa φν] . δS (η2 )} {δQ ( ). 2 4 (7. δK (ΛK )] = {δQ ( 1 ).27) For completeness we also present the corresponding curvature tensors of the superconformal gauge theory. 2 2 4 4 a = Dµ Λa + ΛD eµ + 1 η Γa φµ . K K = ∂µ ΛD . 2 1 δM ( 1 ¯Γab η) + δD (− 4 ¯η) . δQ ( 2 )} {δS (η1 ). ν] 2 [a b] ab ac ¯ 2 ∂[µ ων] − 2 ω[µ ων] c b − 2 f[µ eν] + 1 ψ[µ Γab φν] .e. ab = (∂µ + 1 bµ − 1 ωµ Γab ) . 2¯ δS ( 1 Λ a Γa ) . δK (ΛK )] = [δS (η). 2 K [a b] δP ( 1 ¯2 Γa 1 ) . 4 δK ( 1 η2 Γa η1 ) . 2 2 2 123 = Dµ η + 1 ΛD φµ − 1 fµa Γa + 1 Λa Γa ψµ .. As an example we list some of the commutation relations that can be obtained from (7. Dµ Λa K Dµ ΛD Dµ Dµ η = ∂µ Λa − bµ Λa − ωµ ab ΛKb . δS (η)} = = = b δD ( 1 Λa ξP ηab ) + δM (ΛK ξP ) . K 2¯ a = Dµ − 1 ΛD ψµ − 1 eµa Γa η + 1 ξP Γa φµ .26) Again we suppressed the gauge ﬁelds for the R-symmetry generators. δP (ΛP )] = (7. 2 ¯ 2 D[µ bν] − f a e − 1 ψ[µ φν] . 2 Rµν (D) = a Rµν (K) = . In that case. 2 [δQ ( ). one needs Fierz reorderings to establish the closure of the algebra.B. which depend sensitively on the dimension and on the presence of additional generators (for D = 4. up to the commutators of two supersymmetry transformations acting on the fermionic gauge ﬁelds. de Wit: Supergravity δbµ δfµa δψµ δφµ ¯ = Dµ ΛD + 1 ΛKa eµa − 1 ξP a fµa + 1 ¯φµ − 1 ψµ η . a Rµν (P ) = ab Rµν (M ) = ¯ 2 D[µ ea − 1 ψ[µ Γa ψν] . for example [27]). The above transformation rules close under commutation. 2 4 ab = (∂µ − 1 bµ − 1 ωµ Γab )η . 2 K a δQ ( 1 ξP Γa η) .25) [δP (ξP ). 2 2 2 K (7. see.25) Here we use derivatives that are covariantized with respect to dilatations and Lorentz transformations. a Dµ ξP a a = ∂µ ξP + bµ ξP − ωµ ab ξP b .

just as in (3. Hence.30) a µ where ξ µ = ξP ea . a P -transformation takes the form of a (covariant) general coordinate transformation. making use of the fact that there exists an inverse vielbein ea . (7. based on curvature constraints. when imposing the torsion constraint R(P ) = 0.29) 4 4 1¯ ¯ + 4 φ[ρ Γab Rµν] (Q) + 1 ψ[ρ Γab Rµν] (S) = 0 . the superconformal algebra is not related to symmetries of spacetime. There is a procedure to introducing a nontrivial entangling between the spacetime diﬀeomorphisms and the (internal) symmetries associated with the superconformal gauge algebra. at this stage. (7. except that we are now dealing with arbitrary diﬀeomorphisms. the gauge ﬁelds independently transform as vectors under general coordinate transformations but these transformations have no intrinsic relation with the gauge transformations. Gauge Theory and Strings Rµν (Q) = Rµν (S) = a 2 D[µ ψν] − e[µ Γa φν] . Here one regards the P gauge ﬁeld eµa as a nonsingular vielbein ﬁeld. 4 At this stage. as one would expect for a conventional supersymmetric theory. so that the M gauge ﬁeld ab ωµ becomes a dependent ﬁeld. and one can impose a constraint R(P ) = 0. where ﬂat space was viewed as a coset space. . This is the reason why. To see this. whose inverse will be denoted by eaµ .6). where the P -transformations were also exclusively represented by coordinate changes. a 2 D[µ φν] − f[µ Γa ψν] . Of course. the curvature R(P ) has the interpretation of a torsion tensor. (7. In that case. The eﬀect of this constraint is also that the P gauge transformations are eﬀectively replaced by generalcoordinate transformations.124 Unity from Duality: Gravity. there is no need for the bosonic and fermionic degrees of freedom to match. let us rewrite a P -transformation a µ on eµ . a ab a ¯ D[µ Rρ] (P ) + R[µν (M ) eρ] b − R[µν (D) eρ] − 1 ψ[ρ Γa Rµν] (Q) = 0 .28) These curvature tensors transform covariantly and their transformation rules follow from the structure constants of the superconformal algebra. This interpretation is in line with the interpretation presented in the previous section. a a a a δeµ = Dµ ξP = ∂µ ξ ν eνa − ξ ν Dν eµ + ξ ν Rµν (P ) .22). 2 ab D[µ Rνρ] (M ) + R[µν (K) eρ] + R[µν (P ) fρ] [a b] [a b] a a D[µ Rνρ] (D) + 1 R[µν (K) eρ] a − R[µν (P ) fρ] a 2 ¯ ¯ + 1 φ[ρ Rµν] (Q) − 1 ψ[ρ Rµν] (S) = 0 . As an example and for future reference we list the ﬁrst three identities. This is completely in line with the ﬁeld transformations (7. They also satisfy a number of Bianchi identities which are straightforward to write down.

Indeed. which is proportional to R(Q). by doing so. because the constraint R(P ) = 0 is inconsistent with Q-supersymmetry. Because R(P ) = 0 is consistent with spacetime diﬀeomorphisms. Note that bµ is the only ﬁeld that transforms nontrivially on special conformal boosts and therefore acts as a compensator which induces all the K-transformations for the dependent ﬁelds. however. All other gauge ﬁelds have become dependent.31) = = where. local Lorentz transformations (M ).25). which can easily be eliminated by a ﬁeld redeﬁnition. under Q-supersymmetry. The superalgebra will. Inspection of (7. and the gravitino ﬁeld ψµ associated with Q-supersymmetry. the gauge ﬁeld bµ associated with the scale transformations. one should include possible modiﬁcations of the curvatures due to the changes in the transformation laws of the dependent ﬁelds. R(D) and R(Q) can be conventional and may lead to additional dependent gauge ﬁelds fµa and φµ associated with special conformal boosts and special supersymmetry transformations. 0. local conformal boosts (K). 0. This is also the case for S-supersymmetry.B. The gauge transformations remain with the exception of the P transformations. Because the constraints are consistent with all the bosonic transformations. Other than that. the ﬁeld the ﬁeld ωµab will still transform under these symmetries according to (7. Note that Rµν (D) is a not independent as a result of the ﬁrst Bianchi identity on Rµν (P ) given in (7. the transformation rules of the dependent ﬁelds are determined and they may acquire extra terms beyond the original ones presented in (7.29) and should not be constrained. de Wit: Supergravity 125 A constraint such as R(P ) = 0 is called a conventional constraint. we have diﬀeomorphisms. for reasons of covariance. local scale transformations (D). One of them is that there are potentially more conventional constraints. because it algebraically expresses some of the gauge ﬁelds in terms of the others. At this point we are left with the vielbein ﬁeld eµa . the ﬁeld ωµab acquires an extra term beyond what was presented in (7. takes the form a Rµν (P ) = ab ebb Rµν (M ) Γµ Rµν (Q) 0. Q-supersymmetry and S-supersymmetry. as one can verify by comparing the numbers of . those will not change and will still describe a closed algebra. not close.25). because constraints that diﬀer by the addition of other covariant terms result in the addition of covariant terms to the dependent gauge ﬁelds. the precise form of the constraints is not so important. and the bosonic conformal transformations. but not for Q-supersymmetry. (7. A maximal set of conventional constraints that achieves just that. Of course. We will not elaborate on the systematics of this procedure but concentrate on a number of noteworthy features.28) shows that constraints on R(M ).25).

a a i. Suppose that we couple such a rigidly superconformal matter theory in ﬁrst order to the gauge ﬁelds of conformal supergravity. Hence we write. corresponding 2 to the independent components of a symmetric.22).32) 2 a where hµ denotes the deviation of the vielbein from its ﬂat space value. eµa and bµ . The ﬁrst term denotes the matter Lagrangian in ﬂat space. 7. This theory has an obvious rigid limit. We close this section with a comment regarding the number of degrees of freedom described by the above gauge ﬁelds.126 Unity from Duality: Gravity. the supermultiplet of ﬁelds on which conformal supergravity is based. which are subject to the 1 D2 − 3 D − 1 independent. with the exception of the vielbein which a is equal to the ﬂat vielbein. eµa = δµ . Similarly. comprise the smallest massive supermultiplet whose highest spin coincides with the graviton spin. The independent bosonic ﬁelds. traceless. see [103]).. This representation is the minimal representation that e is required for an oﬀ-shell description of gravitons in D spacetime dimensions. where all the gauge ﬁelds are equal to zero. This construction is limited to theories with Q = 16 supercharges and leads to consistent conformal supergravity theories [22. In this background we may have (matter) theories that are superconformally invariant under rigid transformations. rank-2 tensor in D − 1 dimensions. comprise D2 + D degrees of freedom. which comprise (D − 2)ns degrees of freedom after subtracting the gauge degrees of freedom associate with Q. eµ ≈ δµ + hµa . This leaves us with 1 D(D − 1) − 1 degrees of freedom. 105]. In linearized approximation the above Lagrangian is invariant under local superconformal . 104. (7.3 Matter ﬁelds and currents In the previous section we described how to set up a consistent gauge theory for conformal supergravity. The fact that the ﬁelds can exist oﬀ the mass shell. A practical way to do this makes use of the superconformal multiplet of currents [22]. which constitutes an irreducible representation of the Poincar´ algebra. the conformal framework is set up to reduce the ﬁeld representation to the smallest possible one that describes the leading spin without putting the ﬁelds on shell.1. which we will discuss in the next section.and S-supersymmetry.e. gauge invariances of the conformal 2 2 group. bosonic. described by (7. The current θa µ is the energy-momentum tensor. Here ns denotes the spinor dimension. implies that they must constitute massive representations of the Lorentz group. A similar oﬀ-shell counting argument applies to the fermions. In order to have a consistent superconformal theory one must add additional ﬁelds (for a review. L = Lmatter a ab µ ¯ ¯ µ +hµ θa µ + 1 ωµ Sab + bµ T µ + fµa Ua µ + ψµ J µ + φµ JS . Gauge Theory and Strings bosonic and fermionic degrees of freedom. Hence. This is the background we considered in Section 7.

(7. a a translations: δhµ = ∂µ ξP . not all currents are conserved.25)). = 0. µ ∂µ Ua µ − 1 Saµ − 1 Ta 2 2 ∂µ J µ µ ∂µ JS + 1 Γµ J µ 2 = 0.4). ab a Lorentz: δωµ = ∂µ ab . For S-supersymmetry one can understand the expression of the current by noting that the following combination of a constant S transformation with a spacetime dependent Q-transformation with = 1 xµ Γµ η leaves the 2 gravitino ﬁeld ψµ invariant. Obviously.22) and in (7. ∂µ θa µ µ ∂µ Sab − 2 θ[ab] ∂µ T µ + θµ µ = 0.35) − b 2 θ[aµ xb] ∂µ (T µ + θa µ xa ) − 1 µ a 2T x − 1 µ 2 θb (xa x µ ∂µ JS + b − 1 x2 δa ) 2 ∂µ J µ µ ν 1 2 Γν J x In this result one recognizes the various components in (7. where ξ a are the conformal Killing vectors deﬁned in (7. ignoring variations that are proportional to the superconformal gauge ﬁelds and assuming that the matter ﬁelds satisfy their equations of motion. = 0. 1 δψµ = − 2 Γµ η. .4). δhµ = ab δµ b .32) under the superconformal transformations. One can verify that this leads to a number of conservation equations for the currents. To examine the consequences of this we need the leading (inhomogeneous) terms in the transformations of the gauge ﬁelds (cf. = 0. but we can deﬁne a set of conserved currents by allowing an explicit dependence on the coordinates. δhµ = −δµ ΛD .34) where we used the ﬂat vielbein to convert world into tangent space indices and vice versa. = 0.33) The variations of the action corresponding to (7. for instance. (7. S-supersymmetry: δφµ = ∂µ η . must vanish. ∂µ θa µ ∂µ ∂µ Ua − µ 1 µ b 2 Sab x µ Sab = 0. = 0. K 2 Q-supersymmetry: δψµ = ∂µ . (7. = 0. we employed the notation θab = θa µ eµ b and θµ µ = θa µ eµa . δωµ = ΛK δµ . b] [a (7. a a dilatations: δbµ = ∂µ ΛD . = 0. = 0. = 0 . ab conformal boosts: δfµa = ∂µ Λa . de Wit: Supergravity 127 transformations.B. Observe that the terms involving the energymomentum tensor take the form θa µ ξ a . δbµ = 1 ΛKµ .

38) 2 2 It is conserved by virtue of the ﬁeld equations. It is. and which can be solved by a potential χ. 4(D − 1) DA ∂B χ(φ) = gAB . µ µ suppressing Sab . possible to introduce an improvement term. for which DA χB = DB χA . fµa and φµ can then be written as linear combinations of curls of the independent gauge ﬁelds. moreover it is symmetric. (7. At the linearized level. however. We then obtain the following conditions for the improved currents. K and S. Ua µ and JS . To illustrate the construction of the currents.128 Unity from Duality: Gravity. . 2 (7. Ua µ and JS no longer appear explicitly but are absorbed in the remaining currents as improvement terms.40) the improved energy-momentum tensor is conserved. the currents µ µ Sab . imp θµν = A 1 2 gAB (∂µ φ + When χ(φ) satisﬁes D−2 ηµν ∂ 2 − ∂µ ∂ν χ(φ) .39) (7. ∂ν φB − 1 ηµν ∂ρ φA ∂ ρ φB ) 2 (7. which can be included into the currents without aﬀecting their divergence. After a partial integration.34). to depend on the other ﬁelds. the ﬁelds ωµab .37) Its energy-momentum operator can be derived by standard methods and is equal to θµν = 1 gAB ∂µ φA ∂ν φB − 1 ηµν ∂ρ φA ∂ ρ φB . Hence. This implies that. symmetric and traceless (again. We do not have to work out their explicit form. imp imp ∂ µ θµν = θ[µν] = θimp µ = 0 . we have argued in the previous section that it is possible to choose the gauge ﬁelds associated with the generators M . the currents θa µ . From this result it follows that locally in the target 30 A homothetic vector satisﬁes D χ + D χ A B A B = 2gAB Here we are dealing with an exact homothety. upon using the ﬁeld equations). in accord with the earlier counting arguments e given for the gauge ﬁelds.36) Observe that these equations reduce the currents to irreducible representations of the Poincar´ group. χA = ∂A χ is a homothetic vector30. (7. T µ and J µ are modiﬁed by improvement terms: terms of the form ∂ν A[νµ] . L = 1 gAB ∂µ φA ∂ µ φB . However. but not traceless. µ imp imp ∂ µ Jµ = Γµ Jµ = 0 . Gauge Theory and Strings So far we have assumed that the gauge ﬁelds in (7. because we can simply repeat the analysis leading to (7.32) are independent. let us consider a nonlinear sigma model in ﬂat spacetime with Lagrangian.

43) where the φ-independence of hab can be deduced directly from (7. 2 129 (7.45) where w is the Weyl weight of the scalar ﬁelds which is equal to w = 1 2 (D − 2). when the cone is a K¨hler or hyperk¨hler space. 0) and gAφ = χA = (2χ. (7.39) to gravity must lead to a conformally invariant theory of the nonlinear sigma model and gravity. ϕ) = exp[2φ] χ(ϕ). 4(D − 1) (7.B. From this result one proves directly that the metric takes the form (ds)2 = (dχ)2 + χhab (ϕ) dϕa dϕb . we decompose the target-space coordinates φA into φ and remaining coordinates ϕa . the cone must also be invariant under U (1) a a or SU (2). Using the gauge ﬁelds of the conformal group. e−1 L = 1 gAB g µν (∂µ φA − w bµ χA ) (∂ν φB − w bν χB ) − 1 w fµ µ χ . de Wit: Supergravity space. 2χ (7. In the supersymmetic context it is important to note that. In terms of these new coordinates we have χA = (1. χ can be written as χ = 1 g AB χA χB .40).2. the Lagrangian reads. e−1 L = 1 gAB ∂µ φA ∂ µ φB − 2 D−2 χ(φ) R . To see this. 0. δD φA = wΛD χA .46) 2 2 . where φ is deﬁned by χA ∂ ∂ · = ∂φA ∂φ (7. Coupling the improved energy-momentum tensor (7. gaφ ). The relevant Lagrangian reads. Spaces that have such a homothety are cones. We should also point out that the coupling with the Ricci scalar can be understood in the context of the results of the previous section.44) Indeed.42) It then follows that χ(φ. The transformation of gµν is in accord with the vielbein scale transformation written down in Section 7. These features play an important role when extending to the supersymmetric case. This result shows that the target space is a cone over a base manifold MB parametrized in terms of the coordinates ϕa with metric hab [106]. In that case U (1) or SU (2) must be associated with the R-symmetry of the superconformal algebra. · · · . where χ is an undetermined ˆ ˆ function of the coordinates ϕa . this Lagrangian is invariant under local scale transformations characterized by the functions ΛD (x).41) up to an integration constant. δD gµν = −2ΛD gµν (7.

44). see. Green. Ferrara. u 1989). Supersymmetry. van Nieuwenhuizen. with the dimension D−2 of [mass] . Phys. we refer to [67. this Lagrangian is invariant under local dilatations. Superstring theory. Supergravity – The basics and beyond. S.47) χ = χ0 . References [1] P. Schwarz and E.48) This Lagrangian describes a nonlinear sigma model with the base manifold MB of the cone as a target space. 4(D − 1) (7. for example.31) for the gauge ﬁeld fµ a associated with the conformal boosts and setting bµ = 0 as a gauge condition for the conformal boosts. B. String . [81]. The above example forms an important ingredient in the so-called superconformal multiplet calculus that has been used extensively in the construction of nonmaximal supergravity couplings. I & II (Cambridge Univ. 1989). Substituting the metric (7. Vols. For an introduction to the 4-dimensional N = 1 multiplet calculus. Flume. 1985). A. Rittenberg (Plenum. E. 1987). Press. An obvious one amounts to putting χ equal to a constant χ0 . de Wit and D. The ﬁrst 15-years of superstring theory. [2] M. Salam and E. Superstrings.43) the Lagrangian then acquires the form. L¨st and S. This example thus demonstrates the relation between improvement terms in the currents and constraints on the gauge ﬁelds. Vols. 1 & 2 (World Scientiﬁc. coupled to (nonconformal) gravity. which is still invariant under local dilatations. 1 & 2 (North-Holland/World Scientiﬁc. 1 & 2 (Cambridge Univ. see. 1998). String Theory. G.g. Freedman. von Gehlen and V. R. Kiritsis. 1 & 2 (NorthHolland/World Scientiﬁc. 1985). Theisen.Z. the metric hab should be negative deﬁnite and χ0 must be positive. Vols. D. 419 (1998) 265 [hep-th/9708130]. (7. 1989). NATO ASI B 125 135. 107]. Phys. for 4-dimensional N = 2 vector multiplets and hypermultiplets. the Lagrangian becomes equal to (7. J. Dietz. Theor. Vols. Dine. For a collection of reprints. AIP Conf. Vols. Lectures on string theory (Springer. Supergravities in diverse dimensions. e.25). Rep. in Supersymmetry. Proc. There is an extensive literature on this. Leuven Notes Math. edited by K. Polchinski. The constant χ0 appears as an overall constant and is inversely proportional to Newton’s constant in D spacetime dimensions. J. Press. e−1 L ∝ 1 hab ∂µ ϕa ∂ µ ϕb − 2 D−2 R. Witten. 9 (1997) [hep-th/9709062].H. 68 (1981) 189. Upon using the second constraint (7. M. Schwarz. Observe that in order to obtain positive kinetic terms. conformal boosts and spacetime diﬀeomorphisms. J.130 Unity from Duality: Gravity. Gauge Theory and Strings As one can easily verify from the transformation rules (7. Sezgin. It is possible to also employ a gauge condition for the dilatations.

G. Branes and Dualities. Zumino. Lett. de Wit. lectures given at the 1996 ICTP Summer School in High Energy Physics and Cosmology. R. E. Nicolai. B 451 (1995) 547 [hep-th/9504081]. P. Obers and B. Howe and P. F. Nucl. J. Bergshoeﬀ. P. L. Hawking and M. edited by Dj. Taylor (Cambridge Univ. B. Cremmer. JHEP 0006 (2000) 019 [hep-th/0004086]. Scherk and D. Strathdee. p. Di Francesco. Kounnas. Schwarz. Phys. P. Douglas. J. Nucl. B 322 (1989) 301.H. Supergravities in 5 dimensions. de Roo and B.B. Di Francesco. Sijaˇki. F. Phys. Superstring construction (North-Holland. Scherk. S. M. in Frontiers in Particle ˇ c Physics ’83. Bergshoeﬀ. L¨ st. W. B 226 (1983) 269. West. N. Andrianopoli. L. B 100 (1981) 393. Nicolai and D. Phys. Phys. M. 79 (1981) 1. Pernici. Schwarz. 1999).A. Phys. Phys. D’Auria. H. Roˇek (Cambridge Univ. Lett. Ferrara and C. Nucl. A.G. B. I. 152. B 392 (1993) 3 e [hep-th/9208074]. de Wit. B. Dai. Green and J. p. Rep. J. Branes and Dualities. P. Schellekens. Press. B 122 (1977) 253. E.A. Nucl. Dine. B 76 (1978) 409. Windey (Kluwer. Cremmer. Phys. Phys. Townsend. c C. Windey (Kluwer. 2) Chiral supergravity in six dimensions and solvable Lie algebras [hep-th/9711048]. edited by L. Four lectures on M-theory. Phys. Popoviˇ (World c c c Scientiﬁc. A 4 (1989) 2073. Mod. Press. A. de Wit. Ortin. Representations of extended supersymmetry on one.G. B 243 (1984) 112. Lett. Supersymmetry and dualities in various dimensions. Ferrara. in Strings. Phys. Pioline. Savoy and B.C. Slansky. Nucl. B 466 (1999) 144 u [hep-th/9908169]. Namazie. Lett. edited by S. Giani and M. Phys. Rev. p. M. B 238 (1984) 181.C. de Wit and J. E.K. N. Tollst´n and H. Ferrara and C. Lett. 1988). 1982). in Supergravity ’81. Kazakov. P. R. 318 (1999) 113 [hep-th/9809039]. edited by L. 141 [hep-th/9712004]. S. NATO ASI C520. C. de Wit. D’Auria and S. 267. C. Baulieu. J. 2 (1985) 293. M. V. in Strings. J. 1984). Picco and P.M. Lett. Maximal supergravity. Phys. Lett. M-theory from its superalgebra. Phys. E. Hull and T. edited by S. B 139 (1984) 283. p. Schwarz and P. B. p. Leigh and J. Gliozzi. Nucl. 33 [hep-th/9801132]. B 135 (1978) 149.C. Rep.A. 1981). 1989). N = (4. NATO ASI C520.W. Ferrara. Phys. Nucl. D 30 (1984) 325. Mod. B 122 (1983) 143. Huq and M.H. Douglas. Baulieu. B 126 (1983) 301. M. Bili´.) 55 (1997) 145. Phys. Julia and J. Nucl. Quantum Grav. Scherk and J.A. J. Louis. Polchinski. M. R. Seiberg.K. Campbell and P. Schwarz. Dragoviˇ and D. Nahm. V. Phys. Class. Van Proeyen. B 121 (1977) 77. Huet and N.C. Suppl. 47. A. R.H. J. West. P. Savoy. Phys. S. B 583 (2000) 237 [hep-th/0004195].A. M. Townsend. M. Phys. M. in Superspace & Supergravity. 1999).N. Phys. Ferrara and J.H. Nucl. Ferrara. Phys. Phys. Trieste [hep-th/9612121]. B (Proc. Hull. Brink. B. Abou-Zeid. de Wit: Supergravity 131 [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] Theory in four dimensions (North-Holland. Nucl. B 182 (1981) 173. West. Olive. U-Duality and central charges in various dimensions revisited [hep-th/9612105].B.and two-particle states. S. Picco and P. Nucl. Nucl. Kazakov. . Phys. Tools for supersymmetry [hep-th/9910030]. Phys. Int. A 2 (1987) 273. B.

Adv. van Nieuwenhuizen. B.X. A 13 (1980) 1643. Introduction to supergravity. Duﬀ. . G¨ven. Warner. Pope. 331. 287 [hep-th/9701088]. p. edited by S. Freund. T. P. [35] K. F.K. Phys. 1981). B 129 (1977) 125. van Nieuwenhuizen and B. [31] P. B 69 (1977) 481. Van Proeyen. u [30] B. [44] M. Spin 2 matter-graviton coupling problems. Math. Phys. Olive. J. [38] C. de Wit and H. Freedman (North-Holland. Vanderseypen and A. Ferrara. [42] P. J. S. Ferrara. for instance. D 18 (1978) 3630. 18 (2001) 3095 [hep-th/0011239]. p. and references quoted therein. [32] S. J. Lett. Curtright. Duﬀ and K. B 145 (1984) 368. B 253 (1991) 113. Nucl. Suppl. Particles. Rev. Rep. de Wit. D 18 (1978) 3264. [36] C. Phys. Ferrara. Duﬀ. F. J. M. Berends. Phys. [49] B. Kaku.C. Rev. Lett. Phys. B 85 (1979) 219. 130 (1986) 1. P. B. Phys. Maldacena. B 124 (1983) 45. de Wit. Coquereaux. van Nieuwenhuizen (World Scientiﬁc. S. Chamseddine and P. L. L¨ . Lett. in Supersymmetry and Supergravity ’84. van Holten. Rudaz and M. p. West. Phys. Fang and C. Lectures on supergravity p-branes. High energy physics and cosmology. B 86 (1979) 161. Stelle. Rev. Stelle. 49. edited by S. D 5 (1996) 763 [hep-th/9611024]. [45] R.E. B. B 276 (1992) 49.W. Gauge Theory and Strings [26] R. Proc. u [46] For a review see. Romans and N. Schwinger.R. in Supergravity. [52] Y. B (Proc.J. Phys.S. Phys. Biran. Nucl. D 15 (1977) 2802. Nicolai. Townsend and P. de Wit. F. B 400 (1993) 463 [hep-th/9210068]. J. Int. edited by P. in Uniﬁcation of fundamental particle interactions. K.H. [33] S. T Appelquist. Lu [hep-th/0010292]. Phys.N. D 28 (1983) 2976. Rep. 1980). A. Phys. B 272 (1986) 598. Lett. edited by B. Lett. 137. Zumino. Phys. Proc. Class. B 67 (1977) 439. p. Phys. R. 38 (1977) 1433. Phys. Phys. Lett.A. Zwartkruis. M.W. Townsend. see. Deser. Phys. Page.N. B. in Superspace & Supergravity. de Wit and H. Englert. 1984). J. Class. Rev. J. Lett.Z. Vasiliev. for example. Deser. [43] D. N=8 Supergravity. de Wit and A. [34] B. [47] B. van Nieuwenhuizen and D. [48] For a reprint collection and a review of Kaluza-Klein theory. Bautier. Henneaux and D. de Wit and D. Nicolai. Deser and B. Duﬀ. A 16 (2001) 1002. Phys. 1979). Shifman [hep-th/0103086]. Quantum Grav. Aragone and S. [37] C.T. B 129 (1977) 39. van Nieuwenhuizen (Plenum.J. Phys. Int. [27] S. Mod. de Wit.Z. edited by M. Phys. Phys. “Group disintegrations”. Ellis and P. Freedman. Phys. Quantum Grav.W. [29] M.A. Phys. [41] B. [40] B.J. Nucl. Deser.J. [50] E. Liu and J.P. P. Tanii. D 21 (1980) 358. Modern Kaluza-Klein Theories (Addison-Wesley. Rev. Fronsdal. Khuri and J. Townsend and P. Nucl. Nucl.S.J.) 101 (2001) 154. B 115 (1982) 389. G¨naydin. Strings 2000. edited by K. de Wit. Phys. Lett.A. 1987). Lett. [28] J.K. Rev. J. sources and ﬁelds (Addison Wesley. Phys. 1970). in Trieste 1996. 2 (1998) 231 [hep-th/9711200]. M. [39] M. Phys. c [51] B. 4 (1987) L59. de Wit.O. van Nieuwenhuizen. Phys. A. M. Seminara. B 406 (1997) 49.G.K. Lett. Phys. Thirty Years of Supersymmetry. Chodos and P. u 259 (1995) 213 [hep-th/9412184]. Hawking and M. Roˇek (Cambridge University Press. Cremmer.132 Unity from Duality: Gravity. Fayet and P. Aragone and S. Nilsson and C. Lett. Julia. Mod. Theor. Fronsdal.

Bergshoeﬀ. R. Phys. . Phys. Suppl. B 245 (1984) 89. in Supersymmetry and Supergravity ’82. E. J. B 189 (1987) 75. van Nieuwenhuizen. S. Notes Phys. Van Nieuwenhuizen (World Scientiﬁc. Julia. de Wit. B. E. B 158 (1979) 189. Bergshoeﬀ. Phys. Phys. 25 (1984) 3455. Hull. Phys. Gilmore. Rev. Fr´. C. Lett. Van Proeyen. B. V. B 367 (1996) 97 [hep-th/9510086]. 1974). Towards Quantum Gravity. Townsend. Townsend. Phys.K. 1991).) 46 (1996) 30 [hep-th/9508154]. New gauged N = 8. A 4 (1989) 2475. Phys. Pilch and P. 86 (2001) 1686 [hep-th/0010076]. Nucl. Girardello. edited by L. Cecotti. Schellekens. NATO ASI C520. B. E.H. van Nieuwenhuizen.M. S. B 136 (1984) 38. 79 [hep-th/9908005]. Lett. B 460 (1996) 525 [hep-th/9510169]. JHEP 0104 (2001) 022 [hep-th/0103032].G. Storey. JHEP 0207 (2002) o 010 [hep-th/0203206]. Fr´ and L. L. B (Proc.K. de Wit and A. D’Auria. B 350 (1995) 184 [hep-th/9501068].K. Qunatum Grav. Hawking and G.K. 1999) p. Multiplet calculus. B 143 (1984) 103. Lled´. Gualtieri.J. de Roo.F. Chamseddine. Lett. de Wit and P. Nucl. M.-W. Ann. also in Polanica 1999. B 108 (1982) 285. Nucl. Phys. Lett. R. Lie algebras. B 195 (1982) 97. Phys. Antoniadis. Rev. B. I. Ferrara. Taylor and P. H. P. Nicolai. A. Townsend. Math. P. B 80 (1978) 48. de Wit. B 373 (1996) 68 [hep-th/9512062]. R. 185 (1988) 330. S. 541 (2000) 79. Rev.M. Phys.K. Kazakov. Schwarz. de Wit and I. C. The large scale structure of space-time (Cambridge Univ. de Wit. Phys. Nucl. Phys. Phys. Castellani. E.S. H. M. D 30 (1984) 760. Phys. Phys. Lett. Lect. D’Auria and P. Nucl. M. K. Nucl. Class. Phys. Baulieu. Taylor. Nucl. Pernici. D 18 (1978) 3565. D 24 (1981) 3065.J.H. Lett. B 217 (1983) 143. L. B 142 (1984) 39. Quantum Grav. Nucl. J. Aspinwall. Hull and P. H. F. P. de Wit. Int. M. Nicolai and H. Press. p. B 148 (1984) 297.B. S. 179 [hep-th/9706211]. de Roo and B. G. Trigiante (to appear). 1973). B 208 (1982) 323. Picco and P. Lett. Ellis. Lett. Polchinski and E. Phys. Samtleben. B 120 (1983) 105. J. B. Lett. Lett. edited by S. J. Di Francesco. Phys. Phys. Windey (Kluwer. Nucl. Phys. Chapline and N. Phys. Ferrara and L. 18 e (2001) 395 [hep-th/0009048]. J. Witten. Phys. Samtleben and M. Pilch and P. Bergshoeﬀ. Class. Samtleben. B. 1983). Zumino. Van Proeyen. Phys.A. Townsend. Lett. Rev. L. D = 4 supergravities [hep-th/0204156]. Andreanopoli. Manton. de Wit: Supergravity [53] [54] [55] [56] [57] [58] [59] 133 [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] E. J. Phys. K. B 443 (1995) 85 [hep-th/9503124]. de Wit and H. Phys. Phys. B 438 (1995) 109 [hep-th/9410167].S.A. Phys.R. B 193 (1981) 221.R. Supergravity and superstrings I (World e Scientiﬁc. Ferrara and M. Nucl. P. J. E. Nicolai and H.Andreanopoli. Cremmer and B. M. and some of their applications (Wiley Interscience. Pilch and A. Gaillard and B. K. Cordaro. C. Partouche and T.W. Phys. 2 (1985) 343. Sezgin and P. H. van Holten and A. P. van Nieuwenhuizen. M. Lett. Douglas. Duality and spontaneously broken supergravity in ﬂat backgrounds [hep-th/0204145]. S. A 15 (1982) 3763. B. Witten. Lie groups. Avis.N. Mod. Herger. Isham and D. Phys.

de Wit. D 12 (1975) 3810. Fronsdal. edited by B. Gubser. Nucl. [89] M.M.J. Fronsdal. Mack and A.N. edited by M. Ann. Math. P.M. Oz. G¨naydin u u and N. G¨naydin and C. Quantum Grav. Nucl. M. P. van Nieuwenhuizen. D 12 (1975) 3819. [hep-th/9712239]. B 523 (1998) 145 [hep-th/9802047]. Gibbons and P. Rev. Y. Ferrara and C. Sundell. Lett. 2 (1998) 253 [hep-th/9802150]. Quantum Grav. Rev. Fronsdal and R.M. B 237 (1984) 342. I. Phys. Fronsdal. Kaku. Nicolai. M. B 272 (1986) 99. Saclioglu. [93] S. Phys. Commun. 1983). Mod. Phys. Phys. 1982). Math.K. [103] B. Conformal higher spin theory [hep-th/0207212]. Maldacena. Marcus. H. 4 (1963) 901. Nucl. Phys. P. Singleton and doubleton supermultiplets of space-time supergroups u and inﬁnite spin superalgebras. Witten.A. Phys. [101] S. D 10 (1974) 589. Phys. Commun. C. [102] G. G¨naydin.-W. Freedman. Zagerman. Brown and M. Math. van Holten and A. C. [85] H. 144 (1982) 249. [100] O. Phys. Skenderis and S. Salam. G¨naydin and D. Phys. Balasubramanian. Rychenkova. Math. B 534 (1998) 96 u [hep-th/9806042]. Townsend and P. 442. Van Proeyen. Rev. 1982). D 19 (1979) 3166. in Int. Haugen. C.P. Segal. Taylor (Cambridge Univ.S. Class. J. Henneaux. Polyakov. 15 (1998) 2153 [94] E. Kraus and A. P.P. Klebanov and A. Oscillator like unitary representations of noncompact groups and suu pergroups and extended supergravity theories. Freedman and H. Lecture o u Notes Phys. van Holten. [87] M. Gauge Theory and Strings [82] P. [96] J. Dirac. Phys.R. S. Rev. van Nieuwenhuizen and N. Phys. K. Nucl. [106] G. de Wit. Commun. in Supermembranes and physics of 2+1 dimensions. de Wit and J. Nicolai. Conformal invariance in extended supergravity. Phys. Phys. 91 (1982) 159. Phys. Phys. D 17 (1978) 3179.Z.Z. [91] W. Representations of supersymmetry in anti-de Sitter space.S. Sezgin (World Scientiﬁc. 180 (Springer. [88] M. Rev. Math. van Nieuwenhuizen (World Scientiﬁc. B. J. B 110 (1982) 461. 323 (2000) 183 [hep-th/9905111]. Massless higher spins and holography [hep-th/0205131]. Solodukhin. Phys. Colloq. Warner. Rev. 1990). 37 (1965) 221. G¨naydin and N.134 Unity from Duality: Gravity. Minic and M. Fayet. Phys. u M. Phys. [86] P. u G¨naydin. G¨naydin. [90] M. [104] M. Phys. Nucl. Ferrara and J. Rev. Phys.D.B. Aharony. edited by S. Minic. G¨naydin D. in Supersymmetry and Supergravity ’84. Duﬀ. van Nieuwenhuizen. P.N. Ooguri. 104 (1986) 207. B 155 (1979) 530. . Phys. in Supergravity ’81. Nucl. [99] V. [97] S. [98] E. Adv. Breitenlohner and D. In¨n¨. Lett.G. [95] A. 2 (1985) L11. 53 (1969) 174. Press. Sezgin and P. Nucl. de Haro. B 255 (1985) u 63. Heidenreich. J. Phys.-W. Gubser. [92] M. Rep. edited by M. Lawrence. B 443 (1998) 138 [hep-th/9809158]. Phys. [105] B. Lett. Ann.K. Serdaroglu and E. [83] C. on Group Theoretical Methods in Physics (Istanbul. p. Phys. D 59 (1999) 046003 [hep-th/9805171]. Class. 217 (2001) 595 [hep-th/0002230].W. B 167 (1980) 186. 1984). Theor. [84] D. Townsend and P. B 428 (1998) 105 [hep-th/9802109]. Warner. P. Pope and E. de Wit. Phys.

B.B. de Wit. Kleijn and S. p. Roˇek and S. Vandoren. Vandoren. B. de Wit: Supergravity 135 [107] K. 37 [hep-th/9808160]. . in Dubna 1979. Quantum Grav. JHEP c 0102 (2001) 039 [hep-th/0101161]. M. Galicki. Rigid N = 2 superconformal hypermultiplets. 9 (1992) 27. de Wit. Phys. B. Nucl. B 568 (2000) 475 [hep-th/9909228]. Supersymmetries and Quantum Symmetries. Class.

Jerusalem.LECTURE 2 SUPERSYMMETRIC GAUGE THEORIES E. RABINOVICI Racah Institute. Israel . Hebrew University.

. . . . . . .8 Higgs and conﬁnement phases . . . . . . . . . . . . . . . . . . . . . . . .7 Superconformal quantum mechanics . . . . . . . 3. . . . . 2. . . . . . . . . . . . . . 5. . . 2. . 6 Comments on vacuum energies in scale invariant theories . 5. . . . . . . . . . . . . . . . . .4 Integrating in . . . . . . . . . . . . . . . .6 NF = NC . . . . . . .5 Quantum moduli space for NF ≥ NC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 A nonrenormalisation theorem . . . . . . . . . . . . . . 5. . 5. . . . . . . . . . . . . . . . . . . . . . 5.10 Superconformal invariance in d = 4 . . . . . .1 Symmetry and symmetry breaking . . . . . . . . . . . . . . . . . .2 Superspace and chiral ﬁelds . . . . . . .3 Quantum moduli space for 0 < NF < NC 5. . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 Supersymmetry breaking . . . . . . . .1 The classical moduli space . . . . . . . . . . . . . . . . . .5 Global symmetries . . . . . . . .4 F-terms . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . . . . .1 Kinematics . . . . . 5. . .6 The eﬀective potential . . . . .9 Infra-red duality . . . . . . . . . . . . .7 NF = NC + 1 . .8 Supersymmetric gauge theories 4 Phases of gauge theories 5 Supersymmetric gauge theories/super QCD 5.4 Geometric meaning of the Witten index . . . . 2. . . . . . . 3. . . . . 3. . . 3 Review of supersymmetric models 3. .2 Quantum moduli spaces . . . . . . . . . . . . . . . .6 Conformal quantum mechanics .5 Landau levels and SUSY QM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. . . a 3.Contents 1 Introduction 2 Supersymmetric quantum mechanics 2. . . . . . . . . . . .3 K¨hler potentials . . . . . . . . . . . . . . . . . 2. .3 A two variable realization and ﬂat potentials 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . 141 143 151 152 154 158 159 161 164 165 165 167 169 170 170 172 172 173 179 180 182 185 185 186 190 190 191 192 194 202 207 . . . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . .8 Concluding remarks . . . 7. . . . .7 Obtaining the dual ﬁeld theory . 7. . N = 1. 7. 7. 210 210 211 213 216 226 229 233 235 237 . . . . . . 8 Final remarks . . 7. . .1 Branes in string theory . . . . . . . . . . .2 Branes in IIA and IIB string theories . . . . . . . . . . . . .3 The eﬀective ﬁeld theory on branes . . . . . . . . . .5 An eﬀective D = 4. . . . . . . . . . .4 Eﬀective D = 4 dimensional systems with N = 2 supersymmetry 7. . . . . . 7. . . . . . . . U (NC ) gauge theory with matter . . . . . . .6 More pieces of information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 Supersymmetric gauge theories and string theory 7. . . . . . . . . . .

the phase structure of supersymmetric systems with and without gauge particles. Hebrew University. Jerusalem. the Israel academy of sciences and humanities-Centers of excellence program. little did they know that the constrained interactions [2] will give rise to a multitude of vacua. A discussion of the relation of conformal symmetry to a vanishing vacuum energy (cosmological constant) is included.S. • In their seminal paper Golfand and Likhtman [1] introduced supersymmetry in order to constrain the possible forms of interaction. The weak interactions are in the Higgs phase. Rabinovici1 Abstract We introduce simple and more advanced concepts that have played a key role in the development of supersymmetric systems. This work is supported by BSF. None of these phases exhibits supersymmetry explicitly. Springer-Verlag 2002 . the european RTN network HPRN-CT-2000-00122 and DOE grant DE-FG02-90ER40560. Israel.SUPERSYMMETRIC GAUGE THEORIES D.the American-Israeli Bi-National Science foundation. the force in nature that has yet to be tamed by theoretical physics. The electromagnetic interaction is in the Coulomb phase and the colour interactions are in the conﬁning phase. This is done by ﬁrst describing various supersymmetric quantum mechanics models. 1 Introduction At least three phases of gauge theory are manifest in nature. Berman1 and E. 1 Racah Institute. What is then the motivation to introduce and study supersymmetry? Here is a list of reasons that have motivated people over the years. superconformal theories and infrared duality in both ﬁeld theory and string theory. • The local version of supersymmetry contains automatically gravity [3. 4]. c EDP Sciences. the GermanIsraeli Bi-National science foundation. Topics covered include the basic construction of supersymmetric ﬁeld theories.

The ideas include: supersymmetry and its spontaneous breaking.142 Unity from Duality: Gravity. The construction and analysis of supersymmetric systems requires the use of many diﬀerent concepts and ideas in ﬁeld theory. • Strong coupling problems are generically intractable in ﬁeld theory. and treating theories with no ground states. This problem is much softer in supersymmetric theories [5–7]. In fact. Such models serve as analytical laboratories for physical ideas. This duality is described from both a ﬁeld theoretical and string theoretical point of view. • Perturbative string theory seams to need supersymmetry in order to be stable. index theorems as a tool to answer physical questions. Section three will contain a review of the methods usedto construct supersymmetric ﬁeld theories and supersymmetric gauge theories. • In various supersymmetric models the gauge couplings unify sensibly [8–10]. the impact of conformal symmetries. 15]. This section includes also a discussion of conformal ﬁeld theories and some properties in dimension greater than two including four. called the Hierarchy problem. the vacuum energy of these theories is discussed in relation to the cosmological constant problem. In the process we will discuss connections between string theory in the presence of branes and gauge theories. Gauge Theory and Strings • The weak interaction scale and the Plank scale are widely separated for theoreticians this could be a problem. In particular. In section ﬁve we describe the phenomena of infra-red duality in supersymmetric gauge theories. 4. In section two we will introduce several of these ideas by using a simple quantum mechanical context. In the presence of supersymmetry constraints the problems can be solved in interesting cases. They include the understanding of conﬁnement. In section four. in order for string theories to make sense the number of Bosonic degrees of freedom can diﬀer from those of the Fermionic degrees of freedom by at most those of a d = 2 ﬁeld theory [11]. we review the intricate phase structure of classical and quantum supersymmetric gauge theories. the material in this section is elaborated in similar or greater detail in [12. In particular the idea of monopole condensation as the responsible mechanism for conﬁnement becomes tractable. It may occasionally seem that one sees only the trees and not the forest however note that many of the trees are central problems in ﬁeld theory for years. 12–18]. There are many excellent reviews on the subject. . chiral symmetry breaking and the emergence of massless fermions to mention just a few. we refer the reader to them for a more complete exposition [3.

g.N (2. .4) .3) α 2 2 ∂xα ∂xα ∂xβ α αβ and B is: Bαβ = 1 + [ψ .. Rabinovici: Supersymmetric Gauge Theories 143 Many topics in the study of supersymmetric theories which are as worthy have not been covered for lack of time or because they were covered by other lectures. supersymmetry on the lattice [22] as well as many others.. 2 Supersymmetric quantum mechanics The ideas in this section were introduced in [24. how is SUSY broken spontaneously. ψβ ] 2 α (2. non-supersymmetric deformations of the models described e. and what are nonrenormalisation theorems. xn ) is a general function of the n bosonic variables. The Hamiltonian is the only member of the Poincare group in that case. Hamiltonian is then given by: ∂2W 1 ∂W 2 1 p2 + Bαβ 1L×L − H = {Q. the basic anti-commutation relations that deﬁne the supersymmetric algebra are: {Qi . The questions we wish to examine are: what is supersymmetry. Along the way we will introduce some useful tools such as the Witten index. [20]. j = 1. 25].1) H is the Hamiltonian and Q.. Thus. A rather general N = 1 realization with n bosonic and n Fermionic degrees of freedom is given by: n Q= α=1 + ψα −pα + i ∂W ∂xα . Later we will present a two variable realization where we can introduce the notion of a ﬂat direction.D. (2. Q+ } = 2Hδij j i. supersymmetric matrix models [21]. First we will examine the quantum mechanical realization of the supersymmetry algebra so as to introduce various ideas that will later carry over to ﬁeld theory. Berman and E. Quantum mechanics is a one dimensional ﬁeld theory.2) The W (x. Q+ } = (2. N denotes the number of supersymmetries. Q+ are called the supercharges.S. Now we begin with a one variable realization of N = 1 SUSY. Examples of such topics are: large N gauge theories [19].

The Hamiltonian in these variables then becomes: H= ω N+ 1 2 (2.144 Unity from Duality: Gravity. ψβ } = 0 .12) The spectrum of this operator is given by the nonnegative integers. (2. Gauge Theory and Strings and the ψ variables obey the following anticommutation relations. N is constructed out of the creation and annihilation operators: N = a+ a. p = √ (a+ − a) 2 2 and so we obtain the commutation relation: [a. . H= p2 1 q + mω 2 q 2 2m 2 (2. a+ ] = 1. + + + {ψα .9) (2.13) and the energy spacings are given by the scale ω.6) deﬁne The energy scale is extracted by deﬁning dimensionless variables x and px . a by: 1 i x = √ (a + a+ ) . It is useful to deﬁne creation and annihilation operators. {ψα . px = (mω )− 2 pq .7) This gives the following commutation relations: −i = [pq .11) (2. ψβ } = 0. we ﬁrst will recall some basic facts about the Bosonic harmonic oscillator. 1 (2.10) The number operator.8) ω is the energy scale in the problem. ψβ } = δαβ . (2. a+ .5) The dimension of the Fermionic Hilbert space is L = 2n . the Hamiltonian is now given by: 1 H = ω (p2 + x2 ) 2 x (2. (2. Before moving on to n = 1 realization of this super algebra. x] . q] = [px . x= mω 1 2 q. {ψα .

Rabinovici: Supersymmetric Gauge Theories 145 The energy of the ground state diﬀers from the classical minimum energy 1 2 ω. |0 is obtained by solving the equation: by a|0 = 0. This is essentially due the uncertainty relation. The ground state. This completes the solution of the Bosonic harmonic oscillator. n! (2.18) x2 2 · (2.S. That is: x+ d dx φ(x) = 0 (2. The eigenstates.21) (2.19) So why not obtain a ground state energy of − 1 ω by solving 2 a+ |GS = 0. Using equations (2. Berman and E.D. It is given by: φ(x) = c exp 1 2 x . 2 (2. This is possible only for the ground state.15) yielding the ground state wave function: φ(x) = c exp − The state |n is given by: (a+ )n |n = √ |0 . 2.17) which can now be expressed in x-space.14) Note that this equation is a ﬁrst order as opposed to the second order equation that one would have to solve if one attempted to directly solve the Schroedinger equation.16) (2. (2. Also note that one could have written H as: H = ω aa+ − 1 2 · (2.14): (x + ip)φ(x) = 0.10. |n can be recast in terms of the variable x.20) To ﬁnd a quantum state it is not enough that it be a zero energy solution of the Schr¨dinger equation but it must also be normalizable (by normalizable o we include plane wave normalisability). This ﬁrst order equation always has a solution. .

Consider the Hamiltonian. In such a circumstance the wave function will be exponentially conﬁned to the edge of the universe and thus irrelevant for bulk physics. We will now note some basic facts concerning the validity of a perturbative expansion. m dependence the same type of rescaling used for the harmonic oscillator (2. 2 q (2. g) 1 n+2 (2. g) = The Hamiltonian becomes: n n 1 H = g 1− n+2 m− n+2 (p2 + q n ). There may be situations where one will want to study the physics on the boundary. 2m 2 (2. From now on we will only accept plane wave normalizable states and nothing “worse”.24) pq = 1 px . In such cases the “non-normalizable” states should be kept and may play an important role.25) p2 x + gf (m. Gauge Theory and Strings The above solution is not even plane wave normalizable and hence it is not a quantum state. g) (2.146 Unity from Duality: Gravity. g)x . Also recall that. Physically.23) (2. giving 2H = and so one may choose gf (m. f (m. mf 2 (m.7). To answer this one needs to ﬁnd out if one can remove the g. x that preserve the commutation relations.26) 1 mf (m. H= p2 1 q + gq n . Is it possible to deﬁne a new set of dimensionless canonical variables px . q] = [px . g)n xn . how do we motivate this restriction? After all. the simple harmonic oscillator is a useful approximation for small excitations above the minimum of a generic potential. x] and 1 H = h(m. g)2 · (2. such that: [pq .28) .22) One may wonder whether if one can make a perturbative expansion in small or large g or small or large m. 2 x We will make the following ansatz: q = f (m. the universe may well be ﬁnite. g) (p2 + xn ).27) (2.

(2. a+ } = ωB NB + B 2 2 · (2. F F [aF . we review the harmonic Fermionic Oscillator. HF = 1 1 ωF [a+ . The states are given by |0 . [a+ . a+ |0 . After reviewing the Bosonic harmonic oscillator. σ− = We then identify. They may not serve as perturbation parameters. {aF . Rabinovici: Supersymmetric Gauge Theories 147 The role of g and m is just to determine the overall energy scale. aF } = {a+ . 2 (2. This is essentially a manifestation of the Pauli exclusion principle. which we will return to later. F The spectrum of NF is 0.30) This demonstrates that a+ . F where ωF is the Fermionic oscillator frequency and. H] = −a+ . aF ] = ωF NF − F 2 2 · (2. F Hence the Hamiltonian becomes: H= For the Bosonic case. a+ } = 0.29) The Hamiltonian (that does not have a classical analogue) is taken to be. Hamiltonians of the form p2 + xn can’t be analyzed perturbatively.34) 1 1 (σ1 − iσ2 ) .) Thus. we have HB = 1 1 ωB {aB .32) This algebra can be realized using the Pauli matrices {σ i } as follows. H] = aF . σ+ = (σ1 + iσ2 ) · 2 2 (2.33) 1 This shifts the energy levels by + 2 ω. we choose HF to be. {aF . aF = σ− . Deﬁne. The commutation relations become anticommutation relations. 1. F (2.35) (2.) .S. (This does not apply to the special case of n = −2. a+ = σ+ . Further analysis shows that for Hamiltonians of the form p2 + xn + xm with m < n. aF are creation and annihilation operators. perturbation theory is valid.36) 1 ωF (σ3 + 1) + c. (2.37) .D. By analogy. Berman and E. F f F H = ωF (a+ aF + const.31) (2. a+ } = 1 .

1.48) . 0 ) = NF (Q|E. the vacuum energy may cancel between the Bosons and Fermions. Thus. i. Deﬁne 1 Q = √ (σ1 p + σ2 x) 2 1 {Q. 0 ) 1(Q|E.46) (2.44) hence. There is now a symmetry except for the E = 0 state. Q2 (2. H(Q|E. 0 ) = Q(Q|E. 0 ) = E(Q|E. Gauge Theory and Strings With this choice.e.47) (2. H= 1 ω (p2 + x2 12×2 + σ3 ). We see that the total Hamiltonian becomes: H = ωB NB + 1 2 + ωF NF − 1 2 · (2.45) Q anticommutes with σ3 and commutes with H.42) (2. 2 (2.43) (2. 0 ) E|E.41) = = = 1 2 2 2 σ p + σ2 x2 + σ1 σ2 px + σ2 σ1 xp 2 1 1 2 p + x2 12×2 + iσ3 (px − xp) 2 1 2 1 p + x2 12×2 + σ3 2 2 (2.40) (2. (2.148 Unity from Duality: Gravity. NF = 0. Consider the case when ω = ωF = ωB . 0 denotes a state with energy E and NF = 0. En = ω(nB + nF ).38) The Fermionic and Bosonic number operators both commute with the total Hamiltonian. We will label a state by its energy E and its Fermion number. Q} = Q2 2 (2. 0 . |E. here the vacuum energy precisely vanishes.39) H= Q commutes with H.

For any state |E. Note. Q− } = 0. (i.49) (2.D.52) (2. 1 √ E|E. It can be generalized to more complicated cases.51) H= (2. 0 149 (2.S. NF with E = 0 there is a state with equal energy and diﬀerent NF obtained by the application of Q. σ3 ] = 0 . The 2 generalized Hamiltonian is now: 1 2 1 (p + W (x)2 )1 + W (x)σ3 2 2 [H. 2 (2. Q+ } = 0 .56) In the previous case. Berman and E.57) (2. 2 1 1 Q2 = √ (σ2 p − σ1 x) = √ σ2 (p + iσ3 x) 2 2 and.53) (2. Q− } . (2. with the simple harmonic oscillator.e. The ground state |0. {Q1 . Qi ] = 0 . {Qi . H = Q2 + Q2 = 1 2 Thus E ≥ 0 and for E > 0 the spectrum is paired.55) (2. that there is a Q2 such that also H = Q2 . 0 however is annihilated by Q and so is not necessarily paired.50) which is valid only for E = 0. 1 ) = √ E|E. Q2 } = 0. We can then deﬁne: 1 Q± = √ (Q1 ± iQ2 ) 2 1 {Q+ .54) This was a “free” theory. W (x) = 1 x2 . Rabinovici: Supersymmetric Gauge Theories This implies Q(|E. We introduce new supercharges that depend on a potential W (x) as follows: Q1 Q2 = = 1 √ (σ1 p + σ2 W (x)) 2 1 √ (σ2 p − σ1 W (x)). a simple harmonic oscillator). [H. (2. 0 ) = Q(|E. Q1 = iσ3 Q2 .58) . σ3 } = 0. 2 {Q+ . {Q− .

59) φ0 (x) The zero energy solutions on the other hand. In general one may have any number of zero energy states in each NF sector. (2.64) These can be integrated to give the following solutions: φ1 (x) φ0 (x) = = φ1 (0) exp(−W (0)) exp(W (x)) φ0 (0) exp(W (0)) exp(−W (x)).60) Where we have used that Q is Hermitian. The full spectrum can’t be solved for general potential W (x). thus the mass of a free Boson equals that of a free Fermion. can be found (if they exist) for any W (x) by using the following: Hφ = Eφ .62) (2.67) (2. H = Q2 .68) . So Hφ = 0 ⇔ Qφ = 0. One can now solve the ﬁrst order equation to ﬁnd the zero energy states.61) (2. (2. Proof: 0 = 0|H|0 = 0|QQ|0 = (||Q|0 ||) ⇒ Q|0 = 0. In ﬁeld theory the energy gap between he ﬁrst excited state and the ground state is the particle mass. (2. The energy gap in the Bosonic sector (NF = 0) matches the energy gap in the Fermionic (NF = 1) sector. For W (x) = 1 x2 .66) (2.63) (2. NF = 0) = 1 and n(E = 0. NF = 1) = 0 2 where n denotes the number of states of given E and NF . Gauge Theory and Strings There is an analogue in ﬁeld theory. It is necessary to solve two 2nd order equations: p2 + (W )2 + W 0 0 p2 + (W )2 − W φ1 (x) φ0 (x) = 2E φ1 (x) · (2.150 Unity from Duality: Gravity. Consider the E = 0 case.65) (2. Qφ = 0 1 √ (σ1 p + σ2 W (x))φ = 0. 2 This leads to two independent ﬁrst order diﬀerential equations: − d + W (x) φ1 (x) dx d + W (x) φ0 (x) dx = = 0 0. n(E = 0.

Let |G. In the case of a continuous symmetry. When the symmetry is continuous. g and the symmetry is spontaneously broken if g|G. Is it a physically acceptable solution? If W (x) → ∞ as |x| → ∞ then φ0 (x) is a normalizable solution and one must set φ1 (x) = 0. Berman and E. The symmetry may be broken spontaneously.S.S. it has generators. as it occurs in the Zeeman eﬀect where one introduces a magnetic ﬁeld in the action that does not possess all the symmetries of the original action. For the N previously considered potential. If on the other hand W (x) → −∞ as |x| → ∞ then φ1 (x) is a normalizable solution and one must set φ0 to zero and obtain a Fermionic ground state. such a symmetry breaking may result in massless particles (called Goldstone Bosons). The consequences on the spectrum are as follows. = 0. There is no E = 0 solution if N is odd. Q|G. denote the energy ground state of the system. If the ground state is not invariant under S. This corresponds to a Bosonic ground state.D. then it is said that the symmetry is spontaneously broken. that is. S] = 0. = 0 iﬀ EG. For W (x) → const.1 Symmetry and symmetry breaking One can break a symmetry explicitly. if Q|0 = 0 then there is a supersymmetric ground state and spontaneous supersymmetry breaking does not occur. as |x| → ∞ then both solutions are possible.S. There is at N most one unpaired solution at E = 0.S. This occurs when the ground state does not posses the symmetry of the action. = 0 and Q|G. The fact that nE=0 depends on such global information is very useful.S.S. If supersymmetry is spontaneously broken then the ground state will not have zero energy. = 0. [H. It will appear again in a variety of contexts. = 0 iﬀ EG. one sees that if N is even then there is no spontaneous symmetry breaking since there are zero energy normalizable solutions and conversely if N is odd then there is spontaneous symmetry breaking. if S|G. Previously it was shown that any state with nonzero energy must be paired with at least one other state (its supersymmetric partner). This implies there will be at least two ground states with identical .S.S. Rabinovici: Supersymmetric Gauge Theories 151 A zero energy solution always exists. Q is the generator of a continuous symmetry (supersymmetry). For W (x) = n=0 an xn there is an E = 0 solution if N is even. Assume S is a symmetry of the Hamiltonian. When there is no spontaneous supersymmetry breaking Fermions and Bosons have the same mass.S = |G. 2. Spontaneous symmetry breaking plays an important role in the standard model and in a variety of statistical mechanical systems. The relevant question is whether Q|G.S. = 0. Returning to our model. W (x) = i ai xi . Only the leading asymptotic behavior matters.

70) For N even W (x) has at least one zero where W (x0 ) = 0. Classically therefore the ground state has zero energy. The classical limit is taken by → 0. energies and one must make a choice for the vacuum. For any choice of ground state. Gauge Theory and Strings No Susy E Susy Susy mB mF E=0 mB mF NF=0 NF=1 NF=0 NF=1 NF =0 NF =1 mB=mF >0 mB>mF =0 Fig. For example. taking the ground state to be the state with Fermion number zero implies that there is a Fermionic state with the same energy as the ground state thus the Fermion mass will now be zero while the Boson will remain with mass ∆. This the same as the exact result calculated previously by solving the Schroedinger equation! So how . 2 (2. 2. H→ 1 (W (x))2 ≥ 0. 1. ∂x2 2 (2.152 Unity from Duality: Gravity. After spontaneous symmetry breaking the Boson and Fermion masses are no longer equal. The zero mass Fermion is called the Goldstino. Spectra. The Goldstone Boson and the Goldstino have special low energy couplings. Hamiltonian is H= 1 2 − 2 The ∂2 1 + W (x)2 1 + σ3 W (x). It arises from the breaking of a Fermionic symmetry just as the Goldstone Boson arises from breaking a Bosonic symmetry.2 A nonrenormalisation theorem Consider now the system from a perturbative perspective.69) correc- First we will do a classical analysis and then we will examine the tions to the classical result. there is a zero energy gap between that ground state and the other possible ground state with Fermion number diﬀerent by one.

As classical results are easier to obtain. We will label the vacuum states |+ .S. 2 (2. 2 (2. N is odd and the supersymmetry is broken. (2. this does not work.D. 2 (2. In this case the classical result is not exact. For example. There are nonperturbative eﬀects that provide corrections. there is another source for correction. Rabinovici: Supersymmetric Gauge Theories 153 do classical results become exact? First let us note that in the Bosonic case.71) when x0 is a minimum of V (x) so that Ecl = 1 V (x0 ) we have a perturbed 2 energy given by: Epert = Ecl + 1 V (x0 ).S. and Ecl = 0. the energies of these two states are equal.75) This potential clearly has a vacuum degeneracy. . the ground state of the simple harmonic oscillator has an energy 1 ω above the classical ground state energy.73) Thus the Bosonic correction is canceled by the Fermionic correction and the classical result is exact. there is no ground state with E = 0. However. The result is true to all orders in perturbation theory.72) In a theory with Fermions though. (2. due to the presence of a Yukawa coupling. the potential is: |W (x) |2 = (x2 + a)2 . Perturbatively. Vcl = 1 (3x2 )2 2 G. More 2 generally for the Bosonic Hamiltonian: HB = 1 2 (p + V (x)). Consider the case W (x) = x3 . Calculating the perturbative energy of the ground state for a supersymmetric theory we have: SUSY Epert = 1 ((W (x0 ))2 + W (x0 ) − W (x0 )). |− .74) This also has E = 0 classical solutions and the leading term in W (x) is odd thus there are no exact quantum E = 0 states. In order to see how nonperturbative eﬀects become relevant we will actually consider the following: W (x) = 1 3 x + ax 3 a < 0. This generalizes under some circumstances to supersymmetric ﬁeld theories and is known by the name “nonrenormalisation theorem” [26]. this can be made into a very powerful tool. Berman and E. This is analogous to how the zero point energies of a free oscillator cancel between the Bosons and Fermions.

2. This is a tunneling phenomenon (essentially a nonperturbative eﬀect) that is associated with the presence of instantons.3 A two variable realization and ﬂat potentials Consider a two variable realization of supersymmetry. The supercharge is given by: 2 Q= α=1 + ψα −pα + i ∂W ∂xα (2. its energy gap is: ∆E = a exp − c λ (2.79) where W (x.76) (2.81) The Hamiltonian has the following structure: 0 0 0 H0 0 H11 H12 0 H= 0 H21 H22 0 · 0 0 0 H2 (2.78) a.80) 0 0 0 1 0 0 0 0 0 0 ψ2 = 0 0 0 −1 0 0 0 −1 · 0 0 0 0 0 0 (2.77) E(|A ) − E(|S ) > 0 and the ground state is the symmetric state. y) is a general function of the Bosonic variables x. Note that for a > 0 there are no zero energy solutions already classically.82) . Gauge Theory and Strings The energy eigenstates are: |S |A = = 1 √ (|+ + |− ) 2 √ (|+ − |− ) 2 (2. c are numerical coeﬃcients.154 Unity from Duality: Gravity. One can realize ψ1 and ψ2 by the following matrices: 0 1 0 0 0 0 0 0 ψ1 = (2. y.

84) (2.0 · (2. Classically the ground state is non zero which implies that the supersymmetry is broken. What about n = 1 sector? To ﬁnd these states we need to solve the following pseudo-analytic equations: ∂1 eW φ1.83) (2. the answer is no. In supersymmetric systems the Bosonic and Fermionic ﬂuctuations cancel and the ﬂat directions remain.86) (2. This is a “ﬂat direction”. 0) = (0.88) For any W the n = 0. 1) = (0. 2 sectors.87) H1 = 1 2 (−∆ + (∇W )2 − ∆W )1 + 2 ∂11 W ∂21 W ∂12 W ∂22 W · (2.1 = ∂2 eW φ0. Consider the potential W = x(y 2 + c) then Vcl = 1 [(y 2 + c)2 2 + 4x2 y 2 ]. (1.D. 2 For c > 0. can supersymmetry actually be restored? (There are cases where symmetries have been known to be restored quantum mechanically. 155 (2. 1 [−∆ + (∇W )2 − ∆W ] 2 1 [−∆ + (∇W )2 + ∆W ] 2 a+1 a+2 |00 F F a+1 |00 . 1). these cannot be solved in general.S.89) (2. Such ﬂat directions appear in abundance in supersymmetric models and lead in some contexts to supersymmetry breaking.90) Unfortunately. Flat directions imply the presence of many vacua. Vcl = 1 c2 at y = 0. 2 Note.85) H0 H2 = = (2. Flat directions in purely Bosonic models are lifted by quantum ﬂuctuations.) For n = 0.0 ∂1 e−W φ0. the potential is 2 the same for all values of x. Vcl = 1 c2 > 0 which leads to classical SUSY breaking [27]. We will now show that . What about quantum mechanical eﬀects. 2 sectors can be solved exactly for E = 0 just as before. 0) = With. a+2 |00 F F |00 .1 = −∂2 e−W φ1. neither ground state is normalisable. Berman and E. Rabinovici: Supersymmetric Gauge Theories The possible states are: (1.

d2 x(n · ∇W )φ2 > 0. (2. ∆ = Tr(−1)F = E nF =even (E) − nF =odd (E) .96) (this is the potential used to break supersymmetry spontaneously in scalar ﬁeld theories.92) φ|H|φ · φ|φ (2.156 Unity from Duality: Gravity.93) (2. thus.95) (2.94) where r is some ﬁnite real quantity. Gauge Theory and Strings from more general arguments that supersymmetry cannot be restored by quantum eﬀects. 1 E W (λ) = E W (0) + λ2 + λ 2 d2 x(n · ∇W )φ2 (2. This argument is rather general and also works for the 3 variable potential W = ayz + bx(y 2 + c) (2. nF =even (E). nF =odd (E) denote the number . φ2 ≥ 0 and d2 xφ2 = 1. if we now assume E W (0) = 0 then we obtain a contradiction because by taking λ to be small and negative we obtain E W (λ) < 0.97) The trace indicates a sum over all states in the Hilbert space. So we may write: 1 E W (λ) = E W (0) + λ2 + λr2 2 (2. [25] is deﬁned by.91) where φ denotes the unperturbed solution.) There is a short more elegant argument for non-restoration of supersymmetry that is based on an index theorem.) ∂x W = y 2 + x > 0 . (There is no λ dependence from the Yukawa term. The “Witten index”. EW = Consider perturbing the potential. E W (λ) ≥ 0. F is the fermion number of the state. W λ = W 0 + λxi ni where ni ni = 1 and W 0 = x(y 2 + c) is the unperturbed potential.

Thus for ∆ = 0 one needs more information. ∆! 1 = E (2. Rabinovici: Supersymmetric Gauge Theories 157 of solutions with Fermi number even/odd with energy E.99) (2. if ∆ = 0 for some W then there will be no spontaneous symmetry breaking under any perturbations in W .100) w w+! w δ ∆! !=0 w w+! w δ Q Q Q E=0 Q Q Q Q NF= 0 NF=1 NF=0 NF=1 NF=0 NF=1 NF=0 NF=1 Fig. Returning to the case described above. one obtains: 0 = ∆cl = ∆quantum = nq F =0 (E = 0) + nq F =2 (E = 0) − nq F =1 (E = 0). On the other hand if ∆ = 0 then we do not know whether SUSY may be broken or not since either nF =even = nF =odd = 0 or nF =even = nF =odd = 0. (2.S.D. ∆ is unchanged. Berman and E. In particular. N N N (2.101) One considers turning on quantum corrections as a perturbation. Perturbing the spectrum. Since the Bosons and fermions are paired at all energies greater than zero then ∆ = nF =even (E = 0) − nF =odd (E = 0).98) If ∆ is calculated for some potential W it will not change under perturbations of W (only En will change).102) We have already shown that nq F =0 (E = 0) = nq F =2 (E = 0) = 0 N N (2. (2. Since the index is invariant under perturbations. 2. We have calculated the classical Witten Index to be: ∆cl = 0.103) .

4 Geometric meaning of the Witten index Consider a supersymmetry sigma model with a D dimensional target space whose metric is gij [25].111) . 1.. Rijkl is the curvature.. S= dtgij (x) dxi dxi 1 ¯ + iψγ 0 Dt ψi + Rijkl ψ i ψ k ψ j ψ l . thus.. .. (2. 0 } Dim {|0. {ψi . 2. .. . ψj } = 0 .110) This is identical to the dimension of p-forms in D dimensions. ∗ ∗ ∗ {ψi .. . (2. ψj } = {ψi ..106) D D Q=i i = φ∗ pi .. 1 Dim = = = 1 D 1 2 D(D − 1) 1. there is no supersymmetry restoration. 0 Dim {|0..107) and pi = −Dxi .104) Dt denotes a covariant derivative. .. Gauge Theory and Strings which in turn implies nNF =1 (E = 0) = 0 . (2.... Q acts by adding a Fermion hence is a map Q: φp → φp+1 . D) = D! · (D − p)!p! (2. 1.109) = = Generically the Dimension of a state with p-Fermions is given by: Dim(p. (2. i Q∗ = −i i = φ∗ pi i (2. ψj } = gij .108) The Hilbert space may be graded according to the Fermion occupation number: φ0 φ1 φ2 · · φD = = = |0. .158 Unity from Duality: Gravity. 1. dt dt 4 (2. . 0 } Dim · · |1..105) (2. 1.

112) The dimension of the cohomology is denoted by the Betti number bp = dimH p . 2 (2. The supercharges are: Q1 Q2 = = 1 √ ((px + Ax )σ1 + (py + Ay )σ2 ) 2 1 √ ((px + Ax )σ2 − (py + Ay )σ1 ) 2 [Qi .116) (2. (2. Rabinovici: Supersymmetric Gauge Theories 159 From simple Fermi statistics we see that.D.S. Tr(−1)F = χ. (2. χ= p (−1)p bp . one solved the equation Qφp = 0. we have an isomorphism between p-forms with an exterior derivative. This explains why the Witten index is stable against non-singular perturbations. Q is nilpotent.114) The Euler characteristic is a topological invariant and is independent of geometrical perturbations of the manifold. y and one Fermion. The Euler characteristic is then given by the alternating sum of the Betti numbers. σ3 } = 0. As with any nilpotent operator one can examine its cohomology deﬁned by: Hp = {dφp = 0} · {φp = dφp−1 } (2.115) Two Bosons x. Hence. This is a less familiar realization of supersymmetry as the number of Bosonic and Fermionic oscillators diﬀer. H= 1 2 2 (px + Ax ) 1 + (py + Ay ) 1 + Bz σ3 . (2. d and the space of states with Fermion occupation number p and supercharge Q.5 Landau levels and SUSY QM Consider an electron moving in two dimensions in the presence of a perpendicular magnetic ﬁeld. It is particular to the quantum mechanical system. The Witten index is given a very physical realization in the following example. 2. Berman and E.118) . {Qi .117) (2. Thus one has the remarkable correspondence. hence the E = 0 states are in the cohomology of Q.113) Recall that to ﬁnd the E = 0 states. H] = 0 .

160 Unity from Duality: Gravity. The solution is: φ+1 = c n r exp(inθ). φ−1 = exp(a)g(x − iy) (2.125) The total magnetic ﬂux is a global quantity that is independent of local ﬂuctuations. Multiplying by σ1 gives: ((px + Ax )1 + iσ3 (py + Ay )) One can solve these equations as follows: φ1 (x. n(E=0) = |Φ|. (2. Take Q = Q1 . (2.121) There are an inﬁnite number of E = 0 states.120) φ1 φ−1 = 0 . 2 From translational invariance. take φ1 = 0 and 1 φ−1 = exp − y 2 B g(x − iy). Thus the number of E = 0 states is given by the total ﬂux number. In such a case. The total magnetic ﬂux is: Φ= 1 2π d2 rBz = − 1 2π 2π 0 dθr∂r a |r=∞ . Depend2 ing on the sign of B. In a constant magnetic ﬁeld. i. Now assume that Φ is ﬁnite because B is conﬁned in a solenoid.119) where f and g are arbitrary functions. either φ1 or φ−1 may be normalizable. (Φ is quantized). 0 (2. for example. y) = exp(−a)f (x + iy) . Take a constant negative B ﬁeld with a = − 4 r2 B.122) (2.e.123) Φ = 0 implies a(r = ∞) = 0 and one has only plane wave normalizable 1 states. (2. all n.124) For B constant Φ → ∞ and there are an inﬁnite number of allowed states. (This is true also for any E not just E = 0. taking for example. a convenient choice is g(x − iy) = exp(ik(x − iy)). ∂i Ai = 0 we can write Ai = ij ∂j a Bz = −∇2 a. To ﬁnd E = 0 states we again use Qφ = 0. normalisability requires that n is bounded by [Φ − 1]. The normalisability of these solutions depends on the function a. r|Φ| (2. Again we have shown how the number of zero energy states does not depend on the local details but only upon global information. If B > 0. a = − 1 y 2 B. .) Let us examine this from a topological point of view. Gauge Theory and Strings In Lorentz gauge. Q is just like the Dirac operator.

134) K acts as a special conformal transformation AK = 1 0 δ 1 .128) (2.6 Conformal quantum mechanics 161 Relevant material for this section may be found in [28–30].129) (2. D] = iH . (2. t = t · δt + 1 (2.133) α 0 1 α 0 . [H. t =t+δ (2. are given by: t = at + b 1 . [K. Rabinovici: Supersymmetric Gauge Theories 2. K] = 2iD. and not the Lagrangian L alone. Berman and E. D] = iK . S= dtL. This forms an SO(2. K = x2 4 2 and H is as given above.135) . The Casimir is given by: 1 3 g (HK + KH) − D2 = − · 2 4 16 (2.132) AT = D acts as dilation AD = 1 0 δ 1 .1) algebra where: 1 1 D = − (xp + px) . Recall the Hamiltonian: H= 1 2 (p + gx−2 ) 2 (2. H is part of an algebra: [H.S. x (t ) = x(t) ct + d ct + d .D.126) is special since g has a meaning.127) The meaning of D and K is perhaps clearer in the Lagrangian formalism: L= 1 2 g x − 2 ˙ 2 x . t = α2 t (2.131) A= H acts as translation a c b d (2.130) Symmetries of the action S. detA = ad − bc = 1 (2.

α− < 0. (2. Take the ansatz (2.126 is the open set (0. How do we deal with a system in the absence of a ground state? . α+ > 0. ∞). the spectrum is therefore continuous and bounded from below. is not normalizable either since the function diverges at the origin (a result of the scale symmetry).136) 4 We will now attempt to ﬁnd the zero energy state. 3. φ(x) = xα : H= This implies g = −α(α − 1) solving this equation gives − d2 g + 2 2 dx x xα = 0. there is no normalizable E = 0 solution (not even plane wave normalizable)! H ? ∞ x Fig.139) α=− ± 2 2 This gives two independent solutions and by completeness all the solutions. Thus. The absence of a normalisable ground state for this potential.137) (2.162 Unity from Duality: Gravity. Gauge Theory and Strings The spectrum of the Hamiltonian 2. does not lead to a normalizable solution since the function diverges at inﬁnity.138) √ 1 + 4g 1 · (2. The wave functions are given by: √ √ ψE (x) = xJ√g+ 1 ( 2Ex) E = 0. Most of the analysis in ﬁeld theory proceeds by identifying a ground state and the ﬂuctuations around it.

r << M . The Hamiltonian for this in falling particle in this limit is given by our old friend: g 4l(l + 1) p2 · (2. The blackhole metric is given by: ds2 = − 1 + M r −2 dt2 + 1 + M r 2 (dr2 + r2 dΩ2 ) . we have g > 0 and there is no ground state.S. Take for example.145) g = 8M 2 (m − q) + H= r + 2 . The change in evolution operator is now associated with a change of time coordinate.144) This produces an AdS2 × S 2 geometry.D.143) Now consider the near Horizon limit i. We also wish to keep ﬁxed M 2 (m − q) as we scale M . G= 1 2 1 K + aH a ≡R (2. Surprisingly this question arises in the context of black hole physics. M and charge. ds2 = − r M 2 dt2 + M r 2 dr2 + M 2 dΩ2 . Another possibility is to deﬁne a new evolution operator that does have a ground state. Rabinovici: Supersymmetric Gauge Theories 163 One possibility is to accept this as a fact of life. Consider a particle of mass m and charge q falling into a charged black hole. The black hole is BPS meaning that its mass. 2m 2r M For l = 0.142) One must interpret what this means physically. which we will take by M → ∞ keeping r ﬁxed. 2 (2.e.141) a has dimension of length. At = r · M (2. This means we must scale (m − q) → 0. (2. Any choice explicitly breaks Lorentz and scale invariance. G = uH + vD + wK. That is the particle itself becomes BPS in the limit. Q are related by M = Q. Berman and E. The eigenvalues of R are rn = r0 + n . . r0 = 1 2 1+ g+ 1 4 · (2. This is associated with the coordinate singularity at the Horizon.140) This operator has a ground state if v − 4uw < 0. One for which the world line of a static particle passes through the black hole horizon instead of remaining in the exterior of the space time. Perhaps it is possible to view this as similar to cosmological models that also lack a ground state such those with Quintessence.

153) and we may solve for the full spectrum: √ 1 ψE (x) = x 2 J√ν (x 2E) . Next the supersymmetric version of conformal quantum mechanics is examined to see if supersymmetry is indeed broken. Gauge Theory and Strings 2. neither solution is normalizable. Q+ = ψ ip + dw dx · (2. S + = ψx. The zero energy solutions are exp(±W (x)) = x±g . This breaking has a diﬀerent ﬂavor from that which was discussed earlier. W (x) = yielding a Hamiltonian: H= 1 2 p2 + dw dx 2 1 g log x2 . + (2. (2. {Q .146) 1 − σ3 d2 W dx2 · (2. {Q. Q+ } {S. 2 (2. H factorizes: 2H = p2 + g(g+1) x2 + = 2H .7 Superconformal quantum mechanics The bosonic conformal mechanical system had no ground state. E=0. For this the superpotential is chosen to be.151) (2. S } A realization is: B = σ3 . S + } = g − B + 2iD = 2K . Again 4 4 the spectrum is continuous though there is no normalizable zero energy . {Q. S} = g − B − 2iD.152) 0 p + 2 g(g−1) x2 0 (2. S = ψ + x .149) (2. The absence of a E = 0 ground state in the supersymmetric context leads to the breaking of supersymmetry.148) One now has a larger algebra.164 Unity from Duality: Gravity. the superconformal algebra.147) Representing ψ by 1 σ− and ψ ∗ by 1 σ+ gives the supercharges: 2 2 Q = ψ + −ip + dW dx .150) (2.154) where ν = g(g − 1) + 1 for NF = 0 and ν = g(g + 1) + 1 for NF = 1.

its N = 1 version is: µ ¯˙ ¯˙ ¯˙ {Qα . N } = 4 L± = 1 1 R + B − g ≡ T1 2 2 1 1 R + B + g ≡ T2 2 2 1 L− {M + . gB = gsusy (gsusy − 1). N + } = 4 1 {M. Berman and E. Qα } = 2σαβ Pµ .162) + + M + = Q+ − S + N + (2. Again we must interpret the absence of a normalizable ground state.161) (2. This is the supersymmetric analogue of the situation already discussed. By inspection the operator (2. (2. (2. One can again deﬁne a new operator with a normalizable ground state. N + } = L+ 4 1 − (H − K ∓ 2iD) 2 T2 |0 = 0 .155) (2. Qβ } = 0. One needs to deﬁne a whole new set of operators: M =Q−S N =Q +S which produces the algebra: 1 {M. {Qα . T2 . Qβ } = {Qα . Qα ] = [Pµ .1) (3. 3 Review of supersymmetric models 3.160) (2. (2. The possible particle content of supersymmetric (SUSY) theories is determined by the SUSY algebra.158) =Q+S + T1 . ˙ (3.D.32]. H have a doublet spectra.1 Kinematics For a more detailed discussion of the material presented in this section see [3].141) can be used provided one makes the following identiﬁcations: NF NF = 1 = 0 gB = gsusy (gsusy + 1) .157) (2.2) ¯˙ [Pµ .156) Thus the spectrum diﬀers between the NF = 1 and NF = 0 sectors and so supersymmetry would be broken.S.163) A physical context arises when one considers a supersymmetric particle falling into a black hole [31. Pν ] = 0. .159) (2. Qα ] = [Pµ . “Ground states” are given by: T1 |0 = 0 . M + } = 4 1 {N. H|0 = 0 . Rabinovici: Supersymmetric Gauge Theories 165 state.

˙ ˙ ˙ α ˙ (3. 1. there is again the chiral multiplet which is the same as the massless multiplet but with now massive ﬁelds. 2. . 0. α 2 ˙ transform under the ( 1 . in N = 1. they commute with all other charges (they are antisymmetric in ij). Aµ ) (4.4) In the above table. then the associated component ﬁelds. We will discuss in this section the d = 4 realisations with. Those with α or β indices transform under the ˙ (0.8) . 3 the space-time indices. ν = 0. 2 2 (φ. or β 2 Consider ﬁrst the massless representations of N = 1 supersymmetry. 1 ) representation of the Lorentz group. With N = 2 supersymmetry.e. 2). and ﬁnally are the number of physical degrees of freedom carried by the Bosons and Fermions. ψ1 ) + (φ2 . The massive vector multiplet becomes: −1 1 −2 1 −2 0 0 1 2 11 2 (h. 0 . (3. In four dimensions we have two component Weyl Fermions. The simplest is the the chiral multiplet: 1 1 − . 0. 1 2 (4. 2) (3. φ denotes a complex scalar and ψ a Weyl Fermion. 0) representation. Aµ ) (4. 4 by: µ ¯ {Qi . ψ) + (λα .5) λ is a Weyl Fermion and Aµ is a vector ﬁeld. ﬁrst are written the helicities. ψ2 ) − 1 .7) 0 2 For massive multiplets.6) and a massless hypermultiplet which is given by: 0 1 − 2 0 1 2 (φ1 . Qj α } = 2δ ij σαβ Pµ + δαβ Uij + (γ5 )αβ Vij . λα . 4). Gauge Theory and Strings This algebra can be generalized to include a higher number of supersymmetries N = 2. (3.3) U and V are the central charges i. 1 2 2 (λα . (3. 4). and those with dotted indices.166 Unity from Duality: Gravity. ψ) (2. 4) (3. They are associated with BPS states such as monopoles. Aµ ) (2. there is a massless vector multiplet: −1 1 −2 1 −2 0 0 1 2 11 2 (φ. − . The vector multiplet has: 1 1 −1. ψα . µ.

1 (λa . a = 1.. This can occur dynamically. Rabinovici: Supersymmetric Gauge Theories 167 Where h is a real scalar ﬁeld.9) 1 .10) The following identity will be of use: ¯ d2 θd2 θL = ∂2L d2 θ ¯ ¯ · ∂ θ1 θ2 (3. With N = 4 supersymmetry.6. 3. 8) (3. −2 0 2 1 −2 0 1 2 0 where I = 1. ∂θα dθα = 0. The massive vector multiplet has a diﬀerent ﬁeld content than the massless vector multiplet because a massive vector ﬁeld has an additional physical degree of freedom.12) It will also be useful to deﬁne a supercovariant derivative: Dα = ∂ µ ¯˙ + iσαα θα ∂µ . Berman and E. θα . 1 . the massless vector multiplet is: 0 1 −2 0 1 2 1 1 −2 0 2 −1 . φI .D. ˙ ∂θα ∂ µ ¯˙ Dα = − ¯ − iθα σαα ∂µ . Constructing a Lagrangian out of superﬁelds provides a useful way to construct explicitly supersymmetric Lagrangians. Aµ ) (8. One sees that the massive vector multiplet is composed out of a massless chiral plus massless vector multiplet.4. ˙ ∂ θα ˙ (3.13) A superﬁeld Φ is called “chiral” if: (3.. θα .11) The supercharges can be realized in superspace by generators of supertranslations: Qα = ∂ µ ¯˙ − iσαα θα ∂µ .2 Superspace and chiral ﬁelds ¯˙ Spacetime can be extended to include Grassmann spinor coordinates. (3. (3. ˙ ∂θα ∂ µ ¯˙ Qα = − ¯ + iθα σαα ∂µ . Recall the integration formulas for Grassmann variables: dθα θα = ∂ =1. Superﬁelds are then functions of this superspace.14) . massive vector multiplets may appear by a supersymmetric analogue of the Higgs mechanism. ˙ ∂ θα ˙ ¯˙ Dα Φ = 0. .S.

√ Φ(y) = A(y) + 2θψ(y) + θθF (y). These are called the component ﬁelds of the superﬁeld. ¯¯ (3.19) is a invariant under supersymmetric transformations (up to a total derivative).20) Φ(x) = A(x) + 2θψ(x) + θθF (x) A(x) will be associate with a complex Boson. ¯ y µ = xµ + iθσ µ θ which produces the following expansion for a chiral ﬁeld.22) ¯¯ After expanding and extracting the θθθθ term is (up to total derivatives): i ¯ Fi∗ Ff − |∂µ A|2 + ∂µ ψ¯ µ ψ. The product of two chiral ﬁelds also produces a chiral ﬁeld. The kinetic terms are described by: ¯¯ ¯ d2 θd2 θΦi Φj = Φi Φj |θθθθ .16) The Taylor expansion terminates because of the anticommuting property of the Grassmann coordinates. (3. Therefore.17) 4 i µ¯ − √ θθ∂µ ψ(x)σ θ + θθF (x). This is used to provide a potential for the chiral ﬁeld.23) . ψ(x) will be associated with a Weyl Fermion and F (x) acts as an auxiliary ﬁeld that contributes no physical degrees of freedom.18) 2 The key point is that L= d2 θΦ(x) (3. After the integration some terms will disappear from the expansion of Φ(x) leaving only: √ (3. Gauge Theory and Strings One introduces the variable. As a function of x it may be written as follows: Φ(x) = √ 1 ¯¯ ¯ A(x) + iθσ µ θ∂µ A(x) + θθθ θ A(x) + 2θψ(x) (3.21) is a supersymmetry invariant.15) (3.168 Unity from Duality: Gravity. any polynomial. W (Φ) can be used to construct a supersymmetry invariant as L= d2 θW (Φ) = FW (Φ) (3. (3. σ 2 (3.

At the ground state this must vanish i. Φ) is called the K¨hler potential. these are known as the F terms. 3. this metric is related to the K¨hler a potential by: gij = ∂2K · ¯ ∂ Φi ∂Φj (3.S. (3.28) ¯ K(Φ. K is taken to be given by K = ΦΦ which produces the ∗i µ j −δij ∂µ A ∂ A kinetic terms for the scalars. in fact it is true to all orders in perturbation theory as there exists a non renormalization theorem for the eﬀective potential. . VF ≥ 0). with the target space metric given by equation (3. This in turn implies that Fi = 0 for the ground state. Fi∗ .c.3 K¨hler potentials a To describe the supersymmetric Lagrangian for scalar ﬁelds spanning a more complicated manifold it is convenient to introduce the following supersymmetry invariant: ¯ d4 θK(Φ. For the case of a sigma model with a target space whose metric is gij . The equation of motion ∗ for Fk is as follows: ∗ Fk + λ∗ + mij A∗ + gijk A∗ A∗ .27) where the last term is a potential for A.26) This gives: L 1 1 ∗ ∗ ¯¯ = i∂ ψi σ ψi + A∗ Ai − mij ψi ψj − m∗ ψi ψj i 2 2 ij ∗ ¯ ¯ −gijk ψi ψj Ak − gijk ψi ψj A∗ − Fi∗ Fi k ∗ (3. A) = 0. Φ). Although this is a classical analysis so far. A . (Note. VF (A∗ . k i i j (3. (3. VF (A∗ .e. Rabinovici: Supersymmetric Gauge Theories One thus composes the following Lagrangian: L = = 1 1 ¯ Φi Φi |θθθθ + λi Φi + mij Φi Φj + gijk Φi Φj Φk ¯¯ 2 3 169 (3. Berman and E. A). For the Φ ¯ usual kinetic terms.29) The above supersymmetry invariant (3.28) which previously gave the usual kinetic terms in the action.24) θθ 1 ¯¯ i∂ ψi σ ψi + A∗ Ai + Fi∗ Fi + λi Fi + mij Ai Fj − ψi ψj i 2 +gijk (Ai Aj Fk − ψi ψj Ak ) + h. produces for general K the action of a supersymmetric sigma model.D.25) One must now eliminate the auxiliary ﬁelds Fi . One may add any function of Φ or a ¯ to the integrand since these terms will vanish after integration.29).

5 Global symmetries R-symmetry is a global U (1) symmetry that does not commute with the supersymmetry. The issues we are interested in are whether supersymmetry is spontaneously broken. they have a solution. the R-charges of the component ﬁelds may be read oﬀ as follows: R(lowest component of Φ) = R(A) ≡ n . This implies the classical result is exact. There is no supersymmetry breaking classically. 2 3 To show that the form of W remains invariant. For a superﬁeld Φ with R-character n. The claim is that the form of the eﬀective interacting superpotential Weﬀ will be the same as the classical superpotential Wcl . the following ingredients are used: the holomorphic dependence of W (Φ.30) are n (complex) equations with n (complex) unknowns. Gauge Theory and Strings 3. x). Generically. that for λ = 0.31) This has a solution. λ∗ .32) and hence there will be a choice of vacuum: either A = or A = 0. V = A∗ A|(gA − m)|2 m g (3. (3. that is W is independent of m∗ . and the global symmetries present in the theory [33]. Note. (3.170 Unity from Duality: Gravity. (3. R(ψ) = n − 1 . The action of the R-symmetry on a superﬁeld Φ with R-character n as follows. Φ∗ . 3.4 F-terms In this section we examine the vanishing of the potential generated by the F terms.33) W = mΦ2 + λΦ3 .36) . where F1 = −λ − mA + gA2 . x) = exp(2inα)Φ(exp(−iαθ). Take Wcl to have the form: 1 1 (3. The Witten index Tr(−1)F = 2. λ). RΦ(θ. R(F ) = n − 2.35) Since the R-charge does not commute with the supersymmetry. is there a non renormalization theorem. and are there other internal symmetries broken. VF = 0 ⇔ Fi = 0 ∀i .34) (3. x) = ¯ ¯ RΦ(θ. Take the example of the one component WZ model. x) ¯ exp(−2inα)Φ(exp(iαθ). (3. m. the component ﬁelds of the chiral ﬁeld have diﬀerent R-charges.

38) to have zero R charge requires that R(W ) = 2. barred variables having opposite R charge. R(ψ) = 0. to have vanishing R-charge requires R(Φ) = R(A) = 1 . This is how one treats for example the mass parameters of Fermions in the standard model. This can be viewed as simply a book keeping device or more physically one can view the coupling as the vacuum expectation value of some ﬁeld. and m. To restore R-invariance requires λ is assigned an R-charge of −1.40).40) (3. There is also one other global U (1) symmetry that commutes with the supersymmetry. Rabinovici: Supersymmetric Gauge Theories The R-charge of the Grassmann variables is given by: R(θα ) = 1 . All component ﬁelds are charged the same with respect to this U (1) symmetry. Demanding that the terms in the action maintain this symmetry requires an assignment of U (1) charges to λ. ¯ an R invariant.42) λ 3 Φ 3 (3. Berman and E.D. . W = mψψ + m2 |A|2 . The expectation value inherits the quantum numbers of the ﬁeld. R(dθα ) = −1 171 (3.37) ¯ with.41) (3. For the mass term W = 1 2 2 mΦ . (3. Adding the cubic term: W3 = produces V = |λ|2 |A|4 + λAψψ. (3.43) These symmetries are next used to prove the nonrenormalisation theorem.S. (θ d2 θW (3. The charges are summarized in the following table: Φ m λ W U (1) U (1)R 1 1 −2 0 −3 −1 0 2.) For the potential term. The kinetic term ΦΦ is ¯θθθ is an invariant.39) This term is not R-invariant with the R-charges given by (3.

36] is the ﬁeld theory analogue of the potential given by equation (2. There can therefore be no corrections to the form of Weﬀ [33]. mΦ2 g λΦ m = λn Φn+2 1 1 mΦ2 + Φ3 + an · 2 3 mn−1 n=2 ∞ λΦ m · (3. λΦ3 ) . The presence in the collaboration of a student who paid little respect to the general counter arguements made the discovery possible. for the function to be free of singularities implies that n < 1.45) (3.47) . One could be surprised that the breaking was ﬁrst achieved in the gauge systems. In order to break supersymmetry a minimum of three chiral ﬁelds are needed: LPotential = λΦ0 + mΦ1 Φ2 + gΦ0 Φ1 Φ1 + h.7 Supersymmetry breaking Let us examine now how supersymmetry may be spontaneously broken. λ) = g1 (mΦ2 . if he classical superpotential has unbroken supersymmetry so does the full theory. (3. The Fayet-O’Raifeartaigh potential [35. (3. Consider the limit m → 0.46) As λ → 0 one must recover the classical potential.c. This was done by Fayet and Illiopoulos.6 The eﬀective potential That the eﬀective potential be invariant under the global U (1) implies that Weﬀ (Φ.96) for supersymmetric quantum mechanics. the Kahler potential is renormalized. In particular. m. Fayet went on to discover the breaking mechanism also in supersymmetric scalar theories as did O’Raighfeartaigh. This requires that n ≥ 0. It turns out that a short time after supersymmetry was introduced arguments were published which claimed to prove that supersymmetry cannot be broken spontaneouly. Gauge Theory and Strings 3.) 3. Supersymmetry resisted breaking attmepts for both theories of scalars and gauge theories. The following anecdote may be of some pedagogical value [34].172 Unity from Duality: Gravity.44) combining this with the invariance under the U (1) R-symmetry implies Weﬀ = mΦ2 g Consider expanding in |λ| << 1. (On the other hand.

and C are gauge artifacts and can be gauged away.53) B. It obeys a ¯ ¯ ¯ ¯¯ B + θχθ χ + +θ2 C + θ2 C − θσ µ θAµ 1 ¯ ¯ ¯ 1¯ +iθ2 θ(λ + σ µ ∂µ χ) − iθ2 θ(λ − σ µ ∂µ χ) ¯ 2 2 1 ¯ + θ2 θ2 (D2 + ∂ 2 B) 2 (3.48) (3.51) (3.49) (3. Doing so one discovers that in the ground state A1 = A2 = 0 and A0 is arbitrary.55) . m2 ± 2gλ.54) B. 3. Berman and E.96).8 Supersymmetric gauge theories A vector superﬁeld contains spin 1 and spin ¯ reality condition V = V . To ﬁnd the ground state one must write out the full Lagrangian including the kinetic terms in component ﬁelds and then minimize. from V as follows 1¯¯ ¯˙ Wα = − DDDα V . There will be a local U (1) symmetry with gauge parameter. The arbitrariness of A0 is a ﬂat direction in the potential. and the Fermions have masses: 0. Dβ Wα = 0. 2m. It is possible to construct a chiral superﬁeld.D. The symmetry is actually U (1)C as opposed to the usual U (1)R because although the vector ﬁeld transforms with a real gauge parameter. the other ﬁelds transform with gauge parameters that depend on the imaginary part of Λ. Wα . 4 (3. χ. V = 1 2 component ﬁelds. like in the quantum mechanical example (2.S.50) These cannot be consistently solved so there cannot be a zero energy ground state and supersymmetry must be spontaneously broken. Λ an arbitrary chiral ﬁeld: ¯ V → V + i(Λ − Λ) (3. The zero mass Fermion is the Goldstino.52) (3. Aµ are real and C is complex. m2 m2 . 173 (3. Rabinovici: Supersymmetric Gauge Theories Minimizing the potential leads to the following equations: 0 = 0 = 0 = λ + gΦ1 Φ1 mΦ2 + 2gΦ0 Φ1 mΦ1 . 0. We turn next to supersymmetric theories that are gauge invariant. D. Computing the masses by examining the quadratic terms for component ﬁelds gives the following spectrum: the six real scalars have masses: 0.

λ is the spin 1 ﬁeld and D 2 is an auxiliary scalar ﬁeld.57) ¯ α 2 Where Aµ is the vector ﬁeld.60) Expanding this in component ﬁelds produces. Gauge Theory and Strings One may choose a gauge (called the Wess Zumino gauge) in which B.56) 2 i β ¯ Wα (y) = −iλα + δα D − (σ µ σ ν )β Fµν θβ + (σ µ ∂µ λ)α θ2 . W is gauge invariant.58) Note. D → D. Under the symmetry (3. (3.c.54). (3. C and χ = 0 and then expand in terms of component ﬁelds. L= 1 i θ −1 ¯ Fµν F µν + 2 D2 − 2 λσDλ + (∗F )µν Fµν .64) ¯ + ¯σµ ∂ν λ) − (µ ↔ ν) δ Fµν = i( σµ ∂ν λ (3. λ → λ .67) One may also add to the action a term linear in the vector ﬁeld V . the component ﬁelds transform under a now U (1)R symmetry as: Aµ → Aµ − i∂µ (B − B ∗ ) . The theta term couples to the instanton number density (this vanishes for abelian ﬁelds in a non-compact space).c. (3.63) √ µ δ F = i 2¯σ ∂µ ψ ¯ (3. The following supersymmetric gauge invariant Lagrangian is then constructed: L= d2 θ −iτ 16π W α Wα + h.65) ¯ δλ δD = = i D + σ µν Fµν ¯ ¯σ ∂µ λ − σ ∂µ λ . known as a Fayet-Iliopoulos term [37]: 2K ¯ d2 θd2 θV = KD = dθα Wα + h.62) δA = √ µ √ δ ψ = i 2σ ¯∂µ A + 2 F (3. Fµν its ﬁeld strength. 1 ¯ ¯ ¯¯ ¯ V (y) = −θσ µ θAµ + iθ2 θλ − iθ2 θλ + θ2 θ2 D (3. ¯ µ µ (3. (3.68) . τ= θ 4π +i 2 · 2π g (3.174 Unity from Duality: Gravity. 4g 2 2g g 32π 2 (3.61) D is clearly a non propagating ﬁeld. A monopole in the presence of such a coupling will get electric charge through the Witten eﬀect.66) (3.59) where the coupling constant τ is now complex. The supersymmetries acting on the component ﬁelds are. (up to total derivatives): √ 2 ψ (3.

69) in components produces: L = iq iq ¯σ Aµ φ|2 − iφ¯ ∂µ + qAµ ψ 2 2 iq 1 ¯ ¯¯ ¯ − √ φλψ − φλψ + qDφφ. (3. the D and F ﬁelds. W (Φ) for the matter. The U (1) gauge ﬁelds couple to charged chiral matter through the following term L= i ¯¯ d2 θd2 θΦi exp(qi V )Φi .77) . Φi → exp(−iqi Λ)Φi . Berman and E. Φ2 with opposite U (1) charges and Lagrangian given by: L = 1 ¯ (W α Wα + h. that is if the U (1) is not spontaneously broken then supersymmetry is broken if and only if KF.c. (3. V = i | ˜ ∂ W 2 q2 (2K + | + ∂φi 4 |φi |2 2 . (3. Rabinovici: Supersymmetric Gauge Theories 175 Its role will be discussed later. L= d2 θ W α Wα + ¯¯ ˜ d2 θΦi exp(qi V )Φi + W (Φ) (3. In order that the theory is free from chiral anomalies one requires: qi = 3 qi = 0. (3. 2 2 F ∗ F − |∂µ φ + (3. Some remarks about this potential: Generically the F -terms should vanish since there are n equations for n unknowns.61) for the vector ﬁeld and a potential. = 0. Consider ﬁelds Φ1 . Both these terms need to vanish for unbroken supersymmetry. + 2KV.72) (3.75) The ﬁrst term is the F -term and the second is the D-term. When K = 0 and the F -terms have a solution then so will the D-term and there will be no supersymmetry breaking.D.69) Under the gauge transformation ¯ V → V + i(Λ − Λ) . These ideas are demonstrated by the following example.74) this produces the following potential.73) There are two auxiliary ﬁelds. If φi = 0. ˜ Adding the kinetic term (3.c.S.71) Writing out the term (3.76) (3.70) Since there are chiral Fermions there is the the possibility for chiral anomalies. gives the total Lagrangian.I.) + Φ1 exp(eV )Φ1 4 ¯ +Φ2 exp(−eV )Φ2 + mΦ1 Φ2 + h. (3.

m2 < 1 eK. (For K = 0 though the symmetry is restored. Two Fermions have non-zero mass and there is one massless Fermion. 1 2 1 2 2 2 2 8 (3. The vector ﬁeld has zero mass. at the minimum. m2 > 1 2 eK.176 Unity from Duality: Gravity. Note that a trace of the supersymmetry remains as TrMB = 2 TrMF even after the breaking of supersymmetry. A1 = 0. the Goldstino.85) 1 1 2 2 ( e v ) 2 2 1 ˜ ¯¯ ˜ ¯ m2 + e2 v 2 ψ ψ + ψ ψ + 0 × λλ.) The massless Fermion (the Photino) is now a 2 Goldstino. Since the vector ﬁeld remains massless then the U (1) symmetry remains unbroken. Gauge Theory and Strings This leads to the potential: V = where e D + K + (A∗ A1 − A∗ A2 ) = 0 2 2 1 F1 + mA2 ∗ = 0 F2 + mA∗ = 0. The Higgs ﬁeld. 2 The ﬁrst term implies that supersymmetry is broken for m > 0.78) (3. √2 (A + A∗ ))2 has the same mass as the photon. consider the case. m2 + 1 eK and 2 m2 − 1 eK.81) The scalars have mass.79) (3. The potential expanded around this minimum ˜ becomes. The photon is now massive. m2 = 1 e2 v 2 implying that the U (1) symmetry is broken γ 2 1 ˜ ˜ as well. For K = 0. Two Fermions have mass m 2 and one Fermion is massless.84) (3.83) (3. A2 = 2 v where v 2 ≡ 4 2 e2 . . supersymmetry is broken as the Bosons and Fermions have diﬀerent masses.82) Consider the case. Next.80) (3. 1 Leading to the following expression for the potential: 1 1 1 1 2 V = K 2 + m2 + eK A∗ A1 + m2 − eK A∗ A2 + e2 (A∗ A1 −A∗ A2 ) . 1 2 ∗ ∗ D + F1 F1 + F2 F2 2 (3. with A ≡ A1 and A ≡ A2 − v: V = 2m2 1 (eK − m2 ) + e2 2 +2m2 A∗ A = + 1 2 2 e v Aµ Aµ 2 1 ˜ ˜ √ A + A∗ 2 2 1 eK−m2 (3.

90) .91) (3. Next consider a more generic example where there is U (1) breaking but no supersymmetry breaking. Da .dim(group) and Da = µ ¯ a i φi TR(Φi ) Φi . the presence of a ﬂat direction: a+ → eα a+ .. a− → e−αa− (3.D. Rabinovici: Supersymmetric Gauge Theories 177 In the above example there is both supersymmetry breaking and U (1) symmetry breaking except when m = 0 in which case the supersymmetry remains unbroken.89) leaves a− a+ ﬁxed and the vacuum equations are still satisﬁed. a = − which leads to no U (1) breaking and a+ a− = − 1 8 λ− mµ µ .c.): a+ a− = 0 . Only if there is a U (1) factor will the Fayet-Iliopoulos term be non-vanishing. Note. Inverting this argument leads to the conclusion that supersymmetric breaking in nature cannot be generic since we do not observe such a particle. Φ is neutral under the U (1) while Φ+ has charge +1 and Φ− has charge −1. Aa .87) which breaks the U (1) symmetry. The potential is given by: L= 1 mΦ2 + µΦ+ Φ− + λΦΦ+ Φ− + h. 2 (3. Φ− . λa . Typical generic supersymmetry breaking leads to the breaking of R-symmetry.86) There are two branches of solutions to the vacuum equations (a denotes the vacuum expectation value of A etc. Take the simple example of an abelian theory with no Fayet-Iliopoulos terms and no F -terms. We wish to examine the properties of potentials with ﬂat directions. Four real ﬁelds obeying the constraint: Φ+ = exp(iα)Φ− (3.88) λ m (3. One can also consider non-abelian groups.S. The vanishing of the D-term implies: D = |Φ+ |2 − |Φ− |2 = 0. Berman and E. the index a is the group index. This gives a U (1) theory with oppositely charged ﬁelds Φ+ .a = − g g (3. a = 1. Since there is a broken global symmetry this would lead to the presence of Goldstone Boson corresponding to the broken U (1)R . So far we have only dealt with U (1) vector ﬁelds. The ﬁelds are in an adjoint representation of the group.

The moduli space is the space of all ﬁelds Φi . The metric will then be: ds2 = ¯ 1 dM dM √ · ¯ 2 MM (3. Using Φ− = Φ+ . A convenient gauge ﬁxing is.93) (3. When K = 0.96) (3. This is equivalent to the space of all constant Φ ﬁelds modulo the complex GC gauge transformations. Φ + = Φ− . There is therefore a ﬁxed point at Φ+ = 0. When K = 0 but in the presence of an F -term. modulo GR gauge transformations.95) (3. . Note. However there is still the U (1) gauge symmetry that one must mod out by. (3. GC is spanned by a basis of the independent gauge singlets (such as m = φ+ φ− ). The moduli space is then given by the space of ﬁelds that solve the F -term equations modulo complex GC gauge transformations. then provided there are solutions to the F -terms equations. Thus the singularity in moduli space is a signature of a particle becoming massless. the D-terms automatically vanish. (as is the case above) the moduli space is determined in the following way. What is the physical interpretation of this singularity in moduli space? Let us study the space in terms of gauge invariant variables: M ≡ Φ+ Φ− . the Kahler potential becomes: ¯ ¯ ¯ ¯ K = Φ+ Φ+ + Φ− Φ− = 2Φ+ Φ+ = +2 M M . the classical moduli space is given by the C orbifold. such that Da = 0. which will be a singularity. By expanding the D-term. ones sees that M is a parameter that determines the mass of the matter ﬁelds. There remains a Z2 residual gauge symmetry however: Φ+ → −Φ+ . Φi Moreover the moduli space.94) There is a singularity in moduli space when M = 0.92) After moding out by this action. there is no nonrenormalisation theorem for the Kahler potential. Z2 . This leaves a two dimensional moduli space.178 Unity from Duality: Gravity. In the absence of an F -term then there is always a solution to the D-term equations. Gauge Theory and Strings implies a three dimensional conﬁguration space.

2π (4. this is the dual of electric condensation. (This is called the Higgs phase.D. (This is called the conﬁnement phase. T = N g · 2π For a given theory there is a lattice of electric and magnetically charged operators where the electric charge is denoted by n and the magnetic charge by m.S. T N (4.e.) For intermediate values of T it is possible that there is neither screening of charges nor conﬁnement. magnetic charges are conﬁned. electric charges are conﬁned and the log of the Wilson loop is proportional to the area. This leads to what is called oblique conﬁnement with the charges commensurate with (n0 .2) Keeping N .1) however when the free energy is negative for the operator (n. i. ZN valued gauge ﬁelds with coupling g [38–43] 2 The eﬀective “temperature” of the system will be given by. m0 ). so long as the free energy. m) is perturbative i.) At high T . F > 0. Berman and E. Rabinovici: Supersymmetric Gauge Theories 4 Phases of gauge theories 179 First we will gain our intuition from D = 4 lattice gauge theories. An operator with charges (n. n2 T + C m2 < · T N (4. it is an irrelevant operator and weakly coupled to system. How does the theory change? There are three phases depending which operators condense. C ﬁxed and vary T . that is. magnetic condensation occurs. there is electric condensation which implies that there is electric charge screening. It is diﬃcult to . n2 T + C m2 > . this is the Coulomb phase. an electric charge picks up a magnetic charge and becomes dyonic [44]. In the presence of a theta parameter.e. m).3) This lead to a tilted lattice of dyonic charges and one may condense dyons with charges (n0 . it condenses indicating the presence of a relevant operator and hence an infra-red instability. this occur when. At small T . Magnetic charges are screened. and the log of the Wilson loop is proportional to the length of the perimeter of the loop. n =n+ θ m. How does this relate to QCD? There are ideas that conﬁnement in QCD occurs due the condensation of QCD monopoles [45–47]. m0 ) being screened and all other charges being conﬁned [43].

c. by enriching QCD. adding scalars and making the the theory supersymmetric one can calculate the condensation of monopoles in a four dimensional gauge theory. 4. To this we turn next. however by embedding the monopole in a spontaneously non-abelian theory with an additional scalar ﬁeld one may smooth out the core of the monopole and remove the singularity. ˜ ˜ +Q+ exp(−2V )QF + Q exp(2V )Q+ |θθθθ + mF QF Q|θθ (5. One may proceed analogously. Gauge Theory and Strings m (magnetic) Higgs (T<<1) Possibly Coulomb (T~1) (T>>1) Confinement n (electric) Fig.180 Unity from Duality: Gravity. Novel properties of these theories have been discovered such as new types of conformal ﬁeld theories and new sorts of infra-red duality. The diﬀerent possible phases.1) F F ¯¯ . This has been achieved for gauge theories with N = 1. The Lagrangian is: L = (−iτ )TrW α Wα d2 θ + h. and NF chiral multiplets in the fundamental representation of SU (NC ) and NF chiral multiplets in the antifundamental representation. study this phenomenon directly. Consider the case of an N = 1 vector multiplet with gauge group SU (NC ). The Dirac monopole in a U (1) gauge theory is a singular object. 5 Supersymmetric gauge theories/super QCD The goal will be to examine theories that are simple supersymmetric extensions of QCD. 49]. There are many new methods that have been utilized and the phase structures of these theories have been well investigated [15. 2 supersymmetries. 48.

4) . Qi a ˜ Qa i (5. because 2 ∼ ¯ the global ﬂavor symmetry is enhanced 2.2) Apart from the local SU (NC ) gauge symmetry. while RF is now chosen so that the total anomaly vanishes. There is an anomaly of the U (1)A × U (1)R symmetry. This is denoted as U (1)R and is a full quantum symmetry. This leads to RF = − NC · NF (5. 2NF × RF .S. the ﬁelds are charged under the following global symmetries.D. The adjoint Fermion contributes 2NC × R(λ) to the anomaly. Possible phases of gauge theories (g1 and g2 are some relevant/marginal couplings). where the coupling is: τ= 4π θ +i 2 · 2π g (5. 5. The Chiral Fermions contribute. 2NC + 2RF NF = 0. when NC = 2. SU (NF )L NF 1 1 × SU (NF )R 1 ¯ NF 1 × U (1)V 1 −1 0 × U (1)A 1 1 0 × U (1)RC 1 1 1. to SO(2NF )L × SO(NF )R .3) Wα Note.5) (5. R(λ) = 1. A single U (1) symmetry survives the anomaly. Berman and E. Rabinovici: Supersymmetric Gauge Theories 181 g2 2) SU( GS HIG U(1 ) CO UL OM B SU (3) NT EME NFIN CO g1 Fig.

is given by: R = RC − NC QA .11) For NF = 0 or for NF = 0 and mF = 0.7) where RC is the classical R-charge and A is the classical U (1)A charge.182 Unity from Duality: Gravity. the rank of the group +1. This leaves NC vacua. The gluino condensate is: λλ = exp 2πik NC Λ3 C N (5. Tr(−1)F = NC i. These Fermionic zero modes break through instanton eﬀects the original U (1)R down to Z2NC .10) (5.e. + (5. Note. Further breaking occurs because the gluino two point function acquires a vacuum expectation value which breaks the symmetry down to Z2 . There are 2NC Fermionic zero modes (from the vector multiplet). NF (5. 5.9) (5. the vacuum structure is an infra-red property of the system hence having mF = 0 is equivalent to setting NF = 0 in the infra-red. This indicates there is no supersymmetry breaking. This leads to the following non-anomalous global charges: SU (NF )L Qi a ˜ Qa i Wα NF 1 1 × SU (NF )R 1 ¯ NF 1 × U (1)V 1 −1 0 × U (1)R NF −NC NF NF −NC NF (5. there is no moduli space. One is now ready to identify the classical moduli space.12) .8) 1. Gauge Theory and Strings The Bosons in the chiral multiplet have an R charge one greater than the Fermions in the multiplet. Thus.1 The classical moduli space The classical moduli space is given by solving the D-term and F -term equations: Da ¯ FQF ¯˜ FQF ˜ ˜ = Q+ T a QF − QF T a QF F ˜ = −mF Q = −mF Q.6) The non-anomalous R-charge. Consider the quantum moduli space of the case where NF = 0. The Witten index. RB = 1 − NC NF − NC = · NF NF (5.

.0 0 a2 0. Note. .D..e.16) The matrix ˜ M˜ ≡ Q˜Qj ij i forms a gauge invariant basis.) The gauge symmetry is partially broken: SU (NC ) → SU (NC − NF ).. Chiral symmetry breaking produces a mass gap. (To validate this classical analysis the vacuum expectation values must be much larger than any dynamically generated scale.. The Kahler potential is then. Rabinovici: Supersymmetric Gauge Theories 183 where ΛNC is the dynamically generated scale of the gauge theory and k = 1.15) the number of massless scalar ﬁelds becomes. Further details of quantum moduli spaces will be discussed later. NC −1 label the vacua. There are NF diagonal non-zero real entries.S.. Berman and E. (5. The classical moduli space is determined by the following solutions to the D-term equations: a1 0 0. if NF = NC − 1 then the gauge group is complete broken. This is called the Higgs phase.14) This is for generic values of ai . because chiral symmetry is discrete there are no Goldstone Bosons.0 ˜ Q=Q= · 0. Also. By setting some subset of ai to zero one may break to a subgroup of SU (NC ) that is larger than SU (NC − NF ). ¯ K = 2Tr (M M ) (5.18) (5. detM = 0 this signals the presence of enhanced symmetries. (5.e.. 2 2 2NC NF − (2NC NF − NF ) = NF .0 (5. Consider the case where mF = 0 and 0 < NF < NC . i. The number of massless vector Bosons becomes 2 2 NC − ((NC − NF )2 − 1) = 2NC NF − NF . .0 aNF 0.17) and as before when singularities appear i. (5. ai >> Λ.0 N ×N . ai .13) · 0.. F C Where the row indicates the the ﬂavor and the column indicates the colour.

B and B are classically dependent. B˜ = i F ˜ i1 i1 ···iNC Qj1 ˜ ˜ iN · · · QjNC . one has the following classical moduli space. F iN (5. There is a singular point in the ˜ moduli space where M = 0 = B = B.184 Unity from Duality: Gravity. again there are NF moduli from Mi˜ .25) (5. ﬁrst the classical moduli space for NF = NC .26) 2 giving again an NF + 1 dimension moduli space. consider the case. (The moduli space is not smooth. Nf = NC + 1. 2 Next. as well as the gauge symmetry being completely broken the global ﬂavor symmetry is also broken. 2 NF (5. when ˜ ai = ai = 0 then a subgroup of the SU (NC ) can remain. Generically.20) Generically the SU (NC ) symmetry is completely broken.21) There are degrees of freedom from M˜ and naively one would have two ij further degrees of freedom from: B= i1 i1 ···iNC Qj1 iN ˜ · · · QjNC . B = F ˜ i1 i1 ···iNC Qj1 ˜ · · · QjNC . There are also. We will now consider some special cases.19) with the constraint that a |ai |2 − |˜i |2 = ρ. a a 0 ˜ 0 1 1 0 Q= 0 · 0 a2 · · 0 · 0 ·· ·· ·· ·· ·· ·· aNC 0 0 0 NF ×NC .22) There is however a classical constraint: ˜ detM − B B = 0 (5. 0 ˜= Q 0 · 0 a2 ˜ · · 0 · 0 ·· ·· ·· ·· ·· ·· aNC ˜ 0 0 0 NF ×NC (5. . 2(NC + 1) degrees of freedom given by: j Bi = i1 ii1 ···iNC Qj1 iN ˜ · · · QjNC . However. F (5. The dimension of the moduli space is given by: 2 2 2 2 2NC − (NC − 1) = NC + 1 = NF + 1.24) However there are again the classical constraints: detM M˜ B ji i − = Mi˜ B i B j = 0 j Mi˜ B = 0 j ˜ j (5.23) 2 ˜ which means M . This leaves only NF +1 independent moduli. Gauge Theory and Strings When NF ≥ NC . (5.) There is a generic breaking of gauge symmetry.

There is no Coulomb phase so Wα does not appear. SU(NC ) Qi a Λ ˜ Qa i 3NC −NF M detM NC ¯ NC 1 1 1 SU(NF )L NF 1 1 NF 1 SU(NF )R 1 ¯ NF 1 ¯ NF 1 U(1)V 1 −1 0 0 0 U(1)A 1 1 2NF 2 2NF U(1)Rcl 1 1 2NC 2 2NF U(1)R NF −NC NF NF −NC NF 0 2− 2NC NF 2(NF − NC ) (5. SU (NF )L × SU (NF)R and the U (1)V vanish. the superpotential. Q. One examines the charges of W under the various symmetries. These restrictions ﬁx: W =c Λ3NC −NF detM 1 NC −NF · (5. Berman and E.2 Quantum moduli spaces 185 One is required to examine on a case by case basis the role that quantum eﬀects play in determining the exact moduli space. The symmetries imply. 2 2 NF = 3NC . W . NF > 3NC . Automatically. Rabinovici: Supersymmetric Gauge Theories 5. c is a numerical coeﬃcient. has the following form: W = (Λ3NC −NF )a (detM )b c (5. If c does not vanish.D. The power 3NC − NF is the coeﬃcient in the one loop beta function. the dimension of the moduli space is NF from Q. b are to be determined.27) Λ. If one requires the U (1)A charge to vanish then this implies a = −b. The following table summarizes the charges under the various groups [51–53]. NF = NC + 1. the classical moduli space gets completely lifted by these nonperturbative eﬀects. In what follows we examine the quantum moduli space for the separate cases: 1 ≤ NF ≤ NC − 1. 5. NC + 1 < NF ≤ 3NC . Quantum eﬀects both perturbative and nonperturbative can lift moduli. 3NC < NF < 3NC . .3 Quantum moduli space for 0 < NF < NC 2 ˜ Classically.28) a. NF = NC .29) For non vanishing c. 1 Requiring the U (1)R charge to vanish implies that b = NF −NC . the dynamically generated QCD scale is assigned charges as m and g were before. the charges of W for the ﬂavor symmetries.S. all the moduli are now lifted and there is no ground state.

The potential for 1 < NF < NC . For NF = NC − 1 there is complete symmetry breaking and one can turn to weak coupling. We begin by reviewing the conventional method of integrating out. We will also discuss when these conditions are met.186 Unity from Duality: Gravity. 6.4 Integrating in This method involves the addition of very massive ﬁelds to known eﬀective actions and extrapolating to the case where the additional degrees of freedom are massless.30) One may now go to NF < NC − 1 by adding masses and integrating out the heavy degrees of freedom. see [54–62].31) . It gives results for the phase structure in many cases. We will show that under certain circumstances it is possible to derive in a rather straight forward way the potential for light ﬁelds. This produces. 1 (5. What is the value of c? This is a diﬃcult to calculate directly unless there is complete Brouting. heavy degrees of freedom are integrated out to obtain an eﬀective potential for the light degrees of freedom. It is rather surprising that anything useful can be learned by this ﬂow in the opposite direction to the usual infra-red. M ij 5. Gauge Theory and Strings V det(M) Fig. min = (m−1 )i j (Λ3NC −NF detm) NC . We thus give some of the ﬂavour of this possibility. it has no ground state. From instanton calculations one calculates that c = 0 and the prepotential for the matter ﬁelds is W ∼ Λ2NC +1 detM · (5.

Z.35) . The ui and with them the macroscopic ﬁelds M and Z will be considered as heavy in part of the discussion.34) For the case m → ∞ one may tune the scale Λu so as to replace an ap˜ propriate combination of λ. Rabinovici: Supersymmetric Gauge Theories V 187 MMin M Fig. Λu ) = Wd (X. The da dofs will be kept light throughout all the discussion. The potential with ﬁnite masses has a ground state. Λd ) + WI (X. λ. It is convenient for general m = 0 to write. Berman and E. Λd ) ˜ ˜ (5. m and Λu by Λd the scale of the theory of the ˜ remaining light degrees of freedom. M. One thus obtains: ˜ ˜ ˜ Wd (X.D. Λu ) = (Wu (X. Z. Z · (5. m.33) g i Φi ) gi (5. 7. Consider a theory containing gauge invariant macroscopic light ﬁelds of the following nature: Fields X: built out of dA degrees of freedom Fields M : built out of ui degrees of freedom Fields Z: built out of both da and ui .S. The eﬀective potential is the Legendre transform ˜ Wu (Φ) = (W (gi ) − where ˜ ∂ W (gi ) = Φi · ∂gi (5. m. Mmin → ∞ as m → 0. (Λu is the dynamically generated scale of the theory. ˜ ˜ Wd (X. Λu ) describing all the light macroscopic degrees of freedom. One ends with Wd (X. λ. Λu ) = mM + λZ) M . Λd ). m. Assume that one is given the eﬀective potential Wu (X. λ. λ ˜ to the macroscopic degrees of freedom containing the removed ﬁelds: M and Z.32) Consider next making the microscopic “up” ﬁelds massive and integrating them out retaining only the light degrees of freedom and the couplings m. M.

Λd ) one can obtain: Wu (X. ˜ (5. Z. The quantum numbers of all relevant ﬁelds and parameters are given by: X λ Λb1 m ˜ W∆ U (1)Q 2 −2 2NF 0 0 U (1)Φ 0 −1 4Na −2 0 U (1)R 0 2 4 − 4Na − 2NF 2 2 (5. For the cases corresponding to the colour group SU (2) this indeed turns out to be the case. Imagine however that this term would vanish. WI = Wtree.. W∆ should obeys the constraints: W∆ → 0 .d = (Wtree ) ui (5. For convenience one writes. λ. Writing t as.38) Now one reverses the direction and integrates in.37) has by deﬁnition no scale dependence. and to partition again. Λu ) = (Wd (X.42) where b1 = 6 − 2Na − NF . In this case the calculations become much simpler.36) (5. and quark ﬂavors Qi .39) All the essential complexities of the ﬂow lie in the term W∆ . Λd ) + W∆ + Wtree. one brings down inﬁnitely heavy ﬁelds in the adjoint representation. where M = ΦΦ and Z = QΦQ. one resuscitates the full up theory. .d − Wtree ) m .43) . One starts with SU (2) without adjoint ﬁelds (Na = 0).d + W∆ where Wtree. ˜ (5. Φ is the adjoint ﬁeld we add. Λ) = Wtree. i = 1. M. W∆ (. This can be seen in the following manner. Λd ) is the exact result for inﬁnitely heavy Ui .d f (t) ˜ (5. λ ˜ . (5.NF (NF ≤ 4) for the down theory. given an exact form for Wd (X. or m → ∞ . Gauge Theory and Strings where Wd (X.41) (5. which contains Na adjoint ﬁelds Φα with the superpotential: ˜ Wtree = mX + mM + λZ. that is. t ∼ (Λb1 )a mb X c λd .188 Unity from Duality: Gravity. Namely. when Λu → 0 . m.40) where X = W W and t being any possible singlet of SU (2NF ) × U (1)Q × U (1)Φ × U (1)R .

Rabinovici: Supersymmetric Gauge Theories It is a singlet provided that b = (2Na + 2 − NF )a . For Na + 1. For NF = 0 and Na = 1 b = 4a leading to ∞ (5. (5.49) The constraints imply that r − a indeed vanishes for all a’s. (The role of these elliptic curves in N = 2 theories will be described later.) . W∆ = r(mΛ)b1 ˜ + .47) which implies that for all values of Nf the vanishing of W∆ .44) (5.. The equations of motion of these potentials can be arranged in such a manner as to coincide with the singularity equations of the appropriate elliptic curves derived for systems with N = 2 supersymmetry and SU (2) gauge group. m ˜ (5. ˜ (5. NF = 1. 3. 189 (5.45) as well as the fact that in the Higgs phase one can decompose W as: ∞ W (Λb1 ) = n=1 an (Λb1 )n .50) implying that r = 0 as well. Recall the constraints W∆ → 0 for m → ∞ and W∆ → 0 for ˜ Λ→0 c = (NF − 4)a .46) One shows that Wtree. Starting from known results for Na = NF = 0 one can now obtain the eﬀective potential for all relevant values of Na and NF .D. Berman and E.S. 2. We see this explicitly in the following cases.d ∼ 1 m ˜ (5. d = (2NF − 4)a .. b1 = 4 − NF .48) W∆ = a=1 ra (Λb1 )a ma−1 .

56) . mij for matter one obtains. This implies. On the other ˜ hand. There are allowed soft perturbations. For NF ≥ NC it is possible to have m → 0 while keeping M i j ﬁxed. 1 (5. R(M ) = 0 ⇒ W = 0 .55) (detM )i+j ij then by demanding that there be no singularities at small M or at large ˜ B implies that all cij must vanish and the hence detM − B B = Λ2NC obeys a nonrenormalisation theorem.54) 1 (5. In the presence of a mass matrix. b˜ (5.57) (5.53) since m cancels. B = B = 0 if detm = 0 because all ﬁelds carrying B number are ˜ integrated out.190 Unity from Duality: Gravity.52) This has the eﬀect of resolving the singularity in moduli space. Therefore detM = B B through quantum eﬀects. mass terms given by: W = tr(mM ) + bB + ˜B. Note. (5. detM = Λ2NC (5. In this case. M i j = (m−1 )i j (Λ3NC −NF detm) NC . Mi j = (m−1 )i j (detm) NC Λ2 . Note.6 NF = NC Quantum eﬀects alter the classical constraint to: ˜ detM − B B = Λ2NC . it turned out that m → 0 implied M i j → ∞ thus explicitly lifting the classical moduli space. R(MIJ ) = 0 for NF = NC . Gauge Theory and Strings 5. Writing out an expansion that obeys the R-charge conservation: ˜ (B B)i Λ2NC j ˜ cij detM − B B = Λ2NC 1 + (5. The absence of a singularity means there will not be additional massless particles. thus this also holds in the limit m → 0. 5. for the case of NF < NC .51) Previously.5 Quantum moduli space for NF ≥ NC There is a surviving moduli space.

the global symmetries are broken to: SU (NF )V × UB (1) × UR (1) and there is chiral symmetry breaking. is it in a Higgs/conﬁnement phase? For ˜ ˜ large. Note that M cannot be taken smaller than Λ. M/B/B one is sitting in the Higgs regime. B become dynamically independent. then the group is broken to: SU (NF )L × SU (NF )R × UR (1) (5. This is not a theory of massless gluons but a theory of massless ˜ mesons and baryons. B. where because of strong infra-red ﬂuctuations there are n independent ﬁelds even though there is a classical constraint. This is an interesting situation because there is a dogma that as soon as a system has a bound state there will be chiral symmetry breaking [65]. Rabinovici: Supersymmetric Gauge Theories One check is to integrate out to give the case NF = NC − 1 yielding. 5. The eﬀective potential is: Weﬀ = 1 ˜ (M i j Bi B j − detM ) Λ2NC −1 (5. M = B = B = 0 there is no global symmetry breaking but there is “conﬁnement” with light baryons.58) What is the physics of this theory. The classical and quantum moduli ˜ spaces are the same and hence the singularity when M = B = B = 0 remains.62) ˜ BB = 0 (5. When.D. M. The analogy is from the nonlinear sigma model.60) ˜ BB = 0 (5.63) . Global symmetries need to be broken in order to satisfy the modiﬁed constraint equation. for small M/B/B one is in the conﬁning regime. When.61) (5. At the singular point when. ˜ There is a suggestion that in this situation.S.59) which has chiral symmetry and also has conﬁnement. M ij = 0 . however. M. In both cases the ’t Hooft anomaly conditions [66] are satisﬁed.7 NF = NC + 1 The moduli space remains unchanged. B. Berman and E. W = ˜ Λ2NC +1 mΛ2NC = · detM detM 191 (5. These will be discussed later. M i j = δ i j Λ2 . B = 0 then one is in a Higgs/conﬁning ˜ phase. Consider some examples: With the following expectation value.

F Dual Descriptions 3NC 2 Magnetic IR Free. which is ZN for SU (N ). SU (2) × U (1).192 Unity from Duality: Gravity. Is it in a Higgs or conﬁnement phase? Below we .8 Higgs and conﬁnement phases This section is discussed in [63]. Gauge Theory and Strings the classical limit is taken by: Λ→0 (5. Again the system obeys the ’t Hooft anomaly matching conditions. IR FREE NF 3N C Conformal Two A. This contrasts the situation of pure QCD or super QCD where all the particles are in the adjoint representation which is trivial under the center. The phases of super QCD. So what about the standard model. In such a case there is a phase transition and there is a qualitative diﬀerence between the phases. 5. 8. with matter in a nontrivial representation of the center of the group.64) which in turn imposes the classical constraint. S=1 NC+2 NC+ 1 NC NC-1 No Vacuum 1 0 No Global Chiral Symmetry Breaking Modified Moduli Space !! Confinement < λλ> = 0 Discrete Vacua Fig. the diﬀerence between the Higgs and conﬁnement phases is purely quantitative. There is no phase boundary. While one is discussing the conﬁnement phase in supersymmetric gauge theories one should recall that for gauge theories such as SU (N ) Yang Mills.

Berman and E.74) Unitary gauge is Ω = 1. Along with. At ﬁnite temperature the two phases are qualitatively indistinguishable. + L = trFµν (B)F µν (B) + ∂µ ρ∂ µ ρ + ρ2 (Bµ B µ )|| + V (ρ2 ). Q(ψ) = 1 1 e Q(φ) = e 2 2 Q(W 0 ) = 0. (5.68) (5. φ+ φi i φi Dµ φj ij ψi φj 193 1 2 1 I= 2 I= I=0 φ → φreal (l)L (q)L (l)R (q)R . φ(x) = Ω(x) then Bµ = Ω + D µ Ω leading to the Lagrangian. One may choose a gauge. Examine the charges of the ﬁelds with respect to the unbroken U (1).65) (5.71) ij . ρ(x) 0 (5.D. (q)R are SU (2)L singlets. In the Higgs picture.75) Like conﬁnement but with the scale determined by: |φ|.72) (5. (l)R . Rabinovici: Supersymmetric Gauge Theories present the spectrum in the two pictures. φ+ Dµ φ+ ij i j . W 0 = i Dµ i · |φ| |φ| (5.76) . s= 1 2 s=0 s=1 φ+ ψi .S. i (5.67) (5.66) (5.69) φ+ Dµ φi . |φ| |φ| + φ+ j ˜+ φ Wµ = i Dµ |φ| |φ| ij φ+ φ+ ˜ . ψ2 = .70) (5.73) (5. The Higgs picture also contains the operators: ψ1 = φ+ ψi φi ψj ij i . (5. In the Englert picture: s=0 1 2 1 s= 2 s= In the conﬁnement picture.

78) (5.77) This matches the charges of operators in the conﬁned picture. 9.194 Unity from Duality: Gravity. 5. Gauge Theory and Strings However. The appropraite conserved charge is actually: Q = Q + T3 . this will not work for the Georgi-Glashow model where φ is a triplet in SO(3). This is seen empirically by the absence of radial excitations of the Z particles [64]. they become more and more similar.80) Q (W 0 ) = 0 Q (v) = 0 Q (W ± ) = ±1 Q (ρ) = 0 (5. Then. . when observed at longer and longer length scales. Seiberg has observed and has given very strong arguments that the following set of N = 1 supersymmetric gauge theories are pairwise infra-red dual [88]. the conﬁned objects have integral charge. System I t t t t System II t t t t) • Fig. Is it possible to have strong “weak” interactions at high energies? This is theoretically possible but it is not the course chosen by nature for the standard model.79) (5. Two systems with a diﬀerent ultra-violet behavior ﬂowing to the same infra-red ﬁxed point. (5. Q (e) = 1 and Q (ψφ) = 1 Q (φ+ Dµ φ) = 0 Q (φ + ψ) = 0 Q (φ+ Dµ φ+ ) = 1 Q (φ+ φ) = 0 Q (φDµ φ) = −1.9 Infra-red duality Two systems are called infra-red dual if. Note.

It is infra-red dual to a theory with NF −NC 2 colors and NF ﬂavors and NF color singlets. if the number of ﬂavors is in the interval NC + 1 < NF < 3NC . This comes about in the following way: consider. For example. Nevertheless. the number of ﬂavors. it has been known for quite a long time that two systems which diﬀer by irrelevant operator have the same infra-red behavior. where groups with diﬀerent number of colors are infra-red dual.2 particles are also the infra-red resolution of another theory. We have no indication whatsoever. for which the infra-red duality holds is always large enough so as to make the entries in the table meaningful. Seiberg’s duality allows for the ﬁrst time to ascribe a similar role to massless infra-red free spin-1 particles. Berman and E. that is. for example. QCD. just the infra-red limit of a . one can realize a situation where free 1 massless spin. whose gauge group is SU (NC ) and whose ﬂavor group is SU (NF ) × SU (NF ) × U (1). Rabinovici: Supersymmetric Gauge Theories 195 System Gauge Group SU (NC ) SO(NC ) Sp(NC ) #ﬂavors NF NF 2NF Dual System Gauge Group SU (NF − NC ) SO(NF − NC + 4) Sp(NF − NC − 2) #ﬂavor #singlets NF NF 2NF 2 NF 2 NF 2 NF . Note that the rank of the dual pairs is usually different. that this is the case with Seiberg’s duality.D. QCD whose ultraviolet behavior is described by other particles. is expected to be infra-red dual to a theory of massless pions which are all color singlets. NF . Seiberg’s work has shown that infra-red free massless spin-1 particles(for example photons in a SUSY system) could be. For a given number of colors. Massless spin-1 particles play a very special role in our understanding of the basic interactions. are actually infra-red free in four dimensions. By using supersymmetry. NC . We have thus relearned that free spin-0 massless particles can actually be the infra-red ashes of a stronglyinteracting theory.S. Lets explain why this result is so powerful. the N = 1 supersymmetric model with NC colors and NF ﬂavors. For a given NC . the original theory 2 is strongly coupled in the infra-red. the common wisdom in hadronic physics has already identiﬁed very important cases of infra-red duality. while the dual theory has such a large number of ﬂavors that it becomes infra-red free. Thus the infra-red behavior of the strongly-coupled system is described by infra-red free spin-1 massless ﬁelds (as well as its superpartners). In general. being the spin-0 Goldstone Bosons of the spontaneously broken chiral symmetry. under certain circumstances. The pions.

such that a continuous trajectory relates them. In order to be able to appreciate how that is derived. Fig. Gauge Theory and Strings much more complicated ultraviolet theory. The charges of the matter ﬁelds are given by the table below: SUL (NF ) SUR (NF ) UB (1) Q ˜ Q NF 1 1 ¯F N 1 −1 UR (1) 1− 1− NC NF NC NF · (5. NF fundamental. a continuous path in parameter space relates a pair of two disjoint infra-red-dual ﬁeld theories. We will describe the ingredients of such an embedding [68. 10. In string theory. First we will describe some more details of the Seiberg infra-red duality in ﬁeld theory. From the point of view of string theory the two systems are embedded in a larger space of models. 69] later. The infra-red duality relates two disconnected systems. Consider the example of N = 1 supersymmetric Yang¯ Mills theory with gauge group SU (NC ) and NF . NF fundamental and antifundamental matter and NF .81) The infra-red dual is N = 1 supersymmetric Yang-Mills with gauge group 2 ¯ S(NF − NC ) and NF . Seiberg’s duality has passed a large number of consistency checks under many circumstances. we will need to learn to use some tools of string theory.196 Unity from Duality: Gravity. antifundamental matter.

This looks like there is a solitonic dual for the quarks in this case. ˜ • The Baryons built from Q. • If M is fundamental there should be an associated UM (1) charge which does not appear in the original SU (NC ) theory. If one regards a gauge symmetry as a redundancy in the description of the theory then this is not important.82) 0 −N 2 NFNF C · One must also add an interaction term in the dual theory described by: W = 1 i ˜ M ˜ qi q j . 3NC < NF < 3NC .D.83) • The dual theories have diﬀerent gauge groups.83). • Where are the q. . Q have the same charges as those built from q. Rabinovici: Supersymmetric Gauge Theories gauge singlets. • Note. Q. Berman and E.84) and so it is a relevant operator. In both dual pictures there is an Infra red ﬁxed point. For the case NF = NC + 1 then the Baryon of the SU (NC ) ˜ theory becomes the q in the dual theory (which is a singlet in this case). and for 3 (NF − NC ) < NF < 3(NF − NC ). both are asymptotically free and in the center of moduli space the theories will be a conformal. 2 2 The operator. it is not possible to build a meson out of q. The M ﬁeld in the dual theory ˜ does have the same charges as a meson built from Q. M q q has dimension: ˜ D(M q q ) = 1 + ˜ 3NC <3 NF (5. j µ (5. ˜ For the case. q . q mesons. which implies one may ignore them in the ˜ infra-red. q that has the same ˜ ˜ R-charge as a meson built from Q. q mesons in the original SU (NC ) theory? ˜ • the resolution to the previous two points is provided by the interaction term (5. What does matter is that the two dual theories share the same global symmetries.S. This term breaks the UM (1) symmetry and provides a mass to the q. The charges are given by: SUL (NF ) q q ˜ M ¯ NF 1 NF SUR (NF ) 1 NF ¯ NF UB (1) NC NF −NC NC − NF −NC 197 UR (1) 1− 1− NC NF NC NF (5. Q.

In the above example there are anomalies for: SU (NF )3 . φi is given by: V = ([φ. Let us examine some of the consequences of this duality. • There are the same number of ﬂat directions. The ’t Hooft anomaly matching conditions are determined as follows. For the case 3NC < NF < 3NC the two dual theories are both asymptotically free.this will be discussed later. At ﬁrst we will have no matter in the fundamental representation and scalar multiplets which are adjoint valued. UB (1) UR (1). the theory is an infra-red free gauge theory plus 2 free singlets. (5.85) (5. Perhaps one can more learn about this system since it is a ﬁxed point under duality. Gauge Theory and Strings The checks of the duality are as follows: • They have the same global symmetries. φ])2 . One then calculates their anomalies. UB (1) . This is the ﬁrst example of a weakly interacting theory with spin one particles that in the infra-red one may view as bound states of the dual theory.87) This potential obviously has a ﬂat direction for diagonal φ. • There is the same reaction to a mass deformation. It 2 is symmetric around NF = 2NC . UR (1) . The panorama of these structures is given in Figure 8. • They obey the ’t Hooft anomaly matching conditions. UR (1) .86) All these anomalies match between the dual theories. Let us now enrich the structure of the theory by adding Na particles in the adjoint representation. 3 2 3 2 (5. SU (NF )2 UB (1) . the supersymmetry is now increased to . The potential for the scalars. Adding a mass in one theory is like an Englerting in the other. The gauge invariant macroscopic moduli would be Trφk .198 Unity from Duality: Gravity. One takes the global symmetries in the theory and then make them local symmetries. This will be the subject of the next section. Both dual theories must share the same anomalies. • It is a Z2 operation. At the origin of moduli space one may have obtained a new conformal theory. There was no a priori reason for them to do so. Consider the non generic example of NC = 2 and Na = 1. • There is a construction of the duality by embedding the ﬁeld theory in string theory. SU (NF )2 UR (1) . For NC + 2 ≤ NF ≤ 3 NC . UB (1) UR (1) .

Z) act on τ . At the special points in the moduli space. one would expect that the strong infra-red ﬂuctuations would wash away the expectation value for Trφ2 and the theory would be conﬁning. Rabinovici: Supersymmetric Gauge Theories 199 Fig. new particles will become massless.90) . g.2. the original couplings and the scale. N = 2. B → −E . Berman and E. This theory has an SL(2. Recall the usual electromagnetic duality for Maxwell theory in the presence of charged matter is: E → B . Z) duality symmetry.88) The τeﬀ is the eﬀective complex coupling which is a function of the modulus. τ → τ + 1. naive quantum and exact quantum moduli spaces. This generalizes to a U (1) symmetry by deﬁning: E + iB . deﬁned by 5. SU (2) is broken to U (1) for Trφ2 = 0. e → m . L= d2 θIm(τeﬀ (trφ2 . 11.S. Classically. there are only two special points in moduli space and even there the theory is only on the verge of conﬁnement. τ This is a generalization of the usual U (1) duality that occurs with electromagnetism to the case of a complex coupling. One would expect a singularity at Trφ2 = 0. as follows: 1 (5. We will now examine the eﬀective theory at a generic point in moduli space where the theory is broken down to U (1). 49]. because of the very strong constraints that supersymmetry imposes on the system.D. one would expect the following. The surprising thing is that when SU (2) breaks down to U (1). The exact quantum potential vanishes in this case [48. (5. Λ)Wα W α ). The classical. Trφ2 . m → −e. The generators of the SL(2. e + im. When Trφ2 is of order Λ or smaller. (5. trφ2 .91) (5. Everywhere else the theory is in the Coulomb phase. This is illustrated in the Figure 11. There is a single complex modulus. The Lagrangian is given by. Naively. Λ.89) τ → − .

This reﬂects the presence of massless charged objects. . c. The following complex equation. Gauge Theory and Strings The duality symmetry now acts by: E + iB → exp(iα)(E + iB) . a. (5. When y(x) and y (x) vanish for the same value of x then τ is singular. ˜ ∂uk ∂(q q ) ˜ (5. there is no perturbative or nonperturbative corrections.97) =0 .. The eﬀective prepotential is now: W = M (uk )q q + gk uk ˜ then ∂W ∂W = 0 ⇒ M ( uk ) = 0 .92) Previously for the SU (2) case the moduli were given by u = Trφ2 for SU (NC ) the moduli are given by uk = Trφk . The theory is on the verge of conﬁnement.95) .94) and the eﬀective coupling vanishes.93) determines a torus. e + im → exp(iα)(e + im).96) (5. 2 τeﬀ = i∞ . 62]. One can push this to conﬁnement by adding a mass term: mTrφ2 . The complex structure of the torus. b.. How does one ﬁnd τ as a function of the u? There is a great deal of literature on the subject here we will just sketch the ideas [48–50].200 Unity from Duality: Gravity. (5. k = 2. 59. couplings and scale and so will implicitly determine τtorus . d are holomorphic functions of the moduli. This occurs for deﬁnite values of u in the moduli space. Therefore. or generally for SU (NC ) ˜ the term: δW = gk uk . ∂uk M ( uk ) q q = −gk . For N = 2 supersymmetry that is the best one can do. The monopoles are massless but they have not condensed [54–57. NC . For condensation to occur the monopoles should become tachyonic indicating an instability that produces a condensation. These new massless particles are monopoles or dyons. 60. Again the classical moduli space is singular at times. y 2 = ax3 + bx2 + cx + d (5. τtorus will be identiﬁed with the complex coupling τeﬀ . geﬀ = 0 (5.

1 = = =0 ∂M ∂X ∂Z which imply: m ˜ m 1 √ λ 2 where R−1 ≡ Λ− (P f X) 2 Γ . 2 1 4−b1 (5. j = 1. Rabinovici: Supersymmetric Gauge Theories Since generically.106) = 2Λ−1 (P f X) 2 = R−1 (X −1 − 8Γ−1 X −1 (ZX −1 )2 ) = 4R−1 Γ−1 X −1 ZX −1 1 (5. β = 1.1 ∂W2.104) (5. We deﬁne the quantity.102) (5. Consider the case. Take the case Na = 1.100) (5. The ﬁelds that are the moduli in the system are: XIJ Mαβ Zij = = = a b ab Qi Qj aa aa ab a b bb φα φβ bb (5. This is achieved here by using the integrating in method discussed earlier. The equations of motion from minimizing the superpotential are: ∂W2.108) (5.S. α.105) This respects the necessary symmetries and can be checked semiclassically.107) (5. Z) = Mαβ + Tr2NF (Zα X −1 Zβ X −1 ) which we will use to write the prepotential as follows. Z) = (b1 − 4) Λ−b1 P f X(det(Γαβ ))2 1 +TrNa mM Tr2NF mX ˜ 2 1 + √ Tr2NF λα Zα ...1 ∂W2. i. a. X.103) (5.109) 1 Γ.98) then there will be condensation.110) . ∂uk M ( uk ) = 0 201 (5. 2. NC = 2 with arbitrary NF and Na .. Berman and E.99) (5.Na (M.. 2NF. X ≡ 1 (5. b = 1. WNF . X.101) Qa φa b Qb i α j where Q are fundamental and φ are adjoint ﬁelds. Na .D. We now describe how the complex elliptic curve arises using more physical terms. NF = 2. 2 (5. Γαβ (M.

116) . One can therefore identify the parameters as: a = −M c= α (2M + Tr(µ2 )) 8 b=− α Λ2 + Pfm 4 4 Λ4 .10 Superconformal invariance in d = 4 For the case of 3 NC < NF < 3NC . at the center of moduli space when all 2 expectation values vanish. We now rexamine some special properties of the region 3 NC < NF < 3NC . These are Argyres Douglas points [71]. it is claimed that the theory is described by a non-trivial conformal ﬁeld theory [73].112) Taking y and y to vanish we can compare with the elliptic cure. 2 5.113) Identifying the modular parameter of the torus from the elliptic equation involves standard techniques in algebraic geometry. • These techniques may be extended to obtain curves for other more complicated groups.111) (5. µ ≡ λ−1 m. y 2 = x3 + ax2 + bx + c. This is an indication of the Higgs/conﬁnement complementarity.114) (5. the exact β function: β(g) = − g 3 3NC − NF + NF γ(g 2 ) 2 16π 2 1 − g 2 NC 8π (5. There are several motivations for reaching this conclusion. Examine for example. (5. This modular parameter will then be the eﬀective coupling of the theory.202 Unity from Duality: Gravity.115) (5. are degenerate vacua which are possibly non-local with respect to each other. Gauge Theory and Strings The following equations are then obeyed: X 3 − M X 2 + bX − 1 (c − 8M )˜ = 0 c 128 X 2 − 2M X + b = 0. • As you move in moduli space monopoles turn smoothly into dyons and electric charge. Some comments: • Some points in moduli space when 2 + NF = 4. α≡ 16 (5.

122) Generally. One can prove that the scaling dimension of a ﬁeld is bounded by its R-charge as follows: D≥ 3 |R| 2 (5. Rabinovici: Supersymmetric Gauge Theories where γ(g 2 ) is the anomalous dimension given perturbatively by: γ(g 2 ) = − 2 g 2 NC − 1 + O(g 4 ) . In d = 2 the conformal group is inﬁnite dimensional and so provides very powerful constraints on the theory.117) If one now considers a limit where NF .120) This observatoin will lead to interesting conclusions. is at g ∗2 = 1 8π 2 .D. In d = 4 the conformal group is ﬁnite dimensional and so the conformal symmetry does not constrain the theory in the same way [74.119) ˜ D(QQ) = 2 + γ = 3 NF − NC · NF (5.122). The bound is saturated for (anti-)chiral ﬁelds. Since the coupling at the ﬁxed point is proportional to << 1. The product contains a superposition of operators with diﬀerent dimensions. β(g ∗ ) = 0. The anomalous dimension at the ﬁxed point is: γ(g ∗ ) = 1 − ˜ The dimension of the QQ is 3NC · NF (5. Consider the operator product: O1 (x1 )O2 (x2 ) = i i O12 f i (x1 − x2 ).S. both the existence of a ﬁxed point and the perturbative evaluation for γ is justiﬁed. the dimension of the product of the operator appearing on the righthand side of equation (5. (5.121) where D is the scaling dimension and R is the R-charge. NC are both taken to be large but their ratio is kept ﬁxed then the ﬁxed point of the β function. Nevertheless there are still interesting properties of d = 4 conformal theories arising from the conformal invariance. 75].118) NF where = 3 − NC << 1.122) is not a sum of the dimensions of the two operators appearing on the left hand side of equation (5. For chiral . Berman and E. NC 3 (5. 8π 2 NC 203 (5.

since D(QQ) = 1. There are reasons to expect that at the ﬁxed point. The descendents are obtained by acting on primary states with momentum operators. The deﬁnition of a primary state is that it is annihilated by the generator of special conformal transformations. Non-trivial superconformal N = 2 theories in d = 4 occur for N = 2. Since the R-charge is additive and chiral ﬁelds saturate the bound (5. This NC + 1 < NF < 2 NC it appears that D(QQ) is an indication that one is using the wrong degrees of freedom and a dual description is required. the operator QQ becomes free. the infra-red nonanomalous R-charge equals the non-anomalous R-charge of the ultra violet. D(O) = 1. For 2 3 ˜ < 1 which is forbidden.123) i and f i (x1 − x2 ) are thus all constants.124) The dimension of the Baryon and anti-Baryon are: ˜ D(B) = D(B) = (5. (5. The Lorentz group dej composes into SU (2)L ⊕ SU (2)R with charges j.126) ˜ ˜ When NF = 3 NC . where I is the identity operator and O is an operator = I.120): ˜ D(QQ) = NF − NC 3 ˜ · R(QQ) = 3 2 NF 3NC (NF − NC ) · 2NF (5. D(O) ≥ 1.121) i R(O12 ) = R(O1 (x1 )) + R(O2 (x2 )) ⇒ D(O1 O2 ) = D(O1 ) + D(O2 ) (5. see equation (5.125) For unitary representations (of spin = 0) ﬁelds.128) . This will be elaborated later.204 Unity from Duality: Gravity. SU (3) theories without matter or SU (2) theories with matter [76]. ˜ charge as well as D the dimension. K. For a free ﬁeld the bound is saturated. D(I) = 0 . The closure property of the chiral ﬁelds under the operator product expansion leads to the name “chiral ring”. Gauge Theory and Strings ﬁelds the situation is simpler. By (5. The key point is that a non-trivial conformal theory with vector ﬁelds contains both massless electric and magnetic excitations (these are mutually non-local).127) (5.123) O12 is also a chiral operator. ˜ respectively. ˜ At the ﬁxed point of the dimension of QQ is given by. j Unitary chiral primaries obey: j ˜ = 0. An operator with non-zero spin will be carry j. The dimension of a chiral ﬁled is: j D(O) ≥ j + ˜ + 1 j (5.

j For a free ﬁeld: D(O) = j + ˜ + 1.132) These states are the decendents of F ± .131) One can show that there are states associated with the conserved currents: ± ± Jµ = ∂ ν Fµν . Je ≡ J+ + J− . called the Argyres Douglas point [71]. j 205 (5.136) are non-vanishing [76]. F ± = F ± ∗F. obtained by applying the momentum operator P on the chiral ﬁeld.D.137) (5.133) where the undotted indices are SU (2)L indices. At the point in moduli space (discussed previously) where there are both electric and magnetic charges.130) This generalizes the previous results for scalar chiral ﬁelds. (5. F = F + + F −. Berman and E.135) For D = 2. Jm ≡ J+ − J− (5. dotted indices are SU (2)R i indices and 2 M 12 = J 3 .134) and that K annihilates primary states. K β β ] = ˙ i αβ M 2 αβ ˙ ˙ ¯˙˙ + M αβ αβ +D αβ αβ ˙ ˙ . ˙ (5. The norm is then calculated using the following: the commutator of the d = 4 conformal algebra.129) (5. Fµν decomposes into self-dual and anti self-dual parts that form irreducible representations of SU (2)L ⊕ SU (2)R . then ||J ± | > 0 and therefore both. . After some algebra this produces.S. Rabinovici: Supersymmetric Gauge Theories as compared with a non-chiral operator: D(O) ≥ j + ˜ + 2. (5. the Hermitian conjugate relation. ||J ± | = 2(D − 2). (5. ˙ (P αα )+ = − αβ αβ ˙ ˙ K ββ . ˙ [P αα . If F is not free and D > 2. these are null states and F ± is free. (5. F ± .

the theory has full N = 4 supersymmetry.139) When h = g.206 Unity from Duality: Gravity.φn perturbation one has n βh = h(µ) −dw + k=1 1 d (φk ) + γ(φk ) 2 (5. X3 : Lint = hX1 X2 X3 . For a general hφ1 .. In the case at hand 3 dw = 3 . Hence. g being the gauge coupling.141) and therefore 3 βh = h(µ) γ(φ) 2 and βg = −f (g(µ))((3C2 (G) − k (5.140) where dw is the engineering dimension of the composite perturbing term. One can deform N = 4 theories with marginal operators such that the global symmetries are broken but the theory remains conformal. consider. L = L0 + i gi O i (5. By symmetry all ﬁelds have the same anomalous dimension and so γ(φk ) is independent of k and so is denoted as simply γ(φ).138) where Oi are the set of operators with dimension 4. this is known to be an exact result [77]. d(φk ) is the engineering dimension of the ﬁeld. φk and γ(φk ) is the associated anomalous dimension of the ﬁeld φk . Other conformal theories occur for N = 1 supersymmetry when Na = 3 and the couplings are appropriately tuned. k=1 d(φk ) = 3 .144) = −f (g(µ))3T (R)γ(φ) . X2 . Consider N = 1 with Na = 3 and an interaction for the adjoint ﬁelds X1 . Gauge Theory and Strings this condition is met and it is possible to have a non-trivial conformal ﬁeld theory with spin one particles. (5. For N = 2 supersymmetry. k 1 3 γ(φk ) = γ 2 2 (5.143) (5. this is actually a deformed N = 4 theory.142) T (Rk )) + k T (Rk )γ(φk )) (5. the theory is conformal when NF = 2NC and the one loop β function vanishes.

The problem was originally stated as follows. which is non vanishing. a. At the ﬁxed line. one can ask if the N = 4 theory has any discrete symmetries that relate theories with diﬀerent moduli. 12. Given that there is an eﬀective scale below which the physics is known then (such . 82]. Berman and E. (5.S. RG ﬂow to the ﬁxed line. supersymmetric gauge theories have a very rich phase structure and many outstanding dynamical issues can be discussed reliably in the supersymmetric arena that are hard to address elsewhere. Rabinovici: Supersymmetric Gauge Theories 207 h g Fig. Also the relations between relevant deformations follow a similar pattern [78]. both βh and βg vanish if γ vanishes.2): τ→ aτ + b . The answer is yes [79]. This means there is a ﬁxed line. cτ + d ad − bc = 1 . c. 6 Comments on vacuum energies in scale invariant theories The puzzle to be addressed in this section is that of the cosmological constant problem [81. d ∈ Z. Z) duality [80] that identiﬁes theories with coupling τ given by equation (5. where g = h and the supersymmetry is enhanced to N = 4.) This ﬁxed line is infra-red stable. C2 (G) is the second Casimir of the group G and T (Rk ) is associated with the representation of φ.145) To conclude. b.D. There is a great deal of evidence that N = 4 theories possess an SL(2. Therefore. (This is diﬀerent from the more generic situation where the β functions are not related and there are isolated ﬁxed points. rather than a ﬁxed point.

L = ∂µ φ · ∂ µ φ − g6 |φ|2 3 (6. This result is exact to all orders in perturbation theory and non-perturbatively. βg6 = 0 (6. N → ∞.3) 1 ( N corrections break the conformality). In the limit. If the original theory is scale invariant then the eﬀective action of the spontaneously broken theory should reﬂect this symmetry. The consequence of scale invariance is a zero vacuum energy whether or not scale invariance is spontaneously broken. The potential is given by: V = [φ. described by the Lagrangian.85]. φ]2 (6. g6 is a modulus. An eﬀective potential can be written down for the O(N ) invariant ﬁeld. which is the dilaton (the Goldstone Boson associated with spontaneously broken scale invariance). The vacuum energy remains zero. Giving an expectation value to a ﬁeld will break the scale invariance spontaneously as well as generically breaking the gauge symmetry down to U (1)r where r is the rank of the unbroken gauge group. One should observe some important caveats in this argument [83. When one writes down a low energy eﬀective action for a spontaneously broken theory one should respect all the symmetries that appear in the original action. The presence of a ground state is crucial.84] and also [78. recall the example discussed in quantum mechanics where the action scale invariance and supersymmetry were spontaneously broken but the system had no ground state. As an example consdier N = 4 super Yang-Mills in four dimensions.208 Unity from Duality: Gravity. Thus provided there is a translationally invariant ground state.4) . the vacuum energy calculated in such a manner however would give a cosmological constant many orders of magnitude above the observed value. σ = |φ|2 (6. For either choice of the above physics.1) thus the theory has ﬂat directions. global supersymmetry can’t be spontaneously broken whether or not scale invariance is spontaneoulsy broken. The next example is the O(N ) model in three dimensions.2) where the ﬁelds φ are in the vector representation of O(N ). An analysis of the spectrum shows that a gauge singlet particle emerges. Gauge Theory and Strings as QED or the standard model) integrating out the more energetic degrees of freedom above this scale leads to a vacuum energy that is proportional to the cut oﬀ.

To summarize.0 0 209 (6.2 |φ1 |2 −λ0.6 |φ2 |2 2 3 (6. The possibilities are summarized by the table below: O(N ) + − + − O(N ) + + − − scale massless + all − (N − 1)π s.B.B. D − (N − 1)π s. One has the following possibilities. masses V No 0 0 No 0 0 Yes σ . Rabinovici: Supersymmetric Gauge Theories the eﬀective potential is: V (σ) = f (g)|σ|3 . D massive V none 0 N. (For f (g) < 0 the theory is unstable). in all cases the vacuum energy vanishes.4 |φ1 |2 . σ 0 σ 0.7) 2 |φ2 |2 − λ2. one associated to the break down of a global symmetry and another with the break down of scale invariance.6) where S. Assume a hierarchy of scales where the scale invariance is broken at a scale much above the scale at which the O(N ) symmetries are broken. The vacuum energy remains zero however and not proportional to σ 3 as is the naive expectation. one could write down the eﬀective theory for energy scales below σ and integrate out the physics above that scale.D. One would have argued that one would have had a low energy eﬀective Lagrangian for the pions and dilaton . in the situation where σ = 0 and the scale invariance is spontaneously broken. There are now two possible scales. with Lagrangian: L = ∂µ φ1 · ∂ µ φ1 + ∂µ φ2 · ∂ µ φ2 − λ6. The ﬁnal example will be the O(N ) × O(N ) model with two ﬁelds in the vector representation of O(N ). D − 2(N − 1)π s. the vacuum energy always vanishes.S. (6.8) 3 Again the β functions vanish in the strict N → ∞ limit. indicates spontaneous symmetry breaking of scale invariance and V is the vacuum energy. Berman and E. Note. summarized in the table below: f (g) > 0 f (g) = 0 f (g) = 0 σ 0 σ =0 σ =0 S.5) (6. σ 0 N.9) Again.0 |φ1 |2 −λ4. |φ2 |2 (6.

from which this “comics illustrated” is derived is yet to be written. Our tools for this project will be. It was proposed that the underlying physics of nature is scale invariant and the scale invariance is only removed by spontaneous symmetry breaking [83. This is not true. The scaﬀolding for this construction will be extended objects called branes. Many properties of gauge theories are obtained in this fashion [93. If there are many of them they may mimic a very small cosmological constant [87]. Gauge Theory and Strings with a vacuum energy given by the scale at which the global symmetry is broken. Magnetic monopoles and vortices are examples of solitonic conﬁgurations . This would explain the vanishing of the cosmological constant. (The data in 2001 seems to indicate the presence of a small cosmological constant.1 Branes in string theory Branes are extended object solutions which emerge non perturbatively in string theory in a very similar way that solitons emerge in ﬁeld theory. – with N = 1 supersymmetry (SUSY). – with SU (NF ) × SU (NF ) × U (1) global symmetry. 7. The hope is that this dilaton may somehow be given a mass with a further independent (Higgs like?) mechanism. – with U (Nc ) gauge symmetry. • Identify dualities in a pedestrian way... Such theories may have many superselection sectors (such as magnetic monopoles in N = 4 super Yang-Mills). The total energy of the universe can be augmented even in a scale invariant theory. In the following discussion we will have as our goal to: • Construct: – a D = 4 dimensional eﬀective theory. 7 Supersymmetric gauge theories and string theory We now view the supersymmetric gauge theories from a diﬀerent point of view.) A key problem with this scenario however is that we do not observe a dilaton in nature that would be expected if scale invariance is spontaneously broken.210 Unity from Duality: Gravity. In each sector the total energy will diﬀer from zero by the energy of these particles. the vacuum energy remains zero. 94]. 84] and also [78. The unabridged original novel. 85]. We obtain them as the low energy limit of various string backgrounds. comic strips [95].

2-forms. consisting of one-form and three-form vector potentials.µp+1 . denoted by Aµ and Aµν . The massless spectrum of the open bosonic string consists of photons Aµ . In type-IIB the NS-NS sector also consists of Gµν . in which they can condense and thus take over the control and drive the infra-red dynamics. 2. The RR sector consists of vector potentials which are 0-forms. Φ. whose tension 1 T is: TNS−NS = g2 l6 .. namely Gµν . such regulators have granted a very rapid access to identifying non-perturbative features. In lattice gauge theories. Φ. The exact background corresponding to this conﬁguration is not yet known as the dilaton background seem to contain singularities. µ = 1. The bosonic sector of type-IIA theory consists of two sectors. that in the NS-NS sector.µ8−p−1 . . and a dilaton.. there existence answers a yearning to search for more than meets the eye (or the equations). In the RR sector. there are circumstances. The ﬁrst is called NS-NS sector and consists of the same spectrum as that of the bosonic closed string theory. there are . Similar consequences will occur here. the spectrum consists of a graviton. . the branes will basically serve as regulators for some ﬁeld theories. and the electric ﬁeld Eµ1 . In the context we will discuss here. Bµν . also answers positively the important question of the existence of a magnetic monopole. . In the closed bosonic string theory. we ﬁrst review the spectrum of massless particles in various string theories [97]. where gs is the string coupling and ls is the string scale. there exist solitons. in strong-coupling approximations. and 4-forms.. respectively. derived from the p-form A. Bµν .S. . What about the objects in the RR sector? First we note. To appreciate that. The other sector is called the RR sector. carrying two and four indices. Berman and E. denoted by Gµν . a 5-form in type-IIA and 4-forms and 6-forms in RR sector of type-IIB. It. derived from the (6-p)-form A lead to further vector potentials.. Gravitons and the other particles in the NS-NS sector have a perturbative string realization.. What are they good for? First. the regulator will also learn how to behave from the very theory it regulates. of course. . Rabinovici: Supersymmetric Gauge Theories 211 in gauge theories.D. . playing the same role as string sizes and lattice cut-oﬀs. To each of them is associated an “electric ﬁeld”. A yearning. Φ. duality relations in ten dimensions between the electric ﬁeld Eµ1 . s s These solitons spread over ﬁve space and one time dimensions. 7. However. an anti-symmetric tensor. Bµν .2 Branes in IIA and IIB string theories For some years it has been known that solitons exist in the low-energy effective theory of superstring theory/supergravity [96].. These excitations are often very heavy and have little direct impact on the low-energy dynamics. at this stage. as well as that of other interesting topological excitations. respectively. In string theory. 24. which seems to be engraved in at least part of our community. discovered in supergravity. In addition. such as conﬁnement.

1) contains on its right-hand side the generators of translation. The D-brane solitons in this sector are denoted Dp. . Thus it is not clear a priori if any of the supersymmetric generators. ±Γ0 . It is easier to answer the second question. will survive intact. This is our ﬁrst comic strip (Fig. The massless sector on the NS5 in the type-IIB theory is a non-chiral system consisting of spin-1 particles. Qβ j } = α {Qi . In the presence of the NS5 conﬁguration sixteen (half) of the SUSY charges survive. respectively. The solitonic sector of closed string theory contains conﬁgurations of open strings. residing in the left-hand side of the algebra. x1 . The coordinates x0 . It p+1 s s turns out that open strings. the NS5 brane. and their supersymmetric partners. . in ﬂat space-time. ﬁve scalars. Qβ j } = 0 (7. . The D stands for Dirichlet. The p + 1 dimensional solitons have a tension which is: TRR = g l1 . The other coordinates xp+1 . existing in ten dimensions. x9 are ﬁxed. Qj } = α β µ 2σαβ Pµ δ ij ¯ ¯ { Qα i .. The supersymmetry algebra: ¯ {Qi .212 Unity from Duality: Gravity. and the Γ-matrices are respective ten-dimensional Dirac Γ-matrices. spin-0 particles. . The translational non-invariance of the Dp-brane . and their supersymmetric partners. The answer to the ﬁrst question is more complicated. It turns out that. These are called D-branes. obeying appropriate Dirichlet boundary conditions. whose extension is in p+1 space-time dimensions breaks translational invariance in 9-p directions. One may also inquire what are the supersymmetry properties of that theory. Γ5 L R. One may inquire as to the eﬀective theory on the six-dimensional world volume of the NS-NS soliton. . A soliton.3) where the sign ± corresponds to type-IIA and type-IIB. xp are unconstrained and span the brane’s space-time p + 1 dimensional world volume.2) where L and R obey the constraints: L R = = Γ0 . 13). Let us return now to the RR sector. . . The massless sector on the NS5 in the type-IIA theory is a chiral system consisting of a self-dual anti-symmetric vector potential. . may end on these branes. They are of the form: L QL + R QR (7. where p denotes the number of spatial dimensions of the branes world volume (the spatial volume in which the brane extends). the system has 32 SUSY charges. Γ5 (7. .. Gauge Theory and Strings solitons as well [92].

p T % open strings end on brane 213 −→ xp+1. . and their spin.D. there are sixteen surviving supercharges.1 superpartners.. Newton’s constant. (7. spin-1 massless particles.3 The eﬀective ﬁeld theory on branes The eﬀective theory on the Dp-brane can be identiﬁed in more detail in this case. .4) Thus. Note that. . Similarly. The eﬀec2 tive theory is invariant under local U (1) gauge transformations (Fig. .. For Dp-branes R and L in the surviving supersymmetric charge (Eq. and the gauge symmetry has increased to U (1)NC . It can be shown to contain. Thus GN s “M ” vanishes at weak coupling for Dp branes. Rabinovici: Supersymmetric Gauge Theories D-brane $ x0. GN . one can estimate the eﬀective gravitational 2 coupling GN “M ”.S.. This theory still contains sixteen conserved supercharges. 9-p massless scalars (corresponding to the Goldstone Bosons resulting from the spontaneous breaking of some of the translational invariances). is proportional to gs .2)) obey the constraint L = Γ0 . one can construct a conﬁguration containing NC parallel Dp-branes. 13. 14). “M ” is 1 1 proportional to gs for Dp branes and to g2 for the NS5 brane.1. . 7. Berman and E. unaltering the large distance geometry.. reduces the number of SUSY generators by half for any value of p.. for any p. A Dp brane with an open string ending on it. in addition to the sixteen conserved supercharges.9 Fig. Γp R. (7. The next comic strip describes this conﬁguration for NC = 4 (Fig. applying naively Newton’s law.. 15).

The eﬀective ﬁeld theory describing this D4 brane conﬁguration is a U (4) gauge theory. the mass of a particle has a very simple geometrical meaning. will lead to the emergence of massless . A parameter in ﬁeld theory. Imagine now bringing the parallel branes together (Fig. according to the above relation.214 Unity from Duality: Gravity. A massless state propagating on a Dp brane. It turns out that the masses of the particles can be directly associated with the minimal distance between the end points of the strings stretched between diﬀerent branes. 14. Four massless states (corresponding to open strings ending on the same brane) and six massive states (corresponding to open strings ending on diﬀerent branes) are shown. For example. a string stretched between the i-th and j-th brane represents a particle which has a mass mij = l1 |xi − xj |. 15. spontaneously broken down to U (1)4 . This. Gauge Theory and Strings ¨ © Fig. 16). ¨ © ¨ © ¦¥ §¤ U (1)4 ¨ © ¦¥ ¨ © ¦¥ Fig. 2 s where xi represents the value of the coordinates at which the brane is set.

Fig. Indeed. ˜ ˜ (7. The maximal gauge symmetry is U (NC ). X I ] (7. The Lagrangian describing the bosonic sector of a supersymmetric gauge theory with sixteen supercharges is of the form: L= where Dµ X I = ∂µ X I − i[Aµ . Berman and E. Four parallel D-branes. 16.p+1 is the eﬀective Yang-Mills coupling in p + 1 dimensions. X J ]2 . in a N = 1 supersymmetric gauge theory. The classical potential is ﬂat and allows for an inﬁnite set of vacua. The sixteen massless states formed are not shown in the ﬁgure.5) 4 µ ls ls gYM.5) is generic for supersymmetric theories. the potential is given by: V = (q † q − q † q )2 . the Higgs mechanism in the presence of scalars in the adjoint representation is known to conserve the rank of the group. The system has indeed the same particle content and gauge symmetries as can be inferred from Figure 15. one can show that these massless particles can enhance the gauge symmetry all the way from U (1)4 to U (4).S. only such residual gauge symmetry groups that preserve the rank are allowed according to the comic strip.D. For example.6) X I is a scalar ﬁeld in the adjoint representation of the gauge group and 2 gYM. (7. containing two multiplets with opposite electric charges. it is proportional to gs . Moreover. piled on top of each other. parameterized by those expectation values of the scalar ﬁelds X I for which the potential term vanishes. This has a correspondence in SUSY ﬁeld theory. Rabinovici: Supersymmetric Gauge Theories 215 particles. In fact. The form of the potential term appearing in equation (7.p+1 2 TrFµν + 1 1 D X I D µ XI + 8 2 Tr[X I . The minimal gauge symmetry for NC Dp-branes is U (1)NC .7) 1 2 gYM.p+1 .

the mass formula mij = 1 |x − xj | 2 i ls (7. namely the brane conﬁguration leading to an eﬀective D = 4.4 Eﬀective D = 4 dimensional systems with N = 2 supersymmetry To obtain an eﬀective D = 4 description one can either set up a single D3-brane conﬁguration in type IIB string theory or build a more complicated conﬁguration in type IIA string theory. . . these separation should be sub-stringy. this result is modiﬁed non-perturbatively [15. 7. In this context. . that is. To keep ﬁxed the mass of these “W ” particles in the limit of the decoupling of the string states (ls → 0). so the theory is perceived as a U (1)NC gauge theory on a single p + 1 dimensional world volume.216 Unity from Duality: Gravity. NC ).8) xi on the left-hand side of the equation denotes the location of the i-th brane. we have seen that U (1) symmetries remain unbroken and therefore the theory is said to be in the Coulomb phase. (7. We are by now well on our way to obtain that brane conﬁgurations will help accomplish our goals. Xii on the right-hand side of the equation denotes the component of the Higgs ﬁeld in the i-th element of the Cartan subalgebra of U (NC ). The vacua of the system consist of those expectation values for which q = q = 0. They are denoted moduli and their number is called the dimension of moduli space. the potential remains ﬂat also nonperturbatively. In each of these vacua. the system will have ˜ massless scalar excitations. In this limit. one cannot resolve the NC diﬀerent world volumes. The expectation values of the Higgs ﬁelds in the adjoint representation can be shown to have themselves a very transparent geometrical meaning: xi = Xii (i = 1. Gauge Theory and Strings It is a property of supersymmetric theories that the ﬂat potential is retained to all orders in perturbation theory.9) is just the usual mass obtained by the Higgs mechanism. The number of massless particles is at most NC (in complex notation). . the separations between the branes should vanish themselves in that limit. For some theories with a small number of SUSY charges. The Coulomb phase is that phase of gauge theory for which the force between both electric and magnetic charges is a Coulomb force. It turns out that for our goals . Returning to the conﬁguration of NC Dp-branes. N = 1 supersymmetrical U (NC ) gauge theory containing in addition matter ﬁelds. 52]. For theories with sixteen SUSY charges. This is again apparent from the geometry of the open strings ending on the brane conﬁguration.

x6 . For example. 17. x9 directions and that of the D1-brane extends in x0 . The eﬀective ﬁeld theory on the D4 brane will be four-dimensional for 1 energies much smaller than ∆x6 . in particular in type IIB string theory it can end on a D3-brane. x8 .D. x7 . that is on which branes are the D4-branes actually allowed to end. A fundamental string (F1) can by deﬁnition end on any Dp-brane. we pause now to brieﬂy discuss several types of useful discrete symmetries in string theory. is eﬀectively four dimensional. This is realized by the following conﬁguration: Fig. that is the ﬁeld theory at energy scales much smaller than 1/∆x6 . called S. Before analyzing the various eﬀective theories resulting from the diﬀerent choices of the branes on which the D4-brane ends we discuss which are the allowed “vertices”. in an N = 4 supersymmetric gauge theory in D = 4 dimensions the gauge coupling constant g is a real parameter. In this conﬁguration a D4-brane (whose world volume is 5 dimensional and of the type M 3. The candidates for “heavier” such branes would be either NS5-branes or D6-branes. One constructs a brane conﬁguration which has a world volume of the form M 3.1 × I[∆x6 ] where M 3. Rabinovici: Supersymmetric Gauge Theories 217 the latter is more useful.1 is the four-dimensional Minkowski space-time and I[∆x6 ] is an interval of length ∆x6 . Berman and E. The content of the ﬁeld theory will depend on the branes “?” on which the D4 brane ends. In the ﬁrst example we will show that a D4 is allowed to end on a D6-brane. Establishing that the D1 may end on a D3-brane.1 × I[∆x6 ] ) the two branes between which the D4-brane is suspended would be chosen such that the eﬀective ﬁeld theory has U (NC ) local gauge symmetry and N = 1 supersymmetry. The world-volume of the D3-brane is chosen to extend in the x0 . Performing what is called an S-duality transformation validates that also a D1brane may end on a D3-brane in type IIB theory. the eﬀective ﬁeld theory. S-duality is a symmetry which is familiar already in some ﬁeld theories. that is the vertex appearing in Figure 18 is allowed. For either choice.and T -dualities. The ﬁeld .S.

Type IIB string theory has similar properties with the string coupling playing the role of the gauge coupling. The second term m2 R2 reﬂects the extended nature of the string. There is evidence that the theory with coupling g is isomorphic to the theory with coupling 1/g. an interchange of n and m simultaneously with an interchange of R and 1/R in equation (7. T -duality can . has a generalization involving also the value of the θ parameter in ﬁeld theory and an additional corresponding ﬁeld in string theory. R2 (7. For a small radius R these are very low energy excitations. called S-duality. 18. In its implementation in ﬁeld theory electric and magnetic excitations were interchanged. It is the extended nature of the string which leads to this result. In particular.10) gives an indication of how T -duality works. a D3 is left invariant and a D5 is interchanged with an NS5. It describes those excitations in which the closed bosonic string extends and winds around the compact dimensions m times. This non-perturbative symmetry. All in all. it describes also a point particle in a Kaluza-Klein compactiﬁcation. T -duality is a symmetry which has aspects peculiar to string theory [98]. We have used some of these properties in the derivation above. The mass M of the particles depends on the compactiﬁcation radius through the formula: M2 = n2 + m2 R 2 .218 Unity from Duality: Gravity. is identical to another bosonic string theory whose compact dimension in string units is of radius 1/R. n2 The term R2 is not particular to string theory. the gauge coupling does not run under the renormalization group and thus diﬀerent values of g correspond to diﬀerent theories. Gauge Theory and Strings D4 D6 Fig. An F1 is interchanged with a D1. A “vertex” in which a D4 brane ends on a D6 brane.10) n/R denotes the quantized momentum of the center of mass of the string. similarly in string theory diﬀerent types of branes are interchanged under S-duality. a closed bosonic string theory with one compact dimension whose radius in string units is R. theory is ﬁnite.

19. T Dp Rj ←− Dp T −→ E Rj ←− T T E T Ri D(p-1) Ri ‘Longitudinal Duality’ N −→ D E ‘Transverse Duality’ D(p+1) D −→ N E Fig. and “transverse” dimensions. Berman and E. those dimensions which are not part of the world-volume of the brane. This indicates that it persists non-perturbatively. which are part of the world-volume of the brane. Rabinovici: Supersymmetric Gauge Theories 219 be generalized to an inﬁnite discrete symmetry and can be shown to actually be a gauge symmetry in the bosonic case. 20). Its eﬀect on a NS5-brane is more complicated and we will not need it in this lecture (Figs. In particular the transformation R → 1/R maps a type IIA string theory background with radius R to a type IIB background with radius 1/R and vice versa. T -duality involving a transverse direction transforms a Dp-brane into a D(p+1)-brane. T -duality acting on Dp branes. For supersymmetric string theories T -duality has some diﬀerent manifestations.D. 19. In the presence of D-branes one naturally distinguishes between two types of compact dimensions: “longitudinal” dimensions. D and N denote Dirichlet and Neumann boundary conditions in the compact directions.S. . respectively. A T -duality involving a longitudinal dimension will transform a Dpbrane into a D(p − 1)-brane and will leave an NS5-brane intact.

x1 and x2 to obtain a D3-brane ending on a D5-brane. By performing a T -duality along three directions transverse to both the D3 and the D1branes. we obtain a D4-brane ending on a D6-brane. The construction sketched in this proof shows that any Dp-brane can end on any D(p + 2)-brane. The proof thus rests on the validity of both S. In a somewhat similar manner one can show that a D4-brane can end on a NS5-brane (Fig. Gauge Theory and Strings NS5 T T E NS5 NS5 T −→ E complicated ¡ ¡ ¡ ¡ e e e Fig. Starting from the by now established conﬁguration of a D1-brane ending on a D3-brane in type IIB string theory. one performs T -duality along two transverse direction. The detailed action of the transverse duality is not indicated. 20. for example the directions x1 .and T -duality. and there is ﬁrmer evidence of the validity of T -duality. 22). x2 and x3 . There are many indications that the former is correct. The steps used in the proof are summarized in Figure 21. . Equipped with this information we can continue the proof of the existence of a D4 conﬁguration ending on a D6-brane. Due to the odd number of T -duality transformations. one passes from a IIB background to a type IIA background. T -duality acting on a NS5 brane.220 Unity from Duality: Gravity.

which is transverse to the newly formed D3brane but longitudinal to the NS5-brane. A combination of S and T duality transformations. A T -duality along x3 . 24). This series of S. Rabinovici: Supersymmetric Gauge Theories 221 F1 D3 −→ S D1 D3 −→ D6 06 T(1.and T -duality transformations (called U -duality) is shown in Figure 23. 21.D. 22.2. The spatial extension of the various branes is summarized in the table below. establishing that a D4 brane can indeed end on a D6 brane. Berman and E. D4 NS Fig. A D4 brane ends on a NS5 brane. Let us consider ﬁrst the eﬀective low energy ﬁeld theory related to the conﬁguration in which the D4-brane is suspended between a NS5-brane and a D6-brane (Fig. leads to the desired conﬁguration in which a D4-brane ends on a NS5-brane.S. where + and − denote respectively directions longitudinal and . An S-duality transformation brings us to a D3-brane ending on a NS5brane.3) D4 01236 0789 IIB IIB 0123789 IIA Fig.

Gauge Theory and Strings D1 D3 06 −→ T(1. A D4 brane suspended between and NS5 and D6 branes. The eﬀective ﬁeld theory contains no massless particles and is thus. . 23.and T -duality transformations establish that a D4 brane may end on a NS5 brane. S.222 Unity from Duality: Gravity. some topological theory. 24. at best.2) D5 D3 −→ 0612 S 0345 IIB 034512 IIB 0126 D3 −→ T(3) D4 0612 NS5 012345 IIB NS5 012345 IIA Fig. D4 NS5 D6 Fig.

Therefore now it can be displaced in the x4 and x5 directions. as shown in Figure 26. x8 and x9 . x5 and x6 directions. Fig. Massless degrees of freedom can also be identiﬁed in a geometrical manner. . the D4-brane is frozen.S. x7 . Each possibility to displace the D4-brane along the D6 and the NS5 branes maintaining the shape of the conﬁguration corresponds to a massless particle. x8 and x9 . x8 and x9 directions. 25). Berman and E. x7 . The D4-brane left on its own could be displaced along the directions x4 . x7 . It cannot be displaced in a parallel fashion and therefore the eﬀective theory contains no massless particles whatsoever. Rabinovici: Supersymmetric Gauge Theories 223 transverse to the relevant brane. the D4-brane is locked in the x4 . D4 D6 N S5 x0123 + + + x4 − − + x5 − − + x6 = − − x7 − + − x8 − + − x9 − + −. x5 . and = indicates that the brane is of ﬁnite extent in that direction. The eﬀective ﬁeld theory is N = 2 SUSY U (1) gauge theory. It is at best a topological ﬁeld theory. x6 . 25. On the NS5 side the D4 is locked along x6 . We thus turn to another attempt to build an eﬀective four dimensional ﬁeld theory along the suspended D4-brane. We now suspend it between two NS5-branes which are extended in the same directions as the NS5 of the previous example [95] (Fig. suspended between the D6 and the NS5 it allows no massless degrees of freedom to propagate on it. On the D6 side.11) The eﬀective ﬁeld theory should contain the massless degrees of freedom of the system. Although the D4-brane would have allowed the propagation of 5-dimensional photons had it been left on its own. the D4-branes is locked in the x6 .D. (7. At both ends. All in all. A D4 brane suspended between two NS5 branes.

99]. The D4 brane can be parallely displaced along the x4 and x5 directions. Thus the conﬁguration above describes an eﬀective D = 4. An analysis using (Eqs. 1 The actual symmetry turns out to be SU (NC ). 26. Before pursuing this. (7. x5 directions to the x8 . Rotating one of the NS5-branes from the x4 . x9 directions will lead to the desired D = 4. N = 2 supersymmetry implies the existence of massless spin. (7. That multiplet consists of spin 1. This corresponds to two massless spin-0 particles appearing in the eﬀective ﬁeld 1 theory. . implementing an N = 2 supersymmetry in the eﬀective four dimensional theory. 27). The eﬀective four dimensional gauge coupling constant is related to the eﬀective ﬁve dimensional gauge coupling constant in the usual Kaluza-Klein manner: 2 gYM. U (1) supersymmetric gauge theory. U (NC ) eﬀective gauge theory. 8 supercharges survive as symmetries. As in the single brane case. That is.3)) and (Eq. The two scalar particle identiﬁed geometrically form part of the N = 2 vector multiplet. (7.4)) shows that in this conﬁguration one half of the supersymmetry generators of the single brane conﬁguration survive.5 · ∆x6 (7.224 Unity from Duality: Gravity.4 . the gauge symmetry can be enhanced to U (NC ) by suspending NC D4-branes between the NS5-branes1 (Fig. N = 2.2). it will be useful to study the eﬀective ﬁeld theory on a D4-brane suspended between two D6-branes (Fig. 28). This is discussed in [93. N = 1. spin 1/2 and spin 0 particles.4 = 2 gYM. Thus the eﬀective ﬁeld theory contains at least two massless spin 0 particles.12) 2 Changing the value of the separation ∆x6 amounts to rescaling gYM.2 and spin-1 particles as well. Gauge Theory and Strings Fig.

x8 and x9 . It contains no massless gauge particles. These correspond to the allowed translations in the directions x7 . At the case at hand.S. that is the low energy eﬀective theory is not an unbroken gauge theory. The eﬀective ﬁeld theory is a D = 4. An N = 2 supersymmetric multiplet would require either two or four spin 0 particles. the multiplet is actually the N = 2 hypermultiplet which contains four spin 0 and four spin 1/2 degrees of freedom. later we will show that . D4 D6 D6 Fig. Actually. Similar considerations to the ones used above show that the eﬀective theory contains at least three massless spin 0 particles. Rabinovici: Supersymmetric Gauge Theories 225 D4 }NC NS5 NS5 Fig. N = 2 SUSY ﬁeld theory. (7. It is important to note that the system contains no massless spin 1 degrees of freedom. Berman and E. 27. N = 2 U (NC ) SUSY gauge theory. containg an N = 2 matter hypermultiplet. 28. The spatial extension of the branes is: D4 D6 x0123 + + x4 − − x5 − − x6 = − x7 − + x8 − + x9 − +.D. The eﬀective ﬁeld theory on the D4 brane is a D = 4.13) This conﬁguration also has 8 surviving supercharges and thus the eﬀective ﬁeld theory has N = 2 supersymmetry in D = 4.

U(NC ) gauge theory with matter We construct now the conﬁguration leading to N = 1 supersymmetry. x9 directions corresponds to adding an inﬁnite mass term to the scalar ﬁelds in the adjoint representation. x5 directions. 31). The fact that the geometrical considerations were not suﬃcient to identify all four spin 0 particles shows us one of the limitations of the simple geometrical analysis. This agrees with a description by an eﬀective D = 4. as can be seen from the by now familiar geometrical considerations. This is similar to what was described before in the case of the separations of NC inﬁnitely extended parallel Dp-branes. 7. At this stage of understanding of the gauge theory–brane correspondence. The rotation leads to a conﬁguration with 4 surviving supercharges (Fig. N = 1. The complex number of moduli is immediately read out of the geometrical picture to be NC . The fourth spin 0 particle can be identiﬁed in this case with the component of a compactiﬁed gauge ﬁeld. Returning to the N = 2 U (NC ) gauge conﬁguration. The ﬁnal ingredient needed is to add some ﬂavor to the eﬀective ﬁeld theory. . x5 ).5 An eﬀective D = 4. This coincides with algebraic analysis determining the complex number of massless spin 0 particles surviving the breaking of the gauge group. one notes that a separation of the D4-branes along the directions x4 .226 Unity from Duality: Gravity. The rotation of one of the NS5 from x4 . one ﬁnds parameters in the brane picture with no clearly known ﬁeld theoretical interpretation and vice versa. The gauge symmetry can be enhanced to U (NC ) by suspending NC D4-branes between the two diﬀerent NS5-branes (Fig. 30). x5 to the x8 . The eﬀective four dimensional theory is a U (1) gauge symmetry and has no moduli. Gauge Theory and Strings this conﬁguration will be part of the description of the Higgs phase of the supersymmetric gauge theory. a rotation by diﬀerent angles would have given rise in ﬁeld theoretical language to a ﬁnite mass term. In Figure 29 we summarize the complex number of massless spin 0 particles appearing in any of the four conﬁgurations discussed until now. The mass of the W particles is again proportional to ∆D4(x4 . This is done by distributing NF D6-branes along the x6 direction (Fig. N = 1 supersymmetric U (1) gauge theory. namely A6 . 32). x5 leads to all rank preserving possible breakings of the gauge symmetry. which denotes the separation of two D4-branes in the x4 .

4.14) .D. The spatial extension of the various branes is: x0123 + + + + x4 − − + − x5 − − + − x6 = − − − x7 − + − − x8 − + − + x9 − + − +. Rabinovici: Supersymmetric Gauge Theories 227 D4 D6 2 D6 N=2 Hyper-multiplet D4 NS5 1 NS5 Vector SUSY U(1) D4 NS5 0 D6 ‘Topological’ Fig.S. Berman and E. 29. A summary of the particle content of the low-energy eﬀective theories described in Section 7. D4 D6 N S5 N S5 (7.

The NS5 extending also in the directions x8 . The matter ﬁelds appear as representations of diagonal vectorial subgroup of the ﬂavor group SU (NF ) × SU (NF ) × U (1). 31.228 Unity from Duality: Gravity. £ £ NS5 45 £ £ £ £ £ £ ££ £ D4 }NC NS5’ 89 Fig. N = 1 U (NC ) SUSY gauge theory. The eﬀective ﬁeld theory is thus a four dimensional N = 1 U (NC ) supersymmetric gauge theory which has matter in the fundamental representation of the vector subgroup of the ﬂavor symmetry. Touching the D6-branes endows the open strings with ﬂavor. The conﬁguration leading to a D = 4. x9 is labeled NS5’. Gauge Theory and Strings ££ £ D4 £ £ NS5 45 £ £ £ £ £ £ NS5’ 89 Fig. Geometrically they are associated with strings connecting the D6-branes with the D4-branes. In [93] the reader will ﬁnd references for attempts to construct chiral conﬁgurations and to identify the full ﬂavor group. N = 1 U (1) SUSY gauge theory. and touching the D4-branes endows them with color. Their directionality is responsible for ¯ the appearance of both the NF and NF . 30. The masses of the squarks have a geometrical interpretation: they are proportional to the distance between the D4 and D6-branes along the x4 . The way to obtain the Seiberg dual conﬁguration is essentially to move the position of the NS5-brane residing on the left hand part of the . x5 directions. The conﬁguration leading to a D = 4. This is as close to our goal as we will reach in this lecture.

S. NF color singlet massless particles. 32. NC ) £ £ £ £ £ £ £ £ £ £ £ £ NS5’ D6 D6 D6 D6 D6 ' NF ˜ NC = 2 NS5 Fig. This is the dual conﬁguration to that of a N = 1 SUSY gauge theory with NC = 3 and NF = 5 (Fig. The conﬁguration shown has NC = 5 − 3 = 2. . conﬁguration all the way to the right hand part. 33. Open strings ending on D4 and D6 branes carry both color and ﬂavor. The resulting new conﬁg¯ uration will have SU (NF − NC ) gauge symmetry and NF and NF colored 2 matter as well as NF . Berman and E. Rabinovici: Supersymmetric Gauge Theories ££ £ s d d d d 229 £ £ £ £ £ £ £ £ NC D4 NS5 D6 D6 NS5’ Fig. one needs three more pieces of “information” as to the behavior of branes. The NF ﬁelds are essentially the color-singlet particles appearing in the eﬀective ﬁeld theory corresponding to the brane conﬁguration (Fig.6 More pieces of information In order to be able to perform this displacement of the NS5-brane. in addition. 39). 33). which is exactly the 2 result of Seiberg (Fig. N £ £ £ £ £ £ c c i £ q £ £ £ £ §£¤ £ £ £ £ §£¤ §£¤ £ NF = 5 £ £ §£ ¤ §£ ¤ §£ ¤ £ £ §¤ §¤ §¤ §¤ £ £ £ £ £ £ £ £ £ s d £ d ˜ £ £ £ £ £ £ (NF . E F) ¯ (NF . 2 NF = 5 and. Adding D6 branes allows the existence of matter in the eﬀective ﬁeld theory. 28). ﬂavor adjoint color singlets.D. 7.

A change in the relative position in the D4 and D6-branes in the x4 . Both distances are indicated in the ﬁgure. x5 ) mQ(Q) ˜ T x4 . x5 ) mWi NS5 D6 × s d d D4 d d D4 NS5 ∆D4. x5 ). They are proportional to the distances ∆D4. 34).230 Unity from Duality: Gravity. x5 ) and ∆D4. 34. x5 directions led to a change in the masses of the squarks (it is actually equivalent to a smooth change in the expectation value of the scalar coupled to fermions by a Yukawa coupling) (Fig.D4 (x4 . x5 E x6 Fig. D4 T D4 c ∆D4. respectively.D6 (x4 . x5 directions of the D4-branes in the N = 4 and N = 2 supersymmetric conﬁgurations led to a smooth change of the mass of the W ’s (or equivalently to smooth changes in the Higgs expectation values). Gauge Theory and Strings 1) “Motions of branes” Changes of the positions of branes correspond in some cases to smooth changes in ﬁeld theoretical parameters.D4 (x4 .D6 (x4 . Varying these distances smoothly changes the parameters of the ﬁeld theory. . The values of the masses of the W and squark particles is encoded in this ﬁgure. For example. a change in the relative displacement in the x4 .

35. 36). the two branes have to intersect. Berman and E. has no clear impact on the ﬁeld theory description). Consider an N = 2 . Rabinovici: Supersymmetric Gauge Theories 231 A change in the relative distance between the NS5-branes corresponded to a change in the coupling of the gauge theory on the D4-branes suspended between them.D. x9 ) Fig. Also the Higgs phase has a simple geometrical realization in terms of branes. 3) The brane realization of the Higgs phase We have seen that the Coulomb phase of N = 4 gauge theories is spanned by the separation among the D4-brane. However once they do. What actually can be shown to occur [93. 95] is that as the NS5-brane “crosses” the D6-brane a D4-brane is created and suspended between the NS5 and D6-brane. x8 . 35). Note that there is no particle content to the eﬀective theory on that D4-brane and thus no new degrees of freedom are created as the D4-brane is formed. x8 and x9 directions of the positions of two NS5-branes corresponds to adding a FayetIliopoulos term in an N = 2 ﬁeld theoretical interpretation (Fig. 37). x9 T NS5 × © × NS5 E x6 ∆N S5 (x7 . the relative separation in the x6 direction. The same is true the N = 2 case. 2) Supersymmetry restrictions Supersymmetry allows that only a single D4-brane can be suspended between a NS5-brane and a D6-brane [93. x7 .S. A similar interpretation can be given for the separation along the x7 direction between the NS5 and NS5 in the N = 1 case supersymmetric case. Eventually. Consider a conﬁguration consisting only of a NS5-brane and a D6-brane separated in the x6 direction (Fig. x8 . 95] (Fig. The N = 1 case has no Coulomb phase. The NS5 branes can be separated in the x7 . A change in the relative position along the x7 . x8 . x9 directions. Each such separation corresponds to appropriate components of the Fayet-Iliopoulos term in N = 2 SUSY U (NC ) gauge theory. Displacing the NS5brane for example in the x6 direction is a smooth motion as long as the two branes do not intersect (this is an example where a brane parameter. it is clear that something more singular may result. In other cases changes in the position of branes could have more abrupt consequences.

one starts the journey by setting the D = 4 N = 1 U (NC ) supersymmetric gauge theory with NF ﬂavors in the Higgs phase. as well as comparing the numbers of the relevant operators. In order to be able to trust the pictures as one moves the branes around it is advisable to be in a weak coupling situation as long as possible. Retaining the Higgs phase. This result is true also for the N = 1 conﬁgurations. For the N = 1 case one needs also to realize that the eﬀective theory on a D4-brane suspended between a D6-brane and a NS5 -brane contains one massless complex spin 0 ﬁeld. one is able not only to reproduce the particle content required by Seiberg’s duality. However the eﬀective ﬁeld theory on each of the D4-branes contains no spin 1 massless particles. The D4 brane carries no massless degrees of freedom. from left to right in this ﬁgure. To enforce that. x5 directions maintains the supersymmetry. but also his analysis of comparing the dimensions of the moduli space at both ends of the duality transformation. The separation of the D4-branes on the D6-brane along the x4 . One can show that this breaking indeed represents the Higgs phenomenon. Gauge Theory and Strings £ £ NS5 45 £ £ £ £ £ £ £ £ E ' D6 789 D4 E£ £ £ £ £ £ £ £ £ £ ∆x6 D6 NS5 Fig. 36. x5 directions being proportional to the expectation value of the scalar ﬁeld responsible for the Higgsing. A D4 brane is created when the NS5 brane crosses the D6 brane. the separation among the two D4-branes along the x4 . one massless spin 1/2 ﬁeld and no massless spin 1 ﬁelds. The brane geometrical picture has a classical weakly coupled taste.232 Unity from Duality: Gravity. One is also able to show in the brane picture as Seiberg has done in ﬁeld theory that masses as one end of the duality pair correspond to expectation values on the other side. 38). . conﬁguration in which in addition of having a D4-brane suspended between two NS5-branes a D6-brane is inserted between the NS5-branes (Fig.

Berman and E. 69. 37. Conﬁguration b) is disallowed. 93]). This enabled an easy visualization of the continuous features of the duality transformation. However. I will brieﬂy mention the diﬀerent stages of the journey. 2 An attempt is made to digitalize the movie.S. which has NF = 5 and NC = 3.D. . As a navigational aid. First stage: the NS5-brane is displaced across the D6-branes respecting the rule mentioned in 1) above concerning D4-brane creation. SUSY allows only the existence of the conﬁguration a).7 Obtaining the dual ﬁeld theory At this stage in the lecture a movie2 composed out of the comic strips was shown. conﬁguration b) is allowed. 7. Rabinovici: Supersymmetric Gauge Theories £ ££ 233 D4 £ £ £ £ £ £ £ NS5 a) £ £ D6 ££ D4 £ £ £ £ £ £ £ D4 £ NS5 b) £ £ D6 Fig. The starting point is a conﬁguration depicted in Figure 39. if the NS5 is exchanged with an NS5’. The reader is actually equipped by now to embark on this journey on her/his own (or by reading [68.

28). the appropriate reconnections of the branes shows that the conﬁguration represents the Higgs phase of a U (NF − NC ) gauge theory.2 degrees of freedom. The Higgs phase of ﬁeld theory is realized by the D4 branes connecting the two NS5 branes. the branes are separated along the x7 direction before they are made to coincide in the x6 direction. corresponds in some sense [68. as stated in 1) above. In order to avoid that. The eﬀective ﬁeld theory in this example carries no massless degrees of freedom at all. This indicated distance of separation corresponds to the expectation value of the Higgs matter ﬁeld. but no massless spin-1 degrees of freedom. Gauge Theory and Strings NF = 1 D4 NC = 1 NS5 D6 NS5 D4 T D4 c Q NS5 D6 NS5 Fig. as appropriate for a Higgs phase (Fig. . 69. Third stage: it is realized that if the NS5-brane will directly collide with the NS5 brane the gauge coupling will diverge. Second stage: the system is driven into the Higgs phase according to rule 3) above. This separation. At this stage. In the presence of additional D6 branes. 93] to the turning on of a Fayet-Iliopoulos coupling in ﬁeld theory.234 Unity from Duality: Gravity. the eﬀective theory would 1 contain massless spin-0 and spin. leaving the protected weak coupling regime. breaking and separating on the D6 brane. 38.

7. the gauge theory has NF NC = 5 − 3 = 2 colors. A more detailed analysis of the resulting massless particle spectrum (chiral ring) and its properties is needed to conclude infra-red duality [68. 69. Rabinovici: Supersymmetric Gauge Theories ' £ ££ D6 NC = 3 NF = 5 E 235 £ £ £ £ £ £ £ £ £ NS5 D6 D6 D6 D6 D6 NS5’ Fig. 39.1 massless particles. What has emerged naturally is 2 one gauge group out of another as well as the presence of NF color singlets. Taking a bus ride from Chamonix to Les Houches does not constitute a proof that the two cities are the same. In addition. 93]. NC = 3. Berman and E. This is the realization of the dual the2 ory obtained by Seiberg. O(N ) and product gauge groups can be constructed as well [68. Fourth stage: rejoining the branes in the x7 direction requires some quantum adjustments which we will indicate below. Other groups It can be shown that Sp(N ).D. N = 1 SUSY. . 69. That does not really demonstrate that the two are identical in the infra-red or at all.S. The ﬁnal outcome is shown = 5 ﬂavors and in Figure 33. the theory has 52 singlet complex spin-0 and spin. The result here is simply the outcome of the NS5 moving according to the brane rules from the left of the NS5’ to its right. A D = 4. NF = 5 conﬁguration. 93]. The simple procedure allows one to obtain the infra-red dual of many other brane conﬁgurations.8 Concluding remarks Infrared duality What has actually been shown is that in a continuous motion starting from one N = 1 supersymmetric gauge theory one reaches another.

. Due to the repulsive force between the D4-branes. a stable conﬁguration will form for any ﬁnite value of x3 . M-theory context The brane–ﬁeld theory correspondence obtains even more geometrical features once embedded in M-theory [93. say x3 . that is. these are additional postulates they have however immediate consequences in allowing a uniﬁed description of D = 4 and D = 3 gauge theories with 4 supercharges. 0 ≤ x3 ≤ 2πR. One could ascribe this to a quantum repulsive force between two D4-branes on the same side of a NS5-brane in a non-BPS conﬁguration. Thus a four dimensional IIA system corresponds to a type IIB system with vanishing radius. The agreement between the dimensions of the moduli spaces of the two dual models was obtained in the presence of a quantum attractive interaction between two D4-branes on the opposite sides of a NS5-brane. Only an inﬁnite type IIB radius will correspond to an eﬀective three dimensional gauge theory. Quantum corrections The classical brane picture needs to be amended by quantum considerations [95. 93]. and due to the compact nature of the x3 coordinate. The resulting system is a type IIB string theory whose third direction has radius 1/R. 99. This is indicated in the brane picture in the following manner: compactify one dimension. the three dimensional gauge theory has no vacuum.236 Unity from Duality: Gravity. For any ﬁnite radius. However the same system in D = 3 has no ground state [101]. Once the IIB radius is inﬁnite. recall that two D4-branes were not allowed to be suspended between an NS5 and a D6-brane. the type IIB eﬀective theory corresponds to a ﬁeld theory with more than three dimensions in type IIA. such a vacuum state exists for any eﬀective supersymmetric gauge theory with no ﬂavors in more than three dimensions. At this stage. 100]. consider the D = 4 N = 1 U (NC ) gauge theory with no ﬂavors. 69. For example. 52]. of the world-volume of the D4-brane. Such a system has a ground state [15. 99]. For example. Gauge Theory and Strings Generalized infra-red dualities By allowing conﬁgurations with k NS5-branes and k NS5 -branes one can test new and old generalizations of Seiberg’s duality in the presence of a richer matter content and various Landau-Ginzburg-like interactions [68. the equilibrium state exists no more. It is more convenient to perform a longitudinal T -duality along the x3 direction.

Eksp. 40.P. acknowledges the inspirational atmosphere created by the groups of the Hebrew University and the Weizmann institute. Kutasov and A. . Giveon.D. [3] J.A. acknowledges S. 8 Final remarks In this set of lectures we have had a panoramic vista of the rich structure of supersymmetric gauge theories. Forge. The qualitative understanding of the phase structure of such systems will proceed their complete quantitative analysis. Elitzur. 13 (1971) 452]. [2] J. E. 1990). Berman and E. West. 2nd edition (Princeton University Press. A. B 49 (1974) 52. References [1] Y.B. A. D. Introduction To Supersymmetry And Supergravity.R. Teor. Supersymmetry and Supergravity. Rabinovici: Supersymmetric Gauge Theories 237 gs Black Holes SUGRA 1 v HOLOGRAPHIC BOUND STRINGS V E Fig. Zumino. Likhtman. The references given in this article though detailed are not complete. [4] P. As a prototype for a phase diagram of gravity we propose the following Figure 40. Lett. Wess.C. Singapore (Singapore: World Scientiﬁc. JETP Lett. Perhaps a similar course can be followed for theories whose symmetry includes general coordinate invariance in analogy to Figures 4. Golfand and E. 1992). Bagger and J. Fiz. Wess and B. D. 425. Phys. A prototype of a phase diagram for a gravitational theory. The classiﬁcation of possible phases of gauge theories were understood well before concepts like conﬁnement could have been analysed analytically in four dimensional continuum theories. [102].S. Schwimmer for collobartions and many very valuable discussions. 13 (1971) 323 [Pisma Zh. 5 and 8.

Kogut. Rev. D. Rept. A.H. J. Nucl.P. Rept. Rocek and W. Phys. D 19 (1979) 3698. P. Ferrara. Rev. Nucl. Furstenau. Nucl. Van Proeyen. 32 (1977) 249.x. S. Ooguri and Y. D 32 (1985) 927. Amaldi. Gates. Nucl.T. Kallosh. Suppl. E. U. Horn. Elitzur. Kallosh. Phys. B 245 (1984) 17. Suppl. D 20 (1979) 2619. O. The Quantum Theory Of Fields. Guth. N. Rev. B 255 (1985) 383. 68 (1998) 261 [arXiv:hep-th/9706168]. S.lns. Oz. Nucl. Kelley and D. Rev. Math. Rabinovici and A. Phys. See for example. Grisaru. P. V. Phys. M. Lett. Phys. Weinberg. S. Rabinovici. Myerson and J. K. Rev. Lett. B 318 (1993) 469. J. Schwimmer. H. Witten.M. 62 (1998) 341. B 119 (1982) 165. P. V. de Alfaro. For a review see. Figueroa-O’Farrill. Rev. Phys. Front. P. 58 (1983) 1 [arXiv:hep-th/0108200]. J. Introduction to Supersymmetry http://www. Suppl. O. Phys. Peskin and S. Nilles. Phys. 2000).. Susskind.J. Rev. UK: Univ. Lett. Proc. Rabinovici and A. For a review see. Rev. Kumar. Intriligator and N. .R. L. Akulov and A. Nucl.K. Aharony. Banks. Pearson and J. Pashnev. Phys.cornell. Sonnenschein. Nuovo Cim. S. Phys. Mat. Phys. B 129 (1977) 493. B 260 (1991) 131. Black holes and quantum mechanics [arXiv:hep-th/9902007]. T. Yankielowicz. M.P. R.E. Phys. Claus. M. Vol. Ukawa. 419. Nucl. B 58 (1975) 67. Shigemitsu. Lett. J. Townsend and A. D 44 (1991) 817. Seiberg. Theor. D 19 (1979) 3715. Weinberg. Phys. Phys. Nucl. P. E. Phys. Phys. Rabinovici. Weinberg.M. Kutasov and N. W. D 19 (1979) 1277. R. Forge and E. B 358 (1991) 600. Furlan. D. Proc. B 202 (1982) 253. Nanopoulos. J. Maldacena. S. 56 (1983) 862. Pr.V. O’Raifeartaigh. D 21 (1980) 1013. Phys.I. B 106 (1976) 219. Fayet and J. 110 (1984) 1. R. Lett. Fubini and E. Siegel. Windey and A. N. Fiz. Phys. Phys. Seiberg. Luo. Iliopoulos. Phys. S. Gauge Theory and Strings S. Rev. Nucl. 323 (2000) 183 [arXiv:hep-th/9905111]. S. Phys. P. Phys. Illiopoulos. B 188 (1981) 513. BC 45 (1996) 1 [hep-th/9509066]. Phys. Weinstein and S. Proc. BUSSTEPP lectures on supersymmetry [arXiv:hep-th/0109172]. P.B.B. L. Banks. E. 3: Supersymmetry (Cambridge. S.C. Fayet and S. D 52 (1995) 6157 [arXiv:hep-th/9507013]. Private communication. Nucl. D 13 (1976) 974. Phys. M. Phys. Seiberg. A. S. Rept. Parga.238 [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] Unity from Duality: Gravity. B 96 (1975) 331. Teor. T. Aharony. West. Rev. Phys. B 51 (1974) 461. Derix. Fubini and G. Phys.S. B 260 (1991) 447.edu/ argyres/phys661/ P. 56 (1983) 344. H. P. Ellis. 81 (1998) 4553. Phys. Fayet. de Boer and H. Elitzur. E. R. A 34 (1976) 569. Langacker and M. Phys. Nucl. Phys. Yankielowicz. M. Phys. Witten. Gubser. Phys. Schwimmer. Lett. Lett. Fayet and J.

Witten. Rept. Khoze and M. 23 (1976) 245. E. JHEP 0010 (2000) 011 [arXiv:hep-th/0009009]. A.P. Forge. Casher. B 459 (1996) 160 [hep-th/9509130]. Taylor. Rabinovici. Phys. D 19 (1979) 3682. Seiberg. Phys. Elitzur. [55] S.C. Nucl. D.C. R. Seiberg and E. Nucl. Kutasov. Nucl.G. Abbott and E. Nucl.L. B 125 (1983) 487. [49] N. [66] G. Phys. [68] S. in C79-08-26. Giveon. N. [47] G. B 336 (1994) 409 [hep-th/9407106]. Ahrenshoop Symp. Berman and E. G. Elitzur. Rev. [50] A. [62] S. [59] K. A. Giveon and E. [44] E. Elitzur. Lett. Intriligator and N. Mandelstam. France. B 241 (1984) 493. Rabinovici and A. K. B 426 (1994) 19. Hollowood. Nucl. Phys. ’t Hooft. Giveon. A. Lett. Olesen. [60] K. B 83 (1979) 395. . [65] A. Nucl. B 431 (1994) 484 [arXiv:hep-th/9408099]. B 448 (1995) 93 [hep-th/9505062]. [54] K. Lett. Rev. A. 1979). Fradkin and S. Intriligator.H. Phys. Lett. Phys. [51] T. Forge. [69] S.D. Amati. 430 (1994) 485 [arXiv:hep-th/9407087]. B 444 (1995) 125 [hep-th/9503179]. Phys. [57] S. [63] T. Nucl. [67] N. A. Meurice. D 50 (1994) 1092 [hep-th/9403198]. Intriligator and N. [45] H. Rossi and G. B 86 (1979) 283. Veneziano. [53] I. B 256 (1985) 557. [64] L. Rabinovici. The calculus of many instantons [arXiv:hep-th/0206063]. B 400 (1997) 269 [hep-th/9702014]. Nucl. Elitzur. Intriligator. Rabinovici. Giveon and E. Phys. Farhi. M. Phys. A. B 379 (1996) 121 [hep-th/9603051]. (1995) 174 [hep-th/9512140]. [58] K. Erratum-ibid. Seiberg. Phys. Intriligator. Phys. Lett. Intriligator and E. Nucl. [52] D. A. Phys.R.J. Forge. Lett. Elitzur. Veneziano and S. Argyres and M. [43] J. Aﬄeck. Bilal. Nucl.S. Cardy and E. Rabinovici. Seiberg. Leigh and N. Phys. Phys. Elitzur. Giveon and E. Konishi. Banks and E. Davis. Nucl. Nielsen and P. Dine and N. Seiberg. Elitzur. [72] A. (Cargese. M. Phys. Douglas. Pioline and E.V. Phys. B. [48] N. B 160 (1979) 349. Shenker. Rabinovici. Forge. Yankielowicz. Phys. Phys. Shenker. 162 (1988) 169. Phys. Nucl. Kutasov. Seiberg. ’t Hooft. B 353 (1995) 79 [hep-th/9504080]. Phys. A. Phys. Nucl. [56] S. Seiberg and E. B 61 (1973) 45. Phys. S. Nucl. T. K. Phys.B. B 505 (1997) 202 [hep-th/9704104]. Lett. B 218 (1983) 493. Aug. Nucl. Rept. Y. Witten. Mattis. B 431 (1994) 551 [hep-th/9408155]. Dorey. Schwimmer. V. B 205 (1982) 1. Phys.F. B 342 (1995) 152 [hep-ph/9410203]. Rabinovici. Dine and N. Duality in N = 2 SUSY SU (2) Yang-Mills Theory: A pedagogical introduction to the work of Seiberg and Witten [arXiv:hep-th/9601007]. Witten. A. Nucl. Phys. [71] P. Seiberg. [70] N.4 PRINT-80-0083 (UTRECHT) Lecture given at Cargese Summer Inst. B 189 (1981) 547. [46] S. B 435 (1995) 129 [hep-th/9411149]. Phys. Lett. 8. Rabinovici. [61] K. E. A. G. Seiberg and S. 26 – Sep. Giveon and D. Rabinovici: Supersymmetric Gauge Theories 239 [42] For similar structures in a string theory context see. Phys. Phys. B 138 (1978) 1.

Turin. 220. B 461 (1996) 71. Rabinovici. [87] E. Nucl. Giveon. N. Witten.B. Giveon and M. Petkova. Phys. Phys. Aharony. M. [102] J. Brane Dynamics and Gauge Theory [hep-th/9802067]. Lu. S. see: A. Strassler. [83] D. [89] P. G. Nucl. R. Moshe and M. Singapore: World Scientiﬁc. 61 (1989). Weinberg. Phys. [91] For example. Rev. Dobrev. Nucl. Sov. M. N. Phys. Ronen Plesser. Mack.B. Seiberg.G. 1987).J. Non-perturbative gauge dynamics and strings. Uspekhi 11 (1968) 381. Witten. Rabinovici. Lett. Rept. Phys. Rabinovici. Nucl. [96] M. Nucl. Math. Phys. Bardeen. [77] R. S. Banks and A. February 24-26. Plesser.D. Lambert and P. 1994 (Singapore. 1977). JHEP 9904 (1999) 015 [arXiv:hep-th/9902058]. 1994).240 Unity from Duality: Gravity. B 544 (1999) 104 [arXiv:hep-th/9809033]. Porrati and E. Barbon.D.J. [78] M. [92] J. 53 (1977) 155. West. [84] E. Howe.J.A. and references therein.R. Nucl. Lett. B 499 (1997) 67 [hep-th/9703110]. Witten. Mod. Superstring Theory (Cambridge University Press. A. Commun. [74] G. [79] For some recent string theory results see. Roˇek. J. G. Khuri and J. JHEP 0206 (2002) 039 [arXiv:hep-th/0205090]. Barbon.G. in Proceedings. B 492 (1997) 152 [hep-th/9611230]. D 36 (1987) 562. TASI Lectures on D-Branes [hep-th/9611050]. and references therein. Phys. Hanany and E.T.I. Bander. B 257 (1985) 371.C. [101] O. Nucl. Symposium on The Occasion of Sergio Fubini’s 65th Birthday. [75] V. [80] For a review. Schwarz and E. [81] Ya. B 318 (1993) 469 [hep-th/9309335]. Seiberg and E. [90] N. Yankielowicz. Rev. Basics and highlights in fundamental physics”. Saering and W. Published in “Erice 1999. Phys. Rabinovici. B 500 (1997) 3 [hep-th/9703166]. Harvey [hep-th/9603082]. Goldberg and E.L. Strassler. B 363 (1995) 173 c [hep-th/9508043]. N. Mack. [82] S. B 524 (1998) 141 [arXiv:hep-th/9712040]. Phys. B 418 (1998) 85 [arXiv:hep-th/9710034]. [98] A. N.B.A.A. Argyres. Rabinovici. Rabinovici. Phys. Argyres. Seiberg and M.C. Witten. Bardeen. Gauge Theory and Strings [73] T. [88] N. Phys. [93] A. B 447 (1995) 95 [arXiv:hep-th/9503121]. Phys.R.X.S. Italy. Lett. Phys. Zaks. J. K.H.F. Phys. West. Kogan and E.K. Phys. I. Kol and S. Phys.C.M. Kutasov. Phys. Zeldovich. Green. Symmetry And Simplicity in Physics. Hanany. [85] W. [86] W. Intriligator. [94] E. Einhorn. [100] P. B. Witten. 259 (1995) 213 [hep-th/9412184]. Phys. 284. see: J. O. M. [97] M. Seiberg and E. Todorov. [99] E.L. Nucl. Nucl.I. Nucl. N. Phys. [76] P. Nucl.B. Harmonic Analysis On The N-Dimensional Lorentz Group And Its Application To Conformal Quantum Field Theory (Berlin. Aharony. Polchinski. [95] A. Lett. Phys. Phys. 244 (1994) 77 [hep-th/9401139]. Alberico and S. Rept. Rabinovici. Leigh and M. Giveon and D. Phys. Seiberg. . 52 (1984) 1188. B 461 (1996) 71 [hep-th/9511154]. B 435 (1995) 129 [hep-th/9411149]. Abel. Amit and E. I. unpublished. 280. Sciuto. B 256 (1985) 499. V. Rabinovici. B. Petrova and I. Lambert and P. M. Rev. Kogan and E. B 196 (1982) 189. Nucl. Duﬀ.

SEN Harish-Chandra Research Institute. India .LECTURE 3 AN INTRODUCTION TO DUALITY SYMMETRIES IN STRING THEORY A. Allahabad 211019. Jhusi. Chhatnag Road.

. . . . . . . . 4. . . . . . . . . . 315 . . Z) self-duality of type IIB in D = 10 . . . .2 Self-duality of heterotic string theory on T 6 . . 304 7 Duality in theories with less than sixteen supersymmetry generators 305 7. .1 SL(2. on . . . 301 . . . . .5 Other examples . . . . 6. . . . . . . . . . . . . . . . .2 Interactions . . 251 . . . .4 Recovering higher dimensional dualities from lower dimensional ones . . . . . . . . . . . . . . . . . . . . . . . . . . 296 299 . . 4.2 Test of duality conjectures involving theories with eight supercharges . . . .Contents 1 Introduction 2 A brief review of perturbative 2. . . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299 . . . . . . . . . . . . . . . 5 Precision test of duality: Spectrum of BPS states 5. . . . . . . . . . . . .3 Fiberwise duality transformation . . . .2 SL(2. . . . . . . . . . . . . . . . . . . 254 3 Notion of duality symmetries in string theory 255 3. . . . . . . . . . . . . .1 Type I–SO(32) heterotic duality in D = 10 . . . .2 Compactiﬁcation of M-theory . . . . . . . . . . . 276 moduli . K3 . . . .4 SL(2. . . . 5. . . . . . . . .1 Duality symmetries: Deﬁnition and examples . . . . . Z) duality in type IIB on S 1 and D-branes . . . . 246 . . 299 . . . . . . . . . Z) S-duality in heterotic on T 6 and multi-monopole spaces . . . .3 Compactiﬁcation . . . . . . . . . 2. . . . . . . . . . . . . . . . . .3 Duality between heterotic on T 4 and type IIA 4. . .1 Combining non-perturbative and T -dualities . . . 280 . . 258 4 Analysis of low energy eﬀective ﬁeld theory 4. . . . . . . . . . . .2 Testing duality conjectures . . . . 260 261 264 268 272 274 . . . . . . . . . . . . . . . . . . . . . . . . . 309 8 M-theory 312 8. . . 6. . .3 Massless solitons and tensionless strings . . .1 The spectrum . 255 3. . . . . . . . 243 string theory 245 . . . . . . . .1 Construction of a dual pair of theories with eight supercharges . . . . . . . . . . . . . . . . . . . . . . . 6. . . . . 5. . . . .2 Duality of dualities . . . . 6 Interrelation between diﬀerent duality conjectures 6. . . . . . . . . . . . . . . . . . . . . 4. . . . . . 306 7. . . 312 8. . . .1 M-theory in eleven dimensions . 286 . .

A complete list of references may be obtained by looking at the citations to some of the original papers in spires. see [18]. our understanding of string theory has undergone a dramatic change. which relate the strong and weak coupling limits of apparently diﬀerent string theories. I apologise for the inherent bias in the choice of examples and the topics. Instead I have only included those references whose results have been directly used or mentioned in this article. and an introduction to M-theory. but give us a way to compute certain strong coupling results in one string theory by mapping it to a weak coupling result in a dual string theory.) c EDP Sciences. relationship between diﬀerent duality symmetries. I have also not attempted to give a complete list of references. These symmetries not only relate apparently different string theories. The key to this development is the discovery of duality symmetries. However. instead of surveying all the important developments in this subject I shall try to explain the basic ideas with the help of a few simple examples. Sen Abstract In this review I discuss some basic aspects of non-perturbative string theory.AN INTRODUCTION TO DUALITY SYMMETRIES IN STRING THEORY A. Springer-Verlag 2002 . this is solely due to the varied degree of familiarity that I have with this vast subject. There are also many other reviews in this subject where more references can be found [1–17]. 1 Introduction During the last few years. The topics include test of duality symmetries based on the analysis of the low energy eﬀective action and the spectrum of BPS states. In this review I shall try to give an introduction to this exciting subject. (For an introduction to perturbative string theory. I hope that this review will serve the limited purpose of initiating a person with a knowledge of perturbative string theory into this area.

Duality in theories with <16 supersymmetries: the discussion in Sections 4–6 is focussed on string theories with at least 16 supersymmetry generators. Analysis of the low energy eﬀective action: in this section I shall describe how one arrives at various duality conjectures by analyzing the low energy eﬀective action of string theory. 7. 5. Precision test of duality based on the spectrum of BPS states: in this section I shall discuss how one can device precision tests of various duality conjectures based on the analysis of the spectrum of a certain class of supersymmetric states in string theory. Interrelation between various dualities: in this section I shall try to relate the various duality conjectures introduced in the Sections 3–5 by “deriving” them from a basic set of duality conjectures. This will in no way constitute an introduction to this subject. In this section I consider theories with less number of supersymmetries. . I shall also discuss what we mean by relating diﬀerent dualities and try to formulate the rules that must be followed during such a derivation. Gauge Theory and Strings The review will be divided into seven main sections as described below. Throughout this article I shall work in units where = 1 and c = 1. 3. 2. and the general procedure for testing these duality conjectures. A brief review of perturbative string theory: in this section I shall very brieﬂy recollect some of the results of perturbative string theory which will be useful to us in the rest of this article. at best it will serve as a reminder to a reader who is already familiar with this subject. 4. Speciﬁcally we shall focus our attention on theories with eight supercharges. 1. Notion of duality symmetry: in this section I shall describe the notion of duality symmetry in string theory. which correspond to N = 2 supersymmetry in four dimensions. M-theory: in this section I discuss the emergence of a new theory in eleven dimensions − now known as M-theory − from the strong coupling limit of type IIA string theory. a few examples of duality conjectures in string theory. 6.244 Unity from Duality: Gravity. I also discuss how compactiﬁcation of M-theory gives rise to new theories that cannot be regarded as perturbative compactiﬁcation of a string theory.

. 1. we √ need to choose α to be of the order of 10−33 cm. known as string tension. The energy per unit length of the string. are actually diﬀerent vibrational modes of strings. As we shall describe later. Since α is the only length parameter in the theory.A. 2. a) A closed string. the typical size of a string is of the order √ of α ∼ 10−33 cm − a distance that cannot be resolved by present day experiments. and its appeal lies in its theoretical consistency. which appear as point-like objects to the present day experimentalists. A B Fig. Thus there is no direct way of testing string theory. this theory automatically contains gravitational interaction between elmentary particles. (a) (b) Fig. and b) an open string. Sen: Duality Symmetries 2 A brief review of perturbative string theory 245 String theory is based on the simple idea that elementary particles. Propagation of a closed string. where α has the dimension of (length)2 . but in order to correctly reproduce the strength of this√ interaction. is parametrized as (2πα )−1 .

The weight factor is given by e−S where S is the product of the string tension and the area of the world-sheet. This (1 + 1) dimensional ﬁeld theory is known as the world-sheet theory. 2b). has ﬁnite extent. as we shall describe. It represents a compact circle if the string is closed (Fig. 2. Thus the interesting part of the theory is the one involving the massless states. 2a) and a ﬁnite line interval if the string is open (Fig. the dynamics of a point particle is described by quantum mechanics. The amplitude for the propagation of the string from the space-time position A to space-time position B is given by the weighted sum over all world-sheet bounded by the initial and the ﬁnal locations of the string. which labels the coordinate on the string. However unlike a conventional quantum ﬁeld theory where the spatial directions have inﬁnite extent.1 The spectrum There are ﬁve known fully consistent string theories in ten dimensions. It turns out that this procedure by itself does not give rise to a fully consistent string theory. each world-sheet ﬁeld can be regarded as a collection of inﬁnite number of harmonic oscillators labelled by the quantized momentum along this spatial direction. Since the spatial direction of the world-sheet theory has ﬁnite extent. The leads to ﬁve (apparently) diﬀerent consistent string theories in (9 + 1) dimensional space-time. here the spatial direction.246 Unity from Duality: Gravity. This gives an inﬁnite tower of states. Generalizing this we see that the ﬁrst quantized description of a string will involve a (1 + 1) dimensional quantum ﬁeld theory. Here we give a brief description of the degrees of freedom and the spectrum of massless states in each of these . 1). E8 × E8 heterotic and SO(32) heterotic string theories respectively. The massive string states typically have mass of the order of (10−33 cm)−1 ∼ 1019 GeV and are far beyond the reach of the present day accelerators. type I. Consider propagation of a string from a spacetime conﬁguration A to a space-time conﬁguration B. In the ﬁrst quantized formalism. We shall now brieﬂy describe the spectrum and interaction in various string theories and their compactiﬁcations. type IIB. Diﬀerent states of the string are obtained by acting on the Fock vacuum by these oscillators. In order to get a fully consistent string theory we need to add some internal fermionic degrees of freedom to the string and generalize the notion of area by adding new terms involving these fermionic degrees of freedom. The ﬁelds in this (1 + 1) dimensional quantum ﬁeld theory and the boundary conditions on these ﬁelds vary in diﬀerent string theories. Typically each string theory contains a set of massless states and an inﬁnite tower of massive states. They are known as type IIA. known as the string world-sheet (see Fig. Gauge Theory and Strings The basic principle behind constructing a quantum theory of relativistic string is quite simple. During this motion the string sweeps out a two dimensional surface in space-time.

These eight scalar ﬁelds are in fact common to all ﬁve string theories. 1. Type II string theories: in this case the world-sheet theory is a free ﬁeld theory containing eight scalar ﬁelds and eight Majorana fermions. In all there are four classes of states which need to be included in the spectrum: • NS-NS where we put anti-periodic boundary conditions on both the left. eight of them having left-handed chirality and the other eight having right-handed chirality. we keep only about (1/4)th of the states in each sector by keeping only those states in the spectrum which have in them only even number of left-moving fermions and even number of right-moving fermions. Both the type II string theories contain only closed strings. Sen: Duality Symmetries 247 theories. • NS-R where we put anti-periodic boundary condition on the leftmoving fermions and periodic boundary condition on the rightmoving fermions. It is customary to refer to periodic boundary condition as Ramond (R) boundary condition [123] and anti-periodic boundary condition as Neveu-Schwarz (NS) boundary condition [124]. but at the end two possibilities remain. It turns out that in order to get a consistent string theory we need to include in our theory diﬀerent classes of string states. • R-NS where we put periodic boundary condition on the leftmoving fermions and anti-periodic boundary condition on the right-moving fermions.and the right-moving fermions. • R-R where we put anti-periodic boundary conditions on both the left. It is useful to regard the eight Majorana fermions as sixteen Majorana-Weyl fermions. Finally.and right-moving fermions respectively. some of which have periodic and some of which have anti-periodic boundary condition on the fermions. We shall refer to these as left. These diﬀer from each other in .and the right-moving fermions.A. hence the spatial component of the world-sheet is a circle. The eight scalar ﬁelds satisfy periodic boundary condition as we go around the circle. The fermions have a choice of having periodic or anti-periodic boundary conditions. and represent the eight transverse coordinates of a string moving in a nine dimensional space. The procedure has some ambiguity since in each of the four sectors we have the choice of assigning to the ground state either even or odd fermion number. This is known as the GSO projection [125]. Consistency of string theory rules out most of these possibilities. We shall give the description in the so called light-cone gauge which has the advantage that all states in the spectrum are physical states.

On the other hand. Typically states from the Ramond sector are in the spinor representation of the SO(9. The supersymmetry algebra for type IIB theory is known as the chiral N = 2 superalgebra and that of type IIA theory is known as the non-chiral N = 2 superalgebra. As has already been pointed out before. it is useful to organise the spectrum in terms of ﬁelds.1) Lorentz algebra. In other words the spectrum of massless ﬁelds in string theory is identical to that of a free ﬁeld theory with these ﬁelds. It will be useful to list the massless bosonic states in these two string theories. and those from the NS-R and R-NS sectors are fermionic. whereas those from the NS sector are in the tensor representation. This theory is known as type IIA string theory. the massless states from the RR sector of type IIB string theory consist of a scalar. and a scalar ﬁeld known as the dilaton1 . For the creation operator associated with each oscillator we deﬁne the level as the absolute value of the number of units of world-sheet momentum that it creates while acting on the vacuum. a rank two anti-symmetric tensor ﬁeld. The spectrum of both these theories are invariant under space-time supersymmetry transformations which transform fermionic states to bosonic states and vice versa. In one possibility. In the second possibility the GSO projections in the left.and the right-moving sector diﬀer from each other. This gives type IIB string theory. the assignment of the left. an anti-symmetric rank two tensor ﬁeld.and the right-moving Ramond ground states are carried out in an identical manner. The total oscillator level of a state is then the sum of the levels of all the oscillators that act on the Fock vacuum to create this state. the NS sector bosonic states are the same in the two theories.248 Unity from Duality: Gravity. They constitute a symmetric rank two tensor ﬁeld. and a rank three anti-symmetric tensor. is characterized by several quantum numbers. The RR sector massless states of type IIA string theory consist of a vector.and the right-moving fermion number to the left. Since the product of two spinor representation gives us back a tensor representation. 1 Although from string theory we get the spectrum of states. (The Fock vacuum. in turn. Both superalgebras consist of 32 supersymmetry generators. Gauge Theory and Strings the following way. . and a rank four anti-symmetric tensor gauge ﬁeld satisfying the constraint that its ﬁeld strength is self-dual. Often it is convenient to organise the inﬁnite tower of states in string theory by their oscillator level deﬁned as follows. the states from the NS-NS and the RR sectors are bosonic. the world-sheet degrees of freedom of the string can be regarded as a collection of inﬁnite number of harmonic oscillators. Since the two theories differ only in their R-sector.

or all of them have anti-periodic boundary condition.(right-) moving oscillator level as the contribution to the oscillator level from the left. and the momenta of the (9 + 1) dimensional theory. • E8 × E8 heterotic string theory: in this case we divide the thirty two left-moving fermions into two groups of sixteen each and 2 We should distinguish between world-sheet momentum. Sen: Duality Symmetries 249 which are the momenta conjugate to the zero modes of various ﬁelds − modes carrying zero world-sheet momentum. However. a scalar ﬁeld known as the dilaton and a set of 496 gauge ﬁelds ﬁlling up the adjoint representation of the gauge group SO(32). We have as before NS and R boundary conditions as well as GSO projection involving the right-moving fermions. The latter are the the momenta conjugate to the zero modes of various bosonic ﬁelds in the world-sheet theory. There are two possible boundary conditions on the left-moving fermions which give rise to fully consistent string theories. The massless bosonic states in this theory consist of a symmetric rank two ﬁeld. In each sector we also have a GSO projection that keeps only those states in the spectrum which contain even number of left-moving fermions.1) Lorentz algebra. They are: • SO(32) heterotic string theory: in this case we have two possible boundary conditions on the left-moving fermions: either all of them have periodic boundary condition. an anti-symmetric rank two ﬁeld. Thus bosonic states come from states with NS boundary condition on the right-moving fermions and fermionic states come from states with R boundary condition on the right-moving fermions.A.) We can also separately deﬁne left. . Heterotic string theories: the world-sheet theory of the heterotic string theories consists of eight scalar ﬁelds. in this case the boundary condition on the left-moving fermions do not aﬀect the Lorentz transformation properties of the state. Thus for example the total contribution to L0 will be given by the sum of the right-moving oscillator level and the contribution to L0 from the Fock vacuum. eight right-moving MajoranaWeyl fermions and thirty two left-moving Majorana-Weyl fermions. the NS sector states transform in the tensor representation and the R sector states transform in the spinor representation of the SO(9. unlike in the case of type II string theories. Finally. if E and P denote respectively the world-sheet energy ¯ and momentum2 then we deﬁne L0 = (E + P )/2 and L0 = (E − P )/2. Also as in the case of type II string theories.(right-) moving bosonic and fermionic ﬁelds. 2. ¯ L0 and L0 include contribution from the oscillators as well as from the Fock vacuum.

Type I string theory: the world-sheet theory of type I theory is identical to that of type IIB string theory. we can reformulate both the heterotic string theories by replacing the thirty two left-moving fermions by sixteen left-moving bosons. and also even number of leftmoving fermions from the second group. with the following two crucial diﬀerence. The massless bosonic states in this theory consist of a symmetric rank two ﬁeld. 3. Gauge Theory and Strings use four possible boundary conditions. giving us back the two heterotic string theories. an anti-symmetric rank two ﬁeld. (This symmetry is not present in type IIA theory since the GSO projection in the two sectors are diﬀerent). and has sixteen real generators. • Type IIB string theory has a symmetry that exchanges the leftand the right-moving sectors in the world-sheet theory. 4) all the left-moving fermions in group 1 have anti-periodic boundary conditions and all the left-moving fermions from group 2 have periodic boundary conditions.250 Unity from Duality: Gravity. It turns out that there are only two consistent ways of quantizing the momenta. In order to get a consistent string theory the momenta conjugate to these bosons must take discrete values. The world-sheet degrees of freedom are identical to those in the closed string sector. • In type I string theory we also include open string states in the spectrum. This transformation is known as the world-sheet parity transformation. In constructing type I string theory we keep only those states in the spectrum which are invariant under this world-sheet parity transformation. 3) all the leftmoving fermions in group 1 have periodic boundary conditions and all the left-moving fermions in group 2 have anti-periodic boundary conditions. a scalar ﬁeld known as the dilaton and a set of 496 gauge ﬁelds ﬁlling up the adjoint representation of the gauge group E8 × E8 . 1) all the left-moving fermions have periodic boundary condition 2) all the left-moving fermions have anti-periodic boundary condition. In each sector we also have a GSO projection that keeps only those states in the spectrum which contain even number of left-moving fermions from the ﬁrst group. Specifying the theory requires . The relevant superalgebra is known as the chiral N = 1 supersymmetry algebra. The spectrum of states in both the heterotic string theories are invariant under a set of space-time supersymmetry transformations. Using the bose-fermi equivalence in (1 + 1) dimensions.

we would like to know how to compute a scattering amplitude involving various string states. S being the string tension multiplied by the generalized area of this surface (taking into account the fermionic degrees of freedom of the world-sheet). One such surface is shown in Figure 3. and appropriate boundary conditions on the fermions. Consider for example a scattering involving four external strings. but in order to fully describe the theory we must also describe the interaction between various particles in the spectrum. situated along some speciﬁc curves in space-time. The prescription for computing the scattering amplitude is to compute the weighted sum over all possible string world-sheet bounded by the four strings with weight factor e−S . It turns out that there is a unique way of introducing interaction in string theory. then this diagram represents two strings joining into one string and then splitting into two strings. 2. 3. This spectrum is also invariant under the chiral N = 1 supersymmetry algebra with sixteen real supersymmetry generators. In particular. Fig. If we imagine the time axis running from left to right. The spectrum of massless bosonic states in this theory consists of a symmetric rank two tensor and a scalar dilaton from the closed string NS sector. − the analog of a tree diagram in . Sen: Duality Symmetries 251 us to specify the boundary conditions on the various ﬁelds.2 Interactions So far we have discussed the spectrum of string theory. We put Neumann boundary condition on the eight scalars. A string world-sheet bounded by four external strings. an anti-symmetric rank two tensor from the closed string RR sector. and 496 gauge ﬁelds in the adjoint representation of SO(32) from the open string sector.A.

By deﬁnition this eﬀective action is such that if we compute the tree level scattering amplitude using this action.252 Unity from Duality: Gravity. string theory has a single dimensionless coupling constant. all of these scattering amplitudes in string theory are ultraviolet ﬁnite. As we shall see later. The relative normalization between the contributions from these two diagrams is not determined by any consistency requirement. 4. However. This is the analog of a one loop diagram in ﬁeld theory. In general such an action will have to . The more relevant quantity is the scattering amplitude where the external strings are in the eigenstates of the energy and momenta operators conjugate to the coordinates of the (9 + 1) dimensional space-time. Thus besides the dimensionful parameter α . Gauge Theory and Strings ﬁeld theory. The most convenient way to summarize the result of this computation in any string theory is to specify the eﬀective action. we should reproduce the S-matrix elements involving the massless states of string theory. A more complicated string world-sheet. In practice there is an extremely eﬃcient method of doing this computation using the so called vertex operators. A more complicated surface is shown in Figure 4. which then splits into two and joins again. and ﬁnally splits into two strings. This introduces an arbitrary parameter in string theory. This is done by simply taking the convolution of the above scattering amplitude with the wavefunctions of the strings corresponding to the external states. once the relative normalization between these two diagrams is ﬁxed. the relative normalization between all other diagrams is ﬁxed due to various consistency requirement. It turns out that unlike in quantum ﬁeld theory. This is one of the major achievements of string theory. This represents two strings joining into one string. known as the string coupling constant. Fig. What we have described so far is the computation of the scattering amplitude with ﬁxed locations of the external strings in space-time. both these parameters can be absorbed into deﬁnitions of various ﬁelds in the theory. Out main interest will be in the scattering involving the external massless states.

The perturbative eﬀective action does not have any potential for Φ. g S → eC g S . called type IIA and type IIB supergravity theories respectively. For type I and the SO(32) heterotic string theories the Yang-Mills gauge group is SO(32) whereas for the E8 × E8 heterotic string theory the gauge group is E8 × E8 . but we can organise these terms by examining the number of space-time derivatives that appear in a given term in the action. The low energy eﬀective action for all ﬁve string theories have been found. or equivalently the parameter α . In a similar vain one can argue that in string theory even the string tension. Using this scaling property. combined with the result on ﬁniteness of scattering amplitudes.2) .1) together with possible rescaling of other ﬁelds. (2. gS denotes the string coupling. gµν → λgµν . The emergence of gravity in all the ﬁve string theories is the most striking result in string theory. Since α has the dimension of (length)2 and is the only dimensionful parameter in the theory. gS can be absorbed in Φ. there will be a potential for Φ. On the other hand the actions for the three heterotic string theories correspond to another set of well-known supersymmetric theories in ten dimensions. the eﬀective action will have an invariance under the simultaneous rescaling of α and the metric gµν : α → λα . Sen: Duality Symmetries 253 contain inﬁnite number of terms. One expects that in a realistic string theory where supersymmetry is spontaneouly broken. Here Φ denotes the dilaton ﬁeld.A. The actions for the type IIA and type IIB string theories correspond to those of two well known supergravity theories in ten space-time dimensions. and hence Φ will be determined uniquely. Its origin can be traced to the existence of the symmetric rank two tensor state (the graviton) in all these theories. This. The eﬀective action of all ﬁve string theories are invariant under the transformation Φ → Φ − 2C. We shall explicitly write down the low energy eﬀective action of some of the string theories in Section 4. the dimensionless coupling constant in string theory is related to the vacuum expectation value Φ of Φ. Terms with the lowest number of derivatives constitute the low energy eﬀective action. Put another way. has no physical signiﬁcance. (2. shows that string theory gives us a ﬁnite quantum theory of gravity. − so called because this gives the dominant contribution if we want to evaluate the scattering amplitude when all the external particles have small energy and momenta. − N = 1 supergravity coupled to N = 1 super Yang-Mills theory. and hence Φ can take arbitrary value. and C is an arbitrary constant.

one can also directly describe these compactiﬁed theories as string theories. tori of diﬀerent dimensions. are tori − product of circles.3 Compactiﬁcation So far we have described ﬁve diﬀerent string theories. These states are known as winding states and play a crucial role in the analysis of duality symmetries. For this one needs to modify the string world-sheet action in such a way that it describes string propagation in the new manifold M × Rd.1. and hence consistency of the corresponding string theory order by order in perturbation theory. in the limit when the size of the compact manifold is suﬃciently small so that the present day experiments cannot resolve this distance. Consistency of string theory puts restriction on the kind of manifold on which the string can propagate.g. But that is another novel eﬀect: we now have new states that correspond to strings wrapped around a compact circle. but they all live in ten space-time dimensions. This modiﬁes the world-sheet theory to an interacting non-linear σ-model instead of a free ﬁeld theory. Since our world is (3 + 1) dimensional. Of course we cannot choose any arbitrary manifold M for this purpose.254 Unity from Duality: Gravity. . The eﬀect of this compactiﬁcation is to periodically identify some of the bosonic ﬁelds in the string world-sheet ﬁeld theory − the ﬁelds which represent coordinates tangential to the compact circles. At the end both approaches yield identical results. since this guarantees vanishing of the cosmological constant. However one can construct string theories in lower dimensions using the idea of compactiﬁcation. We shall discuss these two rescalings in detail in Section 4. The simplest class of compact manifolds. Choosing d = 3 will give us a (3 + 1) dimensional theory. Then.1 . One also normally considers only those manifolds which preserve part of the space-time supersymmetry of the original ten dimensional theory. on which we shall focus much of our attention in the rest of this article. we also go once around the compact circle. There are many known examples of manifolds satisfying these restrictions e. Gauge Theory and Strings together with possible rescaling of other ﬁelds. Calabi-Yau manifolds etc. these are not realistic string theories. Instead of going via the eﬀective action. instead of in ﬂat ten dimensional space-time. it must satisfy the equations of motion of the eﬀective ﬁeld theory that comes out of string theory. 2. The idea is to take the (9 + 1) dimensional space-time as the product of a (9 − d) dimensional compact manifold M with euclidean signature and a (d + 1) dimensional Minkowski space Rd. Using this scaling symmetry α can be absorbed into the deﬁnition of gµν . K3. the world will eﬀectively appear to be (d + 1) dimensional. For such a states. as we go once around the string. One eﬀect of this is that the momentum carried by any string state along any of these circles is quantized in units of 1/R where R is the radius of the circle.1 .

This situation is illustrated in Figure 6. • various other background ﬁelds. whereas the white region denotes the strong coupling region. In general this equivalence relation maps the weak coupling region of one theory to the strong coupling region of the second theory and vice versa. • Shape and size of M (information contained in the metric). The shaded region denotes the weak coupling region. there are ﬁve consistent string theories in ten space-time dimensions. Sen: Duality Symmetries 3 Notion of duality symmetries in string theory 255 In this section I shall elaborate the notion of duality symmetries. . Each of these theories is parametrized by a set of parameters known as moduli3 e. String duality provides us with an equivalence map between two diﬀerent string theories. 3. Elsewhere the theory is strongly coupled. A schematic representation of the moduli space of a string theory. 5. We also saw that we can get many diﬀerent string theories in lower dimensions by compactifying these ﬁve theories on appropriate manifold M. This situation has been illustrated in Figure 5.A. Fig. We begin by introducing the notion of duality in string theory. the diﬃculties in testing them. Inside the moduli space of the theory there is a certain region where the string coupling is weak and perturbation theory is valid. and the way of avoiding these diﬃculties. • String coupling constant (related to the vacuum expectation value of the dilaton ﬁeld). 3 In string theory these moduli are related to vacuum expectation values of various dynamical ﬁelds and are expected to take deﬁnite values when supersymmetry is broken.1 Duality symmetries: Deﬁnition and examples As was described in the last section.g.

23–26]4 . Schematic representation of the moduli space of a self-dual theory. let us give a few examples of dual pairs: • Type I and SO(32) heterotic string theories in D = 10 are conjectured to be dual to each other [19–22]. and K. Before we proceed. 7. A on K and B on K .256 Unity from Duality: Gravity.1 denotes (10 − n) dimensional Minkowski space. where A and B are two of the ﬁve string theories in ten dimensions. Gauge Theory and Strings A on K B on K Fig. Under this duality the weak coupling region of the ﬁrst theory (denoted by the shaded region) gets mapped to the strong coupling region of the second theory and vice versa. 4 Throughout this article a string theory on M will mean string theory in the background M × R9−n. Also under duality. Thus for example. and R9−n. tree level results in one theory might include perturbative and non-perturbative corrections in the dual theory.1 where n is the real dimension of M. many of the elementary string states in one theory get mapped to solitons and their bound states in the dual theory. K are two compact manifolds. typically perturbation expansions get mixed up. Under duality. Duality relates weak and strong coupling regions of the same theory. • Type IIA string theory compactiﬁed on K3 and heterotic string theory compactiﬁed on a four dimensional torus T 4 are conjectured to be dual to each other [19. Fig. . 6. A schematic representation of the duality map between the moduli spaces of two diﬀerent string theories.

Examples of T -duality relating a weakly coupled theory to a diﬀerent or the same weakly coupled theory. 9. Although duality in general relates the weak coupling limit of one theory to the strong coupling limit of another theory. duality transformations form a symmetry group that acts on the moduli space of the theory. In each case the shaded region denotes weak coupling region as usual. A schematic representation of the moduli spaces of a chain of theories related by duality. as illustrated in Figure 7.A. Fig. In this case. 8. type IIB . there are special cases where the situation is a bit diﬀerent. For example. Sen: Duality Symmetries 257 Fig. For example. we can have: • Self-duality: here duality gives an equivalence relation between diﬀerent regions of the moduli space of the same theory.

From this discussion we see that the presence of duality in string theory has two important consequences. As a result the duality map does not mix up the perturbation expansions in the two theories. Also. All T -duality symmetries in string theory can be “proved” this way. Z) self-duality group [23]. Gauge Theory and Strings string theory in D = 10 is conjectured to have an SL(2. As we have already emphasized. First of all. (For a review of this subject. it reduces the degree of non-uniqueness of string theory. see [27]. • T -duality: in this case duality transformation maps the weak coupling region of one theory to the weak coupling region of another theory or the same theory as illustrated in Figure 8. Furthermore. due to the equivalence of the SO(32) heterotic and type I string theory in ten dimensions. Thus in order to prove/test duality we must be able to analyze at least one of the theories at strong coupling. Thus for example. at least to all orders in perturbation theory. by relating various apparently unrelated (compactiﬁed) string theories. as illustrated in Figure 9. it allows us to study a strongly coupled string theory by mapping it to a weakly coupled dual theory whenever such a dual theory exists. duality typically relates a weakly coupled string theory to a strongly coupled string theory. 3. For example.258 Unity from Duality: Gravity. type IIA string theory compactiﬁed on a circle of radius R is dual to IIB string theory compactiﬁed on a circle of radius R−1 at the same value of the string coupling.2 Testing duality conjectures Let us now turn to the question of testing duality. but a whole chain of theories. type IIA string theory compactiﬁed on K3 is related to heterotic string theory compactiﬁed on T 4 . But in string theory we only know how to deﬁne the theory perturbatively at weak coupling. SO(32) heterotic string theory compactiﬁed on T 4 is related to type I string theory compactiﬁed on T 4 . This is where supersymmetry comes to our 5 Note that this problem is absent for T -duality transformations which relates two weakly coupled string theories. . Thus it would seem impossible to prove or test any duality conjecture in string theory5 . Thus these three theories are related by a chain of duality transformations. either of the two heterotic string theories compactiﬁed on a circle of radius R is dual to the same theory compactiﬁed on a circle of radius R−1 at the same value of the coupling constant. On the other hand. and hence can be tested using string perturbation theory.) In a generic situation duality can relate not just two theories.

A. Sen: Duality Symmetries

259

rescue. Supersymmetry gives rise to certain non-renormalization theorems in string theory, due to which some of the weak coupling calculations can be trusted even at strong coupling. Thus we can focus our attention on such “non-renormalized” quantities and ask if they are invariant under the proposed duality transformations. Testing duality invariance of these quantities provides us with various tests of various duality conjectures, and is in fact the basis of all duality conjectures. The precise content of these non-renormalization theorems depends on the number of supersymmetries present in the theory. The maximum number of supersymmetry generators that can be present in a string theory is 32. This gives N = 2 supersymmetry in ten dimensions, and N = 8 supersymmetry in four dimensions. Examples of such theories are types IIA or type IIB string theories compactiﬁed on n dimensional tori T n . The next interesting class of theories are those with 16 supersymmetry generators. This corresponds to N = 1 supersymmetry in ten dimensions and N = 4 supersymmetry in four dimensions. Examples of such theories are types IIA or type IIB string theories compactiﬁed on K3 × T n , heterotic string theory compactiﬁed on T n , etc. Another class of theories that we shall discuss are those with eight supersymmetry generators, e.g. heterotic string theory on K3 × T n , type IIA or IIB string theory on six dimensional Calabi-Yau manifolds, etc. For theories with 16 or more SUSY generators the non-renormalization theorems are particularly powerful. In particular, • Form of the low energy eﬀective action involving the massless states of the theory is completely ﬁxed by the requirement of supersymmetry (and the spectrum) [28]. Thus this eﬀective action cannot get renormalized by string loop corrections. As a result, any valid symmetry of the theory must be a symmetry of this eﬀective ﬁeld theory; • These theories contain special class of states which are invariant under part of the supersymmetry transformations. They are known as BPS states, named after Bogomol’nyi, Prasad and Sommerfeld. The mass of a BPS state is completely determined in terms of its charge as a consequence of the supersymmetry algebra. Since this relation is derived purely from an analysis of the supersymmetry algebra, it is not modiﬁed by quantum corrections. Furthermore it can be argued that the degeneracy of BPS states of a given charge does not change as we move in the moduli space even from weak to strong coupling region [29]. Thus the spectrum of BPS states can be calculated from weak coupling analysis and the result can be continued to the strong coupling region. Since any valid symmetry of the theory must be a symmetry of the spectrum of BPS states, we can use this to design non-trivial tests of duality [1].

260

Unity from Duality: Gravity, Gauge Theory and Strings

For theories with eight supersymmetries the non-renormalization theorems are less powerful. However, even in this case one can design non-trivial tests of various duality conjectures. We shall discuss these in Section 7. 4 Analysis of low energy eﬀective ﬁeld theory

In this section I shall discuss tests of various dualities in string theories with ≥16 supersymmetries based on the analysis of their low energy eﬀective action. As has been emphasized in the previous section, the form of this low energy eﬀective action is determined completely by the requirement of supersymmetry and the spectrum of massless states in the theory. Thus it does not receive any quantum corrections, and if a given duality transformation is to be a symmetry of a string theory, it must be a symmetry of the corresponding low energy eﬀective action. Actually, since the low energy eﬀective action is to be used only for deriving the equations of motion from this action, and/or computing the tree level S-matrix elements using this action, but not to perform a full-ﬂedged path integral, it is enough that only the equations of motion derived from this action are invariant under duality transformations. (This also guarantees that the tree level S-matrix elements computed from this eﬀective action are invariant under the duality transformations.) It is not necessary for the action itself to be invariant. Throughout this article we shall denote by Gµν the string metric − the metric that is used in computing the area of the string world-sheet embedded in space time for calculating string scattering amplitudes. For a string theory compactiﬁed on a (9 − d) dimensional manifold M, we shall denote by Φ the shifted dilaton, related to the dilaton Φ(10) of the ten dimensional string theory as Φ = Φ(10) − ln V , (4.1)

where (2π)9−d V is the volume of M measured in the ten dimensional string (10) metric. The dilaton is normalized in such a way that e Φ corresponds to the square of the closed string coupling constant in ten dimensions6 . gµν will denote the canonical Einstein metric which is related to the string metric by an appropriate conformal rescaling involving the dilaton ﬁeld, gµν = e− d−1 Φ Gµν .

2

(4.2)

We shall always use this metric to raise and lower indices. The signature of space-time will be taken as (−, +, · · · +). Finally, all ﬁelds will be made dimensionless by absorbing appropriate powers of α in them.

6Φ

is related to the more commonly normalized dilaton φ by a factor of two: Φ = 2φ.

A. Sen: Duality Symmetries

261

We shall now consider several examples. The discussion will closely follow references [1, 19, 23]. For a detailed review of the material covered in this section, see reference [15]. 4.1 Type I–SO(32 ) heterotic duality in D = 10 In SO(32) heterotic string theory, the massless bosonic states come from (H) the NS sector of the closed heterotic string, and contains the metric gµν , (H) the dilaton Φ(H) , the rank two anti-symmetric tensor ﬁeld Bµν , and gauge (H)a ﬁelds Aµ (1 ≤ a ≤ 496) in the adjoint representation of SO(32). The low energy dynamics involving these massless bosonic ﬁelds is described by the N = 1 supergravity coupled to SO(32) super Yang-Mills theory in ten dimensions [114]. The action is given by [102]: S (H) = 1

2 (2π)7 (αH )4 gH

1 d10 x −g (H) R(H) − g (H)µν ∂µ Φ(H) ∂ν Φ(H) 8

(H) 1 (H) (H) − g (H)µµ g (H)νν e−Φ /4 Tr(Fµν Fµ ν ) 4 (H) 1 (H) (H) − g (H)µµ g (H)νν g (H)ρρ e−Φ /2 Hµνρ Hµ ν ρ , 12

**(4.3) where R(H) is the Ricci scalar, Fµν denotes the non-abelian gauge ﬁeld strength,
**

(H) Fµν = ∂µ A(H) − ∂ν A(H) + ν µ (H)

2 [A(H) , A(H) ] , ν αH µ

(H)

(4.4)

Tr denotes trace in the vector representation of SO(32), and Hµνρ is the (H) ﬁeld strength associated with the Bµν ﬁeld:

(H) Hµνρ

1 1 (H) (H) = ∂µ Bνρ − Tr A(H) Fνρ − µ 2 3 +cyclic permutations of µ, ν, ρ .

2 (H) (H) (H) A [Aν , Aρ ] αH µ (4.5)

2παH and gH are respectively the inverse string tension and the coupling constant of the heterotic string theory. The rescalings (2.1), (2.2) take the following form acting on the complete set of ﬁelds: g H → eC g H ,

(H) (H) Bµν → Bµν ,

Φ(H) → Φ(H) − 2C, A(H)a → A(H)a , µ µ

(H) (H) gµν → eC/2 gµν

(4.6)

262

**Unity from Duality: Gravity, Gauge Theory and Strings αH → λαH ,
**

(H) (H) Bµν → λBµν ,

Φ(H) → Φ(H) ,

(H) (H) gµν → λgµν

A(H)a → λ1/2 A(H)a , µ µ

(4.7)

Since gH and αH can be changed by this rescaling, these parameters cannot have a universal signiﬁcance. In particular, we can absorb gH and αH into the various ﬁelds by setting e−C = gH and λ = (αH )−1 in (4.6), (4.7). This is equivalent to setting gH = 1 and αH = 1. In this notation the physical (H) coupling constant is given by the vacuum expectation value of eΦ /2 , and the ADM mass per unit length of an inﬁnitely long straight string, measured (H) (H) (H) in the metric e Φ /4 gµν that approaches the string metric Gµν far away (H) we can get all possible from the string, is equal to 1/2π. By changing Φ values of string coupling, and using a metric that diﬀers from the one used here by a constant multiplicative factor, we can get all possible values of the string tension. For αH = 1 and gH = 1 equations (4.3)–(4.5) take the form: S (H) = 1 1 d10 x −g (H) R(H) − g (H)µν ∂µ Φ(H) ∂ν Φ(H) 7 (2π) 8 1 (H)µµ (H)νν −Φ(H) /4 (H) (H) − g g e Tr(Fµν Fµ ν ) 4 (H) 1 (H) (H) − g (H)µµ g (H)νν g (H)ρρ e−Φ /2 Hµνρ Hµ ν ρ , 12 (4.8)

(H) Fµν = ∂µ A(H) − ∂ν A(H) + ν µ

√

2[A(H) , A(H) ] , µ ν

(4.9)

(H) Hµνρ

**√ 1 2 (H) (H) (H) (H) (H) A [Aν , Aρ ] = − Tr Aµ Fνρ − 2 3 µ +cyclic permutations of µ, ν, ρ .
**

(H) ∂µ Bνρ

(4.10)

Let us now turn to the type I string theory. The massless bosonic states in type I theory come from three diﬀerent sectors. The closed string Neveu(I) Schwarz − Neveu-Schwarz (NS) sector gives the metric gµν and the dilaton Φ(I) . The closed string Ramond-Ramond (RR) sector gives an anti(I) symmetric tensor ﬁeld Bµν . Besides these, there are bosonic ﬁelds coming from the NS sector of the open string. This sector gives rise to gauge ﬁelds (I)a Aµ (a = 1, . . . 496) in the adjoint representation of the group SO(32). (The superscript (I) refers to the fact that these are the ﬁelds in the type I string theory.) The low energy dynamics is again described by the N = 1 supergravity theory coupled to SO(32) super Yang-Mills theory [115]. But

A. Sen: Duality Symmetries

263

it is instructive to rewrite the eﬀective action in terms of the type I variables. For suitable choice of the string tension and the coupling constant, this is given by [102] S (I) = 1 1 d10 x −g (I) R(I) − g (I)µν ∂µ Φ(I) ∂ν Φ(I) 7 (2π) 8 1 (I)µµ (I)νν Φ(I) /4 (I) (I) − g g e Tr(Fµν Fµ ν ) 4 (I) 1 (I) (I) − g (I)µµ g (I)νν g (I)ρρ eΦ /2 Hµνρ Hµ ν ρ , 12 (4.11) where R(I) is the Ricci scalar, Fµν denotes the non-abelian gauge ﬁeld strength, √ (I) Fµν = ∂µ A(I) − ∂ν A(I) + 2[A(I) , A(I) ] , (4.12) ν µ µ ν and Hµνρ is the ﬁeld strength associated with the Bµν ﬁeld: √ 1 2 (I) (I) (I) (I) (I) (I) A [Aν , Aρ ] Hµνρ = ∂µ Bνρ − Tr A(I) Fνρ − µ 2 3 µ +cyclic permutations of µ, ν, ρ .

(I) (I) (I)

(4.13)

For both, the type I and the SO(32) heterotic string theory, the low energy eﬀective action is derived from the string tree level analysis. However, to this order in the derivatives, the form of the eﬀective action is determined completely by the requirement of supersymmetry for a given gauge group. Thus neither action can receive any quantum corrections. It is straightforward to see that the actions (4.8) and (4.11) are identical provided we make the identiﬁcation: Φ(H) = −Φ(I) ,

(H) (I) Bµν = Bµν , (H) (I) gµν = gµν

A(H)a = A(I)a . µ µ

(4.14)

This led to the hypothesis that the type I and the SO(32) heterotic string theories in ten dimensions are equivalent [19]. One can ﬁnd stronger evidence for this hypothesis by analysing the spectrum of supersymmetris states, but the equivalence of the two eﬀective actions was the reason for proposing this duality in the ﬁrst place. Note the − sign in the relation between Φ(H) and Φ(I) in equation (4.14). Recalling that e Φ /2 is the string coupling, we see that the strong coupling limit of one theory is related to the weak coupling limit of the other theory and vice versa.

264

Unity from Duality: Gravity, Gauge Theory and Strings

From now on I shall use the unit α = 1 for writing down the eﬀective action of all string theories. Physically this would mean that the ADM mass per unit length of a test string, measured in the metric e2 Φ /(d−1) gµν that agrees with the string metric Gµν deﬁned in (4.2) far away from the test string, is given by 1/2π. In future we shall refer to the ADM mass of a particle measured in this metric as the mass measured in the string metric. 4.2 Self-duality of heterotic string theory on T 6 In the previous subsection we have described the massless bosonic ﬁeld content of the ten dimensional SO(32) heterotic string theory. When we compactify it on a six dimensional torus, we can get many other massless scalar ﬁelds from the internal components of the metric, the anti-symmetric tensor ﬁeld and the gauge ﬁelds in the Cartan subalgebra of the gauge group7 . This gives a total of (21 + 15 + 96 = 132) scalar ﬁelds. It turns out that these scalars can be represented by a 28 × 28 matrix valued ﬁeld M satisfying8 M LM T = L, where L = I6 MT = M , −I16 · (4.16) (4.15)

I6

In denotes an n × n identity matrix. We shall choose a convention in which M = I√ corresponds to a compactiﬁcation on (S 1 )6 with each S 1 having 28 radius α = 1 measured in the string metric, and without any background gauge or antisymmetric tensor ﬁelds. We can get another scalar ﬁeld a by dualizing the gauge invariant ﬁeld strength H of the antisymmetrix tensor ﬁeld through the relation: √ H µνρ = −( −g)−1 e2Φ µνρσ ∂σ a , (4.17) where Φ denotes the four dimensional dilaton and gµν denotes the (3 + 1) dimensional canonical metric deﬁned in equations (4.1), (4.2) respectively. It is convenient to combine the dilaton Φ and the axion ﬁeld a into a single complex scalar λ: λ = a + ie−Φ ≡ λ1 + iλ2 . (4.18)

7 Only the sixteen gauge ﬁelds in the Cartan subalgebra of the gauge group can develop vacuum expectation value since such vacuum expectation values do not generate any ﬁeld strength, and hence do not generate energy density. 8 For a review of this construction, see [1].

A. Sen: Duality Symmetries

265

At a generic point in the moduli space, where the scalars M take arbitrary vacuum expectation values, the non-abelian gauge symmetry of the ten dimensional theory is broken to its abelian subgroup U (1)16 . Besides these sixteen U (1) gauge ﬁelds we get twelve other U (1) gauge ﬁelds from components Gmµ , Bmµ (4 ≤ m ≤ 9, 0 ≤ µ ≤ 3) of the metric and the anti-symmetric tensor ﬁeld respectively. Let us denote these 28 U (1) gauge ﬁelds (after suitable normalization) by Aa (1 ≤ a ≤ 28). In terms of these µ ﬁelds, the low energy eﬀective action of the theory is given by [1,30–32,34]9 , S = ¯ √ ∂µ λ∂ν λ 1 µν 1 + g Tr(∂µ M L∂ν M L) d4 x −g R − g µν 2π 2(λ2 )2 8 1 1 a b a b − λ2 g µµ g νν Fµν (LM L)ab Fµ ν + λ1 g µρ g νσ Fµν Lab Fρσ , 4 4 (4.19)

a where Fµν is the ﬁeld strength associated with Aa , R is the Ricci scalar. µ and

F aµν =

1 √ ( −g)−1 2

µνρσ

a Fρσ .

(4.20)

This action is invariant under an O(6,22) transformation10 : M → ΩM ΩT , where Ω satisﬁes: ΩLΩT = L . (4.22) Aa → Ωab Ab , µ µ gµν → gµν , λ → λ, (4.21)

An O(6,22;Z) subgroup of this can be shown to be a T -duality symmetry of the full string theory [27]. This O(6,22;Z) subgroup can be described as follows. Let Λ28 denote a twenty eight dimensional lattice obtained by taking the direct sum of the twelve dimensional lattice of integers, and the sixteen dimensional root lattice of SO(32)11 . O(6,22;Z) is deﬁned to be the subset of O(6,22) transformations which leave Λ28 invariant, i.e. acting on any vector in Λ28 , produces another vector in Λ28 . It will be useful

9 The normalization of the gauge ﬁelds used here diﬀer from that in reference [1] by a factor of two. Also there we used α = 16 whereas here we are using α = 1. 10 O(p, q) denotes the group of Lorentz transformations in p space-like and q time-like dimensions. (These have nothing to do with physical space-time, which always has only one time-like direction.) O(p, q; Z) denotes a discrete subgroup of O(p, q). 11 More precisely we have to take the root lattice of Spin(32)/Z which is obtained 2 by adding to the SO(32) root lattice the weight vectors of the spinor representations of SO(32) with a deﬁnite chirality.

266

Unity from Duality: Gravity, Gauge Theory and Strings

for our future reference to undertstand why only an O(6,22;Z) subgroup of the full O(6,22) group is a symmetry of the full string theory. Since O(6,22;Z) is a T -duality symmetry, this question can be answered within the context of perturbative string theory. The point is that although at a generic point in the moduli space the massless string states do not carry any charge, there are massive charged states in the spectrum of full string theory. Since there are 28 charges associated with the 28 U (1) gauge ﬁelds, a state can be characterized by a 28 dimensional charge vector. With appropriate normalization, this charge vector can be shown to lie in the lattice Λ28 , i.e. the charge vector of any state in the spectrum can be shown to be an element of the lattice Λ28 . Since the O(6,22) transformation acts linearly on the U (1) gauge ﬁelds, it also acts linearly on the charge vectors. As a result only those O(6,22) elements can be genuine symmetries of string theory which preserve the lattice Λ28 . Any other O(6,22) element, acting on a physical state in the spectrum, will take it to a state with charge vector outside the lattice Λ28 . Since such a state does not exist in the spectrum, such an O(6,22) transformation cannot be a symmetry of the full string theory. In order to see a speciﬁc example of a T -duality transformation, let us consider heterotic string theory compactiﬁed on (S 1 )6 with one of the circles having radius R measured in the string metric, and the rest having unit radius. Let us also assume that there is no background gauge or antisymmetric tensor ﬁelds. Using the convention of reference [1] one can show that for this background −2 R I5 . (4.23) M (H) = R2 I5 I16 Consider now the O(6,22;Z) transformation with the matrix: 0 1 I5 . Ω= 1 0 I21 Using equation (4.21) we see that this transforms M (H) to 2 R I5 (H) −2 . R = M I5 I16

(4.24)

(4.25)

A. Sen: Duality Symmetries

267

Thus the net eﬀect of this transformation is R → R−1 . It says that the heterotic string theory compactiﬁed on a circle of radius R is equivalent to the same theory compactiﬁed on a circle of radius R−1 . For this reason R = 1 √ (i.e. R = α ) is known as the self-dual radius. Other O(6,22;Z) transformations acting on (4.23) will give rise to more complicated M (H) corresponding to a conﬁguration with background gauge and/or anti-symmetric tensor ﬁelds. Besides this symmetry, the equations of motion derived from this action can be shown to be invariant under an SL(2, R) transformation of the form [30, 35, 36]

a a b Fµν → (rλ1 + s)Fµν + rλ2 (M L)ab Fµν ,

λ→

gµν → gµν ,

M →M,

pλ + q , rλ + s (4.26)

where p, q, r, s are real numbers satisfying ps − qr = 1. The existence of such symmetries (known as hidden non-compact symmetries) in this and in other supergravity theories were discovered in early days of supergravity theories and in fact played a crucial role in the construction of these theories in the ﬁrst place [30, 113]. Since this SL(2,R) transformation mixes the gauge ﬁeld strength with its Poincare dual, it is an electric-magnetic duality transformation. This leads to the conjecture that a subgroup of this continuous symmetry group is an exact symmetry of string theory [1,36–42]. One might wonder why the conjecture refers to only a discrete subgroup of SL(2, R) instead of the full SL(2, R) group as the genuine symmetry group. This follows from the same logic that was responsible for breaking O(6,22) to O(6,22;Z); however since the SL(2, R) transformation mixes electric ﬁeld with magnetic ﬁeld, we now need to take into account the quantization of magnetic charges. We have already described the quantization condition on the electric charges. Using the usual Dirac-Schwinger-Zwanziger rules one can show that in appropriate normalization, the 28 dimensional magnetic charge vectors also lie in the same lattice Λ28 . Also with this normalization convention the electric and magnetic charge vectors transform as doublet under the SL(2, R) transformation; thus it is clear that the subgroup of SL(2, R) that respects the charge quantization condition is SL(2, Z). An arbitrary SL(2, R) transformation acting on the quantized electric and magnetic charges will not give rise to electric and magnetic charges consistent with the quantization law. This is the reason behind the conjectured SL(2, Z) symmetry of heterotic string theory on T 6 . Note that since this duality acts non-trivially on the dilaton and hence the string coupling, this is a non-perturbative symmetry, and cannot be veriﬁed order by order in perturbation theory. Historically, this is the ﬁrst example of a concrete

268

Unity from Duality: Gravity, Gauge Theory and Strings

duality conjecture in string theory. Later we shall review other tests of this duality conjecture. 4.3 Duality between heterotic on T 4 and type IIA on K3 The massless bosonic ﬁeld content of heterotic string theory compactiﬁed on T 4 can be found in a manner identical to that in heterotic string theory on T 6 . Besides the dilaton Φ(H) , we get many other massless scalar ﬁelds from the internal components of the metric, the anti-symmetric tensor ﬁeld and the gauge ﬁelds. In this case these scalars can be represented by a 24 × 24 matrix valued ﬁeld M (H) satisfying M (H) LM (H)T = L, where L = I4 I4 −I16 . (4.28) M (H)T = M (H) , (4.27)

We again use the convention that M (H) = I24 corresponds to compactiﬁca√ tion on (S 1 )4 with each S 1 having self-dual radius ( α = 1), without any background gauge ﬁeld or anti-symmetric tensor ﬁeld. At a generic point in the moduli space, where the scalars M (H) take arbitrary vacuum expectation values, we get a U (1)24 gauge group, with 16 gauge ﬁelds coming from the Cartan subalgebra of the original gauge group in ten dimensions, and eight other gauge ﬁelds from components Gmµ , Bmµ (6 ≤ m ≤ 9, 0 ≤ µ ≤ 5) of the metric and the anti-symmetric tensor ﬁeld respectively. Here xm denote the compact directions, and xµ denote the non-compact directions. Let us (H)a (H) (1 ≤ a ≤ 24). Finally, let gµν denote these 24 U (1) gauge ﬁelds by Aµ (H) and Bµν denote the canonical metric and the anti-symmetric tensor ﬁeld respectively. In terms of these ﬁelds, the low energy eﬀective action of the theory is given by, SH = 1 1 d6 x −g (H) R(H) − g (H)µν ∂µ Φ(H) ∂ν Φ(H) (2π)3 2 1 µν + g Tr(∂µ M (H) L∂ν M (H) L) 8 (H) 1 (H)b (H)a − e−Φ /2 g (H)µµ g (H)νν Fµν (LM (H) L)ab Fµ ν 4 (H) 1 (H) (H) (4.29) − e−Φ g (H)µµ g (H)νν g (H)ρρ Hµνρ Hµ ν ρ , 12

**A. Sen: Duality Symmetries
**

(H)a (H)a

269

where Fµν is the ﬁeld strength associated with Aµ , R(H) is the Ricci (H) (H) scalar, and Hµνρ is the ﬁeld strength associated with Bµν :

(H) Hµνρ =

1 (H) (H)b ∂µ Bνρ + A(H)a Lab Fνρ 2 µ

+ (cyclic permutations of µ, ν, ρ). (4.30)

**This action is invariant under an O(4,20) transformation: M (H) → ΩM (H) ΩT ,
**

(H) (H) Bµν → Bµν ,

A(H)a → Ωab A(H)b , µ µ

(H) (H) gµν → gµν ,

Φ(H) → Φ(H) ,

(4.31)

where Ω satisﬁes: ΩLΩT = L . (4.32)

Again as in the case of T 6 compactiﬁcation, only an O(4,20;Z) subgroup of this which preserves the charge lattice Λ24 is an exact T -duality symmetry of this theory. The lattice Λ24 is obtained by taking the direct sum of the 8 dimensional lattice of integers and the root lattice of Spin(32)/Z2. Let us now turn to the spectrum of massless bosonic ﬁelds in type IIA string theory on K3. In ten dimensions the massless bosonic ﬁelds in type IIA string theory are the metric gMN , the rank two anti-symmetric tensor BMN and the scalar dilation Φ coming from the NS sector, and a gauge ﬁeld AM and a rank three antisymmetric tensor ﬁeld CMNP coming from the RR sector. The low energy eﬀective action of this theory involving the massless bosonic ﬁelds is given by [90] SIIA = √ 1 1 d10 x −g R − g µν ∂µ Φ∂ν Φ 7 (2π) 8 1 −Φ/2 µµ νν ρρ 1 − e g g g Hµνρ Hµ ν ρ − e3Φ/4 g µµ g νν Fµν Fµ ν 12 4 1 Φ/4 µµ νν ρρ σσ − e g g g g Gµνρσ Gµ ν ρ σ 48 1 √ (4.33) − ( −g)−1 εµ0 ···µ9 Bµ0 µ1 Gµ2 ···µ5 Gµ6 ···µ9 , (48)2

where R is the Ricci scalar, and Fµν Hµνρ Gµνρ = = = ∂µ Aν − ∂ν Aµ , ∂µ Bνρ + cyclic permutations of µ, ν, ρ , ∂µ Cνρσ + Aµ Hνρσ + (−1)P · cyclic permutations , (4.34)

are the ﬁeld strengths associated with Aµ , Bµν and Cµνρ respectively. Upon compactiﬁcation on K3 we get a new set of scalar ﬁelds from the Kahler and

270

Unity from Duality: Gravity, Gauge Theory and Strings

complex structure moduli of K3. These can be regarded as deformations of the metric and give a total of 58 real scalar ﬁelds. We get 22 more scalar ﬁelds φ(p) by decomposing the antisymmetric tensor ﬁeld BMN along the (p) twenty two harmonic two forms ωmn in K3:

22

Bmn (x, y) ∼

p=1

(p) φp (x)ωmn (y) + · · · .

(4.35)

Here {xµ } and {y m } denote coordinates along the non-compact and K3 directions respectively. These eighty scalar ﬁelds together parametrize a coset O(4, 20)/O(4) × O(20) and can be described by a matrix M (A) satisfying properties identical to those of M (H) described in (4.27). This theory also has twenty four U (1) gauge ﬁelds. 22 of the gauge ﬁelds arise from the components of the three form ﬁeld CMNP :

22

Cmnµ (x, y) =

p=1 (p)

(p) ωmn (y)A(p) (x) + . . . . µ

(4.36)

Aµ deﬁned in (4.36) behaves as gauge ﬁelds in six dimensions. One more gauge ﬁeld comes from the original RR gauge ﬁeld Aµ . The last one Aµ comes from dualizing Cµνρ : G∼

∗

(dA) ,

(4.37)

where ∗ denotes Poincare dual in six dimensions. Together we shall denote (A)a for 1 ≤ a ≤ 24. Besides these ﬁelds, the theory these gauge ﬁelds by Aµ contains the canonical metric and the anti-symmetric tensor ﬁeld which we (A) (A) shall denote by gµν and Bµν respectively. The action involving these ﬁelds is given by, SA = 1 1 d6 x −g (A) R(A) − g (A)µν ∂µ Φ(A) ∂ν Φ(A) 3 (2π) 2 1 µν + g Tr(∂µ M (A) L∂ν M (A) L) 8 1 (A) (A)b (A)a − eΦ /2 g (A)µµ g (A)νν Fµν (LM (A) L)ab Fµ ν 4 (A) 1 (A) (A) − e−Φ g (A)µµ g (A)νν g (A)ρρ Hµνρ Hµ ν ρ 12 1 (A) (A)a (4.38) − εµνρδ η ( −g (A) )−1 Bµν Fρδ Lab F (A)b , η 16

(A)a

where Fµν is the ﬁeld strength associated with Aµ , R(A) is the Ricci (A) (A) scalar, and Hµνρ is the ﬁeld strength associated with Bµν :

(A) (A) Hµνρ = ∂µ Bνρ + (cyclic permutations of µ, ν, ρ) .

(A)a

(4.39)

A. Sen: Duality Symmetries

271

In writing down the above action we have used the convention that M (A) = I24 corresponds to compactiﬁcation on a speciﬁc reference K3, possibly with speciﬁc background Bmn ﬁelds. This action has an O(4,20) symmetry of the form: M (A) → ΩM (A) ΩT ,

(A) (A) Bµν → Bµν ,

A(A)a → Ωab A(A)b , µ µ

(A) (A) gµν → gµν ,

Φ(A) → Φ(A) , ΩLΩT = L .

(4.40)

where Ω satisﬁes: (4.41)

An O(4,20;Z) subgroup of this can be shown to be an exact T -duality symmetry of string theory [126]. The lattice Λ24 which is preserved by this O(4,20;Z) subgroup of O(4,20) is not the lattice Λ24 deﬁned earlier, but is in general an O(4,20) rotation of that lattice: Λ24 = Ω0 Λ24 . (4.42) Ω0 depends on the choice of the special reference K3 mentioned earlier. It is now a straightforward exercise to show that the equations of motion and the Bianchi identities derived from (4.29) and (4.38) are identical if we use the following map between the heterotic and the type II variables [23,43]:

(H) (A) gµν = gµν ,

M (H) = ΩM (A) ΩT ,

Φ(H) = −Φ(A) ,

A(H)a = Ωab A(A)a , µ µ 1 µνρδ η (A) (H) (H)µνρ −g (H) exp(−Φ )H = ε Hδ η . 6

(4.43)

where Ω is an arbitrary O(4,20) matrix. This leads to the conjectured equivalence between heterotic string theory compactiﬁed on T 4 and type IIA string theory compactiﬁed on K3 [23]. But clearly the two theories cannot be equivalent for all Ω since in the individual theories the O(4,20) symmetry is broken down to O(4,20;Z). Ω can be found (up to an O(4,20;Z) transformation) by comparing the T -duality symmetry transformations in the two theories. To do this let us note that according to equation (4.43) a transformation M (H) → ΩM (H) ΩT will induce a transformation M (A) → (Ω−1 ΩΩ)M (A) (Ω−1 ΩΩ)T . (4.44) Thus if Ω preserves the lattice Λ24 , Ω−1 ΩΩ should preserve the lattice Λ24 = Ω0 Λ24 . This happens if we choose: Ω = Ω−1 . 0 (4.45) Note again that there is a relative minus sign that relates Φ(H) and Φ(A) , showing that the strong coupling limit of one theory corresponds to the weak coupling limit of the other theory.

or by explicit computation in string theory. Actually it turns out that there is no simple covariant action for this low energy theory. ps − qr = 1 .R) transformation [44]: λ→ gµν pλ + q Bµν p . is 12 Note that this ﬁeld λ has no relation to the ﬁeld λ deﬁned in Section 4.2 for heterotic string theory on T 6 . Z ) self-duality of type IIB in D = 10 As described in Section 2. Bµν (4.48) The existence of this SL(2. the massless bosonic ﬁelds in type IIB string theory come from two sectors. an anti-symmetric tensor ﬁeld Bµν . which are in fact just the equations of motion of type IIB supergravity. and a rank four anti-symmetric tensor ﬁeld Dµνρσ whose ﬁeld strength is self-dual. and so there is no scope for the quantum corrections to change the form of the action.R) symmetry in the type IIB supergravity theory led to the conjecture that an SL(2. q.46) The low energy eﬀective action in this theory can be determined either from the requirement of supersymmetry. r. Basically supersymmetry determines the form of the equations of motion to this order in the derivatives completely. The main point is that these equations of motion are covariant (in the sense that they transform into each other) under an SL(2. and a scalar ﬁeld Φ known as the dilaton. although both transform as modulus under the respective SL(2. we shall not explicitly write down the equations of motion. non-renormalization theorems tell us that this is exact to this order in the space-time derivatives. s to be integers instead of arbitrary real numbers. (4. For the sake of brevity.4 SL(2 .47) where p. but there are covariant ﬁeld equations [44]. Dµνρσ → Dµνρσ . s are real numbers satisfying. Although in string theory this low energy theory is derived from the tree level analysis. It is often convenient to combine the axion and the dilaton into a complex scalar ﬁeld λ as follows12 : λ = a + ie−Φ/2 ≡ λ1 + iλ2 .1.Z) subgroup of this SL(2. (4. another rank two anti-symmetric tensor ﬁeld Bµν . q s Bµν . The NS sector gives the graviton described by the metric gµν . . r.272 Unity from Duality: Gravity. obtained by restricting p. → Bµν r rλ + s → gµν . q. The RR sector contributes a scalar ﬁeld a sometimes called the axion.R). Gauge Theory and Strings 4. − Neveu-Schwarz−Neveu-Schwarz (NS) and Ramond-Ramond (RR).Z) duality transformations in the two theories.

This means that the Bµν charge must be quantized in integer units.49) with p.Z) transformation. leaving the other massless bosonic ﬁelds invariant. Sen: Duality Symmetries 273 a symmetry of the full string theory [23]. it cannot be veriﬁed order by order in string perturbation theory. acting on the massless bosonic sector ﬁelds. • Ω: this is the world-sheet parity transformation mentioned in Section 2.R) transformation gives a state with p units of Bµν charge and r units of Bµν charge. it changes the sign of Bµν . as the spectrum of string theory does not contain fractional strings carrying a fraction of the charge carried by the elementary string. Thus p must be an integer.A. but leaves gµν .R) to SL(2. q. r. it changes the sign of a.1 that exchanges the left. An elementary string is known to carry Bµν charge. Z) matrices r s with p. Besides this non-perturbative SL(2. In particular.Z) is a symmetry of string theory. the SL(2. (4. In particular.and the right-moving sectors of the world-sheet.47) we see that acting on an elementary string state carrying one unit of Bµν charge.R) for which p is always an integer consists of matrices of the form p α−1 r αq s .Z) is the maximal possible subgroup of SL(2. r. type IIB theory has two perturbatively veriﬁable discrete Z2 symmetries. it carries exactly one unit of Bµν charge. R) map q trices consistent with charge quantization are the SL(2. leaving the other ﬁelds invariant. In suitable normalization convention. since SL(2.R) that can be a symmetry of the full string theory. Comparing this with the action of the SL(2. s integers satisfying ps − qr = 1. s integers satisfying (ps − qr) = 1.Z) transformation laws of the . Bµν and Φ invariant. It is easy to see that the maximal subgroup of SL(2. We shall see later how one can ﬁnd non-trivial evidence for this symmetry. whose vacuum expectation value represents the string coupling constant. a and Dµνρσ . and α a ﬁxed constant. we see that the eﬀect of (−1)FL · Ω is to change of sign of Bµν and Bµν .Z) acts non-trivially on the dilaton. They are as follows: • (−1)FL : it changes the sign of all the Ramond sector states on the left moving sector of the world-sheet. From this description. From (4. Absorbing α into a redeﬁnition of Bµν we see that the subgroup of SL(2. but does not prove that SL(2. Bµν and Dµνρσ . Acting on the massless bosonic sector ﬁelds. Note that this argument only shows that SL(2.Z) can be seen as follows. q. The breaking of SL(2.

As mentioned earlier. 101]. (4. Z) × SL(3. The open string states of type I string theory are the analogs of twisted sector states in an orbifold. there is a T -duality that relates type IIA on a circle to type IIB on a circle of inverse radius. we see that (−1)FL · Ω can be identiﬁed with the SL(2. 4. 5. Z) Note that besides the full duality group. 5. 3. Z) SO(8. 8. 7. 6. which must be added to the theory in order to maintain ﬁniteness. 4. The simplest example of an orientifold is type IIB string theory modded out by Ω.274 Unity from Duality: Gravity. Gauge Theory and Strings massless bosonic sector ﬁelds. Z) SO(7. the self-duality groups of type IIA and type IIB theories compactiﬁed on an n-dimensional torus T n will be identical. Z) × SL(2. we are led to many other duality conjectures in theories with 16 or more supersymmetry generators.5 Other examples Following the same procedure. where some of the elements of the group involve Ω. Z) SL(2. The closed string sector of type I theory consists of the Ω invariant states of type IIB string theory. Here we shall list the main series of such duality conjectures. Z) SO(6. Z) E6(6) (Z) E7(7) (Z) E8(8) (Z) E8(8) (Z) T -duality Group − SL(2.50) This information will be useful to us later. This corresponds to type I string theory. Z) SO(5. Thus for n ≥ 1. We begin with the self duality groups of type II string theories compactiﬁed on tori of diﬀerent dimensions. studying symmetries of the eﬀective action together with charge quantization rules. We now list the conjectured self-duality groups of type IIA/IIB string theory compactiﬁed on T n for diﬀerent values of n [23]: D = (10 − n) 9 8 7 6 5 4 3 2 Full Duality Group SL(2. are known as orientifolds [100. Z) SO(5. Theories obtained by modding out (compactiﬁed) type IIB string theory by a discrete symmetry group. Z) SO(4.Z) transformation: −1 −1 . Z) SO(3. namely. Z) SL(5. we have also displayed the T duality group of each theory which can be veriﬁed order by order in string .

Z) O(5. 19. In this case one expects a very large duality symmetry group based on hyperbolic Lie algebra [103].Z) subgroup of this is an exact self-duality symmetry of the type IIB string . 23. Z) O(2. 21. G for any group G denotes the loop group of G based on the corresponding aﬃne algebra and G(Z) denotes a discrete subgroup of this loop group. the second column of the above table also represents the duality symmetry group of type I string theory on T n . upon compactiﬁcation on a circle. these symmetries were discovered in the early days of supergravity theories. Although there are two distinct heterotic string theories in ten dimensions. 24. Z) Since type I and SO(32) heterotic string theories are conjectured to be dual to each other in ten dimensions. the low energy eﬀective ﬁeld theory is invariant under the full continuous group [45].21. Z) × SL(2. Z) O(4.21) symmetry [43]. and En(n) (Z) denotes a discrete subgroup of En(n) . Z) O(3. Z) O(5. there is no perturbatively realised self-duality group (except trivial transformations which are part of the SO(32) gauge group and the group of global diﬀeomorphisms of T n ). 19. and were known as hidden non-compact symmetries. in the case of type I string theory. 7. 8. The eﬀective action of type IIB string theory compactiﬁed on K3 has an SO(5. Note that we have stopped at D = 2. Z) O(8. Z) O(7.A. Z) T -duality Group O(1. 18. Sen: Duality Symmetries 275 perturbation theory. the two heterotic string theories can be shown to be related by a T -duality transformation. 18. Z) O(8. 22. Z) O(8. En(n) denotes a non-compact version of the exceptional group En for n = 6. 20. Z) O(3. upon compactiﬁcation on T n . but charge quantization breaks this symmetry to its discrete subgroup. Z) O(6. Z) O(2. Next we turn to the self-duality conjectures involving compactiﬁed heterotic string theories. 22. As a result. As noted before. 17. which is not well understood to this date. 21. which leads to the conjecture that an SO(5. In each of the cases mentioned. Z) O(4. 24. both of them will have the same self-duality group. Z) O(6. We could in principle continue this all the way to D = 1 where all space-like directions are compactiﬁed. 20. However. 17. 24. We now display this self-duality group in various dimensions: D = (10 − n) 9 8 7 6 5 4 3 2 Full Duality Group O(1.

we can perform a much more precise test by analysing the spectrum of BPS states in the theories. there are other theories with 16 or more supercharges obtained from non-geometric compactiﬁcation of heterotic/type II string theories [46–48]. Its value lies in its simplicity. The conjectured duality between type IIA string theory compactiﬁed on K3 and heterotic string theory compactiﬁed on T 4 has already been discussed before.276 Unity from Duality: Gravity. 5 Precision test of duality: Spectrum of BPS states Analysis of the low energy eﬀective action. But once we have arrived at a duality conjecture based on the analysis of the low energy eﬀective action. type IIA on K3 × T n is equivalent to type IIB on K3 × T n . Gauge Theory and Strings theory on K3. Finally. Thus the self-duality symmetry groups in these theories can be read out from the second column of the previous table displaying the self-duality groups of heterotic string theory on T n . this is not the only part of the full eﬀective action which satisfy such a non-renormalization theorem. Indeed. This has an analog in the theory of representations of the Lorentz group. provides us with only a crude test of duality. Besides the theories discussed here. Later we shall also describe a more systematic way of “deriving” various duality conjectures from some basic set of dualities. and are characterized by two important properties: • They belong to a supermultiplet which has typically less dimension than a non-BPS state. and are known as Green-Schwarz terms. BPS states are states which are invariant under part of the supersymmetry transformation. Adler-Bardeen theorem guarantees that they are not renormalized beyond one loop. The duality symmetry groups of these theories can again be guessed from an analysis of the low energy eﬀective ﬁeld theory and the charge quantization conditions. Quite often the eﬀective action contains another set of terms satisfying non-renormalization theorems. most of the duality conjectures in string theory were arrived at by analysing the symmetries of the low energy eﬀective action. They are required for anomaly cancellation. as discussed in the last section. Due to the equivalence of type IIB on S 1 and type IIA on S 1 . type IIA/IIB on K3 × T n are dual to heterotic string theory on T n+4 for n ≥ 1. Although in this section I have focussed on duality symmetries of the low energy eﬀective action which satisfy a non-renormalization theorem as a consequence of space-time supersymmetry. Thus for example . where massless states form a shorter representation of the algebra than massive states. but I shall not discuss it in this article. due to the conjectured duality between type IIA on K3 and heterotic on T 4 . These terms have also been used eﬀectively for testing various duality conjectures [127].

A. Qβ } = = δαβ . which is 2N/2 dimensional. This relation between the mass and the charge is known as the BPS mass formula. (5. 2. Qβ } = fαβ (m. a spin 1 representation of the Lorentz algebra containing only two states must be necessarily massless. We shall now explain the origin of these two properties [29]. Q denotes various gauge charges carried by the particle. Acting on a single particle state at rest. for 1 ≤ α.) Such states would correspond to non-BPS states. we can bring f into the identity matrix. Q and {y}. β ≤ M . f has (N − M ) zero eigenvalues for some M < N .3) We can form an irreducible representation of this algebra by taking all states to be annihilated by Qα for α > M . Sen: Duality Symmetries 277 a photon has only two polarizations but a massive vector particle has three polarizations. and {y} denotes the coordinates labelling the moduli space of the theory13 . known as central charges. In this case by taking appropriate linear combinations of Qα we can diagonalize f .g. Thus in this basis the supersymmetry algebra has the form: {Qα . (5. We shall now consider the following distinct cases: 1. {y}) . the supersymmetry algebra takes the form: {Qα . appear in the algebra. Here m denotes the rest mass of the particle. Suppose the theory has N real supersymmetry generators Qα (1 ≤ α ≤ N ). e. Thus the single particle states under consideration form a representation of this Cliﬀord algebra. • The mass of a BPS state is completely determined by its charge as a consequence of the supersymmetry algebra. . In that case the states will 13 Only speciﬁc combinations of Q and {y}. by taking linear combinations of the Qα we can bring the algebra into the form: {Qα .2) This is the N dimensional Cliﬀord algebra.1) where fαβ is a real symmetric matrix which is a function of its arguments m. fαβ has no zero eigenvalue. (5. 0 for α or β > M . Q. In this case. (We are considering the case where N is even. Qβ } = δαβ . By a further appropriate rescaling of Qα . This statement also has an analog in the theory of representations of the Lorentz algebra.

This is known as the short representation of this superalgebra. Consider type IIB string theory compactiﬁed on a circle S 1 . these states are invariant under part of the supersymmetry algebra generated by Qα for α > M . the matrix f must have some zero eigenvalues. We can also have states that break 3/4 of the supersymmetries14 . But if we consider states that are invariant under half of the supercharges. These will form a 212 = 256 × 16 dimensional representation. i. Gauge Theory and Strings form a representation of an M dimensional Cliﬀord algebra generated by Qα for 1 ≤ α ≤ M . We can get diﬀerent kinds of BPS states depending on the value of M . This representation is 2M/2 dimensional for M even. Since string coupling is also one of 14 It turns out that these states can exist only for n ≥ 5. These are known as BPS states. As another example. charges Q and the moduli {y}. This in turn. . This constraint arises due to the fact that the unbroken supersymmetry generators must form a representation of the little group SO(9 − n) of a massive particle in (10 − n) dimensional space-time. These are known as long multiplets. we see that these are lower dimensional representations compared to that of a generic non-BPS state. In each case there is a speciﬁc relation between the mass and the various charges carried by the state. Since M < N . This theory also has BPS states breaking half the space-time supersymmetry. and is the origin of the BPS formula relating the mass and the charge of the particle. The original theory has 16 supercharges. Before we proceed. For these states M = 16 and hence we have 28 = 256 dimensional representation of the supersymmetry algebra. depending on the number of supersymmetry generators that leave the state invariant. and are known as short multiplets.278 Unity from Duality: Gravity. From this discussion it is clear that in order to get a BPS state. Thus a generic non-BPS state will belong to a 28 = 256 dimensional representation of the supersymmetry algebra. Furthermore. BPS states are further characterized by the property that the degeneracy of BPS states with a given set of charge quantum numbers is independent of the value of the moduli ﬁelds {y}. Thus a generic non-BPS supermultiplet is 216 = (256)2 dimensional. gives a constraint involving mass m. These belong to a 64 dimensional representation of the supersymmetry algebra known as intermediate states. These states are known as ultra-short multiplets. then they belong to a 24 = 16 dimensional representation of the supersymmetry algebra. We shall discuss this relation as well as the origin of these BPS states in more detail later. We can also have BPS states breaking 3/4 of the space-time supersymmetry (M = 24). let us illustrate the preceeding discussion in the context of a string theory. consider heterotic string theory compactiﬁed on an n-dimensional torus T n .e. The total number of supersymmetry generators in this theory is 32.

3. Sen: Duality Symmetries 279 the moduli of the theory. This is the key property of the BPS states that makes them so useful in testing duality. 62. Thus we see that the degeneracy of ultra-short multiplets cannot change as we change the moduli of the theory. we can now adapt the following strategy to carry out tests of various duality conjectures using the spectrum of BPS states in the theory: 1. Make a conjectured duality transformation.A. Given this result. The question that we shall be asking is: can the ultrashort multiplet become a long (or any other) multiplet as we change the moduli? If we assume that the total number of states does not change discontinuously. as long as it stays ultra-short. its mass is determined by the BPS formula. Thus these states must arise as solitons/composite states. but with quantum numbers that are not present in the spectrum of elementary string states. Suppose the theory has an ultra-short multiplet at some point in the moduli space. but it can be applied to any other theory. Now let us change the moduli. we require that the mass of the BPS state should stay away from the continuum. this implies that the degeneracy at any value of the string coupling is the same as that at weak coupling. Identify BPS states in the spectrum of elementary string states. then this is clearly not possible since other multiplets have diﬀerent number of states. This will provide a non-trivial test of the corresponding duality conjecture. We shall discuss this in the context of the speciﬁc example of type IIB string theory compactiﬁed on S 1 . This requires that the mass of a BPS state should be strictly less than the total mass of any set of two or more particles carrying the same total charge as the BPS state. We shall mainly follow [51. so let us review the argument leading to this property [29]. since otherwise the counting of states is not a well deﬁned procedure. 2. A similar argument can be given for other multiplets as well. The spectrum of these BPS states can be trusted at all values of the coupling even though it is calculated at weak coupling. . Thus as long as the spectrum varies smoothly with the moduli (which we shall assume). Note that for this argument to be strictly valid. Try to explicitly verify the existence of these solitonic states with degeneracy as predicted by duality. We shall now illustrate this procedure with the help of speciﬁc examples. an ultra-short multiplet stays ultra-short as we move in the moduli space [88]. This typically takes a BPS state in the spectrum of elementary string states to another BPS state. 65]. Furthermore.

In this subsection we shall see how one can test this conjecture by examining the spectrum of BPS states. and taking the radius of this S 1 to be the self-dual radius. if we set the internal components of the original ten dimensional gauge ﬁelds to zero. The conjectured SL(2. . heterotic string theory compactiﬁed on T 6 is conjectured to have an SL(2.4) 15 Independently of string theory. As discussed in Section 4. Thus for example. With the choice of suitable normalization convention.2. g and θ are related to the vacuum expectation value of the ﬁeld λ deﬁned in (4. and the vacuum angle θ that multiplies the topological term Tr(F F ) involving the gauge ﬁeld.). we can analyse the spectrum near some particular point in the moduli space. Let us consider a special point in the moduli space where an SU (2) gauge symmetry is restored. at a generic point in the moduli space the unbroken gauge group is U (1)28 .280 Unity from Duality: Gravity. Gauge Theory and Strings 5. Thus by testing the duality invariance of the spectrum of this N = 4 supersymmetric SU (2) gauge theory we can test the conjectured SL(2. The N = 4 supersymmetric SU (2) gauge theory has a vector. But there are special points in this moduli space where we get enhanced non-abelian gauge group [107].2. Z ) S-duality in heterotic on T 6 and multi-monopole moduli spaces As discussed in Section 4. 50]. scalar self-interaction etc. Z) symmetry15 . Z) duality of the heterotic string theory will require the N = 4 supersymmetric SU (2) gauge theory to have this SL(2. six massless scalars and four massless Majorana fermions in the adjoint representation of SU (2) [50]. Yukawa.1 SL(2 . the existence of a strong-weak coupling duality in this theory was conjectured earlier [49. Z) symmetry of heterotic string theory. Since the BPS spectrum does not change as we change the moduli. This can be done for example by taking a particular S 1 in T 6 to be orthogional to all other circles. In that case the eﬀective ﬁeld theory at energies much below the string scale will be described by an N = 4 supersymmetric SU (2) gauge theory.Z) duality symmetry. taking the components of the gauge ﬁelds along this S 1 to be zero. we get unbroken E8 × E8 or SO(32) gauge symmetry. together with a set of decoupled N = 4 supersymmetric U (1) gauge theories and N = 4 supergravity.18) through the relation: λ = 4π θ +i 2 · 2π g (5. The form of the Lagrangian is ﬁxed completely by the requirement of N = 4 supersymmetry up to two independent parameters − the coupling constant g that determines the strength of all interactions (gauge.

we can easily work out its action ne on the charge quantum numbers [1].6) does not break supersymm metry. The answer is nm ne nm → p r q s ne nm .7) for appropriate choice of sign convention for ne and nm . Thus acting on an 1 p state it produces a state. These break half of the 16 space-time supersymmetry generators and hence form a 28/2 = 16 dimensional representation of the supersymmetry algebra. these states become massless and form part of the SU (2) gauge multiplet. m (5. The BPS states in the spectrum of elementary particles in this theory are the heavy charged bosons W ± and their superpartners. These states can be found explicitly in the spectrum of elementary string states from the sector containing strings with one unit of winding and one unit of momentum along the special S 1 that is responsible for the enhanced SU (2) gauge symmetry.6) Vacuum expectation values of φα of the form (5. When SU (2) is broken to U (1) by the vacuum expectation value of φm . the electric charge quantum number ne and the magnetic charge quantum number nm .A. the spectrum of solitons in this theory is characterized by two quantum numbers.5) This vanishes for φα = am δα3 . We ne shall denote such a state by . As we approach the point in the moduli space where this special S 1 has self-dual radius. Sen: Duality Symmetries 281 The potential involving the six adjoint representation scalar ﬁelds φα (1 ≤ m α ≤ 3. 1 ≤ m ≤ 6) is proportional to ( m<n α αβγ β γ 2 φm φn ) . In this notation the elementary W + nm 1 boson corresponds to a state. The parameters {am } correspond to the vacuum expectation values of a subset of the scalar moduli ﬁelds M in the full string theory. Z) 0 transformation (4. but breaks the gauge group SU (2) to U (1). but the√ compared to the string scale (|am | << ( α )−1 ) for all m). so that gravity is still decoupled from this gauge theory. We shall work in a region in the moduli scale of breaking of SU (2) is small space where am = 0 for some m. (5. normalized so that ne and nm are both integers. From the relation ps − qr = 1 satisﬁed 0 r . (5.26) on the gauge ﬁelds. By studying the action of the SL(2.

These can be set to zero by working in the rest frame of the monopole. We can divide this set into the “center of mass” coordinates containing four bosonic and eight fermionic coordinates. The fourth coordinate is periodically identiﬁed and hence its conjugate momentum is quantized in integer units. The total number of fermionic collective coordinates can be computed from an index theorem and is equal to 8r [106]. Each solution also has inﬁnite number of vibrational modes with non-zero frequency. correspond to the spatial location and the U (1) phase of each of the r monopoles. and in fact. exactly as predicted by SL(2. this system has 4r bosonic collective coordinates. and the fourth one is an angular variable describing the U (1) phase of the monopole. Z) [50]. we need to quantize these collective excitations and look for supersymmetric ground states of the corresponding quantum mechanical system. Z) matrix. and the “relative coordinates” containing 4(r − 1) bosonic and 8(r − 1) . There are eight fermionic zero modes. Gauge Theory and Strings by an SL(2. the theory must contain a unique short multiplet with p charge quantum numbers [51]. we can easily see that p and r are relatively prime. These form an eight dimensional Cliﬀord algebra.282 Unity from Duality: Gravity. As has already been said. r We can now directly examine the solitonic sector of the theory to check this prediction. Let us now turn to the analysis of states with r > 1 [51]. Z) duality predicts that for every p and r relatively prime. Thus the ground state has 24 = 16-fold degeneracy. for a given r. The momenta conjugate to the ﬁrst three coordinates correspond to the components of the physical momentum of the particle. which. This has four bosonic collective coordinates. there is a 4r parameter non-singular solution with r units of total magnetic charge [52. three of which correspond to the physical position of the monopole in the three dimensional space. This integer p corresponds to the electric charge quantum number ne . Thus the states obtained by quantizing the bosonic sector of p the theory has charge quantum numbers for all integer p. States with r = 1 come from one monopole solution. we can ﬁnd integers q and s satisfying ps − qr = 1. Thus SL(2. 1 The degeneracy comes from quantizing the fermionic sector. These 4r parameters correspond to the bosonic collective excitations of this system [105]. when the monopoles are far away from each other. In order to study the spectrum of BPS solitons. which describe the result of applying the eight broken supersymmetry generators on the monopole solution. 104]. Furthermore for every p and r relatively prime. but excitations of these modes are not relevant for ﬁnding supersymmetric ground states. The theory contains classical monopole solutions which break half of the supersymmetries of the original theory. These solutions are non-singular everywhere.

55]. although they decouple locally. among all the possible states. (5. This can be understood as follows. we see that the wave-function involving the relative coordinates must pick up a phase of exp(−2πip/r) under this Zr transformation. The quantum mechanics of the bosonic and fermionic relative coordinates can be regarded as that of a supersymmetric particle moving in this moduli space. 53]. 105]. Since the part of the wave-function involving the coordinate of S 1 picks up a phase exp(2πip/r) under this Zr . However. and Mr by the relative coordinates. the total wave-function must be invariant under this Zr transformation. we need to turn to the quantum mechanics of the relative coordinates. • Normally the part of the wave-function involving the relative coordinates will be a function on Mr . Sen: Duality Symmetries 283 fermionic coordinates. known as the relative moduli space of r monopoles [52. p r being the momentum conjugate to the overall U (1) phase. They are listed below: • First of all. S 1 by the overall U (1) phase.A. It turns out that the bosonic coordinates in the relative coordinate system describe a non-trivial 4(r − 1) dimensional manifold. There are several subtleties with this system. not necessarily prime relative to r. consistent with the fact that a short multiplet is 16-fold degenerate. • Finally. There is an identiﬁcation of points in the product space R3 × S 1 × Mr by a Zr transformation that acts as a shift by 2π/r on S 1 and as a diﬀeomorphism on Mr without any ﬁxed point [52. since the total wave-function is a product of the wave-function of the center of mass system and the relative system. The quantization of the center of mass system gives p states carrying charge quantum numbers with 16-fold degeneracy.8) where R3 is parametrized by the center of mass location. the ones saturating Bogomol’nyi bound correspond to harmonic diﬀerential forms on Mr . But it turns out that the eﬀect of the 8(r − 1) fermionic degrees of freedom in the quantum mechanical system makes the wave-function a diﬀerential form of arbitrary rank on Mr [54. This shows that the degeneracy is always a multiple of 16. 53. The full moduli space has the structure [52]: (R3 × S 1 × Mr )/Zr . in order to determine the number of short multiplets for a given value of p. At this stage p can be any integer. Due to this identiﬁcation. It can be shown that the Hamiltonian of the . the center of mass and the relative coordinates do not completely decouple.

Besides the BPS states discussed here. these 28 components of (kL .e. For r > 2 the analysis is more complicated since the metric in the multimonopole moduli space is not known. 59]. since the fermionic degrees of freedom associated with the center of mass coordinates supply the necessary 16-fold degeneracy.and the right-moving sector on the Fock vacuum. The tree level mass formula for an elementary string state in the NS sector is given by16 . Z) duality.57].and the left-moving sectors of the world-sheet. The Fock vacuum. kR ) are just appropriate linear combinations of the charges carried by the state under the 28 U (1) gauge ﬁelds.284 Unity from Duality: Gravity. According to the prediction of SL(2. it must be a harmonic form on Mr . Thus the existence of a short multiplet of charge quantum numbers m2 = 4 λ2 2 4 1 kR + NR − = 2 2 λ2 2 kL + NL − 1 . Also it turns out that the BPS mass formula is saturated by contribution from the center of mass coordinates.9) where NR and NL denote respectively the oscillator levels of the state in the right. a generic state is created by applying oscillators from the left. in turn. . In the world-sheet theory.and 16 In this and all subsequent mass formula λ should really be interpreted as the vacuum 2 expectation value of λ2 . Hence in order to get a BPS state. 56. However general arguments showing the existence of the necessary harmonic forms has been given [58. 2 (5. From the viewpoint of the spacetime theory. kR ) specifying the charges (momenta) associated with the six right-handed and twenty two left-handed currents on the world-sheet. p r would require the existence of a harmonic form on Mr that picks up a phase of exp(2πip/r) under the action of Zr . Z) such a harmonic form should exist only for p and r relatively prime. therby verifying the existence of the states predicted by SL(2. and not for other values of p [51]. Thus for every harmonic diﬀerential form we get a short multiplet. For r = 2 the relevant harmonic form can be constructed explicitly [51. the part of the wave-function involving the relative coordinates must be an eigenstate of the corresponding Hamiltonian with zero eigenvalue i. is characterized by a pair of vectors (kL . In the above equation ¯ the terms in the square bracket denote the total contribution to L0 and L0 from the oscillators. Gauge Theory and Strings relative coordinates correspond to the Laplacian on M. and the vacua in the right. the internal momenta. the spectrum of elementary string states in the heterotic string theory on T 6 contains many other BPS states.

(In the covariant formulation these can be traced to the contributions from the world-sheet ghost ﬁelds).A. The next interesting class of states are the ones with NL = 1. The additive factor of −1/2 and −1 can be interpreted as the contributions ¯ to L0 and L0 from the vacuum. Z) transformation relates these states to appropriate magnetically charged states with r units of magnetic charge and p units of electric charge for p and r relatively prime. . only those states which satisfy the constraint [64] NR = 1 . 2 R (5.12) The BPS states discussed earlier − the ones which can be regarded as the massive gauge bosons of a spontaneously broken non-abelian gauge theory − correspond to the NL = 0 states in this classiﬁcation. An SL(2.10) correspond to BPS states (short multiplets).11) The degeneracy d(NL ) of short multiplets for a given set of kL . Z) self-duality symmetry of the 17 In counting degeneracy we are only counting the number of short multiplets. this is consistent with their description as heavy gauge bosons in an N = 4 supersymmetric gauge theory. Normally we do not have the factor of λ−1 in the mass formula since the formula refers to the ADM mass measured 2 in the string metric Gµν = λ−1 gµν . Sen: Duality Symmetries 285 the left-moving sectors respectively. since it is gµν and not Gµν that remains invariant under a duality transformation. From equations (5. This is more convenient for discussing duality invariance of the spectrum. Thus the SL(2. It turns out that of the full set of elementary string states. 2 (5.9) we see that for these states NL = 1 2 2 (k − kL ) + 1 . From (5.12) we see that they have degeneracy 2417 . But here (and in the rest of the article) 2 we quote the ADM mass measured in the canonical metric gµν . and ignoring the trivial factor of 16 that represents the degeneracy within each short multiplet.12) we see that we have only one short multiplet for states with this quantum number. From equation (5. kR is determined by the number of ways a level NL state can be created out of the Fock vacuum by the 24 left-moving bosonic oscillators (in the light-cone gauge) − 8 from the transverse bosonic coordinates of the string and 16 from the bosonization of the 32 left-moving fermions on the world-sheet − and is given by the formula: ∞ d(NL )q NL = NL =0 1 · (1 − q n )24 n=1 ∞ (5.

and one ﬁnds that these solitons have exactly the correct degeneracy 24. The main problem is that unlike the NL = 0 states. One of the main problems with this approach has been that unlike the example discussed in this section. and hence we cannot unambiguously determine the dynamics of collective coordinates of these solitons just from the low energy eﬀective ﬁeld theory. kL and kR denote the left and the right-moving momenta respectively.2 SL(2 . Z) duality turn out to be singular objects. In this section I shall discuss the consequence of this conjectured symmetry for the spectrum of BPS states in type IIB string theory compactiﬁed on a circle S 1 . Nevertheless. √ 1/4 1/4 kR = (kλ2 /R + wR/λ2 )/ 2 . k/R denotes the momentum along S 1 with k being an integer. 63]. 72–74]. on the Fock vacuum of the world-sheet theory carrying these quantum numbers. most of these other solutions are either singular. to which we now turn. Verifying the existence of these solitonic states turns out to be quite diﬃcult [60]. see [62. 26. 25.13) where R denotes the radius of S 1 measured in the ten dimensional canonical metric. The spectrum of elementary string states in this theory are characterized by two charges kL and kR deﬁned as: √ 1/4 1/4 kL = (kλ2 /R − wR/λ2 )/ 2. 21. 5. The mass formula for any state in this tower. The situation changed after the advent of D-branes. Gauge Theory and Strings heterotic string theory predicts the existence of 24-fold degenerate solitonic states with these charge quantum numbers. As usual we have set α = 1. the problem has now been solved for r = 1 [70. In the world-sheet theory describing ﬁrst quantized string theory. 61]. 23. Z) duality symmetry group. As a result. measured in the ten dimensional . both from the left. There are inﬁnite tower of states with this quantum number. the solitonic states (known as H-monopoles) which are related to the NL = 1 states by SL(2.286 Unity from Duality: Gravity. type IIB string theory in ten dimensions has a conjectured SL(2. obtained by applied appropriate oscillators. analysis based on these solutions has been of limited use. Z ) duality in type IIB on S 1 and D-branes As discussed earlier. also an integer. denotes the number of times the elementary string is wound along S 1 . or has strong curvature at the core where the low energy approximation breaks down. (5. and w.and the right-moving sector of the world-sheet. For details. Similar analysis based on soliton solutions of low energy supergravity theory has been used to test many other duality conjectures [20.

In normal convention. k = 0 or w = 0.14) we see that mass formula for these states takes the form: 2k 2 2k 2 m2 = √ L = √ R · λ2 λ2 (5. 2 R (5. is given by: 2 2 2 2 m2 = √ (kL + 2NL ) = √ (kR + 2NR ) .(right-) moving sector of the world-sheet. (4. From equation (5. but due to space-time supersymmetry we get identical spectrum from the NS and the R sectors.15) This is the BPS mass formula for these ultra-short multiplets.16) There is no further constraint on w and k. equivalently. then the BPS mass formula takes the form: 2k 2 m2 = √ R · λ2 NL is determined in terms of kL and kR through the relation: NL = 1 2 2 (k − kL ) = wk .moving sector of the world-sheet respectively18 . (Note that if we had used the nine dimensional canonical metric as deﬁned in equations (4. . λ2 λ2 287 (5.and the right. one does not have the factors of (λ2 ) in the mass formula. as indicated earlier. It turns out that in order to be invariant under half of the space-time supersymmetry coming from the left. we consider states with NR = 0. Sen: Duality Symmetries canonical metric.) Most of these states are not BPS states as they are not invariant under any part of the supersymmetry transformation.17) (5. but here it comes due to the fact that we are using the ten dimensional canonical metric instead of the string metric to deﬁne the mass of a state. they can also be derived by analyzing the supersymmetry algebra. and will correspond to short multiplets. Although we have derived these mass formulae by directly analysing the spectrum of elementary string states. for deﬁniteness.2).A. Thus a state with NL = NR = 0 will preserve half of the total number of supersymmetries and will correspond to ultra-short multiplets.1). NL (NR ) must vanish [64]. 18 We have stated the formula in the RR sector. a state with either NL = 0 or NR = 0 will break (3/4)th of the total number of supersymmetries in the theory. If.14) where NL . there will be an additional multiplicative factor of R−2/9 in the expression for m2 . This requires kL = ±kR or. NR denote oscillator levels on the left. On the other hand.

and 16 × 16 from the left-moving sector − which is the correct degeneracy of a single short multiplet. There is also an extra degeneracy factor in the left-moving Ramond sector due to the fact that there are many oscillators that can act on the Fock vacuum of the world-sheet theory to give a state at oscillator level NL = 1.288 Unity from Duality: Gravity. for the states with NL = NR = 0. kR ). Let us ﬁrst consider the ultra-short multiplet with k = 0.and right-) sector of the world-sheet. Thus we get a state with total degeneracy 16 × 16 × 16. and eight states by acting with the −1 i transverse fermionic oscillators ψ−1 19 . there is a 16-fold degeneracy of states in each (left. NL = 1. These states have mass m2 = R2 · λ2 (5. − 16 from the right moving sector. It turns out that the total number of short multiplets d(NL ) with NR = 0 for some given value of NL ≥ 1 is given by the formula: d(NL )q NL = NL 1 16 ∞ n=1 1 + qn 1 − qn 8 · (5. The overall factor of (1/16) is due to the fact that the lowest level state is only 256-fold degenerate but a single short multiplet requires 16 × 256 states. Consider for example states with NR = 0. it has to act on the Fock vacua with odd fermion i number in order that the states obtained after acting with ψ−1 on the vacua satisfy GSO projection. For example we get eight states by acting with the transverse bosonic oscillators αi (1 ≤ i ≤ 8). Similar counting can be done for higher values of NL as well. w = 1. − 8 from the NS sector and 8 from the R sector. . showing that there is a unique ultra-short multiplet carrying given charges (kL . Due to supersymmetry. The degeneracy of short multiplets can be found in a similar manner. This gives total degeneracy factor of 8×16 in the left-moving Ramond sector. There is an 8-fold degeneracy from the Ramond sector Fock vacuum of the leftmoving sector. For example. Thus the net degeneracy of such a state is 16×16 = 256.18) The (1+q n )8 and (1−q n )8 factors in the numerator and the denominator are related respectively to the fact that in the light-cone gauge there are 8 leftmoving fermionic ﬁelds and 8 left-moving bosonic ﬁelds on the world-sheet.19) 19 Since ψ i −1 has fermion number one. Gauge Theory and Strings One can easily calculate the degeneracy of these states by analyzing the spectrum of elementary string states in detail. In this case there is a 16-fold degeneracy coming from the right-moving sector of the world-sheet. we get an identical factor from the left-moving NS sector as well.

Thus SL(2. to a state. Ref. a state carrying p B9µ r units of B9µ charge and r units of B9µ charge. B9µ and G9µ charge (momentum along S 1 ) respectively. Z) duality of type IIB string theory predicts that ∀(p.g. Now. r) are relatively prime. the w = 1 state will carry one unit of B9µ gauge ﬁeld charge.A. and is given by. Thus in the (8 + 1) dimensional theory obtained by compactifying type IIB on S 1 . under SL(2. i. d(k.21) Note that this formula is invariant under the SL(2. m2 = R2 |rλ − p|2 . Z) B9µ B9µ → p q r s B9µ B9µ .18). p. r) is independent of p and r and depends on k according to the relation: d(k. the theory must have a unique ultra-short multiplet with p units of B9µ charge and r units of B9µ charge [61]. cλ + d p r → a c b d p r . Z) matrix. The condition ps − qr = 1 implies that the pair of integers (p. p. Such states have degeneracy d(k) given in equation (5. The BPS mass formula for these states can be derived by analysing the supersymmetry algebra. r) relatively prime. as indicated earlier.e. Z) duality predicts that for (p. Sen: Duality Symmetries 289 It is well known that an elementary string acts as a source of the Bµν ﬁeld (see e. Z) transformation: λ→ where aλ + b . In this case the state is characterized by three integers p. p. [64]). Then by following the same logic as before. r) relatively prime. c d A similar prediction for the spectrum of BPS states can be made for short multiplets as well. 1 0 (5. which we shall denote by B9µ p charge carried by the state. r and k reﬂecting the B9µ .20) reﬂecting the This converts the w = 1 state. an SL(2. r) the degeneracy of such short multiplets. r) relatively prime.23) . we see that the SL(2. λ2 (5.22) a b is an SL(2. r)q k = k 1 + qn 1 16 n=1 1 − q n ∞ 8 · (5. (5. Z) transformation relates these to elementary string states with one unit of winding and k units of momentum along S 1 . Let us denote by d(k. For (p.

quantizing its zero modes. a 1-brane a string like object. and seeing if we recover the correct spectrum of BPS states. But we can postulate existence of solitonic extended objects in these theories such that in the presence of these solitons. and a 2-brane a membrane like object. there is a novel way of constructing a soliton solution that avoids this problem. recall that Bµν arises in the RR sector of string theory. 10). partially unbroken supersymmetry. when one constructs the solution carrying Bµν charge. in this theory. r A test of SL(2. one needs to ensure that the soliton deﬁned this way satisfy all the properties expected of a soliton solution in this theory e. in actual practice. Thus a state carrying B9µ charge must arise as a soliton. Gauge Theory and Strings In other words. there can be open string states whose ends lie on these extended objects (see Fig. we shall call a D-brane with Neumann boundary condition in (p + 1) directions (including time) and Dirichlet boundary condition in (9 − p) directions a Dirichlet p-brane. In order to compute the degeneracy of these solitonic states. all elementary string states are neutral under RR gauge ﬁelds as can be seen by computing a three point function involving any two elementary string states and an RR sector gauge ﬁeld. Due to this fact it is diﬃcult to proceed further along this line.290 Unity from Duality: Gravity. In type II theory. Since open strings satisfy Dirichlet boundary condition in directions transverse to these solitons.g. (Thus a 0-brane represents a particle like object. these solitons are called D-branes. Fortunately. we must understand the deﬁnition and some of the the properties of these D-branes. Of course. there should be a Hagedorn spectrum of short multiplets p with charge . To see what such a test involves. In particular we need to determine what boundary condition the modes must satisfy at the singularity.) To be more explicit. This construction uses Dirichlet (D-) branes [65. as identifying the zero modes of a singular solution is not a well deﬁned procedure. let us consider the following . Z) symmetry involves explicitly verifying the existence of these states. In particular. This is the subject to which we now turn. This can in fact be taken to be the deﬁning relation for these solitons. with the open string states with ends lying on the soliton corresponding to the (inﬁnite number of) vibrational modes of the soliton. since it can be regarded as a soliton extending along p space-like directions in which we have put Neumann boundary condition. Normally type IIA/IIB string theory contains closed string states only. it turns out to be singular. The naive approach will involve constructing such a soliton solution as a solution to the low energy supergravity equations of motion.66]. existence of static multi-soliton solutions etc. However.

(5. We call this a Dirichlet p-brane located at xm = xm 0 (p + 1 ≤ m ≤ 9). 0 ≤ µ ≤ p. Let us now summarize some of the important properties of D-branes that will be relevant for understanding the test of SL(2. 10. xp . Note that these boundary conditions break translational invariance along xm . . and extended along x1 .24) using various consistency requirements including world-sheet supersymmetry that relates the world-sheet bosons and fermions. π) = ∂σ X (σ = 0. To see how this property . Open string states with ends attached to a a) Dirichlet membrane. . b) Dirichlet string. the only possible interpretation of such a boundary condition is that there is a p dimensional extended object situated at xm = xm that is responsible for breaking this transla0 tional invariance.24) 0 for where σ denotes the spatial direction on the string world-sheet.A. π) = µ 291 xm 0 for (p + 1) ≤ m ≤ 9 . Since we want the full theory to be translationally invariant. The boundary conditions on the world-sheet fermion ﬁelds are determined from (5. Sen: Duality Symmetries boundary condition on the open string: X m (σ = 0. (a) (b) Fig. Z) duality in type IIB string theory: • The Dirichlet p-brane in IIB is invariant under half of the space-time supersymmetry transformations for odd p. .

26) are compatible only for odd p. (For p = 5 and 7 these correspond to magnetic dual potentials of Bµν and a respectively. .25) and (5. (5. originating in the leftand the right-moving sector of the world-sheet theory respectively. Z) duality 0 . This means that in type IIB on S 1 (labelled by the coordinate x9 ) a D-string wrapped around the S 1 describes a particle charged under B9µ .) This result can be obtained by computing the one point function of the vertex operator for the ﬁeld C in the presence of a D-brane. Γ0 · · · Γ9 L = L. L and R satisfy the chirality constraint: Γ0 · · · Γ9 L = L. We shall not discuss the details of this computation here. It can be shown that a Dirichlet p-brane carries one unit of charge under the RR (p + 1)-form gauge ﬁeld [65]. Gauge Theory and Strings arises. (5.26) It is easy to see that the two equations (5. As we had seen earlier.24) together with the corresponding boundary conditions on the world-sheet fermions give further restriction on L and R of the form [65]: L = Γp+1 .25) is replaced by.27) • Type IIB (IIA) string theory contains a p-form gauge ﬁeld for even (odd) p. More precisely. the rank two anti-symmetric tensor ﬁeld Bµν and the rank four anti-symmetric tensor ﬁeld Dµνρσ . let us denote by L and R the space-time supersymmetry transformation parameters in type IIB string theory. The open string boundary conditions (5. . Γ0 · · · Γ9 R =− R. if we denote by Cµ1 ···µq the q-form gauge potential. Γ9 R . Thus in type IIB string theory Dirichlet p-branes are invariant under half of the space-time supersymmetry transformations for odd p. For example. From this discussion it follows that a Dirichlet 1-brane (D-string) in type IIB theory carries one unit of charge under the RR 2-form ﬁeld Bµν .292 Unity from Duality: Gravity. An identical argument shows that in type IIA string theory we have supersymmetric Dirichlet p-branes only for even p since in this theory equation (5.25) where Γµ are the ten dimensional gamma matrices. Z) duality. (5. then a Dirichlet p-brane extending along 1 · · · p direction acts as a source of C01···p . This then is 0 a candidate soliton carrying charge quantum numbers that is related to 1 1 the state via SL(2. SL(2. in type IIB string theory these p-form gauge ﬁelds correspond to the scalar a. Γ0 · · · Γ9 R = R . The relevant string world-sheet diagram has been indicated in Figure 11.

In particular.A. • A U (1) gauge ﬁeld. It can be shown that the dynamics of these collective coordinates is described by a (1+1) dimensional supersymmetric quantum ﬁeld theory which . for RR state String world-sheet Fig. • Verify if we get an ultra-short multiplet in this quantum theory. The vertex operator of the RR-ﬁeld is inserted at a point on the hemisphere. Since the D-string is a one dimensional object. all the vibrational modes of the D-string are given by the open string states with ends attached to the D-string. The string world-sheet diagram relevant for computing the coupling of the RR gauge ﬁeld to the D-brane. 11. the zero frequency modes (collective modes) of the D-string that are relevant for analyzing the spectrum of BPS states correspond to massless open string states propagating on the D-string. By analyzing the spectrum of these open string states one ﬁnds that the collective coordinates in this case correspond to • 8 bosonic ﬁelds y m denoting the location of this string in eight transverse directions. the dynamics of its collective coordinates should be described by a (1 + 1) dimensional ﬁeld theory. Thus our task now is as follows: 1 • Quantize the collective coordinates of this soliton. Sen: Duality Symmetries D-brane 293 X Vertex op. predicts that there should be a unique ultra-short multiplet carrying charge 0 quantum numbers . It corresponds to a surface of the topology of a hemisphere with its boundary glued to the D-brane. As we had discussed earlier. • 8 Majorana fermions.

we can get states carrying charge k quantum numbers as well.) It can be shown that [62] this momentum. 1 Due to the compactness of the space direction. but by taking py = 2πk/a. Normally in (1 + 1) dimension gauge ﬁelds have no dynamics. • 1 compact bosonic coordinate y. Thus the ﬁrst question we need to ask is: what is the (1 + 1) dimensional quantum ﬁeld theory governing the dynamics of this system? In order to answer this question we need to study the dynamics of r D-strings. we can actually regard this as a quantum mechanical system instead of a (1 + 1) dimensional quantum ﬁeld theory. Thus quantization of the fermionic coordinates gives 28 = 256 -fold degeneracy. This system can be described as easily as a single D-string: instead of allowing open strings to end on a single D-string. Furthermore. It turns out that in looking for ultra-short multiplets. which is precisely the correct degeneracy for a ultra-short multiplet.294 Unity from Duality: Gravity. which represents electric ﬂux along the D-string. This establishes p the existence of all the required states of charge predicted by SL(2. The degrees of freedom of this quantum mechanical system are: • 8 bosonic coordinates y m . y ≡ A1 dl is a physical variable. Thus the momentum py conjugate to y is quantized (py = 2πk/a with k integer. A quantum state is labelled by the momenta conjugate to y m (ordinary momenta) and an integer labelling momentum conjugate to y which can be identiﬁed with the quantum number p labelling B9µ charge. p What about states with r > 1? These carry r units of B9µ charge r and hence must arise as a bound state of r D-strings wrapped along S 1 . is actually a source of B9µ charge! Thus if we restrict to the 0 py = 0 sector then these states carry charge quantum numbers as dis1 cussed earlier. we allow it to end on any of the r . But here since the space direction is compact. Gauge Theory and Strings is the dimensional reduction of the N = 1 supersymmetric U (1) gauge theory from (9 + 1) to (1 + 1) dimensions. • 16 fermionic coordinates. The fermionic coordinates satisfy the sixteen dimensional Cliﬀord algebra. Z) 1 symmetry. we can ignore all modes carrying momentum along S 1 . This corresponds to dimensionally reducing the theory to (0 + 1) dimensions. the compactness of U (1) makes y to be periodically identiﬁed ((y ≡ y + a) for some a).

We shall not go through the details of this analysis. 1 ≤ i ≤ r. It turns out that for r coincident D-strings the dynamics of massless strings on the D-string world-sheet is given by the dimensional reduction to (1 + 1) dimension of N = 1 supersymmetric U (r) gauge theory in ten dimensions. First of all. or equivalently. N = 4 supersymmetric U (r) gauge theory in four dimensions [62]. Sen: Duality Symmetries 295 Fig. Z) [62]! .28) where x(i) denotes the location of the i-th D-string. Possible open string states in the presence of three parallel D-strings. D-strings situated at xm = xm . one can show that the probp lem of computing the degeneracy of states reduces to the computation r of certain Witten index in this quantum theory.A. for each of the r D-strings we get a U (1) gauge ﬁeld. Thus the dynamics of this system will now be described not only by the open strings starting and ending on the same D-string. (5. but also by open strings whose two ends lie on two diﬀerent D-strings. Following a logic similar to that in the case of a single D-string. eight scalar ﬁelds and eight Majorana fermions from open strings with both ends lying on that D-string. For studying the spectrum of BPS states we need to focus our attention on the massless open string states. r) which are relatively prime. but just state the ﬁnal result. It turns out that there is a unique ultra-short multiplet for every pair of integers (p. But we can get extra massless states from open strings whose two ends lie on two diﬀerent D-strings when these two D-strings coincide. (i) 2 ≤ m ≤ 9. precisely as predicted by SL(2. 12. The situation is illustrated in Figure 12.

5. p. This provides us with a test of the conjectured SL(2. The BPS states come from conﬁgurations where only the left. we can no longer dimensionally reduce the (1 + 1) dimensional theory to (0 + 1) dimensions. SU (2) at the self-dual radius etc. the duality between heterotic on T 4 and type IIA on K3 [71]. The calculation of the degeneracy d(k. except that the elementary string is replaced here by the solitonic D-string. k) is done by determining in how many ways the total momentum k can be divided among the various left-moving bosonic and fermionic modes. Thus. The method of using D-branes to derive the dynamics of collective coordinates has been used to verify the predictions of other duality conjectures involving various string compactiﬁcations. Instead we need to take into account the modes of the various ﬁelds of the (1 + 1) dimensional ﬁeld theory carrying momentum along the internal S 1 . Naturally. 63]. In order to get these states from the D-brane spectrum. we do not have the W ± bosons that are required for enhancing a U (1) gauge group to SU (2). for example.296 Unity from Duality: Gravity. However upon closer examination one realises that this cannot really be a problem [116]. r. At this point the would be massless gauge bosons of the non-abelian gauge theory acquire mass by Higgs mechanism. we get back the Hagedorn spectrum for d(k.g. . Gauge Theory and Strings A similar analysis can be carried out for the short multiplets that carry momentum k along S 1 besides carrying the B and B charges p and r [62. At ﬁrst sight this seems to lead to a contradiction. and appear as BPS states in the abelian theory. Thus the answer agrees exactly with that predicted by SL(2.3 Massless solitons and tensionless strings An interesting aspect of the conjectured duality between the heterotic string theory on T 4 and type IIA string theory on K3 is that at special points in the moduli space the heterotic string theory has enhanced non-abelian gauge symmetry e.(or right-) moving modes on S 1 are excited. E8 × E8 or SO(32) in the absence of vacuum expectation value of the internal components of the gauge ﬁelds. etc. r) as well. since the spectrum of elementary string states does not contain any state charged under the U (1) gauge ﬁelds arising in the RR sector. the duality between type I and SO(32) heterotic string theory [22]. Among them are self-duality of type II string theory on T 4 [67–70]. r) of BPS states carrying given charge quantum numbers (p. To see this let us consider a point in the moduli space of heterotic string theory on T 4 where the non-abelian gauge symmetry is broken. Z) duality. p. Perturbative type IIA string theory on K3 does not have any such gauge symmetry enhancement. Z) symmetry of type IIB on S 1 . This counting problem turns out to be identical to the one used to get the Hagedorn spectrum of BPS states in the elementary string spectrum.

after all. This has been possible through the work of [19. we can express this as Ω(A)T Ω(A) for some O(4. These massless D-brane solitons will then provide the states necessary for enhancing the gauge symmetry. one would like to understand this mechanism directly by analysing the type IIA string theory. First of all. It can be shown that in the variables deﬁned in Section 4. 20) matrix. Since the masses of BPS states are determined by the BPS formula.29) as m2 = e−Φ (A) /2 T α LΩ(A)T (I24 + L)Ω(A) Lα .28). If for some Ω(A) it is aligned along one of the eigenvectors of (I24 + L) with zero eigenvalue. we do not encounter massless solitons very often in physics. the vector Ω(A) Lα rotates in the twenty four dimensional space. and rewrite equation (5. and represents the U (1) charges carried by this particular state. 20) matrix Ω(A) . one ﬁnds that at a generic point in the moduli space where SU (2) is broken. Since the two tangential directions on the D-2 brane are directed along the two internal directions of K3 tangential to the 2-cycle. the masses of these states vanish. 117]. For each α we can assign an occupation number n(α) which gives the number of BPS multiplets carrying this speciﬁc set of charges.3 the BPS formula is given by. this object has no extension in any of the ﬁve non-compact spatial directions. (5. Since M (A) is a symmetric O(4. (5. 71.A. (I24 + L) has 20 zero eigenvalues. Although this argument resolves the problem at an abstract level. we shall get massless solitons provided the occupation number n(α) for this speciﬁc α is non-zero. Sen: Duality Symmetries 297 As we approach the point of enhanced gauge symmetry.29) where α is a 24 dimensional vector belonging to the lattice Λ24 .30) As can be seen from equation (4. To see this more explicitly. and hence . As we vary M (A) and hence Ω(A) . Thus if we are able to ﬁnd the images of these BPS states on the type IIA side as appropriate D-brane states. For simplicity let us focus on the case of enhanced SU (2) gauge symmetry. then the masses of these Dbrane states must also vanish as we approach the point in the moduli space where the heterotic theory has enhanced gauge symmetry. m2 = e−Φ (A) /2 T α (LM (A) L + L)α . the + and the − sign of the charge being obtained from two diﬀerent orientations of the D-2 brane. the images of the W ± bosons in the type IIA theory are given by a D-2 brane wrapped around a certain 2-cycle (topologically non-trivial two dimensional surface) inside K3. let us examine the BPS formula. the vanishing of the masses must be a consequence of the BPS formula. since.

It turns out that instead of acquiring enhanced gauge symmetry. and wrapping it on a two cycle of K3. In other words. This gives us the massless solitons that are required for the gauge symmetry enhancement. we get tensionless strings. type IIB string theory acquires tensionless strings at these special points in the K3 moduli space [119]. At this singularity the area of the topologicaly non-trivial 2-cycle mentioned above goes to zero. 20 These states were analyzed in detail in [118]. the tension of the string goes to zero. Thus from the point of view of the (5 + 1) dimensional theory such a conﬁguration will appear as a string. Thus as we approach the singular point on the K3 moduli space where the area of the two cycle vanishes. Does this imply that type IIB on K3 also develops enhanced gauge symmetry at these special points in the K3 moduli space? This does not seem possible. These are precisely the massless gauge bosons required for the gauge symmetry enhancement in type IIB on K3 × S 1 . In fact it turns out that there is a one to one correspondence between the enhanced gauge groups. A similar mechanism works for getting other gauge groups as well.298 Unity from Duality: Gravity. the K3 on which type IIA theory is compactiﬁed becomes singular. This establishes an explicit physical relationship between A-D-E singularities and A-D-E lie algebras. the mass of the wrapped D-2 brane. But type IIA on K3 × S 1 is T-dual to type IIB on K3 × S 1 . and the third direction of the three brane is along one of the non-compact spatial directions. The tension of this string is given by the product of the tension (energy per unit three volume) of the three brane and the area of the two cycle on which the three brane is wrapped. Gauge Theory and Strings behaves like a particle20 . since type IIB string theory does not have any D-2 brane solitons which can be wrapped around the collapsed two cycles of K3. obtained by multiplying the tension of the D-2 brane by the area of the two cycle. These arise from taking a D-3 brane of type IIB string theory. which are also classiﬁed by the A-D-E dynkin diagram [19]. The appearance of enhanced gauge symmetry in type IIA on K3 poses another puzzle. . and the singularity type of K3. It turns out that as we approach the point in the moduli space where the theory on the heterotic side develops enhanced SU (2) gauge symmetry. this theory also has enhanced gauge symmetry when the K3 becomes singular. thus type IIB on K3 × S 1 must also develop enhanced gauge symmetry when K3 develops singularities. Since such a compactiﬁcation does not destroy gauge symmetry. vanishes. As a result. which are classiﬁed by A-D-E dynkin diagram. Thus two of the tangential directions of the three brane are directed along the internal directions of K3. Let us compactify this theory on one more circle. Upon further compactiﬁcation on a circle we get massless particles from conﬁgurations where this tensionless string is wound around the circle.

Z) duality symmetry of the ten dimensional type IIB string theory. but can be “derived” from each other. and in fact generate the full duality symmetry group of type IIB on T n as given in the table of Section 4. Typically part of this self duality group is T -duality.2 Duality of dualities Suppose two theories are conjectured to be dual to each other. Z). Quite often G and H do not commute and together generate a much bigger group [23. We shall discuss them one by one. We shall now see that many of these conjectures are not independent.76]. 78]. In that case. 6.A. Thus we see that the existence of the full duality symmetry group of type IIB on T n can be infered from the SL(2. We shall illustrate this with a speciﬁc example [23]. 6. whose existence can be veriﬁed order by order in string perturbation theory. and each theory in turn has a conjectured self-duality group. the existence of this bigger group of symmetries can be regarded as a consequence of the duality symmetry of A on KA and T -dualities.5. The material covered in this section is taken mainly from [75.1 Combining non-perturbative and T -dualities Suppose a string theory A compactiﬁed on a manifold KA has a conjectured duality symmetry group G.5 we see that type IIB on T n also has a T -duality group SO(n. Then the theory A on KA × M is expected to have the following set of duality symmetries: • It inherits the original duality symmetry group G of A on KA . n. and the perturbatively veriﬁable T -duality symmetries of type IIB on T n . From the table given in Section 4. Sen: Duality Symmetries 6 Interrelation between diﬀerent duality conjectures 299 In the last three sections we have seen many diﬀerent duality conjectures and have learned how to test these conjectures. • It also has a perturbatively veriﬁable T -duality group. Let us call it H. Now further compactify this theory on some manifold M. There are several diﬀerent ways in which dualities can be related to each other. We have seen that in ten dimensions type IIB string theory has a conjectured duality group SL(2. It turns out that typically these two duality groups do not commute. 77. and the rest involves non-trivial transformation of the coupling constant. Z) that acts non-trivially on the coupling constant. But quite often the non-perturbative duality transformations in one theory correspond to T -duality in the dual theory .

Let us start with the conjectured duality between heterotic on T 4 and type IIA on K3.Z) X SL(2.22. and vice versa. we can establish the existence of the self-duality group O(6. the O(6. The embedding of the T -duality groups in the full duality group in heterotic on T 6 and type IIA on K3 × T 2 . Z)×SL(2. Now. Again we shall illustrate this with an example [19.Z) X SL(2. and hence can be veriﬁed order by order in perturbation theory in this theory. 20. Z)×SL(2.22. On the other hand . 24]. On the other hand.Z) TO(6. As a result. Now the question we would like to address is. Now let us compactify both theories further on a two dimensional torus T 2 . the full self duality group in both theories follows from the conjectured duality between the two theories.Z) Fig.22. and the duality between the heterotic on T 4 and type IIA on K3. Z) duality group? This has been illustrated in Figure 13.Z) O(4. Z)×SL(2. Z) that can be veriﬁed using type II perturbation theory. This produces a dual pair of theories: type IIA on K3 × T 2 and heterotic on T 6 . type IIA on K3×T 2 has a T -duality group O(4.Z) factor of the duality group appears as a T -duality symmetry of the heterotic string theory. Z) that can be veriﬁed using heterotic perturbation theory. how are the T -duality symmetry groups in the two theories embedded in the full conjectured O(6. Z) factor of the full duality group is a subgroup of the T -duality group in type IIA on K3 × T 2 . . Z) in heterotic on T 6 and type IIA on K3×T 2. Z) × SL(2. and hence can be veriﬁed in this theory order by order in perturbation theory.Z) X SL(2. Gauge Theory and Strings DUALITY GROUP FULL HETEROTIC ON T6 O(6.20. 13.22. In particular we ﬁnd that the SL(2. The full conjectured duality group in both theories is O(6. Thus assuming that T -duality in either theory is a valid symmetry. 22. Z).Z) X SL(2. 22. 22.300 Unity from Duality: Gravity. 22. heterotic on T 6 has a T -duality group O(6.Z) TYPE II ON K3 X T2 O(6. Z) × SL(2.

IIA on K3 ↔ heterotic on T 4 . The moduli of KA . the independent ones are: 1. Now construct a pair of new manifolds MA . Suppose (Theory A on KA ) has been conjectured to be dual to (Theory B on KB ). MB by starting from some other manifold B. Thus we would expect the theories A on MA and B on MB to be equivalent. SL(2. This then gives rise to a new duality conjecture.3 Fiberwise duality transformation A on KA B on KB B A on MA B B on M B Fig. and 3. we see that so far among all the conjectured non-perturbative duality symmetries. KB are two diﬀerent manifolds (in general). In each local neighbourhood of the base manifold B. . Z) of type IIB in D = 10. We shall now show how to “derive” 3) from 1) and 2). 6. Application of ﬁberwise duality transformation. and erecting at every point on B a copy of KA . and KA . KB . type I ↔ SO(32) heterotic in D = 10. Then we would expect a duality Theory A on MA ↔ Theory B on MB by applying the duality transformation ﬁberwise. In this subsection we shall describe the idea of constructing dual pairs of theories using ﬁberwise duality transformation [77]. This situation has been illustrated in Figure 14. 14. the two theories are equivalent due to the equivalence of the theories living on the ﬁber (× any manifold). Sen: Duality Symmetries 301 Using the results of this and the previous subsection.A. This duality involves a precise map between the moduli spaces of the two theories. Here A and B are two of the ﬁve diﬀerent string theories in D = 10. KB vary slowly over B and are related to each other via the duality map that relates (A on KA ) to (B on KB ). 2.

Q ∈ D) to (hA (p). (We shall see some explicit examples of this later. and compare the two quotient theories (A on KA × D/hA · g) and (B on KB × D/hB · g) (KA ×D/hA ·g) is obtained from the product manifold KA ×D by identifying points that are related by the Z2 transformation hA · g. (KA × D/hA · g) admits a ﬁbration with base D/g and ﬁber KA . then there is no identiﬁcation of the points in the copy of KA that is sitting at Q. Thus we can now apply ﬁberwise duality transformation to derive a new . This situation is illustrated in Figure 15. the ﬁber gets twisted by the transformation hA . Gauge Theory and Strings KA Q KA p g(Q) Z2 hA (p) KA Q=g(Q) D D/g Fig. B on (KB × D)/(hB · g) has an identical structure. we can derive the existence of many new duality symmetries from a given duality symmetry. This shows that the ﬁber is KA and not KA /hA . assuming that this is the case. As shown in this ﬁgure. Although there is no rigorous argument as to why this should be so. As we go from Q to g(Q). The Z2 transformation relates the point (Q. which is a closed cycle on D/g. hA (p)). Now suppose that at some isolated points (or subspaces of codimension ≥1) on B the ﬁbers KA and KB degenerate.302 Unity from Duality: Gravity. this appears to be the case in all known examples. Further suppose that (A on KA ) has a Z2 symmetry generated by hA . Representation of a Z2 orbifold as a ﬁbered space. g(Q)). note that since hA · g takes a point (p ∈ KA . p) on D × KA to the point (g(Q).) Is the duality between (A on MA ) and (B on MB ) still valid? We might expect that even in this case the duality between the two theories holds since the singularities occur on subspaces of “measure zero”. Now compactify both theories on another manifold D with a Z2 isometry generated by g. hA and hB are mapped to each other under duality. 15. Conversely. A special case of this construction involves Z2 orbifolds. In particular. if we focus our attention on a deﬁnite point Q on D. Suppose we have a dual pair (A on KA ) ↔ (B on KB ). The second theory. Then the dual theory must also have a Z2 symmetry generated by hB .

However.A. P0 ). Let us make R → (1/R) duality transformation on one of the circles of T 4 in the theory on the left hand side. This converts type IIB . From this one can explicitly compute the action of S(−1)FL S −1 on these massless ﬁelds. Let S denote the SL(2. The action of S. We take (A on KA ) to be type IIB in D = 10. (6. This gives: S(−1)FL S −1 = Ω . (6. P0 ). as we have discuused earlier. P0 ) and (hB (p ). we still expect the duality to hold since these are points of measure zero on T 4 . but this does not prevent us from applying our method of constructing dual pairs. Z) element 0 1 −1 0 . hB to be Ω. Also there are sixteen ﬁxed points on T 4 under I4 where the application of ﬁberwise duality transformation breaks down. Z) symmetry. but as has been argued before. This action turns out to be identical to that of Ω. P0 ) and (hA (p). A similar result holds for their action on the massless fermionic ﬁelds as well. Similarly in KB × D/hB · g there is an identiﬁcation of points (p . At these points the argument in support of duality between the two theories breaks down. hA to be (−1)FL . This gives the duality: (IIB on T 4 /(−1)FL · I4 ) ↔ (IIB on T 4 /Ω · I4 ). D to be T 4 . We start with type IIB string theory in ten dimensions. Finally. We shall now bring this duality into a more familiar form via T -duality transformation.e. since the action of S on the massive ﬁelds is not known. Sen: Duality Symmetries duality: (A on KA × D/hA · g) ↔ (B on KB × D/hB · g) 303 Note that if P0 ∈ D is a ﬁxed point of g (i. one can deﬁne this action in such a way that the actions of S(−1)FL S −1 and Ω are identical on all states.2) We are now ready to apply our formalism. if g(P0 ) = P0 ) then in KA × D/hA ·g there is an identiﬁcation of points (p. This has a conjectured SL(2. (−1)FL and Ω on the massless bosonic ﬁelds in this theory were described in Section 4. we would expect that the two quotient theories are still dual to each other [78]. (B on KB ) to be type IIB in D = 10 transformed by S.1) Recall that this theory also has two global discrete symmetries (−1)FL and Ω. Thus at P0 the ﬁbers degenerate to KA /hA and KB /hB respectively. We shall now illustrate this construction in the context of a speciﬁc example.4. since these are points of “measure zero” on D. and g to be the transformation I4 that changes the sign of all the coordinates on T 4 . Note that in this case the ﬁbers KA and KB are points.

of the original string theory compactiﬁed on M1 . Suppose a string theory compactiﬁed on a manifold M1 × M2 has a self-duality symmetry group G. acting on this conﬁguration. Thus the theory on the left hand side is T -dual to type IIA on T 4 /I4 . Since type I string theory can be regarded as type IIB string theory modded out by Ω. i. The same argument can be extended to the case of a pair of dual theories. Let us now take the theory on the right hand side and make R → (1/R) duality transformation on all four circles. There are many other applications of ﬁberwise duality transformation. we see that the theory on the right hand side is type I on T 4 . Some of them will be discussed later in this review. the argument can be generalized to establish this duality at a generic point in the moduli space as well [78]. which can be checked by explicitly studying the action of these transformations on the various massless ﬁelds. A generic element of G. This takes type IIB theory to type IIB theory. But this transforms Ω · I4 to Ω. However. acting on a conﬁguration where M2 is big. Z) duality symmetry of type IIB string theory from the conjectured duality between type I and SO(32) heterotic string theories. But we can also proceed in the reverse direction. Now consider the limit when the size of M2 goes to inﬁnity. Thus the theory on the right is T -dual to type IIB on T 4 /Ω. But we shall now show how this procedure can be used to derive the SL(2. will convert this conﬁguration to one where M2 has small or ﬁnite size. A priori this procedure does not appear to be very useful.4 Recovering higher dimensional dualities from lower dimensional ones So far we have discussed methods of deriving dualities involving compactiﬁed string theories by starting with the duality symmetries of string theories in higher dimensions. But by (heterotic – type I) duality in ten dimensions this is dual to heterotic on T 4 . Thus we would expect that H is the duality symmetry group of the theory in the decompactiﬁcation limit. gives us back a conﬁguration where M2 is big. This also transforms (−1)FL · I4 to I4 . Thus we have “derived” the duality (Type IIA on K3) ↔ (Heterotic on T 4 ) from other conjectured dualities in D = 10. This of course is just a special case of type IIA on K3. Gauge Theory and Strings theory to type IIA. which can again be seen by studying the action of these transformations on the massless ﬁelds. and .e.e. 6.304 Unity from Duality: Gravity. i. there may be a subgroup H of G that commutes with this limit. since one normally likes to derive more complicated duality transformations of lower dimensional theories from the simpler ones in the higher dimensional theory. any element of this subgroup. Although this way the duality has been established only at a particular point in the moduli space (the orbifold limit of K3).

We shall describe the main steps in this argument. Z) × SL(2. 7 Duality in theories with less than sixteen supersymmetry generators So far our discussion has been focussed on theories with 16 or more supersymmetry charges. Let us now make an R → (1/R) duality transformation on both the circles of this T 2 . Thus we conclude that SL(2. Z) × SL(2. see [75]. Z). One of these two SL(2. we conclude that SL(2. Z) groups becomes part of the diﬀeomorphism group of type IIB string theory and does not correspond to anything new. (This can be seen by studying the action of various transformations on the massless ﬁelds. Z) × SL(2. for these theories the non-renormalization theorems for the low energy eﬀective action and . Since type I string theory can be regarded as type IIB string theory modded out by the world-sheet parity transformation Ω discussed in Section 4. Thus we see that all the dualities discussed so far can be “derived” from a single duality conjecture. this model describes type IIB string theory compactiﬁed on the surface of a tetrahedron (which is geometrically T 2 /I2 ). this must also be a symmetry of type I on T 2 . Sen: Duality Symmetries 305 T -duality symmetries of the heterotic string theory. In the next section we shall see more examples of dualities which can be derived from the ones that we have already discussed. By the “duality of dualities” argument. Z) subgroup of this T -duality group.4. and the other one is associated with the R → (1/R) duality symmetries on the two circles. As was pointed out in Section 4. where I2 denotes the reversal of orientation of both the circles of T 2 . Since this limit gives us back the decompactiﬁed type IIB string theory. for details.) Geometrically.Z). Z) is a subgroup of the self-duality group of type IIB on T 2 /Ω. − the one between type I and SO(32) heterotic string theories in ten dimensions. Z) is a subgroup of the self-duality group of type IIB on a tetrahedron.18. with an added twist of (−1)FL · Ω as we go around any of the four vertices of the tetrahedron (the ﬁxed points of I2 ). Z) factors is associated with the global diﬀeomorphism of T 2 .A. Now take the limit where the size of the tetrahedron goes to inﬁnity.4. This converts type IIB on T 2 to type IIB theory compactiﬁed on a dual T 2 . Now heterotic string theory on T 2 has a T -duality group O(2. Z) factors commute with this limit. One of these SL(2. It turns out that both the SL(2. and Ω to (−1)FL · Ω · I2 . We shall focus our attention on an SL(2. We start with the duality between type I on T 2 and heterotic on T 2 that follows from the duality between these theories in ten dimensions. Z) factor represents the S-duality transformation discussed in Section 4. but the other SL(2. we conclude that type IIB string theory in ten dimensions has a self-duality group SL(2.

. It is also possible to construct more general class of four dimensional heterotic string theories with the same number of supersymmetries where the background does not have the product structure K3 × T 2 [79. Thus we are left with 8 supersymmetry generators in D = 4. However we shall begin by describing the systematic approach. Type IIA/IIB on Calabi-Yau 3-folds: in our convention an n-fold describes an n complex or 2n real dimensional manifold. giving rise to N = 2 supersymmetry. 79. This makes it easy to test duality conjectures involving these theories. In this section we shall extend our discussion to theories with eight supercharges. again giving N = 2 supersymmetry in four dimensions. and then describe how one tests these dualities. and although the non-renormalization theorems are less powerful here. 2. Heterotic string theory on K3 × T 2: in ten dimensions heterotic string theory has sixteen supersymmetry generators. Compactiﬁcation on a Calabi-Yau 3-fold breaks 3/4 of the supersymmetry. Two of them are: 1. they are still powerful enough to provide us with some of the most striking tests of duality conjectures involving these theories. The steps involved in this construction are as follows: • Start from the conjectured duality (Type IIA on K3) ↔ (Heterotic on T 4 ).306 Unity from Duality: Gravity.1 Construction of a dual pair of theories with eight supercharges For deﬁniteness we shall focus our attention on N = 2 supersymmetric theories in four dimensions. Compactiﬁcation on K3×T 2 breaks half of the supersymmetry. [77. 108]. 80]. We shall see that these theories have a very rich structure. Examples of such theories are provided by N = 2 supersymmetric theories in four dimensions. Thus we have a theory with eight supersymmetry generators. 7. The question we would like to ask is: is it possible to construct pairs of N = 2 supersymmetric type II and heterotic string compactiﬁcations in four dimensions which will be non-perturbatively dual to each other? Historically such dual pairs were ﬁrst constructed by trial and error [79] and then a more systematic approach was developed [77. 80–83]. There are several ways to get theories with N = 2 supersymmetry in four dimensions. The material covered in this section is based mainly on refs. The systematic construction of such dual pairs can be carried out by application of ﬁberwise duality transformation as described in the last section. Gauge Theory and Strings the spectrum of BPS states are particularly powerful. In ten dimensions type II theories have 32 supersymmetry generators.

This construction has been illustrated in Figure 16.A. This gives heterotic string theory on a manifold obtained by varying T 4 on CP 1 according to the duality map. Typically this manifold turns out to be K3 × T 2 or some variant of this. . Thus for a speciﬁc Calabi-Yau. • For type IIA on each such Calabi-Yau 3-fold we can get a dual heterotic compactiﬁcation by replacing the type IIA on K3 by heterotic on T 4 on each ﬁber according to the duality map. Construction of dual pair of N = 2 supersymmetric string theories in four dimensions from the dual pair of theories in six dimensions. Sen: Duality Symmetries 6D Duality map K3 T 4 307 1 1 CP IIA on CY CP Heterotic on K3 X T 2 Fig. 16. • Choose a CP 1 base. Diﬀerent Calabi-Yau manifolds will give rise to diﬀerent gauge ﬁelds on K3 × T 2 . On the heterotic side the moduli involve background gauge ﬁelds on T 4 besides the shape and size of T 4 . • Construct a Calabi-Yau 3-fold by ﬁbering K3 over the base CP 1 . Note that the original duality map gives a precise relationship between the moduli of type IIA on K3 and heterotic on T 4 . One can construct a whole class of Calabi-Yau manifolds this way by choosing diﬀerent ways of varying K3 over CP 1 . This gives the gauge ﬁeld conﬁguration on K3 × T 2 . knowing how K3 varies over CP 1 . Thus we get a duality map (Type IIA on CY) ↔ (Heterotic on K3×T 2). we can ﬁnd out how on the heterotic side the background gauge ﬁelds on T 4 vary as we move along CP 1 . This model is expected to be dual to the type IIA string theory on the Calabi-Yau manifold that we started with.

22 of the gauge ﬁelds come from decomposing the three form ﬁeld along the harmonic two forms on K3. As discussed there. We shall focus our attention on the gauge ﬁelds. By our previous argument relating orbifolds to ﬁbered spaces. The low energy eﬀective action of both the theories and the origin of the various massless ﬁelds in these theories were discussed in Section 4. has known action on the harmonic forms ω (p) . hA corresponds to • exchanging ten of the ω (p) with ten others and • changing the sign of two more ω (p) . hA is a speciﬁc involution of K3 known as the Enriques involution. In order to determine the theory on the heterotic side. K3×T 2/(hA ·g) can be shown to describe a Calabi-Yau manifold. these two theories are expected to be dual to each other via ﬁberwise duality transformation.308 Unity from Duality: Gravity. we need to determine hB .36). Gauge Theory and Strings We shall illustrate this procedure with the example of a pair of Z2 orbifolds of the form [80]: (IIA on K3 × T 2 /hA · g) ↔ (Heterotic on T 4 × T 2 /hB · g) where g acts on T 2 by changing the sign of both its coordinates. in the type IIA on K3. It turns out that the action on the heterotic side is given by: • exchanging the gauge ﬁelds in the two E8 factors. being a geometric transformation on K3. • exchanging (G9µ . For this particular example. coming from the ten dimensional gauge ﬁeld Aµ and the dual of Cµνρ invariant. hA . hA leaves the other two gauge ﬁelds. We can now translate this into an action on the gauge ﬁelds in heterotic on T 4 . We shall now describe this procedure in some detail. B8µ ). Furthermore. • changing the sign of (G7µ and B7µ ). This translates into a similar action on the ﬁelds Aµ deﬁned in equation (4. In order to determine hB . B9µ ) with (G8µ . and hB is the image of this transformation on the heterotic side. we need to study the relationship between the ﬁelds appearing in type IIA on K3 and heterotic on T 4 . Thus the theory on the left-hand side corresponds to type IIA string theory compactiﬁed on this Calabi-Yau manifold.3. Now. and. (p) .

one cannot directly compare the tree level low energy eﬀective action in the two theories. Hence the spectrum computed at weak coupling cannot always be trusted at strong coupling. . Nevertheless there are some non-renormalization theorems which allow us to test these proposed dualities. Using the idea of ﬁberwise duality transformation we can construct many more examples of heterotic – type IIA dual pairs in four dimensions with N = 2 supersymmetry.A. there is no rigorous proof that ﬁberwise duality transformation always produces a correct dual pair of theories.) They are • vector multiplet containing one vector. as we shall now describe. (For a review. one complex scalar. Unlike in the case of theories with sixteen supercharges. Thus this construction gives a type II – heterotic dual pair with N = 2 supersymmetry. • x8 ↔ x9 . It turns out that modding out heterotic string theory on T 6 by the transformation hB · g produces an N = 2 supersymmetric theory. Matter multiplets in N = 2 supersymmetric theories in four dimensions are of two types. Quite often using mirror symmetry [109] we can also relate this to IIB string theory on a mirror Calabi-Yau manifold. This transformation is not visible in perturbative type IIA theory. • x7 → −x7 . as they undergo quantum corrections in general.2 Test of duality conjectures involving theories with eight supercharges Given such a dual pair of theories constructed by application of ﬁberwise duality transformation. and two Majorana fermions. particularly when the ﬁber degenerates at some points/regions in the base. see [88]. in this theory the spectrum of BPS saturated states can change discontinuously as we move in the moduli space [88]. as we have seen. the next question will be: how do we test if these theories are really dual to each other? After all. Sen: Duality Symmetries 309 This translates into the following geometric action in heterotic string theory on T 4 :21 • exchange of two E8 factors in the gauge group. and 21 Here we are regarding this theory as the E × E heterotic string theory compactiﬁed 8 8 on T 4 . this is equivalent to SO(32) heterotic string theory compactiﬁed on T 4 . By the duality between the two heterotic string theories upon compactiﬁcation on a circle. 22 We need to add to this a non-geometrical shift involving half of a lattice vector in Λ24 in order to get a modular invariant theory on the heterotic side. This is hB 22 . Furthermore. 7.

The N = 2 supersymmetry requires that there is no coupling between the vector and the hypermultiplets in those terms in the low energy eﬀective action Seﬀ which contain at most two space-time derivatives [84]. The kinetic terms of the vectors and the fermionic ﬁelds are related to these scalar kinetic terms by the requirement of N = 2 supersymmetry.310 Unity from Duality: Gravity. Thus the hypermultiplet 23 There are however strong restrictions on what kind of metric G V and GH should describe. The dilaton belongs to a vector multiplet. is exact. m¯ α¯ (7. Gauge Theory and Strings • hypermultiplet containing two complex scalars and two Majorana fermions. In type IIA/IIB string theory on Calabi-Yau manifold the dilaton belongs to a hypermultiplet. and ψ denote the complex scalars in the hypermultiplet.1) where GV and GH are appropriate metrics in the vector and the hypermultiplet moduli spaces. 110]. On the other hand in heterotic on K3 × T 2 . Thus quantum corrections to a given term must involve a coupling to the dilaton. 2. Thus in these theories the vector multiplet kinetic term. whereas GH must describe a quaternionic geometry [111]. 1. 79]. the dilaton is in the vector multiplet. In this case the same argument shows that there can be no correction to the hypermultiplet kinetic term. Now consider the following two special cases. Thus the scalar kinetic terms appearing in the Lagrangian density associated with Seﬀ must be of the form: ¯ ¯ GV n (φ)∂µ φm ∂ µ φn + GHβ (ψ)∂µ ψ α ∂ µ ψ β . Let φ denote the complex scalars in the vector multiplet. calculated at the tree level. But in string theory we have some extra ingredient [77.multiplet moduli spaces by itself is not of much help. The dilaton belongs to a hypermultiplet. these restrictions do not ﬁx GV and GH completely. Let us consider a theory at a generic point in the moduli space where the massless matter ﬁelds include only abelian gauge ﬁelds and neutral hypermultiplets. . Recall that the coupling constant in string theory involves the dilaton.and the vector. In particular GV must describe a special Kahler geometry [84. since each term may be independently modiﬁed by quantum corrections23 . However. This decoupling between the hyper. Then there can be no correction to the vector multiplet kinetic term since such corrections will give a coupling between the dilaton and the vector multiplet.

2. since there is no non-perturbative formulation of string theory. Given that the tree and one loop results in the heterotic string theory agree with the expansion of the exact result on the type II side. the agreement between the two answers is a consequence of highly non-trivial mathematical identities. From the exact answer calculated from the type II side we know what this contribution should be. . Indeed. we can eﬀectively get an N = 4 supersymmetric Yang-Mills theory. 3. However. Sen: Duality Symmetries 311 kinetic term. since the one loop calculation is highly non-trivial on the heterotic side. is exact. But we cannot calculate it directly on the heterotic side. So we can now expand the exact answer in powers of eΦH and compare this answer with the explicit calculations in heterotic string perturbation theory. Take a type II – heterotic dual pair and calculate the vector multiplet kinetic term exactly from the tree level analysis on the type II side. Using this information we can adopt the following strategy for testing duality24 . Using the map between the ﬁelds in the type II and the heterotic theory.) Thus one can compare the expected tree and one loop terms. one loop. and involves integrals over the moduli space of the torus. but a similar analysis should be possible with the hypermultiplet kinetic term as well. one can take an appropriate limit in which the stringy eﬀects on the heterotic side disappear and the theory reduces to some appropriate N = 2 supersymmetric quantum ﬁeld theory25 . and non-perturbative terms.86]! This agreement is quite remarkable. one might ask if a similar agreement can be found for the non-perturbative contribution from the heterotic string theory as well.85. The results of the above calculation in heterotic and type II string theories agree in all the cases tested [79. we can rewrite the exact vector multiplet kinetic term in terms of the heterotic variables.A. by going near a special point in the moduli space. (There is no perturbative contribution in the heterotic theory beyond one loop due to some Adler-Bardeen type non-renormalization theorems. calculated explicitly in the heterotic string theory. 1. where. This can be carried out using the method developed by 24 Here we describe the test using the vector multiplet kinetic term. In particular the heterotic variables include the heterotic dilaton ΦH which is in the vector multiplet. 25 This is in the same spirit as in the case of toroidal compactiﬁcation of heterotic string theory. Thus now the calculation of these non-perturbative eﬀects on the heterotic side reduces to a calculation in the N = 2 supersymmetric ﬁeld theory. Typically the expansion involves tree. with the expansion of the exact answer. calculated at the tree level.

One such theory is a conjectured theory living in eleven dimensions.) Let us now compactify this supergravity theory on a circle of radius R(∼ (S) g10.1) where G(S) ∼ dC (S) is the four form ﬁeld strength associated with the three form ﬁeld C (S) . However. The bosonic ﬁelds in (S) N = 1 supergravity theory in eleven dimensions consist of the metric gMN (S) and a rank three anti-symmetric tensor ﬁeld CMNP (0 ≤ M. sometime analysis similar to those that lead to various duality conjectures can also lead to the discovery of new theories. Gauge Theory and Strings Seiberg and Witten [88].1 M-theory in eleven dimensions The arguments leading to the existence of M-theory goes as follows [19. − (12)4 (8. Take type IIA string theory in ten dimensions. 8. In this section we shall give a brief description of this theory following references [19. N ≤ 10). The bosonic part of the action of this theory is given by [112] SSG = 1 1 −g (S) R(S) − G(S)2 d11 x 8 (2π) 48 1 µ0 ···µ10 (S) (S) (S) ε Cµ0 µ1 µ2 Gµ3 ···µ8 Gµ7 ···µ10 . This theory is now known as M-theory. Then the eﬀective action in the dimensionally reduced theory agrees with that of type IIA . 93–98]. In writing down the above equation we have set the eleven dimensional Planck mass to unity (or equivalently we can say that we have absorbed it into a redeﬁnition of the metric.10 ) measured in the (S) supergravity metric gMN and ignore (for the time being) the Kaluza-Klein modes carrying momentum in the internal direction. Besides providing a non-trivial test of string duality. The low energy eﬀective action of this theory is non-chiral N = 2 supergravity in ten dimensions. the relationship between the two theories is as follows. 89]. Again there is perfect agreement with the results from the type II side [87]. this also shows that the complete Seiberg-Witten [88] results (and more) are contained in the classical geometry of Calabi-Yau spaces! 8 M-theory So far we have discussed dualities that relate known string theories.312 Unity from Duality: Gravity. It is well known that this can be obtained from the dimensional reduction of N = 1 supergravity in eleven dimensions [90]. More speciﬁcally.

The situation is illustrated in Figure 17. has been analogous to the evidence for various duality conjectures based on the comparison of their low energy eﬀective action. as discussed so far. One might ask if there are more precise tests involving the spectrum of BPS states. This corresponds to strong coupling limit of the type IIA string theory. These are BPS states. The relationship between M-theory and various other supergravity/string theories. Note that as the radius R ∼ g10. C10µν (S) (0 ≤ µ. powers in ﬁelds.A. 17. g10µ (S) e2Φ/3 Aµ . The evidence for the existence of an eleven dimensional theory. We are using the convention that Φ = 0 corresponds to compactiﬁcation on a circle of unit radius. The non-trivial part of the conjecture is that it describes a Lorentz invariant theory in eleven dimensions. M theory on S 1 will have Kaluza-Klein modes representing states in the eleven dimensional N = 1 supergravity multiplet carrying momentum along the compact x10 direction. whose low energy limit is 11-dimensional N = 1 supergravity.10 approaches ∞. Φ → ∞. and can be shown to belong to the 256 dimensional ultra-short (S) .33) under the identiﬁcation [90]: g10. Sen: Duality Symmetries string theory given in (4. (S) Cµνρ (S) (S) gµν 313 e−Φ/12 gµν Bµν . (8. There are indeed such tests.2) Cµνρ .10 = eΦ/3 . ν ≤ 9) . This theory has been called M-theory. Part of the conjecture is just the deﬁnition of M-theory as the strong coupling limit of type IIA string theory. This leads one to the conjecture [19. Here denotes equality up to additive terms involving second and higher 1 S compactification M-theory Type IIA string theory Low energy limit D=11 supergravity S Type IIA supergravity 1 compactification + low energy limit Low energy limit Fig. 89] that in the strong coupling limit type IIA string theory approaches an 11 dimensional Lorentz invariant theory.

Thus the number of ultra-short multiplets with k-units of Aµ charge is determined in terms of the number of normalizable supersymmetric ground states of this quantum mechanical system. the mass of such a state. and have radii R1 and R2 respectively.IIA duality conjecture. Gauge Theory and Strings representation of the supersymmetry algebra. Thus one should be able to ﬁnd these states in type IIA string theory on . the collective coordinate dynamics of the 0-branes is determined from the dynamics of massless open string states with ends lying on the D0-brane. In M-theory these states carry k units of g10µ charge. This theory has sixteen fermion zero modes whose quantization leads to a 28 = 256 fold degenerate state. If we now recall that in type IIA string theory Aµ arises in the RR sector. Such states are diﬃcult to study. k2 ) relatively prime. In particular the state with k = 1 corresponds to a single Dirichlet zero brane. is k1 R1 2 (S) + k2 R2 2 · (8.IIA duality.314 Unity from Duality: Gravity. as predicted by the M-theory . The analysis can be simpliﬁed by compactifying M-theory on T 2 and considering the Kaluza-Klein modes carrying (k1 . k2 ) units of momenta along the two S 1 ’s. The charge quantum number characterizing such a state is the momentum (k/R) along S 1 . For k = 2 such a bound state with the correct degeneracy has been found [91]. as elementary string excitations are neutral under RR sector gauge ﬁelds. Thus we see that we indeed have an ultra-short multiplet with unit Aµ charge. we see that these states cannot come from elementary string states. Since g10µ gets mapped to Aµ under the M-theory . Assuming that the two S 1 ’s are orthogonal.3) For (k1 . such a state has strictly less energy than the sum of the masses of any other set of states with the same total charge [92]. up to a proportionality factor. The main obstacle to this analysis is that a charge k state has the same energy as k charge 1 states at rest. What about states with k > 1? In type IIA string theory these must arise as bound states of k D0-branes. However Dirichlet 0-branes in this theory do carry Aµ charge. these states must carry k units of Aµ charge in type IIA string theory. As usual. The analysis for higher k still remains to be done. Thus for every integer k we should ﬁnd such BPS states in type IIA string theory in (S) ten dimensions. and in this case is described by the dimensional reduction of N = 1 super-Maxwell theory from (9 + 1) to (0 + 1) dimensions. Thus the bound states we are looking for sit at the bottom of a continuum. Finding these bound states is much more diﬃcult than the bound state problems discussed earlier. Dynamics of collective coordinates of k D0 branes is given by the dimensional reduction of N = 1 supersymmetric U (k) gauge theory from (9 + 1) to (0 + 1) dimensions.

it is dual to IIA on S 1 . But we know that IIA on S 1 is related by T duality to IIB on S 1 . 8. from Lorentz invariance of type IIA string theory we know that we have Lorentz invariance in coordinates x0 . . we can now construct new vacua of the theory by compactifying M-theory on various manifolds. . Z) in type IIB string theory. Now IIB on S 1 has an SL(2. Thus one might ask. Z) of IIB is a non-perturbative symmetry. what does it correspond to in M-theory on T 2 ? One can ﬁnd the answer by using the known map between the massless ﬁelds in the two theories. In the limit when the radius of both the compact circles are taken to be large. x9 are non-compact. • Then from the conjectured SL(2. Consider M-theory on T 2 . . . and the action of SL(2. Sen: Duality Symmetries 315 S 1 (which. Turning this analysis around we see that this also supports the ansatz that M-theory. . x9 when all the coordinates x0 . The argument goes as follows: • First of all.2. (For a review of 26 In fact. Z) symmetry in M-theory is simply the group of global diﬀeomorphisms of T 2 [61. these states can be shown to be in one to one correspondence to a class of supersymmetric vacua in a (1 + 1) dimensional supersymmetric gauge theory compactiﬁed on a circle26 . this would mean that we should have Lorentz invariance in coordinates x0 . All such states have been found with degeneracy as predicted by the M-theory – IIA duality. . Z) duality symmetry of type IIB string theory we know that we have an exchange symmetry between the 9th and the 10th coordinate of M-theory when these coordinates are compact. these states are related via an R → (1/R) duality transformation to the ultra-short multiplets in type IIB on S 1 discussed in Section 5. 94].2 Compactiﬁcation of M-theory Given the existence of M-theory. 93. is a fully Lorentz invariant theory in eleven dimension. By following the same kind of argument. According to M-theory – type IIA duality. . . Z) strong-weak coupling duality inherited from ten dimensional type IIB string theory. It turns out that this SL(2. is equivalent to M-theory on T 2 ) without encountering the diﬃculties mentioned earlier. according to the conjecture. Thus we again have an example of “duality of dualities”. But in M-theory on T 2 it is simply a consequence of the diﬀeomorphism invariance of the 11dimensional theory.A. . The SL(2. Thus we have a duality between M-theory on T 2 and IIB on S 1 . deﬁned as the strong coupling limit of IIA. x10 . There are also other consistency checks on the proposed M-theory – IIA duality. .

we can consider M-theory compactiﬁed on K3.316 Unity from Duality: Gravity. for (S) odd n this is also accompanied by a reversal of sign of CMNP . sits in the bulk. together with a change of sign of the three (S) form ﬁeld CMNP . Each of these duality conjectures satisfy the consistency condition that the theory on the right hand side. These can all be regarded as appropriate strong coupling limits of type IIA compactiﬁcation on the same manifold. is dual to type IIA string theory compactiﬁed on the manifold on the left hand side. The duality between M-theory on S 1 /Z2 and the E8 × E8 heterotic string theory is particularly amusing. the real world and the shadow world live at two ends of the line and interact only via the exchange of supergravity multiplets . on the other hand. S 1 /Z2 denotes a real line segment bounded by the two ﬁxed points on S 1 . Gauge Theory and Strings compactiﬁcation of eleven dimensional supergravity. Calabi-Yau. Now in the conventional heterotic string compactiﬁcation on Calabi-Yau spaces. The essential feature of this strong coupling limit is the emergence of Lorentz invariance in one higher dimension. Such a theory cannot be constructed by conventional compactiﬁcation of type IIA string theory at weak coupling. see [138]. For example. known as the hidden sector or the shadow world. upon further compactiﬁcation on a circle. In the M-theory picture these two sectors are physically separated in space. The second E8 . It turns out that the two E8 gauge multiplets arise from “twisted sector” of the theory and sit at the two ends of this line segment. and various orbifolds. But in general these cannot be regarded as perturbative string vacua. all the observed gauge bosons and charged particles come from one E8 and are neutral under the second E8 [128].) For example. M-theory on a Calabi-Yau manifold gives a ﬁve dimensional theory with N = 1 supersymmetry [137]. The supergravity sector. Of course in many cases these non-perturbative vacua are related to perturbative string vacua by conjectured duality relations. Here the Z2 transformation acts by reversing the orientation of S 1 . Some examples of such conjectured dualities are given below [95–98]: M-theory on ↔ (E8 × E8 ) heterotic in D = 10 S 1 /Z2 K3 T /Z2 5 ↔ Heterotic/Type I on T 3 ↔ IIB on K3 ↔ Type I/Heterotic on T 7 ↔ Type IIB on T 8 /Z2 T 8 /Z2 T 9 /Z2 In each case Z2 acts by reversing the sign of all the coordinates of T n . These duality conjectures can be arrived at by using arguments very similar to those used in arriving at string duality conjectures. is expected to be responsible for supersymmetry breaking. In other words.

x9 ). . and σ denotes the transformation CMNP → −CMNP . D to be T 4 spanned by x6 . Many of the listed duality conjectures involving M-theory (in fact all except the ﬁrst one) can be derived by ﬁberwise duality transformation [98]27 . The Z2 generator is I5 · σ where I5 changes the sign of all ﬁve coordinates (S) (S) (x6 . if we take the theory on the right hand side and make an R → (1/R) duality transformation on one of the circles. 10 Let us express this as (I1 · σ) · I4 where I1 changes the sign of x . we should have inverse cube law for the gravitational force instead of inverse square law. it converts • type IIA theory to type IIB theory. . . hB to be (−1)FL (this can be shown to be the image of hA in the type IIA string theory). x10 ) on T 5 . 27 The duality between E × E heterotic string theory and M-theory on S 1 can be 8 8 “derived” from other known duality conjectures by taking the inﬁnite radius limit of a lower dimensional duality relation [75]. and I4 changes the sign of (x6 . and • (−1)FL · I4 into I4 . which is a special case of type IIB on K3. x9 . . . No such direct limit comes from the inverse square law of gauge interaction. Thus we get the duality: (M-theory on T 5 /Z2 ) ↔ (IIB on K3).A. . since if this dimension is too large. . Let us for example consider the duality (M theory on T 5 /Z2 ) ↔ (type IIB on K3). . hA to be I1 · σ. B on KB to be type IIA string theory. and g to be I4 . On the other hand. . Thus the theory on the right is dual to type IIB on T 4 /I4 . . We now use the result of ﬁberwise duality transformation: (A on KA × D/(hA · g)) ≡ (B on KB × D/(hB · g)) by choosing A on KA to be M-theory on S 1 . The theory on the left hand side is M-theory on T 5 /Z2 . since gauge ﬁelds live on the boundary of S 1 /Z2 and hence do not get aﬀected by the existence of this extra dimension. Thus we get the duality M-theory on (S 1 × T 4 /I1 · σ · I4 ) ↔ IIA on T 4 /(−1)FL · I4 . Sen: Duality Symmetries 317 propagating in the bulk [136]! It has been suggested that this physical separation could be as large as a millimeter [99]! This limit comes from the analysis of the ﬁfth force experiment.

B 357 (1995) 307 [hep-th/9506160]. M. Phys. S. Hull and P. Phys. Phys. Rep. Duﬀ. Phys.] 45B. Townsend [hep-th/9612121. Lett. J. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] A. de Wit and J. B 438 (1995) 109 [hep-th/9410167]. 97].318 Unity from Duality: Gravity. Witten. Duﬀ. Phys. Phys. B 443 (1995) 85 [hep-th/9503124]. B 450 (1995) 103 [hep-th/9504027]. Mukhi [hep-ph/9710470]. Rev. J. Phys. Klemm [hep-th/9705131]. Phys. we can get non-perturbative enhancement of gauge symmetries in M-theory when the compact manifold develops singularities [19. B 78 (1978) 97. Dabholkar. 134. R. Nucl. Witten. J. Strominger. Lett. Duﬀ [hep-th/9611203]. A. 244 (1994) 77 [hep-th/9401139]. Giveon. Sen. 1986). M. Int. Khuri. A. Lu. Polchinski and E. Nucl. Duﬀ and R. Lerche [hep-th/9611190. J. 1989). A. As in the case of type IIA string theory. Phys. C. Harvey and A. Khuri and J. Theisen. Louis [hep-th/9801132]. Nucl. The extra massless states required for this symmetry enhancement come from membranes wrapped around the collapsed two cycles of the singular manifold. Phys. Phys. hep-th/9710246]. Douglas [hep-th/9610041]. Nucl. 259 (1995) 213 [hep-th/9412184]. Fre. B 55 (1) [hep-th/9607201]. Polchinski [hep-th/9602052]. Rep. Phys. M. hep-th/9712004]. Witten and D. B 442 (1995) 47 [hep-th/9501030]. Superstring Theory. Townsend. Phys. Schwarz and E. J. J. edited by D. Louis [hep-th/9612192]. E. Nucl. Mod. C. Polchinski [hep-th/9611050]. Proc. Nucl. Phys. Lett. C. Nucl.C (1996) 59 [hep-th/9512043] and references therein. B 411 (1994) 473 [hep-th/9305142]. W. J. M. [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] . M. Rabinovici. Forste and J. M-theory contains classical membrane and ﬁve-brane soliton solutions carrying electric and magnetic charges of (S) CMNP respectively [8]. Hull. J. B [Proc. Lectures on String Theory (Springer. E. B 460 (1996) 525 [hep-th/9510169]. P. Polchinski [hep-th/9411028]. S. A. Nucl. 68 (1996) 1245 [hep-th/9607050]. B 357 (1995) 545 [hep-th/9506194]. 135]. Lust and S. Kiritsis [hep-th/9708130]. P. A 9 (1994) 3707 [hep-th/9402002]. Phys. Johnson and J. Witten. Vafa [hep-th/9702201]. Vols. 1 and 2 (Cambridge University Press. M. Nucl. Suppl. Phys. E. M. Trigiante [hep-th/9801144]. Gauge Theory and Strings This duality was ﬁrst conjectured in [96. Sup. Sen. C. Chaudhuri. B 449 (1995) 535 [hep-th/9504047]. M. Schwarz. Porrati and E. gr-qc/9707012. Olive. S. Polchinski. Green. Mod. B.

Phys. Phil. B 377 (1996) 67 [hep-th/9505187]. Phys. Phys. G. Phys. J. Lett. Phys. Polchinski. Nucl. Roy. Rev. Kounnas. B 333 (1994) 92 [hep-th/9401133]. Wilczek. Phys. Schwarz. Trivedi and F. Lett. Nucl. N. Osborn. Lett. Nucl. Sen. Phys. Chaudhuri and J. S. M. Nucl. S. Rev. Phys. Dabholkar and J. B. Manton. Witten. Lett. S. B 303 (1993) 22 [hep-th/9209016]. S. Phys. Rev. Ferrara and C. Phys. Sen. 63 (1989) 478. Cremmer. Hitchin. Harvey [hep-th/9407111]. Nucl. N. Schwarz and P. G. Nucl. B 400 (1993) 103 [hep-th/9205008]. Phys. Comm. Hassan and A. Nucl. Nucl. A 6 (1991) 2677. B 404 (1993) 109 [hep-th/9207053]. Gibbons and N. Press. Seiberg. Sen. Ibanez. Phys. Math. Lett. Gauntlett. Lett. Mod. A. Das and S. Terentev. Lett. Green and J.J. Comm. Lett. B 122 (1983) 143. Chaudhuri. Ruback. A. A. Nucl. 115 (1988) 267. J. Gibbons. Phys. Atiyah and N. Phys. Phys. Schwarz and A. Gauntlett and J. Phys. Annals Phys. 177 (1996) 775. B 411 (1994) 443 [hep-th/9305068]. Lett. Nucl. de Roo. S. B 274 (1986) 183. J. Phys. West. Lykken. 49 (1989) 713. J. Phys. B 312 (1993) 105 [hep-th/9305185]. A. A. 1981). West. Mathur. Phys. Olive. Ferrara. G. 225 (1993) 290. in “Uniﬁcation of Fundamental Particle Interactions” (Plenum. C. Blum. Porrati. B 126 (1983) 301. E. Montonen and D. C. in “Superspace and Supergravity” (Cambridge Univ. B 328 (1989) 406. Trans. G. Phys. D. Press). L. Nucl. Nucl. B 340 (1990) 33. B 411 (1994) 35 [hep-th/9304154]. S. M. Phys. B 390 (1993) 3 [hep-th/9207016]. Lust and F. J. Shapere. Phys. Lett. B 238 (1984) 181. J. 75 (1995) 2264 [hep-th/9505054]. . Schwarz. Quevedo. Rev. S. Phys. B 360 (1995) 13 [hep-th/9508143]. Sen. Lett. Hockney and J. B 255 (1985) 515. Lond. J. Harvey and F. J. Nucl. A. Sen: Duality Symmetries [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] 319 [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] M. Gibbons and P. Phys. E. Math. Phys. Schroers. D 43 (1991) 526. J. Phys. 1980). Schwarz. Phys. P. B 375 (1996) 103 [hep-th/9601152]. Phys. B 329 (1994) 217 [hep-th/9402032]. Lett. Lett. Dabholkar. B 460 (1996) 335 [hep-th/9510135]. S. Phys. Nucl. B 181 (1986) 263. A 315 (1985) 459. B 72 (1977) 117. Lett. Lett. Phys. Schwarz. Segal and A. Lett. A. Phys. Ruiz. Harvey. Geometry and Dynamics of Magnetic Monopoles (Cambridge Univ. Maharana and J. Phys. Phys. Porrati.A. B 375 (1992) 103 [hep-th/9109038]. Rev. B 303 (1988) 286. Rey. Phys. Schwarz [hep-th/9209125]. Lett. Manton and B. J. Mod. Soc. M. Selby. Lett. B 83 (1979) 321. A.F. Julia. Phys. B 226 (1983) 269. 75 (1995) 4724 [hep-th/9510017]. J. G. Phys. B 249 (1990) 35. M. A 107 (1985) 21. Sen. Font. Polchinski. Phys. Lett. Howe and P. A 8 (1993) 2023 [hep-th/9303057]. D 52 (1995) 7168 [hep-th/9506048]. H. Kounnas and M. Sov. Phys. Nucl. Sen. J.

B 387 (1996) 492 [hep-th/9607082]. Lett. Witten. B 465 (1996) 399 [hep-th/9512196]. Gauge Theory and Strings [66] J. Mayr. Bershadsky. B 459 (1996) 537 [hep-th/9508155]. Vafa. Phys. W. Kachru and C. Phys. Font. B 243 (1984) 112. 46 (1996) 30 [hep-th/9508154]. [85] V. Phys. [75] A. Nucl. [83] B. Lerche. R. Phys. M. Dai. Huq and M. Sethi and M. Vafa. Mod. Lerche. Phys. [74] J. J. Stern. Rev. B 434 (1995) 179 [hep-th/9408083]. Lett. [78] A. Phys. V. Phys. Campbell and P. B 463 (1996) 435 [hep-th/9512078]. B 357 (1995) 71 [hep-th/9506110]. B 450 ((1995) 69 [hep-th/9605105]. Phys. B 426 (1994) 19 [hep-th/9407087]. [79] S. B 469 (1996) 415 [hep-th/9511088]. [80] S. Phys. J. Taylor. Narain and T. Ibanez and F. B 357 (1995) 313 [hep-th/9506112]. Polchinski. P. [84] B. [73] M. Ferrara. Phys. Witten. Phys. Sen. Giani and M. Sen. Nucl. Blum. A 4 (1989) 2767. Hunt and R. [91] S. Lauwers and A. Nucl. [93] J. Harvey. [94] P. [69] C. A 4 (1989) 2073. P. Townsend. B 460 (1996) 506 [hep-th/9510209]. [92] A. B 255 (1985) 569. L. Witten [hep-th/9507050]. B 350 (1995) 184 [hep-th/9501068]. Witten. Vafa and N. Quevedo. Phys. Phys. D 50 (1994) 6041 [hep-th/9407031]. 2 (1985) 293. Phys. P. Nucl. Class. Nucl. [88] N. [72] E. [67] M. Nucl. . Phys. Mukhi. Suppl. Sethi and M. Nucl. D 54 (1996) 2964 [hep-th/9510229]. Nucl. [81] A. Nucl. B 463 (1996) 398 [hep-th/9510225]. Phys. Hunt. Phys. Phys. Klemm. [90] F. Lett. Phys. Schwarz. Vafa. Sen [hep-th/9609176]. Phys. R. Nucl. Strominger and C. J. W. Phys. B 367 (1996) 97 [hep-th/9510086]. Pernici. D 30 (1984) 325. Phys.320 Unity from Duality: Gravity. Nucl. Sen. Nucl. B 474 (1996) 361 [hep-th/9604070]. B. Porrati. Phys. D 53 (1996) 2874 [hep-th/9711026]. Phys. Lett. Leigh. Seiberg and E. Nucl. B 506 (1997) 223 [hep-th/9705030]. Nucl. van Proeyen. [77] C. [82] G. B 455 (1995) 109 [hep-th/9507115]. Rev. de Wit. B 461 (1996) 537 [hep-th/9510093]. Kaplunovsky. B 381 (1996) 427 [hep-th/9512138]. Sadov and C. West. B 361 (1995) 59 [hep-th/9505162]. B 463 (1996) 420 [hep-th/9511222]. Horava and E. Lett. Phys. Klemm. [70] S. Lett. E. Polchinski. Klemm. Nucl. Phys. A. I. Lett. Schimmtigk [hep-th/9609082]. Vafa. Rev. Aspinwall. Nucl. Schimmrigk. B 475 (1996) 94 [hep-th/9603142]. Nucl. B 477 (1996) 746 [hep-th/9604034]. Phys. [76] A. Vafa. Proc. Leigh and J. Quant. Phys. [96] K. Rev. W. [68] A. Gava. Lett. Phys. B 460 (1996) 541 [hep-th/9511030]. K. [95] P. Grav. Antoniadis. Lett. Nucl. Nucl. Mayr. [86] I. M. Sadov and C. Lett. [89] P. Mayr and C. Louis and S. Aldazabal. Lynker and R. C. Mod. Phys. [87] S. Dasgupta and S. Warner. Bershadsky. Phys. B 431 (1994) 484 [hep-th/9408099]. Stern [hep-th/9705046]. Nucl. [71] M. Vafa and E. Phys. A. A. A. Namazie. Theisen. Phys. Vafa. Phys. Phys. Nucl. Kachru. Lerche and P. B 447 (1995) 62 [hep-th/9503057]. B 398 (1997) 47 [hep-th/9607145]. Nucl. Phys. Sen. V.

[120] M. Friedan and E. S. Rohm. M. S. H. 54 (1985) 502. [99] E. Lett. E. B 208 (1982) 323. D 54 (1996) 1667 [hep-th/9601038]. Banks and L. 133 (1990) 163. B. Nucl. B 55 (1997) 6423 [hep-th/9606165]. Derendinger. Phys. Lett. [109] B. Phys. Strominger. Phys. Sen: Duality Symmetries 321 [97] E. B 169 (1986) 41. J. Rev. J. D 55 (1997) 6382 [hep-th/9610140]. B 307 (1988) 93. de Wit and H. D 20 (1979) 936. Cremmer. Girardello. Douglas [hep-th/9512077. Nucl. Phys. Phys. [119] E. Comm. M. de Roo and B. 1987). [118] M. van Proeyen. B 109 (1982) 444. Phys. Weinberg. Manton. B 267 (1986) 75. Manton. Strominger. Douglas and J. Gimon and J. Sagnotti. A. Nucl. J. M. Greene [hep-th/9702155] and references therein. Sarmadi and E. [115] M. Nucl. [117] A. van Proeyen. Phys. B 216 (1989) 59. [102] A. Phys. Bagger and E. Mod. Phys. Nucl. Callias. [100] A. Sagnotti. B 195 (1982) 97. de Wit and P. Banks. 80 (1981) 575. Rev. B 149 (1984) 117. B 250 (1985) 385. Rev. Phys. Zumino. B 74 (1978) 61. S. de Wit and A. Nucl. B. Nucl. Taubes. Lett. Phys. E. Comm. Bergshoeﬀ. Pradisi and A. Douglas [hep-th/9612126] published in JHEP electronics journal. Lett. C.A. Ferrara. Morrison and A. Galicki. Mandelberg. E. Math. Narain. Nicolai. Phys. Phys. Nucl. B 121 (1977) 393. Phys. B 158 (1979) 189. [98] A. Nucl. Phys. K. Caceres. [105] N. Goddard. B. B 463 (1996) 383 [hep-th/9512219]. Sagnotti. Witten. Pradisi and A. [114] D. B 327 (1989) 461. B 151 (1985) 21. Scherk and B. Phys. Phys. B 299 (1988) 613. Nucl. Kaplunovsky and M. E. L. Cremmer and B.. Dixon. B 256 (1985) 253. Phys. J. Comm. Lett. Bianchi. Nucl. Phys. D. Nucl. B 245 (1984) 89. Phys. Witten. Nucl. Phys. B 222 (1983) 1. Phys. Phys. Math. Johnson and R. Scherk and S. World Scientiﬁc (1989). D 24 (1981) 3065. 91 (1983) 235. Phys. Corrigan and P. Myers. Gaillard and B. D. Julia. B. Phys. Phys. P. Nicolai. de Wit. Phys. Phys. M. G. Nucl. Phys. Open Strings and their Symmetry Groups (Talk at Cargese Summer Inst. K. [103] H. de Wit. Phys. Phys. Lett. E. J. Nucl. Julia. Ferrara. B 120 (1983) 105. Phys. Green and J. [107] K. M. Greene. V. G. Townsend. van Niewenhuizen. D. B 159 (1979) 141. Lett. B 451 (1995) 525 [hep-th/9505073]. 108 (1987) 117. Phys. B 245 (1984) 89. Phys. B 493 (1997) 73 [hep-th/9606036]. B 279 (1987) 369. Horava. Witten. A 11 (1996) 1339 [hep-th/9603113]. Rev. [104] E. [106] C. Phys. Lett. Comm. Math. Lett. B 80 (1978) 48. Dixon. Polchinski. de Wit and A. Phys. B 110 (1982) 54. [112] E. C. B 255 (1985) 93. Narain. Phys. Math. Lett. E. B 451 (1995) 109 [hep-th/9504145]. [101] E. Lett. G. Cremmer and B. Witten [hep-th/9507121]. [116] C. Ferrara. [113] M. B 231 (1989) 251. Phys. Nucl. Chapline and N. . Chamseddine. 62 (1978) 213. Phys. de Roo. Schwarz. van Proeyen. B 376 (1992) 365. Martinec and R. Lett. B 276 (1992) 333. A. Nucl. Hull and P. Martinec. Phys. B 193 (1981) 221. Nucl. Phys. Sen. Nucl. Rev. Zumino. Nucl. Math. B. Nucl. [110] B. C. Kounnas. Diaconescu. Comm. hep-th/9604198]. B. Bergshoeﬀ. Ferrara and A. Strominger. Phys. B 451 (1995) 96 [hep-th/9504090]. Phys. Rev. Phys. Nucl. Strominger. Lett. Phys. Cremmer. Phys. E. [111] J. Gross. Moore [hep-th/9603167]. Harvey. Polchinski. Gomis [hep-th/9712230]. de Wit and L. Douglas and G. Nucl. [108] T. B 217 (1983) 143. Nucl. Phys. T. in “Strings 89”.

Gliozzi. Phys. Blum and A. Phys. Pope. Mukhi. Rev. Witten. C. Nicolai. Luscher and H. . B 258 (1985) 46. Polchinski. Gaberdiel. Ann. Olive. Witten. Phys. Phys. Aspinwall [hep-th/9611137] and references therein. Ferrara. Ferrara [hep-th/9506144]. B 385 (1996) 125 [hep-th/9606044]. [131] K. B 31 (1971) 86. Witten. Nucl. S. Aharony. Class. J. [130] M. Phys. Phys. Vafa and E. Phys. Bergshoeﬀ. Duﬀ. Nucl. B 475 (1996) 115 [hep-th/9605184]. Johnson. Ceresole. Mukhi. R. de Wit. [123] P. P. Freedman and S. Minasian. [126] P. Gauge Theory and Strings [121] L. Horowitz. Phys. Nucl. M. Duﬀ and R. 5 (1988) 189. Park. J. Nucl. M. B 479 (1996) 260 [hep-th/9607057]. [124] A. B. Hauer and B.322 Unity from Duality: Gravity. Sadov and C. Math. B 498 (1997) 135 [hep-th/9702061]. Nucl. Nilsson and C. M. Gopakumar and S. Mukhi. B 479 (1996) 285 [hep-th/9606176]. Candelas. [122] E. B 189 (1987) 75. Nucl. R. Nucl. Sonneschein. Zaﬀaroni. [136] E. 130 (1986) 1. R. Phys. D. Phys. Schwarz. Strominger and E. B 463 (1996) 435 [hep-th/9601036]. Theisen. Nucl. Phys. Nucl. Yankielowicz and S. [132] A. Duﬀ. (NY) 185 (1988) 330. A. Nucl. Nucl. [135] A. Phys. Nucl. B 447 (1995) 261 [hep-th/9505053]. [134] E. [129] E. Nucl. Aspinwall and M. G. Sezgin and P. Duﬀ. Witten. Quant. B 436 (1995) 507 [hep-th/9406198]. Sen [hep-th/9707123] published in JHEP electronic journal. Liu and R. Phys. Aspinwall and D. Minasian. Morrison [hep-th/9404151]. D 55 (1997) 7345 [hep-th/9702165]. Minasian and E. B 305 (1988) 545. Schwarz. Dasgupta and S. Intrilligator. Kachru. Hoppe and H. P. Dabholkar and J. Neveu and J. B 270 (1986) 186. Phys. N. A. Cardy. D 3 (1971) 2415. Phys. Comm. Ann. Lett. [139] J. Phys. E. (NY) 134 (1981) 85. Seiberg and E. Leigh. Phys. Morrison. J. D’Auria and S. Nucl. Johansen. 80 (1981) 443. Bershadsky. Vafa. M. O. J. S. Park [hep-th/9611119]. Phys. Lett. Ferrara and T. B 387 (1996) 735 [hep-th/9605131]. Nucl. B 452 (1995) 261 [hep-th/9506126]. J. Phys. [138] M. Phys. B 122 (1977) 253. B 489 (1997) 139 [hep-th/9611186]. Phys. Freedman. Nucl. [137] A. Rep. E. B 471 (1996) 135 [hep-th/9602070]. Gimon and C. B 493 (1997) 177 [hep-th/9611222]. B 460 (1996) 489. Phys. [128] P. K. A. Grav. T. B. M. Nucl. M. S. I. Nucl. Phys. Phys. Lett. Rev. Gaberdiel and B. B 320 (1989) 135. A. B 65 (1976) 282. Phys. Nicolai. Ramond. D 4 (1971) 1109. Lett. Townsend. Nucl. Phys. Phys. Phys. [127] M. V. Phys. B 395 (1997) 36 [hep-th/9608186]. Berkooz. S. J. Lett. Zwiebach [hep-th/9801205]. Taylor. B 481 (1996) 215 [hep-th/9605200]. Phys. Minasian [hep-th/9602102]. B. Phys. R. B 471 (1996) 195 [hep-th/9603150]. B 500 (1997) 3 [hep-th/9703166]. L. Cadavid. [125] F. Nucl. Nucl. J. Scherk and D. Phys. Khuri and R. B 387 (1996) 71 [hep-th/9607019]. Rev. Witten. Alvarez-Gaume and D. de Wit. Antoniadis. B 394 (1997) 302 [hep-th/9607041]. R. J. Lett. Alvarez-Gaume. D. [133] M. Lett. Witten. Witten. Phys. Phys. Phys. Zwiebach [hep-th/9709013]. Gross. Duﬀ. Sen.

MALDACENA Institute for Advanced Study. .LECTURE 4 LES HOUCHES LECTURES ON LARGE N FIELD THEORIES AND GRAVITY J. NJ.S. U. Princeton.A.

.1 Construction . . . . . . . . . . . . 369 6. . . . . . . . . . . 355 4. . . . . 363 5. . . . . . . . . . . . . . . . . . . . . . . . . .2 Holography . . . . . . . . . . . . . 336 2. . . . . . . . . . . . . . . . . . . . . . 342 3. . . . . . .1 Wilson loops and minimum surfaces . . . . 368 6 Theories at ﬁnite temperature 369 6. . . . . . . . . .1 Two-point functions . . . . . . . . .2 Three-point functions . . . . . . . . . . . . . . . . . . . .Contents 1 General introduction 325 2 The correspondence 330 2. . 338 /CFT correspondence 341 3 Tests of the 3. . . . . .1 The spectrum of chiral primary operators . . . . . .2 Other branes ending on the boundary . . . . . . . . . . . .1 The ﬁeld ↔ operator correspondence . . . . . . . 351 4 Correlation functions 353 4. . . . . . . . . . . . . . . . . . . . . . . .2 Matching of correlation functions and anomalies . . . . . . . . . . 372 Ë . . . .2 Thermal phase transition . . . . . . . . 359 5 Wilson loops 362 5.

We focus on the case of the N = 4 supersymmetric gauge theory in four dimensions. Maldacena Abstract We describe the holographic correspondence between ﬁeld theories and string/M theory. The idea was to view all these particles as diﬀerent oscillation modes of a string. These lecture notes are based on the Review written by Aharony et al. We describe the main results that have been derived from the correspondence in the regime that the ﬁeld theory is approximated by classical or semiclassical gravity. focusing on the relation between compactiﬁcations of string/M theory on Anti-de Sitter spaces and conformal ﬁeld theories. it is not how string theory was originally discovered. This is explained simply by assuming that the mass and angular momentum come from a rotating. Springer-Verlag 2002 . relativistic string of tension T . the mass of the lightest hadron with a given spin obeys a relation like m2 ∼ T J 2 + const. [1]. It was later discovered that hadrons and mesons are actually made of quarks and that they are described by QCD. String theory was discovered in an attempt to describe the large number of mesons and hadrons that were experimentally discovered in the 1960’s. For example. c EDP Sciences. We review the background for this correspondence and discuss its motivations and the evidence for its correctness. The string idea described well some features of the hadron spectrum. Even though though string theory is normally used as a theory of quantum gravity. This work is supported by the Packard foundation and the DOE grant DE-FGO291ER40654. 1 General introduction These lecture notes are taken out of the review [1]. A more complete set of references is given there.LES HOUCHES LECTURES ON LARGE N FIELD THEORIES AND GRAVITY J.

It was suggested by ’t Hooft that the theory might simplify when the number of colors N is large [7]. In particular we could consider a gauge theory where the coupling does not run (as a function of the energy scale). The ’t Hooft argument. One might qualitatively think of this new ﬁeld as the “thickness” of the string. The hope was that one could solve exactly the theory with N = ∞. as described in detail below. the diagrammatic expansion of the ﬁeld theory suggests that the large N theory is a free string theory and that the string coupling constant is 1/N . this theory will contain gravity. reviewed below. At low energies QCD becomes strongly coupled and it is not easy to perform calculations. QCD is asymptotically free. This is sometimes stated by saying that quarks have three colors. if one wants to quantize a four dimensional string theory an anomaly appears that forces the introduction of an extra ﬁeld. We will see that the strings arising in the large N limit of ﬁeld theories are the same as the strings describing quantum gravity. In supersymmetric theories it is sometimes possible to prove exact conformal invariance. as explained in the next section. This is at present the best available tool to do calculations in QCD at low energies. One possible approach is to use numerical simulations on the lattice. including quantum gravity. and then one could do an expansion in 1/N = 1/3. is equivalent (dual) to a ﬁeld theory. If the case with N = 3 is similar to the case with N = ∞ then this explains why the string model gave the correct relation between the mass and the angular momentum. is very general. Strings are not consistent in four ﬂat dimensions. the theory is conformally invariant. In this review we will study this correspondence between string theories and the large N limit of ﬁeld theories. If this is the case. sometimes called the “Liouville” ﬁeld [8]. which will be the main example . Then. so it suggests that diﬀerent kinds of gauge theories will correspond to diﬀerent string theories. string theory in some backgrounds. Gauge Theory and Strings QCD is a gauge theory based on the group SU (3). This ﬁeld on the string worldsheet may be interpreted as an extra dimension. why do we say that the string moves in ﬁve dimensions? The reason is that. Furthermore. The argument that gauge theories are related to string theories in the large N limit is very general and is valid for basically any gauge theory. It is crucial then that the ﬁve dimensional geometry is curved. In this way the large N limit connects gauge theories with string theories. meaning that the eﬀective coupling constant decreases as the energy increases. It is quite hard to ﬁnd quantum ﬁeld theories that are conformally invariant. so that it can correspond to a four dimensional ﬁeld theory.326 Unity from Duality: Gravity. so that the strings eﬀectively move in ﬁve dimensions. like any string theory. and the gravitational theory will live in as many dimensions as the number of ﬁelds we have on the string. Indeed. Namely. A simple example.

2). Here we have presented a very heuristic argument for this equivalence. particle-type soliton solutions. In fact. Four is the maximal possible number of supercharges for a ﬁeld theory in four dimensions. D-p-branes are charged under . Besides the gauge ﬁelds (gluons) this theory contains also four fermions and six scalar ﬁelds in the adjoint representation of the gauge group. we conclude that N = 4 U (N ) Yang-Mills theory could be the same as ten dimensional superstring theory on AdS5 × S 5 [9]. If they have one extended dimension they are called Done-branes or D-strings. D-branes are deﬁned in string perturbation theory in a very simple way: they are surfaces where open strings can end. The Lagrangian of such theories is completely determined by supersymmetry. a gauge ﬁeld living on the brane. Locally there is only one space with SO(4. which describe the oscillations of the branes. There is a global SU (4) R-symmetry that rotates the six scalar ﬁelds and the four fermions. Anti-de Sitter space is the maximally symmetric solution of Einstein’s equations with a negative cosmological constant. S 5 .J. The relationship we described between gauge theories and string theory on Anti-de-Sitter spaces was motivated by studies of D-branes and black holes in strings theory. analogous to the ’t Hooft-Polyakov [11. The simplest way for this to happen is if the ﬁve dimensional geometry has these symmetries. where gs is the string coupling constant. 2) isometries: ﬁve dimensional Anti-de-Sitter space. If they have zero spatial dimensions they are like ordinary localized. Since the gauge theory has an SU (4) SO(6) global symmetry it is rather natural that the extra ﬁve dimensional space should be a ﬁve sphere. Maldacena: Lectures on Large N Field Theories and Gravity 327 in this review. Now that we have added one more dimension it is not surprising any more to add ﬁve more to get to a ten dimensional space. is the supersymmetric SU (N ) (or U (N )) gauge theory in four dimensions with four spinor supercharges (N = 4). 12] monopole in gauge theories. These open strings have some massless modes. later we will be more precise and give more evidence for this correspondence. including the usual Poincar´ transformations as well as scale transformae tions and special conformal transformations (which include the inversion symmetry xµ → xµ /x2 ). This in turn implies that the low energy dynamics is described by a U (N ) gauge theory. These are called D-zero-branes. or AdS5 . D-branes are solitons in string theory [10]. and their fermionic partners. They are much heavier than ordinary fundamental strings when the string coupling is small. the tension of all D-branes is proportional to 1/gs . In this supersymmetric case we expect the strings to also be supersymmetric. So. They come in various dimensionalities. The conformal group in four dimensions is SO(4. These symmetries of the ﬁeld theory should be reﬂected in the dual string theory. We said that superstrings move in ten dimensions. If we have N coincident branes the open strings can start and end on diﬀerent branes. so they carry two indices that run from one to N .

First black hole entropy for extremal black holes was calculated in terms of the ﬁeld theory in [17]. we will see that correlation functions in the gauge theory can be calculated using the string theory (or gravity for large gs N ) description.328 Unity from Duality: Gravity. in the same way that a 0-brane (particle) can be charged under a one-form gauge potential (as in electromagnetism). In this D1-D5 case there were numerous calculations that agreed between the ﬁeld theory and gravity. If we consider a set of N coincident D-3-branes the near horizon geometry turns out to be AdS5 × S 5 . whose geometry is AdS5 × S 5 . On the other hand. by considering the propagation of . On the other hand low energy excitations on the branes are governed by the Yang-Mills theory. More generally. In this case the near horizon geometry involves AdS3 × S 3 and the low energy ﬁeld theory living on the D-branes is a 1+1 dimensional conformal ﬁeld theory. due to the large gravitational potential.19] and for absorption cross sections [20–22]. As an object is brought closer and closer to the black brane horizon its energy measured by an outside observer is redshifted. and they are part of the massless closed string modes. So. which belong to the supergravity (SUGRA) multiplet containing the massless ﬁelds in ﬂat space string theory (before we put in any D-branes). in turn. These calculations. Supergravity is the low-energy limit of string theory. These two pictures of D-branes are perturbatively valid for diﬀerent regimes in the space of possible coupling constants. while the low-energy gravitational description is perturbatively valid when the radius of curvature is much larger than the string scale. Perturbative ﬁeld theory is valid when gs N is small. and then agreement was shown for near extremal black holes [18. Gauge Theory and Strings p+1-form gauge potentials. If we now add D-branes they generate a ﬂux of the corresponding ﬁeld strength. These solutions are very similar to extremal charged black hole solutions in general relativity. and the energy seems to be very small. the low energy dynamics on their worldvolume is governed by a U (N ) gauge theory with N = 4 supersymmetry [13]. Indeed it is possible to ﬁnd solutions of the supergravity equations carrying these ﬂuxes. except that in this case they are black branes with p extended spatial dimensions. and it is believed that these solutions may be extended to solutions of the full string theory. were inspired by similar calculations for coincident D1-D5 branes. it becomes natural to conjecture that Yang-Mills theory at strong coupling is describing the near horizon region of the black brane. Like black holes they contain event horizons. These p + 1-form gauge potentials have p + 2-form ﬁeld strengths. and this ﬂux in turn contributes to the stress energy tensor so the geometry becomes curved. which turns out to imply that gs N should be very large. The ﬁrst indications that this is the case came from calculations of low energy graviton absorption cross sections [14–16]. It was noticed there that the calculation done using gravity and the calculation done using super Yang-Mills theory agreed.

One of the main points of these lectures will be that the strings coming from gauge theories are very much like the ordinary superstrings that have been studied during the last 20 years. but a light ray can go to the boundary and come back in ﬁnite time. One could argue that in general any quantum gravity theory in AdS deﬁnes a conformal ﬁeld theory (CFT) “on the boundary”. Massive particles can never get to the boundary. 26] for reviews. And. the points where operators are inserted [23. Thus. The asymptotically AdS case as well as the asymptotically ﬂat cases are special in the sense that one can choose a natural time and an associated Hamiltonian to deﬁne the quantum theory. Notice that when we say that the theory includes “gravity on AdS” we are considering any ﬁnite energy excitation. but it is certainly well-deﬁned at inﬁnity. 3] for earlier hints of the correspondence. The boundary is at an inﬁnite spatial distance. In some sense the situation is similar to the correspondence between three dimensional Chern-Simons theory and a WZW model on the boundary [27]. The radius of curvature of Anti-de Sitter space depends on N so that large N corresponds to a large radius of curvature. This is a topological theory in three dimensions that induces a normal (non-topological) ﬁeld theory on the boundary. The theory in AdS includes gravity. So in the end we claim that there is an equivalence between a gravitational theory and a ﬁeld theory. Supergravities on AdS spaces were studied very extensively. Maldacena: Lectures on Large N Field Theories and Gravity 329 particles between diﬀerent points in the boundary of AdS. then the U (N ) Yang-Mills theory gives a non-perturbative deﬁnition of string theory on AdS. However. Similarly. In some sense the ﬁeld theory (or at least the set of local observables in the ﬁeld theory) lives on the boundary of spacetime. where asymptotic ﬂatness is required. the mapping between the gravitational and ﬁeld theory degrees of freedom is quite non-trivial since the ﬁeld theory lives in a lower dimension. but the spacetime could have any topology as long as it is asymptotically ﬂat. see [25.J. since any string theory includes gravity. by taking the limit N → ∞. So this is really a sum over all spacetimes that are asymptotic to AdS at the boundary. Since black holes might be present this time coordinate is not necessarily globally well-deﬁned. This is analogous to the usual ﬂat space discussion of quantum gravity. See also [2. by taking N to be large we can make the curvature as small as we want. If we assume that the conjecture we made above is valid. even black holes in AdS. A theory which includes gravity is in some sense topological since one is integrating over all metrics and therefore the theory does not depend on the metric. The only particular feature is that they are moving on a curved geometry (anti-de Sitter space) which has a boundary at spatial inﬁnity. 24]. in a quantum gravity theory we do not have any local observables. we can extract the (ten .

36]. ten dimensional Minkowski space. 2 The correspondence In this section we will present an argument connecting type IIB string theory compactiﬁed on AdS5 ×S 5 to N = 4 super-Yang-Mills theory [9]. and their low-energy eﬀective Lagrangian is that of type IIB supergravity. 30] that quantum gravity theories should be holographic. It is not renormalizable (even for the ﬁelds on the brane). If we consider the system at low energies. Gauge Theory and Strings dimensional string theory) ﬂat space physics. The fact that the ﬁeld theory lives in a lower dimensional space blends in perfectly with some previous speculations about quantum gravity. and their low-energy eﬀective Lagrangian is that of N = 4 U (N ) super-Yang-Mills theory [13. It was suggested [29. 32–35]. plus some higher derivative corrections. in the sense that physics in some region can be described by a theory at the boundary with no more than one degree of freedom per Planck area. String theory on this background contains two kinds of perturbative excitations. The closed strings are the excitations of empty space and the open strings end on the D-branes and describe excitations of the D-branes.1) Sbulk is the action of ten dimensional supergravity. The reason for this bound is that otherwise black hole formation could violate the second law of thermodynamics. The complete eﬀective action of the massless modes will have the form S = Sbulk + Sbrane + Sint . energies lower than the string scale 1/ls .1) involves only the massless ﬁelds but it takes into account the eﬀects of integrating out the massive ﬁelds. We will see that the correspondence between ﬁeld theories and string theory on AdS space (including gravity) is a concrete realization of this holographic principle. a procedure which is in principle (but not in detail) similar to the one used in matrix theory [28]. The open string massless states give an N = 4 vector supermultiplet in (3 + 1) dimensions. (2. Consider N parallel D3 branes that are sitting together or very close to each other (the precise meaning of “very close” will be deﬁned below). and we can write an eﬀective Lagrangian describing their interactions. The closed string massless states give a gravity supermultiplet in ten dimensions. This “holographic” principle comes from thinking about the Bekenstein bound which states that the maximum amount of entropy in some region is given by the area of the region in Planck units [31]. then only the massless string states can be excited. Note that the Lagrangian (2. closed strings and open strings. and it . Other reviews of this subject are [1. Let us start with type IIB string theory in ﬂat. The D3 branes are extended along a (3 + 1) dimensional plane in (9 + 1) dimensional spacetime.330 Unity from Duality: Gravity.

Schematically we have Sbulk ∼ 1 2κ2 √ gR ∼ (∂h)2 + κ(∂h)2 h + · · · . Maldacena: Lectures on Large N Field Theories and Gravity 331 should only be understood as an eﬀective description in the Wilsonian sense. We can expand the bulk action as a free quadratic part describing the propagation of free massless modes (including the graviton). We indicate explicitly the dependence on the graviton. the interaction Lagrangian Sint is proportional to positive powers of κ. In order to see more clearly what happens in this low energy limit it is convenient to keep the energy ﬁxed and send ls → 0 (α → 0) keeping all the dimensionless parameters ﬁxed. If we take the low energy limit. And. i. Next. The leading terms in this interaction Lagrangian can be obtained by covariantizing the brane action. for example terms of the form α 2 Tr(F 4 ). introducing the background metric for the brane [37]. Dbranes are massive charged objects which act as a source for the various supergravity ﬁelds. of . On the one hand we have free gravity in the bulk and on the other hand we have the four dimensional gauge theory. We can ﬁnd a D3 brane solution [38] of supergravity. including the string coupling constant and N . all interaction terms proportional to κ drop out. Similarly. Finally. The brane action Sbrane is deﬁned on the (3 + 1) dimensional brane worldvolume. so that the interaction Lagrangian relating the bulk and the brane vanishes. This is the well known fact that gravity becomes free at long distances (low energies).2) where we have written the metric as g = η + κh. we consider the same system from a diﬀerent point of view. In addition all the higher derivative terms in the brane action vanish. the supergravity theory in the bulk becomes free. In this limit the coupling κ ∼ gs α 2 → 0. involving other ﬁelds. plus some interactions which are proportional to positive powers of the square root of the Newton constant. in this low energy limit we have two decoupled systems. So. which is known to be a conformal ﬁeld theory. leaving just the pure N = 4 U (N ) gauge theory in 3 + 1 dimensions. Sint describes the interactions between the brane modes and the bulk modes. we integrate out all massive degrees of freedom but we do not integrate out the massless ones. can be expanded in a similar way.e.J. but the other terms in the Lagrangian. and it contains the N = 4 super-Yang-Mills Lagrangian plus some higher derivative corrections. (2.

In both cases one of the decoupled systems is supergravity in ﬂat space. There are two kinds of low energy excitations (from the point of view of an observer at inﬁnity). we are led to the conjecture that N = 4 U (N ) super-Yang-Mills theory in . 1 2 3 5 F5 = (1 + ∗)dtdx1 dx2 dx3 df −1 . the low energy theory consists of two decoupled pieces. one is free bulk supergravity and the second is the near horizon region of the geometry. it is natural to identify the second system which appears in both descriptions. or we can have any kind of excitation that we bring closer and closer to r = 0.4) This means that the same object brought closer and closer to r = 0 would appear to have lower and lower energy for the observer at inﬁnity. r4 4 (2.3).332 the form Unity from Duality: Gravity. Now we take the low energy limit in the background described by equation (2. We can have massless particles propagating in the bulk region with wavelengths that becomes very large. the energy Ep of an object as measured by an observer at a constant position r and the energy E measured by an observer at inﬁnity are related by the redshift factor E = f −1/4 Ep . So. r R. In the near horizon region. 1 2 3 5 R2 r (2. Note that since gtt is non-constant. Similarly. In conclusion. where ω is the energy.5) which is the geometry of AdS5 × S 5 . We see that both from the point of view of a ﬁeld theory of open strings living on the brane. and from the point of view of the supergravity description. The bulk massless particles decouple from the near horizon region (around r = 0) because the low energy absorption cross section goes like σ ∼ ω 3 R8 [14. the excitations that live very close to r = 0 ﬁnd it harder and harder to climb the gravitational potential and escape to the asymptotic region. Gauge Theory and Strings ds2 = f −1/2 (−dt2 + dx2 + dx2 + dx2 ) + f 1/2 (dr2 + r2 dΩ2 ) . we can approximate f ∼ R4 /r4 . 15]. (2. and the geometry becomes ds2 = dr2 r2 (−dt2 + dx2 + dx2 + dx2 ) + R2 2 + R2 dΩ2 . This can be understood from the fact that in this limit the wavelength of the particle becomes much bigger than the typical gravitational size of the brane (which is of order R). In the low energy limit these two types of excitations decouple from each other.3) R4 ≡ 4πgs α 2 N . f =1+ R . we have two decoupled theories in the low-energy limit. Thus.

These modes living at the boundary are sometimes called singletons (or doubletons) [43–51].J. as is familiar in Chern-Simons theories [27. so there are no decoupled modes.4) reduces to E ∼ Ep r/ α . For small α (2. so that the metric becomes dU 2 U2 + dΩ2 . Therefore. which correspond to the U (1) degrees of freedom mentioned above. Depending on this choice we could have a correspondence to an SU (N ) or a U (N ) theory. Maldacena: Lectures on Large N Field Theories and Gravity 333 3 + 1 dimensions is the same as (or dual to) type IIB superstring theory on AdS5 × S 5 [9]. This corresponds to giving a vacuum expectation value to one of the scalars in the Yang-Mills theory. The U (1) vector supermultiplet includes six scalars which are related to the center of mass motion of all the branes [39]. Suppose that we take α → 0.6) This can also be seen by considering a D3 brane sitting at r. In fact we were not precise when we said that there were two sets of excitations at low energies. 42]. the excitations in the asymptotic ﬂat space and the excitations in the near horizon region. so this quantity should remain ﬁxed in the decoupling limit. as we did when we discussed the ﬁeld theory living on the brane. is proportional to U = r/α . . When we take the α → 0 limit we want to keep the mass of the “W -boson” ﬁxed. which is the mass of the string stretching between the branes sitting at r = 0 and the one at r. and it looks like we might or might not decide to include them in the AdS theory. There are also some zero modes which live in the region connecting the “throat” (the near horizon region) with the bulk. up to some ZN identiﬁcations (which aﬀect only global issues). so that we can consider arbitrary excited string states there. It is then convenient to deﬁne a new variable U ≡ r/α . the bulk AdS theory is describing the SU (N ) part of the gauge theory. We could be a bit more precise about the near horizon limit and how it is being taken. if one imposes local boundary conditions for these B-ﬁelds at the boundary of AdS one ﬁnds a U (1) gauge ﬁeld living at the boundary [41]. which is the way energies are measured in the ﬁeld theory. A U (N ) gauge theory is essentially equivalent to a free U (1) vector multiplet times an SU (N ) gauge theory. we need to take r → 0 keeping r/α ﬁxed. This implies that √ √ α Ep ∼ ﬁxed. Since we want to keep ﬁxed the energy measured from inﬁnity. This mass. We want to keep ﬁxed the energies of the objects in the throat (the near-horizon region) in string units. In the dual string theory all modes interact with gravity. The U (1) center of mass degree of freedom is related to the topological theory of B-ﬁelds on AdS [40]. From the AdS point of view these zero modes live at the boundary. (−dt2 + dx2 + dx2 + dx2 ) + 4πgs N ds2 = α √ 1 2 3 5 U2 4πgs N (2.

We have written the couplings in this fashion because both the gauge theory and the string theory have an SL(2.1 .7) From the physics of D-branes we know that the Yang-Mills coupling is related to the string coupling through [10. + + 2 gYM 2π gs 2π (2. In fact. Z) selfduality symmetry under which τ → (aτ + b)/(cτ + d) (where a. Thus. We also have some supersymmetries. Gauge Theory and Strings Anti-de-Sitter space has a large group of isometries. We will discuss in more detail the matching between the two sides of the correspondence in Section 3. This can be identiﬁed with the SU (4)R R-symmetry group of the ﬁeld theory. where R is the curvature of the four-dimensional space on which the theory is deﬁned.334 Unity from Duality: Gravity. This formulation of the conjecture is more useful since in the global coordinates there is no horizon. Since the isometries of AdS are in one to one correspondence with the generators of the conformal group of the ﬁeld theory. This is the same group as the conformal group in 3 + 1 dimensions. 52] τ≡ 4πi i θ χ = . the whole supergroup is the same for the N = 4 ﬁeld theory and the AdS5 × S 5 geometry. c. b. The number of supersymmetries is twice that of the full solution (2. lifting the moduli space. (2. we can conclude that this new Hamiltonian 1 (P0 + K0 ) can 2 be associated on AdS to the generator of translations in global time. so there is a term of the form d4 xTr(φ2 )R in the Lagrangian. since the superconformal algebra has twice as many fermionic generators as the corresponding Poincare superalgebra. but we could also think of the conformal ﬁeld theory as deﬁned on S 3 × R by redeﬁning the Hamiltonian. the fact that the low-energy ﬁeld theory on the brane is conformal is reﬂected in the fact that the near horizon geometry is Anti-de-Sitter space. d are . We also have an SO(6) symmetry which rotates the S 5 . In the above derivation the ﬁeld theory is naturally deﬁned on R3. S5 F5 = N. Due to the positive curvature of S 3 this leads to a mass term for the scalars [24].8) where we have also included the relationship of the θ angle to the expectation value of the RR scalar χ. which is SO(4. so both sides of the conjecture have the same spacetime symmetries. When we put the ﬁeld theory on S 3 the Coulomb branch is lifted and there is a unique ground state.3) containing the asymptotic region [39]. This doubling of supersymmetries is viewed in the ﬁeld theory as a consequence of superconformal invariance. The parameter N appears on the string theory side as the ﬂux of the ﬁve-form Ramond-Ramond ﬁeld strength on the S 5 . This is due to the fact that the scalars φI in the ﬁeld theory are conformally coupled. 2) for the case at hand.

In fact. In this fashion we avoid any obvious contradiction due to the fact that the two theories look very diﬀerent. (2.10) We see that the gravity regime (2. If gs > 1 we can perform an SL(2.9. without including stringy eﬀects. The string theory seems to have a parameter that does not appear in the gauge theory. The analysis of loop diagrams in the ﬁeld theory shows that we can trust the perturbative analysis in the Yang-Mills theory when 2 gYM N ∼ gs N ∼ R4 4 ls 1. 2. On the other hand. let us address the question of the validity of various approximations. α corrections to the gravity results give corrections √ which are proportional to powers of 1/ gs N . Notice that in (2. especially when one is doing gravity calculations. only the ratio of the radius of curvature to α is a parameter. to set the radius of curvature to one. will be independent of gs N and will depend only on N . Thus. as we can solve a strongly coupled gauge theory via classical supergravity. which sets the string tension and all other scales in the string theory. and it should also be a symmetry in the present context since all the ﬁelds that are being turned on in the AdS5 × S 5 background (the metric and the ﬁve form ﬁeld strength) are invariant under this symmetry. The connection between the SL(2. Maldacena: Lectures on Large N Field Theories and Gravity 335 integers with ad − bc = 1). With these conventions GN ∼ 1/N 2 and α ∼ 1/ gs N . (2. This can s be achieved by writing the metric as ds2 = R2 d˜2 .9) are perfectly incompatible. since only relative scales are meaningful. SL(2. Z) . However.10) and the perturbative ﬁeld theory regime (2. Z) is a conjectured strong-weak coupling duality symmetry of type IIB string theory in ﬂat space [53].9) 2 2 Note that we need gYM N to be small and not just gYM . the classical gravity description becomes reliable when the radius of curvature R of AdS and of S 5 becomes large compared to the string length. This is the reason that this correspondence is called a “duality”. this is not really a parameter in the theory if we do not compare it to other scales in the theory. In fact.J. This makes the correspondence both hard to prove and useful. but when one side is weakly coupled the other is strongly coupled and vice versa. α will disappear from any ﬁnal physical quantity we compute in this theory. but not α and the radius of curvature independently. ˜ This implies that any quantity calculated purely in terms of the gravity solution. namely α . Now. R4 2 ∼ gs N ∼ gYM N 4 ls 1. The two theories are conjectured to be exactly the same.10) we implicitly assumed that gs < 1. It is sometimes convenient. Z) duality symmetries of type IIB string theory and N = 4 SYM was noted in [54–56]. and rewriting everything √ in terms of g .

black holes. In the interior we can have all kinds of processes. but the gs corrections may not). but the full string theory on AdS might not agree with the ﬁeld theory.. Changing the coupling constant in the ﬁeld theory is related by (2.9. 2. So.1 The ﬁeld ↔ operator correspondence A conformal ﬁeld theory does not have asymptotic states or an S-matrix. so the natural objects to consider are operators.336 Unity from Duality: Gravity.) and gs very large. Another way to 4 see this is to note that the radius of curvature in Planck units is R4 /lp ∼ N . gravitons. as long as spacetime is asymptotically ﬂat. which is then related to the expectation value of the dilaton. More precisely.10) by taking N small (N = 1. So. otherwise we consider φ0 (x) to be the total coeﬃcient of O(x) in the Lagrangian). since in that case the D-string becomes light and renders the gravity approximation invalid. is that the two theories are exactly the same for all values of gs and N .10) but with gs → 1/gs . we cannot get into the gravity regime (2. in N = 4 superYang-Mills we have a deformation by a marginal operator which changes the value of the coupling constant.. It is not a proof because we did not treat the string theory non-perturbatively (not even non-perturbatively in α ). So. We can have black holes and all kinds of topology changing processes. D-branes. In this conjecture the spacetime is only required to be asymptotic to AdS5 ×S 5 as we approach the boundary. . let us denote by O the corresponding operator. etc. to have large N in order to have a weakly coupled supergravity description. changing the gauge theory coupling constant corresponds to changing the boundary value of the dilaton. but not suﬃcient. The Yang-Mills theory is supposed to eﬀectively sum over all spacetimes which are asymptotic to AdS5 × S 5 . 2. We could also consider diﬀerent forms of the conjecture. it is always necessary. The strong form of the conjecture. Even the topology of spacetime can change in the interior. We can consider adding the term d4 xφ0 (x)O(x) to the Lagrangian (for simplicity we assume that such a term was not present in the original Lagrangian. 2. which is the most interesting one and which we will assume here.8) to changing the coupling constant in the string theory. The expectation value of the dilaton is set by the boundary condition for the dilaton at inﬁnity. highly excited fundamental string states. For example. One might wonder why the above argument was not a proof rather than a conjecture. In this case asymptotic ﬂatness is replaced by the asymptotic AdS behavior. In its weakest form the gravity description would be valid for large gs N . According to the . Gauge Theory and Strings duality transformation and get conditions similar to (2. but only in the N → ∞ limit (so that the α corrections would agree with the ﬁeld theory. A not so weak form would say that the conjecture is valid even for ﬁnite gs N . This is completely analogous to the usual conditions of asymptotic ﬂatness.

11) where the left hand side is the generating function of correlation functions in the ﬁeld theory. Therefore. There is a relation between the mass of the ﬁeld φ and the scaling dimension of the operator in the conformal ﬁeld theory. through the left hand side of (2. (2. it is natural to assume that this will change the boundary condition of the dilaton at the boundary of AdS to φ(x.J. z) z=0 = φ0 (x) . . Let us describe this more generally in AdSd+1 . it is natural to propose that Ê e d4 xφ0 (x)O(x) CFT = Zstring φ(x. in order to get consistent behavior for a massive ﬁeld.11) should in general be changed to φ(x. where ∆= d + 2 d2 + R 2 m2 . A formula like (2. as argued in [23. ) = d−∆ φ0 (x). including fermions and tensors on AdS space. Since φ is dimensionless. Notice that φ0 is a function of the four variables parametrizing the boundary of AdS5 . where we will verify that the two-point correlation function of the operator O behaves as that of an operator of dimension ∆ [23. A more detailed derivation of this relation will be given in Section 4. φ0 is an arbitrary function and we can calculate correlation functions of O by taking functional derivatives with respect to φ0 and then setting φ0 = 0. Maldacena: Lectures on Large N Field Theories and Gravity 337 discussion above.11) is valid in general. A similar relation between ﬁelds on AdS and operators in the ﬁeld theory exists also for non-scalar ﬁelds.13) and eventually we would take the limit where → 0.12) Therefore.11). in the coordinate system 2 ds2 = RAdS −dt2 + dx2 + · · · + dx3 + dz 2 1 3 · z2 More precisely.24]. each ﬁeld propagating on AdS space is in a one to one correspondence with an operator in the ﬁeld theory. The wave equation in Euclidean space for a ﬁeld of mass m has two independent solutions. that the associated operator O has dimension ∆ (2. 4 (2. z)|z=0 = φ0 (x). we see that φ0 has dimensions of [length]∆−d which implies.e. 24]. (2. The right hand side is the full partition function of string theory with the boundary condition that the ﬁeld φ has the value φ0 on the boundary of AdS.12). the boundary condition on the ﬁeld in the right hand side of (2. which behave like z d−∆ and z ∆ for small z (close to the boundary of AdS). i. for any ﬁeld φ.

Since the correspondence described above matches the stress-energy tensor with the graviton. Feynman diagrams can be used to compute the interactions of particles in the bulk.2 Holography In this section we will describe how the AdS/CFT correspondence gives a holographic description of physics in AdS spaces. 2. the only diagrams that contribute are the tree-level diagrams of the gravity theory (see for instance Fig. String theory provides such a theory. Fig. Gauge Theory and Strings Correlation functions in the gauge theory can be computed from (2. This method of deﬁning the correlation functions of a ﬁeld theory which is dual to a gravity theory in the bulk of AdS space is quite general. this discussion can be extended to any AdSd+1 space. Any local ﬁeld theory contains the stress tensor as an operator. At tree level. and it applies in principle to any theory of gravity [24]. this implies that the AdS theory includes gravity.11) by diﬀerentiating with respect to φ0 . these diagrams and those related to them by crossing are the only ones that contribute to the four-point function. 1. 1).338 Unity from Duality: Gravity. Correlation functions can be calculated (in the large gs N limit) in terms of supergravity Feynman diagrams. Similarly. In particular. But if a new way of deﬁning quantum gravity theories comes along we could consider those gravity theories in AdS. corresponding to a conformal ﬁeld theory in d spacetime dimensions (for d > 1). Depending on the choice of M 5 we get diﬀerent dual conformal ﬁeld theories. we could consider backgrounds of string theory of the form AdS5 × M 5 where M 5 is any Einstein manifold [63–65]. In the limit where classical supergravity is applicable. Here we see the leading contribution coming from a disconnected diagram plus connected pieces involving interactions of the supergravity ﬁelds in the bulk of AdS. and they should correspond to some conformal ﬁeld theory “on the boundary”. It should be a well deﬁned quantum theory of gravity since we should be able to compute loop diagrams. . Each diﬀerentiation brings down an insertion O. which sends a φ particle (a closed string state) into the bulk.

it is hard to check what the number of degrees of freedom per Planck area is. which states that the maximum entropy in a region of space is Smax = Area/4GN [31]. since we approach the boundary when we do a conformal transformation that localizes objects in the CFT. So. It is clear by studying the action of the conformal group on Poincar´ coordinates that e the radial position plays the role of some energy scale. We saw above that when we calculate correlation functions we have to specify boundary conditions at r = 1 − δ and then take the limit of δ → 0. For this purpose let us write the metric of AdS as ds2 = R2 − 1 + r2 1 − r2 2 dt2 + 4 (dr2 + r2 dΩ2 ) . Suppose that we had a state with more entropy than Smax . (2. Note that this bound implies that the number of degrees of freedom inside some region grows as the area of the boundary of a region and not like the volume of the region. So. the limit δ → 0 corresponds to going to the UV of the ﬁeld theory. Maldacena: Lectures on Large N Field Theories and Gravity 339 Let us start by explaining the Bekenstein bound. has an inﬁnite number of degrees of freedom.15) ds2 = R2 z2 in which the boundary is at z = 0. in order to compare things properly we should introduce a cutoﬀ on the number of degrees of freedom in the ﬁeld theory and see what it corresponds to in the gravity theory. so it looks like holography. In the AdS/CFT correspondence we are describing physics in the bulk of AdS space by a ﬁeld theory of one less dimension (which can be thought of as living on the boundary). and the area of the boundary of AdS space is also inﬁnite. Thus.30] (so that its entropy satisﬁes the Bekenstein bound).J. then we show that we could violate the second law of thermodynamics. However. since the theory. where the area is that of the boundary of the region. The entropy should not decrease. which states that in a quantum gravity theory all physics within some volume can be described in terms of some theory on the boundary which has less than one degree of freedom per Planck area [29. Attempting to understand this behavior leads to the “holographic principle”. If we consider a particle or wave propagating in (2. In standard quantum ﬁeld theories this is certainly not possible. We can throw in some extra matter such that we form a black hole. which is smaller than the area of the boundary of the region (which by our assumption is smaller than the initial entropy). But if a black hole forms inside the region its entropy is just the area of its horizon. (1 − r2 )2 (2.14) In these coordinates the boundary of AdS is at r = 1. When we are close to the boundary we could also use the Poincar´ e coordinates −dt2 + dx2 + dz 2 .15) or (2.14) we see that its motion is independent of R in . being conformal. the second law has been violated.

remain constant. for δ 1. This implies that a cutoﬀ at z∼δ (2. One might get confused by the fact that the surface r = 1 − δ is really nine dimensional as opposed to four dimensional. Consider the case of N = 4 SYM on a three-sphere of radius one. (2. One could be a little suspicious of the statement that gravity in AdS is holographic. if we are in Euclidean space and we have a wave that has some spatial extent λ in the x directions. with no factors of R (δ here is dimensionless. is V 5 R3 δ −3 Area = S ∼ N 2 δ −3 . This can be seen from (2.18) Thus. We get S ∼ N 2 δ −3 . What makes this case diﬀerent is that we have the additional parameter R. So. it will also have an extent λ in the z direction. we see that the theory on this nine dimensional surface becomes eﬀectively four dimensional. any ﬁeld theory in AdS would be holographic in the sense that the number of degrees of freedom within some (large enough) volume is proportional to the area (and also to the volume). Equation (2. in the ﬁeld theory it is measured in terms of the radius of the S 4 or S 3 that the theory lives on). while the other ﬁve. From the form of the full metric on AdS5 × S 5 we see that as we take δ → 0 the physical size of four of the dimensions of this nine dimensional space grow. z → λz. and then we can take AdS spaces of diﬀerent radii (corresponding to diﬀerent values of N in the SYM theory). We can estimate the number of degrees of freedom in the ﬁeld theory with a UV cutoﬀ δ. since we need to multiply the metric by a factor that goes to zero as we approach the boundary in order to deﬁne a ﬁnite metric for the four dimensional gauge theory.14) the area in Planck units of the surface at r = 1 − δ. 4GN 4GN (2. In fact. Furthermore. we see that the AdS/CFT correspondence saturates the holographic bound [66]. .17) since the number of cells into which we divide the three-sphere is of order 1/δ 3 .16) is called the UV-IR relation [66].16) corresponds to a UV cutoﬀ in the ﬁeld theory at distances δ. Gauge Theory and Strings the supergravity approximation. In the gravity solution (2.15) by eliminating λ through the change of variables x → λx. since it does not seem to be saying much because in AdS space the volume and the boundary area of a given region scale in the same fashion as we increase the size of the region. since these have a diﬀerent dependence on R. the S 5 . and then we can ask whether the number of degrees of freedom goes like the volume or the area.340 Unity from Duality: Gravity.

the AdS/CFT correspondence is a strong/weak coupling duality. it relates 2 the region of weak ﬁeld theory coupling λ = gYM N in the SYM theory to the region of high curvature (in string units) in the string theory. since (with our current knowledge) we can only compute most of √ them perturbatively in λ on the ﬁeld theory side and perturbatively in 1/ λ on the string theory side. and it was realized that there are several properties of these theories which do not depend on the coupling. which we do not know how to compare since we can only compute them perturbatively on both sides. . and vice versa. These are: • the global symmetries of the theory. most of the tests described here can be generalized also to cases in other dimensions and/or with less supersymmetry. as described below. 67]1. both theories are believed to have a non-perturbative SL(2. Additional ZN symmetries arise when the theories are compactiﬁed on non-simply-connected manifolds.J. Here we will review the evidence showing that this ﬁeld theory is indeed the same as the conjectured dual ﬁeld theory. A similar situation arises also in many ﬁeld theory dualities that were analyzed in the last few years (such as the electric/magnetic SL(2. this test actually tests the ﬁnite N duality and not just the large N limit. For example. 1 Unlike most of the other tests described here. we can compute the equation of state of the SYM theory and also the quark-antiquark potential both for small λ and for large λ. in the case of the AdS/CFT correspondence we have the same supergroup SU (2. which will be described below. In the ’t Hooft large N limit. We will focus here only on tests of the correspondence between the N = 4 SU (N ) SYM theory and the type IIB string theory compactiﬁed on AdS5 × S 5 . 2) × SU (4)) as the global symmetry of both theories. Also. so they can be compared to test the duality. 2|4) (whose bosonic subgroup is SO(4. a direct comparison of correlation functions is generally not possible. Maldacena: Lectures on Large N Field Theories and Gravity 3 Tests of the AdS/CFT correspondence 341 In this section we review the direct tests of the AdS/CFT correspondence. Z) duality symmetry acting on their coupling constant τ . Z) duality of the N = 4 SYM theory itself). In Section 2 we saw how string theory on AdS deﬁnes a partition function which can be used to deﬁne a ﬁeld theory. and we obtain diﬀerent answers. which cannot change as we change the coupling (except for extreme values of the coupling). As discussed in Section 2. Thus. As described in Section 2. These are the only symmetries of the theory on R4 . and these were also successfully matched in [40.

QβB } = 0. where α is a Weyl-spinor index (in the 2 of the SO(3. are protected from any quantum corrections and do not depend on λ.342 Unity from Duality: Gravity. and it will be compared in Section 3. There are however multicenter solutions so that one might think that there is a background of string theory corresponding to any point in the ﬁeld theory moduli space. We will also not discuss here tests involving the behavior of the theory on its moduli space [60. parametrized by the eigenvalues of six commuting traceless N × N matrices. They obey the algebra A ¯˙ {QA . On the AdS side it is not clear exactly how to deﬁne the moduli space. 69].2 below. • the qualitative behavior of the theory upon deformations by relevant or marginal operators also does not depend on the coupling (at least for chiral operators whose dimension does not depend on the coupling. QαB } = 2(σ µ )αα Pµ δB . . • the spectrum of chiral operators does not change as the coupling varies. α β (3. 3. Gauge Theory and Strings • some correlation functions.1 below. 3) are the Pauli matrices and (σ 0 )αα = −δαα (we use the ˙ ˙ conventions of Wess and Bagger [70]). and in the absence of phase transitions). In the SU (N ) ﬁeld theory the moduli space is R6(N −1) /SN . but it is not clear how to see that this is the exact moduli space on the string theory side (especially since high curvatures arise for generic points in the moduli space). which we will not discuss in this section. • the moduli space of the theory also does not depend on the coupling.1 The ﬁeld theory spectrum The N = 4 supersymmetry algebra in d = 4 has four generators QA (and α ¯˙ their complex conjugates QαA ).1. which are usually related to anomalies. such as the existence of conﬁnement for the ﬁnite temperature theory [68]. There are many more qualitative tests of the correspondence. ˙ α ¯˙ ¯˙ {QA . The matching of these correlation functions will be described in Section 3. 1) Lorentz group) and A is an index in the 4 of the SU (4)R R-symmetry group (lower indices A will be taken to transform in the ¯ 4 representation). 2. QB } = {QαA . 61.1) where σ i (i = 1.1 The spectrum of chiral primary operators 3.

or to an N = 1 theory with three chiral multiplets Φi in the adjoint representation (in the 32/3 of the SU (3) × U (1)R ⊂ SU (4)R which is left unbroken by the choice of a single N = 1 SUSY generator) and a superpotential of the form W ∝ ijk Tr(Φi Φj Φk ). In the ’t Hooft large N limit correlation functions involving multiple-trace operators are suppressed by powers of N compared to those of single-trace operators involving the same ﬁelds. The interactions of the theory include a scalar potential proportional to I. α {QA . Such a ﬁeld theory is equivalent to an N = 2 theory with one hypermultiplet in the adjoint representation. In this section we will focus on local operators which involve ﬁelds taken at the same point in space-time. φJ ]2 ).J. Maldacena: Lectures on Large N Field Theories and Gravity 343 N = 4 supersymmetry in four dimensions has a unique multiplet which does not include spins greater than one. properties of the adjoint representation of SU (N ) determine that such operators necessarily involve a product of traces of products of ﬁelds (or the sum of such products). the multiple-trace operators appear in operator product expansions of products of single-trace operators. ˙ α β ¯˙ [QA . and its ﬁeld content is a vector multiplet in the adjoint of the gauge group. We will discuss here in detail only the single-trace operators. λβB } ∼ (σ µν )αβ Fµν + α ¯˙ {QA . Dµ is the covariant derivative. Dν ].2). 4. For the SU (N ) theory described above. 4→ An N = 4 supersymmetric ﬁeld theory is uniquely determined by specifying the gauge group. It includes a vector ﬁeld Aµ (µ is a vector index of the SO(3. 1) Lorentz group). The classical action of the supersymmetry generators on these ﬁelds is schematically given (for onshell ﬁelds) by [QA . which is the vector multiplet. 4 four complex Weyl fermions λαA (in the ¯ of SU (4)R ).2) . λB } ∼ (σ µ )αβ Dµ φI . It is natural to divide the operators into single-trace operators and multiple-trace operators. and six real scalars φI (where I is an index in the 6 of SU (4)R ). ¯ with similar expressions for the action of the Q’s. where σ µν are the generators of the Lorentz group in the spinor representation. The spectrum of operators in this theory includes all the gauge invariant quantities that can be formed from the ﬁelds described above.J Tr([φI . . Aµ ] ∼ (σµ )αα λA ˙ β α αβ ˙ ˙ I J αβ [φ . and we have suppressed the SU (4) Clebsch-Gordan coeﬃcients corresponding to the products 4 × 6 → ¯ 4× ¯ 1+15 and 4×4 → 6 in the ﬁrst three lines of (3. φ ]. 6). such that the moduli space of the theory is the space of commuting matrices φI (I = 1. the ﬁeld strength Fµν ≡ [Dµ . (3. · · · . φI ] ∼ λαB .

whose range of helicities is 6. Representations of the superconformal algebra are formed by starting with some state of lowest dimension. . and the range of helicities appearing is smaller. which are in short representations of the superconformal algebra and are annihilated by some of the supercharges. These operators form an ultrashort representation of the N = 4 algebra whose range of helicities is from (−1) to 1 (acting with more than two supercharges of the same helicity on any of these states gives either zero or derivatives.e. generated by acting on the lowest state with products of diﬀerent supercharges. for instance. A generic primary representation of the superconformal algebra will thus include 216 primaries of the conformal algebra. This representation is usually called the doubleton representation. and it does not appear in the SU (N ) SYM theory (though the representations which do appear can all be formed by tensor products of the doubleton representation). In particular. In a superconformal theory it is also natural to distinguish between chiral primary operators. on the moduli space of the SU (N ) N = 4 SYM theory [71–78]). in the small representations of the N = 4 superconformal algebra only up to 4 Q’s of the same helicity acting on the lowest dimension operator give a non-zero result. In the U (1) N = 4 SYM theory (which is a free theory). and acting on it with the operators Q and Pµ . For the N = 4 supersymmetry algebra (not including the conformal algebra) it is known that medium representations. it is not clear if such medium representations of the superconformal algebra [79] can appear in physical theories or not (there are no known examples). derivatives). In non-generic representations of the superconformal algebra a product of less than 16 diﬀerent Q’s annihilates the lowest dimension operator. can also exist (they arise. λA and Fµν = ∂[µ Aν] ). All other local gauge invariant operators in the theory involve derivatives or products of these operators. and the range of helicities is between λ − 2 and λ + 2. and non-chiral primary operators. acting with additional supercharges always leads to descendants of the conformal algebra (i. which is annihilated by the operators S and Kµ . Since the supercharges have helicities ±1/2. More details on the structure of representations of the N = 4 superconformal algebra may be found in [79–85] and references therein. which are descendants of the conformal algebra). Gauge Theory and Strings It is natural to classify the operators in a conformal theory into primary operators and their descendants. the only gaugeinvariant “single trace” operators are the ﬁelds of the vector multiplet itself ¯ (which are φI . The N = 4 supersymmetry algebra involves 16 real supercharges.344 Unity from Duality: Gravity. λA . the primary ﬁelds in such representations will have a range of helicities between λ − 4 (if the lowest dimension operator ψ has helicity λ) and λ + 4 (acting with more than 8 supercharges of the same helicity either annihilates the state or leads to a conformal descendant).

In the SUSY algebra (3. Another way to check this is to see if by acting with Q’s on these operators we get the most general possible states or not. ψ} for any ﬁeld ψ. In this way we ﬁnd that the representation is chiral primary if and only if the indices form a symmetric traceless product of n 6’s (traceless representations are deﬁned as those who give zero when any two indices are contracted). There is no known simple systematic way to compute the full spectrum of chiral primary operators of the N = 4 SU (N ) SYM theory. This is a representation of weight (0. There are several diﬀerent ways to answer this question. so we will settle for presenting the known chiral primary operators. The ﬁeld in the 1 representation is Tr(φI φI ). 0) of SU (4)R . Looking at the action of the supersymmetry charges (3. Next. The right-hand side is in the 4 representation. For instance.2) only commutators of φI ’s appear on the right hand side. and to check if this relation is obeyed in the free ﬁeld theory. Maldacena: Lectures on Large N Field Theories and Gravity 345 In the context of AdS space one can think of this multiplet as living purely on the boundary of the space [46. as expected for the U (1) part of the original U (N ) gauge group of the D3-branes (see the discussion in Sect. in this section we will refer to SU (4)R representations either by their dimensions in boldface or by their weights. First we can ask if this operator can be written as {Q. The lowest component of a superconformal-primary multiplet is characterized by the fact that it cannot be written as a supercharge Q acting on any other operator. Let us analyze operators of the form OI1 I2 ···In ≡ Tr(φI1 φI2 · · · φIn ). though in principle there could be representations where “null vectors” appear only at higher levels). namely if the representation contains “null vectors” or not (it turns out that in all the relevant cases “null vectors” appear already at the ﬁrst level by acting with a single Q. Thus. which is 4 . so we see that if some of the indices are antisymmetric the ﬁeld will be a descendant.J. and this turns out to be correct. Using the SUSY algebra (3. One possibility is to use the relation between the dimension of chiral primary operators and their R-symmetry representation [96–100]. 2). Tr(φI φI )] ∼ C AJB Tr(λαB φJ ) where C AIB is a Clebsch-Gordan coeﬃα cient for ¯ × 6 → 4. so one would expect the lowest components of the chiral primary representations to be built only from the scalar ﬁelds. only symmetric combinations of the indices will be lowest components of primary multiplets. we should ask if the multiplet built on such an operator is a (short) chiral primary multiplet or not.2) it is easy to see that for symmetric traceless representations we get “null vectors” while for other representations we do not.2) suggests that generally operators built from the fermions and the gauge ﬁelds will be descendants (given by Q acting on some other ﬁelds). n. where [OI1 I2 ···In ] = n.86–95]. for which [QA . let us analyze in detail the case n = 2. The symmetric product of two 6’s is given by 6 × 6 → 1 + 20 .

as described above. Tr(φ{I φJ} )} ∼ Tr(λαB φK ) with the rightα hand side being in the 20 representation (appearing in ¯ × 6 → 4 + 20).346 Unity from Duality: Gravity. 4 while the left-hand side could in principle be in the 4 × 20 → 20 + 60. as discussed above. we ﬁnd that the short chiral primary representations are built on the operators On = O{I1 I2 ···In } with n = 2. commutators always lead to descendants. but we will not attempt to rigorously prove this here. The product of more than N commuting2 N × N matrices can always be written as a sum of products of traces of less than N of the matrices. It is possible to analyze the chiral primary spectrum also by using N = 1 subalgebras of the N = 4 algebra. The last issue we should discuss is what is the range of values of n. but for a particular choice of the N = 1 subalgebra not all the operators On appear to be chiral (a short multiplet of the N = 4 algebra includes both short and long multiplets of the N = 1 subalgebra). 84. so it does not form an independent operator. for which the indices are in the symmetric traceless product of n 6’s (in a U (N ) theory we would ﬁnd the same spectrum with the additional representation corresponding to n = 1). we ﬁnd that {QA . The 1 representation built on On contains a total of 256 × 12 n2 (n2 − 1) primary states. the operators On include the chiral operators of the form Tr(Φi1 Φi2 · · · Φin ) (in a representation of SU (3) which is a symmetric product of 3’s). Since the 60 does not appear on the right-hand side (it is a “null vector”) we identify that the representation built on the 20 is a short representation of the SUSY algebra. which is the same as their value in the free ﬁeld theory. up to operators including commutators which (as explained above) are descendants of the SUSY algebra. If we use an N = 1 subalgebra of the N = 4 algebra. 3. and we can write any product of matrices as a product of commuting matrices plus terms with commutators. The full chiral primary representations are obtained by acting on the ﬁelds On by the generators Q and P of the supersymmetry algebra. Gauge Theory and Strings the most general representation that can appear in the product 4 × 1. of which half are bosonic and half are fermionic. This means that for n > N we can express the operator OI1 I2 ···In in terms of other operators. . · · · . 101] for more details) one can verify that chiral primary representations correspond exactly to symmetric traceless products of 6’s. N . We argued above that these are the only short chiral primary representations in the SU (N ) gauge theory. if we look at the symmetric traceless product Tr(φ{I φJ} ) ≡ Tr(φI φJ )− 1 δ IJ Tr(φK φK ) in the 20 6 representation. so we ﬁnd no null vectors at this level. On the other hand. Thus. By similar manipulations (see [24. 81. Since these 2 We can limit the discussion to commuting matrices since. The superconformal algebra determines the dimension of these ﬁelds to be [On ] = n.

φK2 ]φL1 · · · φLn−1 ). 0) representation. 2) representation. n − 2.2). ¯ • a real scalar ﬁeld in the (2. n − 1. of the form ˙ ˙ . 0) representation of SU (4)R . In the ﬁrst case we get a Lorentz scalar ﬁeld in the (2. To get a ﬁeld of dimension n + 1 we need to act on On with two supercharges (recall that [Q] = 1 ). On ]} ∼ Tr((σ µν )αβ Fµν φJ1 · · · φJn−1 ) + Tr(λαA λβB φK1 · · · φKn−2 ). ˙ α ˙ At dimension n + 2 (acting with four supercharges) we ﬁnd: • a complex scalar ﬁeld in the (0. [Qβ . n − 2. For example. with the complex conjugate ﬁelds given by ¯ ¯ the action of two Q’s. 1) representation of SU (4)R . 101]. n − 4. 2 of the form Tr(Fµν φI1 · · · φIn−2 ) + · · · . of the form {Q{α . 0) representation of SU (4)R . If we 2 act with two supercharges QA of the same chirality. [Qβ} . in the (0. n−2.J. they will contain primary ﬁelds of helicities between (−2) and 2. The highest dimension primary ﬁeld ¯ in the multiplet is (generically) of the form Q4 Q4 On . see also [102] which checks some components of these superﬁelds against the couplings to supergravity modes predicted on the basis of the DBI action for D3-branes in anti-de Sitter space [4]. given by Q2 Q2 On . It is easy to ﬁnd the form of all the ﬁelds in such a multiplet by using the algebra (3. In the second case we get an anti-symmetric 2-form of the Lorentz group. On ]} ∼ ¯ Tr(λαA λB φJ1 · · · φJn−2 ) + (σ µ )αα Tr((Dµ φJ )φK1 · · · φKn−1 ). and its dimension is n+4. B1 ¯ B2 I1 αβ αβ ˙ ˙ In−4 ¯ Tr(λαA1 λβA2 λα λβ φ · · · φ ) + ···. On ]} ∼ αβ Tr(λαA λβB φJ1 · · · φJn−2 ) + Tr([φK1 . Maldacena: Lectures on Large N Field Theories and Gravity 347 multiplets are built on a ﬁeld of helicity zero. their Lorentz indices α can be either antisymmetrized or symmetrized. Both of these ﬁelds are complex. given by Q4 On . There is an elegant way to write these multiplets as traces of products of “twisted chiral N = 4 superﬁelds” [81. Using an N = 1 subalgebra some of these operators may be written as 2 the lowest components of the chiral superﬁelds Tr(Wα Φj1 · · · Φjn−2 ). of the form ¯˙ {Qα . let us analyze here in detail the bosonic primary ﬁelds of dimension n + 1 in the multiplet. [Qα . which is of the schematic form αβ {Qα . Acting with one Q and one Q on the state On gives a (real) Lorentz-vector ﬁeld in the (1.

φJ2 ]φI1 · · · φIn−2 )+· · · . and a symmetric tensor ﬁeld of dimension 4 which is the energy-momentum tensor (the other currents of the superconformal algebra are descendants of these).1. ﬁelds on AdS5 are in a one-to-one correspondence with operators in the dual conformal ﬁeld theory.1. the second Casimir of these representations is C2 = ∆(∆ − 4) for a primary scalar ﬁeld of dimension ∆ in the ﬁeld theory. 3. of the form Tr(Fµν Dν φJ φI1 · · · φIn−2 ) + · · · . The n = 2 multiplet also includes a complex scalar ﬁeld which is an SU (4)R -singlet. For n = 2.2 The string theory spectrum and the matching As discussed in Section 2.1 should agree with the spectrum of ﬁelds of type IIB string theory on AdS5 × S 5 . the n = 3 multiplet (which contains a scalar in the 50 of dimension 3 and a complex scalar in the 45 of dimension 4). given by ¯ Q3 QOn . n − 4. of the form Tr(Fµν [φJ1 . n − 4. 0) representa¯ tion.348 Unity from Duality: Gravity. n − 2. Note that ﬁelds with more than four Fµν ’s or Tr(Fµν φ · · · φ more than eight λ’s are always descendants or non-chiral primaries. given by ¯ Q2 Q2 On . Single-trace . the spectrum of operators described in Section 3. whose real 2 part is the Lagrangian density coupling to 4g1 (of the form Tr(Fµν ) + · · · ) 2 YM and whose imaginary part is the Lagrangian density coupling to θ (of the form Tr(F ∧ F )). In particular. 0) representation. a complex scalar in the 10 of dimension 3. and C2 = m2 R2 for a ﬁeld of mass m on an AdS5 space with a radius of curvature R. • an complex anti-symmetric 2-form ﬁeld in the (2. and at dimension n + 4 there is a single primary ﬁeld. 1) representation. and the n = 4 multiplet which contains a scalar in the 105 of dimension 4. For later use we note that the chiral primary multiplets which contain scalars of dimension ∆ ≤ 4 are the n = 2 multiplet (which has a scalar in the 20 of dimension 2. The n = 2 representation includes the currents of the superconformal algebra. of the form Tr(D{µ φJ Dν} φK φI1 · · · φIn−2 ) + · · · The spectrum of primary ﬁelds at dimension n + 3 is similar to that of dimension n + 1 (the same ﬁelds appear but in smaller SU (4)R representations). and a complex scalar in the 1 of dimension 4). which is a ¯ real scalar in the (0. It includes a vector of dimension 3 in the 15 representation which is the SU (4)R R-symmetry current. of the form 4 I1 In−4 ) + · · · . Fields on AdS naturally lie in the same multiplets of the conformal group as primary operators. given by Q2 Q2 On . given by Q4 Q4 On . Thus.1. • a symmetric tensor ﬁeld in the (0. Gauge Theory and Strings • a complex vector ﬁeld in the (1. 3 the short multiplets are even shorter since some of the representations appearing above vanish. 0) representation. n − 3. for n = 2 the highestdimension primaries in the chiral primary multiplet have dimension n + 2 = 4.

2. These spherical harmonics are in representations corresponding to symmetric traceless products of 6’s of SU (4)R . while multiple-trace operators correspond to multi-particle states. it is not known how to compute the full spectrum of type IIB string theory on AdS5 ×S 5 . A similar k expansion may be performed for all other ﬁelds. and the equations of motion determine the mass of this ﬁeld to be m2 = k(k + 4)/R2 . Thus. they may be written as Y k (y) ∼ y I1 y I2 · · · y Ik where the y I . which are built on a lowest dimension ﬁeld which is a scalar in the (0. The computation involves expanding the ten dimensional ﬁelds in appropriate spherical harmonics on S 5 . For example. . y) = k=0 τ k (x)Y k (y) where x is a coordinate on AdS5 . for I = 1. and diagonalizing the equations to give equations of motion for free (massless or massive) ﬁelds3 . we ﬁnd a ﬁeld τ (x) on AdS5 in each such (0. Unfortunately. 3. Maldacena: Lectures on Large N Field Theories and Gravity 349 operators in the ﬁeld theory may be identiﬁed with single-particle states in AdS5 . String theory on AdS5 × S 5 is expected to have many additional states. · · · . In fact. plugging them into the supergravity equations of motion. Presumably none of these states are in small multiplets of the superconformal algebra (at least. b. The spectrum of type IIB supergravity compactiﬁed on AdS5 × S 5 was computed in [103]. If we organize the results of [103] into representations of the superconformal algebra [80]. where a. we ﬁnd representations of the form described in the previous section. with masses of the order of the string scale 1/ls or of the Planck scale 1/lp . The lowest dimension scalar ﬁeld in each representation turns out to arise from a linear combination of spherical harmonic modes of the S 5 components of the graviton ha (expanded around the AdS5 × S 5 vacuum) and the 4-form a ﬁeld Dabcd . The scalar ﬁelds of dimension 3 The ﬁelds arising from diﬀerent spherical harmonics are related by a “spectrum generating algebra”. n. 6 and with 6 I 2 5 k I=1 (y ) = 1. k. gs N . Such states would correspond (using the mass/dimension relation described above) to operators in the ﬁeld theory with dimensions of order ∆ ∼ (gs N )1/4 or ∆ ∼ N 1/4 for large N. 0) representation of SU (4)R . so it is clear that they all lie in small multiplets of the superconformal algebra. ∞. d are indices on S 5 . linearized around the AdS5 ×S 5 background. 0) representation of SU (4)R for n = 2. see [104]. These ﬁelds all have helicities between (−2) and 2. this would be the prediction of the AdS/CFT correspondence). the ten dimensional dilaton ∞ ﬁeld τ may be expanded as τ (x. the only known states are the states which arise from the dimensional reduction of the ten-dimensional type IIB supergravity multiplet. and we will describe below how they match with the small multiplets of the ﬁeld theory described above. y is a coordinate on S 5 . are coordinates on S .J. · · · . c. and the Y k are the scalar spherical harmonics on S 5 .

also appears.1. so the results agree within their range of validity. the n = 2 representation is called the supergraviton representation. In particular. and it includes the ﬁeld content of d = 5. In the naive dimensional reduction of the type IIB supergravity ﬁelds. the n = 1 doubleton representation. The ﬁeld/operator correspondence matches this representation to the representation including the superconformal currents in the ﬁeld theory. and that there are no chiral ﬁelds above this scale. Comparing the results described above with the results of Section 3. 3.1. It is not known how to check this prediction. Gauge Theory and Strings n + 1 correspond to 2-form ﬁelds Bab with indices in the S 5 . The ﬁeld theory results suggest that the exact spectrum of chiral representations in type IIB string theory on AdS5 × S 5 actually matches the naive supergravity spectrum up to a mass scale m2 ∼ N 2 /R2 ∼ 2 N 3/2 Mp which is much higher than the string scale and the Planck scale. The supergravity results cannot be trusted for masses above the order of the string scale (which corresponds to n ∼ (gs N )1/4 ) or the Planck scale (which corresponds to n ∼ N 1/4 ). tree-level string theory is certainly not enough for this since when gs = 0 we must take N = ∞ to obtain a ﬁnite value of gs N . The symmetric tensor ﬁelds arise from the expansion of the AdS5 -components of the graviton. This is one of the reasons why it seems more natural to view the corresponding gauge theory as an SU (N ) gauge theory and not a U (N ) theory. Thus. the modes of this multiplet are all pure gauge modes in the bulk of AdS5 . and then classical string theory should lead to exactly the same . and they may be set to zero there. corresponding to a free U (1) vector multiplet in the dual theory. with our current knowledge the matching of chiral primaries of the N = 4 SYM theory with those of string theory on AdS5 × S 5 tests the duality only in the large N limit. · · · . which (as expected) corresponds to the energy-momentum tensor in the ﬁeld theory. but this will not aﬀect most of our discussion in this review so we will ignore this possibility here. It may be possible (and perhaps even natural) to add the doubleton representation to the theory (even though it does not include modes which propagate in the bulk of AdS5 .350 Unity from Duality: Gravity. The dilaton ﬁelds described above are the complex scalar ﬁelds arising with dimension n + 2 in the multiplet (as described in the previous subsection). It includes a massless graviton ﬁeld. N = 8 gauged supergravity. we see that we ﬁnd the same spectrum of chiral primary operators for n = 2. but instead it is equivalent to a topological theory in the bulk) to obtain a theory which is dual to the U (N ) gauge theory. and massless SU (4)R gauge ﬁelds which correspond to (or couple to) the global SU (4)R currents in the ﬁeld theory. N . In some generalizations of the AdS/CFT correspondence the string coupling goes to zero at the boundary even for ﬁnite N . However.

a b c Jµ (x)Jν (y)Jρ (z) − =− N 2 − 1 abc Tr [γ5 γµ ( x− y)γν ( y− z)γρ ( z− x)] id . 32π 6 (x − y)4 (y − z)4 (z − x)4 where dabc = 2Tr(T a {T b . the dimension of all non-chiral operators in the ﬁeld theory. or in terms of the non-conservation of the SU (4)R a current when the theory is coupled to external SU (4)R gauge ﬁelds Fµν . However. in which case the exact spectrum was successfully compared in [105]. In ﬁeld theory this breaking comes from one-loop diagrams and does not receive any further corrections. thus it can be computed also in the strong coupling regime and compared with the results from string theory on AdS space. which is proportional to the number of charged fermions. if we gauge the SU (4)R global symmetry. In other instances of the AdS/CFT correspondence (such as the ones discussed in [106–108]) there exist also additional chiral primary multiplets with n of order N . The reason for this behavior on the ﬁeld theory side is not clear. This happens in particular for the near-horizon limit of NS5-branes. 3.2 Matching of correlation functions and anomalies The classical N = 4 theory has a scale invariance symmetry and an SU (4)R R-symmetry. we will ﬁnd an Adler-Bell-Jackiw anomaly from the triangle diagram of three SU (4)R currents. In the SU (N ) gauge theory this number is N 2 − 1.3) How can we see this eﬀect in string theory on AdS5 ×S 5 ? One way to see it is. These currents are chiral since the fermions λαA are in ¯ the ¯ representation while the fermions of the opposite chirality λA are in 4 α ˙ the 4 representation. such as Tr(φI φI ). grows at 2 least as (gs N )1/4 ∼ (gYM N )1/4 . (3. when the theory is coupled to external gravitational or SU (4)R gauge ﬁelds. these symmetries are broken by quantum eﬀects. to use the general prescription of Section 4 to compute the 3-point . of course. and these have been successfully matched with wrapped branes on the string theory side. The fact that there seem to be no non-chiral ﬁelds on AdS5 with a mass below the string scale suggests that for large N and large gs N . Maldacena: Lectures on Large N Field Theories and Gravity 351 spectrum of chiral operators as the ﬁeld theory. Thus. it is a prediction of the AdS/CFT correspondence.J. (Dµ Jµ )a = N 2 − 1 abc id 384π 2 µνρσ b c Fµν Fρσ . T c }) and we take only the negative parity component of the correlator. and (unlike many other theories) these symmetries are exact also in the full quantum theory. We will begin by discussing the anomaly associated with the SU (4)R global currents. The anomaly can be expressed either in terms of the 3-point function of the SU (4)R global currents.

7) . which we will not discuss here). the action will change by a boundary term of the form − iN 2 384π 2 d4 x ∂AdS5 µνρσ abc d b c Λa Fµν Fρσ . when we have a a coupling of the form d4 xAµ Jµ .3) for large N . to µ a the SU (4)R global currents Jµ of the gauge theory. In the AdS theory we have gauge ﬁelds Aa which couple. as explained above. This eﬀect is precisely reproduced by a Chern-Simons term which exists in the low-energy supergravity theory arising from the compactiﬁcation of type IIB supergravity on AdS5 × S 5 . indicating the breakdown of conformal invariance when the theory is coupled to a curved external metric (there is a similar breakdown of conformal invariance when the theory is coupled to external SU (4)R gauge ﬁelds. In four dimensions.112] and references therein). µν 16π 2 µνρσ 1 1 2 2 I4 = − R2 .4) This term is gauge invariant up to total derivatives.5) From this we can read oﬀ the anomaly in (Dµ Jµ ) since. The conformal anomaly is related to the 2-point and 3-point functions of the energy-momentum tensor [113–116]. but the anomaly means that when we turn on non-zero ﬁeld strengths for these ﬁelds the theory should no longer be gauge invariant.3) from the point of view of the AdS theory [24]. the change in the action under a gauge a 4 a a transformation is given by d x(Dµ Λ)a Jµ = − d4 xΛa (Dµ Jµ ). where E4 = 1 (R2 − 4R2 + R2 ).6) (3. 110] the correct answer to leading order in the large N limit (namely. The other anomaly in the N = 4 SYM theory is the conformal (or Weyl) anomaly (see [111. Gauge Theory and Strings function (3. (3. however.352 Unity from Duality: Gravity. one recovers the term proportional to N 2 ). and indeed one ﬁnds [109. µνρσ − 2Rµν + R 2 16π 3 (3. µ λ σ (3. to consider directly the meaning of the anomaly (3.3). and we ﬁnd exact agreement with (3. the general form of the conformal anomaly is g µν Tµν = −aE4 − cI4 . It is more illuminating. which is of the form iN 2 96π 2 d5 x(dabc AdS5 µνλρσ Aa ∂ν Ab ∂ρ Ac + · · · ). which means that if we take a gauge transformation Aa → Aa + (Dµ Λ)a for which Λ does not µ µ vanish on the boundary of AdS5 .

if we do not require the string theory to have a supergravity approximation then there is no such restriction. so generalizations of the AdS/CFT correspondence which have (for large N ) a supergravity approximation are limited to conformal theories which have a = c in the large N limit. where the conformal anomaly was shown to arise from subtleties in the regularization of the (divergent) supergravity action on AdS space. 5 Computing 4A . such corrections were discussed in [122] and successfully compared in [123–125]5 . µρν and R ≡ Rµ is the scalar curvature. but before summarizing the actual calculations of Green’s functions. Of course. Rµν ≡ Rρ is the Riemann tensor. This approximation amounts to ignoring all the stringy α corrections that cure the divergences of supergravity. This was indeed found to be true in [117–120]4 . so it is protected from any quantum corrections. The result of [117–120] implies that a computation using gravity on AdS5 always gives rise to theories with a = c. 24]. Thus. such corrections tests the conjecture that the correspondence holds order by order in 1/N . In supersymmetric theories the supersymmetry algebra relates g µν Tµν to derivatives of the R-symmetry current. Thus. generalization with more varying ﬁelds may be found in [121]. and to leading order in the large N limit it should be obtained from type IIB supergravity on AdS5 × S 5 . where ISUGRA is the supergravity action evaluated on AdS5 × S 5 (or on small deformations of this space). On the gauge theory side. Maldacena: Lectures on Large N Field Theories and Gravity 353 where Rµνρσ is the curvature tensor. which are of order N 2 . For both of the anomalies we described the ﬁeld theory and string theory computations agree for the leading terms. The basic idea was explained in Section 2. the same result should be obtained in type IIB string theory on AdS5 × S 5 . 4 Correlation functions A useful statement of the AdS/CFT correspondence is that the partition function of string theory on AdS5 × S 5 should coincide with the partition function of N = 4 super-Yang-Mills theory “on the boundary” of AdS5 [23. it seems worthwhile to motivate the methodology from a somewhat diﬀerent perspective. however. which are controlled essentially by the gravitational coupling κ ∼ gst α 2 . It would be interesting to compare other corrections to the large N result.1. For other instances of the AdS/CFT correspondence there are corrections to anomalies at order 1/N ∼ gs (α /R2 )2 . Throughout this section.J. we approximate the string theory partition function by e−ISUGRA . and also all the loop corrections. this is weaker than the statement that the correspondence holds for ﬁnite N . A free ﬁeld computation in the SU (N ) µ N = 4 SYM theory leads to a = c = (N 2 −1)/4. since the 1/N expansion is not expected to converge. they are successful tests of the duality in the large N limit.

then we have more than one saddle point contributing to the string theory partition function. we think of choosing the values φ0 arbitrarily and then extremizing the action ISUGRA [φ] in the region z > subject to these boundary conditions. φ z= (4.2) we have not attempted to be prescient about inserting factors of . In this section.2) =φ0 That is. we arrive at the relation ISUGRA βF . (4.354 Unity from Duality: Gravity. We will discuss the ﬁnite temperature case further in Section 6. and evaluate the action on the solution. The main technical idea behind the bulk-boundary correspondence is that the boundary values of string theory ﬁelds (in particular. In short. we think of closed string states in the bulk as sourcing gauge singlet operators on the brane which originate as composite operators built from open strings. with value φ0 (x) for z = . and we must determine which is most important. Calculating the free energy of a black hole from the Euclidean supergravity action has a long tradition in the supergravity literature [126].1) where W is the generating functional for connected Green’s functions in the gauge theory. so the main claim that is being made here is that the dual gauge theory provides a description of the state of the black hole which is physically equivalent to the one in string theory. W = βF where β is the inverse temperature and F is the free energy of the gauge theory. and the basic relation becomes e−ISUGRA Zstring = Zgauge = e−W . is the on-shell supergravity action. From a D-brane perspective. So. The true boundary of anti-de Sitter space is z = 0. When we apply this relation to a Schwarzschild black hole in AdS5 . we can write Wgauge [φ0 ] = − log e Ê d4 x φ0 (x)O(x) CFT extremum ISUGRA [φ] . If there is more than one solution. We will write the bulk ﬁelds generically as φ(x. the generator of connected Green’s functions in the gauge theory. this approximation amounts to taking both N 2 and gYM N large. which is thought to be reﬂected in the gauge theory by a thermal state at the Hawking temperature of the black hole.2) without modiﬁcation to compute two-point functions of O. Note that in (4. and = 0 serves as a cutoﬀ which will eventually be removed. At ﬁnite temperature. we solve the equations of motion in the bulk subject to Dirichlet boundary conditions on the boundary. supergravity ﬁelds) act as sources for gauge-invariant operators in the ﬁeld theory. gYM N limit. Gauge Theory and Strings as explained in Section 2. Instead our strategy will be to use (4. multiple saddle points will not be a problem. In the supergravity approximation.1. z) (in the coordinate system (2.15)). and then perform a wave-function . and devote the rest of this section to the partition function of the ﬁeld theory on R4 . 2 in the large N.

only the part of the action which is quadratic in the relevant ﬁeld perturbation is needed. As usual. The usual treatment is to split the space into a near region (the throat) and a far region. Equation (4. and the incoming wave impinges upon it. corresponding to a ﬁeld theory which does not have a conformal ﬁxed point in the ultraviolet. the value of the supergravity ﬁeld which the D3-branes feel is the same as the value in the curved space description at the boundary of the near horizon region.J. The bulk-to-boundary propagator is a . The incoming wave from asymptotically ﬂat inﬁnity can be regarded as ﬁxing the value of a supergravity ﬁeld at the outer boundary of the near region. 2 2 (4.2) in a power series in φ0 . We will encounter their explicit form in the next two sections. 127–141]–have undertaken the program of extracting explicit n-point correlation functions of gauge singlet operators by developing both sides of (4. 110.2) to the old approach of extracting Green’s functions from an absorption cross-section [16].3) where η is some normalization which in principle follows from the tendimensional origin of the action. the supergravity description is valid at large N and large ’t Hooft coupling. Because the right hand side is the extremization of a classical action. In the low-energy limit. There is one diﬀerence: in ordinary Feynman graphs one assigns the wavefunctions of asymptotic states to the external legs of the graph. and are called bulk-to-boundary propagators.1 Two-point functions For two-point functions. A remark is in order regarding the relation of (4. At small ’t Hooft coupling. a number of papers–notably [109. In absorption calculations one is keeping the whole D3-brane geometry. For massive scalar ﬁelds in AdS5 . there is a diﬀerent description of the process: a cluster of D3-branes sits at some location in ﬂat ten-dimensional space. They are special limits of the usual gravity propagators in the bulk. but in the present case the external leg factors reﬂect the boundary values φ0 . 24]. this has the generic form S=η 1 √ 1 d5 x g (∂φ)2 + m2 φ2 . Maldacena: Lectures on Large N Field Theories and Gravity 355 renormalization on either O or φ so that the ﬁnal answer is independent of the cutoﬀ. This approach should be workable even in a space (with boundary) which is not asymptotically anti-de Sitter. 4. not just the near-horizon AdS5 × S 5 throat.2) is just a mathematical expression of the fact that the throat geometry should respond identically to the perturbed supergravity ﬁelds as the low-energy theory on the D3-branes. the power series has a graphical representation in terms of treelevel Feynman graphs for ﬁelds in the supergravity. Following [23.

which corresponds to the Poincar´ coordinates of AdS e space. u3 (4. we will employ an explicit cutoﬀ. where the function Z(u) satisﬁes the radial equation u5 ∂u 1 ∂u − u2 − m2 R2 Z(u) = 0 . ( −m2 )φ = 0. Gauge Theory and Strings particular solution of the equation of motion.5).356 Unity from Duality: Gravity. We will always work in Euclidean space6 . then. To avoid divergences associated with the small z region of integration in (4. A complete set of solutions for the linearized equation. . which has proven more convenient for the study of higher point correlators in the conformal case. z) = φ0 (x) = eip·x at z = .5) There are two independent solutions to (4.3). which has special asymptotic properties. z) = Kp (x. A coordinate system in the bulk of AdS5 such that ds2 = R2 dx2 + dz 2 z2 (4. where Iν and Kν are Bessel functions and ∆=2+ 4 + m2 R 2 . we will discuss the position space propagator. is given by φ = eip·x Z(pz). namely Z(u) = u2 I∆−2 (u) and Z(u) = u2 K∆−2 (u). ( −m2 )φ = 0. Imposing the boundary condition φ(x. We will start by considering the momentum space propagator. z) = (pz)2 K∆−2 (pz) ip·x e .4) provides manifest Euclidean symmetry on the directions parametrized by x. we ﬁnd the bulk-to-boundary propagator φ(x. (p )2 K∆−2 (p ) (4. (4.7) 6 The results may be analytically continued to give the correlation functions of the ﬁeld theory on Minkowskian Ê4 . which is useful for computing the two-point function and also in situations where the bulk geometry loses conformal invariance. z ≥ .6) The second solution is selected by the requirement of regularity in the interior: I∆−2 (u) increases exponentially as u → ∞ and does not lead to a ﬁnite action conﬁguration.

8) we have used (4. Only the leading nonanalytic term is essential.3). substituted in (4. and their superpartners . we write O(p)O(q) = ∂ 2 W φ0 = λ1 eip·x + λ2 eiq·x ∂λ1 ∂λ2 λ1 =λ2 =0 = (leading analytic terms in ( p)2 ) −η 2∆−8 (2∆ − 4) Γ(3 − ∆) 4 δ (p + q) Γ(∆ − 1) p 2 2∆−4 (4. the coupling of closed string states to composite open string operators on the brane). In part this is due to the diﬃculty of determining the coupling of bulk ﬁelds to ﬁeld theory operators (or in stringy terms. which amounts to inserting a factor of 4−∆ on the right hand side of (4. Expanding around integer ∆ ≥ 2 one obtains ﬁnite expressions involving log p. as required by conformal invariance. Maldacena: Lectures on Large N Field Theories and Gravity 357 To compute a two-point function of the operator O for which φ0 is a source. in the spirit of diﬀerential regularization [142]. precise matchings between the normalizations in ﬁeld theory and in string theory for all the chiral primary operators have not been worked out. From here on we will take that convention. the graviton. • We have assumed a coupling d4 x φ(x.8) + (higher order terms in ( p)2 ). O(x)O(y) = η 2∆−8 2∆ 1 − 4 Γ(∆ + 1) · 2 Γ(∆ − 2) |x − y|2∆ ∆ π Several explanatory remarks are in order: • To establish the second equality in (4. In fact. The explicit powers of the cutoﬀ in the ﬁnal position space answer can be eliminated by absorbing a factor of ∆−4 into the deﬁnition of O. performed the integral and expanded in . and the higher order terms are unimportant in the limit where we remove the cutoﬀ. z = )O(x) to compute the Green’s functions. The result is a pure power law dependence on the separation |x − y|.7). but they can be deﬁned for generic complex ∆ by analytic continuation and for positive integer ∆ by expanding around a pole and dropping divergent terms. • The Fourier transforms used to obtain the last line are singular.7). We have given the expression for generic real values of ∆. For the dilaton. See [15] for an early approach to this problem. The leading analytic terms give rise to contact terms in position space.J.

2 • vectors: ∆± = 1 (d ± 2 • p-forms: ∆ = 1 (d ± 2 (d − 2)2 + 4m2 ). All other states have a much larger mass. are expected to correspond to operators with dimen2 sion ∆ ∼ (gYM N )1/4 . • The mass-dimension relation (4. the couplings can be worked out explicitly. the literature includes [143–153]. In some of these cases all normalizations have been worked out unambiguously and checked against ﬁeld theory predictions (see for example [23. and typically correspond to dimensions 2 which are ﬁnite (in the large gYM N limit) but irrational. Gauge Theory and Strings (including gauge ﬁelds in AdS5 ). and a table of mass-dimension relations in AdSd+1 with unit radius was made in [154] from the various sources cited above (see also [101]): √ • scalars: ∆± = 1 (d ± d2 + 4m2 ). They still have a mass on the order of 1/R. all correlation functions of ﬁelds in chiral multiplets should follow by application of supersymmetries once those of the chiral primary ﬁelds are known. large N limit to which supergravity applies. (d − 2p)2 + 4m2 ).6) holds even for string states that are not included in the Kaluza-Klein supergravity reduction: the mass and the dimension are just diﬀerent expressions of the second Casimir of SO(4. The remarkable fact is that all the string theory modes with m ∼ 1/R (which is to say. and some of them escape the short multiplets and live in long multiplets of the smaller supergroups. Correlation functions of non-scalar operators have been widely studied following [24]. 2 . The obvious conclusion is that all operators whose dimensions are not algebraically protected have large dimension in the strong ’t Hooft coupling. Conventions seem fairly uniform in the literature. 134]).358 Unity from Duality: Gravity. It is worthwhile to note however that the mass-dimension formula changes for particles with spin. For instance. but the Kaluza-Klein states are roughly as numerous as before. • ﬁrst-order (d/2)-forms (d even): ∆ = 1 (d + 2|m|). with √ m ∼ 1/ α . excited string states. This is no longer true for theories of reduced supersymmetry: the supergroup gets smaller. 2|4). 2 • spinors: ∆ = 1 (d + 2|m|). For N = 4 super-Yang-Mills theory. so in this case it should be enough to study the scalars. 109. all closed string states which arise from massless ten dimensional ﬁelds) fall in short multiplets of the supergroup SU (2. In fact the deﬁnition of mass has some conventiondependence. 2). The operators in short multiplets have algebraically protected dimensions.

the bulk-to-boundary propagator K∆ (x. The unique regular K∆ solving the equation of motion and satisfying (4. . cut oﬀ all bulk integrals at that boundary. z = . we have not attempted to pick which of ∆± is the physical dimension. y) is a continuous function which approximates 4−∆ δ 4 (x − y) better and better as → 0. and only afterwards take → 0. the Fourier transform of (4. one arrives at a diﬀerent answer.8). This apparently innocuous subtlety turned out to be important for two-point functions. (However.J. 4. as discovered in [109]. This is not .) The boundary behavior of position space bulk-to-boundary propagators is speciﬁed in a slightly more subtle way: following [109] we require K∆ (x. A correct prescription is to specify boundary conditions at ﬁnite z = . but for higher point functions position space is preferred because the full conformal invariance is more obvious.10) At a ﬁxed cutoﬀ. z. but in some cases (notably m2 = 15/4 in AdS5 ) there is a genuine ambiguity. Thus at any ﬁnite .2 Three-point functions Working with bulk-to-boundary propagators in the momentum representation is convenient for two-point functions. Usually the choice ∆ = ∆+ is clear from the unitarity bound. but cutting the bulk integrals oﬀ again at . y) → z 4−∆ δ 4 (x − y) as z → 0. z. and instead refer the reader to the (extensive) literature. Maldacena: Lectures on Large N Field Theories and Gravity • spin-3/2: ∆ = 1 (d + 2|m|). y) = Γ(∆) π 2 Γ(∆ − 2) z z 2 + (x − y)2 ∆ · (4.9) Here y is the point on the boundary where we insert the operator. the momentum representation seems uniformly more convenient. for non-conformal examples of the bulk-boundary correspondence. That is what we have done in (4. and (x. In practice this ambiguity is usually resolved by appealing to some special algebraic property of the relevant ﬁelds. such as transformation under supersymmetry or a global bosonic symmetry. z) is a point in the bulk.9).9) is K∆ (x. 2 • massless spin-2: ∆ = d. For brevity we will omit a further discussion of higher spins.8)). Calculating two-point functions directly using the position-space propagators (4.10) only approximately coincides with (4. (4. 359 In the case of ﬁelds with second order lagrangians.7) (modiﬁed by the factor of 4−∆ as explained after (4. and ﬁnally taking the same → 0 answer.

of the form S= √ d5 x g i 1 (∂φi )2 + 1 m2 φ2 + λφ1 φ2 φ3 .e. The Feynman graph for the three-point function as computed in supergravity. and φ3 . The authors of [109] checked that using the cutoﬀ consistently (i. to the evaluation of the graph shown in Figure 2. x2 )K∆3 (x.12) . The legs correspond to factors of K∆i . Gauge Theory and Strings 1 2 λ 3 Fig. via (4. φ2 . 2.11) and if the φi couple to operators in the ﬁeld theory by interaction terms d4 x φi Oi . surprising since the z = boundary conditions were not used consistently. where Jµ is a conserved current. The position of the vertex is integrated over AdS5 . were obeyed. Two-point functions are uniquely diﬃcult because of the poor convergence properties of the integrals over z. x3 ) λa1 . The integrals involved in three-point functions are suﬃciently benign that one can ignore the issue of how to impose the cutoﬀ. That is. O1 (x1 )O2 (x2 )O3 (x3 ) = −λ = √ d5 x gK∆1 (x. |x1 − x2 |∆1 +∆2 −∆3 |x1 − x3 |∆1 +∆3 −∆2 |x2 − x3 |∆2 +∆3 −∆1 (4. and the cubic vertex to a factor of λ.360 Unity from Duality: Gravity. with the momentum space propagators) gave two-point functions O(x1 )O(x2 ) a normalization such that Ward identities involving the three-point function O(x1 )O(x2 )Jµ (x3 ) .2). then the calculation of O1 O2 O3 reduces. 2 i i 2 (4. x1 )K∆2 (x. If one has a Euclidean bulk action for three scalar ﬁelds φ1 .

These operators are the chiral primaries of the gauge theory: all other single trace operators of protected dimension descend from these by commuting with supersymmetry generators. where the parentheses indicate a symmetrized traceless product.15).13) Γ In principle one could also have couplings of the form φ1 ∂φ2 ∂φ3 . The dependence on the xi is dictated by the conformal invariance. φJ ) in ﬁeld theory. For AdS5 × S 5 . 159]. Maldacena: Lectures on Large N Field Theories and Gravity 361 for some constant a1 . The result of [132] is that the equations of motion for the scalar modes sI dual to ˜ I OI = CJ1 . (4.15) requires some explanation. The main technical diﬃculty with three-point functions is that one must ﬁgure out the cubic couplings of supergravity ﬁelds.J Trφ(J1 . . . . In ﬂat ten-dimensional space these terms can be read oﬀ directly from the original type IIB supergravity papers [158. This leads only to a modiﬁcation of the constant a1 . I is an index which runs . . but an appropriate ﬁeld redeﬁnition eliminates them. This was done for a restricted set of ﬁelds in [132].. The ﬁelds considered were those dual to operators of the form Trφ(J1 φJ2 . for 3-point functions it is necessary to go to one higher order of perturbation theory. but the only way to compute a1 is by performing the integral over x.I3 GI1 I2 I3 wI1 wI2 wI3 sI1 sI2 sI3 · ˜ ˜ ˜ 3 (4.14) and (4. one must instead expand in ﬂuctuations around the background metric and ﬁve-form ﬁeld strength.J. it is most straightforward to read oﬀ these “cubic couplings” from quadratic terms in the equations of motion. The result [109] is a1 = − Γ 1 2 (∆1 1 2 (∆1 + ∆2 − ∆3 ) Γ 1 (∆1 + ∆3 − ∆2 ) Γ 1 (∆2 + ∆3 − ∆1 ) 2 2 · 2π 4 Γ(∆1 − 2)Γ(∆2 − 2)Γ(∆3 − 2) + ∆2 + ∆3 ) − 2 . Only the metric and the ﬁve-form are involved in the dual supergravity ﬁelds.14) follow from an action of the form S= 4N 2 (2π)5 d5 x √ g I AI (wI )2 sI −(∇˜I )2 − l(l − 4)˜2 s 2 + I1 .15) Derivative couplings of the form s∂˜∂˜ are expected a priori to enter ˜ s s into (4..I2 . and we are interested only in modes which are scalars in AdS5 . Because of the diﬃculties in writing down a covariant action for type IIB supergravity in ten dimensions (see however [155–157]). The notation in (4. φJ ) (4. The old literature [103] only dealt with the linearized equations of motion.

. ∆1 +∆2 −∆3 |x − x |∆1 +∆3 −∆2 |x − x |∆2 +∆3 −∆1 N |x1 − x2 | 1 3 2 3 where we have deﬁned I I I3 C I1 C I2 C I3 = CJ1 ···Ji K1 ···Kj CJ2 ···Ji L1 ···Lk CK1 ···Kj L1 ···Lk .17) Remarkably.17) is the same result one obtains from free ﬁeld theory by Wick contracting all the φJ ﬁelds in the three operators.2). the three-point function. This is the same strategy as was pursued in comparing matrix models of quantum gravity to Liouville theory. .1) C . To get rid of factors of wI .12).J CJ1 . as calculated using (4. One can choose wI so ˜ ˜ that a two-point function computation along the lines of Section 4. These are the representations I whose Young diagrams are . 3. CJ1 . The Wilson loop operator W (C) = Tr P exp i A (5. Gauge Theory and Strings over the weight vectors of all possible representations constructed as symmetric traceless products of the 6 of SU (4)R . · · · .J is a basis transformation I J matrix.1 leads to OI1 (x)OI2 (0) = δ I1 I2 · x2∆1 (4.. It is worth emphasizing that the normalization ambiguity in the bulk-boundary coupling is circumvented essentially by considering invariant ratios of three-point functions and two-point functions. It is the combination sI = wI sI rather than sI itself which ˜ ˜ has a deﬁnite relation to supergravity ﬁelds. 1 1 (4. into which the “leg factors” wI do not enter.. but such a theorem has not yet been proven (see however comments at the end of Sect. We refer the reader to [132] for explicit expressions for AI and the symmetric tensor GI1 I2 I3 ..362 Unity from Duality: Gravity. . there is generally a normalization ambiguity on how supergravity ﬁelds couple to operators in the gauge theory.. is OI1 (x1 )OI2 (x2 )OI3 (x3 ) = √ 1 ∆1 ∆2 ∆3 C I1 C I2 C I3 .. (4.J = δ IJ . chosen so that CJ1 . As commented in the previous section.16) With this choice. We have taken the coupling to ˜ be d4 x sI OI . and the normalization ambiguity is represented by the “leg factors” wI . This suggests that there is a non-renormalization theorem for this correlation function. 5 Wilson loops In this section we consider Wilson loop operators in the gauge theory. we introduce operators OI = wI OI .

1 Wilson loops and minimum surfaces In QCD.2) 2 where V (L) is the quark anti-quark potential. 5. m = U/2π. viewing T as the time direction. transforming in the N of U (N ). Then. Thus. get a mass proportional to U . which means U should also go to inﬁnity. the string should end on the boundary of AdS space.J. we expect the Wilson loop to be related to the string running from the quark to the antiquark. Maldacena: Lectures on Large N Field Theories and Gravity 363 depends on a loop C embedded in four dimensional space. For this purpose we consider a rectangular loop with sides of length T and L in Euclidean space. Since they are charged they will act as a source for the vector ﬁelds. The coupling to the scalar ﬁelds can be seen qualitatively by viewing the quarks as strings stretching between the N branes and the single separated brane. which act as a source for the various U (N ) ﬁelds. from the point of view of the U (N ) gauge theory. These strings will pull the N branes and will cause a deformation of the branes. which is described by the scalar ﬁelds. For large N and large gYM N . From the expectation value of the Wilson loop operator W (C) we can calculate the quark-antiquark potential. These stretched strings will also act as a source for the scalar ﬁelds. Finally. we will discuss here only the case of the fundamental representation (see [165] for a discussion of other representations). it is clear that for large T the expectation value will behave as e−T E where E is the lowest possible energy of the quark-anti-quark conﬁguration. which will correspond precisely to the Wilson loop operator) we need to take m → ∞. The oﬀ-diagonal states. one can see . we can view these states as massive quarks. (5. We expect this string to be analogous to the string in our conﬁguration. We start with the gauge group U (N + 1). and the coupling to the scalar (Higgs) ﬁelds is determined by supersymmetry. This corresponds. A more formal argument for this coupling is that these states are BPS. The trace is taken over some representation of the gauge group. and which can stretch between two points on the boundary of AdS. So. as discussed in Section 2. the AdS/CFT correspondence maps the computation of W in the CFT into a problem of ﬁnding a minimum surface in AdS [166. which is a superstring which lives in ten dimensions. Thus. In order to motivate this prescription let us consider the following situation. 167]. In order to get a non-dynamical source (an “external quark” with no ﬂuctuations of its own. to having a D3 brane sitting at some radial position U in AdS. and it involves the path-ordered integral of the gauge connection along the contour. and at a point on S 5 . and we break it to U (N ) × U (1) by giving an expectation value to one of the scalars. we have W ∼ e−T V (L) .

is ﬁnite. A detailed derivation of (5. The precise deﬁnition of the Wilson loop operator corresponding to the superstring will actually include also the ﬁeld theory fermions. The area as deﬁned above is divergent. the discussion above implies that in order to compute the expectation value of the operator (5. This area is measured with the AdS metric. . If we evaluate the area up to some radial distance U = r. 6) is a unit vector in R6 (the point on S 5 where the string is sitting). On the other hand. when gs N is large. Generalizing the prescription of section 4 for computing correlation functions. as it should be since a divergence in the Wilson loop would have implied a mass renormalization of the BPS particle. the perturbative computation in the ﬁeld theory shows that W . this will not aﬀect the leading order computations we describe here. The apparent discrepancy between the divergence of the area of the minimum surface in AdS and the ﬁniteness of the ﬁeld theory computation can be reconciled by noting that the appropriate action for the string worldsheet is not the area itself but its Legendre transform with respect to the string coordinates corresponding to θI and the 7 The diﬀerence in the factor of i between the Euclidean and the Minkowski cases can √ be traced to the analytic continuation of x2 .3). The divergence arises from the fact that the string worldsheet is going all the way to the boundary of AdS. In the supergravity regime. and it is generally not the same as the area enclosed by the loop C in four dimensions. for W given by (5. the phase factor associated to the trajectory of the quark has an extra factor “i” in front of θI 7 . So. the ﬁnal conclusion is that the stretched strings couple to the operator W (C) = Tr P exp √ (iAµ xµ + θI φI x2 )dτ ˙ ˙ . as in Figure 3 [166. with the condition that we have a string worldsheet ending on the loop C. we see that for large r it diverges as r|C|. Gauge Theory and Strings this coupling explicitly by writing the full U (N + 1) Lagrangian. which will imply some particular boundary conditions for the worldsheet fermions at the boundary of AdS. · · · . where |C| is the length of the loop in the ﬁeld theory [166. 167]. In the Minkowski signature case.364 Unity from Duality: Gravity. 167]. This is the expression when the signature of R4 is Euclidean. (5. putting in the Higgs expectation value and calculating the equation of motion for the massive ﬁelds [166]. the leading contribution to this partition function will come from the area of the string worldsheet.3) where xµ (τ ) is any parametrization of the loop and θI (I = 1.3) in N = 4 SYM we should consider the string theory partition function on AdS5 × S 5 .3) can be found ˙ in [168]. However.

This is because these string coordinates obey the Neumann boundary conditions rather than the Dirichlet conditions.J.6) This is independent of a as required by conformal invariance. the Legendre transformation simply subtracts the divergent term r|C|. leaving the resulting action ﬁnite. When the loop is smooth. Note that there is a conformal transformation in the ﬁeld theory that maps a line to a circle. .5) where we have regularized the area by putting a an IR cutoﬀ at z = in AdS. Maldacena: Lectures on Large N Field Theories and Gravity 365 Fig. in this case it is easier to use conformal invariance. and we get a contribution to the action S∼ R2 1 A= 2πα 2πα a dθ dza R2 = z2 α a −1 . radial coordinate u [168]. Take C to be a circle of radius a on the boundary. and let us work in the Poincar´ coore dinates. As an example let us consider a circular Wilson loop. e2 are two orthonormal vectors in four dimensions (which deﬁne the orientation of the circle) and 0 ≤ z ≤ a. The Wilson loop operator creates a string worldsheet ending on the corresponding loop on the boundary of AdS. using the coordinates (2. (5. 3. Using the conformal transformation to map the line to a circle we obtain the minimal surface we want.4) where e1 . Subtracting the divergent term we get W ∼ e−S ∼ eR 2 /α =e √ 4πgs N . (5. We could ﬁnd the surface that minimizes the area by solving the Euler-Lagrange equations. x= a2 − z 2 (e1 cos θ + e2 sin θ). which is equivalent to a UV cutoﬀ in the ﬁeld theory [66]. the minimum area surface is clearly a plane that intersects the boundary and goes all the way to the horizon (which is just a point on the boundary in the Euclidean case).15) for AdS5 . (5. We can calculate the area of this surface in AdS. It is. In the case of the line. However.

Gauge Theory and Strings We could similarly consider a “magnetic” Wilson loop. we ﬁnd that the solution is8 1 x = z0 z/z0 dyy 2 1 − y4 . We get the same result as in (5. (5. which is also called a ’t Hooft loop [169]. Using (5.7).9) Deﬁning z0 to be the maximum value of z(x).7) where GMN is the Euclidean AdS5 × S 5 metric. σ = x. we should consider in this case a D-string worldsheet instead of a fundamental string worldsheet. (5.2) it is possible to compute the quark-antiquark potential in the supergravity approximation [166. We put the quark at x = −L/2 and the anti-quark at x = L/2. Since we identify the electric-magnetic duality with the SL(2. Note that the factors of α cancel out in (5. This case is related by electric-magnetic duality to the previous case. which by symmetry occurs at x = 0.366 Unity from Duality: Gravity. Since we are interested in a static conﬁguration we take τ = t. 167].11) integrals in this section can be calculated in terms of elliptic or Beta functions.10) where z0 is determined by the condition L = z0 2 8 All 1 0 dyy 2 1 − y4 = z0 2π 3/2 · Γ(1/4)2 √ (5. Here “quark” means an inﬁnitely massive W-boson connecting the N branes with one brane which is (inﬁnitely) far away. .6) but with gs → 1/gs . Z) duality of type IIB string theory. We take both particles to have the same scalar charge. Since the action does not depend on x explicitly the solution satisﬁes 1 z2 (∂x z)2 + 1 = constant. The classical action for a string worldsheet is S= 1 2πα dτ dσ det(GMN ∂α X M ∂β X N ).8) We need to solve the Euler-Lagrange equations for this action. which means that the two ends of the string are at the same point in S 5 (one could consider also the more general case with a string ending at diﬀerent points on S 5 [166]). (5. In this case we consider a conﬁguration which is invariant under (Euclidean) time translations. as they should. and then the action becomes S= T R2 2π L/2 dx −L/2 (∂x z)2 + 1 · z2 (5.

8). independently of the value of gs . Maldacena: Lectures on Large N Field Theories and Gravity 367 The qualitative form of the solution is shown in Figure 4b. b) Conﬁguration after we consider the coupling to the U (N ) gauge theory.13) We see that the energy goes as 1/L. 8 . This conﬁguration minimizes the action.10) in (5. Some subleading corrections coming from quantum ﬂuctuations of the worldsheet were calculated in [170–172]. Notice that the string quickly approaches x = L/2 for small z (close to the boundary). and we ﬁnd E = V (L) = − 2 4π 2 (2gYM N )1/2 · Γ( 1 )4 L 4 (5. This indicates some screening of the charges at strong coupling. L − x ∼ z3 .J. U= L x U=0 (a) U (b) Fig. 4a). 2 z→0. Note that the energy is proportional to (gYM N )1/2 . We just plug in the solution (5. (5. as opposed 2 to gYM N which is the perturbative result.12) Now we compute the total energy of the conﬁguration. as in Fig. The above calculation makes sense for all distances L when gs N is large. a fact which is determined by conformal 2 invariance. a) Initial conﬁguration corresponding to two massive quarks before we turn on their coupling to the U (N ) gauge theory. 4. The quark-antiquark energy is given by the diﬀerence of the total length of the strings in a) and b). subtract the inﬁnity as explained above (which can be interpreted as the energy of two separated massive quarks.

the dependence on a is very important. In fact. Therefore. Another interesting case one can study analytically is a surface near a cusp on R4 . In the supergravity approximation the zero-brane follows a geodesic. it is associated with the insertion of two local operators at the two points where the trajectory ends. Other aspects of the gravity calculation of Wilson loops were discussed in [175–179]. a . since it leads to a result of the form 1 (5. Therefore.e. and they will connect to form a (1. We ﬁnd a logarithmic divergence when → 0. In this case the fundamental string (ending on the quark) will attach to the D-string (ending on the monopole). and the result will be the same as (5. a particle) of mass m. This can be understood from the fact that when g is small the D-string is very rigid and the fundamental string will end almost perpendicularly on the D-string. In this case.15) e−S ∼ e−2mR log a ∼ 2mR . z a2 − z 2 (5. but in the limit that gYM is small (but still with gYM N large) we get that the monopole-quark potential is just 1/4 of the quark-quark potential. One can also calculate the interaction between a magnetic monopole and a quark. Gauge Theory and Strings In a similar fashion we could compute the potential between two magnetic monopoles in terms of a D-string worldsheet.14) where we took the distance between the two points at the boundary to be L = 2a and regulated the result. In Euclidean space a zero-brane describes a one dimensional trajectory in anti-de-Sitter space which ends at two points on the boundary. Naively we might have thought that (as in the previous subsection) the answer had to be independent of a due to conformal invariance. If we compute the action we get a S=m ds = −2mR adz √ . The area of the minimum surface also contains a logarithmic divergence depending on the angle [168]. the solution for the fundamental string will be half of the solution we had above. proportional to log( /a). 1) string which will go into the horizon. If we subtract the logarithmic divergence we get a residual dependence on a. the perturbative computation in the gauge theory shows a logarithmic divergence with a coeﬃcient depending on the angle at the cusp.13) but with gYM → 4π/gYM . Calculations of Wilson loops in the Higgs phase were done in [174].2 Other branes ending on the boundary We could also consider other branes that are ending at the boundary [180]. The simplest example would be a zero-brane (i. The resulting potential is a complicated function of 2 gYM [173]. 5. Geodesics in the hyperbolic plane (Euclidean AdS) are semicircles. leading to a factor of 1/4 in the potential.368 Unity from Duality: Gravity.

with a three dimensional worldvolume. 6 Theories at ﬁnite temperature As discussed in Section 3. 181]. and the insights we gain from examining the T > 0 physics will be of a more qualitative nature. and it implies that the expectation value of the corresponding operator depends on the overall scale. On general grounds one expects that the subleading corrections are given by surfaces that end on more than one loop. Again.12). 6. Maldacena: Lectures on Large N Field Theories and Gravity 369 which is precisely the result we expect for the two-point function of an operator of dimension ∆ = mR. ending on a two dimensional surface on the boundary of AdS7 . In that case one has to consider a two-brane. 0) ﬁeld theory [180. one gets a logarithmic term. 0) theory. which is proportional to the rigid string action of the two dimensional surface living on the string in the N = (2. the quantities that can be most successfully compared between gauge theory and string theory are those with some protection from supersymmetry and/or conformal invariance – for instance. To leading order in N this will be just the product of the expectation values of each Wilson loop. dimensions of chiral primary operators.1 Construction The gravity solution describing the gauge theory at ﬁnite temperature can be obtained by starting from the general black three-brane solution and taking the decoupling limit of Section 2 keeping the energy density above .2 we explain how a phase transition studied by Hawking and Page in the context of quantum gravity is mapped into a conﬁnement-deconﬁnement transition in the gauge theory. Finite temperature breaks both supersymmetry and conformal invariance. They are no less interesting for that: we shall see in Section 6. then in Section 6. this sort of logarithmic divergence arises when the brane worldvolume is odd dimensional [180]. These thin tubes are nothing else than massless particles being exchanged between the two string worldsheets [165.1 how the entropy of near-extremal D3-branes comes out identical to the free ﬁeld theory prediction up to a factor of a power of 4/3. so we reproduce in the supergravity limit the 2-point function described in Section 4. This is precisely the large mR limit of the formula (2.J. 181]. One limiting case is when the surfaces look similar to the zeroth order surfaces but with additional thin tubes connecting them. In particular one could consider the “Wilson surfaces” that arise in the six dimensional N = (2. In general. One can also compute correlation functions involving more than one Wilson loop.

and one sets the ﬁeld theory temperature equal to the Hawking temperature in supergravity. the Hawking temperature. the entropy of near-extremal D3-branes is just the usual Bekenstein-Hawking result. 3 (6. but nevertheless the two results for the entropy agreed up to a factor of a power of 4/3. and use the relation between the ﬁnite temperature theory and the Euclidean theory with a compact time direction. and it would acquire corrections suppressed by powers of T R if we had considered the full D3brane metric rather than the near-horizon limit.1). The ﬁrst computation which indicated that ﬁnite-temperature U (N ) Yang-Mills theory might be a good description of the microstates of N coincident D3-branes was the calculation of the entropy [182. On the supergravity side. where temperature and volume are the independent variables. There is no problem on the gauge theory side in working at large N . In the canonical ensemble. equation (6.370 Unity from Duality: Gravity. The result is π2 2 N V T 4. and the ADM mass. These corrections do not have an interpretation in the context of CFT because they involve R as an intrinsic scale. The resulting metric can be written as du2 + dΩ2 5 hu2 ds2 = R2 u2 (−hdt2 + dx2 + dx2 + dx2 ) + 1 2 3 u4 h=1− 0 . 183]. The analysis of [182] was limited to a free ﬁeld computation in the ﬁeld theory. Two equivalent methods to evaluate FSUGRA are a) to use F = E − T S together with standard expressions for the Bekenstein-Hawking entropy. u4 u0 = πT.2) The supergravity result is at leading order in ls /R. and it is expected to be a reliable 2 guide to the entropy of the gauge theory at large N and large gYM N . (6. and b) to consider the gravitational action of the Euclidean solution. one identiﬁes the ﬁeld theory volume with the world-volume of the D3-branes.1) It will often be useful to Wick rotate by setting tE = it. . but large 2 gYM N at ﬁnite temperature is diﬃcult indeed. Gauge Theory and Strings extremality ﬁnite. S = A/4GN . 8 FSUGRA = − FSYM 4 = FSUGRA .

. The weak coupling result is a straightforward although somewhat tedious application of the diagrammatic methods of perturbative ﬁnite-temperature ﬁeld theory.. It was pointed 2 out early [185] that the quartic potential gYM Tr[φI . and unlike in the 1+1-dimensional case where the entropy is essentially ﬁxed by the central charge. It is important now to work with the Euclidean solution. S 1/4 S SYM = (4/3) SUGRA . Indeed.3) 2 where f (gYM N ) is a function which smoothly interpolates between a weak coupling limit of 1 and a strong coupling limit of 3/4. which would suggest that f (gYM N ) is monotone decreasing. that f (gYM N ) should remain ﬁnite at strong coupling.2). respectively. the ADM mass above extremality was identiﬁed with the ﬁeld theory energy. (6. The 4/3 factor is a long-standing puzzle into which we still have only qualitative insight.4) 2 for large gYM N . 2π 2 YM ζ(3) 3 45 2 + . φJ ]2 in the N = 4 Yang-Mills action might be expected to freeze out more and more degrees of 2 freedom as the coupling was increased... . The constant term is from one loop. 6 (6. but it is only because the authors worked in the microcanonical ensemble: rather than identifying the Hawking temperature with the ﬁeld theory temperature. f (gYM N ) = + 2 N )3/2 4 32 (gYM 2 f (gYM N ) = 1 − 2 for small gYM N . The strong coupling result follows from considering the leading α corrections to the supergravity action. The results are 3 2 g N + . The gauge theory computation was performed at zero ’t Hooft coupling. 2 The leading corrections to the limiting value of f (gYM N ) at strong and weak coupling were computed in [184] and [187]. and the leading correction is from two loops. and one restricts attention 9 The result of [182]. diﬀers superﬁcially from (6. there is no nonrenormalization theorem for the coeﬃcient of T 4 in the free energy. The relevant one involves a particular contraction of four powers of the Weyl tensor. it was suggested in [184] that the leading term in the 1/N expansion of F has the form 2 F = −f (gYM N ) π2 2 N V T 4. whereas the supergravity is supposed to be valid at strong ’t Hooft coupling.J. based on the non2 renormalization of the two-point function of the stress tensor. Maldacena: Lectures on Large N Field Theories and Gravity 371 with a periodicity in the Euclidean time direction (related to the temperature) which eliminates a conical deﬁcit angle at the horizon9 . An argument has been given [186].

We can think of this as the saddle point approximation to the path integral over supergravity ﬁelds. Despite the conceptual diﬃculties we can use some simple intuition about path integrals to illustrate an important point about the AdS/CFT correspondence: namely. The Weyl curvature comes from the noncompact part of the metric. The action including the α corrections no longer has the Einstein-Hilbert form. which is no longer AdS5 but rather the AdSSchwarzschild solution which we will discuss in more detail in Section 6. and when there is we should sum e−ISUGRA over the classical conﬁgurations to obtain the saddle-point approximation to the gauge theory partition function. the relation I = βF is regarded as the correct one 2 (see [188]). An example of this was studied in [191] long before the AdS/CFT correspondence. we will restrict our attention to AdS5 . generalized. When there are two or more solutions competing to minimize ISUGRA . It has been conjectured that the interpolating function f (gYM N ) is not smooth. and correspondingly the Bekenstein-Hawking prescription no longer agrees with the free energy computed as βI where I is the Euclidean action. Multiple classical conﬁgurations are possible because of the general feature of boundary value problems in diﬀerential equations: there can be multiple solutions to the classical equations satisfying the same asymptotic boundary conditions. In particular. there can be more than one saddle point in the range of integration. 68] as a conﬁnementdeconﬁnement transition in the gauge theory. 6. Gauge Theory and Strings further to the near-horizon limit. String amplitudes (when we can calculate them) render on-shell quantities well-deﬁned.24].2 Thermal phase transition 2 The holographic prescription of [23. It is worth noting however that if the AdS5 geometry is part of a string compactiﬁcation. and subsequently resurrected. The arguments in [189.2. applied at large N and gYM N where loop and stringy corrections are negligible. We regard this as an unsettled question.372 Unity from Duality: Gravity. and reinterpreted in [24. it doesn’t matter what the internal manifold is except insofar . Since the qualitative features are independent of the dimension. but exhibits some phase transition at a ﬁnite value of the ’t Hooft coupling. The solution which globally minimizes ISUGRA is the one that dominates the path integral. That path integral is ill-deﬁned because of the non-renormalizable nature of supergravity. involves extremizing the supergravity action subject to particular asymptotic boundary conditions. 190] seem as yet incomplete. there can be a phase transition between them. In keeping with the basic prescription for computing Green’s functions. where a free energy in ﬁeld theory is equated (in the appropriate limit) with a supergravity action. they rely on analyticity properties of the perturbation expansion which do not seem to be proven for ﬁnite temperature ﬁeld theories.

this space is S 3 × B 2 . or equivalently the radius R of anti-de Sitter space. The one used in [68. Because the S 1 factor is not simply connected. the metric is ds2 = f dt2 + 1 2 dr + r2 dΩ2 . in a saddle-point approximation. Maldacena: Lectures on Large N Field Theories and Gravity 373 as it ﬁxes the cosmological constant. in order to eliminate the conical singularity at r = r+ .6) with µ = 0 and again with periodic time. Actually. we ﬁx the period of Euclidean .5) r2 µ f = 1+ 2 − 2· R r (6. Another space with the same boundary which is also a local extremum of (6. but for X1 . Tr(−1)F e−βH .6) The radial variable r is restricted to r ≥ r+ . 184] is to cut oﬀ both X1 and X2 at a deﬁnite coordinate radius r = R0 . But. which we will call X1 .J. For X2 . There is an embedding of the Schwarzschild black hole solution into anti-de Sitter space which extremizes the action I=− Explicitly. where r+ is the largest root of f = 0. t ∼ t + β. In order to make the comparison of I(X1 ) and I(X2 ) meaningful. corresponding to thermal (anti-periodic) or supersymmetric (periodic) boundary conditions on fermions. but also topologically B 4 ×S 1 rather than S 3 ×B 2 . We will call this space X2 . This space. the period is arbitrary. is not only metrically distinct from the ﬁrst (being locally conformally ﬂat). only X1 contributes. 3 f 1 16πG5 d5 x √ g R+ 12 R2 · (6. For the purpose of computing the twisted partition function. This requires β= 2πR2 r+ · 2 2r+ + R2 (6.7) Topologically. and the boundary is S 3 × S 1 (which is the relevant space for the ﬁeld theory on S 3 with ﬁnite temperature). both I(X1 ) and I(X2 ) are inﬁnite. corresponding to thermal boundary conditions. In contrast. the elimination of the conical deﬁcit angle at the horizon ﬁxes the period of Euclidean time. X2 is simply connected and hence admits a unique spin structure. The Euclidean time is periodically identiﬁed. Tre−βH . and the more important one is the one with the smaller Euclidean action. there are two possible spin structures on X1 . so to compute I(X2 )−I(X1 ) a regulation scheme must be adopted. X1 and X2 make separate saddle-point contributions to the usual thermal partition function.5) is given by the metric in (6.

has the topology B 4 × S 1 . The order parameter for the breaking of the center is the expectation value of the Polyakov (temporal) loop W (C) . In the limit R0 → ∞. one ﬁnds I(X2 ) − I(X1 ) = 3 2 π 2 r+ (R2 − r+ ) 2 + R2 ) . Similar transitions. Gauge Theory and Strings time on X1 so that the proper circumference of the S1 at r = R0 is the same as the proper length on X2 of an orbit of the Killing vector ∂/∂t. Deconﬁnement at high temperature can be characterized by a spontaneous breaking of the center of the gauge group.8) (or more precisely its AdS4 analog) can change its sign was interpreted in [191] as indicating a phase transition between a black hole in AdS and a thermal gas of particles in AdS (which is the natural interpretation of the space X1 ). which immediately implies that W (C) = 0. have been studied in the context of spinning branes and charged black holes in [192–198]. In the deconﬁned phase Fq (T ) is ﬁnite and therefore W (C) = 0. The black hole is the thermodynamically favored state when the horizon radius r+ exceeds the radius of curvature R of AdS. See also [201–204] for other interesting contributions to the ﬁnite temperature literature. Connections with Higgsed states in gauge theory are clearer in [199. This is the expected behavior at low temperatures (compared to the inverse radius of the S 3 ). and the path C wraps around the circle. The boundary of the spaces X1 . Most of these works are best understood on the CFT side as explorations of exotic thermal phenomena in ﬁnite-temperature gauge theories. Since the theory is conformally invariant. Consider ﬁrst the low temperature phase. Therefore C is not a boundary of any D. in order to compute W (C) we have to evaluate the partition function of strings with a worldsheet D that is bounded by the loop C. as discussed above. W (C) ∼ exp (−Fq (T )/T ). . The expectation value of the Polyakov loop measures the change of the free energy of the system Fq (T ) induced by the presence of the external charge q. In the gauge theory we interpret this transition as a conﬁnement-deconﬁnement transition. The fact that (6. X2 is S 3 ×S 1 . where the center of the gauge group is not broken.8) where again r+ is the largest root of f = 0. The contour C wraps the circle and is not homotopic to zero in X1 . and also local thermodynamic instability due to negative speciﬁc heats. 4G5 (2r+ (6.374 Unity from Duality: Gravity. the transition temperature must be proportional to the inverse radius of the space S 3 which the ﬁeld theory lives on. 200]. In a conﬁning phase Fq (T ) is inﬁnite and therefore W (C) = 0. In our case the gauge group is SU (N ) and its center is ZN . The relevant space is X1 which. As discussed in Section 5. The relevance to conﬁnement is explored in [197]. also at r = R0 . An element of the center g ∈ ZN acts on the Polyakov loop by W (C) → g W (C) .

[5] F. B 448 (1999) 37. since the center should not be broken in ﬁnite volume (S 3 ). which contributes iψ to the string worldsheet action.M. Polyakov. Rept. Explicit computations of Polyakov loops at ﬁnite temperature were done in [6. ’t Hooft. B 72 (1974) 461. Lett.R. In contrast. Theor. with vanishing ﬁeld strength. [8] A. I thank O. Park and K. Maldacena.P. Maldacena: Lectures on Large N Field Theories and Gravity 375 For the high temperature phase the relevant space is X2 . The string theory partition function includes now an integral over all values of ψ. which has the topology S 3 × B 2 . Lett. Boonstra. Phys. H. Oz for collaborating on writing that review. This seems to be in agreement with the fact that in the high temperature phase W (C) = 0 and the center of the gauge group is broken. Peeters and K. J. [2] H. J. B 103 (1981) 207. [9] J. References [1] O. Skenderis. [7] G. Phys. S.J.S. Itzhaki. Phys. Some of the non-trivial successes of this approach to QCD are summarized in [1]. [3] K. Ooguri and Y. making W (C) = 0 on S 3 . Phys. the solution X2 is not unique and we can add to it an expectation value for the integral of the NS-NS 2-form ﬁeld B on B 2 . Yankielowicz. and when the S 1 is made small (corresponding to high temperature T ) the theory at distances larger than 1/T eﬀectively reduces to pure Yang-Mills on R3 . 205]. Phys. Gubser. Oz. Maldacena. Phys. Nucl. [4] S. [10] J. ’t Hooft. Gonzalez-Rey. I also thank the organizers for a very nice school. which may be understood as the dependence on the center ZN in the large N limit. Phys. Lett. B. Gubser. In [68] the Euclidean black hole solution (6. Skenderis. on R3 one integrates over the local ﬂuctuations of ψ but not over its vacuum expectation value. Lett. B 517 (1998) 179. Indeed. Rev. Adv. Brandhuber. Nucl. Aharony. . B 411 (1997) 59. S. This work was supported in part by by the Packard foundation and the DOE grant DE-FGO2-91ER40654. 323 (2000) 183. Nucl. I. The contour C is now a boundary of a string worldsheet D = B 2 (times a point in S 3 ). Ooguri and Y. B 434 (1998) 36. Lett. Phys. B 445 (1998) 142. Phys. Polchinski. Schalm. 75 (1995) 4724 [hep-th/9510017]. Trivedi. 2 (1998) 231 [hep-th/9711200]. Sfetsos and K. In the large volume limit the solution corresponds to the N = 4 gauge theory on R3 × S 1 with thermal boundary conditions. N. It was pointed out in [68] that there is a subtlety with this argument. B 79 (1974) 276. [11] G. [6] A. This is an angular parameter ψ with period 2π. Sonnenschein and S. but only in the inﬁnite volume limit (R3 ). Das and S. Now W (C) = 0 and depends on the value of ψ ∈ U (1). Phys. Math. Lett. Aharony. These notes were made by selected pieces of [1].6) was suggested to be holographically dual to a theory related to pure QCD in three dimensions.J. Phys. H.

T.L. Witten. 130 (1986) 1. Phys. Phys. D 55 (1997) 5112 [hep-th/9610043].R. Phys. 2 (1998) 253 [hep-th/9802150]. Phys. G. R. D 49 (1994) 1912 [gr-qc/9307035]. Gubser. E. Introduction to the Maldacena conjecture on AdS/CFT [hep-th/9902131]. Rev.K. Townsend. Theor.R.R. G.P. J.T. P. Rev. E. G. ’t Hooft. Strominger. Rept. M. Mod. G. Phys.376 [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] Unity from Duality: Gravity. Gubser and I. Nucl. I. W.D. Nucl. Klebanov and A. Lett. D.S. B 428 (1998) 105 [hep-th/9802109]. J. Vecchia.R. 71 (1993) 3754 [hep-th/9307049]. Phys. Phys. Lett. 36 (1995) 6377 [hep-th/9409089]. M. Salam and E. Phys. van Nieuwenhuizen and N. Rev. B 360 (1991) 197. Lett. Gauge Theory and Strings A. Gunaydin. Phys. Two lectures on the AdS/CFT correspondence [hep-th/9902022]. Polchinski.S. Phys.W.R. An Introduction to AdS/CFT correspondence [hep-th/9903007]. G. Fischler. Shenker and L.H. I. Klebanov. Horowitz and A. Sezgin.M. C. N. Douglas and S. Phys. Klebanov. B.W. G.D. Strominger and C. Bekenstein. Commun. Strominger.K. Leigh. Seiberg.R. Susskind.J. Nucl. Seiberg. J. private communication (1999). Elitzur. B 413 (1997) 41 [hep-th/9708005]. D 55 (1997) 861 [hep-th/9609026]. Witten. Nucl.N. Phys. Phys. I. S. Nilsson. S. Gunaydin and N. Phys. J. K. S. Callan and J. Klebanov. Math. .G. Gubser.M. S. Nucl. S. B 388 (1996) 51 [hep-th/9605234]. 1989). B 472 (1996) 591 [hep-th/9602043]. S. van Nieuwenhuizen and P. Horowitz and A. Nucl. G. Vafa.K. Randjbar-Daemi. Missing Math. Klebanov and A. Lett.E. Dimensional Reduction in Quantum Gravity [gr-qc/9310026]. Duﬀ. Nucl. Amsterdam. B 379 (1996) 99 [hep-th/9601029]. Nilsson and C.G.E. M. Phys. Netherlands. A.T. Pope. 20 (1974) 194. Fronsdal. B 176 (1986) 45. Witten. Polyakov.P. Phys. Polyakov. B 255 (1985) 63. Mandal and S. B 478 (1996) 561 [hep-th/9606185]. JHEP 12 (1998) 012 [hep-th/9812012].A. Adv. Das and S. A. Phys. Townsend.Z. String Theory (Cambridge University Press. Phys. E. Strominger. C. Lett. JETP Lett.M. Nucl. From three-branes to large N gauge theories [hep-th/9901018]. E. Warner. A 4 (1989) 2767. Phys. Witten. Lett. Banks. Math. Lett. 121 (1989) 351. A. B 496 (1997) 231 [hep-th/9702076]. B 272 (1986) 99. Dhar. Phys. D 26 (1982) 1988. Maldacena and A. 77 (1996) 2368 [hep-th/9602051].R. Townsend. Petersen. Phys. Gibbons and P. M. M. Schwimmer and N. B 237 (1984) 342. Freedman and H. Nucl. Nicolai. Rev. Rev. Nucl. B. B 460 (1996) 335-350 [hep-th/9510135]. 1 and 2 (North-Holland. I. B 242 (1984) 377. B 499 (1997) 217 [hep-th/9703040]. Supergravities in Diverse Dimensions. P. Susskind. Rev.D. Pilch. Warner. Wadia. Mathur. Maldacena. Phys. Vols. Sierra and P. Nucl. Phys. Phys. Moore.R. A. Lett. L. Phys.S. B 326 (1989) 108. Phys.W. Phys. Gunaydin. J. Tseytlin. P. 1998).

Kol. Nucl. [64] S. 2 (1998) 505 [hep-th/9803131].H. Kabat. Nucl. Sasakura. Phys. Li.dimensions [hep-th/9604041]. Ferrara. JHEP 02 (1999) 012 [hep-th/9901004]. Phys.R. B 435 (1998) 337 [hep-th/9805131]. Dine and N. [52] M. Phys. Townsend. Andrianopoli and S. B 153 (1985) 392. Bergman. B 438 (1995) 109 [hep-th/9410167]. [62] S. Witten. Sezgin and Y. Nucl. Susskind and E. Seiberg and E. [67] O. Lett. Tanii. Salam.J.B. [68] E. B 525 (1998) 104 [hep-th/9712211]. Class. [50] E. [73] T. Gunaydin. [57] N. [74] O. [63] A. Math. [71] O. Hata and N.R. Adv. Bagger and J.R. [56] M. [54] A. Douglas. Maldacena: Lectures on Large N Field Theories and Gravity 377 [49] E. Rev. Tong. Nucl. [58] M. [69] F. Minic and M. Shenker. Lett. Das. [59] M. Princeton Series in Physics (Princeton University Press. Phys.J.A. P. Phys. Phys. Lett. . 2 (1985) L11. Sasakura and S. D. Rocek. D-brane realization of N=2 superYang-Mills theory in four. B 224 (1989) 71. Zagermann. JHEP 11 (1998) 018 [hep-th/9807205]. Hashimoto. B 409 (1997) 239 [hep-th/9705057]. B 430 (1998) 248 [hep-th/9803171]. Lett. Bergshoeﬀ.R. Nucl. [80] M. Das. Taylor. Phys.R. M. Marcus. Novel supermultiplets of SU (2. B 431 (1998) 303 [hep-th/9803127]. Phys. Sezgin. Phys. Bergshoeﬀ. B. Kawano and K. Phys. Phys. Douglas. Sugimoto. Duﬀ. 1992). Sasakura. [78] D. Hull and P. B 536 (1998) 149 [hep-th/9804160]. [75] K. Kulik. Gonzalez-Rey. B 544 (1999) 218 [hep-th/9810152]. Grav. Lett. Rev. A. Branes within branes [hep-th/9512077]. Green and M. [61] S.N. B 535 (1998) 83 [hep-th/9804164]. Holograms of branes in the bulk and acceleration terms in SYM eﬀective action [hep-th/9905037]. Lett.K. H. Witten. Tseytlin. B 432 (1998) 338 [hep-th/9804139]. Supersymmetry and Supergravity. Pope and E. Lett. D.R. Lett. Tanii. Nucl. Romans. Witten. Phys. Salam. Phys. [76] K. M theory description of 1/4 BPS states in N=4 supersymmetric Yang-Mills theory [hep-th/9811087]. Quant.J. A. Phys. Nucl. Park and M. Gubser. Gunaydin and N. Phys. [66] L.S. Lett. Theor. Nucl. B 205 (1988) 237. [72] K. Hata and N. I. [60] M. Princeton. Bergshoeﬀ. Yi. Pouliot and S. Douglas and M. [51] E. H. [65] L. D 59 (1999) 025006 [hep-th/9807164]. Nucl. [77] N. [53] C. Seiberg. Gutperle. Witten. Phys. Hashimoto. the Eleven-Dimensional Supermembrane. Bergman and B. Sezgin and Y. Wess. [81] L. B 305 (1988) 497. B 377 (1996) 28 [hep-th/9602077]. Phys. Phys. C. [55] M. B 485 (1997) 85 [hep-th/9608024].Y. Kehagias. A Note on 1/4 BPS states [hep-th/9902005]. B 469 (1996) 51 [hep-th/9602064]. E.M. B 431 (1994) 484 [hep-th/9408099]. E. D 58 (1998) 066005 [hep-th/9804174]. Okuyama. [79] M. 2|4) and the AdS5 /CFT4 duality [hep-th/9810226]. [70] J. Aharony and E. Branes in the bulk of Anti-de Sitter space [hep-th/9807225]. Phys. Lee and P. The Holographic bound in anti-de Sitter space [hep-th/9805114]. Phys. Douglas and W. Phys.

[101] S. Phys. Schalm and R. Phys. Berkooz. Diﬀerential Geometrical Methods In Mathematical Physics. Nucl. Lett. [85] L. Class. 597 [97] V. Adv. [104] P. Nucl. Zagermann. Rev. Nastase. Math. [109] D. [94] S. Class. Fronsdal. Nucl. Correlation functions in the CFTd /AdSd+1 correspondence [hep-th/9804058]. Lett. Phys. [108] S. Phys.Z. L. Lledo and A. Phys. Freedman. 2 (1985) L19. Kutasov and N. Ferrara and A. [107] I. Lett. D 23 (1981) 1278. E. 231 (1994) 311 [hep-th/9307010].B. Phys. Rev. [114] D. Mathur. [92] M. Nucl. Kac. B 536 (1998) 199 [hep-th/9807080]. 46 (1998) 265 [hep-th/9807150]. D 32 (1985) 389. Suppl. [110] G. Quant. B 483 (1997) 431 [hep-th/9605009].J. Matusis and L. M. Dobrev and V. Phys. Phys. Schwimmer. Phys. On Short and Long SU (2. Phys. Minic. Zaﬀaroni. [111] S. Gunaydin and N. D. Bulk gauge ﬁelds in AdS supergravity and supersingletons [hep-th/9807090]. Osborn and A. 2/4) Multiplets in the AdS/CFT Correspondence [hep-th/9812067]. Lett. Phys. D.B. Osborn. Fronsdal. [90] H. [83] S. B 491 (1997) 221 [hep-th/9601023]. Phys. Phys. Lett.J. Lett.K. Fronsdal.G. Aharony. Lett.S. Andrianopoli and S.G. Gauge Theory and Strings [82] L. Ann. Rev.A. Dobrev and V. Ferrara and C. [96] V. Phys. Klebanov and E. H. Angelopoulos. Rastelli. [98] V. Flato. Fronsdal. Math. Gimon and D. Petkos. J. Phys. Duﬀ. Nucl. [115] J. [99] N. Flato. JHEP 07 (1998) 006 [hep-th/9805112]. Flato and C. M.J. B 309 (1993) 279 [hep-th/9302047]. M. [91] M. Flato and C. Phys. Phys. Grav. B 97 (1980) 236. Math. van Nieuwenhuizen. [113] H. [88] M. Nucl. D 58 (1998) 105029 [hep-th/9805082]. Zaﬀaroni. Deser and A. . Fronsdal and M. Witten. Sternheimer. Theor. 2 (1978) 421. Ferrara and C. Fronsdal and D. [103] H. JHEP 10 (1998) 004 [hep-th/9808149]. Grisaru and A. 22 (1981) 1100. Lett. B 143 (1984) 389. Romans and P. C. B 534 (1998) 96 [hep-th/9806042]. Ferrara. Math. Sezgin. Nicolai and E. Ferrara.D. Seiberg. Fronsdal and A.378 Unity from Duality: Gravity. Minic and M. 35 (1987) 537. [102] S. Ferrara. Petkova. Phys. [86] M. 2 (1998) 781 [hep-th/9712074]. Seiberg. Class. Berglund. Quant. A. B 162 (1985) 127. Phys. K. Fortschr. [95] S. Flato and C. 67 (1998) 158 [hep-th/9705117]. Rev. The AdS/CFT correspondence and spectrum generating algebras [hep-th/9905097]. Math. Grav. Gubser and I. Phys. D 58 (1998) 125025 [hep-th/9808075]. Quant. [100] S. [87] C. [89] E. B 532 (1998) 153 [hep-th/9802203]. Phys. Phys. [106] E. Grav. S. Phys. Chalmers.Ii (Berlin 1977). [93] S. Witten. 11 (1994) 1387 [hep-th/9308075]. 15 (1998) 2153 [hep-th/9712239]. Ferrara.R. Representations of classical Lie superalgebras. Erdmenger and H. C. [105] O. 46 (1998) 157 [hep-th/9806072]. Fronsdal. [112] M. Johansen. Phys.K. Proceedings. Nucl. Lett.R. Fronsdal. Anselmi. Zaﬀaroni. Gunaydin. Kim. Klebanov. B 172 (1986) 412. B 433 (1998) 19 [hep-th/9802126]. Siebelink. M. B 540 (1999) 247 [hep-th/9805105]. Minwalla. Ferrara and C. Petkova. Andrianopoli and S. Marcus. [84] M. Proc.

D. ﬁeld theories in anti-de Sitter space and singletons [hep-th/9803028]. I. [138] W.Z.M. Johnson and J. [119] V. Ghezelbash. [137] G. [136] E. Subleading corrections and central charges in the AdS/CFT correspondence [hep-th/9812092]. M. Phys.Z. Nucl. Theisen and S. Counterterms for the Dirichlet prescription of the AdS/CFT correspondence [hep-th/9905046]. Adv. [120] W. Freedman. Rev. Schalm. Aref’eva and I. [135] D. B 533 (1998) 88 [hep-th/9804083]. Latorre. Lett. Chalmers and K. [133] H.D. Kaviani. Rev. Rev. [118] M.S. Henningson and K. K. Freedman. M.S. Minces and V. Theor. Mueck and K. Phys. Henningson and K. [140] P. D 58 (1998) 041901 [hep-th/9804035]. 2 (1999) 697 [hep-th/9806074].A. B 431 (1998) 63 [hep-th/9803251]. Three-Point Green Function of the Stress-Energy Tensor in the AdS/CFT Correspondence [hep-th/9901121]. Nucl.D. B 371 (1992) 353. D 59 (1999) 045008 [hep-th/9807098]. General scalar exchange in AdSd+1 [hep-th/9811257]. Phys. Viswanathan. Viswanathan. Gauge boson exchange in AdSd+1 [hep-th/9809179]. Blau. Seiberg. Liu and A. JHEP 07 (1998) 023 [hep-th/9806087]. Phys. [125] S. Lett. D. [139] E. Minwalla. K.W. The Large Nc Limit of Four-Point Functions in N = 4 Super Yang-Mills Theory from Anti-de Sitter Supergravity [hep-th/9810051]. Chern-Simons theories in the AdS/CFT correspondence [hep-th/9902123]. Nucl. [131] S. Freedman. Tseytlin. [128] W. Muck and K. Phys. Nojiri and S. Anselmi and A. On subleading contributions to the AdS/CFT trace anomaly [hep-th/9904179]. Comments on 4 point functions in the CFT/AdS correspondence [hep-th/9808006].S. Freedman and W. Phys. Aharony. J. Liu and A. Holography and the Weyl anomaly [hep-th/9812032]. [121] S. A Note on Anomalies in the AdS/CFT Correspondence [hep-th/9901134]. Shahrokh AF Tehran Parvizi and A. Muck and K. Hawking. Phys. D 15 (1977) 2752. Skenderis. B 526 (1998) 543 [hep-th/9708042]. [117] M. Skenderis. Rivelles. [132] S. [123] O. Sfetsos.A.Z. Henningson and K. [124] M. A Stress Tensor for Anti-de Sitter Gravity [hep-th/9902121]. Pawelczyk. Anselmi. Phys.T. Phys. [122] D. [144] A. Nojiri and S. Rangamani and N. D’Hoker and D. [130] W.O. Skiba. Tseytlin. Freedman. . Solodukhin. Gibbons and S.S.S.Y. [129] H. On the conformal anomaly from higher derivative gravity in AdS/CFT correspondence [hep-th/9903033].Z. [143] M. [127] I. Gava. Muck and K.Z. Odintsov. Yankielowicz. Odintsov. Lee. [126] G. S. On large N conformal theories. D’Hoker and D. [134] E. Freedman. B 444 (1998) 92 [hep-th/9810008].N. A.H.D. Math. Phys. The Graviton in the AdS-CFT correspondence: Solution via the Dirichlet boundary value problem [hep-th/9810151]. Maldacena: Lectures on Large N Field Theories and Gravity 379 [116] D. Nucl. Phys.I. D 58 (1998) 106006 [hep-th/9805145]. Arutyunov and S. D’Hoker. Johansen. K. D 59 (1999) 086002 [hep-th/9807097]. [141] G. [142] D. Frolov. Phys. Viswanathan.Z. Narain and E. Phys.V. B 539 (1999) 403 [hep-th/9806004]. B 435 (1998) 291 [hep-th/9805162]. Rev. Fatollahi.J. S. Viswanathan. Kehagias. Rev. Lett. Volovich. Matusis and L. Kraus. Mathur. Rastelli. S.A. Grisaru and A. Balasubramanian and P.W.

S. Volovich. Nucl. Naik. [167] S. Note on the boundary terms in AdS/CFT correspondence for Rarita-Schwinger ﬁeld [hep-th/9904098]. K. Renormalization group ﬂows from holography supersymmetry and a c theorem [hep-th/9904017]. D. [169] G. B 441 (1998) 173 [hep-th/9807046]. B 450 (1999) 368 [hep-th/9812238]. D’Hoker. Schwarz. JHEP 09 (1998) 022 [hep-th/9809009]. Chalmers and K. B 448 (1999) 218 [hep-th/9811097]. [175] I. Mathur. B 442 (1998) 136 [hep-th/9807223].A. Rastelli. JHEP 06 (1998) 005 [hep-th/9805104]. D’Hoker.Z. Sorokin.380 Unity from Duality: Gravity. . Stringy corrections to the Wilson loop in N=4 superYang. Schalm.A.S. [151] A. 80 (1998) 4859 [hep-th/9803002].A. Quant.P. [148] A. [153] A. Freedman. Lechner and D. Frolov. Arutyunov and S.J. Scattering in anti-de Sitter space and operator product expansion [hep-th/9811152]. Phys. Mathur. [165] D. 2 (1998) 559 [hep-th/9803111]. Grav. l’Yi. [152] R. D. Frolov. [159] P. Gauge Theory and Strings [145] G. Lett. Tonin. Warner. [162] H. JHEP 07 (1998) 017 [hep-th/9806140]. Gross and H. S.S. Nucl. Minahan. Lett. Greensite and P. Class. Maldacena.C. Gross and H. Freedman and L. Ooguri. [155] G.Mills theory [hep-th/9903042]. [160] E.S. [173] J. [170] S. Matusis and L. Holographic normal ordering and multiparticle states in the AdS/CFT correspondence [hep-th/9901144]. Phys. Lechner and M.I. [149] W. Pilch and N. [172] J. Solovev. Frolov. Math. Nucl. ’t Hooft.H. Ooguri. Koshelev and O. Graviton exchange and complete four point functions in the AdS/CFT correspondence [hep-th/9903196]. [156] G. Rev. A. K. Phys. 14 (1997) L195 [hep-th/9707044]. Phys. Rastelli. Lett. Phys. Dall’Agata. [174] J. Arutyunov and S. AdS/CFT four point functions: How to succeed at Z integrals without really trying [hep-th/9905049]. Yee. Nucl. Gubser. Arutyunov and S. Liu. Phys.A. Adv.S. Olesen. Graviton and gauge boson propagators in AdS(d+1) [hep-th/9902042].E.E. Freedman. Drukker. Phys. Macroscopic strings as heavy quarks in large N gauge theory and anti-de Sitter supergravity [hep-th/9803001]. Lett.J. Rev. Phys. B 544 (1999) 576 [hep-th/9806216]. Kogan and O.Z. B 226 (1983) 269. D 58 (1998) 106002 [hep-th/9805129]. [158] J. Phys.E.A. [166] J. Freedman. [164] E. Massive symmetric tensor ﬁeld on AdS [hep-th/9905048]. S. Improved heavy quark potential at ﬁnite temperature from Anti-de Sitter supergravity [hep-th/9904147]. West.D. Ghoshal and S. Generating functionals of correlation functions of p form currents in AdS/CFT correspondence [hep-th/9809132]. Quadratic action for Type IIB supergravity on AdS5 × S 5 [hep-th/9811106]. Rashkov.C. l’Yi. Howe and P. [147] W.A.D. D’Hoker. Dall’Agata. [171] S. D. D. [146] G.P. Lett. B 153 (1979) 141. Minahan and N. K. [157] G. l’Yi. Forste. Rastelli. [154] D. A. [168] N.Z.S.A. Matusis and L. D. B 238 (1984) 181. Rytchkov. S. Polishchuk.-J. [161] E. Phys. Warner. Worldsheet Fluctuations and the Heavy Quark Potential in the AdS/CFT Approach [hep-th/9901057]. [163] G. Theisen.Z. Phys. Wilson Loops and Minimal Surfaces [hep-th/9904191]. Theor. Rey and J. Holographic projection of massive vector ﬁelds in AdS/CFT correspondence [hep-th/9808051]. [150] W.

Maldacena: Lectures on Large N Field Theories and Gravity 381 [176] I. B 541 (1999) 145 [hep-th/9809032]. Hunter and M. [183] A. [204] A. Li. Phases of R charged black holes. Landsteiner. Running gauge coupling and quark . Fischler and J. [201] D.R.M. [184] S. Thermodynamic stability and phases of general spinning branes [hep-th/9903132]. [205] S.W. Taylor-Robinson.C. Fotopoulos and T. Wilson loop correlator in the AdS/CFT correspondence [hep-th/9904149]. Hawking and D. Class. 16 (1999) 1197 [hep-th/9808032]. [189] M. Tseytlin and S.Mills theory at ﬁnite temperature [hep-th/9811224].J. Rev. Page. Phys. Math. Nucl. [194] R. Graham and E.R. Comment on two loop free energy in N=4 supersymmetric Yang. [197] M. Taylor. Chamblin. spinning branes and strongly coupled gauge theories [hep-th/9902195]. Phys. Hawking. Trivedi. B 545 (1999) 217 [hep-th/9806075].A.S. Lett. Phys. Phys.-J. Cvetic and S. Caldarelli and D. Critical behavior in the rotating D-branes [hep-th/9812121]. Gubser.S. Tseytlin. Berenstein. Gomez and T.S. Myers. Grav. B 534 (1998) 202 [hep-th/9805156]. [198] M. I.W. Peet and S. D 54 (1996) 3915 [hep-th/9602135].-G. . Soh. Rev. Witten.V. Wald. D 59 (1999) 064005 [hep-th/9811056]. Conformal anomaly of submanifold observables in AdS/CFT correspondence [hep-th/9901021].anti-quark potential from dilatonic gravity [hep-th/9904036]. Polchinski. Peet.J.I. D 55 (1997) 6189 [hep-th/9612146]. [203] S. Emparan. [179] E. Solovev. R. Yankielowicz. Kogan and O.R. I. Mod. A Comment on the entropy of strongly coupled N=4 [hep-th/9904035]. Birmingham.W. Nucl. Rey. Odintsov. Thermodynamics of spinning D3-branes [hep-th/9810225].W. Gubser.S. JHEP 03 (1999) 003 [hep-th/9811120]. [195] M. Cai and K.N. [178] K. [177] S. [193] K. [188] R. Nucl. Cvetic and S. Corrado. Marolf.M.D. JHEP 12 (1998) 020 [hep-th/9810200]. Phys. [186] N. [199] P. Phys. Nucl. Rev. A 14 (1999) 379 [hep-th/9901143]. 87 (1983) 577. Phys. B 527 (1998) 171 [hep-th/9803135]. S.M. Alvarez. C. C. Rev. Strominger. [202] J.R. Phys. On zigzag invariant strings [hep-th/9901131]. [181] D. Larsen and S. Ortin. Gubser.-S.F. Quant. [190] Y. Li. [185] G. Kraus. Rev.T. F. Klebanov and A. Maldacena. D 59 (1999) 066002 [hep-th/9808081]. Louko and D. W.A. [182] S. C. Charged AdS Black Holes and Catastrophic Holography [hep-th/9902170]. Yee. JHEP 03 (1999) 004 [hep-th/9807196]. Ross. Phys.S. M theory and stringy corrections to Anti-de Sitter black holes and conformal ﬁeld theories [hep-th/9903078].P. D 48 (1993) 3427 [gr-qc/9307038]. [192] S. [187] A. [191] S. Gubser. Large N strong/weak coupling phase transition and the correspondence principle [hep-th/9810053].Commun. Klebanov and A.A. Horowitz and J. [180] C. Johnson and R. The Operator Product Expansion for Wilson Loops and Surfaces in the Large N Limit [hep-th/9809188]. unpublished notes (1997). Nojiri and S. [196] A. Klemm. Zarembo. Phys. [200] A. R. Phys. Theisen and J. Itzhaki. Gubser. hong Gao and M.

.A.S.R. KLEBANOV Department of Physics. NJ 08544.LECTURE 5 D-BRANES ON THE CONIFOLD AND N = 1 GAUGE/GRAVITY DUALITIES I. U. Princeton University Princeton.

. .3 Chiral symmetry breaking and gluino condensation . . . . . . . 6. . . . . . .2 SO(4) invariant expressions for the 3-forms . . . . . . . . . . . . . . . . . . . . . . . . . . 400 4 The chiral anomaly 402 4.Contents 1 Introduction 385 2 D3-branes on the conifold 388 2. . . . . . . . 391 2.3 Other ways of wrapping D-branes over cycles of T 1. . . . .1 The anomaly as a classical eﬀect in supergravity . . . . . . . . . . . . . . . . . . . . . . . . . . 393 2. . .1 Dimensions of chiral operators . . . . .1 Matching of the β-functions . 6. . . . . . 413 6 Infrared physics 6. . . . 394 3 The RG cascade 397 3. . . . . . . . . .1 . . . .2 The anomaly as spontaneous symmetry breaking in AdS5 . . . . . 414 414 416 418 . . . . .1 The ﬁrst-order equations and their solution . . . . . . . . 412 5. . . . . . . . . . . 403 4. . . . . . . . . .1 Dimensional transmutation and conﬁnement . .2 Tensions of the q-strings . . .2 Wrapped D3-branes as “dibaryons” . . . 405 5 Deformation of (KS) the conifold 410 5. . . . .

We consider ﬁrst the conformal case. In particular. These notes are primarily devoted to extensions of the AdS/CFT correspondence to theories with N = 1 supersymmetry. 1 Introduction Comparison of a stack of D3-branes with the geometry it produces leads to a formulation of duality between N = 4 supersymmetric Yang-Mills theory and type II strings on AdS5 × S5 [1–3]. Herzog and P. Next. Ouyang Abstract We review extensions of the AdS/CFT correspondence to gauge/ gravity dualities with N = 1 supersymmetry. changing the gauge group and introducing a logarithmic renormalization group ﬂow. This may be accomplished through placing a stack of D3-branes at the apex of a Ricci ﬂat 6-dimensional cone [4–7]. we break the conformal symmetry by adding a stack of partially wrapped D5-branes to the system. the eﬀect of these wrapped D5-branes is to turn on the ﬂux of 3-form ﬁeld strengths. Klebanov. with discussions of chiral primary operators and wrapped D-branes. C. The associated RR 2-form potential breaks the U (1) R-symmetry to 2M and we study this phenomenon in detail. Then we show how to break the conformal invariance in this set-up and to introduce logarithmic RG ﬂow into the ﬁeld theory.P. This extra ﬂux also leads to deformation of the cone near the apex.D-BRANES ON THE CONIFOLD AND N = 1 GAUGE/GRAVITY DUALITIES I. In the gravity dual. It is of obvious interest to consider more general dualities between gauge theories and string theories where some of the supersymmetry and/or conformal invariance are broken. Springer-Verlag 2002 . we describe the gauge/gravity dualities that emerge from placing D3-branes at the apex of the conifold.R. We ﬁrst show how to break some of the supersymmetry without destroying conformal invariance. A convenient way to make the coupling constants run logarithmically is to introduce fractional D3-branes c EDP Sciences. which describes the chiral symmetry breaking and conﬁnement in the dual gauge theory.

(1. we consider primarily one particular example of a cone. the duality cascade in the UV. which describes the chiral symmetry breaking and conﬁnement in the dual gauge theory [11]. Gauge Theory and Strings at the apex of the cone [8–10]. where the conifold is replaced with R6 . the review papers [12. At the same time. In the gravity dual the eﬀect of these wrapped D5-branes is to turn on the ﬂux of 3-form ﬁeld strengths. a solution of type IIB supergravity that is dual to a certain N = 1 supersymmetric SU (N + M ) × SU (N ) gauge theory in the limit of strong ’t Hooft coupling. the conifold. We will start the notes with a very brief review of some of the basic facts about the AdS/CFT correspondence. The conifold has enough structure that many new aspects of AdS/CFT correspondence emerge that are not immediately visible for the simplest case. and the selfdual 5-form ﬁeld strength is given by F5 = F5 + F5 . Note that the dilaton. On the other hand.2) r4 This 10-dimensional metric may be thought of as a “warped product” of the R3.1 along the branes and the transverse space R6 . 1 2 3 5 where dΩ2 5 is the metric of a unit 5-sphere and h(r) = 1 + (1.386 Unity from Duality: Gravity. This solution encodes various interesting gauge theory phenomena in a dual geometrical language. There are two reasons for this focus. the curved background produced by the stack is ds2 = h−1/2 −dt2 + dx2 + dx2 + dx2 + h1/2 dr2 + r2 dΩ2 . To make the discussion more concrete. This program eventually leads to the warped deformed conifold [11]. and chiral symmetry breaking and conﬁnement in the IR. for example. The SYM theory is the low-energy limit of the gauge theory on the stack of D3-branes. F5 = 16π(α )2 N vol(S5 ). For more background the reader may consult.4) . however. The duality between N = 4 supersymmetric SU (N ) gauge theory and the AdS5 × S5 background of type IIB string theory [1–3] is usually motivated by considering a stack of a large number N of D3-branes. is constant.13]. the conifold is simple enough that we can follow the program outlined in the paragraph above in great detail. First. the logarithmic running of couplings.3) The normalization above is dictated by the quantization of Dp-brane tension which implies S8−p Fp+2 = 2κ2 τp N . such as the chiral anomaly. This extra ﬂux may lead to deformation of the cone near the apex. we review the original AdS/CFT correspondence. Φ = 0.1) L4 · (1. gs (1. these fractional branes may be thought of as D5-branes wrapped over 2-cycles in the base of the cone.

6) S5 which is consistent with (1. Note that the 5-form ﬁeld strength may also be written as gs F5 = d4 x ∧ dh−1 − r5 dh vol(S5 ). 2 (1.R.3) since the volume of a unit 5-sphere is Vol(S5 ) = π 3 . The low-energy limit corresponds to r → 0. The radial coordinate r is related to the scale in the dual gauge theory.10) in agreement with the preceding paragraph. (1. Klebanov et al. (1. for p = 3 we have F5 = (4π 2 α )2 N.3) that dr L4 = 4πgs N α .11) . dr (1.9) L Vol(S ) = 2 κ κ This way we ﬁnd L4 = κN 2 = 4πgs N α 2π 5/2 (1.I. 5 z2 (1. In particular.7) Then it is not hard to see that the Einstein equation RMN = 2 gs FMPQRS FN PQRS 96 is satisﬁed.8) A related way to determine the scale factor L is to equate the ADM tension of the supergravity solution with N times the tension of a single D3-brane [14]: √ 2 4 π 5 N.5) and κ = 8π 7/2 gs α is the 10-dimensional gravitational constant. Since −r5 dh = 4L4 .: N = 1 Gauge/Gravity Dualities where √ π τp = (4π 2 α )(3−p)/2 κ 2 387 (1. In this limit the metric becomes ds2 = L2 −dt2 + dx2 + dz 2 + L2 dΩ2 . we ﬁnd by comparing with (1.

This describes the direct product of 5-dimensional Anti-de r Sitter space.1 .12) L = 2Vol(X5 ) Vol(X5 ) an important normalization formula which we will use in the following section. The dual gauge theory is the IR limit of the world volume theory on a stack of N D3-branes placed at the orbifold singularity of R6 /Γ. 18]. An interesting generalization of the basic AdS/CFT correspondence [1–3] is found by studying branes at conical singularities [4–7]. The simplest examples of X5 are the orbifolds S5 /Γ where Γ is a discrete subgroup of SO(6) [4]. AdS5 . 2 D3-branes on the conifold The conifold may be described by the following equation in four complex variables. Constructions of the dual gauge theories for Einstein manifolds X5 which are not locally equivalent to S5 are also possible. In these cases X5 has the local geometry of a 5-sphere. Consider a stack of D3-branes placed at the apex of a Ricci-ﬂat 6-d cone Y6 whose base is a 5-d Einstein manifold X5 . and the 5-dimensional sphere.1) Since this equation is invariant under an overall real rescaling of the coordinates. 6 T (2. Comparing the metric with the D-brane description leads one to conjecture that type IIB string theory on AdS5 × X5 is dual to the low-energy limit of the world volume theory on the D3-branes at the singularity.17]. the base of this cone is precisely the space T 1. this space is a cone. In fact. with equal radii of curvature L. The simplest example is the Romans compactiﬁcation on X5 = T 1. 4 a=1 2 za = 0.2) .388 Unity from Duality: Gravity. Gauge Theory and Strings 2 where z = L .1 [6. Let us explain this connection in more detail. The dual gauge theory is the conformal limit of the world volume theory on a stack of N D3-branes placed at the singularity of a Calabi-Yau manifold known as the conifold [6]. Such theories typically involve product gauge groups SU (N )k coupled to matter in bifundamental representations [16]. the metric on the conifold may be cast in the form [18] ds2 = dr2 + r2 ds2 1. S5 .1 . (1. The equality of tensions now requires [15] √ πκN π3 2 4 = 4πgs N α .1 = (SU (2)× SU (2))/U (1) [6. (2. Remarkably. which is a cone over T 1.

1 is 16π [8]. in the (N. (2.6) where σ n are the Pauli matrices for n = 1.R. 3 and σ 4 is i times the unit matrix.12) it then follows 27 that L4 = 4πgs N (α )2 27κN 27 = · 16 32π 5/2 (2. The A’s transform as a doublet under one of the global SU (2)s while the B’s transform as a doublet under the other SU (2). Using the metric (2. Bj is to rewrite the deﬁning equation of the conifold. N) representation.I. Here ψ is an angular coordinate which ranges from 0 to 4π.1 . From (1.j 1 zij = √ 2 n σij zn n (2. A simple way to motivate the appearance of the ﬁelds Ai . This ﬁeld theory was constructed in [6]: it is SU (N ) × SU (N ) gauge theory coupled to two chiral superﬁelds.: N = 1 Gauge/Gravity Dualities where ds2 1. Therefore. in the (N.1 . (2. φ1 ) and (θ2 .3) it is not 3 hard to ﬁnd that the volume of T 1.5) The same logic that leads us to the maximally supersymmetric version of the AdS/CFT correspondence now shows that the type IIB string theory on this space should be dual to the infrared limit of the ﬁeld theory on N D3-branes placed at the singularity of the conifold.3) is the metric on T 1. Now placing N D3-branes at the apex of the cone we ﬁnd the metric ds2 = 1+ L4 r4 −1/2 −dt2 + dx2 + dx2 + dx2 1 2 3 dr2 + r2 ds2 1. Bj .1 = T 1 cos θi dφi dψ + 9 i=1 2 2 389 + 1 6 2 i=1 2 dθi + sin2 θi dφ2 i (2. as det zij = 0.7) .1 T (2.1 is an S1 bundle over S2 × S2 . Klebanov et al. i. This quadratic constraint may be “solved” by the substitution zij = Ai Bj . Ai . φ2 ) parametrize two S2 s in a standard way.4) + 1+ L4 r4 1/2 whose near-horizon limit is AdS5 × T 1. N) representation and two chiral superﬁelds.1). Since Calabi-Yau spaces preserve 1/4 of the original supersymmetries we ﬁnd that this should be an N = 1 superconformal ﬁeld theory. while (θ1 . this form of the metric shows that T 1. 2.

let us add to the superpotential of this Z2 orbifold a relevant term. chiral multiplets Φ and Φ in the adjoint representations of the two U (N )’s. . it was proposed in [6] that the SU (N ) × SU (N ) SCFT with this superpotential is dual to type IIB strings on AdS5 × T 1. N) ⊕ (N. this model has been identiﬁed [4] as an AdS dual of a U (N ) × U (N ) theory with hypermultiplets transforming in (N. −1) and B1 .1 . The importance of this choice is that this particular Z2 orbifold of AdS5 × S5 has N = 2 superconformal symmetry. i (2. This proposal can be checked in an interesting way by comparing to a certain AdS5 × S5 /Z2 background. the hypermultiplets correspond to chiral multiplets Ak . Bl . l = 1. A2 with charges (1.1 ). 2 in the (N. then we ﬁnd a U (1) × U (1) gauge theory coupled to ﬁelds A1 . x6 . then the Z2 acts as −1 on four of the xi and as +1 on the other two. Since a marginal superpotential has R-charge equal to 2 it must be quartic. If we place a single D3-brane at the singularity of the conifold.8) Therefore. . (2. The model also contains. For consistency of the duality it is necessary that we add an exactly marginal superpotential which preserves the SU (2) × SU (2) × U (1)R symmetry of the theory (this superpotential produces a critical line related to the radius of AdS5 × T 1. In constructing the generalization to the non-abelian theory on N D3-branes. m ˜ TrΦ2 − TrΦ2 . N) and (N. Using orbifold results for D-branes [16].10) It is straightforward to see what this does to the ﬁeld theory. From an N = 1 point of view. cancellation of the anomaly in the U (1) R-symmetry requires that the A’s and the B’s each have R-charge 1/2. 2 (2. k. . Gauge Theory and Strings where Ai . 1). . and the symmetries ﬁx it uniquely up to overall normalization: W = ij kl trAi Bk Aj Bl . N). N) representations respectively. from an N = 1 point ˜ of view. to ﬁnd the superpotential − g2 [Tr(A1 B1 A2 B2 ) − Tr(B1 A1 B2 A2 )] . Bj are unconstrained variables. Now. m .390 Unity from Duality: Gravity. If S5 is described by an equation 6 i=1 x2 = 1. The superpotential is ˜ gTrΦ(A1 B1 − A2 B2 ) + gTrΦ(B1 A1 − B2 A2 ). We simply ˜ integrate out Φ and Φ. B2 with charges (−1.9) with real variables x1 .

Thus. (For a more extensive analysis of the spectrum of the model.11) [15].1 . Using the fact that the chiral superﬁelds carry R-charge equal to 1/2.1 case [15. on the supergravity side 27 cIR Vol (S5 /Z2 ) = · = cUV Vol (T 1.1 Dimensions of chiral operators There are a number of further convincing checks of the duality between this ﬁeld theory and type IIB strings on AdS5 × T 1. these modes are mixtures of the conformal factors of the AdS5 and S5 and the 4-form ﬁeld.I. In fact. This is a striking and highly sensitive test of the N = 1 dual pair constructed in [6. 2. The perturbation in (2.10) is odd under exchange of the two U (N )’s. see [19]. The twisted sector mode which is a relevant perturbation of the ﬁeld theory is the blowup of the orbifold singularity of S5 /Z2 into the smooth space T 1. so the Z2 orbifold with relevant perturbation (2. 7].1 . It is interesting to examine how various quantities change under the RG ﬂow from the S5 /Z2 theory to the T 1.1 theory. Let us try to understand why this works from the point of view of the geometry of S5 /Z2 . The behavior of the conformal anomaly (which is equal to the U (1)3 anomaly) was studied R in [15].12) Thus.11) On the other hand. Klebanov et al. on the ﬁeld theory side it was found that 27 cIR · = cUV 32 (2. 19. Therefore we associate this perturbation with a twisted sector mode of string theory on AdS5 × S5 /Z2 . all 3-point functions calculated from supergravity on AdS5 × X5 carry normalization factor inversely proportional to Vol(X5 ). Here we discuss the supergravity modes which correspond to chiral primary operators.R.10) apparently ﬂows to the T 1.1 model associated with the conifold. A somewhat diﬀerent derivation of the ﬁeld theory on D3-branes at the conifold singularity. which is based on blowing up a Z2 × Z2 orbifold. The exchange of the two U (N )’s is the quantum symmetry of the AdS5 × S5 /Z2 orbifold – the symmetry that acts as −1 on string states in the twisted sector and +1 in the untwisted sector.: N = 1 Gauge/Gravity Dualities 391 This expression is the same as (2. was given in [7].8).) For the AdS5 × S5 case.1 ) 32 (2. we may keep the discussion of such modes quite general and consider AdS5 × X5 where X5 is any Einstein manifold. the supergravity calculation is in exact agreement with the ﬁeld theory result (2. The same has been shown to be true for the T 1. . 20].

First.. (2. It is now well-known that this family of operators with (k) dimensions ∆ = k. Aik Bjk ).20].14) We will be primarily interested in the modes which correspond to picking the minus branch: they turn out to be the chiral primary ﬁelds. However. Here there is a family of wave functions labeled by non-negative integer k. let us recall the S5 case where the spherical harmonics correspond (k) to traceless symmetric tensors of SO(6). Thus. X ik ).392 Unity from Duality: Gravity. The absence of k = 1 is related to the gauge group being SU (N ) rather than U (N ). The eigenvalues of this matrix are √ m2 = 16 + E ± 8 4 + E. .13) where E ≥ 0 is the eigenvalue of the Laplacian on X5 . . Gauge Theory and Strings The diagonalization of such modes carried out in [22] for the S5 case is easily generalized to any X5 .17) Since the F -term constraints in the gauge theory require that the i and the j indices are separately symmetrized.16) In [6] it was argued that the dual chiral operators are tr(Ai1 Bj1 .1 . For such modes there is a possibility of m2 falling in the range −4 < m2 < −3 (2.ik . and it seems that the bound (2. di1 . . The situation is diﬀerent for T 1.. Thus. The corresponding eigenvalues of the Laplacian are E(k) = 3 k(k + 2) − k2 4 · (2. this is precisely the special case where the corresponding mode is missing: for k = 1 one of the two mixtures is the singleton [22].19. all chiral primary operators in the N = 4 SU (N ) theory correspond to the conventional branch of dimension. and with U (1)R charge k [15..15) is satisﬁed for k = 1. k = 2. The mixing of the conformal factor and 4-form modes results in the following mass-squared matrix. (2. . k/2). in this case we do not encounter operator dimensions lower than 2. is di1 . we ﬁnd that their SU (2)×SU (2)× U (1) quantum numbers agree with those given by the supergravity analysis. Here E = k(k + 4). ∆+ . transforming under SU (2) × SU (2) as (k/2. .15) where there is a two-fold ambiguity in deﬁning the corresponding operator dimension [21]. . . 3. .ik Tr(X i1 . m2 = E + 32 8E 4/5 E (2..

from (2. For large N the dimensions of such operators calculated from the supergravity are found to be 3N/4 [8]. Let us also note that substituting E(1) = 33/4 into (2. For example.: N = 1 Gauge/Gravity Dualities 393 In the ﬁeld theory the A’s and the B’s have dimension 3/4. 20] that the supergravity analysis based on (2. and ∆− for the other. therefore. in this case we have to pick the unconventional ∆− branch [21]. Choosing this branch for k = 1 and ∆+ for k > 1 we indeed ﬁnd following [15. Klebanov et al. the supersymmetry requires that we pick dimension ∆+ for one of the conformally coupled scalars. we need to calculate the mass of a D3-brane wrapped over a minimal volume 3-cycle.17). (2. hence the dimensions of the chiral operators are 3k/2. 19. Indeed.16) reproduces the dimensions 3k/2 of the chiral operators (2.2 Wrapped D3-branes as “dibaryons” It is of further interest to consider various branes wrapped over the cycles of T 1. In studying the dimensions from the supergravity point of view. While for k > 1 only the dimension ∆+ is admissible. √ 8π 5/2 L3 π = · (2. Thus. Thus. An example of such a 3-cycle is the subspace at a constant value of (θ2 . In fact.15).R. Since the supersymmetry requires the corresponding dimension to be 3/2.1 and attempt to identify these states in the ﬁeld theory [8]. for k = 1 one could pick either branch. To show how this works in detail. while the other corresponds to the upper component which has dimension 5/2. while ∆+ = 5/2. which has dimension 3/2. One of these scalar ﬁelds corresponds to the lower component of the superﬁeld Tr(Ai Bj ).16) we have E(1) = 33/4 which falls within the range (2. wrapped D3-branes turn out to correspond to baryon-like operators AN and B N where the indices of both SU (N ) groups are fully antisymmetrized.14) we ﬁnd m2 = −15/4 which corresponds to a conformally coupled scalar in AdS5 [22]. one encounters an interesting subtlety discussed in [21]. Here we ﬁnd that ∆− = 3/2. φ2 ).14). 2.I.18) m = V3 κ 9κ . the short chiral supermultiplet containing this scalar includes another conformally coupled scalar and a massless fermion [19]. and its volume is found to be V3 = 8π 2 L3 /9 [8]. The mass of the D3-brane wrapped over the 3-cycle is. the conifold theory provides a simple example of AdS/CFT duality where the ∆− branch has to be chosen for certain operators. This is in complete agreement with the fact that the dimension of the chiral superﬁelds at the ﬁxed point is 3/4 and may be regarded as a direct supergravity calculation of an anomalous dimension in the dual gauge theory.

βN Dl i=1 β Bkiiαi . our discussion may be generalized to wrapped (p. the dimension of the “dibaryon” operators agrees exactly with the supergravity calculation. Similarly. The tension of such a “fat” string scales as L2 /κ ∼ N (gs N )−1/2 /α ..kN i=1 Aαiiβi . φ1 ).21) Under the duality these operators map to D3-branes classically localized at a constant (θ1 . 2.5).3 Other ways of wrapping D-branes over cycles of T 1 . Let us construct the corresponding operators in the dual gauge theory. 1). we can construct color-singlet “dibaryon” operators by antisymmetrizing completely with respect to both groups: N B1l = α1 .. the collective coordinate for the wrapped D3-brane has to be quantized.. 2...20) k where Dl 1 . N B2l = α1 .. carry an index α in the N of SU (N )1 and kβ an index β in the N of SU (N )2 . At the quantum level.βN k Dl 1 . we will limit explicit statements to D5-branes: since a (p. Z) . Gauge Theory and Strings For large mL. the existence of two types of “dibaryon” operators is related on the supergravity side to the fact that the base of the U (1) bundle is S2 × S2 . the corresponding operator dimension ∆ approaches mL = 8π 5/2 L4 3 = N.1 There are many other admissible ways of wrapping branes over cycles of T 1. (2. see [23]). The most basic check on the operator identiﬁcation is that.αN β1 . N + 1). Thus the SU (2) × SU (2) quantum numbers of B1l are (N + 1.kN is the completely symmetric SU (2) Clebsch-Gordon coeﬃcient corresponding to forming the N + 1 of SU (2) out of N 2’s. q) 5-branes using the SL(2. k (2. Since the ﬁelds Aα .... k = 1.. and this explains its SU (2)×SU (2) quantum numbers [8]. The non-trivial dependence of the tension on the ’t Hooft coupling gs N indicates that such a string is not a BPS saturated object..αN k1 . which produces a string in AdS5 . since the exact dimension of the A’s and the B’s is 3/4. In discussing wrapped 5-branes. Z) transform of a D5-brane.kN β1 .. a D3-brane may be wrapped over a 2-cycle.19) where in the last step we used (2. For example. 9κ 4 (2..394 Unity from Duality: Gravity. q) 5-brane is an SL(2. This should be contrasted with the tension of a BPS string obtained in [24] by wrapping a D5-brane over RP4 : T ∼ N/α .. Thus.1 (for a complete list. we can construct “dibaryon” operators which transform as (1.

1 then. a fundamental string stretched between them is created. that the domain wall positioned at some arbitrary AdS5 radial coordinate r is not stable: its energy scales as r3 . The domain wall is tensionless there. we show how to construct the SU (N ) × SU (N + M ) theories mentioned above. the eﬀect in ﬁeld theory is to change the gauge group from SU (N ) × SU (N ) to SU (N + 1) × SU (N + 1). with its remaining two dimensions parallel to R3. correspondingly. The matter ﬁelds Ak and Bk are still bifundamentals. We will see.25]. it serves as a vertex connecting N fundamental strings. One piece of evidence for this claim is the way the D3-branes wrapped over the S3 behave when crossing the D5-brane domain wall. If a D5-brane is wrapped over an S3 .I. This is the BPS domain wall which separates adjacent inequivalent vacua distinguished by the phase of the gluino condensate.1 . N + 1). This theory is conﬁning and. Consider a D5-brane wrapped over the 2-cycle. but for deﬁniteness let us wrap the D3-brane around a particular three-sphere S3 which (1) is invariant under the group SU (2)B under which the ﬁelds Bk transform. Finally. according to the arguments in [24. Therefore. then on the other side the theory is SU (N ) × SU (N + 1) [8]. The simplest domain wall is a D3-brane which is not wrapped over the compact manifold.R. Through an analysis of the ﬁve-form ﬂux carried over directly from [24] one can conclude that when one crosses the domain wall. . however. then we ﬁnd a domain wall in the dual ﬁeld theory. then it is a domain walls in AdS5 .1 . the resulting ﬁeld theory state is a baryon built out of external quarks. If a D5-brane is wrapped over the entire T 1. The origin of this eﬀect is creation of fundamental strings by crossing D5 and D3 branes. In this case the wrapped D5-brane again falls to the minimum value of the radial coordinate. as shown in [26. Consider positioning a “fat” string made of a wrapped D3-brane orthogonally to the domain wall. 27]. that the domain wall made of a wrapped D5-brane deﬁnitely exists in the SU (N ) × SU (N + M ) generalization of the gauge theory. the only stable position is at r = 0 which is the horizon. The ﬁeld theory interpretation of a D5-brane wrapped around S2 is more interesting: if on one side of the domain wall we have the original SU (N ) × SU (N ) theory.21). ﬁlling out 2 (N. the dual background does not have a horizon. N + 1) ⊕ 2(N. We should note. Since each string ends on a charge in the fundamental representation of one of the SU (N )’s. and it is unlikely that this object really exists in the dual CFT. If this object is located at some ﬁxed r. but its tension there is non-vanishing. In homology there is only one S3 . with its remaining directions ﬁlling R3.: N = 1 Gauge/Gravity Dualities 395 symmetry of the type IIB string theory. Klebanov et al. The corresponding state in the SU (N ) × SU (N ) ﬁeld theory is B1 of (2. As the string is brought through the membrane. however.

the domain wall is not stable.. 9. In the case of the conifold.αN β1 . In [10] this solution was completed to all orders. It was further suggested that the strong dynamics of this gauge theory would resolve the naked singularity in the metric. The D5-branes wrapped over 2-cycles are examples of a more general phenomenon. i. 29]. decreases logarithmically as well. 9]..αN β1 . In [10] it was conjectured that this solution corresponds to a ﬂow in which the gauge group factors repeatedly drop in size by M units. and the relative gauge coupling runs logarithmically at all scales. one has instead α1 . The D3-brane charge. AαN β β or α1 . the 5-form ﬂux. For many singular spaces Y6 there are fractional D3-branes which can exist only within the singularity [8. The ﬂow is in fact an . Indeed. after a wrapped D3-brane has passed through the wrapped D5-brane domain wall. These fractional D3-branes are D5-branes wrapped over (collapsed) 2-cycles at the singularity. . but in fact wants to move towards r = 0.. the relative gauge coupling −2 −2 g1 − g2 runs logarithmically. until ﬁnally the gauge groups are perhaps SU (2M ) × SU (M ) or simply SU (M ). the singularity is a point. Instead. Gauge Theory and Strings In the SU (N ) × SU (N + 1) theory. Either the upper index βN +1 . The addition of M fractional branes at the singular point changes the gauge group to SU (N + M ) × SU (N ). now in the representation (N + M. it emerges with a string attached to it due to the string creation by crossing D-branes which together span 8 dimensions [26. where the supergravity equations corresponding to this situation were solved to leading order in M/N . the four chiral superﬁelds remain. . 27]. .e. the conifold suﬀers logarithmic warping. 28..22) where we have omitted SU (2) indices.. How can this be in supergravity? The answer is simple: the wrapped D3-brane must have a string attached to it. N) and its conjugate. Calculating the tension of a wrapped D5-brane as a function of r shows that it scales as r4 /L2 . This gives a well-deﬁned way of constructing the SUGRA duals of the SU (N ) × SU (N + M ) gauge theories. Hence.. the D3-brane charge eventually becomes negative and the metric becomes singular. We will assume that the wrapped D5-branes “fall” behind the horizon and are replaced by their ﬂux in the SUGRA background. the logarithm in the solution is not cut oﬀ at small radius.βN +1 1 N Aα1 . indicating a fundamental of SU (N + 1). AαN AβN +1 β β α (2.βN +1 N +1 1 N Aα1 . However. . is free..396 Unity from Duality: Gravity. indicating a fundamental of SU (N ). as pointed out in [9]. as does the superpotential [8. The theory is no longer conformal. or the upper index αN +1 ..

as in pure N = 1 Yang-Mills theory. We will show that these gauge theories have an exact anomaly-free Z2M R-symmetry. In the supergravity. In terms of this basis.1) The coeﬃcients above follow from the quantization rule (1. leading to conﬁnement [11]. a warped deformed conifold. 1 (4π 2 α )2 T 1.1 F5 = N.: N = 1 Gauge/Gravity Dualities 397 inﬁnite series of Seiberg duality transformations – a “duality cascade” – in which the number of colors repeatedly drops by M units [11].3) (g i )2 .I. Once the number of colors in the smaller gauge group is less than M.4) . which is broken dynamically. e3 ≡ cos ψ sin θ2 dφ2 − sin ψdθ2 . 2 1 e + e3 g3 = √ . the Einstein metric on T 1. e5 ≡ dψ + cos θ1 dφ1 + cos θ2 dφ2 . e2 ≡ dθ1 . e4 ≡ sin ψ sin θ2 dφ2 + cos ψdθ2 .1 . It will be useful to employ the following basis of 1-forms on the compact space [30]: e1 − e3 √ . to Z2 . the resolution of the naked singularity found in [10] occurs through the chiral symmetry breaking of the gauge theory. this occurs through the deformation of the conifold. Klebanov et al. g2 = (3.4).1 = T 1 5 2 1 (g ) + 9 6 4 i=1 e2 − e4 √ . 3 The RG cascade The addition of M fractional 3-branes (wrapped D5-branes) at the singular point changes the gauge group to SU (N + M )× SU (N ). The D5-branes serve as sources of the magnetic RR 3-form ﬂux through the S3 of T 1. 2 5 g = e5 . in addition to N units of the 5-form ﬂux: 1 4π 2 α S3 F3 = M. the supergravity dual of this ﬁeld theory involves M units of the 3-form ﬂux. 2 g1 = where e1 ≡ − sin θ1 dφ1 . 2 2 e + e4 g4 = √ .2) (3. The warped conifold (KT) solution with such ﬂuxes was constructed in [10]. (3. The resulting space. non-perturbative eﬀects become essential.R. Let us consider the eﬀect on the dual supergravity background of adding M wrapped D5-branes. Therefore.1 assumes the form ds2 1. is completely nonsingular and without a horizon. In short. (3.

Φ = 0. the 6 complex 3-form G3 satisﬁes the self-duality condition 6 G3 = iG3 .1).12) µνλ which implies that the dilaton is constant.10) where 6 is the Hodge dual with respect to the metric ds2 . gs (3. (3. the RR scalar vanishes as well. . 2 B2 = 3gs M α ω2 ln(r/r0 ). Gauge Theory and Strings Keeping track of the normalization factors.10) that 2 2 2 gs F3 = H3 . the quantization condition for RR 3-form ﬂux is obeyed. (2. It follows from (3.9) where the S 2 is parametrized by ψ = 0. (3.2).7) 2 2 1 5 g ∧ g1 ∧ g2 + g3 ∧ g4 .398 Unity from Duality: Gravity.11) Note that the self-duality ﬁxes the relative factor of 3 in (3.8) S2 ω2 = 4π.5) H3 = dB2 = where ω2 = 3gs M α dr ∧ ω2 . Note also that gs 6 F3 = H3 .5) (see (2. in order to be consistent with the quantization conditions (3. G3 = F3 − i H3 . We will see that this geometrical factor is crucial for reproducing the well-known factor of 3 in the N = 1 beta functions. S3 ω3 = 8π 2 (3. Since F3µνλ H3 = 0. 2 (3. Thus. F3 = Mα ω3 . (3.3)). and the S 3 by θ2 = φ2 = 0. As a result. Both ω2 and ω3 are closed. 2r (3. gs F3 = − 6 H3 .6) 1 1 1 g ∧ g 2 + g 3 ∧ g 4 = (sin θ1 dθ1 ∧ dφ1 − sin θ2 dθ2 ∧ dφ2 ) . θ1 = θ2 and φ1 = −φ2 . 2 ω3 = One can show that [31] (3.

The non-conservation of the ﬂux is due to the type IIB SUGRA equation ˜ dF5 = H3 ∧ F3 .1 ).18) The novel phenomenon in this solution is that the 5-form ﬂux present at the UV scale r = r0 may completely disappear by the time we reach a scale where Neﬀ = 0.14) Integrating this diﬀerential equation.17) and ω2 ∧ ω3 = 54vol(T 1.1 ). 10 T (3. we ﬁnd that h(r) = 27π(α )2 [gs N + a(gs M )2 ln(r/r0 ) + a(gs M )2 /4] 4r4 (3.19) F5 = 27πα Neﬀ (r)vol(T 1. (3. Note from (3.1 ). This is because ˜ F5 = F5 + B2 ∧ F3 . Thus. 24 3 12 3 (3.R.15) 1 1 2 2 2 (H 2 + gs F3 ) = H . 2π (3.13) The solution for the warp factor h may be determined from the trace of the Einstein equation: R= This implies −h−3/2 1 d 5 1 2 (r h ) = H3 . .19) that for a single cascade step Neﬀ (r) → Neﬀ (r) − M the radius changes by a factor r2 /r1 = exp(−2π/3gs M ).: N = 1 Gauge/Gravity Dualities 399 The 10-d metric found in [10] has the structure of a “warped product” of R3. agreeing with a result of [32]. 2 (3. ˜ An important feature of this background is that F5 acquires a radial dependence [10].16) with a = 3/(2π).1 and the conifold: ds2 = h−1/2 (r)dxn dxn + h1/2 (r)(dr2 + r2 ds2 1. F5 = dC4 . and Neﬀ (r) = N + 3 gs M 2 ln(r/r0 ). we may write ˜ F5 = F5 + F5 . (3. Klebanov et al.I. Neﬀ (r) → Neﬀ (r) − M which has the eﬀect of decreasing the 5-form ﬂux by M units. 5 dr r 6 (3.20) A related fact is that S2 B2 is no longer a periodic variable in the SUGRA solution once the M fractional branes are introduced: as the B2 ﬂux goes through a period.

B2 and C2 . 2 g1 g2 gs eΦ 1 Exactly the same relations apply to the N = 2 supersymmetric ory [4. we believe that the continuous logarithmic variation of Neﬀ (r) is related to continuous reduction in the number of degrees of freedom as the theory ﬂows to the IR. We may identify this quantity with Neﬀ deﬁning the gauge group SU (Neﬀ + M ) × SU (Neﬀ ) only at special radii rk = r0 exp(−2πk/3gsM ) where k is an integer. Writing h(r) = L4 ln(r/rs ).400 Unity from Duality: Gravity. We will work near r = r0 where Neﬀ may be replaced by N .23) orbifold the- .1 of the NS-NS and R-R 2-form potentials. Some support for this claim comes from studying the high-temperature phase of this theory using black holes embedded into an asymptotic KT geometry [33]. this SUGRA solution is reliable for suﬃciently large radii 1.22) Since T 1.20). in agreement with (3. 2 2 (3.1 Matching of the β-functions In order to match the two gauge couplings to the moduli of the type IIB theory on AdS5 ×T 1. T (3. the curvatures decrease there so that corrections to the SUGRA become negligible. 34]. one notes that the integrals over the S2 of T 1. Gauge Theory and Strings 1 ˜ Due to the non-vanishing RHS of (3.1 expands slowly toward large r. r4 L2 = 9gs M α √ . 3. Thus. Furthermore. The metric (3.21) we ﬁnd a purely logarithmic RG cascade: ds2 = r2 L2 ln(r/rs ) dxn dxn + L2 ln(r/rs ) 2 dr + L2 r2 ln(r/rs )ds2 1. The eﬀective number of degrees of freedom computed from the Bekenstein–Hawking entropy grows logarithmically with the temperature.1 . 7]1 : 4π 2 4π 2 π + 2 = .1 F5 is not quantized. the two gauge couplings are determined as follows [6. so that the SUGRA calculation of the β-functions may be compared with SU (Neﬀ + M ) × SU (Neﬀ ) gauge theory. Therefore.13) has a naked singularity at r = rs where h(rs ) = 0. Neﬀ = N − kM . In particular. 2 (3.19). even if gs M is very small. are moduli. In this regime the separation between the cascade where gs Neﬀ (r) steps is very large. (4π2 α )2 T 1.1 .

2 g1 g2 (3.I. 35.26) Since ln(r/rs ) = ln(Λ/µ).24) S2 1 From the quantization condition on H3 . Klebanov et al. We choose the normalization so that the relevant kinetic term in Ê 1 the ﬁeld theory action is 2g2 d4 xd2 θTr(W α Wα )+ h. = 3N − 2(N + M )(1 − γ).25) In this section we adopt this UV/IR relation. . (3. Let us compare with the Shifman–Vainshtein β-functions [37]2 : d 8π 2 2 d log(Λ/µ) g1 d 8π 2 d log(Λ/µ) g2 2 = 3(N + M ) − 2N (1 − γ). The constancy of the dilaton 2 2 translates into the vanishing of the β-function for 8π + 8π . There are diﬀerent ways of establishiing the precise relation. This periodicity is crucial for the cascade phenomenon. As we mentioned earlier. With this convention the additional factor in the β-function does not appear. this choice is dictated by the form of the supergravity action and diﬀers from the canonical normalization by a factor of 1/g 2 . Now we are ready to interpret the solution of [10] in terms of RG ﬂow in the dual SU (N + M ) × SU (N ) gauge theory. 10].R. For all metrics of the form (3. gs M well-separated.. (3. The simplest one is to identify the ﬁeld theory energy scale Λ with the energy of a stretched string ending on a probe brane positioned at radius r.c. for instance. Substituting 2 2 g1 g2 the solution for B2 into (3.24) we ﬁnd 8π 2 8π 2 − 2 = 6M ln(r/rs ) + const.27) (3. (3. 401 (3.26) implies a logarithmic running of g12 − g12 1 2 in the SU (N + M ) × SU (N ) gauge theory. In gauge/gravity duality the 5-dimensional radial coordinate deﬁnes the RG scale of the dual gauge theory [1–3. 2πα ( S2 B2 ) must be a periodic variable with period 2π. 1 so that the cascade jumps are We may consider.13) this gives Λ ∼ r. this SUGRA result is reliable for any value of gs M provided that gs N 1.: N = 1 Gauge/Gravity Dualities 4π 2 1 4π 2 gs eΦ = 2 − g2 g1 2πα 2 B2 −π (mod 2π). These equations are crucial for relating the SUGRA background to the ﬁeld theory β-functions when the theory is generalized to SU (N + M ) × SU (N ) [9. which typically corresponds to the Wilsonian renormalization group.28) 2 These expressions for the β-functions diﬀer from the standard NSVZ form [38] by a factor of 1/(1 − g 2 Nc /8π 2 ). The diﬀerence comes from the choice of normalization of the vector superﬁelds. 36]. A nice review of the derivation of the exact β-functions is in [39].

however. the fact that the solution of [10] is singular tells us that it has to be modiﬁed in the IR. Anomalies are especially interesting creatures for the gauge/gravity duality. The essential mechanism is the β-functions contribute to the trace anomaly. Gauge Theory and Strings where γ is the anomalous dimension of operators TrAi Bj . The SU (N +M ) gauge factor has 2N ﬂavors in the fundamental representation. In the dual quantum ﬁeld theory there are chiral fermions charged under the U (1)R . which is related by supersymmetry to the divergence of the U (1)R current. (3. Here we show that the chiral anomaly can be read oﬀ the solution as well. Tii . The essential observation is that 1/g1 and 1/g2 ﬂow in opposite directions and. because the . and so we can understand the R-symmetry breaking as an eﬀect of the chiral anomaly. there is a scale where the SU (N + M ) coupling.26) found on the SUGRA side.29) Remarkably. this becomes an SU (2N − [N + M ]) = SU (N − M ) gauge group. the coeﬃcient 6M is in exact agreement with the result (3. according to (3. In the previous section we showed how the logarithmic running of the gauge couplings manifests itself in the dual supergravity solution of [10]. Under a Seiberg duality transformation. 4 The chiral anomaly In theories with N = 1 supersymmetry. As the theory ﬂows to the IR. namely. we perform a N = 1 duality transformation on this gauge group factor. ∂i J i . we should not be able to continue the solution (3. g1 . The necessary modiﬁcation proceeds via the deformation of the conifold. To summarize. Seiberg dual2 2 ity [40]. diverges.402 Unity from Duality: Gravity. 1 require that γ = − 2 +O[(M/N )2n ] where n is a positive integer [11].22) to the region where Neﬀ is negative. To continue past this inﬁnite coupling. and is discussed in Section 5. We may also trace the jumps in the rank of the gauge group to a wellknown phenomenon in the dual N = 1 ﬁeld theory.27) we then ﬁnd 8π 2 8π 2 − 2 2 g1 g2 = = M ln(Λ/µ)[3 + 2(1 − γ)] 6M ln(Λ/µ)(1 + O[(M/N )2n ]). the full supergravity solution is only invariant under a Z2M subgroup of this U (1). Thus. Although the metric has a continuous U (1)R symmetry. the cascade must stop. β-functions are related to chiral anomalies [37]. The conformal invariance of the ﬁeld theory for M = 0. including its normalization. This consitutes a geometrical explanation of a ﬁeld theory β-function. Taking the diﬀerence of the two equations in (3. because negative N is physically nonsensical.27). Thus we obtain an SU (N ) × SU (N − M ) theory which resembles closely the theory we started with [11]. and symmetry under M → −M .

there is no smooth global expression for C2 .13. the symmetry breaking is a classical eﬀect on the gravity side. . a Z2M subgroup of the U (1) leaves ﬁxed the asymptotic values of the ﬁelds.2) Since S2 C2 is deﬁned modulo 4π 2 α . normalized so that β has period 2π. however. In this section we will study some aspects of the anomaly in detail for the cascading gauge theory. This is the R-symmetry of the dual gauge theory. but it is single-valued up to a gauge transformation. although this result is hardly surprising. which is a good thing as they do not appear anywhere explicitly in the gravity dual. 2. This Z2M is a symmetry since it respects the asymptotic values of the ﬁelds. Second. There is no need to appeal to instantons. so β → β + is a symmetry precisely if is an integer multiple of π/M . (4.1) This expression is not single-valued as a function of the angular variable β. Because of the explicit β dependence. Finally. a gauge transformation can shift C2 /(4π 2 α ) by an arbitrary integer multiple of ω2 /(4π). It is crucial. with no radiative corrections.R. C2 → C2 + M α ω2 . 4. Indeed.: N = 1 Gauge/Gravity Dualities 403 Adler-Bardeen theorem [41] guarantees that anomaly coeﬃcients computed at one loop are exact. Because is anyway only deﬁned mod 2π.3) has a U (1) symmetry associated with the rotations of the angular coordinate β = ψ/2. First. that the background value of the R-R 2-form C2 does not have this continuous symmetry. (4. the anomaly coeﬃcients computed on each side of the duality agree exactly. although F3 is U (1) symmetric. the R-symmetry is broken spontaneously in the supergravity solution – the bulk vector ﬁeld dual to the R-current of the gauge theory acquires a mass. In fact. Locally.I. the signiﬁcance of this fact is that we can compute anomaly coeﬃcients in the ﬁeld theory at weak coupling. Under the transformation β → β + . The symmetry breaking then appears “anomalous” if one insists on a fourdimensional description. even for our non-conformal cascading theory with only N = 1 supersymmetry. we may write C2 → M α βω2 . F3 is completely independent of β. C2 is not U (1)-invariant. so that F3 = dC2 is single-valued. There are three lessons that we can take away from this analysis [42].1 The anomaly as a classical eﬀect in supergravity The asymptotic UV metric (3. then extrapolate the results to strong coupling. and thus corresponds to a symmetry of the system. where we can use dual gravity methods to check the calculation. Klebanov et al. it is a nice check of the duality.

Varying with respect to . Using the fact that S2 ω2 = 4π. (4.5) are d4 x − ∂i J i + M ˜ ˜ (F a F aij − Gb Gbij ) .1 is a modulus. ij 16π 2 (4. (4.404 Unity from Duality: Gravity.) As pointed out in [6]. the Noether current associated with that symmetry is not generally conserved but instead obeys the equation ∂i J i = 1 32π 2 a ˜ nm Rm Fij F aij m (4. it is natural that the integral of C2 is dual to the diﬀerence of Θ-angles (it is possible to check this statement explicitly in orbifold backgrounds). The Θ-angles are given by Θ1 − Θ2 = 1 πα C2 .8) .7) This anomaly equation.4) the terms linear in in the dual gauge theory (4. we ﬁnd that the small U (1) rotation β → β + induces Θ1 = −Θ2 = 2M . Because the integral of B2 was dual to the diﬀerence of gauge couplings for the two gauge groups. the integral of the RR 2-form potential C2 over the S2 of T 1.6) where J i is the chiral R-current. we therefore obtain ∂i J i = M a ˜ ˜ Fij F aij − Gb Gbij . derived from supergravity. The appearance of the second term is due to the non-invariance of C2 under the U (1) rotation. Θ1 + Θ2 ∼ C. agrees exactly with our expectations from the gauge theory.5) If we assume that is a function of the 4 world volume coordinates xi .4) With a conventional normalization.3) S2 where C is the RR scalar. the Θ terms appear in the gauge theory action as d4 x Θ2 b ˜ bij Θ1 a ˜ aij F F + G G 32π 2 ij 32π 2 ij · (4. A standard result of quantum ﬁeld theory is that in a theory with chiral fermions charged under a global U (1) symmetry of the classical Lagrangian. which vanishes for the case under consideration. Gauge Theory and Strings Let us compare the above analysis with the gauge theory. ij 16π 2 ij (4. then under the U (1) rotation (4. (A similar comparison for the case of an N = 2 orbifold theory appeared in [43].

I. Klebanov et al. No additional fractional D-instanton eﬀects are needed to explain the anomaly. We will show that symmetry breaking through anomalies can also have the bulk Higgs mechanism as its dual.7). it becomes a local symmetry in the full gravity 3 The connection between anomalies in a D-brane ﬁeld theory and spontaneous symmetry breaking in string theory was previously noted in [45] (and probably elsewhere in the literature). there a are two gauge groups. so the U (1) R-symmetry is a true symmetry of the ﬁeld theory. so let us deﬁne Fij and Gb to be the ﬁeld strengths ij of SU (N + M ) and SU (N ) respectively. but in the gravity dual it as usual becomes a gauge symmetry. An equivalent calculation for the SU (N ) gauge group with 2(N + M ) ﬂavors produces the opposite anomaly. On the gauge theory side.: N = 1 Gauge/Gravity Dualities 405 where nm is the number of chiral fermions with R-charge Rm circulating in the loop of the relevant triangle diagram. In the case of interest. Therefore. Rather. we will ﬁnd that the gauge symmetry is spontaneously broken: the 5-d vector ﬁeld dual to the R-current of the gauge theory “eats” the scalar dual to the diﬀerence of the theta angles and acquires a mass3 . In these cases it was shown [46. A closely related mechanism was observed in studies of RG ﬂows from the dual gravity point of view [46. so the anomaly equation as computed from ﬁeld theory is just (4. 48]. The upshot of the calculation presented above is that the chiral anomaly of the SU (N + M ) × SU (N ) gauge theory is encoded in the ultraviolet (large r) behavior of the dual classical supergravity solution. the anomaly M coeﬃcient is 16π2 . . Similar methods have been used to describe chiral anomalies in other supersymmetric gauge theories [42–44]. a quantum eﬀect on the gauge theory side turns into a classical eﬀect in supergravity. or the theory would not make sense at all.R. 4. Bj contribute 2N ﬂavors to the gauge group SU (N + M ). and each one carries R-charge 1/2. The chiral fermions which are their superpartners have Rcharge −1/2 while the gluinos have R-charge 1. the R-symmetry is global. Now. There R-current conservation was violated not through anomalies but by turning on relevant perturbations or expectation values for ﬁelds. Because the R-symmetry is realized geometrically by invariance under a rigid shift of the angle β. Thus. In the absence of fractional branes there are no background three-form ﬂuxes. the chiral superﬁelds Ai . which must not be anomalous. as often occurs in the gauge/gravity duality.2 The anomaly as spontaneous symmetry breaking in AdS5 Let us look for a deeper understanding of the anomaly from the dual gravity point of view. 48] that the 5-d vector ﬁeld dual to the R-current acquires a mass through the Higgs mechanism.

406 Unity from Duality: Gravity. the unperturbed F5 of (3. (4.9). which is invariant under the combined gauge transformations β → β + λ. and L4 = 27 (4πα gs N ). indeed.9) where h(r) = L4 /r4 . It is convenient to deﬁne 16 the one-form χ = g 5 − 2A. It is straightforward to show that ˜ the supergravity ﬁeld equation dF5 = 0 implies that the ﬁeld A satisﬁes the equation of motion for a massless vector ﬁeld in AdS5 space: d 5 dA = 0. An appropriate ansatz to linear order in A is ˜ F5 = dC4 = 1 4 πα N χ ∧ g1 ∧ g2 ∧ g3 ∧ g4 d x ∧ dh−1 + gs 4 −dA ∧ g 5 ∧ dg 5 + 3 L 5 2 dA ∧ dg 5 .14) . and the associated gauge ﬁelds A = Aµ dxµ appear as ﬂuctuations of the ten-dimensional metric and RR four-form potential [19. (4. RMN = 2 gs ˜ ˜ PQRS . A → A + dλ. (4. (4.13) Using the identity dg 5 ∧ dg 5 = −2g 1 ∧ g 2 ∧ g 3 ∧ g 4 .10) The equations of motion for the ﬁeld Aµ appear as the χµ components of Einstein’s equations.11) The ﬁve-form ﬂux will also ﬂuctuate when we activate the Kaluza-Klein ˜ gauge ﬁeld. The natural metric ansatz is of the familiar Kaluza-Klein form: ds2 = h(r)−1/2 (dxn dxn ) +h(r) 1/2 2 4 r 1 5 dr2 g − 2A + r2 9 2 2 1 + (g r )2 . Gauge Theory and Strings theory. FMPQRS FN 4 · 4! (4.22].18) is not self-dual with respect to the gauged metric (4. we can check that the expression for C4 is C4 = 1 −1 4 πα N 1 βg 1 ∧ g 2 ∧ g 3 ∧ g 4 − A ∧ dg 5 ∧ g 5 h d x+ gs 2 2 − 3 −1/4 h 2r 5 2 dA ∧ g 5 . 6 r=1 (4.12) The ﬁve-dimensional Hodge dual 5 is deﬁned with respect to the AdS5 metric ds52 = h−1/2 dxn dxn + h1/2 dr2 .

9) R = R(A = 0) − h1/2 r2 Fµν F µν 9 (4. The story changes when we add wrapped D5-branes. A more heuristic approach is to consider the type IIB supergravity action to . we ﬁnd that up to terms of order gs M 2 /N the three-form equation implies d 5 W =0 ⇒ L2 1 ∂ W i + 5 ∂r r5 Wr = 0 2 i r r (4.I.: N = 1 Gauge/Gravity Dualities 407 Another way to see that A is a massless vector in AdS5 is to consider the Ricci scalar for the metric (4. and the additional Wr term in (4. To a four dimensional observer.R.16) so that F3 is invariant under the gauge transformation β → β +λ. To restore the gauge invariance.17) we can immediately see how the anomaly will appear in the gravity dual. the new wrinkle is that the RR three-form ﬂux of (3. however. 2 (4.10). In terms of the gauge invariant forms χ and W = Wµ dxµ . Klebanov et al. we should derive the χµ components of Einstein’s equations. As described in Section 2. Thus in the ﬁeld theory one cannot interpret the U (1) symmetry breaking as being spontaneous. Another way to see that the vector ﬁeld becomes massive is to compute its equation of motion.6). This approach is somewhat nontrivial.15) so that on reduction from ten dimensions the ﬁve-dimensional supergravity action will contain the action for a massless vector ﬁeld. a massive vector ﬁeld would satisfy ∂i W i = 0. the 5-branes introduce M units of RR ﬂux through the threecycle of T 1.17) From (4. we introduce a new ﬁeld θ ∼ S 2 C2 : F3 = dC2 = Mα 2 g 5 + 2∂µ θdxµ ∧ ω2 (4. F3 = Mα (χ + 2W ) ∧ ω2 .1 . and also ﬁnd the appropriate expressions for the ﬁve-form and metric up to quadratic order in gs M and linear order in ﬂuctuations.18) which is just what one would expect for a massive vector ﬁeld in ﬁve dimensions. θ → θ−λ.18) appears in four dimensions to be an anomaly. Assuming that the NS-NS three form is still given by (3. Now. Let us also deﬁne Wµ = Aµ + ∂µ θ. To do this calculation precisely.5) is not gauge-invariant with respect to shifts of β (4.

To relate the 5-d metric (4. In [46–48] it was shown that the 5-d vector ﬁeld associated with a U (1)R symmetry acquires a mass in the presence of a symmetry-breaking relevant perturbation. Doing this carefully we ﬁnd ˜ Gµν dxµ dxν = hr4 /L4 5/6 (h−1/2 ηij dxi dxj + h1/2 dr2 ). . ignores the subtlety of the type IIB action in presence of the self-dual 5-form ﬁeld.22) 2h3/2 r6 This result.24) to the 10-d metric (5. It is conventional to write the 5-d gauged supergravity metric in the form ˜ Gµν dxµ dxν = e2T (q) ηij dxi dxj + dq 2 . which takes the mixing into account. A more precise calculation [49].23) This shows that the 10-d mass actually appears at a higher order in perturbation theory compared to the result (4. 12 (4. however.408 Unity from Duality: Gravity.24) The result of [46] is that m = −2T . (4.2) we must normalize the 5-d metric so that the graviton has a canonical kinetic term. and that this mass is related in a simple way to the warp factor of the geometry4 . DeWolfe and K. Let us compare this result to earlier work. ignoring the 5-form ﬁeld strength contributions: S = ∼ − − − 1 2 2κ10 1 2 2κ10 d10 x d10 x 2 −G10 R10 − −G10 − gs 2 |F3 |2 + .22) that ignores the mixing with the 5-form. we see that the mass-squared is given by m2 ˜ = (gs M α ) 2 (4. Skenderis for pointing out the relevance of this work to the present calculation. which has as its equation of motion ∂µ (hr7 F µν ) = m2 hr7 W ν ˜ ˜ which in diﬀerential form notation is d(h7/4 r7 5 dW ) = −m2 h7/4 r7 From the action (4. gives instead the following equation for the transverse vector modes: (9M α )4 1 ∂ hr7 ∂r + h∂i ∂i − 7 r hr 64h2 r10 Wi = 0.19) h1/2 r2 Fµν F µν 9 (4. Gauge Theory and Strings quadratic order in W . 2 (4.20) gs M α 2 36 Wµ W µ + . 81 · (4.25) 4 We are grateful to O.20). hr4 This is clearly the action for a massive four-dimensional vector ﬁeld.21) 5 W. . (4. . . .

: N = 1 Gauge/Gravity Dualities 5/6 409 The factor hr4 /L4 arises due to the radial dependence of the size of T 1. 50] where anomalies arose from boundary terms. The appearance of a mass implies that the R-current operator should acquire an anomalous dimension.R. by log(r) ∼ q gs M 2 − L 2πN q L 2 · (4.26) We can also show that −2T = −2 log(r)+(terms which do not aﬀect the mass to leading order in gs M 2 /N ).30) Using the AdS/CFT correspondence (perhaps naively. the anomalous dimension of the current is ∆ − 3 ≈ (mL)2 /2 = (gs M )2 · π(gs N ) (4.23) after a rescaling [49] (4. The radial variables q and r are related. (4.I.1 through the usual Kaluza–Klein reduction.32) . For these calculations it is convenient to work with the transverse 4-d vector modes Vi and to decouple the longitudinal modes such as Vr . at leading order in gs M 2 /N .29) The nonvanishing vector mass is consistent with gauge invariance because the massless vector ﬁeld A has eaten the scalar ﬁeld θ. so now computing the mass-squared by the prescription of [46] we obtain m2 = (gs M )2 4 · α (3π)3/2 (gs N )3/2 (4. From (4. as the KT metric is not asymptotically AdS) we ﬁnd that the dimension of the current J µ dual to the vector ﬁeld W µ is ∆=2+ 1 + (mL)2 . (4. The equation of motion of V is e−2T ∂ 2T ∂ e + e−2T ∂i ∂i − m2 Vi = 0. in contrast to some earlier examples [3. this equation follows from (4.27) where this mass applies to a vector ﬁeld V with a canonical kinetic term for the metric (4.27) it follows that (mL)2 = 2(gs M )2 · π(gs N ) (4.28) In fact. It is interesting that the anomaly appears as a bulk eﬀect in AdS space. as advertised.31) Therefore. spontaneously breaking the gauge symmetry.25). Klebanov et al. ∂q ∂q Vi = (hr4 /L4 )2/3 Wi .

Our supergravity calculation predicts that this anomalous dimension is corrected at large gs N by an extra factor of 1/(gs N ). but with the conifold replaced by the deformed conifold as the transverse space. It is diagonal in the basis (3. and the resulting anomalous dimension γJ is quadratic in M and N . The leading correction to the current-current two-point function comes from the three-loop Feynman diagram composed of two triangle diagrams glued together. to remove the naked singularity found in [10] the conifold (2. which simply interchanges the two gauge groups. 5 Deformation of (KS) the conifold It was shown in [11] that. the lowest order piece of the anomalous dimension will be of order (gs M )2 .1) should be replaced by the deformed conifold 4 i=1 2 zi = ε2 .13). Of course.1 . it would be interesting to understand this result better from the gauge theory point of view. 10 6 (5. N → N + M .3) 2 . and it must be invariant under the map M → −M .2): ds2 = 6 τ 1 1 4/3 ε K(τ ) (dτ 2 + (g 5 )2 ) + cosh2 [(g 3 )2 + (g 4 )2 ] 2 3K 3 (τ ) 2 + sinh2 where K(τ ) = (sinh(2τ ) − 2τ )1/3 · 21/3 sinh τ (5. (5. Gauge Theory and Strings We can obtain a rough understanding of this result by considering the relevant weak coupling calculation in the gauge theory. This is the same 6 type of “D-brane” ansatz as (3. Thus.1) in which the singularity of the conifold is removed through the blowing-up of the S3 of T 1. The 10-d metric of [11] takes the following form: ds2 = h−1/2 (τ )dxn dxn + h1/2 (τ )ds2 . 30.3). γJ must vanish when M = 0. The metric of the deformed conifold was discussed in some detail in [18.4) τ [(g 1 )2 + (g 2 )2 ] .410 Unity from Duality: Gravity.2) where ds2 is the metric of the deformed conifold (5. (5. 51].

6 T At τ = 0 the angular metric degenerates into dΩ2 = 3 1 4/3 1 ε (2/3)1/3 (g 5 )2 + (g 3 )2 + (g 4 )2 .13) (5. the self-dual 5-form ﬁeld strength may be decomposed as F5 = F5 + F5 . shrink as 1 4/3 ε (2/3)1/3 τ 2 [(g 1 )2 + (g 2 )2 ].7) with F (0) = 0 and F (∞) = 1/2.5) and in terms of this radial coordinate ds2 → dr2 + r2 ds2 1. and F5 = 4gs M 2 (α )2 ε−8/3 dx0 ∧ dx1 ∧ dx2 ∧ dx3 ∧ dτ K 2 h2 (τ ) · sinh2 (τ ) (5. 30]. We have F5 = B2 ∧ F3 = where = f (1 − F ) + kF. Klebanov et al.8) (5. The additional two directions.R.6) which is the metric of a round S3 [18.12) gs M 2 (α )2 (τ )g 1 ∧ g 2 ∧ g 3 ∧ g 4 ∧ g 5 . (5.: N = 1 Gauge/Gravity Dualities For large τ we may introduce another radial coordinate r via r2 = 3 25/3 ε4/3 e2τ /3 . 2 = gs M α dτ ∧ (f g 1 ∧ g 2 + k g 3 ∧ g 4 ) 2 1 + (k − f )g 5 ∧ (g 1 ∧ g 3 + g 2 ∧ g 4 ) . and B2 = H3 = dB2 gs M α [f (τ )g 1 ∧ g 2 + k(τ )g 3 ∧ g 4 ].9) (5.11) .10) ˜ As before. 8 The simplest ansatz for the 2-form ﬁelds is F3 = = Mα 2 Mα 2 g 5 ∧ g 3 ∧ g 4 + d[F (τ )(g 1 ∧ g 3 + g 2 ∧ g 4 )] g 5 ∧ g 3 ∧ g 4 (1 − F ) + g 5 ∧ g 1 ∧ g 2 F +F dτ ∧ (g 1 ∧ g 3 + g 2 ∧ g 4 ) .1 .I. 2 (5. 4 (5. corresponding to the S2 ﬁbered over the S3 . 2 2 (5. 411 (5.

2 sinh τ τ coth τ − 1 (cosh τ − 1). 4 sinh2 τ (5. = 2 (5. Note that the ﬁrst three of these equations (5.412 Unity from Duality: Gravity.14) These equations follow from a superpotential for the eﬀective radial problem [52]. Luckily. For large τ we impose. The solution is F (τ ) f (τ ) k(τ ) = = = sinh τ − τ . these equations imply the self-duality of the complex 3-form with respect to the metric of the deformed conifold: 6 G3 = iG3 . this is possible for our ansatz [11]: f k F and h = −α where α = 4(gs M α )2 ε−8/3 . K 2 (τ ) sinh2 τ (5. the warp factor may be determined by integrating (5.19) . as usual. 4 (5. form a closed system and need to be solved ﬁrst.17) Now that we have solved for the 3-forms on the deformed conifold.16) f (1 − F ) + kF .15) = (1 − F ) tanh2 (τ /2). = F coth2 (τ /2). First we note that (τ ) = f (1 − F ) + kF = τ coth τ − 1 (sinh 2τ − 2τ ). 2 sinh τ (5. 1 (k − f ).15). (5. Gauge Theory and Strings 5. the second order equations should be replaced by a system of ﬁrst-order ones.18) This behaves as τ 3 for small τ . The resulting integral expression for h is h(τ ) = α 22/3 I(τ ) = (gs M α )2 22/3 ε−8/3 I(τ ).14). 2 sinh τ τ coth τ − 1 (cosh τ + 1). In fact.1 The ﬁrst-order equations and their solution In searching for BPS saturated supergravity backgrounds. the boundary condition that h vanishes.

sinh2 x (5. and lifting the result to M-theory maps our background to a Becker-Becker solution supported by a G4 which is a (2. the curvatures are small and the SUGRA approximation is reliable. This is true even far in the IR. Klebanov et al. Performing a T -duality along one of the longitudinal directions. The small τ behavior follows from the convergence of the integral (5.I. I(τ → ∞) → 3 · 2−1/3 τ − 1 4 e−4τ /3 .2 SO(4 ) invariant expressions for the 3-forms In [53.21).71805. As long as this is large. from (5.21) (5. since the radius-squared of the S3 at τ = 0 is of order gs M in string units. Very importantly. (5. .1 times the deformed conifold: ds2 10 → ε4/3 1/2 21/3 a0 gs M α dxn dxn + a0 6−1/3 (gs M α ) 1/2 1 2 1 5 2 dτ + (g ) 2 2 (5.23) 1 +(g 3 )2 + (g 4 )2 + τ 2 [(g 1 )2 + (g 2 )2 ] · 4 This metric will be useful in Section 6 where we investigate various infrared phenomenon of the gauge theory. This I(τ ) is nonsingular at the tip of the deformed conifold and. for small τ the ten-dimensional geometry is approximately R3. 5. Perhaps the easiest way to see the supersymmetry of the deformed conifold solution is through a T -duality. 2) form on T 2 × CY. G-ﬂux of this type indeed produces a supersymmetric background [55].20) We have not succeeded in evaluating this integral in terms of elementary or well-known special functions. Thus. matches the form of the large-τ solution (3. 54] it was shown that the warped background of the previous section preserves N = 1 SUSY if and only if G3 is a (2.22) where a0 ≈ 0. for large gs M the curvatures found in our solution are small everywhere.R. but it is not hard to see that I(τ → 0) → a0 + O(τ 2 ).: N = 1 Gauge/Gravity Dualities where I(τ ) ≡ τ ∞ 413 dx x coth x − 1 (sinh(2x) − 2x)1/3 . while at large τ the integrand becomes ∼ xe−4x/3 .5).19). This is the ’t Hooft coupling of the gauge theory found far in the IR. 1) form on the CY space.

we will now demonstrate that there is conﬁnement. Gauge Theory and Strings While writing G3 in terms of the angular 1-forms g i is convenient for some purposes.1) .1 Dimensional transmutation and conﬁnement The resolution of the naked singularity via the deformation of the conifold is a supergravity realization of the dimensional transmutation. Other stringy approaches to infrared phenomena in N = 1 SYM theory have recently appeared in [57–59]. 6 Infrared physics We have now seen that the deformation of the conifold allows the solution to be non-singular. Below we write the G3 found in [11] in terms of the obvious 1-forms on the deformed conifold: dz i and d¯i .24) ¯ z We also note that the NS-NS 2-form potential is an SO(4) invariant (1. In particular. and so is the metric. i = 1. 3. From (5. 4: z G3 = 2ε6 sinh4 Mα τ sinh(2τ ) − 2τ ( sinh τ ijkl zi zj dzk ∧ d¯l ) ∧ (¯m dzm ) ¯ z z +2(1 − τ coth τ )( ijkl zi zj dzk ∧ dzl ) ∧ (zm d¯m ) · (5.5) we see that ε2/3 has dimensions of length and that τ = 3 ln(r/ε2/3 ) + const. 6.414 Unity from Duality: Gravity. The scale necessary to deﬁne the logarithm transmutes into the the parameter ε that determines the deformation of the conifold. In the following sections we point out some interesting features of the SUGRA background we have found and show how they realize the expected phenomena in the dual ﬁeld theory. Our expressions for the gauge ﬁelds are manifestly SO(4) invariant. 1) was demonstrated in [56] with the help of a holomorphic basis. we saw that turning on the R-R 3-form ﬂux produces the logarithmic warping of the KT solution. that there is a gluino condensate that breaks the Z2M chiral symmetry down to Z2 and that there are domain walls separating inequivalent vacua. 2. 1) form: B2 = igs M α τ coth τ − 1 2ε4 sinh2 τ ijkl zi zj dzk ∧ d¯l . the (2. That G3 is indeed (2. 1) nature of the form is not manifest. ¯ z (5. While the singular conifold has no dimensionful parameter. (6.25) The derivation of these formulae is given in [31]. that the theory has glueballs and baryons whose mass scale emerges through a dimensional transmutation.

Note that the 3-cycle has the minimal volume near τ = 0.6) . Note that for M = 0 the D3-brane wrapped on the S3 gave a dibaryon [8]. we ﬁnd that a fundamental string with this surface will have a ﬁnite tension. The masses of glueball and Kaluza-Klein (KK) states scale as mglueball ∼ mKK ∼ ε2/3 · gs M α (6. On the other hand.2) The reason the theory is conﬁning is that in the metric for small τ (5. If the contour has a very large area A.4) Comparing with the string tension. From the fact that the coeﬃcient of dxn dxn is ﬁnite at τ = 0. which sets the tension of the conﬁning ﬂux tubes. and so the resulting Wilson loop satisﬁes the area law. we see that Ts ∼ gs M (mglueball )2 . and calculate the expectation value of the Wilson loop using the prescription [60.: N = 1 Gauge/Gravity Dualities 415 Thus.R. the full SUGRA background has a ﬁnite 3-cycle. The minimal area surface bounded by the contour bends towards smaller τ .I.23) the function multiplying dxn dxn approaches a constant. Klebanov et al. We may interpret various branes wrapped over this 3-cycle in terms of the gauge theory. A simple estimate shows that the string tension scales as Ts = 1 1/2 24/3 a0 π ε4/3 · (α )2 gs M (6. then most of the minimal surface will drift down into the region near τ = 0.5) Due to the deformation. Studying a probe D3-brane in the background of our solution show that the mass of the baryon scales as Mb ∼ M ε2/3 · α (6. A wrapped D3-brane plays the role of a baryon vertex which ties together M fundamental strings. (6. hence all the wrapped branes will be localized there. This should be contrasted with the AdS5 metric where this function vanishes at the horizon. the form of the IR metric (5. the scale rs entering the UV solution (3.21) should be identiﬁed with ε2/3 .3) We will return to these conﬁning strings in the next section. Consider a Wilson contour positioned at ﬁxed τ . 61].23) makes it clear that the dynamically generated 4-d mass scale. is ε2/3 √ · α gs M (6. and therefore joins M external charges. the connection between these two objects becomes clearer when one notes that for M > 0 the dibaryon has M uncontracted indices. or with the singular metric of [10] where it blows up.

[68].9) where Λ is the overall IR scale. An interesting generalization is to consider Wilson loops in antisymmetric tensor representations with q indices where q ranges from 1 to M − 1.7) in the limit of large area. This type of behaviour is also found in the supergravity duals of N = 1 gauge theories [66]. It is interesting to ask how the tension of this class of conﬁning strings depends on q. 2. Tq+q < Tq + Tq . M q = 1. These Wilson loops can be thought of as conﬁning strings which connect q probe quarks on one end to q corresponding probe anti-quarks on the other. hence the corresponding Wilson loop does not have an area law. for brevity we build on a closely related result of Bachas et al. . Although the blow-up can be shown directly. Here the conﬁning q-string is described by q coincident fundamental strings placed at τ = 0 and oriented along the R3. 65]: Tq = Λ2 sin πq . In the S-dual of 5 Qualitatively similar conﬁning ﬂux-tubes were examined in [67] where the authors use the near horizon geometry of non-extremal D3-branes to model conﬁnement.416 Unity from Duality: Gravity. the appearance of composite conﬁning strings. . This is precisely the situation found by Douglas and Shenker (DS) [62] in softly broken N = 2 gauge theory. but it is important that there are M units of F3 ﬂux through the S3 . . The classic criterion for conﬁnement is that this Wilson loop obey the area law − ln W1 (C) = T1 A(C) (6.1 . M − 1 (6. To explain what they are. If it is a convex function. (6. q = 1 corresponds to the fundamental representation as denoted above. and later by Hanany et al. let us recall that the basic string corresponds to the Wilson loop in the fundamental representation. (HSZ) [63] in the MQCD approach to conﬁning N = 1 supersymmetric gauge theory [64. . and there is a symmetry under q → M − q which corresponds to replacing quarks by anti-quarks. In fact.8) then the q-string will not decay into strings with smaller q.5 In the deformed conifold solution analyzed above both F5 and B2 vanish at τ = 0. Gauge Theory and Strings 6. For q = M the probe quarks combine into a colorless state (a baryon).2 Tensions of the q-strings The existence of the blown up 3-cycle with M units of RR 3-form ﬂux through it is responsible for another interesting infrared phenomenon. this R-R ﬂux blows up the q fundamental strings into a D3-brane wrapping an S2 inside the S3 . .

. We are now using the standard round metric on S3 so that ψ is the azimuthal angle ranging from 0 to π. at τ = 0 the metric is ε4/3 2 21/3 a0 gs M α 1/2 1/2 dxn dxn + bM α dψ 2 + sin2 ψdΩ2 .15) Note that under q → M − q. 2 (6. 2 (6.1 .12) Following [68] closely we ﬁnd that the tension of a D3-brane which wraps an S2 located at the azimuthal angle ψ is sin(2ψ) πq − b sin ψ + ψ − 1/3 π 2 g 2 α 2 b 2 M 12 s 2 4 4/3 2 1/2 .11) while the world volume ﬁeld is q F = − sin θdθ ∧ dφ.14) The tension of the wrapped brane is given in terms of the solution of this equation by 4/3 Tq = 2 121/3 π 2 gs α 2 sin ψ 1 + (b2 − 1) cos2 ψ. After applying S-duality to the KS solution. This is a crucial property needed for the connection with the q-strings of the gauge theory. The NS-NS 2-form ﬁeld at τ = 0 is B2 = M α ψ− sin(2ψ) 2 sin θdθ ∧ dφ.93266. (6. Klebanov et al.R. T -dualizing along the D1-brane direction we ﬁnd q D0-branes on an S3 with M units of NS-NS ﬂux. (6. We will ﬁnd the same phenomenon.10) where b = 2a0 6−1/3 ≈ 0. we ﬁnd ψ → π − ψ.13) Minimizing with respect to ψ we ﬁnd ψ− πq 1 − b2 = sin(2ψ). but our probe brane calculation is somewhat diﬀerent from [68] because the radius of our S3 is diﬀerent.: N = 1 Gauge/Gravity Dualities 417 our type IIB gravity model.1 × S3 geometry with M units of NS-NS H3 ﬂux through the S3 and q coincident D1-branes along the R3.I. at τ = 0 we ﬁnd the R3. M 2 (6. so that TM−q = Tq . (6. This geometry is very closely related to the setup of [68] whose authors showed that the q D0-branes blow up into an S2 .

now we ﬁnd sin πq Tq M = . M. This simpliﬁes the probe calculation and makes it identical to that of [68]. If we ignore the RHS of this equation. 6. In this background only the F3 ﬂux is present. we expect interactions among the strings to become negligible and the q-string tension to become just q times the ordinary string tension (6. In Section 3 we showed that the corresponding symmetry of the UV (large τ ) limit of the metric is ψ→ψ+ 2πk . Our argument applied to the MN background leads very simply to the DS–HSZ formula for the ratios of q-string tensions (6.7% of that in the b = 1 case.17).14) is not exactly solvable. . all the tensions are multiplied by gs . It is interesting to compare (6.16) The deviations from this formulae are small: even when ψ = π/4 and correspondingly q ≈ M/4. An analogous calculation for the MN background [57] proceeds almost identically. this formula also holds approximately for the KS background.06507 is small numerically. the tension in the KS case is approximately 96. then ψ ≈ πq/M and Tq ∼ sin πq · M (6. . 2.17) without making any approximations. we ﬁnd that gs Tq = qTs in the large M limit. The value of B2 at the minimal radius is again given by (6. As we have shown earlier. Tq sin πq M (6. When we S-dualize back to the original background with RR-ﬂux and q F -strings. (6.17). There is a subtle diﬀerence however from the calculation for the KS background in that now the parameter b entering the radius of the S3 is equal to 1. In particular. Gauge Theory and Strings Although (6. .418 Unity from Duality: Gravity.16) with the naive string tension (6. It is interesting to note that recent lattice simulations in non-supersymmetric pure glue gauge theory [69] appear to yield good agreement with (6. we note that (1−b2 )/2 ≈ 0.11). Indeed.3). In the large M limit. M k = 1. The extra gs appears because we have been computing tensions in the dual background. hence after the S-duality we ﬁnd only H3 = dB2 .3 Chiral symmetry breaking and gluino condensation Our SU (N + M ) × SU (N ) ﬁeld theory has an anomaly-free Z2M R-symmetry. .18) .3) we obtained in the previous section.

we see that the full solution. as in the basic solution derived in the preceding sections.I. Such a domain wall is obviously a stable object in the KS background and crossing it takes us from one ground state of the theory to another. hence B F3 = S2 ∆C2 . we ﬁnd that the domain wall tension is Twall ∼ 1 ε2 · gs (α )3 (6. there are M inequivalent vacua: there are exactly M diﬀerent discrete orientations of the solution. Therefore from (3. Klebanov et al. the wrapped D5-brane produces a discontinuity in B F3 . Also.20) for large τ . a fundamental string can end on the wrapped D5-brane. has the Z2 symmetry generated by ψ → ψ + 2π. If to the left of the domain wall B F3 = 0. Indeed.21) In supersymmetric gluodynamics the breaking of chiral symmetry is associated with the gluino condensate λλ . A holographic calculation of the condensate was carried out by Loewy and Sonnenschein in [71] (see also [72] for previous work on gluino condensation in conifold theories. Let us consider domain walls made of k D5-branes wrapped over the ﬁnite-sized S3 at τ = 0. then to the right of the domain wall F3 = 4π 2 α k. with remaining directions parallel to R3.1). This change in C2 is produced by the Z2M transformation (6. As a result.9) it is clear that to the right of the wall ∆C2 → πα kω2 (6. baryons can dissolve in them. Indeed.18) on the original ﬁeld conﬁguration (4. B (6.19) as follows from the quantization of the D5-brane charge.R. which depends on ψ through cos ψ and sin ψ.: N = 1 Gauge/Gravity Dualities 419 Recalling that ψ ranges from 0 to 4π. By studying a probe D5-brane in the metric.) They looked for the deviation of the complex 2-form ﬁeld C2 − gis B2 from its asymptotic large τ form that enters the KT solution: δ C2 − i B2 gs ∼ M α −τ τ e [g1 ∧ g3 + g2 ∧ g4 − i(g1 ∧ g2 − g3 ∧ g4 )] 4 . The B-cycle is bounded by a 2-sphere at τ = ∞. where B is the cycle dual to the S3 .1 . It is expected that ﬂux tubes can end on these domain walls [70]. corresponding to breaking of the Z2M UV symmetry through the IR eﬀects. The domain walls constructed out of the wrapped D5-branes separate these inequivalent vacua.

Klebanov and A. Phys. Klebanov and N. From the angular dependence of the perturbation we see that the dual operator is SU (2) × SU (2) invariant and carries R-charge 1. Klebanov. Maldacena. Strassler.R.Phys. Peet. N. [8] S. Math.M. Phys. B 428 (1998) 105 [hep-th/9802109]. A. [2] S. 2 (1998) 231 [hep-th/9711200]. Vafa. [3] E. M. [5] A.W. Ooguri and Y. The presence of an extra ln(r/ε2/3 ) factor is presumably due to the fact that the asymptotic KT metric diﬀers from AdS5 by such logarithmic factors. Phys. J. Adv. Nekrasov. Phys. B 578 (2000) 123 [hep-th/0002159]. r3 (6. Phys. Rev.R. Nucl. Klebanov and E. 3 (1999) 1 [hep-th/9810201]. Nekrasov and C. Klebanov and A. Phys. Math. I. B 435 (1998) 337 [hep-th/9805131]. Thus. Rev. Adv. [10] I.420 ∼ Unity from Duality: Gravity.R.R. Phys. 80 (1998) 4855 [hep-th/9802183]. These are precisely the properties of λλ. Tseytlin and E. Lawrence. Rept. Theor. Nucl.S. [6] I. [9] I. Nucl. Phys. I.R. Klebanov and A. N. Witten. [7] Non-Spherical Horizons. Phys.S. Strassler. B 536 (1998) 199 [hep-th/9807080]. Klebanov and M. B 533 (1998) 199 [hep-th/9803015].K.R. I.23) Thus. [14] S. Math. Gubser. Nekrasov.R. Witten. is grateful to S.R. [4] S. S. Aharony. Nucl. Polyakov. Lett. D 58 (1998) 125025 [hep-th/9808075]. A. Silverstein.22) In a space-time that approaches AdS5 a perturbation that scales as r−3 corresponds to the expectation value of a dimension 3 operator. TASI Lectures: Introduction to the AdS/CFT Correspondence [hep-th/0009139]. H. [12] O. Phys. Witten for collaboration on parts of the material reviewed in these notes and for useful input. Lett. Kachru and E. Tseytlin. Gubser. D 54 (1996) 3915 [hep-th/9602135]. Theor.R. Theor. This work was supported in part by the NSF grant PHY-9802484. the holographic calculation tells us that λλ ∼ M ε2 · (α )3 (6. B 574 (2000) 263 [hep-th/9911096]. 2 (1998) 253 [hep-th/9802150]. References [1] J. Adv. Lett.S. Gubser. Phys. JHEP 0008 (2000) 052 [hep-th/0007191]. Oz. . Klebanov. 323 (2000) 183 [hep-th/9905111]. Maldacena. Phys. Gubser and I. [13] I. I. [11] I. Gubser. Gauge Theory and Strings M α ε2 ln(r/ε2/3 )eiψ (dθ1 − i sin θ1 dφ1 ) ∧ (dθ2 − i sin θ2 dφ2 ). the parameter ε2 which enters the deformed conifold equation has a dual interpretation as the gluino condensate6 . Rev.S. Kehagias. 6 It would be nice to understand the relative factor of gs M between Twall and λλ .

[34] J. B 600 (2001) 219 [hep-th/0011146]. Novikov. Class.S. Rev. Phys. Shifman. quivers. Klebanov. B 435 (1995) 129 [hep-th/9411149]. Zakharov. Nucl. [44] M. S. Polchinski. C. Lerda and R.R. [32] S. S. Vainshtein.J. A. [26] C. Gross and H. Romans.B. 2 (1985) L11. B 488 (1997) 159 [hep-th/9610205]. [20] D. More Anomalies from Fractional Branes [hep-th/0202195].J. Ferretti and I. B 460 (1999) 281 [hep-th/9904187]. Randjbar-Daemi. [38] V. [21] I. Nucl. D 54 (1996) 1667 [hep-th/9601038]. [39] P. JHEP 007 (1997) 004 [hep-th/9612126]. Tseytlin. de la Ossa. Herzog. [27] U. A.L. Douglas and G.R. Kachru and J. Candelas and X. A 16 (2001) 707 [hep-th/0011193]. Polchinski. A. [30] R. [33] A. P. Marotta. u [23] S.A. Ouyang and E. D 61 (2000) 066001 [hep-th/9905226]. M. P. Witten. [37] M. [24] E. Nucl. Nucl. R. Rev. Vainshtein and V. Lerda and R. .R. Marotta. Herzog. Herzog.: N = 1 Gauge/Gravity Dualities 421 [15] S. Phys. [35] A. Marcus. Phys.I.J. Dall’Agata.G. Peet and J. Bachas. I. Phys. Rev. 182 (1969) 1517.R. B 556 (1999) 89 [hep-th/9905104]. Moore. D 65 (2002) 105007 [hep-th/0202056]. [22] H. D-branes. Minasian and D. Witten. B 153 (1985) 392.A. Quant. D 58 (1998) 106002 [hep-th/9805129]. Suryanarayana. Gubser.I.A. Phys. Shifman and A. Phys. Ferrara. Douglas and M. M. Rev. Ceresole. [45] O. 79 (1997) 1984 [hep-th/9705084]. Phys. L. Lett. Phys. Rev. Rev. Susskind and E. I. Nucl. Phys. Phys. Phys. Witten. Phys. [42] I. Rev. Phys. G. Pando Zayas and A. [41] S. B 572 (2000) 499 [hep-th/9911042]. Nucl.C. Buchel. Gimon and J. Douglas. The holographic bound in anti-de Sitter space [hep-th/9805114]. Phys. Aharony. B 229 (1983) 381. Nucl. Klebanov and A. G. I.P. [28] E. Silverstein. [31] C. [19] A. Romans and P. Seiberg. D 32 (1985) 389. Klebanov. Tsimpis. [29] M.R. Frau. Tseytlin. Phys. B 277 (1986) 456. B 342 (1990) 246. Int. Witten. D 59 (1999) 065011 [hep-th/9809022]. Jatkar and S. Giddings. [25] D. Mod. JHEP 0104 (2001) 033 [hep-th/0102105]. P.W. Gubser. Danielsson. M. L. JHEP 9807 (1998) 006 [hep-th/9805112].R. Phys. Nucl. Adler and W. Frau. Polchinski. van Nieuwenhuizen. Phys. Ooguri.R. Lett. [43] M. Bertolini. Remarks on the warped deformed conifold [hep-th/0108101]. Lett. Di Vecchia. Grav. J. Buchel. Klebanov. M. Argyres. Nucl. Polchinski. D 59 (1999) 025006 [hep-th/9807164]. B 621 (2002) 157 [hep-th/0107057]. Hierarchies from ﬂuxes in string compactiﬁcations [hep-th/0105097].I. Ouyang.P. JHEP 9707 (1997) 002 [hep-th/9705074]. [16] M. [17] L. [36] L. D’Auria and S. Phys. Kim. Klebanov et al. Phys. Stable Non-BPS States and Their Holographic Duals [hep-th/0011185].S. Bertolini. Rev. M. Lecture Notes. Klebanov and E. Bardeen. Phys. C. Mukhi and N. G¨naydin and N.R. and ALE instantons [hep-th/9603167]. S.R.P. Phys. Kachru and E. [40] N. Klebanov and P. Green. [18] P. JHEP 0105 (2001) 028 [hep-th/0102172]. Di Vecchia.A. A.A. Rev.

Supersymmetry and F-theory realization of the deformed conifold with 3-form ﬂux [hep-th/0010010]. Harmonic c u Forms and Brane Resolutions [hep-th/0012011]. [48] M. c u M. Maldacena. H. . Phys. K. D. Brandhuber and K. Phys. 1 (1998) 1 [hep-th/9706082]. Phys. JHEP 0012 (2000) 014 [hep-th/0010048]. Yankielowicz. N. Henningson and K. C 22 (2001) 379 [hep-th/9803001].J. [64] E. paper to appear. Phys.N.Z.P. Skenderis. Freedman and K. Skenderis.G. Bigazzi. A. O. [58] C. Rev. 86 (2001) 588 [hep-th/0008001]. Dvali and M. [63] A.Z.N. Cvetiˇ. Park and R. Superstrings and Topological Strings at Large N [hep-th/0008142]. Becker and M. Girardello and A. Sonnenschein.G. [61] S. Phys. Tseytlin. Yokono. Brandhuber. Gibbons. Nucl.J. Lett. Hanany. JHEP 0005 (2000) 048 [hep-th/0003037]. Lett. [49] M. [55] K. M. Phys. B 396 (1997) 64 [hep-th/9612128]. Sonnenschein and S. [52] L. Phys. Theor. Phys. Rey and J. Panagopoulos. E. Rev. [54] S. Del Debbio. Douglas and S. Phys. B 447 (1995) 271 [hep-th/9503163]. D. JHEP 0101 (2001) 021 [hep-th/0009156]. Ricci-ﬂat Metrics. H. B 410 (1997) 27 [hep-th/9706127].422 Unity from Duality: Gravity. [70] G. Polchinski. Tatar. Tafjord. Phys. [66] C. k-string tensions in SU (N ) gauge theories [hep-th/0106185]. Maldacena and C. Vicari. JHEP 0011 (2000) 028 [hep-th/0010088]. K. [65] K. G. Zaﬀaroni. Phys. Hori. Ohta and T. Phys. J. M. Lett. B 507 (1997) 658 [hep-th/9706109]. A. Herzog and I. B 555 (1999) 183 [hep-th/9902197]. Krasnitz. Pope. Zaﬀaroni. [62] M. [57] J. JHEP 9807 023 (1998) [hep-th/9806087]. [72] F. Oh. 80 (1998) 4859 [hep-th/9803002]. Nucl. Gubser.H. Rev. [50] M. [56] M. Strassler and A. Becker. Ooguri and Y. Rev. J. Lucini and M. Oz. B 513 (1998) 87 [hep-th/9707244]. Kaplunovsky. D 64 105019 (2001) 105019 [hep-lat/0107007]. Phys. [71] A. Phys. DeWolfe. Lett. Gauge Theory and Strings [46] M. Nucl. Savvidy and O. Phys. Pando Zayas and A. L. Nucl. J. [51] K. Freedman and K. B 600 (2001) 103 [hep-th/0011023]. Grana and J. Rossi and E. L¨ and C. Nucl. Phys. Dasgupta. Shenker. [53] M.R. Yee. [59] K.S. [60] J. Nucl. B 500 (1997) 3 [hep-th/9703166]. Pope. H. Eur. [69] L.R. Nucl. JHEP 0201 (2002) 031 [hep-th/0110050]. Klebanov. V. Loewy and J. Holographic Renormalization [hep-th/0112119]. Callan. P. Phys. Bianchi. Douglas and C. Teper. Bachas. Vafa. Pilch. B 526 (2002) 388 [hep-th/0111078].W. D 63 (2001) 026001 [hep-th/0009211]. Itzhaki. Bianchi. [47] A. L¨ and C. Guijosa. Shifman. JHEP 0002 (2000) 023 [hep-th/9912266]. JHEP 0108 (2001) 007 [hep-th/0103163]. B. Witten.h. Lett. Sfetsos. Witten. Nucl. Schweigert. Cvetiˇ. Nunez. B 598 (2001) 530 [hep-th/0011041]. [68] C. Adv. B 477 (1996) 155 [hep-th/9605053]. H. [67] C. Math.

MA 02138.A. Harvard University. . STROMINGER Department of Physics.S. U. Cambridge.LECTURE 6 DE SITTER SPACE A.

. . .2 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . 447 A Calculation of the Brown-York stress tensor 451 . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Entropy . . 435 2. . . . . . . . . . . . 437 3. . 443 4 Quantum gravity in de Sitter space 446 4. . . . .1 Coordinate systems and Penrose diagrams . .3 Geodesics . . . . . . . 428 2. . . . . 436 3 Quantum ﬁeld theory on de Sitter space 437 3. . . . . . . . . . .2 Schwarzschild-de Sitter . . . . . . . . . . . . . . . .1 Asymptotic symmetries . . . 446 4. . . . . . . . . . . . . . . .1 Green functions and vacua . . . . . . . . . . . . .Contents 1 Introduction 425 2 Classical geometry of de Sitter space 427 2. . . . . . .2 De Sitter boundary conditions and the conformal group . . . . . . . . . . . . . . . . . . . . . . . . . 440 3. . . . . . . . . . . . .

and to C.DE SITTER SPACE M. 1 Introduction We begin these lectures with one of our favorite equations S= A · 4G (1. Bachas. Moscow. Landau Institute for Theoretical Physics. Russia c EDP Sciences. which says that the entropy S associated with an event horizon is its area A divided by 4G.V. McAllister for comments on the manuscript. 2]. is also supported by INTASOPEN-97-1312. F. M.2 . The ﬁnal lecture contains a pedagogical discussion of the appearance of the conformal group as an asymptotic symmetry group. Nekrasov for organizing a very pleasant and productive summer school and for arranging ﬁnancial support. This is a macroscopic formula.1) This is the Bekenstein-Hawking area-entropy law. Strominger1 and A. This work was supported in part by DOE grant DE-FG02-91ER40655. Maloney for useful discussions. A. A. Cambridge. Spradlin1. MA. M.3 Abstract These lectures present an elementary discussion of some background material relevant to the problem of de Sitter quantum gravity. 1 Harvard University. Princeton NJ. U. Volovich1 . U.S.A.S. We are grateful to R.S. Springer-Verlag 2002 . especially the temperature and entropy of de Sitter space. Bousso and A.A. is also supported by DOE grant DE-FG02-91ER40671. Herzog and L.D. 2 Princeton University. which is central to the dS/CFT correspondence. where G is Newton’s constant [1. It should be viewed in the same light as the macroscopic thermodynamic formulae that were ﬁrst studied in the 18th and 19th centuries. The ﬁrst two lectures discuss the classical geometry of de Sitter space and properties of quantum ﬁeld theory on de Sitter space. David. and A. and N. A (previously lacking) derivation of asymptotically de Sitter boundary conditions is also given. Douglas. 3 L. It describes how properties of event horizons in general relativity change as their parameters are We would like to thank C. Bilal.

Since then we’ve managed to creep away from the supersymmetric limit a little bit. Such an interpretation for the entropy (1. ultimately culminating in the AdS/CFT correspondence [5]. as well as to black strings and to all of the strange new objects we’ve found in string theory. Black holes can be supersymmetric. There has been some deﬁnite but still limited progress in understanding the microscopic origin of (1. but not very far. such as the horizon in de Sitter space.1) is still very limited. but almost no progress in understanding the entropy of de Sitter space? One reason is that one of the principal tools we’ve used for understanding black hole entropy is supersymmetry. and certainly we never managed to get all the way to Schwarzschild black holes. In some stringy cases this agreement has been achieved. One of the surprising and impressive features of this formula is its universality. and it is diﬃcult even to see where the quantum microstates that we would like to count are supposed to be. So supersymmetry is a crutch that we will need to throw away before we can do anything about de Sitter space. Nevertheless the progress towards a complete understanding of (1. In some ways cosmological horizons are much more puzzling than black hole horizons because in the black hole case one may expect that the black hole is a localized object with some quantum microstates. Gauge Theory and Strings varied. and indeed the ﬁrst black holes whose entropy was counted microscopically were supersymmetric. and in particular of the statistical origin of this law. like the event horizon in de Sitter space [3]. is undoubtedly one of the main keys to understanding what quantum gravity is and what the new notions are that replace space and time in quantum gravity. shapes and rotation. It applies to all kinds of black holes with all kinds of charges.1) that we have managed to understand has led to all kinds of interesting insights. That little piece of (1.1) was not given at the time that the law was discovered in the early 70s.426 Unity from Duality: Gravity. Why has there been signiﬁcant progress in understanding black hole entropy. It also applies to cosmological horizons.1). After Boltzmann’s work we tend to think of entropy in microscopic statistical terms as something which counts the number of microstates of a system. This behavior can be succinctly summarized by ascribing to them an entropy given by (1. On the other hand in de Sitter space the event horizon is observer dependent. A complete understanding of this law. Indeed there is a very simple . Then if you could provide the correct description of that localized object. because we only understand special kinds of black holes–among which Schwarzschild black holes are not included–and we certainly don’t understand much about cosmological event horizons.1) in very special cases of black holes which can be embedded into string theory [4]. you would be able to count those microstates and compare your result to the Bekenstein-Hawking formula and see that they agree.

it has recently been receiving more attention [7– 42]. in Section 4 we turn to some recent work on de Sitter quantum gravity. 1) symmetry. related. 1 See. where is a parameter with units of length called the de Sitter radius. A second. Finally. These lectures are mostly an elementary discussion of the background material relevant to the problem of de Sitter quantum gravity. (2. The Appendix contains a calculation of the asymptotic form of the Brown-York stress tensor. One reason for this is the recent astronomical observations which indicate that the cosmological constant in our universe is positive [43–46].1 .A. as shown in Figure 1. missing in previous treatments. which is the isometry group of dS d . 2 Classical geometry of de Sitter space The d-dimensional de Sitter space dS d may be realized as the hypersurface described by the equation 2 2 2 −X0 + X1 + · · · + Xd = 2 (2. So perhaps de Sitter quantum gravity is a nut ready to be cracked. Scalar quantum ﬁeld theory in a ﬁxed de Sitter background is in Section 3.1 . The embedding (2.: De Sitter Space 427 observation [6] that de Sitter space is inconsistent with supersymmetry in the sense that there is no supergroup that includes the isometries of de Sitter space and has unitary representations1. Strominger et al. The de Sitter metric is the induced metric from the standard ﬂat metric on Md. which leads to the dS/CFT correspondence [27]. While the importance of understanding de Sitter quantum gravity has been evident for decades. of the asymptotically de Sitter boundary conditions on the metric.1) is a nice way of describing de Sitter space because the O(d.1) in ﬂat d+1-dimensional Minkowski space Md. is manifest. obstacle to progress in understanding de Sitter space is that so far we have not been able to embed it in a fully satisfactory manner into string theory. The classical geometry of de Sitter space is described in Section 2. Furthermore one can show that dS d is an Einstein manifold with positive scalar curvature. A second reason is that recent progress in string theory and black holes provides new tools and suggests potentially fruitful new angles. . This section also contains a derivation.2) however [7]. This hypersurface in ﬂat Minkowski space is a hyperboloid. A pedagogical derivation is given of the appearance of the the two dimensional conformal group in three dimensional de Sitter space. and the Einstein tensor satisﬁes Gab + Λgab = 0.

where (d − 2)(d − 1) 2 2 is the cosmological constant. Hyperboloid illustrating de Sitter space. . = = cos θ1 . sin θ1 · · · sin θd−2 sin θd−1 . θi ). but 0 ≤ θd−1 < 2π.5) a.3) = 1.1 Coordinate systems and Penrose diagrams We will now discuss a number of coordinate systems on dS d which give diﬀerent insights into the structure of dS d . .428 Unity from Duality: Gravity. Gauge Theory and Strings X 0 Fig.4) where 0 ≤ θi < π for 1 ≤ i < d − 1. (2. sin θ1 cos θ2 . We will frequently make use of coordinates on the sphere S d−1 . (2. Henceforth we will set Λ= 2. which is conveniently parametrized by setting ω1 ω2 ω d−1 ωd = = . The dotted line represents an extremal volume S d−1 . Global coordinates (τ. sin θ1 · · · sin θd−2 cos θd−1 . . Then it is clear d that i=1 (ω i )2 = 1. and the metric on S d−1 is d dΩ2 = d−1 i=1 2 2 2 (dω i )2 = dθ1 + sin2 θ1 dθ2 + · · · + sin2 θ1 · · · sin2 θd−2 dθd−1 . This coordinate system is obtained by setting X0 = sinh τ. (2. 1.

which lie on the north or south pole respectively.1) for any point (τ. The surfaces marked I − . cos T (2.8) In these coordinates dS d looks like a d−1-sphere which starts out inﬁnitely large at τ = −∞.9) so that we have −π/2 < T < π/2. The Penrose diagram 2 contains all the information about the causal structure of dS d although distances are highly distorted.: De Sitter Space Xi = ω i cosh τ. It is not hard to check that these satisfy (2. . .6) we obtain the induced metric on dS d .11) So from the point of view of analyzing what null geodesics do in dS d we are free to work with the metric (2. s d−1 (2. If a geodesic is null with respect to the metric (2.7) plugging in (2. then it is also null with respect to the conformally related metric d˜2 = (cos2 T )ds2 = −dT 2 + dΩ2 .A. Conformal coordinates (T. ds2 = −dτ 2 + (cosh2 τ )dΩ2 . 429 (2.4). These coordinates are related to the global coordinates by cosh τ = 1 .10). . Note that a light ray which starts at the north pole at I − will exactly reach the south pole by the time it reaches I + inﬁnitely far in the future. Light rays travel at 45◦ angles in this diagram. From the ﬂat metric on Md.6) where −∞ < τ < ∞ and the ω i are as in (2. In this diagram each point is actually an S d−2 except for points on the left or right sides. then grows again to inﬁnite size as τ → +∞.10). They are the surfaces where all null geodesics originate and terminate. b. ωi ).10) This is a particularly useful coordinate system because it enables us to understand the causal structure of de Sitter space. . then shrinks to a minimal ﬁnite size at τ = 0. θi ). i = 1.1 2 2 2 ds2 = −dX0 + dX1 + · · · + dXd . d. .11). The metric in these coordinates takes the form ds2 = 1 (−dT 2 + dΩ2 ). I + are called past and future null inﬁnity. d−1 cos2 T (2. while timelike surfaces are more vertical than horizontal and spacelike surfaces are more horizontal than vertical. d−1 (2. which looks simpler than (2. (2. Strominger et al.

Gauge Theory and Strings I South Pole North Pole I Fig. 2. where a timelike observer will eventually have the entire history of the universe in his/her past light cone. This is qualitatively diﬀerent from Minkowski space.430 Unity from Duality: Gravity. One of the peculiar features of de Sitter space is that no single observer can access the entire spacetime. Penrose diagram for dS d . The dashed lines are the past and future horizons of an observer at the south pole. for example. Therefore in attempting to develop de Sitter quantum gravity we should be aware of what can and cannot be observed. Also shown in Figure 3 is the region O+ . We often get into trouble in physics when we try to describe more than we are allowed to observe–position and momentum in quantum mechanics. I South Pole North Pole North Pole I South Pole O O I I Fig. These diagrams show the regions O− and O+ corresponding respectively to the causal past and future of an observer at the south pole. for example. which is the only part of de Sitter space that an observer on the south pole will ever be able to send . while every point in the interior represents an S d−2 . 3. A horizontal slice is an S d−1 . This region is marked as O− in Figure 3. The conformal time coordinate T runs from −π/2 at I − to +π/2 at I + . A classical observer sitting on the south pole will never be able to observe anything past the diagonal line stretching from the north pole at I − to the south pole at I + . The north and south poles are timelike lines.

.13) These coordinates do not cover all of de Sitter space. Strominger et al. For example if she/he wishes to know the weather anywhere in the southern diamond. I South Pole North Pole O I Fig. Note that each slice is an inﬁnite ﬂat d−1-dimensional plane which extends all the way down to I − . The northern diamond on the left of 3 is completely inaccessible to an observer on the south pole. but only the region O− and are therefore appropriate for an observer on the south pole. d − 1. It is this region that is fully accessible to the observer on the south pole. The surfaces of constant t are spatial sections of de Sitter space which are inﬁnite volume d−1-planes with the ﬂat metric. . xi ). c. . 2 xi e−t . . i = 1. 2 (2. i = 1. . To deﬁne this coordinate system we take X0 Xi X d = = = 1 sinh t − xi xi e−t . . (2. The slices of constant t are illustrated in Figure 4. The dashed lines are slices of constant t in planar coordinates. d − 1. 4. or the upper diamond of O+ . The intersection O+ ∩ O− is called the (southern) causal diamond. This is not possible in the lower diamond of O− . to which a query can never be sent. 1 cosh t − xi xi e−t . Planar coordinates (t. From the diagram it .12) The metric then takes the form ds2 = −dt2 + e−2t dxi dxi . .: De Sitter Space 431 a message to. from which a response cannot be received.A. a query can be sent to the appropriately located weather station and the response received before I + is reached. .

e. This happens because I − is very large. . In the top and bottom diamonds. = 1−r (2. 1 − r2 (2. d−2 1 − r2 (2. with the south pole at r = 0. Gauge Theory and Strings is clear that every surface of constant t intersects I − at the north pole. The horizons are at r2 = 1. . r. d. . Eddington-Finkelstein coordinates (x+ . so that it becomes null. This coordinate system is constructed to have an explicit timelike Killing symmetry. If we write X0 Xa Xd = 1 − r2 sinh t. One of the reasons to want a timelike Killing vector is so that we can use it to deﬁne time evolution. This coordinate system is the de Sitter analog of the (outgoing) Eddington-Finkelstein coordinates for a Schwarzschild black hole. θa ).14) then the metric takes the form ds2 = −(1 − r2 )dt2 + dr2 + r2 dΩ2 . . The absence of a globally timelike Killing vector in de Sitter space has important implications for the quantum theory. = rω a .15) In this coordinate system ∂/∂t is a Killing vector and generates the symmetry t → t + constant. θa ). . but we are running out of letters! Note also that for these coordinates and the following ones we will need a parametrization of S d−2 . and you can get there along a spatial trajectory from anywhere in O− . a = 1. The t in these coordinates is not the same as the t in planar coordinates. It may seem puzzling–and is certainly one of the salient features of de Sitter space–that a spatial plane can make it to the inﬁnite past. r.432 Unity from Duality: Gravity. . 2 cosh t. Static coordinates (t. ∂/∂t is spacelike. In these coordinates the time t is not a Killing vector. not S d−1 . But from (2. we deﬁne x+ by the equation dx+ = dt + dr . . Starting from the static coordinates. while in the northern diamond the vector is pointing towards the past! Thus ∂/∂t in static coordinates can only be used to deﬁne a sensible time evolution in the southern diamond of de Sitter space. a = 1.16) . and the only manifest symmetries are translations and rotations of the xi coordinates. In Figure 5 we illustrate what the Killing vector ﬁeld ∂/∂t is doing when extended to the various diamonds of the Penrose diagram. . d − 2. d − 1.15) we see that at r = 1 the norm of ∂/∂t vanishes. or in other words to deﬁne the Hamiltonian. and the southern causal diamond has 0 ≤ r ≤ 1.

20) −x− · 2 f.: De Sitter Space 433 Fig. which we can solve to obtain x+ = t + In these coordinates the metric is ds2 = −(1 − r2 )(dx+ )2 + 2dx+ dr + r2 dΩ2 . Lines of constant x+ are the null lines connecting I − with the south pole depicted in Figure 3. We can also deﬁne x− = t − so that the metric takes the form ds2 = −(1 − r2 (x+ . and the southern causal diamond is the region with 0 ≤ r ≤ 1 on the right hand side.19) (2. Kruskal coordinates (U. This Penrose diagram shows the direction of the ﬂow generated by the Killing vector ∂/∂t in static coordinates.17) The same symmetries are manifest in this coordinate system as in the static coordinates since ∂/∂t at ﬁxed r is the same as ∂/∂x+ at ﬁxed r. d−2 where r = tanh x + 1 1+r ln .A. These coordinates cover the causal past O− of an observer at the south pole while still keeping the symmetry manifest. (2. x+ = − ln(−V ). d−2 (2. Finally we take x− = ln U.18) 1 1+r ln · 2 1−r (2. Strominger et al. 5. Past and future null inﬁnity I ± are at r = ∞. θa ). 2 1−r (2. V. x− ))dx+ dx− + r2 dΩ2 . The horizons (dotted lines) are at r 2 = 1.21) .

Gauge Theory and Strings in which case r= Then the metric takes the form ds2 = 1 (−4dU dV + (1 + U V )2 dΩ2 ). V UV=1 U UV=1 V= 0 UV=1 Fig.23) 1 + UV · 1 − UV (2. The north and south poles correspond to U V = −1. The southern diamond is the region with U > 0 and V < 0.22) These coordinates cover all of de Sitter space. .24) (2. d−2 and surfaces of constant τ are d−1-dimensional hyperbolic planes. 6. and I ± correspond to U V = 1.25) in which the metric takes the form ds2 = −dτ 2 + sinh2 τ (dψ 2 + sinh2 ψ dΩ2 ). and planar coordinates foliate de Sitter space with planes. UV=1 U =0 (2. the horizons correspond to U V = 0. In this Penrose diagram the coordinate axes U = 0 and V = 0 are the horizons. The arrows denote the directions of increasing U and V . θa ). and U V = 1 are I + and I − . V . ω a sinh τ sinh ψ.434 Unity from Duality: Gravity. . One can also foliate de Sitter space with spaces of constant negative curvature by using the hyperbolic coordinates X0 Xa Xd = = = sinh τ cosh ψ. The Kruskal coordinate system covers all of de Sitter space. U V = −1 are the north and south poles. . and θa for the Kruskal coordinates. d−2 (1 − U V )2 (2. X d as functions of U. . Exercise 1. Hyperbolic coordinates (τ . cosh τ . Find X 0 . . g. Global coordinates foliate de Sitter space with spheres. ψ.

(2. φ = rH φ. In d dimensions in static coordinates the SdS d metric takes the form ds2 = − 1 − 2m − r2 rd−3 dt2 + 1 1− 2m r d−3 − r2 dr2 + r2 dΩ2 . Note that the two horizons approach each other as m is increased. (2.26) where m is a parameter related to the black hole mass (up to some d-dependent normalization constant). which we abbreviate as SdS.29) In the rescaled coordinates the metric (2. This solution represents a black hole in de Sitter space.30) .: De Sitter Space 2. Strominger et al. One reason to introduce SdS is that it plays an important role in the work of Gibbons & Hawking [3] determining the entropy of pure de Sitter space.3. We can learn a little more about the solution (2.2 Schwarzschild-de Sitter 435 The simplest generalization of the de Sitter space solution is Schwarzschildde Sitter. In general there are two horizons (recall that these are places where the timelike Killing vector ∂/∂t becomes null).28) is simply ds2 = −dt 2 + dr 2 + r 2 dφ 2 . The fact that there is only a de Sitter horizon and not a black hole horizon is not surprising in light of the fact that in three dimensional ﬂat space there are no black holes. r = r/rH . at rH = 1 − 8GE.27) where we have normalized the energy E of the Schwarzschild black hole appropriately for√ three dimensions.27) by looking near r = 0. In three dimensions there is only one horizon. which will be reviewed in Section 3. one of which is the black hole horizon and the other of which is the de Sitter horizon. so that there is a maximum size black hole which can ﬁt inside de Sitter space before the black hole horizon hits the de Sitter horizon. and as E goes to zero this reduces to the usual horizon in empty de Sitter space. (1 − 8GE − r2 ) (2. For this purpose it will be convenient to focus on the three dimensional Schwarzschild-de Sitter solution [49] ds2 = −(1 − 8GE − r2 )dt2 + dr2 + r2 dφ2 . rH (2. where ds2 behaves like 2 ds2 ∼ −rH dt2 + dr2 2 2 2 + r dφ . d−2 (2.A.28) Now we can rescale the coordinates by deﬁning t = rH t.

this is related to the geodesic distance D(X. In general P = −1 when the antipodal point of X lies on the light cone of X . X ) between X and X by P = cos D. rather than a black hole. we note a few important properties of P for later use. but we can do something similar for dS d . We can easily write explicit formulas for P (X.3 Geodesics Our last topic in the classical geometry of de Sitter space is geodesics. It is useful to deﬁne a quantity P by R2 P (X. . so the two points X and X coincide or are separated by a null geodesic. That is the geodesic distance D. n + 1. 2 (2. For points in a common causal diamond. If the solution is maximally extended one ﬁnds there is also point-like mass of the same size at the north pole [49]. which are related by D = Rθ. Exercise 2. y i ) = cosh(t − t ) − e−t−t δij (xi − y i )(xj − y j ). with the embedding equation δij X i X j = R2 . t . then there is only one independent SO(n + 1)-invariant quantity that we can associate to the two points. It is a little harder to visualize. . X ) ≡ δij X i X j = R2 cos θ. Gauge Theory and Strings This looks like ﬂat space. 2. 1. To conclude. . at the south pole of dS 3 . φ is identiﬁed modulo 2πrH . 1) (2. . in planar coordinates we have 1 P (t. In general. in which case P = −1. Let us think of the sphere as being embedded in ﬂat Euclidean space. We can also consider taking antipodal points X = −X. X ) in the various coordinate systems discussed above. Show that SdS 3 is a global identiﬁcation of dS 3 . This quantity P will turn out to be a more convenient invariant to associate to two points in de Sitter space. but it is not quite ﬂat space because while φ was identiﬁed modulo 2π. the geodesic separating X and X is spacelike for P < 1 and timelike for P > 1. then the geodesic distance is equal to zero. j = 1. . . There we can deﬁne P (X.32) The expression for P is simple in terms of the X’s but can get complicated when written in a particular coordinate system. For example. It is clear that if we take two points on the sphere S n of radius R. . . You may be familiar with the fact that if you put a point-like mass in ﬂat three dimensional Minkowski space you would also get a conical deﬁcit angle at the location of the particle.27) as a pointlike mass.436 Unity from Duality: Gravity. .31) (recall that we have set the de Sitter radius to one). i. while for P < −1 the geodesic between X and the antipodal point of X is timelike. or equivalently the angle θ between them. ηij = diag(−1. Therefore there is a conical singularity with a positive deﬁcit angle at the origin. xi . X ) = ηij X i X j . Hence we recognize (2. If P = 1.

3. a complete understanding of the entropy-area relation (1. Y ) = P (Y. This turns out to be a very rich subject which has been studied by many authors [3. Strominger et al. . can not be written as a function of P alone. and/or by multiplying by a step function in time. 1)) invariant. which obeys the free ﬁeld equation (∇2 − m2 )G(X.5) 2 P (X.3) (3.: De Sitter Space 3 Quantum ﬁeld theory on de Sitter space 437 Ultimately. In this section we will take a baby step in that direction by considering a single free massive scalar ﬁeld on a ﬁxed background de Sitter spacetime. Y ) = 0|φ(X)φ(Y )|0 . Y ) will be de Sitter invariant. 50–60]. Hadamard and so on. (3.1 Green functions and vacua Let us consider a scalar ﬁeld in dS d with the action S=− 1 2 √ dd x −g (∇φ)2 + m2 φ2 . Y ) = 0. (3. Let us assume that the state |0 in (3. Y ) between X and Y 2 . There are other two point functions that one can discuss: retarded. but these can all be obtained from the Wightman function (3. which is SO(d.2) is invariant under the SO(d. X) is insensitive to the time ordering between points. advanced. 1) (but not O(d. Y )).2) where ∇2 is the Laplacian on dS d . and so at generic points can only depend on the de Sitter invariant length P (X. Y ) = G(P (X. Because of this the i prescription for G. We will study the Wightman function G(X.4) With the change of variable z = equation z(1 − z)G + 1+P 2 this becomes a hypergeometric d − dz G − m2 G = 0. Then G(X. (3.1) in de Sitter space will require an understanding of quantum gravity on de Sitter space. all information is encoded in the two-point function of φ. Writing G(X.3) reduces to a diﬀerential equation in one variable P 2 (1 − P 2 )∂P G − dP ∂P G − m2 G = 0. 2 (3. Feynman. for example by taking the real or imaginary part. as discussed below.A. 1) de Sitter group.2). (3.1) Since this is a free ﬁeld theory.

7) The hypergeometric function (3. .10) is nonsingular everywhere within an observer’s horizon. z .d F d h+ . This singularity sounds rather unphysical at ﬁrst.10) The singularity is now at P = −1. We can use this fact to ﬁx the normalization constant cm. The second linearly independent solution to (3.d = 41−d/2 Γ( d ) Γ(h+ )Γ(h− ) Γ(h+ )Γ(h− ) 2 × · = d d d/2 Γ( 2 )Γ( 2 − 1) 2(d − 2)π (4π)d/2 Γ( d ) 2 (3. Near z = 1 the hypergeometric function behaves as F d 1+P h+ . . or P = 1.9) The prescription for going around the singularity in the complex plane is also the same as in Minkowski space. Γ(h+ )Γ(h− ) (3. The singularity occurs when the points X and Y are separated by a null geodesic. h− . which corresponds to X being null separated from the antipodal point to Y . and a branch cut for 1 < P < ∞. and (3. Γ( d ) 2 Comparing with the usual short-distance singularity 2(d−2)πd/2 (D2 )1−d/2 ﬁxes the coeﬃcient to be cm.6) where cm. Hence the “unphysical” singularity can not be detected by any experiment. so if G(P ) is a solution then G(−P ) is also a solution. but we should recall that antipodal points in de Sitter space are always separated by a horizon. The Green function (3. At short distances the scalar ﬁeld is insensitive to the fact that it is in de Sitter space and the form of the singularity is precisely the same as that of the propagator in ﬂat d-dimensional Minkowski space.d is a normalization constant to be ﬁxed shortly. (3.4) clearly has a P → −P symmetry.d . 2 2 ∼ D2 4 1−d/2 Γ( d )Γ( d − 1) 2 2 . 2 (3.438 Unity from Duality: Gravity.10) can be thought of as arising from an image source behind the horizon. 2 2 · (3.6) has a singularity at z = 1. namely replacing X 0 −Y 0 with X 0 −Y 0 −i . h− . The equation (3. h− . . and h± = 1 (d − 1) ± 2 (d − 1)2 − 4m2 . Gauge Theory and Strings whose solution is G = cm. .8) where D = cos−1 P is the geodesic separation between the two points.4) is therefore F d 1−P h+ .

which when analytically continued back to de Sitter space yields (3. (3. k k (3.13) and are normalized with respect to the invariant Klein-Gordon inner product (uk .: De Sitter Space 439 De Sitter space therefore has a one parameter family of de Sitter invariant Green functions Gα corresponding to a linear combination of the solutions (3. we write an expansion for the scalar ﬁeld in terms of creation and annihilation operators of the form φ(X) = k ak uk (X) + a† u∗ (X) . u∗ ∼ eiEt f ∗ (x). u ∼ e−iEt f (x). (3. l ↔ (3.12) and which do we associate with annihilation operators? In Minkowski space we take positive and negative frequency modes. Let us say a few more words about the vacuum states. the dS d metric becomes the metric on the sphere S d .6). ul ) = −i dΣµ uk ∂ µ u∗ = δkl . De Sitter Green functions are often discussed in the context of analytic continuation to the Euclidean sphere. 59].10).12) where ak and a† satisfy k [ak .11) That is. On the sphere there is a unique Green function. (3. Y ) = α|φ(X)φ(Y )|α . However. If we work in static coordinates and take t → iτ .14) (3. as we try to understand the quantum theory of de Sitter space these funny extra vacua will surely turn out to have some purpose in life. a† ] = δkl . Strominger et al. Corresponding to this one-parameter family of Green functions is a one-parameter family of de Sitter invariant vacuum states |α such that Gα (X. l The modes uk (X) satisfy the wave equation (∇2 − m2 )uk = 0. These vacua are discussed in detail in [58. which modes do we associate with creation operators in (3. A vacuum state |0 is deﬁned as usual by saying that it is annihilated by all annihilation operators an |0 = 0.16) .A.6) and (3.15) where the integral is taken over a complete spherical spacelike slice in dS d and the result is independent of the choice of this slice. The question is. but are usually discarded as somehow “unphysical”.

(3. Thus we will conclude that de Sitter space is naturally associated with a temperature [53]. Let H denote the detector Hamiltonian. which we will calculate. It turns out [51. i. This is an important and general feature of quantum ﬁeld theory in curved spacetime. we would like to ﬁnd a way to divide the modes into u and u∗ that is invariant under SO(d. In order to preserve classical symmetries of dS d in the quantum theory. All observers will agree on whether or not the detector has binged.17) where g is the strength of the coupling and τ is the proper time along the observer’s worldline. If we make a choice of time coordinate.18) . by emitting or absorbing scalar particles. 3. The detector could for example be constructed so that it emits a “bing” whenever its internal energy state changes. But if we had made some other choice of time coordinate then we would have a diﬀerent vacuum |0 . Since the notion of a particle is observer-dependent in a curved spacetime. we can get a vacuum state |0 and then the state (a† )n |0 ≡ |n is said to have n particles in it. and a corresponding family of Green functions such as Gα . Hence the question “How many particles are present?” is not well-deﬁned independently of a choice of coordinates. 58. we must be careful to give a coordinate invariant characterization of the temperature. 59] that there is a family of such divisions. although they may disagree on whether the bing was caused by particle emission or absorption. (3. A good way to achieve this is to consider an observer equipped with a detector.e. But in a general curved spacetime there is no canonical choice of a time variable with respect to which one can classify modes as being positive or negative frequency. H|Ej = Ej |Ej . Such a detector is called an Unruh detector and may be modeled by a coupling of the scalar ﬁeld φ(x(τ )) along the worldline x(τ ) of the observer to some operator m(τ ) acting on the internal detector states ∞ g −∞ dτ m(τ )φ(x(τ )).440 Unity from Duality: Gravity.2 Temperature In this section we will show that an observer moving along a timelike geodesic observes a thermal bath of particles when the scalar ﬁeld φ is in the vacuum state |0 . Gauge Theory and Strings respectively to multiply the annihilation and creation operators. The resulting vacuum will then be de Sitter invariant. 1). The detector will have some internal energy states and can interact with the scalar ﬁeld by exchanging energy. which we could express as a linear combination of the |n ’s. with energy eigenstates |Ej .

The Green function is a function only of the geodesic distance P (x(τ ). where β| is any state of the scalar ﬁeld.21) gmji −∞ dτ ei(Ej −Ei )τ β|φ(x(τ ))|0 · (3. we should square the amplitude (3.22) Since we are only interested in the probability for the detector to make the transition from Ei to Ej . We can divide out this factor and discuss the transition probability per unit proper time along the detector worldline.20) m(τ ) = eiHτ m(0)e−iHτ . x(τ )). which will not be measured.24) has something to do with a thermal response is that the function G is periodic in imaginary time under τ → τ + 2πi. we ﬁnd the probability P (Ei → Ej ) = g 2 |mij |2 ∞ −∞ dτ dτ e−i(Ej −Ei )(τ −τ ) G(x(τ ). Using β |β β| = 1.A. (3. x(τ )) is the Green function (3.19) We will calculate the transition amplitude from a state |0 |Ei in the tensor product of the scalar ﬁeld and detector Hilbert spaces to the state Ej | β|. and Green functions which are periodic in imaginary time are thermal Green functions. which is then given by ˙ P (Ei → Ej ) = g 2 |mij |2 ∞ −∞ dτ e−i(Ej −Ei )τ G(cosh τ ).24) The ﬁrst hint that (3. To ﬁrst order in perturbation theory for small coupling g. the desired amplitude is ∞ g Using −∞ dτ Ej | β|m(τ )φ(x(τ ))|0 |Ei · (3.22) and sum over the ﬁnal state |β of the scalar ﬁeld.23) depends only on τ − τ and we get an inﬁnite factor from integrating over τ + τ .: De Sitter Space and let mij be the matrix elements of the operator m(τ ) at τ = 0: mij ≡ Ei |m(0)|Ej · 441 (3. x(τ )). this can be written as ∞ (3. then P is given in static coordinates by P = cosh(τ − τ ). . and if we consider for simplicity an observer sitting on the south pole.23) where G(x(τ ). Therefore everything inside the integral (3. (3. Strominger et al.2).

(3. so that (3. Ni = N e−βEi .26). Gauge Theory and Strings To investigate the nature of a thermal state. Consider integrating the function e−i(Ej −Ei )τ G(cosh τ ) around the contour shown in Figure 7.25) holds for the transition probabilities calculated in (3. Let us now show that (3.28). (3. Since the total integral around this contour is zero. (3. The integrand in (3.28) .26) where N is some normalization factor. The integrand in (3. 7. we have ∞ dτ −∞ −i(Ej −Ei )τ −∞−iβ G(cosh τ ) + +∞−iβ dτ e−i(Ej −Ei )τ G(cosh τ ) = 0.24).442 Unity from Duality: Gravity. This ﬁgure shows the contour C used in the integral (3. So (3.24) has singularities in the complex τ -plane at τ = 2πin for any integer n. Then it is clear that the total transition rate R from Ei to Ej is the same as from Ej to Ei : ˙ R(Ei → Ej ) = N e−βEi P (Ei → Ej ) = R(Ej → Ei ).24) has singularities in the complex τ -plane at τ = 2πin for any integer n. if the transition probabilities are related by (3.27) which is the principle of detailed balance in a thermal ensemble.25) and that the energy levels of the detector were thermally populated. In other words.25) describes the transition probabilities of a system in a thermal bath of particles at temperature T = 1/β.25) and the population of the states is thermal as in (3. The dotted lines signify the closure of the contour at inﬁnity. let us suppose it were true (as will be demonstrated shortly) that ˙ ˙ P (Ei → Ej ) = P (Ej → Ei )e−β(Ej −Ei ) . τ τ=0 C τ=−2πi Fig. then there is no change in the probability distribution for the energy levels with time.

The constant of integration is determined by requiring that a black hole of zero mass has zero entropy.31) to ﬁnd the entropy SdS .30) to ﬁnd the entropy. For the case of black holes one can use similar methods as those in the previous section to calculate the temperature TBH of the black hole. whereas in the black hole case there is a whole one parameter family of solutions labeled by the mass of the black hole. We will restrict our attention to dS 3 . The contour in Figure 7 corresponds to the pole prescription for the Wightman function as discussed in Section 3. Strominger et al. The problem in de Sitter space is that once the coupling constant of the theory is chosen there is just one de Sitter solution. what is EdS in (3.25).1.30) where EBH is the energy or mass of the black hole.: De Sitter Space 443 where β = 1/2π. We conclude that any geodesic observer in de Sitter space will feel that she/he is in a thermal bath of particles at a temperature TdS = 1 . where the analysis simpliﬁes considerably. The black hole entropy SBH can then be found by integrating the thermodynamic relation 1 dSBH = . the result for the temperature is the same for any observer moving along a timelike geodesic.A. all timelike geodesics in de Sitter space are related to each other by the SO(d. for ﬁxed coupling constant. In other words. Although we performed this calculation only for an observer stationary at the south pole. 3. So if you know the value of the temperature just for one value of EBH you will not be able to get the entropy. So for de Sitter space one would expect to use the relation dSdS 1 = dEdS TdS (3. 2π (3.31)? One might try to vary the cosmological .29) by dimensional where we have restored the factor of the de Sitter radius analysis. dEBH TBH (3. 1) de Sitter isometry group.3 Entropy In this subsection we will associate an entropy to de Sitter space. Now redeﬁning the variable of integration in the second integral as τ = −τ − iβ we get precisely the desired relation (3. Since the Green function used in this calculation is de Sitter invariant. but if you know it as a function of the black hole mass then you can simply integrate (3.

38) . dE TSdS (3.36) This diﬀers by a minus sign from the famous formula (1. The SdS 3 solution in static coordinates is ds2 = −(1 − 8GE − r2 )dt2 + dr2 + r2 dφ2 . 1 − 8GE = 2π one ﬁnds that the entropy is equal to SSdS = − AH · 4G (3.33) From the exercise and the discussion in the previous section we conclude that the temperature associated with the Schwarzschild-de Sitter solution is √ 1 − 8GE · (3. Show that this Green function is periodic in imaginary time with periodicity τ →τ+√ 2πi · 1 − 8GE (3.35) and writing the result in terms of the area AH of the de Sitter horizon at √ rH = 1 − 8GE which is given by √ AH . Let us instead follow Gibbons & Hawking [3] and use the one parameter family of Schwarzschild-de Sitter solutions to see how the temperature varies as a function of the parameter E labeling the mass of the black hole.35) but instead dSSdS 1 · = d(−EdS ) TSdS (3.444 Unity from Duality: Gravity.37) (3.34) TSdS = 2π Using the formula 1 dSSdS = . Exercise 3. (1 − 8GE − r2 ) (3.1)! What did we do wrong? Gibbons and Hawking suggested that to get the de Sitter entropy we should use not (3. but that is rather unphysical as it is the coupling constant. Gauge Theory and Strings constant.32) Find a Green function for SdS 3 by analytic continuation from the smooth Euclidean solution. One would be going from one theory to another instead of from one conﬁguration in the theory to another conﬁguration in the same theory.

Strominger et al. If we put something with positive energy on the south pole. In (3. 8. but we saw in one of the exercises that in the SdS 3 solution there is a positive deﬁcit angle at both the north and south poles. The de Sitter entropy. Consider a closed surface in de Sitter space such as the one shown in Figure 8. we are forced to ascribe negative energy to the positive deﬁcit angle at the north pole. If we ascribe positive energy to a positive deﬁcit angle at the south pole. is supposed to correspond to the entropy of the stuﬀ behind the horizon which we can’t observe. and ended up with the wrong sign in (3. If we ascribe positive energy to the positive deﬁcit angle at the south pole. The energy associated to the Killing vector ∂/∂t (indicated by the arrows) along the spacelike slice t = 0 (solid line) must vanish. then because the Killing vector ∂/∂t used to deﬁne the energy changes direction across the horizon.38). Therefore the northern singularity of Schwarzschild-de Sitter behind the horizon actually carries negative energy. although we don’t know exactly how to think about it. then we must ascribe negative energy to a positive deﬁcit angle at the south pole since the Killing vector ∂/∂t runs in the opposite direction behind the horizon.37). This can be seen quite explicitly in the Schwarzschild-de Sitter solution.A. I South Pole North Pole t= 0 I Fig. but if we more sensibly vary with respect to the energy at the north pole. then necessarily there will be some negative energy on the north pole.: De Sitter Space 445 This looks funny but in fact there is a very good reason for using this new formula. and hence vanishes on any closed surface.35) we varied with respect to the energy at the south pole. With no black hole. Then we arrive at the entropy for . which reduces to a surface integral on the boundary of the surface. Now in general relativity the expression for the energy on a surface is the integral of a total derivative. then we should use the formula (3. the spacelike slice in Figure 8 is an S 2 .

1) expresses the dS/CFT correspondence. Equation (4. . A gauge transformation which goes to zero at spatial inﬁnity will annihilate physical states (this is just the statement that a physical state is gauge invariant). which were assumed/guessed without derivation in [27].39) The integration constant has been chosen so that the entropy vanishes for 1 the maximal energy E = 8G at which value the deﬁcit angle is 2π and the space has closed up.1) where Oφ is an operator in the CFT associated to the ﬁeld φ. 11. while a gauge transformation which goes to a constant at spatial inﬁnity will act nontrivially on the states.1) indeed applies to three dimensional Schwarzschild-de Sitter. 14. about which little is established or understood. 26. Now we turn to the more challenging problem of quantum gravity in de Sitter space. in the sense that correlation functions of an operator φ inserted at points xi on I − or I + are generated by a two dimensional Euclidean CFT: φ(x1 ) · · · φ(xi ) dS 3 ↔ Oφ (x1 ) · · · Oφ (xi ) S2 .446 Unity from Duality: Gravity. The results of this section are largely contained in [27] except for the derivation of the asymptotic boundary conditions for dS 3 . In this section we will give a pedagogical discussion of several aspects of some recent eﬀorts in this direction [27] (some similar ideas appeared in [7. 47. 48]). 4 Quantum gravity in de Sitter space So far we have discussed well established and understood results about classical de Sitter space and quantum ﬁeld theory in a ﬁxed de Sitter background. In fact the generator will be proportional to the charge operator. 4G 2G (3. 4. by Noether’s theorem. (4. but the three dimensional case is the richest because of the inﬁnite dimensional nature of the I ± conformal group. The tool which will allow us to reach this conclusion is an analysis of the asymptotic symmetry group for gravity in dS 3 .1 Asymptotic symmetries Consider a simple U (1) gauge theory in ﬂat Minkowski space. Parallel results pertain in arbitrary dimension. In conclusion we see that the area-entropy law (1. Gauge Theory and Strings Schwarzschild-de Sitter SSdS = π √ AH = 1 − 8GE. We will argue that quantum gravity in dS 3 can be described by a two dimensional conformal ﬁeld theory.

and in principle there could be more than one choice. However if the boundary conditions are too restrictive. In this section we will see that the ASG of quantum gravity in dS 3 is the Euclidean conformal group in two dimensions. The states and correlators of the theory clearly must lie in representations of the ASG.: De Sitter Space 447 It is useful therefore to consider the so-called asymptotic symmetry group (ASG). but in some cases a knowledge of the ASG already places strong constraints on the theory. The result for de Sitter diﬀers only by a few signs. 4. Although we will not learn anything about the details of this theory. this means that correlators with points on I ± are those of a conformal ﬁeld theory.A. the fact that the conformal group in two dimensions is inﬁnite dimensional already strongly constrains the physics. In quantum gravity the relevant gauge symmetry is diﬀeomorphism invariance.2 De Sitter boundary conditions and the conformal group Our ﬁrst task is to specify the boundary conditions appropriate for an asymptotically dS 3 spacetime. Then the energy . ASG = Allowed Symmetry Transformations · Trivial Symmetry Transformations (4. and the correspondence (4. one might try to demand that the metric fall oﬀ spatially as r12 . However the physical interpretation in the dS 3 case is very diﬀerent from that of AdS 3 . Since this group acts on I ± . The analogous problem for three dimensional anti-de Sitter space was solved long ago by Brown & Henneaux [61]. This would allow only zero energy conﬁgurations and hence the theory would be trivial. which is deﬁned as the set of allowed symmetry transformations modulo the set of trivial symmetry transformations. Therefore we need to consider diﬀeomorphisms in dS 3 which preserve the boundary conditions on the metric at I ± but do not fall oﬀ so fast that they act trivially on physical states. the theory will become trivial. In general speciﬁcation of the boundary conditions is part of the deﬁnition of the theory. For example in 4d gravity. and in de Sitter space the only asymptotia are I ± . Strominger et al.1) is simply an expression of diﬀeomorphism invariance of the theory. and remains to be fully understood. and “trivial” means that the generator of the transformation vanishes after we have implemented the constraints–for example asymptotically vanishing gauge transformations in the example of the previous paragraph. Of course one must know the details of the theory to know which representations of the ASG actually appear. On the 1 other hand one might try to demand that it fall oﬀ as √r .2) Here “allowed” means that the transformation is consistent with the boundary conditions that we have speciﬁed for the ﬁelds in the theory.

= 3 Brown and York mainly consider a timelike boundary. z (4. 2 = −1 + O(e2t ). 66]). 63] (and applied to AdS3 in [65. but their results can be extended to the spacelike case. and hence all the symmetry generators. (4. We will deﬁne an asymptotically dS 3 spacetime to be one for which the associated stress tensor. and it is unlikely any sense can be made of the theory. 4 We caution the reader that the generalization of (4. 4G (4. = O(1).3) vanishes identically for vacuum dS 3 in planar coordinates ds2 = −dt2 + e−2t dzd¯.4) For a perturbed metric gµν + hµν we obtain the Brown-York stress tensor Tzz = Tz z = ¯ 1 1 hzz − ∂z htz + ∂t hzz + O(h2 ). Hence we need to understand the surface integrals which generate the diffeomorphisms of dS 3 . Gauge Theory and Strings and other symmetry generators are in general divergent. . A convenient and elegant formalism for this purpose was developed by Brown & York [62. The Brown-York stress tensor for dS 3 with = 1 is given by4 Tµν = 1 [Kµν − (K + 1)γµν ] .448 Unity from Duality: Gravity. are ﬁnite.3) Here γ is the induced metric on the boundary I − and K is the trace of the extrinsic curvature Kµν = −∇(µ nν) = − 1 Ln γµν with nµ the outward2 pointing unit normal. They showed that bulk diﬀeomorphisms are generated by appropriate moments of a certain stress tensor which lives on the boundary of the spacetime3 . Requiring the stress tensor to be ﬁnite evidently leads to the boundary conditions gz z ¯ gtt gzz e−2t + O(1).5) Details of this calculation are given in Appendix A. So the idea is to make the falloﬀ as weak as possible while still maintaining ﬁniteness of the generators. 4G 2 1 1 −2t 1 e htt − hzz + (∂z htz + ∂z ht¯ − ∂t hzz ) ¯ ¯ z ¯ 4G 4 2 + O(h2 ). (4.3) to d > 3 or to theories with matter is not entirely straightforward [66].

7) where U = U (z) is holomorphic in z 5 . 2 = −1 + O(e2t ).6) It is not hard to see that the most general diﬀeomorphism ζ which preserves the boundary conditions (4.8). 65].A. In other words. then applying ¯ the diﬀeomorphism ζ = e2t f ∂z gives δζ gtz = O(e2t ). = O(1).: De Sitter Space gtz = O(1). Strominger et al.6) is violated very near the singularity). Therefore one can ¯ always set the component gtz of the metric to be O(e2t ) with a trivial diﬀeomorphism.6) may be written as 1 ζ = U ∂z + U ∂t + O(e2t ) + complex conjugate. The third term is the familiar linearization of the anomalous Schwarzian derivative term corresponding to a central charge c= 3l . 8G (4.9) where we have restored the power of 6 . The resolution of this apparent discrepancy comes from noting that if gtz → f on the boundary where f = f (z.7) do not contribute in (4. = (4. Note that the O(e2t ) terms in (4. 2G (4. . = O(e2t ).7) is the conformal group of the Euclidean plane. In principle this should be carefully justiﬁed (as (4. Therefore they are trivial diﬀeomorphisms. The last boundary condition (4.6) diﬀers from the condition gtz = O(e2t ) assumed in [27] and obtained by analytically continuing the AdS 3 boundary conditions of Brown & Henneaux [61] from anti-de Sitter to de Sitter space. Exploiting this freedom one can impose the asymptotic boundary conditions gz z ¯ gtt gzz gtz e−2t + O(1).10) 5 We allow isolated poles in z. 449 (4. 2 (4. 6 Parallel derivations of the central charge for AdS were given in [64. then in fact one can always choose a gauge in which gtz = O(e2t ). z) is an arbitrary function.7) acts on the Brown-York stress tensor as δζ Tzz = −U ∂Tzz − 2U Tzz − 1 U .8) The ﬁrst two terms are those appropriate for an operator of scaling dimension two. A parallel issue arises in AdS3 /CFT2 . We conclude that the asymptotic symmetry group of dS 3 as generated by (4. in the sense described above. and we have not done so here. if gtz = O(1). A diﬀeomorphism of the form (4.

γ are complex constants. A special case of (4. C) subgroup of the asymptotic symmetry group. and a normal piece. A three dimensional diﬀeomorphism is thereby equivalent to a two dimensional conformal transformation.11) where α. This is particular to the three dimensional case. Now the form of the metric (4.12) followed by a Weyl transformation (4. In particular. Gauge Theory and Strings as given in [27]. (4. Where do conformal transformations come from? Recall that a conformal transformation in two dimensions is a combination of an ordinary diﬀeomorphism and a Weyl transformation.4) makes it clear that this can be compensated by a shift in t. we conclude that the asymptotic symmetry group of gravity in dS 3 is the conformal group of the complex plane. So 2 a diﬀeomorphism in dS 3 splits into a tangential piece. since in higher dimensional de Sitter space the ASG is the same as the isometry group SO(d. ¯ ¯ (4.12) A conformal transformation is just an ordinary diﬀeomorphism (4. Now if we look at what the diﬀeomorphism ζ deﬁned in (4.7) is the choice U = α + βz + γz 2 .13) with φ chosen so that gzz → gz z ¯ ¯ under the combined transformation. ¯ dz d¯ z (4. These transformations generate the SL(2. β. In this case U vanishes. (a) Find an example of string theory on de Sitter space. 1). a fact which highly constrains quantum gravity on dS 3 . In two dimensions a diﬀeomorphism acts as gz z → ¯ and a Weyl transformation acts as gzz → e2φ gzz .450 Unity from Duality: Gravity. which acts like an ordinary diﬀeomorphism of the complex plane.13) z dz d¯ gz z . (b) Find the dual conformal ﬁeld theory. This accounts for the second term 1 U ∂t in ζ. C) global isometries of dS 3 . the ASG is inﬁnite dimensional. We conclude these lectures with a last Exercise 4. which from the point of view of the z-plane is a Weyl transformation. we see that the ﬁrst term U ∂z generates a holomorphic diﬀeomorphism of the plane. and the dS 3 metric is therefore invariant. .7) does. which acts like a Weyl transformation. Since U (z) was arbitrary. The isometry group is the SL(2.

Strominger et al.2) (A.3) and the induced metric on the boundary I − is γzz = hzz . z ¯ N z = 2e2t htz .1) and we will always drop terms of order O(h2 ). ¯¯ ¯¯ (A. 2e2t htz z 2 · (A.8) into (4.: De Sitter Space A Calculation of the Brown-York stress tensor 451 We wish to calculate the Brown-York stress tensor (4.A.5).7) and (A. N z = N 1 −1 − htt . 0.5) Upon lowering the indices. We can put (A.4) The outward pointing unit normal vector to the boundary is nµ = 1 ¯ −1. = = 1 −∂z htz + ∂t hzz . z. we have nµ = 1 1 − htt . where the lapse and shift functions are given by 1 N = 1 − htt .7) . 2e2t ht¯. ¯ z ¯ 2 2 2 (A.3) gives the desired result (4. We write the metric in planar coordinates (t. ¯ ¯ z ¯ (A. γzz = ¯ 1 −2t e + hz z .1) into the form ds2 = −N 2 dt2 + γij (dxi + N i dt)(dxj + N j dt). 2 1 1 1 − e−2t (1 + htt ) − (∂z htz + ∂z ht¯ − ∂t hzz ) .8) Plugging (A. ¯ 2 γzz = hzz .6) and we use the formula Kµν = − 1 (∇µ nµ + ∇ν nµ ) to obtain 2 Kzz Kz z ¯ The trace is K = g µν Kµν = γ ij Kij = −2 − htt + 4e2t hzz − 2e2t (∂z htz + ∂z ht¯ − ∂t hzz ) . N z . z) as ¯ ds2 = gµν dxµ dxν = −dt2 + e−2t dzd¯ + hµν dxµ dxν . (A. 2 N z = 2e2t ht¯. 0 2 (A. z (A.3) for a metric which is a small perturbation of dS 3 .

Hull. D 59 (1999) 047503 [hep-th/9810169]. Waldron. JHEP 0105 (2001) 043 [hep-th/0103171]. Wu. Quantum gravity in de Sitter space [hep-th/0106109]. N. Vafa. A. Rev.in/strings/Proceedings. Nojiri and S. Fischler. Some aspects of the de Sitter/CFT correspondence [hep-th/0106247]. Khuri. B 575 (2000) 231 [hep-th/9911082]. and de Sitter gravity [hep-th/0107067]. J. Witten. Hawking. Horava and D. Nucl. Klemm. Y. Phys. Strominger. B 607 (2001) 577 [hep-th/0103198]. Phys. . R. Sohnius. T.I. Brotz and M.W. J. Banks and W. D.C. Gao. Phys. Maldacena. Balasubramanian. A. Phys. Hawking. Phys. B 453 (1999) 222 [hep-th/9901147].M. F. quantum mechanics and the cosmological constant [hep-th/0107031]. Quantum gravity in de Sitter space. Phys. M. T. Phys. Lett. Phys.P. Annals Phys. M. van Nieuwenhuizen and M. B 379 (1996) 99 [hep-th/9601029]. JHEP 0105 (2001) 001 [hep-th/0002145]. J. Math. matrices. Volovich. Int. R. 38 (1998) 1113 [hep-th/9711200]. Matrix model for de Sitter [hep-th/0106184].T. Strominger. Woodard. Theor. Banks. Waldron. Lett. Phys. Bros. Commun. (A)dS backgrounds from asymmetric orientifolds [hep-th/0106209]. Minic. S. C. Bousso. Epstein and U. The dS/CFT correspondence [hep-th/0106113]. Deser and A. Bekenstein. Lett. Theor. B 513 (2001) 137 [hep-th/0105181]. Zero-branes. P. P. Hull and R. G. Rev. D 15 (1977) 2738.W. A. D. Gibbons and S. 253 (1997) 1 [hep-ph/9602316]. Phys. Silverstein. 98 (1985) 105. Bousso. Strominger. R. J. Strominger and C. Deser and A. S. B 440 (1998) 275 [hep-th/9806119].W. K. E. Lett. Kim. Adv.tifr. Witten. S.452 Unity from Duality: Gravity. Tsamis and R.res. Pilch. JHEP 0011 (2000) 038 [hep-th/0010252]. Cosmological breaking of supersymmetry or little Lambda goes back to the future. JHEP 9807 (1998) 021 [hep-th/9806146]. C. S. M. Maldacena and A. Banados. Nucl. S. Gauge Theory and Strings 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] J. J. V. Phys. Chamblin and N. R.M. Math. Symmetries. D 7 (1973) 2333. 43 (1975) 199.D. JHEP 0104 (2001) 035 [hep-th/0012052]. JHEP 9802 (1998) 014 [gr-qc/9801096]. Rev. Maldacena and A. A. Ortiz. H. Phys. Moschella. E. E. W. M-theory observables for cosmological space-times [hep-th/0102077]. Discreteness in deSitter space and quantization of Kaehler manifolds [hep-th/0101176]. JHEP 9906 (1999) 028 [hep-th/9906022]. Hawking. Conformal anomaly from dS/CFT correspondence [hep-th/0106191]. Lambert. Math. Phys. The asymptotic symmetry of de Sitter spacetime [hep-th/0107091].F. Bousso. Rev. T. II [hep-th/0007146]. Odintsov. Commun. Phys. 2 (1998) 231. Lin and Y. Strings 2001 online proceedings http://theory. D 59 (1999) 046002 [hep-th/9807216]. Li. Park.E. D.

Horizon holography [hep-th/0107173]. [63] J. Solodukhin and K. Ann. Phys. Mazur and E. 114 (1978) 332. Solodukhin. York. Deser and R. S. D 30 (1984) 1733. [38] A. D 31 (1985) 710. Nucl. A 9 (1968) 109. Urbantke. Phys. Phys. Strominger et al. Quantization of the massless minimally coupled scalar ﬁeld and the dS/CFT correspondence [hep-th/0108119]. Brown and J. Brown. Folacci. Henneaux.H. 153 (1984) 405. [36] E. Chernikov and E. Phys. B 195 (1982) 76.D. Perlmutter et al.R.. Phys.O. [62] J. Hoegh-Krohn and C. Math. de Haro. Skenderis. Nappi and E.W. ApJ 517 (1999) 565 [astro-ph/9812133].D. [58] E.H. Commun.A. Dolan. Conformal operators for partially massless states [hep-th/0109096]. [42] C. Nojiri and S. 44 (1975) 265. Rev. Phys. Allen. R. D 47 (1993) 1407. [64] M. Kraus. Phys. D. Odintsov. 1982). Mazur and E. inﬂationary brane-world creation and dS/CFT correspondence [hep-th/0107134]. Henningson and K. Rumpf and H. N = 2 supersymmetry and de Sitter space [hep-th/0109168]. Rev. Ida and T. Skenderis. Mottola. D 64 (2001) 104022 [hep-th/0106151]. S. C. Int. Antoniadis. Commun. Birrell and P. Najmi and A. Phys. Ottewill. [37] I. Cambridge. De Sitter Space in Supergravity and M Theory [hep-th/0109213]. Rev. D.W. [54] H.O.R. D 31 (1985) 754. Allen and A. [59] B.A.A. [46] S.A. [45] S. [61] J. Phys. Commun. [39] T. Quantum cosmology. Figari. JHEP 9807 (1998) 023 [hep-th/9806087].D. D 35 (1987) 3771.: De Sitter Space 453 [35] S. Mod. York. 76 (1973) 561. [50] N. [41] R. Perlmutter. 104 (1986) 207. Math. [49] S.F. Kallosh. Theor. J. Ann. Phys. Deser. [57] L. [47] I. Ann.G.W. A 15S1 (2000) 715 [eConfC 990809 (2000) 715].P.J. Quantum Fields in Curved Space (Cambridge University Press. dS/CFT correspondence and cosmic no-hair [hep-th/0109057]. [66] S. [55] L. [53] R. Phys. Annals Phys. Commun. Mottola.C.M. [43] B. 217 (2001) 595 [hep-th/0002230]. Turok. Nappi. Lau and J. ApJ 507 (1998) 46 [astro-ph/9805200]. Tolley and N. [48] P. Poincare Phys. Phys. P. D 32 (1985) 3136. Comment on Nongaussian isocurvature perturbations from inﬂation [astro-ph/9705200]. Ford. Tagirov. Phys.D. [52] E.C. Halyo. Brown and M. Shiromizu. [56] A. [65] V. Rev. Jackiw. Balasubramanian and P. Tagirov. Math. Davies.K. Phys. 208 (1999) 413 [hep-th/9902121]. ApJ 116 (1998) 1009 [astro-ph/9805201]. Math. Schmidt et al. [60] B. Gravitational energy.N. Rev. Witten. Mottola. Torii. Action and Energy of the Gravitational Field [gr-qc/0010024]. . Sachs and S.. Phys. [51] N. Rev. De Sitter entropy and strings [hep-th/0107169]. Rev.R. Riess et al. [44] A.. Abbott and S. [40] L. N. Hull.

35 route de Chartres. France .LECTURE 7 STRING COMPACTIFICATION WITH N = 1 SUPERSYMMETRY M.R. 91440 Bures-sur-Yvette. DOUGLAS Rutgers and IHES.

Contents 1 Introduction and the question of vacuum selection 2 From ten dimensional geometry to four dimensional eﬀective ﬁeld theory 3 D-branes with stringy corrections 4 Quantum corrections 5 Towards the low energy theory of everything 457 461 466 470 472 .

This situation is very good from the point of view of the c EDP Sciences. 1 Introduction and the question of vacuum selection The last Les Houches summer school which treated the fundamental theories of physics: gauge theory.R. and the more recent Dirichlet brane paradigm. This changed at the end of 1995 with the Strominger-Vafa derivation of the Bekenstein-Hawking entropy formula from string theory. We now have some non-perturbative understanding of the theory. the subject has changed dramatically. Many of the central topics of our 2001 school. Although compared to 1994–97. and some new and radically diﬀerent pictures of space-time. it is not in obvious crisis. The new ideas have still not been completely understood and absorbed. solutions of gauge theory using brane techniques. string theory today appears relatively stable. In the six years since. it had said little about the main physical questions identiﬁed in previous work. and the ﬁeld remains very active and lively. gravity and the string and M theories which contain both of these. and the relevance of noncommutative ﬁeld theory. following the paradigms of heterotic string compactiﬁcation on a Calabi-Yau manifold. there is still much to do. Much of the attention has shifted towards possible application to the real world. F theory. in what is often called the “second superstring revolution”. Springer-Verlag 2002 . By 1997 the main ideas of this second revolution had emerged.STRING COMPACTIFICATION WITH N = 1 SUPERSYMMETRY M. the AdS/CFT correspondence. were in 1995 either curiosities known to few (such as Dirichlet branes) or had not been conceived of at all. such as M(atrix) theory. was held in 1995 [29]. Although string theory was by then widely accepted as the leading candidate theory of quantum gravity. Douglas Abstract We give a very broad overview of problems and recent work in the study of string compactiﬁcation with N = 1 supersymmetry. Even more surprising discoveries were to follow.

a particular theory and internal space. each check for a possible conﬂict with observation could be regarded as a vacuum selection principle. which states that if choices enter in deriving predictions from string/M theory. At present there are not even good speculations as to how many four-dimensional compactiﬁcations of string/M theory there might be. Early hopes that many compactiﬁcations would be inconsistent at the nonperturbative level now seem ill-founded. but instead check whether observations can be ﬁt by any of the possible choices. Many new ways were found to get gauge symmetry. Just as importantly. So let us try to look ahead a bit to the “third superstring revolution”. where opinions diﬀer is on how. the heterotic bound on the rank of the gauge group is easy to exceed. in introducing material of lasting importance (we think). but attempts to make more speciﬁc statements foundered on the problem that these depend on the choice of manifold.458 Unity from Duality: Gravity. we know weak coupling nonperturbative consistency conditions for backgrounds. the speakers could feel that their time was well spent. and we have strong coupling duals for enough examples which satisfy these conditions to believe they are suﬃcient. and from this point of view one might ﬁnd the school a bit too cut and dried. But there need be no single principle which in itself governs the problem. From this point of view. It seems to me that the third superstring revolution will have to address the problem of vacuum selection. the second superstring revolution only seems to make this problem worse. At the other extreme is the null hypothesis. as the outstanding barrier to contact with observation. The original scenario of E8 × E8 heterotic string theory compactiﬁed on a six dimensional Calabi-Yau manifold led quite directly to theories resembling the standard model. Gauge Theory and Strings organizers of a school. one simply needs to sort through all possibilities and apply a series of tests to each. which clearly enlarge the number of models. one should not rule out possibilities a priori. after the primary test of consistency. Perhaps the most basic question from the ﬁrst revolution left untouched by the developments of the second. I think many would agree with this. It does seem to me however that too much importance . as it allowed us to make a well motivated choice of topics. For example. Far from helping. and there are thousands if not millions (if indeed the number is ﬁnite!) of such manifolds. On the other hand. which will uniquely ﬁx four macroscopic dimensions. Of course these are two extreme points of view and the truth is likely to lie in between. students (and everyone else