A Detailed Introduction To String Theory

A Detailed Introduction to String Theory
Andy Svesko
June 2, 2013
2
Contents
Acknowledgements 8
Preface 9
I Fundamentals and the Bosonic String 15

1 Introduction 17
1.1 A Brief History of Modern Physics and the Mark of String Theory . . . . . . . . . . 17
1.2 A Qualitative Look at Renormalization Theory . . . . . . . . . . . . . . . . . . . . . 19
1.3 Fundamental Units and the Planckian System . . . . . . . . . . . . . . . . . . . . . . 21
1.4 Planck Length in Arbitrary Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.5 Compactification and Large Extra Dimensions . . . . . . . . . . . . . . . . . . . . . 23
1.6 Verifying String Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2 Special Relativity and Light Cone Coordinates 29

2.1 Special Relativity and Einstein’s Postulates . . . . . . . . . . . . . . . . . . . . . . . 29
2.2 Space and Time, and Space-time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3 A Quick Glance at Four-Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4 Vectors, Dual Vectors, and Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.5 Coordinate Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.6 The Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.7 Index Gymnastics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.8 Light Cone Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3 Non-Relativistic Strings and the Relativistic Point Particle 51

3.1 Basic Lagrangian Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2 Lagrangian Field Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.3 Gauge Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.4 Classical String Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.5 Lagrangian for a Non-Relativistic String . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.6 The Relativistic Point-Particle in Minkowski Space . . . . . . . . . . . . . . . . . . . 65
3.7 The Relativistic Point Particle in a Curved Space-time . . . . . . . . . . . . . . . . . 68
3.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3
4 CONTENTS
4 Classical Relativistic Strings 73

4.1 Preparation for the Action of a Relativistic String . . . . . . . . . . . . . . . . . . . 73
4.2 The Nambu-Goto Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3 Parameterizing the String . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.4 Transverse Velocity and the String Action . . . . . . . . . . . . . . . . . . . . . . . . 82
4.5 Motion of Open and Closed Strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5 Relativistic Strings and Mode Expansions 93

5.1 A More General Parameterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.2 Constraining the Wave Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.3 Solving the Wave Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.4 Solutions to the Wave Equation in the Light-Cone Gauge . . . . . . . . . . . . . . . 102
5.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6 Charges, Currents, and Symmetries 107

6.1 Conserved Quantities and Noether’s Theorem . . . . . . . . . . . . . . . . . . . . . . 107
6.2 Worldsheet Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.3 Lorentz Charges and Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
7 A Crash Course on Quantum Field Theory 117

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
7.2 The Klein-Gordon Equation and Scalar Fields . . . . . . . . . . . . . . . . . . . . . . 118
7.3 Quantization of Free Scalar Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
7.4 Constructing the State Space for Scalar Fields . . . . . . . . . . . . . . . . . . . . . 125
7.5 Charged Scalar Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
7.6 Time-Ordering and the Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.7 Light-Cone Coordinates and Scalar Fields . . . . . . . . . . . . . . . . . . . . . . . . 131
7.8 The Dirac Equation and Spinor Fields . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7.9 Free Spinor Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.10 The Dirac Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
7.11 Light-Cone Coordinates and Photon States . . . . . . . . . . . . . . . . . . . . . . . 139
7.12 Gravitational Fields and Gravitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.13 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
8 Quantizing the Relativistic Point Particle 147

8.1 Quantization and Point Particle States . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.2 Light-cone Momentum Operators and Symmetry Transformations . . . . . . . . . . 155
8.3 The Lorentz Charges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
8.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
9 Light-Cone Quantization of the String 165

9.1 Quantizing the String . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
9.2 Commutation Relations for the Oscillators of Open and Closed Strings . . . . . . . . 169
9.3 Strings and Harmonic Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
9.4 The Transverse Virasoro Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
CONTENTS 5
9.5 Lorentz Generators and the Dimension of Space-Time . . . . . . . . . . . . . . . . . 190

9.6 Constructing the State Space of Open String Theory . . . . . . . . . . . . . . . . . . 193
9.7 Constructing the State Space of Closed String Theory . . . . . . . . . . . . . . . . . 195
9.8 Unoriented Strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
9.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
10 Lorentz Covariant Quantization 203

10.1 The Covariant Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
10.2 Virasoro Operators and Quantum Constraints . . . . . . . . . . . . . . . . . . . . . . 204
10.3 Constructing the Lorentz Covariant State Space . . . . . . . . . . . . . . . . . . . . . 211
10.4 The Polyakov Action and “Modern” Covariant Quantization . . . . . . . . . . . . . . 214
10.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
II Advanced Topics and Superstrings 223

11 Conformal Field Theory and BRST Quantization 225
11.1 Introduction to Group Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
11.2 Conformal Transformations and the Conformal Group . . . . . . . . . . . . . . . . . 229
11.3 The 2-Dimensional Conformal Group . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
11.4 Propagators and Operator Product Expansions . . . . . . . . . . . . . . . . . . . . . 238
11.5 Primary Fields and the Verma Module . . . . . . . . . . . . . . . . . . . . . . . . . . 247
11.6 BRST Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
11.7 BRST Symmetry and String Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
11.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
12 D-Branes 257
12.1 Some Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
12.2 Quantizing the String on Dp-Branes . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
12.3 Stretched Strings and Parallel Dp-Branes . . . . . . . . . . . . . . . . . . . . . . . . 262
12.4 Multiple D-Branes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
12.5 Strings Between D-branes of Different Dimension . . . . . . . . . . . . . . . . . . . . 266
12.6 String Charge and D-Brane Charges . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
12.7 Tachyons and D-brane Decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
12.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
13 T-Duality, Symmetries, and Compactification 277

13.1 Quantizing the Closed String on a Compactified Space . . . . . . . . . . . . . . . . . 279
13.2 Constructing the State Space of Compactified Closed Strings . . . . . . . . . . . . . 286
13.3 T-Duality of Closed Strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
13.4 T-Duality of Open Strings on D-Branes . . . . . . . . . . . . . . . . . . . . . . . . . 290
13.5 A Closer Look at U(1) Gauge Transformations . . . . . . . . . . . . . . . . . . . . . 293
13.6 The Aharonov-Bohm Effect and Wilson Lines . . . . . . . . . . . . . . . . . . . . . . 295
13.7 T-duality of Open Strings in the Presence of Wilson Lines . . . . . . . . . . . . . . . 297
13.8 A Brief Aside on Real and Complex Manifolds . . . . . . . . . . . . . . . . . . . . . 300
13.9 Orbifolds and the Twisted Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
13.10 Orientifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
6 CONTENTS
13.11 Calabi-Yau Manifolds and Mirror Symmetry . . . . . . . . . . . . . . . . . . . . . . 307

13.12 String Theory, Particle Physics, and the Multiverse . . . . . . . . . . . . . . . . . . 309
13.13 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
14 A Crash Course in Supersymmetry 313

14.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
14.2 A Review of Weyl Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
14.3 Lorentz Transformations of Weyl Spinors . . . . . . . . . . . . . . . . . . . . . . . . 317
14.4 The Spinor ‘dot product’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
14.5 Charge Conjugation and Weyl Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . 321
14.6 Massive Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
14.7 Building a Simple Supersymmetric Lagrangian . . . . . . . . . . . . . . . . . . . . . 325
14.8 A Review of Charges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
14.9 The Supersymmetric Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
14.10 Classifying Quantum States Using Algebra . . . . . . . . . . . . . . . . . . . . . . . 339
14.11 Classifying States with Supercharges . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
14.12 Some More Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
14.13 Superspace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
14.14 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
15 An Introduction to Superstrings 357

15.1 Adding Fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
15.2 SUSY Transformations of the World-sheet and Conserved Currents . . . . . . . . . 359
15.3 The World-Sheet and Superspace . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
15.4 Boundary Conditions and Mode Expansions . . . . . . . . . . . . . . . . . . . . . . 367
15.5 Canonical Quantization of RNS Superstrings . . . . . . . . . . . . . . . . . . . . . . 369
15.6 Constructing the State Space for RNS Superstrings . . . . . . . . . . . . . . . . . . 371
15.7 Generating Functions and GSO Projection . . . . . . . . . . . . . . . . . . . . . . . 375
15.8 A Summary of Superstring Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
15.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
III General Relativity and String Theory 383
16 The Thermodynamics of Strings 385

16.1 A Brief Review of Thermal and Statistical Physics . . . . . . . . . . . . . . . . . . . 385
16.2 Density Operators and von Neumann Entropy . . . . . . . . . . . . . . . . . . . . . 389
16.3 Partitions and the Non-Relativistic String . . . . . . . . . . . . . . . . . . . . . . . . 392
16.4 The Hagedorn Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398
16.5 Partition Function of the Relativistic Point Particle . . . . . . . . . . . . . . . . . . 399
16.6 Partition Function of a Bosonic String . . . . . . . . . . . . . . . . . . . . . . . . . . 402
16.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
CONTENTS 7
17 Elements of Differential Geometry 407

17.1 Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
17.2 The Covariant Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
17.3 Curvature, Parallel Transport and Geodesics . . . . . . . . . . . . . . . . . . . . . . 415
17.4 Killing Vectors and Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422
17.5 Exterior Algebra and Exterior Differentiation . . . . . . . . . . . . . . . . . . . . . . 426
17.6 Holonomic vs. Non-Holonomic Bases . . . . . . . . . . . . . . . . . . . . . . . . . . 432
17.7 Cartan’s Structure Equations and Tetrad Methods . . . . . . . . . . . . . . . . . . . 433
17.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440
18 A Crash Course on General Relativity 443

18.1 The Energy-Momentum Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
18.2 A Physical Derivation of Einstein’s Field Equations . . . . . . . . . . . . . . . . . . 446
18.3 Lagrangian Formulation of Einstein’s Field Equations . . . . . . . . . . . . . . . . . 448
18.4 The Schwarzschild Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 450
18.5 The Friedmann Equations and Cosmology . . . . . . . . . . . . . . . . . . . . . . . 453
18.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
19 Black Holes in General Relativity 459

19.1 The Schwarzschild Black Hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
19.2 Eddington-Finklestein Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
19.3 Rindler Space-Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
19.4 Kruskal-Szekeres Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 466
19.5 Conformal Compactification and Penrose Diagrams . . . . . . . . . . . . . . . . . . 468
19.6 Charged Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472
19.7 Rotating Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474
19.8 The Unruh and Hawking Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
19.9 Black Hole Thermodynamics and Beyond . . . . . . . . . . . . . . . . . . . . . . . . 479
19.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482
20 Black Holes in String Theory 485

20.1 Black Holes in Higher Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
20.2 Entropy of the Schwarzschild Black Hole in D=d+1 . . . . . . . . . . . . . . . . . . 489
20.3 Microscopic Entropy of an Extremal Black Hole . . . . . . . . . . . . . . . . . . . . 491
20.4 The Laws of Nature and Black Hole Complementarity . . . . . . . . . . . . . . . . . 494
20.5 UV/IR Connection and Bounds on Entropy . . . . . . . . . . . . . . . . . . . . . . . 497
20.6 AdS/CFT Correspondence and Holography . . . . . . . . . . . . . . . . . . . . . . . 499
20.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
21 Alternative Approaches toward Quantum Gravity 503

21.1 A Critique of String Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
21.2 Twistor Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
21.3 Loop Quantum Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
21.4 Causal Dynamical Triangulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
21.5 Final Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
8 CONTENTS
A Van der Waerden Notation for Weyl Spinors 511

A.1 Lorentz Invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 511
A.2 Index Notation for the Pauli Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . 512
A.3 The SUSY Algebra in van der Waerden Notation . . . . . . . . . . . . . . . . . . . . 514
B Grassmann Variables and Grassmann Integration 517
C Solutions to Exercises 521
Bibliography 555
Acknowledgements
This project was a tremendous undertaking, one which required the assistance and feedback of
several of my peers. I would like to thank all of those who gave such feedback, including the
students of 2012-2013 Physics thesis class. In particular, I would like to thank my mentor Professor
Albert Stetz for giving me the idea of writing this text in the first place, supporting my efforts,
and providing me with deeper insight when needed. I would also like to recognize Teal Pershing
who has acted as an editor, figure designer, student, and friend through the entire process. I can
honestly say that this text would not be the same without their help.
9
10 CONTENTS
Preface
In high school I had a vision. I held that if I could understand the mathematics behind physics
problems, the physical interpretation would naturally follow. Rather quickly I realized the naivety
of my perspective toward physics. After I purchased texts on relativity, quantum mechanics, and
string theory, it was apparent that I held an ill-conceived approach to physics, as nearly each text
I bought was almost unreadable. The content that lay within each text was no different than a
foreign language. Although I didn’t understand the material in each text, I found what was needed
to understand relativity, quantum mechanics, and even string theory; uncovering that the road to
understanding string theory was a long one.
After enrolling at Oregon State University to study both math and physics, I continued to study
the texts I collected as a high school student, gaining further insight into the mathematical tools
and physical theories formulating the necessary background for string theory. It became clear in my
independent studies that the texts I referred to had a specific audience in mind, graduate students,
and advanced ones at that. Searching for the ultimate source on string theory, I began filling my
personal library with a plethora of texts on the subject, some of which became more accessible
as my academic career continued, however still out of reach. And the reason for this was simple:
each text assumed that the reader had background in quantum field theory, general relativity, and
supersymmetry, along with a fair understanding of a variety of branches of mathematics. Of course,
each of these subjects require background in other physical theories and mathematical tools. The
task of learning string theory was a daunting one; at times I felt like Sisyphus. Just as I reached a
pinnacle in my studies, I would come across a new but crucial detail necessary for understanding
the basics of string theory; the boulder I pushed would roll back down the hill. During this process
I could only dream of finding that sole text which had everything I needed to know about string
theory, and all in one location. I was aiming for a shortcut; a text which would not only provide
the details of string theory, but would also present the necessary background material. Of course
I would not find such a text, as such a text had not been written yet, particularly at the level an
undergraduate could hope to understand.
At the time this text is being written, there are three main graduate level books on the topic
of superstring theory, and one text aimed for undergraduates. The graduate texts include the
two volume series Superstring Theory by Green, Schwarz, and Witten (GSW); String Theory and
M-Theory: A Modern Introduction by Becker, Becker, and Schwarz, and the two volume series
Bosonic String Theory and Superstring Theory, by Polchinski. The text aimed for undergraduates is
Zwiebach’s A First Course in String Theory. As I have been independently studying this subject for
quite some time, I have found that each of these sources have both benefits and deficiencies. Firstly,
the graduate texts make it difficult for the undergraduate to follow for the given reasons above,
assuming the reader has a formidable background in both physics and mathematics. Moreover, as
11
12 CONTENTS
the texts are graduate level, they rarely detail the results, leaving most of the derivations to the
reader, a serious problem for someone new to the subject. Nonetheless, these texts summarize the
more difficult topics of the subject, and include fairly modern avenues for research in string theory,
and, more broadly, quantum gravity.
Alternatively, Zwiebach’s text is particularly thorough in discussing the bosonic string, making it
a widely accessible book for undergraduates, especially those aiming for self-study. The cost of this
however is that it does not provide a rigorous treatment of superstrings, nor does it go into grave
detail on the subjects which are necessary for further study in string theory, marking its deficiency.
In this sense, after a student completes a first reading of Zwiebach’s text, they will have the tools
to only understand the first chapter of the mainstream graduate level texts, limiting the reader’s
ability to grasp fundamental topics, let alone advanced topics in each of these texts.
The overarching goal of this text is to resolve this issue. Here I aim to bridge the material
presented in Zwiebach’s text and the graduate level texts on string theory. In a sense, this text
presents only a few new topics in string theory. A student studying string theory, depending on their
level, can already find the necessary information from other resources, including the ones above,
in which case this text gives nothing new. What is different about this text is the presentation
of the material. Synthesizing the material already available, this book also augments the current
information with additional computational details to aid the reader as they study this text. As
Zwiebach provides, from a pedagogical standpoint, a suitable approach toward the fundamentals of
bosonic string theory and D-branes, much of the first half of this text closely follows his methods.
In the second half of this text, particularly the work done on superstrings, much of the material
was synthesized from the three mainstream texts as well as other books and online sources. For
the reader who is interested in this subject, it is strongly encouraged that they review the source
materials which this text is based off of as the authors of those texts and papers have a far better
understanding of the subject than the author of this text does.
One of the chief aims of this text is act as a bridge between the books already available. To do this
involves the arduous task of providing the necessary background material so that an undergraduate
physics student with minimal background can begin to study the entirety of string theory. As noted
in Zwiebach, one can get pretty far without having to ever introduce the hairy details of quantum
field theory, general relativity, or supersymmetry. As one moves away from bosonic string theory
however, it is a must that the student be at least familiar with the tenets of quantum field theory
and supersymmetry. For that reason this text has a few select chapters in which string theory is
not discussed at all, but rather preliminary material is presented so that the following chapters
make sense. I call these chapters “crash course” chapters as their aim is to provide the background
material necessary for the student to continue with the text, however in a timely fashion. By all
means, the crash course chapters are not intended to deliver every detail of the subject, but rather
give a flavor and some of the terminology of each subfield of physics while framing it in the context
of string theory. Ultimately, this text aims to be a sole resource on the basics of string theory: a
detailed introduction assuming minimal background knowledge of the reader.
String theory is a controversial area of study amongst physicists due to its lack of experimental
evidence. A final goal of this text is to provide not experimental evidence, but arguments that
suggest string theory is at least on the right track, as well as its place among competing theories of
quantum gravity. One of the biggest results of string theory (some say the only) is the statistical
derivation of the entropy of a black hole, matching Hawking’s and Bekenstein’s own predictions.
To provide a more quantitative discussion on this result, the third part of this book is devoted to
general relativity and black holes as interpreted in string theory. Moreover, within the past year
CONTENTS 13
some interesting discussions on the modern principles of black holes have been taking, some of
which suggest a radical change in the way we are to think about black holes. An analysis of these
discussions is presented in this text giving an up to date report on black holes in string theory.
Due to the level of the material presented in this text, the intended audience of this text ranges
from undergraduate to first year graduate students. It is crucial that the reader is assumed to have
a fair background in: ordinary and partial differential equations, some complex analysis, quantum
mechanics, and special relativity. This is all that is technically required to read this text, however
some of the material would be difficult for a first time reader. For that reason a track system has
been devised to offer the reader suggested routes of study that would prove most beneficial. The
first time reader should go through the entirety of chapters 1 and 2, and pay special attention to
the mathematics of four vectors and the metric in chapter 2. From here the first time reader should
chapter 3, however may skip section 3.7, The Relativistic Point Particle in a Curved Space-time,
on a first read. From here the first time reader should read chapters 4-6. Chapter 7, the first
of the crash course chapters, the first time reader should at least read through sections 7.1-7.7
and 7.12-7.13. Although the material on quantizing spinor fields and the electromagnetic field is
important for understanding superstrings, the basic idea is presented using scalar fields, in which
the quantization procedure of the other fields follow similarly.
From here chapters 8-10 should be read entirely by the first time reader as they pose no serious
difficulties. Past chapter 10 the difficulty of the text increases as this is the advanced topics section.
The first time reader should skip chapter 11 on BRST quantization altogether as it requires a
fair understanding of the calculus of complex valued functions. Of course, an undergraduate with
sufficient training in contour integration and residue calculus could easily comprehend the material
presented in chapter 11. On the other hand, chapters 12-13 could likely be understood by the first
time reader, however a second read through of chapters 1-10 would be helpful. Chapter 14 is the
second crash course chapter, this time on supersymmetry. From here on the first time reader track
ends as, though the material isn’t incredibly harrowing, it does require a full reading of chapter 7
and perhaps outside insight.
The last section of this text, general relativity and string theory, is not out of the realm of the first
time reader assuming they have had experience with tensors and relativity. Some of the chapters in
this section are actually based on class notes taken during a two course undergraduate level series
on differential forms and relativity at Oregon State University presented by Professor Tevian Dray.
The second reading track would be one for first year graduate students, in which this entire text
is likely quite accessible to students at this level. Readers with more experience would likely find
this book not containing enough information and are therefore pointed to the references for further
details.
Some may view string theory as a subject which cannot be taught to students at the undergrad-
uate level. I challenge this notion, and this book is hopefully the resource which will allow other
undergraduates to challenge this belief. Regardless of the outcome, a student who has intently
studied the material in this text and others like it will gain insight into the program of modern
theoretical physics and the marvels it postulates.
14 CONTENTS
Part I
Fundamentals and the Bosonic

String
15
Chapter 1
Introduction
1.1 A Brief History of Modern Physics and the Mark of

String Theory
At the turn of the 19th century, it was thought that the studies of physics were coming to a
close. Newtonian mechanics described the motion of everything from the trails of the comets to the
trajectories of falling apples. Maxwell’s equations elegantly summarized the relationship between
stationary and moving charges, and even going as far as providing a mathematical treatment of
electromagnetic radiation. The work completed by Boltzmann lent insight to chemical kinetics and
the laws of thermodynamics, supplying a near complete understanding of chemical reactions and
the microscopic world. Physicists at the time found that the natural world seemed to have fewer
and fewer mysteries. Even Lord Kelvin, purportedly, in 1900 announced to the British Association
of the Advancement of Science that “There is nothing new to be discovered in physics now. All
that remains is more and more precise measurement” [16]. Whether Kelvin said it or not is unclear,
but the same tone was felt among many of his contemporaries, and they couldn’t have been more
wrong.
Five years after Kelvin’s address, a bright young physicist, going by Albert Einstein working at
a Swiss Patten office developed his theory of special relativity, a work which revolutionized the laws
of physics, and our entire conception of reality. Space and time were no longer the rigid objects
that Newtonian mechanics required. In order to provide an accurate description of the universe,
absolute space and absolute time were thrown out to be replaced by a malleable structure known
as space-time. For the laws of physics to remain invariant, our notion of time and space had to
become dynamic, a notion which was met with much friction, however gave way to the scientific
method.
If the physics community wasn’t disturbed by relativity, they were almost certainly horrified of
the second revolution of 20th century physics. Around the same time Einstein’s theory of special
relativity was overturning several classical ideas in Newtonian mechanics, Max Planck (and, of
course, Einstein) developed a theory of the quantum, radically changing the view of the microscopic
world. Even Einstein abhorred his own creation, never being able to reconcile that the determinism
promised by classical physics had to be supplanted by a non-deterministic description. Nonetheless,
the physics community had undergone a second revolution; the natural world was once again teeming
17
18 CHAPTER 1. INTRODUCTION
with mystery.
As quantum mechanics and special relativity came on the scene around the same time, the natural
progression was to proceed in unifying both theories, a theory of relativistic quantum mechanics.
Upon the consummation, quantum field theory posited the existence of antiparticles, particles which
have the same mass but have opposite additive quantum numbers. For example, if the positron,
or anti-electron, were to come into contact with the electron, the two would abruptly annihilate
each other, leaving only a radiation signature behind. As time went on, quantum field theory was
felt to be the correct candidate in accurately describing the entirety of the interactions of particles,
supplying a deeper understanding of the microscopic world.
Not long after the birth of quantum mechanics, Einstein developed his general theory of relativity,
correcting Newton’s ideas about gravity; providing an elegant depiction of the fabric of the cosmos,
once again overturning the classical, long upheld ideas about the mechanics of the universe. Not only
did the theory propose that gravity was the warping of space-time, but this warping could become
so extreme that it allowed for the creation of black holes, regions of space-time which are cut-off
from communicating with the rest of the universe. Even still, the principles of general relativity
seemed to provide a method of traversing through the cosmos via shortcuts known as wormholes,
traveling not only through space, but even time! While physicists scratched their heads, science
fiction writers picked up their pens.
Just as physicists had unified quantum mechanics and special relativity with quantum field
theory, the natural expectation was to unify quantum mechanics with general relativity. To their
dismay however, a successful theory of quantum gravity was out of reach. As we will see in the next
section, quantizing the gravitational field within this framework leads to unphysical results. In short,
general relativity and quantum mechanics were the two great pillars of twentieth century physics,
but the two theories appear to almost entirely disjoint. General relativity describes the macroscopic
world; the motion of planets, galaxies, and the evolution of the cosmos. Alternatively, quantum
mechanics specializes in describing the microscopic world: the motion of elementary particles,
particle decay,and the fundamental properties of matter. The theories, by themselves, perform
well in their realms, but in the microscopic limit, general relativity breaks down, and the instant
quantum mechanics attempts to absorb gravity, the theory blows up. This is one of the goals of
string Theory, and all other theories of quantum gravity: to develop a coherent framework in which
quantum mechanics and gravity can coexist.
Historically, string theory was first introduced in the 1960’s as an attempt to understand the
strong nuclear force through the interactions of hadrons. It turned out that a theory based on
one-dimensional extended objects, referred to as strings, solved some of the issues that the point-
like behavior of hadronic interactions introduced. The crucial idea of using strings is that specific
particles would correspond to oscillation modes (quantum states) of a string, much like the oscilla-
tion modes on a violin string producing a variety of musical notes. With the string description, a
single one-dimensional object retains the ability to explain the differences of the myriad of observed
hadrons (Becker, Schwarz, 2). In the early 1970’s however, another theory, Quantum Chromody-
namics (QCD), was developed to describe the strong nuclear force. With this development, as well
as the technical issues of using strings, string theory fell out of favor with the masses of the physics
community. Some, however, weren’t ready to abandon the elegance string theory seemed to offer.
In 1974, around the same time QCD was being fine tuned, physicists Julius Wess and Bruno
Zumino developed a solution to eliminate tachyons from the models of particle physics. Not only did
the theory eliminate undesired tachyons it also provided a symmetry between bosons and fermions.
Such a symmetry is formally known as supersymmetry. Not long after the presentation of global
1.2. A QUALITATIVE LOOK AT RENORMALIZATION THEORY 19
supersymmetry, work was done in extending the theory to local supersymmetry or supergravity,
an extension which includes General Relativity. Unification seemed to be approaching (Green,
Schwarz, Witten,16).
The year 1984 became known as the first superstring revolution amongst theoretical physicists.
Due to the importance of supersymmetry, it is expected that string theory should contain local
supersymmetry. The revolution was marked by the discovery that to maintain quantum mechanical
consistency with a ten-dimensional supersymmetry requires a local Yang-Mills gauge symmetry with
one of two possible Lie algebras: SO(32) and E8 × E8 . Without getting into the details, the fact
that only two Lie algebras allow consistency suggests that string theory is fairly restrictive and
therefore might be predictive.
It turns out that the superstring formalism gives rise to five distinct theories: type I, type IIA,
type IIB, and two Heterotic string theories (more on this later). This realization that there were
five different string theories posited an intuitive problem: if there is only one universe, why are
there so many theories? In the late 1980s, a property known as T-duality was found to relate the
type II theories and the two heterotic theories, suggesting that they shouldn’t be viewed as distinct
theories (Becker, Schwarz, 10-12). But what of the other theories?
The mid-1990s for theoretical physicists has become known as the second superstring revolution.
Similar to T-duality, another type of duality, called S-duality was discovered, which relates the type
I theory and one of the heterotic string theories and the type II B theory to itself. Remarkably, it
was found that if an 11th dimension was introduced, a quantum theory called M-theory emerges.
Together with the S and T dualities, the five superstring theories and 11-dimensional supergravity
are connected by a web of dualities. That is, each theory separately can be viewed as different
corners of the same theory, a single model of string theory describing quantum gravity. To this
date however, there is not yet a complete, or compelling enough, formulation of M-theory. As it
stands, the physics community is waiting for the upcoming theoretical physicists to either complete
the description, or abandon M-theory altogether.
1.2 A Qualitative Look at Renormalization Theory

Calculations in quantum field theory can be done using perturbative expansions. The terms in
the expansion describes a possible particle process, each of which can be represented by Feynman
diagrams . The diagrams themselves are representations of terms in a series, having various orders
of magnitude describing the amplitude for the particle process to occur. In quantum field theory,
it was found that photons can turn into an electron-positron pair, only to subsequently annihilate
producing another photon. Such a process, an interior process is often referred to as virtual, and
particles involved in such processes are called virtual particles. To determine the amplitude for a
particle interaction, one must draw a Feynman diagram for every possible virtual process, in effect,
expanding the entire series of terms. More specifically, to calculate the amplitude of an interaction
which includes all virtual processes, one must compute a loop integral, which can be written in the
form (McMahon, String Theory Demystified, 7)
Z
I∼ p4J−8 dD p (1.1)
where p is momentum, J is the spin of the particle, and D is the dimension of space-time. Let
us also define
λ = 4J + D − 8
If p → ∞ and λ < 0, then the loop integral I is finite. Alternatively, if p → ∞ and λ > 0,
then I diverges, leading to infinities in calculating the amplitudes of particle processes. If however
I → ∞ slowly, renormalization, a mathematical technique that often seems like a sleight of hand,
can be used to obtain finite results. This procedure works well for electromagnetic, weak, and strong
interactions, however renormalization doesn’t in fact work when trying to quantize the gravitational
field.
General relativity includes gravitational waves that carry an angular momentum J = 2. Then,
by analogy, it is deduced that the graviton, the quantum carrying the gravitational force is a spin-2
particle. If we let the dimension of space-time D be equal to four, as proposed by relativity and
quantum field theory, and using J = 2, we find
λ = 4J − 8 + D = 8 − 8 + 4 = 4 > 0
and therefore have
p4J−8 → p0 = 1
yielding
Z
I∼ d4 p → ∞ (1.2)
when integrated over all momenta. This means that gravity cannot be renormalized, indicat-
ing that gravity cannot be incorporated using the standard framework of quantum field theory.
Essentially, this issue arises because QFT is a point-particle theory, and therefore has particle in-
teractions occurring at a single point in space-time. String theory is governed by one-dimensional
objects rather than point-like objects getting rid of the issue of interactions occurring at a single
point.
The problem of point-like behavior isn’t a new one. Consider Heisenberg’s uncertainty relation
∆x∆p ∼ ~ (1.3)
Point-like particle interactions provide ∆x = 0, which implies ∆p → ∞, leading to divergent
loop integrals. In string theory, ∆x 9 0 but instead cuts off at a small but non-zero value. The
cutoff turns out to be defined by the length of the string, in turn modifying the uncertainty relation
(Hossenfelder, 23-28). It is found in string theory that the generalized uncertainty in position ∆x
is approximately given by
~ ∆p
∆x = + α0 (1.4)
∆p ~
where α0 is related to the string tension T
1
α0 = (1.5)
2πT
Altogether, as long as α0 6= 0, the problems resulting from point-like interactions can be avoided,
yielding physical results when calculating the amplitudes of particles processes.
1.3. FUNDAMENTAL UNITS AND THE PLANCKIAN SYSTEM 21
1.3 Fundamental Units and the Planckian System

Now that we have a basic history of the subject, and a bit of motivation to pursue this theory, let
us explore some of the fundamentals of string theory and modern theoretical physics. Let’s first
make sure we are all on the same page on what unit system we will use in this book, and is used in
other, more advanced texts.
When one first learns physics, or any branch of science for that matter, they are often taught unit
analysis. There the student learns a plethora of constants, and their corresponding units. It turns
out that (nearly) every physical constant or ’unit’ may be expressed in terms of more fundamental
units: length L, time T , and mass m.
When studying gravitation, particularly Newtonian gravity, we frequently come across two so-
called fundamental constants: G, Newton’s gravitational constant, and c, the speed of light. A
simple question to ask is: what are the units of G and c in terms of fundamental units. From
Newton’s law of gravitation in four dimensions
Gm1 m2
F~ (4) = (1.6)
r2
we find that the units of Newton’s gravitational constant G are
L2 M L L2 L2
[G(4) ] = [F orce]= =
M2 T2 M2 MT2
Moreover, the speed of light in fundamental units is simply given by
L
[c] =
T
When we study quantum theory we come across another fundamental constant, famously known
as Planck’s constant, ~. From De Broglie’s relation we can easily determine the units of ~:
[E] M L2
E = 2π~ν ⇒ [~] = ⇒ [~] =
[ν] T
These constants, G, ~, c, are often referred to as the fundamental constants as they can be
expressed solely in terms of fundamental units. The numerical value of each have been determined
experimentally to be (Zwiebach)
m3 8m −34 kg · m
2
G = 6.67 × 10−11 , c = 2.998 × 10 , ~ = 1.055 × 10 (1.7)
kg · s2 s s
Anyone who has studied relativity is well aware of the annoyance of caring factors of c and G
throughout calculations. To avoid this mess, it is convenient to use the Planckian system of units,
in which we let G, c, and ~ all equal one. To do this however, we must find new units of mass,
length, and time such that G = c = ~ = 1. These units are know as the Planck, mP ; Planck length,
`P , and Planck time, tP . Using this system and our desire for G = c = ~ = 1, we find
`3P `P mP `2P
G=1· 2 ,c = 1 · ,~ = 1 ·
m P tP tP tP
We can solve for these units numerically [61]:
r
G~
`P = = 1.616 × 10−33 cm (1.8)
c3
r
`P G~
tP = = = 5.391 × 10−44 s (1.9)
c c5
r
~c
mP = = 2.176 × 10−5 g (1.10)
G
By brief observation, we easily recognize the difficulty of doing experiments on the Planck scale.
However, a true theory of quantum gravity, including string theory, should be able to explain the
universe on its smallest scales and earliest times. It should be further noted that the Planck mass
is enormous compared to the mass of most other particles observed at particle accelerators. This
is strange since, if nature depends on the fundamental constants, why are the mass of most other
particles such smaller than that of the Planck mass? This problem is formally known as the hiearchy
problem.
1.4 Planck Length in Arbitrary Dimensions

If string theory ends up being true, then the universe is really higher dimensional. One immediate
consequence is that of Newton’s gravitational constant and the Planck length change depending
on the space-time dimension. To determine this modification, we first recall the expression of the
gravitational potential in four dimensions:
∇2 V (4) = 4πGρm (1.11)

where V (4) = − GM
r is the gravitational potential in four dimensions, and ρm is the mass density.
In arbitrary dimensions however, it turns out that the units of V remain the same, but that of ρm
change. To have the expression remain invariant in arbitrary dimensions, we require that Newton’s
gravitational constant change with the dimension of space-time, resulting in the expression
∇2 V (D) = 4πG(D) ρm (1.12)

To see how Newton’s constant changes with dimension, we first compare five dimensions to four
dimensions:
M M
∇2 V (4) = ∇2 V (5) ⇒ [G(5) ] 4
= [G] 3 ⇒ [G(5) ] = L[G]
L L
[c]3 L2
But since [G] = [~] , we find
[c]3 L3
[G(5) ] = (1.13)
[~]
Then, from our expression for Planck length in four dimensions, we may write the Planck length
for five dimensional space-time in the following way
3
(5) ~G(5)
`P = (1.14)
c3
1.5. COMPACTIFICATION AND LARGE EXTRA DIMENSIONS 23
Multiplying by one in an interesting way gives
G(5) (5)
3
2 G
~G
(5)
`P = = (`P )
c3 G G
The above statement can be generalized to D space-time dimensions (Zwiebach, 63):

(D)
D−2 ~G(D) 2 G
(D)
`P = = (`P ) (1.15)
c3 G
When it comes to calculating the higher dimensional gravitational constant, a more useful ex-
pression is (Zwiebach, 67)
G(D)
= VC (1.16)
G
where VC is the volume of the space of compactified extra dimensions, a topic we move to now.
1.5 Compactification and Large Extra Dimensions

When Einstein first proposed that our universe was four dimensional, a three spatial dimensional
joined with one time dimension began to make intuitive sense, based on human experience and
observation. In string theory, among its many revolutionary ideas, requires that we live in a
universe with extra spatial dimensions, a notion which has yet to be observed. To get around the
issue, theorists have proposed that we don’t readily observe these extra dimensions because they
have been compactified.
is a highly complex subject and goes well beyond the scope of this text. However, the basic idea
of compactification is simple to follow. Imagine that you’re a small creature who lives on a thin
cylinder (string theorist Brian Greene gives a similar analogy with an ant travelling along a wire
in his The Fabric of the Cosmos). Further suppose that you cannot leave the cylinder, but must
traverse along it which ever way you’d like. You would quickly discover that you can only travel in
two directions, thereby suspecting your world is two dimensional.
Now suppose a large being observes this thin cylinder from very far away. To the being, the
cylinder would appear one-dimensional. They would presume that the universe of the cylinder is a
one-dimensional line. This, of course, is not true, as you are living on the cylinder and can attest
that it is truly two-dimensional. In this very basic sense, there is an extra dimension which has been
“compactified”, giving the appearance of a one dimensional world when it really is two dimensional.
The creature might even be able to determine that they live in a world with a compact dimen-
sion. As the creature moves along the width of the cylinder, first marking their initial position, it
would find eventually arrive back to its initial location after a moving a distance of 2πR. In an
analogous sense, someone walking along the inside of a circle would come to a similar conclusion.
Mathematically, we can view a circle as an open line with an identification, i.e. points which differ
by 2πR are in fact the same point. In other words, two points P1 and P2 are said to be equivalent
if
Figure 1.1: Notice that points P1 and P2 are identified using (1.17). To a person walking along this
line, the points P1 and P2 look exactly the same. One can imagine bending the line into a circle,
connecting points P1 and P2 .
P1 ∼ P2 ←→ x(P1 ) = x(P2 ) + 2πRn, n ∈ Z (1.17)
as illustrated in figure 1.1. String theory wasn’t the first model to propose a higher spatial di-
mensional universe. In 1921, was able to extend Einstein’s general relativity to a five dimensional
space-time (four spatial dimensions and one time dimension); unifying gravitation and electromag-
netism. In a sense, this was one of the first models seeking to unify the fundamental forces of nature,
and it was done through the addition of higher dimensions. Just as string theorists do, Oskar Klein,
in 1926, proposed that the extra spatial dimension was curled up into a circle, however with a very
small radius.
To gain further insight for extra dimensions, we consider a simple model presented by Zwiebach.
Consider quantum square well where we introduce an extra curled up dimension (Zwiebach, 38).
Let us make the identification,
(x, y) ∼ (x, y + 2πR)
indicating that y is an extra dimension curled up into a small circle of radius R. Since the
the coordinate y represents a circle of circumference 2πR, the space where the particle moves is a
cylinder with length a and a circumference of 2πR. The Schrödinger equation in two dimensions
becomes
~2 ∂ 2 ψ ∂ 2 ψ

− + = Eψ (1.18)
2m ∂x2 ∂y 2
To solve this partial differential equation, we use the method of separation of variables, supposing
that ψ(x, y) is separable, i.e. ψ(x, y) = X(x)Y (y). Substituting this back into Schrödinger’s
equation, we find
~2 ∂ 2 X(x) ∂ 2 Y (y)

− Y (y) + X(x) = EX(x)Y (y)
2m ∂x2 ∂y 2
leading to
1.5. COMPACTIFICATION AND LARGE EXTRA DIMENSIONS 25
~2 1 d2 X(x) 1 d2 Y (y)

− + =E
2m X(x) dx2 Y (y) dy 2
Since the left hand side is equal to a constant, then each term on the left hand side must equal
a constant. That is, we may write
d2 X(x)
= kx2 X(x)
dx2
and
d2 Y (y)
= ky2 Y (y)
dy 2
Using the boundary conditions for the square well, and the identification for y, we find that

kπx `y `y
X(x) = ak sin Y (y) = b` sin + c` cos
a R R
where k and ` are both integers, however k ranges from 1 onward, while ` ranges from zero
onward. Then, where ψk,` = X(x)Y (y), we find that the energy eigenvalues are found to be,
" 2 #
2
~2 kπ `
Ek,` = + (1.19)
2m a R
In the case where ` 6= 0, the energies are two-fold degenerate since the eigenenergies correspond
to two linearly independent solutions.
First notice that when ` = 0 the energy spectrum for a one-dimensional square well emerges.
This new energy spectrum includes the spectrum for that of the original system, but also contains
energies as a product of the additional curled up dimension. The next lowest energy level occurs
when k = 1 and ` = 1 " 2 #
~2 π 2 1
E1,1 = +
2m a R
If we then consider the case when R a, a scenario stating that the curled up dimension is much
smaller than the size of the well, we find that the lowest order energy is approximately,
2
~2

1
E1,1 ∼
2m R
Since R is much smaller than a, the first new energy level appears to be at a much higher energy
than the low energy levels of the original system, indicating that this extra dimension has remain
hidden from the current energy levels at which our particle accelerators can reach. In principle then,
as particle accelerators become more powerful, our ability to test the existence of extra dimensions
will improve.
It turns out the circumference of the compact dimension can be calculated in general by
2
! D−4
(D)
(D) `P
`C = `P (1.20)
`P
If we consider a 5-dimensional universe, the circumference of the compact dimension is approx-

imately given by `C ∼ 107 km, over an order of magnitude greater than the distance between the
earth and the moon. This dimension size should have been detected long ago simply due to its size.
For the sake of curiosity, let us suppose a six dimensional universe. Then,
`C ∼ 10−3 cm (1.21)
Remember that particle accelerators can currently probe on the scale of 10−16 cm, and therefore it
might be possible that these ’large’ extra dimensions simply haven’t been experimentally discovered
yet, but do in fact exist. Lastly, according to some superstring theories, we live in a 10 dimensional
space-time, in which we have `C ∼ 10−13 cm, a distance which could, in principle, be searched for
using particle accelerators.
1.6 Verifying String Theory

Before we get too far ahead of ourselves, it should be mentioned that string theory has yet to be
experimentally verified. But any good experiment requires a precisely defined hypothesis. This is
the crucial issue with string theory: it is still being developed, making it difficult to formulate a
sound prediction. Nonetheless, there are some elements of string theory which theorists are hopeful
might be experimentally verifiable using the Large Hadron collider.
If string theory ends up being the correct model for quantum gravity, then our universe is higher
dimensional. For consistency, superstring theories require at least ten space-time dimensions; M-
theory requires 11 space-time dimensions. Thus, our three dimensional universe must be embedded
in at least a nine-dimensional space. That is to say, our universe is a hypersurface, or a multi-
dimensional membrane, often called a . String theory contains two different types of strings: open
and closed. Open strings are attached to D-branes, while closed strings are not; instead they prop-
agate through the higher dmensional ’bulk’ universe, traveling between other D-branes. According
to string theory, the quantum state of the graviton is a closed string, and therefore is not attached
to a D-brane. Many string theorists propose that this is the reason why gravity appears to be so
weak compared to the other forces in our universe. All of these analyses leads some theorists to
predict that large extra dimensions might be detected using gravitational experiments. Moreover, it
can also be shown that these extra dimensions are on the size of a tenth of a millimeter, something
which be able to be detected in the lab!
Another possible avenue for verification lies with the detection of cosmic strings. A cosmic string
is a string that was present at the early stages of the universe. As the universe expanded, so did the
string, stretching over the entirety of the observable universe. In that sense, cosmic strings might
be detected through gravitational lensing, or via gravitational waves. But cosmic strings aren’t
unique to string theory either. Conventional theories of particle physics also propose the existence
of cosmic strings. Therefore, the experiments would have to examined carefully in order to clarify
whether the detected strings are in fact the cosmic strings predicted by string theory.
Since string theory is manifestly supersymmetric, any experiment which seeks to verify the exis-
tence of superymmetric particles would be indirect evidence for superstring theory. If we take a ten
dimensional superstring theory and compactify six of the dimensions, the remaining 4-dimensional
theory is, in many cases, supersymmetric (Zwiebach, 9). Therefore, any experimental discovery of
supersymmetry would at least indicate that string theory is on the right track.
1.6. VERIFYING STRING THEORY 27
String theory, if a true unified theory, shouldn’t just predict new phenomena, but also phenomena
we are already aware of. The Standard Model of particle physics contains all of the forces except for
gravity, and an entire zoo of particles, most of which have been experimentally verified. In effect,
String theory should be able to be reduced to the Standard model in the low energy limit. In other
words, the Standard model should be, in a sense, predicted by String theory. In other words, the
consistency of string Theory with other theories with provided experimental evidence is another
route in which string theory might be tested, though indirectly.
Some physicists claim however that a new paradigm of experimental physics must be considered,
as the conventional methods will never lead to successful tests for string theory. That is, the method
of using higher energy supercolliders will not allow us to probe the scales of any theory of quantum
gravity, including string theory. To understand this argument, start from Heisenberg’s uncertainty
principle
~
∆x∆p ≥ (1.22)
2
Using the De Broglie relation λ = h/p and keeping in mind that quanta have an energy E = hf ,
we can approximate that the size ∆x can be probed with an energy of
~c
E≈ (1.23)
∆x
The traditional and current line of thinking is that we collide particles head on with a center
of mass energy equal to that above, and using the higher energy products of collision we hope to
reconstruct short distance events, distances which approach the Planck scale. The problem with this
thinking is that at energies far above the Planck mass, high energy collisions will create mini-black
holes of mass on the order of energy E above. Moreover, interesting short distance effects that we
wish to study will be hidden by these mini-black hole event horizons, with a radius of approximately
R ≈ 2GE/c2 ; meaning that the distances we hope to probe are completely inaccessible. Even still,
some theorists go on to say that even if we increase the energy on particle accelerators, scales smaller
than the Planck scale `P will be left unprobed, no matter how high the energy of the supercollider.
This is why some researchers studying string theory have moved away from proposing experiments
with colliders but instead by other non-conventional means. One non-conventional approach is to
search for cosmic strings as discussed above.
Some string theorists are bold enough to claim one successful prediction: string theory predicts
gravity! Certainly the our knowledge of gravity has existed well before the studies of string theory,
but string theory is a theory about quantized, relativistic strings, originally developed for investi-
gating the issues of the strong nuclear force, intially it had no hopes of being a theory of quantum
gravity. There was never any thought of including gravity, yet, the graviton naturally arises as a
quantum vibration of a relativistic string. In that sense, gravity was never put into the theory by
hand; rather it is required for string theory to remain consistent. It is this emergence of quantum
gravity which has led so many theorists to claim that string theory is the grand unified theory
physicists have long been waiting for.
Physicists can be, rather obtusely, be broken up into two main factions: experimentalists and
theorists. Both are necessary for sound theories, and our ability to make accurate judgements of
the nature of the universe. In the case of string theory, experimentalists tend to despise the theory
as it claims so much, yet, so far, has nothing to prove for itself. Even some theorists fall into this
category. And the skeptics are right about this. String theory, at some point should be a testable
theory with transparent predictions and reasonable hypotheses. But before this can happen, string
theorists must face the challenge of describing nature and formulating the theory, a task which
will likely take much more time. String theorist Ed Witten once wrote that string theory is a 21st
century physical theory which happened to appear in the 20th century. If the present challenges
are surmounted, we will have a theory which can describe all interactions, giving us the ability to
understand the origin and fate of space-time, and unlocking the mysteries of a quantum mechanical
universe. When the stakes are so high, physicists are likely to continue this endeavor until definite
answers have been discovered.
Chapter 2
Special Relativity and Light Cone

Coordinates
2.1 Special Relativity and Einstein’s Postulates

In 1905, the physics community was shocked by the proposal of a young physicist going by the
name of Albert Einstein. In his seminal paper, Einstein formulated the special theory of relativity,
a theory which challenged the basic, assummingly underlying notions of space and time. The
crucial notion was that the speed of light, measured in a vacuum, was constant and finite, in
consequence overturning the commonplace ideas about space and time; at the same time, lending
further verification that indeed the aether did not pervade the universe as once thought.
It is assumed that the reader has had some experience with special relativity, and for that reason
we will only briefly review the basic tenets and mathematics of the theory. The reader might not
be familiar with the mathematical notation we will introduce, and therefore the reader which is not
used to tensors and index notation should pay close attention to this chapter. For those that are
comfortable with this material, they can skip the reading until the section on light-cone coordinates.
A good place to start with studying relativity is classical Galilean relativity. In short, relativity
is the study of how various physical phenomena appear to different observers. Before Einstein,
this type of analysis was done using Galilean transformations. These transformations are simple
mathematical relations that give a way to connect measurements in different frames of reference.
Given two reference frames F and F 0 , the first with coordinates (t, x, y, z), and the second with
(t0 , x0 , y 0 , z 0 ) moving with a velocity v with respect to F , the Galilean transformations between these
two frames are simply
t = t0 x = x0 + vt y = y0 z = z0 (2.1)
With these relations, a measurement in one reference frame can be related to a measurement in
the other reference frame. These expressions are rather intuitive, which is part of the reason why
so many physicists had a difficult time coming to terms with Einstein’s special relativity. The first
of these transformations is of particular interest, and is perhaps the most intuitive. Suppose we
have two frames: one with an observer standing at the starting line of a race track, and the other
29
30 CHAPTER 2. SPECIAL RELATIVITY AND LIGHT CONE COORDINATES
frame belongs to an observer in a racecar traveling toward the finish line. Assuming both observers
can see the finish line clearly, as the racecar crosses the finish line, they would almost certainly
agree (up to experimental error) on when the racecar passes the finish line. That is an example
of the first Galilean transformation. In classical Galilean relativity, events in different reference
frames happen simultanesouly. By an event, we mean anything that can happen in space-time, or
in pre-relativity, something which occurs at some spatial location at some time. As we will see,
events in special relativity are very similar to the events in classical relativity, however contain a
subtle but imperative difference. Einstein was courageous in the sense that he took these intuitive
notions and turned them on their heads.
Let us summarize by stating the two postulates [38]:
(1) The laws of physics are the same in all intertial reference frames.
(2) The speed of light is invariant. That is, all observers in inertial frames will measure the same
speed for light, regardless of their state of motion.
At the time Einstein came out with his theory, nearly all physicists agreed with the first postulate,
however many had a difficult time with the second postulate. Einstein noted that the second
postulate had simply been looked over, as the constant value of the speed of light came out naturally
from Maxwell’s theory of electromagnetism.
Let’s seek a replacement for the Galilean transformations by applying the postulates Einstein
came up with, paying special attention to the constant value of the speed of light. As before, let
us use frames two, F and F 0 , where F 0 is moving relative to F with velocity v. We will write the
coordinates of F in the column vector:
 
ct
x
 
y (2.2)
z
where we have restored the constant c for now (typically we will set c equal to one). The
coordinates of F 0 are related via
 0  
ct ct
 x0  x
 0 = L  (2.3)
y  y
z0 z
where L is some 4 × 4 transformation matrix. We make it our goal to determine the matrix
representation of L. Assuming that frames have their y and z axes coincident, we have that
y = y0 z = z0
To get the form of the transformation L, we rely on the fact that the speed of light is invariant,
i.e. the same in both reference frames. Now consider a light bulb that emits a flash from the origin
at time t = 0. Light moves in spherical wavefronts that are described by
c2 t 2 = x 2 + x 2 + z 2
2.1. SPECIAL RELATIVITY AND EINSTEIN’S POSTULATES 31
Or, equivalently,
c2 t 2 − x 2 − y 2 − z 2 = 0
Invariance of the speed of light means that for F 0 observer will describe the flash of light using
her coordinate system, in which case the flash is described as
c2 t02 − x02 − y 02 − z 02 = 0
where we have insisted the speed of light is the same in both frames. This spherical wavefront
is observed to be the same in both frames, and hence
c2 t2 − x2 − y 2 − z 2 = c2 t02 − x02 − y 02 − z 02
Since y = y 0 and z = z 0 , the above becomes
c2 t2 − x2 = c2 t02 − x02
Provided that the transformation from one reference frame to the other should be linear, the
linearity means that the transformation must have the form
x0 = Ax + Bct ct0 = Cx + Dct (2.4)

This fact allows us to write the transformation matrix as
 
D C 0 0
 B A 0 0
L=  0 0 1 0
 (2.5)
0 0 0 1
We leave it as an exercise for the reader to solve for A, B, C, D, and show that the transformation
matrix takes the form
 
cosh(φ) − sinh(φ) 0 0
− sinh(φ) cosh(φ) 0 0
L=  (2.6)
 0 0 1 0
0 0 0 1
From the transformation matrix we have
x0 = x cosh(φ) − ct sinh(φ) ct0 = −x sinh(φ) + ct cosh(φ) (2.7)

Using the first of these we may solve for the rapidity φ. First note that when the origin of
both frames coincide, we have x0 = 0 and x = vt. Using this condition along with our linear
transformation we find
x0 = 0 = x cosh(φ) − ct sinh(φ) = vt cosh(φ) − ct sinh(φ) = t(v cosh(φ) − c sinh(φ))

Thus,
v
v cosh(φ) − c sinh(φ) = 0 ⇒ tanh(φ) = (2.8)
c
Focusing on the second expression in (2.7), we find that

sinh(φ) v
−x sinh(φ) + ct cosh(φ) = cosh(φ) −x + ct = c cosh(φ) t − 2 x
cosh(φ) c
Therefore,
v
t0 = cosh(φ) t − 2 x (2.9)
c
Similarly,
x0 = cosh(φ)(x − vt) (2.10)

Making use of the identity
1
cosh(φ) = q
1 − tanh2 (φ)
We find
1 1
cosh(φ) = q =q ≡γ (2.11)
2 v 2

1 − tanh (φ) 1− c
where we have defined the Lorentz factor γ. Altogether, we may write the transformations we
seek, the Lorentz transformations, which take the familiar form (McMahon, 10)
vx
t0 = γ t − 2 x0 = γ(x − vt) y0 = y z0 = z (2.12)
c
Based on these relations, it is no wonder that the physicists before Einstein thought that Galilean
transformations accurately described the measurements between two frames, one moving relative
to the other. The speed of light is so fast, that for normal moving speeds here on earth, one would
never suspect the more accurate Lorentz transformations. Notice for objects moving much slower
than the speed of light, v c, γ ≈ 1, and the reduce to the Galilean transformations given in (2.1).
Briefly observing the transformations (2.12), one finds two immediate consequences the reader
is likely familiar with, so we will only briefly discuss the implications. The first is time dilation.
Imagine two frames such that F 0 moves at a uniform velocity with respect to frame F . An interval
of time ∆t0 is measured by an observer in F 0 as seen by F to be
1
∆t = γ∆t0 = q ∆t
0
(2.13)
v 2
1− c
In other words, the clock of an observer whose frame is F 0 runs slow relative to the clock of an
observer whose frame is F . In short, moving clocks run slow. Most of all, the notion of simultaneity
goes out the window in special relativity (more on this shortly). Time dilation is the basis of the
famous twin paradox.
A second physical consequence is length contraction. Again consider two frames as in the previous
scenario. At a fixed time t, the measured distances along the direction of motion are related by
2.2. SPACE AND TIME, AND SPACE-TIME 33
1
∆x0 = q ∆x (2.14)
v 2

1− c
This expression tells us that distances in F 0 , along the direction of motion appear to be contracted
in the direction of motion. Length contraction is the basis of the famous pole-in-barn paradox.
It is important to point out that for speeds v > c, we obtain strange imaginary results. These
results correspond to tachyons, theoretical particles which have imaginary mass and move faster
than the speed of light. Most physicists abhor tachyons as they interupt our preconceived notion
of causality, which, as far as we can tell, appears be fundamental to our universe and the laws of
physics. On another note, it’s important to point out that many make the mistake of assuming
nothing can travel faster than the speed of light. Particles traveling faster than the speed of light
itself is not an issue, but rather particles that move slower than photons and then exceed the speed
of light is a violation of special relativity (this is why faster than light neutrinos would have been
interesting to the least). The same goes on the opposite side of the spectrum. Tachyons traveling
faster than the speed of light can never move slower than the speed of light, doing so would violate
the tenets of special relativity for the same reason. This won’t be the last time we see tachyons, as
they show up when we examine the spectrum of bosonic string theory.
2.2 Space and Time, and Space-time
Let’s take a step back for a moment and examine the actual structure of space-time, as interpreted
in both Newtonian theory and special relativity. Special relativity, just like Newtonian theory,
is a theory of both the physical and mathematical structure of space-time, the background where
particles move and fields evolve. As we have pointed out, takes the notions of absolute space and
absolute time that are fundmanetal to Newtonian theory and replaces them with an absolute space-
time. By absolute space and absolute time we mean that if the universe was void of all matter and
energy, one would be left with a rigid structure of space and time; absolute in the sense that space
and time are forever unchanging.
A more technical definition of space-time is a four dimensional set, with the elements of the set
labeled by three spatial dimensions and one time dimension. The path or trajectory of a particle
moving in space-time is a curved called a worldline. Figure 2.1 gives a diagram of space-time and
a worldline. In either special relativity or Newtonian theory, space and time appear to be different.
In either theory particles always move forward in time, never backward, while particles can travel
in any which way spatially (Carroll, 4).
Figure 2.1: An example of a world line of a particle. Notice time is on the vertical axis, and space
is on the horizontal axis. This means if a world line is located below the c = 1, it has a slope less
than c = 1, and is therefore moving faster than the speed of light (more space covered in less time).
There is a crucial difference between the theories however, and it is related to the Galilean
transformation t = t0 . In Newtonian mechanics, one divides space-time into slices of all space
occuring at one fixed point in time; in other words, the Newtonian mechanics physicists had become
so comfortable with until Einstein, unambigously defines simultaneity. As Einstein showed in his
special relativity, there is no well defined notion of two events occuring at the same time. The idea
of simultaneity is linked to the absolute time of Newton’s universe. Since the universe, when it is
bare naked, is only composed of absolute space and time, meaning that one can split space up into
slivers where all events take place at the same time, according to Newtonian theory.
Alternatively, at any event in space-time, as viewed in special relativity, one defines a light-cone,
a collection of paths through space-time which can be taken by rays of light. This gives space-time a
structure which replaces the absolute division of space and time in Newtonian mechanics. In other
words, the division of space-time into unique slices of space given at various points in time (slices
that are parameterized by time), is changed by the fact that all physical particles must always
remain inside a light-cone, that all physical particles move slower than the speed of light (Carroll,
5). An illustration of a light-cone is given in figure 2.2.
2.3. A QUICK GLANCE AT FOUR-VECTORS 35
Figure 2.2: A lightcone. All known physical particles live inside these lightcones in the time-like
region.
The invariance of the speed of light also leads to another interesting invariant. Just as one uses
the Pythagorean theorem to calculate distances in ordinary Euclidean space, the space-time interval
yields the distance between two events in space-time via
(∆S)2 = −(c∆t)2 + (∆x)2 + (∆y)2 + (∆z)2 (2.15)

Since the speed of light c is the same in all reference frames, observers in other inertial frames
would measure the same space-time interval. That is,
(∆S)2 = −(c∆t0 )2 + (∆x0 )2 + (∆y 0 )2 + (∆z 0 )2 (2.16)

Therefore, the space-time interval is invariant under Lorentz transformations.The space-time
interval is designated to be space-like if ∆S 2 > 0; time-like if ∆S 2 < 0, and null or light-like if
∆S 2 . Sometimes, depending on the calculation, one will instead use proper time which the time
measured by an observer’s own clock and is defined to be
∆τ 2 = −∆S 2 (2.17)
In summary, the Newtonian view of space-time is different than Einstein’s interpretation, based
on his two postulates. Simply put, we view special relativity as a theory of a four dimensional
space-time known as Minkowski space.
2.3 A Quick Glance at Four-Vectors

In order to describe the mathematical structure of special relativity, we must make use of four-
vectors. A four-vector is more than an object with three spatial components and a single time
component. It is a vector with four such components, and has a magnitude invariant under Lorentz
transformations. That is, a four-vector dotted to itself will be the same in all frames related by a
Lorentz transformation. In relativity, there are two four-vectors of particular importance. The first
is the four-velocity u given by

dt dx dy dz
u= , , , (2.18)
dτ dτ dτ dτ
It is a four-vector since not only does it have three spatial components and one time component,
it’s magnitude is invariant under Lorentz transformations. To see what we mean by this, suppose
that we have two frames, F, F 0 , each with a four-velocity, u, u0 . These four-velocities are four-vectors
since
u · u = u0 · u0
Another four-vector that is used often in relativity is the four-momentum, p, defined in terms of
the four-velocity as
p = γmu (2.19)
where m is the particle’s rest mass. In a more familiar form, the four-momentum is

E
p= , px , py , pz (2.20)
c
Remember that for a four-vector, the dot product with itself is Lorentz invariant, i.e., the dot
product is the same in all intertial frames. This means that we can calculate the value of the dot
product in any frame. Choosing the rest frame of the particle for simplicity, the momentum p~
(spatial components of the four-momentum p) are zero, and thus, by Lorentz invariance
2
Erest
p2 = p · p = = m2 c2
c2
What this means is m2 c2 is another Lorentz invariant, a constant that is the same in all intertial
frames. Later on we will see that, in one convention, the generic dot product of the four-momentum
is
E2 2 2 2 E2
p·p= − px − py − p z = − p~2
c2 c2
Putting these two results together we find
E 2 = p~2 c2 + m2 c4 (2.21)
the famous Einstein relation for energy.

2.4. VECTORS, DUAL VECTORS, AND TENSORS 37
2.4 Vectors, Dual Vectors, and Tensors

String theory is a highly mathematical theory, which is part of the reason why many find it difficult,
if not impossible, to bring the subject to undergraduates. At the same time, many undergraduates
are discouraged to learn it due to the amount of mathematics involved. Just to learn the mathe-
matics alone can be daunting, let alone comprehending the physics. To have a fair understanding
of string theory, one should be familiar with quantum field theory, supersymmetry, and general
relativity, each of which have their own mathematical formulation. For that reason, this text,
which is meant to be a bridge between the popularized books on the subject and the rigorous, often
abstract, language of the graduate texts, goes through each of these theories in details, providing a
basic background in both the physics and mathematics.
In this section we will go into the details on vectors, dual vectors, and tensors, placing it in
the context of special relativity, as this is typically the first physical theory where one first meets
these concepts. Though we focus on using the mathematics to provide a mathematical description
of special relativity, these topics are fundamental to studying general relativity, and most other
advanced fields of physics. For the most part, the next few sections of this chapter are meant to
slowly introduce the reader to some fundamental mathematical notation that will be used exten-
sively throughout the text. In fact, many of these calculations can be easily solved and understood
if one has a fair understanding of index notation, and knowing how to manipulate expressions with
indices. This is true here, in general relativity, and supersymmetry. For the reader that is new to
the subject, the next few sections are crucial in understanding the notation used throughout the
rest of the book.
In the last section we introduced four-vectors. Perhaps the most pertinent detail to emphasize
about four-vectors is that each vector is located at a given point in space-time. When we introduce
curvature in a later chapter, one loses the ability to draw preferred curves from one point to another
as one can do in ordinary flat space; in a curved space, we cannot draw the straight lines we are
used to when studying flat space geometry. Instead, at each point p in space-time we associate a
set of all possible vectors located at that point. We call this set the tangent space located at p,
which we denote by Tp . The picture to go with this is imagine we have some point. Any vector
that is tangent to that point then gets thrown into a set of all the other vectors tangent to that
point. Without getting into too many of the details, the we may view Tp as an abstract vector
space located at each point p in space-time. Figure 2.3 gives a suggestive depiction of a tangent
space. Just for reference, the reader might come across a tangent bundle, which is simply the union
of all the tangent spaces located at different points p.
Recall that a real vector space is a collection of objects (namely, vectors) that can be added
together, and multiplied by real numbers in a linear fashion. That is, for any two real vectors, U, V ,
and real numbers a, b we have (Carroll, 15)
(a + b)(U + V ) = aU + bU + aV + bV (2.22)
Moreover, every vector space has a zero vector, which acts as the identity element under vector
addition (i.e. the zero vector added to any non-zero vector yields the non-zero vector). Let us
imagine that for each tangent space we set up a basis of four-vectors eµ , where µ = 0, 1, 2, 3. This
means that we may write any abstract vector V as a linear combination of basis vectors in the
following way
V = V µ eµ (2.23)
Figure 2.3: A suggestive depiction of a tangent space, the collection of vectors tangent to the surface
at point p.
The coefficients V µ are the components of the vector V , and, as will be done in this text, are
often loosely referred to as the vector V µ . For more details on this, the reader is pointed to the
section on abstract index notation in Wald’s General Relativity.
In relativity, since we work with four-vectors, we have an additional time component as well as
three spatial components. Thus,
V = (V 0 , V 1 , V 2 , V 3 ) (2.24)
where we identify the zeoreth component of the four-vector with the time component. We are
used to writing out vectors in terms of basics vectors and the components of vectors in a sum. That
is, typically we expand a vector as
3
X
V = V i ei = V 1 e1 + V 2 e2 + V 3 e3 (2.25)
i=1
Comparing to (2.23), we realize that there is something implicit going on with the expression.
We call it the Einstein summation convention. This is simply a way to write sums in an easy
format. We use the Einstein summation convention by writing an expression with repeated indices.
In other words
3
X
V i ei → V i ei = V 1 e1 + V 2 e2 + V 3 e3 (2.26)
i=1
In the case of (2.23), since V is a four-vector, which means that there is an implicit sum over
the index µ, ranging from µ = 0, 1, 2, 3. That is,
V = V µ eµ = V 0 e0 + V 1 e1 + V 2 e2 + V 3 e3 (2.27)
When one studies differential geometry, the basis vectors ei are tangent to the coordinate lines.
For this reason, one will often write basis vectors as partial derivatives in a particular coordinate
direction (McMahon, 29):
∂
ei = ≡ ∂i (2.28)
∂xi
For instance, in Cartesian coordinates where we have basis vectors ex , ey , and ez , we may also
represent them as
∂ ∂ ∂
ex = ey = ez = (2.29)
∂x ∂y ∂z
We refer to this type of basis as a coordinate basis. Using this notation allows us to think of
vectors as operators, mapping a function into a new function that is related by a derivative.
We wrote our vector V as V µ eµ , with the indices on the components upstairs. But we very
well could have chosen to write the indices downstairs instead. Vectors with upper indices, V µ ,
are often called contravariant vectors, while vectors with lower indices are called covariant vectors
(to remember the difference, just think co is for low). Where one puts the indices does matter,
however contravariant vectors and covariant vectors are simply different representations of the same
mathematical entity. Vectors with covariant components are also called dual vectors or 1-forms.
These are imperative in relativity so let’s go into some detail on what dual vectors are.
Once we set up a vector space, we can define another associated vector space called a dual vector
space,which is typically denoted by an asterisk. Adopting this notation means that the dual space
to the tangent space Tp , called the cotangent space, is denoted by Tp∗ . The dual space is the space
of all linear maps from the original vector space to the real numbers. More precisely, if ω ∈ Tp∗ is a
dual vector, it acts as a map such that (Carroll,18)
ω(aU + bV ) = aω(U ) + bω(V ) (2.30)

which lives in the real numbers R. Here U and V are real vectors and a, b ∈ R. The collection
of all of these dual vectors form a vector space themselves. That is, if ω and η are dual vectors we
have
(aω + bη)(V ) = aω(V ) + bη(V ) (2.31)

We can also introduce a set of dual basis vectors θν by demanding
θν eµ = δ νµ (2.32)
where δ νµis the Kronecker-Delta, which is defined to 0 if µ 6= ν and one when µ = ν. In a
coordinate representation the dual basis vectors, or basis 1-forms take the form
θν = dxν (2.33)
In a later a chapter it will become clear, based on this representation, why we call dual basis
vectors basis 1-forms. Moreover, since we have a set of basis vectors, we can expand every dual
vector in terms of dual basis vectors using the Einstein summation convention
ω = ωµ θµ (2.34)
Again, the ωµ are the components of the dual vector and will often be referred to as the entire
dual vector. Moreover, one will often see the elements of the tangent space Tp (what are called
vectors in the ordinary sense) referred to as contravariant vectors. On the other hand, elements
of the dual space Tp∗ (what we have been calling dual vectors) are often referred to as covariant
vectors.
Using index notation, we can observe the action of a dual vector on a vector readily:
ω(V ) = ωµ θµ (V ν eν ) = ωµ V ν θµ eν = ωµ V ν δ µν = ωµ V µ (2.35)
where we used the fact that the components of a vector or dual vectors are just scalars and can
be pulled through the basis vectors and dual basis vectors. Moreover, since the components ωµ and
V µ are just real numbers, they live in R.
A rather straightfoward generalization of vectors and dual vectors is the mathematical object
called a tensor. Just as a dual vector is a linear map from vectors to the reals, a tensor T of rank
(k, `) is a multi-linear map sending a collection of dual vectors and vectors to the real numbers
(Carroll, 21). That is,
T : Tp∗ × ... × Tp∗ × Tp × ... × Tp → R (2.36)

| {z } | {z }
k−times `−times
where × denotes the Cartesian product such that Tp × Tp is the space of ordered pairs of vectors
(analogous to taking the Cartesian product R × R to come up with R2 ). By multilinearity we mean
that the tensor acts linearly for each argument. For example, consider a tensor of rank (1, 1), we
have
T (aω + bη, cU + dV ) = acT (ω, U ) + adT (ωV ) + bcT (η, U ) + bdT (η, V )
From here it’s easy to see that a tensor of rank (0, 0) is a scalar; a tensor of rank (1, 0) is a vector,
and a tensor of rank (0, 1) is a dual vector.
Just like vectors and dual vectors, tensors also form a vector space in which a basis may be
constructed. For this one must define the tensor product ⊗. If T is a (k, `) tensor and S is an
(m, n) tensor, we define a (k + m, ` + n) tensor T ⊗ S by
T ⊗ S(ω 1 , ..., ω k , ..., ω k+m , V 1 , ..., V ` , ...V `+n )

= T (ω 1 , ..., ω k , V 1 , ..., V ` ) × S(ω k+1 , ..., ω k+m , V `+1 , ...V `+n )
(2.37)
From here one constructs a basis space for all the (k, `) tensors by taking tensor products of
basis vectors and basis dual vectors, consisting of tensors of the form
eµ1 ⊗ ... ⊗ eµk ⊗ θν1 ... ⊗ θν` (2.38)

Thus, in component notation we can write an arbitrary tensor T as
T = T µ1 ...µkν1 ...ν` eµ1 ⊗ ... ⊗ eµk ⊗ θν1 ... ⊗ θν` (2.39)

where T µ1 ...µkν1 ...ν` are the components of the tensor T , and, as usual, we often refer to the
components of the tensor as the tensor.
Let’s now summarize a few basic algebraic operations that can be carried out with tensors to
produce new tensors. These operations essentially mimic the operations one does with ordinary
vectors. First of all, we may add tensors of the same type (rank) to get a new tensor. For example,
Rijk = S ijk + T ijk

In order to add tensors it is imperative that they are the same rank. This idea is analogous to
adding matrices. If they are not of the same size, the addition of matrices simply doesn’t make
sense. The same holds for tensor addition. We may also multiply tensors by scalars, for example
Sij = aTij
where a is some scalar. Both of these operations come from the fact that tensors form a vector
space. Moreover, using the addition of tensors, we can define the symmetric and antisymmetric
parts of a tensor. A tensor is symmetric if, upon exhange of any of its indices, does not change
sign. In the simplest case, the tensor Tij = Tji is symmetric, as the exchange of the indices did
not change sign of the tensor. On the other hand, a tensor is said to be antisymmetric if under
the exchange of any of its indices results in a change in sign. Therefore, we say that the tensor
Tij = −Tji is antisymmetric. We can then write the symmetric part of a tensor as
1
T(ij) = (Tij + Tji ) (2.40)
2
and the antisymmetric part of a tensor is
1
T[ij] = (Tij − Tji ) (2.41)
2
These expressions can be generalized to include more than two indices, however we will rarely
see tensors with more than two indices in this text and will therefore not go into the details in this
chapter. One of the most important examples of a symmetric tensor is the Kronecker-Delta, while
one of the most important examples of an antisymmetric tensor is the electromagnetic field tensor
Fµν . We will go into more detail on this tensor in the next chapter.
One final algebraic operation one can do with tensors is contraction. This can be used to turn a
(k, `) tensor into a (k − 1, ` − 1) tensor, which is done by setting a raised and lowered index equal
(McMahon, 55), e.g.
Rij = Rkikj
All of these mathematical objects, vectors, dual vectors, and tensors, belong to the subject of
Linear Algebra, and are appropriate when we have an abstract vector space to work with. In
relativity, we have a vector space at each point in space-time, and typically care about tensor fields,
which can be thought of as tensor-valued functions on space-time.
For the reader that is new to this material, don’t stress too much about about the technical
mathematical details of these objects. Certainly, it is important to understand what these objects
are and how they are defined, however what will be more important in this text is how to make
use of these objects from an operational view. That’s enough abstract mathematics for now. Let’s
move on to some more physical ideas that will be imperative in our study of relativity and string
theory.
2.5 Coordinate Transformations

In relativity, it is often necessary to change from one coordinate system to another. A transformation
of this kind is implemented with a transformation matrix, that we denote by Λµν 0 . Ordinary vectors
transform as (McMahon, 31)
0 0
V µ → V µ = Λµν V ν (2.42)
while dual vectors transform as
ωµ → ωµ0 = Λν µ0 ων (2.43)
In both of these expressions we notice that we have repeated indices, and therefore there is an
implicit sum by the Einstein summation convention. Based on the transformations in (2.42) and
(2.43), we can determine how the basis vectors and basis 1-forms transform. Expanding a vector V
we find
0 0
V = V µ eµ → V ν eν 0 = Λν µ V µ eν 0
which means that the basis vectors transform as
0
eµ = Λν µ eν 0 (2.44)
Similarly, basis 1-forms are found to transform as
0 0
θρ = Λρ σ θσ (2.45)
Using the fact that basis vectors are given in terms of partial derivatives in the coordinate frame,
0
we can find the form of the coordinate transformation matrix Λµν . This can be done by applying
the chain rule to a basis vector. We have
ν
∂ ∂ ∂x ∂xν ∂
eµ0 = 0 = 0 = = Λν µ0 eν
∂x µ ∂x µ ∂x ν ∂xµ0 ∂xν
Thus,
∂xν
Λν µ0 = (2.46)
∂xµ0
For concreteness, let us consider an example with polar coordinates. We are familiar that the
relationship between polar and cartesian coordiantes is given by
x = r cos θ y = r sin θ (2.47)

Using our current definition in (2.46) we find the components of the transformation matrix to be
∂x ∂y
Λxr = = cos θ Λy r = = sin θ
∂r ∂r
∂x ∂y
Λxθ = = −r sin θ Λy θ = = r cos θ
∂θ ∂θ
2.6. THE METRIC 43
From here we can the write the basis vectors in polar coordinates. Using (2.44) and making note
that we are dealing with an implicit sum, we find
er = Λxr ex + Λy r ey = cos θex + sin θey
eθ = Λxθ ex + Λy θ ey = −r sin θex + r cos θey

which are the expected basis vectors in polar coordinates.
2.6 The Metric

In general relativity, and therefore string theory, the metric and associated metric tensor are so
crucial it is imperative that we spend some time developing our physical and mathematical intuition
for them. The metric is also fundamental to mathematicians, particularly in the field differential
geometry. Let us first consider the mathematical definition before moving to its physical significance.
Suppose we have some abstract set M and we would like to define a distance function on M .
This distance function could, for example, tell us the distance between two points on a number line
or the distance between two pairs of points on a plane. Suppose that we have points x, y ∈ M , and
let g be defined as a function taking the set of all ordered pairs (x, y), M × M , and map them to
the positive reals, i.e. M × M → R+ . Such a function, satisfying the following properties is called
a metric on M (Carothers, 38):
(1) 0 ≤ g(x, y) < ∞
(2) g(x, y) = 0 if and only if x = y
(3) g(x, y) = g(y, x)
(4) g(x, y) ≤ g(x, z) + g(z, y)
for all x, y, z ∈ M . Formally, we call the couple (M, g) a metric space. When we write g(x, y),
what we mean is the distance between the points x and y. With this in mind, the first property says
that the distance between two points is positive definite (a property which we relax in relativity).
The second property points out that the distance between a point and itself is zero, and (3) is a
statement about symmetry: the distance between x and y is the same as the distance between y and
x. Finally, (4) is properly known as the triangle inequality and is the embodiment of the shortest
distance between two points is a straight line. This last notion changes slightly when we consider
spaces that have curvature.
In differential geometry as well as general relativity, we view the metric as a tensor. In fact,
the metric tensor is a symmetric second rank tensor, i.e. gµν = gνµ . It makes sense that it is a
symmetric tensor, as the metric, as defined above, is symmetric. Reminding ourselves that tensors
are maps from vectors and 1-forms to the real numbers, we can view the metric in another way.
In particular, as a second rank tensor that accepts two vector arguments, the output must also be
a real number. But we know of an operation which takes two vectors and sends them to the real
numbers: the inner product between two vectors. We also know that the inner product between
two vectors is symmetric, i.e. V · W = W · V for vectors V and W . With this as motivation, we
can convince ourselves that (McMahon, 43)
g(V, W ) = V · W (2.48)
indicating that the metric tensor gives us the traditional dot product.
Up to this point we have defined the metric arbitrarily as a distance function, but we also know
it is a symmetric second rank tensor. The tensor itself is formally written as
g = gij dxi ⊗ dxj (2.49)

where gij are the components of the metric tensor g. This definition is often used interchangeably
with the line element amongst physicsts, which is defined as
ds2 = gµν dxµ dxν (2.50)

where we still have the components of the metric which can be represented as a matrix. Written
like this, the line element further illustrates that the metric is like a distance function. The simplest
line element familiar to us from elementary calculus is
ds2 = dx2 + dy 2 + dz 2 (2.51)

This formula reminds us of the Pythagorean theorem for an infinitesimal line in cartesian co-
ordinates, and we wouldn’t be wrong. In some sense, the line element is simply the infinitesimal
version of the Pythagorean theorem. The reason why we care about the line element is with it we
can determine the components of the metric tensor. Using the summation convention in (2.50), we
write out the line element in cartesian coordinates
ds2 = gxx dx2 + gyy dy 2 + gzz dz 2 (2.52)

Comparing (2.51) to (2.52) we find gxx = gyy = gzz = 1, while all of the cross terms gxy = gxz =
gyz = 0. This allows us to write the matrix of the components of the metric
 
1 0 0
[gij ] = 0 1 0 (2.53)
0 0 1
For spherical coordinates, the line element is just
ds2 = dr2 + r2 dθ2 + r2 sin2 θdφ2 (2.54)

which allows us to write the components of the metric as
 
1 0 0
[gij ] = 0 r2 0  (2.55)
0 0 r sin2 θ
2
In the case of Special Relativity, we are dealing with flat Minkowski space, which has the line
element in cartesian coordinates
ds2 = −dt2 + dx2 + dy 2 + dz 2 (2.56)

2.7. INDEX GYMNASTICS 45
Sometimes this will be written instead with the identification (t, x, y, z) → (x0 , x1 , x2 , x3 ). From
the line element, we see that the matrix of the components of the metric can represented by
 
−1 0 0 0
 0 1 0 0
[gµν ] = 
 0 0 1 0
 (2.57)
0 0 0 1
Keeping the convention of the literature, we denote the Minkowski metric by ηµν . In an arbitrarily
curved space-time, gµν is often used.
Since the components of the metric may be represented as a matrix, we can find the inverse of
the metric, g µν . Specifically, it is defined by
gij g jk = δi k (2.58)
where δi k is the Kronecker-delta. When the metric is diagonal, it is rather simple to find the
inverse metric. Consider the metric in spherical coordinates as shown in (2.55). Notice then that
g rr grr = 1 ⇒ g rr (1) = g rr = 1
1
g θθ gθθ = g θθ r2 = 1 ⇒ g θθ =
r2
1
g φφ gφφ = g φφ r2 sin2 θ = 1 ⇒ g φφ =
r2 sin2 θ
Hence the components of the inverse metric can be arranged as
 
1 0 0
[gij ] = 0 r12 0  (2.59)
1
0 0 r2 sin 2θ
For line elements with cross terms, finding the inverse metric is a more laborious task.
2.7 Index Gymnastics

We have already seen how the metric is a type of distance function. The metric also have an
important role in computations in relativity theory. As the reader will become comfortable with,
in relativity the metric is used to manipulate expressions by raising and lowering the indices of the
object. First of all, we can use the metric to obtain the covariant components of a vector as
Vi = gij V j (2.60)
To obtain contravariant components of a vector, we instead use
V i = g ij Vj (2.61)
The ability to raise and lower indices using the metric is so crucial to the computations in
relativity, one should practice doing so to the point of exhaustion, and then a little more. In this
text, aside from the abstract mathematical definitions we come across, the most difficult task, as
far as computations are concerned, is to raise and lower indices. For that reason, let’s go through
an example, and leave more for the reader to work through.
Suppose again that we are working in spherical coordinates where we have a contravariant vector
X i = (1, r, 0), and a covariant vector Yi = (0, −r2 , cos2 θ). Using
Xi = gij X j
we find the components of the vector dual to X i :
Xr = grr X r = (1)(1) = 1 Xθ = gθθ X θ = (r2 )(r) = r3
Xφ = gφφ X φ = (r2 sin2 θ)(0) = 0

Altogether we have Xi = (1, r3 , 0). In the second case we use
Y i = g ij Yj (2.62)
Using the components of the inverse metric we find the vector dual to Yi :

1
Y r = g rr Yr = (1)(0) = 0 Y θ = g θθ Yθ = (−r2 ) = −1
r2
1 cot2 θ
Y φ = g φφ Yφ = 2
2 cos θ =
r2 sin θ r2
Typically we will raise and lower indices with the metric in a more abstract fashion. For instance,
we might write
X i Y k = g ij Xj Y k
Or,
Rijkl = gim Rmjkl

Following these examples, we can realize a help tip in analyzing the validity of expressions with
indices. Consider the above expression. Notice that the location and indices used on either side of
the equation are equal. That is, on the left hand side, there are three lowered indices, and on the
right hand side, again, there are the same three lowered indices. One might object to this as on the
right hand side there is an implicit sum over the index m. But that is just it: m is a dummy index
that is summed over. Therefore, if one explicitly wrote out the sum, the expression would only
contain the the three lower indices. This method is sometimes referred to as balancing of indices,
and it allows for a quick and dirty method of evaluating expressions with several indices.
Lastly, as we discussed before, the metric tells us how to compute the dot product of two vectors.
Using index notation, the dot product between vectors V and W is written as
V · W = Vi W i (2.63)
Then, using the components of the metric
2.7. INDEX GYMNASTICS 47
V · W = Vi W i = gij V j W i = g ij Vi Wj (2.64)
Remember from before when we looked at four-vectors. In particular, recall that we wrote the
dot product of the four-momentum with itself as
p2 = p · p = pµ pµ = p0 p0 + p1 p1 + p2 p2 + p3 p3 (2.65)
In special relativity, we work in flat space-time with the Minkowski metric ηµν . Using this metric
to raise and lower indices, we find that we introduce a minus sign into our expression for the dot
product:
η00 p0 p0 + η11 p1 p1 + η22 p2 p2 + η33 p3 p3
= −p0 p0 + p1 p1 + p2 p2 + p3 p3
E2
=− + p2x + p2y + p2z
c2
It is important to point out the most frustrating convention in all of physics. The signature of
the metric, as far as we are concerned, is defined as the sum of the diagonal elements of the matrix
of gµν . In our present case of the Minkowski metric, we have that the sig(ηµν ) = +2. Another way
of saying this is we are working with the mostly plus convention. This is the convention typically
taken in Special and General Relativity. However, when one studies particle physics, the Minkowski
metric is chosen to be ηµν = diag(1, −1, −1, −1), which changes all of the calculations in relativity
by a minus sign. Using this, mostly minus convention, we find that the dot product of the energy
momentum tensor with itself becomes
E2
p2 = − p2x − p2y − p2z
c2
In the majority of this text, we will use the mostly plus convention of the Minkowski metric.
However, in order to maintain the convention used in the literature, we will use the mostly minus
convention in the chapter on supersymmetry. It is also important to note that with the Minkowski
metric, we see that the dot product between two vectors is not always positive, but sometimes
zero (corresponding to a null vector),or negative. Hence, in relativity, the metric is semi-positive
definite, thereby relaxing the first property of the metric given in the definition earlier on.
We started with a basic mathematical definition of the metric, which can be viewed as a distance
function, allowing one to calculate the length between points on a number line or the distance
between pairs of points in a plane. In fact, the metric is even more than that. The metric tensor
also encodes what type of space we are in, whether it is curved or flat. Therefore, the metric g,
in a sense, gives the geometry of the space we are dealing with. In general relativity this notion
of the metric is imperative as the curvature of our background space-time leads to a description of
gravity.
2.8 Light Cone Coordinates

In string theory, one of the ways we will quantize the string is through something called the light-
cone gauge. Here we make use of a convenient coordinate system called light-cone coordinates. This
coordinate system will be used extensively in this text, so let us take some time to become familiar
with it.
Let us again consider the Minkowski line element, this time with the mostly minus convention:
2 2 2 2
−ds2 = − dx0 + dx1 + dx2 + dx3
Although this line element looks quite tidy (in the sense that there are no cross-terms, the squared
terms (and the minus sign) can be challenging to work with in some cases. Since the physics is the
same in all coordinate frames, let us use a coordinate change to remove the squares. We impose
the following coordinate transformations:
1
x0 → x+ = √ (x0 + x1 ) (2.66)
2
1
x1 → x− = √ (x0 − x1 ) (2.67)
2
The new coordinates described are called light cone coordinates, also referred to as null coordi-
nates. Our line element will change according to the new choice of coordinates; notice that if we
multiply our + and - coordinates together, we arrive at
1 0 1
x+ x− = (x + x1 )(x0 − x1 ) = (x0 )2 − (x1 )2
2 2
and multiplying by −2 gives
−2x+ x− = −(x0 )2 + (x1 )2
Directly substituting this into our original line element, the null coordinate line element will take
the form
2 2
−ds2 = −2dx+ dx− + dx2 + dx3 (2.68)
Figure 2.4 gives a diagram of space-time in light-cone coordinates. Now, we need to build the
metric to describe this new coordinate system. Notating the null coordinate metric as η̂ab , we can
use
−ds2 = η̂ab dxa dxb
to read off the null metric components from the line element. The null coordinate line element’s
−2dx+ dx− will now give the metric off-diagonal components η̂+− and η̂−+ . Unfortunately, since
dx+ dx− = dx− dx+ , it is unclear which components get what from the −2 term in the line element.
Since we know the metric must be symmetric, we choose to distribute the −2 evenly between the
elements, giving
2.9. EXERCISES 49
 
0 −1 0 0
−1 0 0 0
η̂ab =
0
 (2.69)
0 1 0
0 0 0 1
Using the components of η̂ ab , we can also write that the scalar product in null coordinates as
V · S = η̂ab V b S a = −V + S − − V − S + + V 2 S 2 + V 3 S 3 (2.70)
Figure 2.4: A space-time diagram using light-cone coordinates.
2.9 Exercises
1. Derive the Lorentz transformation matrix given in (2.6).
2. State and explain how the pole-in-barn paradox is resolved.
3. Let the tensor Qij be symmetric, and let Rij be an antisymmetric tensor. Prove that
Qij Rij = 0
4. Recall that cartesian coordinates are related to spherical coordinates by
x = r sin θ cos φ y = r sin θ sin φ z = r cos θ

(a) Find the components of the transformation matrix Λν µ0 .
(b) Detemine the basis vectors in spherical coordinates.

5. Consider the line element associated with the Bondi metric:

f 2β
ds2 = e − g 2 r2 e2α du2 + 2e2β dudr + 2gr2 e2α dudθ − r2 (e2α dθ2 + e−2α sin2 θdφ2 )
r
Write out the components of this metric and represent it in matrix form. This line element is
used in the study of gravitational waves [38].
6. Consider the metric in plane polar coordinates, with g11 = 1 and g22 = r2 , and g12 = g21 = 0.
Moreover, let V i = (1, 1) be a contravariant vector, and Wi = (0, 1) be a covariant vector. Calculate
Vi , W i , and find V · W .
7. (a) Determine x+ and x− by using the Minkowski metric in light-coordinates.
(b) Using (2.66) and (2.67) as motivation, come up with expressions for the light-cone momenta
p+ and p− . Calculate p+ , p− , and determine p · p in light-cone coordinates.
Chapter 3
Non-Relativistic Strings and the

Relativistic Point Particle
To build our intuition for String Theory, a good place to start is to review the dynamics of classical
strings. That is, we aim to first understand the system of a classically vibrating string, particularly
to determine the equations of motion which govern the behavior of the system. In the next chapter
we begin exploring the behavior of relativistic strings. Therefore we must first be familiar with
the dynamics of a relativistic point particle. Here, we develop our understanding for both of these
systems, laying the background necessary for the rest of the text.
Before we analyze a vibrating string, let us first examine the basics of Lagrangian mechanics
and Lagrangian field theory. Lagrangian mechanics will be used later on in this chapter, while
field theory will be fundamental in later chapters. The discussion on field theory follows naturally
from Lagrangian mechanics, and the notation used follows directly from the previous chapter,
however, for those who are new to this subject can skip the discussion on field theory and gauge
transformations for now, however should keep in mind that these topics will become imperative for
understanding the material in later chapters.
3.1 Basic Lagrangian Mechanics

A Lagrangian L of a system is defined as,
L=T −V (3.1)
where T is the kinetic energy of the system and V is the potential energy of the system. Let us
work out a simple example to recall the details of Lagrangian mechanics. Consider a point particle
of mass m moving in the x-direction with time independent potential energy. The Lagrangian for
this system takes the form,
1
L= mẋ(t)2 − V (x(t))
2
The action S is defined as
51
52CHAPTER 3. NON-RELATIVISTIC STRINGS AND THE RELATIVISTIC POINT PARTICLE
Z
S= L(t)dt (3.2)
C
where C is some path x(t) between the initial position and initial time (xi ,ti ), and the final position
and final time (xf ,tf ). More specifically, the action is a functional. That is, the action takes
functions as inputs and returns scalars. For our system, the action is written as
Z tf
1
S[x] = mẋ(t)2 − V (x(t))dt
ti 2
Recall from classical mechanics that by Hamilton’s Principle, particles follow the path of least
action. Thus, if we vary the action, the path the particle takes can be determined. More precisely,
if the curve C is varied infinitesimally, the action does not change to first order in variation. The
perturbed path can be written as
x(t) + δx(t)
where δx(t) is the variation. Moreover, we consider variations such that the initial and final
positions remain the same, i.e.
δx(ti ) = δx(tf ) = 0
This variation then changes the action in the following way: S → S + δS. The path the particle
takes occurs when δS = 0. Let’s calculate the action S[x + δx] for our given system:
Z tf
m d
S[x + δx] = [ ( [x + δx])2 − V (x(t) + δx(t))]dt (3.3)
ti 2 dt
The potential energy term can be approximated by a Taylor series expansion,
dV (x)
V (x(t) + δx(t)) ≈ V (x(t)) + δx(t)
dx
Then,
Z tf
m 2 d d dV (x)
S[x + δx] = [ (ẋ + δx(t)2 + 2ẋ δx(t)) − (V (x(t)) + δx(t) )]dt
ti 2 dt dt dx
Z tf
d dV (x)
= S[x] + (mẋ δx(t) − δx(t) )dt
ti dt dx
To get to the second line, we assumed that δx(t) << 1, thereby getting rid of terms of δx(t) of
orders higher than one. Therefore, we have that
Z tf
d dV (x)
δS = (mẋ δx(t) − δx(t) )dt (3.4)
ti dt dx
We can rewrite the above using integration by parts on the first term,
Z tf Z tf
d tf
mẋ(t) δx(t)dt = mẋδx(t)|ti − δx(t)mẍ(t)dt
ti dt ti
3.1. BASIC LAGRANGIAN MECHANICS 53
Then, using the constraint δx(ti ) = δx(tf ) = 0, the variation of the action becomes
Z tf
dV
δS = δx(t)(−mẍ(t) − )dt (3.5)
ti dx
The action is said to be stationary if δS = 0 for every variation δx(t). This only occurs when
the integrand multiplying δx(t) is identically zero. That is,
dV
mẍ(t) + =0 (3.6)
dx
Equation (3.6) is easily recognized as Newton’s second law for the motion of a particle in some
time independent potential V (x).
Of course, we may generalize this result. Consider now N generalized coordinates qi (t) for
i ∈ 1...N . Consider a Lagrangian L expressed in terms of qi (t) and q̇i (t). The action then takes the
form,
Z t2
S= L(q1 , q2 , ...qN , q̇1 , q̇2 , ...q̇N )dt (3.7)
t1
To find the equations of motion of the system, the principle of least action is applied in the same
way as before. We find that,
Z t2 Z t2 X ∂L ∂L
δS = δ L(q1 , q2 , ...qN , q̇1 , q̇2 , ...q̇N )dt = ( δqi + δ q̇i )dt
t1 t1 i
∂qi ∂ q̇i
And since
d d d
(δq) = q δ + δ q̇(t) = δ q,
dt dt dt
we write
Z t2 X ∂L ∂L d
δS = ( δqi + (δqi ))dt (3.8)
t1 i
∂qi ∂ q̇i dt
Using integration by parts yields,

Z t2 X ∂L X ∂L Z t2
t2 d ∂L
( δqi )dt + ( δqi |t1 − δqi )dt
t1 i
∂qi i
∂ q̇i t1 dt ∂ q̇i
Z t2 X ∂L d ∂L
= ( δqi − δqi )dt (3.9)
t1 i
∂qi dt ∂ q̇i
To get from the first line to the second line, we again used the constraint that δqi (t1 ) = δqi (t2 ) = 0
∀i. Just as before, for the action to be stationary, it must be that δS = 0 for any variation δqi ,
yielding,
∂L d ∂L
− =0 (3.10)
∂qi dt ∂ q̇i
which are the familiar Euler-Lagrange equations for N generalized coordinates.
From here we define the canonical momentum to be

∂L
pi = (3.11)
∂ q̇i
allowing us to write the Hamiltonian as
X
H(p, q) = pi q̇i − L (3.12)
i
3.2 Lagrangian Field Theory

In most texts, including this one, one works with functions of space-time, or rather, fields, and will
thereby generalize the techniques above to include fields φ(x). Then, instead of working directly
with a Lagrangian L, we work with the Lagrangian density, L, which is related to the Lagrangian
in the following way,
Z
L = Ld3 x (3.13)
The action is then easily defined to be

Z Z
S= Ldt = Ld4 x (3.14)
Typically we have interest in fields that are local. That is, given some point in space-time x, the
Lagrangian density L depends on the fields and their first derivatives at that point. In an analogous
way as above, the variation of the action becomes
Z Z
4 4 ∂L ∂L
δS = δ Ld x = d x δφ + δ(∂µ φ) (3.15)
∂φ ∂(∂µ φ)
Then, again using the fact that δ(∂µ φ) = ∂µ (δφ), and using integration parts on the second term
in the integral, we find
Z
4 ∂L ∂L
δS = d x δφ − ∂µ δφ (3.16)
∂φ ∂(∂µ φ)
To find the equations of motion, the variation of the action δS goes to zero, yielding

∂L ∂L
− ∂µ =0 (3.17)
∂φ ∂(∂µ φ)
The above can be interepted simply as the Euler-Lagrange equations for a field φ. It is important
to note that the Einstein summation convention is in effect on the second term.
In an analogous way, the canonical momentum density is given by
∂L ∂L
Π(x) = = (3.18)
∂(∂0 φ) ∂ φ̇
Then the Hamiltonian density is defined as
H = Π(x)φ̇(x) − L (3.19)
3.2. LAGRANGIAN FIELD THEORY 55
Lastly, to obtain the Hamiltonian, simply integrate the Hamiltonian density H over all space
Z
H = Hd3 x (3.20)
To become more familiar with the notation and elegance of field theory, consider the Maxwell
tensor (often referred to as the Electromagnetic Field tensor) given by
 
0 −Ex −Ey −Ez
Ex 0 −Bz By 
F µν =   (3.21)
Ey Bz 0 −Bx 
Ez −By Bx 0
As can be inferred from above, the Maxwell tensor encodes all of the necessary information of
the electromagnetic field of a physical system. Moreover, in this form, it is easy to see that Maxwell
tensor is antisymmetric, i.e., F µν = −F νµ (or with covariant indices, Fµν = −Fνµ ). Alternatively,
the Maxwell tensor in field theory notation takes the form
F µν = ∂ µ Aν − ∂ ν Aµ (3.22)
~ It can be easily checked that the inhomoge-

where Aµ is the usual 4-vector potential, Aµ = (φ, A).
neous Maxwell’s equations can be summarized with the simple expression
∂µ F µν − J ν = 0 (3.23)
For practice, consider the Lagrangian density L = − 14 Fµν F µν − J ν Aν . The action is then,
Z
1
S = d4 x(− Fµν F µν − J ν Aν ) (3.24)
4
Let us vary the action by only varying the potential Aµ . Then,
Z
1 1
δS = d4 x(− (δFµν )F µν − Fµν (δF µν ) − J ν Aν ) (3.25)
4 4
To see what varying this action gives, we carefully consider each term in the integral. Using the
definition of the Maxwell tensor, the first term can be rewritten as
1 1
− (δFµν )F µν = − (∂µ δAν − ∂ν δAµ )F µν
4 4
Integrating by parts yields
Z Z
4 1 µν 4 1 µν µν
d x(− (δFµν )F ) = d x (∂µ F δAν − ∂ν F δAµ )
4 4
Since repeated indices are dummy variables, we swap µ and ν in the second term, resulting in
Z
1
d4 x (∂µ F µν δAν − ∂µ F νµ δAν )
4
Then, using the antisymmetry of F µν , we find that the above becomes
Z Z
1 1
d x (2∂µ F µν δAν ) =
4
d4 x ∂µ F µν δAν (3.26)
4 2
Let’s move on to the second term in (3.25). Using the definition of the Maxwell tensor, we find
Z Z
1 1
d x(− Fµν (δF )) = d4 x(− Fµν (∂ µ δAν − ∂ ν δAµ )
4 µν
4 4
Using an arbitrary metric, we can raise the indices on Fµν , yielding
Z
4 1 ρσ µ ν ν µ
d x − gµρ gνσ F (∂ δA − ∂ δA )
4
Z
1 1
= d4 x gµρ gνσ ∂ µ (F ρσ )δAν − gµρ gνσ ∂ ν (F ρσ δAµ )
4 4
where we used integration by parts. Then, by lowering indices, we find that the above becomes
Z
4 1 ρσ 1 ρσ
d x ∂ρ F δAσ − ∂σ F δAρ
4 4
Relabeling dummy indices gives,
Z
1 1
d x ∂µ F µν δAν − ∂µ F νµ δAν
4
4 4
Z
1
= d4 x ∂µ F µν δAν (3.27)
2
Altogether we have
Z
1 1
δS = d x 4
∂µ F µν δAν + ∂µ F µν δAν − J ν δAν
2 2
Z
= d4 x (∂µ F µν − J ν ) δAν (3.28)
Requiring the action to vanish, by the usual argument we find the compact expression for
Maxwell’s equations
∂µ F µν − J ν = 0
3.3 Gauge Transformations

As physicists, we often care about symmetry. In particular, we care whether or not two systems
are physically equivalent. One method of determining this equivalence is through gauge transfor-
mations.
To gain a basic understanding of what gauge transformations are, recall Maxwell’s equations in
the Heaviside-Lorentz unit system:
3.3. GAUGE TRANSFORMATIONS 57
~
∇ ~ = − 1 ∂B
~ ×E ~ ·B
∇ ~ =0
c ∂t
~
~ ·E
∇ ~ =ρ ~ ×B
∇ ~ = 1~j + 1 ∂ E
c c ∂t
Moreover, recall from any introductory course on vector calculus that any divergenceless vector
~ ·B
field can be written as the curl of some vector field. Then, since ∇ ~ = 0, we may write
~ ×A
∇ ~=B
~ (3.29)
~ is often called the vector potential. By direct substitution we have
where A
!
1 ∂B~ ~
1 ∂A
~
∇×E+ ~ ~
=∇× E+ ~ =0
c ∂t c ∂t
Another familiar result from vector calculus is that the curl of any gradient vector field is zero.
By the above, we may therefore write the term inside the parentheses, E ~ + 1 ∂ A~ = −∇Φ, where Φ
c ∂t
is the scalar potential. We may write then
~
~ = − 1 ∂ A − ∇Φ
E (3.30)
c ∂t
As it turns out, the potentials A~ and Φ associated with E

~ and B~ are not unique. To witness
~0 = A
this, consider the transformation, A ~ + ∇,
~ where is some small function of space-time. Then

~0 = ∇
B ~ ×A ~0 = ∇~ A~ + ∇
~ ~ ×A
=∇ ~+∇ ~ × ∇
~
~ ×A
=∇ ~=B
~
In other words, this transformation leaves the magnetic field invariant. Note however that this
~ does not leave E
change in A ~ invariant:
~0
~ 0 = − 1 ∂ A − 1 − ∇φ = − 1 ∂ A

E ~ + ∇ − ∇φ
c ∂t c c ∂t
1 ∂ ~ 1 ∂ ~
=− A− ∇ − ∇φ 6= E
c ∂t c ∂t
Rather, consider the transformation
1 ∂
φ0 → φ − (3.31)
c ∂t
Then,

~0 = − 1 ∂ A ~ + ∇ − ∇ φ − 1 ∂ = − 1 ∂ A~ − 1 ∂ ∇ − ∇φ + 1 ∇ ∂

E
c ∂t c ∂t c ∂t c ∂t c ∂t
1 ∂ ~ ~
=− A − ∇φ = E
c ∂t
The transformations A ~0 → A ~ + ∇,

~ and φ0 → φ − 1 ∂ leave both E~ and B
~ invariant. It is this
c ∂t
pair of transformations in the potentials which are formally known as gauge transformations, and
where is often referred to as the gauge parameter.
Moreover, using the notation we have been developing, we can summarize the gauge transfor-
mations for the electromagnetic fields neatly with index notation. Namely,
Aµ → Aµ + ∂µ (3.32)
where = (x) is a function of spacetime coordinates, and where we write A0 = φ. Furthermore,

since this gauge transformation leaves E ~ and B
~ invariant, it is expected, and can be easily verified,
that the field strength, Fµν , the ordinary Maxwell tensor, is also gauge invariant.
Gauge invariance isn’t a sole property of the physical fields themselves. In fact, we often im-
plement gauge invariance on the action S, and thereby we seek a gauge invariant Lagrangian L or
Lagrangian density L. To provide a general notion of a gauge invariant action, we start from a
theory which possesses a global U (1) invariance. For now we needn’t worry about the technicalities
of U (1) invariance. It is fundamental in quantum field theory, a necessary prerequisite for fully
understanding string theory, however, we avoid those messy details as much as possible. With-
out loss of generality, and pulling it from a quantum field theorist’s hat, we consider the complex
Klein-Gordon Lagrangian density
L = ∂µ φ∗ ∂ µ φ − m2 φ∗ φ (3.33)
This Lagrangian density is composed of a complex field φ, with its real and imaginary parts
individually satisfying the Klein-Gordon equation
+ m2 φ = ∂µ ∂ µ + m2 = 0

(3.34)
As a reminder, this equation was one of the first attempts at a relativistic equation that seeks
to describe quantum mechanics. We then consider the following transformation
φ → eiqθ φ (3.35)
We immediately notice that if θ is constant, then L is left invariant
L0 = ∂µ e−iqθ φ∗ ∂ µ eiqθ φ − m2 φφ∗

= ∂µ φ∗ ∂ µ e−iqθ eiqθ − m2 e−iqθ eiqθ φ∗ φ

= ∂µ φ∗ ∂ µ − m2 φ∗ φ = L
Transformations where θ is constant are called global gauge transformations. Though important,
there is more interesting physics in local gauge transformations, in which θ = θ(x). That is, the
gauge parameter θ depends on space-time coordinates. Notice however that this generic transfor-
mation no longer leaves L invariant the derivative ∂µ operates on eiqθ , and does not cancel with
e−iqθ .
To fix the lack of invariance, we define two other types of transformations, one which we are
familiar with. Recall that before we had
Aµ → Aµ + ∂µ
3.4. CLASSICAL STRING DYNAMICS 59
It is also true that the electromagnetic fields are invariant under the transformation
Aµ → Aµ − ∂ µ θ
Then, we simply define the covariant derivative of φ as
Dµ φ = (∂µ + iqAµ ) φ (3.36)
Using these two transformations we verify that
Dµ φ → Dµ eiqθ φ = (∂µ + iqAµ ) eiqθ φ = iq∂µ θeiqθ φ + eiqθ ∂µ φ + iqeiqθ Aµ φ

→ iq∂µ θeiqθ φ + eiqθ ∂µ φ + iqeiqθ Aµ φ − iq∂µ eiqθ φ = eiqθ (∂µ φ + iqAµ φ) = eiqθ Dµ φ
Summarizing, we notice that with both of these transformations the covariant derivative trans-
forms in the same way as the fields
Dµ eiqθ φ → eiqθ Dµ φ (3.37)
In other words, Dµ transforms analogously as φ, even if the gauge parameter depends locally on
space-time. From here we can easily construct a Lagrangian with local U (1) invariance by replacing
all partial derivatives ∂µ with covariant derivatives Dµ , promoting the symmetry from global to
local. For example, if we substitute Dµ in for ∂µ , the complex Klein-Gordon Lagrangian density
we find
L = Dµ φ∗ Dµ φ − m2 φ∗ φ
2
= ∂µ φ∂ µ φ∗ + iqAµ (φ∂µ φ∗ − φ∗ ∂µ φ) + q 2 |φ| Aµ Aµ − m2 φ∗ φ
It turns out this Lagrangian describes scalar electrodynamics (Maggiore, 71). More interestingly,
2
the term q 2 |φ| Aµ Aµ has a role with the Higgs mechanism, which is imperative to the Standard
Model and most branches of theoretical particle physics.
Now that we have had a (rather lengthy) review of Lagrangian mechanics and classical field
theory, a pertinent question arises: why use the Lagrangian formulation in the first place? There
are a multitude of reasons for this but perhaps the most important is that we can show that the
Lagrangian density is Lorentz invariant. In fact, it is a Lorentz scalar, transforming as a scalar
under Lorentz transformations. The fact that the Lagrangian density is Lorentz invariant is crucial
since it means the consequent action S will also be invariant under Lorentz transformations, and
we may therefore construct a Lorentz invariant theory. This is particularly important in quantum
field theory, and string theory, as we desire Lorentz invariance.
3.4 Classical String Dynamics

Consider a thin string stretched horizontally via a tensional force To . Suppose that this string can
be deformed from its original shape by some force however small compared to To . It is safe to
assume then that the slope of the deformation is small everywhere during its motion. Lastly, we
further suppose that the string is not subject to any horizontal forces other than the tension T . To
aid in this discussion, consider the diagram in figure 3.1 (Butkov, 289).
Figure 3.1: A stretched string.
From the diagram, we can readily see that for an arbitrary segment of the string AB we have
T1 cos α = T2 cos β (3.38)

By our above restrictions (3.38) holds at all times. Since we are dealing with a small deformation,
we may make the approximation:
1 1
cos α = p ≈1− tan2 α ≈ 1
2
1 + tan α 2
Similarly, cos β ≈ 1, yielding
T1 = T2 = To (3.39)
The equations of motion are satisfied by deflection y of the string at any point along the string,
and at any time t. Considering an infinitesimal element dx, and with an external deflection force
F (force per unit length at a point x), we have by Newton’s Second Law,
∂2y
To (sin β − sin α) + F dx − µgdx = µdx (3.40)
∂t2
2
where µ is the mass density of the string, and ∂∂t2y is the acceleration of the string in the transverse
direction. Then, using the approximation

tan α ∂y
sin α = p ≈ tan α =
1 + tan2 α ∂x x+dx
and similarly
∂y
sin β ≈
∂x x
3.4. CLASSICAL STRING DYNAMICS 61
we find that (3.40) becomes

!
∂y ∂y

To ∂x − ∂x
To (sin β − sin α) ∂2y x+dx x ∂2y
+ F − µg = µ 2 ≈ + F − µg = µ
dx ∂t dx ∂t2
∂2y ∂2y
≈ To + F − µg = µ
∂x2 ∂t2
This is the general expression for a vibrating string under our the proposed assumptions. We
can further simplify the expression by considering the case when µg F , in which,
∂2y ∂2y
To + F = µ(x)
∂x2 ∂t2
Moreover, if there is no external force F and if the mass density is constant(i.e. µ(x) = µ), then
we have
∂2y ∂2y
To 2 = µ 2 (3.41)
∂x ∂t
Recall the one dimensional wave equation,
∂2y 1 ∂2y
2
= 2 2 (3.42)
∂x c ∂t
By comparison we find that the speed of the wave along a vibrating string is given by
To
c2 = (3.43)
µ
General solutions of this partial differential equation take the form:
y(t, x) = h+ (x − vt) + h− (x + vt) (3.44)
where h+ and h− denote right moving and left moving waves along the string. Furthermore,
since the wave equation is a partial differential equation involving both space and time, we require
boundary conditions and initial conditions (Zwiebach, 75). The most common types of boundary
conditions are Dirichlet and Neumann boundary conditions.
Figure 3.2: Dirichlet and Neumann boundary conditions.

On a string, Dirichlet boundary conditions state that the positions of the endpoints of the string
are fixed. That is,
y(t, x = 0) = y(t, x = a) = 0 (3.45)
where a is the ending position of the string. Alternatively, imagine we were to attach massless loops
to each end of the string endpoints with the loops connected to two frictionless poles. Neumann
∂y
boundary conditions specify values of ∂x at the endpoints. Since the poles are frictionless and
the loops are massless, we require that Neumann boundary conditions, in this particular example,
satisfy,
∂y
=0 (3.46)
∂x x=0,a
If this condition was not satisfied, the slope of the string at the pole would be nonzero, thus giving
a component of the string tension which would accelerate the rings in the y direction. However,
since the rings are massless, this acceleration would be infinite. Therefore, we apply Neumann
boundary conditions as defined above so that the endpoints of the string are allowed to move freely
in the y-direction. For an illustration of these conditions, refer to figure 3.2. In general, Dirichlet
boundary conditions are defined to be such that the values of the function y are specified at each
point along the boundary, and Neumann boundary conditions are defined to be such that the values
∂y
of ∂n (the normal derivative of y) are specified at each point on the boundary( Riley,752). Both
of these boundary conditions have important consequences when it comes to the interactions of
strings with branes as we will see later on.
3.5 Lagrangian for a Non-Relativistic String

Another method of finding the equations of motion, in a more fundamental way, is to use Lagrangian
mechanics. The kinetic energy T of the entire string is simply the sum the kinetic energies of each
infinitesimal segment of the string. Therefore,
Z a 2
1 ∂y
T = µdx (3.47)
0 2 ∂t
where the endpoints of the string are located at x = 0 and x = a. The potential energy isn’t as
obvious. We can determine the potential energy however from the work done on a small segment
of the string. Consider the segment (x, 0) → (x + dx, 0). If we momentarily stretch the string from
(x, y) to (x + dx, y + dy), then the change in length of the small segment is given by
s 
2
p ∂y
∆` = dx2 + dy 2 − dx = dx  1 + − 1
∂x
2
∂y
Since we are still assuming that the string under goes small oscillations, ∂x 1. Therefore,
we may use a series approximation, yielding
2
1 ∂y
∆` ≈ dx (3.48)
2 ∂x
3.5. LAGRANGIAN FOR A NON-RELATIVISTIC STRING 63
The work done by stretching a string an infinitesimal amount is given by To ∆` (Zwiebach, 81).
Thus, the potential energy is given by,
Z a 2
1 ∂y
V = To dx (3.49)
0 2 ∂x
Since a Lagrangian is defined as T − V , the Lagrangian for the string is given by
Z a" 2 2 #
1 ∂y 1 ∂y
L(t) = µ − To dx (3.50)
0 2 ∂t 2 ∂x
Note that the Lagrangian density is given by

2 2
∂y ∂y 1 ∂y 1 ∂y
L=L , = µ − To
∂t ∂x 2 ∂t 2 ∂x
The action for the string is then
Z tf Z tf Z a " 2 2 # Z tf Z a
1 ∂y 1 ∂y ∂y ∂y
S= L(t)dt = dt µ − To dx = dt dxL , (3.51)
ti ti 0 2 ∂t 2 ∂x ti 0 ∂t ∂x
Let us define the following quantities

∂L x ∂L
Pt ≡ ,P ≡
∂ ẏ ∂y 0
∂y ∂y
where ẏ = ∂t and y 0 = ∂x . A brief calculation shows that for our system
∂y x ∂y
Pt = µ , P = −To
∂t ∂x
To find the equations of motion, we vary the action:
Z tf Z a Z tf Z a
∂L ∂L
δ ẏ + 0 δy 0 = dx P t δ ẏ + P x δy 0

δS = dt dx dt (3.52)
ti 0 ∂ ẏ ∂y ti 0

Then, since δ ∂y∂t
∂
= ∂t (δy) and δ ∂x ∂y ∂
= ∂x (δy), we may write the variation as
Z tf Z a Z tf Z a
∂ ∂
δS = Pt (δy)dtdx + Px (δy)dtdx
ti 0 ∂t ti 0 ∂x
Integrating by parts yields,

Z a tf tf x=a tf a
∂P t Px
Z Z Z
t x

P δy dx +
P δy
dt − dt dx + δy
0 ti ti x=0 ti 0 ∂t ∂x
Specifying intial and final configurations of our system, we can set the first term of the above
expression equal to zero by allowing the variation in y vanish at these times. The second term
vanishes as long as we assume Neumann boundary conditions. The reason we choose Neumann
boundary conditions here is that momentum is conserved this way. On the other hand, momentum
is generally not conserved with Dirichlet boundary conditions [72]. This leaves us with,
tf a
∂P t ∂P x
Z Z
δS = − dt dx + δy
ti 0 ∂t ∂x
To find the equations of motion, we require that the variation δS = 0. Thus,

tf a
∂P t ∂P x
Z Z
dt dx + δy = 0
ti 0 ∂t ∂x
The above must hold for any variation on y, thereby indicating that the expression in the inte-
grand must be identically zero,
∂P t ∂P x
+ =0 (3.53)
∂t ∂x
Plugging in the explicit expressions for P t and P x , we find that the equations of motion for our
string is the wave equation
∂2y ∂2y
To 2 = µ 2 (3.54)
∂x ∂t
just as we had found earlier.
Before we move on to the system of a relativistic point-particle, let us briefly summarize our
review of classical strings. We began by considering a model of a stretched string which we then
slightly perturbed. Using Newton’s basic laws, we were able to determine the equations of motion
of the system. That is, we determined the expression which allows us to predict the behavior of the
string. From here we decided to use Lagrangian mechanics and varied the action to determine same
equations of motion. The reason for this is because, in some sense, the Lagrangian formulation
of classical mechanics is more fundamental then using Newton’s second. One reason is because
from the Lagrangian we can almost immediately determine the Hamiltonian (i.e. the energy of the
system). This is the method used in most advanced physics courses. In classical mechanics, to
gain insight about the behavior of an unknown system, we first attempt to determine the equations
of motion. As learned from classical physics, the best way of achieving this goal is through the
Lagrangian formulation. Adapting this trend to other areas of physics (e.g. quantum field theory,
general relativity), we can determine the expressions and necessary constraints to yield information
about the system in question. This trend is used so often, and will be used in this text, because it
relates back to our understanding of classical physics. And once we arrive to quantization, having
a classical analog is most beneficial.
As a final note, in deriving the equations of motion via the Lagrangian formulation, we found
that we had to employ Neumann boundary conditions. At the time we went this rather quickly,
however our reason for choosing these boundary conditions was that Neumann boundary conditions
conserve momentum. Alternatively, Dirichlet boundary conditions, where the endpoints of the
string are fixed to a ”wall” , do not conserve momentum in general. As we will see later, this has
imperative consequences for open ended strings in string theory, as it means the objects that open
strings are attached to (D-Branes) must carry away the momentum/energy of the string in some
way, thereby conserving the momentum of the entire system. But let’s not get too far ahead.
3.6. THE RELATIVISTIC POINT-PARTICLE IN MINKOWSKI SPACE 65
3.6 The Relativistic Point-Particle in Minkowski Space

Instead of jumping straight to relativistic strings, a background in the relativistic point-particle is
necessary. We consider a point particle with non-vanishing mass m, that moves freely in flat space-
time (later on we will consider a particle that moves in a curved space-time). Since the classical
motion of a point particle is to move along geodesics, the action S should be proportional to the
invariant length ds of the particle’s trajectory (Becker and Schwarz, 18). That is,
Z
SRel = −α ds (3.55)
The minus sign has been introduced so that ds is real for a time-like trajectory. We can determine
what this proportionality constant is by comparing the above action to that of a free non-relativistic
point-particle, Z
1
Snr = mv(t)2 dt
2
If we assume only one spatial dimension, then the relativistic action can be written as,
s 2
Z Z p Z Z p
2 2
dx
SRel = −α ds = −α dt − dx = −α 1− dt = −α 1 − v 2 dt
dt
Expanding we find, Z
1
SRel ≈ −α 1 − v 2 dt
2
By comparing the second term of the expansion with the action for the non-relativistic particle,
we notice that for the expressions to agree in the classical limit, we require α = m, resulting in
Z
SRel = −m ds (3.56)
Now that we have the action, we can determine the equations of motion. If we let the particle’s
trajectory xµ (often referred to as its world-line) be paramterized by some parameter τ , then we
have a parameterization of the world-line given by xµ (τ ). As shown in Appendix B, the line element
ds2 , in a flat space-time can be written as
ds2 = −ηµν dxµ dxν
Since we are parameterizing the path with τ , we can parameterize the line element as well, allowing
us to write,
dxµ dxν 2
ds2 = −ηµν dτ (3.57)
dτ dτ
Using the above line element, the action for relativistic point particle takes the form
Z r
dxµ dxν
S = −m −ηµν dτ (3.58)
dτ dτ
Often this is rewritten in a neater form (Becker, Schwarz, 19)
Z r
dxµ dxν
Z q Z p
S = −m −ηµν dτ = −m dτ −ηµν x˙µ x˙ν = −m dτ −ẋ2 (3.59)
dτ dτ
The Lagrangian is then q p

L = −m −ηµν x˙µ x˙ν = −m −ẋ2 (3.60)
It should be noted that the above action is reparameterization invariant. That is, suppose an
observer decides to measure the aciton via a parameter τ 0 rather than τ . Changing this parameter,
we have by the chain rule that
dxµ dxµ dτ 0
=
dτ dτ 0 dτ
If we substitute this reparameterization into the action, we find
Z r Z r
dxµ dxν dτ 0 dxµ dxν 0
S = −m −ηµν 0 dτ = −m −η µν dτ (3.61)
dτ dτ 0 dτ dτ 0 dτ 0
Notice that the above action assumes the same form as before. This thereby verifies that the
action is reparameterization invariant. On a similar note, the action is also Lorentz invariant,
however we leave it to the reader to fill in these details explicitly.
As usual, to find the equations of motion, we vary the action. That is,
Z
δS = −m δ(ds)
To determine how the interval ds is varied, we instead choose to vary ds2 . First notice
2 2 2
d(s + δs)2 = (d(s + δs)) = (ds + d(δs)) = ds2 + (d(δs)) + 2dsd(δs) = ds2 + 2dsd(δs) = ds2 + δds2
where we used the approximation that terms of higher order than δ are approximately zero. This
leaves us with,
δds2 = 2dsd(δs) (3.62)
. We must also vary the right hand side of equation (3.57). Since we are considering a flat space-time,
the Minkowski metric ηµν doesn’t change under variation of xµ . Rather, applying this variation,
we find,
d d
−ηµν (xµ + δxµ ) (xν + δxν ) dτ 2
dτ dτ
µ ν
dx d µ dx d ν
= −ηµν + (δx ) + (δx ) dτ 2
dτ dτ dτ dτ
µ ν
dxµ d dxν d

dx dx d d
= −ηµν + (δxµ ) (δxν ) + (δxν ) + (δxµ ) dτ 2
dτ dτ dτ dτ dτ dτ dτ dτ
µ ν
dxµ d dxν d

dx dx ν µ
= −ηµν + (δx ) + (δx ) dτ 2
dτ dτ dτ dτ dτ dτ
where we again made use of the fact that δ 1. To write this expression more compactly,
consider the last two terms written out explicitly,
dxµ d dxν d

ν µ
− ηµν (δx ) + ηµν (δx )
dτ dτ dτ dτ
3.6. THE RELATIVISTIC POINT-PARTICLE IN MINKOWSKI SPACE 67
Focusing on the first term, notice that we are summing over two dummy indices, thus we may
write
dxµ d dxν d
ηµν (δxν ) = ηνµ (δxµ )
dτ dτ dτ dτ
Using the fact that the metric is symmetric, ηµν = ηνµ , we find
dxµ d dxν d dxν d

ν µ µ
− ηµν (δx ) + ηµν (δx ) = − 2ηµν (δx )
dτ dτ dτ dτ dτ dτ
Putting everything together, we immediately see that the variation of the right hand side becomes
dxν d
−2ηµν (δxµ )dτ 2 (3.63)
dτ dτ
Altogether then,
dxν d
2dsd(δs) = −2ηµν (δxµ )dτ 2 (3.64)
dτ dτ
Simplifying yields,
dxν d dxµ d
δ(ds) = −ηµν (δxµ )dτ = − (δxµ )dτ (3.65)
ds dτ ds dτ
Finally, the variation of the action is
Z
dxµ d
δS = m (δxµ )dτ (3.66)
ds dτ
Recognizing that the 4-momentum
dxµ
m = pµ
ds
is in our expression, we rewrite variation of the action as
Z
d d
δS = (δxµ ) (δxµ )dτ (3.67)
dτ dτ
Introducing explicit limits of integration and performing integration by parts, we find that the
above becomes τf Z τf
µ dpµ
dτ δxµ (τ )

δS = pµ δx − (3.68)
τi τi dτ
Since we may fix the coordinates at the boundary, the first term vanishes. Moreover, the for the
variation of the action to be zero, the second term must vanish for any variation δxµ (τ ), leaving
the equation of motion for the relativistic point particle,
dpµ
=0 (3.69)
dτ
The equation of motion indicates that the momentum of the point particle moving along its
world-line is constant. Moreover, since we worked with a Lorentz invariant and reparameterization
invariant action, the equation of motion is also Lorentz invariant and reparameterization invariant.
Before moving on to investigating the relativistic point particle in a curved space-time, its impor-
tant to first note that the action above is not the most efficient it could be. Notice that the action
does not hold for massless particles. Another issue is that the square root introduces complications
when it comes time to quantization. To avoid these problems, we introduce and auxiliary field,
denoted by a(τ ), and consider the Lagrangian (Becker, Schwarz, 20)
1 2 m2
L= ẋ − a (3.70)
2a 2
The action then becomes
Z
1 1 2
S= dτ ẋ − m2 a (3.71)
2 a
as an exercise, the reader will work out that the equation of motion for the auxiliary field is
given by
ẋ2 + m2 a2 = 0 (3.72)
which yields the expression
r
ẋ2
a= − (3.73)
m2
The equation of motion above is often referred to as the equation for the mass shell. Moreover,
a brief calculation shows that the action with the auxiliary field may be recast in the form of our
original action without the auxiliary field (McMahon, 27).
3.7 The Relativistic Point Particle in a Curved Space-time

Since general relativity assumes that mass and energy distort space-time, we desire to know how a
particle moves through a warped space-time. A similar procedure as performed above leads to the
equation of motion of a relativistic point particle in a curved space-time. Consider a point particle
of mass m traveling through a curved space with that has some associated metric gµν (x). The
motion of the particle in this space is investigated using the same action as before (Zwiebach, 99).
Z
S = −m ds (3.74)
This time however, the invariant interval is given in terms of the metric of the curved space
ds2 = −gµν (x)dxµ dxν (3.75)
Analogous to the point particle in flat space-time, we may then parameterize our space-time
coordinate with a parameter τ and write
dxµ dxν 2
ds2 = −gµν (x(τ )) dτ (3.76)
dτ dτ
The action becomes Z r
dxµ dxν
S = −m −gµν (x(τ )) dτ (3.77)
dτ dτ
3.7. THE RELATIVISTIC POINT PARTICLE IN A CURVED SPACE-TIME 69
We find the equation of motion by varying the action in the same way as before,
Z
δS = −m δ(ds)
In the flat space-time case we found that

dxν d
2dsd(δs) = −2ηµν (δxµ )dτ 2 (3.78)
dτ dτ
This time however the metric gµν changes with the variation. Let’s first focus on the right hand
side of (3.76). Varying it we find

d d
−gµν (x + δx) (xµ + δxµ ) (xν + δxν ) dτ 2
dτ dτ
dxµ dxν dxµ d

ν
= − gµν (x + δx) + 2gµν (x + δx) (δx ) dτ 2
dτ dτ dτ dτ
Therefore we have
dxµ dxν dxµ d

2 2 ν
d(s + δs) = ds + 2dsd(δs) = − gµν (x + δx) + 2gµν (x + δx) (δx ) dτ 2
dτ dτ dτ dτ
µ ν
Then, subtracting ds2 = −gµν (x(τ )) dx dx 2
dτ dτ dτ from both sides yields
dxµ dxν dxµ d

ν
= − (gµν (x + δx) − gµν (x)) + 2gµν (x + δx) (δx ) dτ 2
dτ dτ dτ dτ
Let’s multiply by 1 in a suggestive way,
(gµν (x + δx) − gµν (x)) ρ dxµ dxν dxµ d

ν
− δx + 2gµν (x + δx) (δx ) dτ 2
δxρ dτ dτ dτ dτ
If we take the limit where δ gets very small, then the above is approximated by
∂gµν ρ dxµ dxν dxµ d

ν
− δx + 2gµν (x) (δx ) dτ 2
∂xρ dτ dτ dτ dτ
Therefore, altogether we have,

∂gµν ρ dxµ dxν 2 dxµ d
2dsd(δs) = − ρ
δx dτ − 2gµν (x) (δxν ) dτ 2 (3.79)
∂x dτ dτ dτ dτ
Simplifying yields
1 ∂gµν ρ dxµ dxν dxµ d
δ(ds) = − δx dτ − gµν (δxν ) dτ (3.80)
2 ∂xρ ds dτ ds dτ
This gives us
1 ∂gµν ρ dxµ dxν dxµ d
Z
δS = m δx dτ + gµν (δxν ) dτ (3.81)
2 ∂xρ ds dτ ds dτ
Since ν is a dummy variable in the second term of the integral, we simply make the exchange
ν → ρ. Then using integration by parts on the second term, we find
τ
1 ∂gµν dxµ dxν dxµ dxµ ρ f
Z
d ρ
δS = m − gµρ δx dτ − gµν δx
2 ∂xρ ds dτ dτ ds ds
τi
Since we may choose that the variation vanishes at the boundary, we are left with
1 ∂gµν dxµ dxν dxµ
Z
d
δS = m − gµρ δxρ dτ
2 ∂xρ ds dτ dτ ds
If we let the world-line be parameterized by s, we find that,
Z
d
δS = m − gµρ δxρ ds (3.82)
2 ∂xρ ds ds ds ds
δS must vanish for an arbitrary variation, hence the equation of motion of a point particle in
curved space-time is given by

d
− gµρ =0 (3.83)
2 ∂xρ ds ds ds ds
Or more compactly,
1 ∂gµν µ ν d
x˙ x˙ − (gµρ x˙µ ) = 0 (3.84)
2 ∂xρ ds
This is known as the geodesic equation, and we will see it again during our discussion on general
relativity. Note that if we again consider flat space-time, then the metric gµν is constant, and we
recover the equation of motion for a point particle in Minkowski space,
ẍµ = 0 (3.85)
3.8 Exercises
1. Determine the Lagrangian for the system of a simple harmonic oscillator. From the Lagrangian,
find the equations of motion.
2. It can be shown that variation of the action δS for a single coordinate (3.9) may be recast
using the functional derivative
Z
δS
δS = δq(t)dt
δq(t)
from which we identify the functional derivative as
δS ∂L(t, q, q̇) d ∂L(t, q, q̇)

δq(t) = −
δq(t) ∂q dt ∂ q̇
(a) This means we have an alternate method in finding the equations of motion for a system. Cal-
culate the functional derivative for the system of the one-dimensional harmonic oscillator. Compare
to your answer in problem 1.
3.8. EXERCISES 71
(b) To calculate the functional derivative, somtimes one has to convert the functionals to an
integral form. If the original functional does not contain any integration, the Dirac delta function
can be used to get the functional in the correct form. Keeping this in mind, calculate the functional
derivative of the following functional
p
S[q] = 3 q(1) + cos(q(2))
3. Using (3.21), show that the following two Maxwell equations may be summarized by (3.23):
~
∇ ~ = 1ρ
~ ·E ~ = 1~j + 1 ∂ E
~ ×B
∇
c c c ∂t
4. Using the gauge transformation given in (3.32), show that the Maxwell tensor is invariant
under the this particular gauge transformation.
5. From the action given in (3.71), work out the equation of motion for the auxiliary field, as
shown in (3.72).
Chapter 4
Classical Relativistic Strings
With a fair background in the classical mechanics of non-relativistic strings and relativistic point-
particles, we are ready to explore the behavior of relativistic strings, our first big step in under-
standing the fundamentals of string theory. In the next chapter we will apply light-cone coordinates
to the analysis of the relativistic string, demanding that we first become familiar with the classical
mechanics of relativistic strings.
4.1 Preparation for the Action of a Relativistic String

The action for a relativistic string must be a functional of the string trajectory. As a particle traces
out a world-line as it moves through space-time, a string would trace out a surface, the world-sheet.
An open string would trace out a flat sheet while a closed string would trace out a closed tube-like
surface, as depicted in figure 4.1.
In the last chapter we saw that the action for a relativistic particle is proportional to the proper
distance of the world-line. For strings, we define a proper area of the world-sheet, and require that
the string action be proportional to this proper area. This string action is known as the Nambu-
Goto action. Before we examine the proper area of a world-sheet, let us first develop the area
functional of a spatial surface.
A curve in ordinary space can be parameterized using a single parameter, e.g. x = x(τ ). A
surface in space is two-dimensional, thereby requiring two parameters, which we will call ξ 1 and ξ 2 .
Given a surface, we may parameterize using our parameters, embedding it with a grid of constant
lines of ξ 1 and ξ 2 . The parameterized surface can then be described by the collection of functions
(Zwiebach, 101)
~x(ξ 1 , ξ 2 ) = x1 (ξ 1 , ξ 2 ), x2 (ξ 1 , ξ 2 ), x3 (ξ 1 , ξ 2 )

(4.1)
If we zoom in on our surface and analyze it locally, we can view our parameters ξ 1 and ξ 2 as
coordinates on the surface which will help us develop the area functional for any arbitrary spatial
surface.
73
74 CHAPTER 4. CLASSICAL RELATIVISTIC STRINGS
Figure 4.1: Worldsheet of an open and a closed string.
Remember that the action for a point particle is proportional to a differential distance (dx for
example). For a spatial surface, we have two dimensions to work with, therefore we expect that the
action for an object moving along a spatial surface will be proportional to a differential area dA,
which, in terms of our parameters, is often written as dA = dξ 1 dξ 2 . As written, this differential area
is limited as it only describes an infinitesimal rectangle. In general however, dA is a parallelogram.
Let’s call the sides of the parallelogram as d~v1 and d~v2 , and write them as
∂~x 1 ∂~x
d~v1 = dξ d~v2 = 2 dξ 2 (4.2)
∂ξ 2 ∂ξ
The differential area dA is simply the cross product of the two sides of the parallelogram
dA = d~v1 × d~v2
Making use of the expression
q
2
|x × y| = (x · x) (y · y) − (x · y)
we find that q
2
dA = (d~v1 · d~v1 ) (d~v2 · d~v2 ) − (d~v1 · d~v2 )
Substituting in the expressions for the sides d~v1 and d~v2 , the differential area dA becomes
s 2
1 2 ∂~x ∂~x ∂~x ∂~x ∂~x ∂~x
dA = dξ dξ · · − · (4.3)
∂ξ 2 ∂ξ 2 ∂ξ 2 ∂ξ 2 ∂ξ 2 ∂ξ 2
Now consider the line element
ds2 = d~x · d~x
4.1. PREPARATION FOR THE ACTION OF A RELATIVISTIC STRING 75
where ~x = ~x(ξ 1 , ξ 2 ). The total differential of ~x is
∂~x i
d~x = dξ
∂ξ i
where an implicit sum over i is assumed. Using this fact, we may write the line element as
∂~x ∂~x i j
ds2 = · dξ dξ ≡ gij (ξ)dξ i dξ j (4.4)
∂ξ i ∂ξ j
where we have defined the induced metric

∂~x ∂~x
gij (ξ) = · (4.5)
∂ξ i ∂ξ j
Written in matrix form,  

∂1 ~x · ∂1 ~x ∂1 ~x · ∂2 ~x
[gij ] =   (4.6)
∂2 ~x · ∂1 ~x ∂2 ~x · ∂2 ~x
where ∂i = ∂/∂ξ i . As written in (4.6), notice that the determinant shows up the differential area
dA. Denoting g = det(gij ), we may write
√
Z
A = dξ 1 dξ 2 g (4.7)
Since the induced metric is reparameterization invariant, it can be shown that the area A above is
manifestly reparameterization invariant.
We can use the above derivation now as a guide for computing the action of relativistic strings,
since the world-sheets traced out by strings are really just surfaces in space-time. Just as before,
since a world-sheet is a two-dimensional surface, we require two parameters. Instead of using ξ 1
and ξ 2 , it is convention that we call our parameters τ and σ. Given the usual D dimensional
space-time coordinates xµ = (x0 , x1 , ..., xd ), a surface in space-time is described by the so-called
mapping functions X µ (τ, σ). Given a point (τ, σ) in parameter space, we map it to a point with
space-time coordinates
X 0 (τ, σ), X 1 (τ, σ), ..., X d (τ, σ)

We call X µ string coordinates. Similar to our parameters for spatial surfaces, (τ, σ) can be
interpreted as coordinates on the world-sheet as long as we examine the world-sheet locally.
It turns out, as we will see later on, that τ is related to time (hence the choice of notation), and
σ is related to the position on the string. The world-lines of the string endpoints have constant
σ, and are therefore parameterized by τ (Zwiebach, 107). As τ flows, time also flows, yielding, at
least at the endpoints,
∂X 0

6= 0 (4.8)
∂τ endpoint
We will make the assumption that the above relation holds for other values of σ.
To find the differential area element, we use the same recipe as before, however using index
notation displaying the fact that it is the relativistic analog. Instead of a parallelogram, let us
denote the vectors dv1µ and dv2µ spanning the quadrilateral (instead of a rectangle) by
∂X µ ∂X µ
dv1µ = dτ dv2µ = dσ (4.9)
∂τ ∂σ
Following the same idea as before, we would assume that
q
2
dA = (dv1 · dv1 ) (dv2 · dv2 ) − (dv1 · dv2 )
This turns out to be incorrect because the expression under the square root is negative since
we have assumed that τ is time-like and σ is space-like (Hatfield, 491). To avoid an imaginary
differential area, we introduce a minus sign yielding the proper area
s 2
∂X µ ∂Xµ ∂X µ ∂Xµ ∂X µ ∂Xµ
Z
A = dτ dσ − (4.10)
∂τ ∂σ ∂σ ∂σ ∂τ ∂τ
4.2 The Nambu-Goto Action

Now that we have a well-defined notion of proper area, we may introduce an action which is
proportional to the proper area,
s 2
∂X µ ∂Xµ ∂X µ ∂Xµ ∂X µ ∂Xµ
Z
S = −T dτ dσ −
∂τ ∂σ ∂σ ∂σ ∂τ ∂τ
The constant T is in fact the tension of the string, as we will see briefly. Making the following
definitions
∂X µ ∂X µ
Ẋ µ ≡ X 0µ ≡ (4.11)
∂τ ∂σ
we may write the above in a slightly simpler form,
Z q
2 2 2
S = −T dτ dσ Ẋ · X 0 − Ẋ (X 0 ) (4.12)
Or if we were explicitly putting in c, we had have

Z
T
q
2 2 2
S=− dτ dσ Ẋ · X 0 − Ẋ (X 0 ) (4.13)
c
As an exercise, the reader verify that this action is also reparameterization invariant. As usual,
we may write the line element as
∂X ν ∂X ν α β
−ds2 = dX µ dXµ = ηµν dX µ dX ν = ηµν dξ dξ
∂ξ α ∂ξ β
Where indices α and β can run from 1 to 2, and ξ 1 = τ and ξ 2 = σ. Let us further define the
induced metric similar to before,
∂X ν ∂X ν
γαβ = ηµν α (4.14)
∂ξ ∂ξ β
4.2. THE NAMBU-GOTO ACTION 77
Or, in matrix notation,  

2 0
Ẋ Ẋ · X
[γαβ ] =  0
0 2  (4.15)
Ẋ · X X
Denoting γ = det (γαβ ), we may write the action as

√
Z
S = −T dτ dσ −γ (4.16)
The above action is known as the Nambu-Goto string action, and it will allow us to find the
equations of motion of a classical relativistic string. It should be noted that, written in the form
above, we can readily generalize the Nambu-Goto action to describe the dynamics of objects which
are higher dimensional, for instance, D-Branes, which are the higher dimensional analog of strings.
Unfortunately, this action is somewhat a pain to deal with when we quantize the string. For this
reason, later on, when we quantize the string using the covariant quantization procedure, we will
introduce another action, the Polyakov action to obtain a more elegant method for quantization.
Now that we have an action for a relativistic string, we can determine the equations of motion.
Let us rewrite the Nambu-Goto action in terms of the Lagrangian density L
Z τf Z σ1
dσL Ẋ µ , X 0µ

S= dτ (4.17)
τi 0
where the Lagrangian density is

q
µ 0µ
2 2 2
Ẋ · X 0 (X 0 )

L Ẋ , X = −T − Ẋ (4.18)
We vary the action to find the equations of motion, which is simply,

Z τf Z σ1
∂L ∂(δX µ ) ∂L ∂(δX µ )

δS = dτ dσ + (4.19)
τi 0 ∂ Ẋ µ ∂τ ∂X 0µ ∂σ
Computing the derivative ∂L/∂ Ẋ µ we find
q q
∂L ∂ 0
2 2 0 2 ∂ 2 2 0 2
µ
= −T Ẋ · X − Ẋ (X ) = −T Ẋ µ Xµ0 − Ẋµ Ẋ µ (X )
∂ Ẋ ∂ Ẋ µ ∂ Ẋ µ
2
Ẋµ X 0µ Xµ0 − (X 0 ) Ẋµ

= −T q 2 2 2
Ẋ µ Xµ0 − Ẋµ Ẋ µ (X 0 )
2
Ẋ · X 0 Xµ0 − (X 0 ) Ẋµ

= −T q 2 2 2
Ẋ µ Xµ0 − Ẋµ Ẋ µ (X 0 )
Similarly,
2
Ẋ · X 0 Ẋµ − Ẋ Xµ0

∂L
= −T q
∂X 0µ 2 2 2
Ẋ µ Xµ0 − Ẋµ Ẋ µ (X 0 )
Since these terms show up often enough, we define

2
Ẋ · X 0 Xµ0 − (X 0 ) Ẋµ

∂L
Pµτ ≡ = −T q (4.20)
∂ Ẋ µ 2 2
Ẋ µ Xµ0 − Ẋµ Ẋ µ (X 0 )
2
2
Ẋ · X 0 Ẋµ − Ẋ Xµ0

∂L
Pµσ ≡ = −T q (4.21)
∂X 0µ 2 2 2
Ẋ µ Xµ0 − Ẋµ Ẋ µ (X 0 )
Using this notation, the variation of the action becomes
Z τf Z σ1
∂(δX µ ) ∂(δX µ )

δS = dτ dσ Pµτ + Pµσ
τi 0 ∂τ ∂σ
τ
∂Pµσ
Z τf Z σ1
∂Pµ

∂ µ τ
∂ µ σ
µ
= dτ dσ δX Pµ + δX Pµ − δX +
τi 0 ∂τ ∂σ ∂τ ∂σ
Carrying out the integrals on the first two terms yields
τ f σ1 ∂Pµτ ∂Pµσ
Z σ1 Z τf Z τf Z σ1
µ τ µ σ µ
δS = dσ δX Pµ + dτ δX Pµ − dτ dσδX +
0 τi τi 0 τi 0 ∂τ ∂σ
Since we are assuming that τ flows with time, we may specify that δX µ (τi , σ) = δX µ (τf , σ) = 0,
causing the first term on the right hand side to vanish. The second term vanishes via two types of
boundary conditions: Dirichlet boundary conditions, or the free endpoint condition. Remember that
Dirichlet boundary conditions mean we fix the endpoint of the string, meaning that the variation
δX µ vanishes at the fixed endpoints. Alternatively it means
∂X µ

=0 (4.22)
∂τ endpoints
whenever µ 6= 0.
On the other hand, the free endpoint condition is that

Pµσ

=0 (4.23)
endpoints
Specifically,
P0σ (τ, 0) = P0σ (τ, σ1 ) = 0
We will see later on that the free endpoint conditions are actually Neumann boundary conditions
(Zwiebach,115). Using the Dirichlet boundary conditions and free endpoint condition, we are left
with τ
∂Pµσ
Z τf Z σ1
∂Pµ

µ
δS = − dτ dσδX +
τi 0 ∂τ ∂σ
The variation of the action must be zero for every variation, yielding the equations of motion
∂Pµτ ∂Pµσ
+ =0 (4.24)
∂τ ∂σ
It is important to note that we never assumed that the string was open or closed. Therefore, the
above expression is the equation of motion for a relativistic string, open or closed.
4.3. PARAMETERIZING THE STRING 79
4.3 Parameterizing the String
Since the string action is manifestly reparameterization invariant, we may parameterize the world-
sheet in a useful way. Doing so will allow us to pull out physical information about the action, and
the equations of motion. We first consider a partial parameterization, where we fix lines of constant
τ by relating τ to X 0 = ct. Let us do this by declaring that for any point P on the world-sheet we
have
τ (P ) = t(P ) (4.25)
A diagram of this parameterization is given by figure 4.2. This parameterization is called the
static gauge since the lines of constant τ represent static strings. With the static gauge, we imme-
diately notice that
X 0 (τ, σ) ≡ ct(τ, σ) = cτ
where we have explicitly put in the factors of c. The above therefore implies that
τ =t (4.26)
We may also describe the set of coordinates X µ (τ, σ) as

~ σ)
X µ (τ, σ) = X µ (t, σ) = ct, X(t,
~ are the spatial string coordinates. Using this notation, we find that
where X
! !
∂X µ ~
∂X 0 ∂ X ~
∂X
= , = 0, (4.27)
∂σ ∂σ ∂σ ∂σ
And
! !
∂X µ ~
∂X 0 ∂ X ~
∂X
= , = c, (4.28)
∂τ ∂t ∂t ∂t
From here, its easy to see that the static gauge parameterization separates time and space
components in a simple fashion.
Figure 4.2: The static gauge. At time t = t0 the string world-sheet intersects with the hyperplane
at t = t0 . In the static gauge this would be equivalent to the τ = t0 segment AB. (Figure motivated
by Zwiebach)
Using this parameterization, we are now in a position in which we may relate T in the Nambu-
Goto action to the tension of the string. Consider the endpoints of a string to be fixed at X 1 = 0
and X 1 = a but zero at all other values of X µ , implying that the string is only stretched along one
spatial coordinate. A schematic is given in figure 4.3. Using the static gauge, we may write
X 1 (t, σ) = f (σ), X 2 = X 3 = ... = X d = 0
Moreover, we may also assume that f (0) = 0, f (σ1 ) = a and f (σ) is montone increasing. Then,
setting c = 1, we find that
Ẋ µ = (1, 0, ~0), X 0µ = (0, f 0 , ~0)
Therefore,
2 2 2
Ẋ = −1, (X 0 ) = (f 0 ) , Ẋ · X 0 = 0
4.3. PARAMETERIZING THE STRING 81
Figure 4.3: A string that is parameterized by X 1 (t, σ) = f (σ), with all other string coordinates set
to zero.Function f is also assumed to be monotone increasing.
Substituting this into the Nambu-Goto action, we find that

Z tf Z σ1 Z tf Z σ1
p df
S = −T dt dσ −(−1)(f 0 )2 = −T dt dσ
ti 0 ti 0 dσ
Z tf Z tf
= −T dt (f (σ1 ) − f (0)) = dt (−T a)
ti ti
But the Lagrangian for a static object is simply L = −V , from which the action can cast as
Z tf Z tf
S= dtL = dt(−V )
ti ti
Comparing to the action of a static string, we find that
V = Ta (4.29)
V is simply the potential energy of a string stretched a length a. Verily, we conclude that the T
is precisely the tension of the string. It is important to note that we assumed the static string
satisfied the equation of motion derived earlier. A quick calculation, left for the reader, verifies this
assumption.
Now that we have seen the benefit of parameterizing τ , let us choose a σ parameterization. For
open strings we choose an edge of the world-sheet to be σ = 0 and another at σ = σ1 . That is, for
an open string, σ is parameterized by
σ ∈ [0, σ1 ]
For closed strings, we must make an identification such that the σ direction is made into a circle,
making (τ, σ) into a cylinder. Let σc denote the circumference of the σ circle. Then,
(τ, σ) ∼ (τ, σ + σc )
Using this identification, closed strings are parameterized by
σ ∈ [0, σc ]
Without presenting the details, a σ parameterization of a given string can be used to construct
lines of constant σ which are always orthogonal to the lines of constant t (for a more rigorous
description of this parameterization, visit Zwiebach, 130-131). With this paramaterization of the
~ ~
string surface, the tangent to the strings, ∂∂σ
X
, and the tangent to the lines of constant σ, ∂∂t
X
, are
orthogonal. That is,
∂X~ ∂X ~
· =0 (4.30)
∂σ ∂t
~
∂X
Moreover, since the velocity ∂t is perpendicular to the string, the transverse velocity of the string,
~v⊥ , satisfies
∂X~
~v⊥ =
∂t
at all points. This discovery is important since we may define the string action in terms of the
transverse velocity.
4.4 Transverse Velocity and the String Action

~
It would be nice to define a type of string velocity. Initially, we might assume that ∂∂t X
is the
best choice for a string velocity, but we would be wrong. This velocity depends on the coordinates
X µ (τ, σ), which is variable under a non-unique choice of a σ parameterization. This would imply
that the string velocity, if we use our naive assumption, is non-unique and therefore physically
meaningless. It turns out the reparameterization invariant string velocity we seek is the transverse
velocity, ~v⊥ . This velocity is a vector tangent to the string and to the string spatial surface (a
surface that is composed of the strings we observe at all times)(Zwiebach, 121).
To define the transverse velocity, we require a unit vector tangent to the string. Let us define a
parameter s which measures the length along the string. Further, declare that s(σ) to be the length
of the string on the interval [0, σ]. The infinitesimal lenght ds is then
~

∂X
~
ds = |dX| = |dσ|
∂σ
Note that
~ ∂X~ ~ ∂X~ dσ 2 ∂ X
~ 2 dσ 2

∂X ∂X
· = · = =1
∂s ∂s ∂σ ∂σ ds ∂σ ds
~
∂X
Hence ∂s is a unit vector. Moreover,
~
∂X ~ dσ
∂X
=
∂s ∂σ ds
4.4. TRANSVERSE VELOCITY AND THE STRING ACTION 83
Figure 4.4: A small piece of the world-sheet showing the transverse velocity ~v⊥ .
∂X~
Therefore, ∂s is a unit tangent vector to the string. We then Go on and define ~v⊥ to be the
∂X~
component of ∂s perpendicular to the string (figure 4.4). As an exercise, the reader will show that
using our parameterizations of σ and τ , the Nambu-Goto action may recast using the transverse
velocity, taking the form:
r
~v 2
Z Z
ds
S = −T dt dσ 1− ⊥ (4.31)
dσ c2
where we have again explicitly put in constants for c. The associated Lagrangian is then,
r
~v 2
Z
L = −T ds 1 − ⊥ (4.32)
c2
In this form we may recognize this Lagrangian as the generalization of the relativistic particle
Lagrangian.
We can also simplify the expressions for Pµτ and Pµσ by using P τ µ and P σµ and our choices of
parameterizations. In terms of ~v⊥ , we find
2
~ ~ ~
∂X
∂σ · ∂t
∂X
Ẋ − −c + ∂∂t
µ 2 X
X 0µ
σµ T
P =− q
c ds v2
~
c dσ 1 − c⊥2
2 µ
~
∂X ~
∂X µ 2 ~
∂X ∂X
· Ẋ − −c +
T ∂s ∂t ∂t ∂s
=− 2 q
c v2
~
1 − c⊥2
2 µ
~
c − ∂∂t
2 X ∂X r
T ∂s 2
~v⊥ ∂X µ
=− 2 = −T 1 −
c2 ∂s
q
c v2
~
1 − c⊥2
To get to the last line, we used the definition of the transverse velocity in our chosen σ parame-
terization. Similarly, we also have
µ
µ ~
∂X ∂X~ ∂X
T ds Ẋ − ∂σ · ∂t ∂s T ds ~v
τµ
P = 2 q = 2 q ⊥ 2
c dσ ~
v⊥2 c dσ ~
v
1 − c2 1 − c⊥2
With the way we have chosen our parameterizations and found the reparameterization invariant
string velocity, there is no ambiguity in defining the velocity of the string endpoints. In fact, as the
reader will show, by our choice of σ parameterization
~
∂X
~v⊥ =
∂t
at all points. Therefore ~v⊥ = ~v at all points along the string, including the endpoints. Moreover,
we just found that r
σµ ~v 2 ∂X µ
P = −T 1 − 2 (4.33)
c ∂s
Looking only at spatial coordinates, µ = 1, ..., d, and applying the boundary condition for open
strings, we find r
~v 2 ∂ X~
~ σ
P = −T 1 − 2 =0
c ∂s
But the only way this condition is satisfied is if
v 2 = c2
In other words, open string end points move at the speed of light!
Using the action in terms of transverse velocity, it is easy to calculate the energy of the string.
To do this we wish to find the Hamiltonian H for our string. We can do this by ascertaining the
Hamiltonian density H and integrating over spatial coordinates σ,
Z
H = Hdσ (4.34)
Recall from before that hamiltonian density is defined by
H = Π(x, t) · ẋ − L (4.35)
~ σ) ≡
where Π(x, t) is the canonical momentum. In our case, the cannonical momentum is simply P(t,
~ Earlier we saw the Lagrangian density L in our chosen parameterization of σ and τ
∂L/∂(∂t X).
was given by v
u !2
T ds u
tc2 − ∂ ~
X
L=− (4.36)
c dσ ∂t
4.4. TRANSVERSE VELOCITY AND THE STRING ACTION 85
Then,
∂L T ~
ds ∂ X T ~v⊥ ds
~ σ) ≡
P(t, =r
~
∂(∂t X) 2 dσ ∂t = c2 q ~
v⊥2 dσ
~ 1−
c2 − ∂∂t
X
c2
Hence, the Hamiltonian density H is simply,

r
~ 2 2
~v⊥
~ · ∂ X − L = T q ~v⊥
H=P
ds
−T 1− =
∂t c 2 ~
v⊥2 dσ c2
1− c2
T ds
=q
~
v⊥2 dσ
1− c2
The Hamiltonian is then Z Z

ds
H= Hdσ = Tq 2
(4.37)
~
v⊥
1− c2
Finally, just as we simplified our expressions for Pµτ and Pµσ , we can also gain a fair physical
interpretation of the equation of motion using our parameterizations. Using the static gauge, the
equation of motion becomes
∂P τ µ ∂P σµ
=− (4.38)
∂t ∂σ
To intepret these equations, we consider time and space coordinates of Pµτ and Pµσ separately. First
consider when µ = 0. Then, by the boundary conditions of an open string, P σ0 = 0. Moreover,
T 1 ds
Pτ0 = q =0
c ~
v⊥2 dσ
1− c2
To understand this physically, consider

q a small piece of the string associated with dσ. Multiplying
v2
~
the above by dσ we find that T ds/ 1 − c⊥2 is constant in time (Zwiebach, 133). But this is the
integrand of our expression for the Hamiltonian of our string,
Z
ds
H = Tq
v2
~
1 − c⊥2
Therefore,
∂H
=0 (4.39)
∂t
implying
q that the energy of the entire string is conserved. But it does more than that. Since,
~
v⊥2
T ds/ 1 − c2 is constant in time,the energy of each small piece of the string is conserved, and
therefore the energy of the entire string is actually constant.
Now let us consider the space components of the string equation of motion. We may then write
~ τ = T q ~v⊥
P
ds
(4.40)
c2 1 − v2 dσ
c2
and, r
~
v2 ∂ X
~ σ = −T
P 1− (4.41)
c2 ∂s
Substituting these expressions into (4.38) we find that
" r #  
∂ ~
v2 ∂ X ∂ T 1 ds  T ~v⊥ ds ∂~v⊥
T 1− 2 = 2
q = 2q (4.42)
∂σ c ∂s ∂t c 1 − vc2 dσ
2 c 1− v2 dσ ∂t
c2
To take the time derivative, we used the fact that qT ds 2 is constant in time. It is possible to
~
v
1− ⊥
c2
interpret the above equation loosely in terms of an ”effective” non-relativistic string. Considering
small oscillations and comparing to the equations of motion for a non-relativistic string, we find
the effective tension and effect mass density (Zwiebach, 134)
r
v2
Tef f = T 1 − 2 (4.43)
c
T 1
µef f = (4.44)
c2 1 −
q
v2
c2
At the endpoints, ~v⊥ = c, therefore Tef f → 0 at the endpoints. This makes sense since this is
the only way in which an open relativistic string can have tension yet free endpoints. Moreover,
it would seem we have a diverging issue with the effective mass density at the endpoints. This is
turns out not to be the case since we are describing a string with finite energy.
4.5 Motion of Open and Closed Strings

ds
If we move dσ from the right hand side of (4.42) to the left hand side, we find that
" r #
T ~v⊥ ∂~v⊥ ∂ ~
v2 ∂ X
= T 1− 2 (4.45)
c2 1 − v2 ∂t
q
∂s c ∂s
c2
~
∂X ~ dσ
∂X ~
∂X
Then using ∂s = ∂σ ds and ~v⊥ = ∂t , we may change the s-derivatives of (4.45) to σ-derivatives,
yielding q 2
q 2

v⊥ v⊥
1 ∂2X~ 1− c2 ∂  1− c2
~
∂X
2 2
= ds ds
 (4.46)
c ∂t dσ
∂σ dσ
∂σ
Notice if
ds 1
q =1 (4.47)
dσ 2
v⊥
1− c2
we attain the familiar wave equation
1 ∂2X~ ~
∂2X
2 2
= (4.48)
c ∂t ∂σ 2
4.5. MOTION OF OPEN AND CLOSED STRINGS 87
To acheive (4.47), we must parameterize σ in such a way that it gives the wave equation yet
maintains our previous σ parameterization, keeping constant lines of σ on the string spatial surface.
To find such a σ, assign σ = 0 at one endpoint of the string, and then move along the string, assigning
each infinitesimal piece ds of the string a dσ given by
ds
dσ = q 2
(4.49)
v⊥
1− c2
This parameterization gives us (4.47). Put another way, we may modify (4.47) to give
2
ds 1 2
+ 2 v⊥ =1 (4.50)
dσ c
which allows us to write !2 !2
~
∂X 1 ~
∂X
+ 2 =1 (4.51)
∂σ c ∂t
Using this method of parameterization, we can also reexamine the boundary conditions. Recall
that r
v2 ∂ X~
~
P = −T 1 − ⊥
σ
=0
c2 ∂s
Using condition (4.47), we find then
r ! !
2
v⊥

dσ ∂X~ ∂X~
~ σ
P = −T 1 − 2 = −T =0 (4.52)
c ds ∂σ ∂σ
which indicates that

~

∂X =0 (4.53)
∂σ endpoints
But this is simply the Neumann boundary condition! That is, the free endpoint boundary
condition, in the right σ parameterization is simply the Neumann boundary condition.
To find the motion of the relativistic string, we have four constraints: the wave equation, two
parameterization conditions, and the Neumann boundary condition. Written out explicitly, they
are
1 ∂2X~ ∂2X ~ ∂X~ ∂X ~
− = 0 · =0 (4.54)
c2 ∂t2 ∂σ 2 ∂t ∂σ
!2 !2
~ ~ ~

∂X 1 ∂X ∂X
+ 2 =1 =0 (4.55)
∂σ c ∂t ∂σ endpoints
With these constraints, we are now in a position to determine the general motion of open and
closed relativistic strings. Let us first consider an open string. The most general solution to the
wave equation is given by
~ σ) = 1 f~(ct + σ) + ~g (ct − σ)

X(t, (4.56)
2
Applying the boundary condition σ = 0 we find that
f~0 (ct) − ~g 0 (ct) = 0

Substituting this back into our general solution, we find that
~ σ) = 1 f~(ct + σ) + f~(ct − σ)

X(t, (4.57)
2
Now let us consider the same boundary condition, however at the endpoint σ = σ1 , yielding
f~0 (ct + σ1 ) − f~0 (ct − σ1 ) = 0
If we let u = ct − σ1 , we may write
df~ df~
(u + 2σ1 ) = (u) (4.58)
du du
The above implies that f~(u) is periodic with a period 2σ1 . Integrating with respect to u yields
~v
f~(u + 2σ1 ) = f~(u) + 2σ1 (4.59)
c
where we have conveniently chosen our integration constant, implying that f~ is in fact quasi-periodic
(Zwiebach, 137).
Now let us consider the parameterization conditions. Convince yourself that if we add and
subtract the second expression (4.54) to the first expression in (4.55) we obtain
!2 !2
~
∂X ~ ∂X
1 ∂X ~ 1 ~
∂X
±2 + 2 =1
∂σ c ∂σ ∂t c ∂t
Or, written more compactly,

!2
~
∂X ~
1 ∂X
± =1 (4.60)
∂σ c ∂t
~ we find that
If we use our explicit expression for X,
~
∂X 1 ~0
= f (ct + σ) − f~0 (ct − σ)
∂σ 2
and
~
1 ∂X 1 ~0
= f (ct + σ) − f~0 (ct − σ)
c ∂t 2
providing
~
∂X ~
1 ∂X
± = ±f~0 (ct ± σ)
∂σ c ∂t
Applying (4.60), we find that
f~0 · f~ = 1 (4.61)
~0
implying that f is a unit vector. Notice that at σ = 0 we have
~ 0) = f~(ct)
X(t,
4.5. MOTION OF OPEN AND CLOSED STRINGS 89
indicating that f~(u) is the position of the σ = 0 endpoint at a time u

c. Similarly, from (3.67),
~v
f~(2σ1 ) = f~(0) + 2σ1
c
Altogether, notice that
~ = t + 2σ1 , σ) = 1 f~(ct + 2σ1 + σ) + f~(ct + 2σ1 − σ) = 1 f~(ct + σ) + f~(ct − σ) + 2 σ1 ~v

h i h i
X(t
c 2 2 c
~ σ) + 2 ~vσ 1
= X(t,
c
which yields that ~v is the average velocity of any point σ during the time interval [0, 2 σc1 ] (Zwiebach,
140).
In an analogous manner, we may examine the motion of a closed string. Again,from the wave
equation, we have the general solution
~ σ) = 1 f~(u) + ~g (v)

X(t, (4.62)
2
where we have defined u = ct + σ and v = ct − σ. Taking derivatives we find that
1 ∂X~ 1 ~0
= f (u) + ~g 0 (v)
c ∂t 2
∂X~ 1 ~0
= f (u) − ~g 0 (v)
∂σ 2
Thus we have
~
∂X ~
1 ∂X
+ = f~0 (u) (4.63)
∂σ c ∂t
and
∂X ~ ~
1 ∂X
− = −~g 0 (v) (4.64)
∂σ c ∂t
Just as before, the parameterization constraints give
0 2 0 2

f~ (u) =~g (v) = 1 (4.65)

The crucial difference open and closed strings is that closed strings require the identification con-
dition
σ ∼ σ + σ1
Therefore,
~ σ + σ1 ) = X(t,
X(t, ~ σ)
which yields
f~(u + σ1 ) + ~g (v − σ1 ) = f~(u) + ~g (v)
f~(u) and ~g (v) can be best described as two parameterized closed curves on the surface of a unit
2-sphere (Zwiebach, 143). It might be however that the slopes f~0 (u) and ~g 0 (v) intersect for some
values uo and vo , i.e.
f~0 (uo ) = ~g 0 (vo )
It follows then for corresponding values to and σo that

~
1 ∂X 1 ~0
(to , σo ) = f (uo ) + ~g 0 (vo ) = f~0 (uo )
c ∂t 2
But since f~0 (u) is a unit vector
!2
1 ~
∂X ~
∂X
(to , σo ) =1⇒ (to , σo ) = c
c2 ∂t ∂t
In other words, the point σ = σo on the string reaches the speed of light at a time t = to .
Before moving onto the next chapter, let us briefly summarize what we have accomplished in this
chapter. We started off with the goal of determining the action of a relativistic string. Using the
developments of the relativistic point particle as a guide, we required that our action be proportional
to a proper area, eventually resulting in the Nambu-Goto string action. From there we ascertained
the equations of motion of by varying the action in the usual way. To analyze these equations, we
chose the static gauge parameterization and a clever parameterization for σ, allowing us to write the
string action, and equations of motion in terms of the transverse velocity ~v⊥ . Finally, we applied
various contraints to determine the general motion of both an open an closed string.
It should be noted that the analysis done here is not found in most of the graduate texts on
the subjects. That being said, it was determined that this material be included in our text as it
supplies the new student with a better physical intuition of relativistic strings. In the next chapter
we consider another approach, light-cone relativistic strings, that of which can be found in some
other graduate texts (most notably Polchinski’s String Theory).
4.6 Exercises
p
1. Derive the expression |x × y| = (x · x)(y · y) − (x · y)2 using the conventional rules of dot
products and cross products.
2. (a) Show that the area A given in (4.7) is reparameterization invariant. This is actually a
property of the metric itself. (Hint: Start by showing that for another set of parameters ξ˜ the
equality
˜ ξ˜p dξ˜q
gij (ξ)dξ i dξ j = g̃pq (ξ)d
must be satisfied. Take the chain of the right hand side, and defining the matrix M̃ij = ∂ ξ˜i /∂ξ j ,
rewrite gij (ξ). Take the determinant of both sides, and the result should follow).
(b) Argue that (4.12) is reparameterization invariant.
3. Show that the string given in the static gauge satisfies the equation of motion in (4.24).
4. In this exercise the reader will rewrite the Nambu-Goto action in terms of the transverse
velocity, ~v⊥ . (a) Start with the fact that for any vector ~u, its component orthogonal to a unit
vector ~n is ~u⊥ = ~u − (~u · ~n)~n. Using this show
4.6. EXERCISES 91
!
~
∂X ~ ∂X
∂X ~ ~
∂X
~v⊥ = − ·
∂t ∂t ∂s ∂s
~
which reduces to ~v⊥ = ∂ X/∂t using our σ parameterization.
(b) Using the static gauge, calculate (Ẋ)2 , (X 0 )2 , and Ẋ ·X 0 . Therefore show that the Lagrangian
takes the form in (4.31).
5. Show that (4.56) indeed satisfies the wave equation (4.54).

Chapter 5
Relativistic Strings and Mode

Expansions
5.1 A More General Parameterization

In the last chapter we worked in the static gauge, where we let
X 0 (τ, σ) = cτ (5.1)
In general however, we can choose a gauge where the string coordinates X µ can be written in a
linear combination proportional to τ . That is,
nµ X µ (τ, σ) = λτ (5.2)
for a constant λ. Notice that if nµ = (1, 0, 0...) and λ = c, we recover the static gauge parame-
terization. To understand (5.1) more fully, consider instead the space-time coordinates xµ and the
same gauge condition
nµ xµ (τ, σ) = λτ
Specifically, consider two points xµ1 and xµ2 which satisfy this gauge condition. It follows that
nµ (xµ1 − xµ2 ) = 0
which implies that any vector joining points in the space nµ xµ is orthogonal to nµ . The set of all
such points forms a hyperplane normal to nµ , as shown in figure 5.1. Therefore, X µ (τ, σ) are string
coordinates and form a hyperplane orthogonal to nµ [60].
On another note, we would prefer that our strings to be space-like. The reason for this is we desire
points on the string to be spatially separated, or perhaps null in some limit, but never time-like.
Therefore, more accurately we desire ∆X µ to be space-like, or null. To guarantee that the interval
∆X µ is space-like or null, we require that nµ is null. For rigor, we simply prove this statement.
93
94 CHAPTER 5. RELATIVISTIC STRINGS AND MODE EXPANSIONS
Figure 5.1: The hyperplane orthogonal to nµ by the string coordinates X µ (τ, σ) using the general
gauge condition n · X = λτ (Motivated by Zwiebach [64]).
Let nµ = (n0 , ~n) be a null vector in a D-dimensional Minkowski space. Therefore, it satisfies
nµ nµ = 0. Moreover, let bµ = (b0 , ~b) be a vector satisfying nµ bµ = 0. For our null vector, we have
that |n0 |2 = |~n|2 , yielding n0 = ±|~n|. Without loss of generality, we assume that n0 = |~n| = 1. By
the requirement on bµ , we have
nµ bµ = b0 − ~n · ~b = 0
if and only if b0 = ~n ·~b. With |~n| = 1, then |~n ·~b| ≤ |~b|. So, |b0 | ≤ |~b|. This yields two possibilities for
bµ : (1) bµ is space-like (|b0 | < |~b|), or (2) bµ is null (|b0 | = |~b|). In fact, if bµ is null, then bµ = λnµ .
Therefore, in our case, if nµ is a null vector, and nµ ∆X µ = 0, then ∆X µ is space-like or null in
some limit. It can be further shown that if nµ is strictly time-like, and nµ ∆X µ = 0, then ∆X µ
is strictly space-like. We choose the more general case when nµ is null, thereby allowing for the
possibility that the interval ∆X µ is occasionally null.
We can further modify our gauge condition as
nµ X µ (τ, σ) = λ̃(nµ pµ )τ (5.3)
where λ̃ is some new constant replacing λ, and, as we will see later, pµ is a well-defined conserved
momentum for open strings with free endpoints. Therefore, for pµ to be conserved, we must have
that have that nµ pµ = constant. By unit analysis, we notice that λ̃ ∼ Tc . Letting c = 1, then
1
λ̃ ∼ ≡ 2πα0
T
where α0 is known as the slope parameter. It turns out for open strings that λ̃ = 2α0 yielding
n · X(τ, σ) = 2α0 (n · p)τ (5.4)

5.1. A MORE GENERAL PARAMETERIZATION 95
Now that we have an accurate τ parameterization, let us generalize our σ parameterization. In

the static gauge, σ is found by constructing a constant energy density P τ 0 over the strings. In
general then, we demand that nµ P τ µ = n · P τ = constant. For open strings, we also require a
parameterization range of σ ∈ [0, π]. Using the constancy condition of the energy density, we may
write that
n · P τ (τ, σ) = a(τ )
Notice however if we integrate both sides over the range of σ
Z π Z π
dσ (n · P τ (τ, σ)) = n · p = dσa(τ ) = πa(τ )
0 0
Asserting
n·p
a(τ ) =
π
Thus, for open strings,
n·p
n · Pτ = (5.5)
π
From this parameterization, and making use of the equations of motion
∂Pµσ ∂Pµτ
+ =0 (5.6)
∂σ ∂τ
we may write
∂ ∂
(nµ P τ µ ) + (nµ P σµ ) = 0
∂τ ∂σ
We already established that nµ P τ µ = constant, leaving us with
∂
(n · P σ ) = 0
∂σ
implying that n · P σ is a constant. However, to maintain the conservation of n · p, we may assume
that n · P σ = 0.
For closed strings we perform a similar analysis. The τ parameterization becomes

n · X(τ, σ) = α0 (n · p)τ (5.7)
where we have eliminated the factor of 2 for convenience. The crucial difference for our σ pa-
rameterization is that we require the parameterization range of σ as σ ∈ [0, 2π] for closed strings.
Analogously then,
n·p
n · Pτ = (5.8)
2π
∂
Similar to before, we also have ∂σ (n · P σ ) = 0. We again may choose n · P σ = 0 for closed strings,
however the reason for this turns out to be quite subtle, relating to the nature of closed strings
themselves (Zwiebach, 180).
To summarize, we may write our gauge conditions as
n · X(τ, σ) = βα0 (n · p)τ (5.9)
2π
n·p= n · Pτ (5.10)
β
where β = 1 for closed strings, and β = 2 for open strings.
5.2 Constraining the Wave Equation

From our choices of parameterization, we have constraints on Ẋ and X 0 . Recall that,
2 2
Ẋ · X 0 Xµ0 − (X 0 ) Ẋµ Ẋ · X 0 Ẋµ − Ẋ Xµ0

Pµτ = −T q 2 2 Pµσ = −T q 2 2 (5.11)
2 2
Ẋ µ Xµ0 − Ẋµ Ẋ µ (X 0 ) Ẋ µ Xµ0 − Ẋµ Ẋ µ (X 0 )
Then,
1 Ẋ · X 0 ∂τ (n · X) − Ẋ 2 ∂σ (n · X)

σ
n·P =−
2πα0
q 2 2 2
Ẋ µ Xµ0 − Ẋµ Ẋ µ (X 0 )
But from n · X(τ, σ) = βα0 (n · p)τ ,

∂σ (n · X) = 0
yielding
Ẋ · X 0 ∂τ (n · X)

σ1
n·P =− =0 (5.12)
2πα0
q 2 2 2
Ẋ µ Xµ0 − Ẋµ Ẋ µ (X 0 )
where have used our assumption that n · P σ = 0. Moreover, since ∂τ (n · X) = βα0 (n · p) =

constant, n · P σ = 0 is satisfied only if
Ẋ · X 0 = 0 (5.13)
Using this we find
2
Ẋ · X 0 Ẋµ − Ẋ Xµ0

∂L
Pµσ ≡ = −T q (5.14)
∂X 0µ 2 2 2
Ẋ µ Xµ0 − Ẋµ Ẋ µ (X 0 )
1 (X 0 )2 Ẋµ
= 0
q
2πα
−(Ẋ)2 (X 0 )2
From (5.10) we may write
2π 2π 1 (X 0 )2 ∂τ (n · X)
n·p= n · Pτ =
β 2πα0
q
β
−(Ẋ)2 (X 0 )2
However we know that

∂τ (n · X) = βα0 (n · p)
yielding
(X 0 )2 (X 0 )2
n·p= q (n · p) ⇒ 1 = q
−(Ẋ)2 (X 0 )2 −(Ẋ)2 (X 0 )2
which leads to
Ẋ 2 + X 02 = 0 (5.15)
5.3. SOLVING THE WAVE EQUATION 97
Together, both constraints (5.13) and (5.15) give

2
Ẋ ± X 0 =0 (5.16)
Also since X 02 > 0, then q √

−(Ẋ)2 (X 0 )2 = X 02 X 02 = X 02
Therefore,
1
Pτµ = Ẋ µ (5.17)
2πα0
Similarly,
0 0
σµ 1 Ẋ 2 X µ 1 Ẋ 2 X µ
P = =
2πα0 2πα0 X 02
q
−(Ẋ)2 (X 0 )2
Using (5.15), the above simplifies to
1 0
P σµ = − Xµ (5.18)
2πα0
The equation of motion then becomes,
∂Pµσ ∂Pµτ 00
+ = Ẍ µ − X µ = 0 (5.19)
∂σ ∂τ
which is just the wave equation!
5.3 Solving the Wave Equation

Here we will solve the equations of motion for two possible types of a relativistic string: an open
string with free endpoints, and a closed string. As an exercise, the reader will work out the equations
of motion for an open string with fixed endpoints. We start by considering an open string with
free endpoints, however the analysis for each type of string will be analogous. The most general
solution of the wave equation is
1 µ
X µ (τ, σ) = (f (τ + σ) + g µ (τ − σ)) (5.20)
2
Since we are considering an open string with free endpoints, we use Neumann boundary conditions
∂X µ

=0 (5.21)
∂σ σ=0,π
The boundary conditions at σ = 0 yield
∂X µ 1 µ0 0

(τ, 0) = f (τ ) − g µ (τ ) = 0
∂σ 2
0 0
⇒ gµ = f µ ⇒ gµ = f µ
Therefore we may rewrite (5.20) as

1 µ
X µ (τ, σ) = (f (τ + σ) + f µ (τ − σ)) (5.22)
2
Now consider the boundary conditions when σ = π
∂X µ 1
(τ, π) = (f µ (τ + π) + f µ (τ − π)) = 0
∂σ 2
0
Since this holds for all τ , we conclude that f µ is periodic with a period of 2π. Moreover, since
0 0
f is periodic, we may write f µ as a general Fourier series
µ
∞
0 X
f µ (u) = f1µ + (aµn cos(nu) + bµn sin(nu)) (5.23)
n=1
where we are using u to stand in as any general argument. Integrating with respect to u gives
us
∞
X
f µ (u) = f0µ + f1µ u + (Aµn cos(nu) + Bnµ sin(nu)) (5.24)
n=1
where we have replaced aµn and bµn with Aµn and Bnµ to denote that the integration of (5.23) picks
up extra constants which are absorbed giving the coefficients Aµn and Bnµ . Then by substitution,
1 µ
X µ (τ, σ) =
(f (τ + σ) + f µ (τ − σ))
2
∞
!
1 µ µ
X
µ µ
= f0 + f1 (τ + σ) + [An cos(n(τ + σ)) + Bn sin(n(τ + σ))]
2 n=1
∞
!
1 X
+ f0µ + f1µ (τ − σ) + [Aµn cos(n(τ − σ)) + Bnµ sin(n(τ − σ))]
2 n=1
∞
X
= f0µ + f1µ τ + [Aµn cos(nτ ) + Bnµ sin(nτ )] cos(nσ) (5.25)
n=1
We wish to change the coeffcients Aµn and Bnµ so that they have more of a physical interpretation.
First we do this by introducing
i µ
Aµn cos(nτ ) + Bnµ sin(nτ ) = − (Bn + iAµn ) einτ − (Bnµ − iAµn ) e−inτ

2
√
−i 2α0 µ inτ
ān e − aµn e−inτ

≡ √ (5.26)
n
where āµn is the complex conjugate of aµn . As we will see later, āµn and aµn are the creation and
annihilation operators familiar to quantum mechanics and quantum field theory. The constant f1µ
has a simple physical interpretation as well. Notice that
∞
!
τµ 1 µ 1 µ
X
µ µ
P = Ẋ = f1 + [Bn n cos(nτ ) − An n sin(nτ )] cos(nσ) (5.27)
2πα0 2πα0 n=1
Integrating yields Z π
1 µ
pµ = dσP τ µ = f π ⇒ f1µ = 2α0 pµ (5.28)
0 2πα0 1
where pµ is the total string momentum. Lastly, we simply declare
f0µ = xµ0 (5.29)
where xµ0 is the center-of-mass position (Becker, Becker, Schwarz, 34). Altogether then, we have
∞
√ X cos(nσ)
X µ (τ, σ) = xµ0 + 2α0 pµ τ − i 2α0 āµn einτ − aµn e−inτ √ (5.30)
n=1
n
Notice if āµn = aµn = 0, then we obtain the equation of motion for a point particle. Moreover, if
we take the complex conjugate,
∞
√ X cos(nσ)
X̄ µ (τ, σ) = xµ0 + 2α0 pµ τ + i 2α0 aµn e−inτ − āµn einτ √ = X µ (τ, σ)
n=1
n
Therefore X µ is real. This is only true however as long as we assume that xµ0 and pµ are both real.
As we will see shortly, this is not always the case.
Let us also introduce further notation to simplify our expression. We make the following defini-
tions √
α0µ = 2α0 pµ (5.31)
µ µ
√
αn = an n (5.32)
µ √
α−n = āµn n (5.33)
µ
Obviously, (5.32) and (5.33) holds when n 6= 0. Moreover, notice that ᾱnµ = α−n . Using these
definitions, we may rewrite (5.30) as
∞
√ √ X 1 µ inτ
X µ (τ, σ) = xµ0 + 2α0 α0µ τ − i 2α0 α−n e − αnµ einτ cos(nσ)

n=1
n
Or,
√ √ X1
X µ (τ, σ) = xµ0 + 2α0 α0µ τ + i 2α0 αµ e−inτ cos(nσ) (5.34)
n n
n6=0
It is a rather trivial exercise to find the spatial and time derivatives of X µ . We simply give the
expressions for each √ X
Ẋ µ = 2α0 αnµ cos(nσ)e−inτ (5.35)
n∈Z
0 √ X
Xµ = −i 2α0 αnµ sin(nσ)e−inτ (5.36)
n∈Z
And together we have √

0 X
Ẋ µ ± X µ = 2α0 αnµ e−in(τ ±σ) (5.37)
n∈Z
Before moving on to the closed string, we first note that often we will see X µ written in terms
of left moving and right moving solutions. That is,
X µ (τ, σ) = XLµ (τ + σ) + XR
µ
(τ − σ) (5.38)
which, after doing a similar Fourier expansion, take the form
xµL,0
r
µ α0 µ α0 X α̃nµ −in(τ +σ)
XL (τ, σ) = + p̃ (τ + σ) + i e (5.39)
2 2 2 n
n6=0
xµR,0
r
µ α0 α0 X αnµ −in(τ −σ)
XR (τ, σ) = + pµ (τ − σ) + i e (5.40)
2 2 2 n
n6=0
Here we have used ∼ to distinguish between the left and right modes. It is still true that
µ µ ¯ nµ (Becker, Becker, Schwarz, 34). Notice then
α−n = ᾱnµ and α̃−n = α̃
∂XLµ
r
α0 α0 X µ −in(τ +σ)
= p̃µ + α̃n e
∂σ 2 2
n6=0
and
µ
r
∂XR α0 α0 X µ −in(τ −σ)
= − pµ − αn e
∂σ 2 2
n6=0
Using Neumann boundary conditions (5.21), we find that
∂XLµ µ
r
∂XR 0 α0 X µ
+ µ µ
= α (p − p̃ ) + (αn − α̃nµ )e−inτ = 0
∂σ σ=0
∂σ σ=0
2
n6=0
which leads to
pµ = p̃µ (5.41)
αnµ = α̃nµ (5.42)
(5.42) reveals something interesting. It indicates that open strings with free boundary points have
the same left and right moving modes. Physically this means that for an open string with free
endpoints force the left and right moving modes to combine into standing waves. Moreover, using
(5.41) and (5.42), we notice that (5.38) reduces to (5.34).
Lastly, we consider the closed string. In the case of closed strings, we must use the condition of
periodicity, namely,
X µ (τ, σ) ∼ X µ (τ, σ + 2π) (5.43)
.
The most general solution to the wave equation satisfying periodicity condition is similar to that
of the open string,
xµL,0
r
µ α0 µ α0 X α̃nµ −in(τ +σ)
XL (τ, σ) = + p (τ + σ) + i e (5.44)
2 2 2 n
n6=0
xµR,0
r
µ α0 µ α0 X αnµ −in(τ −σ)
XR (τ, σ) = + p (τ − σ) + i e (5.45)
2 2 2 n
n6=0
Notice here that there isn’t any term with p̃µ . This is because the identification condition of
a closed string forces pµ = p̃µ . It turns out that if there are compact extra dimensions, pµ = p̃µ
is no longer satisfied (McMahon, 47). We will discuss compactification and the presence of extra
dimensions in a later chapter. Moreover, using the fact that the total momentum is
√
Z 2π Z 2π Z 2π
r
1 1 2 µ
µ
p = P τµ
(τ, σ)dσ = µ
dσ Ẋ (τ, σ) = dσ 2α0 α0µ = α (5.46)
0 2πα0 0 2πα0 0 α0 0
Thus,
r
µ 2 µ
p = α (5.47)
α0 0
Putting everything together, noting that α0µ = α̃0µ and using (5.47) we may write
√
r
1 µ µ µ α0 X e−inτ
µ
αnµ einσ + α̃nµ e−inσ

0
X (τ, σ) = (xL,0 + xR,0 ) + 2α α0 τ + i (5.48)
2 2 n
n6=0
Since there is only one momentum variable, pµ , in the corresponding quantum theory there will
only be one momentum operator. There is also one conjugate coordinate to the zero mode, therefore
xµL,0 = xµR,0 ≡ xµ0 . Altogether then we have
√
r
α0 X e−inτ
µ
xµ0 2α0 α0µ τ αnµ einσ + α̃nµ e−inσ

X (τ, σ) = + +i (5.49)
2 n
n6=0
Moreover, we have
0 0 √ X
Ẋ µ + X µ = 2XLµ (τ + σ) = 2α0 α̃nµ e−in(τ +σ) (5.50)
n∈Z
µ 0 0 √ X
Ẋ µ − X µ = 2XR (τ − σ) = 2α0 αnµ e−in(τ −σ) (5.51)
n∈Z
These relations will be important when we move to quantize the closed string in a later chapter.
Indeed we have solved the wave equation for open and closed strings using mode expansions,
however we must also check that the constraints (5.16) are satisfied. Arbitrarily specifying the
constants αnµ will not satisfy the constraints. Instead, we will use the light-cone gauge to find
solutions to the wave equation and satisfy the constraints.
5.4 Solutions to the Wave Equation in the Light-Cone Gauge

To solve the constraints of the wave equation, we will use the light-cone gauge to specify worldsheet
coordinates. This means imposing the condition
n · X = X+ (5.52)
. Letting nµ = ( √12 , √12 , 0, 0) we find
X0 + X1
n·X = √ = X+ (5.53)
2
and for momentum,
p0 + p1
n·p= √ = p+ (5.54)
2
Using these relations, we find
X + (τ, σ) = βα0 p+ τ (5.55)
2π τ +
p+ = P (5.56)
β
The idea behind using the light-cone gauge as defined is to show that there are no dynamics of the
string on X − but rather all of the dynamics are on the transverse coordinates X I = (X 2 , X 3 , ..., X d ).
We consider the constraints (5.16). To evaluate using light-cone coordinates, recall from chapter
two that the dot product is given by
a · b = −a+ b+ − a− b− + a2 b2 + a3 b3 + ... (5.57)
Therefore, (5.16) becomes

0 0 0
−2(Ẋ + ± X + )(Ẋ − ± X − ) + (Ẋ I ± X I )2 = 0 (5.58)
0
Since X + = 0 and Ẋ + = βα0 p+ , its easy to see
0 1 0
Ẋ − ± X − = (Ẋ I ± X I )2 (5.59)
2α0 βp+
where we assume that p+ > 0. If p+ were in fact zero, the light-cone formalism does not work.
0
We can determine both Ẋ − and X − in terms of X I , so we can find X − up to an integration
constant. Notice from (5.59) we write it as two equations rather than one
0 1 h
I0 2 I I0
i
Ẋ − + X − = (Ẋ I 2
) + (X ) + 2 Ẋ X (5.60)
2α0 βp+
0 1 h 0 0
i
Ẋ − − X − = (Ẋ I )2 + (X I )2 − 2Ẋ I X I (5.61)
2α0 βp+
It follows that
1 h
I0 2
i
Ẋ − = ( Ẋ I 2
) + (X ) (5.62)
2α0 βp+
5.4. SOLUTIONS TO THE WAVE EQUATION IN THE LIGHT-CONE GAUGE 103
0 1 0
X− = Ẋ I X I (5.63)
2α0 βp+
Notice then
∂X I ∂ 2 X I ∂X I ∂ 2 X I

∂ 1
(Ẋ − ) = 0 + +
∂σ α βp ∂τ ∂σ∂τ ∂σ ∂σ 2
and similarly
∂X I ∂ 2 X I ∂X I ∂ 2 X I

∂ −0 1
(X ) = 0 + +
∂τ α βp ∂τ ∂σ∂τ ∂σ ∂τ 2
∂ −0 ∂ −
We would like the consistency condition ∂τ (X ) = ∂σ (Ẋ ) to hold. This is only possible
however if
∂2X I ∂2X I
2
= (5.64)
∂τ ∂σ 2
That is, we require that X I satisfy the wave equation to maintain consistency condition above.
As long as we assume that X I satisfies the wave equation, by our consistency condition we may
integrate the total differential
∂X − ∂X −
dX − = dτ + dσ
∂τ ∂σ
along any path, for open strings, to determine X − .
For closed strings, there is a further complication. A contour integral over a path beginning and
ending at the same point might not give one the expected zero as we integrate over X − , which is
neccessary for X − to be well defined. Therefore, we will require
2π
∂X −
Z
dσ =0 (5.65)
0 ∂σ
Now that we have satisfied the constraints, we may move on an explicitly solve the wave equation
in the light-cone gauge. We will only focus on open strings, however the analysis is similar for closed
strings. Using β = 2 and (5.34) we have the following:
√
X + (τ, σ) = 2α0 p+ τ = 2α0 α0+ τ (5.66)
∞
√ √ X 1 − −inτ
−
X (τ, σ) = x− + 2α0 α0− τ + i 2α 0 α e cos(nσ) (5.67)
0
n n
n6=0
∞
√ √ X 1 I −inτ
X I (τ, σ) = xI0 + 2α0 α0I τ + i 2α0 α e cos(nσ) (5.68)
n n
n6=0
Using (5.37) we find

0 √ X
(Ẋ − ± X − ) = 2α0 αn− e−in(τ ±σ) (5.69)
n∈Z
0 √ X
(Ẋ I ± X I ) = 2α0 αnI e−in(τ ±σ) (5.70)
n∈Z
If we make use of (5.67) and (5.68), along with (5.59), we may solve for the αn− oscillators:
√ X 1 X I I −i(p+q)(τ ±σ)
2α0 αn− e−in(τ ±σ) = αp αq e
2p+
n∈Z p,q∈Z
 
1 X 1 X X
= αpI αn−p
I
e−in(τ ±σ) = +  αpI αn−p
I 
2p+ 2p
n,q∈Z n∈Z p∈Z
Comparing the left and right hand sides, we conclude that

√ 1 X I I
2α0 αn− = αp αn−p (5.71)
2p+
p∈Z
Equation (5.71) comes up often enough that it is referred to as the transverse Virasoro mode L⊥
n
√ 1 ⊥
2α0 αn− = L (5.72)
p+ n
1X I I
L⊥
n = αp αn−p (5.73)
2
p∈Z
√
Notice that for n = 0, and using α0µ = 2α0 pµ ,
√ 1 ⊥ 1
2α0 α0− = 2α0 p− = L → 2p+ p− = 0 L⊥
p+ 0 α 0
Thus,
0 1 X ⊥ −in(τ ±σ) 1 0
(Ẋ − ± X − ) = Ln e = + (Ẋ I ± X I )2 (5.74)
p+ 4p
n∈Z
Moreover, by substitution, we may rewrite (5.69) as

1 ⊥ i X 1 ⊥ −inτ
X − (τ, σ) = x−
0 + L0 τ + + L e cos(nσ) (5.75)
p + p n n
n6==
which indicates that the Virasoro modes are in fact expansion modes of X − (τ, σ) [64].
Lastly, the mass of the string can be ascertained via
M 2 = −pµ pµ = −p2 = 2p+ p− − pI pI (5.76)
Using the fact that 2p+ p− = 1 ⊥

α0 L0 and
∞ ∞
1X I 1 X 1 X
L⊥
0 = α0−p αpI = α0I α0I + I
α−n αnI = α0I α0I + ᾱnI αnI
2 2 n=1
2 n=1
p∈Z
we find
∞
!
1
+ − 1 I I X I I
2p p = 0 α0 α0 + ᾱn αn
α 2 n=1
5.5. EXERCISES 105
√ √
But since α0µ = α0 pµ , and αnµ = aµn n its easy to see that
∞
1 X I I
2p+ p− = pI pI + nā a
α0 n=1 n n
Thereby yielding
∞
1 X I I
M2 = nā a (5.77)
α0 n=1 n n
It turns out that our expressions for M 2 will not survive when we quantize the string. Rather,
it will be come quantized itself, causing the string states to not exhibit a continuous spectrum
of masses. A continuous mass spectrum is completely viable in classical physics, however this is
not the case in quantum mechanics. Moreover, we don’t observe particles that take on continuous
values for mass. Most of all, (5.77) does not yield any massless states. As we will see later on, an
extra constant will be added to (5.77) allowing for massless states, an imperative requirement if
string theory is to explain all particles, including photons and gravitons.
5.5 Exercises
1. Going through the details, derive (5.17) and (5.18).
2. Using (5.34), prove (5.35) and (5.36), and hence (5.37).
3. In this exercise you will come up with the solutions to the wave equation for a string with
fixed endpoints. (a) Show that Dirichlet boundary conditions are satisfied only if p̃µ = −pµ and
α̃nµ = αnµ .
(b) Using part (a), show that

√ √ X αµ
X µ (τ, σ) = xµ0 + 2 α0 pµ σ − 2α0 n −inτ
e sin(nσ)
n
n6==0
(c) Using the result from part√(b), compute the length of the string, ∆X µ . Write it in terms of
the so-called winding term, w = 2α0 pµ .
4. Prove that (5.50) and (5.51) hold.
5. Assuming all of the constraints are satisfied, use the analogous computation for the open string
to solve the wave equation in the light-cone gauge for a closed string. Make sure to find the closed
string Virasoro modes.
Chapter 6
Charges, Currents, and

Symmetries
6.1 Conserved Quantities and Noether’s Theorem

From classical physics, we are familiar with the conservation laws of energy and momentum. In
classical field theory, one is able to ascertain conserved quantities by examining the symmetries of
a Lagrangian. As we will see shortly, energy and momentum are such conserved quantities, deter-
minable by the Lagrangian. Energy and momentum are only some of these conserved quantities.
In general, we find the conserved charge or the conserved current, using a classical result known as
Noether’s Theorem. We’d like to use this technique in string theory so that we can determine the
conserved quantities associated with our string Lagrangian, as it will be useful later on when we go
to quantize the string. Before examining conserved quantities in string theory however, we must
first become familiar with the notion of a conserved charge and a conserved current classically.
From electromagnetism, we have the conserved current 4-vector j µ = (cρ, ~j). We say that the
electric current is conserved because it satisfies
∂µ j µ = 0 (6.1)
The four vector current isn’t the only quantity which satisfies this equation however. Due to
analogy, any four vector which satisfies this equation is referred to as a conserved current. (6.1)
says more than current is conserved; it is also a statement about conservation of charge. Writing
out the (6.1) explicitly, we have,
∂j 0
∂0 j 0 + ∂i j i = + ∇ · ~j = 0
∂x0
From classical electrodynamics, the charge is the integral of the charge density ρ over the volume
of the space in which the charge is spread throughout
Z Z
3 j(t, ~x) 3
Q(t) = ρ(t, ~x)d x = d x (6.2)
V V c
107
108 CHAPTER 6. CHARGES, CURRENTS, AND SYMMETRIES
Taking the time derivative gives

∂j 0 3
Z Z
dQ
= d x=− ∇ · ~jd3 x
dt V ∂x0 V
By the divergence theorem, we may write

Z
dQ ~j · d~a
=− (6.3)
dt ∂V
The above suggests that charge inside the region V can only change if there is a flux of current
across the boundary ∂V . Often in in electrodynamics, we take the case where the current ~j vanishes
at the boundary, leaving
dQ
=0 (6.4)
dt
The charge Q is then said to be conserved. Charge, it turns out is also Lorentz invariant. This
property is not true for all conserved quantities however. For example, energy is conserved, but is
not Lorentz invariant.
We have already seen some of the important uses of Lagrangians, however one of the most
useful property is that Lagrangians can help find conserved quantities. It is the symmetries of the
Lagrangian which allow for the deduction of the existence of conserved quantities. By symmetry we
typically mean a transformation which leaves the equations of motion invariant. Mathematically,
a symmetry is a variation to the fields or to the Lagrangian that leave the equations of motion
invariation. To witness this, consider the variation in space-time coordinates
x µ → x µ + µ
with µ as a small constant. The field φ(x) then changes as
φ(x) → φ(x + )
Expanding φ in a Taylor series, we arrive to
φ(x + ) ≈ φ + µ ∂µ φ (6.5)
Varying the field in general means

φ → φ + δφ
This suggests that our variation is
δφ = µ ∂µ φ (6.6)
Now let’s see what happens when we vary the Lagrangian density L = L(φ, ∂µ φ). Using some
results from chapter three, the variation is given by
∂L ∂L
δL = δφ + δ(∂µ φ) (6.7)
∂φ ∂(∂µ φ)
Recall that the Euler-Lagrange equation is given by

∂L ∂L
= ∂µ (6.8)
∂φ ∂(∂µ φ)
6.1. CONSERVED QUANTITIES AND NOETHER’S THEOREM 109
Using this, we may rewrite (6.7) as

∂L ∂L
δL = ∂µ δφ + ∂µ (δφ)
∂(∂µ φ) ∂(∂µ φ)
where we used the fact δ(∂µ φ) = ∂µ (δφ). The above can be further simplified by writing it as a
total derivative
∂L
δL = ∂µ δφ (6.9)
∂(∂µ φ)
Applying δφ = ν ∂ν φ, we have

∂L ν ∂L
δL = ∂µ ∂ν φ = ∂µ ∂ν φ ν
∂(∂µ φ) ∂(∂µ φ)
Analogous to the field variation given in (6.6), the Lagrangian also varies with the displacement,
given by
δL = ∂µ (L)µ = δνµ ∂µ (L)ν (6.10)

where δνµ is the familiar Kronecker-delta function. Equating both results for the variation of the
Lagrangian, we find
∂L
δL = δνµ ∂µ (L)ν = ∂µ ∂ν φ ν
∂(∂µ φ)

∂L
⇒ ∂µ ∂ν φ − δν L ν = 0
µ
∂(∂µ φ)
Since this holds for any arbitrary ν , we require

∂L
∂µ ∂ν φ − δνµ L = 0 (6.11)
∂(∂µ φ)
We call the expression in the parantheses the energy-momentum tensor

µ ∂L µ
Tν = ∂ ν φ − δν L (6.12)
∂(∂µ φ)
yielding
∂µ Tνµ = 0 (6.13)
Notice that if we consider the µ = ν = 0 component, we have
∂L
T00 = φ̇ − L = H (6.14)
∂ φ̇
which we recognize as the Hamiltonian density. Therefore,
∂µ T00 = 0 (6.15)
is a statement about the conservation of energy. Moreover, the components of momentum are
given by Ti0 with i running over spatial indices. The components of momentum are simply given
by Z
pi = d3 xTi0 (6.16)
Consider what we have just done. We assumed some space-time translation which in turn varied
our field and our Lagrangian. Both variations, a type of continuous symetry, allowed us to construct
two conserved quantities we are most familiar with: energy and momentum!
We may find conserved currents in an analogous way as done above. Consider the general
variations
φ → φ + δφ L → L + δL
Further suppose that we require L to be invariant under some variation, i.e. some symmetry, we
then require that δL = 0. We saw earlier that

∂L ∂L ∂L
δL = δφ + δ(∂µ φ) = δL = ∂µ δφ = 0
∂φ ∂(∂µ φ) ∂(∂µ φ)
We call the term in the parantheses the conserved current,
∂L
Jµ = δφ (6.17)
∂(∂µ φ)
which allows us to write the conservation of current equation (6.1)
∂µ J µ = 0 (6.18)
The charge associated with the conserved current J µ is given by (Hatfield,22)
Z
Q = d3 xJ 0 (6.19)
We are now in a position to precisely state Noether’s theorem: For any continuous symmetry
of the Lagrangian, that is, a variation in the field that leaves the Lagrangian invariant, there is a
conserved current which can be found using
∂L
Jµ = δφ
∂(∂µ φ)
This result is so fundamental that it has found its way into quantum field theory and string
theory.
6.2 Worldsheet Currents

For each string that moves freely, we would like to have an associated conserved momentum pµ . We
will see that there are conserved currents living on the worldsheet which give rise to the conserved
string momentum. To begin, recall the action
Z
dξ 0 dξ 1 L(∂0 X µ , ∂1 X µ ) (6.20)
6.2. WORLDSHEET CURRENTS 111
with ξ 0 = τ and ξ 1 = σ. To find the conserved currents, we require that the variation δX µ
does not change L. The current, under the constant space-time relation δX µ (τ, σ) = µ is given by
(Zwiebach, 159)
∂L
Jµa = (6.21)
∂(∂a X µ )
Briefly, here we use a to denote the world coordinates ξ a . Notice then, with the above definition,
we have
∂L ∂L
(Jµ0 , Jµ1 ) = , 0 = (Pµτ , Pµσ ) (6.22)
∂ Ẋ µ ∂X µ
Moreover, using the equation for conservation of current, we have that
∂Pµτ ∂Pµσ
∂a Pµa = + =0 (6.23)
∂τ ∂σ
But this is just the wave equation! In other words, the wave equation for the relativistic string is
also a statement about the conservation of the currents Pµa living on the worldsheet. If we continue
our analysis just as we did before, the charges are found by integrating the zeroeth components of
Pστ of the currents over space. Presently this means we integrate over σ
Z σ1
pµ (τ ) = Pµτ (τ, σ)dσ (6.24)
0
This equation by itself leads us to conclude that Pµτ is the σ density of the space-time momentum
carried by the string. To check conservation of charge, we differentiate pµ with respect to τ and
apply the wave equation:
σ1
∂Pµτ
Z σ1 Z σ1
dpµ ∂P σ µ σ

= dσ = − dσ = −Pµ
dτ 0 ∂τ 0 ∂σ 0
This vanishes for both open and closed strings, leaving us to conclude that
dpµ
=0 (6.25)
dτ
Since we differentiated with respect to τ rather than time t, a natural question arises: is pµ
conserved in world-sheet time or Minkowski time? There is a subtlety here, however, it turns out
pµ is conserved in both (Zwiebach, 161). Moreover, there are instances where the open string
momentum is not conserved. For open strings with fixed endpoints, we employ Dirichlet boundary
conditions. This means the strings are attached to D-branes (D for Dirichlet) which are not space
filling, i.e. the D-brane does not extend over all space. Though the string momentum is not
conserved in this case, it is true in fact that the total momentum of the string and the D-brane is
conserved. More on D-branes will covered in a later chapter.
Lastly, (6.24) isn’t the whole story. The integral assumes we are integrating over a curve where
τ is constant, which is not always the case. Since we trust the reparameterization invariance of
our physics, we can imagine a case where we integrate over a curve where τ is not constant. In
order to maintain the conservation of current and charge, we must modify (6.24) such that charge
is conserved for any arbitrary curve and for any arbitrary parameterization of the world-sheet. This
is the subject of one of the exercises at the end of this chapter.
6.3 Lorentz Charges and Currents

We would like to maintain Lorentz invariance with our conserved charges. To do this we will have to
construct conserved charges (and therefore currents) with a Lorentz symmetry. First, we recall that
Lorentz transformations are coordinate transformations on X µ which leave ηµν X µ X ν invariant. A
Lorentz transformation is an infinitesimal transformation
X µ → X µ + δX µ (6.26)
with δX µ = µν Xν , where µν is antisymmetric (Griffiths, 83; Zwiebach, 165). With this defini-
tion, we can show that the Lagrangian density L is invariant under Lorentz transformations. Every
term that appear in the Lagrangian density for the string are of the form
∂X µ ∂X ν
ηµν
∂ξ α ∂ξ β
Varying this term gives us
∂X µ ∂X ν ∂(δX µ ) ∂X ν ∂X µ ∂(δX ν )

δ ηµν α = η µν +
∂ξ ∂ξ β ∂ξ α ∂ξ β ∂ξ α ∂ξ β
ν µ
∂X ρ ∂X ν ∂X µ ∂X ρ

µρ ∂Xρ ∂X νρ ∂X ∂Xρ
= ηµν + = νρ + µρ
∂ξ α ∂ξ β ∂ξ α ∂ξ β ∂ξ α ∂ξ β ∂ξ α ∂ξ β
Focusing on the second term and letting µ → ρ and ρ → ν, we find that
∂X µ ∂X ν ∂X µ ∂X ν

δ ηµν α = ( νρ + ρν ) =0 (6.27)
∂ξ ∂ξ β ∂ξ α ∂ξ β
where we used the antisymmetry of νρ .
Moreover, using this Lorentz symmetry, we are able to construct conserved currents. Analogous
to (6.17) we have that
a ∂L
jµν = δX µ = Pµa µν Xν
∂(∂a X µ )
Recall that we may break up an antisymmetric tensor as
1
Tµν = (Tµν − Tνµ )
2
. Using the antisymmetry of µν , we may write
a 1
jµν = − µν (Xµ Pνa − Xν Pµa ) (6.28)
2
If we were substitute this into the conservation of current equation, the coeffcient − 12 µν would
not play a role. We therefore define the conserved currents to be
a
Jµν = Xµ Pνa − Xν Pµa (6.29)
The equation for conservation of current is then given by

τ σ
∂Jµν ∂Jµν
+ =0 (6.30)
∂τ ∂σ
6.3. LORENTZ CHARGES AND CURRENTS 113
Moreover, analogous to (6.29), the conserved charges are given by

Z
τ σ
Jµν = (Jµν dσ − Jµν dτ ) (6.31)
γ
Notice that the conserved charges are are antisymmetric (Jµν = −Jνµ ). Lastly, we may also
compute the Lorentz charges using constant lines of τ , yielding
Z Z
τ
Jµν = Jµν (τ, σ)dσ = (Xµ Pντ − Xν Pµτ )dσ (6.32)
Since P~ τ can be viewed as the momentum density, then from how we have defined our currents
a
Jµν , we notice that Jija , where i and j run over spatial indices, is the angular momentum density.
As an application, we may use (6.32) to determine mode expansion for the angular momentum
J µν of an open bosonic string. In the last chapter we saw that
Pµτ = T Ẋ µ (6.33)
1
where T = 2πα0 . Using (6.32), we may write the angular momentum as
Z π Z π
µν µν
J = Jτ dσ = T (X µ Ẋ ν − X ν Ẋ µ )dσ (6.34)
0 0
Moreover, in the last chapter we derived that solution to the wave equation for an open string
is given by
∞
√ X 1 µ inτ
X µ (τ σ) = xµ0 + 2α0 pµ τ − i 2α0 (α−n e − αnµ e−inτ ) cos(nσ) (6.35)
n=1
n
Taking the τ derivative yields
∞
√ X µ
Ẋ µ (τ, σ) = 2α0 pµ + 2α0 (α−n einτ + αnµ e−inτ ) cos(nσ) (6.36)
n=1
Substituting (6.35) and (6.36) into our expression for the angular momentum, we have
∞
!
µν
Z π
µ 0 µ
√ X 1 µ inτ µ −inτ
J =T x0 + 2α p τ − i 2α 0 (α e − αn e ) cos(nσ)
0 n=1
n −n
∞
!
0 ν
√ X
ν inτ ν −inτ
× 2α p + 2α 0 (α−n e + αn e ) cos(nσ) dσ
n=1
∞
!
Z π √
0 ν
X 1 ν inτ
−T xν0
+ 2α p τ − i 2α0 (α−n e − αnν e−inτ ) cos(nσ)
0 n=1
n
∞
!
0 µ
√ X µ
inτ µ −inτ
× α p + 2α 0 (α−n e + αn e ) cos(nσ) dσ
n=1
Noting the fact that Z π

cos(σ)dσ = 0
0
we multiply out our terms, leaving only those which will contribute to the integral. This yields,
Z π
µν
J =T [(xµ0 + 2α0 pµ τ ) (2α0 pν ) − (xν0 + 2α0 pν τ ) (2α0 pµ )] dσ
0
Z π ∞
X 1 ν inτ µ
i2α0 (α−n e − αnν e−inτ )(α−n einτ + αnµ e−inτ ) cos2 (nσ)dσ

+T
0 n=1
n
Z π ∞
X 1 µ inτ
i2α0 (α−n e + αnµ e−inτ )(α−n
ν
einτ − αnν e−inτ ) cos2 (nσ)dσ

−T
0 n=1
n
Simplifying gives us
Z " ∞
#
π X 1
µν
(2α0 xµ0 pν 2α0 xν0 pµ ) 0 ν µ ν µ 2

J =T − + i2α 2α−n αn − 2αn α−n cos (nσ) dσ
0 n=1
n
∞
X 1 ν µ
= xµ0 pν − xν0 pµ + i µ
α−n αn − αnν α−n (6.37)
n=1
n
1
where we used T = 2πα 0 . The angular momentum of the open string therefore depends on the
µ µ
modes αn and α−n . Alternatively, the angular momentum, depends on the creation and annihilation
operators of quantum field theory. Computing the τ derivative we find that
d µν
J =0 (6.38)
dτ
Therefore, the angular momentum of an open bosonic string with free endpoints is conserved.
6.4 Exercises
1. Consider the Lagrangian density
i~ ∗ ~2
(ψ ψ̇ − ψ̇ ∗ ψ) −
L= (∇ψ ∗ · ∇ψ) − V (r)ψ ∗ ψ
2 2m
Show that the space part of the conserved current is
2
~j = i~ (ψ∇ψ ∗ − ψ ∗ ∇ψ)
2m
and that the time part is
−~ ∗
(ψ̇ ψ − ψ̇ψ ∗ )
j0 =
2
These expressions will be useful in the next chapter.
2. In this exercise we modify (6.24) such that charge is conserved for any arbitrary curve and
for any arbitrary parameterization of the world-sheet. For this, reconsider the integral in (6.24),
which really describes a flux of current across the curve where τ is constant. To generalize this
integral, consider an infinitesimal (dτ, dσ) along an oriented closed curve Γ that encloses a simply
6.4. EXERCISES 115
connected region R of the world-sheet (figure 6.1). We have that (dτ, dσ) is parallel to the tangent
to the infinitesimal segment, indicating that (dσ, −dτ ) is normal to the segment. We then define
the infinitesimal flux of current across the segment as
(Pµτ , Pµσ ) · (dσ, −dτ ) = Pµτ dσ − Pµσ dτ
(a) Show that the outgoing flux across the curve Γ is

Z τ
∂Pµ ∂P σ µ

pµ (Γ) = + dτ dσ = 0
R ∂τ ∂σ
using the two-dimensional divergence theorem, and the fact that Pµa are conserved currents,
thereby satisfying the conservation of current equation.
(b) The above flux integral can be further generalized. Pick any arbitrary curve γ, starting and
ending at the boundary points of the world-sheet. Moreover, consider curves ξ,α, and β, where ξ
is a curve of constant τ , and α and β are paths oriented in such a way that we may define a closed
contractible curve Γ as Γ = ξ − β − γ + α, where α and β are curves where dσ is constant. Sketch
Γ. Using this and part (a), show pµ (Γ) = 0, allowing you to also show that
Z
pµ = (Pµτ dσ − Pµσ dτ )
γ
for any arbitrary curve γ. This allows one to conclude that string charge conservation (conser-
vation of momentum) is an integral of the current over any curve on the world-sheet.
Figure 6.1: A closed string worldsheet with two closed curves γ and γ̄ at constant τ forming a
region R. (Motivated by Zwiebach [72])
Chapter 7
A Crash Course on Quantum Field

Theory
7.1 Introduction
To have a complete understanding of string theory, one should have a fair understanding of quantum
field theory. Unfortunately, most undergraduates have not likely had much experience with this
subject. Though one could scrape by without much knowledge of quantum field theory when
studying bosonic string theory, it is absolutely necessary for more advanced topics in string theory,
as it is fundamental to the study of superstrings. Since this text is aimed for undergraduates, we
seek to provide the minimal set of knowledge of quantum field theory so that later chapters in this
text and other graduate level texts are more accessible. We therefore present discussions on free
quantum fields, avoiding the technical details of interacting quantum fields. For further details on
these topics and more, the reader is urged to resort to the bibliography to seek more comprehensive
references.
Simply put, quantum field theory is the marriage of Einstein’s special relativity with ordinary
quantum mechanics. This union leads to profound physical insights, as well as rigorous mathe-
matical detail. When we introduce relativity into the quantum world, what we are really doing is
introducing Lorentz symmetry causing us to restrict to a local field description for point particles.
Remember that a field is a function of space variables and a single time variable. The quantum
states of these fields can be interpreted in terms of corresponding particles. Additionally, and
perhaps more importantly, when relativity and quantum mechanics merge, we obtain a notion of
causality; the fact that no measurable signal goes faster than the speed of light has far reaching
consequences.
This speed limit on measurable signals in turn forces local operators that represent physical
observables to commute at space-like separations. A consequence is that we must include negative
energy states. That is, in order to maintain causality, we are required to introduce antiparticles. If
we do not introduce negative energy solutions, then we violate causality. Since we appear to live
in a universe which obeys causal relationships, we insist that the negative energy solutions exist,
however interpret them in such a way so that they fit with laboratory observation. Worse still,
117
118 CHAPTER 7. A CRASH COURSE ON QUANTUM FIELD THEORY
to satisfy causality not only must we allow the existence of antiparticles, but (mostly as a useful
mathematical description), we must allow backwards-in-time propagation!
These two consequences of quantum field theory come from the fact that we seek a local field de-
scription of the quantum point particle along with a notion of causality. What ’s more is relativistic
quantum mechanics is the language that describes the standard model of particle physics. Quantum
field theory therefore has the ability to give an effective description of bosons and fermions, and
all known forces, aside from gravity. In this chapter, we will provide the rigor behind free bosons
and fermions, and briefly mention free photon fields. As the reader is most likely aware, when a
quantum field description was applied to gravity, the attempt was riddled with intractable infinities
plaguing the theory. This is the goal of string theory: to provide a theory unifying all of the forces
of nature, as well as a quantum theory of gravity. But before we get to quantizing the relativistic
string we have been developing the past few chapters, let’s first cover the basics of the theory of
quantum fields.
7.2 The Klein-Gordon Equation and Scalar Fields

The first attempts to merge relativity with quantum mechanics involved the relativistic general-
ization of the Schrödinger equation. Schrödinger himself actually came with this equation, known
as the Klein-Gordon equation, however he abandoned it because it gave solutions with negative
energy (which, as noted earlier, must stay), and gave the incorrect energy spectrum for hydrogen.
It turns out the Klein-Gordon equation is successful in describing spin-0 bosons, and is therefore a
tool we must become familiar with.
In relativity, time and space are on equal footing. To make a relativistic wave equation, we seek
to make the Schrödinger equation on equal footing with time and space. Recall that the Schrödinger
equation may take the form
∂ψ −~2 ~ 2
i~ = ∇ ·ψ+Vψ (7.1)
∂t 2m
Immediately we see that space and time are not on equal footing; there is a first derivative in
time while there is a second derivative in space. Therefore, we cannot start with the Schrödinger
equation, but instead must use a different method. Recall Einstein’s famous formula
E 2 = p2 c2 + m2 c4 (7.2)
Then, using the more general form of the Schrödinger equation
∂ψ
i~ = Êψ (7.3)
∂t
we decide to promote the energy E to become an operator. That is, let
∂
E → i~ (7.4)
∂t
Using (7.4) and the usual definition the quantum mechanical momentum operator, the Einstein
relation for energy becomes
∂2 ~ 2 + m2 c4
−~2 2 = −~2 c2 ∇ (7.5)
∂t
7.2. THE KLEIN-GORDON EQUATION AND SCALAR FIELDS 119
Applying (7.5) to a function of space and time, φ(~x, t) and using natural units, we have
∂2φ ~ 2
− ∇ φ + m2 φ = 0 (7.6)
∂t2
which also takes the form
( + m2 )φ = 0 (7.7)
where
∂2 ~2
≡ −∇
∂t2
Since both and m2 are scalars, the operator ( + m2 ) is a scalar as well. Therefore, the
Klein-Gordon equation is said to apply to scalar fields, which have been found to represent spin-0
particles [35]. We may also write the Klein-Gordon equation as
(∂µ ∂ µ + m2 )φ = 0 (7.8)
As written, the Klein-Gordon equation describes a free particle. Therefore it has a classical
plane-wave solution, namely
φ(~x, t) = e−ip·x
Remember that when we are working in relativity p and x are actually 4-vectors and therefore
the scalar product is given by
p · x = pµ xµ = Et − p~ · ~x
Notice then
∂φ ∂
= e−i(Et−~p·~x) = −iEφ
∂t ∂t
∇φ = ∇e−i(Et−~p·~x) = i~
pφ
Together, we find p
(E 2 − p~2 − m2 ) = 0 → E = ± p~2 + m2 (7.9)
where we have kept both signs of the energy for a reason. We will return to the consequences of
(7.9) shortly. A typical analysis of a classical field equation often uses the Fourier transformation
of the scalar field φ(x)
dD p ip·x
Z
φ(x) = e φ(p) (7.10)
(2π)D
where φ(p) is the the Fourier transform of φ(x). In classical field theory, the plane wave-solution is
expected to be real, φ(x) = φ∗ (x), allowing us to write
dD p ip·x dD p −ip·x ∗
Z Z
e φ(p) = e φ (p)
(2π)D (2π)D
If we change the integration variable on the left hand side of the equation, p → −p, we find
dD p −ip·x
Z
e (φ(−p) − φ∗ (p)) = 0
(2π)D
Applying the Klein-Gordon equation yields
dD p
Z
(−p2 − m2 )φ(p)eipx = 0
(2π)D
Since this relation holds for all values of x, we require that
(p2 + m2 )φ(p) = 0 (7.11)
for all p. In momentum space, the hypersurface p2 + m2 = 0 is called the mass-shell. Where
pµ = (E, p~), the mass-shell is therefore the locus of points in momentum space where E 2 = p~2 + m2
[72].
Let’s now return to (7.9). Equation (7.9) implies that the energy of the particle takes on both
positive and negative energy states. The issue of negative energy states turns out to not be a
problem, as it is also a consequence of imposing causality in our theory, which we desire. There
is another issue however which also troubled Schrödinger when he arrived to the Klein-Gordon
equation. The Klein-Gordon equation leads to a negative probability density in the free particle
case. To see this explicitly, consider one spatial dimension for simplicity and assume that the
probability current takes the usual form [59]:
∂φ ∂φ∗
J = −iφ∗ + iφ (7.12)
∂x ∂x
Taking the spatial derivative of the probability current yields
∂J ∂2φ ∂2φ
= −iφ∗ 2 + iφ 2
∂x ∂x ∂x
Using the Klein-Gordon equation in one dimension
∂2φ ∂2φ
2
= 2 + m2 φ
∂x ∂t
we find
∂2φ ∂2φ ∂2φ ∂ 2 φ∗

∂J
= −iφ∗ 2 + iφ 2 = −i φ∗ 2 − φ 2 (7.13)
∂x ∂x ∂x ∂t ∂t
A fundamental result from quantum mechanics is that the probability density ρ and the proba-
bility current J satisfy the conservation of probability equation [49]:
∂ρ ∂J
+ =0 (7.14)
∂t ∂x
Hence,
2
∂ 2 φ∗

∂ρ ∗∂ φ
=i φ −φ 2
∂t ∂t2 ∂t
Leading to
∂φ∗

∗ ∂φ
ρ=i φ −φ (7.15)
∂t ∂t
7.3. QUANTIZATION OF FREE SCALAR FIELDS 121
Using our plane wave solution, we find
∂φ∗

∗ ∂φ
ρ=i φ −φ = 2E (7.16)
∂t ∂t
p
But E = ± p~2 + m2 . Therefore, allowing the negative energy solution to exist yields a negative
probability density p
ρ = −2 p~2 + m2 < 0
which doesn’t make any sense at all! A first step to solve this problem is to quantize the fields
by promoting φ to become an operator.
7.3 Quantization of Free Scalar Fields

The process of quantizing a field basically involves us imposing commutation relations. Canonical
quantization refers to the process of imposing the fundamental commutation relations in position
and momentum
[x̂, p̂] = i (7.17)
A similar procedure holds for quantizing classical fields. This method is formally known as
second quantization, although it is a misleading name. To quantize fields we must continue to place
space and time on equal footing. In quantum field theory, momentum and position revert back to
parameters, just as they were in ordinary classical mechanics, and instead we promote the fields
to operators, imposing equal time commutation relations on fields and their conjugate momentum
fields. The fields are operators in the sense that they act on quantum states to destroy or create
particles, which is important since particle number is not fixed in relativity theory [45].
But changing the number of particles has its roots in the simple harmonic oscillator from ordinary
quantum mechanics. Let’s briefly review. Recall that the Hamiltonian for a simple harmonic
oscillator in quantum mechanics is [42]:
p̂2 mω 2 2
Ĥ = + x̂ (7.18)
2m 2
Let us define the creation and annihilation operators (also known as the raising and lowering
operators):
r
mω i
â = x̂ + p̂ (7.19)
2 mω
r
† mω i
â = x̂ − p̂ (7.20)
2 mω
Using the commutation relation of x̂ and p̂, as the reader will show, one finds
[â, â† ] = 1 (7.21)
In terms of the creation and annihilation operators, the Hamiltonian takes the form
1
Ĥ = ω(â† â + ) (7.22)
2
We define the number operator as

N̂ = â† â (7.23)
which satisfies the eigenvalue equation,
N̂ |ni = n|ni
The Hamiltonian can then be written as
1
Ĥ = ω(N̂ + ) (7.24)
2
The eigenstates of the Hamiltonian satisfy
1
Ĥ|ni = ω(n + )|ni (7.25)
2
Implying the eigenenergy is
1
En = ω(n + ) (7.26)
2
The annihilation operator has its name because it drops the eigenstate |ni by one unit in the
following way:
√
â|ni = n|n − 1i (7.27)
Alternatively, the creation operator increases the eigenstate |ni by one unit as
√
â† |ni = n + 1|n + 1i (7.28)
There is a lowest lying energy state in quantum mechanics called the ground state or vacuum
state. We enforce the condition that the vacuum state is annihilated:
â|0i = 0 (7.29)
But â† raises the energy of the system without limit. Therefore, we can obtain a generic state
from the vacuum state as
(â† )n
|ni = √ |0i (7.30)
n!
We call the collection of all states spanned by the states formed by operating on the vacuum
state with any number of the creation operators a Fock space [32].
In quantum field theory, we take the notion of the number operator literally. The state |ni is
not a state of a single particle, but rather a state of a field with n particles. The creation operator
adds one particle to the field while the annihilation operator removes one particle from the field.
Moreover, as we will see, the physical vacuum |0i has no particles present, however the fields remain,
indicating that the vacuum state is not entirely void of everything.
Let’s move on to quantizing the free scalar field. For now, consider a real scalar field that satisfies
the Klein Gordon equation
∂2φ
− ∇2 φ + m2 φ = 0 (7.31)
∂t2
7.3. QUANTIZATION OF FREE SCALAR FIELDS 123
The free field solution of the Klein-Gordon equation is
φ(x, t) = e−i(Et−px)
If we use the wave number instead, then we let E → k 0 = ωk and p~ → ~k, allowing us to write
0
−~
φ(x) = e−i(ωk x k·~
x)
(7.32)
Doing this allows us to write the general solution of the Klein-Gordon equation in terms of a
Fourier expansion [35]
d3 k
Z h i
0 ~ 0 ~
φ(x) = 3√ φ(~k)e−i(ωk x −k·~x) + φ∗ (~k)ei(ωk x −k·~x) (7.33)
(2π) 2 2ωk
We now promote the field φ(x) to become an operator by having φ(~k) → â(~k) and φ∗ (~k) → â† (~k).
Therefore, the field operator is given by
d3 k
Z h i
~ −i(ωk x0 −~
k·~
x) † ~ i(ωk x0 −~
k·~
x)
φ̂(x) = 3 √ â(k)e + â ( k)e (7.34)
(2π) 2 2ωk
To impose the commutation relations, we also require a conjugate momentum to the field. The
Klein-Gordon Lagrangian is
1 1
L = ∂µ ∂ µ φ − m2 φ (7.35)
2 2
The conjugate momentum of the field is then
∂L
Π(x) = = ∂0 φ (7.36)
∂(∂0 φ)
A brief calculation yields the conjugate momentum operator:
d3 k
Z h i
~ −i(ωk x0 −~
k·~
x) † ~ i(ωk x0 −~ k·~x)
∂0 φ̂(x) = ∂0 3 √ â(k)e + â (k)e
(2π) 2 2ωk
d3 k
Z h i
~ −i(ωk x0 −~
k·~
x) † ~ i(ωk x0 −~k·~
x)
= 3 √ â(k)(−iω k )e + â ( k)(iω k )e
(2π) 2 2ωk
d3 k
Z r
ωk h ~ −i(ωk x0 −~k·~x) † ~ i(ωk x0 −~ k·~x)
i
= −i 3 â( k)e − â ( k)e (7.37)
(2π) 2 2
The commutation relations we impose follow from the canonical commutation relations from
non-relativistic quantum mechanics:
[x̂i , p̂j ] = iδij (7.38)
[x̂i , x̂j ] = [p̂i , p̂j ] = 0 (7.39)
For fields we impose the equal time commutation relations:
[φ̂(x), Π̂(y)] = iδ(~x − ~y ) (7.40)
[φ̂(x), φ̂(y)] = [Π̂(x), Π̂(y)] = 0 (7.41)

We call these the equal time operators since although ~x 6= ~y , we assume that we looking at the
fields at the same time, i.e. x0 = y 0 . Rather than just stating the equal time commutator (7.40),
let’s explicitly calculate it. The reader should convince themselves that the field and associated
conjugate momentum operators may also take the form
d3 p
Z
p)eipx + a† (~
p)e−ipx

φ(x) = p a(~ (7.42)
3
(2π) 2p 0
r
d3 p p0
Z
∂φ
p)eipx − a† (~
p)e−ipx

Π(x) = = i a(~ (7.43)
∂x0
p
(2π) 3 2
We have dropped the hat notation indicating operators, however it is still implied. Let us now
compute the equal time commutator given in (7.40). Explicitly, the commutator is
[φ̂(x), Π̂(y)] = φ(x)Π(y) − Π(y)φ(x)
The first term is

r
d3 p d3 p0 p00 h ~0 ip0 x
Z Z i
0
ipx † −ipx
− a† (p~0 )e−ip x

φ(x)Π(y) = p a(~
p)e + a (~
p)e i p a(p )e
(2π)3 2p0 (2π)3 2
Collecting terms and using brute force,

s
d3 p d3 p0 1 p00 h
Z
0 0
φ(x)Π(y) = i p p 0
a(~p)a(p~0 )eipx eip y − a(~p)a† (p~0 )eipx e−ip y +
(2π) 3 (2π) 3 2 p (7.44)
0 0
i
p)a(p~0 )e−ipx eip y a† (~
a† (~ p)a† (p~0 )e−ipx e−ip y
Similarly,
s
d3 p d3 p0 1 p00 h ~0
Z
0 0
Π(y)φ(x) = i p p 0
a(p )a(~ p)eipx eip y − a† (p~0 )a(~p)eipx e−ip y +
(2π)3 (2π)3 2 p (7.45)
0 0
i
a(p~0 )a† (~
p)e−ipx eip y a† (p~0 )a† (~
p)e−ipx e−ip y
Now we put everything together, however noting that the creation and annihilation operators
obey the commutation relations
p), a† (p~0 )] = δ(~
[a(~ p − p~0 ) (7.46)
p), a(p~0 )] = [a† (~
[a(~ p), a† (p~0 )] = 0 (7.47)
Using these commutation relations, noting that x0 = y 0 , and δ(p − p0 )f (p) = δ(p − p0 )f (p0 ), we
have that
s
d3 p d3 p0 1 p00 h
Z i
[φ̂(x), Π̂(y)] = i p δ(~
p − ~0 )ei~p(~x−~y) + δ(~
p p − ~0 )e−i~p(~x−~y)
p
(2π)3 (2π)3 2 p0
p
d3 p 1 i~p(~x−~y)
Z
=i p e + e−i~p(~x−~y) (7.48)
(2π)3 2
7.4. CONSTRUCTING THE STATE SPACE FOR SCALAR FIELDS 125
One definition of the Dirac delta distribution is [68]
d3 p
Z
δ(~x − ~y ) = p ei(~x−~y)~p (7.49)
(2π)3
Then, using the above identity and the symmetry of the delta distribution, we have
d3 p 1 i~p(~x−~y)
Z 1
[φ̂(x), Π̂(y)] = i p e + e−i~p(~x−~y) = i (δ(~x − ~y ) + δ(~y − ~x))
(2π) 2
3 2
= iδ(~x − ~y ) (7.50)
Before moving on, let us briefly interpret our choice of commutation relations. We started with a
classical theory which was relativistic. To quantize the theory we imposed equal time commutators.
However, our chosen commutators turn out to not be Lorentz covariant. Therefore, in order to
quantize we must specify a particular Lorentz frame [32]. Moreover, no signal can travel faster
than the speed of light, therefore two events which are separated by a space-like distance cannot
affect each other. The creation and destruction of a particle is such an event, so for our field theory
to remain relativistic, we require the commutator [φ̂(x), φ̂(y)] to vanish when the separation of
the points x and y is space-like. Performing a similar calculation as above, at different times the
commutator takes the form
d3 k
Z
[φ̂(x), φ̂(y)] = p e−ik(x−y) − eik(x−y) ≡ i4(x − y) (7.51)
(2π)3 2ωk
which is in fact Lorentz invariant because of the appearance of the dot products which are
inherently Lorentz invariant. Since 4 is Lorentz invariant, it follows that 4 vanishes for space-like
separations as desired. If this were not true, then signals could propagate faster than the speed of
light, violating one of the tenets of relativity.
7.4 Constructing the State Space for Scalar Fields

Now that we know how to write the field operators in terms of the creation and annihilation
operators, we can see how the operators act on the state of the fields. Due to our understanding of
the simple harmonic oscillator from ordinary quantum mechanics, we already have an idea on how
the operators behave. Let’s begin by considering the vacuum state |0i. Analogous to the harmonic
oscillator, the vacuum state is destroyed by the annihilation operator
a(~k)|0i = 0 (7.52)
where we use the wave vector ~k to notate our states. On the other hand
|~ki = a† (~k)|0i (7.53)
which describes a one-particle state. If we use multiple creation operators of different modes, we
may construct a Fock space
|~k1 , ~k2 , ...~kn i = a† (~k1 )a† (~k2 )...a† (~kn )|0i (7.54)
The accurate interpretation of the action of creation operator is that each creation operator
a† (~ki ) creates a single particle with momentum ~~ki and energy ~ω~ki where
q
ω~ki = ~k 2 + m2
i
Alternatively, the annihilation operator destroys particles with the same momentum and energy.
We can construct the number operator from the creation and annihilation operators analogously
to the number operator defined in non-relativistic quantum mechanics:
N (~k) = a† (~k)a(~k) (7.55)
The eigenvalues of the number operator, n(~k) are called occupation numbers and are integers,
telling us how many particles there are of momentum ~k for a given state. Therefore (7.53) is a state
consisting of n particles, with a single particle with momentum ~k1 , a single particle of momentum
~k2 and so on. We can also have states where there are multiple particles of the same momentum.
Consider for example the state
a† (~k1 )a† (~k1 ) † ~

|~k1 , ~k1 , ~k2 i = √ a (k2 )|0i
2
We may rewrite this state as
|~k1 , ~k1 , ~k2 i = |n(~k1 )n(~k2 )i
where n(~k1 ) = 2 and n(~k2 ) = 1. Therefore,
~ ~
(a† )n(k1 ) (a† )n(k2 )!
|n(~k1 )n(~k2 )i = q q |0i
n(~k1 )! n(~k2 )
In general, the Fock space takes the form [35]:

n ~
Y (a† )n(kj )
|n(~k1 )n(~k2 )...n(~kn )i = q |0i (7.56)
j=1 n(~kj )!
As written, the number N (~k) is actually a number density. To get the total number of particles,
we integrate over all states in momentum space
d3 k
Z
N̂ = p â† (~k)â(~k) (7.57)
(2π)3 2ωk
Moreover, in terms of the number operator, it can be shown that the Hamiltonian and momentum
take the form
d3 k
Z
~ 1
Ĥ = p ωk N̂ (k) + (7.58)
(2π)3 2ωk 2
d3 k
Z
p̂ = p ~k N̂ (~k) + 1 (7.59)
(2π)3 2ωk 2
7.4. CONSTRUCTING THE STATE SPACE FOR SCALAR FIELDS 127
Now that we know how to construct the states of our system, we would like to ensure that they
are normalized. We begin by assuming that the vacuum state is normalized to unity
h0|0i = 1
To compute the normalization of a generic state we proceed by using the creation and annihilation
operators and the commutation relations associated with them. Recall that the creation operator
acts on the vacuum state as
a† (~k)|0i = |~ki
The adjoint of this expression is given by
h~k| = h0|a(~k)
Therefore,
h~k|k~0 i = h0|a(~k)a† (~k)|0i = h0|a† (~k)a(~k) + δ(~k − k~0 )|0i

= h0|a† (~k)a(~k)|0i + h0|δ(~k − k~0 )|0i = δ(~k − k~0 )
Therefore,
h~k|k~0 i = δ(~k − k~0 ) (7.60)
With our assumption on the normalization of the vacuum, notice what happens when we try to
compute the energy of the vacuum, the sum of the zero point energies of all the oscillators:
d3 k d3 k
Z Z
~ 1 † ~ ~ 1
h0|H|0i = h0| p ωk N (k) + |0i = h0| p a (k)a(k) + |0i
(2π)3 2ωk 2 (2π)3 2ωk 2
d3 k
Z Z
ωk ωk
= d3 kh0|0i = p →∞
2 2 (2π)3 2ωk
since we are integrating over all space. We might think that we are doomed because of this
divergence, however we remember that in physics we don’t measure pure energy, but rather energy
differences. Therefore, we may set the energy of the vacuum state to zero. How we do this is we
renormalize the theory by simply subtracting away the infinity (a seeming mathematical sleight of
hand). We redefine the Hamiltonian by substracting this infinite constant [32]
H → H − h0|H|0i (7.61)
This is formally accomplished by normal ordering the operators. Normal ordering essentially
means that we order the creation and annihilation operators such that the creation operators
appear to the left of the annihilation operators. It is typically denoted by placing colons on both
sides of the operator product:
: a(~k)a† (~k) := a† (~k)a(~k) (7.62)
Then, the normal ordered Hamiltonian is just
d3 k
Z
: H := p ωk a† (~k)a(~k) (7.63)
(2π)3 2ωk
And subsequently,
h0| : H : |0i = 0 (7.64)
In most texts, the Hamiltonian is assumed to be normal ordered.
7.5 Charged Scalar Fields

So far we have only considered real scalar fields which in fact does not allow the distinction between
particles and antiparticles. To distinguish between particles and antiparticles we must have that
the particle has some charge quantum number while the antiparticle has the opposite charge. It is
important to note that by ”charge”, we don’t necessarily mean electric charge, just as discussed in
the last chapter. To incorporate a conserved charge, we need a conserved current. The Klein-Gordon
equation does in fact provide a conserved current, yielding the charge density to be [22]
∂φ ∂φ∗

i
ρ= φ∗ − φ (7.65)
2m ∂t ∂t
Notice that the charge density depends explicitly on φ∗ , the Hermitian conjugate of φ. Therefore,
if φ is Hermitian (i.e. real), the charge density ρ vanishes. Thus, to incorporate a charge into our
theory, we must work with a non-Hermitian field, that is, a complex scalar field.
When we introduce antiparticles, it is common to use a† (~k) and a(~k) for particles, and b† (~k)
and b(~k) for antiparticles. With this notation, a† (~k) can be interpreted as creating a particle with
momentum ~~k and energy ~ωk , while b† (~k) creates an antiparticle with the same momentum and
energy. A similar interpretation holds for a(~k) and b(~k).
We then write the field operators using a positive and negative frequency decomposition, asso-
ciating the particles with the positive frequency portion of the field and the antiparticles with the
negative frequency portion of the field [35], yielding
d3 k
Z
0 ~ 0 ~
φ̂(x) = p â(~k)e−i(ωk x −k·~x) + b̂† (~k))ei(ωk x −k·~x) (7.66)
(2π)3 2ωk
The adjoint field is given by taking the Hermitian conjugate (which, since we are dealing with
scalar fields, is equivalent to the complex conjugate) of φ̂(x):
d3 k
Z
0
−~ 0
−~
φ̂† (x) = p â† (~k)ei(ωk x k·~
x)
+ b̂(~k))e−i(ωk x k·~
x)
(7.67)
(2π)3 2ωk
The conjugate momentum associated with the fields are Π̂(x) = ∂0 φ̂(x) and Π̂† (x) = ∂0 φ̂∗ (x).
Analogous to the real scalar field, we have similar equal time commutation relations
[φ̂(x), Π̂(y)] = [φ̂∗ (x), Π̂∗ (y)] = iδ(~x − ~y ) (7.68)

For a charged complex field, we have two number operators to consider. First, we have the
familiar number operator that corresponds to the number of particles
d3 k
Z
N̂a = p â∗ (~k)â(~k) (7.69)
(2π)3 2ωk
The second corresponds to antiparticles
d3 k
Z
N̂b = p b̂† (~k)b̂(~k) (7.70)
(2π)3 2ωk
7.6. TIME-ORDERING AND THE PROPAGATOR 129
The total energy and total momentum are then given by

d3 k
Z
Ĥ = p ωk (â† (~k)â(~k) + b̂† (~k)b̂(~k)) (7.71)
(2π)3 2ωk
d3 k
Z
p̂ = p ~k(â† (~k)â(~k) + b̂† (~k)b̂(~k)) (7.72)
(2π)3 2ωk
Using the methods laid out in the last chapter, we can compute the conserved charge using
normal ordering
Z Z
Q = d3 x : ρ := d3 x : (φ̂† Π̂ − Π̂† φ̂) : (7.73)
Subsitituting in the expansions for the fields (7.66), (7.67), the charge can be written as
d3 k
Z
Q= p (â† (~k)â(~k) − b̂† (~k)b̂(~k)) = N̂a − N̂b (7.74)
(2π)3 2ωk
Therefore, type “a” particles are created by φ̂∗ and destroyed by φ̂ and have a charge of +1, while
“b” particles are created φ̂ and destroyed by φ̂∗ , carrying a charge of −1. Moreover, in any state
|na nb i where nb > na , the total charge is less than zero. Therefore, we reinterpret the probability
density as the charge density, avoiding the issue of negative probability densities discussed earlier.
7.6 Time-Ordering and the Propagator

Let’s begin by considering an ordinary quantum mechanical propagation of a particle at x =
(~x, t) to x0 = (x~0 , t0 ). Let the state describing the particle located at x be |ψ(~x, t)i and the state
corresponding to the particle located at x0 be |ψ(x~0 , t0 )i. The quantum mechanical amplitude for
the particle to start at x and propagate to x0 is the overlap between the two states
hψ(x~0 , t0 )|ψ(~x, t)i
Now let’s consider the propagation of charge in the charged scalar field theory. The state corre-
sponding to the particle of charge +1 at x is φ̂∗ (x)|0i, while the state corresponding to the particle
at x0 is φ̂∗ (x0 )|0i. The amplitude to transport the charge from x to x0 is then
h0|φ̂(x0 )φ̂∗ (x)|0i (7.75)
We may interpret the propagation of charge as the creation of a particle of +1 charge out of
the vacuum at x, it is then transported from x to x0 where it is reabsorbed into the vacuum.
Remember, when we introduced relativity we inevitably introduced causality, therefore we cannot
have a particle destroyed until it is created. Thus, for the sake of causality we require that t0 > t.
This is not the total amplitude for the propagation of a +1 charge. The total amplitude in fact is
the sum of all of the amplitudes of different processes that give equivalent physical results [22]. A
process which gives the same physical result is the creation of a particle with −1 charge at x0 and
is transported to x where it is then annihilated. The amplitude describing this process is given by
h0|φ̂∗ (x)φ̂∗ (x0 )|0i (7.76)

Again, if we impose causality, for this scenario, we require that t > t0 . We write the total
amplitude for this propagation, G(x, x0 ) as the sum of these two amplitudes
G(x0 , x) = Θ(t0 − t)h0|φ̂(x0 )φ̂∗ (x)|0i + Θ(t − t0 )h0|φ̂∗ (x)φ̂∗ (x0 )|0i (7.77)
Often we write this using the Dyson time ordering product:
φ(x)φ∗ (x0 ) : t0 < t

T φ(x)φ∗ (x0 ) =
φ∗ (x0 )φ(x) : t < t0
The operator T orders the operators by their time-ordering. Operators occurring at later times
appear on the left of the operators that occur at earlier times. Using the time ordering operator T ,
the propagator is
G(x0 , x) = h0|T φ(x)φ∗ (x0 )|0i (7.78)
In classical field theory, the propagators are Green’s functions. The same is true in quantum
field theory. With a bit of work, it can be shown that the propagator takes the form [22]
0
d4 k e−ik(x −x)
Z
G(x0 , x) = lim+ i ≡ i4F (x0 − x) (7.79)
→0 (2π)4 k 2 − m2 + i
In this form, one may check that 4F (x0 − x) satisfies
0
(∂µ0 ∂ µ + m2 )4F (x0 − x) = −δ (4) (x0 − x)
indicating that G(x0 , x) is a Green’s function for the Klein-Gordon equation (Note: for the reader
who has not seen Green’s functions before, don’t worry as we will not focus on these functions in
this text; we simply refer to the fact that the propagator is such a special function).
In essence, 4F (x0 − x) describes the propagation of a particle from x to x0 when t0 > t, and
the propagation of an antiparticle from x0 to x when t > t0 . If we fix t0 > t, 4F (x0 − x) will only
propagate positive energy states forward in time and negative energy states backward in time. Thus,
antiparticles of positive energy propagating forward in time can be interpreted as negative energy
particles propagating backward in time. To observe this, consider the simple Feynman diagram in
figure 7.1.
7.7. LIGHT-CONE COORDINATES AND SCALAR FIELDS 131
Figure 7.1: A simple Feynman diagram of a particle and anti-particle pair exchanging a Z 0 boson.
Notice the direction of time: antiparticles of positive energy propagating forward in time can be
interpreted as negative energy particles propagating backward in time.
7.7 Light-Cone Coordinates and Scalar Fields

It will be useful later on to consider scalar fields in terms of light-cone coordinates. Let ~xT denote
a vector whose components are transverse coordinates xI
~xT = (x2 , x3 , ...xd ) (7.80)
The collection of space-time coordinates then becomes (x+ , x− , ~xT ). Using light-cone coordinates,
the Klein-Gordon equation is written as

∂ ∂ ∂ ∂
−2 + − + − m φ(x+ , x− , ~xT ) = 0
2
(7.81)
∂x ∂x ∂xI ∂xI
To simplify, we Fourier transform the spatial dependence of the field, changing x− into p+ and
x into pI . Similarly then,
I
p~T = (p2 , p3 , ...pd ) (7.82)

The Fourier transform of the field is given by [72]:
Z D−2
dp+
Z
+ − d p~T −ix− p+ +i~xT ·~pT
φ(x , x , ~xT ) = e φ(x+ , p+ , p~T ) (7.83)
2π (2π)D−2
Substituting this into the Klein-Gordon equation, (7.81), we find

∂ I I
−2 + (−ip ) − p p − m φ(x+ , p+ , p~T ) = 0
+ 2
∂x
Dividing by 2p+ we find

∂ 1 I I
i + − + (p p + m ) φ(x+ , p+ , p~T ) = 0
2
(7.84)
∂x 2p
To make things even simpler, let’s introduce a new time parameter, τ , which is related to x+ in
the following way:
p+ τ
x+ = (7.85)
m2
allowing us to write

∂ 1 I I
i − (p p + m ) φ(τ, p+ , p~T ) = 0
2
(7.86)
∂τ 2m2
Before moving on, let’s briefly consider what we have just shown. We started with the Klein-
Gordon equation, a second order equation in space and time, and rewrote it using light-cone co-
ordinates to yield a first order differential equation in light-cone time. This expression appears to
have the same structure as the Schrödinger equation, a feature we will exploit in the next chapter.
Moving right along, to describe quantum states of scalar fields in light-cone coordinates we label
the oscillators (creation/annihilation operators) with p+ and p~T . Single particle states are then
constructed via
a†p+ ,~pT |0i (7.87)
In light-cone coordinates, one can show that the momentum operator becomes three operators
[72]:
X
p̂+ = p+ a∗p+ ,~pT ap+ ,~pT (7.88)
p+ ,~
pT
X
p̂I = pI a∗p+ ,~pT ap+ ,~pT (7.89)
p+ ,~
pT
X
p̂− = p− a∗p+ ,~pT ap+ ,~pT (7.90)
p+ ,~
pT
If we use the mass-shell condition (p2 + m2 ) = 0, and the momentum in light-cone coordinates,
yielding p2 = (−p+ p− − p− p+ + pI pI ), we find that
1
2p+ p− = (pI pI + m2 ) → p− = (pI pI + m2 )
2p+
which allows us to rewrite (7.90) as
X 1
p̂− = (pI pI + m2 )a∗p+ ,~pT ap+ ,~pT (7.91)
2p+
p+ ,~
pT
These expansions are important in the analysis of the relativistic quantum point particle which
we consider in the next chapter.
7.8. THE DIRAC EQUATION AND SPINOR FIELDS 133
7.8 The Dirac Equation and Spinor Fields

One the issues with the Klein-Gordon equation is that it does not give the correct spectra of
the hydrogen atom, which was one of the original reasons why Schrödinger abandoned the Klein-
Gordon equation. Moreover, the Klein-Gordon equation included negative energy solutions. We
were able show that these negative energy solutions correspond to antiparticles, but in 1928, the
physics community was unaware of antiparticles. Paul Dirac approached the difficulties of the
Klein-Gordon equation by inventing his own equation, which we will go on to discuss briefly here.
The Klein-Gordon equation is second-order in space and time. Dirac went the other way and
instead chose an equation which is first order in space and time. His reason for this is that it
eliminated the negative probability density which appeared to plague the Klein-Gordon equation.
Moreover, to maintain Lorentz covariance, space and time should be treated on equal footing,
another reason for having both derivatives in space and time be of the same order. A possible
candidate for a relativistic wave equation fitting this description is given by
∂ψ ∂ψ
i = −i~
α· + βmψ (7.92)
∂t ∂~x
We would like the components of ψ to satisfy the Klein-Gordon equation so that the relativistic
relation of energy and momentum for a free particle holds. In the non-relativistic case, ψ can have
two components, spin up and spin down, thereby requiring that αi and β be matrices, making the
∂ ∂
Dirac equation a matrix equation. If we apply the operators E = i ∂t and pj = −i ∂x twice to the
Dirac equation, the Klein-Gordon equation should fall out. If we demand this to be true, one can
show that the matrices must satisfy [22]
{αi , αj } = 2δij (7.93)
{αi , β} = 0 (7.94)
β2 = 1 (7.95)
Moreover, since αi2 = 1, the eigenvalues of αi and β are ±1. It follows then that αi and β must
be traceless:
tr(αi ) = tr(αi β 2 ) = tr(βαi β) = −tr(β 2 αi ) = −tr(αi ) → tr(αi ) = 0

Since the eigenvalues are ±1 and the matrices are traceless, αi and β must be even dimensional.
If we picked two dimensions we would simply choose the Pauli-spin matrices. Our next option is
dimension four with the Dirac matrices:

0 σi
αi = (7.96)
σi 0

I 0
β= (7.97)
0 −I
where the σi ’s are the familiar Pauli-spin matrices. It turns out to be more convenient to work
with the gamma matrices defined as γ 0 = β and γ i = βαi , and they satisfy
{γ µ , γ ν } = 2η µν (7.98)
In terms of the gamma matrices, the Dirac equation is of the form
(i6∂ − m)ψ = 0 (7.99)
where we have employed Feynman’s slash notation, 6∂ ≡ γ µ ∂x∂ µ . Finally, so that we don’t spend
too much time on the Dirac equation, without proof, however well established, the Dirac equation
in fact describes spin- 12 particles, i.e. fermions.
With the Dirac equation we can also find the conserved current that it produces. However, we
must also know the conjugate equation. Taking the complex conjugate of (7.92), we find
∂ψ † ∂ψ †
−i α·
= i~ + mβψ † (7.100)
∂t ∂~x
since αi and β are Hermitian. To recover the form with the gamma matrices, we multiply the
right hand side of (7.100) by 1 = β 2 , in turn causing ψ † to be everywhere multiplied by β. We then
define
ψ̄ = ψ † β = ψ † γ 0 (7.101)
Therefore,
∂ ψ̄ 0 ∂ ψ̄
−i γ =i · ~γ + ψ̄m (7.102)
∂t ∂~x
Or,
(i6∂ + m)ψ̄ = 0 (7.103)

To find the current, we multiply (7.99) by ψ̄, multiply (7.103) by ψ and sum the two, yielding,
∂µ ψ̄γ µ ψ + ψ̄γ µ ∂µ ψ = ∂µ (ψ̄γ µ ψ) = 0 (7.104)

Thus, the conserved current is
j µ ψ̄γ µ ψ (7.105)
Moreover, the charge density is given by
j 0 = ψ† ψ (7.106)
Remember, the Dirac equation describes spin- 12 fermions, but ψ is 4-component column vector
called a spinor. Since we forced the components of ψ to satisfy the Klein-Gordon equation, we,
inevitably, introduced negative energy solutions, one that is spin-up and one that is spin-down.
Therefore, the components of the spinor describes an electron with both spin states, and a positron
(antielectron) with both spin states. Let’s denote the positive energy solution by
ψ+ = e−ip·x u(p) (7.107)
where u(p) is a 4-component spinor. Similarly, we denote the negative energy solution as
ψ− = eip·x v(p) (7.108)
Subsituting both ψ+ and ψ− into the Dirac equation, we find
(6p − m)u(p) = 0 (7.109)

7.8. THE DIRAC EQUATION AND SPINOR FIELDS 135
(6p + m)v(p) = 0 (7.110)

0
For particles at rest, p~ = 0 and therefore p = E = m. The positive energy solution then becomes
(β − 1)u(0) = 0 (7.111)
Giving rise to two independent solutions:

   
1 0
0 1
u1 (0) =  2
   
0 , u (0) = 0 (7.112)

0 0
The solution u1 (0) describes a positive energy particle which is spin up, while u2 (0) describes a
positive energy particle which is spin down. Similarly, for the negative energy solution we have
(β + 1)v(0) = 0 (7.113)
giving rise to
   
0 0
1
0 2 0
v (0) =   , v (0) =  
   (7.114)
1 0
0 1
where v 1 (0) describes a negative energy particle with spin up, and v 2 (0) describes a negative
energy particle that is spin down. For particles not at rest, we solve (7.109) and (7.110) noting
that (6p − m)(6p + m) = p2 − m2 = 0 such that ui (p) = N (6p + m)ui (0),where N is a normalization
constant, satisfies (7.109), and v i (p) = N (−6p + m)v i (0) satisfies (7.110). With a bit of algebra,
one can show that the resulting spinors for the positive and negative energy solutions to the Dirac
equation take the form
   
1 0
 0   1 
u1 (p) =  pz 
 E+m
2
 , u (p) =  E+m
 p− 
 (7.115)
p+ pz
E+m − E+m
 pz   p− 
E+m E+m
 p+  2 − pz 
v 1 (p) =  E+m   E+m 
 1  , v (p) =  0  (7.116)
0 1
1
where p± = px ± ipy , and E = (p2 + m2 ) 2 .
Dirac succeeded in producing a non-negative current density, getting rid of the negative prob-
ability densities that seemed to plague the Klein-Gordon equation, however, he could not get rid
of the negative energy solutions. Faced with the problem of having an energy spectrum which is
unbounded from below, Dirac supposed that all negative energy states, the negative electron sea,
were filled. Since the equation describes fermions, by Pauli’s exclusion principle no two electrons
can occupy the same state. Therefore, the positive energy electrons are prevented from falling into
the negative energy sea because there are no vacancies. Solving this issue implicitly assumed that
one was no longer dealing with a one particle system. Moreover, the ground state with no posi-
tive energy electrons is no longer empty since the negative energy electrons would still be present.
Rather, the bare vacuum, is the vacuum void of positive and negative energy electrons. It turns out
however that the bare vacuum is unstable.
There are quantum fluctuations in the bare vacuum, making it unstable. Therefore, if we start
with an empty vacuum, eventually a fluctuation will create a pair of electrons, one with negative
energy, and one with positive energy, thereby causing the total energy of the new state to be zero.
This system can lower its energy however by letting a particle pair ”fall” to the bottom of the
sea. This means that the bare vacuum is no longer empty, nor is the energy zero since it has been
lowered. These fluctuations continue to occur until the negative energy sea is full. The fluctuations
will no longer be able to create a zero energy pair because there will be no place to put the negative
energy pair (the sea is ”full”). This is what we call the physical vacuum, a vacuum which is stable
against quantum fluctuations. What’s more is the physical vacuum is not empty, as one might
assume.
Particles and antiparticles can be distinguished by their charge. For scalar fields, we introduced
a complex field φ∗ to distinguish between particles and antiparticles. Suppose we did not want
to describe spin- 21 fermions with charge. This is analogous to our theory of real scalar fields. We
therefore wish to make ψ real. This is possible if we work in a different representation of the gamma
matrices. Instead we use the Majorana representation where we take

0 0 σ2 1 iσ3 0
γ = , γ = (7.117)
σ2 0 0 iσ3

2 0 −σ2 3 −iσ3 0
γ = , γ = (7.118)
σ2 0 0 −iσ3
If we multiply γ µ by i, all four matrices are real. The Dirac equation then is
(i6∂ − m)ψ = (iγ µ ∂µ − m)ψ = 0 (7.119)

which has solutions that are real. Therefore, with Majorana fermions, we cannot expect to
distinguish between particles and antiparticles since now we are dealing with a real spinor field.
Finally, a last case to consider is when we are dealing with massless fermions. The Dirac equation
is simply,
i6∂ ψ = 0 (7.120)
By a proper choice of representation of the gamma matrices, we may decouple the 4-component
spinor into 2-component spinors. Such a choice of representation is known as the Weyl representa-
tion or chiral representation, where the gamma matrices take the form

0 0 −1 i iσi 0
γ = , γ = (7.121)
−1 0 0 −iσi
When we restrict ψ to two components, we lose the antiparticles, but with a real four component
we cannot distinguish the particles from antiparticles since there is no charge. We can distinguish
particles from antiparticles through their chirality however. When the projection of spin of the
particle onto the direction of motion is positive, we say that the particle has positive chirality. If
7.9. FREE SPINOR FIELDS 137
the opposite hold, the particle has negative chirality. It turns out that particles and antiparticles
have opposite chirality [22]. The operator which distinguishes chirality is the γ 5 matrix,
γ 5 = γ5 = iγ 0 γ 1 γ 2 γ 3 (7.122)
In the Weyl representation, the chirality operator takes the form

1 0
γ5 = (7.123)
0 −1
If this is your first time seeing spinor fields, chirality, and helicity, don’t be too alarmed, as when
we quantize the bosonic string one doesn’t need to work with fermions. However, in a later chapter
when we examine the basics of supersymmetry, and move on to superstring theories these notions
are fundamental, and will be examined in more detail there.
Now that have explored the Dirac equation in detail, let us move on to quantizing the spinor fields
of the Dirac equation, constructing a relativistic quantum mechanical theory for spin- 21 particles.
7.9 Free Spinor Fields

If we treat the Dirac spinor field as a classical field, we may proceed to canonically quantize it, similar
to how we quantized the free scalar fields. The difference is since we are dealing with fermions, we
impose anticommutators rather than commutators. Recall the Dirac equation presented above
(i6∂ − m)ψ = 0 (7.124)
Just as we did for scalar fields, we promote the fields to operators and, for fermions, impose
equal time anticommutators
{ψα (~x, t), ψβ† (~y , t)} = δαβ δ (3) (~x − ~y ) (7.125)
{ψα (~x, t), ψβ (~y , t)} = {ψα† (~x, t), ψβ† (~y , t)} = 0 (7.126)
The solutions to the Dirac equation give plane waves, which take the form
2
d3 p m X
Z
ψ̂(~x, t) = (ai (p)ui (p)e−ip·x + b†i (p)v i (p)eip·x ) (7.127)
(2π)3 E i=1
2
d3 p m X †
Z
†
ψ̂ (~x, t) = (a (p)ui† (p)γ 0 eip·x + bi (p)v i† (p)γ 0 e−ip·x ) (7.128)
(2π)3 E i=1 i
In order for our anticommutation relations to hold, we require that
E (3)
{aα (p), a†β (p0 )} = (2π)3 δ (~p − p~0 )δαβ (7.129)
m
E (3)
{bα (p), b†β (p0 )} = (2π)3 δ (~p − p~0 )δαβ (7.130)
m
Similar to the quantized scalar fields, the a†i are the creation operators for the positive energy
electrons. The indices i stand for the electron’s spin: i = 1 corresponds to a spin-up electron, while
i = 2 corresponds to a spin-down electron. Alternatively, the b†i are the creation operators for the
positive energy antielectrons.
We are in a position to go on and rigorously define the physical vacuum, the state where all the
negative energy states are filled. Let |0i denote the physical vacuum, and let |0ib denote the bare
vacuum. Then, since bi (p) creates negative energy electrons, the physical vacuum is given by
2 Y
Y
|0i = bi (p)|0ib (7.131)
i=1 p
Moreover, since the physical vacuum has all of the negative energy states filled, it follows then
bi (p0 )|0i = 0 (7.132)
Alternatively, b†i (p0 )|0i destroys a negative energy particle, creating a positive energy positron.
Using the creation and annihilation operators we may also wrtie the Hamiltonian as
2 Z
d3 p
m(a†i (p)ai (p) − bi (p)b†i (p))
X
Ĥ = (7.133)
i=1
(2π)3
In this form, we again run into the problem where the expectation value of the Hamiltonian
on |0i, i.e., the energy of the physical vacuum, diverges. To fix this issue, we normal order our
Hamiltonian, yielding, [22]
2 Z
d3 p
m(a†i (p)ai (p) + b†i (p)bi (p))
X
: H := (7.134)
i=1
(2π)3
which implies that the physical vacuum is a state of lowest energy for the system to be stable.
Finally, we may write the conserved charge in terms of the creation and annihilation operators
2 Z
d3 p m †
(a (p)ai (p) − b†i (p)bi (p))
X
Q=
i=1
(2π)3 E i
2 Z
X d3 p m
= (Ni (p)+ − Ni− (p)) (7.135)
i=1
(2π)3 E
where Ni+ are the number operators for positive energy electrons, while Ni− are the number
operators for the positive energy positrons.
7.10 The Dirac Propagator

We construct the propagator for fermions in the same manner as we did for spin-0 particles. The
amplitude for a positive energy electron to propagate from x to x0 is given by
hψ(x0 )|ψ(x)i (7.136)

7.11. LIGHT-CONE COORDINATES AND PHOTON STATES 139
The state |ψ(x)i describes one positive energy particle. On the other hand, the state containing
one electron at position x is ψ † (x)|0i. The amplitude is then
h0|ψ(x0 )ψ † (x)|0i (7.137)

Similar to that of a charged scalar particle, the propagation process is interpreted as a positive
charged electron being created out of the physical vacuum at point x, only to be annihilated at x0 .
Just as last time, the same process can also be described by the amplitude
h0|ψ † (x)ψ(x0 )|0i (7.138)
which may interpreted as creating a positron at x0 , where it is then destroyed at x. We then

have time ordering to account for both processes
iSF (x0 , x)γ 0 = h0|ψ(x0 )ψ † (x)|0iΘ(t0 − t) − h0|ψ † (x)ψ(x0 )|0iΘ(t − t0 ) (7.139)

Or,
iSF (x0 , x) = h0|T ψ(x0 )ψ̄(x)|0i (7.140)

where
ψ(x)ψ(x0 ) : t0 < t

0
T ψ(x)φ(x ) =
−ψ(x0 )ψ(x) : t < t0
Before moving on, we briefly summarize the past two sections. In 1928, Paul Dirac attempted
to overcome the seeming issues of the Klein-Gordon equation, inventing his own relativistic wave
equation, the Dirac equation. We saw that this equation included spinors, which we went on to write
in a variety of representations, each with their own benefits. We also established that the Dirac
equation in fact describes spin- 21 fermions. Moreover, when we constructed the Dirac equation, we
saw that it inherently introduced negative energy solutions, thereby including both particles and
antiparticles. Dirac invented this notion of the negative energy sea to fix the issue of an energy
spectrum that is unbounded from below, giving rise to the physical vacuum which is stable from
quantum fluctuations and is, contrary to its name, is not empty.
We proceeded in the usual way to quantize the spinor fields, promoting the spinor fields to
operators, and imposing equal time anticommutation relations. With our expansions and anti-
commutation relations were when able to construct the physical vacuum out of the bare vacuum,
also allowing us to show that the expectation value of the normal ordered Hamiltonian acting
on the physical vacuum is state of lowest energy, and is thereby stable. Lastly, we briefly exam-
ined the Dirac propagator and time-ordered operator for fermions, which follows similarly as the
time-ordered operator for scalar particles.
7.11 Light-Cone Coordinates and Photon States

An introduction on the quantization of free fields is incomplete without the free electromagnetic
field. However, due to gauge invariance, the quantization procedure of the free EM field is far more
involved than the other two procedures, and, frankly, isn’t worth the trouble considering that we
will not need the results to move forward in our study of string theory. Nonetheless, the reader
who chooses to persue this subject will need to know this quantization procedure, and are therefore
pointed to the bibliography for wonderful sources providing detail on this subject. We will however
assume that the electromagnetic field can be quantized, and make use of some of the results that
one would find if they went through the entire procedure.
Since it will be useful later on, let us find the quantum states describing a photon using light-cone
coordinates. The gauge field Aµ can then be broken up as {A+ (p), A− (p), AI (p)}. Then, consider
the gauge transformations [60]:
δA+ = ip+ δA− = ip− δAI = ipI (7.141)

+
These gauge transformations above make it clear that A may be set zero with the correct choice
of epsilon. Indeed 0
A+ → A+ = A+ + ip+
+ 0
Therefore, if = i A
p+ , then A
+
= 0. Therefore, we make our defining light-cone gauge
A+ (p) = 0
With a little work, the fields equations of Maxwell’s theory can be written as
∂ν F µν → p2 Aµ − pµ (p · A) = 0 (7.142)
under the gauge transformation δAµ = ∂µ . Therefore, using our light-cone gauge condition
(7.142) is recast as
p+ (p · A) = 0 ⇒ p · A = 0 (7.143)
Put another way,
−p+ A− − p− A+ + pI AI = 0 (7.144)
And since A+ = 0, we find that
1 I I
A− = (p A ) (7.145)
p+
Now let us briefly describe the photon states in terms of the light-cone gauge. Each of the
independent classical fields AI can be expanded in the same way as scalar fields. Therefore, by
analogy we introduce the creation and annihilation operators aI† I
p and ap , where the subscripts p
+
represent the values of p and p~T . The one photon states are then written as
aI† pT |0i
p+ ,~ (7.146)
The label I is the polarization label. In D-dimensions, we have (D-2) possible polarizations, and
hence have (D-2) linearly independent one-photon states. For example, when D = 4 we have two
single one photon states, one for each transverse direction. In general, a one photon state is actually
a linear superposition of the above states [60]
D−1
X
I aI† pT |0i
p+ ,~ (7.147)
I=2
7.12. GRAVITATIONAL FIELDS AND GRAVITONS 141
where I is the transverse polarization vector.
At this point we have discussed the basic elements of quantum field theory. We have examined,
briefly, the quantization of free scalar fields, free spinor fields, and the electromagnetic field, in
light-cone coordinates. Therefore, we now have a relativistic quantum theory which describes spin-
0 particles, spin- 21 fermions, and spin-1 photons. We have yet to introduce a theory which combines
each of these fields. That is, we have yet to discuss interacting field theories. We will not cover
interacting fields as it goes beyond the scope of this text. The real triumph of quantum field
theory however lies in its ability to describe the interactions of these particles, giving rise to the
Feynman rules and Feynman diagrams that nearly every person, even foreign to physics, has seen.
However, to save time, we will not go into the details of interacting field theories, as it is not crucial
in understanding the fundamentals of quantizing the relativistic strings. For a more advanced
treatment of string theory however, it is necessary that one is at least familiar with interacting
field theories, and we therefore highly recommend that the reader take a look at the references for
further details on this topic.
7.12 Gravitational Fields and Gravitons

Finally, let us briefly discuss how we can construct a state space for gravitons that way we can
recognize them in the theory when we quantize the string. Gravitation emerges in string theory in
the language of Einstein’s general relativity. This time however the dynamical field variable is the
space-time metric gµν (x). Often one makes the perturbative approach for weak gravitational fields
by
gµν (x) = ηµν + hµν (x) (7.148)
That is, we assume the background space-time is mostly flat, however with some flucutations
given by hµν (x). This is actually the method used when studying gravitational waves. It turns
out that the field equations for gµν can be used to derive a linearized equation of motion for the
fluctuations hµν . In the abscence of the sources, the resulting equation is [60]
∂ 2 hµν − ∂α (∂ µ hνα + ∂ ν hµα ) + ∂ µ ∂ ν h = 0 (7.149)

where hµν ≡ η µα η νβ hαβ and h ≡ η µν hµν . Then, if we define hµν (p) to be the Fourier transform
of hµν (x), the momentum space version of the linearized equation of motion becomes
S µν (p) = p2 hµν − pα (pµ hνα + pν hµα ) + pµ pν h = 0 (7.150)

Since every term here contains two derivatives, the flucutations hµν are associated with massless
excitations [60].
Let us consider the following gauge transformation
δ0 hµν (p) = ipµ ν (p) + ipν µ (p) (7.151)

Here the infinitesimal gauge parameter µ (p) is a vector. Let’s first show that (7.150) is invariant
under this gauge transformation. First notice that
δ0 h = ηµν δ0 hµν = iηµν (pµ ν + pν µ ) = 2ip ·

The resulting variation on S µν is
δS µν = ip2 (pµ ν + pν µ ) − ipα pµ (pν α + pα ν ) − ipα pν (pµ α + pα µ ) + 2ipµ pν p ·

which can be written as
δ0 S µν ip2 (pµ ν + pν µ ) − ipµ pν (p · ) + 2ip2 pµ ν − ipµ pν (p · ) − ip2 pν µ + 2ipµ pν p · = 0 (7.152)
Therefore the gauge transformation (7.151) indeed leaves our equation of motion S µν invariant.
We can also write the flucutations hµν using light cone coordinates. Explicitly, the metric can
be broken up as (hIJ , h+I , h−I , h+− , h++ , h−− ). The gauge we will consider is to set all indices
with + to zero. To do this, we use (7.151) in light-cone coordinates, yielding
δ0 h++ = 2ip+ + (7.153)
δ0 h+− = ip+ − + ip− + (7.154)
δ0 h+I = ip+ I + ipI + (7.155)

+ ++
From here, we readily see that a particular choice of will allow us to gauge away h , thereby
fixing + . Similarly, choices of − and I gauge away h−+ and h+I . Altogether then, the light-cone
gauge for the gravitational field becomes
h++ = h+− = h+I = 0 (7.156)
The remaining degrees of freedom are (hIJ , h−I , h−− ). Notice now what happens to the equation
of motion. First considering the case when µ = ν = +, we find
(p+ )2 h = 0 (7.157)
Remember that in the light-cone gauge we had that p+ 6= 0. Therefore, (7.157) implies h = 0.
More explicitly,
h = ηµν hµν = −2h+− + hII = 0 → hII = 0

This implies that all of the diagonal elements of the matrix hIJ is traceless. Thus, with h = 0,
the equation of motion reduces to
S µν = p2 hµν − pµ (pα hνα ) − pν (pα hµα ) = 0 (7.158)
Furthermore, notice that when µ = +,we have that
p+ (pα hνα ) = 0 → pα hνα = 0
The final form of the equation of motion is
p2 hµν = 0 (7.159)
Let’s further explore pα hνα = 0. Consider when ν = I. Then,
7.13. EXERCISES 143
pα hIα = −p+ hI− − p− hI+ + pJ H IJ = 0

leading to
1
hI− = pJ hIJ (7.160)
p+
Similarly, if we let ν = −, we find that
1
h−− = pI h−I (7.161)
p+
To obtain the graviton states, each of the independent classical fields hIJ is expanded in terms of
IJ†
creation and annihilation operators aIJ pT and ap+ ,~
p+ ,~ pT . A one graviton state is written analogously
as a one spin-0 particle state [60]:
aIJ† pT |0i
p+ ,~ (7.162)
which describes a single graviton with momentum (p+ , p~T ). A multi-graviton state is given by
D−1
X
I,J aIJ† pT |0i
p+ ,~ (7.163)
I,J=2
with II = 0. Here IJ is the graviton polarization tensor. As we will see later, graviton states
correspond to closed string theories.
We made it! As mentioned at the beginning of this endeavor, this chapter is meant to provide a
little insight on quantum field theory, a theory of physics which is fundamental to the study of string
theory or any other advanced physical theories requiring a quantum mechanical interpretation. We
have only scratched the surface, and the interested reader should visit the end of the text for
references providing more details than we were able to do here. To summarize, in this chapter
we have discussed the basics of quantum field theory: free scalar fields, free spinor fields, and free
photon fields. We also went on to construct scalar particle states, photon states and graviton states
in light-cone coordinates, which will be very useful later on when we quantize the string using the
light-cone gauge. Now that we have a basic understanding of the quantum theory of fields we are
ready to take on the relativistic quantum point particle.
7.13 Exercises
1. Using the canonical commutation relations of X and P , show that the creation and annihilation
operators associated with the simple quantum harmonic oscillator satisfy the commutation relation
[a, a† ] = 1.
2. Consider a simple quantum harmonic oscillator. Calculate hai, ha† i, ∆X, and ∆P in an
arbitrary eigenstate |ni. Show that ∆X∆P obeys the Heisenberg uncertainty relation. Show that
in the ground state the uncertainty bound is saturated.
3. Here the reader solves for the eigenfunctions of the harmonic oscillator. (a) Start by projecting
the equation defining the ground state of the oscillator
a|0i = 0
in the X basis. Solve the differential equation, and show that one finds the ground state
mωx2
mω 1/4
ψ0 (x) = exp −
π~ 2~
where ~ has been restored.
(b) Deduce that

(a† )n
|ni = |0i
(n!)1/2
(c) By projecting part (b) on the X-basis, show that one arrives to the general eigenstate of the
harmonic oscillator:
1/4
mωx2

mω mω 1/2
ψn (x) = exp − Hn x
π~22n (n!)2 2~ ~
where Hn are Hermite polynomials.
4. Using the X basis, show that ψ0 (x) saturates the bound given in the Heisenberg uncertainty
relation. This actually comes from a general result which says Gaussian wavepackets will saturate
this bound in uncertainty.
5. Prove that the other equal time commutation relations for scalar fields are satisfied with the
fields and conjugate momenta found.
6. Consider the Hermitian matrices M 1 , M 2 , M 3 , M 4 that obey {M i , M j } = 2δ ij I for i, j =

1, ...4. (a) Show that the eigenvalues of M i are ±1. (Hint: Try going to the eigenbasis of M i , and
use the above relation when i = j).
(b) Consider when i 6= j. Show that M i are traceless. Using this result and part (a), show that
i
M cannot be odd dimensional matrices. The proof follows similarly for the Dirac gamma matrices,
however one must consider separately the time component and spatial components of the flat space
metric ηµν .
7. In the next two problems, the reader will derive the Dirac equation in a slightly different light,
and examine its solutions. An equivalent strategy to deriving the Dirac equation is to “factor”
Einstein’s energy momentum relation:
pµ pµ − m2 c2 = 0
(a) Start with the case where all of the spatial components of the 4-momentum are zero, p~ = 0,
and show that the momentum-energy relation may be factored into
7.13. EXERCISES 145
(p0 + mc)(p0 − mc) = 0

Using this as motivation, when one includes the spatial components, the momentum-energy
relation can be factored as
pµ pµ − m2 c2 = (β k pk + mc)(γ λ pλ − mc)
(b) Factor out the right hand side of the above. Since linear terms in pk are undesired, show
that β k = γ k , and therefore
pµ pµ = γ k γ λ pk pλ
.
(c) Write out both sides of this equation in explicit components using the convention ηµν =
Diag(1, −1, −1, −1). Once factored, show that for the equation to be satisfied the components γ k
must be matrices which satisfy the definition of the Dirac gamma matrices given in (7.98).
(d) Almost done! Using parts (a)-(c), factor the energy-momentum relation. The Dirac equation
is found by considering only one of these factors. The convention is to take (γ µ pµ − mc) = 0. Make
the correct substitution for the operator pµ and show that what follows matches the Dirac equation
as given in (7.99).
8. Here the reader will examine the solutions of the Dirac equation derived in the previous
problem. (a) Consider when the particle is at rest, and restore ~ and c. Show that the Dirac
equation reduces to
i~ 0 ∂ψ
γ − mcψ = 0
c ∂t
Using the Bjorken-Drell convention for the gamma matrices:

0 1 0 i 0 σi
γ = , γ =
0 −1 −σi 0
Recast the above in matrix form using

ψ ψ
ψA = 1 ψB = 3
ψ2 ψ4
Come up with two differential equations for ψA and ψB , showing that the Dirac equation for the
case p~ = 0 admits four independent solutions:
       
1 0 0 0
2 0 2 1 2 0 −i(mc ~)t 0
2
ψ 1 = e−i(mc ~)t  2 −i(mc ~)t   3 −i(mc ~)t   4
       
0 ψ = e 0 ψ = e 1 ψ = e

0
0 0 0 1
What is the physical meaning of these solutions?
(b) Motivated by the solution to (a), consider the plane wave solutions ψ(x) = ae−ik·x u(k) where
k · x = kµ xµ = kct − ~x · ~r. Putting this into the Dirac equation, show that one attains
(~γ µ kµ − mc)u = 0
(c) Using γ µ kµ = γ 0 k 0 − ~γ · ~k, show that the solution in part (b) becomes
" #
µ (~k 0 − mc)uA −~~k · ~σ uB
(~γ kµ − mc)u = =0
~~k · ~σ uA −(~k 0 + mc)uB
Therefore show that for this equation to be satisfied it must be
(~k · ~σ )2 (~k)2
uA = uA = 0 2 uA
(k 0 )2 − (mc/~) 2 (k ) − (mc/~)2
And thus deduce k µ = ±pµ /~.
(d) Show that for

1 p~ · ~σ 1 c pz
uA = : uB = 0 =
0 p + mc 0 E + mc 2 p+
Show this holds similarly for uB (there should be a total of four similar equations). Therefore,
using this result and the result from (a) derive (7.115) and (7.116). Discuss the physical meaning
of these four solutions. Lastly, don’t completely redo the calculation, but consider how the results
might have changed had we decided that the Dirac equation in problem 7 be the non-conventional
factor.
Chapter 8
Quantizing the Relativistic Point

Particle
8.1 Quantization and Point Particle States

This is the final chapter of prepartion. Here we will quantize the relativistic point particle developed
in chapter 3 using the light-cone gauge. Along the way we will apply our knowledge of conserved
charges and the method of quantization discussed in the last chapter. All in all, a rigorous de-
velopment of the quantization of the relativistic point particle in the light-cone gauge will help us
understand how to go about quantizing the string.
First, recall that the action of the relativistic point particle is
τf
r
dxµ dxν
Z
S = −m −ηµν dτ (8.1)
τi dτ dτ
which may be recast as
Z τf Z τf p
S= Ldτ = −m −ẋ2 dτ (8.2)
τi τi
dxµ
where we used ẋ = dτ . In terms of the Lagrangian, the momentum pµ is simply
∂L mẋµ
pµ = =√ (8.3)
∂ ẋµ −ẋ2
The Euler-Lagrange equations are then easily found to be
dpµ
=0 (8.4)
dτ
Moreover, given the form of the momentum pµ , its easy to see that we have the constraint
p2 + m2 = 0 (8.5)
147
148 CHAPTER 8. QUANTIZING THE RELATIVISTIC POINT PARTICLE
Since we are going to quantize the point particle using the light-cone gauge, we must first define
our choice of gauge. Our definition will be that the light-cone coordinate x+ is proportional to τ
in the following way
1 +
x+ = p τ (8.6)
m2
Using our expression for momentum, consider
m 1 p+
p+ = √ ẋ+ = √ (8.7)
−ẋ 2 −ẋ2 m
yielding
1
ẋ2 = − (8.8)
m2
allowing us to write
pµ = m2 ẋµ (8.9)
Let’s see what we might find from our constraint (8.5). Expanding using light-cone coordinates
we find
p2 + m2 = −2p+ p− + pI pI + m2 = 0
1
⇒ p− = + (pI pI + m2 ) (8.10)
2p
Making use of (8.9), we find that
dx− 1
= 2 p− (8.11)
dτ m
We integrate to find that
p−
x− (τ ) = x−
0 + τ (8.12)
m2
where x−
0 is an integration constant. Similarly one can find that
pI
xI (τ ) = xI0 + τ (8.13)
m2
It is assumed that we know p+ and pI , allowing us to find the complete momentum. Moreover,
we presume to know the integration constant x− I I
0 , and since we know p , we presume to know x .
I − I +
Therefore, the independent dynamical variables for the point particle are (x , x0 , p , p ).
Before we move to quantize the point particle, let us briefly review the two fundamental represen-
tations of quantum mechanics, the Heisenberg and Schrödinger pictures.We use both representations
when we quantize the string, so it is imperative that we become familiar with these ideas.
When one first learns quantum mechanics, they typically learn the Schrödinger representation.
In this representation, the states of the system are time dependent, while the operators are not. In
the Heisenberg picture, the opposite is true: the states are time independent, while the operators
are time dependent. In either representation, the results are equivalent. To transform between both
8.1. QUANTIZATION AND POINT PARTICLE STATES 149
pictures, we use the time evolution operator. Recall that the solution to the Schrödinger equation
is
|ψ(t)i = e−iHt |ψ(0)i (8.14)
|ψ(0)i is the initial state of the system at a fixed time t = 0; therefore it has no time dependence
and can be used in the Heisenberg representation. But |ψ(t)i is time dependent, so it belongs to
the Schrödinger picture. Applying eiHt to both sides, we have
eiHt |ψ(t)i = |ψ(0)i (8.15)

Since the right hand side is time-independent, it must be that the left hand side is time indepen-
dent as well. Therefore, use of the time evolution operator has converted a state in the Schrödinger
picture to a state in the Heisenberg picture. To find the transformation law for operators, we note
that the physical results should be the same in either representation. In particular, the expectation
of any operator Z should be the same in both pictures. Starting in the Schrödinger picture we have
hψ(t)|Zs |ψ(t)i = hψ(0)|eiHt Zs e−iHt |iψ(0)i ≡ hψ(0)|Zh (t)|ψ(0)i (8.16)
where we have defined the time dependent Heisenberg operator as
Zh (t) = eiHt Zs e−iHt (8.17)
In the Heisenberg picture, the states are no longer time dependent, and therefore do not have
to obey the Schrödinger equation. Instead, the dynamics of the system are linked to the time
dependent operators. For consistency between the Schrödinger picture and Heisenberg picture, we
require that the operators satisfy an operator equation of motion. To derive this equation of motion,
consider the small time translation t → t + with 1. Expanding Zh we find
Zh (t + ) = eiH Zh (t)e−iH ≈ (1 + iH)Zh (t)(1 − iH) = Zh (t) + i[H, Zh ] + ...

If we expand the left hand side with a Taylor series to the same order, we find
dZh
Zh (t + ) = Z − h(t) + + ...
dt
Equating powers of yields
dZh
= i[H, Zh ] (8.18)
dt
Notice that if Zh commutes with H, then Z is a constant of motion. To quantize a classical
system in the Heisenberg picture, we elevate the observables to time dependent operators and
specify a Hilbert space containing time independent state vectors [25]. To compute the dynamics
of the system, we must solve the operator equations of motion.
We are now ready to quantize the point particle. As mentioned above, in quantum mechanics
there are two different representations to describe the time evolution of a system: The Schrödinger
picture and the Heisenberg picture. The crucial difference is that in the Schrödinger picture,
the states themselves may be time dependent while the operators are time independent, contrary
to the Heisenberg picture where the operators exhibit time dependence and the states are time-
independent. For the point particle, let us choose our time independent Schrödinger operators to
be the dynamical variables, (xI , x− I +

0 , p , p ). At the same time, let our Heisenberg operators be
the time dependent (by time dependent we actually mean τ dependent) version of the dynamical
variables, i.e. (xI (τ ), x− I +
0 (τ ), p (τ ), p (τ )) [64]. It’s important to note that we have refrained from
using the typical hat notation to denote operators. Since it will be clear when we are using the
operators and the variables, we have decided to make this choice. The unsuspecting reader should
remain weary however, making sure to understand the context in which the the variables are being
used.
Let us go on postulate the commutation relations in accordance with ordinary quantum mechan-
ics. For the Schrödinger operators we have
[xI , pJ ] = iη IJ = iδ IJ (8.19)
[x− +
0 ,p ] = iη −+
= −i (8.20)
and all other commutators are zero. Similarly, the same commutation relations hold for the
Heisenberg operators
[xI (τ ), pJ (τ )] = iδ IJ (8.21)
[x− +
0 (τ ), p (τ )] = iη
−+
= −i (8.22)
+ − −
Again, all other commutators are zero. Any other operators, x , x , p , are defined in terms of
the quantum analog of the dynamical variables. That is,
p+ τ
x+ (τ ) ≡ (8.23)
m2
p− τ
x− (τ ) ≡ x−
0 + (8.24)
m2
1
p− ≡ + (pI pI + m2 ) (8.25)
2p
Notice that p− is not time dependent.
We also wish to find the Hamiltonian as a quantum operator. Remember from the last chapter,
to become an operator, we imposed that the energy E take the form
∂
E→ (8.26)
∂t
Here we wish to do the same thing, however using light-cone coordinates as it will end up being
convenient To see how this is done, let us first refer to non-relativistic quantum mechanics. In this
case, a point particle of energy E and momentum p~ has a wavefunction given by
ψ(t, ~x) = exp (−i(Et − p~ · ~x)) = exp i(p0 x0 + p~ · ~x) = eip·x

(8.27)
One readily sees that this wavefunction satisfies Schrödinger’s equation in one dimension
∂ψ
i = Eψ (8.28)
∂x0
Similarly, with light-cone evolution and light-cone energy ELC , we would expect that, where x+
is light-cone time, Schrödinger’s equation in light-cone coordinates is
∂ψ
i = ELC ψ (8.29)
∂x+
To determine the light-cone energy explicitly, first we must write the wavefunction ψ in light-cone
coordinates. Using (8.27) we find that
ψ(x) = exp i(p+ x+ + p− x− + p2 x2 + p3 x3 )

(8.30)
Equation (8.29) gives us then
∂ψ
i= −p+ ψ → −p+ = ELC
∂x+
However, since −p+ = p− . The light-cone energy is therefore,
p− = ELC (8.31)
Then, analogous to (8.26) we insist that
∂
→ p− (8.32)
∂x+
But our Heisenberg operators are parameterized with τ . Therefore, we would expect that the
+
Hamiltonian H generates τ evolution. Since x+ = pm2τ , we expect that the τ evolution be generated
by
∂ p+ ∂ p+
= 2 + → 2 p− (8.33)
∂τ m ∂x m
Putting all of this together, we postulate that the Heisenberg Hamiltonian for the point particle
in light-cone coordinates takes the form [64]:
p+ (τ ) − 1
H(τ ) = p (τ ) = (pI (τ )pI (τ ) + m2 ) (8.34)
m2 2m2
From our above discussion on the Heisenberg representation, we know that the dynamics of the
system in the Heisenberg picture are governed by
dZh
= i[H, Zh ] (8.35)
dt
Or,
dZh
i = [Zh , H] (8.36)
dt
where Zh is a Heisenberg operator. In our case, we have a total of seven Heisenberg operators
to consider, and each operator will give an equation describing the time evolution of the system.
First we start with p+ (τ ). Using (8.36) we have that
dp+ (τ )
i = [p+ (τ ), H(τ )] = p+ (τ )H(τ ) − H(τ )p+ (τ )
dτ
p+ (τ ) I 1
= (p (τ )pI (τ ) + m2 ) − (pI (τ )pI (τ ) + m2 )p+ (τ )
2m2 2m2
p+ (τ ) I I p+ (τ ) I
= (p (τ )p (τ ) + m2
) − (p (τ )pI (τ ) + m2 ) = 0
2m2 2m2
since [p+ , pI ] = 0. Thus,
dp+ (τ )
i = [p+ (τ ), H(τ )] = 0 (8.37)
dτ
Similarly,
dpI (τ )
i = [pI (τ ), H(τ )] = 0 (8.38)
dτ
Equations (8.37) and (8.38) imply that pI and p+ are constants of motion and commute with
the Heisenberg operator. Now consider the operator xI (τ ). The time evolution equation is simply
dxI (τ ) 1
i = [xI (τ ), H(τ )] = [xI (τ ), (pJ pJ + m2 )]
dτ 2m2
1 1
xI (τ )(pJ pJ + m2 ) − (pJ pJ + m2 )xI (τ ) = [xI (τ ), pJ pJ ]

=
2m2 2m2
Using the identity
[A, BC] = [A, B]C + B[A, C] (8.39)
we find
[xI (τ ), pJ pJ ] = [xI (τ ), pJ ]pJ + pJ [xI (τ ), pJ ] = iδ IJ pJ + ipJ δ IJ = 2ipI
Thus,
dxI (τ ) pI
=i = [xI (τ ), H(τ )] = i 2 (8.40)
dτ m
Finally, since [x− I
0 (τ ), p ] = 0, it is easy to see that
dx−
0 (τ ) 1
i = [x− −
0 (τ ), H(τ )] = [x0 (τ ), (pI pI + m2 )] = 0 (8.41)
dτ 2m2
which implies that x− 0 is a constant of motion. Now let’s move on to the other three operators
defined in (8.23), (8.24), (8.25). First notice that p− is a function of pI only. It is easy to see then
that the commutator with the Hamiltonian H vanishes. That is,
dp−
i = [p− , H(τ )] = 0 (8.42)
dτ
Next consider x− (τ ). By analogy, we have that
dx− (τ )
i = [x− (τ ), H(τ )] = 0 (8.43)
dτ
p− τ
since x− (τ ) ≡ x−
0 + m2 and both x− − I
0 and p commute with p . But, we expected that
dx− (τ ) p−
i =i 2
dτ m
So what went wrong? Looking at the forms of (8.23) and (8.24) we notice that these Schrödinger
operators have explicit time dependence. But how we derive (8.36) is by assuming that the
Schrödinger operator is totally time independent (indeed it could depend on time, however, im-
plicitly). This then changes our analysis of the equation governing the time evolution of the sys-
tem. This time, let Z be a Schrödinger operator with explicit time dependence. Then, using
Zh (t) = eiHt Ze−iHt , we find that

dZh (t) −iHt ∂Z
iHt
= iHe Ze +e iHt
e−iHt − ieiHt ZH −iHt
dt ∂t

−iHt ∂Z −iHt ∂Z
iHt
= ie (HZ − ZH)e +e iHt
e = i[H, Zh (t)] + e iHt
e−iHt
∂t ∂t
Consider the identity
1
eY Ze−Y = Z + [Y, Z] + [Y, [Y, Z]] + ... (8.44)
2!
Taking a first term approximation, we find that
dZh (t) ∂Zh (t)

= i[H, Zh (t)] + (8.45)
dt ∂t
Which may also be written as,
dZh (t) ∂Zh (t)

i =i + [Zh (t), H] (8.46)
dt ∂t
Therefore, going back to our case, we have that
dx− (τ ) ∂x− (τ ) ∂x− (τ ) p−

i =i + [x− (τ ), H(τ )] = = 2 (8.47)
dτ ∂τ ∂τ m
which is what we expected. Similarly,
dx+ (τ ) ∂x+ p+
= = 2 (8.48)
dτ ∂τ m
another expected result.
For a quantum theory of the relativistic point particle, we must develop a state space for the
particle. In general, the time-independent states of a quantum theory are labeled by the eigenvalues
of a maximal subset of commuting operators. In our case, this maximal subset of commuting
operators can include only one element from the pair (x− + I +
0 , p ) and one from (x , p ) [64]. Since it
is usually easier to work in momentum space, we denote the quantum state of the point particle as
|p+ , p~T i (8.49)
where p+ is the eigenvalue of the p+ operator and p~T is the eigenvalue of the pI operator. More
explicitly,
p̂+ |p+ , p~T i = p+ |p+ , p~T i (8.50)

p̂I |p+ , p~T i = pI |p+ , p~T i (8.51)

where we have briefly restored hat notation to denote the operators from their associated eigen-
values.
We also have
1
p̂− |p+ , p~T i = (pI pI + m2 )|p+ , p~T i (8.52)
2p+
Moreover, the Hamiltonian acts on the particle states as
1
(pI pI + m2 )|p+ , p~T i
H|p+ , p~T i = (8.53)
2m2
One readily sees then that the time dependent states take the form

1
exp −i 2 (pI pI + m2 )τ |p+ , p~T i (8.54)
2m
More generally, we may consider the time dependent superposition of basis states as
Z
|ψ, τ i = dp+ d~ pT ψ(τ, p+ p~T )|p+ , p~T i (8.55)
where indeed ψ(τ, p+ p~T ) is the momentum space wavefunction associated with the state |ψ, τ i.
More explicitly, notice that since
hp0+ , p~0 T |p+ , p~T i = δ(p0+ − p+ )δ(p~0 T − p~T ) (8.56)

we have that
hp+ , p~T |ψ, τ i = ψ(τ, p+ p~T ) (8.57)

Moreover, where the Schrödinger equation is
∂
i |ψ, τ i = H|ψ, τ i (8.58)
∂τ
we use (8.55) to find
Z
∂ 1 I I
dp+ d~pT i ψ(τ, p+ p~T ) − (p p + m2
)ψ(τ, p +
p
~ T ) |p+ , p~T i = 0 (8.59)
∂τ 2m2
Since the basis vectors |p+ , p~T i are linearly independent, the expression within the brackets must
vanish for all momenta. Therefore, Schrödinger’s equation becomes
∂ 1
i ψ(τ, p+ p~T ) = (pI pI + m2 )ψ(τ, p+ p~T ) (8.60)
∂τ 2m2
After reading the previous chapter, one easily notices a similarity between the quantum states
of a scalar particle in light-cone coordinates and the quantum states we have been developing here.
Indeed, there is a natural identification of the quantum states of a relativistic point particle of mass
8.2. LIGHT-CONE MOMENTUM OPERATORS AND SYMMETRY TRANSFORMATIONS155
m with the one-particle states of the quantum theory of a scalar field of mass m. That is, we may
write
|p+ , p~T i = a∗p+ ,~pT |0i (8.61)

But why scalar field theory? The answer is because the Schrödinger equation for a point par-
ticle wavefunctions has the form of the field. Part of the correspondence between the quantum
point particle and scalar field is that the classical field equation for the scalar field. In light-cone
coordinates, this equation takes the form

∂ 1 I I 2
i − (p p + m ) φ(τ, p+ p~T ) = 0 (8.62)
∂τ 2m2
Comparing (8.62) to (8.60), we are led to make the identification φ(τ, p+ p~T ) ←→ ψ(τ, p+ p~T ).
That is, the momentum space wavefunctions for the quantum point particle are identified with the
classical plain wave solutions of the scalar field. We therefore have that the scalar field theory de-
scribed in the last chapter in effect describes the free relativistic quantum point particle. Moreover,
now that we know that (8.61) describes the single particle states for the point particle, further
action of the creation operator allows for the description of multiple particle states.
8.2 Light-cone Momentum Operators and Symmetry Trans-

formations
Since the Lagrangian L depends on τ derivatives of the space-time coordinates, it is invariant under
the translation δxµ (τ ) = µ with µ being constant. Using the language we developed in chapter
six, one can easily show that the conserved charge associated with this symmetry transformation
is the momentum pµ . Additionally, in quantum theory conserved charges become quantum op-
erators where, via commutation relations, generate a quantum mechanical version of symmetry
transformations [64].
This is most easily observed if we use a framework in which manifest Lorentz invariance of the
classical theory is preserved after quantization. Unfortunately, in the light-cone gauge we lose
manifest Lorentz invariance since we have chosen to use light-cone coordinates. In a later chapter
we will consider the Lorentz covariant quantization of the string theory, but for now let’s briefly
consider Lorentz covariance of a point particle.
In Lorentz covariant quantization of a point particle, we have the Heisenberg operators xµ (τ )
and pµ (τ ) with commutation relations that remind us of ordinary quantum mechanics
[xµ (τ ), pν (τ )] = iη µν (8.63)
[xµ (τ ), xν (τ )] = [pµ (τ ), pν (τ )] = 0 (8.64)

Let’s now check to see whether pµ (τ ) generates translations. That is, let’s see if iρ pρ (τ ) gives
the translations δxµ = µ . Notice then
δxµ (τ ) = [iρ pρ (τ ), xµ (τ )] = iρ (−iη ρµ ) = µ (8.65)

Thus, the quantum operators, via a commutation relation indeed generate a symmetry transfor-
mation; in this case, a translation. Indeed this is an elegant result, however we are unsure whether
this elegance carries over when we use light-cone gauge quantization. To check, let’s test (8.65)
with light-cone coordinates. For this we expand the generator in light-cone coordinates:
iρ pρ (τ ) = −i− p+ − i+ p− + I pI (8.66)

Particularly, let’s try the case where I 6= 0 and + = − = 0. Then we consider the symmetry
transformation,
δxµ (τ ) = iI [pI , xµ ] (8.67)

J J + −
From here we see that δx (τ ) = and δx (τ ) = δx (τ ) = 0. These are all expected symmetry
transformations had we used the commutation relations given by (8.21) and (8.22) and the explicit
forms of x+ (τ ) and x− (τ ) given in (8.23) and (8.24).
But p− gives rise to further complications since it is a non-trivial function of momentum, shown
in (8.25). Let’s again consider (8.65) except this time let’s use + 6= 0 and − = I = 0. Therefore
we have
δxµ (τ ) = −i+ [p− , xµ (τ )] (8.68)

Notice then,
τ
δx+ (τ ) = −i+ [p− , p+ ]=0 (8.69)
m2
The expected result however is + . Moreover, notice
I
1 +p
δxI (τ ) = −i+ [p− , xI (τ )] = −i+ (−2ip I
) = − (8.70)
2p+ p+
Lastly, consider
p− τ

δx (τ ) = −i p , x0 + 2 = −i+ [p− , x−
− − +−
0] (8.71)
m
Using the fact that
1
p− = (pI pI + m2 )
2p+
we find

− + 1 I I 2 − + 1 I I 2 − − 1 I I 2
δx (τ ) = −i (p p + m ), x0 = −i (p p + m )x0 − x0 + (p p + m )
2p+ 2p+ 2p
And since x− I
0 commutes with p , we have

− 1 I I 2 1 −
δx (τ ) = (p p + m ) + , x0
2 p
But
8.2. LIGHT-CONE MOMENTUM OPERATORS AND SYMMETRY TRANSFORMATIONS157

− 1 1 1 −
x0 , + = x−
0 + − + x0
p p p
1 + − 1 1 1 1 1 i
= p x0 + − + x− p+ + = + [p+ , x−
0 ] + = +2
p+ p p 0 p p p p
Thus,
1 − −i
, x = +2
p+ 0 p
Altogether then,
p−
δx− (τ ) = −+ (8.72)
p+
All (8.69), (8.70), and (8.72) indicate that p− does not generate the expected transformations. It
turns out, however, that p− generates both translations and reparameterizations of the world-line of
the point particle [64]. Earlier we saw that the action is invariant under changes in parameterization.
But symmetries can be exhibited as changes in the dynamical variables. It turns out a change in
parameterization can be described in this way as well. Consider the change in parameterization
τ → τ 0 = τ + λ(τ ) where λ(τ ) is infinitesimal. Therefore we have,
xµ (τ ) → xµ (τ + λ(τ )) = xµ (τ ) + λ(τ )∂τ xµ (τ )

Leading to
δxµ (τ ) = λ(τ )∂τ xµ (τ ) (8.73)

We claim that these are the symmetries of the point particle [64]. Let’s now show that p−
generates a transformation plus a reparameterization. The expected translation was δx+ = + .
But using (8.73) we instead find,
δx+ = λ∂τ x+
Remember (8.69) gives δx+ (τ ) = 0. Therefore the expected translation plus reparameterization
give zero variation
λp+
δx+ (τ ) = + + λ∂τ x+ (τ ) = + + =0
m2
−m2 +
→λ=
p+
Therefore λ is constant. Moreover, notice that for coordinates xI and x− we have from using
(8.73)
−m2 + pI pI
δxI (τ ) = λ∂τ xI (τ ) = +
2
= −+ + (8.74)
p m p
−m2 + p− +p
−
δx− (τ ) = λ∂τ x− (τ ) = = − (8.75)
p+ m2 p+
But this is just what we found in (8.70) and (8.72). Thus for coordinates xI and x− there is no
translation, but rather only a reparameterization.
Lorentz covariant momentum operators generate simple translations and commute with each
other. This fact follows directly from √us using light-cone coordinates and that the corresponding
momentum operators p± = (p0 ± p1 )/ 2 commute with pI . It is important to note that the light-
cone gauge momentum operators are different, however obey the same commutation relations that
the covariant operators do when using light-cone coordinates.
8.3 The Lorentz Charges

Earlier we determined that there are conserved charges associated with the Lorentz invariance of
the relativistic string Lagrangian. Similar charges exist for the relativistic point particle. The
infinitesimal Lorentz transformations of the coordinates of the point particle take the form
δxµ (τ )µν xν (τ ) (8.76)

where µν is antisymmetric. The associated Lorentz charges are then
J µν = xµ (τ )pν (τ ) − xν pµ (τ ) (8.77)
As discussed in the last section, the quantum analog of charges yield symmetry transformations.
Here, the quantum charges are expected to generate Lorentz transformations of the coordinates of
the point particle. This is actually rather straight forward to see using the operators of Lorentz
covariant quantization as developed in the last section, which is one of the benefits of using Lorentz
covariance: the Lorentz symmetry is manifest. In this case, the quantum charges are given by
(8.77) with xµ (τ ) and pµ (τ ) as Heisenberg operators. Notice then
[J µν , xρ (τ )] = J µν xρ (τ ) − xρ (τ )J µν
= (xµ (τ )pν (τ ) − xν (τ )pµ (τ )) xρ (τ ) − xρ (τ )(xµ (τ )pν (τ ) − xν (τ )pµ (τ ))
= xν (τ )[xρ (τ ), pµ (τ )] − xµ (τ )[xρ (τ ), pν (τ )] = iη µρ xν (τ ) − iη νρ xµ (τ ) (8.78)
This helps us check that the quantum Lorentz charges indeed generate Lorentz transformations:

i
δxρ (τ ) = − µν J µν , xρ (τ )
2
1 1 1
= µν (η µρ xν (τ ) − η νρ xµ (τ )) = ρν xν (τ ) + ρµ xµ (τ )
2 2 2
= ρν xν (τ ) (8.79)
Moreover, (8.78) can used with light-cone indices:
[J −I , x+ (τ )] = iη −+ xI (τ ) − iη I+ x− (τ ) = −ixI (τ ) (8.80)
The operator J −I here is a Lorentz covariant generator expressed using light-cone coordinates.
It is not the light-cone gauge Lorentz generator. We will get to the light-cone Lorentz generators
shortly. Similar to (8.78) we also have the commutator of J µν with pµ (τ ). Explicitly,
8.3. THE LORENTZ CHARGES 159
[J µν , pρ (τ )] = [xµ (τ ), pρ (τ )]pν (τ ) − [xν (τ ), pρ (τ )]pµ (τ ) = i(η µρ pν (τ ) − η νρ pµ (τ )) (8.81)
Typically in quantum mechanics we are given a set of quantum operators. It is often interesting to
calculate the commutators of these operators as it can lead to interesting physical insights (a prime
example are commutators of the angular momentum operators from ordinary quantum mechanics).
We would like to know the commutator of two Lorentz generators as given in (8.77). Let’s explicitly
compute the commutator here. Consider two Lorentz charges J µν and J ρσ . For simplicity we drop
the explicit τ dependence of the Heisenberg operators. The commutator is then
[J µν , J ρσ ] = [J µν , xρ pσ − xσ pρ ]
= [J µν , xρ ]pσ + xρ [J µν , pσ ] − [J µν , xσ ]pρ − xσ [J µν , pρ ]
Then, using (8.78) and (8.81) we have
[J µν , J ρσ ] = i (η µρ xν pσ − η νρ xµ pσ + η µσ xρ pν − η νσ xρ pµ − η µσ xν pρ + η νσ xµ pρ − η µρ xσ pν + η νρ xσ pµ )
= i (η µρ J νσ − η νρ J µσ + η νσ J ρν − η νσ J ρµ )
= i (η µρ J νσ − η νρ J µσ − η νσ J νρ + η νσ J µρ ) (8.82)
(8.82) actually defines the Lorentz Lie algebra, a topic which will be discussed in more detail
later on [64]. Moreover, (8.82) must be satisfied by similar operators J µν of any Lorentz invariant
quantum theory. If not, the theory is not Lorentz invariant. Now that we have the commutator
(8.82), we can determine the commutators of the Lorentz charges in light-cone coordinates. Using
light-cone indices, the Lorentz generators are given by J IJ , J +I , J −I , J −+ . Using (8.82), we find
that the relevant commutators are
[J ±I , J JK ] = i(δ IK J ±J − δ IJ J ±K ) (8.83)
[J ±I , J ∓J ] = i(−J IJ ± δ IJ J +− ) (8.84)
[J +− , J ±I ] = ±iJ ±I (8.85)
[J ±I , J ±J ] = 0 (8.86)
Up to this point we have only considered covariant Lorentz charges in light-cone coordinates.
Let’s now move on to determining the Lorentz charges in the light-cone gauge. We are faced with
a series of problems in finding these charges: defining the charges, the type of transformations the
charges generate, and the commutation relations the charges satisfy. To have a complete theory
describing the Lorentz charges in the light-cone gauge, we solve each of these problems. Let’s start
by ascertaining the form of the Lorentz charges in the light-cone gauge.
Our method will be to “guess” the correct form of the charges. The simplest guess we could
make for the light-cone gauge generators is to use the light-cone coordinates in the covariant formula
(8.77), this time replacing x+ (τ ), x− (τ ), and p− using the light-cone gauge condition. Therefore we
might guess for J +− ,
p+ τ − p+ τ − +

−
J +− + − − +
= x (τ )p (τ ) − x (τ )p (τ ) = 2 p − x0 + 2 p p = −x− 0p
+
m m
However there is a minor problem. Since xµ (τ ) and pµ (τ ) are Hermitian operators, the covariant
Lorentz charges J µν are automatically Hermitian. As written, J +− is not Hermitian, which is
something we would like to keep in our quantum theory. Therefore, we are motivated to define the
Hermitian version of the generator J +− as
1
J +− ≡ − (x− p+ + p+ x−
0) (8.87)
2 0
We will take (8.87) to be the light-cone Lorentz generator. Now let’s consider the Lorentz
generator J −I . Here we might guess that
p− pI

J −I
= x (τ )p − x (τ )p = x0 + 2 τ p − x0 + 2 τ p− = x−
− I i − − I I I I −
0 p − x0 p
m m
We again run into the same problem as we did for J +− . We therefore define the Hermitian
verison as
1 I −
J −I ≡ x− I
x0 p + p− xI0

0p − (8.88)
2
If the light-cone gauge Lorentz generators are to satisfy the Lorentz algebra, we must require
that the generators satisfy the commutator given in (8.86). Let’s verify this is true. Notice that we
may rewrite (8.88) as
1 I −
J −I = x− I I −
0 p − x0 p + [x0 , p ]
2
Since [xI0 , pJ ] = iδ IJ and p− = 1 I I
2p+ (p p + m2 ) we find that the Lorentz generator takes the form
i pI
J −I = x− I I −
0 p − x0 p + (8.89)
2 p+
Let’s now compute the commutator between two Lorentz charges J −I and J −J .
pJ pI

i − I i − J
[J −I , J −J ] = [x− I I − − J J −
0 p − x0 p , x0 p − x0 p ] + x0 p − xI0 p− , + − x0 p − xJ0 p− , +
2 p 2 p
Since pI , pJ , p+ all commute, the last two terms cancel. Therefore we are left with
[J −I , J −J ] = [x− I I − − J J −
0 p − x0 p , x0 p − x0 p ]
This may be expanded as

8.3. THE LORENTZ CHARGES 161
[J −I , J −J ] = xI0 [x− J J J − − I I J − − J I − − − J I − − I J −
0 , p ]p −x0 [x0 , p ]p −x0 [x0 , p ]p +x0 [x0 , p ]p +x0 [x0 , p ]p −x0 [x0 , p ]p
−
But [x− − p
0 , p ] = i p+ and
pJ pJ + m2 pI
[xI0 , p− ] = [xI0 , +
]=i +
2p p
Therefore the commutator is just
xI0 p− pJ xI pJ p− xJ p− pI xJ pI p−

[J −I , J −J ] = i +
− 0 + + δ IJ x− p− − 0 + + 0 + − δ JI x− p− =0 (8.90)
p p p p
All in all, the light-cone gauge Lorentz generators do in fact satisfy the Lorentz algebra. Though
we only considered the commutator of one particular Lorentz charge, our conclusion is then that the
light-cone gauge Lorentz generators satisfy the commutation relations that the covariant operators
in light-cone coordinates do. Thus, we have established that Lorentz symmetry holds in the light-
cone theory of the relativistic quantum point particle. What’s more is that the calculation of
[J −I , J −J ] in quantum string theory leads to some interesting consequences. The commutator is
zero only if the string propagates in a space-time of some particular dimension, and if the definition
of mass is changed in a way such that we can find massless gauge fields in the spectrum of the open
string [64]. All in all, string theory is constrained enough that it is only Lorentz invariant for a
fixed space-time dimensionality. In other words, choosing a space-time of a higher dimensionality
isn’t simply a mathematical convenience, it is required in order to maintain Lorentz invariance.
Lastly, before moving on to quantizing the string, let us briefly see what transformations are
generated from the light-cone gauge Lorentz charges. In particular, let’s consider the generators
J +− and J −I . First let’ s seek the transformations of J +− . To do this we must compute the
commutators of J +− with the Heisenberg operators x+ (τ ), x− (τ ), andxI (τ ). Using (8.87), the form
of x+ (τ ) and the fact that [x− +
0 , p ] = −i, we find that
+

+− 1 − + + − p τ
[J +
, x (τ )] = − (x0 p + p x0 ), 2 = ix+ (τ ) (8.91)
2 m
p− τ
Additionally, with x− (τ ) = x−
0 + m2 , we have
τ − − + τ p−
[J +− , x− (τ )] = x− − +
0 [x0 , p ] − 2
[x0 , p ]p = −ix− −
0 − i 2 + = −ix (τ ) (8.92)
m m p
Similarly,
pI τ
[J +− , xI (τ )] = [−x− + I
0 p , x0 + ]=0 (8.93)
m2
Using (8.79) we find that J +− generates transformations by
δxµ = [−i+− J +− , xµ ] (8.94)

We notice then that J +− yields the expected translations. Let’s now move on to the generator
J . Just as before we must first calculate the commutator of J −I with the Heisenberg operators
−I
x+ (τ ), x− (τ ), and xI (τ ). Using (8.88) we find that

τ − + I iτ
[J −I , x+ (τ )] =
[x , p ]p = − 2 pI = −i(xI xI0 ) (8.95)
m2 0 m

− i 1 − I τ τ
−I − I −
[J , x (τ )] = −x0 [p , x0 ] + , x p + 2 [x− , p− ]pI − 2 [xI0 , p− ]p−
2 p+ 0 m 0 m
p− i i I iτ p− I −iτ pI −
= ixI0 +
− 2 p + p − 2 +p
p 2p+ m2 p+ m p
1 i I −
(x p + p− xI0 ) (8.96)
2 p+ 0
And,
τ I J − i 1
[J −I , xJ (τ )] = x− I J
[x , p ]p − xI0 [p− , xJ0 ] + [pI , xJ0 ] +
0 [p , x0 ] −
m2 0 2 p
τ − pJ 1 δ IJ
= −iδ IJ x−
0 + 2
p + ixI0 + +
m p 2 p+
i
(xI pJ + pJ xI0 ) − iδ IJ x− (τ ) (8.97)
2p+ 0
The expected Lorentz transformations are
[J −I , x+ ] = iη −I xI = −ixI (8.98)
[J −I , x− ] = 0 (8.99)
[J −I , xJ ] = −iη IJ x− = −iδ IJ x− (8.100)

Therefore, we see that there are extra terms appearing for each of the commutators (8.95),(8.96),
and (8.97). The extra term for [J −I , x+ ] is
ixI0
For [J −I , x− ],
i
(xI p− + p− xI0 )
2p+ 0
And the extra term for [J −I , xJ ] is just
i
(xI pJ + pJ xI0 )
2p+ 0
Then, using the fact that
dxµ pµ
= 2 (8.101)
dτ m
for indices µ = +, −, I, we find that the extra terms take the “Hermiticized”
8.4. EXERCISES 163
1
δxµ = (λ∂τ xµ + ∂τ xµ λ) (8.102)
2
where
m2 xI0
λ= (8.103)
p+
Therefore we see that the Lorentz charge J −I generates the expected Lorentz transformations
together with a compensating reparameterization of the world-line [64].
Now that we have detailed the method of quantizing the relativistic quantum point particle
using the light-cone gauge, we are finally ready to quantize the relativistic string using the light-
cone gauge.
8.4 Exercises
1. (a) Derive Ehrenfest’s Theorem. To do this consider an operator Ω, and compute
d d
hΩi = hψ|Ω|ψi
dt dt
for arbitrary, time-dependent state vectors |ψi. Assume that Ω is not time dependent. Using
the Schrödinger equation, arrive to the general form of Ehrenfest’s theorem:
˙ = i h[H, Ω]i
hΩi
~
Compare this to the correpsonding differential equation for time-dependent operators.
(b) Let Ω, Λ, and Θ be arbitrary operators. Derive the following useful identities:
[Ω, ΛΘ] = Λ[Ω, Θ] + [Ω, Λ]Θ
[ΛΩ, Θ] = Λ[Ω, Θ] + [Λ, Θ]Ω
(c) Consider the case where H = P 2 /2m+V0 sin(2πX/a). Compute hXi ˙ and hP˙ i. What happens
to each result when we consider the translation x → x+ma, where m is an integer? (Hint: Consider
writing the potential V (X) in terms of a power series and then compute the commutator term by
term).
˙ and hP˙ i with H = P 2 /2m + V (X) for any arbitrary

(d) Using part (c) as a guide, compute hXi
V (X). In general, what does Ehrenfest’s theorem say? What does this mean for the translation
above?
2. In a sense, when we canonically quantize a classical theory we are really promoting our
observables (dynamical variables) to operators, and insist that they satisfy an appropriate set of
commutation relations that have a classical correspondence. Here we will explore this classical
correspondence. (a) Let ω(p, q) be some function of the state variables p, q with no explicit time
dependence. Show that the time variation of ω is

dω X ∂ω ∂H ∂ω ∂H
= − ≡ {ω, H}
dt i
∂qi ∂pi ∂pi ∂qi
where H is the classical Hamiltonian, and where we have defined the Poisson bracket {·, ·}.
(b) Show that
q̇i = {qi , H} ṗi = {pi , H}

Compare this result to what was considered in problem 1. What can we say about Ehrenfest’s
theorem?
(c) Prove that
{qi , qj } = {pi , pj } = 0 {qi , pj } = δij

Compare this to the canonical commutation relations for the operators Xi and Pj . Write a
brief statement describing what is meant by canonical quantization, i.e. how does one choose the
canonical commutation relations?
3. Check that (8.52) and (8.53) hold.
4. Deduce (8.62). (Hint: It may help to use a Fourier transform to exchange x− into p+ and xI
into pI .)
5. Using (8.82), check the commutators given in (8.83)-(8.86).

Chapter 9
Light-Cone Quantization of the

String
We are finally in a position to quantize the string. In this chapter we will quantize the string using
the light-cone gauge. Light-cone quantization is, though not as elegant, is a quantization procedure
that reveals the physics much quicker than some of the other, more elegant approaches. We will
examine these approaches in later chapters, however not in as much detail.
9.1 Quantizing the String

Using the quantization of the relativistic point particle as motivation, we must interpret the classical
equations of motion in the light-cone gauge for the Heisenberg operators. Recall the worldsheet
parameterizations
n · X(τ, σ) = βα0 (n · p)τ (9.1)
α0 π
n·p= n · Pτ (9.2)
β
where β = 2 for open strings and β = 1 for closed strings. We used these parameterizations to
find the wave equations
Ẍ µ − X 00µ = 0 (9.3)
along with the constraints
2
Ẋ ± X 0 =0 (9.4)
We then found simple expressions for the momentum operators:
1 1
P σµ = − X 0µ , P τ µ = Ẋ µ (9.5)
2πα0 2πα0
165
166 CHAPTER 9. LIGHT-CONE QUANTIZATION OF THE STRING
Equations (9.1)-(9.5) hold in all gauges, including the light-cone gauge, as well as for both open
and closed strings. The difference is for open strings we found σ ∈ [0, π], while for closed strings
σ ∈ [0, 2π]. For now, let’s focus on open strings. In the light-cone gauge, for open strings, we set
X + = 2α0 p+ τ and solved for X − in terms of the transvers coordinates X I using
1 1 2
Ẋ − ± X 0− = 0 +
Ẋ I ± X 0I (9.6)
βα 2p
Using β = 2 we find
1 1
Ẋ − = Ẋ I Ẋ I + X 0I X 0I

(9.7)
2α0 2p+
yielding
X 0I X 0I

1 1 1
Pτ− = Ẋ −
= (2πα 0 2
) P τI τI
P +
2πα0 2πα0 4α0 p+ (2πα0 )2
X 0I X 0I

π τI τI
= P P + (9.8)
2p+ (2πα0 )2
For a quantum theory we must find the Schrödinger operators. Motivated by the Schrödinger
operators of the quantum point particle, we choose the τ independent operators to be
(X I (σ), x− τI +
0 , P (σ), p )
The associated Heisenberg operators are then
(X I (τ, σ), x− τI +
0 (τ ), P (τ, σ), p (τ ))
It’s important to note that since the Schrödinger operators do not exhibit any explicit time depen-
dence, the Heisenberg operators have no explicit time dependence either.
Let’s now set up the commutation relations for Schrödinger and Heisenberg operators. The
operators have σ dependence, so we demand that the operators fail to commute only if they are at
the same point along the string. That is, we do not expect simultaneous measurements at different
points on the string to intefere with each other. Therefore, we assume the commutators for the
Schrödinger operators satisfy [64]:
[X I (σ), P τ J (σ 0 )] = iη IJ δ(σ − σ 0 ) (9.9)
[X I (σ), X J (σ 0 )] = [P τ I (σ), P τ J (σ 0 )] = 0 (9.10)

We also have
[x− +
0 , p ] = −i (9.11)
Operators x− +
0 and p commute with all other operators. Similarly, for the Heisenberg operators,
we set up equal time commutators just as we did for scalar fields:
[X I (τ, σ), P τ J (τ, σ 0 )] = iη IJ δ(σ − σ 0 ) (9.12)

9.1. QUANTIZING THE STRING 167
And also, we have the τ dependent analog of (9.11)
[x− +
0 (τ ), p (τ )] = −i (9.13)
All other commutators vanish. Let’s now determine the Hamiltonian. The Hamiltonian should
generate τ translation for the string. From the analysis of the point particle, we know that p−
generators X + translation. In the light-cone gauge, X + = 2α0 p+ τ , so
∂ ∂X + ∂ ∂
= +
= 2α0 p+
∂τ ∂τ ∂X ∂X +
The Hamiltonian that generates τ evolution is then
Z π
H = 2α0 p+ p− = 2α0 p+ dσP τ − (9.14)
0
Using (9.8), the Hamiltonian takes the form

π
X 0I (τ, σ)X 0I (τ, σ)
Z
0 τI τI
H = πα dσ P (τ, σ)P (τ, σ) + (9.15)
0 (2πα0 )2
As noted from ordinary quantum mechanics, the Hamiltonian must generate the quantum equa-
tions of motion that are operator versions of the classical equations of motion. Recall that the
Heisenberg operators, when their Schrödinger analogs have no explicit time dependence, the oper-
ator equation of moton takes the form
dZh
i = [Zh , H] (9.16)
dτ
Presently, in our case Z = Z(τ, σ) and H = H(τ ). Therefore,
dZh (τ, σ)
i = [Zh (τ, σ), H(τ )] (9.17)
dτ
Since the Hamiltonian is built from Heisenberg operators with no explicit time dependence, we
conclude that H(τ ) = H is constant in τ . Moreover, it is easy to see that [x− +
0 , H] = [p , H] = 0;
− − + +
thus x0 (τ ) = x0 and p (τ ) = p are τ independent. Therefore the Heisenberg commutator given
in (9.13) is equivalent to the Schrödinger operator given in (9.11). The Heisenberg equation of
motion for X I (τ, σ) is
Z π
I I I 0 0 τJ 0 τJ 0
iẊ (τ, σ) = [X (τ, σ), H] = X (τ, σ), πα dσ P (τ, σ )P (τ, σ ) (9.18)
0
We dropped the second term in (9.15) since it commutes with X I since
∂
[X I (τ, σ), X 0J (τ, σ 0 )] = [X I (τ, σ), X J (τ.σ)] = 0 (9.19)
∂σ 0
To evaluate (9.18) we use (9.12), yielding
Z π
0
I
iẊ (τ, σ) = 2πα dσ 0 P τ J (τ, σ 0 )iη IJ δ(σ − σ 0 ) (9.20)
0
Performing the integral leads us to conclude that
Ẋ I (τ, σ) = 2πα0 P τ I (τ, σ) (9.21)

1
Of course, we could have expected this result since P τ µ = µ
2πα0 Ẋ . If we were to calculate Ṗ τ I
we could show that
Ẍ I − X 00I = 0 (9.22)
is the quantum equation of motion.
Going from classical string theory to quantum theory the classical boundary conditions become
constraints on our operators. For instance, the Neumann boundary condition

∂σ X I (τ, σ)

=0 (9.23)
σ=0,π
can be literally taken as the condition that the operator ∂σ X I vanishes at open string endpoints.
In an earlier chapter, we saw that (Ẋ I ± X 0I ) was a particularly useful relation. The same is true
when we quantize the string. For this reason, let’s calculate the commutators of these derivatives.
Using Ẋ J = 2πα0 P τ J (τ, σ 0 ), we have
[X I (τ, σ), Ẋ J (τ, σ 0 )] = 2πα0 iη IJ δ(σ − σ 0 ) (9.24)

Taking a σ derivative of this commutator we find
d
δ(σ − σ 0 )
[X 0I (τ, σ), Ẋ J (τ, σ 0 )] = 2πα0 iη IJ (9.25)
dσ
Differetiating [X I (τ, σ), X J (τ, σ)] with respect to both σ and σ 0 and recalling [P τ I (τ, σ), P τ J (τ, σ 0 )] =
0, we find that the τ and σ derivatives of the coordinates separately commute among themselves:
[X 0I (τ, σ), X 0J (τ, σ 0 )] = [Ẋ I (τ, σ), Ẋ J (τ, σ 0 )] = 0 (9.26)

Let’s now examine the commutator
0
[(Ẋ I + X I )(τ, σ), (Ẋ J + X 0J )(τ, σ 0 )] (9.27)
Using (9.26), we find that
[(Ẋ I + X 0I )(τ, σ), (Ẋ J + X 0J )(τ, σ 0 )] = [Ẋ I (τ, σ), X 0J (τ, σ 0 )] + [X 0I (τ, σ), Ẋ J (τ, σ 0 )] (9.28)
The second term is just

d
[X 0I (τ, σ), Ẋ J (τ, σ 0 )] = 2πα0 iη IJ δ(σ − σ 0 ) (9.29)
dσ
The first term is just
d d
[Ẋ I (τ, σ), X 0J (τ, σ 0 )] = −[X 0J (τ, σ 0 ), Ẋ I (τ, σ)] = −2πα0 iη IJ 0
δ(σ 0 − σ) = 2πα0 iη IJ δ(σ − σ 0 )
dσ dσ
(9.30)
9.2. COMMUTATION RELATIONS FOR THE OSCILLATORS OF OPEN AND CLOSED STRINGS169
Therefore, (9.27) becomes
d
[(Ẋ I + X 0I )(τ, σ), (Ẋ J + X 0J )(τ, σ 0 )] = 4πα0 iη IJ δ(σ − σ 0 ) (9.31)
dσ
Or, more generally,
d
δ(σ − σ 0 )
[(Ẋ I ± X 0I )(τ, σ), (Ẋ J + X 0J )(τ, σ 0 )] = ±4πα0 iη IJ (9.32)
dσ
Moreover, since only the cross terms contribute, we also have that
[(Ẋ I ± X 0I )(τ, σ), (Ẋ J ∓ X 0J )(τ, σ 0 )] = 0 (9.33)

0 0
which holds for σ, σ ∈ [0, π]. In fact, all of these commutation relations hold for σ, σ ∈ [0, 2π],
and thus the commutation relations leading up and including (9.32) and (9.33) hold for the closed
string as well.
9.2 Commutation Relations for the Oscillators of Open and

Closed Strings
So far we have examined commutation relations for the field operators, coming up with an infinite set
of commutation relations involving delta functions. Often it is useful to recast these commutation
relations in a discrete form. For this reason we will examine the mode expansions of our operators
and explicitily determine the commutation relations for the classical modes αnI as they become
quantum operators.
Let’s first examine the open string. Recall our solution to the wave equation with Neumann
boundary conditions:
√ √ X1
X I (τ, σ) = xI0 + 2α0 α0I τ + i 2α0 αI cos nσe−inτ (9.34)
n n
n6=0
We also found
√ X
(Ẋ I + X 0I )(τ, σ) = 2α0 αnI e−in(τ +σ) (9.35)
n∈Z
√ X
(Ẋ I − X 0I )(τ, σ) = 2α0 αnI e−in(τ −σ) (9.36)
n∈Z
with σ ∈ [0, π]. Let’s no construct a function of σ that has a period of 2π. To do this, we evalute
(9.36) at −σ:
√ X
(Ẋ I − X 0I )(τ, −σ) = 2α0 αnI e−in(τ +σ) (9.37)
n∈Z
where now σ ∈ [−π, 0]. Now define the function AI (τ, σ) [64]:
√ X
AI (τ, σ) ≡ 2α0 αnI e−in(τ +σ) (9.38)
n∈Z
along with
AI (τ, σ + 2π) = AI (τ, σ) (9.39)

Notice then
0
(Ẋ I + X I )(τ, σ)

: σ ∈ [0, π]
AI (τ, σ) = 0
(Ẋ I − X I )(τ, σ) : σ ∈ [−π, 0]
AI will be useful in determining the commutation relations for the αnI . To do this, we must
compute the commutator [AI (τ, σ), AJ (τ, σ 0 )] over a variety of ranges for σ and σ 0 . In particular,
we must compute
[(Ẋ I + X 0I )(τ, σ), (Ẋ J + X 0J )(τ, σ 0 )], σ, σ 0 ∈ [0, π] (9.40)
[(Ẋ I + X 0I )(τ, σ), (Ẋ J − X 0J )(τ, −σ 0 )], σ ∈ [0, π], σ 0 ∈ [−π, 0] (9.41)
[(Ẋ I − X 0I )(τ, −σ), (Ẋ J + X 0J )(τ, σ 0 )], σ ∈ [−π, 0], σ 0 ∈ [0, π] (9.42)
[(Ẋ I − X 0I )(τ, −σ), (Ẋ J − X 0J )(τ, −σ 0 )], σ, σ 0 ∈ [−π, 0] (9.43)

We have already determined the first and last commutators, (9.40), and (9.43). They are just
d
[(Ẋ I + X 0I )(τ, σ), (Ẋ J + X 0J )(τ, σ 0 )] = 4πα0 iη IJ δ(σ − σ 0 ) (9.44)
dσ
Similarly,
d
[(Ẋ I − X 0I )(τ, −σ), (Ẋ J − X 0J )(τ, −σ 0 )] = −4πα0 iη IJ δ(−σ + σ 0 )
d(−σ)
d
= 4πα0 iη IJ
δ(σ − σ 0 ) (9.45)
dσ
To evalute (9.41) and (9.42), we make use of (9.33), noticing that they vanish. Therefore all four
commutators can be summarized as
d
[AI (τ, σ), AJ (τ, σ 0 )] = 4πα0 iη IJ δ(σ − σ 0 ) (9.46)
dσ
with σ, σ 0 ∈ [−π, π]. Using our mode expansion of AI (τ, σ), we notice that (9.46) may be written
as
X 0 0 0 d
e−im (τ +σ) e−in (τ +σ ) [αm
I J
0 , αn0 ] = 2πiη
IJ
δ(σ − σ 0 ) (9.47)
dσ
n0 ,m0 ∈Z
To extract information about the commutators of the modes, we apply on both sides
Z 2π Z 2π
1 1 0
dσe imσ
· dσ 0 einσ
2π 0 2π 0
For the left hand side, the integrals only pick up the terms where m0 = m and n0 = n, resulting
in
 
Z 2π Z 2π
1 1 0 X 0 0 0
dσeimσ · dσ 0 einσ  e−im (τ +σ) e−in (τ +σ ) [αm
I J 
0 , αn0 ]
2π 0 2π 0 n0 ,m0 ∈Z
= e−i(m+n)τ [αm
I
, αnJ ] (9.48)
Evaluating the right hand side of (9.47) yields
Z 2π Z 2π
1 d 0
iη IJ dσeimσ dσ 0 eiσ δ(σ − σ 0 )
2π 0 dσ 0
Z 2π Z 2π
IJ 1 imσ d inσ 1
= iη dσe e = −nη IJ dσei(m+n)σ
2π 0 dσ 2π 0
−nη IJ δm+n,0 = mη IJ δm+n,0 (9.49)

Altogether then we have that
I
[αm , αnJ ] = mη IJ δm+n,0 ei(m+n)τ = mη IJ δm+n,0 (9.50)
where we used the fact that we may use the Kronecker delta to set m = −n. Thus,
I
[αm , αnJ ] = mη IJ δm+n,0 (9.51)
√
Take note that α0I commutes all other oscillators since α0I = 2α0 pI . Now that we have the
commutation relation between the oscillators α0I , let’s find the commutators between xI0 and the
oscillators. Consider
[X I (τ, σ), Ẋ J (τ, σ 0 )] = 2πα0 iη IJ δ(σ − σ 0 ) (9.52)

Integrating both sides over the range σ ∈ [0, π] yields
√
[xI0 + 2α0 α0I τ, Ẋ J (τ, σ 0 )] = 2α0 iη IJ (9.53)
Since Ẋ J is a sum of terms containing αnJ , [α0I , Ẋ J ] = 0. Moreover, if we use the mode expansion
of Ẋ J we find that (9.53) becomes
X 0 √
[xI0 , αnJ 0 ] cos n0 σ 0 e−in τ = 2α0 iη IJ (9.54)
n0 ∈Z
Reorganizing gives
∞ h
X 0
in0 τ
i √
[xI0 , α0J ] + xI0 , αnJ 0 e−in τ + α−n
J
0e cos(n0 σ 0 ) = 2α0 iη IJ (9.55)
n0 =1
If we apply to both sides Z π

1
dσ cos(nσ)
π 0
we find that
I J −inτ J
einτ = 0

x0 , αn e + α−n (9.56)
which can also be written as
[xI0 , αnJ ]e−inτ + [xI0 , α−n

J
]einτ = 0 (9.57)
Equation (9.57) must hold for all values of τ . Notice then, that if we consider the case where
τ = 0 in (9.57)
lim [xI0 , αnJ ]e−inτ + [xI0 , α−n

J
]einτ → [xI0 , αnJ ] + [xI0 , α−n
J

]=0 (9.58)
τ →0
We require that
[xI0 , α−n
J
]=0 (9.59)
and
[xI0 , αnJ ] = 0 (9.60)

It is important to note that (9.59) and (9.60) hold when n 6= 0. When n = 0, we instead have
√
[xI0 , α0J ] = 2α0 iη IJ (9.61)
√
Using the fact that α0I = 2α0 pI , we attain the expected commutator
[xI0 , pJ ] = iη IJ (9.62)
The commutation relations for the oscillators αnµ
have a familiar physical meaning: they are
equivalent to a set of creation and annihilation operators. Recall that we defined our oscillators as
√
αnµ = naµn (9.63)
µ √
α−n = naµ∗
n (9.64)
In quantum theory, as we’ve already seen, these oscillators become operators. For light-cone
modes µ = I, we have
√
αnI = naIn (9.65)
I
√
α−n = naI∗
n (9.66)
By ∗ we mean the Hermitian conjugate. Sometimes this is written as † to avoid ambiguity
with the classical variables and their corresponding complex conjugate. However, since we will be
working with operators in this chapter, this ambiguity isn’t present. Furthermore, note that
(αnI )∗ = α−n
I
(9.67)
for n ∈ Z. The important consequence of the Hermiticity properties is that X I (τ, σ), used in
classical theory, becomes a Hermitian operator, (X I (τ, σ))∗ = X I (τ, σ). Moreover, we can rewrite
(9.51) as
I J
[αm , α−n ] = mδm,n η IJ (9.68)
When the m and n are integers of opposite signs, the right hand side vanishes. Therefore we
learn that
[aIm , aJn ] = [aI∗ J∗

m , an ] = 0 (9.69)
However, if both m and n are positive integers, we have that
√ √ m
[ maIm , naJ∗
n ] = mδm,n η
IJ
→ [aIm , aJ∗
n ]= √ δm,n η IJ
mn
Using the fact that the right hand side vanishes unless m = n, we arrive to the commutation
relation
[aIm , aJ∗
n ] = δm,n η
IJ
(9.70)
Equation (9.70) indicates that the operators (aIm , aI∗
m ) satisfy the commutation relations of the
canonical annihilation and creation operators from the quantum mechanical harmonic oscillator.
This suggests that the operators (aIm , aI∗
m ) are the creation and annihilation operators. We will see
this for certainty in a moment. Moreover, we find that there is a pair of creation and annihilation
operators for each value m ≥ 1 of the mode number and for each transverse light-cone direction I.
Additionally, the oscillators corresponding to different modes and light-cone coordinates commute.
If the mode numbers and light-cone coordinates agree then the commutator is one. In summary,
we find that αnI may be interpreted as annihilation operators, while α−n I
may be interpreted as
I
creation operators. Notice then that we may write the expansion of X as
∞
√ X cos(nσ)
X I (τ, σ) = xI0 + 2α0 pI τ + i 2α0 (αnI e−inτ − α−n
I
einτ ) (9.71)
n=1
n
which can also take the form
∞
√ X
inτ cos(nσ)
X I (τ, σ) = xI0 + 2α0 pI τ + i 2α0 (aIn e−inτ − aI∗
n e ) √ (9.72)
n=1
n
We are also led to identical commutation relations for the oscillators corresponding to the closed
string. This time however, we have commutation relations for each of the left and right moving
oscillators. Using the general expansion for a closed string
√
r
µ µ α0 X e−inτ
µ

0
X (τ, σ) = x0 + 2α α0 τ + i (9.73)
2 n
n6=0
and using
√
Ẋ µ + X 0µ = 2XL0µ (τ + σ) =
X
2α0 α̃nµ e−in(τ +σ) (9.74)
n∈Z
0µ
√ X
Ẋ µ − X 0µ = 2XR (τ − σ) = 2α0 αnµ e−in(τ −σ) (9.75)
n∈Z
we complete the same computations as done for the open string, leading us to
r
α0 IJ
[xI0 , α0J ] = [xI0 , α̃0J ] =i η (9.76)
2
This leads us to
[xI0 , pJ ] = iη IJ (9.77)
Moreover, we also find
I
[α̃m , α̃nJ ] = mδm+n,0 η IJ (9.78)
I
[αm , αnJ ] = mδm+n,0 η IJ (9.79)
Separately, the right and left oscillators correspond to open strings. In other words, in closed
string theory, one has operators that contain two copies of open string theory, excluding the zero
modes of course. The left and right oscillators do not mix however, leading to
I
[αm , α̃nJ ] = 0 (9.80)
We can define canonical creation and annihilation operators just as we did for open strings:
√ √
αnI = naIn I
α−n = naI∗
n (9.81)
√ √
α̃nI = nãIn I
α̃−n = nãI∗
n (9.82)
The non-vanishing commutation relations are as expected
[ãIm , ãJ∗ I J∗
n ] = [am , an ] = δm,n η
IJ
(9.83)
The fact that the form of the commutation relations are that of the canonical creation and
annihilation operators are identical does not completely justify that they are the creation and
annihilation operators. To be able to make this claim rightfully, let us consider a more physical
analysis of the oscillators.
9.3 Strings and Harmonic Oscillators

Here we will show that open strings, and therefore closed strings, are actually harmonic oscillators.
This will justify our claim that the oscillators examined in the last section are indeed the canonical
creation and annihilation operators from ordinary Quantum Mechanics. Let’s begin by briefly re-
viewing the basic properties of the quantum simple harmonic oscillator. Let qn (t) be the coordinate
of the simple harmonic oscillator and consider the action
Z Z
1 2 n 2
Sn = Ln (t)dt = dt q̇ (t) − qn (t) (9.84)
2n n 2
9.3. STRINGS AND HARMONIC OSCILLATORS 175
Then, where the canonical momentum is
∂L 1
pn = = q̇n (9.85)
∂ q̇n n
the Hamiltonian is written as

n 2
(p + qn2 )
Hn (pn , qn ) = pn q̇n − Ln = (9.86)
2 n
which is the Hamiltonian describing a simple harmonic oscillator. By comparing to the more
conventional expression of a simple harmonic oscillator, we find that here n takes the place of the
angular frequency ω. As we switch to quantum theory, we impose the canonical commutation
relation
[qn , pn ] = i (9.87)
The creation and annihilation operators are introduced as
1
an = √ (pn − iqn ) (9.88)
2
1
a∗n = √ (pn + iqn ) (9.89)
2
From which we find the commutation relation
[an , a∗n ] = 1 (9.90)

We can recast the Hamiltonian using these operators as
1
Hn = n(a∗n an + ) (9.91)
2
Now consider the Heisenberg operators (an (t), a∗n (t)) associated with the Schrödinger operators.
They satisfy the same commutation relation:
[an (t), a∗n (t)] = 1 (9.92)

The Heisenberg equation of motion for an (t) is simply
ȧn = i[Hn (t), an (t)] = in[a∗n (t)an (t), an (t)] = −inan (t) (9.93)
This differential equation is easily solved, yielding
an (t) = an e−int (9.94)

where an is a constant. Similarly,
a∗n (t) = a∗n eint (9.95)

From here we can easily see that n is indeed the angular frequency ω. Let’s now consider the
action for X I (τ, σ). We claim it is given by [64]:
Z Z Z π
1
S= dτ dσL = dτ dσ(Ẋ I Ẋ I − X 0I X 0I ) (9.96)
4πα0 0
It’s easy to convince oneself that the first term in the action represents the kinetic energy while
the second term represents the potential energy. The canonical momentum is
∂L 1
= Ẋ I = P τ I (9.97)
∂ Ẋ I 2πα0
The equations of motion for X I , as usual, follow by varying the action:
Z Z π
1
dσ ∂τ (δX I )Ẋ I − ∂σ (δX I )X 0I

δS = 0
dτ (9.98)
2πα 0
By restricting our variations where the initial and final positions of the transverse coordinate are
fixed, we find that we may write the variation of the action as
Z π Z π
1 0I I
dσδX I (Ẍ I − X 00I )

δS = − 0
dτ [X δX ] + (9.99)
2πα 0 0
For δS = 0 for any variation δX I we require the usual boundary conditions and that
Ẍ I − X 00I = 0 (9.100)
which is just the wave equation for the transverse coordinate. Moreover, the Hamiltonian is
simply
Z π Z π
H= dσH = dσ(P τ I Ẋ I − L) (9.101)
0 0
I
Writing Ẋ in terms of the canonical momentum, we find
Z π
0 τI τI 1 0I 0I
dσ πα P P + X X (9.102)
0 4πα0
Let’s use the action to quantize our theory. We can do easily by replacing the dynamical variable
X I by a collection of other dynamical variables qnI . That is, we write X I as
∞
√ X cos(nσ)
X I (τ, σ) = q I (τ ) + 2 α0 qnI (τ ) √ (9.103)
n=1
n
This is the most general expression that satisfies the Neumann boundary conditions. We also
have
∞
√ X cos(nσ)
Ẋ I (τ, σ) = q̇ I (τ ) + 2 α0 q̇nI (τ ) √ (9.104)
n=1
n
∞
0 √ X √
X I (τ, σ) = −2 α0 qnI (τ ) n sin(nσ) (9.105)
n=1
Using these expansions, we may write the action given in (9.96) as

9.3. STRINGS AND HARMONIC OSCILLATORS 177
∞ ∞
" ! !
1
Z Z π
I
√ X cos(nσ) √ X cos(mσ)
S= dτ dσ q̇ (τ ) + 2 α0 q̇nI (τ ) √ I
q̇ (τ ) + 2 α0 I
q̇m (τ ) √
4πα0 0 n=1
n m=1
m
∞ X
∞
#
X √ √
−4α0 I
qnI (τ ) n sin(nσ)qm (τ ) m sin(mσ)
n=1 m=1
(9.106)
∞ X
∞
"
Z Z π
1 X cos(nσ) cos(mσ)
= I I
dτ q̇ (τ )q̇ (τ )π + dσ 4α0 q̇n (τ )q̇m (τ ) √
4πα0 0 n=1 m=1
mn (9.107)
√
−4α0 qnI qm
I

mn sin(nσ) sin(mσ)
Then, making use of the fact that

Z π Z π
cos(nσ) cos(mσ)dσ = sin(nσ) sin(mσ)dσ = 0 (9.108)
0 0
when m 6= n, we find that the action is

∞
Z " #
1 I X 1 n
S = dτ q̇ (τ )q̇ I (τ ) + q̇nI (τ )q̇nI (τ ) − qnI (τ )qnI (τ ) (9.109)
4α0 n=1
2n 2
By comparing to the action given in (9.86), we see that qnI (τ ) are the coordinates of the simple
harmonic oscillators. This is the physical interpretation of the coefficients of X I . The Hamiltonian
is then given by
∞
X n I I
H = α0 pI pI + pn pn + qnI qnI

(9.110)
n=1
2
where we used
∂L 1 I
pI = = q̇ (9.111)
∂ q̇ I 2α0
The Heisenberg operators lead us to qn (τ ), for which we have
i
qnI (τ ) = √ aIn e−inτ − aI∗ inτ

n (τ )e (9.112)
2
where (aIn , aI∗
n ) are the canonically normalized creation and annihilation operators. The Heisen-
berg equation for q I (τ ) is simply
q̇ I (τ ) = i[H(τ ), q I (τ )] = iα0 [pJ (τ )pJ (τ ), q I (τ )] = 2α0 pI (τ ) (9.113)

Solving the differential equation yields
q I (τ ) = xI0 + 2α0 pI τ (9.114)

Substituting everything into (9.103), we find that

∞
√ X cos(nσ)
X I (τ, σ) = xI0 + 2α0 pI τ + i 2α0 (αnI e−inτ − α−n
I
einτ ) (9.115)
n=1
n
in direct agreement with (9.71). Therefore, we have identified the classical variables becoming
oscillators. We are therefore justified in claiming that the oscillators αnI and α−n
I
are the annihilation
and creation operators as the string itself behaves as a harmonic oscillator.
9.4 The Transverse Virasoro Operators

Let’s first focus on the open string. We have written the mode expansions for X I and have seen
their relationship to the oscillators. Let’s now consider the other light-cone coordinates, X + (τ, σ)
and X − (τ, σ). X + is simple:
√
X + (τ, σ) = 2α0 p+ τ = 2α0 α0+ τ (9.116)
where we see that x+ +
0 = 0 and αn = 0 for n 6= 0. For X
−
we use a mode expansion
√ √ X1
X − (τ, σ) = x−
0 + 2α0 α0− τ + i 2α0 α− e−inτ cos(nσ) (9.117)
n n
n6=0
Earlier we had used constraints which allowed us to write the oscillators αn− in terms of the αnI
modes,
√ 1 ⊥
2α0 αn− = L (9.118)
p+ n
with
1X I
L⊥
n = αn−p αpI (9.119)
2
p∈Z
We had called L⊥ I
n transverse Virasoro modes. We just saw that the αn modes became operators
⊥
during our quantization procedure, thus, we expect that Ln will become transverse Virasoro oper-
ators in our quantum theory of the string. Since to α operators fail to commute when their mode
numbers add to zero, the two operators in L⊥ n fail to commute only when n = 0. For that reason,
let’s begin by considering the operator L⊥ n .
In a previous chapter, we saw briefly that L⊥0 is important in determining the mass of the strings.
In fact, adding quantum theory brings in a subtlety: normal ordering the operator. Note that we
may write the operator as
∞ ∞
1X I I 1 1X I I 1X I I
L⊥
0 = α−p αp = α0I α0I + α α + α α (9.120)
2 2 2 p=1 −p p 2 p=1 p −p
p∈Z
Notice that the first sum on the right hand side is normal-ordered: the annihilation operators
appear to the right of the creation operators. The last sum is not normal-ordered. Remember we
introduced the notion of normal-ordering when we examined the Hamiltonian acting on the vacuum
9.4. THE TRANSVERSE VIRASORO OPERATORS 179
state. Typically in quantum theory we normal order our operators so that they act on the vacuum
state in a simple way. For this reason, we rewrite the last sum so that it is normal-ordered:
∞ ∞
1X I I 1 X I I I I
αp α−p = α α + αp , α−p
2 p=1 2 p=1 −p p
∞ ∞
1 X I I 1 X IJ
= α α + pη
2 p=1 −p p 2 p=1
∞ ∞
1X I I 1 X
= α−p αp + (D − 2) p (9.121)
2 p=1 2 p=1
Now we have run into a problem: the last sum in (9.121) is infinite! We will correct this in a
moment. For now, we have that
∞ ∞
1 I I X I I 1 X
L⊥
0 = α0 α0 + α−p αp + (D − 2) p (9.122)
2 p=1
2 p=1
√
Then, where 2α0 α0− = 2α0 p− = p1+ L⊥ ⊥
0 , let’s define L0 to be the normal-ordered operator given
in (9.122), without including the ordering constant. That is,
∞ ∞
1 I I X I I X
L⊥
0 ≡ α0 α0 + α−p αp = α0 pI pI paI∗ I
p ap (9.123)
2 p=1 p=1
Let’s also introduce the ordering constant a as

1
2α0 p− ≡ (L⊥ + a) (9.124)
p+ 0
Since we want to normal-order L⊥
0 , we require that the constant be
∞
1 X
a= (D − 2) p (9.125)
2 p=1
This choice turns out to be correct and will have physical significance. It will actually turn out
to have a finite value. Before determining a, let’s first see how the introduction of this constant
changes the mass squared operator M 2 . Using
∞
!
2 2 + − I I 1 ⊥ I I 1 X
I∗ I
M = −p = 2p p − p p = 0 (L0 + a) − p p = 0 a + nan an (9.126)
α α n=1
We notice then that a introduces a shift to the mass-squared operator. Let’s now move on to
interpret (9.125) and determine the value of a. There is an important result in from mathematics
that suggest that the sum in (9.125) has a finite value. Consider the zeta function ζ(s) defined as
∞
X 1
ζ(s) = (9.127)
n=1
ns
which converges for R(s) > 1, where the argument s is assumed to be complex. We can use
analytic continuation to define the zeta function for all possible values of s except s = 1. To see
this, consider the definition of the gamma function
Z ∞
Γ(s) = dte−t ts−1 (9.128)
0
Now let t → nt, yielding

Z ∞ Z ∞
Γ(s) = d(nt)e−nt ns−1 ts−1 = dte−nt ns ts−1 (9.129)
0 0
Which allows us to write [64]

∞
X ∞ Z
X ∞ Z ∞ ∞
X
Γ(s)ζ(s) = Γ(s)n−s = dte−nt ts−1 = dtts−1 e−nt (9.130)
n=1 n=1 0 0 n=1
P∞
But n=1 e−nt is just a geometric series,
∞
X e−t 1
e−nt = = t (9.131)
n=1
1 − e−t e −1
Therefore, we find that

∞
ts−1
Z
Γ(s)ζ(s) = dt (9.132)
0 et − 1
Moreover, if we expand the denominator about t = 0 we find that
1 1 1 t
= − + + O(t2 )
et − 1 t 2 12
1
where we used 1+x = 1 − x + x2 + .... We can therefore recast (9.130) as
1 ∞
ts−1 ts−1
Z Z
Γ(s)ζ(s) = dt + dt (9.133)
0 et − 1 1 et − 1
Only the first integral might diverge near t = 0. Let’s rewrite this term as
1 Z 1 Z 1
ts−1
Z
s−1 1 1 1 t s−1 1 1 t
dt t = dtt − + − + dtt − +
0 e −1 0 et − 1 t 12 12 0 t 2 12
Z 1
1 1 1 t 1 1 1
= dtts−1 t−1
− + − + − + (9.134)
0 e t 12 12 s − 1 2s 12(s + 1)
Putting everything together we find
1 ∞
ts−1
Z Z
s−1 1 1 1 t 1 1 1
Γ(s)ζ(s) = dtt − + − + − + + dt (9.135)
0 et − 1 t 12 12 s − 1 2s 12(s + 1) 1 et − 1
We notice there is at least one pole at s = −1. By using the calculus of residues, one can show
that
1
ζ(−1) = − (9.136)
12
1
But this is just what we need to determine the ordering constant a. Using ζ(−1) = − 12 , we find
that
1
a = − (D − 2) (9.137)
24
Soon we will see that for the quantum theory of the open string to include massless photon
states, we will require that D = 26, and therefore a = −1.
Before we get to constructing the state space for the quantum open string, let’s first consider the
−
other transverse Virasoro operators. First, since (αn− )∗ α−n , we’d expect a similar relation to hold
for the transverse Virasoro operators.
1X I 1X I I
(L⊥ ∗
n) = (αn−p αpI )∗ = α−p α−n+p
2 2
p∈Z p∈Z
Then, by letting p → −p, we find that

1X I
(L⊥ ∗
n) = α−n−p αpI = L⊥
−n (9.138)
2
p∈Z
Let’s now move on to the construction of the commutation relations for the transverse Virasoro
operators. First consider
1 X I
L⊥ J
αm−p αpI , αnJ

m , αn =
2
p∈Z
1X I
[αpI , αnJ ] + [αm−p
I
, αnJ ]αpI

= αm−p (9.139)
2
p∈Z
Recall that η IJ = δ IJ . Then, using the commutation relations for the oscillators, we find that
⊥ J 1 X J
+ (m − p)δm−p+n,0 αpJ

Lm , αn = pδp+n,0 αm−p (9.140)
2
p∈Z
Making use of the Kronecker delta, we find that the commutator is reduced to
⊥ J 1 J J J
Lm , αn = (−nαm+n − nαm+n ) = −nαm+n (9.141)
2
It’s unclear whether this commutator holds for m = 0. For that reason, let’s also consider the
commutator of L⊥ J
0 with the oscillators αn .
∞
" #
⊥ J 1 I I X I I J
L0 , αn = α α + α α ,α
2 0 0 p=1 −p p n
∞
1 J I I I J I
X
[αnJ , α−p
I
]αpI + α−p
I
[αnJ , αpI ]

= − [αn , α0 ]α0 + α0 [αn , α0 ] −
2 p=1
∞
1 X
(nη IJ δn,0 )α0I + α0I (nη IJ δn,0 ) − (nη IJ δn−p,0 )αpI + α−p
I
(nη IJ δn+p,0 )

=−
2 p=1
∞
X ∞
X
nη IJ δn−p,0 αpI − I
(nη IJ δn+p,0 )

=− α−p
p=1 p=1
= −nδ IJ αnI = −nαnJ (9.142)

But, had we used (9.143), we would have come up with the same result,
⊥ J
L0 , αn = −nαnJ (9.143)
Therefore, the commutator given in (9.139) holds for all values of m. Let’s now consider the
commutator of two transverse Virasoro operators L⊥ ⊥
m and Ln . We will compute this commutator
step by step, making sure to check that along the way the expressions are normal-ordered, doing so
will lead to some interesting mathematical results imperative to calculations in string theory. To
maintain an operator that is normal-ordered, we start off by splitting L⊥
m as
1X I 1X I I
L⊥
m = αm−k αkI + αk αm−k (9.144)
2 2
k≥0 k<0
It’s easy to see now that for any value of m the right hand side of (9.144) is normal-ordered.
Now evalute
⊥ ⊥ 1 X I 1 X I I
αm−k αkI , L⊥ αk αm−k , L⊥

Lm , Ln = n + n
2 2
k≥0 k<0
1 X I I 1X I I
αm−k , L⊥ αk αm−k , L⊥

= n αk + n
2 2
k≥0 k<0
(9.145)
1X I 1 X I ⊥ I
αm−k αkI , L⊥

+ n + αk , Ln αm−k
2 2
k≥0 k<0
Evaluting each of the commutators yields
1X 1X
L⊥ ⊥ I
αkI + (m − k)αkI αm+n−k
I

m , Ln = (m − k)αm+n−k
2 2
k≥0 k<0
(9.146)
1X I I 1X I I
+ kαm−k αk+n + kαk+n αm−k
2 2
k≥0 k<0
The terms on the first line of (9.146) are always normal-ordered, however the last two terms
require normal-ordering depending on the values of m and n. In fact, there are two possibilities:
(m + n) = 0 and (m + n) 6= 0. Let’s first consider when (m + n) 6= 0. Since all pairs of the oscillators
commute, we may write
⊥ ⊥ 1 X I 1X I
Lm , Ln = (m − k)αm+n−k αkI + I
kαm−k αk+n
2 2
k∈Z k∈Z
If we let k → k − n, the above becomes simplified

⊥ ⊥ 1 X I 1X
Lm , Ln = (m − k)αm+n−k αkI + I
(k − n)αm+n−k αkI
2 2
k∈Z k∈Z
1X I
= (m − n) (m − k)αm+n−k αkI (9.147)
2
k∈Z
Since we maintain that m+n 6= 0, this final operator require no further ordering and is recognized
to be nothing more than L⊥ m+n . Therefore, when m + n 6= 0, we have that
⊥ ⊥
Lm , Ln = (m − n)L⊥
m+n (9.148)
Therefore, for this particular case, the commutator of two Virasoro operators simply returns
another Virasoro operator with a mode that is a sum of the modes of the individual Virasoro
operators. This algebra is formally known as the Virasoro algebra without central extension or is
sometimes referred to as the Witt algebra [64].
Let’s now move on to the case where (m + n) = 0. Here n = −m, giving us
∞
⊥ ⊥ 1X I 1X
Lm , L−m = (m − k)α−k αkI + (m − k)αkI α−k
I
2 2
k=0 k<0
∞
(9.149)
1X I I 1X I I
+ kαm−k αk−m + kαk−m αm−k
2 2
k=0 k<0
To be able to relate the terms to one another, we each of the sums so that the rightmost oscillator
is always αkI . Thus, for the second term we let k → −k, for the third term, k → m + k, and for the
fourth term, k → m − k. We then write
∞ ∞
⊥ ⊥ 1X I 1X
Lm , L−m = (m − k)α−k αkI + I
(m + k)α−k αkI
2 2
k=0 k=1
∞ ∞ (9.150)
1 X
I 1 X
+ (m + k)α−k αkI + I
(m − k)α−k αkI
2 2
k=−m k=m+1
Let’s assume that m > 0. Then all of the terms are normal-ordered except for the third term.
Splitting the sum we find
∞ m ∞
1 X I 1X 1X
(m − k)α−k αkI = (m − k)αkI α−k
I
+ I
(m + k)α−k αkI
2 2 2
k=m+1 k=0 k=1
m m ∞
1X 1X 1X
= (m − k)[αkI , α−k
I
]+ I
(m − k)α−k αkI + I
(m + k)α−k αkI
2 2 2
k=0 k=0 k=1
Now that the term is normal-ordered, we substitute it back in to find that

∞ ∞
⊥ ⊥ X I
X
Lm , L−m = (m − k)α−k αkI + I
(m + k)α−k αkI + (D − 2)C(m) (9.151)
k=0 k=1
where
m m
1X 1 X 1X 2
C(m) = k(m − k) = m k− k (9.152)
2 2 2
k=0 k=1 k=1
By mathematical induction, one can show that

m
X 1
k2 = (2m3 + 3m2 + m) (9.153)
6
k=1
Therefore,
1 2 1 1
C(m) = m (m + 1) − (2m3 + 3m2 + m) = (m3 − m) (9.154)
4 12 12
Then, expanding our sum in (9.151) and substituting in C(m), we find that
∞
!
⊥ ⊥ 1 I I X I I 1
Lm , L−m = 2m α0 α0 + α−k αk + (D − 2)(m3 − m)
2 12
k=0
1
= 2mL⊥
0 + (D − 2)(m3 − m) (9.155)
12
The general result in both cases is then given by
⊥ ⊥ D−2 3
Lm , Ln = (m − n)L⊥
m+n + (m − m)δm+n,0 (9.156)
12
Therefore, a set of transverse Virasoro operators L⊥ n satisfying (9.156) defines the centrally
extended Virasoro algebra. The second term in (9.156) is known as the central extension. This
algebra is arguably the most important algebra in string theory. For that reason let’s explore a
little more of the mathematical structure of this algebra.
The Virasoro Algebra is in fact a Lie Algebra. To show this explicitly, first recall one of the
“definitions” of a Lie algebra. A vector space L with elements x, y, z and a bilinear bracket [...]
that takes two elements of L and returns another element of L is a Lie algebra if the following two
conditions hold [64]:
(i) Antisymmetry: [x, y] = [−y, x], ∀x, y ∈ L;
(ii) The Jacobi Identity: [x, [y, z]] + [y, [z, x]] + [z, [x, y]] = 0∀x, yz ∈ L.
Let’s consider a vector space L spanned by the Virasoro operators. First notice for the Virasoro
algebra without central extension, satisfied by (9.148) gives us
L⊥ ⊥ ⊥ ⊥
⊥ ⊥
n , Lm = (n − m)Ln+m = −(m − n)Lm+n = − Lm , Ln
Thus, the bilinear bracket is antisymmetric. Moreover, notice that

⊥ ⊥ ⊥ ⊥ ⊥ ⊥ ⊥ ⊥ ⊥
Lm , Ln , Lk + Ln , Lk , Lm + Lk , Lm , Ln
= L⊥ ⊥
⊥ ⊥
⊥ ⊥

m , (n − k)Ln+k + Ln , (k − m)Lk+m + Lk , (m − n)Lm+n
= [(n − k)(m − n − k) + (k − m)(n − k − m) + (m − n)(k − m − n)]L⊥

m+n+k = 0
Therefore the Jacobi identity is satisfied. Therefore, by our definition, the Virasoro algebra
without central extension is indeed a Lie algebra. Let’s move on to the Virasoro algebra with
central extension, governed by (9.154). Its rather trivial to show that certainly the commutator is
antisymmetric,
⊥ ⊥
Ln , Lm = − L⊥ ⊥

m , Ln
Moreover,
⊥ ⊥ ⊥ ⊥ ⊥ ⊥ ⊥ ⊥ ⊥
Lm , Ln , Lk + Ln , Lk , Lm + Lk , Lm , Ln
= L⊥ ⊥
⊥ ⊥
⊥ ⊥

m , (n − k)Ln+k + Ln , (k − m)Lk+m + Lk , (m − n)Lm+n
=[(n − k)(m − n − k) + (k − m)(n − k − m) + (m − n)(k − m − n)]L⊥ m+n+k

1 (9.157)
[(n − k)(m3 − m) + (k − m)(n3 − n) + (m − n)(k 3 − k)]δm+n+k,0 = 0
12
Hence, the Virasoro algebra with central extension is also a Lie algebra. Though we won’t prove
it here, the operators L⊥ ⊥ ⊥
0 , L1 and L−1 actually generate a subalgebra of the Virasoro algebra.
This subalgebra is the SL(2, R) subalgebra, which is the noncompact form of the SU (2) algebra
imperative to the standard model [5].
Let’s move on and consider the commutators of the transverse Virasoro operators with the string
coordinates. First of all, consider
⊥ I 1 X I
Lm , x0 = [αm−p αpI , xJ0 ]
2
p∈Z
1X J I
[x0 , αm−p ]αpI + αm−p
I
[xJ0 , αpI ]

=−
2
p∈Z
1 √ √ √
=− 2α0 iδ IJ αm
I
+ 2α0 iδ IJ αm
I I
= −i 2α0 αm (9.158)
2
From here we can compute the commutator of L⊥ I
m with X (τ, σ). Notice
⊥ I √ X1
Lm , X (τ, σ) = L⊥ cos(nσ) L⊥
I
I
−inτ
0
m , x0 + i 2α m , αn e
n
n6=0
√ √ X
I
−i 2α0 αm − i 2α0 cos(nσ)e−inτ αm+n
I
(9.159)
n6=0
which may be written as a single sum

⊥ I √ X
cos(nσ)e−inτ αm+n
I

Lm , X (τ, σ) = −i 2α0
n∈Z
√ 1 X −in(τ −σ)
= −i 2α0 e + e−in(τ +σ) αm+n
I
2
n∈Z
√ 1 X −i(n−m)(τ −σ)
= −i 2α0 e + e−i(n−m)(τ +σ) αnI (9.160)
2
n∈Z
where we let n → n − m. Finally then,
⊥ I i √ X i √ X
Lm , X (τ, σ) = − eim(τ −σ) 2α0 e−in(τ −σ) αnI − eim(τ +σ) 2α0 e−in(τ +σ) αnI

(9.161)
2 2
n∈Z n∈Z
To interpret this result we write the above using the derivative expressions, (9.35) and (9.36),
yielding
⊥ I i i
Lm , X (τ, σ) = − eim(τ −σ) (Ẋ I − X 0I ) − eim(τ +σ) (Ẋ I + X 0I )

2 2
= −ieimτ cos(mσ)Ẋ I + eimτ sin(mσ)X 0I (9.162)

Taking the form
⊥ I τ
Ẋ I + ξm
σ
X 0I

Lm , X (τ, σ) = ξm (9.163)
with
τ
ξm (τ, σ) = −ieimτ cos(mσ) (9.164)
σ
ξm (τ, σ) = eimτ sin(mσ) (9.165)
As it turns out, the Virasoro operators generate reparameterizations of the worldsheet [64].
Briefly, notice that when m = 0 we have ξ0τ = −i and ξ0σ = 0, so
⊥ I
L0 , X = −i∂τ X I (9.166)
I
which is just the Heisenberg equation of motion for X . Indeed, as we will see shortly, the
Virasoro operator L⊥0 is the string Hamiltonian up to an additive constant, and therefore must
generate time translations.
Let’s now move on to the closed string Virasoro operators. The analysis is quite similar to the
open string, so most of the computations are the same. We will therefore quickly go through the
analysis to reach a few interesting differences from the open string.
To begin, we start by relating X − to the transverse coordinates X I . Since we are dealing with
closed strings, we make note of the fact that β = 1, from our gauge conditions for the closed string.
Then, we find
1 1 2
Ẋ − ± X 0− = 0 +
Ẋ I ± X 0I (9.167)
α 2p
We define the closed string Virasoro operator in a similar manner just as we had for the open
string:
 
0I 2
X 1X X
Ẋ I + X = 4α0 α̃pI α̃n−p
I  e−in(τ +σ) ≡ 4α0 L̃⊥ −in(τ +σ)

ne (9.168)

2
n∈Z p∈Z n∈Z
 
0I 2
X 1 X X
Ẋ I − X = 4α0 αpI αn−p
I  e−in(τ −σ) ≡ 4α0 L⊥ −in(τ −σ)

ne (9.169)

2
n∈Z p∈Z n∈Z
where
1X I I
L̃⊥
n = α̃p α̃n−p (9.170)
2
p∈Z
1X I I
L⊥
n = αp αn−p (9.171)
2
p∈Z
Substituting (9.170) and (9.171) back into (9.167), we find
2 X ⊥ −in(τ +σ)
Ẋ − + X 0− = L̃n e (9.172)
p+
n∈Z
2 X ⊥ −in(τ −σ)
Ẋ − − X 0− = Ln e (9.173)
p+
n∈Z
Alternatively, we may also write (9.172) and (9.173) using (9.74) and (9.75), yielding
√ X
Ẋ − + X 0− = 2α0 α̃n− e−in(τ +σ) (9.174)
n∈Z
√ X
Ẋ − − X 0− = 2α0 αn− e−in(τ −σ) (9.175)
n∈Z
Comparing these expressions leads us to find

√ 2 ⊥
2α0 α̃n− = L̃ (9.176)
p+ n
and
√ 2 ⊥
2α0 αn− = L (9.177)
p+ n
Noptice for n = 0, since α0− = α̃0− , we have the level matching condition
L⊥ ⊥
0 = L̃0 (9.178)
To interpret this condition, we have that any state |λ, λ̃i of closed string theory must satisfy
L⊥ ⊥
0 |λ, λ̃i = L̃0 |λ, λ̃i (9.179)
Thus, the level matching condition is actually a constraint on the state space of the theory of
closed strings. Moreover, just as we had for the open string, there is an ambiguity in the ordering of
the Virasoro operator. To fix this, we define the Virasoro operators L⊥ ⊥
0 and L̃0 to be normal-ordered
without any additional constants:
α0 I I
L̃⊥
0 = p p + Ñ ⊥ (9.180)
4
α0 I I
L⊥
0 = p p + N⊥ (9.181)
4
where we have introduced the familiar number operators Ñ ⊥ , N ⊥ as
∞
X
Ñ ⊥ ≡ nãI∗ I
n ãn (9.182)
n=1
∞
X
⊥
N ≡ naI∗ I
n an (9.183)
n=1
This ordering constraint requires that the ordering constants for L⊥ ⊥

0 and L̃0 are the same and
⊥
equal to the L0 from open string theory. But this makes intuitive sense since the left and right
sectors of closed string theory behave like open strings. Shortly, after we detemine the ordering
constant and space-time dimension for the open string, we will see how the ordering of L⊥ 0 and L̃0
⊥
modifies the mass spectrum of the closed string and closed string Hamiltonian. Before we get there
however, let’s first ascertain the commutators of the Virasoro operators.
Since closed string theory is really a theory describing the left and right sectors of open string
theory, it’s no surprise that both L⊥ ⊥
m and L̃m satisfy the Virasoro algebra we examined in detail
earlier. First notice that we have the commutators
⊥ J J
L̃m , α̃n = −nα̃m+n (9.184)
⊥ J J
Lm , αn = −nαm+n (9.185)
Moreover, since the left and right moving sectors don’t mix, we also have
⊥ J ⊥ J
L̃m , αn = Lm , α̃n = 0 (9.186)
Moreover, we can compute the commutation relations of the Virasoro operators with the trans-
verse string coordinates. First of all, notice
 
⊥ I 1 X
L̃m , x0 =  α̃pI α̃m−p
I
, xI0 
2
p∈Z
1X I I I
[x0 , α̃p ]α̃m−p + αpI [xI0 , α̃m−p
I

=− ]
2
p∈Z
q
0
Then, making use of [xI0 , α̃nJ ] = 0 and [xI0 , α̃0J ] = i α2 δ IJ , we find that the above simply becomes
1
[xI0 , α̃0I ]α̃m
I I
[xI0 , α̃0I ]

=− + αm
2
r
α0 I
= −i α̃
2 m
Therefore,
r
⊥ I α0 I
L̃m , x0 = −i α̃ (9.187)
2 m
A similar calculation shows
r
⊥ I α0 I
Lm , x0 = −i α (9.188)
2 m
Let’s now focus on the zero mode transverse Virasoro operators L⊥ ⊥
0 and L̃0 . The commutator
⊥ I
of L̃0 with the transverse string coordinate X is just
 
√
r
⊥ I α 0 X e−inτ
L̃0 , X (τ, σ) = L̃⊥ I
2α0 α0I τ + i (αnI einσ + α̃nI e−inσ )

0 , x0 +
2 n
n6=0
⊥ I h ⊥ √
r
i α0 X e−inτ ⊥ I
L̃0 , x0 + L̃0 , 2α0 α0I τ + i L̃0 , α̃n
2 n
n6=0
r r
α0 I α0 X −in(τ +σ) I
= −i α̃ − i e α̃n
2 0 2
n6=0
i √
r
α0 X −in(τ +σ) I X
= −i e α̃n = − √ 2α0 e−in(τ +σ) α̃nI
2 2
n∈Z n∈Z
i 0
= − (Ẋ I + X I )
2
Therefore,
⊥ I i 0
L̃0 , X (τ, σ) = − (Ẋ I + X I )

(9.189)
2
Similarly,
⊥ I i 0
L0 , X (τ, σ) = − (Ẋ I − X I )

(9.190)
2
If we add (9.189) and (9.190) together we find that
L̃⊥ ⊥ I I

0 + L0 , X (τ, σ) = −i∂τ X (9.191)
which is consistent with the Heisenberg operator equation of motion. Moreover, subtracting
(9.189) and (9.190) yields
L⊥ ⊥ I I

0 − L̃0 , X (τ, σ) = i∂σ X (9.192)
Equation (9.191) generates τ translations and therefore the operator L̃⊥ ⊥
0 + L0 can be viewed as
⊥ ⊥
the world-sheet energy. Moreover, (9.192) indicates that the operator L0 − L̃0 generates constant
translations along the string. More generally, one can show that
X I (τ, σ + σ0 ) = e−iP σ0 X I (τ, σ)eiP σ0 (9.193)

where P ≡ L⊥
0− L̃⊥
0.(9.193) is simply a finite translation along the string. Additionally,
since L⊥
0 − L̃⊥
0 generates translations along the world-sheet coordinate σ, it can be viewed as the
world-sheet momentum [64].
9.5 Lorentz Generators and the Dimension of Space-Time

In an earlier chapter we calculated the angular momentum generators J µν to be
∞
X 1 µ ν
J µν = xµ0 pν − xν0 pµ − i ν
(α−n αn − α−n αnµ ) (9.194)
n=1
n
We will use this expression to suggest the form of the quantum Lorentz generators. In the light-
cone gauge, as we saw in the last chapter, the most delicate quantum Lorentz generator is J −I
since it is a nontrivial function of X I . A consistent J −I must generate Lorentz transformations on
the string coordinates, accompanied by the commutator
[J −I , J −J ] = 0 (9.195)
Requiring this to be true in string theory leads to something remarkable: the dimension of space-
time! First however, let’s find the correct form of J −I . Using (9.194), we’d suspect that J −I takes
the form
∞
X 1 − I
J −I = x− I I −
0 p − x0 p − i
I
(α−n αn − α−n αn− ) (9.196)
n=1
n
But we notice that the second term is not Hermitian since xI0 and p− don’t commute. Therefore,
we rewrite the term as
∞
1 I − X 1 − I
J −I = x−
0 pI
− (x 0 p + p − I
x0 ) − i I
(α−n αn − α−n αn− ) (9.197)
2 n=1
n
9.5. LORENTZ GENERATORS AND THE DIMENSION OF SPACE-TIME 191
Now remember that p− can be written terms of the Virasoro operator L⊥

0 and the ordering
constant a. Making this substitution we find that
∞
1 i X 1 ⊥ I
J −I = x− I I ⊥ ⊥
L⊥
I I

0p − 0 +
(x 0 L 0 + a + L0 + a x 0 ) − √ L−n αn − α−n n (9.198)
4α p 2α p n=1 n
0 +
After a long, tedious calculation, the commutator of two of such Lorentz charges is [64]
∞
1 X I
[J −I , J −J ] = − (α αJ − α−m
J I
αm )
α0 p+2 m=1 −m m
(9.199)
1 1 1
× m 1 − (D − 2) + (D − 2) + a
24 m 24
We immediately notice that the only way for this commutator to be zero, we require that

1 1 1
m 1 − (D − 2) + (D − 2) + a = 0 (9.200)
24 m 24
∀m ∈ Z+ . Moreover, just by considering when m = 1 and m = 2, we conclude that each term
must zero. Thus,
1
1− (D − 2) = 0 (9.201)
24
1
(D − 2) + a = 0 (9.202)
24
The first of these equations fixes the dimension of space-time:
D = 26 (9.203)
Thus, in order to maintain Lorentz invariance in bosonic string theory, we require that the di-
mension of space-time is 26, i.e., 25 spatial dimensions and one time dimension. The high dimension
of space-time does seem a little absurd, and more like a bit of mathematical magic than physical
insight. However, remember that we are dealing with bosonic string, so already our theory is not
realistic. We must also include fermions. When we do this however, the number of space-time
dimensions is reduced to 10, and according to M-theory, the space-time dimension is 11. As of
now, the superstring theories and M-theory seem to be the most promising apporaches toward a
unifying theory, particularly amongst those in the string theory community. At some point one has
to ask whether or not the dimension of space-time, indeed a radical change in perspective, is an
accurate description of our own reality. A lot of time spent by some string theorists is to solve this
very problem by compactifying the extra dimensions in such a way that the observed universe is 4
dimensional, though there might be hidden dimensions. We will discuss this more later on. Keep in
mind however, that the presence of extra dimensions has turned many theoretical physicists away
from string theory.
The second equation yields the value of the ordering constant:
a = −1 (9.204)
Thus,
1
2α0 p− = (L⊥ − 1) (9.205)
p+ 0
The open string Hamiltonian is therefore
H = L⊥
0 −1 (9.206)
Now that we know the ordering constant, we can apply the result to closed string theory as well.
Using (9.176) and (9.177) for n = 0, we include the ordering constants, yielding
√ 2
2α0 α̃0− = (L̃⊥ − 1) (9.207)
p+ 0
√ 2
2α0 α0− = (L⊥ − 1) (9.208)
p+ 0
Notice then we have
√ 1
2α0 α0− ≡ (L⊥ + L̃⊥ 0 −
0 − 2) = α p (9.209)
p+ 0
The mass-squared operator in closed string theory then becomes
2 ⊥
M 2 = −p2 = 2p+ p− − pI pI = (L + L̃⊥ I I
0 − 2) − p p (9.210)
α0 0
Or, in terms of the number operators of the left and right sectors,
2
M2 = (N ⊥ + Ñ ⊥ − 2) (9.211)
α0
Lastly, the closed string Hamiltonian is simply
H = α0 p+ p− = L⊥ ⊥
0 + L̃0 − 2 (9.212)
which is just the sum of open string Hamiltonian for the right moving and left moving operators.
In terms of the number operators, the closed string Hamiltonian is written as
α0 I I
H= p p + N ⊥ + Ñ ⊥ − 2 (9.213)
2
Though we won’t prove it here, the dimension of space-time is also found to be D = 26 for the
closed bosonic string. The fact that string theory is only a Lorentz invariant quantum theory in a
particular dimension of space-time exemplifies the fact string theory is very constrained. Whether
this fact of string theory is correct is yet to be determined, though if it were found to be true, our
notion of space-time would undergo yet again a radical change.
9.6. CONSTRUCTING THE STATE SPACE OF OPEN STRING THEORY 193
9.6 Constructing the State Space of Open String Theory

The classical open string does not provide an accurate physical picture since is suggests the mass
of the string states take on a continuous range values. Moreover, only the ground state is massless,
that is, it does not allow for any massless excitations. This would mean that photons couldn’t exist.
Quantum string theory solves this problem, and correctly identifies the photon states.
Let’s begin by introducing the ground states of the open quantum string. The quantum string
actually shares the same zero modes with the quantum point particle developed in the previous
chapter. Moreover, we have the canonical pairs (xI0 , pI ) and (x− +
0 , p ). Just as we did for the
quantum point particle, we introduce the ground states of the string as |p+ , p~T i. These ground
states are also declared to be the vacuum states for all oscillators in string theory. Therefore, by
definition of a vacuum state, we have that
aIn |p+ , p~T i = 0 (9.214)

for n ≥ 1 and for I = 2...25. We can create string states from |p+ , p~T i just as we did when
created a Fock space during our excursion in quantum field theory: we act on |p+ , p~T i using the
creation operators to obtain string states. Just like the multi-particle states from scalar fields, the
general basis state is
∞ Y
Y 25
|λi = (aI∗
n )
λn,I +
|p , p~T i (9.215)
n=1 I=2
where the non-negative number λn,I denotes the number of times the creation operator aI∗ n
appears. The string Hilbert space then is an infinite dimensional vector space spanned by an
infinite set of linearly independent basis states |λi [64]. To physically interpret the state space, let’s
consider the mass-squared operator using the ordering constant a = −1,
∞
!
2 1 X I∗ I
M = 0 na a − 1 (9.216)
α n=1 n n
Or, written in terms of the number operator N ⊥ ,

1
(N ⊥ − 1)
M2 = (9.217)
α0
Since the number is normal-ordered, it annihilates the ground states:
N ⊥ |p+ , p~T i = 0 (9.218)

Moreover, we may also write the zero mode transverse Virasoro operator in terms of the number
operator
L⊥ 0 I I
0 =α p p +N
⊥
(9.219)
In general, the action of N ⊥ on a basis string state is [64]:
N ⊥ |λi = Nλ⊥ |λi (9.220)

where
∞ X
X 25
Nλ⊥ = nλn,I (9.221)
n=1 I=2
Moreover, for each basis state |λi there is also an associated Hermitian conjugate bra vector hλ|,
written as
∞ Y
Y 25
hλ| = hp+ , p~T | (aIn )λn,I (9.222)
n=1 I=2
Computing the inner product of any two basis states, one can show that the inner product gives
positive norm, allowing for a probablistic interpretation. In addition, for each basis state, we expect
that the time dependent state appear as
exp(−iHτ )|λi = exp(−i(L⊥

0 − 1)τ )|λi (9.223)
which indeed satisfies the Schrödinger equation with using the string Hamiltonian. Now that we
have seen how to construct a general state space of the open string, let’s examine in detail some of
the particular states. First consider the ground states, in which N ⊥ = 0. These are the states of
the scalar particle. The mass of these states are simply
1 1
M 2 |p+ , p~T i = (N ⊥ − 1)|p+ , p~T i = − 0 |p+ , p~T i (9.224)
α0 α
Therefore we find that M 2 = − α10 < 0. Imaginary mass? The scalar field which has negative
mass-squared corresponds to a tachyon. It turns out that the presence of tachyons, aside from their
strange nature when it comes to causality, represent an instability. It turns out these instabilities
are actually the instability of a D-Brane, causing D-Brane decay [64]. Again, so far we have only
considered bosonic strings, and therefore we aren’t examining a realistic theory. But the problem of
existing open tachyon states do show up when studying some superstring theories. In particular, in
string cosmology one of the proposed ideas in describing the early universe is through the collision
of branes and anti-branes. It turns out that the open strings streching between the D-brane/anti-
D-brane pairs contain a superstring tachyon. Therefore the study of open string tachyons is still
underway as it might play a prominent role in study of the beginning and evolution of our universe.
Let’s now consider excited states with the lowest possible mass-squared. For this we let N ⊥ = 1,
in turn yielding M 2 = 0. We get these such states when we act with any of the transverse oscillators
a1I∗ on the ground states. In fact, we have D − 2 = 24 such states. A general massless state is
simply a linear combination of these basis states:
25
X
ξI aI∗ +
1 |p , p
~T i (9.225)
I=2
But these massless states look familiar. Recall the photon states we derived in chapter six:
D−1
X
ξI aI∗ pT |0i
p+ ,~ (9.226)
I+2
By matching the basis states

9.7. CONSTRUCTING THE STATE SPACE OF CLOSED STRING THEORY 195
aI∗ +
~T i ←→ aI∗
1 |p , p pT |0i
p+ ,~ (9.227)
and since both states are massless and have the same Lorentz labels carrying the same momenta,
we have correctly identified photon states. That is, open string theory quantum states include
photon states! The Lorentz index here is important as it implies that the states given in (9.227)
may transform into one another via the correct Lorentz transformation.
Each state |λi of quantum string theory represents a one-particle state of fixed momentum.
Multi-particle states are described using quantum fields. The total quantum field theory which
gives a description of the quantum fields associated with the one-particle states of string theory
is known as string field theory. Though interesting, we won’t have time to discuss the details of
string field theory, however the reader is encouraged to refer to the references for more details on
this subject.
9.7 Constructing the State Space of Closed String Theory

Let’s build the state space of closed string theory just as we did for open string theory. This time
the ground states |p+ , p~T i are annihilated by both the left and right moving annihilation operators.
To generate all of the basis states we must act on the ground state with the creation operators aI∗ n
and ãI∗
n . The general basis state is then [64]:
∞ Y ∞ Y
25
! 25
!
Y Y
I∗ λn,I J∗ λ̃m,J
|λ, λ̃i = (an ) × (am ) |p+ , p~T i (9.228)
n=1 I=2 m=1 J=2
Again, λ̃m,J and λn,I are the non-negative occupation numbers. Just as for the open string, the
number operators act on the general basis state with eigenvalues
∞ X
X 25
Nλ⊥ = nλn,I (9.229)
n=1 I=2
∞ X
X 25
Ñλ̃⊥ = nλ̃n,I (9.230)
n=1 I=2
It’s important to note that note all of the states given in (9.228) are part of the closed string
state space. The level matching condition yields that N ⊥ = Ñ ⊥ . States of the closed string state
space must satisfy this condition, and hence only the states where this condition holds are part of
the close string state space. Some of the states given in (9.228) do not satisfy the level matching
condition and are therefore not part of the closed string state space.
To understand the state space physically, let’s identify some of the states in particular. Just as
they were in open string theory, the ground states happen to be one particle states of a quantum
scalar field. Such states satisfy the level matching condition N ⊥ = Ñ ⊥ = 0. Thus, using
1 0 2
α M = N ⊥ + Ñ ⊥ − 2 (9.231)
2
we find that for the ground states we again have a negative mass-squared: M 2 = − α40 < 0.
These the closed string tachyons. Again, the presence of these tachyons indicate instabilities,
however these tachyons aren’t that well known, however appear to represent instabilities in space-
time itself! Some have even claimed that to solve to issue of singularities occurring in general
relativity can be resolved using the phase of a closed string tachyon condensate [54]. We won’t go
into detail on this here, however an open problem in string theory is understanding closed string
tachyons and the instabilities they give rise to.
The next excited states must be built with two oscillators acting on the ground state: an oscillator
from sector is require. Moreover, since we must obey the level matching condition, the next excited
state occurs when N ⊥ = Ñ ⊥ = 1, in turn yielding M 2 = 0. The general massless state of fixed
momentum is simply written as
XX
RIJ aI∗ J∗ +
1 ã1 |p , p
~T i (9.232)
I J
where RIJ is a square matrix of size (D − 2). This matrix can be decomposed into a symmetric
part and antisymmetric part:
1 1
RIJ = (RIJ + RJI ) + (RIJ − RJI ) = SIJ + AIJ
2 2
The symmetric part can be decomposed even further:

1 1
SIJ = SIJ − δIJ S + δIJ S
D−2 D−2
where
S = S II = δ IJ SIJ
Notice that the first term on the right hand side is traceless:

IJ 1 1
δ SIJ − δIJ S = S − δIJ δ IJ S = 0
D−2 D−2
since δIJ δ IJ = D−2. Thus, the matrix SIJ is decomposed intro a traceless matrix and a multiple
of the unit matrix. By letting ŜIJ denote the traceless part of SIJ and S 0 = D−2
S
, the matrix RIJ
is decomposed as
RIJ = ŜIJ + AIJ + S 0 δIJ (9.233)

Each of piece of the decomposition above and be specified independently when writing RIJ ,
thereby splitting the states in (9.232) into three different groups:
XX
ŜIJ aI∗ J∗ +
1 ã1 |p , p
~T i (9.234)
I J
XX
AIJ aI∗ J∗ +
1 ã1 |p , p
~T i (9.235)
I J
S 0 aI∗ I∗ +
1 ã1 |p , p
~T i (9.236)
Again, we some familiarity with the above states in (9.234) and states we studied in chapter 7.
Recall the one-particle graviton states. In the quantum theory of the gravitational field we found
that the quantum states are given by
9.7. CONSTRUCTING THE STATE SPACE OF CLOSED STRING THEORY 197
D−1
X
ξIJ aIJ∗ pT |0i
p+ ,~ (9.237)
I,J=2
where ξIJ is an arbitrary symmetric,traceless matrix. Since ŜIJ is also a symmetric, traceless
matrix, we may identify these two sets of states by identifying the basis states:
aI∗ J∗ +
~T i ←→ aIJ∗
1 ã1 |p , p pT |0i
p+ ,~ (9.238)
Indeed we are allowed to make this identification since the the two sets of states have the same
Lorentz labels, carry the same momentum, and have the same mass. But what did we just show?
Closed string theory includes graviton states! Gravity has emerged naturally in string theory. We
did not have to put gravity in by hand, but rather, when we quantized the closed string, the quantum
states associated with the gravitational field popped into existence. When this was discovered many
physicists realized that string theory might not just be a theory to better understand hadrons, but
a theory of quantum gravity, and at the same time, be a theory which unifies the fundamental
forces of nature! No doubt this is a big result. This result alone has encouraged some physicists
to claim that string theory already makes predictions: predicting gravity. As important of a result
as it may be, the authors of this text maintain that it alone does not justify string theory as being
the ultimate theory physicists have been waiting for. Rather, it seems to only suggest that string
theory appears to be on the right track.
Moving on, the set of states in (9.235) correspond to the one-particle states of the Kalb-Ramond
field, an antisymmetric tensor field Bµν . This field can be viewed, in a sense, as the tensor gener-
alization of the Maxwell gauge field Aµ . Though we won’t discuss it in this text, it turns out that
strings carry Kalb-Ramond charge [64].
Equation (9.236), since it has no free indices, represents a single state, corresponding to a one-
particle state of a massless scalar field called the dilaton. It turns out that the dilaton field has the
interesting property that its expectation value controls the string coupling constant, a dimensionless
number which sets the strength of string interactions.
Before moving on, let’s briefly summarize the findings of the last two sections. When we con-
structed the state space of the open string, we identified open string tachyons and the one-particle
photon states. For closed string theory, the constructed state space revealed closed string tachyons,
one-particle graviton states, one-particle states of the Kalb-Ramond field, and the dilaton. Based
on this brief analysis it is no wonder that string theorists often give the analogy of a guitar string.
As the guitar string vibrates, different modes refer to different notes. In a figurative, the various
particle states of string theory are just the different vibrational modes of a string.
There is one issue that we haven’t discussed yet. In the past couple sections we only considered
a couple of string states, the two lowest energy states. We could have continued this analysis ad
infinitum, revealing a state space which has far more particles then we have even come close to
observing in the lab. In order to reconcile string theory with our experiments and the standard
model, we need to figure out how to deal with such an enormous state space.
9.8 Unoriented Strings

Up to this point we have only considered oriented strings. Unoriented strings do appear in string
theory as well as oriented strings, so let’s briefly detail what we mean by oriented and unoriented
strings. First consider the open string. The open string X µ (τ, σ), with σ ∈ [0, π] and fixed τ is a
parameterized curve in space-time. The orientation of a string is the direction of increasing σ on
this string.
Consider now the open string X µ (τ, π − σ) at the same fixed τ . Notice then that as σ goes
from 0 → π, the coordinate π − σ goes from π → 0. Therefore it’s easy to see that the string
X µ (τ, π − σ) covers the same set of space-time points as the string X µ (τ, σ), however the endpoints
are interchanged and the orientation is reversed.
Let’s now introduce an orientation reversing twist operator Ω such that
ΩX I (τ, σ)Ω−1 = X I (τ, π − σ) (9.239)

Let’s also declare that the variables x−
0 and p+ are invariant under the action of this twist
operator [64]:
Ωx−
0Ω
−1
= x−
0 (9.240)
Ωp+ Ω−1 = p+ (9.241)

Using the open string expansion of the transverse coordinates (9.34), and keeping in mind that
cos(n(π − σ)) = (−1)n cos(nσ), it’s easy to find that
ΩxI0 Ω−1 = xI0 (9.242)
Ωα0I Ω−1 = α0I (9.243)
ΩαnI Ω−1 = (−1)n αnI (9.244)

Moreover, provided Ωx−
0Ω
−1
= x−
0, and
1 1 ⊥ −1
Ωαn− Ω−1 = Ω √ Ln Ω
2α0 p+
1 1 X
=√ (−1)n−p αn−p
I
(−1)p αpI = (−1)n αn−
2α0 2p+ p∈Z
we also find that
ΩX − (τ, σ)Ω−1 = X − (τ, π − σ) (9.245)

Also, since
ΩX + (τ, σ)Ω−1 = X + (τ, π − σ) (9.246)

we see that (9.239) holds for all string coordinates. We say that orientation reversal is a symmetry
of open string theory because it also leaves the string Hamiltonian invariant:
9.8. UNORIENTED STRINGS 199
ΩHΩ−1 = H (9.247)
We may also assume that the twist operator leaves the ground states invariant
Ω|p+ , p~T i = Ω−1 |p+ , p~T i = |p+ , p~T i (9.248)

Notice then that if we operator on a general basis state from open string theory we have that
∞ Y
25
Y
−1
λn,I
Ω|λi = ΩaI∗
n Ω Ω|p+ , p~T i
n=1 I=2
P
λ ⊥
= (−1) n,I n,I |p+ , p~T i = (−1)Nλ |λi
Notice then that the states with N ⊥ = 0, are ground states with twist +1, and with N ⊥ = 1 are
ξI aI∗ +
1 |p , p
~T i with twist (-1). Why do we care about the twist? We say that a state is unoriented if
its is invariant under the twist operator. Therefore, if we were to construct a theory of unoriented
strings we would only keep the states where Ω|ψi = +|ψi. In that sense, we notice that for an
unoriented open string theory, we would not keep the photon states. But if we lose these photon
states, why would we even care to build an unoriented string theory? It turns out that we can
actually construct consistent interacting string theories using only the unoriented states.
Let’s now move on to consider unoriented closed string theories. A closed string X µ (τ, σ), with
σ ∈ [0, 2π] and fixed τ is a parameterized closed curve in space-time. The orientation of a closed
string, just like the open string, is the direction of increasing σ on the curve.
Now consider the closed string X µ (τ, 2π − σ) with the same τ . Analogous to the open string,
the two strings are related in the sense that they cover the same points, however going in opposite
directions (orientation reversed). Just as before, let’s introduce an orientation reversing twist
operator Ω such that
ΩX I (τ, σ)Ω−1 = X I (τ, 2π − σ) (9.249)

also with [64]
Ωx−
0Ω
−1
= x−
0 (9.250)
Ωp+ Ω−1 = p+ (9.251)

If we use the oscillator expansion of the closed string (9.73), a short exercise reveals
ΩxI0 Ω−1 = xI0 (9.252)
Ωα0I Ω−1 = α0I (9.253)
ΩαnI Ω−1 = (−1)n αnI (9.254)
Ωα̃nI Ω−1 = (−1)n α̃nI (9.255)

Moreover, just as shown for the open string, a short calculation indicates that ΩX − (τ, σ)Ω−1 =
X (τ, 2π − σ) and ΩX + (τ, σ)Ω−1 = X + (τ, 2π − σ) illustrating that (9.249) actually holds for all
−
string coordinates. Again, orientation reversal is a symmetry of closed string theory since it leaves
the string Hamiltonian invariant.
When we construct a theory of closed unoriented strings, we only include states that are invariant
under the twist operator Ω. When we do this we end up losing the Kalb-Ramond states, since when
the twist operator acts on the Kalb-Ramond states it has a twist of -1. Lastly, when we construct
interacting string theories, we can have only closed strings, closed and open strings, but not only
open strings. This is because in the scattering of open strings closed strings naturally emerge, i.e.,
open strings can form closed strings. A second constraint we have on interacting string theories is
that oriented strings must couple with oriented strings and unoriented strings can only couple to
unoriented strings [45].
As a summary, below we list the possible combinations of interacting theories where we include
the Gµν -graviton, Bµν particle of Kalb-Ramond field, Φ-dilaton, and Aµ -photon:
1. Closed Oriented Bosonic string includes: Gµν , Bµν , Φ
2. Closed Unoriented Bosonic string includes: Gµν , Φ
3. Closed and Open Oriented Bosonic string includes: Gµν , Bµν , Φ, Aµ
4. Closed and Open Unoriented Bosonic string includes: Gµν , Φ
We have finally completed the task of quantizing the string. When we quantized the string we
observed several remarkable (perhaps even troubling) consequences. First and foremost, when we
quantized the closed string, we inherited gravity for free. It naturally emerged rather than us putting
it in by hand. A second, just as interesting consequence, is that when we quantized the open string
and required Lorentz invariance, the price we paid is that we require a 26-dimensional background
space-time. Again, we only require a 26-dimensional space-time in bosonic string theory, which
is a non-realistic theory as it only includes bosons. However, when we introduce fermions into
our theory the number of dimensions is reduced to 10 dimensions or 11 dimensions if we consider
M-theory. So the issue of extra dimensions is a little better but still present. Later on, we will
briefly discuss the efforts many physicists have taken to solve the issue of hidden extra dimensions
via a topological tool known as compactification.
In this chapter we quantized the string using the light-cone gauge. We chose this gauge since it
readily gives us physical results, however it does not imply manifest Lorentz invariance. Typically
in string theory one instead learns Lorentz Covariant quantization at first as it achieves manifest
Lorentz invariance. We will detail this quantization procedure in a little detail in the next chapter.
9.9 Exercises
1. Verify equation (9.33). Then, using σ ∈ [0, 2π], go through the details to show that (9.32) and
(9.33) are also satisfied for closed strings.
2. Verify (9.53).
9.9. EXERCISES 201
3. A more complete definition for a Lie algebra is: L, a real or complex vector space with a law
of composition given by the bracket [X, Y ], is a Lie algebra if the following are satisfied:
(i) [X, Y ] = −[Y, X]

(ii) [X, aY + bZ] = a[X, Y ] + b[X, Z]
(iii) [X, [Y, Z]] + [Y, [Z, X]] + [Z, [X, Y ]] = 0
(a) Check the Virasoro operators for both the open and closed string satisfy property (ii).
(b) Determine whether the Poisson bracket given in the last chapter constitutes a Lie algebra
(Hint: Don’t bother showing property (iii), although it does hold. Rather check the first two
properties. To show that the Poisson bracket satisfies the Jacobi identity takes more work with
mathematical concepts not yet discussed.).
4. Consider the operators L⊥ ⊥ ⊥

0 , L1 , and L−1 . Use (9.156) to determine the commutators of each
of these operators. Do they form a subalgebra? How does each operator act on the vacuum |0i?
Chapter 10
Lorentz Covariant Quantization
10.1 The Covariant Formalism

Let’s start off by saying that, just as there are a few different quantization procedures in quantum
field theory, string theory too has various methods one may use to quantize the string. So far,
there are essentially three main methods: Light-cone gauge quantization, covariant quantization
and BRST quantization. In the last chapter we quantized the string using the light-cone gauge.
Certainly this quantization procedure resulted in a Lorentz invariant quantum theory, however the
Lorentz symmetry was not made manifest using light-cone coordinates. Covariant quantization
solves this problem by making Lorentz invariance manifest. As mentioned before, the covariant
approach is really the “proper” approach to quantizing the string, however the physical results aren’t
as readily observed. Now that we have seen the physical consequences of quantizing the bosonic
string in the light-cone gauge, we are ready to tackle the more elegant approach in quantizing the
string.
Let’s begin by recalling some facts about the parameterization of the world-sheet. Earlier we
described a class of gauges, including the static and light-cone gauge, characterized by nµ . We were
able to show that the string coordinates satisfy the constraints
0
2
Ẋ ± X =0 (10.1)
From these constraints we also showed that the wave equations take the form
Ẍ µ − X 00µ = 0 (10.2)
Implying the momentum densities
1
P σµ = − X 0µ (10.3)
2πα0
1
Pτµ = Ẋ µ (10.4)
2πα0
In the covariant formalism we still use the constraints but we do not completely fix the parame-
terization of the world-sheet, as we had done in the light-cone gauge. To get to the wave equations
203
204 CHAPTER 10. LORENTZ COVARIANT QUANTIZATION
(10.2) we had to do work with the Nambu-Goto action. However this action is a pain to deal with
when we quantize the string using the covariant formalism, mostly because of the square root. For
that reason, we are motivated to seek an action better for quantizing the string. Using the work
we did in the last chapter, we consider an action which is very similar to the action used for the
transverse string coordinates X I :
Z Z
1
S = dτ dσL = dτ dσ (∂τ X µ ∂τ Xµ − ∂σ X µ ∂σ Xµ ) (10.5)
4πα0
This action will end up giving us a more physical derivation of the commutation relations for
the oscillators. Moreover, notice that, unlike the Nambu-Goto action, this action does not have a
square root, which makes for an easy time quantizing the string. We know this action works becasue
a variation in X µ immediately produces the wave equation. Moreover, the conjugate momenta are
easily computed, matching those in (10.3) and (10.4).
The Hamiltonian is also easier to deal with
Z Z Z
µ µ 1 0 2 0
2
H = dσH = dσ(Pµ Ẋ − L) = dσ Pµ Ẋ − (2πα ) − X
4πα0
X0 · X0
Z
= πα0 dσ P · P + (10.6)
(2πα0 )2
which is analogous to the light-cone Hamiltonian derived in the last chapter. Lastly, in the co-
variant formalism, it is rather natural to introduce the Heisenberg operators, X µ (τ, σ) and P µ (τ, σ)
and postulate the commutation relations
[X µ (τ, σ), P ν (τ, σ)] = iη µν δ(σ − σ 0 ) (10.7)

µν
where now η is the usual Minkowski metric.
10.2 Virasoro Operators and Quantum Constraints

The covariant quantization procedure of X µ is quite similar to the quantization of the transverse
light-cone coordinates X I . Let’s start by considering the open string. Recall the oscillator expansion
√ √ X1
X µ (τ, σ) = xµ0 + 2α0 α0µ + i 2α0 αµ e−inτ cos(nσ) (10.8)
n n
n6=0
which led us to
√ X
Ẋ µ ± X 0µ = 2α0 αnµ e−in(τ ±σ) (10.9)
n∈Z
The commutation relations for the oscillators are computed in an analogous manner as in the
last chapter. We avoid the explicit computations here and just quote the results [64]:
µ
[αm , αnν ]m = η µν δm+n,0 (10.10)
[aµm , aν∗
n ] = δm,n η
µν
(10.11)
10.2. VIRASORO OPERATORS AND QUANTUM CONSTRAINTS 205
√
Moreover, with α0µ = 2α0 pµ this time the zero mode operators satisfy
[xµ0 , pν ] = iη µν (10.12)
0
It should be noted that even now the time coordinate x becomes an operator, suggesting that
the states have built in time dependence, thereby implying that the role of the Schrödinger equation
changes. The constraints (10.1) now take the form
0
2 X
Ẋ ± X = 4α0 Ln e−in(τ ±σ) (10.13)
n∈Z
where we have the covariant open string Virasoro operator:

1X µ
Ln = αn−p αp,µ (10.14)
2
p∈Z
in which we notice that on the classical level, in order for our constraints (10.1) be satisfied
we require Ln = 0, ∀n ∈ Z. The covariant Virasoro operators are different from the transverse
Virasoro operators L⊥ n because the covariant formalism include contributions from all of the string
coordinates. Just as when we quantized the string in the light-cone gauge, the only Virasoro operator
with ambiguous ordering is the zeroeth mode L0 . Again, we will define L0 to be a normal-ordered
operator without any additional ordering constant.
In the covariant treatment, we obtain similar commutation relations between the Virasoro oper-
ators as we did in the light-cone gauge:
D 3
[Lm , Ln ] = (m − n)Lm+n + (m − m)δm+n,0 (10.15)
12
As from before, the Virasoro operators generate world-sheet parameterizations that preserve the
gauge conditions of the Lorentz-covariant formalism.
In the covariant quantization procedure, unlike light-cone quantization, the quantum constraints
are not solved for but instead imposed on the states of the theory. Now that we have the Virasoro
operators for the open string, let’s move on and select the quantum constraints of our theory. Based
from classical theory, the physical states of the quantum theory should be annihilated by all of the
Virasoro operators. It turns out this is too stringent as this would actually eliminate all states from
the theory. Therefore some constraints must be imposed. If we did not, then we would end up with
a quantization procedure that results in a different physical theory, a consequence we certainly do
not want. The method of quantization should not give a different physical theory, rather, different
insight into the same physical theory.
Let’s start by exploring what the constraint on the Virasoro operator L0 would impose on the
states of our theory. As states above, we have defined L0 as normal ordered with no additional
constant. But there is no real reason why L0 should annihilate the physical states (Zwiebach, 572).
Rather, we would expect the operator (L0 − a), where a is some constant, should annihilate the
physical states of our theory. We therefore impose the constraint
(L0 − a)|φi = 0 (10.16)

for any physical state |φi. Explicitly the operator (L0 − a) in terms of the oscillator expansion
is just
∞ ∞
1 µ X µ
X
L0 − a = α0 α0,µ + α−p − a = α0 p2 − a + naµ∗
n an,µ = 0 (10.17)
2 p=1 n=1
Moreover, since M 2 = −p2 the mass-squared operator is easily determined to be

1
M2 = (N − a) (10.18)
α0
with
∞
X
N= naµ∗
n an,µ (10.19)
n=1
where N is the covariant version of the number operator. Just as the was true for the light-cone
gauge number operator N ⊥ , a short calculation making use of [an,0 , a∗n,0 ] = −1 and the commutator
[−na∗n,0 an,0 , a∗n,0 ] = na∗n,0

reveals that indeed N is non-negative [64]. We do run into a problem however using covariant
quantization: many of the states do not have positive norm. To see this, consider an eigenstate |φi
of N with an eigenvalue N0 and a positive norm hφ|φi > 0. Then, by definition of the creation and
annihilation operators, we have
an,0 |φi = 0
for n > N0 . Now consider the state |χi = a∗n,0 |φi, along with the Hermitian conjugate hχ| =
hφ|an,0 . The norm of the state is then
hχ|χi = hφ|an,0 a∗n,0 |φi = hφ|[an,0 , a∗n,0 ]|φi = −hφ|φi < 0

Since |χi has negative norm, our probability interpretation goes out the window. In our brief
survey on quantum field theory, we encountered similar ghost states. To interpret any good quantum
theory we must try an rid our theory of these ghost states. To do this in covariant quantization,
we establish more constraints on the states of our theory. We cannot have (L0 − a) along with
all of the other Virasoro operators annihilate the physical states. Let’s show this by an example.
Consider three operators (L0 − a), L1 and L−1 , let |φi be a general state and impose the condition
(L0 − a)|φi = 0 (10.20)
L1 |φi = 0 (10.21)
L−1 |φi = 0 (10.22)

Assuming (10.21) and (10.22) hold, then it follows that the commutator of L1 with L−1 must
annihilate the general state as well:
[L1 , L−1 ]|φi = (L1 L−1 )|φi − (L−1 L1 )|φi = 0

But
D
[L1 , L−1 ] = (1 − (−1))L0 + ((−1)3 − (−1)) = 2L0
12
Therefore we’d also have
(2(L0 − a) + 2a)|φi = 2a|φi = 0

However, since we are assuming that a is some non-zero constant, it would have to be that
|φi ≡ 0. Thus, we cannot include that all Virasoro operators, Lm ∀m 6= 0 annihilate the physical
states as no states would survive and it is incompatible with the Virasoro algebra.
We have already seen that the constraint (L0 − a) = 0 is necessary to fix the mass spectrum. The
next simplest constraint to try is to set to zero either all of the Virasoro operators with a positive
mode number or those with a negative mode number. When working in the light-cone gauge we
saw that the positively moded oscillators were the onces that functioned as annihilation operators.
Therefore it is naturl to associate annihilation of the physical states with the positively moded
Virasoro operators. We declare then that all positively moded Virasoro operators in the covariant
formalism annihilate physical states. Such states states are called primary states. Being a primary
state is necessary but not sufficient to guarantee that a quantum state is truly physical. We require
two additional conditions. The first is that the state must be annihilated by the operator (L0 − a).
States that are primary and satisfy this first condition are called admissible states. In summary, a
quantum state |φi is said to be admissible if and only if
(Ln − aδn,0 )|φi = 0 (10.23)

for all n ≥ 0.
The second condition required for physical states has to do with a class of states called Virasoro
descendents. If a state is admissible and satisfies this condition it is said to be physical. WARNING:
This terminology is not standard. Here admissible states are sometimes called physical states while
our physical states are called real physical states.
A Virasoro descendent of a given primary is a state that can be written as a finite linear combina-
tion of products of negatively moded Virasoro operators acting on the primary state. For instance,
if |pi denotes a primary state, the state L−1 |pi is a descendent. Moreover, a state that is both pri-
mary and a descendent is called a null state. We call it a null state because it has an inner product
of zero with itself , any primary state, and any descendent state. If we alter a primary state by the
addition of a null state, the new primary state has the same inner products with primary states as
the original one. A null state will behave as a pure gauge if the physical operators in the quantum
theory map null states to null states, in which case the addition of the null states to the primary
states cannot affect any of the physical expectation values. Based on this, we may now introduce
a definition of a physical state: A non-vanishing state is said to represent a physical state if it is
admissible and it is not a descendent [64]. To prove that this definition is in fact a good one, we
will use it in the next section.
Before moving on, are we certain we relieved our theory of negative norm states? To require a
spectrum free of negative norm states, we will impose further constraints on our theory, namely
the value of a and the space-time dimension D. At the boundary of where positive norm states
turn into negative norm states an increased number of zero norm states appear [5]. It is therefore
an effective strategy to search for zero norm states that satisfy the physical state conditions to
determine a and D.
A spurious state |ψi is a state that satisfies
(L0 − a)|ψi = 0 (10.24)

and is orthogonal to all physical states |φi
hφ|ψi (10.25)
An example of a spurious state is
∞
X
|ψi = L−n |χn i (10.26)
n=1
with
(L0 − a + n)|χn i = 0 (10.27)

Based on this definition we see that a spurious state and a Virasoro descendent are related.
Moreover, from our orthogonality relation we find that [21]:
∞ ∞
∗
X X
hφ|ψi = hφ|L−n |χn i = (hχn |Ln |φi) (10.28)
n=1 n=1
If a state |ψi is spurious and physical it is orthogonal to all physical states including itself
∞
X
hψ|ψi = hχn |Ln |χn i = 0 (10.29)
n=1
As a result, such a state has zero norm.
We are now in a position to to determine the ordering constant a. When a is suitably chosen, a
class of zero norm spurious states has the form |ψi = L−1 |χ1 i with
(L0 − a + 1)|χ1 i = 0 (10.30)

and
Ln |χ1 i = 0 (10.31)
for n > 0. If we then demand that |ψi is physical implies that
Ln |ψi = (L0 − a)|ψi = 0 (10.32)

for n = 1, 2.... Moreover, the Virasoro algebra implies the identity
L1 L−1 = 2L0 + L−1 L1 (10.33)

Leading us to
L1 |ψi = L1 L−1 |χ1 i = (2L0 + L−1 L1 )|χ1 i = 2(a − 1)|χ1 i = 0

Which implies that we require
a=1 (10.34)
Now that we know a we can determine the space-time dimension D. To do this, let us construct
zero-norm spurious states of the form
|ψi = (L−2 + γL2−1 )|χ̃i (10.35)

which has zero-norm for a particular value of γ. |ψi is a spurious state if the state |χ̃i satisfies
(L0 + 1)|χ̃i = Ln |χ̃i = 0 (10.36)

for n > 0. We now impose that |ψi is a physical state. Thus L1 |ψi = 0 and L2 |ψi = 0. We can
further examine the condition that L1 |ψi = 0 using
[L1 , L−2 + γL2−1 ] = 3L−1 + 2γL0 L−1 + 2γL−1 L0 = (3 − 2γ)L−1 + 4γL0 L−1 (10.37)
yielding
L1 |ψi = L1 (L−2 + γL2−1 )|χ̃i = ((3 − 2γ)L−1 + 4γL0 L−1 ) |χ̃i (10.38)
3
The first term will vanish for γ = 2 and the second term vanishes in general since
L0 L−1 |χ̃i = L−1 (L0 + 1)|χ̃i = 0

Let’s next consider the condition L2 |ψi = 0, this time using
3 D
[L2 , L−2 + L2−1 ] = 13L0 + 9L−1 L1 + (10.39)
2 2
which yields
3 D
L2 |ψi = L2 (L−2 + L2−1 )|χ̃i = (−13 + )|χ̃i = 0 (10.40)
2 2
For this to hold true, we require that the space-time dimension be 26. Hence, just as found in
the light-cone gauge, we require that for quantized bosonic string theory we are working with a
D = 26-dimensional background space-time.
The non-zero spurious states are unphysical. Since they are spurious ensures that they decouple
from all physical processes, and in fact have all the negative norm states decouple, leaving all the
physical states with positive norm. Therefore the complete physical spectrum is free of negative
norm states where a = 1 and D = 26 [5].
Looking back, we now have the quantum constraint
(L0 − 1)|φi = 0 (10.41)

and for n > 0
Ln |φi = 0 (10.42)
which means that an admissible state satisfies
(Ln − δn,0 )|φi = 0 (10.43)

for n ≥ 0. Finally, the mass squared operator becomes
1
(N − 1) M2 = (10.44)
α0
where N is the same number operator as before.
Before we move on to constructing the covariant state space, let’s first quickly write down the
Virasoro operators for the closed string. Let’s assume no compactification. The mode expansion of
the closed string is then
√
r
µ µ α0 X e−inτ
µ

0
X (τ, σ) = x0 + 2α α0 τ + i (10.45)
2 n
n6=0
q
α0
where α0µ = pµ 2 . Moreover, we also have that
 
0µ 2
X 1X X
Ẋ µ + X = 4α0 α̃pµ α̃n−p,µ  e−in(τ +σ) ≡ 4α0 L̃n e−in(τ +σ)

 (10.46)
2
n∈Z p∈Z n∈Z
 
0µ 2
X 1 X X
Ẋ µ − X = 4α0 αpµ αn−p,µ  e−in(τ +σ) ≡ 4α0 Ln e−in(τ +σ)

 (10.47)
2
n∈Z p∈Z n∈Z
where we have defined the closed string Virasoro operators
1X µ
L̃n = α̃p α̃n−p,µ (10.48)
2
p∈Z
1X µ
Ln = αp αn−p,µ (10.49)
2
p∈Z
In closed string theory we have two sets of Virasoro operators. Then, in analogy with the open
string quantum constraints, we demand that for the physical closed states |ψi we have
(Ln − δn,0 )|ψi = 0 (10.50)
(L̃n − δn,0 )|ψi = 0 (10.51)

for n ≥ 0. For n = 0 notice
α0 2

(L0 − 1)|ψi = p + N − 1 |ψi = 0 (10.52)
4
10.3. CONSTRUCTING THE LORENTZ COVARIANT STATE SPACE 211
α0 2

(L̃0 − 1)|ψi = p + Ñ − 1 |ψi = 0 (10.53)
4
It follows that the operator L0 − L̃0 annihilates all physical states. Moreover, we find two
expressions for the mass-squared operator
4 4
M 2 = −p2 (N − 1) = 0 (Ñ − 1) (10.54)
α0 α
yielding N = Ñ , a result we could have guessed. We may also write the mass-squared operator
as
2
M2 = (N + Ñ − 2) (10.55)
α0
Now that we have all that we need, let’s move on to construct the state space of the covariantly
quantized string.
10.3 Constructing the Lorentz Covariant State Space

In the covariant formalism, all the components pµ of the momentum are independent commuting
operators. We therefore label the ground states using the full momentum vector |pi = |p0 , p1 , ...p25 i.
Similarly, the conjugate variables X µ are labeled as |Xi = |X 0 , X 1 , ...X 25 i. Thus, since the con-
jugate variables include the time component, given a state |ψi, the associated wavefunction hX|ψi
will have time dependence before the Schrödinger equation is introduced. To understand the con-
sequences of this, let’s first examine the covariant Hamiltonian. We may write it as
Z π Z π
1 02
1 1
2
dσ (Ẋ + X 0 )2 + (Ẋ − X 0 )2

H= dσ Ẋ + X =
4πα0 0 4πα0 0 2
1 X π
Z
= dσLn (e−inσ + einσ )e−inτ (10.56)
2π 0
n∈Z
We note that all integrals for n 6= 0 vanish and for n = 0, the right hand side yields L0 . Therefore
we have that H ∝ L0 . Recall that the light-cone open string Hamiltonian was H = L0 −1, suggesting
that the covariant Hamiltonian must be chosen as
H = L0 − 1 = α0 p2 + N − 1 (10.57)
But now remember one of the quantum constraints that we selected. Since the operator L0 −
1 annihilated all physical states, we conclude that the Hamiltonian also annihilates all physical
states. This implies then that, as the Hamiltonian is defined, it is not possible to introduce a time
variable and generate non-trivial time evolution through the Schrödinger equation. The Schrödinger
equation has now turned into the constraint
H|φi = 0 (10.58)
where |φi denotes a physical state. Basis vectors of covariant state space are constructed by
acting on the ground states |pi with all the possible creation operators:
∞ Y
Y 25
|bi = (aµ∗
n )
λn,µ
|pi (10.59)
n=1 µ=0
Now that we have constructed a general basis state, let’s examine some of the physical states
of the theory. First consider the ground states of |pi. Using our quantum constraints, all of the
ground states are automatically annihilated by Ln for n ≥ 1. The only non-trivial constraint is
1
0 = (L0 − 1)|pi = (α0 p2 − 1)|pi → p2 = (10.60)
α0
But where M 2 = −p2 , we see that the ground states again yield the tachyon states of the theory.
With a slight generalization we can deal with tachyon fields, in which the constraint then gives us
the classical field equation for a tachyon. A general tachyon state |T i can be constructed using a
superposition of the momentum ground states
Z
|T i ≡ dD pφ(p)|pi (10.61)
Though physicists tend not to like tachyons, we would still prefer the tachyon state to be physical.
For this to be true we require
Z Z
(L0 − 1)|T i = dD pφ(p)(L0 − 1)|pi = dD pφ(p)(α0 p2 − 1)|pi = 0 (10.62)
For this equation to hold for any of the momentum ground states, we require that the integrand
must vanish
(α0 p2 − 1)φ(p) = 0 (10.63)

This is just the field equation for a tachyon, in which the function φ(p) turns out to be the
tachyonic field.
Now that we examined the tachyon states, let’s consider the photon states. Analogous to the
quantized string in the light-cone gauge, we consider the first set of excited states with N = 1 along
with fixed momentum p and a polarization ξµ . These take the form
µ
ξµ α−1 |pi (10.64)
Using our constraint, L0 − 1 = 0, we find
µ µ
0 = (α0 p2 + N − 1)ξµ α−1 |pi = α0 p2 ξµ α−1 |pi (10.65)
which leads to
p2 = 0 (10.66)
Moreover, with L1 = α0 · α1 + (α−1 · α2 + α−2 · α3 + ...), we have that
µ µ
√
L1 ξµ α−1 |pi = α0 · α1 ξµ α−1 |pi = 2α0 pµ ξµ |pi = 0 (10.67)
yielding
10.3. CONSTRUCTING THE LORENTZ COVARIANT STATE SPACE 213
p·ξ =0 (10.68)
Moreover, it can be shown that for all Virasoro operators Ln n ≥ 2 vanish automatically [64].
Therefore, physical photon states exist only for p2 = 0 and p · ξ = 0. For any pµ which satisfies
p2 = 0 we can choose a Lorentz frame where pµ = (p0 , p0 , 0, 0...) and the constant p · ξ = 0 gives
ξ0 + ξ1 = 0 (10.69)
This leaves us with D-1 independent polarizations, but when we quantized the string in light-
cone coordinates we had D-2 independent polarizations. To have the covariant analysis match that
of the light-cone analysis, we end up imposing the condition that physical states are defined up to
null states. By imposing this condition, we mean that there is an equivalence relation between a
µ
state and the same state modified by a null state. For instance, suppose we have the state ξµ α−1 |pi.
2
Moreover, suppose that we consider a frame where p = 0. By (10.68), we would then have the
equivalence relation
µ µ
ξµ α−1 |pi ∼ (ξµ + ipµ )α−1 |pi
Moreover, we also find an equivalence in polarizations: ξ µ ∼ ξ µ + ipµ for p2 = 0. Ultimately,

this allows us to choose a polarization such that [64]
ξ0 − ξ1 = 0 (10.70)
With both (10.69) and (10.70), we can conclude that the physical states have D-2 independent
polarizations, just as we had found in the light-cone analysis. Analogous to the tachyon state, one
can also work with a general superposition of momentum ground states by defining the gauge field
state |Ai as
Z
µ
|Ai = dD pAµ (p)α−1 |pi (10.71)
The closed string state space is constructed in the same way as the open string. The general
basis vector is given by
∞ Y ∞ Y
25
! 25
!
Y Y
|b, b̃i = (aµ∗
n )
λn,µ
× (aν∗
m)
λm,ν
|pi (10.72)
n=1 µ=0 m=1 ν=0
To uncover the closed tachyon states we consider the ground states and apply the constraints
L0 − 1 = L̃0 − 1 = 0. To uncover the massless string states (i.e. Kalb-Ramond, graviton, and
dilaton states) we examine the condition also examine the conditions L1 = L̃−1 = 0. Again, as
in correspondence with the light-cone analysis, the quantized closed string naturally has graviton
states.
10.4 The Polyakov Action and “Modern” Covariant Quan-

tization
In this chapter we have been exploring the more traditional approach of quantizing the string
using the covariant formalism. We already noted that using the Nambu-Goto action is difficult in
quantizing the string, in which case we introduced the action (10.5). This string action is indeed
elegant, however, in the “modern” approach to covariant quantization one uses another action, the
Polyakov action, from which the constraint equations emerge naturally. To do this we rewrite the
action (10.5) in a more suggestive form
Z
1
S=− dτ dση αβ ∂α X µ ∂β X ν ηµν (10.73)
4πα0
where α and β run over two corresponding to the world-sheet coordinates τ and σ. We have also
introduced
∂
∂α = (10.74)
∂ξ α
with ξ α = (ξ 1 , ξ 2 ) = (τ, σ). Moreover, we introduced the two-dimensional Minkowski metric η αβ .
This two-dimensional metric should remind the reader of the one that emerged while we made the
manifest reparameterization invariance of the Nambu-Goto action, where we came up with
√
Z
1
S=− dτ dσ −γ (10.75)
2πα0
where γ = det(γαβ ) and γαβ is the world-sheet metric explicitly given by
∂X ∂X
γαβ = · = ∂α X · ∂β X (10.76)
∂ξ α ∂ξ β
Similarly, the Polyakov action involves a new worldsheet metric, hαβ (τ, σ), which also the type
of metric used in two-dimensional general relativity [64]. The metric hαβ is a dynamical variable in
the action, so it leads to its own field equations. This metric ends up entering the Polyakov action
in a way that is analogous to η αβ in (10.73). In all of its glory, the Polyakove action is
√
Z
1
S=− 0
dτ dσ −hhαβ ∂α X µ ∂β X ν ηµν (10.77)
4πα
Here we have h = det(hαβ ) and
hαβ hαγ = δγα hαβ hαβ = 2 (10.78)

As it turns out, variation of hαβ in the Polyakov action will give us the Virasoro constraints.
Moreover, since hαβ is a symmetric 2 × 2 matrix, it has three independent entries, and therefore we
would actually expect three constraints. But there are only two Virasoro conditions.
√ To reconcile
this, we first note that the metric hαβ enters the action through the combination −hhαβ . As we
will see, this combination is only determined by two numbers at any point on the world-sheet, not
three. This fact is actually a property of two-dimensional metrics. To witness this, we define the
matrix
10.4. THE POLYAKOV ACTION AND “MODERN” COVARIANT QUANTIZATION 215
√
M αβ = −hhαβ (10.79)
αβ
and let n be the size of the metric h . Notice then
n
√ (−h) 2
det(M αβ ) = ( −h)n det(hαβ ) =
det(hαβ )
n
(−h) 2 n
= = −(−h) 2 −1 (10.80)
h
Then, when n = 2, det(M αβ ) = −1. A 2 × 2 symmetric matrix with a determinant equal to
a fixed constant is only determined by two parameters, giving us only two independent constants
when we vary the action. Now that we have the constraint issue resolved, lets move to vary the
Polyakov action by first varying the string coordinate X µ , which yields
√
Z
1
δS = − 0
dτ dσ −hhαβ ∂α (δX µ )∂β X ν ηµν
2πα
1
Z √
µ αβ ν
dτ dσδX ∂α −hh ∂β X η µν (10.81)
2πα0
As always, to find the equations of motion we require that δS = 0, in which case we have the
resulting equations of motion
√
∂α −hhαβ ∂β X µ = 0 (10.82)
Let’s now vary the Polyakov action by varying the metric hαβ . Its easy to check that for a 2 × 2
matrix A we have δdetA = (detA)T r(A−1 δA) [64]. By denoting the variation of the metric as δhαβ
we find
δh = δdet(hαβ ) = h(hαβ δhαβ ) (10.83)

αβ
Making use of h hαβ = 2, we also have
δhαβ hαβ + hαβ δhαβ = 0 → hαβ δhαβ = −δhαβ hαβ (10.84)

Using this result we find that
δh = −hδhαβ hαβ (10.85)

√
But what we really need to vary is −h. Therefore we have
√ 1 δh 1 (−h)δhαβ hαβ 1√
δ( −h) = − √ =− √ =− −hδhαβ hαβ (10.86)
2 −h 2 −h 2
Varying the action with the variation in hαβ we find
√
Z
1 1 αβ γδ αβ
δS = dτ dσ −h − δh hαβ (h ∂γ X · ∂δ X) + δh ∂α X · ∂β X
4πα0 2
(10.87)
√
Z
1 αβ 1 γδ
=− dτ dσ −hδh ∂α X · ∂β X − hαβ (h ∂γ X · ∂δ X)
4πα0 2
The equation of motion for δhαβ is then

1
∂α X · ∂β X − hαβ (hγδ ∂γ X · ∂δ X) = 0 (10.88)
2
But ∂α X · ∂β X = γαβ , allowing us to write the equations of motion as
1
γαβ − hαβ (hγδ γγδ ) = 0 (10.89)
2
And we may therefore write
hαβ = f 2 (ξ)γαβ (10.90)

where f (ξ) is some function on the world-sheet. Metrics that are realted by a positive factor
of proportionality are said to be conformal to each other. Hence, the world-sheet metric hαβ
is conformal to the induced metric γαβ . Moreover, notice from our conformal relation, we have
√ √
−hhαβ = −γγ αβ , allowing us to rewrite the Polyakov action as
√
Z
1
S=− dτ dσ −γγ αβ γαβ (10.91)
4πα0
Again making use of hαβ hαβ = 2, the above becomes
√
Z
1
S=− dτ dσ −γ (10.92)
2πα0
which is exactly the Nambu-Goto action! Therefore, we can conclude that the Polyakov action
is equivalent to the Nambu-Goto action in the classical limit.
It’s important to note that while we have shown the Polyakov is invariant under the rescaling of
f 2 (ξ), we have also shown that the Polyakov action is invariant under Weyl transformations:
hαβ (τ, σ) → Ω2 (τ, σ)hαβ (τ, σ) (10.93)

where Ω2 is some arbitrary function on the world-sheet. This type of transformation crucial in
the study of general relativity and we will come across it again. In this case, the invariance of the
Polyakov action under a Weyl transformation indicates that on the world-sheet, distances measured
with hαβ have no physical significance [64].
Now that we have varied the action and found the equations of motion, let’s determine the Vira-
soro constraints. First of all, a well known result from two-dimensional geometry that a coordinate
reparameterization allows the metric hαβ on a surface to be written locally in the form
hαβ = ρ2 (ξ)ηαβ (10.94)

where ρ is a conformal factor. The statement above says that the metric is conformally flat. By
restricting to a class of world-sheet coordinates resulting in a conformally flat metric
√we are making
the partial gauge choice called the conformal gauge. In the conformal gauge we have −hhαβ = η αβ ,
and therefore
∂α (η αβ ∂β X µ ) = η αβ ∂α ∂β X µ = 0 (10.95)
Expanding yields
1
∂α X · ∂β X − ηαβ (η γδ ∂γ X · ∂δ X) = 0 (10.96)
2
But this is just
1
∂α X · ∂β X − ηαβ (−Ẋ 2 + X 02 ) = 0 (10.97)
2
If we then take α = β = 1 we find
1
Ẋ 2 + (−Ẋ 2 + X 02 ) = 0 (10.98)
2
leading us to one of the constraints
02
Ẋ 2 + X =0 (10.99)
Morover, notice that when α = β = 2 we acquire the same constraint. Finally, with α = 1 and
β = 2 we get
Ẋ · X 0 = 0 (10.100)
Altogether, the equations of motion for the world-sheet metric reduce to the familiar Virasoro
constraints in the conformal gauge. Moving on, let’s consider some more interesting properties of
the Polyakov action. First we rewrite the action using the line element
ds2 = hαβ dσ α dσ β (10.101)

1
where σ 0 = τ and σ 1 = σ. Then, using our expression for the string tension T = 2πα0 , we may
rewrite the Polyakov action as
√
Z
T
S=− d2 σ −hhαβ ∂α X µ ∂β X ν ηµν (10.102)
2
Let’s define the energy-momentum tensor Tαβ to be [37]
2 1 δS
Tαβ ≡ − √ (10.103)
T −h δhαβ
Equation (10.103) has a strange looking derivative in it. This type of derivative is known as a
functional derivative, and is, for our purposes, a streamlined way of calculating the variation of
the action S. In fact we may recgonize this is just another way of calculating the variation of the
Polyakov action by varying the worldsheet metric hαβ . Thus
2 1 δS 1
Tαβ = − √ = ∂α X · ∂β X − ηαβ (η γδ ∂γ X · ∂δ X) = 0 (10.104)
T −h δhαβ 2
Therefore the energy-momentum tensor of the world-sheet is zero. Moreover if we consider a flat
world-sheet, hαβ = ηαβ , then, restoring the Minkowski metric to make to lower the indices in the
dot products, we find that
1
Tαβ = ∂α X µ ∂β Xµ − ηαβ (η γδ ∂γ X µ ∂δ Xµ ) (10.105)
2
Since the energy-momentum tensor is zero, we also have T r(Tαβ ) = 0. Therefore,
T r(Tαβ ) = Tτ τ + Tσσ = 0 (10.106)

To be explicit, notice that τ τ component of the energy-momentum tensor is just
1
Tτ τ = ∂τ X µ ∂τ Xµ − ητ τ (η τ τ ∂τ X µ ∂τ Xµ + η σσ ∂σ X µ ∂σ Xµ )
2
1 1 0
= (∂τ X µ ∂τ Xµ + ∂σ X µ ∂σ Xµ ) = (Ẋ µ Ẋµ + X µ Xµ0 ) (10.107)
2 2
Similarly,
1 µ 0
Tσσ =(Ẋ Ẋµ + X µ Xµ0 ) (10.108)
2
Then, statement T r(Tαβ ) = 0 is nothing more than the Virasoro constraint
02
Ẋ 2 + X =0 (10.109)
In other words, the Virasoro constraints we have been working with are nothing more than a
statement about the energy-momentum of the world-sheet being zero.
We’ve already already seen that the Polyakov action is invariant under Weyl transformations
and reparameterizations of the world-sheet. There is one more symmetry of the Polyakov action
which results in the Lorentz charges. This symmetry is related to Poincaré transformations. Before
we get to finding the Lorentz charges, first recall that the Poincaré group consists of translations
in space-time and Lorentz transformations [37] (with that definition we can see that we have been
working with Poincaré transformations this entire time). First consider a space-time translation
X µ → X µ + bµ (10.110)
with δX µ = bµ . Moreover, an infinitesimal Lorentz transformation takes the form
X µ → X µ + ωνµ X ν (10.111)
with δX µ = ωνµ X ν and where ωνµ is antisymmetric. We can combine both the space-time
translations and infinitesimal Lorentz transformations as
δX µ = ωνµ X ν + bµ (10.112)
Let’s see what we get when we consider a space-time translation of the Polyakov action. First
note that the Lagrangian density of the Polyakov action is just
T√
L=− −hhαβ ∂α X µ ∂β X ν ηµν (10.113)
2
Then considering the space-time translation given in (10.110) our Lagrangian density changes to
T√
L→− −hhαβ ∂α (X µ + bµ )∂β (X ν + bν )ηµν
2
T√
=− −hhαβ (∂α X µ ∂β X ν + ∂α X µ ∂β bν )ηµν
2
where we dropped terms quadratic in bµ as it is assumed to be small. Expanding we find that

the Lagrangian density changes to
T√ T√
− −hhαβ ∂α X µ ∂β X ν ηµν − −hhαβ (∂α X µ ∂β bν + ∂α bµ ∂β X ν )ηµν = L + δL (10.114)
2 2
With a little bit of index gymnastics we find that the variation of the Lagrangian density is
equivalent to
√
δL = −T −hhαβ (∂α Xµ )∂β bµ (10.115)
By demanding that the variation of the Lagrangian density vanish, we find that the conserved
current is
√
Pµβ = −T −hhαβ (∂α Xµ ) = 0 (10.116)
Using reparameterization invariance and a Weyl rescaling to let hαβ → ηαβ , we find that the
conservation equation for current is
∂α Pµα = 0 (10.117)
Or,
∂τ Pµτ + ∂σ Pµσ = 0 (10.118)

matching the conservation of current equation we derived early on. A similar exercise reveals
the Polyakov Lagrangian density is invariant under Lorentz transformations and thus the Polyakov
action is invariant under Poincaré transformations. Finally, note that
∂L T√
= − −hhαβ ∂β X ν ηµν = Pµα (10.119)
∂(∂α X µ ) 2
Then, using the fact that Noether currents can be found in general using
∂L
µν Jαµν = δX µ (10.120)
∂(∂α X µ )
we find that under a Lorentz transformation δX µ = ωνµ X ν and using the antisymmetry of ωνµ
the Lorentz current is
Jαµν = T (X µ Pαν − X ν Pαµ ) (10.121)

just as we had found before.
Aside from its elegance, why do we care about the Polyakov action? The reason we care is
because this action is the starting point of the so-called modern covariant approach. We don’t have
all of the tools at this point to full appreciate this method of covariant quantization, however let
us very briefly discuss the basic notion.
Figure 10.1: (a) Path integrals in QFT is a sum of Feynman diagrams over loops; (b) Path integrals
in string theory (and GR) is a sum of surfaces including those with holes and handles.
In quantum field theory, path integrals offer an alternative quantization procedure. In the path
integral method, the propagation of a quantum point particle from (~x0 , t0 ) to (~x, t) is viewed as a
sum over all possible paths or world-lines of the particle starting at (~x0 , t0 ) and ending at (~x, t).
Moreover, if interactions are included, the possible paths to be summed involve loops and vertices.
To see this diagrammatically, consider figure 10.1.
Path integrals offer an altenative approach to quantizing the string as well. When the point par-
ticle is replaced by a string the paths connecting the intial string configuration with the final string
configuration are world-sheets swept out by the string. Thus, the path integral representation of
string propagation is a sum over all possible world-sheets, or a sum over random surfaces. The path
integral approach also allows one to describe string interactions. When these are included, similar
to path integral quantization of interacting fields, the surfaces summed over include those with
holes and handles. A perturbative evaluation of the path integral for a closed string is illustrated
diagrammatically in figure 10.1.
The expansion is in terms of the genus of the surface (the number of handles or holes, e.g. a
torus has a genus of 1). As it turns out, each surface in the sum possesses a complex structure. A
connected surface with a complex structure is called a Riemann surface. Thus the path integral
viewpoint of string propagation is a sum over random Riemann surfaces [25]. As one can imagine,
this involves a fair amount of mathematics which we will not cover. For that reason we will not go
through the details of the path integral approach, however the reader is pointed to the references
for full details on this method of quantization.
The path integral approach makes heavy use of the Polyakov string. Using the Polyakov action
√
Z
T
S0 = − d2 σ −hhαβ ∂α X µ ∂β X ν ηµν (10.122)
2
10.5. EXERCISES 221
one attempts to quantize the string using something called the Euclidean path integral formula-
tion, which is nothing more than assuming that you’re working with a Euclidean metric (we discuss
how one does this more in the next chapter)[21].
From here one applies the so-called Faddeev-Popov procedure, allowing one to develop the Vi-
rasoro algebra for the string. In this approach however, the algebra contains an anamolous term
related to the presence of ghost states. Just as before, these ghost states and the anamoly are
eliminated as long as the space-time dimension D=26, the critical space-time dimension of bosonic
string theory. In the next chapter we will work out some of these mathematical details, and ex-
plicitly solve this anamoly. Again, the path integral approach goes beyond the scope of this text,
and we will not cover it in this book. For the interested reader, Polchinski’s two volume set String
Theory, makes heavy use of the path integral approach, though it is quite rigorous.
In summary, we have used the Lorentz covariant approach, the more traditional approach to
quantizing the string, to help reveal the manifest Lorentz invariance of the theory. We saw that
when we did this we had to introduce the quantum constraints by hand to maintain consistency.
In doing so we also saw that, just as was proved in the light-cone analysis, we require that the
dimension of space-time be 26. Finally, after examining the Polyakov action and its symmetries in
detail, we briefly outlined the method of modern covariant quantization which employs the path
integral procedure using the Polyakov path integral. Though we didn’t prove it here, we noted that
the path integral approach, in order to have a theory with no ghosts, one must be working in a
26-dimensional space-time. We also saw that gravity naturally emerges when we consider the path
integral approach using an arbitrary space-time metric gµν .
10.5 Exercises
1. Show that [an,0 , a∗n,0 ] = −1, and [−na∗n,0 an,0 , a∗n,0 ] = na∗n,0 . Using both of these relations, show
that the number operator N is indeed non-negative.
2. Verify (10.58).
3. Using an arbitrary 2 × 2 matrix A, prove that δdetA = (detA)T r(A−1 δA).
4. Prove (10.115) and (10.116). Then, using the antisymmetry of ωνµ , show that the Lorentz
current associated with the Polyakov action takes the form as given in (10.121).
Part II
Advanced Topics and Superstrings
223
Chapter 11
Conformal Field Theory and

BRST Quantization
In this chapter we briefly cover the final quantization procedure, BRST quantization, for quantizing
the bosonic string. BRST quantization makes heavy use of complex variables, a subject the reader
is assumed to know a little of, as it is a two-dimensional conformal field theory. We will cover
the basics of conformal field theory before moving on to BRST symmetry. Due to the difficulty of
this procedure, even the basics are rigorous, and therefore this chapter may be skipped on a first
reading. But before we discuss the fundamentals of conformal field theory, let’s first consider a
more rigorous development of Lie groups and Lie algebra.
11.1 Introduction to Group Theory

Here we will briefly examine the basics of a Lie group and Lie algebra as it is particularly crucial
in the study of BRST quantization Let’s begin by considering the definition of a discrete group. A
group G is a set of elements {a, b, c...} with a binary operation × such that if a ∈ G and b ∈ G then
the product a × b is also an element of the group G, i.e.
a×b∈G (11.1)
This first property is known as closure. A group G must also satisfy the following three properties:
(i) Associativity: For elements a, b, c ∈ G the associative law with the binary operation × must
be satisfied:
a × (b × c) = (a × b) × c (11.2)
(ii) Identity Element: There exists a unique element e ∈ G such that for every element a ∈ G
e×a=a×e=a (11.3)
225
226 CHAPTER 11. CONFORMAL FIELD THEORY AND BRST QUANTIZATION
(iii) Inverse: For every element a ∈ G there exists an element a−1 ∈ G such that
a × a−1 = a−1 × a = e (11.4)

If these three properties and the closure property are satisfied then G with the binary operation ×
constitutes a group [39]. Usually the first property to be checked when testing whether a collection
of elements with a particularly binary operation is a group is closure. Moreover, when we say
binary operation we mean nearly any mathematical operation. For instance, the addition and
multiplication we are all familiar with are examples of binary operations. Moreover, a group G
where every element satisfies a × b = b × a is said to be an abelian group under the given binary
operation. Alternatively, a group with elements that are not commutative in the binary operation
is non-abelian. we won’t go into the details here, but it’s easy to check that the integers Z form a
group under addition. It also suffices to show that Z is in fact an abelian group under addition.
Now that we have a working definition of a group, let’s consider Lie groups. We’ve already seen
that Lie algebra plays a vital role in string theory: the Virasoro algebra is an example of a Lie
algebra. In general, Lie groups are imperative in the study of modern physics, particularly particle
physics. For all intensive purposes, a Lie group is a group whose elements g depend in a continuous
and differentiable way on a set of real parameters θa , with a = 1...N . For the reader familiar with
differential geometry (if not, you will be soon!), it’s through this definition that one can show that
a Lie group is at the same time a group and a differentiable manifold [33]. We can write a generic
element as g(θ).
A linear representation R of a group is an operation that assigns to a generic element g of a
group a linear operator DR (g) defined on a linear space
g 7→ DR (g) (11.5)
with the properties:
(i) DR (e) = 1, where 1 is the identity operator.
(ii) DR (g1 )DR (g2 ) = DR (g1 g2 ), such that the mapping preserves the group structure (sometimes
called a homomorphism).
The space on which the operators DR act on is called the basis for the representation R. A
typical example of a representation is a matrix representation, in which case the basis is a vector
space of finite dimension n and an abstract group element g is represented by an n × n matrix
(DR (g))ij with i, j = 1...n. The dimension n of the representation is defined as the dimension n
of the base space. Moreover, if we write a generic element of the base space as (φ1 , ...φn ), a group
element g induces a transformation of the vector space:
φi 7→ (DR (g))ij φi (11.6)

A representation R is said to be reducible if the action of any DR (g) on the vectors in the subspace
gives another vector in the subspace. Such a representation is said to have an invariant subspace
[33]. Alternatively, a representation without an invariant subspace is said to be irreducible. When
we change the representation of a group, in general the explicit form of the matrices DR (g) will
change. But there is an important property of a Lie group that is independent of the representation:
its Lie algebra.
11.1. INTRODUCTION TO GROUP THEORY 227
By the assumption of smoothness for θa infinitesimal we have that
DR (θ) ≈ 1 + iθa TRa (11.7)

where

∂DR
TRa ≡ −i (11.8)
∂θa θ=0
The TRa are called the generators of the group in the representation R. It can shown, with an
appropriate choice of parameterization away from the identity, the generic group elements g(θ) can
always be written in the form
a
DR (g(θ)) = eiθa TR (11.9)
If one is studying Lie groups from pure mathematics typically the factor of i in the definition
of the generators is not present. We include this factor so that the if in the representation R the
generators are Hermitian (which we like when we study Quantum Mechanics), then the matrices
DR (g) are unitary. In this case, R is a unitary representation.
Let’s now develop the associated Lie algebra. Consider two matrices DR (g1 ) = exp(iαa TRa ) and
DR (g2 ) = exp(iβa TRa ). Their product obeys the homomorphism property and therefore
a a a
DR (g1 g2 ) = eiαa TR eiβa TR = eiδa TR (11.10)
with δa = δa (αa , βa ). We make this distinction because in general for matrices A and B eA eB 6=
A+B
e . Thus in general δa 6= αa + βa . Solving for iδa TRa yields

1 1
iδa TRa = ln (1 + iαa TRa + (iαa TRa )2 )(1 + iβa TRa + (iβa TRa )2 )
2 2

a 1 a 2 1 a 2 a b
= ln 1 + i(αa + βa )TR − (αa TR ) − (βa TR ) − αa βb TR TR (11.11)
2 2
x2
Then, using the Taylor series expansion ln(1 + x) ≈ x − 2 , we find [33]
αa βb TRa , TRb = iγc (α, β)TRc

(11.12)
with
γc (α, β) = −2(δc (α, β) − αc − βc ) (11.13)

This must be true for all α and β, and therefore γc must be linear of the general form γc =
αa βb f ab c . Thus we find
a b
T , T = if ab c T c (11.14)
This is the Lie algebra of the Lie group being considered. One important point to be noted here
is that the structure constants f ab c are independent of representation.
But what about the definition we used earlier when examining the mathematical structure of
the Virasoro algebra? The two are equivalent, however the definition we used then came from the
general definition of an algebra. Formally, an algebra consists of a vector space A over a field K
together with a law of composition or product of vectors denoted
(A, B) 7→ AB ∈ A
for vectors A, B ∈ A which also satisfies a pair of distributive laws:
(i) A(aB + bC) = aAB + bAC
(ii) (aA + bB)C = aAC + bBC

for all scalars a, b ∈ K and vectors A, B, C ∈ A.
With this definition in hand, a Lie algebra L is a real or complex vector space with a law of
composition of bracket product [X, Y ] satisfying:
(i) Antisymmetry: [X, Y ] = −[Y, X]
(ii) Distributive Law: [X, aY + bZ] = a[X, Y ] + b[X, Z]
(iii) Jacobi Identity: [X, [Y, Z]] + [Y, [Z, X]] + [Z, [X, Y ]] = 0
This is an alternative though equivalent definition of a Lie algebra [62]. We used this definition
ealier to examine the Virasoro algebra, however we will use the other definition of a Lie algebra
when we study BRST quantization. Before moving on, let’s consider two of the most fundamental
groups of modern physics: the Lorentz group and the Poincaré group. The Lorentz group is defined
as the group of linear coordinate transformations
xµ → x0µ = Λµν xν (11.15)

which leave the invariant quantity
x · x = ηµν xµ xν = −t2 + x2 + y 2 + z 2 (11.16)

The group of transformations of a space with coordinates (y1 , ..., ym , x1 , ..., xn ) which leaves
(y12 + ... + ym
2
) − (x21 + ... + x2n )
invariant is called the orthogonal group O(n, m). Therefore, by analogy, the Lorentz group is
O(1, 3). Moreover, transformations where the determinant of Λ is unit, detΛ = +1, are called
proper transformations. The subgroup of O(1, 3) with detΛ = +1 is called the special orthogonal
group denoted SO(1, 3). A generic element A of the Lorentz group is written as
i µν
A = e− 2 ωµν J (11.17)
where ωµν is an antisymmetric matrix and J µν are the generators with a pair of antisymmetric
indices. By definition, one cane show that the Lie algebra of the special Lorentz group SO(1, 3) is
[33]
[J µν , J ρσ ] = i(η νρ J µσ − η µρ J νσ − η νσ J µρ + η µσ J νρ ) (11.18)
11.2. CONFORMAL TRANSFORMATIONS AND THE CONFORMAL GROUP 229
It is often convenient to rearrange the six components of J µν into two spatial vectors
1 ijk ik
Ji = J (11.19)
2
K i = J i0 (11.20)
Then the Lie algebra of the Lorentz group becomes
[J i , J j ] = iijk J k , [J i , K j ] = iijk K k , [K i , K j ] = −iijk J k (11.21)

The first commutator in (11.21) is actually the Lie algebra for the special unitary group SU (2) in
which the generators J i are the angular momenta we are familiar with [33]. In the last chapter we
saw that Lorentz transformations combined with space-time translations yield a Lie group called
the Poincaré group. Together with the Lorentz algebra, we can define the Poincaré algebra as:
[J i , J j ] = iijk J k , [J i , K j ] = iijk K k , [J i , P j ] = iijk P k (11.22)
[K i , K j ] = −iijk J k , [K i , P j ] = iHδ ij , [K i , H] = iP i (11.23)
[P i , P j ] = [J i , H] = [P i , H] = 0 (11.24)
0
where we have denoted P = H. As mentioned before Lie groups and Lie algebras are essential
tools in modern physics, including conformal field theory and BRST quantization.
11.2 Conformal Transformations and the Conformal Group

For our final quantization procedure, BRST quantization, we will exploit the conformal symmetry
of the world-sheet using the techniques of conformal field theory. In particular, since the world-
sheet can be described using two coordinates (τ, σ), we will use two dimensional conformal field
theory. As we mentioned on the outset, conformal field theory makes extensive use of the calculus
of complex variables making it a particularly mathematically elegant approach to quantizing the
string. Moreover, though we won’t cover it here, conformal field theory is particularly useful in
perturbative string theory.
Let’s begin by first developing a more concise definition for conformal transformation. Rather
simply, a conformal transformation is a transformation that maps a region of the complex plane to
another region while preserving angles but not lengths. That is, suppose there were two vectors in
the complex plane with an angular spread θ. As this region of the complex plane was conformally
mapped to another region, the lengths of the vectors might not be preserved however the angle
θ between the two vectors would remain invariant. The preservation of angles has a geometric
interpretation leading to the notion of conformal transformation of space-time coordinates. Consider
the coordinate transformation x → x0 . In general, the metric gµν (x) is found to transform as [37]
∂xα ∂xβ
0
gµν (x0 ) =
gαβ (x) (11.25)
∂x0µ ∂x0ν
Earlier we noted that a two metrics were conformally related if for a function of space-time
coordinates Ω(x) we have
0
gµν (x0 ) = Ω2 (x)gµν (x) (11.26)
Indeed, (11.26) denotes the conformal transformation of the metric. Essentially the conformal
factor Ω2 (x) is simply a scaling factor preserving angles but not lengths. To see how a conformal
transformation preserves angles, consider the dot product between two vectors ~u and ~v which may
written using the metric
~u · ~v g(~u, ~v )
cos(θ) = p =p (11.27)
(~u · ~u)(~v · ~v ) g(~u, ~u)g(~v , ~v )
Using the transformation given in (11.26) we find that
g 0 (~u, ~v ) Ω2 (x)g(~u, ~v ) g(~u, ~v )

cos(θ) = p =p =p = cos(θ) (11.28)
0 0
g (~u, ~u)g (~v , ~v ) 2 2
Ω (x)g(~u, ~u)Ω (x)g(~v , ~v ) g(~u, ~u)g(~v , ~v )
Hence, conformal transformations preserve the angles between vectors.

A conformal field theory is a quantum field theory that is invariant under conformal transfor-
mations. Such theories are sometimes called Euclidean Quantum Field Theories since, as we will
see in a moment, a Euclidean metric is used. We won’t go into the details here but conformal field
theories are useful when studying quantum field theory because it is scale invariant (has no length
scale, nor mass scale). Scale invariance is important because in regular quantum field theory scale
invariance of some actions are broken upon renormalization introducing a term which will give rise
to infinities.
As it turns out, two-dimensional conformal field theory is important in the study of world-
sheet dynamics. Remember the guitar analogy of a string: different vibrational modes of the string
correspond to different particle states. That is, different ways in which the string vibrates against the
background space-time determines what kind of particle the string is observed as. These vibrational
modes can be studied by exploring the dynamics of the world-sheet where the modes of the string
are described by conformal field theory. If the string is closed, it has two vibrational modes, each
of which can be described by a conformal field theory. Since these modes have direction conformal
field theories describing the closed string are sometimes called chiral conformal field theories [37].
In each of the quantization procedures up to this point, we have assumed that the string world-
sheet has a Lorentzian signature metric. If we make a Wick rotation τ → −iτ , the usual world-sheet
line-element
ds2 = −dτ 2 + dσ 2 (11.29)

becomes a line element associated with a Euclidean metric (positive definite)
ds2 = dτ 2 + dσ 2 (11.30)
An immediate consequence of this Wick rotation is that the Polyakov action under a Euclidean
metric is
Z
1
S= dτ dσ(∂τ X µ ∂τ Xµ + ∂σ X µ ∂σ Xµ ) (11.31)
4πα0
Moreover, the designated Wick rotation causes the light-cone coordinates (+, −) are replaced
with complex coordinates (z, z̄). To do this we define the complex coordinates as
z = τ + iσ, z̄ = τ − iσ (11.32)
Using these coordinates, the Polyakov action will simplify even further. First notice that we may
write the world-sheet coordinates τ and σ as
z + z̄ z − z̄
τ= , σ= (11.33)
2 2i
It follows then
∂τ ∂τ 1 ∂σ 1 ∂σ 1
= = , = , =− (11.34)
∂z ∂ z̄ 2 ∂z 2i ∂ z̄ 2i
Thus we see that

∂ ∂τ ∂ ∂σ ∂ 1 ∂ ∂
= + = −i
∂z ∂z ∂τ ∂z ∂σ 2 ∂τ ∂σ
Succinctly we have

1 ∂ ∂
∂z ≡ ∂ = −i (11.35)
2 ∂τ ∂σ

1 ∂ ∂
∂z̄ ≡ ∂¯ = +i (11.36)
2 ∂τ ∂σ
Using these coordinates and making use of (11.25) a brief exercise shows that the world-sheet
metric becomes [37]:
0 1

hαβ = 1 2 (11.37)
2 0
Volume elements in integrals can be written using coordinate transformations by including the
determinant of the world-sheet metric
p
|h|d2 z = dτ dσ → d2 z = 2dτ dσ (11.38)
Notice then that the term in parentheses of (11.31) can be simplified:
1
(∂τ X µ ∂τ Xµ + ∂σ X µ ∂σ Xµ )
4
1
= (∂τ X µ ∂τ Xµ + i∂τ X µ ∂σ Xµ − i∂σ X µ ∂τ Xµ + ∂σ X µ ∂σ Xµ )
4

1 1 1
= (∂τ X µ − i∂σ X µ )(∂τ Xµ + i∂σ Xµ ) = (∂τ − i∂σ )X µ (∂τ + i∂σ )Xµ
4 2 2
¯ µ
= ∂X µ ∂X (11.39)
The Polyakov action complex using the designated complex coordinates is simply
Z Z
1 µ µ 1 ¯ µ
S= dτ dσ(∂τ X ∂ X
τ µ + ∂ σ X ∂ X
σ µ ) = 2dτ dσ∂X µ ∂X
4πα0 2πα0
Z
1 ¯ µ
= d2 z∂X µ ∂X (11.40)
2πα0
As usual, let’s vary the action to determine the equations of motion. Explicitly,
Z Z
1 µ¯ 1 ¯ µ + ∂δX
¯ µ)
S→ 2
d z∂X ∂(Xµ + δXµ ) = d2 z∂X µ (∂X
2πα0 2πα0
Z Z
1 µ¯ 1 ¯ µ)
= 2
d z∂X ∂Xµ + d2 z∂X µ (∂δX
2πα0 2πα0
Therefore the variation of the action δS is just
Z Z
1 2 µ ¯ 1 ¯ µ (δXµ )
δS = d z∂X (∂δXµ ) = − d2 z∂ ∂X (11.41)
2πα0 2πα0
where we used integration by parts. For the variation to vanish we require that
¯ µ (z, z̄) = 0
∂ ∂X (11.42)
As one would recall from a basic course on complex variables, a holomorphic function (sometimes
called analytic) is a function f (z, z̄) satisfying
∂f
=0 (11.43)
∂ z̄
Thus f = f (z), a function which depends entirely on the complex coordinate z, not at all on the
conjugate z̄. Alternatively, an antiholomorphic function is one that satisfies
∂f
=0 (11.44)
∂z
¯
And therefore f = f (z̄). In string theory, if ∂(∂X µ
) = 0, then ∂X µ is holomorphic and is
said to describe a string that is left moving. On the other hand, if ∂(∂X¯ µ ) = 0, then ∂X
¯ µ is
antiholomorphic and is said to describe a string that is right moving.
Now that we have a basic understanding of the change in coordinates of the world-sheet, let’s
briefly discuss the generic D-dimensional conformal group. With brevity, the conformal group is
the subgroup of the group of general coordinate transformations (diffeomorphisms) that preserves
conformal flatness of the metric [5]. To fully understand the conformal group and the Lie algebra it
generates let’s study the generators of conformal transformations. First consider the infinitesimal
transformation of coordinates
x0µ = xµ + µ (11.45)
Now recall that an infinitesimal conformal transformation takes the form
0
gµν (x0 ) = Ω2 (x)gµν (x) (11.46)
where we suppose Ω2 (x) = 1 − f (x), with f (x) is some small departure from the identity.
Therefore we have
0
gµν (x0 ) = (1 − f (x))gµν (x) = gµν (x) − f (x)gµν (x) (11.47)
Alternatively, based on our infinitesimal coordinate transformation (11.45), one can show the
metric transforms as [37]
0
gµν = gµν − (∂µ ν + ∂ν µ ) (11.48)
Comparing both (11.47) and (11.48) we are led to
∂µ ν + ∂ν µ = f (x)gµν (11.49)
To determine the form of f (x), we multiply both sides by the inverse metric g µν . In a D-
dimensional space-time, the product g µν gµν = D. Therefore,
2
g µν (∂µ ν + ∂ν µ ) = 2∂µ µ = Df (x) → f (x) = ∂µ µ (11.50)
D
This allows us to write
2 2
∂µ ν + ∂ν µ = gµν ∂ρ ρ = gµν (∂ · ) (11.51)
D D
Taking a ∂ µ derivative of the left hand side (11.51) we have
∂ µ ∂µ ν + ∂ µ ∂ν µ = ν + ∂ν (∂ · ) (11.52)
using the standard ∂ µ ∂µ = . Taking the same derivative on the right-hand side yields

2
ν + 1 − ∂ν (∂ · ) = 0 (11.53)
D
Notice that if we move to two dimensions, D = 2, (10.53) reduces to
ν = 0 (11.54)
µ
The infinitesimal parameter can represent four different types of transformations: translations,
scale transformations, rotations, and special conformal transformations. Translations take the usual
form with µ → xµ + aµ where aµ is some constant. A scale transformation is one of the form
µ = λxµ . For rotations we write µ → xµ + ωνµ xν with ωνµ begin antisymmetric. The special
conformal transformations are less obvious however can be easily derived using the inversion element
of the conformal group [5]
xµ
xµ → (11.55)
x2
The special conformal transformation can then be determined by the sequence of transformations:
inversion, translation, inversion. Taking note of this we find
xµ xµ xµ + bµ xµ + bµ
xµ → = → =
x2 xµ xµ (xµ + bµ )(xµ + bµ ) x2 + 2(x · b) + b2
x µ + bµ x 2
→
1 + 2(b · x) + b2 x2
Assuming that bµ is infinitesimal, we find
xµ + bµ x2
xµ → ⇒ xµ + 2(b · x)xµ → xµ + bµ x2
1 + 2(b · x)
Leading to
δxµ = bµ x2 − 2(b · x)xµ (11.56)

In summary, the following infinitesimal transformations are conformal
δxµ = aµ + ωνµ xν + λxµ + bµ x2 − 2(b · x)xµ (11.57)

where aµ , bµ , ωνµ , and λ are infinitesimal constants. These operations can be combined with the
Poincaré group to form the conformal group. Using Pµ to denote translations, and Jµν to denote
rotations, we can write the generators of the conformal group as
Pµ = −i∂µ , D = −ix · ∂, Jµν = i(xµ ∂ν − Xν ∂µ ), Kµ = −i[x2 ∂µ − 2xµ (x · ∂)] (11.58)
With that, let’s move on to the conformal group of interest, the 2-dimensional conformal group.
11.3 The 2-Dimensional Conformal Group

Earlier we said that a holomorphic function is one that only depends on the complex coordinate z
and not on its complex conjugate z̄. Labeling our coordinates with the usual complex coordinates
(z, z̄), conformal transformations in two dimensions are implemented using the analytic functions
z → f (z) z̄ → f¯(z̄) (11.59)

To obtain the generators of the group, consider the infinitesimal coordinate transformations of
the form
z → z 0 = z − n z n+1 z̄ → z̄ 0 = z̄ − ¯n z̄ n+1 (11.60)

Finding an explicit form of the generators in two dimensions requires that we take derivaties of
the transformed coordinates and search for terms containing the derivatives ∂n and ∂¯ ¯n . In the
first case we have
∂
∂z 0 = (z − n z n+1 ) = 1 − n (n + 1)z n − z n+1 ∂n (11.61)
∂z
Allowing us to identify the generator [5]
`n = −z n+1 ∂ (11.62)
A similar procedure to complex conjugate coordinate yields
11.3. THE 2-DIMENSIONAL CONFORMAL GROUP 235
`¯n = −z̄ n+1 ∂¯ (11.63)

Notice then that if we take the commutator of these generators we find
[`m , `n ]f = (`m `n − `n `m )f = −z m+1 ∂(−z n+1 ∂)f − (−z n+1 ∂)(−z m+1 ∂)f
= −z m+1 −(n + 1)z n ∂f − z n+1 ∂ 2 f + z n+1 −(m + 1)z m ∂f − z m+1 ∂ 2 f

= (n + 1)z m+n+1 ∂f + z m+n+2 ∂ 2 f − (m + 1)z m+n+1 ∂f − z m+n+2 ∂ 2 f
= (m − n)(−z m+n+1 ∂)f = (m − n)`m+n f

Therefore,
[`m , `n ] = (m − n)`m+n (11.64)

A similar calculation shows
[`¯m , `¯n ] = (m − n)`¯m+n (11.65)
[`m , `¯n ] = 0 (11.66)

which is just the classical Virasoro algebra we examined in an earlier chapter. The 2-dimensional
conformal group is infinite dimensional in two dimensions however it contains a finite subgroup
generated by `0,±1 and `¯0,±1 . The corresponding infinitesimal transformations are
`−1 : z →z− (11.67)
`0 : z → z − z (11.68)
`1 : z → z − z 2 (11.69)
Comparing to the transformations we came up with earlier, we notice that (11.67) corresponds
to a translation, (11.68) a scaling, and (11.69) a special conformal transformation. In fact, `−1
and `¯−1 generate translations, (`0 + `¯0 ) generates scalings, i(`0 − `¯0 ) generates rotations, and `1
and `¯1 generates special conformal transformations. All together, the finite form of the group
transformations is
az + b
z → γ(z) = (11.70)
cz + d
where a, b, c, d ∈ C with ad − bc = 1 and γ(z) is a Möbius transformation. This finite group, as
mentioned in an earlier chapter, is in fact SO(3, 1 [5].
Figure 11.1: Using (11.73) to map the cylinder to the complex plane.
In the previous chapter we described the energy-momentum tensor of the world-sheet Tαβ . The
energy-momentum quantity meaning
∂ α Tαβ = 0 (11.71)
After the Wick the rotation and keeping in mind that the energy-momentum tensor was found to
be traceless, the non-vanishing components are Tzz and Tz̄z̄ . Therefore, the conservation condition
is changed to
¯ zz = 0
∂Tz̄z̄ = ∂T (11.72)
Therefore the energy-momentum tensor is composed of holomorphic and antiholomorphic func-
tions, Tzz = T (z) and Tz̄z̄ = T (z̄). Moving on, let’s see what happens when the describe the
world-sheet of the closed string (an infinitely long cylinder) using conformal field theory by making
the conformal transformations
z = eτ +iσ z̄ = eτ −iσ (11.73)

This transformation effectively maps the cylinder to the complex plane (see figure 11.1). The
radial coordinate plays the role of time τ with the infinite past at the origin. Moreover, now that we
have moved to the complex plane, any spatial integrals become contour integrals about the origin
in the complex plane. Now recall the left and right moving modes of the closed string:
r r
µ 1 µ α0 µ α0 X α̃nµ −in(τ +σ)
XL (τ, σ) = x0 + α̃0 (τ + σ) + i e (11.74)
2 2 2 n
n6=0
r r
µ 1 µ α0 µ α0 X αnµ −in(τ −σ)
XR (τ, σ) = x + α (τ − σ) + i e (11.75)
2 0 2 0 2 n
n6=0
Using our conformal transformation the left and right movers become
r r
µ 1 µ α0 µ α0 X α̃nµ −n
XL (z) = x0 − i α̃0 ln(z) + i z (11.76)
2 2 2 n
n6=0
11.3. THE 2-DIMENSIONAL CONFORMAL GROUP 237
r r
µ 1 µ α0 µ α0 X αnµ −n
XR (z̄) = x0 − i α0 ln(z̄) + i z̄ (11.77)
2 2 2 n
n6=0
Though we won’t go into the details, from here one can compute the holomorphic component of
the energy-momentum tensor
∞
X Ln
Tzz (z) = n+2
(11.78)
n=−∞
z
From studying complex variables, we are able to recognize that (11.78) is a Laurent expansion.
Recall from standard complex analysis that a Laurent expansion of a holomorphic function f takes
the form
∞
X
f (z) = bn (z − a)n (11.79)
n=−∞
Using Cauchy’s integral formula,

I
1 f (z)
f (a) = dz (11.80)
2πi γ (z − a)
where γ is some rectifiable path in the complex plane, one can show that the coefficient bn of
the Laurent expansion can be determined by
I
1 f (z)
bn = dz (11.81)
2πi γ (z − a)n+1
This allows us to invert (11.78) yielding the Laurent coefficient Ln :

I
1
Ln = dzz n+1 Tzz (z) (11.82)
2πi γ
Similarly, the antiholomorphic component is given as
∞
X L̄n
Tz̄z̄ (z̄) = n+2
(11.83)
n=−∞
z̄
Or,
I
1
L̄n = dz̄ z̄ n+1 Tz̄z̄ (z̄) (11.84)
2πi γ
As we will see momentarily, the Laurent coefficient Ln is actually a Virasoro operator as it

satisfies the Virasoro algebra.
11.4 Propagators and Operator Product Expansions

To gain some physical insight let’s begin by calculating the propagators for the closed string theory.
Let’s start by considering the propagator or Wick expansion for X µ (τ, σ) associated with the closed
string. Recall from our review of quantum field theory that the propagator explains how a particle
is moved within the vacuum, popping into existence at some location and then transported to some
other point before it is annihilated. Another way of computing the propagator is done by calculating

0 0
µ ν
X (τ, σ), X (τ , σ ) = T (X µ (τ, σ), X ν (τ 0 , σ 0 ))− : X µ (τ, σ)X ν (τ 0 , σ 0 ) : (11.85)
where T indicates that our expression is time-ordered. That is,
T (X µ (τ, σ), X ν (τ 0 , σ 0 )) = X µ (τ, σ)X ν (τ 0 , σ 0 )Θ(τ − τ 0 ) + X ν (τ 0 , σ 0 )X µ (τ, σ)Θ(τ 0 − τ ) (11.86)
Moreover notice that we have used a colon in (11.85) denoting that the propagator is normal-
ordered, which puts all creating operators to the left of the annihilation operators. Our goal here
is to determine the vacuum expectation of the time-ordered product. In closed string theory, X µ
can be written in terms of left and right movers. Therefore, notice the vacuum expectation value
of the product of two different sting coordinates is
h0|X µ (τ, σ)X ν (τ 0 , σ 0 )|0i = h0|(XLµ + XR

µ
)(XLν + XR
ν
)|0i
= h0|XLµ XLν + XLµ XR

ν µ ν
+ XR µ ν
XL + XR XR |0i (11.87)
µ µ
Moving to the complex plane via our conformal transformation, we have XR = XR (z̄) and
XLµ = XLµ (z). Hence,
h0|X µ (τ, σ)X ν (τ 0 , σ 0 )|0i → h0|X µ (z, z̄)X ν (z 0 , z̄ 0 )|0i (11.88)

What we have next is a lengthy calculation in computing the vacuum expectation of the product
of string coordinates. From (11.87) we have four terms to consider. Here we will calculate the
µ ν
XR XR term and leave the rest for the reader’s pleasure. We begin by computing the product of
the explicit expressions for the right movers:
 r r  r r 
1 α0 α0 X αµ 0
α ν α 0 ν
αm 0−m 
n −n   1 ν
µ ν
X
XR XR =  xµ0 − i α0µ ln(z̄) + i z̄ x −i α ln(z̄ 0 ) + i z̄
2 2 2 n 2 0 2 0 2 m
n6=0 m6=0
r r r
1 i α0 µ ν i α0 X 1 µ ν 0−m i α0 µ ν
= xµ0 xν0 − x α ln(z̄ 0 ) + x α z̄ − α x ln(z̄)
4 2 2 0 2 2 m 0 m 2 2 0 0
m6=0
r
α0 α0 X 1 µ i α0 X 1 µ ν −n
− ln(z̄)ln(z̄ 0 )α0µ α0ν + ln(z̄) α αν z̄ 0−m + α x z̄
2 2 m 0 m 2 2 n n 0
m6=0 n6=0
11.4. PROPAGATORS AND OPERATOR PRODUCT EXPANSIONS 239
α0 X1 α0 X 1 µ ν −n 0−m
+ ln(z̄ 0 ) αnµ α0ν z̄ −n − α α z̄ z̄
2 n 2 nm n m
n6=0 n,m6=0
µ
Using αm |0i= 0 for m > 0; h0|αnµ = 0 for n < 0, and α0ν |0i = 0, we find that the vacuum
expectation value of this product is
r r
µ ν 1 i α0 X 1 i α0
h0|XR XR |0i = h0|xµ0 xν0 |0i + h0|xµ0 αm
ν
|0iz̄ 0−m − ln(z̄)h0|α0µ xν0 |0i
4 2 2 m<0 m 2 2
r
α0 X 1 i α0 X 1 α0 X X 1
+ ln(z̄) h0|α0µ αm
ν
|0iz̄ 0−m + h0|αnµ xν0 |0iz̄ −n − h0|αnµ αm
ν
|0iz̄ −n z̄ 0−m
2 m<0
m 2 2 n>0 n 2 n>0 m<0 nm
Recall that α0mu commutes with all other of the oscillators, which gets rid of the fourth term
above. Moreover, αnµ commutes with xν0 , in turn cancelling the second term and fifth term. Thus,
we are left with
r
µ ν 1 i α0 α0 X X 1
h0|XR XR |0i = h0|xµ0 xν0 |0i − ln(z̄)h0|α0µ xν0 |0i − h0|αnµ αm
ν
|0iz̄ −n z̄ 0−m (11.89)
4 2 2 2 n>0 m<0 nm
√
Using [xµ0 , α0ν ] = 2α0 iη µν we find
√ √
h0|α0µ xν0 |0i = h0|xµ0 α0ν − 2α0 iη µν |0i = − 2α0 iη µν
Moreover, looking at the last term in (11.89), we may add zero in a way that will simplify this
expression:
XX 1 XX 1
h0|αnµ αm
ν
|0i = h0|αnµ αm
ν ν µ
− αm αn |0i
n>0 m<0
nm n>0 m<0
nm
ν
Making use of [αnµ , αm ] = nη µν δn,−m = −[αm
ν
, αnµ ] = −mη µν δm,−n we find that the last term
becomes
XX 1 X1
h0|αnµ αm
ν
|0i = − η µν
n>0 m<0
nm n>0
n
Altogether we have
µ ν 1 α0 α0 X 1 µν −n 0n
h0|XR XR |0i = h0|xµ0 xν0 |0i − η µν ln(z̄) + η z̄ z (11.90)
4 2 2 n>0 n
Now using the Taylor series expansion

∞
X xn
= −ln(1 − x)
n=1
n
we find that (11.90) becomes

µ ν 1 α0 α0 z̄ 0
h0|XR XR |0i = h0|xµ0 xν0 |0i − η µν ln(z̄) + ln(1 − )
4 2 2 z̄
1 α0
h0|xµ0 xν0 |0i − η µν ln(z̄ − z̄ 0 )
= (11.91)
4 2
If we complete similar calculations, adding up each contribution from (11.87), we find
1 α0
h0|X µ (z, z̄)X ν (z 0 , z̄ 0 )|0i = h0|xµ0 xν0 |0i − η µν (ln(z − z 0 ) + ln(z̄ − z̄ 0 )) (11.92)
2 2
The vacuum expectation value of the time-ordered product then becomes
1 α0
h0|T (X µ (z, z̄)X ν (z 0 , z̄ 0 ))|0i = h0|xµ0 xν0 |0i − η µν (ln(z − z 0 ) + ln(z̄ − z̄ 0 )) (11.93)
2 2
Let’s return to the propagator given in (11.85). The normal-ordering puts all of the destruction
operators to the right of the creation operator, thereby annihilating the vacuum. Thus, the vacuum
expectation of the normal-ordered piece in (11.85) reduces to
1
h0| : X µ (τ, σ)X ν (τ 0 , σ 0 ) : |0i = h0|xµ0 xν0 |0i (11.94)
2
Hence, the propagator (11.85) becomes
α0 µν α0
h0|T (X µ (z, z̄)X ν (z 0 , z̄ 0 ))|0i = −
η ln(z − z 0 ) − η µν ln(z̄ − z¯0 ) (11.95)
2 2
From here it’s easy to calculate other propagators. Two other propagators which show up often
in conformal field theory. The first is
α0 α0

∂X µ (z, z̄), X ν (z 0 , z̄ 0 ) = ∂ − η µν ln(z − z 0 ) − η µν ln(z̄ − z¯0 )
2 2
α0 µν 1
=− η (11.96)
2 (z − z 0 )
The other propagator is
α0 µν

1
∂z X µ (z, z̄), ∂z0 X ν (z 0 , z̄ 0 ) =− η (11.97)
2 (z − z 0 )2
A note to the reader: When perusing other texts one will observe that these same expressions
for the propagators are different by a factor of two. The subtle reason for this is because
√ some
researchers in the field work with the definition of the string length as being ` ≡ 2α 0 , while
√ s
others use `s ≡ α0 . In this text we use latter definition. Had we used the alternative definition
0 0
our results would be in exact agreement, with α4 rather than α2 .
Let’s now move on to another important tool in conformal field theory: operator product expan-
sions or OPE’s. Essentially, an operator product expansion is a series expansion of a product of
two operator valued fields. Specifically, an OPE is a product of local operators in a given quantum
field theory defined at nearby locations which can be expanded in a series of the local operators
at one of their positions [5]. Let’s denote these field operators as Ai and consider two space-time
points z and w. Then, in some region R that does not contain w we may write the product of these
two field operators as [37]
X
Ai (z)Aj (w) = cijk (z − w)Ak (w) (11.98)
k
The coefficients cijk here are actually holomorphic functions in the region R, while the Ak (w)
are operator valued fields. Now define a conformal transformation as z → w(z). A conformal field
or primary field Φ is one that transforms as
h h̄
∂w ∂ w̄
Φ(z, z̄) = Φ(w, w̄) (11.99)
∂z ∂ z̄
We say that (h, h̄) are the conformal weights or conformal dimension of the field Φ. If this
transformation holds, we notice
Φ(z, z̄)(dz)h (dz̄)h̄ = Φ(w, w̄)(dw)h (dw̄)h̄
leading us to conclude that Φ(z, z̄)(dz)h (dz̄)h̄ is invariant under conformal transformations.
Remember that with our conformal transformation (11.73) the time coordinate τ became the
radial coordinate in the complex plane. Therefore, when working in the complex plane, time-
ordering is transformed into radial ordering since the radial direction encodes the flow of time in
the complex plane. More rigorously, consider two operators defined in the complex plane A(z) and
B(w). The radial ordering operator R fixes the order of the operators based on which has the larger
radius in the complex plane,(equivalently, which has had more time pass). That is,

A(z)B(w) : |z| > |w|
R[A(z)B(w)] =
(−1)f B(w)A(z) : |w| > |z|
Therefore, the time ordering operator is completely analogous to the radial ordering operator
introduced above. As it turns out, if the operators are fermionic, then f = 1 [37]. If the operators
are bosonic, then f = 0, yielding the time-ordering operator behavior we are familiar with One
operator product expansion of particular interest involves the energy-momentum tensor:
Tzz (z) =: ηµν ∂X µ ∂X ν : (11.100)

Let’s work out the OPE of the radially ordered product Tzz (z)∂X ρ (w). Before we do this
however, let’s first briefly introduce Wick’s theorem.
In quantum field theory, Wick’s theorem is a very useful tool for reducing the expectation value
of a generic time-ordered product of fields
h0|T (φ(x1 )...φ(xn ))|0i (11.101)

to a combination of Feynman propagators. It generalizes the identity
T (φ(x)φ(y)) =: φ(x)φ(y) : +hφ(x), φ(y)i (11.102)

to an n-point function and states that T (φ(x1 )...φ(xn )) is equal to the normal ordered product
: φ(x1 )...φ(xn ) : plus all combinations of normal-ordering and contractions of fields, where a con-
traction of two fields φ(x1 ), φ(x2 ) is defined to equal the Feynman propagator hφ(x1 ), φ(x2 )i. For
example (Maggiore, 123):
T (φ1 φ2 φ3 φ4 ) =: φ1 φ2 φ3 φ4 : +hφ1 , φ2 i : φ3 φ4 : +hφ1 , φ3 i : φ2 φ4 :

+hφ1 , φ4 i : φ2 φ3 : +hφ2 , φ3 i : φ1 φ4 : +hφ2 φ4 i : φ1 φ3 : +hφ3 , φ4 i : φ1 φ2 :
hφ1 , φ2 ihφ3 , φ4 i + hφ1 , φ3 ihφ2 , φ4 i + hφ1 , φ4 ihφ2 , φ3 i
In our case, we use the radial-ordering operator, which is the time-ordering operator in the
complex plane. In that sense, we can still use Wick’s theorem in a general sense. Therefore, using
the radial-ordering operator and (11.100), one can show that it follows from Wick’s theorem that
R(Tzz (z)∂X ρ (w)) = R(: ηµν ∂X µ ∂X ν : ∂X ρ (w))

= ηµν µ ρ ν
∂z X (z), ∂w X (w) ∂z X (z) + ηµν ∂z X (z), ∂w X (w) ∂z X µ (z)
ν ρ
(11.103)
Then using our result from (11.97), we have that the above becomes
α0 1 α0 1
R(Tzz (z)∂X ρ (w)) = − η µν η µρ
∂z X ν
(z) − ηµν η νρ ∂z X µ (z)
2 (z − w)2 2 (z − w)2
1
= −α0 ∂z X ρ (z)
(z − w)2
By expanding ∂z X ρ (w) in a power series about z = w we find
(z − w)2 3 ρ
∂z X ρ (z) = ∂w X ρ (w) + (z − w)∂z2 X ρ (z) + ∂z X (z) + ...
2!
Hence,
1
R(Tzz (z)∂X ρ (w)) = −α0 ∂z X ρ (z)
(z − w)2
1 1 (z − w)2 3 ρ
= −α0 2
∂w X ρ (w) − α0 ∂z2 X ρ (z) − α0 ∂z X (z) + ...
(z − w) (z − w) 2!
The singular terms are what is of interest so typically we only write the singular terms and
represent all of the other terms in the series by +.... Therefore,
1 1
R(Tzz (z)∂X ρ (w)) = −α0 ∂w X ρ (w) − α0 ∂ 2 X ρ (z) + ... (11.104)
(z − w) 2 (z − w) z
Another OPE of interest is the energy-momentum tensor with itself as it will lead to the Virasoro
algebra. Using radial ordering, and making use of Wick’s theorem, it can be shown that
R(Tzz (z)Tww (w)) = R (: ηµν ∂z X µ (z)∂z X ν (z) : : ηρσ ∂w X ρ (w)∂w X σ (w) :)

= 2ηµν ηρσ ∂z X (z), ∂w X (w) ∂z X (z), ∂w X (w) −4ηµν ηρσ ∂z X (z), ∂w X (w) : ∂z X ν (z), ∂w X σ (w) :
µ ρ ν σ µ ρ
Then, using previous work we find that
α0 η µρ α0 η νσ

1 1
∂z X µ (z), ∂w X ρ (w) ∂z X ν (z), ∂w X σ (w) = − −
2 (z − w)2 2 (z − w)2
α0 η µρ η νσ 1
= (11.105)
4 (z − w)4
Also note that ηµν ηρσ η µρ η νσ = δ ρν δ νρ = D. Moreover if we use the same Taylor expansion on the
normal-ordered term, after a bit of messy algebra we find that the OPE of the energy-momentum
tensor with itself is just
α0 D 2α0 α0
R(Tzz (z)Tww (w)) = + Tww (w) + ∂w Tww (w) + ... (11.106)
2(z − w)4 (z − w)2 (z − w)
We are now in a position to derive the Virasoro algebra using the OPE of the energy-momentum
tensor with itself. Using (11.82), we see that the commutator we wish to compute is really
I I
dz m+1 dw n+1
[Lm , Ln ] = z Tzz (z), w Tww (w)
γ 2πi γ 2πi
I I I I
dz m+1 dw n+1 dw n+1 dz m+1
= z Tzz (z) w Tww (w) − w Tww (w) z Tzz (z)
γ 2πi η 2πi η 2πi γ 2πi
(11.107)
where γ and η are rectifiable contours in the complex plane.
Let’s start with the first term by first computing the z integral keeping w fixed. When doing the
integral we will assume that the integrand is radially ordered, allowing us to use our OPE given
in (11.106). As a consequence, we consider the contour integral along a small path encircling w
(figure 11.2). Using the operator expansion, the first integral becomes
α0 D 2α0 α0
I I
dw n+1 dz m+1
w z + Tww (w) + ∂w Tww (w) + ... (11.108)
η 2πi C 2πi 2(z − w)4 (z − w)2 (z − w)
To evaluate this integral we need a result from standard complex analysis. Recall Cauchy’s
integral formula (11.80). Differentiating with respect to a once yields,
Figure 11.2: The contour we are considering to evaluate the contour integral given in the OPE.
(Motivated by BBS)
I I
df (a) 1 d f (z) 1 f (z)
= dz = dz
da 2πi γ da (z − a) 2πi γ (z − a)2
For higher-order derivatives, by induction one can prove that
k I
d k! f (z)
f (a) = dz (11.109)
da 2πi γ (z − a)k+1
Now, let F (z) be holomorphic inside and on a simple closed curve C except for a pole of order
f (z)
k at z = w. Then F (z) = (z−w) k where f (z) is holomorphic inside and on C and f (w) 6= 0. Then,
by our generalized Cauchy’s integral formula we have that
f (k−1) (w)
I I
1 1 f (z)
F (z)dz = dz =
2πi C 2πi C (z − w)k (k − 1)!
1 dk−1
(z − w)k F (z)

= lim (11.110)
z→w (k − 1)! dz k−1
Using this result we can evaluate the integral at hand. First of all, we can deal with the non-
singular terms easily using Cauchy’s Integral Theorem, which states that if f is a holomorphic
function on an open disc W in the complex plane, and γ : [a, b] → W is a closed C 1 curve in W ,
then
I
f (z)dz = 0 (11.111)
γ
Therefore, since all of the non-singular terms are composed of holomorphic functions and our
chosen contour C is a simple closed curve, (11.108) is reduced to only a contour integral over the
singular terms shown. To compute the singular terms we make use of (11.110). The lowest ordered
singular term is simply
z m+1 0
I
1
dz α ∂w Tww (w) = lim α0 (z m+1 )∂w Tww (w) = α0 wm+1 ∂w Tww (w) (11.112)
2πi C (z − w) z→w
Similarly, the other two terms are just
z m+1
I
1
dz α0 2Tww (w) = 2(m + 1)α0 wm Tww (w) (11.113)
2πi C (z − w)2
z m+1 d3 m+1
I
1 0D 0D D
dz α = α lim (z ) = α0 (m3 − m)wm−2 (11.114)
2πi C (z − w)4 2 12 z→w dz 3 12
This leaves us with the integral over w:
I
0 dw D 3 m+n−1 m+n+1 m+n+2
α (m − m)w + 2(m + 1)w Tww (w) + w ∂w Tww (w) (11.115)
2πi 12
We can break this integral up into three terms. From Cauchy’s integral theorem and (11.110),
the last term simply integrates to zero. The second term, is easily recognized as a Virasoro operator
with a mode that is a sum of the modes m and n. That is, using (11.82) we find that
I
dw
α0 2(m + 1)wm+n+1 Tww (w) = 2α0 (m + 1)Lm+n
2πi
Lastly, the first term vanishes upon integration except for when m + n 6= 0, allowing us to write
I
0 dw D 3 D
α (m − m)wm+n−1 = α0 (m3 − m)δm+n,0
2πi 12 12
Hence we have that
I I
dz m+1 dw n+1 D
z Tzz (z) w Tww (w) = 2α0 (m + 1)Lm+n + α0 (m3 − m)δm+n,0
γ 2πi η 2πi 12
If we complete a very similar computation for the other part of the commutator (holding z
constant first, then using a contour of integration nearly identical to the one showed in figure
(11.2)), we would find
I I
dw n+1 dz m+1 D
w Tww (w) z Tzz (z) = 2α0 (n + 1)Lm+n + α0 (n3 − n)δm+n,0
η 2πi γ 2πi 12
Subtracting the two results yields
D 3
[Lm , Ln ] = 2α0 (m − n)Lm+n + 2α0
(m − m)δm+n,0 (11.116)
12
This is almost the commutator that we expect for the Virasoro algebra, however we are off by a
factor of two. What happened? Recall earlier
√ that we were using a different definition of α0 . Notice
that if we use the definition where `s = 2α = 1, we find that α0 = 21 . Thus, when using conformal
0
symmetry of the world-sheet for quantizing the string, we are required to use this definition of the
slope parameter α0 . Doing so yields the expected commutator and Virasoro algebra:
D 3
[Lm , Ln ] = (m − n)Lm+n + (m − m)δm+n,0 (11.117)
12
µ
We can also compute the commutation relations of the oscillators αm . Using the fact that
i X µ −n−1
∂X µ (z) = − α z (11.118)
2 n n
Just as before with the Virasoro operator Ln , it follows from Cauchy’s integral formula that the
oscillator takes the form
I
µ 1
αn = z n ∂X µ (z) (11.119)
π γ
µ
Therefore, the commutator between two oscillators αm and αnν is
I I
µ 1
[αm , αnν ] = 2 dzdwz m wn (∂X µ (z)∂X ν (w) − ∂X ν (w)∂X µ (z)) (11.120)
π C C
Again, let’s focus on the first term, computing the z integral first while keeping w constant and
using the same contour of integration as shown in figure 11.1. Earlier we found the radially ordered
expansion of ∂X µ (z)∂X ν (w) was just
1 1
R(∂X µ (z)∂X ν (w)) = − η µν (11.121)
4 (z − w)2
√
where we have adopted the convention that `s = 2α0 . Therefore, the first integral is just
I I
1 n m 1 1
dww dzz − η µν
π2 C C 4 (z − w)2
I I
1 d 1 1 m µν
= dwwn lim − η µν = − η dwwm+n−1
π2 C z→w dz 4 (z − w)2 4π 2 C
m µν
= η δm+n,0
2
Similarly, the other integral yields
I I
1 n
2
dzdwz m wn ∂X ν (w)∂X µ (z) = η µν δm+n,0
π C C 2
Subtracting the two results and using the Kronecker delta to our advantage we attain the ex-
pected commutation relations for the oscillators
µ
[αm , αnν ] = mη µν δm+n,0 (11.122)
A very similar calculation shows that
µ
[α̃m , α̃nν ] = mη µν δm+n,0 (11.123)
as expected.
11.5. PRIMARY FIELDS AND THE VERMA MODULE 247
11.5 Primary Fields and the Verma Module

Given a holomorphic primary field Φ(z) of dimension h, one can associate a state |Φi that satisfies
L0 |Φi = h|Φi Ln |Φi = 0 (11.124)

for n > 0. Such a state is called a highest-weight state. In fact, the primary field may be written
in a mode expansion as [5]
∞
X Φn
Φ(z) = (11.125)
n=−∞
z n+h
in which we have
Φn |0i = 0 (11.126)
for n > −h and
Φ−h |0i = |Φi (11.127)

Here we use |0i to denote the conformal vacuum. A highest-weight state |Φi, taken with the
infinite collection of descendant states in the form L−n1 L−n2 ...L−nk |Φi gives a representation of
the holomorphic Virasoro algebra known as the Verma module.
Recall from the last chapter that we had the admissible states of open string satisfy
(L0 − 1)|φi = 0, Ln |φi = 0 (11.128)

for n > 0. Therefore, we see that the admissible open string states of bosonic string theory
correspond to the highest-weight states with dimension h = 1. This construction takes a straight
foward generalization to primary fields Φ(z, z̄) of dimension (h, h̄) in which case
(L0 − h)|Φi = (L̃0 − h̄)|Φi = 0 Ln |Φi = L̃n |Φi = 0 (11.129)

with n > 0. It follows then that the admissible closed string states of bosonic string theory
correspond to the highest-weight states with h = h̄ = 1.
This concludes our brief sruvey of conformal field theory. There are several more topics in
which one can and should cover if they are interested in continuing on in this subject (Kac-Moody
algebras, coset theories, and minimal models are a few). In the next section we move to quantize
the string using the BRST procedure. Several of the results we include come from the path integral
approach and will therefore just be stated rather than derived. For derivations of some of these
results, Polchinski’s text as well as Green, Schwarz, and Witten’s text give a thorough discussion
on this approach through the use of path integrals.
11.6 BRST Quantization

Up to this point we have really studied two methods of quantizing the string: light-cone quantization
and covariant quantization. In the next couple sections we develop and introduce the third and
final quantization procedure we will discuss in this text. Remember that in light-cone quantization
the physical results were easy to interpret, however the manifest Lorentz invariance was lost. On
the other hand, covariant quantization maintained manifest Lorentz invariance but negative-norm
states were introduced as a consequence, which we had to eliminate in order to understand the
physical states of the theory. BRST quantization takes the middle ground of these two approaches:
it is manifestly Lorentz invariant but includes ghost states, however, as we will see, the physical
states are easier to identify than in the traditional covariant approach.
Let’s begin by determining the BRST operators. Consider any physical system with symmetry
operators Ki that form a closed Lie algebra,
[Ki , Kj ] = fijk Kk (11.130)

where fijk are the structure constants satisfying the identity [21]
fijm fmk
l
+ fjkm fmil + fkim fmjl = 0 (11.131)
which actually follows from (11.130) via the Jacobi identity. The BRST quantization procedure
begins by first introducing two ghost fields denoted by bi and cj and satisfying the anticommutation
relations
{ci , bj } = δ ij {ci , cj } = {bi , bj } = 0 (11.132)

Here the ci are the ghost fields while the bj are the ghost momenta. Since anticommutation
relations are satisfied by our ghost fields, we know that from our studies of quantum field theory
that these fields are fermionic.
Moving on, there are two operators that are constructed out of the ghost fields and the operators
Ki . The first is the BRST operator given by
1
Q = ci Ki − fijk ci cj bk (11.133)
2
where the operator Q is Hermitian. We say that the BRST operator is nilpotent of degree two
meaning that
Q2 = 0 (11.134)
Or, equivalently,
{Q, Q} = 0 (11.135)
To prove the nilpotency of the BRST operator, one makes use of the identity given, (11.131).
We have anticipated that this operator is in fact a conserved charge of the system. Therefore, often
Q is called the BRST charge.
The second operator used in BRST quantization is composed solely of the ghost fields and is
called the ghost number operator U given by
U = ci bi (11.136)
where there is an implicit sum over the index i. This operator has with it integer eigenvalues.
We say that a state |ψi has a ghost number m if U |ψi = m|ψi. What’s more is the BRST charge
raises the ghost number one. To see this, first notice
11.6. BRST QUANTIZATION 249
X X X
U ci Ki = cr br ci Ki = cr (δ ir − ci br )Ki = ci Ki − ci cr br Ki
r r r
X
i i r i i
= c Ki + c Ki c br = c Ki + c Ki U
r
Now consider an arbitrary state |ψi with ghost number m. We have then

i 1 k i j
U (Q|ψi) = U c Ki − fij c c bk |ψi
2

1
= U ci Ki − cr br fijk ci cj bk |ψi
2

1
= ci Ki + ci Ki U − fijk cr br ci cj bk |ψi
2
We have dropped the sum over r as it is implied. From here we use the anticommutation relations
to our advantage and power through the algebra. The above becomes

i i 1 k r i i j
c Ki + c Ki U − fij c (δr − c br )c bk |ψi
2

i i 1 k i j 1 k i r j
= c Ki + c Ki U − fij c c bk + fij c c br c bk |ψi
2 2

1 1
= ci Ki + ci Ki U − fijk ci cj bk + fijk ci cr (δrj − cj br )bk |ψi
2 2

1
= Q + ci Ki U + fijk ci cr (δrj − cj br )bk |ψi
2

i 1 k i j 1 k i j r
= Q + c Ki U + fij c c bk − fij c c c bk br |ψi
2 2

i 1 k i j 1 k i j k i k r
= Q + c Ki U + fij c c bk − fij c c (δr − bk c (δr − bk c )br |ψi
2 2

i 1 k i j r
= Q + c Ki U − fij c c c br |ψi
2
= (Q + QU )|ψi = (1 + m)Q|ψi (11.137)

Therefore the BRST charge Q raises the ghost number m by 1.
A state |ψi is said to be BRST invariant if it is annihilated by the BRST operator:
Q|ψi = 0 (11.138)
As it turns out, BRST invariant states are the physical states of the theory [37]. Since Q2 = 0
it follows that any state of the form |ψi = Q|χi =
6 0 is BRST invariant since
Q|ψi = Q2 |χi = 0
We call the state |ψi = Q|χi a null state. Now, suppose that |φi is an arbitrary physical state.
Notice then
hφ|ψi = hφ|Q|χi = (hφ|Q)|χi = 0

Hence, any amplitude taken between a physical state and a null state vanishes. This result is
similar to the notion of a phase factor in ordinary Quantum Mechanics. Two states that differ by
a phase a factor represent the same physical state since they yield the same amplitude. Similarly,
any inner product between and a physical state and a null state vanishes, and thus adding a null
state |ψi = Q|χi to a physical state |φi generates a new state which is physically equivalent to |φi.
Moreover, since Q raises the ghost number by 1, if the ghost number of |ψi is m, it follows that
the ghost number of |χi is m − 1. A case of particular interest is when a state |ψi has a ghost
number of zero:
U |ψi = 0 (11.139)
This implies that
bk |ψi = 0 (11.140)
That is, the a state with zero ghost number is annihilated by all of the bk . Furthermore, a state
annihilated by the ghost fields bk cannot be annihilated by the ghost fields ck since
U = ci bi = δ ij − bi ci = n − bi ci
where n is the dimension of the Lie algebra. We see then
U |ψi = (n − bi ci )|ψi = n|ψi − bi ci |ψi

Therefore, if bk |ψi = 0 the ghost fields ck cannot annihilate the state as it would result in a
contradiction. Looking at the BRST charge, it is clear that if bk |ψi = 0 we have
Q|ψi = ci Ki |ψi (11.141)

Since the ci cannot annihilate the state, for |ψi to be a BRST invariant state we instead require
that
Ki |ψi = 0 (11.142)
for i = 1...n. In summary, the state which is BRST invariant with a ghost number of zero is also
invariant under the symmetry described by the generators Ki . What’s more is that if the state has
a ghost number of zero it tells us that the state is actually not a ghost state. For the most part
this is what we want as it allows us to isolate the states invariant under the Lie algebra that do
not contain ghosts, which is imperative in determining the physical states of our theory. For this
reason, this approach turns out to be useful in string theory.
11.7. BRST SYMMETRY AND STRING THEORY 251
11.7 BRST Symmetry and String Theory

Here we will consider two approaches to BRST quantization, one of which makes heavy use of path
integrals, and where we simply state some of the results avoiding the lengthy, technical details. The
first approach we consider however makes heavy use of conformal field theory. The advantage of
this approach is that the critical space-time dimension D arises in a relatively straightfoward way.
With this second method we work in the conformal gauge hαβ = ηαβ . Moreover, in this approach
we introduce the ghost fields as functions of a complex variable z as follows. We define
1 1
b(z)c(w) = b̄(z̄)c̄(w̄) = (11.143)
(z − w) (z̄ − w̄)
Next, one can show that the energy-momentum tensor Tgh (z) for the ghost fields takes the form
Tgh (z) = −2b(z)∂z c(z) − ∂z b(z)c(z) (11.144)

Using the energy-momentum tensor along with the energy-momentum tesnor associated with
the ghost fields it can be shown that the BRST current takes the form [37]

1
j(z) = c(z) Tzz (z) + Tgh (z) = c(z)Tzz (z) + c(z)∂z c(z)b(z) (11.145)
2
From here the BRST charge is calculated by
I
dz
Q= j(z) (11.146)
γ 2πi
Now recall that the OPE of the energy-momentum tensor with itself took the form given in
(11.106). The first term of the OPE was found to be
α0 D
2(z − w)4
The presence of this term is called the conformal anomaly as it ends up preventing the Virasora
algebra from closing. To get rid of this anomaly, we consider the total-energy momentum tensor,
which is composed of the string energy-momentum tensor and the ghost energy momentum tensor,
T = Tzz (z) + Tgh (z). It can be shown that the OPE of the ghost energy-momentum tesnor is
−13α0 2Tgh (w) 1

Tgh (z)Tgh (w) = 4
− 2
− ∂w Tgh (w) (11.147)
(z − w) (z − w) (z − w)
0
−Dα
Notice then that if we take the leading term of this OPE in the form 2(z−w) 4 , we see that the
ghost fields contribute a central charge of −26, precisely canceling the conformal anomaly. Actually,
this result comes from the nilpotency of Q [21].
Let’s move on to the alternative approach to BRST quantization by considering a set of BRST
transformations. This method actually stems from the Faddeev-Popov procedure we briefly dis-
cussed in the last chapter. This procedure makes heavy use of path integrals and ends up introducing
ghost and anti-ghost fields often referred to as the Faddeev-Popov ghosts. Working in light-cone
coordinates, we define the ghost field c and the anti-ghost field b where b and c have the components
gh
c+ , c− and b++ , b−− . We also introduce the energy-momentum tensor for the ghost fields T±± given
by
gh 1
T++ = −i( c+ ∂+ b++ + ∂+ c+ b++ ) (11.148)
2
gh 1
T−− = −i( c− ∂− b−− + ∂− c− b−− ) (11.149)
2
The BRST transformations, using a small anticommuting operator are just
δX µ = i(c+ ∂+ + c− ∂− )X µ (11.150)
δc± = ±i(c+ ∂+ + c− ∂− )c± (11.151)
gh
δb±± = ±i(T±± + T±± ) (11.152)
The action of the ghost field in the conformal gauge is just
Z
i
SF P = d2 σ(c+ ∂− b++ + c− ∂+ b−− ) (11.153)
π
One can in fact prove that (11.153) coincides with the Faddeev-Popov action that would be used
in the Faddeev-Popov procedure [21]. From here the equations of motion are calculated to be
∂− b++ = ∂− c+ = 0 (11.154)
∂+ b−− = ∂+ c− = 0 (11.155)
Open string boundary conditions imply that c− = c+ and b++ = b−− , allowing us to determine
the mode expansions
∞
X ∞
X
c+ = cn e−in(τ +σ) c− = cn e−in(τ −σ) (11.156)
n=−∞ n=−∞
∞
X ∞
X
b++ = bn e−in(τ +σ) b−− bn e−in(τ −σ) (11.157)
n=−∞ n=−∞
The canonical anticommutation relations for the modes are just
{cm , bn } = δm+n (11.158)
{cm , cn } = {bm , bn } = 0 (11.159)

The Virasoro operators are defined for the ghost fields using the modes. Using normal-ordered
expansions, we have
X
Lgh
m = (m − n) : bm+n c−n : (11.160)
n
11.7. BRST SYMMETRY AND STRING THEORY 253
X
L̃gh
m = (m − n) : b̃m+n c̃− : (11.161)
n
The total Virasoro operator is simply the sum of a real Virasoro operator and the ghost field
Virasoro operator. That is
Ltot gh
m = Lm + Lm − aδm,0 (11.162)
where aδm,0 is a normal-ordering constant. It can be shown that the commutator of two total
Virasoro operators is
[Ltot tot tot

m , Ln ] = (m − n)Lm+n + A(m)δm+n,0 (11.163)
where have the conformal anomaly keeping us from closing the algebra:
D 3 1
A(m) = (m − m) + (m − 13m3 ) + 2am (11.164)
12 6
Notice that if D = 26 and a = 1 we find that
65m3 − 77m
+ 2m
6
which vanishes for all m. The BRST charge is given by the mode expansion
X 1X
Q= cn L−n + (m − n) : cm cn b−m−n : −c0 (11.165)
n
2 n,m
Using the nilpotency of Q, after some tedious algebra one can show that
1X 1
Q2 = [Ltot tot tot

m , Ln ] − (m − n)Lm+n c−m c−n = (D − 26) (11.166)
2 m,n 12
Therefore, in order for the nilpotency condition to be satisfied, Q2 = 0, we require the dimension
of space-time be D = 26.
Let’s move on to briefly examine the physical spectrum of the open string, as the analysis will
be similar for closed strings. The states in BRST quantization are built up from the ghost vacuum
state |χi, which is annihilated by all of the positive ghost modes
bn |χi = cn |χi = 0
for n > 0. As in BRST invariant states, the zero modes yield interesting results as they are used
to build the physical states of the theory. Using the anticommutation relations, the zero modes
satisfy
{b0 , c0 } = 1 (11.167)
We also require that
b0 |ψi = 0 (11.168)
for physical states |ψi. Let’s now construct tow state systems from the zero modes. The basis
states are denoted by | ↑i and | ↓i. We may also choose that
b0 | ↓i = c0 | ↑i = 0 (11.169)
in which case we find
b0 | ↑i = | ↓i c0 | ↓i = | ↑i (11.170)
We define the ghost vacuum state to be | ↓i. We may think of these states as containing no ghosts
or anti-ghosts. Acting on physical states of these forms, we find that BRST invariance becomes
!
X
Q|ψi = c0 (L0 − 1) + c−n Ln |ψi = 0
n>0
Therefore on the condition that Q|ψi = 0, we have reproduced all of the physical state conditions
of the covariant quantization we examined in the previous chapter.
We have finally finished quantizing the bosonic string! We have studied two of the methods
extensively, light-cone quantization and classical covariant quantization, and have developed a va-
riety of tools to study BRST quantization using conformal field theory. Plenty of material has been
overlooked in this chapter to maintain the undergraduate feel, but to go on in this subject it is
crucial that one is comfortable with these more technical topics. For this text however, we will not
use much of the material in this chapter as it is not fundamental for where we are headed. Though
we won’t discuss the topic in this text, if one were to study the interactions of strings, conformal
field theory and the methods of BRST quantization are imperative.
11.8 Exercises
1. (a) Prove that the set of integers Z form a group under the binary operaton of addition, ‘+0 .
(b) One particularly important symmetry of physics is rotational invariance. In the rest of this
problem and the next problem we will check to see that the set of rotations in three dimensional
space form a group (in fact a Lie group). One of the most convenient ways of characterizing a
rotation is to a real 3 × 3 matrix representation R that is orthogonal with a determinant of one.
The orthogonality condition gives us RRT = RT R = I, or using component form:
X
Rji Rik = δjk
i
Using this, show that the following differential equation is rotation invariant:
2 2
∂2 ∂2

~ ∂ ∂
+ + − V (r) + i~ ψ(~x, t) = 0
2m ∂x21 ∂x22 ∂x23 ∂t
√
with r = ~x2 , and where we may write rotated coordinates x0i as
X
x0i = Rij xj
i
11.8. EXERCISES 255
(c) Using the fact that we are represnting our rotations as real 3 × 3 orthogonal matrices with
determinant 1, show that this set of matrices form a group under matrix multiplication. Is this
group abelian? In fact, the rotation matrices are elements of the special orthogonal group in three
dimensions, SO(3).
2. (Continuation) Rotations intuitively seem to be continuous operations. To study this, let us

consider infinitesimal rotations corresponding to a parameter α. For simplicity, consider rotations
around the z-axis, then the coordinates under a rotation transform as:
x → cos(α)x + sin(α)y y → cos(α)y − sin(α)x z→z

(a) Write out the transformation matrix characterized by this operation.
(b) This matrix encodes a rotation by an angle α. One way to interpret this rotation is that it
is the result of n successive rotations over an angle α/n. If we take n to be large, the rotation by a
small angle will vary from the identity infinitesimally. Ignoring terms quadtratic in α/n, show that
R(α/n) takes the form:
 
0 1 0 2
α α
R(α/n) = I + −1 0 0 + O
n n2
0 0 0
(c) Taking the limit where n → ∞ and making use of the identity:
n
A 1
e = lim I + A
n→∞ n
where A is a matrix, show R(α) = R(α/n)n = exp(αT ) for some matrix T in the limit n → ∞.
From here show that
 
0 1 0
R(α) = I + (cos(α) − 1) −1 0 0 + sin(α)T
0 0 0
which coincides with the matrix for R(α) given above.
(d) We can consider a very similar analysis for any generic rotation in three dimensions which
leads to a relation similar to (11.14):
[Li , Lj ] = ck ij Lk
where Li are the following Hermitian matrices:
     
0 0 0 0 0 i 0 −i 0
L1 = 0 0 −i L2 =  0 0 0 L3 =  i 0 0
0 i 0 −i 0 0 0 0 0
With explicit matrix multiplication, determine the structure constants ck ij (try using the Levi-
Civita tensor, ijk ) Then, using the Jacobi identity show that the structure contants obey
cmij cnmk + cmjk cnmi + cmki cnmj = 0

which leads to the identity for the Levi-Civita tensor ijk :
ijm mkn + jkm min + kim mjn

(Hint: The following identity helps ijm mkl = δik δjl − δil δjk ).
3. Using (11.25) verify that the world-sheet metric takes the form given in (11.37).
4. Prove (11.65) and (11.66).
5. Prove (11.123).
Chapter 12
D-Branes
12.1 Some Notation

An interesting and more recent development in string theory is the realization that the theory
can incorporate higher dimensional extended objects called Dp-branes. Throughout this text we
have been hinting at the existence of such objects. In this chapter we will give a detailed analysis
of Dp-branes in the context of light-cone quantization and qualitatively review the Dp-branes of
superstring theory. The first few sections of this chapter is based on Zwiebach’s text, in which case
the reader is urged to review his work for further details.
A Dp-brane is an extended object with p spatial dimensions. In bosonic string theory, the
number of spatial dimensions is p = 25 and hence the D-branes of bosonic string theory are really
D25 branes. Since D = 26 in bosonic string theory, the D25 is a space-filling brane as it fills out
all of the spatial dimensions. The distinguishing characteristic of D-branes is that open strings
attach to them, i.e. the endpoints of open strings lie on the D-brane, where they end up satisfying
Dirichlet boundary conditions, hence the D in D − brane. Therefore D-branes are extended objects
with particular properties; not all extended objects in string theory are D-branes. For instance,
strings themselves are 1-branes since they are one-dimensional extended objects, but they are not
D1 branes. Generally, branes with p spatial dimensions are called p-branes. In that sense, a 0-brane
can be interpreted as some type of particle.
Before we begin quantizing the string in the presence of D-branes, let’s first develop some notation
to describe D-branes. Here will we use d to denote the total number of spatial dimensions, in which
case the total number of space-time dimensions is D = d + 1.We will mostly be concerned with
bosonic strings, and therefore a Dp-brane with p < 25 extends over a p-dimensional subspace of
the 25 dimensional space. For now we will analyze Dp-branes that are p-dimensional hyperplanes
in a d dimensional space. To describe such hyperplanes we need as many conditions as there are
spatial coordinates normal to the brane. Consider such a Dp brane with space-time coordinates xµ
with µ = 0...25 and let’s split these coordinates into two groups:
{x0 , x1 , ...xp } {xp+1 , xp+2 , ...xd } (12.1)
where the first set of coordinates are those that are tangential to the brane world-volume while
the second set consists of the (d − p) coordinates that are normal to the brane world-volume. The
257
258 CHAPTER 12. D-BRANES
actual location of the Dp-brane is specified by fixing the values of the coordinates normal to the
D-brane. We write this as xa = x̄a for a = p + 1, ...d, where x̄a are a set of (d − p) constants.
The string coordinates X µ (τ, σ) are defined in an analogous way as the space-time coordinates:
{X 0 , X 1 , ...X p } {X p+1 , X p+2 ...X d } (12.2)

Since the endpoints of the open string lie on the Dp brane, the string coordinates normal to the
brane satisfy Dirichlet boundary conditions:

a
= x̄a

X (τ, σ) (12.3)
σ=0,π
a
for a = p + 1, ...d. The string coordinates X are called DD coordinates since both ends of the
string satisfy Dirichlet boundary conditions [64]. Moreover, the open string endpoints can move
freely along the directions tangential to the D-brane. Therefore, the string coordinates tangential
to the D-brane satisfy Neumann boundary conditions
∂X m

=0 (12.4)
∂σ σ=0,π
for m = 0, 1...p. These string coordinates are called N N coordinates. To make our quantization
procedure easier we will make use of the light-cone gauge. For this we need at least one spatial
N N coordinate that can be used together with X 0 to define the coordinates X ± . We therefore
are required to assume that p ≥ 1, in which case our analysis will not include strings attached
to D0-branes. To study these requires Lorentz covariant quantization. We label the light-cone
coordinates as
X + , X − , {X i }, {X a } (12.5)
+ − i a
where the X , X , {X } the N N coordinates with i = 2...p, and the {X } are the DD coordi-
nates with a = p + 1...d.
12.2 Quantizing the String on Dp-Branes

First recall that
1 1 I 0 2
0

Ẋ − ± X − = 0 +
Ẋ ± X I (12.6)
2α 2p
0
Back in chapter five we found that the mode expansion of the Ẋ I ± X I was just
0 √ X
Ẋ I ± X I = 2α0 αnI e−in(τ ±σ) (12.7)
n∈Z
With these expressions, along with a similar expression for X − , we were led to
25 25
1 I I X I I X
L⊥
0 ≡ α0 α0 + α−p αp = α0 pI pI + paI∗ I
p ap (12.8)
2 p=1 p=1
when we quantized the open string in the light-cone gauge. Moreover, from here we defined
12.2. QUANTIZING THE STRING ON DP-BRANES 259
1
2α0 p ≡ L⊥

+ 0 +a (12.9)
p
where a is the ordering constant that we found to equal −1. Summarizing these results we have
∞
!
+ − 1 1 I I X I I
2p p ≡ 0 α α + α α +a (12.10)
α 2 0 0 n=1 −n n
The light-cone labels I = 2...25 end up taking the values of the NN coordinates labeled by i and
a, in which case we have
1 1 h i 0i 2
i
0a 2
Ẋ − ± X 0− = a

Ẋ ± X + Ẋ ± X (12.11)
2α0 2p+
where
√ X
Ẋ i ± X 0i = 2α0 αni e−in(τ ±σ) (12.12)
n∈Z
Let’s now quantize open strings that have their endpoints attached to Dp-branes. The coordi-
nates X a normal to the D-brane satisfy the wave equation in which the general solution
1 a
X a (τ, σ) = (f (τ + σ) + g a (τ + σ)) (12.13)
2
At σ = 0 we find
1 a
X a (τ, 0) = (f (τ ) + g a (τ )) = x̄a
2
where we see
g a (τ ) = −f a (τ ) + 2x̄a
yielding
1 a
X a (τ, σ) = x̄a + (f (τ + σ) − f a (τ − σ)) (12.14)
2
Moreover, at σ = π we find f a (τ + π) = f a (τ − π), indicating that f a (u) is periodic with a period
of 2π. We are then allowed to come up with the expansion [64]
∞
X
f a (u) = f˜0a + fña cos(nu) + g̃na sin(nu)

(12.15)
n=1
Upon subsitution we find

∞
X
X a (τ, σ) = x̄a + −fña sin(nτ ) sin(nσ) + g̃na cos(nτ ) sin(nσ)

(12.16)
n=1
Redefining the expansion coefficients allows us to write this as

∞
X
X a (τ, σ) = x̄a + fna cos(nτ ) + fña sin(nτ ) sin(nσ)

(12.17)
n=1
To quantize a theory in a general, as we have done so far in this text, one looks for the dynamical
variables of the system and promote them to operators. For ordinary quantum mechanics, we
upgraded the classical dynamical variables x and p to become the position and momentum operators.
In quantum field theory, the dynamical variables were the fields and conjugate momenta and were
promoted to become field operators. Here, since we are attempting to quantize a string to a fixed
Dp-brane, the x̄a are not dynamical variables, but rather the coefficients f a , fã are. Therefore,
upon quantization, the constants x̄a remain unchanged while f a , fã are promoted to becoming
operators. Writing X a (τ, σ) in terms of oscillators we find
√ X1
X a (τ, σ) = x̄a + 2α0 αa e−inτ sin(nσ) (12.18)
n n
n6=0
It’s important to note that from this expansion, and since x̄a is merely a constant, the zero mode
oscillator α0a simply does not exist. The τ and σ derivatives can be combined to give
√ X
X 0a ± Ẋ a = 2α0 αna e−in(τ ±σ) (12.19)
n6=0
From here the quantization is relatively straight forward. The non-vanishing commutators,
motivated by the other equal time commutators, are postulated to be
[X a (τ, σ), Ẋ b (τ, σ 0 )] = 2πα0 iδ ab δ(σ − σ 0 ) (12.20)

Comparing to the commutators we came up with when we first quantizd the string using light-
cone coordinates we notice that the expressions are identical except the light-cone indices (I, J)
have been changed to (a, b). Moreover, the oscillators satisfy the usual commutation relation
a
[αm , αnb ] = mδ ab δm+n,0 (12.21)
for m, n 6= 0. Using the oscillators we can split up (12.10) as
∞
!
+ − 1 0 i i
X
i
2p p ≡ 0 αpp + (α−n αni + a
α−n αna ) −1 (12.22)
α n=1
where we have made the substitution a = −1. Moreover, since pa ∼ α0a ≡ 0, we notice that the
term 21 α0I α0I becamr α0 pi pi . From here the mass-squared operator is found to be
∞
!
2 2 + − i i 1 X i i a a
M = −p = 2p p − p p = 0 (α α + α−n αn ) − 1 (12.23)
α n=1 −n n
Or, using creation and annihilation operators instead
p
∞ X ∞ ∞
!
1 X X X
M2 = 0 −1 + nai∗ i
n an + maa∗ a
m am (12.24)
α n=1 i=2 m=1 a=p+1
12.2. QUANTIZING THE STRING ON DP-BRANES 261
Now that we have the mass-squared operator we can physically interpret the quantum states of
our theory. First let’s construct the state space. Recall that the ground states with the space-filling
D25 brane were denoted as |p+ , p~T i with p~T = (p2 ...p25 ). Similarly, we denote the ground states
for the case of an open string attached to a Dp-brane as
|p+ , p~i (12.25)
with p~ = (p2 , ...pp ). As usual, we build the excited states by acting on the groud state wth the
creation operators. This time we use the creation operators ai∗ n , i = 2...p, tangential to the brane,
and the creation operators aa∗ n , a = p + 1, ...d, normal to the brane. Using both of these operators,
general states take the form
p
∞ Y ∞
" #" d
#
Y Y Y
(ai∗
n)
λn,i
(aa∗
m)
λm,a
|p+ , p~i (12.26)
n=1 i=2 m=1 a=p+1
In field theories describing the states of the string, the fields take the same form as the string
Schrödinger wavefunctions. All together, the fields end up depending on the space-time coordinates
x+ , x− , and xi with i = 2, ...p, which are just the coordinates spanning the world-volume of the
Dp brane. In that sense, it seems reaonable to conclude that the fields live on the Dp brane. To
prove this hypothesis we would have to study the interactions of closed string scattering off of open
strings attached to the D-branes. Through this scattering we can examine whether the interactions
between fields from the closed string and open string sectors occur on the D-brane world-volume.
So far, the answer appears to be that indeed the fields live on the world-volume of the D-branes
[64].
Let’s now examine some of the fields satisfying M 2 ≤ 0, all of which live on the Dp-brane. The
simplest states are the ground states |p+ , p~i which, from our mass-squared formula (12.24), have
M 2 = − α10 . These states are the tachyon states on the D-brane. Notice that they have the same
mass as the tachyon states found on the D25 brane. The next excited states have on creation
operator acting on them. First consider the ai∗ 1 operator
ai∗ +
1 |p , p
~i
with i = 2, ...p. Here in this case we have M 2 = 0. For any momenta, these are (p + 1)-2 massless
states that transform as a Lorentz vector; allowing us to identify these as the photon states. The
associated Maxwell field is a gauge field living on the brane. Therefore, we have found that a Dp
brane has a Maxwell field living on its world-volume. Lastly, using the aa∗ 1 creation operator we
find
aa∗ +
1 |p , p
~i
with a = p + 1, ...d and M 2 = 0. For any momenta, these are (d-p) states living on the brane
which transform as Lorentz scalars. Hence, a Dp-brane has a massless scalar for each normal
direction. Before moving on, its important to note that often the open string states attached to
Dp-branes are interpreted as D-brane excitations.
12.3 Stretched Strings and Parallel Dp-Branes

Let’s consider quantization of open strings that extend between two parallel Dp-branes, which
creatively call brane 1 and brane 2. These D-branes are assumed to have the same p dimensionality,
in which case that also have the same set of longitudinal and normal coordinates as defined before.
An illustration of two parallel D2-branes is given in figure 12.1. The configuration of parallel Dp-
branes supports up to four different types of strings. The first two classes of strings are those
that begin and end on the same brane. The other two classes of strings are those which begin
on one brane and end on the other. That is, strings that are stretched between the two parallel
Dp-branes.These different classes of open strings are often called sectors. It is important to note
that strings starting on brane 1 and ending on brane 2 is different than strings starting on brane 2
and ending on brane 1; they are oppositely oriented.
Let’s consider a sector consisting of open strings beginning on brane 1 and ending on brane 2.
The NN coordinates are quantized in the same way just as in the previous section. The DD string
coordinates now have an additional boundary condition

X a (τ, σ) = x̄a1 X a (τ, σ) = x̄a2

(12.27)
σ=0 σ=π
The rest of the analysis is just as before except we denote x̄a as x̄a1 , yielding the DD string
coordinates
1 a
X a (τ, σ) = x̄a1 + (f (τ + σ) − f a (τ − σ)) (12.28)
2
where we have anticipated the boundary condition at σ = 0. At σ = π we find
f a (τ + π) − f a (τ − π) = 2(x̄a2 − x̄a1 ) (12.29)
Or, with u = τ − π we have
f a (u + 2π) − f a (u) = 2(x̄a2 − x̄a1 ) (12.30)
After integrating the function f a (u) we can write it as the expansion
∞
X
f a (u) = f0a u + (han cos(nu) + gna sin(nu)) (12.31)
n=1
where
1 a
f0a = (x̄ − x̄a1 ) (12.32)
π 2
is fixed by our boundary conditions. By substitution we find that the string coordinates X a take
the form
12.3. STRETCHED STRINGS AND PARALLEL DP-BRANES 263
Figure 12.1: Open strings stretching across parallel Dp branes. Notice the four different sets of
stretched strings: (1) From brane 1 to itself; (2) From brane 2 to itself; (3) From brane 1 to brane
2; (4) From brane 2 to brane 1.
∞
σ X a
X a (τ, σ) = x̄a1 + (x̄a2 − x̄a1 ) + (f cos(nτ ) + fña sin(nτ )) sin(nσ) (12.33)
π n=1 n
If we were asked to describe strings extending from brane 2 to brane 1 instead, all we would have
to do is exchange x̄a1 and x̄a2 . In terms of oscillators, (12.33) becomes
σ √ 0 X 1 a −inτ
X a (τ, σ) = x̄a1 + (x̄a2 − x̄a1 ) + 2α α e sin(nσ) (12.34)
π n n
n6=0
It is imperative that the reader understand that the oscillators of different sectors, though denoted
the same way, are fundamentally different. Thus, the expansion above would appear very similar
to strings beginning are brane 2 and ending at brane 1, however it cannot be physically interpreted
in the same way. Combining the derivatives we have
0 √ X
X a ± Ẋ a = 2α0 αna e−in(τ ±σ) (12.35)
n∈Z
where we have
√ 1 a
2α0 α0a = (x̄ − x̄a1 ) (12.36)
π 2
Equation (12.36) is consistent with our description of open strings in the previous section since
x̄a1 and x̄a2 are constants and therefore the strings don’t carry momentum in x̄a . Moreover, letting
the light-cone index I → (i, a) just as before, we find that
∞
!
+ − 1 0 i i1 X
2p p = 0 α p p + α0a α0a + i
(α−n αni + α−n
a
αna ) − 1 (12.37)
α 2 n=1
Therefore, the mass-squared operator is just

∞
!
1 a a 1 X
M 2 = 2p+ p− − pi pi = α α + i
a
αna ) − 1 (12.38)
2α0 0 0 α0 n=1
Using (12.36), we have

2
x̄a2 − x̄a1

1
M =2
+ (N ⊥ − 1) (12.39)
2πα0 α0
where we have defined the number operator
X p
∞ X ∞
X d
X
N⊥ ≡ nai∗ i
n an + maa∗ a
m am (12.40)
n=1 i=2 m=1 a=p+1
1
The first term in (12.39) is new. Recall that we had defined the string tension as T = 2πα 0.
Therefore we recognize this first term in (12.39) as the square of energy of a classical static string
stretched between two D-branes. When the D-branes coincide, this additional term vanishes, re-
covering the mass-squared operator we are used to workin with. In other words, the tnesion of the
string between the two D-branes gives rise to a modified mass spectrum for the quantum states of
our theory.
let’s consider the ground states of the theory including all four sectors. To distinguish between
the sectors we include additional ground state labels [i, j], each of which take the values 1, 2. The
first integer denotes the brane on which σ = 0 while the second denotes the brane on which σ = π.
Since there are four sectors, we have four different ground states |p+ , p~; [ij]i to consider. Namely,
|p+ , p~; [11]i, |p+ , p~; [22]i, |p+ , p~; [12]i, |p+ , p~; [21]i (12.41)
The states in the [ij] sector are constructed in the usual way with creation operators from
the associated sectors. Earlier we noted that the fields associated with the [11] and [22] sectors
live on the world-volume of the D-brane. Interestingly, the fields corresponding to the [12] string
states live on both D-branes. This really means that the fields exist on a p + 1 dimensional space
(not necessarily one of the D-branes) and have non-local interactions exemplifying the separation
between the two D-branes. More work is being done on this topic via a branch of mathematics
called noncommutative geometry.
The simplest states are the ground states, |p+ , p~; [12]i with
2
x̄a2 − x̄a1

2 1
M =− 0 + (12.42)
α 2πα0
The separation of the D-branes goes√to zero, these states are easily recognized as the tachyon
states. Notice that if |x̄a2 − x̄a1 | = 2π α0 , the ground states represent a massless scalar field.
For larger separations, the ground states represent massive scalar fields. The next excited states
have one creation operator acting on them. The creation operator associated with the normal
coordinates, aa∗ a∗ +
1 yields a the excited state a1 |p , p~; [12]i which yields
2
x̄a2 − x̄a1

2
M = (12.43)
2πα0
12.4. MULTIPLE D-BRANES 265
For any momenta, these are (d−p) massive states corresponding to the (d−p) massive scalar fields.
Alternatively, from the tangential coordinates we have the excited state ai∗ +
1 |p , p
~; [12]i yielding the
same mass-squared operator as above. For any momenta, this are the (p + 1 − 2) = p − 1 massive
states. We might assume that that these states associate themselve with massive Maxwell gauge
fields but this would be incorrect. It turns out that a massive gauge field has more degrees of
freedom than a massless gauge field, and in fact has one more state for each momenta than a
massless gauge field.
On another note, as the separation between the D-branes go to zero, though coincident, they
are still distinguishable. In fact where the two branes coincide there is a massless gauge field
for each sector. The gauge fields end up interacting with each other by a process of the strings
joining endpoints. Theories of interacting gauge fields are called Yang-Mills theories. On the world-
volume of the two coincident D-branes we indeed obtain a U (2) Yang-Mills theory with additional
interactions that become negligible at low energies [64].
12.4 Multiple D-Branes

Suppose we have N Dp-branes. The sectors will be labeled by pairs [ij] where i and j are integers
running from 1...N . A simple counting argument leads one to conclude that there are N 2 sectors.
With this configuration string interactions can be visualized rather simply. In a typical process, an
open string joins with a second open string to form a new open string. That is, the end of the first
string at σ = π joins the beginning of the second string at σ = 0. If the open strings are stretched
between the D-brane, the first string from the [ij] sector can be joined to a second open string from
a [jk] sector to give a produce an open string in the [ik] sector. The result is a single string which
does remain attached to the j D-brane since the joining point is no longer an endpoint. The new
string belongs to the [ik] sector. To summarize this possible interaction we may write
[ij] ∗ [jk] = [ik] (12.44)

If N Dp-branes are coincident, we have N 2 interacting massless gauge fields defining a U (N )
Yang-Mills theory on the world-volume of the N coincident D-Branes. We will go in more detail
with this later on but the U here stands for unitary as U (N ) is a unitary group. The group U (1) is
associated with the U (1) field which describes Maxwell theory. The electroweak theory is described
by a U (2) Yang-Mills theory. Though we won’t discuss it here, models of intersecting D-branes can
be used to construct models of particle physics.
Here the labels i, j label the D-branes and are sometimes called Chan-Paton indices. This is
because historically string theory was invented to describe the interactions of hadrons and the
Chan-Paton indices were introduced obtain Yang-Mills theories. In string theory, these labels are
now recognized as the labels of the D-branes that the string is stretched between.
There is another possible string interaction we haven’t considered. Open strings might interact in
such a way that they become closed strings. But closed strings don’t have endpoints and therefore
cannot attach themselves to D-branes. In other words, since closed strings do not attach to D-
branes, they are free to move in space-time.On a similar note, one of the approaches of compactifying
extra spatial dimensions to attempt to coincide with out 4-dimensional perspective of the universe
is that our universe is actually one of these D-branes in which the gauge fields, such as the Maxwell
field, we observe in our every day lives exist on the world-volume of our D-brane.In short, we, on
the physics describing our universe, live on a D-brane. Since closed strings are not attached, they
are free to escape the D-brane and are free to move in space-time, possibly interacting with other
strings attached to other D-branes.
This brings up an interesting suggestion that has become popularized by individuals such as
Brian Greene and Michio Kaku. Out of the four forces, gravity on the shorter scales is by far the
weakest. In string theory, gravitons are represented by closed strings, and therefore do not attach
to D-branes. One speculative answer to why gravity is so weak is because our universe, some D-
brane, only experiences portions of the gravitational field. The entire gravitational field does not sit
on the world-volume of the D-brane since the gravitons do not attach themselves to the D-brane.
An even more interesting, border-line science fiction, notion is that the entirety of space-time is
composed of a collection of these D-branes, each one a universe like ours. This entire collection
yields a multiverse so to speak. What’s more is since closed strings do not attach themselves to the
D-branes, we might in fact be able to communicate with these parallel universes using the closed
strings of string theory. That is, gravitons, for instance, might be sent to interact with open strings
on other D-branes, thereby “contacting” a parallel universe. As mentioned before, most of this
is pure speculation and fun to consider, nonetheless, it is indeed true that closed strings traverse
space-time, interacting anything in its path, including other D-branes.
12.5 Strings Between D-branes of Different Dimension

We have already studied strings stretched between D-branes of the same dimensionality: parallel
Dp-branes. Here we will quantize the string that is attached to parallel Dp-branes and Dq-branes,
we 1 ≤ q < p ≤ 25. The branes are coincident where the world-volume of the Dq-brane is a subset
of the world-volume of a Dp-brane. We say that the branes are parallel in the sense that if the Dp-
brane and Dq-brane are separated, there is a p-dimensional hyperplpane parallel to the Dp-brane
that contains the Dq-brane. With this set up, its easy to see that some of the coordinates in this
configuration will be mixed. Let’s separate the coordinates as
{x0 , ...xq } {xq+1 , ...xp } {xp+1 , ...xd } (12.45)

where the first set corresponds to the common tangential coordinates, the second set is the set
of mixed coordinates, and the third set is the set of common normal coordinates.
Let’s consider strings that are stretched from the Dp-brane to the Dq-brane. The N N coordinates
are easily recognized to be the common tangential coordinates, while the DD coordinates are the
common normal coordinates. The mixed coordinates can be either tangential or normal and we
therefore label them as DN or N D. The string coordinates are analogous to the coordinates given
in (12.44) however correspond to the string. In the light-cone gauge, we use three indices to label
the string coordinates. Namely,
X + , X − , {X i } {X r } {X a } (12.46)
with i = 2, ...q; r = q + 1, ...p; a = p + 1, ...d. From here it’s easy to see that the first set given
in (12.45) corresponds to the N N coordinates; the second set, N D coordinates, and the third set,
DD coordinates. In our analysis we will view the Dp-brane as the brane where an open string
starts, andthe Dq-brane will be where the string ends. The position of the Dp-brane is given by
x̄a1 , while the position of the Dq-brane is given by x̄r2 and x̄a2 . From here we find that the boundary
conditions associated with the ND coordinates are
12.5. STRINGS BETWEEN D-BRANES OF DIFFERENT DIMENSION 267
∂X r

r
= x̄r2

(τ, σ) =0 X (τ, σ) (12.47)
∂σ σ=0 σ=π
Just as before, the N D coordinates satisfy the wave equation in which we arrive to the general
solution
1 r
X r (τ, σ) = (f (τ + σ) + g r (τ − σ)) (12.48)
2
The boundary condition at σ = 0 yields
0 0
f r (u) = g r (u) → g r (u) = f r (u) + cr0 (12.49)
Keeping in mind the second boundary condition, X r = x̄r2 we choose the integration constant to
be cr0 = 2x̄r2 . Altogether we have
1 r
X r (τ, σ) = x̄r2 + (f (τ + σ) + f r (τ − σ)) (12.50)
2
From here we see that the boundary condition at σ = π gives us f r (u+2π) = −f r (u). Therefore,
the function f r (u) goes to minus itself when its argument increases by 2π. In order to acheive this
sign change, we need to use exponentials of the form exp(iku) with k as a half-integer. Or, using
the correct trigonometric functions, we find
X nu nu
f r (u) = fnr cos( ) + hrn sin( ) (12.51)
+
2 2
n∈Zodd
Substituting this back into our expression for the N D string coordinates we find
X nτ nτ nσ
X r (τ, σ) = x̄r2 + Arn cos( ) + Bnr sin( ) cos( ) (12.52)
+
2 2 2
n∈Zodd
To quantize the string we define oscillators with half-integer moding, which leads us to [64]:
√ X 2 inτ nσ
X r (τ, σ) = x̄r2 + i 2α0 αrn e− 2 cos( ) (12.53)
n 2 2
n∈Zodd
As usual, the x̄r2 are constants are therefore not promoted to becoming operators upon quanti-
zation. The derivatives of the N D string coordinates can be combined yielding
√ X in
Ẋ r ± X 0r = 2α0 αrn2 e− 2 (τ ±σ) (12.54)
n∈Zodd
It can be shown that the commutation relation for the string coordinates take the form
[X r (τ, σ), Ẋ s (τ, σ 0 )] = 2πα0 iδ(σ − σ 0 )δ rs (12.55)

which implies a commutator we have seen before:
d
[(Ẋ r ± X 0r )(τ, σ), (Ẋ s ± X 0s )(τ, σ 0 )] = ±4πα0 iη rs δ(σ − σ 0 ) (12.56)
dσ
Using the commutator expansion we obtain

X im0
(τ +σ) − in 0 0 h i d
e− 2 2 (τ +σ )
e αrm0 , αsn0 = 2πiη rs δ(σ − σ 0 ) (12.57)
2 2 dσ
m0 ,n0 ∈Zodd
To extract the commutators we apply the following to both sides of (12.57)

Z 2π Z 2π
1 im 1 in 0
dσe 2 σ · dσ 0 e 2 σ
2π 0 2π 0
where m, n ∈ Zodd . Moreover, recall that

Z 2π
ik ik0
e 2 σe 2 σ
=0
0
with k, k 0 ∈ Zodd and k + k 0 6= 0 since k + k 0 is an even integer. The left hand side becomes
Z 2π Z 2π
X 1 im im0
σ − im
0 1 in 0 in0 0
σ 0 − in
h i
dσe 2 σ e− 2 e 2 τ · dσ 0 e 2 σ e− 2 e 2 τ αrm0 , αsn0
2π 0 2π 0 2 2
m0 ,n0 ∈Zodd
h i
= e−i(m+n)τ αrm2 , αsn2
The right hand side is just

Z 2π Z 2π
im d 1 in 0
iη rs dσe 2 σ dσ 0 e 2 σ δ(σ − σ 0 )
0 dσ 2π 0
Z 2π
n 1 i n m
= − η rs dσe 2 (m+n)σ = − η rs δm+n,0 = η rs δm+n,0
2 2π 0 2 2
Hence, we have
h i m
e−i(m+n)τ αrm2 , αsn2 = δ rs δm+n,0
2
since δ rs rs
= η . Using δm+n,0 to our advantage we find
h i m
αrm2 , αsn2 = δ rs δm+n,0 (12.58)
2
Let’s now calculate the mass-squared operator. Just as the mas operator for the Dp-brane we
first looked at, the present mass-squared operator includes contributions from the N N and DD
coordinates. What’s more, since we are dealing with mixed coordinates, this mass-squared operator
includes contributions from the N D coordinates as well. The expression for 2p+ p− takes a form
similar to before
 
∞
1 1 X X
2p+ p− = 0 α0 pi pi + α0a α0a + i
a
αna ) + r
α− r
m α m + a (12.59)
α 2 n=1 +
2 2
m∈Zodd
12.5. STRINGS BETWEEN D-BRANES OF DIFFERENT DIMENSION 269
where we have written in an ordering constant a. With the addition of the N D coordinates
we will no longer have a 6= −1. We must still normal orderour expression. This can be done by
rearranging the last sum above as
1X r X 1 Xh r i
α− m2 αrm2 = r
α− r
m αm + α r
m,α m
−2
2 +
2 2 2 + 2
Zodd m∈Zodd Zodd
Since we have (p − q) ND coordinates and making use of the commutation relations we may write
the above as
X
r r 1 X
= α− m αm + (p − q) m
2 2 4
m∈Z+
odd
+
m∈Zodd
To evalute this result notice that the we can find the sum of all integers using
∞
X X X X ∞
X
m= m+ m= m+2 m
m=1 m∈Z+ m∈Z+
even m∈Z+ m=1
odd odd
Therefore,
∞
X X 1
m=− m=
m=1
12
m∈Z+
odd
where we used the zeta function to determine ζ(−1). Putting all of this together, we are able to
conclude that the ordering constant is just
1 X h r r
i 1 X 1
α m2 , α− m = (p − q) m= (p − q) (12.60)
2 +
2 4 +
48
m∈Zodd m∈Zodd
1
This shows that each ND coordinate contributes a a factor of 48 to the ordering constant. The
total ordering constant is given by the above along with the (24 − (p − q)) coordinates that are
either N N orDD. Since the ordering constant for N N and DD coordinates is the same ordering
1
constant we had come up with in our analysis for open string theory with a = − 24 , we find that
the total ordering constant a is just
1 1 1
a=− (24 − (p − q)) + (p − q) = −1 + (p − q) (12.61)
24 48 16
With the ordering constant in hand, we can now ascertain the mass-squared operator:
2
x̄a2 − x̄a1

1 1
M2 = + N⊥ − 1 + (p − q) (12.62)
2πα0 α0 16
where we have defined the number operator N ⊥ to be
q
∞ X p ∞ ∞
X X X k r∗ r X X
N⊥ = nai∗ i
n an + ak ak + maa∗ a
m am (12.63)
n=1 i=2 r=q+1
2 2 2
m=1 a=p+1
k∈Z+
odd
Briefly, let’s examine the state space and the associated fields of the two lowest mass levels. The
ground states are labeled using the notation we develope in the previous section, however, since
we are only considering two parallel D-branes, we denote them as |p+ , p~; [12]i, p~ = (p2 , ...pq ). For
all intensive purposes, the fields associated with these states can be viewed as living on the world-
volume of the Dq-brane, though strictly speaking the fields live on a q + 1 dimensional space-time.
We can construct a general state by letting three types of creation operators act on the ground
states: ai∗ r∗ a∗
p , a k and am . In full generality, states take the form
2
" p
∞ Y
# p
"
∞ d
#
r∗ λ k
Y Y Y Y Y
(ai∗
n)
λn,i 
(a k ) 2
,r 
(aa∗
m)
λm,a
|p+ , p~; [12]i (12.64)
2
n=1 i=2 k∈Z+ r=q+1 m=1 a=p+1
odd
The ground states have N ⊥ = 0, corresponding to a single scalar field on the Dq-brane. In
general this scalar field is massive howver it can be tachyonic depending on the separation of the
branes and the difference in dimensionality of the D-branes (p − q).The next excited states come
from one creation operator acting on the ground states with N ⊥ = 12 . All other states are massive
since N ⊥ ≥ 1 along with our assumption p > q, M 2 > 0.
12.6 String Charge and D-Brane Charges

As we well know, a point particle can carry electric charge. When we first learn that particles
can be charged we take the classical approach and presume that certain particles have an inherent
property known as charge and charged particles interact with other charged particles via Coulomb’s
force. However, the more elegant, and perhaps more accurate, field description of a charged particle
is that there is an interaction such that the particle can couple to the Maxwell field, gaining electric
charge. Hence, a particle attains charge upon coupling with a particular gauge field, in this case
the Maxwell field. The interacting field picture does more than describe how particles gain charge.
Take the most famous gauge field for instance, the Higg’s field. Without any details, as the reader
is most likely aware, particles gain mass an interaction such that the particles couple to the Higg’s
field, as opposed to the classical perspective which is that mass is an inherent property of the
particle. All in all, when studying fiel theory, we realize the charge, and mass, and other properties
that were presumed to be innate properties of a particle can be more accurately interpreted as an
interaction which allows these particles to couple to a specific gauge field.
Now, recall that the world-line of the particle is one-dimensional, in which the trajectory has a
µ
tangent vector given by dxdτ(τ ) where we use τ to parameterize the world-line of the particle. Then,
denoting the Maxwell gauge field as Aµ one can show that the interaction for a point particle of
charge q takes the form [64]
dxµ (τ )
Z
q Aµ (x(τ )) dτ (12.65)
dτ
As it turns out, the endpoints of strings carry electric charge from the associated Maxwell field.
This is not too surprising as if point particles can gain electric, the endpoints of strings should
also be able to gain charge. However, though these endpoints carry electric charge, the string itself
is a fundamentally different object than a point particle and therefore carries a different kind of
charge. To see this, let’s construct the interaction for a string carrying charge. At any point along
12.6. STRING CHARGE AND D-BRANE CHARGES 271
the string trajectory there are two linearly independent tangent vectors, which we have previously
µ
∂X µ
chosen to be ∂X∂τ and ∂σ . But what is the correct field corresponding to string charge? In closed
string theory we saw that some of the strings can be physically interpreted as states describing the
antisymmetric Kalb-Ramond field Bµν , so let’s use that in place of the Maxwell gauge field Aµ .
With the two tangent vectors and the Kalb-Ramond field, we can construct
∂X µ ∂X ν
Z
− dτ dσ Bµν (X(τ, σ)) (12.66)
∂τ ∂σ
which, analogous to our expression describing the interaction of charged point particle, the above
describes how the string couples to the anti-symmetric Kalb-Ramond field. Indeed, we can say that
integral is the electric coupling as it is the natural generalization of the point particle coupling to
the Maxwell field. Therefore, we are allowed to say that the string carries electric Kalb-Ramond
charge. There is more work one must to prove that the the endpoints of strings carry charge and
the dynamics of the string itself. We won’t go into those details here however point the reader to
Zwiebach’s text which does an excellent job in covering this material in depth.
Based on this analysis, and keeping in the mind the topic of this chapter, a natural question
arises to whether the other higher dimensional objects in string theory can carry charge. That is
can the Dp-branes of string theory carry charge?
The short of it is yes, Dp-branes do carry charge, however it depends on the string theory one
is studying. In bosonic string theory, the theory we have been studying so far, this is not the case,
Dp-branes do not carry charge. On the other hand, as we will discuss in a moment, type IIA and
type IIB superstring theories indeed have D-branes that are charged. Before we examine these
details, let’s first come up with the analogous expression describing the coupling of the D-brane.
The world-volume of a Dp-brane is parameterized by our time coordinate τ and a set of other co-
ordinates σ 1 , ...σ p , the natural generalization of the parameterization of the one dimensional string.
Thus, the space-time coordinates that describe the position of the Dp-brane are just X µ (τ, σ 1 , ...σ p ).
Moreover, presuming we have the general antisymmetric tensor field Aµ,µ1 ,...µp , the coupling is just
a generalized version of (12.66):
∂X µ ∂X µ1 ∂X µp
Z
− dτ dσ1 ...dσp ... Aµ,µ1 ,...µp (X(τ, σ 1 , ...σ p )) (12.67)
∂τ ∂σ 1 ∂σ p
So far in our studies we have only considered bosonic string theory in which we had the Kalb-
Ramond field is the only massless antisymmetric tensor field that we know of, which is the field
describing the electric charge of the string, and therefore cannot be the field we seek to describe
the charge the Dp-brane. In the absence of any other massless antisymmetric tensor fields we are
left to conclude that the Dp-branes from bosonic string theory do not couple to any other massless
antisymmetric tensor fields and therefore do not carry charge.
But not all is lost. In type IIA and type IIB closed superstring theories there additional anti-
symmetric fields. The additional fields associated with type IIA superstring theory are Aµ , Aµνρ ,
while the additonal fields associated with type IIB superstring theory are A, Aµν , Aµνρσ (Zwiebach,
371). It turns out these are the gauge fields which couple electrically to the D-branes corresponding
to each theory. In type IIA superstring theory, the Aµ couples to the D0 branes, yielding charged
supersymmetric point particles; and Aµνρ couples to the D2 branes. In type IIB superstring theory
Aµν ends up coupling to the D1 branes while the gauge field Aµνρσ couples with the D3 branes. The
field A is a strange one. This gauge ends up coupling electrically with to the D(-1)-brane, known
as the D-instanton. It is called an instanton because it is an object forever fixed in time, hence no
space-time flows for the instanton. All together than, the charged D-branes are the D0-branes and
D2-branes of type IIA superstring theory and the D1-branes and D3-branes of type IIB superstring
theory.
Charge and energy conservation together imply that the charged object cannot decay if there are
no lesser mass objects in which the charged object can decay in to. Here, the D-branes described in
type IIA and type IIB are in fact in stable. As we will see shortly the D-branes from bosonic string
theory, which are not charged, actually decay. Moreover, it is known that in type IIA superstring
theory the only stable Dp-branes are those with even spatial dimension but are unstable in type
IIB superstring theory. Alternatively, the only stable Dp-branes in type IIB superstring are those
of odd dimension but are unstable in type IIA superstring theory. Lastly, we have only mentioned
a few D-branes, what of the D-branes of other dimensions? We won’t go into detail here in this
text, but the D4, D6, and D8 branes from type IIA superstring, and the D5, D7, and D9 branes of
type IIB superstring turn out to have magnetic charge.
The electric charge of a Dp-brane actually has a simple description when p-spatial dimensions
are curled up into circles and the Dp-brane is then wrapped around this compact space. This
would then mean that the directions of the p compact space lie along the D-brane while the other
space-time directions define an effective lower dimensional space-time normal to the D-brane. If
the D-brane is wrapped around these compact dimensions, as far as an observer living in the lower
dimensional space-time is concerned, the D-brane appears as a point particle. But we have already
seen that a point particle carries a charge associated with a Maxwell field originating from the
antisymmetric tensor field Aµµ1 ...µp .
Let x1 , ...xp denote the p compact spatial dimensions and let X 1 , ...X p denote the corresponding
brane coordinates. Since we have chosen the compact directions to be circles, we associate with
each compact dimension radii R1 , ...Rp . Then, to represent the Dp-brane being wrapped around
the compact space, we can use the brane coordinates
X k (τ, σ 1 , ...σ p ) = Rk σ k (12.68)

with σ k ∈ [0, 2π] and k = 1, ...p. Based on the range of σ k , it’s easy to see that as σ k moves
from 0 to 2π we have X k moving from 0 to 2πX k , indicating that it traverses once around a
circle (the kth circle to be exact). Moreover, let X m (τ, σ 1 , ...σ p ) = xm (τ ) denote the non-compact
dimensions. Comparing to our earlier expression characteristic to a point particle, we notice that
this statement really means that one living in the lower-dimensional space-time views the D-brane
as a point particle. Using this notation we can apply it our integral expression describing the
electric coupling . Since the X k only depend on σ k , the only non-zero contributions to (12.67) are
those with µk = k, k = 1, ...p. In other words, (12.67) changes to
∂X m 1 2
Z
− dτ dσ1 ...dσp R R ...Rp Am12...p (X(τ, σ 1 , ...σ p )) (12.69)
∂τ
where we have note µ = m since µ can only take values of the non-compact directions as all of the
compact indices have been used. Now, let us focus on the field that is independent of the compact
spatial coordinates, i.e. is only concerned with the non-compact coordinates, Am12...p (xm (τ )). The
above integral becomes
dxm
Z
−R1 R2 ...Rp dτ dσ1 ...dσp Am12...p (x(τ )) (12.70)
dτ
12.7. TACHYONS AND D-BRANE DECAY 273
The σ integrals contribute a total factor of (2π)p . Using the radii of the circles of the compact
dimensions, we obtain the volume of the compact space Vp = (2πR1 )...(2πRp ). Moreover, let’s
define
1
p Ãm (x(τ )) ≡ Am12...p (x(τ )) (12.71)
α0 2
where we have introduce the slope parameter α0 to give the gauge field Ām the expected dimension
of mass. We claim that Ãm is the Maxwell field arising from the antisymmetric tensor field Aµ12...p
through dimensional reduction. Using this definition, the electric coupling becomes
dxm
Z Z
Vp Vp
− 0p dτ Ãm (x(τ )) = − p Ãm dxm (12.72)
α2 dτ `s
√
where we used the convention `s = α0 . Comparing to (11.65), we easily recognize the integral
above as the coupling a point particle to a Maxwell field Ãm . Therefore, the Dp-brane, in the
compact space we have built, appears as a charged point particle in which the Maxwell charge of
the brane is
Vp
Q= (12.73)
`ps
12.7 Tachyons and D-brane Decay

In the last section we discussed that the D-branes in bosonic string theory are not charged. We
also noted that the charged D-branes of type IIA and type IIB superstring theories have stable D-
branes partly because they are charged. As a consequence we might link the fact that the D-branes
from bosonic string theory are unstable because they are not charged. Part of this is true but the
instability of the D-branes of bosonic string theory arise from the tachyons of bosonic open string
theory. To see this, first consider the action for a massive scalar field
Z
1
S = dD x(∂µ φ∂ µ − λφ2 ) (12.74)
2
When looking at scalar fields, we identify the mass of the scalar field by examining the terms in
the scalar field potential that are quadratic in the field. Hence, in the present case we identify the
mass term as
λ = m2
We make this identification by comparing it to the Klein-Gordon Lagrangian
1 1
L= (∂µ )2 − m2 φ2 (12.75)
2 2
where we see that the mass term is associated with the term quadratic in the field φ. Notice that
the quadratic terms indicate a harmonic potential. Using this we can see how a tachyon represents
an instability of the vacuum. The potential of the scalar field is just V (φ) = 12 m2 φ2 . Notice then
that if m2 > 0, we see that V (φ) is just an upward parabola with a minimum at φ = 0. On
the contrary, if m2 < 0,the case we have with tachyons, the parabola opens downward with the
maximum being located at φ = 0, which is an unstable point. Illustrations of these potentials are
given in figure 12.2.
Figure 12.2: The potential of scalar field indicating its point of instability.
We can also expand the potential energy V (φ) about its critical points φ∗ telling us where the
maxima and minima are to determine its behavior. Using the potential given in the action (12.74),
we notice that the critical point occurs at φ∗ = 0. Hence, using a Taylor series approximation we
find that the potential to second order is
1
V (φ) = V (φ∗ ) + V 0 (φ∗ )(φ − φ∗ ) + V 00 (φ∗ )(φ − φ∗ )2 + ... = V (φ∗ ) + λ(φ − φ∗ )2 + ... (12.76)
2
The second term in our expression is quadratic in the fields and so we recognize it as our mass
term. In the case of a D-brane from bosonic string theory, it turns out the leading term of the
expansion of the potential is given by the tension T . In fact, if a tachyon lives on the D-brane, then
the field potential looks something like [37]
1 2
V (φ) = T − φ + ... (12.77)
2α0
Therefore, if the potential strays away from φ = 0, this shows that the D-brane is losing energy.
What ends up happening is the D-brane decays away into closed string states. In that sense, the
D-branes of bosonic string are composed of closed string states, even tachyons. One could argue
that certainly the D-branes of bosonic string decay and are unstable, but who cares anyway since
we well know that bosonic string theory is an unrealistic theory.
Indeed, in superstring theory there are stable D-branes. However in superstring theory, as
there are charged D-branes, there also exist oppositely charged anti-D-branes which can become
coincident with a D-brane. Just like the interactions of particles and anti-particles, branes and
anti-branes can annihilate one another when they coincide. This ends up happening because there
are tachyons stretched between them. To see see this, consider the simple tachyon potential for a
D1-brane coincident with an anit-D1-brane
λ 2
V (φ) = (φ − φ20 )2 (12.78)
2
12.8. EXERCISES 275
where φ0 is some constant. Let’s first look for the critical points of the potential. The first
derivative yields
dV (φ)
= 2λ(φ2 − φ20 )φ
dφ
Setting this equal to zero, we find that the critical points are just φ∗ = 0, ±φ0 . The second
derivative is just
V 00 (φ) = 2λ(φ2 − φ20 ) + 4λφ2

Expanding the potential to second order about the critical point φ∗ = 0, we find
λφ40
V = − λφ20 φ2 + ...
2
The mass term is then identified to be
m2 = −2λφ20 (12.79)
which is imaginary mass, the mass of a tachyon. Therefore, we may conclude that the critical
point φ∗ = 0 corresponds to a tachyon. In summary, a D-brane coincident with an anti-D-brane
contains a superstring tachyon and therefore represents an instability in the D-brane/anti-D-brane
pair. Do to the importance of D-brane/anti-D-brane collisions in string cosmology, the issues of
the instability of the D-branes of bosonic string theory are quite relevant.
12.8 Exercises
1. Argue for (12.15), then derive (12.16). Using oscillators, go through the details that yield (12.18)
and (12.19).
2. (a) Using (12.51), derive (12.52), and then (12.53).
(b) Prove (12.57).
3. Consider the following potential:

2
φ2

1 2
V (φ) = φ − 1
8α0 0 φ20
(a) Plot V (φ) as a function of φ and find its critical points.
(b) Using the Klein-Gordon Lagrangian, at each critical point φ̃ expand the Lagrangian for
fluctuations of φ around this point, i.e. let φ = φ̃ + for small . Determine the mass term.
Note: this potential is a crude model for the superstring tachyon potential on the world-volume of
a D-brane and a coincident anti-D-brane.
Chapter 13
T-Duality, Symmetries, and

Compactification
In this chapter we focus on some of the most intriguing aspects of string theory. We explore some
methods of compactifying the extra dimensions string theory requires and examine the phyiscal
consequences of spatial compactifications. One of the most interesting consequences is that of T-
Duality, which reveals that the physics of two seemingly different physical systems actually yield
the same physical result. We will examine T-duality in detail in this chapter, which will naturally
lead us to other, more elegant methods of compactification and string geometry. Due to the level of
this text, this more elegant methods of compactification are examined rather qualitatively, however
will allow us to discuss string theory models of particle physics and the string theory landscape,
both of which are topics that are being actively researched today.
We begin by slowly motivating the definition of T-duality using closed strings, and then move
on to looking at T-duality applied to open strings attached to Dp-branes. Before we get into those
details however, let’s first discuss the notion of duality symmetries in the context of classical physics.
Duality symmetries are symmetries that relate two systems with entirely different physical de-
scriptions yet maintain identical physics. To see this, first recall Maxwell’s equations in the absence
of sources
~
1 ∂E
~ =0
∇·E ~ =
∇×B (13.1)
c ∂t
~
~ =0
∇·B ∇×E ~ = − 1 ∂B (13.2)
c ∂t
We immediately notice that Maxwell’s equations are invariant under the duality transformation
~ B)
(E, ~ → (−B,
~ E).
~ Notice however that the Lagrangian density is not invariant under this change.
By expanding out the Lagrangian density in terms of the electric field and magnetic field we find
1 1 1 1
L = − Fµν F µν = − (2F0k F 0k + Fij F ij ) = − (−F0k F0k + Fij Fij )
4 4 2 2
1 2 2
= (E − B )
2
277
278 CHAPTER 13. T-DUALITY, SYMMETRIES, AND COMPACTIFICATION
If we make the above duality transformation we notice that L → −L indicating that the La-
grangian density in not invariant under the duality transformation. This is because the Lagrangian
is an object which is defined as the kinetic energy minus the potential energy. The duality symmetry
here essentially exchanges these types of energy, which does not leave the Lagrangian invariant. On
the other hand, both kinetic and potential energies enter the Hamiltonian in the same way (both
types of energies are summed together). For this reason, duality symmetries are typically exhibited
using the Hamiltonian rather than the Lagrangian. In the present case, th Hamiltonian density H
is proportional to (E 2 + B 2 ), and by applying the duality symmetry we see that the Hamiltonian
density, and hence the Hamiltonian remains invariant under the specific duality transformation.
A second, more interesting example to consider is the simple harmonic oscillator from classical
mechanics. Here the oscillator consists of a mass m attached to a spring with a characteristic spring
constant k. The Hamiltonian is simply
p2 1
H(m, k) = + kx2 (13.3)
m 2
q
k
The oscillator has an angular frequency of ω = m , which motivates the duality transformation
(m, k) → ( k1 , m
1
). With this duality transformation we see that the new Hamiltonian becomes
1 1 1 1 2
H( , ) = kp2 + x (13.4)
k m 2 2m
revealing that the Hamiltonian is not invariant. It’s similar, but not quite the same. This is
fine since, as we have learned from our studies of quantum mechanics, the actual physics of the
systems lies within the commutation relations of the dynamical variables. Certainly the form of
the Hamiltonian changed, but the dynamical variables continue to yield the expected commutation
relations, i.e. the expected physics. To exhibit this connection between the Hamiltonians we
use canonical transformations. Essentially, these transformations are equivalent to changing the
canonical variables in such a way that all of the commutation relations are preserved, rather, that
the physics is preserved. For instance, consider the canonical transformation that takes x → p and
p → −x. Indeed this is a canonical transformation since we still have
[x, p] → [p, −x] = −(−i) = i
yielding the expected commutation relations. Under this canonical transformation, as we see
that the dual transformed Hamiltonian becomes
1 1 1 1 2
H( , ) → k(−x)2 + p = H(m, k)
k m 2 2m
Therefore, the two Hamiltonians, under the duality transformation is observed to be the same
via our canonical transformation. That is, the Hamiltonian with the dual parameters ( k1 , m1
) is
canonically equivalent to the original Hamiltonian describing a simple oscillating spring system,
thereby exemplifying that the underlying physics of the two systems is unchanged by the duality
transformation (m, k) → ( k1 , m
1
)
13.1. QUANTIZING THE CLOSED STRING ON A COMPACTIFIED SPACE 279
13.1 Quantizing the Closed String on a Compactified Space

Before we move on to exploring T-duality of closed string theory, it is imperative to first understand
the effect of compactifying one spatial dimension has on closed strings. Up to this point we have
been assuming our strings live in a background Minkowskian space-time. When we compactify
spatial dimensions, as we will see shortly, closed strings can actually “wrap” themselves around
these compact dimensions.
To ease ourselves into this notion, let’s begin by imagining a world with two spatial dimensions
(x, y), one of which has been compactified into a circle. This world can be thought of as an infinitely
long cylinder. We will let the spatial coordinate x be compactified via the usual identification
x ∼ x + 2πR, where R is the radius of the circle x has been compactified into. We also let y denote
the coordinate that extends along the length of the cylinder. Now consider closed strings that live
on the two dimensional surface of the cylinder. Figure 13.1 illustrates a variety of possible strings
that live on the 2D surface and how they appear in the so-called covering space, the plane in which
the string lives before it has been wrapped around a cylinder (one could imagine taking a piece of
paper, drawing closed strings and wrapping it tightly around a paper towel roll).
The simplest strings are those which do not wrap around the compact dimension, such as string 1.
These strings have zero winding number because they do not wind around the compact dimension.
In this case the string coordinate X in the covering space satisfies
X(τ, σ + 2π) − X(τ, σ) = 0
Now consider the string drawn in the figure that is oriented in the increasing x direction wrapping
around the cylinder just once, string 2. In this case the string cannot be contracted to a point
because it is no longer simply-connected. That is, the string cannot contract to a point without
being cut first. This string has a winding number of +1 since it wraps around the cylinder, the
compact dimension, once. From the figure, we recognize that this string is closed due to our
identification. The string coordinate in the covering space corresponds to
X(τ, σ = 2π) − X(τ, σ = 0) = 2πR
Similarly, string 3 has a winding number −1 since it traverse in the opposite direction to that of
string 2. Moreover, the string coordinate in the covering space satisfies
X(τ, σ = 2π) − X(τ, σ = 0) = −2πR
Continuing with this logic, we can convince ourselves that a string that wraps twice around the
cylinder with identified points on the cylinder has a winding number of +2, and a string coordinate
on the covering space satisfying
X(τ, σ = 2π) − X(τ, σ = 0) = 2(2πR)
To generalize this notion, we say that a string has a winding number m, where m is an integer,
if it wraps around the cylinder, the compact dimension, m times in the direction of positive x. The
string coordinates in the covering space then satisfy
X(τ, σ + 2π) = X(τ, σ) + m(2πR) (13.5)

Figure 13.1: Strings wrapped around a compactified dimension, living on the surface of a two
dimensional surface. To the right are the same strings as shown in the “covering” space of the
cylinder.
It is important to note that strings with different winding numbers cannot be continuously
deformed into each other. A mathematician would say that strings of different winding numbers
are not homeomorphic to one another. Therefore, we call the winding number of a closed string a
topological property.
Looking ahead, we define the winding w in terms of the winding number m and the radius of
the compactified space
mR
w≡ (13.6)
α0
Notice that the winding has units of inverse length, or momentum. It will turn out that the
winding can interpreted as a type of momentum. Lastly, in terms of the winding, we may write
(13.5) as
X(τ, σ + 2π) = X(τ, σ) + 2πα0 w (13.7)
Now that we have a fair understanding of the effects compactifying spatial dimensions has on
closed strings, we are ready to begin quantizing the closed string on a compactified space. As usual,
let’s consider closed bosonic strings. As we well know, these strings propagate in 26-dimensional
space-time that has space-time coordinates x0 , ...x25 . Let’s assume that the x25 space-time direction
is curled up into a circle of radius R. Since it is the easiest to work with, we will use the light-cone
gauge, in which the string coordinates are arranged as X + , X − , {X i } = {X 2 , ...X 24 }, X 25 = X. All
that has been done here is we have isolated the string coordinate corresponding to the compactified
dimension from the rest of the transverse light-cone coordinates.
The perodicity condition for X is simply given by (13.6). Just as the other string coordinates do,
X satisfies the wave equation, in which we can write the general solution in terms of left and right
movers (as we had done when we examined the mode expansions of the closed string). Therefore
X(τ, σ) = XL (τ + σ) + XR (τ − σ) = XL (u) + XR (v) (13.8)

where u = τ + σ and v = τ − σ. Applying the periodicity condition (13.6) we find
XL (u + 2π) + XR (v − 2π) = XL (u) + XR (v) + 2πα0 w

yielding
XL (u + 2π) − XL (u) = XR (v) − XR (v − 2π) + 2πα0 w (13.9)

To continue, we can use some of the details of our original analysis of the closed string to help
us determine the mode expansions for the closed string with winding. The derivatives XL0 (u) and
0
XR (v) are still periodic functions, and hence the same expansions as before still hold:
r
0 α0 X µ −inu
XLµ (u) = α̃n e (13.10)
2
n∈Z
r
µ 0 α0 X µ −inv
XR (v) = αn e (13.11)
2
n∈Z
We can integrate these expressions just as we had done before to find

r r
1 L α0 α0 X α̃n −inu
XL (u) = x0 + α̃0 u + i e (13.12)
2 2 2 n
n6=0
r r
1 α0 α0 X αn −inv
XR (v) = xR + α0 v + i e (13.13)
2 0 2 2 n
n6=0
Before we had α0 = α̃0 . This time however we see that by the periodicity condition we have
r r
α0 α0
2π α̃0 = 2π α0 + 2πα0 w
2 2
From which we find
√
α̃0 − α0 = 2α0 w (13.14)
We notice then that if the winding vanishes, we attain the familiar result α0 = α̃0 . Moreover,
we can calculate the momentum p of the string along the compact dimension:
Z 2π
1 1
p= (ẊL + ẊR )dσ = √ (α0 + α̃0 ) (13.15)
2πα0 0 2α0
where we used the fact that only terms linear in arguments u and v contribute to the integral.
We notice that (13.14) and (13.15) are different only by a minus sign, from which find
r r
α0 α0
α0 = (p − w) α̃0 = (p + w) (13.16)
2 2
Due to the similarity of the momentum p and the winding w, we are allowed to view the winding
as a type of momentum. Without compactification, as studied previously, we found that the zero
mode oscillators of the right and left moving sectors of closed string theory were in fact equal,
α0 = α̃0 . In this case we also only had one momentum p since there was only one zero mode. In
the present case, where we assume that one of the spatial directions is compactified, we no longer
have that the zero mode oscillators are equivalent, and we have two different momenta, p and the
winding w. This leads us to presume that the coordinates associated with the zero modes are not
equivalent, i.e. xL R L
0 6= x0 . For this reason we choose to rewrite x0 as
xL
0 = x0 + q 0
and similarly,
xR
0 = x0 − q0
where we have defined x0 as the average of the left and right moving zero mode coordinates,
x0 = 21 (xL R
0 + x0 ), and q0 as the difference of the left and right moving zero mode coordinates,
1 L R
q0 = 2 (x0 − x0 ). Using this change we can write the right and left moving string coordinates XL
and XR as
r
1 α0 α0 X α̃n −in(τ +σ)
XL (τ + σ) = (x0 + q0 ) + i (p + w)(τ + σ) + i e (13.17)
2 2 2 n
n6=0
r
1 α0 α0 X αn −in(τ −σ)
XR (τ − σ) = (x0 − q0 ) + i (p − w)(τ − σ) + i e (13.18)
2 2 2 n
n6=0
Adding the right and left movers together we find the string coordinate associated with the
compactified dimension x is
r
0 0 α0 α0 X e−inτ
X(τ, σ) = x0 + α pτ + α wσ + i (p − w)(τ − σ) + i (α̃n e−inσ + αn einσ ) (13.19)
2 2 n
n6=0
For future reference, we write down the usual combination of derivatives of X

√ X
Ẋ + X 0 = 2α0 α̃n e−in(τ +σ) (13.20)
n∈Z
√ X
Ẋ − X 0 = 2α0 αn e−in(τ −σ) (13.21)
n∈Z
With the mode expansions of the string coordinates in hand (all of the other directions take the
usual form given back in our previous analysis of quantizing the closed string), we can proceed with
quantizing the string coordinate X associated with the compactified dimension x.We start by using
the familiar set of commutation relations. In particular, just as we had before, we have
[X(τ, σ), P τ (τ, σ 0 )] = iδ(σ − σ 0 ) (13.22)

Moreover, we also have that
[X µ (τ, σ), X ν (τ, σ 0 )] = [P µτ (τ, σ), P ντ (τ, σ 0 )] = 0 (13.23)

The rest of the quantization procedure follows exactly from our previous analysis where we
quantized the closed string in the light-cone gauge without compactification. For that reason we
only consider the string coordinates, and oscillators associated with the compact direction. When
we quantized the closed string we ended finding the non-trivial commutation relation
h 0 0
i
(Ẋ I ± X I )(τ, σ), (Ẋ J ± X J )(τ, σ 0 ) = ±4πα0 iη IJ δ(σ − σ 0 ) (13.24)
from which we found the commutation relations of the oscillators
I
[α̃m , α̃nJ ] = [αm
I
, αnJ ] = mδm+n,0 η IJ (13.25)
We can still use these commutation relations for our present analysis for the string coordinate X
[αm , αn ] = [α̃m , α̃n ] = mδm+n,0 (13.26)
[αm , α̃n ] = 0 (13.27)

for all m, n ∈ Z. A particularly interesting commutation relation is the one between the zero
mode oscillators of the left and right moving sectors, [α0 , α̃0 ] = 0. Plugging in the expressions for
the zero mode oscillators, this commutation relation becomes
[p, w] = 0 (13.28)
Moreover, since both α0 and α̃0 commute with all of the other oscillators αn , α̃n , we find that p
and w also commute with all of the other oscillators:
[p, αn ] = [p, α̃n ] = [w, αn ] = [w, α̃n ] = 0

We are almost done with the new additions to the commutation relations. All that remains is
determining the commutation relations with the zero mode coordainte x0 with the other operators.
Notice that if we combine (13.22), and the σ 0 derivative of (13.23) we obtain
X(τ, σ), (Ẋ ± X 0 )(τ, σ 0 ) = 2πα0 iδ(σ − σ 0 )

(13.29)
Observing our expansions given in (13.19), (13.20), and (13.21), we notice the terms including p
and w do not contribute since they commute with all of the other oscillators. Using this fact and
integrating over the range σ ∈ [0, 2π], we find that
[x0 , (Ẋ ± X 0 )(τ, σ 0 )] = α0 i (13.30)

It follows then that
r
α0
[x0 , α0 ] = [x0 , α̃0 ] = i (13.31)
2
Upon using (13.16), the above leads to
[x0 , p] = i [x0 , w] = 0 (13.32)
Since we have compactified the x = x25 dimension, the zero coordinate mode x0 is coordinate
living on a circle that is the compactified dimension. As we see from the commutation relations,
the momentum operator p along the x-direction is the momentum conjugate to x0 . Using a result
from ordinary quantum mechanics, this means that the possible values of the momentum p carried
by the string states are quantized [64].
In our present case, consider the operator e−iap which translates the string states along the
x-direction by an amount a. Due to our identification, we find that the operator e−i2πR behaves as
an identity operator, allowing us to conclude that the string states of the theory having momentum
along the x-direction is quantized by
n
p= (13.33)
R
with n ∈ Z. Another quantization condition is the periodicity condition
X(τ, σ + 2π) = X(τ, σ) + m(2πR)

Due to its similarity with the momentum p, we interpret the winding w as an operator that has
eigenvalues corresponding to the various possible windings the closed string may take when it is
wrapped around the compact dimension:
mR
w= (13.34)
α0
with m ∈ Z. Therefore both the momentum p and the winding w have discrete spectra. Before
moving on, let’s briefly remark on the consequences of compactifying one spatial dimension. First
of all, before compactification, there was no winding term, and the momentum operator had a
continous spectrum. Upon compactification, we lost the continuous spectrum, thereby losing some
of the string states we had before. However, with compactification we gained a winding term, and
hence gained states associated with various windings. All in all, we lost some quantum states, and
gained others. Most interestingly, the compactification of the one of the spatial dimensions caused
the momentum p to become quantized, having a discrete spectrum of possible values.
When we quantized the closed string in the light-cone gauge before, we found that L⊥ ⊥
0 − L̃0
⊥
annihilated the states of the theory. This result still holds however its implication, namely N −
Ñ ⊥ = 0 is no longer valid. To see this, recall the transverse Virasoro operators from our analysis
without compactification:
1X I I 1X I I
L̃⊥
n = α̃p α̃n−p L⊥
n = αp αn−p (13.35)
2 2
p∈Z p∈Z
α0 I I α0 I I
L̃⊥
0 = p p + Ñ ⊥ L⊥ 0 = p p + N⊥ (13.36)
4 4
In the present case, the zero mode transverse Virasoro operators change as
1 I I α0 1
L̃⊥
0 = α̃0 α̃0 + Ñ ⊥ = pi pi + α̃0 α̃0 + Ñ ⊥ (13.37)
2 4 2
1 I I α0 1
L⊥
0 = α0 α0 + N ⊥ = pi pi + α0 α0 + N ⊥ (13.38)
2 4 2
Notice then that the difference of the two Virasoro operators yields
1
L⊥ ⊥
0 − L̃0 = (α0 α0 − α̃0 α̃0 ) + N ⊥ − Ñ ⊥ = −α0 pw + N ⊥ − Ñ ⊥ (13.39)
2
It follows then that the constraints on the physical states take the form
N ⊥ − Ñ ⊥ = α0 pw (13.40)
Using our quantization conditions of the momentum and winding, we find
n mR
N ⊥ − Ñ ⊥ = α0 = mn (13.41)
R α0
To gain physical insight into the states of our theory, it is useful to have the mass-squared
operator M 2 . If we take the point of view of an observer living in a 25-dimensional Minkowski
space-time that does not include the compactified dimension we have that
2 ⊥
M 2 − p2 = 2p+ p− − pi pi = (L + L̃⊥ i i
0 − 2) − p p (13.42)
α0 0
where we made use of
√ 1
2α0 α0− ≡ (L⊥ + L̃⊥ 0 −
0 − 2) = α p
p+ 0
Then, if we make use of (13.37) and (13.38) we find
1 2
M2 = 0
(α0 α0 + α̃0 α̃0 ) + 0 (N ⊥ + Ñ ⊥ − 2) (13.43)
α α
Lastly, if we use our expressions for the zero mode oscillators, we are left with the mass-squared
operator being in terms of the winding w and momentum p,
2
M 2 = p2 + w2 + (N ⊥ + Ñ ⊥ − 2) (13.44)
α0
Let’s briefly examine some key features of M 2 . First suppose that M 2 = p2 . This leads to
M = |p|, illustrating that the internal momentum of the string contributes to the rest energy of
the string in the same way as the momentum contributes to the energy of a massless particle. Now
consider the case where M 2 = w2 , in which we have M = |w|. Compare this to a string state that
is wound |m| times around the compact dimension. The length of such a string is just |m|2πR, and
1
since the tension of a string is T = 2πα 0 , the rest energy is equal to
1 |m|R
M= |m|2πR = = |w|
2πα0 α0
We may therefore conclude that the contribution of the winding to the mass is understood as
the energy associated with the stretching required to wrap the string around the compactified
dimension.
Finally, recall that the closed string Hamiltonian is written as
H = α0 p+ p− = L⊥ ⊥
0 + L̃0 − 2 (13.45)
We leave the calculation as an exercise for the reader to show that in the present case the
compactified closed string Hamiltonian becomes
α0 i i
H= (p p + p2 + w2 ) + N ⊥ + Ñ ⊥ − 2 (13.46)
2
13.2 Constructing the State Space of Compactified Closed

Strings
Now that we have the mass-squared operator, let’s construct the state space of the closed string in
the presence of a single compactified dimension. This time we will label the ground states using p+
n
and pi for i = 2...24. Moreover, making note of the fact that p = p25 = R and w = mRα0 , we choose
+
to label the ground states of our theory as |p , p~T ; n, mi.
To find the excited states we act on the ground state using creation operators. A basis state for
the state space is constructed by applying creation operators to the ground states. Therefore, a
general candidate basis state is
" ∞ Y
24
#  ∞ 24 "
∞
#" ∞ #
Y YY Y Y
(ai∗
r )
λi,r 
(ãj∗
s )
λ̃j,s 
(a∗k )λk (ã∗` )λ̃` |p+ , p~T ; n, mi (13.47)
r=1 i=2 s=1 j=2 k=1 `=1
where we separated the creation operators that arise form the compact dimension. The number
operators act on the state above yielding
∞ X
X 24 ∞
X ∞ X
X 24 ∞
X
N⊥ = rλi,r + kλk Ñ ⊥ = sλ̃j,s + `λ̃` (13.48)
r=1 i=2 k=1 s=1 j=2 `=1
It is important to note that the candidate state (13.47) is a member of the string state space if
and only if we have N ⊥ − Ñ ⊥ = mn. The mass-squared of the candidate state is just (13.43)
n 2 2
mR 2
2
M = + + (N ⊥ + Ñ ⊥ − 2) (13.49)
R α0 α0
To gain some physical insight, let’s examine some of the closed string states. First consider
the case where m = n = 0. These states have zero momentum and zero winding in the compact
dimension. From our constraint (13.41) we see that this means N ⊥ = Ñ ⊥ , in which case we must
13.2. CONSTRUCTING THE STATE SPACE OF COMPACTIFIED CLOSED STRINGS 287
match the number of right moving oscillators to the number of left moving oscillators. The vacuum
state is simply
|p+ , p~T ; 0, 0i
with a corresponding mass-squared M 2 = − α40 , indicating this is a closed string tachyon state.
The massless states have N ⊥ = Ñ ⊥ = 1. Since we are dealing with a theory where one spatial
dimension is compactified, both left and right sectors of the theory have two kinds of oscillator:
those that belong to the compact dimension and those that do not. This gives a total of four ways
of combining the oscillators to form the possible massless states
a∗1 ã∗1 |p+ , p~T ; 0, 0i a∗1 ai∗ +

1 |p , p
~T ; 0, 0i (13.50)
∗ + j∗ +
ai∗
1 ã1 |p , p
~T ; 0, 0i ai∗
1 ã1 |p , p
~T ; 0, 0i (13.51)
The first state in (13.50) carries no index and is therefore contains a single state corresponding
to a massless scalar field. The next state in (13.50) and the first state in (13.51) carry a light-cone
index for 25-dimensional space-time, allowing us to identify them as the photon states corresponding
to a Maxwell field. This is incredibly interesting as in the ordinary closed string theory we studied
previously did not yield a single Maxwell field or photon states. Therefore, compactifying a single
spatial dimension allows for closed string theory to include photon states. Finally, the remaining
states given in the (13.51) have the same structure as the massless closed string states of the theory
we constructed in a previous chapter. These states are simply the strings states corresponding to
gravitons, the dilaton, and the states corresponding to a Kalb-Ramond field.
Let’s move on to states with n = 0 or m = 0 but not zero at the same time. That is, let’s
consider states with either momentum p or winding w. Since either n = 0 or m = 0, we still have
N ⊥ = Ñ ⊥ . The ground states are simply
n2 4
|p+ , p~T ; n, 0i M2 = 2
− 0 (13.52)
R α
m2 R2 4
|p+ , p~T ; 0, mi M2 =
− 0 (13.53)
α02 α
Depending on the values of n and m, the above states may be interpretated as tachyons, massless
particles, or massive particles. Acting on these ground states with oscillators yields heavier states
since such states have N ⊥ + Ñ ⊥ ≥ 2.
Finally, let’s consider states with n = m = ±1 or n = −m = ±1. In this case we have that the
strings have momentum and winding along the compact direction and therefore N ⊥ 6= Ñ ⊥ . We will
only consider two cases: N ⊥ − Ñ ⊥ = 1, and N ⊥ − Ñ ⊥ = −1. The lowest mass states associated
with N ⊥ − Ñ ⊥ = 1 happens when N ⊥ = 1 and Ñ = 0, in which case we attain the two states
a∗1 |p+ , p~T ; ±1, ±1i ai∗ +

1 |p p~T ; ±1, ±1i (13.54)
Similarly, the lowest mass states associated with N ⊥ − Ñ ⊥ = −1 happen when N ⊥ = 0 and
⊥
Ñ = 1, yielding the two states
ã∗1 |p+ , p~T ; ±1, ∓1i ãi∗ +

1 |p p~T ; ±1, ∓1i (13.55)
In either case, the states have a mass of

2
R2

1 2 1 R
M2 = + 02 − 0 = − (13.56)
R 2 α α R α0
1 R
From here we can see that a particular radius will yield massless states, namely when = α0 →
√ R
R = α0 = `s We call this radius the self-dual radius.
13.3 T-Duality of Closed Strings

We are now in a position to define T-duality of closed string theory. In the last section we saw
that the mass spectrum of the compactified string is dependent on the radius of compactification
R. Consider (13.49). Notice then that if we use the compactification radius R or the dual radius,
0
R̃ ≡ αR we obtain the same mass spectra:
n 2 2
mR 2
2
M (R; n, m) = + + (N ⊥ + Ñ ⊥ − 2) (13.57)
R α0 α0
n2 R2 m2 2
M 2 (R̃; n, m) = + + 0 (N ⊥ + Ñ ⊥ − 2) (13.58)
α02 R2 α
The difference between these two expressions is rather superficial since the integers n and m
range over all possible integers. Therefore, for all m, n ∈ Z we have that
M 2 (R; n, m) = M 2 (R̃; m, n) (13.59)

Simply put, the closed string spectrum for a compactification with radius R is identical to a
0
closed string spectrum for a compactification with a dual radius of R̃ = αR . As we will prove
shortly, we can already see from (13.59) that these two compactifications are physically equivalent.
This is what we call the T-duality of closed string theory.
Let’s summarize our findings. On a circle of radius R, the momentum p is quantized in units of R1 .
In the limit where R >> 1, the spacing between the discretized momenta become small, yielding a
continuous spectrum. When the radius is very small, the momentum is largely spaced. What’s more
is that the winding is quantized by units of αR0 . Therefore when the spacing between momentum
is small, the spacing between the winding is large and vice versa. In other words, suppose we are
looking at a theory with a very large compactification radius, the momentum has a near continuous
spectrum while the winding has a discretized spectrum. The T-dual picture however, one that
yields the same physics, presents a theory where the winding has a near continuous spectrum while
the momentum has a discretized spectrum. The physics of these systems at first glance come off as
entirely different but they in fact yield the same physical spectrum.
It is important to note that in the present analysis, the compactification radius R is an adjustable
parameter. Our theory did not fix R. Moreover, the compactification radius is not a parameter of
string theory, but rather a parameter of space-time itself (this makes sense as we are compactifying a
spatial dimension of space-time, not a string coordinate). We typically call an adjustable parameter
a moduli, and the set of values of parameters the moduli can take is called a moduli space. We will
discuss this more later on in this chapter.
13.3. T-DUALITY OF CLOSED STRINGS 289
Let’s now go on to prove that T-duality is indeed a quantum symmetry of closed string theory.
Doing so will prove that T-duality imples that the physics of a compactified closed string theory
is identical at dual radii. To show this equivalence we will approach the problem in two different
ways. At first we will show that T-duality comes about as an “ambiguity” in the interpretation of
a single closed string theory. In the second approach we will show that T-duality is exhibited as an
equivalence between two distinct theories. This will allow us to make a one to one correspondence
between the operators of both theories which preserve the commutation relations, and maps one
Hamiltonian into the other. As discussed in the beginning of this chapter, satisfying these conditions
will effectively prove that T-duality is really a duality symmetry of closed string theory.
Beginning with the first approach, we note that the winding w had vanishing commutators with
all of the operators that appeared in our expansion for X. Then, just as (x0 , p) were a conjugate
pair, it is rather intuitive to make (q0 , w) a conjugate pair of variables. Motivated by this fact, we
introduce a dual coordinate operator defined by
X̄(τ, σ) ≡ XL (τ + σ) − XR (τ − σ) (13.60)
from which we find
r
0 0 α0 X e−inτ
X̄(τ, σ) = q0 + α wτ + α pσ + i (α̃n e−inσ − αn einσ ) (13.61)
2 n
n6=0
Moreover, we define the momentum P̄ τ conjugate to X̄ just as we are used to

1
P̄ τ ≡
∂τ X̄ (13.62)
2πα0
Using the dual coordinate operator and its conjugate momentum, we postulate the commutators
[64]
[X̄(τ, σ), P̄ τ (τ, σ 0 )] = iδ(σ − σ 0 ) (13.63)
[X̄ µ (τ, σ), X̄ ν (τ, σ 0 )] = [P̄ µτ (τ, σ), P̄ ντ (τ, σ 0 )] = 0 (13.64)

Moreover, since the conjugate pair (q0 , w) appear in X̄ in the same way as (x0 , p) does for X,
we find the standard commutator between q0 and w
[q0 , w] = i (13.65)
We are already familiar with the fact that x0 is a coordinate that lives on a circle of radius R and
n
has an associated conjugate momentum p with quantized eigenvalues R . Similarly, from w = mRα0 ,
α0
we are left to infer that q0 is an associated coordinate living on a circle of radius R̃ = R . Lastly,
the Hamiltonian (13.46), which was derived from (X, P τ ), coincides with the Hamiltonian derived
from (X̄, P̄ τ ), as the exchange of p and w has no effect. All together then, T-duality emerges
as an ambiguity in interpretation, arising from the possibility of replacing X = XL + XR with
X̄ = XL − XR .
Let’s move on to the second approach toward revealing T-duality. For this we would like to
describe a map which takes (X, P) → (X̄, P̄ τ ). This map is effectively the map which takes
(XL , XR ) → (XL , −XR ), which may be implemented through the following
{x0 → q0 ; q0 → x0 } {p → w; w → p} {αn → −αn ; α̃n → α̃n } (13.66)

In order to find the explicit map, we consider two theories, one with a compactification radius
0
R and another with a compactification radius αR . For a dual symmetry we are required to find an
operator map which takes the operators associated with each theory and yield the same commuta-
tion relations, as the physics lies within the commutation relations. Based on (13.66), we come up
with the operator map
{x0 → q̄0 ; q0 → x̄0 } {p → w̄; w → p̄} ¯n}

{αn → −ᾱn ; α̃n → α̃ (13.67)
From here it is straightfoward but tedious to check that with this operator map all of the com-
mutation relations between the operators of each theory and the Hamiltonian of each theory are
mapped into each other. Moreover, as we have already observed, this map yields two seemingly
different systems with identical spectra. Therefore, we may conclude that this map indeed yields
physical equivalence of these two theories, thereby proving that T-duality is a full quantum sym-
metry of free closed string theory on a compactified circle. T-duality of closed string theory may
in fact be summarized as
α0
R ←→ (13.68)
R
It it should be mentioned that in this analysis we have only been working with free closed strings.
Therefore we have proven T-duality for free closed strings. Though we won’t go into the details in
this text, T-duality does in fact hold for interacting closed strings as well, indicating that T-duality
is a symmetry of quantum closed string theory.
13.4 T-Duality of Open Strings on D-Branes

Up to this point we have shown that T-duality is a symmetry of closed string theory. Though this
is interesting, it is not too useful unless T-duality is a symmetry of open string theory as well. For
that reason, let’s consider the propagation of open strings in a space-time in which again one spatial
dimension has been curled up into a circle. Just as we chose in our analysis of closed strings, let’s
choose x25 = x to be compactified into a circle via the identification
x ∼ x + 2πR
All open string coordinates including X 25 satisfy Neumann boundary conditions and are all
therefore NN type coordinates. For open strings with endpoints attached to a space-filling D25
brane, they are free to move and can actually be shrunk to a point. Therefore, contrary to closed
strings winding around the compact dimension, open strings do not attain additional states. More-
n
over, the open string momentum along x is quantized in units of p25 = p = R . Since the string
does not wind around the compact dimension, the open string does not have any winding w.
Let’s now consider another theory, one with an open string in a space-time that has the same
0
spatial dimension compactified into a circle with a radius of R̃ = αR . T-duality of closed strings
would say that these two space-times are identical from the point of view of closed strings. Now
the open string has a momentum quantized in units of nR α0 . Again, since the string does not wind
around the compact dimension, there is no winding. In the first theory, the mass spectrum would
13.4. T-DUALITY OF OPEN STRINGS ON D-BRANES 291
2
n
gain a contribution of R 2 from the quantized momentum along the compact direction, while in the
2 2
second theory, the mass spectrum would gain a contribution of nαR 02 . Neither theory has a winding
term and therefore the two spectra do not yield the same physics. On the outset, it appears that
T-duality does not work upon the inclusion of open strings.
Don’t worry! Our previous work has not been for nothing. There is a solution which ends up
preserving T-duality in the presence of open strings. As we will see, for open strings T-duality ends
up finding a correspondence between a space-time with a compactification radius R and a D25-
0
brane and a space-time with a compactification radius of R̃ = αR , and a D24-brane. The physics
is equivalent for both open and closed strings if the D25-brane of the original theory is replaced by
a D24-brane in the dual theory (of course this would hold true for closed strings, as they do not
attach to D-branes). In summary, we have that T-duality for closed and open bosonic string theory
with one spatial dimension being compactifed is
α0
(D25; R) ←→ (D24; R̃ = ) (13.69)
R
All in all, by allowing T-duality to modify D-branes, we can preserve the T-duality symmetry
with the inclusion of open strings.
To begin our analysis, let’s first recall the expansion of NN type open string coordinates
√ √ X1
X I (τ, σ) = xI0 + 2α0 τ + i 2α0 αI cos(nσ)e−inτ (13.70)
n n
n6=0
Then, for the string coordinate X 25 (τ, σ) ≡ X(τ, σ) we have

√ √ X1
X(τ, σ) = x0 + 2α0 α0 τ + i 2α0 αn cos(nσ)e−inτ (13.71)
n
n6=0
Moreover, we also have that

√ n √
α0 = 2α0 p = 2α0 (13.72)
R
since the momentum along the circle is quantized. The Hamiltonian of the open string is just
1 I I 1
H = L⊥
0 −1= α α + N ⊥ − 1 = α0 pi pi + α0 α0 + N ⊥ − 1 (13.73)
2 0 0 2
where i = 2...24. Let’s now separate the string coordinate X into left and right moving compo-
nents just as we did for the closed string
X(τ, σ) = XL (τ + σ) + XR (τ − σ)
where, similar to before,
i√ 0X 1
r
1 α0
XL = (x0 + q0 ) + α0 (τ + σ) + 2α αn e−in(τ +σ) (13.74)
2 2 2 n
n6=0
i√ 0X 1
r
1 α0
XR = (x0 − q0 ) + α0 (τ − σ) + 2α αn e−in(τ −σ) (13.75)
2 2 2 n
n6=0
Motivated by the T-duality of closed strings, we define
X̄(τ, σ) ≡ XL − XR (13.76)
from which we find
√ √ X αn
X̄(τ, σ) = q0 + 2α0 α0 σ + 2α0 e−inτ sin(nσ) (13.77)
n
n6=0
But this is just the expansion for a string that stretches between two parallel D-branes:
√ √ X αa
n −inτ
X a (τ, σ) = x̄a1 + 2α0 α0a σ + 2α0 e sin(nσ) (13.78)
n
n6=0
where
√ 1 a
2α0 α0a =
(x̄ − x̄a1 ) (13.79)
π 2
By comparing (13.77) to (13.78), we rightfully identify x̄1 = q0 where we have purposefully
deleted the superscript a. The string coordinate X̄ is of the DD type since the endpoints are fixed:

∂τ X̄ = 0
σ=0,π
Moreover, notice that when σ goes from 0 to π, the open string stretches over the interval
√ α0 n
X̄(τ, π) − X̄(τ, 0) = 2α0 (α0 (π − 0)) = 2πα0 p = 2π = 2π R̃n (13.80)
R
where n ∈ Z. The above implies that we have an infinite collection of D24 branes with a uniform
spacing of 2π R̃ along the x direction. Furthermore, also note that the duality changes the boundary
conditions:
∂σ X = ∂τ X̄ ∂τ X = ∂σ X̄ (13.81)
This allows us to summarize the T-duality of open strings as
X = XL + XR ←→ X̄ = XL − XR (13.82)
∂σ X = ∂τ X̄ ←→ ∂τ X = ∂σ X̄ (13.83)
Indeed, the map X → X̄ is a quantum symmetry. Our work X a and P a from the last chap-
1
ter prove that X̄ and P̄ = 2πα 0 ∂τ X̄ satisfy the canonical commutation relations, indicating that
the duality transformation does not change the commutation relations. Moreover, recall that the
Hamiltonian for the sector including X a is
∞
1 X
H = 2α0 p+ p− = α0 pi pi + α0a α0a i
a
αna ) − 1 (13.84)
2 n=1
If we again delete the superscript a, we notice that this Hamiltonian in also derived from X̄ and
P̄, matching (13.73). All in all then, T-duality for open string theory is a full quantum symmetry.
13.5. A CLOSER LOOK AT U(1) GAUGE TRANSFORMATIONS 293
We can in fact extend the above analysis to include space-times where more than one spatial
dimension is compactified. Consider a D25-brane in a world where k spatial dimensions are curled
up into circles. A simulataneous T-duality transformation on each circle gives a physically equivalent
world where we have a D(25 − k)-brane and each circle with a radius of R is replaced by a circle
with a a dual radius R̃. In general, if a Dp-brane stretches around a compact dimension, T-duality
along the direction of compactification will yield in a D(p − 1) brane at some fixed point on a circle
of a dual radius.
13.5 A Closer Look at U(1) Gauge Transformations

As they will be used later on, it behooves us to review Maxwell gauge transformations in detail. In
this section, we heavily follow the methods laid out in Zwiebach’s text. For more details on this
notion and holonomies, the reader is urged to review Zwiebach’s text, A First Course in String
Theory.
Just as D-branes change under T-duality, gauge field configurations living D-branes also change
under T-duality. So far our study of gauge transformations has not included a world where spatial
dimensions are compactified. For this reason we must reexamine gauge transformations in the
presence of a compactified dimension, that way we can study its topological effects on gauge fields.
Since we are going to consider gauge transformations associated with the Maxwell field, U(1) gauge
transformations, let us consider gauge transformations in the presence of charges. Let’s start with
the action of a relativistic, charged point particle, which is just a generalization of the action we
came up with when we studied the relativistic point particle [64]
Z Z
S = −m ds + q Aµ (x)dxµ (13.85)
where we have set c = 1, and Aµ (x) is the Maxwell gauge field. For simplicity, consider the
action of the nonrelativistic, charged point particle, which is changed from the action above by
replacing the first term with the nonrelativistic action for a free particle
dxµ
Z Z
1
S= mv 2 dt + q Aµ (x) dt (13.86)
2 dt
~ and ~v = dxi
We can rewrite this action using the fact that Aµ = (Φ, A), dt for i = 1, 2, 3, in which
case we find
Z Z Z
1 ~ · ~v dt − q Φdt
S= mv 2 dt + q A (13.87)
2
The minus sign on the last term came from lowering the µ = 0 index with the Minkowski metric
ηµν . Using the above we can extract the Lagrangian:
1 ~ · ~v − qΦ
L= mv 2 + q A (13.88)
2
From which we find the momentum p~
∂L ~
p~ = = m~v + q A (13.89)
∂~v
The Hamiltonian then is just
~ · ~v − 1 mv 2 − q A
H = p~ · ~v − L = mv 2 + aA ~ · ~v + qΦ = 1 mv 2 + qΦ (13.90)
2 2
~ the Hamiltonian simply becomes
p − q A),
Using m~v = (~
1 ~ 2 + qΦ
H= p − q A)
(~ (13.91)
2m
The analysis of gauge transformations with a charge q is rather simple if we choose to examine
the Schrödinger equation for the nonrelativistic charged particle. We just found the Hamiltonian,
leaving us the Schrödinger equation
2
∂ψ 1 ∇ ~
i = Hψ = − qA ψ + qΦψ (13.92)
∂t 2m i
where we have turned the Hamiltonian into an operator, which effectively changes the momentum
p~ to the operator −i∇. It turns out that the Schrödinger equation is not invariant under just a
change in the vector potential Aµ . Rather, in quantum mechanics, a gauge transformation involes
changes in with the potentials of the wavefunctions. We leave it as an exercise for the reader to
show that the Schrödinger equation is invariant under the simultaneous changes
~→A
A ~ + ∇χ Φ → Φ0 = Φ − ∂χ
~0 = A ψ → ψ 0 = eiqχ ψ ≡ U ψ (13.93)
∂t
where χ(x) is a function of space-time and we have defined U (x) ≡ eiqχ(x) . With these simulta-
neous changes, the Schrödinger equation is found to be invariant since the primed coordinates yield
the same Schrödinger equation as before:
2
∂ψ 0

1 ∇ ~0
i = Hψ = − qA ψ 0 + qΦ0 ψ 0 (13.94)
∂t 2m i
Up to this point we have been viewing the vector potential gauge transformations as A0µ =
Aµ + ∂µ χ where χ(x) is the gauge parameter. It turns out this definition of the gauge parameter is
not sufficient in spaces with compact dimensions. U (x) however, is sufficient in compact dimensions,
and is equal to χ(x) in non-compact dimensions. Therefore, from now on we call U the gauge
parameter. If this the case, we must be able to write the gauge transformation of the vector
potential in terms of U . In order to match our gauge transformation in non-compact dimensions,
we can see that the gauge transformation becomes
i
A0µ = Aµ − (∂µ U )U −1 (13.95)
q
since (∂µ U )U −1 = iq∂µ χ. We call Maxwell theory a U (1) gauge theory because the gauge
parameter U can be viewed, for any fixed value of x, as an element of the group U (1), which is
literally defined as the group of one by one unitary matrices. A one by one matrix has a single
entry, and the unitarity condition, u∗ u = uu∗ = 1 is equivalent to taking the complex conjugate of
the single entry. This unitarity condition allows us to write u as a phase factor, u = eiθ . We leave it
as an exercise for the reader to prove that U (1), as presented, forms a group under multiplication.
13.6. THE AHARONOV-BOHM EFFECT AND WILSON LINES 295
It is often helpful to think about the gauge parameter U (x) is that for every space-time point x,
we attain a group element of the U (1) group.
We care about groups because gauge transformations performed in sequence combine by a rule
defined under group multiplication. That is, a gauge transformation with parameter U2 followed by
a gauge transformation with parameter U1 is equivalent to a gauge transformation with parameter
U1 U2 . In particular, the gauge field Aµ also transforms under the same sequence:
i i i
Aµ → Aµ − (∂µ U2 )U2−1 − (∂µ U1 )U1−1 = Aµ − (∂µ (U1 U2 ))(U1 U2 )−1 (13.96)
q q q
13.6 The Aharonov-Bohm Effect and Wilson Lines

Now that we have a more general understanding of U (1) gauge transformations, let us apply it to
the case where we work in the presence of a compactified dimension. It turns out this method is
very similar to the Aharonov-Bohm effect, so let’s first briefly review this effect.
Essentially, the Aharonov-Bohm effect is a quantum mechanical phenomenon in which an electri-
cally charged particle is placed in a region where the electric and magnetic fields are zero yet is still
is affected by the electromagnetic field. It turns out that the underlying mechanism of this effect
is that the electromagnetic potential is coupled with the complex phase of the charged particle’s
wavefunction. For simplicity we consider the Aharonov-Bohm effect when applied to a solenoid.
From our studies of electromagnetism, we know that the solenoid produces a magnetic field solely
on the interior of the solenoid, while the magnetic field is zero in the region outside the interior of
the solenoid. The Aharonov-Bohm effect tells us that although a charged particle moving outside
the solenoid is moving in a region where the magnetic field B ~ = 0, the wavefunction of the particle
~ Since the solenoid yields a non-zero magnetic field in its
is still affected by the vector potential A.
the interior, the vector potential outside the solenoid cannot vanish. If we choose a simple gauge to
work in, one can show that the potential outside the solenoid actually goes around the solenoid. It
is important to note that if there was no magnetic field anywhere, then the Aharonov-Bohm effect
would cease to exist as the interference effects would vanish. As we will see, in the case where we
have a compact dimension, there are physical effects from vector potentials even when the magnetic
field is zero everywhere.
Moving on, let us suppose that the x dimension is compactified into a circle. Moreover, let us
assume that the vector potential is zero except for its component along the x-direction, Ax . It is
important to point that when we compactify the x spatial dimension, there is no space inside the
circle. Therefore, the magnetic field is zero everywhere since there is no space in which the magnetic
field could exist on. Let us consider a vector potential that depends on the curled up dimension x,
Ax (x). Under the gauge transformation we are used to where χ is the gauge parameter, we have
∂χ
Ax → Ax +
∂x
At first glance, one might suspect that the usual identification x ∼ x + 2πR would yield χ(x +
2πR) = χ(x). To see what the consequences of this is, let us explore the line integral of the vector
potential around the circle
I
θ ≡ q Ax dx (13.97)
We have defined the line integral as θ for purposes which will become clear momentarily. More-
over, let us also define the holonomy W of the gauge field
I
W ≡ exp(iθ) = exp iq Ax dx (13.98)
We call W a Wilson line. We will come back to the definitions of a holonomy and a Wilson line
in shortly. For now we take the definition for face value: the Wilson line associated with a closed
curve is simply the phase factor that depends on the values of the gauge field along the curve. Now
notice that the gauge transformation above yields
I
0 ∂χ
θ = q dx( A + = θ + q (χ(x0 + 2πR) − χ(x0 )) (13.99)
∂x
where x0 is some arbitrary point on the circle. Notice then that if we use to suspected periodicit
condition, the gauge transformation yields θ = θ0 , which turns out to not be correct (Zwiebach,
408). Of course we now know why the suspected periodicity condition is wrong: it’s because here
we have used χ(x) as the gauge parameter. Instead we must use the gauge parameter U (x). Using
this as our gauge parameter, the periodicity condition becomes U (x + 2πR) = U (x). However since
we have defined U = exp(iqχ), we find that the periodicity condition really becomes
qχ(x + 2πR) = qχ(x) + 2πm (13.100)

with m ∈ Z. Put another way,
q(χ(x + 2πR) − χ(x)) = 2πm (13.101)

This is the periodicity condition since
U = exp(iqχ(x + 2πR)) = exp(iq(χ(x) + 2πm)) = exp(iqχ(x))

Moreover, if we use this periodicity condition with our gauge parameter U , we see that the gauge
transformation now forces
θ0 = θ + 2πm (13.102)
We are therefore allowed to make the identification θ ∼ θ + 2πm. Or, equivalently,
I I
q Ax dx ∼ q dxAx + 2πm (13.103)
We have chosen to use θ to denote the line integral of the vector potential since its most natural
interpretation is that it is an angle. Therefore, in the presence of compact dimensions, line integrals
of the vector potential are simply angular variables, in which case gauge transformations act as
θ → θ + 2πm. It is important to note that the Wilson line W = exp(iθ) is gauge invariant, and
that θ that are gauge equivalent yield the same holonomy W .
Before moving on, let us briefly discuss the basic notion of a holonomy. We have used the notion,
and made a definition, however we didn’t really explain what a holonomy is. Consider Maxwell
theory with a vector potential A. ~ Moreover, consider the circulation of the vector potential about
a curve C. Assuming that the curve is closed, we may invoke Stoke’s theorem, in which case
13.7. T-DUALITY OF OPEN STRINGS IN THE PRESENCE OF WILSON LINES 297
Z Z
~ · d~s =
A ~ · ~nd2 x
∇×A (13.104)
C S
where S os any surface where its boundary is C, and n is the unit vector normal to the surface at
the point of integration. As we are familiar with electromagnetism, if one specified this circulation
~ This
for all possible curves, it would be equivalent to specifying the curl of the vector potential A.
has physical importance since the curl of the vector potential is proportional to the electromagnetic
field tensor Fµν (e.g. the a component of the curl of the vector potential is related to the field
tensor as abc Fbc = 2(∇ × A)~ a ). What’s more is we know that this result is gauge invariant since
performing a gauge transformation is tantamount to modifying the vector potential by a gradient
of a function, which vanishes over the circulation of a closed curve.
A problem arises when one considers Yang-Mills theory. In Yang-Mills theory, the field tensor
is more than just the curl of a vector potential: it contains some non-linear terms that are the
consequence of the vector potential interacting with iteself (i.e. there are non-linear interactions
among the vector potential’s components. A result of this is the gauge transformations in Yang-Mills
theory become more complicated, and the circulation of the vector potential is not gauge invariant.
One therefore requires a more abstract, yet equally important, object to play the role of circulation
of a vector potential. A holonomy does just that. It is a quantity that is a gauge invariant and
constructs physical observables of the theory [48]. It is for this reason why we introduced the notion
of a holonomy and a Wilson line, it allows for us to identify gauge invariant physical observables of
our theory upon compactification.
In our present state, this is about as far as we can get to defining a holonomy in the strict
mathematical sense. It does have a strict definition, one that is intimately linked the geometric
notion parallel transport, however we are not in a position to explore the notion of a holonomy in
detail. For the interested reader, Rodolfo Gambini’s and Jorge Pullin’s A First Course in Loop
Quantum Gravity gives a fairly easy introduction into this concept, however still requires a little
background with differential geometry.
Moving on, based on our definition, with constant Ax , a Wilson line is realized to be
θ
qAx = (13.105)
2πR
where we made use of the identification x ∼ x + 2πR.
13.7 T-duality of Open Strings in the Presence of Wilson

Lines
We are now in a position to study the physics of open strings attached to D-branes that have
gauge fields described by holonomies. We will then use T-duality of open strings to yield a physical
interpretation.
Let us consider a Dp-brane that wraps around a single compact dimension x. By wrapping we
mean that one of the spatial dimensions along the Dp-brane is the compactified x-direction. As
we discovered in the last chapter, a gauge field lives on the world-volume of the Dp-brane. Using
the last section as motivation, let Hus assume that this gauge field on the world-volume is one such
that it can be described by θ = q Ax dx. Remember that T-duality alters the Dp-brane, the dual
world has a D(p − 1)-brane located at some position on a circle of a dual radius. It turns out that
the angle θ describes the position of the D(p − 1)-brane on the dual circle. Figure 13.2 gives a
representation of this dual picture.
Figure 13.2: Shown is a Dp wrapped on a circle, with vector potential A. The T-dual picture
is shown to the right with angle describing the position of the D(p-1) brane on the dual circle.
(Motivated by Zwiebach [64])
Let us now consider a string that is stretched between two Dp-branes, each one with their own
Maxwell field. From the last chapter we know that the endpoints of the string are charged. It
turns out that though the string endpoints are charged by the Maxwell field, the string itself is
neutral because the two endpoints have equal and opposite Maxwell charges. In the our present
configuration, let us suppose that the negatively charged endpoint lies on the first Dp-brane with
a Wilson line parameter θ1 , and the positively charged endpoint lies on the second Dp-brane with
a Wilson line parameter θ2 . Figure 13.3 illustrates this configuration.
13.7. T-DUALITY OF OPEN STRINGS IN THE PRESENCE OF WILSON LINES 299
Figure 13.3: Two Dp branes wrapped on the same circle, one with Wilson parameter 1 , and the
other with 2 . On the left is the T-dual picture where now the two D(p-1) branes are separated by
the difference in the angles 2 − 1 . (Motivated by Zwiebach [64])
As the reader will discover in the exercises, for a particle, the addition of a Wilson line results
in changing the momentum p to p − qA and the energy levels to shift by R` → R` − 2πR θ
. Since we
have two endpoints on the string, each one acting as a charged point particle, the addition of the
Wilson lines causes the momentum to shift as
p → p + qA1 − qA2
The energy levels shift as
` ` θ2 θ1
→ − +
R R 2πR 2πR
Now recall that if a Dp-brane wraps around a compact dimension, the mass-squared operator
for the open string states is
2
1 ` 1
M 2 = p2 + (N ⊥ − 1) = + (N ⊥ − 1) (13.106)
α0 R α0
Including the Wilson lines, which yield in an enegy shift, cause the mass-squared operator to be
shifted as
2
2π` − (θ2 − θ1 ) 1
M2 = + (N ⊥ − 1) (13.107)
2πR α0
Notice then that if θ1 = θ2 , the effects of the holonomies cancel each other out. Remember that
the angles θ1 and θ2 effectively give the positions of the two Dp-branes. Therefore, when these two
angles are equivalent, what we are really saying is that the Dp-branes are coincident to one another,
implying that they have the same gauge field. The T-dual picture to the two Dp-branes discussed
here would consist of two D(p − 1)-branes with different positions corresponding to two values of
θ. Finally, though we won’t go into the details in this text, it turns out that T-duality arguments
can be used to show that the electromagnetic fields living on the world-volumes of Dp-branes are
governed by Born-Infeld theory, a non-linear theory of electrodynamics which generalizes the linear
Maxwell theory [64].
In this chapter we have solely explored T-duality of both closed and open strings. We saw that it
0
can be summarized as (D25; R) → (D24; R̃ = αR ), for T-duality along the compactified dimension
x25 = x. We saw that T-duality is a quantum symmetry for both closed and open strings. The
most interesting consequence of T-duality is that by altering the D-branes that open strings are
attached to, it also alters the gauge fields living on D-branes, which saw as through the addition of
Wilson lines and holonomies.
In this chapter we have only considered T-duality of open and closed bosonic strings. As we
will see later, T-duality can also be applied to theories of superstrings, making it one of the most
useful symmetries in string theory. It is important to point out that T-duality is not the only
symmetry of string theory. The two other prevalent dual symmetries in the string theory literature
are S-duality and U-duality. In short, S-duality is a transformation that relates a string theory
with a coupling constant gs to a (perhaps) different theory with a coupling constant gs , becoming
crucial when one studies the interactions of strings [5]. U-duality, standing for unified -duality is a
duality transformation which combines both S and T-dualities, which theorists have found use for
in the study of M-theory.
13.8 A Brief Aside on Real and Complex Manifolds

We observe the world in four dimensions: three spatial dimensions and one time dimension. We
have studied bosonic string theory which requires that our universe is actually 26-dimensional.
Even still, the more realistic superstring theories require 10 or 11 space-time dimensions. In order
to make contact with our reality, we must be able to explain why we only experience four space-time
dimensions. In this chapter we have only considered compactifying a single spatial dimension. To
be able to make sense of a higher dimensional universe, we must compactify more than one spatial
dimension at a time. For the rest of the chapter we move to the topic of compactification, a topic
that researchers are still working on. discussion on compactification will yield some interesting and
controversial implications toward the construct of our universe.
Though this discussion is mostly qualitative, we must go over some basic elements of differential
geometry and topology. One of the most fundamental objects of differential geometry is a manifold.
The concept of manifolds are heavily used in General Relativity as we will see later. They are also
of tremendous importance in string theory, particularly in the methods of compactification. For
that reason, let’s spend a little bit of time go gain a physical intuition and a mathematical definition
of a a manifold.
A real d-dimensional manifold is a space which locally looks like Euclidean space Rd . More
precisely, a manifold of dimension d is defined by introducing a covering of open sets in which local
coordinate systems are introduced. These open sets are pasted together to explicitly construct the
manifold. A covering of open sets have strict mathematical meaning, however for our purposes we
can simply view them as coordinate patches, which when joined together locally look like Euclidean
space. Some of the most simple examples of a manifold include Rd and Cd . These examples are
called non-compact manifolds. We will go over the notion of compactness momentarily.
The most illustrative example of a manifold is the surface of a sphere. Consider the n-sphere,
which is characterized by
13.8. A BRIEF ASIDE ON REAL AND COMPLEX MANIFOLDS 301
n+1
X
(xi )2 = 1 (13.108)
i=1
The surface of a sphere is actually an example of a compact manifold. There is a more rigorous
mathematical definition of compactness, but for our purposes here, we will treat a compact manifold
as one with finite extension. With this loose language, we can interpret the non-compact examples
above as manifolds with infinite extension.
With the example of a surface of a sphere we can easily gain an intuition for a real manifold. Take
the surface of Earth for example. As one zooms in really close to the surface, it would appear the
Earth is flat (as long as we don’t consider mountain ranges or deep valleys). That is, locally (i.e.,
really zoomed in) the surface of the Earth, or the surface of any sphere, looks like flat, Euclidean
space. It’s no wonder that our ancestors thought the world was flat, it’s a manifold!
A manifold is a mathematical concept that one would meet in their studies in topology as well
as differential geometry. In fact, when one studies manifolds, they also study how to classify
different surfaces (two-dimensional manifolds, or 2-manifolds). One way to classify surfaces is
through the topological invariant Euler characteristic χ, unique to the surface. That is, one can
distinguish surfaces through the Euler-characteristic. For 2-manifolds, the Euler-number is given
by χ = 2 − 2g, where g is a non-negative integer called the genus. For all intensive purposes, we can
view the genus as the number of holes or handles a 2-manifold might have. For example, consider
the 2-sphere, which is the two dimensional surface of a sphere. This surface has no holes or handles,
and therefore has g = 0. Hence, χ = 2. This is important for classification purposes because nearly
any 2-manifold that has an Euler characteristic of 2 is said to be homeomorphic to the 2-sphere.
Formally, a homeomorphism is a continuous deformation of a topological object into a new shape
stretching it or compressing it without ever tearing it.
Figure 13.4: The coffee mug and donut can be continuously deformed into each other, and there-
fore said to be topologically equivalent, or homeomorphic. More precisely we say that there is a
continuous mapping taking the mug to the donut, and the inverse mapping from donut to mug is
also continuous.
Another famous example is a two-dimensional torus. A two-dimensional torus looks similar to

a doughnut having a single, thereby has a genus g = 1 and a corresponding Euler characteristic of
χ = 0. Nearly any 2-manifold that has an Euler characteristic of 0 is said to be homeomorphic to
the 2-torus. This notion yields a humorous result. A coffee mug has a hole: the hole created by
the handle! Therefore, coffee mugs and donuts are one in the same (conveinently paired for police
officers)! More precisely, a donut can be continuously deformed into a coffee mug and vice versa.
Consider figure 13.4 to find illustrations of these surfaces.
In differential geometry, a manifold gains a little more structure. It is endowed with a metric
g, a mathematical object we are at least a little familiar with by this point. If the metric is
positive definite, then the pair (M, g) where M is a set, is said to be a Riemannian manifold. If
it has indefinite signature, such as the cases we have been using so far and those found in general
relativity, the pair is called a pseudo-Riemannian manifold. In either case, the metric is a symmetric
tensor characterized by an infinitesimal line element
ds2 = gµν (x)dxµ dxν (13.109)

allowing one to compute the length of a curve by integration. Here gµν (x) is referred to as the
metric tensor, an object we are also familiar with by this point.
By describing a manifold with a metric we can describe distances and the overall geometry en-
dowed on the manifold. That is, by adding a metric to the manifold we are really endowing the
manifold with a type of geometry, a geometry which may be changed under general coordinate
transformations, or diffeomorphisms (more specifically, these are viewed as differentiable homeo-
morphisms, i.e. a homeomorphism with a differentiable inverse).
When one first learns differential geometry and general relativity, they learn about real manifolds.
In string theory real manifolds and complex manifolds are used extensively. A complex manifold
of dimension n is nothing more than a special case of a real manifold of dimension d = 2n. It is
defined in an analogous way using complex local coordinates z a and their complex conjugates z̄ a ,
with a = 1, ..., n. Essentially, an n-dimensional complex manifold is a space which locally looks like
the complex plane Cn . A crucial difference is that a complex manifold admits a tensor J with one
covariant and one contravariant index, which in complex coordinates has components
Jab = iδab Jāb̄ = −iδāb̄ Jab̄ = Jāb = 0 (13.110)

Let us now consider a complex Riemannian manifold. In terms of complex coordinates, the line
element associated with the metric is
ds2 = gab dz a dz b + gab̄ dz a dz̄ b̄ + gāb dz̄ ā dz b + gāb̄ dz̄ ā dz̄ b̄ (13.111)
A Hermitian manifold is a special case of a complex Riemannian manifold, which is characterized
by
gab = gāb̄ = 0 (13.112)

One of the most important types of manifolds in string theory is a Kähler manifold. A Kähler
manifold is defined to be a Hermitian manifold on which the Kähler form J is said to be closed,
meaning that dJ = 0. Here d is the exterior derivative. The concept of an exterior derivative and a
differential form will be discussed in detail in a later chapter. For now the reader should be aware
that a Kähler manifold is a special type of complex Riemannian manifold.
With these key mathematical concepts we are ready to proceed with our fairly qualitative dis-
cussion on compactification.
13.9. ORBIFOLDS AND THE TWISTED SECTOR 303
13.9 Orbifolds and the Twisted Sector

In the current research there are essentially two different methods to compactification. The first
approach is based around the work Kaluza and Klein did to present a five dimensional space-time
incorporating both gravitation as described by Einstein’s General Relativity and Maxwell’s theory of
electromagnetism. In this approach, the extra dimensions form a compact manifold of some critical
size, that are essentially invisible for observations at the current energies being used in particle
accelerators. The second approach is the brane world scenario, in which the four dimensions we
experience everyday are identified with a defect in a higher dimensional space-time. This defect
is given by a collection of coincident or intersecting D-branes. This approach appears promising
since Yang-Mills gauge fields reside on the world-volume of the D-branes, allowing one to contruct
(incomplete) string theory models of particle physics.
In the next few sections we will qualitatively discuss the first approach, and the physical impli-
cations of this approach. But before we examine the complexity of the compact manifolds used in
this approach, let us first discuss the simpler compact spaces, orbifolds
So far we have been considering compactifications where we take a single dimension and compact-
ify it into a circle of radius R using the identification x ∼ x+2πR. We call the interval 0 ≤ x < 2πR
the fundamental domain for the given identification. In general, a fundamental domain is a subset
of the entire space which satisfies two conditions: (1) No two points in the fundamental domain are
identified, (2) Any point in the entire space is in the fundamental domain are identified to some
point in the fundamental domain [64].
Figure 13.5: Two examples of orbifolds: (Top) The R1 Z2 , where R1 is the real line, and Z2 describes
the basic property of the identification x ∼ −x. (Bottom) The two-dimensional cone orbifold,
C/ZN , which is characterized by the identfication z ∼ e2πi/N z, for N ≥ 2 and z = x + iy is a
complex coordinate. (Motivated by Zwiebach [64])
Though we haven’t run into them explicitly yet, sometimes identifications have points that are
related to themselves by the identification. To see this, consider the real line when it is subject
to the identification x ∼ −x. In this case the point x = 0 is a fixed point because it is related to
itself via the identification. A fundamental domain can be chosen for this space however. It turns
out the fundamental domain is x ≥ 0, which is an example of an orbifold. An orbifold can be best
described as a space which is obtained by identifications with fixed points. Moreover, an orbifold
1
is singular at these fixed points. This particular orbifold is called an RZ2 , where R1 is the real line,
and Z2 describes the basic property of the identification x ∼ −x. If the identification is applied
twice, then one returns to the original coordinate.
2πi
Another basic example of an orbifold is ZCN , which is characterized by the identfication z ∼ e N z,
for N ≥ 2 and z = x+iy is a complex coordinate. This particular orbifold is actually a 2-dimensional
cone, in which the apex of the cone is the fixed (singular) point z = 0. Figure 13.5 illustrates both
1
the ZCN and RZ2 orbifolds.
As it turns out, orbifolds, despite their singular points, are tractable when we examine string
1
propagation in these spaces. To see this let us consider closed strings propagating on the RZ2 orbifold.
In this case we let x25 = x be the dimension that is compactified via the identification x ∼ −x.
This orbifold will end up restricting the spectrum of the original parent theory, the closed string
theory before orbifolding.
Let us write the corresponding string coordinate X 25 (τ, σ) = X(τ, σ). In the light-cone gauge, we
then have the collection of string coordinates as X + , X − , X i , and X, with i = 2, ..., 24. Moreover,
let us introduce the operator U which defines the identification of this orbifold. Therefore, acting
on X we have
U X(τ, σ)U −1 = −X(τ, σ) (13.113)

But U should transform any of the other coordinates meaning
U X i (τ, σ)U −1 = X i (τ, σ) U X ± (τ, σ)U −1 = X ± (τ, σ) (13.114)

As the reader will show, U turns out to be a symmetry of closed string theory. It’s important to
note that orbifold closed stirng theory only retains the U − invariant states of the original closed
string theory. To implement this constraint, it turns out to be conveinent to determine the action
of U on the oscillators. The coordinate X has the usual mode expansion
r
0 α0 X e−inτ
X(τ, σ) = x0 + α pτ + i (αn einσ + α̃n e−inσ ) (13.115)
2 n
n6=0
Using (13.112), we find that
U x0 U −1 = −x0 U pU −1 = −p U αn U −1 = −αn U α̃n U −1 = −α̃n (13.116)
From here we can define some of the states of the orbifold theory. We denote the ground states
of the parent theory by |p+ , p~, pi where p denotes the momentum along the compact direction. It
turns out that the U operator yields
U |p+ , p~, pi = |p+ , p~, −pi (13.117)

allowing us form U-invariant states of the theory as linear combinations of the orbifold ground
states [64]:
13.9. ORBIFOLDS AND THE TWISTED SECTOR 305
|p+ , p~, pi + |p+ , p~, −pi (13.118)
As usual, to build excited states we let creation operators act on the above superposition of
ground states.
The interesting feature of orbifold theories is that additional states exist aside from the ones
mentioned above. It contains the so-called twisted sector with a new type of closed string. These
closed strings can be viewed as open strings in the original theory, however with their endpoints
identified with the orbifold identification condtion. In the present analysis this means that
X(τ, σ + 2π) = −X(τ, σ) (13.119)
Equation (13.118) yields X(τ, 2π) = −X(τ, 0), meaning that the orbifold identification makes
them the same point, effectively closing the string. To develop a quantum theory of the twisted
sector we must first find an appropriate oscillator expansion for the string coordinate X. It turns
out the analysis is very similar to that of open strings stretched between parallel D-branes of
different dimensionality, so we will only quote the results here, and leave the details for the reader
to complete. After combining the left and right moving parts of the X coordinate, the mode
expansion for the twisted sector of closed strings is given by
r
α0 X 2 −i n τ n n
X(τ, σ) = i e 2 (α̃ n2 e−i 2 σ + α n2 ei 2 σ ) (13.120)
2 n
n∈Zodd
Using an analysis similar to that given in the previous chapter, we find that the commutation
relations of the oscillators are just
m
α̃ m2 , α̃ n2 = α m2 , α n2 = δm+n,0 (13.121)
2

α m2 , α̃ n2 = 0 (13.122)
By implementing the the orbifold identification and hence (13.112), we find that
U α n2 U −1 = −α n2 U α̃ n2 U −1 = −α̃ n2 (13.123)
Since there is no zero mode in the expansion of X we conclude that the twisted states do not
have a conserved momentum along the X-direction. Therefore, the U-invariant ground states of
the twisted sector are |p+ , p~i. The excited states in the twisted sector are again constructed using
the creation operators. To be able to physically interpret the states, one needs a formula for the
mass-squared operator M 2 that is similar to one built in the last chapter to describe the states of
strings stretched between parallel D-branes of different dimensionality.
Orbifold string theories always acquire a twisted sector. Orbifolding is therefore said to make a
double strike as it loses some states in the parent theory that are not invariant under the orbifold
action but then gains a new sector of states satisfying twisted boundary conditions. In that sense,
we see how twisted string states enclose the singular points of the orbifold, making it a tractable
theory with the inclusion of string propagation.
13.10 Orientifolds
Orientifolds are fundamental in interacting and unoriented string theories, and some string theory
models of particle physics. They also provide a physical system which is similar to an important
concept in compactification known as mirror symmetry.
An orientifold Op-plane is a hyperplane with p-spatial dimensions arising when we perform a
truncation which keeps only the closed string states that are invariant under symmetry transfor-
mations which simultaneously reverse the string orientation and reflects the coordinates to the
Op-plane. Similar to the coordinates describing a D-brane, let x1 , ...xp be directions along the
Op-plane and let xp+1 , ..., xd with d = 25 be the directions orthogonal to the Op-plane. We define
the position of the Op-plane by xa = 0 where a = p + 1, ..., d. Moreover, motivated by the Op-
plane directions and position, we arrange the string coordinates as X + , X − , {X i }, and {X a }, with
i = 2, ..., 24. Finally, let Ωp denote the operator that generates the transformation
Ωp X a (τ, σ)Ω−1 a
p = −X (τ, 2π − σ) Ωp X i (τ, σ)Ω−1 i
p = X (τ, 2π − σ) (13.124)
Moreover, let us assume that
Ωp x − −1 −
0 Ωp = x 0 Ωp p+ Ω−1
p =p
+
(13.125)
Using these transformation properties will allow us to calculate the action of a and i component
operators from the general expansion of a closed string coordinate
√
r
µ µ α0 X e−inτ µ inσ
µ 0
X (τ, σ) = x0 + 2α α0 τ + i (αn e + α̃n e−inσ ) (13.126)
2 n
n6=0
First of all, notice that if we change σ with 2π − σ, we effectively interchange the oscillators
αn and α̃n . Using this fact and (13.123), we find that the action of Ωp changes all of the signs
of the coefficients given in the expansion for X a , while it does not change any of the signs of the
coefficients given in the expansion for X a . That is,
Ωp xa0 Ω−1 a
p = −x0 Ωp pa Ω−1
p = −p
a
Ωp αna Ω−1 a
p = −α̃n Ωp α̃na Ω−1 a
p = −αn
Ωp xi0 Ω−1 i
p = x0 Ωp pi Ω−1
p =p
i
Ωp αni Ω−1 i
p = α̃n Ωp α̃ni Ω−1 i
p = αn
Moreover, due to the form of their mode expansions we have
ΩX ± (τ, σ)Ω−1 ±
p = X (τ, 2π − σ) (13.127)
Most importantly, since the closed string Hamiltonian is
H = α0 p+ p− = L⊥ ⊥
0 + L̃0 − 2 (13.128)
we find that
Ωp HΩ−1
p =H (13.129)
implying the action of the operator Ωp is a symmetry of closed string theory [64]. Let us denote
the space-time coordinates {x0 , ...xp } = xm . It turns out that the presence of an orientifold plane,
13.11. CALABI-YAU MANIFOLDS AND MIRROR SYMMETRY 307
the values of the fields at (xm , xa ) will determine the values of the fields located at (xm , −xa ).
Therefore we can see that the Op-plane acts as a type of mirror that relates the physics reflected
points. This means that one obtains the full set of fields of oriented closed string theory on one half
of the space and away from the orientifold. If we consider an O25-plane, which is space-filling, it has
no normal directions, and therefore orientifold symmetry includes only string orientation reversal,
a topic we discussed in a previous chapter.
Orientifold symmetry doesn’t just apply to closed strings. It also effects open strings attached
to D-branes. It turns out that the O25-plane truncates the open string spectrum to a set of states
that are invariant under the action of the operator Ωp , which reverses the orientation and reflects
the string coordinates normal to the Op-plane. When we have open strings attached to a Dp-brane,
the orientifold operator Ωp acts on the open string coordinates as
Ωp X a (τ, σ)Ω−1 a
p = −X (τ, π − σ) Ωp X i (τ, σ)Ω−1 i
p = X (τ, π − σ) (13.130)
Just as before, we demand that
Ωp x − −1 −
0 Ωp = x 0 Ωp p+ Ω−1
p =p
+
(13.131)
The resulting spectrum of possible states is interesting. Remember we only keep states that are
invariant under the action of the operator Ωp , i.e. the theory keeps states that have Ωp = +1. For
any state |ψi it can be shown that
⊥
Ωp |ψi = (−1)N |ψi (13.132)
Therefore we only keep the states where N ⊥ is even. But for massless states, M 2 = 0, we have
that N ⊥ = 1, yielding states that are not invariant under the action of the orientifold operator.
This implies that there are no massless states in the modified spectrum (Zwiebach, 351). Moreover,
recall from that the last chapter that a Dp-brane has a massless scalar for each normal direction.
The modified mass spectrum has no massless scalars and therefore has no independent directions
in which the Dp-brane can be moved. This is exactly the case for a space-filling D25-brane, which
has no massless scalars, consistent with the fact that it cannot be displaced either. In short, the
Dp-spectrum, in the presence of an Op-plane, is modified in such a way that it forces the Dp-brane
to stay with the Op-plane; it cannot move away from the Op-plane.
13.11 Calabi-Yau Manifolds and Mirror Symmetry

We have already examined a type of compactification, orbifolding, which lead to the twisted sector
states. This method is along the lines with the Kaluza-Klein approach to compactify the extra
dimensions in order to make contact with our 4-dimensional universe. Though orbifolds are an
interesting class of compactification spaces, researchers in the field have been mostly working with
Calabi-Yau manifolds as an attempt to curl up the extra dimensions of the more realistic super-
string theories. The mathematical background necessary to have a full understanding of these
compactification spaces goes beyond the scope of this text, however let us briefly discuss the basics.
The definition of a Calabi-Yau n-fold is a Kähler manifold having n-complex dimensions that
has a SU (n) holonomy [64]. It is the holonomy which had physicists intrigued by the Calabi-Yau
manifolds to begin with; it helps provide some models of particle physics. The standard non-
compact Calabi-Yau 1-fold is simply the complex plane C, which is actually described in terms of
the flat metric
ds2 = |dz|2 (13.133)

2
The only compact Calabi-Yau 1-fold is the two torus T (indeed the 2-torus has a two dimensional
surface, however it turns out it can be described in one complex dimension, which has two real
dimensions embedded in it). The 2-torus can also be described by the flat metric above, and can
be best thought of as a parallelogram with opposite sides identified.
Some examples of non-compact Calabi-Yau 2-folds, which have two complex dimensions, or four
real dimensions, can be obtain by taking products of the previous two manifolds. That is, non-
compact examples include: C2 = C×C, and C×T 2 . The most natural guess for a compact example
is the 4-torus, T 4 .
The Calabi-Yau manifolds of greatest interest are the Calabi-Yau 3-folds. First and foremost, in
superstring theory, the dimension of space-time is 10, and therefore 6 spatial dimensions must be
compactified in order to reach our four dimensional universe. A Calabi-Yau 3-fold has three complex
dimensions, or six real dimensions. Therefore, if the compactified space is a Calabi-Yau 3-fold, it is a
space compactifying six of the ten space-time dimensions in superstring theory. However, in contrast
to the lower dimensional Calabi-Yau manifolds, there are thousands of known Calabi-Yau-3 folds,
making it difficult to discern which Calabi-Yau 3-fold(s) would correspond the compactification
space associated with our universe. In fact, and this is one of the more troubling issues of the
Kaluza-Klein method to compactification, it is still an open question to whether or not the number
of Calabi-Yau 3-folds is actually finite!
A lot of effort has been spent in trying to determine the total number of such manifolds, that
we one can determine which, if any, Calabi-Yau 3-fold works for compactifying the extra spatial
dimensions we don’t experience everyday. Another, rather elegant, approach to this issue is based
on the T-duality of string theory. As we saw earlier, T-duality shows that geometry probed by
point particles is far different from the geometry probed by strings. In string geometry, a circle of
0
radius R can yield a physically equivalent theory associated with a circle of radius αR . A similar
phenomenon happens with Calabi-Yau 3-folds, known as mirror symmetry.
Maintaining our qualitative discussion, mirror symmetry is a mathematical conjecture that re-
lates two different Calabi-Yau 3-folds, call them M and W . This is crucial as in general M and W
are topologically different, meaning that they are not homeomorphic to one another. Nonetheless,
the conjecture provided by mirror symmetry supposes that the physics of these two Calabi-Yau
3-folds yields identical physics. Based on this definition, we can already recognize an example of
mirror symmetry: T-duality itself! A compactified circle with a radius R is certainly different from
0
a compactified circle with the dual radius αR . However, as we are well aware by now, the physics
of these two theories is identical.
A more involved example is the mirror symmetry of the 2-torus, T 2 = S 1 ×S 1 . Let the first circle
have a compactification radius of R1 , and let the second circle have a compactification radius of R2 .
The torus is ends up being characterized by complex structure and Kähler structure parameters
R2
τ =i ρ = iR1 R2 (13.134)
R1
If we then perform T-duality on the first circle, which sends R1 → R11 (where we have set α0 = 1),
we see that the resulting mirror torus has complex structure and Kähler structure parameters
13.12. STRING THEORY, PARTICLE PHYSICS, AND THE MULTIVERSE 309
R2
τ̃ = iR1 R2 ρ̃ = i (13.135)
R2
ultimately yielding that two different Calabi-Yau manifolds would end up yielding the same
physical spectrum (Becker, Schwarz, 413). A real triumph for the researchers working on this topic
was that they were able to show using the conjecture of mirror symmetry that type IIA superstring
theory compactified on a Calabi-Yau 3-fold was physically equivalent to type IIB superstring theory
compactified on another, topologically different Calabi-Yau 3-fold, indicating that the once thought
distinct theories are actually describe the same physics on different compactification spaces. Plenty
of more work lies ahead for string theorists to iron out all of the issues of using Calabi-Yau 3-folds
as the compactification spaces to describe the string theories developed so far. It is the hope of
many that mirror symmetry will provide further insights into unifying all of the string theories,
including M-theory, into a single theory, while at the same time give reason to the multitude of
Calabi-Yau 3-folds.
13.12 String Theory, Particle Physics, and the Multiverse

String theory was originally considered as a theory to describe the interactions of hadrons. It was
found that Quantum Chromodynamics is the correct theory to describe the interaction of hadrons,
however some theoreticians working in particle physics saw that string theory could do much more.
A lot of work done in string theory has been motivated to solve the shortcomings of the Standard
Model and to be able to describe all of particle physics (quantizing gravity just happened to be a
bonus!). This path is not yet complete, and a couple of approaches are being currently researched.
Each string theory model of particle physics relies heavily on the five supersymmetric string
theories and M-theory. Each of these five theories have been used to see how the Standard Model
might emerge as the particle physics model at lower energies. One particular method being currently
pursued is the model of intersecting D-branes, which makes use of type IIA superstrings. This
method turns out to be rather simple, however will not be discussed in detail in this text. For more
details on this approach, the reader is urged to review Zwiebach’s text.
Since each of the five string theories and M-theory are believed to constitute a single unified
theory, it is useful to consider the models of particle physics based on starting from the other
superstring theories. String phenomenology was originally based on the Heterotic E8 × E8 super-
string theory. In this particular theory, Calabi-Yau 3-folds were used to compactify six of the nine
spatial dimensions of Heterotic string theory. This compactification, still retaining what is a called
N = 1 supersymmetry (more on this later), the original gauge symmetry described by the group
E8 × E8 was broken to E6 × E8 . The group E6 contains SU (3) × SU (2) × U (1), the group defining
the Standard Model, as a subgroup and therefore, through further symmetry breaking, Calabi-Yau
compactification of heterotic strings gave the first models of semi-realistic particle physics. More
recently, string theorists have been studying the phenomenological properties of type II and type I
superstring theories using intersecting D-branes, orbifolds, and orientifolds [64].
Due to the second string theory revolution in the 90’s, there has much recent effort in exploiting
M-theory to come up with realistic models of particle physics. M-theory has a background space-
time of 11-dimensions, in which case we must compactify seven spatial dimensions as to make
contact with our 4 dimensional world. It turns out, to obtain a four dimensional theory with N=1
supersymmetry, the seven dimensional manifold (compactified space) must have something called
a G2 holonomy, a geometrical property which ends up constraining the curvature of the space.
As we have already mentioned, one of the chief goals of string theory is to be able to derive the
Standard Model of elementary particles from first principles. We would like string theory to be
unique, as then it would be the Grand Unified Theory that physicists have been searching for. The
uniqueness would lie in the existence of a single consistent quantum vacuum capable of predicting all
of the elements of the Standard Model. When it was discovered that several consistent superstring
vacua exist in ten dimensions, many became troubled by string theory as a unifying theory, and as
a model of particle physics. It was further found that several of these vacua were unrealistic. To
understand this better in terms of what we have studied in this chapter, consider the moduli space
of compactifications of closed string theory, the space of possible choices of the modulus R. The
fact that R is a parameter is a modulus means that the potential V (R) for R vanishes. This results
in the potential being unable to contain a mass term, ultimately causing the theory to include a
massless scalars, a conjecture that is inconsistent with observation [64]. Moreover, the vacua end
up having unbroken supersymmetry. This has dire consequences for cosmologists trying to under
the why the value of the cosmological constant is incredibly small but nonzero (the cosmological
constant, Einstein’s greatest blunder has recently found its way back into literature as a possible
explanation for the existence of dark energy). Both of these problems have been addressed by
moduli stabilization and flux compactifications. We won’t go into detail on these subjects in this
text, however for a detailed analysis of both of these topics, the text by Becker, Becker and Schwarz
is technical and thorough.
One approach in literature to deal with the high number of string vacua is with something called
the anthropic principle. In this approach, theorists argue that there are a plethora of nonsupersym-
metric vacua such that the typical spacing between adjacent values for the cosmological constant is
much smaller than the observed value. Using this logic, one argues that it would be reasonable to
assume that some of the vacua have, approximately the observed value of the cosmological constant.
This approach does not tell us why the cosmological constant, attributing to the dark energy (or
vacuum energy) is present in our universe, but rather implies that it can be small. If string theory
incorporates cosmic inflation then the anthropic principle is invoked. As presently understood,
inflation is forever into the future, which causes bubble universes to nucleate eternally. All of the
vacua of the string theory landscape, the potential function describing all of the vacua (the vacua
are actually viewed as the local minima of the potential function, i.e. landscape), are eventually
found to be physical bubble universes, one of which is a bubble universe like ours where the vacuum
energy is so small. We find ourselves in such a universe because a significantly larger cosmological
constant than observed would not lead to the formation of galaxies, and life as we know wouldn’t
exist. Put simply, we live in a universe where the cosmological is small because had it been any
other way, we would not exist.
Myself as well as others don’t find this argument too convincing, mostly based on the attitude that
the explanation should be more elegant. The anthropic principle seems a bit contrived. However,
this has had some success. Before a non-zero cosmological constant was observed in 1999, theorists
believed for a long time that it should be zero, however were never able to do so. Moreover, nobel
laureate Steven Weinberg showed that structure formation in the universe led to an upper bound
on the vacuum energy of the universe. That is, if the vacuum energy was significantly greater than
observed, nothing would have formed in the universe and we would not exist to discuss the matter!
Based on these findings, and the building of evidence for inflation, it is at least conceivable that
the landscape idea is the correct notion and our universe is just a part of a larger multiverse, one
that is not particularly natural. All in all, despite some of these fantastic ideas proposed by string
13.13. EXERCISES 311
theorists, it is certainly more constructive to continue studying the theories at hand. Perhaps we
should take the position John Wheeler took when he learned of Beckenstein’s discovery of black
hole entropy: “Your idea is crazy enough, it just might be right.”
13.13 Exercises
1. (a) Derive equation (13.15) to prove that the momentum p along the compactified direction
is a superposition of the zero modes α0 and α̃0 .
(b) Go through the details to prove the commutators (13.29), (13.30), and (13.31).
(c) Derive the compactified closed string Hamiltonian given in (13.46).
2. Show that one gets (13.94) upon using the gauge transformations given in (13.93).
3. Following the discussion presented after (13.95), prove that U(1) forms a group under multi-
plication.
4. (a) Complete the derivation necessary to come up with (13.119), (13.120), and (13.121).
(b) Based on your solution to (a), build the mass squared operator M 2 crucial to physically
interpret the states of the twisted sector.
Chapter 14
A Crash Course in Supersymmetry
Up to this point we have only been working with the bosonic string. That is, we have developed
a quantum theory of a relativistic string that only yields bosons, particles of integer spin. But
as we know from basic atomic and particle physics, there are other particles, half-integer spin
particles known as fermions. Fermions are the particles which constitute matter, and therefore any
quantum theory that is unable to include fermions is unrealistic. Therefore, as we have reminded
ourselves throughout this text, bosonic string theory is unrealistic. Certainly it yields particularly
interesting results, however in order to have a realistic quantum theory that aims to unify all
of the known forces, we must include fermions. There are string theories which do just this:
superstring theories. They are called super because the fermions are introduced into the theory
through supersymmetry. For that reason, in order to have a fair understanding of the superstring
theories, even at a qualitative level, it is necessary to first understand the basics of supersymmetry.
The focus of this chapter is to lend insight to the reader who is unfamiliar with supersymmetry,
both the mathematics involved and phenomenological implications. We pull out all of the stops
in this chapter, slowly introducing the reader to supersymmetry. Many might be weary of this
proposal, however the mathematics of supersymmetry isn’t too difficult (mostly using algebra to
find anticommutation relations, and plenty of index gymnastics). The real issue is the amount of
notation one meets when they learn supersymmetry. All in all, with only a little motivation from
our studies of quantum field theory, supersymmetry is a rather self-contained theory.
The bulk of this chapter is based around Patrick Labelle’s Supersymmetry Demystified, a real
tour de force seeking to provide a basic background in this subject. If the reader is interested in the
topics presented in this chapter, they are urged to peruse Labelle’s text. Another fair text on the
subject, though a bit more advanced, is Aitchison’s Supersymmetry in Particle Physics. Before we
get to laying the groundwork of supersymmetry, the reader is encouraged to go through this chapter
slowly, taking plenty of coffee breaks, as the notation can be overwhelming at times. Moreover,
though not all of the concepts developed in this chapter will be used explicitly in future chapters,
it is important to have a well rounded understanding as to leave no loose ends. As a final warning
to the reader, in this chapter we use the mostly minus convention of the Minkowski metric as this
is the convention most physicists use when studying supersymmetry. Without any further delay,
let us proceed. Carefully.
313
314 CHAPTER 14. A CRASH COURSE IN SUPERSYMMETRY
14.1 Motivation
In a previous chapter we examined the basics of quantum field theory. The motivation for this
was to have a handle on how physicists construct the state space or physical spectrum, and the
involvement of quantum fields, providing some background into a necessary prerequisite crucial in
fully appreciating string theory. What we did not mention is that the theories constructed from
the notions we discussed are incomplete. That is, many physicists maintain that all quantum field
theories we know should really be treated as effective field theories, theories that are approximations
to a more fundamental theory. We have already encountered this notion in string theory. Earlier we
briefly discussed that there were five seemingly plausible, and seemingly separate, string theories,
however were found to be connected to a more fundamental theory, M-theory. In that sense, the
five superstring theories are simply approximations to the more fundamental unifying M-theory.
What this means for quantum field theories is that there is nothing wrong with non-renormalizable
theories. Back in the first chapter we saw how when we introduce gravity into a quantum field the-
ory, the resulting theory is plagued with infinities, making it impossible to extract any helpful
physical information. However, if the quantum field theory is an effective field theory, we can con-
sider a finite cutoff in the integrals one must perform, thereby avoiding the infinities, while knowing
all along that something more must be done to discover the more fundamental theory.
These issues also led us to the fact that the standard model of particle physics is incomplete.
One key reason for this incompleteness is the so-called Hiearchy problem, which originates from
the Higg’s particle receiving large corrections from loop diagrams in particle processes. These
corrections can in principle be canceled by fine-tuning the parameters of the standard model. But
this solution is contrived and is rather unnatural, leading physicists to believe that the standard
model is incomplete. String theory is a theory beyond the standard model which aims to resolve
these issues as well as unify all of the known forces. Supersymmetry is a theory which aims to
resolve the issues of the standard model through a symmetry that hasn’t been observed yet.
The basic idea of supersymmetry, SUSY, is simple: it is a symmetry which involves changing the
bosonic and fermionic fields into one another. Therefore, SUSY solves the hiearchy problem: the
fermionic and bosonic loop corrections to particle masses exactly cancel one another, ridding the
theory of large loop corrections; eliminating the need for fine tuning the parameters of the Standard
Model. As it turns out, almost all of the ultraviolet (high momentum) divergences of conventional
quantum field theory vanish when SUSY is introduced.
Unfortunately, there is a catch: exact invariance under supersymmetry requires that for each
boson there is an associated fermion, its supersymmetric partner (superpartner), and for each
fermion there is an associated superpartner that is a boson. So far these sparticles have yet to be
observed, however experimentalists at the LHC are working on it!
In a different vein, performing two successive SUSY transformations on a quantum field gives the
same field however evaluted at a different point in space-time. That is, SUSY transformations are
intimately linked to space-time translations. In fact, when one imposes that a field is invariant under
local SUSY transformations, they are forced to introduce a new set of fields that automatically
reproduce Einstein’s general relativity! This theory is called supergravity (SUGRA). We won’t
explore this theory in this text, but supergravity has been found to be the low energy limit of M-
theory and has therefore found its place in the realm of string theory. The reason why we care about
learning SUSY is because when we build a string theory that includes both bosons and fermions,
supersymmetry naturally emerges!
14.2. A REVIEW OF WEYL SPINORS 315
14.2 A Review of Weyl Spinors

As we mentioned earlier, the real difficulty of supersymmetry is not the mathematics but rather
the notation. What’s more is in quantum field theory, one becomes most familiar with Dirac
spinors, the spinors we first introduced when we examined the Dirac equation. In supersymmetry
however, Weyl and Majorana spinors are used more often, meaning that the student must become
comfortable with this, likely less familiar representation. Indeed, we introduced the notion of Weyl
and Majorana spinors during our studies on the Dirac equation, however, we will spend a bit more
time examining the details of the theory as they are crucial in being able to read the literature on
supersymmetry.
In the chapter on quantum field theory, we found that the Dirac equation could be written in
the form
γ µ Pµ ψ = mψ (14.1)
where Pµ = i∂µ , and ψ is a four component Dirac spinor. Remember, the Dirac equation and
spinor fields describe fermions. Moreover, recall the Dirac Lagrangian density
L = ψ̄(γ µ Pµ − m)ψ (14.2)

where ψ̄ = ψ † γ 0 . We are going to be using dagger notation in this chapter to denote the
Hermitian conjugate, as we will use the asterisk symbol to denote typical complex conjugation. As
we mentioned before in the chapter on quantum field theory, there are alternative representations
for the matrics γ 0 and ~γ . When we study supersymmetry, we will make use of the representation

0 0 I 0 −~σ
γ = ~γ = (14.3)
I 0 ~σ 0
where I is the 2 × 2 identity matrix and ~σ are the Pauli matrices we are familiar with from
ordinary quantum mechanics

1 0 1 2 0 −i 3 1 0
σ = σ = σ = (14.4)
1 0 i 0 0 −1
One property of the Pauli matrices is that the product of any two Pauli matrices is
σ i σ j = δ ij + iijk σ k (14.5)
ij ijk
where δ is the Kronecker delta, and is the totally antisymmetric Levi-Civita tensor, having
that 123 = 231 = 312 = 1, and 321 = 213 = 132 = −1, and has all other components equal
to zero. Equation (14.5) allows us to compute the commutator and anticommutator between two
Pauli matrices:
[σ i , σ j ] = σ i σ j − σ j σ i = 2iijk σ k (14.6)
{σ i , σ j } = σ i σ j + σ j σ i = 2δ ij (14.7)
We can use (14.3) to define

γ µ = (γ 0 , ~γ ) (14.8)
Moreover, using the mostly minus convention of the Minkowski metric yields the covariant version
of (14.3):
γµ = ηµν γ ν = (γ 0 , −~γ ) (14.9)

We had also defined the γ5 matrix which is important for defining chirality

I 0
γ5 = (14.10)
0 −I
The γ5 matrix comes in handy when we define Weyl spinors. Let’s do that now. First let us
write the four component Dirac spinor in terms of two two component spinors. That is,

η
ψ= (14.11)
χ
These two component spinors, η and χ are called Weyl spinors. For reasons which will become
clear shortly, we decompose the Dirac spinor into two spinors because the Weyl spinors separately
transform under Lorentz transformations, which will help us build Lorentz invariants. More pre-
cisely, we say that a Dirac spinor is a reducible representation of the Lorentz group while Weyl
spinors form an irreducible representation, which, in a sense, suggests that Weyl spinors are more
fundamental than Dirac spinors.
It is important to note that if we set either Weyl spinor equal to zero in the Dirac spinor, we
find the eigenstates of the γ5 matrix. In particular,

η η 0 0
γ5 =+ γ5 =−
0 0 χ χ
We call the eigenvalue of γ5 the chirality of the spinor. A Weyl spinor with positive chirality is
sometimes referred to as a right-chiral spinor, while a Weyl spinor with negative chirality is referred
to as a left-chiral spinor. In that sense, we see that η is a right-chiral spinor and χ is a left-chiral
spinor. We therefore sometimes denote η by ηR , and χ by χL . We will only use η to mean right-
chiral spinors, and χ to mean left chiral spinors in this text, so we avoid using the subscript. As
we will see in this chapter, left-chiral spinors, though mostly through convention, are of particular
importance in supersymmetry.
We leave it to the reader to show that substituting (14.11) into the Dirac equation yields two
coupled equations
(EI − ~σ · P~ )η = mχ (EI + ~σ · P~ )χ = mη (14.12)
Before moving on, let us introduce some further notation that will prove useful later on. Let us
first define
σ µ ≡ (I, ~σ ) σ̄ µ ≡ (I, −~σ ) (14.13)

Then, using the fact that (σ 2 )2 = I, we find
14.3. LORENTZ TRANSFORMATIONS OF WEYL SPINORS 317
σ 2~σ σ 2 = −~σ ∗ σ 2~σ T σ 2 = −~σ

from which gives us the useful identities
σ 2 (σ µ )T σ 2 = σ̄ µ σ 2 (σ̄ µ )T σ 2 = σ µ (14.14)
2 T 2
Or, taking the transpose, and using (σ ) = −σ , we find
σ 2 σ µ σ 2 = (σ̄ µ )T σ 2 σ̄ µ σ 2 = (σ µ )T (14.15)
Another identity that is rather trivial in proving is
σ̄ µ σ ν + σ̄ ν σ µ = σ µ σ̄ ν + σ ν σ̄ µ = 2η µν (14.16)
µ µ
Finally, using the definitions of σ and σ̄ , the coupled equations become
Pµ σ µ η = mχ Pµ σ̄ µ χ = mη (14.17)
µ
Moreover, we may write the gamma matrices γ as
0 σ̄ µ

γµ = µ (14.18)
σ 0
In terms of Weyl spinors, the Dirac Lagrangian can be written, as the reader will prove, in the
form
L = η † σ µ i∂µ η + χ† σ̄ µ i∂µ χ − mη † χ − mχ† η (14.19)
14.3 Lorentz Transformations of Weyl Spinors

It is imperative that we know how Weyl spinors transform under Lorentz transformations. The
inifinitesimal transformations of right-chiral and left-chiral Weyl spinors are [30]:

i 1
η → η 0 = I + ~ · ~σ − β~ · ~σ η (14.20)
2 2

i 1
χ → χ0 = I + ~ · ~σ + β~ · ~σ χ (14.21)
2 2
~ is the infinitesimal boost parameter. Based
where ~ is the infinitesimal rotation vector, and β
on these expressions, we notice that indeed the left and right-chiral spinors transform independent
from one another. For comparison, consider the transformation of the four-momentum P µ . The
zeroeth component, P 0 = E and spatial components P~ transform as
E → E − β~ · P~ ~
P~ → P~ − ~ × P~ − βE
To build Lorentz invariant quantities out of Weyl spinors, we focus on the Dirac Lagrangian den-
sity. We know that Lagrangians and therefore Lagrangian densities are Lorentz invariant quantities.
Therefore, the mass term in the Dirac Lagrangian density, ψ † γ 0 ψ, is Lorentz invariant. Writing the
mass term with Weyl spinors we find that
ψ̄ψ = ψ † γ 0 ψ = η † χ + χ† η
is invariant under Lorentz transformations. To show this, consider the transformation η † χ →
†0
η χ0 . Using (14.20) and (14.21), this transformation is simply

† i 1~ i 1~
η I − ~ · ~σ − β · ~σ I + ~ · ~σ + β · ~σ χ
2 2 2 2
Making note of the fact that all terms of higher order then the linear terms drop since they we
are dealing with infinitesimal parameters, we are left with

i 1~ i 1~
η I − ~ · ~σ − β · ~σ + ~ · ~σ + β · ~σ = η † χ
†
2 2 2 2
Therefore, η † χ is invariant under the transformations. Similarly, one can show that χ† η is
invariant under the transformations given in (14.20) and (14.21), proving that the mass term,
written in Weyl spinors is Lorentz invariant.
Let’s also consider the kinetic energy term of the Dirac equation, ψ̄γ µ Pµ ψ. What’s more is since
Pµ is a four-vector, it must also be that ψ̄γ µ ψ transforms as a four-vector. Writing this in terms of
Weyl spinors, and using our defintion for γ µ given in (14.18), we have
ψ̄γ µ ψ = η † σ µ η + χ† σ̄ µ χ
For this to transform as a four-vector under Lorentz transformations, we must have that each
term seperately transforms as a four-vector. To show that this indeed true, consider the η † σ µ η,
and first consider when µ = 0. We have then

0 i 1~ i 1
η † σ 0 η → η † η 0 = η † I − ~ · ~σ − β · ~σ + ~ · ~σ − β~ · ~σ η
2 2 2 2
= η † η − β~ · (η †~σ η) = η † σ 0 η − β~ · (η †~σ η
which we recognize as the transformation of the zeroeth component of the four-momentum Pµ .
If we consider the spatial components, η † σ j η, with a little work one can show that
0
~ † σ0 η
η † ~σ η 0 = η †~σ η − ~ × (η †~σ η) − βη
indicating that the spatial components transform as the spatial components of the four-momentum.
A similar exercise shows that the other term χ† σ̄ µ χ transforms as a four-vector, and therefore al-
together, the kinetic energy term, when written in Weyl spinors, transforms as a four-vector, as
expected.
Since we will mostly only consider left-chiral spinors when we study supersymmetry, we wonder
whether it is possible to construct Lorentz invariants purely out of left-chiral spinors. That is, can
we construct a mass term purely out of χ. In short, the answer is yes, however it is definitely
non-trivial to do so. Remember the mass term has two terms in it: η † χ and χ† η, each of which are
Lorentz invariant. Therefore, for the collective mass term to be written in terms of only left-chiral
spinors χ, we must build something out of χ such that it transforms as η, making it so the mass
14.3. LORENTZ TRANSFORMATIONS OF WEYL SPINORS 319
term we build out of solely left-chiral spinors does not spoil the Lorentz invariance. First of all,
notice that if we take the transpose of the Hermitian conjugate of χ we have
∗
i 1 i 1
(χ† )T → (χ0 )†T = I + ~ · ~σ + β~ · ~σ χ†T = I − ~ · ~σ ∗ + β~ · ~σ ∗ χ†T (14.22)
2 2 2 2
Comparing to (14.20), we notice this does not transform as a right-chiral spinor because minus
signs are switched and we are working with the complex conjugate of the Pauli matrices. Both of
these issues are resolved if we use σ 2~σ = −~σ σ 2 . Let us now consider iσ 2 χ† , in which case we find
that
∗
i 1 i 1
iσ 2 χ†T → (iσ 2 ) I + ~ · ~σ + β~ · ~σ χ†T = iσ 2 I − ~ · ~σ ∗ + β~ · ~σ ∗ χ†T
2 2 2 2

i 1~
= I + ~ · ~σ − β · ~σ iσ 2 χ†T
2 2
Comparing this to (14.20), we see that indeed iσ 2 χ†T transforms like a right-chiral spinor. Sim-
ilarly, one can check that −iσ 2 η †T transforms like a left-chiral spinor. With these facts in hand,
let’s proceed in constructing Lorentz invariants only in terms of left-chiral spinors.
Since η † χ is invariant under Lorentz transformations, and iσ 2 χ†T transforms as like η, we also
know that
(iσ 2 χ†T )† χ = χT (iσ 2 )† χ = χT (−iσ 2 )χ (14.23)

is invariant under Lorentz transformations. This will also help us construct a mass term solely
in terms of left-chiral spinors that is Lorentz invariant. Moreover, since χ† η is Lorentz invariant,
we also find that
χ† iσ 2 χ†T (14.24)
2 †T
is Lorentz invariant. Moreover, since −iσ η transforms like a left-chiral spinor, one can show
that we have the Lorentz invariants [30]
χ† η η† χ (14.25)
−χT iσ 2 χ χ† iσ 2 χ†T (14.26)
−η † iσ 2 η †T η T iσ 2 η (14.27)
Let us move on to terms which contain derivatives of the spinor fields and see if we can construct
more Lorentz invariants. If we look at the Dirac Lagrangian density shown in (14.19), we see that
σ µ i∂µ η transforms like a a left-chiral spinor, while σ̄ µ i∂µ χ transforms as a right-chiral spinor. We
know that χ† η is Lorentz invariant, which means that
χ† σ̄ µ i∂µ χ
is Lorentz invariant. Alternatively, we saw that η † χ is a Lorentz invariant. Using the above
discussion, we can convince ourselves that
(σ̄ µ i∂µ χ)† χ = −i(∂µ χ† )σ̄ µ χ

is also Lorentz invariant. As it turns out, as the reader will prove, one can use integration by
parts to show that these two expressions are infact equivalent. All in all, we have terms containing
derivatives of spinor fields which are indeed Lorentz invariant.
14.4 The Spinor ‘dot product’

In this section we will discuss some more notation which will be rather useful when writing down
expressions containing Weyl spinors. In practice, physicists studying SUSY work almost entirely
with left-chiral spinors. For that reason, we really only worry about the Lorentz invariants given
in (14.26). We use these Lorentz invariants to define two types of a spinor dot product:
χ · χ ≡ χT (−iσ 2 )χ χ̄ · χ̄ ≡ χ† (iσ 2 )χ†T (14.28)

To make practical use of these definitions, let us write out the spinor dot product in terms of
the components of the left-chiral spinor. That is, if we define the two component Weyl spinor as

χ
χ≡ 1 (14.29)
χ2
we can work out the spinor dot product in terms of components to be
χ · χ = χ2 χ1 − χ1 χ2 (14.30)
But let’s keep in mind that spinors are Grassmann quantities, i.e. anticommuting variables.
Therefore χ1 χ2 = −χ2 χ1 . Using this fact we see that
1
χ · χ = 2χ2 χ1 ⇒ χ1 χ2 = − χ · χ (14.31)
2
The second spinor dot product can be worked out to give us
χ̄ · χ̄ = χ†1 χ†2 − χ†2 χ†1 = 2χ†1 χ†2 (14.32)

yielding
1
χ†1 χ†2 =
χ̄ · χ̄ (14.33)
2
Both (14.31) and (14.33) can be summarized into a rather useful identity
1 1
χa χb = − (iσ 2 )ab χ · χ χ†a χ†b = (iσ 2 )ab χ̄ · χ̄ (14.34)
2 2
where the subscripts a and b stand for the components of the spinor, and the components of the
Pauli matrix. For instance, (σ 2 )11 = 0.
Now recall that in general (AB)† = B † A† . But what if A and B were Grassmann variables, such
as the spinor fields we are dealing with presently? Shouldn’t the exchanged of A and B introduce
14.5. CHARGE CONJUGATION AND WEYL SPINORS 321
a minus sign? The convention in the literature is that the Hermitian conjugate of two spinor
fields does not introduce an extra minus sign [30]. Instead, we define that the complex conjugate
of a product of two Grassmann variables change their order without introducing a minus sign.
Effectively, each exchange would cancel out the minus signs, fixing this subtle ambiguity. Using
this convention we find
(χ · χ)† = 2(χ2 χ1 )† = 2χ†1 χ†2 = χ̄ · χ̄ (14.35)

Hence,
(χ · χ)† = χ̄ · χ̄ (χ̄ · χ̄)† = χ · χ (14.36)

We can actually generalize these expressions to include more than one left-chiral spinor. That
is, suppose that we have two left-chiral spinors χ and λ. Notice then
λ · χ = λT (−iσ 2 )χ = −χT (iσ 2 )T λ = χT (−iσ 2 )λ = χ · λ (14.37)

where we made use of the identity
αT Aβ = −β T AT α (14.38)
For two Grassmann variables α and β and a matrix A. Similarly one can prove that
λ̄ · χ̄ = χ̄ · λ̄ (14.39)
From here we may generalize (14.36)
(λ · χ)† = λ̄ · χ̄ (λ̄ · χ̄)† = λ · χ (14.40)

All of these identities will prove quite useful when we work out the supersymmetric algebra.
14.5 Charge Conjugation and Weyl Spinors

When one studies particle physics, they come across the operation known as charge conjugation.
This effectively turns a field describing a particular particle into a field describing the associated
antiparticle. To study this effect, consider the Dirac spinor written in terms of two Weyl spinors

η
ψp = p
χp
We will denote the antiparticle Dirac spinor upon operation of the charge conjugation operator
as

η
ψpc ≡ ψp̄ ≡ p̄
χp̄
where we have used the common bar notation to indicate an antiparticle (e.g. ē is the antielec-
tron). The operation of charge conjugation is defined using the charge conjugation operator C in
the following way
ψpc ≡ C ψ̄pT (14.41)

where the charge conjugation matrix takes the representation [30]
2
2 0 iσ 0
C = −iγ γ = (14.42)
0 −iσ 2
This can be recast in a more convenient form
ψpc = C ψ̄pT = C(ψp† γ 0 )T = C(γ 0 )T ψp†T = Cγ 0 ψp†T ≡ C0 ψp†T

where we made use of the fact that (γ 0 )T = γ 0 , and introduced
iσ 2

0
C0 = Cγ 0 = −iγ 2 γ 0 γ 0 = −iγ 2 = (14.43)
−iσ 2 0
If we write the Dirac spinor in terms of the Weyl spinors and apply C0 , we find that the charge
conjugated state is
2 †T
c iσ χp
ψp = (14.44)
−iσ 2 ηp†T
Comparing this to the antiparticle spinor, we find
ηp̄ = iσ 2 χ†T
p χp̄ = −iσ 2 ηp†T (14.45)
indicating that the right-chiral spinor corresponding to an antiparticle is written in terms of a
left-chiral spinor, while a left-chiral spinor corresponding to an antiparticle is written in terms of a
right-chiral spinor. It is rather simple to prove that applying the charge conjugation operator twice
to a spinor state yields the same state. We may therefore conclude that ψp = C(ψ̄pc )T , which leads
us to
ηp = iσ 2 χ†T
p̄ χp = −iσ 2 ηp̄†T (14.46)
What we may conclude then is that we may always write left-chiral and right-chiral states of
an antiparticle in terms of left-chiral and right-chiral states of a particle, and vice versa. This is
important for us since in supersymmetry we only care to use left-chiral spinors. Therefore, we may
write the Dirac spinor of a particle as
2 †T
η iσ χp̄
ψ= p = (14.47)
χp χp
The importance of charge conjugation should be stressed. Realistic relativistic quantum field
theories must invariant under the CPT operator. This is the combination of charge conjugation C,
parity P , which for all intensive purposes can viewed as spatial inversion, and time reversal T . The
so-called CPT theorem is important in the sense that it shows that Lagrangian densities are real,
meaning that if a theory contains a left-chiral spinor, it must also contain its Hermitian conjugate,
otherwise it would not be invariant. What this means for us is that for a CPT invariant theory, when
we have a left-chiral spinor describing a particle, it must also contain the corresponding antiparticle
spinor, which, as we have shown, will be a right-chiral spinor.
14.6. MASSIVE SPINORS 323
14.6 Massive Spinors

Let us construct a theory with a left-chiral Weyl spinor χp that is massive. This is one step up from
the massless fermion case, however is a more realistic theory. We will have to keep in mind that we
wish to maintain CPT invariance, meaning that we are forced to introduce the charge conjugate
state to χp , namely, ηp̄ . What’s more is, when we consider massive fermions described by χp , one
can show that Lorentz transformations require that we must also introduce a right-chiral spinor ηp .
As we will see, these two right-chiral states have an interesting interpretation which will lead us
to the concept of Majorana spinors, the other type of spinor a physicist studying supersymmetry
uses.
It turns out we have two options: the right chiral states represent the same physical state, or
they do not. Let us first consider the latter case, where the right-chiral states represent different
physical states, i.e. ηp 6= ηp̄ . If this is the case then we are left with four distinct states: χp , ηp ,
which describe some particle, and χp̄ , ηp̄ , which describe the associated antiparticle. As one might
guess, we can combine the left and right-chiral states into a four component Dirac spinor

η
ψD = p (14.48)
χp
Since, at the moment, we are assuming that the spinors describing the antiparticle state is distinct
c
from the particle state we may conclude that ψD 6= ψD . Using the Dirac Lagrangian density, we
can write out the mass term for the Dirac spinor. We recognize the mass term to be
−mψ̄D ψD = −mψ † γ0 ψ = −m(ηp† χp + χ†p ηp ) (14.49)

Had we used the form of the Dirac spinor we found in the last section, (14.47), we would instead
write the mass term as
−mψ̄D ψD = −m(χTp̄ (−iσ 2 )χp + χ†p iσ 2 χ†T

p̄ ) (14.50)
Or using the definitions of the spinor dot products (14.28), we instead write the mass term as
−mψ̄D ψD = −m(χp̄ · χp + χ̄p · χ̄p̄ ) (14.51)
Let’s now consider the other case, when the two right-chiral states ηp , and ηp̄ do infact describe
the same physical state. Such a notion is not foreign to particle physics. Experiments have found
that some particles can also be identified as their antiparticles. When one studies particle physics,
they are able to distinguish particles from their antiparticles by examining the particle’s quantum
numbers. An antiparticle has quantum numbers that are exactly opposite to the quantum numbers
of the particle. Thus, when we say that a particle is also identified as its own antiparticle, what we
mean is that there is no quantum number allowing us to distinguish the particle from its antiparticle.
What this means for us is that
ηp = ηp̄ = iσ 2 χ†T
p
It’s also easy to check that the left-chiral antiparticle state is the same as the left-chiral particle
state:
χp̄ = −iσ 2 ηp†T = −iσ 2 (iσ 2 χ†T

p )
†T
= −iσ 2 (iσ 2 χp ) = χp
In summary, we have that
χp̄ = χp ηp̄ = ηp = iσ 2 χ†T

p (14.52)
Using this we can construct a four component spinor formally called a Majorana spinor ψM
2 †T
η iσ χp
ψM = p = (14.53)
χp χp
An important property of Majorana spinors to consider is that by construction they are invariant
c
under charge conjugation, ψM = ψM . Now that we have a form for Majorana spinors, we can easily
write down a Lagrangian for a free particle described by Majorana spinors. By substitution, we
have the Dirac equation applied to Majorana spinors
1
LM = ψ̄M (γ µ i∂µ − m)ψM (14.54)
2
where the factor of 12 is a normalization factor which will play its role shortly. We can expand
this out using the Weyl spinors:
1 † µ m T
χ σ̄ i∂µ χTp (−iσ 2 )σ µ (iσ 2 )i∂µ χ†T χp (−iσ 2 )χp + χ†p iσ 2 χ†T

LM = p −
2 p 2
1 † µ 1 m
= χ σ̄ i∂µ χp + χTp σ̄ µT i∂µ χ†T − (χp · χp + χ̄p · χ̄p )
2 p 2 2
where we made use of the fact that σ 2 σ µ σ 2 = σ̄ µT . We can clean up this expression a bit further
if we use integration by parts on the second term, allowing us to write it as
1 T µT 1
χp σ̄ i∂µ χ†T = − (i∂µ χTp )σ̄ µT χ†T
p
2 2
Then, if we make use of the identity, αT Aβ = −β T AT α, we find that the above is simply equal
to the first term of our Lagrangian density above:

1 1 1
− (i∂µ χTp )σ̄ µT χ†T
p = −(χ †T T
p ) (σ̄ µT T
) − (i∂ χ
µ p
T T
) = σ̄ µ χ†p i∂µ χp
2 2 2
Hence, our Lagrangian density becomes
m
LM = χ†p σ̄ µ i∂µ χp − (χp · χp + χ̄p · χ̄p ) (14.55)
2
We can appreciate why the factor of 21 was introduced in the first place. It yields a Lagrangian
density which can be compared the Lagrangian densities we are used to, allowing for more recog-
nizable terms. For instance, this factor appearing with the mass is important as it allows to us
more easily verify that the term is indeed a mass term [30]. Moreover, it turns out that since we
can easily write out a mass term for a single left-chiral spinor field, the Lagrangian for a massive
left-chiral Weyl fermion is identical to the above Lagrangian. In other words, there is no physical
difference between the theory of massive Weyl fermions and massive Majorana particles. Due to
14.7. BUILDING A SIMPLE SUPERSYMMETRIC LAGRANGIAN 325
this equivalence, we can always write any expression containing Weyl spinors as one with Majorana
spinors and vice versa. As an exercise, the reader will work out some of the relations between
Majorana and Weyl spinors.
14.7 Building a Simple Supersymmetric Lagrangian

In this chapter we are not aiming to learn all of SUSY (even though sometimes it might feel that
way). All we really care about is to become familiar with the general idea of SUSY. For that reason,
we can get away with studying the simplest possible supersymmetric theory there is: one which
does not include mass or interactions. We will discuss these details qualitatively later on, however
they will not be a prime focus in this chapter or text. Moreover, by simple we certainly don’t mean
trivial. Rather, the theory is so simplistic that some might even say it’s boring. Nonetheless, this
simple theory will give us what we need when it comes to understanding the basics of SUSY.
The simplest supersymmetric theory we may consider is composed of two massless fields, one
that is bosonic and one that is fermionic. For our purposes, we will consider a Weyl spinor and
a complex scalar field. Since we are only concerned with left-chiral spinors, we start with the
Lagrangian density
L = ∂µ φ∂ µ φ† + χ† iσ̄ µ ∂µ χ (14.56)
We recognize that this is the correct starting point as it is simply the sum of the usual scalar field
part we are now familiar with from studying quantum scalar fields, and the appropriate part of the
Dirac Lagrangian. We must now introduce SUSY transformations that will leave this Lagrangian
invariant. We could just state the transformations that get the job done, however the process of
coming up with these supersymmetric transformations is important and non-trivial enough that we
should spend a little bit of detail on them.
Let us consider the transformation of the scalar field, as this will be the easier one to deal with
first. We will consider a transformation that it is proportional to some space-time independent,
infinitesimal parameter ζ. By space-time independent we mean that the parameter is not a function
of space-time, thereby vanishing under a derivative, ∂µ ζ = 0. If we think back to our earlier
discussion on gauge transformations, we recognize that the transformation for our scalar field is
global, contrary to local transformations which do depend on space-time. In short, we will only
consider global SUSY transformations. Some call this rigid supersymmetry. Had we decided to
examine local SUSY transformations, where ζ does depend on space-time, we would be forced to
introduce a gauge field that has the properties of a graviton. Theories with local SUSY invariance
are supergravity theories, and are a modern research avenue for theoretical physicists studying
various approaches to quantizing gravity [30].
Let us assume that the variation of the scalar field is proportional to the Weyl spinor χ. The
reason for this choice is because scalar fields, which describe spin-0 bosons, should transform into
fermionic fields. That is the fundamental consequence of SUSY. Therefore, we consider the trans-
formation
φ → φ0 = φ + δφ (14.57)
with δφ ≈ ζχ. Notice we have not written an ’equal’ sign. This is because when we write down
transformations we must make sure that both sides of the equation have the same dimension, and
behave the same under Lorentz transformations. Considering this second requirement first, since φ
is a scalar field, we must build a Lorentz invariant out of ζ and χ. What’s more is since χ is a Weyl
spinor, we are forced to make ζ a Weyl spinor. Hence, the constant parameter of our global SUSY
transformation is infact a Weyl spinor independent of space-time, and, by convention, we declare
ζ is a left-chiral spinor. Since this is the case, we already know of a Lorentz invariant between
left-chiral spinors. It’s simply the spinor dot product between the two spinors, ζ · χ. We are then
tempted to write the SUSY transformation of the scalar field as
δφ = ζ · χ (14.58)
To be certain however, we must check that the dimensions of both sides match up. In natural
units we have c = ~ = 1. Moreover, recall that the action S is the integral over all four dimensional
space of the Lagrangian density L, and is dimensionless (since we are working with natural units).
In such a system, we only have one independent dimension left, that of energy, or mass M . We say
that mass has dimension 1 (M 1 ). On the other hand, since c = 1, length L and time T have the
same dimension of M −1 , since ~ = 1. What this means is for the action to remain dimensionless,
we require that the Lagrangian density have dimension M 4 (coming from the fact that the action
is an integral over four dimensional space-time). Since gradients, ∂µ have dimension M , we can
read off the dimensions of the scalar field φ and spinor field χ from looking at the Klein-Gordon
and Dirac Lagrangian densities, yielding
3
[φ] = M [χ] = M 2
What this means for us is in order to make the dimensions of our SUSY transformation work
1
out, we require that [ζ] = M − 2 . Let us now move on to the SUSY transformation of the Weyl
spinor χ. We would like the transformation to be linear in the infinitesimal parameter ζ and either
φ or φ† , since dealing with non-linear transformations is much too difficult. For reasons which will
become clear shortly, we will use the complex scalar field, allowing us to make our first guess for
the SUSY transformation of χ:
δχ ≈ Cζφ†
where C is some constant yet to be determined. Just as before, we must ensure that this guess
has both sides transforming in the same way under Lorentz transformations. The left hand side
would transform as a left-chiral field, and since φ† is a scalar field, the right hand side too transforms
as a left-chiral spinor. There is a problem however: the dimensions don’t match up: the left hand
3 1
side has a dimension of M 2 , while the dimensions of the right hand side are M 2 , meaning we need
to raise the dimension by M . To keep a linear transformation, we introduce ∂µ which has dimension
M . Therefore, our next guess is
δχ ≈ Cζ∂µ φ†
The dimensions are certainly correct, however introducing ∂µ caused a discrepancy in the Lorentz
properties of both sides. In short, the indices don’t match, meaning we must apply another object
such that it contracts with ∂µ . It must also be an object that is dimensionless. A natural choice
is one of the Pauli matrices σ µ or σ̄ µ . For reasons which will become clear momentarily, we choose
σ̄ µ , and make another guess
δχ ≈ C(∂µ φ† )σ̄ µ ζ
14.7. BUILDING A SIMPLE SUPERSYMMETRIC LAGRANGIAN 327
But this is still incorrect! Recall that σ̄ µ ∂µ χ transforms like a right-chiral spinor. This is exactly
our same issue since, although the derivative is acting on φ† , it does not actually affect the behavior
of the expression under a Lorentz transformation. Meaning that the right-hand side transforms like
a right-chiral spinor while the left hand side still transforms like a left-chiral spinor. Luckily, this
problem can be easily resolved.
Remember that we showed −iσ 2 η †T transforms as a left-chiral spinor. Applying this to our latest
guess, we have
δχ = −iσ 2 (C(∂µ φ† )σ̄ µ ζ)†T

Since the operation of †T on anything other than a quantum field, including ζ, is nothing more
than simple complex conjugation ∗, the above becomes
δχ = −iσ 2 C ∗ (∂µ φ)σ̄ µ∗ ζ ∗ = −C ∗ (∂µ φ)iσ 2 σ̄ µ∗ ζ ∗

Then, if we use the fact that σ̄ µ∗ = σ̄ µT and (σ 2 )2 = I, the above becomes
−C ∗ (∂µ φ)iσ 2 σ̄ µT σ 2 σ 2 ζ ∗
Lastly, using σ 2 σ̄ µT σ 2 = σ µ we find that the correct SUSY transformation for χ is
δχ = −C ∗ (∂µ φ)σ µ iσ 2 ζ ∗ (14.59)
It still remains to be seen whether we may choose a value for C. The easiest way to go about
doing so is ensure that our Lagrangian is indeed invariant under the SUSY transformations we have
developed, (14.58) and (14.59). Let us vary our Lagrangian density, (14.56):
δL = ∂µ (δφ)∂ µ φ† + ∂µ φ∂ µ δφ† + (δχ† )iσ̄ µ ∂µ χ + χ† iσ̄ µ ∂µ δχ
= ∂µ (δφ)∂ µ φ† + ∂µ φ∂ µ (δφ)† + (δχ)† iσ̄ µ ∂µ χ + χ† iσ̄ µ ∂µ δχ

where we made use of δφ† = (δφ)† and δχ† = (δχ)† . Taking Hermitian conjugates of our
transformations (14.58) and (14.59), we find that
(δφ)† = χ̄ · ζ̄ = χ† (iσ 2 )ζ ∗ (δχ)† = C(∂µ φ† )ζ T iσ 2 σ µ (14.60)

If substitute everything in, we find that the variation of the Lagrangian is
δL = (∂ µ χ† )iσ 2 ζ ∗ ∂µ φ−(∂ µ φ† )ζ T (iσ 2 )∂µ χ+C(∂ν φ† )ζ T (iσ 2 )σ ν iσ̄ µ ∂µ χ−C ∗ χ† iσ̄ µ σ ν (∂µ ∂ν φ)(iσ 2 ζ ∗ )
As an exercise, the reader will prove that
σ µ σ̄ ν ∂µ ∂ν χ = ∂ µ ∂µ χ = χ (14.61)
which will help us show that the Lagrangian is invariant under our SUSY transformations. To
see this explicitly, consider the sum of the first and last term of our variation above:
(∂ µ χ† )iσ 2 ζ ∗ ∂µ φ − iC ∗ χ† iσ 2 ζ ∗ φ
where we moved φ around to the end of the second term since it is not a matrix quantity. We
can make these two terms cancel if we use integration by parts, allowing us to get the derivatives
to act on the same fields. Applying integration by parts to the first term, keeping in mind that ζ
is space-time independent, we find that the above becomes
−χ† iσ 2 ζ ∗ φ − iC ∗ χ† iσ 2 ζ ∗ φ
At this point it is easy to see that indeed these two terms cancel as long as C = −i. Completing
a similar analysis shows that the second and third term of the variation also cancel when C = −i.
In summary, we have introduced SUSY transformations
δφ = ζ · χ δφ† = ζ̄ · χ̄ (14.62)
δχ = −i(∂µ φ)σ µ iσ 2 ζ ∗ δχ† = −i(∂µ φ† )ζ T iσ 2 σ µ (14.63)

that leave the Lagrangian
L = ∂µ φ∂ µ φ† + χ† iσ̄ µ ∂µ χ
invariant. We have completed our first supersymmetric theory! However, we are not done yet.
As we will see, when we determine the supersymmetric algebra, we will be forced to introduce a
new field to maintain consistency. Let us proceed to examining the supersymmetric algebra now.
14.8 A Review of Charges

As we have done so far in this text, when we have a Lagrangian that possesses a symmetry, we
find the charges or generators that generate this symmetry. What ends up mattering most is
the algebra of these generators: the commutation relations or anticommutation relations of the
generators (when an algebra involves anticommutators, it is often called a graded Lie algebra).
Before we get to examining the algebra of SUSY, let us first review some more elementary results
of charges and their resulting algebra. More specifically, let us review the algebra of the Poincaré
group. We will examine charges from a different perspective, in part because the SUSY algebra
deals with spinors, something which we are not yet used to.
In a sense, the fundamental object in quantum mechanics is the expectation value, as it leads to
the physical results of the system. That is, suppose we have two states |ai and |bi, and consider
the scalar field φ. If a transformation is a symmetry of the theory, then it must leave the physics of
the system invariant. In other words, we must have the expectation value of φ between these two
states be invariant:
ha|φ|bi
Any transformation which does not do so, is not a symmetry of the theory. For example, space-
time transformations, φ0 (x) = φ(x + δx) leave the classical scalar field unchanged and therefore the
expectation value is unaffected as well
ha|φ0 (x)|bi = ha|φ(x)|bi

14.8. A REVIEW OF CHARGES 329
The point is, physically meaningful transformations preserve the physics of the system, or from
a quantum mechanical perspective, the norm of the states. In the last chapter we saw that such
symmetries can be written in the form of unitary operators U † = U −1 . For example,
φ0 (x) = U φ(x)U † (14.64)

From here we can introduce the charges generations the actual transformations. We will assume
that the transformations are parameterized by a finite set of infinitesimal parameters 1 , 2 , ...n , in
which case we also have a finite set of charges, Q1 , Q2 , ...Qn . The charges are defined by [30]
U ≡ exp(±ii Qi ) = exp(±i · Q) (14.65)

Substituting this into (14.64) and expanding to first order, we find
φ0 (x) ≈ (1 ± i · Q)φ(x)(1 ∓ i · Q) = φ(x) ± i[ · Q, φ(x)]

We are more used to writing the space-time translations as φ0 (x) = φ(x) + δφ(x). If we use this
instead, by comparison we find that
±[ · Q, φ] = −iδφ (14.66)

Moreover, we see by combining both forms of space-time symmetries, we have
δφ = φ(x + δx) − φ(x)

Let’s apply this result to a simple case. Consider a Lagrangian density L which only depends
on a single scalar field φ. Indeed the action is invariant under the space-time translation φ(xµ ) →
φ0 (xµ ) = φ(xµ + aµ ) where aµ is a constant displacement 4-vector. Using the above, we find that
the variation of the field δφ under this transformation is simply
δ(xµ ) = φ(xµ + aµ ) − φ(xµ ) ≈ aµ ∂µ φ (14.67)

Since we are dealing with a four-vector aµ , we have have four parameters, and therefore introduce
four charges, which we denote by P µ . Hence, the unitary operator motivated by (14.65) takes the
form
U ≡ exp(iaµ Pµ ) (14.68)
From which we find
aµ [Pµ , φ] = −iaµ ∂µ φ (14.69)

Moreover, since this must hold for all displacement four-vectors aµ , we may conclude that
[Pµ , φ] = −i∂µ φ (14.70)

As we are already familiar with these from a previous chapter, we call Pµ the generators of space-
time translations. As a second, slightly more complicated example, consider a theory containing
a single complex scalar field described by a Lagrangian density which is invariant under the U(1)
transformation
φ → eiα φ
where α is some constant. We call this type of symmetry and internal symmetry, one that is
not related to a specific change of coordinates, but rather an equivalence between different fields
at the same space-time point. The variation of the field is easily recognized to be δφ = iαφ. Then,
writing the unitary operator as U = exp(iαQ), we find that
[Q, φ] = φ (14.71)
The reader is invited to work out a more complicated example where they first consider Lorentz
transformations and work to prove
[Jµν , φ] = i(xν ∂µ − xµ ∂ν )φ (14.72)
As we noted in the previous chapter, the physics of the system really lies within the commutation
or anticommutation relations, the algebra. Afterall, it is the algebra which yields the physically
meaningful symmetries which leave the expectation value of fields invariant. For that reason, we
try hard to determine the algebra. As we will see, that are two different ways to go about doing
this. The first approach is to find the field transformations, which will lead us the the currents and
hence charges of the quantum fields, in which case we may then construct the algebra.
We consider the other approach now. In the case of space-time symmetries, the charges may also
be explicitly represented as differential operators acting on the quantum fields. This representation
is defined by
φ(x0 ) ≡ (±ii Q̂i )φ(x) ≈ φ(x) ± ii Q̂i φ(x) (14.73)

As a quick example, consider again space-time translations, where we use P̂µ to denote the
differential operator representation of the charges associated with the symmetry. The unitary
operator is then
Û = exp(−iaµ P̂µ ) (14.74)

which yields
φ(x) + aµ ∂µ φ(x) = φ(x) − iaµ P̂µ φ(x)

allowing us to correctly identify
P̂µ = i∂µ (14.75)

It is imperative to distinguish between the two generators Pµ and P̂µ . The first is a quantum
field operator, while the second is a differential operator. For completeness, let us further define

i µν ˆ
φ(x0 ) ≡ exp ω Jµν φ(x) (14.76)
2
Notice then, that if we make a change in coordinates xµ + ω µν xν , we see
1
φ(xµ + ω µν xν ) ≈ φ + ω µν (xν ∂µ − xµ ∂ν )φ
2
14.8. A REVIEW OF CHARGES 331
If we then set this equal to φ(x0 ), we find that

i µν ˆ i
exp ω Jµν φ(x) ≈ φ + ω µν Jˆµν φ
2 2
which leads us to
Jˆµν = i(xµ ∂ν − xν ∂µ ) (14.77)

as expected.
To summarize, this approach first uses the transformations of coordinates to represent the charges
as differential operators, which will lead us to the algebra. This approach is indeed relevant to
supersymmetry as SUSY and Lorentz transformations are connected via an extension of space-
time called superspace, which contain Grassmann coordinates as well as the common space-time
coordiantes. We will discuss this in more depth later on as it is the approach most of the literature
and string theorists use.
We can also use field transformations to determine the algebra of the generators without explicitly
finding the charges. For this approach we will again assume that the charges are viewed as quantum
field operators, and also consider two successive field transformations. For simplicity, we assume
that we are dealing with a scalar field φ. Moreover, to distinguish between the field transformations,
we attribute to the first transformation parameters αi , and to the second transformation, parameters
βi . Motivated by (14.64), we have
Uβ Uα φUα† Uβ† = exp(iβ · Q)exp(iα · Q)φexp(−iα · Q)exp(−iβ · Q)
≈ φ + i[β · Q, φ] + i[α · Q, φ] − [β · Q, [α · Q, φ]] + ...

where we are neglecting terms of order α2 and β 2 . Alternatively, by defintion, all we have really
done here is write a variation with parameter of α on a scalar field followed by a second variation
with parameter β. Meaning that we may also write
Uβ Uα φUα† Uβ† = δβ δα φ
Notice then that if we consider the commutator of these two transformations, δβ δα − δα δβ , we
find using our above result that this is just
Uβ Uα φUα† Uβ† − Uα Uβ φUβ† Uα† ≈ −[β · Q, [α · Q, φ]] + [α · Q, [β · Q, φ]]

Therefore,
δβ δα − δα δβ = −[β · Q, [α · Q, φ]] + [α · Q, [β · Q, φ]] (14.78)

As the reader will show, if we expand this out carefully, paying close attention to the order of
the elements we find that half of the terms cancel, leaving us with
δβ δα − δα δβ = [[α · Q, β · Q], φ] (14.79)

To become acquainted with this formula, lets us consider the case of the algebra beteen the
charges P µ . Looking at the left side first, we apply a second translations to δa φ(xµ ) with an
infinitesimal displacement bµ . By simple plug and chug we find that this is just
δb δa φ(xµ ) = δb (φ(xµ + aµ ) − φ(xµ )) = δb φ(xµ + aµ ) − δb φ(xµ )
= φ(xµ + aµ + bµ ) − φ(xµ + aµ ) − φ(xµ + bµ ) − φ(xµ + bµ ) + φ(xµ )

We notice that if we exchanged aµ with bµ the result would be equivalent, leading us to conclude
δb δa φ(xµ ) − δa δb φ(xµ ) = 0
Let’s move on to the right hand side.The unitary transformations are written as
Ua = exp(iaµ Pµ ) Ub = exp(ibν Pν )
We compare to (14.79) and notice that we simply replace α · Q with aµ Pµ and replace β · Q with
ν
b Pν , yielding
[[aµ Pµ , bν Pν ], φ] = aµ bν [[Pµ , Pν ], φ] = 0 ⇒ [[Pµ , Pν ], φ] = 0

An obvious solution is that the commutator between the generators Pµ and Pν is zero, [Pµ , Pν ] =
0. It could however be more general than that. For example, it could be that
[Pµ , Pν ] = Cηµν
where C is some constant. It’s easy to see that this choice would also solve the commutator in
question. However, remember that [Pµ , Pν ] = −[Pν , Pµ ]. This is effectively saying that if we switch
the indices around in our commutator, we pick up a minus sign. Remember however that the metric
is symmetric under exchange in indices, concluding that the more general solution forces C = 0.
Hence,
[Pµ , Pν ] = 0 (14.80)
as hoped for. The reader will try their luck with the Lorentz generators Jµν . What is crucial
to point out is that we never once knew the explicit form of the generators Pµ , yet we were still
able to construct the algebra, i.e. the commutation relations of the generators. Such will be the
approach when we search for the algebra of the supersymmetric charges, which we discuss in the
next section.
14.9 The Supersymmetric Algebra

We will now make use of (14.79) to find the algebra of supersymmetric charges. First of all, since we
are using a two component spinor ζ and its complex conjugate ζ ∗ , we have a total of four charges,
denoted as Q1 , Q2 , Q†1 , Q†2 , and may form a Weyl spinor, which we shall simply label as Q, and
it has a Hermitian conjugate Q† . These are our supersymmetric charges which are often called
supercharges.
14.9. THE SUPERSYMMETRIC ALGEBRA 333
We require that the argument in the exponential of our unitary operator be Lorentz invariant.
What’s more is we may also choose that Q is a left-chiral spinor, in which case we have the two
possible Lorentz invariants
Q · ζ = Q(−iσ 2 )ζ Q̄ · ζ̄ = Q† iσ 2 ζ ∗ (14.81)
Using this in our unitary operator U that generates SUSY transformations is given by
Uζ = exp(iQ · ζ + iQ̄ · ζ̄) (14.82)

If we then apply (14.66), and our SUSY transformations given in (14.62) and (14.63), we find
[ζ · Q + ζ̄ · Q̄, φ] = −iζ · χ [ζ · Q + ζ̄ · Q̄, χ] = −i(∂µ φ)σ µ σ 2 ζ ∗ (14.83)

which imply
[ζ · Q, φ] = −iζ · χ [ζ̄ · Q̄, χ] = −i(∂µ φ)σ µ σ 2 ζ ∗ (14.84)
[ζ̄ · Q̄, φ] = [ζ · Q, χ] = 0 (14.85)

Just as we did near the end of the last section, let us consider two successive SUSY trans-
formations in the same way as before. This time β as the infinitesimal parameter of the second
transformation, which is generated by Uβ = exp(iQ · β + iQ̄ · β̄). If we apply (14.79) to a scalar
field φ our supersymmetric Lagrangian in (14.56), we find
δβ δζ φ − δζ δβ = [[Q · ζ + Q̄ · ζ̄, Q · β + Q̄ · β̄], φ]
= [[Q · ζ, Q · β], φ] + [[Q · ζ, Q̄ · β̄], φ] + [[Q̄ · ζ̄, Q · β], φ] + [[Q̄ · ζ̄, Q̄ · β̄], φ]
Let us first examine the right hand side and the commutators between the charges first. Using
our results of the spinor dot product (14.28) and (14.34), the first commutator is just
[Q · ζ, Q · β] = [Q(−iσ 2 )ζ, Q(−iσ 2 )β] = −[Qa (σ 2 )ab ζb , Qc (σ 2 )cd βd ]

As a warning to the reader, we remind ourselves that the ζi and βi are Grassmann numbers
while the Q are Grassmann operators, meaning that as we pass them through each other, we will
pick up extra minus signs as they anticommute with each other. Therefore, the above becomes
−(σ 2 )ab (σ 2 )cd (Qa ζb Qc βd − Qc βd Qa ζb ) = (σ 2 )ab (σ 2 )cd ζb βd (Qa Qc + Qc Qa )
= (σ 2 )ab (σ 2 )cd ζb βd {Qa , Qc } (14.86)

Rather than commutation relations, we have instead found that anticommutators between the
supercharges arise. However this could be expected as we are dealing with fermionic variables
instead of ordinary commuting numbers. Using the same method, the three other commutators
are:
[Q · ζ, Q̄ · β̄] = −(σ 2 )ab (σ 2 )cd ζb βd∗ {Qa , Q†c } (14.87)

[Q̄ · ζ̄, Q · β] = −(σ 2 )ab (σ 2 )cd ζb∗ βd {Q†a , Qc } (14.88)
[Q̄ · ζ̄, Q̄ · β̄] = (σ 2 )ab (σ 2 )cd ζb∗ βd∗ {Q†a , Q†c } (14.89)
In summary, we have
δβ δζ φ − δζ δβ = [O, φ]
where we have defined the operator O as
O = (σ 2 )ab (σ 2 )cd (ζb βd {Qa , Qc } − ζb βd∗ {Qa , Q†c } − ζb∗ βd {Q†a , Qc } + ζb∗ βd∗ {Q†a , Q†c }) (14.90)
Let’s take a break from the right hand side for now and move on to the left hand side. We leave
it as an exercise for the reader to show that
δβ δζ φ = −iζ T σ 2 σ µ σ 2 β ∗ ∂µ φ = iβ † σ̄ µ ζ∂µ φ (14.91)

and
δζ δβ φ = −iβ T σ 2 σ µ σ 2 ζ ∗ ∂µ φ = iζ † σ̄ µ β∂µ φ (14.92)

Combining both results yields
δβ δζ φ − δζ δβ φ = −i(ζ T σ 2 σ µ σ 2 β ∗ − β T σ 2 σ µ σ 2 ζ ∗ )∂µ φ (14.93)

Using ∂µ φ = i[Pµ , φ], this just becomes
δβ δζ φ − δζ δβ φ = (ζ T σ 2 σ µ σ 2 β ∗ − β T σ 2 σ µ σ 2 ζ ∗ )[Pµ , φ]
= (ζ T σ 2 σ µ σ 2 β ∗ − β T σ 2 σ µ σ 2 ζ ∗ )Pµ , φ

Our task now is to compare

T 2 µ 2 ∗
(ζ σ σ σ β − β T σ 2 σ µ σ 2 ζ ∗ )Pµ , φ = [O, φ]

Since the operator O is written in terms of spinor components, we must also do the same for the
left hand side of the above commutator, thereby allowing us to compare the two sides more easily.
Therefore, we make note of the fact that we may write
ζ T σ 2 σ µ σ 2 β ∗ = ζb (σ 2 )ba (σ µ )ac (σ 2 )cd βd∗

and
β T σ 2 σ µ σ 2 ζ ∗ = βd (σ 2 )dc (σ µ )ca (σ 2 )ab ζb∗

Moreover, recall that σ 2 is antisymmetric. Hence, (σ 2 )ba = −(σ 2 )ab and (σ 2 )dc = −(σ 2 )cd . Also,
if we move ζb∗ to βd , the above two expressions become
ζ T σ 2 σ µ σ 2 β ∗ = −ζb βd∗ (σ 2 )ab (σ 2 )cd (σ µ )ac

and
β T σ 2 σ µ σ 2 ζ ∗ = ζb∗ βd (σ 2 )ab (σ 2 )cd (σ µ )ca

Using both these results, we are finally able to identify
O = −(σ 2 )ab (σ 2 )cd (ζb βd∗ (σ µ )ac + ζb∗ βd (σ µ )ca )Pµ (14.94)
Now, if we cancel out the components of this expression with the components of (14.90), we are
left with
ζb βd {Qa , Qc } − ζb βd∗ {Qa , Q†c } − ζb∗ βd {Q†a , Qc } + ζb∗ βd∗ {Q†a , Q†c } =
− (ζb βd∗ (σ µ )ac + ζb∗ βd (σ µ )ca )Pµ
(14.95)
Since α and β and their complex conjugates are completely arbitrary, we are led to four anti-
commutation relations [30]:
{Qa , Qb } = 0 {Q†a , Q†b } = 0 (14.96)
{Qa , Q†c } = (σ µ )ac Pµ {Q†a , Qc } = (σ µ )ca Pµ (14.97)

These last two anticommutators are of particular interest as it shows that the anticommutator
of two supercharges yields space-time translations, further indicating the deep connection SUSY
has with space-time transformations.
We are close now, but not quite done. To complete the supersymmetric algebra, we must work
out the commutators of the supercharges with the Poincaré generators P µ and J µν . It’s easy to
see that
[Qa , Pµ ] = 0 [Q†a , Pµ ] = 0 (14.98)

The more interesting case is the commutator between the supercharge Qa and the Lorentz
generator Jµν . We won’t detail it here however as, though relatively straightforward, it is a rather
tedious calculation. For a full explanation, the reader is pointed to review Patrick Labelle’s text.
The result is [30]
[Qa , Jµν ] = (σµν )ab Qb (14.99)

where we have defined
i
σµν ≡ (σµ σ̄ν − σν σ̄µ ) (14.100)
4
We have indeed found the the supersymmetric algebra as applied to the scalar field φ. We must
now check to make sure that this algebra also holds for the spinor χ. If the algebra is still satisfied,
we will know that we have come up with the correct algebra for our SUSY transformations.
Consider the variation of the spinor fields that we came up with in the previous section. We find
then that two successive such variations yield
δβ δζ χ = δβ (−iσ µ iσ 2 ζ ∗ ∂µ φ) = (σ µ σ 2 ζ ∗ )∂µ (δβ ) = (σ µ σ 2 ζ ∗ )β · ∂µ χ

where
β · ∂µ χ = β T (−iσ 2 )∂µ χ
Of course, we really have an identity for each component of χ. This means we may write
δβ δζ χa = (σ µ σ 2 ζ ∗ )a β · ∂µ χ
To simplify this expression, let us define α ≡ σ µ σ 2 ζ ∗ , in which case the above simply becomes
δβ δζ χa = αa β · ∂µ χ
Our task at hand is to verify that the commutator of two SUSY transformations on the spinor
are of the same form as before. Using the identity [30]
αa (β · γ) + βa (γ · α) + γa (α · β) = 0 (14.101)
where α, β, and γ are obviously spinors. If we use this, by defining ∂µ χ = γ, we find that
αa β · ∂µ χ = −βa (∂µ χ · α) − ∂µ χa (α · β)
Moreover, since the dot product is commutative, this just becomes
αa β · ∂µ χ − ∂µ χ · αβa − β · α∂µ χa
Using this, we may write the two variations on χa as
δβ δα χa = −(∂µ χ) · αβa − α · β∂µ χa = −(∂µ χ) · αβa − β · α∂µ χa
= −((∂µ χ)T (−iσ 2 )σ µ σ 2 ζ ∗ )βa − (β T (−iσ 2 )σ µ σ 2 ζ ∗ )∂µ χa
= i((∂µ χ)T σ 2 σ µ σ 2 ζ ∗ )βa + i(β T σ 2 σ µ σ 2 ζ ∗ )∂µ χa
If we make use of αT Aβ = −β T AT α and σ 2 σ µ σ 2 = σ̄ µT , we find the above simplifies to
δβ δα χa = i((∂µ χ)T σ̄ µT ζ ∗ )βa + i(β T σ̄ µT ζ ∗ )∂µ χa
= −i(ζ † σ̄ µ ∂µ χ)βa − i(ζ † σ̄ µ β)∂µ χa (14.102)

On the other hand, if we switched α and β, we would have found
δζ δβ χa = −i(β † σ̄ µ ∂µ χ)ζa − i(β † σ̄ µ ζ)∂µ χa (14.103)

Therefore, the commutator between the two variations is simply
(δβ δζ − δζ δβ )χa = −i(ζ † σ̄ µ ∂µ χ)βa + i(β † σ̄ µ ∂µ χ)ζa − i(ζ † σ̄ µ β − β † σ̄ µ ζ)∂µ χa (14.104)
If we compare to (14.93), we notice that the last term in the above expression is what we want.
Unfortunately, the first two terms present are not terms we want. We do notice that if iσ̄ µ ∂µ χ = 0,
these unwanted terms drop out. This equation is actually a familiar one: it’s called the Weyl
equation. Hence, if the spinor satisfies its own equation of motion (objects we call on-shell spinors),
then we recover the desired SUSY algebra. However, off -shell spinors, those which do not satisfy the
Weyl equation, would not yield the SUSY algebra we seek, we therefore say that the SUSY algebra
is non-closed for off-shell spinors. We would like to think that we could live with this, however,
if we do not include off-shell spinors, the SUSY transformations will actually become non-linear,
which nobody wants. For that reason, we must modify our theory we constructed in the previous
section with an auxiliary field.
Remember, SUSY requires that the bosonic and fermionic degrees of freedom are equal (as they
change into one another). Indeed the number of on-shell degrees of freedom match: two degrees of
freedom from the complex scalar field, and two from the Weyl spinor. However, the off-shell degrees
of freedom between the bosonic field and spinor field do not match (we only have one bosonic field
with two on-shell degrees of freedom). This suggests that we must add another bosonic field that
provides two degrees of freedom off-shell, but no on-shell degrees of freedom. We call such a field
an auxiliary field. Therefore, let us add to our original theory a complex scalar auxiliary field F .
To make sure F does not possess any on-shell degrees of freedom, we assume that the equation of
motion of both F and F † vanish. Since we seek to add a real term to our Lagrangian density, the
simplest such possibility with F and F † is F F † . Hence, our modified Lagrangian density takes the
form
L = ∂µ φ∂ µ φ† + χ† iσ̄ µ ∂µ χ + F F † (14.105)
If we use the same logic as we did before to construct our SUSY transformations for the scalar
and spinor fields, we find that the transformation of the auxiliary field F is
δF = Kζ † σ̄ µ ∂µ χ (14.106)
where K is some constant that shall be determined shortly. Moreover, taking the Hermitian
conjugate yields
δF † = K ∗ (∂µ χ)† σ̄ µ ζ (14.107)

This is fine, however we entered the auxiliary field into the Lagrangian density as the product
F F † . Therefore, we must check that this product indeed leaves the Lagrangian density invariant
under our given SUSY transformations. First of all, the variation δF F † is
δF F † = Kζ ∗ σ̄ µ (∂µ χ)F † + K ∗ (∂µ χ† )σ̄ µ ζF

which, as one can check, does not leave our modified Lagrangian invariant. One way to fix
the problem is to modify our SUSY transformation of either the scalar field φ or χ, in hopes of
canceling the extra terms the variation of the product of the auxiliary term gives. Keeping this in
mind, notice that the variation of the spinor kinetic energy is
δ(χT iσ̄ µ ∂µ χ) = (δχ† iσ̄ µ ∂µ χ + χ† iσ̄ µ ∂|mu (δχ)

What’s more is if we perform integration by parts on the second term of δF F † we find
δ(F F † ) = (K ∗ ζF )† σ̄ µ ∂µ χ − χ† σ̄ µ ∂µ (K ∗ ζF )
Then, if our variation of χ is modified to
δχ = −iσ µ (iσ 2 ζ ∗ )∂µ φ − iK ∗ ζF

We then notice that the variation of the spinor kinetic energy will cancel the extra terms of from
the variation of F F † . Notice that we have yet to impose any constraints on the constant K. But
why make things more complicated than they already? We make the variation of χ slightly simpler
by letting K = −i [30]. All in all, we have that the Lagrangian
L = ∂µ φ∂ µ φ† + χ† iσ̄ µ ∂µ χ + F F †
is invariant under the SUSY transformations
δφ = ζ · χ δχ = −iσ µ (iσ 2 ζ ∗ )∂µ φ + F ζ δF = −iζ † σ̄ µ ∂µ χ (14.108)

Let’s not forget why we had to introduce the auxiliary field in the first place. We noted that
our SUSY algebra would not close for spinors that did not satisfy the Weyl equation. In order to
have an algebra which closed under both spinors which did and did not satisfy the Weyl equation,
we were forced to modify our Lagrangian by adding an auxiliary field, and then modify our SUSY
transformation of the left-chiral spinor χ such that the Lagrangian was still invariant by our SUSY
transformations. We have done this. Now one must still check that the SUSY algebra is the same
for each field φ, χ, and F . That is, it still remains to be verified that
δβ δζ A − δζ δβ A = −i(ζ † σ̄ µ β − β † σ̄ µ ζ)∂µ A (14.109)

where A is φ, χ, or F . In short, the answer is yes, each field does satisfy the SUSY algebra,
and therefore, the introduction of our auxiliary field was not in vain: the algebra closes for off-shell
fields as well. To make this concrete, let us examine the case when we are dealing with the spinor
field χ. Using the modified spinor transformation, we have that
δβ δζ χ = δβ (−iσ µ (iσ 2 ζ ∗ (∂µ φ) + F ζ) = σ µ σ 2 ζ ∗ ∂µ (δβ φ) + (δβ F )ζ
= σ µ σ 2 ζ ∗ β · ∂µ χ − iβ † σ̄ µ (∂µ χ)ζ
Using the same logic as we did to arrive to (14.102) gives us
δβ δζ χa = −i(ζ † σ̄ µ ∂µ χ)βa − iζ † σ̄ µ β∂µ χa − i(β † σ̄ µ ∂µ χ)ζa (14.110)

Similarly,
δζ δβ χa = −i(β † σ̄ µ ∂µ )ζa − i(β † σ̄ µ ζ)∂µ χa − i(ζ † σ̄ µ ∂µ χ)βa (14.111)

Putting everything together we have
14.10. CLASSIFYING QUANTUM STATES USING ALGEBRA 339
(δβ δζ − δζ δβ )χa = −i(ζ † σ̄ µ ∂µ χ)βa − iζ † σ̄ µ β∂µ χa − i(β † σ̄ µ ∂µ χ)ζa

(14.112)
+ i(β † σ̄ µ ∂µ )ζa + i(β † σ̄ µ ζ)∂µ χa + i(ζ † σ̄ µ ∂µ χ)βa
= i(β † σ̄ µ ζ − ζ † σ̄ µ β)∂µ χa
which takes the form we had hoped for. Using similar methods, one can show that the scalar
field φ and auxiliary field F share the same form. Hence, the SUSY algebra, the anticommutation
relations of the supercharges, we derived earlier holds in general.
14.10 Classifying Quantum States Using Algebra

Before we get to examining the physical consequences of the algebra of the supercharges, let us
first how we can classify quantum states (in particular, single particle states) using the Poincaré
algebra. For this we must introduce the concept of Casimir operators. These operators are defined
to be operators which commute with all of the generators of a group, allowing us to use their
eigenvalues to classify states since they share the same basis with the generators. Specific to the
Poincaré group, one operator which commutes with both Pµ and Jµν is the momentum squared
squared operator P µ Pµ . It is rather easy to convince yourself that certainly the momentum squared
operator commutes with Pµ , and a brief calculation shows that the momentum squared operator
shows that it commutes with Jµν . Thus, since P µ Pµ commutes with both generators of the Poincaré
group, P µ Pµ as the same basis as the generators.
The second Casimir operator for the Poincaré is built out of the so-called Pauli-Lubanski operator
Wµ defined as
1 µνρσ
Wµ ≡ Jρσ Pν (14.113)
2
where µνρσ is the totally antisymmetric Levi Civita tensor. Immediately we find that
1 µνρσ
W µ Pµ =
Jρσ Pν Pµ = 0
2
Since if exchange in the indices of Pµ is symmetric, however, exchange of the indices in the
Levi-Civita tensor introduces a minus sign, meaning that W µ Pµ = −W µ Pµ = 0. Moreover, notice
that
1 µνρσ
Pµ W µ = W µ Pµ + [Pµ , W µ ] = [Pµ , W µ ] = [Pµ , Jρσ ]Pν
2
i
= − µνρσ (ηνσ Pρ − ηµρ Pσ )
2
Again, since the metric is symmetric, if we exchange its indices we don’t pick a minus sign.
When we do the same with the Levi-Civita tensor, we pick up an extra minus sign, this time
yielding [Pµ , W µ ] = 0. We won’t go through it here, but it is rather straightforward to show that
the operator Wµ W µ commutes with the generators of the Poincaré group and hence has the same
basis as the generators.
To gain physical insight into the Pauli-Lubanski operator, consider massive particles. In this
case, we may choose to work in the rest frame of the particle, where the four-momentum is pµ =
pµ = (m, ~0) Using pµ to denote the eigenvalues of Pµ . Therefore, writing the momentum state of
the particle as |pi, we find that
Pµ |pi = p0 |pi = m|pi

If we apply the Pauli-Lubanski operator to the same state, we find
1 µνρσ m
W µ |pi = Jρσ Pν |pi = µ0ρσ Jρσ |pi
2 2
Since the zeroeth component, the time component, has already been used, we find the only
indices left over in the Levi-Civita tensor are spatial indices, implying that
W 0 |pi = 0
and
m i0jk m m
W i |pi = Jjk |pi = − 0ijk Jjk |pi = ijk Jjk |pi
2 2 2
where we have made the identification 0ijk = −ijk , and where we have used Latin symbols
i, j, k to denote spatial indices. A tedious excercise shows that the operator 12 ijk Jjk is the total
angular momentum of the particle [30]. That is,
1 ijk
Jjk = Li + S i (14.114)
2
where Li is the usual angular momentum and S i is the spin of the particle. Therefore, we have
that
W i |pi = m(S i + Li )|pi

Of course, if we are working in the particle’s rest frame, there is no orbital angular momentum,
only spin, thus in our case
W i |pi = mS i |pi
All in all, notice that the Casimir operator Wµ W µ acting on the single particle state gives us
~ 2 |pi = −m2 s(s + 1)|pi

Wµ W µ |pi = W 0 W 0 − W i W i |pi = −m2 S (14.115)
Therefore, we see that this second Casimir operator, at least for massive particles, yields the
spin of the particle. Before moving on to the case of a massless particle, in Quantum Mechanics we
are used to adding the z-component of the spin operator to the set of observables we work with.
Notice then, the zth component of the Pauli-Lubanksi operator is simply
W 3 = m(Lz + Sz ) (14.116)
Again, working in the rest frame means that Lz = 0, and therefore W 3 = mSz . With the
addition of this operator we may classify massive particle states, denoted by |p, s, sz i, in their rest
frame as
14.11. CLASSIFYING STATES WITH SUPERCHARGES 341
P µ |p, s, sz i = m|p, s, sz i Wµ W µ |p, s, sz i = −m2 s(s + 1)|p, s, sz i W 3 |p, s, sz i = msz |p, s, sz i

(14.117)
Let’s move on to the slightly more complex case of massless particles. The reason for the
additional complexity is because we cannot simply use their rest frame to analyze the particle
states. This is because all of the known massless particles move at the speed of light, which, in a
sense, don’t have a frame that they are at rest in. Instead however, we may choose a rest frame such
that the four-momentum is given by pµ = (E, 0, 0, E), where we immediately see that pµ pµ = 0 a
null vector, which describes particles moving at the speed of light. Let us denote the state associated
with this four-momentum by |pi0 . This means that
P µ |pi0 = (E, 0, 0, E)|pi0 (14.118)

In components, this means P 0 = P 3 = E. As it turns out, the two Casimir operators Pµ P µ , Wµ W µ
yield zero when they act on a massless particle state. Moreover, since Wµ W µ yields zero when ap-
plied to a massless particle, and using the fact that W µ Pµ = 0, we have that
W µ |pi0 = hP µ |pi0 (14.119)

where h is a proportionality constant which will have its physical significance revealed momen-
tarily. As an exercise, the reader will be asked to show that the action of W µ on the massless
particle state is
W µ |pi0 = (hE, 0, 0, hE)|pi0 (14.120)

The physical significance of h can now easily be extracted. Consider the cas when µ = 3, in
which case
W 3 |pi0 = E(Sz + Lz )|pi0 = ESz |pi0

where we have assumed the particle has no orbital angular momentum. Using this, upon com-
parison with (14.120), we identify the constant h to be h = sz . However, since our state, by
construction has its momentum along the z-direction, we may replace this with
h = ~s · p~ (14.121)
Now, remember in our study on quantum field theory, we had defined the helicity of a particle
as the projection of the spin of the particle along the direction of the momentum. Thus, h is simply
the helicity of the particle! What this tells is us the massless particle states can be entirely specified
by their energy and their momentum.
14.11 Classifying States with Supercharges

In this section we will explore the physical details of the SUSY algebra. We will only consider
massless supersymmetric states, classifying them using the supercharges we came up with earlier.
The reason we will only consider massless states is because these are the only states that are
relevant in the Minimal Supersymmetric Standard Model (MSSM), a supersymmetric extension

of the Standard Model. In the MSSM, just like the SM, massless particles gain mass through the
Higg’s mechanism, and therefore all one really needs are the massless states. Though we won’t study
it here in this text, supersymmetry breaking leads to the massless states of the MSSM becoming
massive. With that, let us move forward.
We already noted that the supercharges commute with the four-momentum operator P µ , however
they do not commute with the Lorentz generators Jµν . Let us compute the commutators of the
supercharges with the Pauli-Lubanski operator W µ . Again, since we considering massless particles,
we really only need to consider the time component of W µ , since W 0 |pi0 = W 3 |pi0 , and W 1 |pi0 =
W 2 |pi0 = 0. Notice then
1 0νρσ 1 1
[Qa , W 0 ] = [Qa , Jρσ Pν ] = 0νρσ [Qa , Jρσ ]Pν = 0νρσ (σρσ )ab Qb Pν
2 2 2
where we have made use of (14.99). For the sake of simplicity, we will drop the indices, keeping
in mind all along that the supercharge Q is a column vector. Again, since we are considering a
massless particle state, we have that the only components of the four-momentum which do not
vanish are when ν = 0, 3. Hence, choosing the case when ν = 0, the above is simply zero due to
the Levi-Civita tensor. Thus, we are left with considering when ν = 3, in which case
1 03ij 1
[Q, W 0 ] = σij QP3 = − 3ij σij QP3
2 2
A short exercise shows that ijk σjk = σ i , which allows us the change the above to
1
− σ 3 QP3
2
Therefore, if we put the indices back into the expression we find that
1
[Qa , W 0 ] = − (σ 3 )ab Qb P3 (14.122)
2
Looking at the matrix representation of σ 3 , we find that the only non-zero elements are (σ 3 )1 1 =
−(σ 3 )2 2 = 1, in which case we find
1 1
[Q1 , W0 ] = − Q1 P3 [Q2 , W0 ] = Q2 P3
2 2
where we used the fact that [Q, W 0 ] = [Q, W0 ]. Keeping in mind that for massless particle states
we have (14.120), we have that
W0 |p, hi = hE|p, hi
where we have exchanged |pi0 with |p, hi, for reasons that will become clear momentarily. To
the effect the supercharges has on these states, first consider the case when we have
Pµ Q1 |p, hi = Q1 Pµ |p, hi = pµ (Q1 |p, hi)

indicating that the state |p, hi has a four-momentum pµ . Now consider the more interesting case:

1
W0 (Q1 |p, hi) = [W0 , Q1 ]|p, hi + Q1 W0 |p, hi = E h + Q1 |p, hi
2
14.11. CLASSIFYING STATES WITH SUPERCHARGES 343
in which we find a remarkable result

1
Q1 |p, hi = |p, h + i (14.123)
2
In words, the action of the supercharge Q1 on the massless particle state raised its helicity by
a factor of 21 . Since bosons are particles with a helicity differing from the helicity of fermions by
a factor of 12 , we see that the supercharges transform bosons into fermions! Using a very similar
analysis, one can show that the operation of Q2 lowers the helicity of the particle by 12 :
1
Q2 |p, hi = |p, h − i (14.124)
2
We won’t go into the details here, but, as on emight expect, the Hermitian conjugate supercharges
Q†1 and Q†2 have the opposite effect of Q1 , and Q2 .
Now that we have seen how the supercharges act on massless particle states, we can use this
information to build the SUSY multiplets, or supermultiplets which are irreducible representations
of SUSY, the states that appear in the MSSM. To complete this is rather straightforward, we will
use methods similar to before to classify the supersymmetric massless states. Therefore, the starting
point is to use the SUSY algebra we have developed.
Using the same frame as before, notice that we have

µ 0 0 3 3 0 0
σ Pµ |p, hi = (P σ − P σ )|p, hi = |p, hi
0 2P 0
Then, making use of the SUSY algebra derived earlier, we find
{Q1 , Q†1 }|p, hi = {Q1 , Q†2 }|p, hi = {Q2 , Q†1 }|p, hi = 0

and
{Q2 , Q†2 }|p, hi = 2P 0 |p, hi

From here we find that the expectation value of {Q1 , Q†1 } is just
hp, h|{Q1 , Q†1 }|p, hi = hp, h|Q1 Q†1 |p, hi + hp, h|Q†1 Q1 |p, hi
= ||Q1 |p, hi||2 + ||Q†1 |p, hi||2 = 0

Remember that when we only consider physical states, we are only considering states with
positive definite norm. Therefore, individually, we have
Q1 |p, hi = Q†1 |p, hi = 0

Thus, we will only consider the super charges Q2 and Q†2 . It’s imperative to notice that since Q2
lowers the helicity of a particle, there exists such a state where the helicity can no longer be lowered.
That is, there is a state with minimum helicity. From this minimum state, we can only use Q†1 to
generate a higher helicity state. One might think we could do this indefinitely, generating higher
helicity states, however, this cannot be done since the supercharges anticommute, i.e. {Q†2 , Q†2 } = 0.
Therefore, we may only operate with the supercharge Q†2 once on the lowest helicity state. This
means that we only have two states the work with: the lowest helicity particle state, and the
next highest helicity state. All in all, we have a massless supermultiplet only contains two states,
each with the same four-momentum, but different helicities, a boson state and a fermion state.
Formally, this result is known as N = 1 supersymmetry. We say it is N = 1 because we only
have one supercharge (and its Hermitian conjugate). Indeed, there are supersymmetric theories,
called extended supersymmetric theories, which have more than one supercharge, however, we won’t
explore these theories in this text.
In N=1 SUSY we only have two state, however if we want to have a realistic theory that’s an
extension of the Standard Model, we must pay attention to CPT invariance. the CPT theorem
changes the sign of the helicity of the particle, and therefore we require two more conjugate states
since the two states we came up with are not conjugate states to each other. In summary, a realistic
N=1 supersymmetric theory has four states: two states making up the supermultiplet, and its CPT
conjugate state.
Let’s look at at some of the phenomenological implications of our N=1 supersymmetric theory.
Let us first consider the case where we have a particle state that has a minimum helicity of 0,
hmin = 0, which corresponds to a scalar field, or spin-0 particles. As we noted above, SUSY would
also require states with helicity of h = 12 . Moreover, CPT invariance requires that we introduce
another h = 0 state and h = − 12 state. Both h = 0 helicity states can be assembled into a complex
scalar field, while the two fermion states are easily identified as a particle with its associated
antiparticle. We call this type of multiplet a chiral multiplet [30].
In the MSSM all of the known fermions are members of chiral multiplets, and therefore each
fermion has a spin-0 supersymmetric partner. The convention to add an s to the names of the
known fermions designating their spin-0 superpartners. That is, in SUSY, for each fermion there
is a sfermion. For instance, the spin-0 superpartner of the electron is the selectron. Moreover, the
Higgs field is also in the chiral multiplet, being part of the scalar component. What’s more is in
the MSSM, there are two doublets for the Higgs field, compared to one doublet in the ordinary
Standard Model. Therefore, in SUSY, there is more than one Higgs particle!
Just as the fermions are paired with sfermions, all of the scalar fields which belong to chiral
multiplets have fermionic superpartners. The naming convention for these particles is to add the
sufix ino at the end of the name of the known scalar boson. For instance, the spin- 12 superpartners
of the Higgs field are known as Higgsinos. But the fun doesn’t stop there! For higher helicity states,
one finds that their are gauginos, the fermionic superpartner to the gauge bosons. For instance, we
have the photino, the winos, and the zino. Just when particle physicists thought they had a zoo on
their hands, supersymmetric theorists brought in a whole jungle!
Now let’s keep in mind that so far, despite its success in resolving some of the issues of the
Standard Model, SUSY has yet to be experimentally verified. One chief goal of the LHC is to test
whether or not they can find these superpartners. If these particles were indeed observed, SUSY
would be far more than a theory. What’s more, since realistic string theories rely on SUSY, it
would be a good for the validity of string theory.
14.12 Some More Notation

We are near the end of our brief introduction to supersymmetry. However, now comes the news that
no student wants to hear: there is in fact another, more elegant approach to supersymmetry, one
which is most often used in the literature, and particularly in superstring theory. This is not to say
14.12. SOME MORE NOTATION 345
all that we have done has been in vain, quite the contrary actually. We have used the more brute
force approach as it allows us to gain deep physical insight into the subject without introducing
more elegant, and more advanced mathematics. Moreover, in the more modern approach, the
superspace formalism, we require a further layer of notation, called van de Waerden notation. We
won’t go through all of the details of this new layer of notation, however, appendix A is devoted
to working out some of the results we have already become familiar with in this chapter, as to give
further insight into the notation. With that, let us proceed.
The plan is to introduce further notation so that we may construct Lorentz invariants more easily.
Our motivation will come from the Einstein summation convention. This notation can be quite
tedious at times, however, actually simplifies many of the calculations. Let us start by defining the
components of a left-chiral Weyl spinor as
χ ≡ χa (14.125)
Now recall that so far we have constructed a Lorentz invariant out of η † χ. In order to maintain
this convention as well as introduce the summation convention notation, we are motivated to build
a Lorentz invariant by
η † χ ≡ η a χa (14.126)
which means that η a are the components of the Hermitian conjugate of the right-chiral spinor
η. To represent the components themselves, we use van de Waerden notation, which has the
components of η as (Labelle, 45)
η ≡ η̄ ȧ (14.127)
Therefore, η̄ ȧ = (η a )† . This means we would write the Lorentz invariant
η † χ = (η̄ ȧ )† χa = η a χa
We also have the Lorentz invariant
χ† η = (χa )† η̄ ȧ
To use similar notation, we write
χ† η ≡ χ̄ȧ η̄ ȧ (14.128)
where we have defined χ̄ȧ ≡ (χa )† . Our goal is to define quantities with the same type of indices
to have the same Lorentz transformations, therefore we want χ̄ȧ to transform in the same way as
η̄ ȧ . Moreover, earlier we showed that iσ 2 χ†T transforms as a right-chiral spinor, therefore we also
define χ̄ȧ as the components of iσ 2 χ†T . To match indices we require that
χ̄ȧ = (iσ 2 )ȧḃ χ†b (14.129)

Or, instead
χ̄ȧ = (iσ 2 )ȧḃ χ̄ḃ

Using the matrix representation of σ 2 , we notice that χ̄1̇ = χ̄2̇ and χ̄2̇ = −χ̄1̇ . What this
means is the matrix iσ 2 may be used to raise dotted indices, similar to the operation of the metric.
Moreover, we also define
χa = (χȧ )† (14.130)
from which one finds
χa = ((iσ 2 )ȧḃ χ̄ḃ )† = ((iσ 2 )ȧḃ )∗ (χ̄ḃ )† = (iσ 2 )ab (χ†b )† = (iσ 2 )ab χb
where we used the fact that (iσ 2 )ȧḃ is real so complex conjugation has no effect, and we changed
dotted indices into undotted indices in order to keep the indices of the entire equation balanced (for
more on these indices mutating, refer to appendix A). Now recall that the inverse of iσ 2 is simply
−iσ 2 . Therefore we also have that
χ̄ȧ = (−iσ 2 )ȧḃ χ̄ḃ χa = (−iσ 2 )ab χb

which shows that −iσ 2 can be used to lower both dotted and undotted indices. Moreover, if we
impose the condition that ηa transform like χa , and the fact that χ transforms like −iσ 2 η †T , we
find
ηa ≡ (−iσ 2 )ab (η̄ ḃ )† = (−iσ 2 )ab η b (14.131)

If w define η̄ȧ ≡ ηȧ† we can also show that
η̄ȧ = (−iσ 2 )ȧḃ η̄ ḃ (14.132)

To ease the reader on the morphing of dotted to undotted indices in the Pauli matrix, we keep in
mind that the components of σ 2 are always the same, whether undotted or undotted. For instance,

0 1
(iσ 2 )12 = (iσ 2 )1̇2̇ = =1
−1 0 12
Before moving on to the superpace formalism of the supersymmetry, let us write out some of
the important results we have come up with using van de Waerden notation. The most notable
ones involve the algebra of the supercharges. Using dotted notation, the anticommutator of the
supercharges with themselves and the Hermitian conjugate supercharges with themselves is nicely
summarized as
{Q̄ȧ , Q̄ḃ } = 0 (14.133)

Then, the only non-vanishing anticommutation relation of the supercharges is written as
{Qa , Q̄ḃ } = (σ µ )aḃ Pµ (14.134)

Finally, the commutation relations with the Lorentz generators Jµν in the dotted notation is
given by
[Q̄ȧ , Jµν ] = −Q̄ḃ (σ̄µν )ḃ ȧ [Q̄ȧ , Jµν ] = (σ̄)ȧḃ Q̄ḃ (14.135)
14.13. SUPERSPACE 347
Details of these results and more will be given in appendix A for the reader desiring further
practice with this new notation. With that we are ready to confront the alternative approach to
supersymmetry. On to superspace!
14.13 Superspace
In this section we will explore the details of the superspace approach. Here we will explore the more
conventional approach to supersymmetry, and the approach used in superstring theory. What’s
more, in using the superspace formalism, we make special note of the deep relationship supersym-
metry has with space-time translations. Let’s move to that first.
We start with the notion that we may write the SUSY charges as differential operators rather
than functions on fields. Moreover, as we have already noted, the anticommutator of two of the
SUSY charges give us the four-momentum P µ , which effectively translates field about in space-time.
What ends up happening is that when we choose to represent the charges of SUSY as differential
operators, we mathematically extend space-time to include Grassmann coordinates. This extension
of space-time is formally known as superspace. In this section we follow closely to Labelle, as it is
a rather straightforward approach. For more details, the reader is urged to examine his text.
Recall from before that we had the unitary transformation
U (a) ≡ exp(iaµ Pµ )
where aµ is an infinitesimal displacement four-vector. From here we had that we could represent
a translation of a field as
φ(x0 ) = U (a)φ(x)U † (a) = φ(x + aµ )

That is, x → x + aµ . Moreover, had we started with a field that is evaluted at the space-time
origin, x = 0, we can generate the field at any position in space time x by applying this unitary
transformation, this time letting aµ = xµ .
This is the basis for superspace. We view the supercharges as producing a translation of the
fields. We end up having parameters (ζ, ζ̄) that are dotted to the supercharges. Using the above as
motivation, we will choose to view these Grassmann quantities as displacement vectors. Doing this
however requires that we must add more coordinates to the usual set of four space-time coordinates.
This extension of space-time, superspace, is more of a mathematical entity than anything else, as
the addition of the Grassmann coordinates does not alter our previous notion of space-time.
Four Grassmann coordinates are added to the original four space-time coordinates, which we will
denote as θ1 and θ2 , and their complex conjugates. Since the SUSY parameter ζ is a left-chiral Weyl
spinor, we continue to work with left-chiral Weyl spinors. Therefore, we let the extra coordinates
be left-chiral Weyl spinors, and define
1̇
θ θ̄
θ≡ 1 θ̄ ≡ 2̇ (14.136)
θ2 θ̄
Using van der Waerden notation, we have that θ̄1̇ = (θ1 )† and similarly for the other coordinate.
Since this Grassmann coordinates are not physical quantum fields, one can view the action of the
Hermitian conjugate as a complex conjugate, no transpose occurs.
Let us now introduce a superfield which is simply a function of all space-time coordinates xµ ,
as well as θ and θ̄. We denote the superfield as Φ(x, θ, θ̄). Let us apply a SUSY transformation
U (a, ζ, ζ̄), yielding
U (a, ζ, ζ̄)Φ(x, θ, θ̄)U † (a, ζ, ζ̄) = Φ(x0 , θ0 , θ̄0 ) (14.137)

Remember, the plan is to find a differential operator representation of the supercharges that
act on superspace. Suppose we start at the origin of superspace, described by the superfield Φ(0).
Using a SUSY transformation, we can generate a superfield at any point in superspace, namely,
U (x, θ, θ̄)Φ(0)U † (x, θ, θ̄) = Φ(x, θ, θ̄)

One the other hand, we also could have also defined
U (x0 , θ0 , θ̄0 )Φ(0)U † (x0 , θ0 , θ̄0 ) = Φ(x0 , θ0 , θ̄0 )

If we substitute both of these expressions into (14.137), we find that
U (a, ζ, ζ̄)U (x, θ, θ̄) = U (x0 , θ0 , θ̄0 ) (14.138)

Our next step is to determine expressions for the unitary transformation U in terms of the
supercharges. As usual, we will write them as exponentials of Lorentz invariants. The correct
choice ends up being (Labelle, 245)
U (x, θ, θ̄) = exp(ix · P + iQ̄ · θ̄ + iQ · θ) (14.139)

This will allow us to write (14.138) as
exp(ia · P + iζ̄ · Q̄ + iζ · Q)exp(ix · P + iθ̄ · Q̄ + iθ · Q) = exp(ix0 · P + iθ̄0 · Q̄ + iθ0 · Q)
To evalute this expression we require the Baker-Cambell-Hausdorff formula:

1
exp(A)exp(B) = exp A + B + [A, B] + ...
2
where the terms not written are higher order commutators (i.e. commutators within commuta-
tors). Using this we see that the left hand side of the transformations become

1
exp i(x + a) · P + i(θ̄ + ζ̄) · Q̄ + i(θ + ζ) · Q − [ζ̄ · Q̄ + ζ · Q, θ̄ · Q̄ + θ · Q] + ... (14.140)
2
This commutators can be easily calculated using van der Waerden notation discussed in the last
section and in appendix A. First consider
[ζ · Q, θ · Q] = [ζ a Qa , θb Qb ] = ζ a Qa θb Qb − θb Qb ζ a Qa
= −ζ a θb {Qa , Qb } = 0
where we used the algebra of the supercharges developed earlier. Similarly,
[Q̄ · ζ̄, Q̄ · θ̄] = [Q̄ȧ ζ̄ ȧ , Q̄ḃ θ̄ḃ ] = −ζ̄ ȧ θ̄ḃ {Q̄ȧ , Q̄ḃ } = 0
Alternatively, consider
[ζ · Q, Q̄ · θ̄] = [ζ a Qa , Q̄ḃ θ̄ḃ ] = ζ a θ̄ḃ {Qa , Q̄ḃ } = ζ a θ̄ḃ (σ µ )aḃ Pµ

Without explicit indices,
[ζ · Q, Q̄ · θ̄] = ζσ µ θ̄Pµ
Similarly, we also have
[Q̄ · ζ̄, θ · Q] = −θσ µ ζ̄Pµ

We might suspect that we are not finished since we still have to deal with the higher order
commutators not explicitly written. However, since the commutators we have found all yield the
momentum operator P µ , which commutes with everything, all high order commutators vanish.
Therefore, all that we are left with is
1 1
exp(i(x+a)·P +i(θ̄+ζ̄)·Q̄+i(θ+ζ)·Q− ζσ µ θ̄Pµ + θσ µ ζ̄Pµ ) = exp(iθ0 ·Q+iθ̄0 ·Q̄+ix0 ·P ) (14.141)
2 2
In this form we can easily read off the transformations of the superspace coordinates
i i
x0 = x + a + ζσ µ θ̄ − θσ µ ζ (14.142)
2 2
θ0 = θ + ζ θ̄ = θ̄ + ζ̄ (14.143)
From here we will begin to look at the physics of superfields. To do this, more mathematics is
necessary: one requires the calculus of Grassmann variables. A short introduction to this topic is
presented in appendix B. The most important notion to keep in mind is that Grassmann variables
are anticommuting. This has consequences when it comes to computing derivatives of Grassmann
variables. Moreover, Grassmann integration is not the integration based off of Riemann sums one
is familiar with from real variable calculus. Rather, it is more like an abstract algebraic operation,
which behaves similarly to computing derivatives of Grassmann variables. If the reader is new to
this mathematics, they are urged to first review appendix B.
Moving on, recall that the action for a differential operator P̂µ is defined through
exp(−iaµ P̂µ )φ(x) = φ(x + δx)

In the present case, the analogous expression that contains a space-time translation as well as a
supersymmetric transformation is given by [30]
exp(−iaµ P̂µ − iζ · Q − iζ̄ · Q̄)S(x, θ, θ̄) = S(x + δx, θ + δθ, θ̄ + δ θ̄) (14.144)
where S denotes the superfield, and Q denotes the supercharge on superspace. If we plug in the
transformed coordinates given in (14.142) and (14.143), and Taylor expand the right hand side of
the above expression, we find
S(x + δx, θ + δθ, θ̄ + δ θ̄) ≈ S(x, θ, θ̄) + δxµ ∂µ S(x, θ, θ̄) + δθa ∂a S(x, θ, θ̄) + δ θ̄ȧ ∂ ȧ S(x, θ, θ̄)
i i
= S + (aµ + ζσ µ θ̄ − θσ µ ζ̄)∂µ S + ζ a ∂a S + ζ̄ȧ ∂¯ȧ S
2 2
where we have defined
∂S ∂S
∂a S ≡ ∂¯ȧ S ≡
∂θa ∂ θ̄ȧ
Making use of the spinor dot product notation using dotted and undotted indices, we may write
i i ¯
S(x + δx, θ + δθ, θ̄ + δ θ̄) ≈ S + (aµ + ζσ µ θ̄ − θσ µ ζ̄)∂µ S + ζ · ∂S + ζ̄ · ∂S
2 2
Then, expanding the exponential above to first order, we find that
i i ¯
(−iaµ P̂µ − iζ · Q − iζ̄ · Q̄)S = (aµ + ζσ µ θ̄ − θσ µ ζ̄)∂µ S + ζ · ∂S + ζ̄ · ∂S (14.145)
2 2
In this form, we may readily read off the differential operator representations of P̂ , Q, and Q† .
First examining terms proportional to aµ , we, as expected, find
P̂µ = i∂µ (14.146)

Now, putting in explicit indices, we have ζσ µ θ̄ = ζ a (σ µ )aḃ θ̄ḃ . Then, if we only keep the terms
that contain the SUSY parameter ζ, and if we write all expressions with indices, we find
i µ
−iζ · Q = ζσ θ̄∂µ S + ζ · ∂S
2
i a µ
ζ (σ )aḃ θ̄ḃ ∂µ S + ζ a ∂a S
⇒ −iζ a Qa =
2
which implies that the differential operator representation of Q is
1
Qa = i∂a − (σ µ )aḃ θ̄ḃ ∂µ (14.147)
2
A similar calculation we find
i
−iζ̄ȧ Q̄ȧ = ζ̄ȧ (σ̄ µ )ȧb θb ∂µ + ζ̄ȧ ∂¯ȧ
2
where we made use of the identity χσ µ λ̄ = −λ̄σ̄ µ χ. From, it’s easy to see that
1
Q̄ȧ = i∂¯ȧ − (σ̄ µ )ȧb θb ∂µ (14.148)
2
As an exercise, the reader is asked to show the differential operator representation of the lower
indexed version of Q̄ is just
1
Q̄ȧ = −i∂¯ȧ + θb (σ µ )bȧ ∂µ (14.149)
2
Now that we have differential operator representations of the supercharges, obey the anticom-
mutation relations of the ordinary supercharges:
{Qa , Qb } = {Q̄ȧ , Q̄ḃ } = 0 {Qa , Q̄ḃ } = (σ µ )aḃ P µ

Since the first two of these anticommutators are rather trivial, we will only spend time on
the third anticommutator. Looking ahead, we know that we will have to compute at least three
anticommutators between the derivatives and the Grassmann numbers. Specifically,
{∂a , θ̄ḃ } {∂a , θb } {∂¯ȧ , θ̄ḃ }

To evaluate these anticommutators, we imagine that we apply the anticommutator to some
arbitrary function, f . Keeping this in mind, the first anticommutator is just
{∂a , θ̄ḃ }f = ∂ θ̄ḃ f θ̄ḃ ∂a f = (∂a θ̄ḃ )f − θ̄ḃ ∂a f + θ̄ḃ ∂a f = 0 ⇒ {∂a , θ̄ḃ } = 0
Similarly,
{∂a , θb } = δ b a {∂¯ȧ , θ̄ḃ } = δ ḃ ȧ

Using these results, we find that

i µ ċ i d µ
{Qa , Q̄ḃ } = i∂a + (σaċ θ̄ P̂µ , −i∂ḃ − θ (σ )dḃ P̂µ
2 2
1 1
= {∂a , θd }(σ µ )dḃ P̂µ + {θ̄ċ , ∂ḃ }(σ µ )aċ P̂µ
2 2
1 d µ 1
= δ (σ )dḃ P̂µ + δ ċ ḃ (σ µ )aċ P̂µ = (σ µ )aḃ P̂µ
2 a 2
Therefore, we have correctly determined the differential operator representation of the super-
charges. Moreover, the supercharges we have derived obey the anticommutation relations, and
hence satisfy the SUSY algebra. All in all, we have found a representation for the supercharges
which behave as operators on superspace.
Now that we have the explicit representation of the supercharges, let’s explore some of the
consequences, particularly the constraints on the superfields. Suppose we looked at a superfield
F that was only a function of space-time coordinates x and Grassmann coordinates θ. That is,
F = F(x, θ). Expanding this function out, we would have then
i i
F 0 (x, θ) = F(x + δx, θ + δθ) = F(x, θ) + (aµ + ζσ µ θ̄ − θσ µ ζ̄)∂µ F(x, θ) + ζ · ∂F(x, θ)
2 2
We immediately see an issue here. Although our superfield does not depend on the superspace
coordinate θ̄, the transformed field does depend on θ̄. What this means is our superfield, when just
a function of x and θ is inconsistent with SUSY. To figure out what the issue is, let’s consider a
superfield S on which we impose the following constraint
∂
S(x, θ, θ̄) = 0
∂ θ̄ȧ
This is equivalent to saying that S is simply a function of x and θ, i.e. S(x, θ, θ̄) = F(x.θ). Let’s
test to see whether this constraint is inconsistent with our SUSY transformations. In essence, we
will check to see that our SUSY transformed field F 0 satisfies the same constraint. Starting from
the left hand side of (14.144), the constraint is just
∂ 0 ∂
ȧ
F = ȧ (1 − iaµ P̂µ − iζ · Q − iζ̄ · Q̄)F
∂ θ̄ ∂ θ̄
∂ ∂ ∂
= F − i ȧ (aµ P̂µ + ζ · Q + ζ̄ · Q̄)F = −i ȧ (aµ P̂µ + ζ · Q + ζ̄ · Q̄)F
∂ θ̄ȧ ∂ θ̄ ∂ θ̄
where to get to the final step we applied the constraint. Notice then that if we can pass the
derivative through this term to apply to F, we find that indeed the constraint is satisfied by the
SUSY transformation because this action would result in zero, as it would be acting on F. The
real question then is whether or not the derivative commutes with the term in front of F, namely,
aµ P̂µ + ζ · Q + ζ̄ · Q̄
We can immediately see that the derivative commutes with aµ P̂µ :
∂ µ ∂ ∂
ȧ
a P̂µ = aµ ȧ i∂µ = aµ i∂µ ȧ
∂ θ̄ ∂ θ̄ ∂ θ̄
Therefore, all that we must check is

∂
, ζ · Q + ζ̄ · Q̄
∂ θ̄ȧ
Reminding ourselves that the SUSY parameter ζ is a constant Weyl spinor, we have, using van
der Waerden notation, that the commutator is just

∂ b ḃ ∂ ∂
, ζ Qb + ζ̄ ḃ Q̄ = ȧ (ζ b Qb + ζ̄ḃ Q̄ḃ ) − (ζ b Qb + ζ̄ḃ Q̄ḃ ) ȧ
∂ θ̄ȧ ∂ θ̄ ∂ θ̄

b ∂ ∂ ∂ ḃ ḃ ∂
= −ζ Qb + Qb ȧ − ζ̄ḃ Q̄ + Q̄
∂ θ̄ȧ ∂ θ̄ ∂ θ̄ȧ ∂ θ̄ȧ

∂ ∂
= −ζ b ȧ
, Qb − ζ̄ḃ , Q̄ḃ
∂ θ̄ ∂ θ̄ȧ
Therefore, the imposed constraint will be consistent as long as

∂ ∂ ḃ
, Qb = 0 , Q̄ = 0 (14.150)
∂ θ̄ȧ ∂ θ̄ȧ
Looking at (14.147) and (14.148), we see that the second of the anticommutators is indeed
satisfied, however the first is not satisfied. What this means is the constraint is not satisfied by
SUSY transformations and therefore does not work for us. But what we have done is not in vain!
Rather, now we have a simple way of determining whether a constrain on a superfield is consistent
with SUSY transformations. Let us write this constraint as
D̄ȧ S(x, θ, θ̄) = 0 (14.151)

where D̄ȧ is a differential operator which acts on superspace. As we just saw, this is only a
valid constraint as long as it anticommutes with the supercharges. Moreover, we already know that
the second anticommutator in (14.150) holds but the first anticommutator does not. Starting from
here, we can construct D̄ȧ . Let’s first try
∂
D̄ȧ =+ Cȧc θc (14.152)
∂ θ̄ȧ
where Cȧc is included to balance indices. By construction, we already see that (14.152) anti-
commutes with Q̄ḃ . It still remains to see whether it will anticommute with Qb . By enforcing this
condition, we can figure out what form Cȧb . That is, consider when

∂ c
+ Cȧc θ , Q = 0
∂ θ̄ȧ
Using (14.147), we see that this becomes

∂ c ∂ i µ d˙
+ Cȧc θ , i b + (σ )bd˙θ̄ P̂µ
∂ θ̄ȧ ∂θ 2

∂ ∂ i ∂ µ d˙ c ∂ i c µ d˙
=i , + , (σ ) bd˙ θ̄ P̂ µ + i C ȧc θ , + Cȧc θ , (σ ) bd˙ θ̄ P̂ µ =0
∂ θ̄ȧ ∂θb 2 ∂ θ̄ȧ ∂θb 2
If we assume that Cȧc has no space-time dependence, we see that the last anticommutator above
vanishes. Moreover, we already know the first anticommutator vanishes identically. All that remains
is to show the second and third anticommutator vanish. Setting the sum of these two terms equal
to zero, we have

i ∂ µ d˙ c ∂
, (σ )bd˙ θ̄ P̂ µ + i Cȧc θ , =0
2 ∂ θ̄ȧ ∂θb
The left hand side is simply
i ˙ i
iCȧc δ c b + (σ µ )bd˙P̂µ δ dȧ = iCȧb + (σbµȧ P̂µ
2 2
Setting this equal to zero, we find that in order for (14.152) to satisfy the constraint, we must
have that
1 i
Cȧb = − (σ)bȧ P̂µ = − (σ µ )bȧ ∂µ (14.153)
2 2
Therefore, we may rewrite (14.152) as
∂ i
D̄ȧ = − θc (σ µ )cȧ ∂µ (14.154)
∂ θ̄ ȧ 2
Now that we have a form for D̄ȧ , we can go all of the way back to our initial assumption on the
constraint. That is, from the beginning we could have instead insisted that the constraint be with
a derivative with respect to θa rather than θ̄ȧ . Had we gone this route, a very similar treatment
would have yielded [30]
∂ i
Da = − (σ µ )aḃ θ̄ḃ ∂µ (14.155)
∂θa 2
A rather simple exercise also yields the raised index version of these derivative operators
∂ i ∂ i
Da = − + θ̄ḃ (σ̄ µ )ḃa ∂µ D̄ȧ = − + (σ̄ µ )ȧb ∂µ (14.156)
∂θa 2 ∂ θ̄ȧ 2
These differential operators remind us of the covariant derivatives from quantum field theory (see
chapter 2, the section on gauge transformations), and are therefore sometimes called supercovariant
derivatives. Now that we have derivatives which satisfy the constraint we have been examining,
the next step is to look for superfields which satisfy the constraint. We won’t go into details in this
text on finding these fields, however the reader is urged to peruse Labelle’s or Aitchison’s text for
more information on the subject.
This concludes our crash course on supersymmetry, and what a ride it has been! We started
with the fundamentals of the Dirac equation, chose to write the Dirac spinors in terms of two com-
ponent Weyl spinors, constructed Lorentz invariants and a supersymmetric Lagrangian, computed
the SUSY algebra, and finally took the superspace approach. Our motivation for studying super-
symmetry was that in more realistic string theories,we introduce fermions through supersymmetry,
this will be the subject of the next couple chapters. Therefore, we haven’t seen the real triumphs of
SUSY in action. Remember, initially SUSY was built to solve the fine tuning problem, and seek to
fix some of the problems of the Standard Model. In a normal course on SUSY, one would go through
calculations showing the improvement of the Standard Model SUSY brings. On a similar note, the
superspace approach is much more elegant indeed but is certainly more esoteric. Moreover, when it
comes to studying the MSSM in detail, one often decides to use the superspace approach because
it makes calculations much simpler. Since this book is of finite length, we will not go through these
beautiful consequences of these approaches. That is, we won’t see all of the beauty SUSY has to
offer.
Our motives are more selfish than that. We desire to understand more realistic string theories,
superstring theories. For that reason we must be fairly familiar with SUSY. Therefore, we now have
(most of) the tools necessary in deciphering the elements of superstring theory. With out further
delay, let us continue forward with our study of string theory.
14.14 Exercises
1. (a) The goal of this exercise is for the reader to become more familiar with the notation of Weyl
spinors. Prove equations (14.12) and (14.19).
(b) Show that
χ† σ̄ µ i∂µ χ − i(∂µ χ† )σ̄ µ χ

are both Lorentz invariants. Moreover, using integration by parts, show that these two invariants
are actually equivalent.
2. In this exercise the reader is asked to establish some basic relations between Weyl and Majorana
spinors. First show that
ψ̄M ψM = χ · χ + χ̄ · χ̄
From here, using the fact that
ψ̄M γ5 ψM = −χ · χ + χ̄ · χ̄
show that
χ · χ = ψ̄M PL ψM χ̄ · χ̄ = ψ̄M PR ψM
where
1 + γ5 1 − γ5
PR = PL =
2 2
Of course, these relations can be generalized. Consider another left-chiral spinor λ, and an
associated Majorana spinor ΛM . Go on to show that
ψ̄M γ µ ΛM = χ† σ̄ µ λ − λ† σ̄ µ χ ψ̄M γ 5 γ µ ΛM = χ† σ̄ µ λ + λ† σ̄ µ χ
3. Prove (14.61).
4. In this exercise the goal is for the reader to prove the commutator given in (14.72). Consider
the Lorentz transformation
xµ → Λµν xν
Considering an infinitesimal Lorentz transformation, show that one may write
Λµν xν ≈ xµ + ω µν xν = xµ + ω µν xν
where ω µν is an antisymmetric tensor. The unitary operator which implements the Lorentz
transformation is given by
i µν
U = ei 2 ω Jµν
where Jµν are the Lorentz charges, which are also antisymmetric. Show that the transformation
of a scalar field φ is
δφ = φ(xµ + ω µν xν ) − φ(xµ )
which can be rewritten as

1 µν
δφ = ω (xν ∂µ − xµ ∂)φ
2
After showing this, prove (14.72).
5. Prove (14.79). Completing this calculation gives a basic idea on how to determine the algebra
of the supercharges without explicitly writing down the form of the supercharges.
6. Using a similar argument leading to (14.80), show that the commutator between the Lorentz
charges Jµν and the momentum operator Pλ is
[Jµν , Pλ ] = i(ηνλ Pµ − ηλµ Pν )
7. Using (14.108), show that two consecutive SUSY transformations on the scalar field φ and the
auxiliary field F take the same form as (14.122). Completing this calculation as well as (14.112),
show that the SUSY algebra developed holds in general for fields φ, F , and χ.
8. Using (14.118) and (14.119), prove (14.120). This equation allows us to extract the physical
significance of the proportionality constant h.
9. Prove that the first anticommutator in (14.150) does not hold, while the second anticommu-
tator does hold. Doing this indicates that the constraint on the superfield is not consistent with
the SUSY transformations (14.142) and (14.143).
10. (a) Prove that the raised index versions of the supercovariant derivatives take the form given
in (14.156).
(b) Work out the following anticommutation relations:
{Da , Db } {Da , Db } {D̄ȧ , D̄ḃ } {D̄ȧ , D̄ḃ } {Da , D̄ḃ }

Chapter 15
An Introduction to Superstrings
Up to this point we have only been considering the relativistic quantized bosonic string. As the
reader is well aware by now, the bosonic string is a non-realistic theory, and it is often viewed as
more of a toy theory. We know bosonic string theory is not a theory which describes the observable
universe because it does not include fermions, particles of half integer spin which make up matter.
In this chapter we develop the basic notions of constructing a realistic string theory, focused on
including fermions. As we will see, when we include fermions, they come in through supersymmetry,
therefore we call the realistic string theories supersymmetric string theories or superstring theories.
This was our prime motivation of learning supersymmetry in the first place, such that we have the
base knowledge to grasp superstrings. In previous chapters, most notably the chapters on T-duality
and D-branes, we have briefly discussed other superstring theories such as type IIA and type IIB,
as these theories have interesting insights into particle physics and the construction of D-branes. In
this chapter we will only briefly examine the basic ideas of these theories as a full analysis requires
concepts and techniques that go beyond the scope of this text.
It should be clarified that the work we have done up to this point has not gone in vain. That is,
bosonic string theory is not a totally useless theory. From a pedagogical standpoint, bosonic string
theory is imperative as it provides the context and goals of string theory without getting into the
mess of supersymmetry. Moreover, from a physical point of view, bosonic string theory offers deep
insight into the possible structure of our universe (extra dimensions, compactification), as well as
present a base theory that attempts to unify all of the known forces of nature. For this reason we
don’t want to lose all of the work we have done so far. Rather, when we add in fermions, we add
it to our bosonic theory in such a way that the results we attained are only improved on. As we
will see, when we introduce supersymmetry, we rid our theory of the unwanted tachyon, as well as
reduce the number of space-time dimensions from D=26 to D=10. With that, let’s proceed.
15.1 Adding Fermions

We can modify our theory by adding fermions in a rather straightforward way using the Ramond-
Neveu-Schwarz (RNS) formalism. This approach adds supersymmetry to the worldsheet. Later on
we will discuss an alternative approach known as the Green-Schwarz (GS) formalism, which uses
space-time supersymmetry.
357
358 CHAPTER 15. AN INTRODUCTION TO SUPERSTRINGS
As usual, we will introduce an action, determine the equations of motion, and move on to quantize
the theory. Let’s start with the Polyakov action in the conformal gauge
Z
T
S=− d2 σ∂α X µ ∂ µ Xµ (15.1)
2
where α = 0, 1. To include fermions using the RNS formalism, we modify the Polyakov action by
including a kinetic energy term for a Dirac field, a fermion field. In fact, to generalize, we introduce
D Majorana fermions in the following way [21]:
Z
T
S=− d2 σ(∂α X µ ∂ µ Xµ − iψ̄ µ ρα ∂α ψµ ) (15.2)
2
where ρα are the two dimensional Dirac matrices which, in an appropriate basis, can be repre-
sented as

0 0 −i 1 0 i
ρ = ρ = (15.3)
i 0 i 0
As an exercise, the reader will show that the Dirac matrices on the worldsheet obey the anti-
commutation relations known as the Dirac algebra, or sometimes called the Clifford algebra:
{ρα , ρβ } = −2η αβ (15.4)

αβ
where η is the two dimensional Minkowski metric. A note to the reader: in some of the
literature, the convention is to write the Dirac matrices without a factor of i, in which case the
above anticommutation relations will be off by an overall minus sign. For completion, in our
modified action (15.2) we use the Dirac conjugate, which may be written now as ψ̄ µ† ρ0 , where we
notice ρ0 takes place of the usual γ 0 matrix that we are familiar with from our studies of the Dirac
equation and fermion fields.
Notice what we have done. We started with the action for a bosonic string, the Polyakov action,
and proceeded to add in fermions by hand via a kinetic energy term using Majorana fermions.
Therefore, we did not throw out the work we have done with the bosonic string, and rather simply
modified our theory to include fermions. Momentarily we will see that this action is supersym-
metric, at least on the worldsheet. The Majorana fermions we introduced ψ µ = ψ µ (τ, σ), where
µ = 0, 1, ...D − 1,since they live on the two dimensional worldsheet, are two component Majorana
fermions, which are often written as the column vector

ψ
ψ= − (15.5)
ψ+
It is important to note that the Majorana fermions we use are real, and therefore obey ψ µ∗ = ψ µ
[5]. Using (15.5), we may actually rewrite the modified term in a way that will be helpful in a
moment. Using the summation convention over α we may write

0 −i 0 i
ψ̄ µ ρα ∂α ψµ = ψ̄ µ (ρ0 ∂0 + ρ1 ∂1 )ψµ = ψ̄ µ ∂τ + ∂ ψµ
i 0 i 0 σ

0 −i(∂τ − ∂σ ) 0 −2i∂−
= ψ̄ µ ψµ = ψ̄ µ ψµ
i(∂τ + ∂σ ) 0 2i∂+ 0
15.2. SUSY TRANSFORMATIONS OF THE WORLD-SHEET AND CONSERVED CURRENTS359
where we have used ∂0 = ∂σ and ∂1 = ∂τ and defined

1 1
(∂τ − ∂σ ) = ∂− (∂τ + ∂σ ) = ∂+ (15.6)
2 2
Using (15.5), the above may expanded out as

µ µ 0 −i 0 −i(∂τ − ∂σ ) 0 −2i∂− ψ−
= ψ− ψ+ ψµ = ψ̄ µ
i 0 i(∂τ + ∂σ ) 0 2i∂+ 0 ψ+

µ µ 2∂+ ψ−µ µ µ
= ψ− ψ+ = 2ψ− ∂+ ψ−µ + 2ψ+ ∂− ψ+µ
2∂− ψ+µ
= 2(ψ− · ∂+ ψ− + ψ+ · ∂− ψ+ )
This means we may write the fermionic part of the modified action in (15.2) as
Z Z
T µ α
SF = − d σ(−iψ̄ ρ ∂α ψµ ) = iT d2 σ(ψ− · ∂+ ψ− + ψ+ · ∂− ψ+ )
2
(15.7)
2
As we will see momentarily, writing the action like this will simplify some of the computations
we will do when we vary the action.
15.2 SUSY Transformations of the World-sheet and Con-

served Currents
Since we are dealing with fermion fields, the equal τ anticommutation relations for fermions are
decided to be [22]
µ
{ψA ν
(τ, σ), ψB (τ, σ 0 )} = πη µν δAB δ(σ − σ 0 ) (15.8)
where A, B = ±. It’s imperative to point out that, using the mostly minus convention of the
Minkowski metric, we have η 00 = −1. Therefore the time-like fermions ψA 0
(τ, σ) end up creating
ghost states, quantum states with negative norm. We found the same thing when we looked at the
method of covariant quantization of the bosonic string. We were able to rid our theory of these
ghost states using the Virasoro constraints. As it turns out, to solve the analogous problem for
time-like fermions, we are required to introduce another type of symmetry. As one might guess,
this symmetry happens to be supersymmetry.
In this section we will introduce supersymmetric transformations for our modified action and
examine the consequences of requiring the Lagrangian to be invariant under these transformations.
Later on, we will derive the SUSY transformations we look at here, and yield evidence that the
action (15.2) is indeed supersymmetric.
We begin by introducing a constant infinitesimal Majorana spinor , our SUSY parameter. In
terms of components, we write it as

= − (15.9)
+
Then, as the reader will show, the modified action is invariant under the infinitesimal transfor-
mations
δX µ = ¯ψ µ δψ µ = −iρα ∂α X µ (15.10)

Just so these transformations seem completely arbitrary, in the last chapter on our discussion
of SUSY transformations we found that the commutator of two SUSY transformations yield a
translation in space-time. In the present context, by spatial translation we mean a translation
of the string world-sheet. To see this explicitly, consider two SUSY transformations, δ1 and δ2 ,
each with SUSY parameters 1 and 2 respectively. Let’s see what happens when we apply the
commutator of these two SUSY transformations to the string coordinates (or, now, bosonic fields)
X µ . First notice
δ1 δ2 X µ = δ1 (¯
2 ψ µ ) = ¯2 (−iρα ∂α X µ 1 )
δ2 δ1 X µ = δ2 (¯
1 ψ µ ) = ¯1 (−iρα ∂α X µ 2 )
Therefore,
[δ1 , δ2 ]X µ = (δ1 δ2 − δ2 δ1 )X µ = −i¯

2 ρα 1 ∂α X µ + i¯
1 ρ α 2 ∂ α X µ
We may rewrite the first term in the above expression, so that the commutator of the SUSY
transformations can be recast in a more elegant form. In the first term we have the expression
¯2 ρα 1 . Let’s see if we may rewrite this using what we know about spinors. First of all, using the
Dirac algebra in (15.4), we have
¯2 ρα 1 = 2 ρ0 ρα 1 = [(ρα 1 )(2 ρ0 )]T = 1 ρα ρ0 2 = 1 (−2η α0 − ρ0 ρα )2
= −2η α0 1 2 − 1 ρ0 ρα 2 = −2η α0 1 2 − ¯1 ρα 2

On the other hand, we could have instead written
¯2 ρα 1 = 2 ρ0 ρα 1 = 2 (−2η 0α − ρα ρ0 )1 = −2η 0α 2 1 − 2 ρα ρ0 1
= −2η 0α 2 1 − [(ρ0 1 )(2 ρα )]T = −2η 0α 2 1 − ¯1 ρα 2

Adding these two expressions we find
2 ρα 1 = −2η 0α (1 2 + 2 1 ) − 2¯
2¯ 1 ρα 2
Since 1 , 2 are Grassmann quantities, they anticommute, 1 2 = −2 1 . Therefore, the above
yields
¯2 ρα 1 = −¯
1 ρα 2 (15.11)
Using this result, we find that the commutator of two SUSY transformations on the bosonic field
X µ is
[δ1 , δ2 ]X µ = 2i¯
1 ρα 2 ∂α X µ ≡ aα ∂α X µ (15.12)
Therefore, as expected, the commutator of two SUSY transformations yields a translation of the
string coordinates along the world-sheet. Similarly, the reader will show that
[δ1 , δ2 ]ψ µ = aα ∂α ψ µ (15.13)
Moreover, we can relate the individual components of the SUSY transformations. For instance,
notice that
µ
0 −i ψ− µ µ
δX µ = ¯ψ µ = − +

µ = i(+ ψ− − − ψ+ ) (15.14)
i 0 ψ+
Moreover, the variation in the fermionic field can be calculated fairly simply. First of all, note
that

α µ 0 −2i∂−
ρ ∂α X = X µ
2i∂+ 0
Then, from (15.10), we have that
µ
µ δψ− 0 −2i∂−
δψ = µ = X µ
δψ+ 2i∂+ 0
Yielding,
µ µ
δψ− = −2∂− X µ + δψ+ = 2∂+ X µ + (15.15)
Now that we have reason to believe the transformations given in (15.10) are SUSY transforma-
tions (we will prove this shortly), let us find the conserved currents associated with our modified
Lagrangian density given in (15.2):
T
L = − (∂α X µ ∂ µ Xµ − iψ̄ µ ρα ∂α ψµ ) (15.16)
2
Let’s first consider translation invariance in the string coordinates. That is, let’s consider when
X µ → X µ + aµ where aµ is some constant infinitesimal parameter. Since the fermionic portion of
the Lagrangian does not depend on X µ , it will be unchanged by this translation. By substitution,
our Lagrangian becomes
T
L→− [∂α (X µ + aµ )∂ α (Xµ + aµ ) + LF ]
2
T
=− [∂α X µ ∂ α Xµ + ∂α X µ ∂ α aµ + ∂α aµ ∂ α Xµ + ∂α aµ ∂ α aµ + LF ]
2
T
=L− [∂α X µ ∂ α aµ + ∂α aµ ∂ α Xµ ]
2
where we have eliminated any terms that are quadratic in the infinitesimal parameter. Therefore,
the variation in our Lagrangian is
T
δL = − (∂α X µ ∂ α aµ + ∂α aµ ∂ α Xµ ) (15.17)
2
As an exercise in index gymnastics, the reader will show that this variation takes the form
δL = T ∂α X µ ∂ α aµ (15.18)
µ
Since this variation holds for any translation parameter a , we recognize the string momentum:
Pαµ = T ∂α X µ (15.19)
just as before. We know that this is indeed the string momentum as we started with a translation
in string coordinates X µ . Using this calculation as a reference we can also find the current associated
with the supersymmetric transformations from (15.10). The variation of the full Lagrangian density
is
T
2∂α (δX µ )∂ α Xµ − i(δ ψ̄ µ )ρα ∂α ψµ − iψ̄ µ ρα ∂α (δψµ )

δL = −
2
T
ψ µ )∂ α Xµ + (¯ ρβ ∂β X µ )ρα ∂α ψµ − ψ̄ µ ρα ∂α (ρβ ∂β Xµ )

=− 2∂α (¯
2
ψ µ )∂ α Xµ − ∂α (¯ρβ ∂β X µ )ρα ψµ

= −T ∂α (¯
ψ µ )∂ α Xµ − ∂α ¯(ρβ ∂β X µ )ρα ψµ − ¯ρβ ρα (∂α ∂β X µ )ψµ

= −T ∂α (¯
ψ µ ∂ α Xµ ) − ∂α ¯(ρβ ρα ψ µ ∂β Xµ )

= −T ∂α (¯
The first term in our final expression above is a total derivative, and therefore does not actually
contribute to the variation of the action, the object we are really concerned with. From here we
may identitfy the conserved Noether current, the so-called supercurrent:
1 β
Jα = ρ ρα ψ µ ∂β Xµ (15.20)
2
1
The factor of 2 is a normalization factor that has been chosen for convenience in the literature.
We may also calculate the energy-momentum tensor with our modified action (15.2). In the
present case, the energy-momentum tensor is associated with translation invariance on the string
world-sheet. For this reason, consider the infinitesimal translation parameter α which varies the
world-sheet coordinates σ α = (τ, σ) as σ α → σ α + α . As one might suspect, the bosonic fields X µ
and fermionic fields ψ µ change as according to Taylor expansions. One may write the changes of
these fields as [37]
X µ → X µ + α ∂α X µ ψ µ → ψ µ + α ∂α ψ µ (15.21)
To determine the energy-momentum tensor, we will apply the usual Noether method. For now
we will assume that the infinitesimal parameter α depends on the world-sheet coordinates as it
will help us later identify the expression we seek. We also drop the factor of T . Let’s first examine
the fermionic part of the Lagrangian density.
i
LF = − ψ̄ µ ρα ∂α ψµ (15.22)
2
Varying using (15.21) leads us to

i i
δLF = − (δ ψ̄ µ )ρα ∂α ψµ − ψ̄ µ ρα ∂α (δψµ )
2 2
i i i
= − (β ∂β ψ̄ µ )ρα ∂α ψµ − ψ̄ µ ρα ∂α β ∂β ψµ − ψ̄ µ ρα β ∂α ∂β ψµ
2 2 2
Now, remember what we are really concerned with is the variation of the action S, which is an
integral of the Lagrangian density L over all space-time. Therefore, we may apply integration by
parts to this expression to manipulate it in such a way that is advantageous for us. Assuming the
boundary terms vanish and applying integration by parts to the last term in the above expression
we find
i i i
δLF = − (β ∂β ψ̄ µ )ρα ∂α ψµ − ψ̄ µ ρα ∂α β ∂β ψµ + ∂β (ψ̄ µ ρα β )∂α ψµ
2 2 2
i i i i
= − (β ∂β ψ̄ µ )ρα ∂α ψµ − ψ̄ µ ρα ∂α β ∂β ψµ + β ∂β ψ̄ µ ρα ψµ + ψ̄ µ ρα ∂β β ∂α ψµ
2 2 2 2
Since the divergence term ∂β β is not going to contribute to the overall variation of the action
we may simply drop it, leaving us with us

β i µ α
δLF = ∂α − ψ̄ ρ ∂β ψµ (15.23)
2
Since we would also like the energy-momentum tensor Tαβ to be symmetric under the indices α
and β, we rewrite the above variation as

i i
δLF = ∂α β − ψ̄ µ ρα ∂β ψµ − ψ̄ µ ρβ ∂α ψµ (15.24)
4 4
As an exercise, the reader will compute the variation of the bosonic part of the Lagrangian
density LB to be
δLB = ∂α β ∂ α Xµ ∂β X µ (15.25)
Adding the two variations and using a bit of index gymnastics, one can show that the energy-
momentum tensor takes the form [21]
i i
Tαβ = ∂α X µ ∂β Xµ + ψ̄ µ ρα ∂β ψµ + ψ̄ µ ρβ ∂α ψµ − (T race) (15.26)
4 4
where we have subtracted off the trace of the tensor by hand to maintain the fact that the
energy-momentum tensor is traceless. Using light-cone coordinates, the components of the energy-
momentum tensor become
i µ i µ
T++ = ∂+ Xµ ∂+ X µ + ψ+ ∂+ ψ+µ T−− = ∂− Xµ ∂− X µ + ψ− ∂− ψ−µ (15.27)
2 2
Moreover, just as in the bosonic case, the components T+− = T−+ = 0. On a similar note, in
light-cone coordinates the components of the supercurrent are
µ µ
J+ = ψ+ Xµ J− = ψ− ∂− Xµ (15.28)
Moreover, as a consequence of the identity ρα ρβ ρα = 0, one can show that
ρα Jα = 0 (15.29)
15.3 The World-Sheet and Superspace

As we have discussed throughout this text, the string world-sheet is a two dimensional space,
parameterized by the coordinates (τ, σ). In the last chapter we saw that supersymmetry makes
itself manifest when we extend space-time to include Grassmann coordinates called superspace.
Let’s apply this same notion to the world-sheet. That is, let us extend the two dimensional field
theory we have been working with so far by including Grassman coordinates θA , which form a two
component Majorana spinor. In this case we will denote our superfield as [21]
1
Y µ (σ, θ, θ̄) = X µ (σ) + θ̄ψ µ (σ) + θ̄θB µ (σ) (15.30)
2
where we have used σ to stand in for both world-sheet coordinates. To see how supersymmetry
becomes manifest, we use the differential operator representation of the supercharge. Based on the
expression developed in the previous chapter, we can convince ourselves that the correct operator
for the present case is
∂
QA = + i(ρα θ)A ∂α (15.31)
∂ θ̄A
Notice then that the supercharge operator acts on the world-sheet coordinates σ α as

α ∂
Qσ = α
+ i(ρ θ)∂α σ α = iρβ θδβα = iρα θ
∂ θ̄
where we have explicitly left out indices. Therefore, we find that SUSY transformation of the
world-sheet coordinates is simply
δσ α = [¯
Q, σ α ] = −i¯
ρα θ (15.32)
Moreover, in similar way one may also show that SUSY transformation of the Grassmann coor-
dinates is
δθA = [¯
Q, θA ] = A (15.33)
In this sense, we notice that supersymmetry can be realized as a type of geometrical trans-
formation, a result we are familiar with based on the studies of the last chapter. Moreover, the
supercharge can also be used to show that the variation of the superfield is just
δY µ = [¯
Q, Y µ ] = ¯QY µ (15.34)
Using (15.34) we can derive the SUSY transformations given in (15.10). Putting explicit indices
in, we find
15.3. THE WORLD-SHEET AND SUPERSPACE 365

µ µ ∂ 1 B B µ
δY = ¯QA Y = ¯ + i(ρα θ)A ∂α µ B µ
X (σ) + θ̄ ψ (σ) + θ̄ θ B (σ)
∂ θ̄A 2

µ ∂ 1 B B µ
= ¯ψ (σ) + ¯ A ρα θA ∂α X µ (σ) + i¯
θ̄ θ B (σ) + i¯ ρα θA θ̄B ∂α ψ µ (σ) + i¯
ρα θA θ̄B θB ∂α B µ (σ)
∂ θ̄ 2
We can actually drop the last term in this expression since the Grassmann quantities anticom-
mute. If we make use of the Fierz identity
1
θA θ̄B = − δAB θ̄C θC (15.35)
2
we may write the variation of the superfield Y µ as
δY µ = ¯ψ µ + θ(i¯
ρα ∂α X µ + ¯B µ ) + θθ̄(i¯
ρα ∂α ψ µ ) (15.36)
where we have intentionally suppressed indices. If we compare this to (15.30) we can simply read
off the variations of the fields constituting the superfield (sometimes called the component fields):
δX µ = ¯ψµ δψ µ = −iρα ∂α X µ + B µ δB µ = −i¯

ρα ∂α ψ µ (15.37)
These are just the SUSY transformations we came up with from before, with the minor ex-
ception of the additional component field B µ . If we set this field to zero, we obtain the previous
transformations.
Now let’s consider multiple superfields, Y1 ...Yk , obeying the same transformation law, δYk =
¯QYk . it turns out, any product of such superfields transforms in the same way as a single superfield
[21]. For example,
δ(Y1 Y2 ) = ¯Q(Y1 Y2 ) (15.38)

At first it doesn’t appear obvious that this is true, however since ¯Q is a first order differential
operator in superspace, it still obeys the Leibniz rule
¯Q(Y1 Y2 ) = ¯Q(Y1 )Y2 + Y1 ¯Q(Y2 ) (15.39)

All in all, we are allowed to conclude that the product of two superfields also transforms as a
single superfield.
As we mentioned briefly in the previous chapter, superspace is the more elegant approach to
studying supersymmetric theories. The reason for this is superspace makes it easier to look at
gauge fields and construct Lagrangians that are manifestly invariant under SUSY transformations.
This is also the same reason why we came up with the supercovariant derivative. In the present
case, the supercovariant derivative takes the form
∂
− iρα θ∂α
D= (15.40)
∂ θ̄
It satisfies the following anticommutation relations
{DA , QB } = 0 (15.41)
{DA , D̄B } = 2iρα

AB ∂α {DA , DB } = 2i(ρα ρ0 )AB ∂α (15.42)
Since the supercovariant derivative anticommutes with the supercharge, we can see why the
superspace approach is useful in constructing Lagrangians. Remember we had a superfield Y
transforming as δY = ¯QY . Since D anticommutes with Q , then we find that DA Y transforms
in the same way as the superfield. That is, the supercovariant derivative of a superfield is simply
another superfield, allowing us to construct Lagrangians with derivatives in them, as we know it
will remain invariant under SUSY transformations.
To put all of this to use, consider a particularly interesting action given by
Z
i
S= d2 σd2 θD̄Y µ DYµ (15.43)
16πα0
Remember, since θ is a Grassmann coordinate, the above action must make use of Grassmann
integration. To be able to evaluate the θ integrals we first note that the supercovariant derivatives
are simply
i
DY µ = ψ µ + θB µ − iρα θ∂α X µ + θ̄θρα ∂α ψ µ (15.44)
2
i
D̄Y µ = ψ̄ µ + B µ θ̄ + i∂α X µ θ̄ρα − θ̄θ∂α ψ̄ µ ρα (15.45)
2
where we have again made use of the Fierz identity. Therefore, D̄Y µ DYµ contains the following
terms quadratic in θ
i
∂α X µ ∂β Xµ θ̄ρα ρβ θ + B µ Bµ θ̄θ + (ψ̄ µ ρα ∂α ψµ − ∂α ψ̄ µ ρα ψµ )θ̄θ
2
With a bit of index gymnastics, the above just becomes
= (−∂ α Xµ ∂ α Xµ + iψ̄ µ ρα ∂α ψµ + B µ Bµ )θ̄θ

Then, using the fact that θ̄θ = −2iθ1 θ2 , we see that from our rules of Grassmann integration
Z
d2 θ̄θ = −4i
Evaluating the action in consideration, we are left with

Z
T
S=− d2 σ(∂α X µ ∂ α Xµ − iψ̄ µ ρα ∂α ψµ − B µ Bµ ) (15.46)
2
We notice that if the component field B µ = 0, we are left the action in (15.2), back to where we
started. Not is lost however. We readily see now why the action we constructed from the beginning is
in fact a supersymmetric action: it can be derived from superfields and the supercovariant derivative
using the superspace formalism of SUSY. Therefore, we now have a much better grasp to why the
action we chose at the beginning was automatically invariant under SUSY transformations. This
is because we created a Lagrangian that was manifestly invariant under SUSY transformations.
15.4. BOUNDARY CONDITIONS AND MODE EXPANSIONS 367
15.4 Boundary Conditions and Mode Expansions

In this section we will examine the boundary conditions of the RNS superstring and the mode
expansions of the fields ψ µ . Since X µ are the same bosonic fields from before, they obey the same
commutation relations, mode expansions, and boundary conditions that we have already developed
in the chapter on covariant quantization of the bosonic string. For this reason we will not go
over the details of the bosonic fields in this section. The method of finishing the classical physics
of the fermionic coordinates follows the same program as before: apply boundary conditions to
the fermionic fields and write down the mode expansions. For simplicity, let us use light-cone
coordinates, and examine the fermionic action, which, aside some unimportant constants, took the
form
Z
SF = d2 σ(ψ− ∂+ ψ− + ψ+ ∂− ψ+ ) (15.47)
For now, let’s consider the second term of the above action and vary it. We find
Z Z
δ d σψ+ ∂− ψ+ = d2 σ(δψ+ ∂− ψ+ + ψ+ ∂− (δψ+ ))
2
Integration by parts on the second term yields

Z Z ∞ σ=π Z
2
− d2 σ∂− ψ+ (δψ+ )

d σψ+ ∂− (δψ+ ) = dτ ψ+ δψ+
−∞ σ=0
A similar expression arises from the other term. Altogether, one may write the boundary terms
of the entire fermionc action as
Z ∞

δSF = dτ (ψ+ δψ+ − ψ− δψ− ) − (ψ+ δψ+ − ψ− δψ− ) (15.48)
−∞ σ=π σ=0
Remember that for open strings the boundary terms σ = 0 and σ = π must vanish independently
from one another. All in all, we require that ψ+ δψ+ − ψ− δψ− vanish at each endpoint of the string.
This is only satisfied when ψ+ = ±ψ− (and, naturally, δψ+ = ±δψ− ), at each end point. The
overall relative sign is really a matter of convention for the first endpoint at σ = 0. The literature
often sets
µ µ
ψ+ (τ, 0) = ψ− (τ, 0) (15.49)
We still must choose the relative sign between the other string endpoint. Making this choice
becomes physically meaningful, as we shall see momentarily. There are two cases to be considered.
The first choice would likely be the most natural one to make,
µ µ
ψ+ (τ, π) = ψ− (τ, π) (15.50)
We call this choice Ramond boundary conditions, and the strings that obey this boundary con-
dition are said to belong to the Ramond sector. We won’t go through the details here, however
µ
from these boundary conditions one can show that the mode expansions of the fermionic fields ψ±
is simply [21]
µ 1 X µ −in(τ −σ) µ 1 X µ −in(τ +σ)

ψ− =√ dn e ψ+ =√ dn e (15.51)
2 n∈Z 2 n∈Z
The alternative choice of boundary conditions is
µ µ
ψ+ (τ, π) = −ψ− (τ, π) (15.52)
These boundary conditions are formally known as the Neveu-Schwarz boundary conditions, and
strings obeying this set of boundary conditions are said to belong to the Neveu-Schwarz sector.
We have already seen how to create the mode expansion of fields that obey these type of boundary
conditions: they have half integer moding. Hence, the mode expansions of the fermionic fields
obeying NS boundary conditions are simply
µ 1 X µ −ir(τ −σ) µ 1 X µ −ir(τ +σ)

ψ− =√ br e ψ+ =√ br e (15.53)
2 1 2 1
r∈Z+ 2 r∈Z+ 2
As we will see later, the Ramond and Neve-Schwarz boundary conditions give rise to different
physical particles. Moreover, following the usual notation, the negatively moded coefficients are
creation operators. Therefore, bµ− 1 , bµ− 3 , ... are creation operators while the positively moded os-
2 2
cillators, bµ1 , bµ3 , ... are annihilation operators. These operators act on the Neveu-Schwarz vacuum,
2 2
which we choose to denote by |N Si. An identical treatment of the oscillators corresponding to
the Ramond sector yield that the negatively moded coefficients are creation operators while the
positively moded creation operators correspond to the annihilation operators.
For closed strings, the boundary terms will vanish when the boundary conditions are periodic or
antiperiodic for each component of ψ, from which we obtain the following mode expansions:
µ
X
ψ− = dµn e−2in(τ −σ) (15.54)
n∈Z
or
µ
X
ψ− = bµr e−2ir(τ −σ) (15.55)
r∈Z+ 12
Moreover, we also have
µ
X
ψ+ = d˜µn e−2in(τ −σ) (15.56)
n∈Z
or
µ
X
ψ+ = b̃µr e−2ir(τ +σ) (15.57)
r∈Z+ 12
Each of these boundary conditions for closed strings will allow us to pair up the different left
and right moving sectors of the strings to the R and NS sectors. All in all, there are four distinct
closed-string theory sectors that can be constructed which we will simply list for now: NS-NS,
NS-R, R-NS,R-R. We will go over more details on the physical meaning of these sectors shortly.
15.5. CANONICAL QUANTIZATION OF RNS SUPERSTRINGS 369
When we examined the classical physics of the bosonic string we ended up defining the Virasoro
operator Lm . In superstring theory we still have Virasoro operators, however when we quantize the
theory we require the so-called super-Virasoro operators [37]
∞
1 X 1 X
Lm = αm−n · αn + (2r − m)b−r · bm+r (15.58)
2 n=−∞ 4 1
r∈Z+ 2
In addition to the super Virasoro operators, we obtain other generators from the NS and R
sectors. For the NS sector, a generator arises from the supercurrent:
√ Z π ∞
2 X
Gr = dσeirσ J+ = αm · br+m (15.59)
π −π m=−∞
while for the R sector one obtains the generator

X
Fm = α−n · dm+n (15.60)
n∈Z
As one might suspect, these additional generators extend the Virasoro algebra in becoming the
Super-Virasoro algebra, which we will briefly discuss in the next section.
15.5 Canonical Quantization of RNS Superstrings

When we quantized the bosonic string, we used the commutation relations of X µ to come up with
the commutation relations of the oscillators αnµ :
µ
[αm , αnν ] = mδm+n,0 η µν (15.61)
We can use an almost identical procedure to determine the, this time, anticommutation relations
µ
of the oscillators bµr and dµm . Recall that the fermion fields ψA , A = ±, obey anticommutation
relations since they describe fermions:
µ
{ψA ν
(τ, σ), ψB (τ, σ 0 )} = πη µν δAB δ(σ − σ 0 ) (15.62)
As an exercise, the reader will show that a procedure similar to the one used to determine (15.58)
may be used to show that
{bµr , bνs } = η µν,0 {dµm , dνn } = η µν δm+n,0 (15.63)

One immediate consequence of this is that we find an interesting representation of the zero modes
dµ0 . Notice we have
{dµ0 , dν0 } = η µν
This algebra is almost the Dirac algebra we mentioned at the beginning of the chapter. Requiring
that the gamma matrices Γµ satisfy
{Γµ , Γν } = −2η µν
It follows that the zero modes dµ0 may be represented as
i
− √ Γµ = dµ0
2
As mentioned before, the Virasoro operators introduced in the previous section generate the
so-called super-Virasoro algebra. However the algebra is slightly different for each sector. For the
NS sector, we have the following relations [21]:
c 3
[Ln , Lm ] = (n − m)Lm+n + (n − n)δn+m,0 (15.64)
12
1
[Ln , Gr ] = (n − 2r)Gn+r (15.65)
2
c
{Gr , Gs } = 2Lr+s + (4r2 − 1)δr+s,0 (15.66)
12
where c is the central charge
D
c=D+
2
For the R sector, we instead have the relations
D 3
[Lm , Ln ] = (m − n)Lm+n + m δm+n,0 (15.67)
8
m
[Lm , Fn ] = − n Fm+n (15.68)
2
D 2
{Fm , Fn } = 2Lm+n + m δm+n,0 (15.69)
2
As one might suspect based on our experience with covariant quantization, we can find the
allowed states, the quantum constraints, for each sector. Let’s first consider the NS sector. Let |ψi
be a physical state in the NS sector. The NS sector super-Virasoro constraints end up being
(L0 − a)|ψi = 0 Ln |ψi = 0 Gr |ψi = 0 (15.70)

where a is some normal ordering constant and for n, r > 0. Alternatively, the constraints on the
physical states in the R sector are
(L0 − a0 )|ψi = 0 Ln |ψi = 0 Fm |ψi = 0 (15.71)

where a0 is some other ordering constant, and n > 0, m ≥ 0. If we compare this to our previous
conditions on physical states, we see that we merely have an extension of the constraints for the
covariantly quantized bosonic string.
We are actually in a position to determine the normal ordering constant a0 . Consider the anti-
commutator (15.69) with m = n = 0. We have then
{F0 , F0 } = F0 F0 + F0 F0 = 2F0 F0 = 2L0 ⇒ L0 = F02

15.6. CONSTRUCTING THE STATE SPACE FOR RNS SUPERSTRINGS 371
However, based on the constraints of a physical state, we have that F0 |ψi = 0. Therefore, it
follows
F0 (F0 |ψi) = 0 ⇒ L0 |ψi = 0

However, we also have
0 = (L0 − a0 )|ψi = L0 |ψi − a0 |ψi = −a0 |ψi

We may conclude then that it must be the ordering constant a0 for the Ramond sector is identi-
cally zero, a0 = 0.
Finally, we won’t go through the details here, but just as before we must do something to get
rid of ghost states that arise. This can be accomplished by looking for zero-norm spurious states.
As it happens, additional zero norm states appear when D = 10, and when the ordering constants
take the values a = 12 , a0 = 0. Therefore, in order to eliminate negative-norm states, we find that
the dimension of space-time, according to superstring theory is ten-dimensional instead of the 26-
dimensional space-time described in bosonic string. Remember this is the reason why an active
area of research is to find compactified spaces which curl up six spatial dimensions of a higher ten-
dimensional universe in order to make contact with our everyday experience with four-dimensional
space-time. For more details on this determination of D, the reader is pointed to text by Becker
and Schwarz [5].
15.6 Constructing the State Space for RNS Superstrings

Rather than using the covariant quantization approach to construct a state space, let us instead
use the light-cone gauge as it will be easier to identify meaningful physical states. This section is
heavily based on Zwiebach’s chapter on superstrings, so for more details the reader should review
his chapter.
Using light-cone analysis, the mode expansion of the open string Neveu-Schwarz fermions ψ I is
given by
1 X I −ir(τ −σ)
ψ I (τ, σ) = √ br e (15.72)
2 1
r∈Z+ 2
Again, since ψ I is anticommuting, the expansion coefficients bIr satisfy a very similar set of
anticommutation relations. Namely,
{bIr , bJs } = δr+s,0 δ IJ (15.73)

As usual, negatively moded expansion coefficients represent the creation operators while the
positively moded expansion coefficients represent the annihilation operators. When it comes to
constructing a state space for the NS fermions, it is imperative to point out that since all of the
creation operators anticommute with one another and square to zero, each creation operator bI−r
for positive r can appear at most once in any string state. Moreover, since X I (τ, σ) are quantized
I
in same way as before, we still maintain that α−n represents a creation operator for the bosonic
string. Putting everything together, it’s easy to see that the states of the NS sector are of the form
[64]
9 Y
Y 8 9
Y 8
Y
I
|λi = (α−n )λn,I (bJ−r )ρr,J |N Si ⊗ |p+ , p~T i (15.74)
I=2 n=1 J=2 r= 21 , 32 ,...
where ρr,J are either zero or one. To interpret the states of the physical spectrum, it is useful
to use the mass-squared operator. For the NS sector, before normal ordering, the mass-squared
operator is just
 
1 1 X
I 1 X
M2 = 0  α−p αpI + rbI−r bIr  (15.75)
α 2 2 1
p6=0 r∈Z+ 2
To determine the ordering constant, we can use the same method as before where we exploit the
Zeta function. For the bosonic oscillators αI we already know that for each coordinate there is a
1 1
contribution of − 24 to the ordering constant a. Let’s denote this by saying aB = − 24 . Looking at
the mass-squared formula, we readily see that the term which requires reordering is
1 X 1 X
rbI−r bIr = (−r)bIr bI−r
2 2
r=− 21 ,− 32 ,... r= 12 , 32 ,...

1 X 1 1 3 5
= rbI−r bIr − (D − 2) + + + ...
2 2 2 2 2
r= 21 , 23 ,...
In the first step we made the substitution of r → −r and in the second step we made use of the
anticommutation relations in (15.73). Moreover, earlier we saw that the sum of the positive odd
1
integers yields an overall factor of 12 . Therefore, we may write
1 X 1 X 1
rbI−r bIr = rbI−r bIr − (D − 2)
2 2 48
r=− 12 ,− 23 ,... r= 12 , 32 ,...
This allows use to conclude that for each fermion in the NS sector contributes to the overall
1
normal ordering constant a factor of aN S = − 48 . The full ordering constant for the mass-squared
operator then is

1 1 1
a = (D − 2) − − = −(D − 2)
24 48 16
Since we know with SUSY we are dealing with a ten dimensional space-time, D=10, resulting in
a = − 21 . Therefore, the mass-squared operator becomes

1 1
M2 = 0 − + N⊥ (15.76)
α 2
where we have defined the number operator
∞
X 1 X
N⊥ = I
α−p αpI + rbI−r bIr (15.77)
p=1
2
r= 21 , 32 ,...
15.6. CONSTRUCTING THE STATE SPACE FOR RNS SUPERSTRINGS 373
To see the physical implications, let’s examine some of the specific states. As usual, first suppose
N ⊥ = 0. We therefore have a single quantum state, |N Si ⊗ |p+ , p~T i, with a mass-squared valued
of α0 M 2 = − 21 . Since this state describes a particle with imaginary mass, we recognize that this
state represents a tachyon.
The first excited state is the massless state, where N ⊥ = 12 , and therefore, α0 M 2 = 0. In this
state, there is only one creation operator, bI− 1 , and therefore our excited state is bI− 1 |N Si⊗|p+ , p~T i.
2 2
There are in fact eight massless states since we have a creation operator for each value of the light-
cone index I.
For completeness, consider the next excited state, where N ⊥ = 1. In this case the mass-squared
operator takes on the value α0 M 2 = 12 , and the possible states include α−1 I
|N Si ⊗ |p+ , p~T i, and
I J +
b− 1 b− 1 |N Si ⊗ |p , p~T i.
2 2
For the purposes of particle physics, it is typically useful to have an operator which distinguishes
between states representing bosons and fermions. We call this operator (−1)F , where F is the
so-called fermion number. If F is any odd integer, we notice that the operator yields minus one.
We say that if the operator takes on a value of −1 after acting on some arbitrary quantum state the
state is fermionic. On the other hand, if the operator yields +1, the state is said to be fermionic.
This operator is sometimes referred to as G-parity due t its historical significance in trying to adapt
string theory to explain strong interactions. To explicitly calculate (−1)F on any given state, we
must first know the eigenvalue of the G-parity operator on the ground state |N Si⊗|p+ , p~T i. Mostly
due to convention, let’s define it to be
(−1)F |N Si ⊗ |p+ , p~T i = −|N Si ⊗ |p+ , p~T i (15.78)

F
With this declaration, the eigenvalue of (−1) on an arbitrary state is equal to minus one times
the sequence of factors of minus one, one for each fermionic oscillator which appears to act on the
arbitrary state. Therefore, for the state |λi, we find that the eigenvalue of the G-parity operator is
simply
P
ρ
(−1)F |λi = −(−1) r,J r,J |λi (15.79)
Let’s consider again the few states we examined above. We saw that fermionic oscillators, as
they are moded by half-integers, contribute half-integers to the number operator N ⊥ . Moreover,
as we noticed was the case for N ⊥ = 1, states with integer values of the number operator must
have an even number of fermionic oscillators. It follows then that all states with integer values of
the number operator have a G-parity of −1, and are therefore fermionic. On the other hand, all
states with half integer values of N ⊥ have an odd number of fermionic oscillators, and therefore
have a G-parity of +1, and are therefore bosonic states. We will come back to this discussion on
bosonic and fermionic, but first let’s examine the physical spectrum of the Ramond sector using
the light-cone gauge.
With Ramond boundary conditions, we can expand the Ramond fermions as

1 X I −in(τ −σ)
ψ I (τ, σ) = √ dn e (15.80)
2 n∈Z
Again, we use the notation that negatively moded expansions coefficients are creation opera-
tors. Moreover, since ψ I is anticommuting, the creation and annihilation operators satisfy the
anticommutation relations:
{dIm , dJn } = δm+n,0 δ IJ (15.81)

As it happens, the fermions of the Ramond sector are slightly more complicated since there
are actually eight fermionic zero modes, dI0 . We end up grouping these eight operators into linear
combinations of four creation operators ξ1 , ξ2 , ξ3 , ξ4 , and four annihilation operators, which we won’t
care about labeling for our present discussion. Since we are dealing with zero mode oscillators, these
creation operators don’t actually contribute to the mass-squared operator. Rather, let us suppose
some alternative vacuum, |0i, from which the creation operators construct 24 = 16 degenerate
Ramond ground states. It’s easy to work out that eight of these states have an even number of
ξ creation operators acting on the vacuum |0i, while the other eight states have an odd number
of creation operators acting on the vacuum. We denote the eight states with an even number of
creation operators as |Ra i, while we label the other eight states by |Rā i [64]. Putting together
each set of states together, we find a full set of Ramond sector ground states, |RA i. Therefore, an
arbitrary state |λi in the Ramond sector takes the form
9 Y
Y 8 ∞
9 Y
Y
I
|λi = (α−n )λn,I (dJ−m )ρm,J |RA i ⊗ |p+ , p~T i (15.82)
I=2 n=1 J=2 m=1
Similar to the NS sector, the Ramond sector also has a G-parity operator, however with the
additional piece
(−1)F |0i = −|0i

As an exercise, the reader prove that the eight states comprising |Ra i are fermionic, while the
eight states comprising |Rā i are bosonic. Moreover, the mass-squared operator associated with the
Ramond sector, before normal ordering, is given by
 
1  1X I 1 X
M2 = α−p αpI + ndI−n dIn  (15.83)
α0 2 2
p6=0 n∈Z
From this expression we readily see that the term which requires normal ordering is
1 X 1 X
ndI−n dIn = − ndI dI
2 n=−1,−2,... 2 n=1,2,... n −n
1 X 1
= ndI−n dIn + (D − 2)
2 n=1,2,... 24
1
Therefore we find that for each Ramond fermion there is a contribution of 24 to the ordering
constant of M . This number exactly cancels the contribution from the bosonic fields X I , hence
2
the ordering constant of the Ramond sector is zero, as expected from our analysis in the previous
section. Therefore, the mass-squared operator becomes
1 X I I
M2 = α−n αn + ndI−n dIn

(15.84)
α0
n≥1
15.7. GENERATING FUNCTIONS AND GSO PROJECTION 375
What this means is the Ramond ground states are massless, α0 M 2 = 0, contrary to the ground
states of the NS sector. Hence, the Ramond sector is void of a tachyon! This result is imperative
in constructing a realistic string theory since most physicists abhor the existence of such particles.
The next mass level has α0 M 2 = 1, coming from the states α−1
I
|Ra i, dI−1 |Ra i, and α−1
I
|Rā i, dI−1 |Rā i.
Notice that we have an equal number of states between the two sets |Ra i and |Rā i. Hence, at the
first two mass levels, we have that there are an equal number of bosonic and fermionic states. As an
exercise the reader will list the possible states for the mass level α0 M 2 = 2. There the reader will
find that indeed there are the same number of bosonic and fermionic states. Remember from the
last chapter what we said was one of the most characteristic features of supersymmetry: at every
mass level, there are an equal number of bosonic and fermionic states. It would seem then, though
we have not proved it, the Ramond sector points out we indeed have a supersymmetric theory. This
of-course is the world-sheet supersymmetry we have been working with this entire time. We would
also like our superstring theory to be space-time supersymmetric. As we will see momentarily, this
only happens if we can combine the states from the NS and R sectors in a physically meaningful
way.
15.7 Generating Functions and GSO Projection

Before we get into some of the finer points of GSO projection and space-time supersymmetry, we
first must have a brief math lesson in generating functions. Essentially, a generating function is
a power series where the coefficients themselves encode information about a sequence of numbers.
Specifically, an ordinary generating function is a power series,
∞
X
f (x) = an xn (15.85)
n=0
there gives information about a sequence of numbers through the coefficients an . In short, a
generating function is a type of power series where we care about the coefficients themselves rather
than the indeterminate x. For example, consider the sequence 1, 1, 1, 1, 1.... The generating function
for this sequence must have all of its coefficients equal an = 1 for all n. Explicitly, we would write
∞
X
f (x) = an xn = a0 + a1 x + a2 x2 + a3 x3 + ...
n=0
Since we are considering the sequence 1, 1, 1, ..., we have that a0 = a1 = a2 = ... = 1, and
therefore the generating function for the sequence of numbers 1, 1, 1, 1... is simply
1
f (x) = 1 + x + x2 + x3 + ... =
1−x
Suppose we first considered this power series. We would immediately recognize that the value of
all the coefficients was 1, and therefore we would see that this function is the ordinary generating
function for the sequence 1, 1, 1, 1, .... As another example, consider the sequence 1, 0, 1, 0, 1, 0, ....
The generating function describing this sequence must have that the coefficients of the power series
alternate between 1 and 0. Explicitly we would write
1
(1)1 + (0)x + (1)x2 + (0)x3 + (1)x4 + ... = 1 + x2 + x4 + x6 + ... =
1 − x2
1
Therefore, the ordinary generating function describing the sequence 1, 0, 1, 0, 1, ... is 1−x2 .
In string theory we actually use generating functions to count the number of states we have
in the NS and R sectors. Specifically, we use ordinary generating functions where the coefficients
actually encode the number of states of in the R and NS sectors at each mass level. Consider the
power series
∞
X
f (x) = a(n)xn
n=0
where a(n) is the number of states and we let n = N ⊥ . To see how this works, suppose we have
a single creation operator a†1 . We therefore would have one state with N ⊥ (|0i, the vacuum), one
state with N ⊥ = 1 (a†1 |0i, first excited state), and so forth. In fact, there is only one state where
N ⊥ = k, some integer ((a†1 )k |0i). Therefore, in a system with a single creation operator, for every
integer value of N ⊥ there is one state. Since we started by saying our generating functions have
the coefficients encoding the number states of our physical system, and where n = N ⊥ , we have
that the generating function describing this specific example is just
∞
X ⊥ 1
f1 (x) = a(N ⊥ )xN = 1 + x + x2 + x3 + ...
1−x
N ⊥ =0
Now consider the case when we have just the creation operator a†2 with mode number 2. Again,
we have one vacuum state for N ⊥ = 0, but, since we only have a single mode number two creation
operator, there are no states for N ⊥ = 1. There is however one state with N ⊥ = 2. In fact, there
is one state for each even value of N ⊥ . The ordinary generating function describing this system is
therefore
1
f2 (x) = 1 + x2 + x4 + x6 + ... =
1 − x2
Now suppose we wanted to come up with the generating function where the coefficients encode
the number of states built using both creation operators a†1 and a†2 . To obtain this generating
function we would form products where the first factor is a state built solely with a†1 creation
operators and the second factor is a state built solely out of a†2 creation operators. One reason for
this is remember for N ⊥ = 2, states are built either out of two a†1 oscillators or one a†2 oscillator.
Symbolically then, we would imagine the product

1 + a†1 x + (a†1 )2 x2 + (a†1 )3 x3 + ... 1 + a†2 x2 + (a†2 )2 x4 + (a†2 )3 x6 + ...
Multiplying out a few terms we see that we have
1 + a†1 x + ((a†1 )2 + a†2 )x2 + ((a†1 )3 + a†1 a†2 )x3 + ...

If we continue with this muliplication, we would see indeed that the product of the generating
functions f1 (x) and f2 (x) yields the generating function that encodes the number of states for a
system with two creation operators a†1 and a†2 . Hence,
1 1
f12 (x) = f1 (x)f2 (x) =
1 − x 1 − x2
To generalize this notion, suppose we used oscillators a†1 , a†2 , a†3 , ... of all modes, we would find
that the ordinary generating function describing this system is just [64]
∞
Y 1
f (x) = (15.86)
n=1
1 − xn
Let’s use these ideas to see if we can count the number of states of bosonic open string theory.
When we used the light-cone analysis we had a creation operator of all modes have they came in 24
species, one for each light-cone index (aI†
n for all n). Since each species would give its own ordinary
generating function, we find that the full generating function for bosonic open string theory is given
by
∞
Y 1
(15.87)
n=1
(1 − xn )24
This is a good start as we can count the number of states using the number operator N ⊥ ,
however we would have an easier time interpretating the number of type of particle states if we
could use a generating function based on the mass squared operator α0 M 2 , where the coefficients
of our generating function counts the number of states for a given mass level. Recall that for open
bosonic string theory, we had that α0 M 2 = N ⊥ − 1. Therefore, our generating function would take
the form
∞ ∞
X ⊥ 1 X ⊥
f (x) = a(N ⊥ − 1)xN −1
= a(N ⊥ − 1)xN
x
N ⊥ −1=0 N ⊥ −1=0
Hence, for open bosonic string theory, the generating function is

∞
1 Y 1
f (x) = (15.88)
x n=1 (1 − xn )24
Writing this out we find [72]

1
+ 24 + 324x + 3200x2 + ...
f (x) = (15.89)
x
Remember, the coefficients give us the number of states for a given mass-level, given by the
exponents of the indeterminant x. From the above, we see that in open bosonic string theory there
is one tachyon state, M 2 = − α10 , 24 massless states, and 324 massive states with α0 M 2 = +1.
These results exactly match what we found from before when we first quantized the string.
That’s enough background for what we really would like to use generating functions to do:
counting the states of the NS and R sectors. To do this we must first become familiar with counting
states using fermionic (anticommuting) oscillators. Again, let’s start with the number operator N ⊥
and a single fermionic creation operator f−r . Due to the anticommuting nature of we only have two
states: |0i, and f−r |0i. Therefore, the ordinary generating function for a single fermionic oscillator
is simply
f (x) = 1 + xr (15.90)
Remember, since N ⊥ in the RNS formalism may take on half-integer values it is accurate to say
that the generating function for a system consisting of only the oscillator f− 12 is
√
1+ x
In the NS sector we have the creation operators bI− 1 , bI− 3 , ... which come in eight different species
2 2
(for each light-cone index). Hence, the generating function associated with all of these creation
operators is merely
h ∞
i8 Y 8
1 3 1
(1 + x 2 )(1 + x 2 )... = 1 + xn− 2 (15.91)
n=1
Again, since having a generating function associated with the mass-squared operator is more
useful, we remind ourselves that for the NS sector α0 M 2 = N ⊥ − 21 , as well as keep in mind that
we still have eight bosonic coordinates to deal with, ultimately yielding the ordinary generating
function for the NS sector:
∞ 1
!8
1 Y 1 + xn− 2
fN S (x) = √ (15.92)
x n=1 1 − xn
If we were to expand this out we would find
1 √
fN S (x) = √ + 8 + 36 x + 128x + ... (15.93)
x
From here we see that we have a single tachyon state with α0 M 2 = − 21 , eight massless states, and
so forth. For the Ramond sector, we have that α0 M 2 = N ⊥ . Though it seems reasonable enough, as
an exercise the reader is asked to provide detailed arguments why the ordinary generating function
associated with the Ramond sector is
∞ 8
Y 1 + xn
fR (x) = 16 (15.94)
n=1
1 − xn
Expanding out this series we find
fR (x) = 16 + 256x + 2304x2 (15.95)

If we compare this expansion to the expansion for fN S , we notic that for each of the integer
values of α0 M 2 , there are twice as many states in the Ramond sector than there are in the NS
sector. Now that we have the generating functions associated with each sector of RNS superstrings,
we are in a position where we may discuss space-time supersymmetry.
As noted in the previous section, we saw that the Ramond sector of superstring theory exhibits
the fact we are dealing with world-sheet supersymmetry: we had an equal number of fermionic and
bosonic states at each mass level. It turns out however, the RNS model, as described so far, is an
inconsistent quantum theory unless we impose further conditions. We must actually truncate some
of the states in the spectrum in a very specific way using a method proposed by Gliozzi, Scherk,
and Olive [21]. We call this method GSO projection.
There are a few rather obvious arguments that suggest this truncation. First, in the NS sector we
continue to see the persistence of a tachyon, a particle representing an instability in the vacuum and
creating all sorts of problems with causality that we would like to rid our theory of it. Moreover,
from our light cone analysis, the zero modes dI0 carry a Lorentz index and therefore form a Lorentz
vector, transforming as such. However, under Lorentz transformations the states |Ra i and |Rā i
transform as spinors, which is the appropriate way for space-time fermions to transform. But this
means we have two different types of fermions in ten-dimensional space-time. This is odd since each
set of states has a different G-parity (−1)F , and a different commuting character. More importantly,
with two space-time fermions we immediately don’t get space-time SUSY, as space-time SUSY is
comprised of having bosons with their fermionic superpartners, and fermions with their bosonic
superpartners.
Therefore the strategy to truncate the spectrum comes into light. First, we see that all fermions
must arise from string states with the same G-parity. This is the first part of GSO projection:
we truncate the Ramond sector down to only include states with (−1)F . These states, initially
recognized to be the world-sheet fermions are now recognized as space-time fermions. We call this
the R− sector. Upon GSO projection, the generating function for the Ramond sector is modified
to
∞ 8
Y 1 + xn
fR− (x) = 8 (15.96)
n=1
1 − xn
Let’s now move on to the states in the NS sector. We recognize that the ground states are
tachyonic with a G-parity of (−1)F = −1 and therefore desire to truncate away these states.
Therefore, GSO projection calls for truncating the NS sector to the set of states with (−1)F = +1,
comprising the so-called N S+ sector. Moreover, since now states in the NS sector carry a spinor
index, we recognize that, after GSO projection, we are left with a spectrum that include space-time
bosons.
Remember our initial goal was to determine whether or not we have space-time SUSY. First we
had to truncate the states in the sector in order to have a consistent quantum theory, but now we
can determine whether we have space-time SUSY. First we need the generating function for the
states in the N S+ sector. Then, if the two generating functions fN S+ and fR− are equivalent, we
know that we have and equal number of space-time bosons and fermions are each mass level. To
come up with the generating function for the N S+ sector, we change the sign inside each factor in
the numerator. Doing so only changes the sign of each term in the generating function who states
arise from an odd number of fermions. But these are just the states we want to keep. Subtracting
off all of the other states and dividing by two to avoid double counting, we find that the generating
function of the N S+ sector is simply
 !8 !8 
∞ 1 ∞ 1
1  Y 1 + xn− 2 Y 1 − xn− 2
fN S+ (x) = √ −  (15.97)
2 x n=1 1 − xn n=1
1 − x n
As mentioned above, in order to have space-time SUSY we require that fR− = fN S+ . Or,
explicitly,
 !8 !8 
∞ n− 12 ∞ n− 12 ∞ 8
1 Y 1+x Y 1−x Y 1 + xn
√  − =8 (15.98)
2 x n=1 1 − xn n=1
1 − xn n=1
1 − xn
Just when one might think all hope is lost, there is a shinning light from the world of mathematics!
The mathematician Carl Gustav Jacob Jacobi in 1829, when writing his treatsie on elliptic curves
actually proved this very statement, referring to it as a rather odd and seemingly out of place
identity. Little did he know, his work has made a formidable contribution to the world of superstring
theory.
As it happens the GSO projection and the counting of states is certainly intriguing and seems to
point to space-time SUSY in string theory, it is not an actual proof. To be clear, string theory does
invoke space-time SUSY through an alternative, more elegant method known as the Green-Schwarz
formalism. In this formalism one applies Grassmann coordinates to space-time rather than the
world-sheet coordinates, just as we did in the last chapter when we studied superspace. Ultimately
this is done by adding in new superfields which map the world-sheet to fermionic coordinates,
mapping the entire world-sheet to superspace. Unfortunately, the technical details require the use
of a type of symmetry known as Kappa symmetry, a local fermionic symmetry, which goes beyond
the scope of this text. All in all, using the GS formalism, one is able to construct a string theory that
is space-time supersymmetric rather than the RNS formalism, which relies on world-sheet SUSY.
For more details on the GS approach, the reader is pointed to the text by Becker and Schwarz.
15.8 A Summary of Superstring Theory

For the most part, we have only been talking about open superstrings. Since string theory also in-
cludes closed strings, we must also design a supersymmetric string theory which includes supersym-
metric closed strings. As one might suspect, this is done roughly by combing by tensor producting
left and right moving copies of open superstring theory. As mentioned before, we have four possible
combinations of the R and NS sectors: (NS,NS), (NS,R), (R,NS), and (R,R). In order to get the a
closed string theory with supersymmetry we must apply the same GSO projection to truncate the
four combinations. A consistent truncation occurs if we truncate the left and right sectors sepa-
rately. For instance, if have the left sector be composed of the combination (N S+, R− ) and have
the right sector composed of (N S+, R+). If we combine these states multiplicatively we find four
sectors which constitute type IIA superstring theory: (N S+, N S+), (N S+, R+), (R− , N S+), and
(R− , R+). Type IIA superstring theory has no tachyon but keeps the graviton, the Kalb-Ramond
field, and the dilaton.
Alternatively, if we compose the left sector out of (N S+, R− ), and the right sector out of
(N S+, R− ), we end up with four sectors constituting type IIB superstring theory: (N S+, N S+), (N S+, R− ),
(R− , N S+), and (R− , R− ). In both type IIA and type IIB superstring theories, the sector (N S+, N S+)
describe bosons, and actually are the same between each theory. However, the (R, R) bosons in
type IIA superstring theory include a Maxwell field and a three indexed gauge field Aµνρ , while the
(R, R) bosons in type IIB superstring theory include a scalar field A, a Kalb-Ramond field Aµν ,
and a four indexed gauge field Aµνρσ . The point is, the bosons arising from the (R, R) sectors
in each type II theory are different. The difference actually lies in the fact that partners in type
IIA superstring theory have opposite chirality, while in type IIB superstring theory have the same
chirality.
Let’s now summarize what we have found. We knew all along that bosonic string theory had
deficiencies as it did describe fermions. To overcome this, we first used to RNS formalism, which
makes use of world-sheet supersymmetry, to include fermions. This eventually led us look for a
string theory which included space-time supersymmetry. We saw evidence of this through GSO
15.9. EXERCISES 381
projecting the states of the Ramond and Neveu-Schwarz sectors. Two important consequences
happened when we introduced SUSY. First, to eliminate ghost states we were required to introduce
a space-time that is ten dimensional. Second, we were able to eliminate the tachyon from our a
theory, making a far more realistic theory.
There is an issue that remains however. As we just saw, we were able to construct two closed
superstring theories, type IIA and type IIB. In summary, type IIA theory includes closed oriented
strings with N=2 SUSY and has a U (1) gauge symmetry. Since it does not include SU (2) gauge
symmetry, type IIA string theory cannot actually describe the weak or strong nuclear forces. Type
IIB superstring theory is another N=2 supersymmetric theory describing oriented closed strings
however can only describe gravity.
Though we won’t cover the details here, there are two other superstring theories known as the
heterotic superstring theories. At first glance this theory is rather bizarre. It also describes closed
superstrings, with N=1 SUSY, however combines a left sector comprised out of 26-dimensional
bosonic string theory with a 10-dimensional superstring right sector using the GS formalism. In
order reconcile the difference between space-time dimensions one takes the extra 16 dimensions of
bosonic string theory and regards them as abstract mathematical entities, similar to superspace.
There are in fact two heterotic string theories, one correpsonding to the gauge group SO(32), and
the other corresponding to the group E8 × E8 . Finally, there is a fifth superstring theory, known
as type I superstring theory, which describes open unoriented superstrings.
All in all, by the late 1980’s, there were five superstring theories. Each had its accomplishments
and deficiencies, but there were five individual, seemingly separate theories. What were physicists to
do? Back in chapter 13 we examined duality symmetries, particularly T-duality. It was found that
T-duality actually realted the two type II superstring theories, and two heterotic string theories.
Moreover, in the early 1990’s, Ed Witten came onto the scene revealing that each of the five
superstring theories were really all part of a more underlying theory, M-theory. In a way similar to
T-duality, Witten showed that each theory was simply a part of M-theory, and all of the five string
theories were related to one another through M-theory. It is imperative to point out that M-theory
itself, as far as we understand it now, is not a string theory, but rather a more fundamental, a mostly
mysterious theory linking all of the string theories into a unified theory. M-theory remains to be
one of the most active areas of modern string theory as it seems to hold the key to in discovering
a unifiying theory.
15.9 Exercises
1. Using (15.3), prove the Dirac algebra given in (15.4).
2. (a) Prove (15.13), further indicating that SUSY transformations we came up with do behave
as we expect.
3. Prove that the variation of the bosonic part of the Lagrangian density given in (15.2) takes
the form in (15.25). Proving this helps us write the energy-momentum tensor in the form given in
(15.26).
µ
4. (a) Give arguments why the mode expansions for ψ± take the form as expressed in (15.51)
and (15.53).
(b) Using the anticommutation relations given in (15.62), prove that the expansion coefficients
satisfy the anticommutation relations presented in (15.63).
5. (a) Show that the G-parity anticommutes with the oscillators:
{(−1)F , dIn } = {(−1)F , bIr } = 0
(b) Prove that the eight states comprising |Ra i are fermionic while the eight states comprising
|Rā i are bosonic.
(c) List all of the possible states in the Ramond sector for the mass level α0 M 2 = 2. Prove
that there are the same number of bosonic and fermionic states. What does this reveal about the
Ramond sector?
6. (a) Prove that the generating function describing the Ramond sector fR is given by (15.94).
(b) Discuss the motivations for GSO projection and what are the consequences of this method.
Part III
General Relativity and String

Theory
383
Chapter 16
The Thermodynamics of Strings
The goal of the next handful of chapters is to provide a thorough introduction to the issues of
black holes from the standpoint of classical general relativity, and one of the (some say the only)
triumphs of string theory. As the reader is likely aware, black holes, when examined from a quan-
tum mechanical perspective, thermally radiate. We will go into details on the consequences of this,
but for now since black holes emit thermal radiation there is a set of thermodynamical laws on the
mechanics of black holes. Moreover, as we are well aware from our preliminary studies of thermody-
namics, there is a second approach, statistical mechanics, which provides a more microscopic view
of thermodynamic entities. Therefore, if we believe that black holes obey a set of thermodynamical
laws, there should also be a statistical approach in determining, for instance, the entropy of a black
hole. Strominger and Vafa did just that in the context of superstring theory.
Our path to understanding the basics of the derivation of these laws along with the microscopic
origin of black hole entropy is a lengthy one as we will not cut corners in understanding the funda-
mentals of general relativity and black holes. We will examine the basics of differential geometry,
including tensor calculus and differential forms, and then move to the motivation and construction
of Einstein’s field equations, aiming to have a background that will allow us to investigate black
holes. Before we get to the details of these subjects however, let us first look at the thermodynamics
of strings, as it is through string theory we will be able to acheive the wonderful result of black hole
entropy from a statistical approach.
16.1 A Brief Review of Thermal and Statistical Physics

The goal of this chapter is to come up with the partition function for a single string. The majority of
this chapter primarily considers bosonic strings in the light-cone gauge as this is what we are most
familiar with. We will however briefly examine superstrings. Once we find the string partiation
function, we will be able to use our basic understanding of stastical mechanics to come up with
an expression for the entropy of a string. Before we do that, let’s first review some of the basic
elements of thermal and statistical physics.
First, a physicist’s favorite subject, thermodynamics.Thermodynamics is the subject aiming to

understand the macroscopic quantities of a thermal system, such as temperature, heat, entropy,
385
386 CHAPTER 16. THE THERMODYNAMICS OF STRINGS
and so forth. In a typical course on thermodynamics, one of the first rules we learn is the first law
of thermodynamics which is often given in the form of the total differential
dE = T dS − pdV (16.1)
where E is the energy of the system, T is the temperature, S is the entropy, p is the pressure,
and V is the volume. Sometimes we write the internal energy U instead of E, however in our
context it will be clear by what we mean. Each term in the first law has a physical meaning. T dS
is often viewed as the heat transferred into the system, while the term −pdV is the work done on
the system. Moreover, we see that our total differential on E is expressed in terms of two other
total differentials, namely dS and dV . Therefore, it is safe to conclude that E = E(S, V ). If we
take this position, from a mathematical standpoint we would write the total differential of E as

∂E ∂E
dE = dS + dV (16.2)
∂S V ∂V S
When we compare this to (16.1) we readily identify

∂E ∂E
T = p=− (16.3)
∂S V ∂V S
From here we see that we could measure the temperature of a system if we change the energy
of the system by changing the entropy while keeping the volume constant. Similarly, we could
measure the pressure if we change the energy of the system through a change in volume, keeping
the entropy constant. In principle this task might sound trivial, but in reality these measurements
might involve elaborate experiments. Often one will use a Maxwell relation to find an alternative
measurement one can make to calculate these macroscopic thermodynamical quantities.
Another useful quantity is the Helmholtz free energy F . We can derive this energy in the following
way: notice that the first law may also be written as
dE = d(T S) − SdT − pdV (16.4)

yielding
d(E − T S) ≡ dF = −SdT − pdV (16.5)

where we have defined the Helmholtz free energy
F ≡ E − TS (16.6)
Moreover, using a similar strategy as above, one can show that

∂F ∂F
S=− p=− (16.7)
∂T V ∂V T
We will come back to the free energy momentarily, but first let’s briefly review statistical me-
chanics. In statistical physics, particularly when we study strings, there are two ensembles we use
to describe the thermodynamical system: microcanonical and canonical. A microcanonical ensem-
ble consists of an isolated system A of fixed volume, and a fixed number of particles where the
macrostate is at a fixed energy and each microstate has an equal probability [53]. By macrostate
16.1. A BRIEF REVIEW OF THERMAL AND STATISTICAL PHYSICS 387
we mean a state of a system which refers to the macroscopic properties of the system, and by
microstate we mean a particular configuration the system is allowed to take with some probability.
We denote the possible number of microstates of the system at a fixed energy E by Ω(E). The
entropy of the system is then defined in terms of the number of microstates in Boltzmann’s famous
formula
S(E) = klnΩ(E) (16.8)

where k is Boltzmann’s constant. What this equation tells us is that as the number of microstates
of the system increases, the entropy increases logarithmically.
A canonical ensemble consists of a system A which has a fixed volume, however is in contact with
a thermal reservoir at some temperature T . Contrary to the microcanonical ensemble, a canonical
ensemble is a distribution of microcanonical ensembles, which has a constant avergae energy rather
than a constant energy. To look at the average energy of the canonical ensemble, and find the
connection between the statistical mechanics and thermodynamics, we require another function the
reader is likely familiar with: the parition function. To begin, suppose we have two states. The
probability of finding an atom in any particular state is directly proportional to the number of
microstates accessible to the reservoir in contact with our system. The ratio of the probabilities of
any two states is [53]
P(α2 ) ΩR (α2 )
= (16.9)
P(α1 ) ΩR (α1 )
Using Boltzmann’s formula (16.8), we may write
SR (α2 )
P(α2 ) e k (SR (α2 )−SR (α1 ))
= SR (α1 ) = e k
P(α1 ) e k
Notice that the exponent has the change in entropy between the two states. Rearranging the
first law (16.1) and noting tht typically pdVR is much smaller than dE, we have that
1
dSR = dE (16.10)
T
This allows us write the difference in entropies as
1
SR (α1 ) − SR (α1 ) = (E(α2 ) − E(α1 ))
T
The ratio of probabilities then becomes
E(α2 )
P(α2 ) e− kT
= E(α1 )
(16.11)
P(α1 ) e− kT
Each of these exponential factors are called the Boltzmann factor. Manipulating this expression
we find
P(α2 ) P(α1 )
E(α )
=E(α2 )
(16.12)
− kT2
e e− kT
Notice that the left hand side of the above equation is independent of α1 . For the equation to be
true we therefore require the right hand side is also independent of α1 . A similar argument holds
for α2 . Both sides then are independent of both α1 and α2 and in fact must be equal to a constant
for all states. We call this constant Z1 . In general, we may write the probability of any state α as
1 − E(α)
P(α) =
e kT (16.13)
Z
Moreover, using the fact that the sum of all of the probabilities of system must equal one, we
find
X 1 X − E(α)
1= P(α) = e kT
α
Z α
which gives us
X E(α)
Z= e− kT (16.14)
α
We call this function the partition function Z. By definition, the parition function depends on
the temperature T and external parameters of a system. In this chapter, the only systems we
consider have the volume occupied by the system as the only external parameter. Therefore we will
only consider parition functions which depend on both the temperature and volume of the system.
We can calculate the average energy of the system E by computing a weighted sum of the energies
of each microstate
X
E= Pα Eα (16.15)
α
where we use subscripts to ease on the notation. If we subsitute the parition function and
probability Pα in the above expression, we find that the average energy may also be calculated
using
X X e−βEα 1 X
E= Pα Eα = Eα = P −βE
e−βEα Eα
α α
Z α e α
α
!
∂ X ∂lnZ
= − ln e−βEα =− (16.16)
∂β α
∂β
1
where we used β = kT .
The Hemlholtz free energy may also be written in terms of the partition function. Using (16.6)
and (16.7) we may write

∂F F −E
= (16.17)
∂T V T
Now, let us suppose another function, F̃ = −kT lnZ. Notice that
∂ F̃ ∂ ∂
= (−kT lnZ) = −klnZ − kT lnZ
∂T ∂T ∂T
If we use the chain rule to rewrite the derivatives in terms of β, we find
16.2. DENSITY OPERATORS AND VON NEUMANN ENTROPY 389
∂ ∂β ∂ 1 1 ∂Z E
lnZ = lnZ = − 2 =
∂T ∂T ∂β kT Z ∂β kT 2
Therefore,
∂ F̃ E F̃ − E
= −klnZ − kT 2
= (16.18)
∂T kT T
What we have then is that both F̃ and F satisfy the same differential equation. We cannot say
that F = F̃ just quite however as we must also consider initial conditions. If F = F̃ for at least one
initial condition we may conclude that they are equivalent. Let’s consider the case when T = 0.
This means that F = E. Moreover, since temperature is zero, this energy must also be the lowest
possible energy, which we choose to denote as E0 . Moreover, at zero temperature Z(T = 0) = e−βE0
since all of the other Boltzmann factors showing up in the parition function (the excited states) are
infinitely suppressed. Therefore,
F̃ (0) = −kT lnZ(0) = E0 = F (0)

Since we have at least one initial condition where F̃ = F , we have that this statement holds for
all values of temperature T [52]. All in all,
F = −kT lnZ (16.19)

This concludes our brief review of thermal physics. Before moving on to the thermodynamics of
strings, let us first review a topic which will prove useful in a later chapter.
16.2 Density Operators and von Neumann Entropy

There is an alternative method of defining the entropy of a system which makes use of the density
operator from ordinary quantum mechanics. Before moving to this alternative definition, let us
briefly examine the basics of the density operator.
When one first learns quantum mechanics, they typically only learn how to take measurements
of a single quantum state. Such a state |ψi may be expanded in terms of basis kets, taking the form
|ψi = c1 |1i + c2 |2i + ... + cn |ni (16.20)

Sometimes this state is referred to as a coherent superposition of the basis states |ii [39]. When
we study most realistic systems, rather than considering a single state, we consider an ensemble of
quantum states prepared in a statistical mixture. The members of this ensemble are denoted by
the states |ψ1 i, |ψ2 i, ...|ψn i, while the probability of finding each state is denoted by p1 , p2 , ...pn . To
describe this ensemble of states we use the density operator, given by
n
X
ρ= pi |ψi ihψi | (16.21)
i=1
In the case we work with a pure state, all of the pi = 0 except for one. That is, a pure state is
described by a density matrix in the following way
ρ = |ψihψ|
When we compare this to the summation expression, we notice that indeed all of the probabilities
pi equal zero, except for one which takes on the value of 1. A mixed state is simply a collection
of pure states, each with their own probability. Therefore, in general the density matrix describes
a mixed state. Moreover, given the above representation, it’s easy to see that for pure states, the
density matrix satisfies ρ2 = ρ. Another property of the density matrix is that it has a trace equal
to one:
X XX XX XX X
T r(ρ) = hψj |ρ|ψj i = hψj |pi |ψi ihψi |ψj i = pi hψj |ψi ihψi |ψj i = pi δji δij = pi
j j i j i j i i
=1
where we used the fact that the sum of all the probabilities pi equals one. As an exercise, the
reader will show the following additional properties of the density matrix: the density matrix is
Hermitian; T r(ρ2 ) < 1 for mixed states, and for an ensemble uniformly distributed over n states,
ρ = n1 I.
A typical calculation in quantum theory is to calculate the expectation value of an operator Ω,
given by h|Ω|i. In the case we are considering a statistical ensemble, we can compute the ensemble
average of an operator. The expectation value of Ω, using the density matrix can be shown to be
given by T r(Ωρ) [53]. Notice then
X XX X
T r(Ωρ) = hψj |Ωρ|ψj i = hψi |ψj ihψj |Ω|ψi ipi = hψi |Ω|ψi ipi ≡ hΩ̄i
j i j i
Therefore, the expectation value of the ensemble can be interpreted as the sum of the quantum
expectation values of each state constituting the ensemble. Moreover, using the density operator
still allows us to make the usual predictions of quantum theory. If we let a measurement result,
denoted by m, be represented by the projection operator Pm = |mihm|, the probability of obtaining
the result m is given by
T r(ρPm ) = T r(ρ|mihm|) = hm|ρ|mi

Let’s take a closer look at pure states. Consider pure quantum state given in (16.20). Written
using the density matrix we find
n
X n
X
ρ = |ψihψ| = |ci |2 |iihi| + ci c∗j |iihj| (16.22)
i=1 i6=j
Hence, the density operator of a pure state can be split up into two terms. The first term gives
us the probability to find a system in the state |ii. The second term exhibits quantum interference.
To see this, recall that the expansion coefficients are complex numbers, which take the polar form
representation, cj = |cj |eiφj . Therefore, in the second sum we find
hi|ρ|ji = ci c∗j = |ci ||cj |ei(φi −φj )

16.2. DENSITY OPERATORS AND VON NEUMANN ENTROPY 391
This phase difference expresses the coherence of terms in the state to interfere with each other.
Since the density operator may be represented by a matrix, we find that the off diagonal elements
of a density matrix represent the ability of a system to have quantum interference. Since we are
considering a pure state, it follows that the density matrix representing a pure state has off diagonal
elements. This is contrary to mixed states. As one can show, the density matrix representing a
mixed state does not have elements off the diagonal (the elements are zero), which indicates that a
mix state has no coherence. Related to the notion of coherence is decoherence, which we will cover
shortly.
A completely mixed state is a state where the probability for each state is equal to all others. An
example of a completely mixed state is represented by
1 1
ρ= |0ih0| + |1ih1|
2 2
To measure the purity of a state, we compute T r(ρ2 ), which is bounded by n1 ≤ T r(ρ2 ) ≤ 1
where n is the dimension of the Hilbert space. For a totally mixed state, T r(ρ2 ) = n1 , while for
a pure state T r(ρ2 ) = 1. Any state that has a trace within this bound is called a partially mixed
state.
For concreteness, consider a spin- 21 system. Suppose the pure state ρ1 = |+y ih+y | and the
completely mixed state
1 1
ρ2 = |+ih+| + |−ih−|
2 2
Writing the |+y i ket in the Sz basis, |+y i = √1 (|+i + i|−i), we find
2
1 1 1 i i 1
ρ1 = √ (|+i + i|−i) √ (h+| − ih−|) = |+ih+| − |+ih−| + |−ih+| + |−ih−|
2 2 2 2 2 2
Or, in their the matrix representations, the density operators ρ1 and ρ2 take the form
− 2i
1 1
2 2 0
ρ1 = i ρ2 =
2
1
2
0 21
From this representation it’s easy to see that T r(ρ21 ) = 1 and T r(ρ22 ) = 12 . Now consider a
measurement in Sz . The probability of measuring + ~2 is done usng the projection operator |+ih+|:
1
T r(ρ1 |+ih+|) = h+|ρ1 |+i =
2
Similarly, T r(ρ2 |+ih+|) = 12 . Had we made a measurement in Sy , we would have found the
probabilty of obtaining + ~2 to be T r(ρ1 |+y ih+y |) = 1, indicating that ρ1 is a pure state. On the
other hand, T r(ρ2 |+y ih+y |) = 12 , indicating that ρ2 is a mixed state.
As we noted before, the density matrix lends information about the entropy of a system. In
quantum information theory, one defines the von Neumann entropy by [63]
X
S = −kT r(ρlnρ) = − λi lnλi (16.23)
i
where λi are the eigenvalues of the given density matrix. Interestingly, with this definition of
entropy, it can be proven mathematically that ∆S > 0 for an adiabatically isolated system. More-
over, with this definition of entropy we can better understand the notion of quantum decoherence.
In a very rudimentary sense, decoherence is the loss of quantum coherence. The way this can be
characterized is through pure and mixed states. A pure state is one that has coherence, and also
has a vanishing von Neumann entropy (S = 0). A mixed state on the hand, which has the quantum
interference information erased has a non-zero von Neumann entropy. The maximum value of von
Neumann entropy is S = 1. For example, if we took the pure state |ψi = √12 (|+i + |−i), which has
the corresponding density matrix

1 1 1
ρ=
2 1 1
we would find that the entropy is equal to zero. This is contrary to the state given by

1 1 0
ρ=
2 0 1
which has an entropy of S = ln2 = .69. The crucial difference between these two states is that
one has coherence while the other has lost quantum interference information. Roughly speaking,
the entropy S measures the purity of a quantum state, or the amount of disorder a quantum state
has.
Quantum decoherence occurs when the system interacts with the surrounding environment in
a thermodynamically irreversible way. Put another way, decoherence can be viewed as the loss of
information from the system to the environment. Decoherence is imperative to quantum mechanics
as it seems to point to the location of the boundary between classical physics and quantum physics;
also giving an interpretation of the notion of the wavefunction collapse. We care about this formu-
lation as it is imperative in understanding the consequences Hawking radiation, and determining
the resolution given by string theorist Leonard Susskind, a topic which we will discuss in detail in
a later chapter.
16.3 Partitions and the Non-Relativistic String

Rather than jumping with both feet into the partition function of the bosonic string, we will build
up to this function in parts. First we will consider the non-relativistic quantum string, the topic
of this section. From here we will briefly examine the relativistic point particle. After that we will
finally have the necessary tools to confront the quantum relativistic string.
Consider a quantum mechanical non-relativistic string with fixed endpoints. We may idealize
this quantum string as a collection of simple quantum harmonic oscillators with an infinite set of
frequencies ω0 , 2ω0 , .... Each of these harmonic oscillators would have their own pair of creation
and annihilation operators, and Hamiltonian. In general, the Hamiltonian for a harmonic oscillator
in terms of the creation and annihilation operator takes the form
Ĥ = ~ω0 a†n an (16.24)

The quantum string we are considering is the union of each harmonic oscillators, we have the
Hamiltonian for the quantum string is simply
16.3. PARTITIONS AND THE NON-RELATIVISTIC STRING 393
∞
X
Ĥ = ~ω0 a†n an (16.25)
n=1
Or, writing the Hamiltonian in terms of the number operator N̂ ,

∞
X
Ĥ = ~ω0 N̂ N̂ = a†n an (16.26)
n=1
As usual, we would build an arbitrary quantum state by letting the collection of creation operators
act on the vacuum state:
|ψi = (a†1 )m1 (a†2 )m2 ...(a†n )mn |0i

From the above, it is easy to see that we may specify the state by the set of occupation numbers
{m1 , m2 , ...mn }. By specifying the state, we notice that the action of the number operator on the
above arbitrary state is simply
∞
X
N̂ |ψi = m1 + 2m2 + 3m3 + ...|ψi = nmn |ψi ≡ N |ψi (16.27)
n=1
We therefore the energy E of the state of the string is E = ~ω0 N . It would be nice to know
how many states are associated with N . Some results from elementary partition theory, a branch
of mathematics known as number theory will help us with our goal. Specifically, a partition of a
non-negative integer n is a representation composed from the sum of positive integers that are equal
to n. For example, the partitions of 5 are:
5, 4 + 1, 3 + 2, 3 + 1 + 1, 2 + 2 + 1, 2 + 1 + 1 + 1, 1 + 1 + 1 + 1 + 1
Often, we will write these partitions as a collection of sets, namely
{5}, {4, 1}, {3, 2}, {3, 1, 1}, {2, 2, 1}, {2, 1, 1, 1}, {1, 1, 1, 1, 1}
Moreover, we denote the number of partitions by p(n), from which we readily see that p(5) = 7.
In 1917, an exact formula was obtained by mathematicians G.H. Hardy and S. Ramanujan, in which
the first term is
 q 
2π 1
1 d  exp √
6
n − 24 
√ q (16.28)
2π 2 dn
 
1
n − 24
As mentioned above, we care about partitions as it will allow us to determine the number of
states of N , from the number of partitions p(N ). Since we will be looking at systems with large N ,
we will use the asymptotic expansion of Hardy’s and Ramanujan’s result:
r !
1 N
p(N ) ≈ √ exp 2π (16.29)
4N 3 6
Moreover, although we won’t go into the details here, there is a generating function for the
number of partitions p(n), given by
∞ ∞
X Y 1
p(n)q n =
n=0 j=1
1 − qj
Finding an explicit expression for p(N ) is rather difficult (afterall, it took two brilliant mathe-
maticians to derive it), however we will consider a formula for lnp(N ) that is accurate for high N .
E
For this, for a given energy E, we have that N = ~ω 0
, and the number of microstates Ω(E) is equal
to the number of partitions of N , p(N ). By Boltzmann’s famous expression, we therefore have

E
S(E) = klnp(N ) = kln p (16.30)
~ω0
From here we see that if we only knew S(E), we could calculate a function for p(N ). One
straightforward method in calculating S(E) is by finding the partition function Z, using (16.19) for
the Helmholtz free energy, and then (16.7) to yield the entropy S. Following the basic recipe, the
partition for the quantum string is just
XX X
X Eα ~ω0
Z= exp − = ... exp − (m1 + 2m2 + ...)
α
kT m m m
kT
1 2 n
X ∞ X ∞
X ~ω0 ~ω0 Y ~ω0
= exp − m1 exp − 2m2 ... = exp − nmn (16.31)
m1
kT m
kT n=1 m =0
kT
2 n
As an exercise, the reader will show that the sum over each mn is simply a geometric series,
allowing us to write the partition function as
∞
Y 1
Z= −~ω0 n
(16.32)
n=1
1 − exp kT
Then, if we use (16.19), we find that the Helmholtz energy for the quantum string is
∞
X ~ω0 n
F = −kT lnZ = kT ln 1 − exp − (16.33)
n=1
kT
The form that the Helmholtz energy is in now is rather useless, so let’s look in the high temper-
ature regime to see if we can get F in a more workable form. For high enough temperatures we
have that ~ω
kT 1. In this case, each successive term in our sum above differs slightly from the
0
previous term, which allows us to approximate the sum by the integral

Z ∞
~ω0 n
F ≈ kT dnln 1 − exp − (16.34)
1 kT
~ω0 n
Making the substitution x = kT ,F reduces to
(kT )2 ∞
Z
F ≈ dxln(1 − e−x ) (16.35)
~ω0 0
To evaluate this integral, recall the expansion

1 2 1 3
ln(1 − x) = − x + x + x + ... (16.36)
2 3
for 0 ≤ x < 1. Therefore, we may write F as
∞
(kT )2 (kT )2
Z
−x 1 −2x 1 −3x 1 1 1
F ≈− dx e + e + e + ... = − 1+ + + + ...
~ω0 0 2 3 ~ω0 4 9 16
Notice that the sum in the parentheses is the familiar Zeta function. Specifically,
1 1 1 π2
ζ(2) = 1 + + + + ... =
4 9 16 6
Altogether then, the Helmholtz free energy F in the higher temperature limit is approximated
by
(kT )2 π 2 1 π2 1
F ≈− =− (16.37)
~ω0 6 ~ω0 6 β 2
We are now in a position to calculate the entropy S using (16.7), from which we find
pi2

∂F kT
S=− =k (16.38)
∂T 3 ~ω0
Moreover, recall that we may express the energy E in the following way
2
π2 1 ∂ π 2 kT

∂lnZ ∂ 1
E=− = (βF ) = − = ~ω0 (16.39)
∂β ∂β 6 ~ω0 ∂β β 6 ~ω0
Hence, as a function of energy, the entropy is given by
r r
2 E N
S(E) = πk = 2πk (16.40)
3 ~ω0 6
From (16.30) we identify
r
N
ln(p(N )) ≈ 2π (16.41)
6
Equation (16.41) holds for higher temperatures, which is equivalent to saying that it holds for
large values of N . As it turns out, our result is the leading term of the asymptotic expansion of
Hardy’s and Ramanujan’s result in (16.29)
Now, we are still not done, as we may further generalize our system. Let us consider a string
which can vibrate in b directions. It follows then that for each frequency nω0 we have b harmonic
oscillators, each representing the possible polarization of motion [64]. In this case our partition
function Zb will be the sum over all states, however we must sum over all the possible values of the
(q)
occupation number mk , with k = 1, 2, ...∞ and q = 1, 2, ...b. Therefore, the new partition function
takes the form
∞
" b
#
X ~ω X X
Zb = exp − nm(q)
n (16.42)
(1) (b)
kT n=0 q=1
mk ,...mk
We can actually factorize the sums over the occupation numbers. Therefore,
! !
X ~ω0 X (1)
X ~ω X (b)
Zb = exp − nmn ... exp − nmn (16.43)
(1)
kT n=0 (b)
kT n=0
mk mk
Now we notice that each sum above is equal to the partition function Z that we calculated
above. We may conclude then, as the reader will show, that Zb = (Z)b . This allows us to write the
Helmholtz free energy Fb associated with Zb as
Fb = −kT ln(Zb ) = −kT ln(Z) = bF (16.44)

Similarly, Sb = bS and Eb = bE, allowing us to write
r r
1 E Nb
Sb = 2πkb = 2πk (16.45)
6 ~ω0 6
Moreover, the number of partitions of N associated with these transverse directions is denoted by
pb (N ). This means that we partition N into integers that may carry any of the b labels. Therefore,
using Boltzmann’s formula, we find
Sb = kln (pb (N )) (16.46)

Allowing us to conclude
r
Nb
ln (pb (N )) ≈ 2π (16.47)
6
If we use the expression derived by Hardy and Ramanujan, we find that the partitions pb (N )
take the form
b+1 r !
1 b 4
− b+3 Nb
pb (N ) ≈ √ N 4 exp 2π (16.48)
2 24 6
from which we immediately see that when b = 1 we simply have pN as given in (16.29). Unfortu-
nately, we are still not quite done. So far we have assumed that our harmonic oscillators commute.
We could also imagine the same system of simple harmonic oscillators, with frequencies ω0 , 2ω0 , ...,
however where the occupation numbers mn can only take on values of 0 and 1. Oscillators of this
type are said to be fermionic. This notion is related to the fact that fermionic creation and annihi-
lation operators anticommute, in which case we may only have a fermionic operator acting once on
a state. If the same operator acted again on the state, it would yield zero, suggesting occupation
numbers of 0 or 1. Thus, let us consider a non-relativistic string composed entirely out of fermionic
oscillators. We wish to determine the number of partitions q(N ). Since the occupation numbers
may only take on values of 1 or 0, we can see that the total number N of any state is effectively
split into unequal parts. Therefore, more refined, q(N ) denotes the number of partitions of N in
unequal parts.
Just as before, we will come up with an approximation for the ln(q(N ). The most direct route is
to use the Helmholtz free energy in finding the entropy, from which we may extract ln(q(N )) using
the Boltzmann formula for entropy. To find F , let us begin with the partition function
Y ∞ X ∞
X Eα ~ω0 nmn
Z= exp − = exp − (16.49)
α
kT n=1 m =0
kT
n
Remember we are dealing with fermionic oscillators, and therefore each occupation number mn
may on take on the values 0 or 1, leaving us with
∞
Y ~ω0 nmn
Z= 1 + exp − (16.50)
n=1
kT
It follows then that the Helmholtz free energy F is simply
∞
X ~ω0 n
F = −kT lnZ = −kT ln 1 + exp − (16.51)
n=1
kT
As before, we choose to take the high temperature limit, allowing us to approximate F as an
integral. We leave the details as an exercise for the reader. Upon integration we find that the
Helmholtz energy is approximated by
(kT )2

1 1 1
F ≈− 1− + − + ... (16.52)
~ω0 4 9 16
Notice that the sequence given in the parentheses can be written as
π2

1 1 1 1 1 1 1 1 1 ζ(2)
1− + − + ... = 1 + + + + ... − 2 + + + ... = ζ(2) − =
4 9 16 4 9 16 4 16 36 2 12
Therefore, in the high temperature limit, the Helmholtz free energy is approximated by
π 2 (kT )2
F ≈− (16.53)
12~ω0
From here it’s a simple matter to compute the entropy of our string in the high energy limit:
∂F k2 π2 T
S=− = (16.54)
∂T 6~ω0
Moreover, we can also compute the energy E as
2
π2

∂lnZ ∂ kT
= (βF ) = ~ω0 (16.55)
∂β ∂β 12 ~ω
which tells us that
r r
12E kπ 2 1 12E
kT = 2
~ω0 ⇒ S = ~ω0
π 6 ~ω0 π2
And since E = ~ω0 N , we find that
r r
1 E N
S(E) = πk = 2πk (16.56)
3 ~ω0 12
Comparing this expression to S(E) = kln(q(N )), we conclude that
r
N
ln(q(N )) = 2π (16.57)
12
Just as before, when we computed p(N ) when it carried b labels, we may also compute qf (N ),
with f labels by letting N → N f in (16.57). Lastly, we may also consider partitions of both
ordinary numbers (as indicated by the b labels) and fermionic numbers (as indicated by the f
labels). We denote these partitions as p(N ; b, f ). This statement gives the number of partitions of
N into ordinary and fermionic numbers, with the first having b possible labels and the latter having
f possible labels. For example, p(2; 1, 2) is the collection of sets
{2}, {21 }, {22 }, {1, 1}, {1, 11 }, {1, 12 }, {11 , 12 }

Therefore, p(2; 1, 2) = 7. It can be shown that the large N behavior of p(N ; b, f ) is given by [64]
s
N f
lnp(N ; b, f ) ≈ 2π b+ (16.58)
6 2
The above expression is rather useful when looking at the thermodynamics of superstrings, as
those strings are built out of both bosonic and fermionic oscillators. Moreover, this type of state
counting is useful in counting the microstates of a supersymmetric black hole, something we will
examine in a later chapter.
16.4 The Hagedorn Temperature

Before moving on to partition function of the relativistic point particle, let us introduce another
topic which is important for the thermodynamics for strings, and will be important later on. Let
us consider bosonic open strings (in the light-cone gauge) that carry no momentum, which is true
only when the string end points end on D0 branes. Recall that the mass-squared of a given state
of an open string is
1
M2 = (N ⊥ − 1) (16.59)
α0
⊥
In the large N ⊥ limit, this simply reduces to M 2 = Nα0 . A string which does not carry any
spatial momentum has energy levels that are given by the rest masses of its quantum states, similar
to particles
√ which √ carry no spatial momentum. Hence, using c = 1, we have that E = M , which
means N ⊥ = α0 E.
Since we are considering bosonic strings in the light-cone gauge, there are 24 transverse light-
cone directions and therefore we are dealing with 24 oscillators. This means that the number of
microstates Ω(E) = p24 (N ⊥ ), and therefore
S(E) = kln p24 (N ⊥ )

(16.60)
16.5. PARTITION FUNCTION OF THE RELATIVISTIC POINT PARTICLE 399
For large N ⊥ , or, equivalently, high energy, this becomes
√
r
24N ⊥ √
S(E) = 2πk = 4πk N ⊥ = 4πk α0 E (16.61)
6
The entropy is proportional to the energy E. Using this, notice what happens to the temperature,
1 1 ∂S √
= = 4π α0
kT k ∂E
In the high energy limit, the temperature of our bosonic string is constant. We call this temper-
ature the Hagedorn temperature, denoted by TH :
1 1
= kTH = √ (16.62)
βH 4π α0
This is an intriguing result as, using the high energy approximation, when we increase the energy
of the string, the temperature of the string stays constant, at the Hagedorn temperature. Plenty
of modern research aims toward determining whether the Hagedorn is the maximum temperature,
or is instead a phase transition. Some researchers studying four-dimensional superstring theory,
have found that the Hagedorn temperature is simply a phase transition of a superstring, and seems
to lend insight toward the cold to hot transition of the primordial universe. It turns out that, as
the reader will show, the energy associated with the Hagedorn temperature, kTH is relatively small
small compared to the rest energy of the string. As an exercise, the reader will also show that the
Hagedorn temperature also arises for closed strings.
Had we considered the relativistic fermionic string, we would have had Ω(E) = q24 (N ⊥ ). There-
fore, using (16.56) we have
r
24N ⊥ √
S(E) = 2πk = 2πk 2α0 E
12
Yielding the Hagedorn temperature to be
1
kTH = √ (16.63)
2π 2α0
which is simply √1 times the Hagedorn temperature of the bosonic string.
2
16.5 Partition Function of the Relativistic Point Particle

It behooves us to first review the process of determining the partition function of the relativistic
point particle before we consider the slightly more complicated case of the bosonic string. In this
set up we consider a relativistic particle of mass m which is free to move in D = d + 1 dimensional
space-time. For simplicity, we will assume that the particle is confined to a box of volume V
where v = L1 L2 ...Ld , the lengths of each side of the box. Moreover, let us assume that the box
is in thermal contact with some resevoir held at a temperature T . Since we are dealing with a
relativistic particle, it has an energy obeying the Einstein relation
p
E = p~2 + m2 (16.64)
It follows then that the parition function of the particle, when quantized, is simply
X
Z(m2 ) = exp(−βE(~
p)) (16.65)
p
~
We sum over the momentum p~ because each quantum state is denoted by the particle’s quantized
momentum. The wavefunctions describing these particles, carrying a momentum p~ = ~~k, have a
spatial dependence given by exp(i~k · ~x). Since we have assumed our particle is confined within the
walls of our box, the spatial part of the wavefunctions will obey periodic boundary conditions, such
that for each spatial index we have ki Li = 2πni where i = 1, ...d and ni ∈ Z. Using this boundary
conditions give that spatial part of the wavefunction is equivalent at each end of the box. Using
this fact we easily see that we may write
Li
ni = pi
2π~
Therefore, as the momenta are proportional to ni , summing the over all of the various momenta
is equivalent to summing over the ni . Moreover, if we consider a box large enough such that the
momenta of the particle changes slightly when the ni shift by a single unit, we may approximate
the sum in our partition function with an integral. In particular,
dd p~
X Z Z p
Z(m2 ) = p)) ≈
exp(−βE(~ dn1 dn2 ...dnd exp(~
p(n)) = V d
exp(−β p~2 + m2 ) (16.66)
(2π~)
p
~
Generalizing the momentum four vector to D spatial dimensions, we have that p~ = m~u where
~u is the velocity vector in D spatial dimensions, our expression for the partition function of the
relativistic point particle is approximately given by
dd ~u
Z p
Z(m2 ) = V md d
exp(−βm 1 + ~u2 ) (16.67)
(2π)
where we have also taken the liberty to set ~ = 1. As is, (16.67) is not a very helpful expression.
We can make it more accessible if we use modified Bessel functions however. To do this, it is useful
to know how to determine the surface area of a sphere in d spatial dimensions. This can be done
by first dividing up the surface of a sphere living in 3-space into a collection of loops with a width
rdθ and a circumference of A2 (rsinθ), where A2 is the circumference of the unit sphere in 2-space
(the circumference of the unit circle). The surface area of the sphere in three spatial dimensions is
then the integral of all of these loops. Continuing this process, one can show by induction that the
surface area of a d dimensional hypersphere with radius r is given by [52]
d
2π 2 rd−1
Ad (r) = (16.68)
Γ( d2 )
This result will help us write (16.67) using modified Bessel functions. We choose to use the
modified Bessel functions in their integral representation [64]
√ 1 vZ ∞
π( 2 z)
Kv (z) = e−zcosht sinh2v tdt (16.69)
Γ(v + 21 ) 0
To utilize (16.68), first notice that we may write the measure of integration as
16.5. PARTITION FUNCTION OF THE RELATIVISTIC POINT PARTICLE 401
dd u = ud−1 dΩdu
where dΩ is the solid angle generalized for d dimensions. To convince oneself, one can check the
cases for d = 2 and d = 3, which yield the expected results. Substituting this into (16.63) gives us
m d Z p
Z(m2 ) = V ud−1 exp(−βm 1 + u2 )dΩdu (16.70)
2π
Integrating dΩ first gives us
Z d
2π 2
dΩ =
Γ( d2 )
Yielding
m d 2π d2 Z p
2 d−1
Z(m ) = V u exp(−βm 1 + u2 )du
2π Γ( d2 )
Our goal is to have this expression match (16.69).
√ We see that this can be achieved if we make
the substitutions u = sinht; du = coshtdt and 1 + u2 = cosht. Putting all of this together brings
our partition function to the form
m d 2π d2 Z
Z(m2 ) = V sinhd−1 (t)cosh(t)exp(−βmcosht)dt (16.71)
2π Γ( d2 )
We notice that the above has a cosht in the integrand, which is no where to be found in the
modified Bessel function expression given in (16.69). Notice however that
√ 1 vZ ∞ √
π( 2 z) π v z v−1 ∞ −zcosht
Z
∂ −zcosht 2v
Kv (z) = −cosh(t)e sinh (t)dt+ e sinh2v (t)dt
∂z Γ(v + 12 ) 0 Γ(v + 12 ) 2 2 0
Going back to expression (16.71), let us write βm = z and d − 1 = 2v, which gives us
2v+1 1
2π v+ 2
Z
z
Z(m2 ) = V sinh2v (t)cosh(t)exp(−zcosht)dt
2πβ Γ(v + 12 )
v v 1
2π v+ 2
Z
z z z
=V sinh2v (t)cosh(t)exp(−zcosht)dt
2πβ 2πβ 2πβ Γ(v + 12 )
v v 1 Z
z z 1 z v 2π 2
=V 2π v sinh2v (t)cosh(t)exp(−zcosht)dt
2πβ 2πβ πβ 2 Γ(v + 21 )
v v
z z v 1 v 2 0
=V 2π Kv − Kv
2πβ 2πβ πβ 2 z
Substituting back in βm = z and d − 1 = 2v and doing a small bit of algebra yields
r d2
2mβ m d−1
2
Z(m ) = V K d−1 (βm) − K 0d−1 (βm) (16.72)
π 2πβ 2βm 2 2
where K 0 stands for the derivative of K. This expression of the partition function, though
containing modified Bessel functions, is slightly more useful as one just has to refer to their favorite
manual on special functions to determine further explicit forms of the partition function.
Let us go one step further and consider the case where βm 1, the low temperature limit. We
may then use the asymptotic expansion of the modified Bessel function [64]
4v 2 − 1
r
π
Kv (z) ≈ e−z 1+ + ... (16.73)
2z 8z
from which we also gain
4v 2 + 3
r
π
Kv0 (z) ≈e −z
1+ + ... (16.74)
2z 8z
Using these asymptotic expansions, keeping only the first two terms we may approximate our
partition function in the low temperature limit. Noting that the first term in our expression will
vanish in this limit, we find that
r d2
2mβ m
2
Z(m ) ≈ V −K 0d−1 (βm)
π 2πβ 2
d2 d−1 2
r !!
4 +3
r
2mβ m −βm π 2
≈V −e 1+
π 2πβ 2βm 8βm
d2
m
≈V e−βm (16.75)
2πβ
where we decided to only keep the first term in the asymptotic expansion of K 0 . Therefore, the
partition function of the relativistic point particle in the low temperature limit is given by (16.75).
This function is actually pretty close to the low temperature limit of the partition function for a
non-relativistic particle, which does not have the factor e−βm , as this factor obviously accounts for
the contribution of the rest energy of the relativistic particle.
16.6 Partition Function of a Bosonic String

We are now in a position to develop the partition function of the bosonic string, which will ultimately
allow us to determine the energy associated with the string in the limit where the temperature is
close to the Hagedorn temperature. To do this, we will consider a system similar to the one
considered in the last section. Rather than a particle enclosed by a box with volume V , we will
work with a single open string in a box of volume V . We will need the quantum states of this
string, which we recall are given by, in the light-cone gauge,
∞ Y
Y 25
|λ, pi = (aI†
n )
λn,I +
|p , p~T i (16.76)
n=1 I=2
16.6. PARTITION FUNCTION OF A BOSONIC STRING 403
where λn,I are the usual occupation numbers. Moreover, recall the mass-squared operator for
the bosonic string:
M 2 = −p2 = 2p+ p− − pI pI (16.77)

and
1 X
M2 = (N ⊥ − 1) N⊥ = nλn,I (16.78)
α0
n,I
It follows then the energy, which still obeys the Einstein energy relation, is given by
p
E = M 2 + p~2 (16.79)
As usual, to find the partition function of the bosonic string we must sum over all of the quantum
states |λ, pi. This equivalent to summing over all of the spatial momenta p~, and occupation numbers.
Hence,
X XX h p i
Z= exp(−βEα ) = exp −β M 2 + p~2 (16.80)
α λn,I p
~
Comparing this sum to the partition function we had for the relativistic point particle, we see
that the sum over momenta above simply yields the partition function for a relativistic point particle
with a mass squared M 2 . Therefore, the string partition function simply becomes
X
Z= Z(M 2 ) (16.81)
λn,I
Since we are summing over the occupation numbers, we may also write the above sum as a
sum over the number operator N ⊥ . Remember however that there is also degeneracy of states to
consider since there are p24 (N ⊥ ) states with the eigenvalue N ⊥ . Therefore, the above sum may
instead be rewritten as
∞
X
Z= p24 (N ⊥ )Z(M 2 (N ⊥ ) (16.82)
N⊥
We would like to make use of our large N approximation of p24 (N ⊥ ). To do this, we follow
a method as proposed by Zwiebach. Let√N0⊥ denote an integer for which we may approximate
27
p24 (N ⊥ ) by p24 (N ⊥ ≈ √12 (N ⊥ )− 4 exp(4π N ⊥ ), which follows from (16.48). Then, we may break
up the partition function of the single string into two sums as follows:
0 −1
NX ∞
X
Z= p24 (N ⊥ )Z(M 2 (N ⊥ )) + p24 (N ⊥ )Z(M 2 (N ⊥ )) (16.83)
N ⊥ =0 N ⊥ =N 0
⊥
It is actually rather difficult to calculate the sum for N < N0 since we cannot use our approx-
imate expression for p24 (N ⊥ ). Therefore, to calculate the string partition function, this is as far
as we can get unless we make an approximation such that the first sum above can be neglected.
This was our reason for choosing the integer N0 in the first place. Aside from mathematical con-
venience, there is a physical reason why we desire to ignore the first sum. For the bosonic string,
the first sum will include the the tachyonic state, in which M 2 < 0, giving imaginary contributions
to Z. By making an approximation so that we may neglect the first sum, we are also ridding our
thermodynamic theory of tachyons, a physically reasonable assumption.
Let’s take the case where we may approximate the second sum by an integral, yielding partition
function as
0 −1
NX Z ∞
Z≈ p24 (N ⊥ )Z(M 2 (N ⊥ )) + dN ⊥ p24 (N ⊥ )Z(M 2 (N ⊥ )) (16.84)
N0
N ⊥ =0
To compute this integral, it will be useful to define the density of states as a function of M ,
ρ(M ). That is, we will make the subsitution, p24 (N ⊥ )dN ⊥ = ρ(M )dM . Since we are taking the
large N ⊥ limit for the integral, we have that α0 M 2 ≈ N ⊥ , and therefore,
√ √
dN ⊥ = 2α0 M dM = 2( α0 M )d( α0 M ) (16.85)
Moreover, since we are going to look at the single string partition function in the regime where
temperatures are near the Hagedorn temperature, we find that
1 √ 27
p24 (N ⊥ ) ≈ √ ( α0 M )− 4 exp(βH M ) (16.86)
2
Putting everything together we find that the density of states function is given by
√ √ 25 √
ρ(M )dM = 2( α0 M )− 2 exp(βH M )d( α0 M ) (16.87)
Therefore,
0 −1
NX
⊥
√ Z
⊥
∞ √ 25 √
Z≈ 2
p24 (N )Z(M (N )) + 2 ( α0 M )− 2 exp(βH M )Z(M )2 d( α0 M ) (16.88)
M0
N ⊥ =0
where we have defined α0 M0 = N0 . Moreover, using the fact that
M √ √ TH
= 2( α0 M )kT kTH βM = 4π α0 M
2πβ T
allows us to approximate the relativistic point particle partition function Z(M 2 ) by
25 √ √

2 25
0
25
0
TH
Z(M ) ≈ 2 V (kT kTH ) ( α M ) exp −4π α M
2 2 2 (16.89)
T
The partition function for the bosonic string then becomes
0 −1
NX ∞ √ √
Z
25 TH
Z≈ p24 (N ⊥ )Z(M 2 (N ⊥ )) + 213 V (kT kTH ) 2 √
d( α 0 M )exp −4π α0 M − 1
M0 T
N ⊥ =0
√ (16.90)
To compute this integral, it is useful to make the subsititution x = α0 M , from which we obtain
16.7. EXERCISES 405
0 −1
NX Z ∞
⊥ 2 ⊥ 13 25 TH
Z≈ p24 (N )Z(M (N )) + 2 V (kT kTH ) 2
√
dxexp −4πx −1 (16.91)
N0 T
N ⊥ =0
We immediately notice that when T > TH , the exponential factor in the integral becomes positive
cause the integral to blow up. Therefore, we consider the convergent case when T < TH . Computing
this integral leaves us with
0 −1
NX
211
p TH
25 T
Z≈ p24 (N ⊥ )Z(M 2 (N ⊥ )) + V (kT kTH ) 2 exp −4π N0 −1
π TH − 1 T
N ⊥ =0
Now notice that as the temperature T approaches the Hagedorn temperature from below, the
exponential goes to one while the factor multiplying the exponential blows up. This term is much
larger than the sum in our above expression and therefore the sum, which is difficult to calculate,
can be neglected, leaving us with
211

T
Z≈ V (kTH )25 (16.92)
π TH − 1
which is the partition function for a single bosonic string in the limit where T → TH . Using
this approximation of the partition function, it is readily easy to determine the average energy E
of the string near the Hagedorn temperature. As an exercise, the reader will show that the average
energy is given by [64]

∂lnZ 1 TH
E=− ≈ ≈ kTH (16.93)
∂β β − βH TH − T
from which we see that as the temperature T approaches the Hagedorn temperature, the average
energy blows up. This seems to tell us that as the bosonic string reaches the Hagedorn temperature,
it will carry an infinite average energy.
The next step would be to determine the partition function of a single superstring and compare
the results. We won’t do this here as it is a problem that goes beyond the scope of this text. It
should be mentioned however that string thermodynamics is not well understood, as the results from
the microcanonical ensemble don’t exactly match the results of the canonical ensemble. In order
to have a working theory of string thermodynamics, both of these ensembles should yield the same
results. Nonetheless, the thermodynamics of strings is imperative for our physical understanding
and is therefore an active area of research. The thermodynamics of strings might also lead to
interest insights in the thermodynamics of black holes. More recently, researchers have begun to
examine the thermodynamics of space-like D-branes, s-Branes, which has a time dimension as one
of its transverse directions. These s-branes have connections to unstable D-branes, and time-like
holography. For more on this subject, the reader is urged to review the references.
16.7 Exercises
1. Prove the following properties of the density operator: (a) ρ = ρ† (the density operator is
Hermitian) (b) T r(ρ2 ) < 1 for partially mixed states.
2. Show the partition function of the non-relativistic string takes the form given in (16.32).
3. Our goal here is to prove the area of a d dimensional hypersphere as given in (16.68). This
will be a proof by induction. (a) Given that A1 (r) = 2, A2 (r) = 2πr, and A3 (r) = 4πr2 , prove that
(d−1) Z π
2π 2
Ad (r) = rd−1 (sinθ)d−2 dθ
Γ d−1
2 0
(b) Prove the following by induction:

√
π
πΓ n2 + 12
Z
n
(sinθ) dθ =
Γ n2 + 1

0
(Hint: First check that the above holds for n = 0 and n = 1. Next show that
Z π Z π
n−1
(sinθ)n dθ = (sinθ)n−2 dθ
0 n 0
Combining both of these results should yield the proof by induction one seeks.)
(c) From parts (a) and (b), prove (16.68).
4. Starting from (16.67), show that by taking the low temperature limit, 1 βm, one arrives to
the result given in (16.75). (Hint: Try looking at the one-dimensional case first. Also, one should
arrive to a Gaussian function which can be easily integrated to yield the sought out result.)
5. Show that the average energy of a single bosonic string is given by (16.93).
Chapter 17
Elements of Differential Geometry
To study general relativity, one is required to have a fair understanding of tensor calculus, exterior
differentiation, and differential forms, all of which are aspects of differential geometry. In this chap-
ter we aim to provide the mathematical background necessary for understanding the mathematical
model of space-time. As with the rest of this text, we will aim to be more pragmatic in our study
and therefore leave out the rigorous details of differential geometry (by rigorous, we mean the finer
points which require the mathematician’s flavor of analysis). This is not to say that our task will
be any easier; it will be particularly computationally intensive, the cost of giving a detailed intro-
duction to the subject. For more thorough discussions of the mathematics behind general relativity,
the reader is urged to examine the references. Lastly, since this material builds heavily from the
discussions in chapter two, the reader should review chapter two before reading the current chapter.
17.1 Manifolds
We are already familiar with manifolds, as we had discussed them briefly back when we examined
the notion of compactification. As one might have guessed by this point, in general relativity
space-time is modeled by a differentiable manifold. This fact actually comes from the equivalence
principle and that gravity is a manifestation of the curvature of space-time, a detail which we will
explore further in the next chapter. As we have previously noted, a differentiable manifold is a space
that locally looks like flat space, even though globally it might have a drastically different geometry
than that of flat Euclidean space. Again, the typical example is that of the surface of a sphere.
For relativity this means we pick a model of space-time that locally looks like Minkowski space, the
space-time model of special relativity. In terms of the equivalence principle, this translates to the
laws of physics on small regions of space-time match the laws given by special relativity. There is,
of course, a rigorous mathematical definition of a manifold, however it won’t be necessary for the
goals of our survey.
Now that we have our general model of space-time, we can introduce structure on our manifold
by introducing different objects. We have already seen many of these objects, namely vectors and
tensors. Starting with vectors, we had defined the tangent space Tp , as the set of all vectors at
a single point p in our space-time. A more refined definition of a tangent space is that it can be
identified with the space of directional derivatives along curves passing through the point p [11]. To
do this we must first establish that the space of directional derivatives indeed form a vector space,
407
408 CHAPTER 17. ELEMENTS OF DIFFERENTIAL GEOMETRY
and, more subtle, it ends up being the vector space we desire. As an exercise, the reader will prove
first of these requirements. The second bit is slightly trickier, however one can show that the set of
partials {∂µ } represent a basis for the vector space of directional derivatives; ultimately allowing us
to identify it with the tangent space Tp . This basis is called the coordinate basis, taking the form
we saw back in chapter two in equation (2.28). From here our analysis of the objects on manifolds,
and therefore manifold structure, follows exactly to the one given in chapter two, allowing us to
define vectors and tensors in the way we did before.
Before moving on, let us briefly consider vector fields. We have developed a notion of vectors at
a particular point in space-time, one that can be viewed as a directional derivative along a path
through that specific point. It follows rather naturally then that a vector field defines a type of
map, one which takes a derivative at each point on the manifold, taking smooth functions to smooth
functions. Therefore, we define a vector field as an operator. In the coordinate basis, we may define
the vector field X as [14]
X ≡ X a ∂a (17.1)
where the Einstein summation convention is in effect. From this representation we notice that
the vector field indeed acts as an operator, namely, for some arbitrary real-valued function f , we
have
X(f (xa )) = X a ∂a (f (xa )) = X a (∂a f )

This definition of a vector field is independent under coordinate transformations. To see this,
consider some other coordinate system, x0a . Then, using our expressions for coordinate transfor-
mations as given in (2.44) and (2.45), we find
∂ ∂x0a b ∂xc ∂
X 0a ∂a0 = X 0a 0a
= X
∂x ∂xb ∂x0a ∂xc
where in the last step we used the chain rule. Rearraning the above yields
∂xc ∂x0a b ∂ ∂ ∂
0a b
X c
= δ c b X b c = X b b = X a ∂a
∂x ∂x ∂x ∂x ∂x
where we used the fact that the b was a dummy index as it was being summed over. Altogether,
we have that the coordinate basis representation of the vector field is independent, or irrespective
of the coordinate system chosen. We can also construct another vector field from two vector fields
X and Y called the commutator or Lie bracket of X, Y :
[X, Y ] = XY − Y X (17.2)
To observe that this is indeed another vector field, consider acting the commutator on an arbitrary
function f . Then,
[X, Y ]f = (XY − Y X)f = X(Y (f )) − Y (X(f )) = X b ∂b (Y a ∂a f ) − Y b ∂b (X a ∂a f )
= X b (Y a ∂b ∂a f + ∂b Y a (∂a f )) − Y b (X a ∂b ∂a f + ∂b X a (∂a f )) = X b ∂b Y a (∂a f ) − Y b ∂b X a (∂a f )

17.2. THE COVARIANT DERIVATIVE 409
where the two other terms cancel by relabeling dummy indices and using the fact that partials
derivatives commute. Since f is arbitrary, we are left with
[X, Y ]a = X b ∂b Y a − Y b ∂b X a ≡ Z a (17.3)
Moreover, from the definition of the commutator, as the reader will prove, it follows that
[X, X] = 0 (17.4)
[X, Y ] = −[Y, X] (17.5)
[X, [Y, Z]] + [Z, [X, Y ]] + [Y, [Z, X]] = 0 (17.6)

But we have already seen these properties before. Operators satisfying these properties form
a Lie Algebra; we considered this definition of a Lie Algebra when we looked at the transverse
Virasoro operators for the first time in chapter nine. It is from this definition that we can begin to
understand why a Lie group is at the same time a manifold.
17.2 The Covariant Derivative

Now that we have objects on a manifold, we would like to be able to perform calculus with these
objects, particularly differentiation, as this will yield the curvature of our space. A problem arises
however: when we compute the ordinary partial derivative of a tensor, we don’t get a tensor back
in general. Partial differentiation of tensors does not yield tensorial objects. To see this, consider
the contravariant vector V 0a and let us differentiate this vector with respect to x0c . We see that,
0a 0a
0 0a ∂ ∂x b ∂xd ∂ ∂x
∂c V = 0c b
V = 0c d
Vb
∂x ∂x ∂x ∂x ∂xb
∂x0a ∂xd b ∂ 2 x0a ∂xd b

= ∂d V + V
∂xb ∂x0c ∂xb ∂xd ∂x0c
From our transformation law of tensors, we can easily show that a (1,1) tensor transforms as
0
a ∂x0a ∂xd b
T 0c = T
∂xb ∂x0c d
Therefore we see that the first term above transforms like a (1,1) tensor, however the presence
of the second term makes it so the resulting object is not tensorial since as a whole it does not
transform as a tensor. All in all, a partial derivative of a tensor does not yield a tensor in general,
a result which we strongly desire. To fix this problem, we end up introducing an auxiliary field
onto the manifold, similar to the auxiliary field we introduced in the supersymmetric Lagrangian to
make it so the SUSY algebra would close. In effect, we will come up with a new type of derivative,
the covariant derivative which will allow us to differentiate tensors and get back tensors.
For ease, let us consider the case of taking the derivative of a vector. In cartesian coordinates,
taking a derivative of a vector is not a problem because the basis vectors x̂, ŷ, ẑ are constant. This
is not the case in general. When dealing with curved space, and hence curved space-time, the basis
vectors are likely to change from point to point. In general the derivative of an ordinary vector V
is
∂V ∂ b ∂V b ∂eb
= (V e b ) = eb + V b a (17.7)
∂xa ∂xa ∂xa ∂x
If we were in cartesian coordinates, the second term would vanish, yielding the common under-
standing of the derivative of a vector. However this is not the case in general for curved space since
often the basis vectors in curved space are not constant. As an example, the reader will consider
spherical coordinates and show that
∂er ∂er 1 ∂er 1
=0 = eθ = eφ (17.8)
∂r ∂θ r ∂φ r
It turns out that when differentiating basis vectors there is a general relationship that gives
another basis vector in terms of a weighted sum with coefficients Γc ab [38]. Namely,
∂ea
= Γc ab ec (17.9)
∂xb
For the spherical coordinate case, we readily find that
∂er 1 1 ∂er 1 1
= eθ ⇒ Γθ rθ = = eφ ⇒ Γφrφ =
∂θ r r ∂φ r r
Based on this example, we can see that the coefficients Γabc are functions of the coordinates in a
given coordinate system. These coefficient functions are known as Christoffel symbols or sometimes
an affine connection. The Christoffel symbols act as the correction terms we seek in the case of
differentiating tensors. To see this explicitly, again consider the case of a vector V . It follows then
that (17.7) may be rewritten using (17.9):
∂V ∂V b
a
= eb + Γc ba V b ec (17.10)
∂x ∂xa
Exchanging dummy indices by letting Γc ba V b ec → Γb ca V c eb , we find
∂V b

∂V b c
= + Γ ca V eb ≡ ∇a V b eb (17.11)
∂xa ∂xa
where we have defined the covariant derivative to be
∂V b
∇a V b = + Γb ca V c (17.12)
∂xa
Let’s check that this object transforms as a (1,1) tensor. To do this we must first determine how
the Christoffel symbols transform, which can readily figure out from (17.9)
0 ∂ea0
Γc a0 b0 ec0 = (17.13)
∂xb0
Moreover, recall that basis vectors transform as
∂xd
ec0 = ed
∂xc0
Hence, we may write the left hand side of (17.13) as
0 0 ∂xd
Γc a0 b0 ec0 = Γc a0 b0 ed
∂xc0
The right hand side of (17.13) takes a little bit more work:
m
∂xn ∂
m
∂xn
2 m
∂xm ∂em

∂ea0 ∂ ∂x ∂x ∂ x
= em = em = em +
∂xb0 ∂xb0 ∂xa0 ∂xb0 ∂xn ∂xa0 ∂xb0 ∂xn ∂xa0 ∂xa0 ∂xn
∂em
But ∂xn is just another Christoffel symbol. Therefore the above becomes
∂xn
2 m
∂xm `

∂ea0 ∂ x
= em + Γ e`
∂xb0 ∂xb0 ∂xn ∂xa0 ∂xa0 mn
By letting m → d in the first term and ` → d in the second term, we have
∂xn ∂ 2 xd ∂xm d

∂ea
= + Γ ed
∂xb0
∂xb0 ∂xn ∂xa0 ∂xa0 mn
Setting both the right hand side and left hand side of (17.13) equal to each, we have
∂xd ∂xn ∂ 2 xd ∂xn ∂xm d

0
Γc a0 b0 ed = + Γ ed
∂xc0 ∂xb0 ∂xn ∂xa0 ∂xb0 ∂xa0 mn
Rearranging we find that the Christoffel symbols transform as
0 0
0 ∂xc ∂xn ∂ 2 xd ∂xc ∂xn ∂xm d
Γc a0 b0 = + Γ
∂xd ∂xb0 ∂xn ∂xa0 ∂xd ∂xb0 ∂xa0 mn
0 0
∂xc ∂xn ∂xm d ∂xc ∂ 2 xd
= 0 0 Γ mn + (17.14)
d b
∂x ∂x ∂x a ∂xd ∂xb0 ∂xa0
Seeing how the Christoffel symbol transforms we find two things: first, it indeed acts as a
correction term, canceling out the unwanted term that prevents a differentiated tensor from being
tensorial, and second, we find that Christoffel symbols don’t actually transform as tensors and are
therefore not tensors themselves. The latter of these two properties is a rather obvious one since
the term we sought to correct was a non-tensorial object, and would therefore take a non-tensorial
object to cancel it.
Equation (17.12) is the expression of a covariant derivative of a contravariant vector, but in

general relativity we don’t just work with contravariant vectors. There is in fact a more general
definition of a covariant derivative or an arbitrary tensor given by [14]
∇c T a... a... a d... d b...

b... = ∂c T b... + Γ dc T b... + ... − Γ bc T d... − ... (17.15)
Instead of working with this general expression, we provide below some of the more common
covariant derivative expressions one works with, each of which is relatively simple to prove
∇c Va = ∂c Va − Γb ac Xb (17.16)
∇c T ab = ∂c T ab + Γacd T db − Γd bc T ad (17.17)
∇c Tab = ∂c Tab − Γd ac Tdb − Γd cb Tad (17.18)
∇c T ab = ∂c T ab + Γacd T db + Γb dc T ad (17.19)
Looking at these few identities, (17.16)-(17.19), a pattern emerges. For every upper index in the
tensor that we differentiate, we obtain a positive contribution from a Christoffel symbol, where the
upper indices between the Christoffel symbol and the tensor are cyclic, and the lower indices of the
Christoffel symbol switch. Moreover, for every lower index in a tensor we differentiate, we obtain
a negative contribution from a Christoffel symbol. Seeing these expressions also gives a reason for
the name of the covariant derivative: it takes the derivative of a tensor of type (p, q) and becomes
a tensor of type (p, q + 1), adding one extra covariant rank.
Finally, one can show that the antisymmetric part of a Christoffel symbol Γabc may be written
as
T abc = Γabc − Γacb (17.20)

The tensor T abc is called the torsion tensor, which can viewed as measuring the twist of a frame
around a curve. Typically when one studies relativity it is assumed the torsion tensor vanishes,
from which we have that the Christoffel symbols are symmetric under the exchange of their lower
indices, Γabc = Γacb . Though we won’t persue it in this text, there are relativity theories which do
not make this assumption called Einstein-Cartan theories. These models allow the coupling of a
spin angular momentum to the metric, and are still being actively researched.
There is another assumption one often makes when studying general relativity, yielding a unique
connection on a manifold endowed with some metric gµν . We obtain this connection by introducing
two additional properties: the torsion free condition as discussed above, and metric compatiability,
∇ρ gµν = 0. This second condition ends up being incredibly useful in the manipulation and differ-
entiation of tensors. Put in words, we say that a connection is metric compatible if the covariant
derivative of the metric with repsect to that specific connection vanishes everywhere.
It turns out that deriving the existence and uniqueness of this connection comes from deriving
the unique expression for the connection coefficients in terms of a metric. Let us suppose that we
have the metric compatible condition, ∇ρ gµν = 0, and permute the indices. Then, using (17.18),
we have the three expressions:
∇ρ gµν = ∂ρ gµν − Γλρµ gλν − Γλρν gµλ = 0
∇µ gνρ = ∂µ gνρ − Γλµν gλρ − Γλµρ gνλ = 0
∇ν gρµ = ∂ν gρµ − Γλνρ gλµ − Γλνµ gρλ = 0

Notice then, subtracting the last two from the first expression we find
∇ρ gµν − ∇µ gνρ − ∇ν gρµ = ∂ρ gµν − ∂µ gνρ − ∂ν gρµ − Γλρµ gλν − Γλρν gµλ
+Γλµν gλρ + Γλµρ gνλ + Γλνρ gλµ + Γλνµ gρλ
= ∂ρ gµν − ∂µ gνρ − ∂ν gρµ + 2Γλνµ gρλ = 0

where we used the symmetry of the two lower indices of the Christoffel symboles. Rearranging
leads us to
1
Γλνµ gρλ =
[∂µ gνρ + ∂ν gρµ − ∂ρ gµν ]
2
Multiplying both sides by g σρ and using the fact that g σρ gρλ = δ σλ , we find that the metric
compatible connection coeffcients can be written as
1 σρ
Γσνµ =
g [∂µ gνρ + ∂ν gρµ − ∂ρ gµν ] (17.21)
2
As an exercise, the reader will prove the condition of metric compatiability from the other
direction.
It is interesting to point out that, unlike ordinary partial differentiation, covariant differentiation
is not commutative in general. For any general tensor T a... b... , we define its commutator as
∇c ∇d T a... a...
b... − ∇d ∇c T b... (17.22)
To explicitly show that the commutator does not vanish in general, let us work out the case for
some vector V c . That is, we will work out
∇a ∇b V c − ∇ b ∇a V c
First recall that the covariant derivative of a contravariant vector is simply
∂V c
∇b V c = + Γc eb V e
∂xb
Earlier we noted that the covariant derivative of a contravariant vector returns a tensor of rank
(1, 1), therefore it is also useful to recall the covariant derivative of a (1, 1) tensor, namely
∇c T ab = ∂c T ab + Γacd T db − Γd bc T ad (17.23)
Using this we see that
∇a ∇b V c = ∇a (∂b V c + Γc eb V e )
Since the term inside is a (1, 1) tensor, we may use (17.23) yielding
∇a ∇b V c = ∇a (∂b V c + Γc eb V e ) = ∂a (∂b V c + Γc eb V e ) + Γc ad (∂b V d + Γd eb V e )
−Γd ba (∂d V c + Γc ed V e )
Similarly,
∇b ∇a V c = ∂b (∂a V c + Γc ea V e ) + Γc bd (∂a V d + Γd ea V e ) − Γd ab (∂d V c + Γc ed V e )

Now we subtract the last two expressions term by term. Examining the first term in each
expression, the difference is
∂a (∂b V c + Γc eb V e ) − ∂b (∂a V c + Γc ea V e ) = ∂a (Γc eb V e ) − ∂b (Γc ea V e )
= V e (∂a Γc eb − ∂b Γc ea ) + Γc eb ∂a V e − Γc ea ∂b V e
For the other terms, we will assume that we are considering torsion free connections, therefore
Γabc = Γacb . Subtracting these terms yields
Γc ad (∂b V d + Γd eb V e ) − Γd ba (∂d V c + Γc ed V e ) − Γc bd (∂a V d + Γd ea V e ) + Γd ab (∂d V c + Γc ed V e )
= Γc ad ∂b V d − Γc bd ∂a V d + Γc ad Γd eb V e − Γc bd Γd ea V e
Relabeling indices and again using the fact we are considering torsion free connections, the above
just becomes
Γc ea ∂b V e − Γc eb ∂a V e + Γc ad Γd eb V e − Γc bd Γd ea V e
Altogether, after some cancellation, we find that the commutator of V c is
∇a ∇b V c − ∇b ∇a V c = V e (∂a Γc eb − ∂b Γc ea ) + Γc ad Γd eb V e − Γc bd Γd ea V e
= (∂a Γc db − ∂b Γc da + Γc ae Γe db − Γc be Γc da )V d
Defining the term in the parentheses as the tensor
Rc dab ≡ ∂a Γc db − ∂b Γc da + Γc ae Γe db − Γc be Γe da (17.24)
we see that the commutator is
[∇a , ∇b ]V c = ∇a ∇b V c − ∇b ∇a V c = Rc dab V d (17.25)

The tensor Rc dab defined in (17.24) is known as the Riemann tensor or sometimes the curvature
tensor. We will look at this tensor in more detail in the next section as it is imperative in charac-
terizing the curvature of our space. Had we not assumed that our connections are torsion free, the
commutator would have ended up with a term proportional to the torsion tensor. In summary, the
only way for covariant derivatives to commute is to have the Riemann tensor (and torsion tensor if
included) vanish.
So far we have only discussed the covariant derivative, which takes a (p, q) tensor to a (p, q + 1)
tensor. A natural question arises about whether a derivative of a (p, q) tensor will give back a
(p, q) tensor. There in fact is such a derivative, the Lie derivative. Suppose we wish to differentiate
17.3. CURVATURE, PARALLEL TRANSPORT AND GEODESICS 415
a...
a tensor field Tb... (x) with respect to a tangent vector field X a . Intuitively, we seek to find the
change of our tensor field between two points. The idea is we drag our tensor at some point p on
our manifold along a curve passing through p, to another point q and compare the ’dragged along’
tensor with the tensor already at point q. Since both of these tensors are of the same rank, we are
allowed to subtract the two tensors and define a derivative in the usual way, taking the limit as the
point q tends to p.As an example, with a little effort one can define the Lie derivative of a tensor
T ab with respect to a tangent vector field X a as [14]
T ab (x) − T 0ab (x0 )

LX T ab = lim (17.26)
δu→0 δu
where T ab is our tensor field at point p, with components T ab (x) at p, and the dragged-along
tensor is T 0ab at point q with tensor components T 0ab (x0 ) at point q. This definition can actually
be expressed in a more useful form:
LX T ab = X c ∂c T ab − T ac ∂c X b − T cb ∂c X a (17.27)
There are several properties that one can prove about the Lie derivative, most of which are to be
expected of a differential operator. First, from the definition one can prove in a straightfoward way
that the Lie derivative is linear and obeys the Leibniz rule. Two other important tensorial properties
the Lie derivative has is that it is type preserving and that it commutes with contraction. By type
preserving we mean that the Lie derivative of a tensor of type (p, q) is again a tensor of type (p, q).
Moreover, the Lie derivative of a contravariant vector field Y a is given by the Lie bracket of X, Y :
LX Y a = [X, Y ]a = X b ∂b Y a − Y b ∂b X a (17.28)
Another useful identity is the Lie derivative of the metric given in terms of the covariant derivative
[11]:
LV gµν = V σ ∇σ gµν + (∇µ V λ )gλν + ∇ν V λ gµλ = ∇µ Vν + ∇ν Vµ (17.29)
We will use this definition later on when we look at Killing vectors which satisfy Killing’s equation:
LK gµν = 0 (17.30)
17.3 Curvature, Parallel Transport and Geodesics

As one reading this text is likely aware, in general relativity gravity makes itself manifest in the
curvature of space-time. The common analogy is that one can imagine placing a bowling ball or
some heavy object on a malleable sheet, say a trampoline. The bowling ball will warp the trapoline
in such a way that if one were to release the marble in a particular way, the marble would ‘orbit’
the bowling ball. Using this oversimplified toy model, we can imagine that the trampoline is the
background space-time while the bowling ball is a sun or Neutron star and the marble is a planet.
To be able to quantify this interpretation of gravity we must have a way to mathematically analyze
the curvature of our space-time. In this section our aim is to do just that.
Figure 17.1: Parallel Transport. (Left) As we parallel transport a vector along a closed curve
in flat space the vector returns to the same position it started in. (Right) A vector is parallely
transported along a closed curved along the surface of a sphere, notice the ambiguity of keeping
the vector constant at the north pole.
If we picked up a globe and asked the question “Is this object curved?”, the answer would almost
immediately be yes; the globe does not look flat. However, suppose we were a creature living on the
surface of this sphere, it would be a more difficult task to answer this question; at least, it wouldn’t
be as obvious as before. This difference in perspective is the intuitive idea behind extrinsic curvature
and intrinsic curvature. The obvious reason to why we know the globe is curved is because we live
in the space in which the globe is embedded, from which it is easy to tell when an embedded
object is curved or not. In the alternative case, which is such the case for general relativity, we are
interested in the notion of intrinsic curvature, which is defined at each point on the manifold we
are considering. This approach makes sense for the study of relativity as we cannot easily abstract
ourselves from our universe, our ‘manifold’, and must therefore be able to characterize the curvature
of our space-time in an intrinsic way. The way in which we may quantify the curvature of our space
or space-time is through a notion called parallel transport.
When we work in flat space, it is natural for us to compare vectors at different points in the space,
in which case we either forget or are unaware that vectors are really elements of tangent spaces
defined at specific points. The reason for this is because in flat space we may move any vector from
point to point in our space without the vector changing on us. In curved spaces however, this is
not true in general. Consider figure 17.1. There we have a square curve drawn in real flat space,
and we move a vector V along the curve. Notice as we ’parallely transport’ this vector around the
entire curve, the vector returns to the same position it started in; we have kept the vector constant
as we moved through our flat space.
Now consider the right side of figure 17.1, where we have a curve drawn on the surface of the
sphere. As we parallely transport a vector V around the drawn curve we notice that we run into
an issue when we reach the north pole. It becomes unclear how to keep the vector constant. In
fact, in curved space the result of parallel transporting a vector from point to point depends on the
path taken between points. We notice this is not the case for flat space. In an intuitive sense then,
we find that parallel transport is nothing more than the curved generalization of keeping vectors
constant as they are moved through space point to point. This notion of parallel transport is one
way in which we may, firstly, determine whether a space is curved, and secondly, characterize the
curvature of the space.
Of course, for our purposes, parallel transport is only helpful if we can quantify the result. Let us
generalize the constancy of a vector to the constancy of a tensor. Given some parameterized curve
xµ (λ), requiring the tensor T µ1 ,µ2 ,...µkν1 ,ν2 ,...ν` to remain constant along this curve in flat space is a
simple task: the tensor components themselves must be constant with respect to the parameter λ
[11]. That is,
d µ1 ,µ2 ,...µk dxµ ∂ µ1 ,µ2 ,...µk
T ν1 ,ν2 ,...ν` = T ν1 ,ν2 ,...ν` = 0 (17.31)
dλ dλ ∂xµ
In the spirit of keeping this statement clearly tensorial, we define the directional covariant deriva-
tive
D dxµ
= ∇µ (17.32)
dλ dλ
and replace the partial derivative in (17.31). Therefore, we define parallel transport of a tensor
T along a parameterized curve xµ (λ) as
dxσ

D µ1 ,µ2 ,...µk
T ν1 ,ν2 ,...ν` = ∇σ T µ1 ,µ2 ,...µkν1 ,ν2 ,...ν` = 0 (17.33)
dλ dλ
(17.33) is formally known as the equation of parallel transport. For a vector, one can show that
it takes the form:
DV µ dV µ dxσ ρ
= + Γµσρ V =0 (17.34)
dλ dλ dλ
From this expression we notice that the definition of parallel transport relies on the connection.
In the case of flat space, one can choose a coordinate system such that the connection vanishes,
in which we return to the definition given in (17.31). We then see that again the connection, in a
sense, yields information about whether a space is curved or not. In this text we will assume that
the connection is metric compatible, in which case we find that
Dgµν dxσ
= ∇σ gµν = 0 (17.35)
dλ dλ
since ∇σ gµν = 0. Therefore, when we consider a metric compatible connection, the metric is
always parallel transported. Moreover, with (17.35) we can see that the inner product of two parallel
transported vectors V and W , along a curve xσ (λ) is preserved:

D µ D D D µ D ν
(V Wµ ) = (gµν V µ W ν ) = gµν V µ W ν + gµν V Wν + V µ W =0
dλ dλ dλ dλ dλ
(17.36)
With parallel transport we are able to give a rigorous definition of a geodesic, the curved space
generalization of a straight line, i.e. the ’straightest possible curve’. In Euclidean space, a straight
line is a very intuitive concept: the shortest path between two points. A more strict definition is a
‘straight line is a path that parallel transports it’s own tangent vector’[11]. Using (17.34), we arrive
to the geodesic equation
D dxµ d dxµ dxσ dxρ

= + Γµσρ
dλ dλ dλ dλ dλ dλ
d2 xµ σ
µ dx dx
ρ
= + Γ σρ =0 (17.37)
dλ2 dλ dλ
Again, if we assume we are in flat Euclidean space, the connection vanishes, leaving us with
d2 xµ
= 0 ⇒ xµ (λ) = mλ + b
dλ2
the equation for a straight line.
In general relativity we care about geodesics since freely falling particles happen to follow such
paths, allowing us to explore the properties of the gravitational field around some astronomical
body without considering the field created by the particle following such a path. Sean Carroll, in
his Spacetime and Geometry, explains that, from a physical perspective, the geodesic equation is the
curved generalization of Newton’s second law. We will explore this idea more in the next chapter as
it will help motivate us to Einstein’s field equations. Moreover, an important consequence of (17.36)
in the context of space-time is that the character of a geodesic, whether it is time-like, space-like
or null, never changes because the equation of parallel transport preserves inner products, which is
how we define the character of such geodesics in the first place.
For concreteness, let us show how one would go about finding the geodesic equations for cylin-
drical coordinates. Recall that the line element for cylindrical coordinates is given by
ds2 = dr2 + r2 dφ2 + dz 2

A brief exercise shows that the Christoffel symbols are
1
Γr φφ = −r Γφrφ = Γφφr =
r
Then, using (17.37), for µ = r, σ = φ and ρ = φ, we find that one of the geodesic equations for
a cylinder is
2
d2 r

dφ
−r =0
dλ2 dλ
We also have one for µ = φ, σ = r and ρ = φ:
d2 φ 1 dr dφ
+ =0
dλ2 r dλ dλ
Lastly, we also have that Γz σρ = 0, yielding the geodesic equation
d2 z
=0
dλ2
which, as we have already seen, is the familiar equation for a straight line.
So far we have seen how parallel transport might characterize a space as curved; the parallel
transport of a vector around a closed loop in a curved space yields a transformation of that vec-
tor. This transformation in fact depends on the curvature of the space enclosed within that loop.
What’s more is there is a relationship between the commutator of a covariant derivative and parallel
transport. The commutator of two covariant derivatives measures the difference between parallel
transporting a tensor along the same two curves, however in a different order (to help, think of the
flat space case, where the commutator of the two covariant derivatives would become the commu-
tator of two partial derivatives, which commute). But we have already computed the commutator
of two covariant derivatives, we saw that it was proportional to the Riemann tensor, (17.25). It is
for this reason we identify the Riemann tensor as the curvature tensor: the failure of the covariant
derivatives to commute comes from the fact that the tensor is being parallel transported around
a closed loop enclosing some of the curvature of the specific space. In this sense, the Riemann
curvature tensor measures the curvature of the space point by point.
It should be mentioned that typically when one studies studies general relativity, they are pri-
marily concerned with finding the Christoffel symbols which is associated with a specific metric,
in which case, pragmatically, the curvature of the space is associated with the metric. In the back
of one’s mind however, it is understood that the actual curvature of the space is connected to the
Riemann tensor. For example, suppose we took a model space-time from a given line element (i.e. a
given metric) and found that the Riemann tensor vanishes, we know that the metric is ‘flat’, a met-
ric which describes a flat space-time. On the other hand, if we studied a model space-time in which
the Riemann tensor did not vanish, we would say that the metric describes a curved space-time.
Since Einstein’s equations require manipulations of the Riemann tensor, some properties of the
Riemann tensor include:
Rρσµν = −Rσρµν Rρσµν = −Rρσνµ Rρσµν = Rµνρσ (17.38)

The Riemann tensor also satisfies the Bianchi identity [14]:
∇a Rdebc + ∇c Rdeab + ∇b Rdeca = 0 (17.39)

If we take a contraction of the Riemann tensor we form the Ricci tensor
Rµν = Rλµλν (17.40)

Sometimes the Ricci tensor is viewed as the average curvature of a particular space, and was
crucial in proving the Poincaré conjecture. Even more, the trace of the Ricci tensor yields the Ricci
scalar
R = Rµµ = g µν Rµν (17.41)

To each of these tensors in action, let us work them out in the case for a 2-sphere, which has a
line element of
ds2 = a2 (dθ2 + sin2 θdφ2 )

where a is the radius of the sphere. A brief exercise shows that
Γθ φφ = − sin θ cos θ Γφθφ = Γφφθ = cot θ

One component of the Riemann tensor then is simply
Rθ φθφ = ∂θ Γθ φφ − ∂φ Γθ θφ + Γθθλ Γλφφ − Γθ φλ Γλθφ
where the index λ is a dummy index as it is being summed over. Substituting in the Christoffel
symbols above yields
Rθ φθφ = (sin2 θ − cos2 θ) − 0 + 0 − (− sin θ cos θ) cot θ = sin2 θ
Lowering the θ index yields
Rθφθφ = gθλ Rλφθφ = gθθ Rθ φθφ = a2 sin2 θ
It is a simple task to show that all of the other components of the Riemann tensor either vanish
or are related to the above using the symmetries given in (17.38). From here we may compute the
components of the Ricci tensor using Rµν = g αβ Rαµβν :
Rθθ = g φφ Rφθφθ = 1 Rφφ = g θθ Rθφθφ = sin2 θ Rθφ = Rφθ = 0
Lastly, the Ricci scalar is simply
2
R = g θθ Rθθ + g φφ Rφφ =
a2
in which case we see that the Ricci scalar for a 2-sphere is a constant inversely proportional to the
sphere’s radius. We note that the scalar curvature decreases as the radius of our sphere increases.
This makes intuitive sense as we would say that the curvature of a marble is greater than the
curvature of a basketball.
The Riemann curvature tensor also shows up when one studies geodesics. At some point it
is likely that the reader has heard the defining property of Euclidean geometry characterizing its
‘flatness’: the parallel postulate. The parallel postulate states that, in flat space, initially parallel
lines remain parallel. Intuitively however we know that this isn’t in general true for curved spaces.
To see this readily examine any ordinary map of the earth showing curved longitude lines that run
from the north pole to the south pole. At the equator these lines are parallel, but as one moves
north or south, neighboring lines begin to move together.
The disconnect here is that the notion of parallel does not naturally extend from flat space to
curved space. This problem, observing the behavior of two initially parallel geodesic curves, is
formally known as geodesic deviation. To determine such behavior, consider a congruence, a set
of curves such that each point p on a manifold lies on a single curve. In the case of relativity, we
consider a congruence of time-like geodesics. Let us call the tangent vector to such a curve ua .
Moreover, let us define the connecting vector η a as a vector that points from one geodesic to its
neighbor. That is, a connecting vector joins two points on neighboring parameterized curves at the
same value of the parameter [37]. Figure 17.2 gives an illustration of this set up. Since η a points
to neighboring geodesics, we are motivated to define the relative velocity between geodesics as
Figure 17.2: Two curves γ and µ on some particular manifold. V, the deviation vector, connects
curves at two points p on γ and η on µ.
a
V a = (∇u η) = ub ∇b η a
and the relative acceleration between neighboring geodesics as
D2 a a
Aa = η = (∇u V ) = ub ∇b V a
dτ 2
Moreover, since the tangent vector ua and the connecting vector η a are basis vectors adapted to
some coordinate system, one can show that the commutator [u, η] vanishes. Or, in terms of the Lie
bracket, we have
[u, η]a = ub ∇b η a − η b ∇b ua = 0 (17.42)

Then, for intertial geodesics with a tangent vector ua and a parameter τ we have that the
acceleration of the neighboring geodesics is given by
D2 ηa
= ub ∇b (uc ∇c η a ) = ub ∇b (η c ∇c ua )
dτ 2
where we used (17.42). Using the Leibniz rule we find
D2 ηa
= ub ∇b η c ∇c ua + ub η c ∇b ∇c ua
dτ 2
Using (17.25) we have ∇b ∇c ua = ∇c ∇b ua + Radbc ud . Therefore,
D2 ηa
= ub ∇b η c ∇c ua + ub η c (∇c ∇b ua + Radbc ud )
dτ 2
= ub ∇b η c ∇c ua + η c ub ∇c ∇b ua + η c ub ud Radbc
= η b ∇b uc ∇c ua + η c ub ∇c ∇b ua + η c ub ud Radbc
Relabeling dummy indices and setting η c ub ∇c ∇b ua = η b uc ∇b ∇c ua we find
D2 ηa
= η b ∇b uc ∇c ua + η b uc ∇b ∇c ua + η c ub ud Radbc
dτ 2
Using the Leibniz rule once more, ∇b (uc ∇c ua ) = ∇b uc ∇c ua + uc ∇b ∇c ua , gives us
D2 ηa
= η b (∇b (uc ∇c ua )) + η c ub ud Radbc
dτ 2
But, as the reader will show, a tangent vector to a geodesic satisfies
uc ∇c ua = 0
Therefore, after a little rearranging and relabeling of dummy indices we arrive to the geodesic
deviation equation
D2 ηa
= Rabcd ub uc η d (17.43)
dτ 2
In words this equation tells us that the relative acceleration between neighboring geodesics is
proportional to the curvature of the space. Put another way, the Riemann curvature tensor measures
the failure for initially parallel geodesics to remain parallel. From here we notice that if our space
is flat, in which case the Riemann curvature is zero, we recover the parallel postulate of Euclidean
geometry. If one were to interpret this result physically, as we will do in the next chapter, they
would see that geodesic deviation exemplifies that gravity exhibits itself through tidal forces.
17.4 Killing Vectors and Symmetries

As the reader is likely aware, the laws of physics are built around a set of conservation laws, coming
from symmetries innate in the physical system (e.g. space translations, Lorentz transformations,
etc.). When we study relativity, a natural question arises: how does one find symmetries in relativity
when the theory is so geometric? The method of finding symmetries is not immediately obvious.
It turns out that to find a symmetry geometrically one looks to find when the metric is the same
as when one moves point to point in the manifold. Without going into the rigorous details, when
the metric remains the same as we move from point to point in a manifold we say it is isometric
between these two points. As it happens, we may find such geometric symmetries of a specific
17.4. KILLING VECTORS AND SYMMETRIES 423
space by determining the Killing vectors associated with that space. More specifically, if X is some
Killing vector, as we move along the direction of X it can be shown that the metric does not change,
leading to a conserved quantitiy. The way we quantify this notion is if all of the metric components
vanish, i.e.
∂σ∗ gµν = 0 (17.44)

∗
for some fixed σ , there will be a symmetry (as the metric does not change). From this expression
we can see why we might care about this notion of a symmetry when we study relativity theory. As
the reader will prove in an exercise, the motion of test particles may be described by the geodesic
equation
pλ ∇λ pµ = 0 (17.45)
By metric compatibility we are free to lower the µ and expand the covariant derivative to yield
that the above becomes
pλ ∂λ pµ − Γσλµ pλ pσ = 0
Using the definition of four-velocity, we find that the first term in the above expression becomes
dxλ dxλ d dpµ

pλ ∂λ pµ = m
∂λ pµ = m pµ = m
dτ dτ dxλ dτ
Meanwhile, the second term can be written using the definition of the Christoffel symbols:
1 σν
Γσλµ pλ pσ = g (∂λ gµν + ∂µ gνλ − ∂ν gλµ )pλ pσ
2
1 1
=(∂λ gµν + ∂µ gνλ − ∂ν gλµ )pλ pν = (∂µ gνλ )pλ pν
2 2
where we used the symmetry of pλ pν to eliminate two of the derivative terms in the above line.
Altogether then
dxλ d 1
m pµ = (∂µ gνλ )pλ pν
dτ dxλ 2
Therefore, if we insist that the metric satisfies (17.44) we find then
dxλ d
pµ = 0
dτ dxλ
which we recognize as the conservation law for momentum.
For practice, let us examine the familiar space of a two 2-sphere and find the associated Killing
vectors. Using the line element for a sphere
ds2 = a2 dθ2 + a2 sin2 θdφ2

recall that we had said the Christoffel symbols are
Γθ θθ = Γφθθ = Γθ φθ = Γφφφ = 0 Γφφθ = Γφθφ = cot θ Γθ φφ = − sin θ cos θ

Earlier we had said that Killing vectors satisfy Killing’s equation:
∇b Xa + ∇a Xb = 0
Moreover, recall that the covariant derivative of a covariant vector is given by
∇b Va = ∂b Va − Γc ab Vc
Then, if we first consider when a = b = θ, we find
∇θ Xθ + ∇θ Xθ = 0 ⇒ ∇θ Xθ = 0
⇒ ∂θ Xθ − Γc θθ Vc = ∂θ Xθ = 0 ⇒ Xθ = f (φ)
Now consider the case when a = b = φ from which ∇φ Xφ = 0 from which we find
∂φ Xφ = − sin θ cos θXθ = − sin θ cos θf (φ)

Integrating yields,
Z
Xφ = − sin θ cos θ f (φ0 )dφ0 + g(θ)
where g(θ) is an integration constant. Lastly, when a = θ and b = φ we end up with
∇θ Xφ + ∇φ Xθ = 0
which yields
∂θ Xφ + ∂φ Xθ = 2 cot θXφ
But,
Z Z
∂θ Xφ = ∂θ sin θ cos θ f (φ0 )dφ0 + g(θ) = (sin2 θ − cos2 θ) f (φ0 )dφ0 + ∂θ g(θ)
Moreover, ∂φ Xθ = ∂φ f (φ). Putting everything together we find

Z
∂θ Xφ + ∂φ Xθ = (sin2 θ − cos2 θ) f (φ0 )dφ0 + ∂θ g(θ) + ∂φ f (φ) = 2 cot θXφ
Z
= −2 cos2 θ f (φ0 )dφ0 + 2 cot θg(θ)
Rearranging and adding 2 cos2 θ f (φ0 )dφ0 to both sides yields

R
Z
f (φ0 )dφ0 + ∂φ f (φ) = 2 cot θg(θ) − ∂θ g(θ) (17.46)
Now using a method from the study of partial differential equations, we suppose we change the
right hand side by a small amount in θ; we must therefore also change the left hand side by the
same amount θ. But the left hand side does not depend on θ and so nothing changes. A similar
argument holds if we were to imagine altering the left hand side by a small change in φ. If this
17.4. KILLING VECTORS AND SYMMETRIES 425
equation is to be true for all values of φ and θ, we require that it is actually equal to a constant k.
Notice that if we differentiate the left hand side with respect to φ we find
d2 f d
+ f (φ) = k = 0 ⇒ f (φ) = A cos φ + B sin φ
dφ2 dφ
which implies
Z
f (φ0 )dφ0 = A sin φ − B cos φ ∂φ f (φ) = −A sin φ + B cos φ
from which we find

Z
f (φ0 )dφ0 + ∂φ f (φ) = 0 = k ⇒ k = 0
Shifting our focus to the right hand side of (17.46), using the method of integrating factors one
can show that [38]
g(θ) = sin2 θ(k cot θ + C)

where C is some integration constant. But we just saw that k = 0, hence g(θ) = C sin2 θ. From
these last two results we may finally write down the Killing vectors associated with the 2-sphere:
Xθ = A cos φ + B sin φ Xφ = − sin θ cos θ(A sin φ − B cos φ) + C sin2 θ (17.47)
Using the metric to raise the indices of our Killing vectors we find X θ = Xθ and sin2 θX φ = Xφ .
Then, if we expand our Killing vector in the coordinate basis using partial derivatives to represent
the basis vectors we find
X = X θ ∂θ + X φ ∂φ = (A cos φ + B sin φ)∂θ + [C − cot θ(A sin φ − B cos φ)]∂φ
A cos φ∂θ − A cot θ sin φ∂φ + B sin φ∂θ + B cot θ cos φ∂φ + C∂φ
= −ALx + BLy + CLz

where we have identified the angular momentum operators:
Lx = − cos φ∂θ + cot θ sin φ∂φ Ly = sin φ∂θ + cot θ cos φ∂φ Lz = ∂φ
As the reader will show in an exercise, the covariant derivatives of a Killing vector can related
to the Riemann tensor in the following way:
∇µ ∇σ K ρ = Rρ σµν K ν (17.48)
Moreover, contraction leads to a relation with the Ricci tensor
∇µ ∇σ K µ = Rσν K ν (17.49)
Returning to relativity for a moment, as one might expect, the existence of a time-like Killing
vector leads to the conservation of energy for the entire space-time. Given a Killing vector Kν and a
conserved energy-momentum tensor Tµν , we can construct a current J µ = Kν T µν that is conserved
by construction since
∇µ J µ = (∇µ Kν )T µν + Kν (∇µ T µν ) = 0 + 0 = 0
where we used the fact that Kν is a Killing vector, and the conserved energy-momentum tensor
satisfies ∇µ T µν = 0, as we will examine in the next chapter. From here one can define the total
energy of the space-time by integrating this current over a space-like hypersurface [11].
17.5 Exterior Algebra and Exterior Differentiation

As one might imagine, tensor calculus can be quite exhausting, especially if we are to determine
the components of the Riemann tensor of given a particularly nasty line element. There is in fact a
slightly more practical way to go about computing such information, one which involves some messy
calculations, however not nearly as bad as the ones using the covariant derivative and Christoffel
symbols. Rather, as we will develop over the next few sections, there is a slightly more practical
method, both mathematically and physically, which makes heavy use of dual vectors, or as we
will learn to call them, differential forms. Before we get to computing curvature using this more
pragmatic method, let us first outline the basics of the algebra and calculus of forms.
In a general, a differential p-form is a (0, p) rank tensor that is totally antisymmetric. Practically
this means that any exchange of the indices of a form yields a minus sign. We often denote the
space of all the p-forms as Λp (M ) where M is a given manifold. By definition, a scalar is a 0-form,
and hence lives in the space Λ0 (M ), while a dual vector is a 1-form, thereby living in Λ1 (M ). In
general, an arbitrary p-form can be written as a combination of wedged basis forms ω a
1
α= αa a ...a ω a1 ∧ ω a2 ∧ ... ∧ ω ap
p! 1 2 p
We can do both algebra and calculus with differential forms, all of which can be summarized
by three operators. First, given a p-form α and a q-form β, we can form a (p + q) form using the
exterior product ∧, the so-called wedge product, which, for practical purposes, takes the form:
α ∧ β = (−1)pq β ∧ α (17.50)
from which we immediately see that for any odd dimensional p-form we have α ∧ α = −α ∧ α = 0.
In short, the wedge product is defined by the map:
∧ : Λp (M ) × Λq (M ) → Λp+q (M ) (17.51)
One can actually show from the definition that the wedge product is both linear and associative,
i.e. given forms α, β, γ and constants a, b, we have that [38]
(aα + bβ) ∧ γ = aα ∧ γ + bβ ∧ γ
There is another important algebraic operator which will consider momentarily, but first let us
do some calculus with differential forms. We call this operator the exterior derivative denoted by d,
17.5. EXTERIOR ALGEBRA AND EXTERIOR DIFFERENTIATION 427
which acts very similar to the familiar total derivative. Exterior derivatives allow us to differentiate
p-forms to give us p + 1-forms. That is, the exterior derivative is defined by the map:
d : Λp (M ) → Λp+1 (M ) (17.52)
As an exercise, the reader will show that the exterior derivative satisfies a modified form of the
Leibniz rule when applied to the product a p-form ω and a q-form η:
d(ω ∧ η) = (dω) ∧ η + (−1)p ω ∧ (dη) (17.53)
Unlike the ordinary partial derivative, the exterior derivative, due to its antisymmetric character,
automatically transforms as a tensor. This is one of the reasons why we like to work with differential
forms, particularly in relativity: we are not required to introduce a connection to act as a correction
term when we take the derivative of a tensor. This is one of the reasons why differential geometers
like to work with forms. One can integrate and differentiate forms without being forced to introduce
additional geometric structure on the specific space one is considering. In fact, one of the only
necessary geometric objects is the metric.
We said that the exterior derivative acts like the total derivative. This is precisely the definition
of the exterior derivative of some scalar function f , i.e.
∂f
df = dxa
∂xa
We also see that the total derivative here operates like the exterior derivative. We started with a
scalar function, a 0-form, and obtained a 1-form. One interesting fact about the exterior derivative
is that for any form α, we have that it satisfies Poincaré’s lemma:
d(dα) = 0 (17.54)
Typically we will refer to this property as d2 = 0. This lemma is actually a consequence of the
definition of d and the fact that partial derivatives commute. Just for completeness, we say that a
p-form α is closed if dα = 0, and exact if α = dβ for some (p − 1) form β. From these definitions we
immediately see that all exact forms are also closed, however the converse is not necessarily true.
The Poincaré lemma and the definitions closed and exact come into play when one examines the
differential form formulation of Maxwell’s equations. There are also some familiar vector calculus
identities that one obtains from the Poincaré lemma, as we will see momentarily.
Another useful identity one can work out is
d(fa dxa ) = dfa ∧ dxa (17.55)
where fa is some scalar function. For concreteness, let us work out a few examples. Consider
the 1-form α = ef (r) dt. We know α is a one form as it is being expressed in terms of a basis dual
vector, which we now know lives in the space Λ1 , and hence is a 1-form. Notice then that if we
take the exterior derivative we find
∂ f (r) ∂f f (r)
dα = d(ef (r) dt) = (e )dr ∧ dt = e dr ∧ dt
∂r ∂r
Notice here we used the identity given in (17.55). If we were to take another exterior derivative,
we would find that, by the Poincaré lemma, d(dα) vanishes. For a slightly more involved computa-
tion, consider the 1-form β = eg(r) sin θ cos φdr. This time the exterior derivative is slightly more
complicated:
dβ = d(eg(r) sin θ cos φdr) = g 0 (r)eg(r) sin θ cos φdr∧dr+eg(r) cos θ cos φdθ∧dr−eg(r) sin θ sin φdφ∧dr
= eg(r) cos θ cos φdθ ∧ dr − eg(r) sin θ sin φdφ ∧ dr

where we used the fact that dr ∧ dr = 0. Now consider a slightly more interesting example.
Suppose we have the 1-form α = αi dxi = Adx + Bdy + Cdz. Taking the exterior derivative we find
dα = (dA) ∧ dx + (dB) ∧ dy + (dC) ∧ dz

∂A ∂A ∂A ∂B ∂B ∂B ∂C ∂C ∂C
= dx + dy + dz ∧dx+ dx + dy + dz ∧dy+ dx + dy + dz ∧dz
∂x ∂y ∂z ∂x ∂y ∂z ∂x ∂y ∂z

∂B ∂A ∂C ∂B ∂A ∂C
= − dx ∧ dy + − dy ∧ dz + − dz ∧ dx
∂x ∂y ∂y ∂z ∂z ∂x
where we again used dxi ∧ dxi = 0, and dxi ∧ dxj = −dxj ∧ dxi for i 6= j. The final expression we
have come up with looks strikingly similar to the curl of a vector field. The above is not the curl,
but it is quite close. We can actually write the expression for the curl of a vector field in terms of
the exterior derivatives, however we need the third and final operator that is useful when working
with differential forms, the Hodge star operator.
The Hodge star operator is defined as follows: given an n-dimensional manifold M , there is an
operator, the Hodge star operator, which maps p-forms to (n − p)-forms. That is,
∗ : Λp (M ) → Λn−p (M ) (17.56)
Interestingly, if we were to apply the Hodge star twice, on any form α, we would find [11]
∗ ∗ α = (−1)s+p(n−p) α (17.57)
where s is the number of minus signs of the eigenvalues of the metric, or the signature. The
above identity leads us to the notion of Hodge duality: using the star operator twice will return us
to the same space, while using it once we move between different ranked form spaces. As a quick
example, suppose we were working in three dimensional ordinary Euclidean space. We therefore
have the basis 1-forms, dx, dy, dz ∈ Λ1 (R3 ). Using the definition given (17.56) we can determine
the Hodge star of each of these 1-forms:
∗dx = dy ∧ dz ∗ dy = dz ∧ dx ∗ dz = dx ∧ dy
We started with a 1-form, and went to a (3 − 1)-form, yielding for us 2-forms. Had we worked in
Euclidean 2-space, we would have found that the Hodge star operating on our basis 1-forms gives
us other 1-forms.
The definition we have for the Hodge star operator is true but not very useful for practical
purposes. One can work out a more pragamatic definition of the Hodge star operator, namely:
α ∧ ∗α = g(α, α)ω (17.58)

where α ∈ Λp (M ), ∗α ∈ Λn−p (M ), and ω is the volume form, sometimes denoted by dn x =
dx0 ∧ dx1 ∧ ... ∧ dxn−1 . To see how this works, consider the case with spherical coordinates, where
our volume element is simply ω = r2 sin θdr ∧ dθ ∧ dφ. As we will see in the next section, we can
come up with an orthonormal basis {dr, r sin θ, rdθ}, such that
g(dr, dr) = g(r sin θ, r sin θ) = g(rdθ, rdθ) = 1

We now have everything we need to calculate the Hodge dual of each of the basis forms at each
rank (i.e. both 1-forms, 2-forms, and 3-forms). For example, notice that we may compute the
Hodge dual of the basis 1-form dr using (17.58):
dr ∧ ∗dr = g(dr, dr)r2 sin θdr ∧ dθ ∧ dφ = r2 sin θdr ∧ dθ ∧ dφ

For both the left hand and right hand sides to hold, we conclude that it must be
∗dr = r2 sin θdθ ∧ dφ (17.59)

Similarly,
rdθ ∧ ∗rdθ = g(rdθ, rdθ)ω = r2 sin θdr ∧ dθ ∧ dφ ⇒ ∗rdθ = −r sin θdr ∧ dφ (17.60)
and
∗r sin θdφ = rdr ∧ dθ (17.61)

An interesting calculation is finding the Hodge dual of ω itself:
ω ∧ ∗ω = g(ω, ω)ω
One can work out g(ω, ω) = 1, leading us to
∗ω = 1 (17.62)
Although we were implicitly working in spherical coordinates, this result actually holds for any
volume form. Moreover, by (17.57) it follows that
∗1 = ω (17.63)
which also holds in general. As an exercise, the reader will work out the rest of the Hodge duals
in spherical coordinates.
One reason why we care about the Hodge star operator is that we may write nearly all of the
important vector identities in terms of ∧, d, and ∗. Given two 1-forms, α, β, one can prove that the
dot product and cross product can be written as:
α · β = ∗(α ∧ ∗β) α × β = ∗(α ∧ β) (17.64)

As an example, consider two general 1-forms in spherical coordinates. Let α = Adr + Brdθ +
Cr sin θdφ, and β = Edr + F rdθ + Gr sin θdφ, where A, B, C, E, F, G are all real vector valued
functions. The Hodge dual of β is simply
∗β = Er2 sin θdθ ∧ dφ − F r sin θdr ∧ dφ + Grdr ∧ dθ

Wedging α with ∗β yields
α ∧ ∗β = (AE + BF + CG)r2 sin θdr ∧ dθ ∧ dφ
⇒ ∗(α ∧ ∗β) = AE + BF + CG = α · β
A similar exercise proves that the cross product satisfies the expression given above. The case of
the cross product is actually an interesting one. It says that the Hodge dual of the wedge product
of any two 1-forms gives another 1-form. In Euclidean space we have the cross product operation
which takes two vectors and maps them to another vector. This says something about the existence
of the cross product: it can only exist in three dimensions. This is because only in three dimensions
is there a map which takes two dual vectors to a third dual vector. That is, it is due to the dual
space, the space of dual vectors or 1-forms which forces the cross product to work only in three
dimensions. As it turns out, for other reasons, the cross product works in 7-dimensions as well, but
we won’t go into the reasons here.
The Hodge star operator also allows us to define the three famous vector calculus operators: div,
grad, curl. Given a 1-form field F , one can prove that the divergence and curl of F is given by
∇ · F = ∗(d ∗ F ) ∇ × F = ∗dF (17.65)

Moreover, given any scalar function f , one can show that the gradient of f and Laplacian of f
can be written as
∇f · d~r = df ∆f = ∗d ∗ (df ) (17.66)

From these identities it is straightforward to prove that, as a consequence of the Poincaré lemma
d2 = 0, we obtain the familiar rules
∇ · (∇ × F ) = 0 ∇ × (∇ · F ) = 0 (17.67)
Conversely, if one assumed these rules, one could work backwards and show that the Poincaré
lemma must hold.
For concreteness, let us work out the Laplacian in spherical coordinates. First let us work out
the gradient of some scalar function f in spherical coordinates. This isn’t too bad since all we have
to do is apply a total differential to f :
∂f ∂f ∂f ∂f
df = a
dxa = dr + dθ + dφ
∂x ∂r ∂θ ∂φ
which may be rewritten as
∂f 1 ∂f 1 ∂f
df = dr + rdθ + r sin θdφ
∂r r ∂θ r sin θ ∂φ
Then, taking the Hodge dual gives us
∂f 2 1 ∂f 1 ∂f
∗df = r sin θdθ ∧ dφ − r sin θdr ∧ dφ + rdr ∧ dθ
∂r r ∂θ r sin θ ∂φ
Taking the exterior derivative of ∗df yields

∂ ∂f 2 ∂ ∂f 2 ∂ ∂f 2
d ∗ df = r sin θ dr + r sin θ dθ + r sin θ dφ ∧ dθ ∧ dφ
∂r ∂r ∂θ ∂r ∂φ ∂r

∂
1 ∂f ∂ 1 ∂f ∂ 1 ∂f
− r sin θ dr + r sin θ dθ + r sin θ dφ ∧ dr ∧ dφ
∂r
r ∂θ ∂θ r ∂θ ∂φ r ∂θ

∂ 1 ∂f ∂ 1 ∂f ∂ 1 ∂f
+ r dr + r dθ + r dφ ∧ dr ∧ dθ
∂r r sin θ ∂φ ∂θ r sin θ ∂φ ∂φ r sin θ ∂φ
Since α ∧ α = 0, we can see that we will only get one term from each of these lines. In the first
line, the only term we get is
2
∂ ∂f 2 ∂ f 2 ∂f
r sin θ dr ∧ dθ ∧ dφ = (r sin θ) + (2r sin θ) dr ∧ dθ ∧ dφ
∂r ∂r ∂r2 ∂r
2
∂ f 2 ∂f
= + r2 sin θdr ∧ dθ ∧ dφ
∂r2 r ∂r
In the second line, the only non-zero term is

2
∂ 1 ∂f ∂ f ∂f
− r sin θ dθ ∧ dr ∧ dφ = sin θ + cos θ dr ∧ dθ ∧ dφ
∂θ r ∂θ ∂θ2 ∂θ
1 ∂2f

1 ∂f
= + cos θ r2 sin θdr ∧ dθ ∧ dφ
r2 ∂θ2 r2 sin θ ∂θ
Finally, in the third line, the only non-vanishing term is
∂2f

∂ 1 ∂f 1
r dφ ∧ dr ∧ dθ = r 2 dr ∧ dθ ∧ dφ
∂φ r sin θ ∂φ r sin θ ∂φ
1 ∂2f 2
= 2 ∂φ2 r sin θdr ∧ dθ ∧ dφ
r2 sin θ
Summing each of these results together and taking the Hodge dual yields
2
1 ∂2f ∂2f

∂ f 2 ∂f cotθ ∂f 1
∗d ∗ df = + + + + (17.68)
∂r2 r ∂r r2 ∂θ2 r2 ∂θ r2 sin2 θ ∂φ2
which we recognize as the Laplacian given in spherical coordinates. As an exercise, the reader
will work out similar expressions for the divergence and curl of an arbitrary 1-form field in spherical
coordinates.
17.6 Holonomic vs. Non-Holonomic Bases

Up to this point we have been working in a coordinate basis, which is sometimes called a holonomic
basis. In the coordinate basis we represent the basis vectors as ordinary partial derivatives and basis
1-forms as differentials, i.e. ea = ∂a , ω a = dxa . There is one specific issue with using a holonomic
basis however, it is not necessarily orthonormal. Take, for example, spherical coordinates in the
coordinate basis. There we have
er · er = grr = 1 eθ · eθ = gθθ = r2 eφ · eφ = gφφ = r2 sin2 θ

where it is obvious that the unit vectors are not normalized. As mentioned earlier, we can in
fact define a basis that is orthonormal such that the inner product we construct satisfies
g(ea , eb ) = ηab
where ηab is the flat space metric (or, in relativity theory, the Minkowski metric). The reason
we choose such a basis is because in most cases where we have to calculate the curvature of a
space, using an orthonormal basis is most conveinent. We call this type of basis a non-coordinate
or non-holonomic basis, and we denote that we are working in such a basis using hat notation, i.e.
eâ , ω â .
A non-holonomic basis also has physical significance. A coordinate basis represents the global
coordinate system of a space-time while a non-coordinate basis is one used by an observer in a local
Lorentz frame [38]. Since a non-holonomic basis represents a local frame, the flat metric ηâb̂ ends
up being the Minkowski metric.
The remaining question is how to actually construct a non-holonomic basis. Luckily, it’s a rather
a simple procedure. Above we noticed that the unit vectors in spherical coordinates were not
normalized (aside from er ). To make them normalized, all we have to do is multiply by the correct
factor, which just so happens to be the coefficients of the differentials in the line element. Therefore,
the non-coordinate basis vectors in spherical coordinates is simply
1 1
er̂ = ∂r ∂θ
eθ̂ = eφ̂ = ∂φ
r r sin θ
These basis vectors can be easily checked to yield the unit normalization we seek, e.g.:
1 1 r2
eθ̂ · eθ̂ = ∂θ · ∂θ = eθ · e θ = =1
r2 r2 r2
This example yields the general method of coming up with a non-holonomic basis: divide each
of the coordinate basis vectors by the right correction, which in most cases ends up being one of
the coefficients in the line element.
There is also a relatively straightforward way of determining whether a basis is holonomic or
not. This is done by calculating the so-called commutation coefficients. We are familiar with the
fact that partial derivatives commute
17.7. CARTAN’S STRUCTURE EQUATIONS AND TETRAD METHODS 433
[∂x , ∂y ] = 0
Therefore, given the representation of the unit vectors in the coordinate basis, all commute with
each. For example, in the case of spherical coordinates we have
[er , eθ ] = [er , eφ ] = [eθ , eφ ] = 0

Alternatively, let us calculate two basis vectors in the non-holonomic representation, namely
[er̂ , eθ̂ ]. To determine the action of this commutator, let us have it act on some test function f .
Therefore

1 1 1
[er̂ , eθ̂ ]f = ∂r , ∂θ f = − 2 ∂θ f = − eθ̂ f
r r r
from which we can glean
1
[er̂ , eθ̂ ] = − eθ̂
r
What this calculation indicates is that the commutator of non-holonomic basis vectors do not
always vanish. In general, the commutator of two basis vectors is given by
[ei , ej ] = Cij k ek (17.69)

where the Cij k are our commutation coefficients. Therefore, we can figure we have a holonomic
basis when all of the commutation coefficients vanish for a given set of basis vectors. We can also
determine if a basis is holonomic or not using basis 1-forms. In this case, one can calculate the
commutation coefficients using [38]
1
dω a = − Cbc a ω b ∧ ω c (17.70)
2
Notice however, since in the coordinate basis we have that ω a = dxa , it follows from the Poincaré
lemma that
1
− Cbc a dxb ∧ dxc = 0
2
Therefore, when working in the holonomic basis, the basis 1-forms satisfy dω a = 0. In the non-
holonomic basis this is not in general true, allowing one to determine the commutation coefficients.
There is of course a method in which one can transform back and forth from the coordinate basis
and non-coordinate basis using something called a tetrad or vierbein (or, in dimensions greater than
four, a vielbein), however we will not care to do this in this book.
17.7 Cartan’s Structure Equations and Tetrad Methods

In general relativity, the name of the game is to solve Einstein’s fields equations. As we will see in
the next chapter, these equations involve the components of the Ricci tensor, and the Ricci scalar.
To find them we must compute the components of the Riemann tensor. Therefore, a typical problem
in relativity is, given some metric, hence some line element modeling a specific space-time, we seek
to calculate the curvature of the space-time by computing the components of the Riemann tensor.
We have already discussed one method of computing the components of the Riemann tensor, one
which makes heavy use of Christoffel symbols. In this section we will consider a simpler approach,
one which makes use of differential forms in a non-holonomic basis.
The essential approach is, given a set of basis 1-forms in a non-coordinate frame ω â , we calculate
dω . Though we won’t derive it here, it is relatively straightforward to prove that dω â satisfy
â
dω â + Γâb̂ ∧ ω b̂ = 0 (17.71)
where Γâb̂ are called the curvature one forms which also satisfy
Γâb̂ = Γâb̂ĉ ω ĉ (17.72)
where Γâb̂ĉ are called Ricci rotation coefficients which are related to Christoffel symbols. Equa-
tion (17.71) is referred to as Cartan’s first structure equation. Moreover, from the definition of the
Christoffel symbols, one can work out to show that
Γâb̂ + Γb̂ â = 0
Now recall the form of the Riemann tensor we came up with before
Rabcd = ∂c Γabd − ∂d Γabc + Γe bd Γaec − Γe bc Γaed (17.73)

As we saw, to find the components of the Riemann tensor one must find all of the Christoffel
symbols, making this calculation a pain to do, especially with some of the metrics one is given
in relativity theory. There is a more esoteric approach developed by Eli Cartan. Starting from
Cartan’s first structure equation one can derive Cartan’s second structure equation
Ωâb̂ = dΓâb̂ + Γâĉ ∧ Γĉ b̂ (17.74)
where Ωâb̂ are called the curvature 2-forms. It turns out one can directly relate the Riemann
curvature tensor to the curvature two forms in the following way:
1 â ˆ
Ωâb̂ = R ω ĉ ∧ ω d (17.75)
2 b̂ĉdˆ
All in all, using both of Cartan’s structure equations, (17.71) and (17.73), and using (17.74) we
can calculate the components of the Riemann curvature tensor in a local Lorentz frame, opposed
to finding them in a global frame. This will tell us what the curvature of the space or space-time
is locally. Moreover, if we were to choose to do so, after we find the components in the local frame,
we could always transform back to the global frame and see how the curvature of the space appears
globally. This method, sometimes called the method of tetrads, is a common technique in analyzing
different models of space-time, and is crucial to understand when studying general relativity.
To become acquainted with this technique we will consider two examples, first the metric of the
unit sphere which is likely the simplest non-trivial case. We will then move on to the Scwarzschild
metric, one which we will study in more depth in chapters to come as it yields the solution of a
non-rotating, uncharged black hole. Just a warning, these calculations are quite lengthy. Therefore,
the reader should sit back with a pen and plenty of paper in hand, and prepare for an endurance
test.
Recall the line element for the unit sphere:
ds2 = dθ2 + sin2 θdφ2

From here we realize that our non-coordinate basis of 1-forms is {ω θ̂ , ω φ̂ } = {dθ, sin θdφ}. Let
us compute the Ricci scalar of this metric using Cartan’s structure equations. To do this we must
first use Cartan’s first structure equation to find the curvature one forms. Using (17.71), we see
that for ω θ̂ we get
d(ω θ̂ ) = d(dθ) = 0 ⇒ Γθ̂ b̂ ∧ ω b̂
This doesn’t give us much information, so let’s move on to the other basis 1-form, ω φ̂ . This time
we have
dω φ̂ = d(sin θdφ) = cos θdθ ∧ dφ = cotθω θ̂ ∧ ω φ̂ = −cotθω φ̂ ∧ ω θ̂
= −Γφ̂θ̂ ∧ ω θ̂ − Γφ̂φ̂ ∧ ω φ̂
But since Γâb̂ = −Γb̂ â it follows that Γφ̂φ̂ = 0, leaving us with
−cotθω φ̂ ∧ ω θ̂ = −Γφ̂θ̂ ∧ ω θ̂ ⇒ Γφ̂θ̂ = −cotθω φ̂

One can check that this is the only non-zero curvature 1-form. From here we use Cartan’s second
structure equation and our newly found result to calculate the curvature 2-forms. Using (17.74) we
have
Ωφ̂θ̂ = dΓφ̂θ̂ + Γφ̂ĉ ∧ Γĉ θ̂ = dΓφ̂θ̂ = d(cos θdφ) = − sin θdθ ∧ dφ
= ω φ̂ ∧ ω θ̂
One can quickly check that this is the only non-zero curvature 2-form. From here we can calculate
the components of the Riemann tensor using (17.75). We find then
1 φ̂ ˆ
Ωφ̂θ̂ = R ω ĉ ∧ ω d
2 θ̂ĉdˆ
ˆ ˆ This yields
The only non-zero terms are those containing ω ĉ ∧ ω d when ĉ 6= d.
1 1 1 φ̂
Ωφ̂θ̂ = Rφ̂θ̂θ̂φ̂ ω θ̂ ∧ ω φ̂ + Rφ̂θ̂φ̂θ̂ ω φ̂ ∧ ω θ̂ = R θ̂θ̂φ̂ − Rφ̂θ̂φ̂θ̂ ω θ̂ ∧ ω φ̂
2 2 2
Using the fact that Rabcd = −Rbacd = −Rabdc , we find Rφ̂θ̂φ̂θ̂ = −Rφ̂θ̂θ̂φ̂ . Therefore,
Ωφ̂θ̂ = ω φ̂ ∧ ω θ̂ = Rφ̂θ̂θ̂φ̂ ω θ̂ ∧ ω φ̂
from we which we realize
Rφ̂θ̂θ̂φ̂ = −1 Rφ̂θ̂φ̂θ̂ = +1
Moreover, using the flat space metric ηab = diag(−1, 1, 1, 1) we have that
1 = Rφ̂θ̂φ̂θ̂ = η φ̂φ̂ Rφ̂θ̂φ̂θ̂ = Rφ̂θ̂φ̂θ̂ = Rθ̂φ̂θ̂φ̂ = ηθ̂θ̂ Rθ̂ φ̂θ̂φ̂ = Rθ̂ φ̂θ̂φ̂
Now recall that the Ricci tensor is just the contraction of the Riemann tensor
Rc acb = Rab
We find then that the only non-zero components of the Ricci tensor are
Rθ̂θ̂ = Rφ̂θ̂φ̂θ̂ + Rθ̂ θ̂θ̂θ̂ = Rφ̂θ̂φ̂θ̂ = 1
Rφ̂φ̂ = Rθ̂ φ̂θ̂φ̂ + Rφ̂φ̂φ̂φ̂ = Rθ̂ φ̂θ̂φ̂ = 1
Finally, the Ricci scalar is simply
R = Rââ = η âb̂ Râb̂ = η θ̂θ̂ Rθ̂θ̂ + η φ̂φ̂ Rφ̂φ̂ = 1 + 1 = 2

In summary, we have found that the Ricci scalar for the unit sphere, according to a local observer,
is R = 2 > 0. Since R > 0 we say that the space locally looks like a sphere, as it has positive
curvature. We cannot say that it is exactly a sphere, but does look like one locally. Had we found
R = 0 we would say the space locally looks flat, or if R < 0 we would say that the space locally
looks like a saddle. Therefore, the Ricci scalar characterizes whether the space is positively curved,
negatively curved or flat. This calculation also exemplifies the fact that curvature is also a local
phenomenon.
Let’s move on to a more involved example. Here we will examine a particular form of the
Schwarzschild line element, one which we will become more familiar within chapters to come as it
is the metric which describes the simplest possible black hole. Here we won’t go to the end and
compute the components of the Ricci tensor, we will do that in a later chapter. Here we will only
calculate the components of the Riemann tensor.
We start with the more general form of the Schwarzschild line element:
ds2 = −e2v(r) dt2 + e2λ(r) dr2 + r2 (dθ2 + sin2 θdφ2 ) (17.76)

from which we extract our basis of orthonormal 1-forms
{ω t̂ , ω r̂ , ω θ̂ , ω φ̂ } = {ev(r) dt, eλ(r) dr, rdθ, r sin θdφ}

Using Cartan’s first structure equation we may calculate the Ricci rotation coefficients. Starting
with ω t̂ we have
dω t̂ + Γt̂ r̂ ∧ ω r̂ + Γt̂ θ̂ ∧ ω θ̂ + Γt̂ φ̂ ∧ ω φ̂ = 0
dv(r) v(r)
= e dr ∧ dt + Γt̂ r̂ ∧ ω r̂ + Γt̂ θ̂ ∧ ω θ̂ + Γt̂ φ̂ ∧ ω φ̂ = 0
dr
dv −λ(r) r̂
= e ω ∧ ω t̂ + Γt̂ r̂ ∧ ω r̂ + Γt̂ θ̂ ∧ ω θ̂ + Γt̂ φ̂ ∧ ω φ̂ = 0 (17.77)
dr
For ω r̂ we have
dω r̂ + Γr̂ t̂ ∧ ω t̂ + Γr̂ θ̂ ∧ ω θ̂ + Γr̂ φ̂ ∧ ω φ̂ = 0
dλ(r) λ(r)
= e dr ∧ dr + Γr̂ t̂ ∧ ω t̂ + Γr̂ θ̂ ∧ ω θ̂ + Γr̂ φ̂ ∧ ω φ̂ = Γr̂ t̂ ∧ ω t̂ + Γr̂ θ̂ ∧ ω θ̂ + Γr̂ φ̂ ∧ ω φ̂ = 0 (17.78)
dr
We also have for ω θ̂
dω θ̂ + Γθ̂ t̂ ∧ ω t̂ + Γθ̂ r̂ ∧ ω r̂ + Γθ̂ φ̂ ∧ ω φ̂
e−λ(r) r̂
= ω ∧ ω θ̂ + Γθ̂ t̂ ∧ ω t̂ + Γθ̂ r̂ ∧ ω r̂ + Γθ̂ φ̂ ∧ ω φ̂ = 0 (17.79)
r
Lastly,
dω φ̂ + Γφ̂t̂ ∧ ω t̂ + Γφ̂r̂ ∧ ω r̂ + Γφ̂θ̂ ∧ ω θ̂ = 0
e−λ(r) r̂ cotθ θ̂
= ω ∧ ω φ̂ + ω ∧ ω φ̂ + Γφ̂t̂ ∧ ω t̂ + Γφ̂r̂ ∧ ω r̂ + Γφ̂θ̂ ∧ ω θ̂ = 0 (17.80)
r r
To determine each of the Ricci rotation coefficients we simply compare (17.77), (17.78), (17.79),
and (17.80). By deduction we see from (17.77) that it must be
dv −λ(r) r̂
e ω ∧ ω t̂ + Γt̂ r̂ ∧ ω r̂ = 0
dr
Because this Ricci rotation coefficient is the only one which matches the form of the dω t̂ , leading
us to say that Γt̂ θ̂ = Γt̂ φ̂ = 0 and
dv −λ(r) t̂
Γt̂ r̂ = e ω
dr
It is important to point out that it could have been that Γt̂ θ̂ ∝ ω θ̂ and Γt̂ φ̂ ∝ ω φ̂ , as this would
have still allowed the expression to be satisfied. This is part of the difficulty of the tetrad method, it
requires a little bit of guesswork. Here we guessed that the other Ricci rotation coefficients vanish.
There is physical motivation for making this choice, a motivation which we will be discussed in
detail in a later chapter. For now, simply note that if our guess ends up being bad, the other
expressions will tell us so as some inconsistency will arise.
Moving on to (17.79), using the same logic we see that
e−λ(r) r̂ e−λ(r) θ̂
ω ∧ ω θ̂ + Γθ̂ r̂ ∧ ω r̂ = 0 ⇒ Γθ̂ r̂ = ω
r r
From (17.80) we have
e−λ(r) r̂
ω ∧ ω φ̂ + Γφ̂r̂ ∧ ω r̂ = 0
r
and
cotθ θ̂
ω ∧ ω φ̂ + Γφ̂θ̂ ∧ ω θ̂ = 0
r
yielding
e−λ(r) φ̂ cotθ φ̂
ω Γφ̂θ̂ =
Γφ̂r̂ = ω
r r
Altogether we have that the non-zero Ricci rotation coefficients are
dv −λ(r) t̂ e−λ(r) θ̂ e−λ(r) φ̂ cotθ φ̂

Γt̂ r̂ = e ω Γθ̂ r̂ = ω Γφ̂r̂ = ω Γφ̂θ̂ = ω (17.81)
dr r r r
Using these coefficients, we may calculate the curvature 2-forms Ωâb̂ . By Cartan’s second struc-
ture equation we have
Ωt̂ r̂ = dΓt̂ r̂ + Γt̂ θ̂ ∧ Γθ̂ r̂ + Γt̂ φ̂ ∧ Γφ̂r̂ = dΓt̂ r̂
" 2 2 #
dv −λ(r) v(r) d v dv dλ dv −λ(r) v(r)
=d e e dt = + − e e dr ∧ dt
dr dr2 dr dr dr
" 2 #
d2 v dv dλ dv −2λ(r) r̂
= + − e ω ∧ ω t̂ (17.82)
dr2 dr dr dr
We also have
Ωθ̂ r̂ = dΓθ̂ r̂ + Γθ̂ t̂ ∧ Γt̂ r̂ + Γθ̂ φ̂ ∧ Γφ̂r̂ = dΓθ̂ r̂ + Γθ̂ φ̂ ∧ Γφ̂r̂
cotθ e−λ(r) φ̂ dλ
= d(e−λ(r) dθ) − ω ∧ ω φ̂ = − e−2λ(r) ω r̂ ∧ ω θ̂ (17.83)
r r dr
Next,
e−λ(r)

Ωφ̂r̂ = dΓφ̂r̂ + Γφ̂θ̂ ∧Γ θ̂
r̂ =d r sin θdφ + cos θdφ ∧ e−λ(r) dθ
r
dλ −λ(r)
=− e sin θdr ∧ dφ + cos θe−λ(r) dθ ∧ dφ + cos θe−λ(r) dφ ∧ dθ
dr
dλ e−2λ(r) r̂
=− ω ∧ ω φ̂ (17.84)
dr r
Moreover,
e−2λ(r) φ̂
Ωφ̂θ̂ = dω φ̂θ̂ + Γφ̂r̂ ∧ Γr̂ θ̂ = d(cos θdφ) − ω ∧ ω θ̂
r2
e−2λ(r) φ̂
= − sin θdθ ∧ dφ − ω ∧ ω θ̂
r2
1 φ̂ e−2λ(r) φ̂
2
ω ∧ ω θ̂ −
= ω ∧ ω θ̂ (17.85)
r r2
The last two curvature 2-forms can be easily solved to give
dv e−2λ(r) t̂ dv e−2λ(r) t̂
Ωt̂ θ̂ = − ω ∧ ω θ̂ Ωt̂ φ̂ = − ω ∧ ω φ̂ (17.86)
dr r dr r
We are almost there! Using (17.75) we may calculate the components of the Riemann tensor
using the curvature 2-forms. Starting with Ωt̂ r̂ we have
1 t̂ 1 1 t̂
Ωt̂ r̂ = R r̂r̂t̂ ω r̂ ∧ ω t̂ + Rt̂ r̂t̂r̂ ω t̂ ∧ ω r̂ = R r̂r̂t̂ ω r̂ − Rt̂ r̂t̂r̂ ω t̂ ω r̂ ∧ ω t̂
2 2 2
Since Rt̂ r̂t̂r̂ = −Rt̂ r̂r̂t̂ we find
Ωt̂ r̂ = Rt̂ r̂r̂t̂ ω r̂ ∧ ω t̂

Comparing to (17.82) leads us to conclude
" 2 #
t̂ d2 v dv dλ dv −2λ(r)
R r̂r̂t̂ = + − e
dr2 dr dr dr
Moving right along,
Ωθ̂ r̂ = Rθ̂ r̂θ̂r̂ ω r̂ ∧ ω θ̂

Comparing to (17.83) we find that
dλ e−2λ(r)
Rθ̂ r̂θ̂r̂ =
dr r
Similarly,
Ωφ̂r̂ = Rφ̂r̂φ̂r̂ ω φ̂ ∧ ω r̂
Looking at (17.84) leads us to
dλ e−2λ(r)
Rφ̂r̂φ̂r̂ =
dr r
Using this same procedure we can find all of the other non-vanishing components of the Riemann
tensor. Altogether we have
" 2 #
d2 v

t̂ dv dλ dv −2λ(r)
R r̂r̂ t̂
= + − e (17.87)
dr2 dr dr dr
dλ e−2λ(r) dv e−2λ(r) 1 − e−2λ(r)

Rθ̂ r̂θ̂r̂ = Rφ̂r̂φ̂r̂ = Rt̂ φ̂φ̂t̂ = Rt̂ θ̂θ̂t̂ = Rφ̂θ̂φ̂θ̂ = (17.88)
dr r dr r r2
We have now arrived to a point common in many problems in relativity theory. Notice we have a
few different differential equations, some of which are non-linear. When we study relativity, rather,
Einstein’s field equations, often the equations are non-linear or at least difficult to solve. Also note
that we would not be finished if this were a typical relativity problem. There we would continue
to solve for the components of the Ricci tensor and then also solve for the Ricci scalar. In a later
chapter we will do just this and solve another similar set of differential equations, which will yield
the more familiar form of the Schwarzschild line element
−1
2 2m 2 2m
ds = − 1 − dt + 1 − dr2 + r2 (dθ2 + sin2 θdφ2 ) (17.89)
r r
17.8 Exercises
1. Working in spherical coordinates in a holonomic basis, prove (17.8). (Hint: Start by writing out
each of the basis vectors in spherical coordinates in terms of cartesian basis vectors. Calculate the
derivatives from there.)
2. Using (17.21), show the metric compatible condition, i.e. ∇ρ gµν = 0
3. Calculate the Christoffel symbols of the 2-sphere, which has the line element ds2 = a2 (dθ2 +
sin2 θdφ2 ) where a is some constant radius.
4. Show that a tangent vector to a geodesic satisfies uc ∇c ua = 0. Using this one can also show
that the motion of test particles may be described by the geodesic equation:
pλ ∇λ pµ = 0
5. Prove equation (17.48).
6. (a) Using (17.58), work out all of the Hodge duals of each possible form in spherical coordinates.
(b) Using (17.65), calculate the curl and the divergence of an arbitrary 1-form field in spherical
coordinates. The result should look familiar.
7. Compute the components of the Riemann tensor for the Robertson-Walker metric
a2 (t)
ds2 = −dt2 + dr2 + a2 (t)r2 dθ2 + a2 (t)r2 sin2 θdφ2
1 − kr2
17.8. EXERCISES 441
which describes a homogenous, isotropic, expanding universe. Moreover, k = −1, 0, 1 pertains

to a universe that is open, flat, or closed. As a check, one should find
ȧ2 + k ä
Rθ̂ r̂θ̂r̂ = Rr̂ θ̂r̂θ̂ = Rθ̂ φ̂θ̂φ̂ = Rφ̂r̂φ̂r̂ = Rt̂ θ̂t̂θ̂ = Rt̂ φ̂t̂φ̂ = Rt̂ r̂t̂r̂ =
a2 a
Chapter 18
A Crash Course on General

Relativity
In this chapter we put the mathematics developed in the previous chapter to use. Here we will
derive the Einstein field equations, the basis of general relativity, in two different ways. All in all,
the point of this chapter is to provide a basic background in the tenets of general relativity, which
will help us in later chapters when we examine black holes in the context of relativity and string
theory. Again, just as the other crash course chapters are, this chapter is not meant to provide a
comprehensive overview of the subject, rather provide a minimal, though complete, background in
the subject. For more information of this subject and on some of the topics in this chapter, the
reader is strongly encouraged to review the references at the the end of this text.
18.1 The Energy-Momentum Tensor

Before we get to deriving the Einstein field equations, we must first review one more tensorial object,
one which we are already familiar with, however not in the context of relativity theory. As we will
show, it is the energy-momentum tensor, sometimes called the stress-energy tensor, Tµν which
acts as the source of the gravitational field in general relativity. As noted throughout this text,
the energy-momentum tensor is a symmetric tensor which describes the energy density or spatial
momentum density of a system. Back when we explored the covariant quantization approach of
the bosonic string, we examined the energy-momentum tensor of the world-sheet. A more common
view, including the view taken in relativity, is that the energy-momentum tensor may describe a
fluid.
To explore this more, let us examine each component of the energy-momentum tensor. Consider
a surface defined by a constant xν , then T µν can be interpreted as the flux of the µ component
of the momentum crossing the surface defined by xν [38]. Since we are studying relativity, by
momentum we mean the 4-momentum. Therefore, if µ = t, we would say that T tν describes the
flow of energy across the surface xν . From this viewpoint, we say that the T tt component of the
energy momentum density describes the energy density, or the flow of energy as it crosses a surface
of constant time (xt ≡ t). We denote this component by T tt = ρ. If we have one spatial component
and one time component, T it = T ti ≡ Πi , which we interpret as the momentum per unit volume,
443
444 CHAPTER 18. A CRASH COURSE ON GENERAL RELATIVITY
the momentum density. If we have only spatial components, we intepret T ij as the flux of force
per unit area across a surface with a normal direction given by the unit vector ej . This is what is
often referred to as stress.
As one would hope for, the energy-momentum tensor indeed satisfies a conservation law. In
curved space it takes the form
∇ν T µν = 0 (18.1)
If we were considering a local frame, in which we may approximate our curved space by a flat
space, the conservation law reduces to
∂T µν
=0 (18.2)
∂xν
From here, notice that if we let µ = 0, the time component, the conservation equation in the
local frame becomes
∂T 00 ∂T 0i ∂ρ
+ = +∇·Π=0 (18.3)
∂t ∂xi ∂t
which is simply the conservation of energy we are used to.
In general relativity, there are two different types of fluids which come up often, a perfect fluid
and dust, each of which can be described by the energy-momentum. A perfect fluid is characterized
solely by pressure and energy density (or, equivalently, since we will be considering this in the
context of relativity, mass density), while dust is a perfect fluid with a vanishing pressure. Since
dust is a simpler case to study, let’s examine it first.
Dust can be viewed as a collection of particles which move with some momentum and have some
energy density ρ. To make things easier, let us suppose we are in a co-moving frame alongside
the dust particles. In this dust cloud there is some number of dust particles per unit volume n.
Therefore, in a co-moving frame, we can describe the energy density ρ by
ρ = mn (18.4)
where m is the mass of the dust particles. The only other quantity describing the dust particles
is a velocity, in this case a 4-velocity uµ . What ends up happening is we construct the energy-
momentum tensor such that the components are propotional to the energy density and the 4-
velocity, i.e.
T µν = ρuµ uν (18.5)
If we are in a co-moving frame we know that u = (1, 0, 0, 0), which allows us to write the
energy-momentum tensor as
 
ρ 0 0 0
µν
0 0 0 0
T = 0 0
 (18.6)
0 0
0 0 0 0
Had we assumed we were not in a co-moving frame but rather assumed the role of a stationary
observer, the 4-velocity would be written as u = (γ, γux , γuy , γuz ), in which case one can show that
the energy momentum tensor takes the form [38]
18.1. THE ENERGY-MOMENTUM TENSOR 445
ux uy uz
 
1
 x
2 u (ux )2 ux uy ux uz 
T µν = ργ  y  (18.7)
u uy ux (uy )2 uy uz 
uz uz ux uz uy (uz )2
If we again consider the conservation equation with µ = 0 we find
∂ρ ∂ρux ∂ρuy ∂ρuz ∂ρ

+ + + = + ∇ · (ρ~u) = 0 (18.8)
∂t ∂x ∂y ∂z ∂t
which one may recognize as the continuity equation for a fluid.
Now that we have a handle on dust, let us move on to the slightly more complex case of a perfect
fluid. One can work out that in a local frame, the energy momentum tensor describing a perfect
fluid takes the form [38]
 
ρ 0 0 0
0 P 0 0
T µν =  (18.9)
0 0 P 0
0 0 0 P
where we use P to denote pressure. We can actually construct the energy-momentum tensor
describing a perfect fluid in a general frame by first considering it in flat space. Realizing that the
fluid only depends on the 4-velocity uµ and the Minkowski metric tensor ηµν , we may write that
components of T µν as
T µν = Auµ uν + Bη µν (18.10)
From (18.9), we find that the only spatial components are T ii = P , or T ij = δ ij P . Moreover, in
the rest frame we have that u0 = 1 and all other components vanish. This means (18.10) can be
written as T ij = Bη ij . Moreover, from (18.9) we have that T 00 = ρ, which implies that
ρ = Au0 u0 + B 00 = A − P ⇒ A = ρ + P
Therefore, the general form of the energy-momentum tensor for a perfect fluid in a local frame
is
T µν = (ρ + P )uµ uν + P η µν (18.11)
From here it is rather straightforward to show that for any metric g µν we have for a perfect fluid
T µν = (ρ + P )uµ uν + P g µν (18.12)
It’s important to note that dust and a perfect fluid are not the only fluids that the energy-
momentum tensor can describe. Some who study relativity include viscosity and shear terms to
T µν , however we won’t examine these cases here.
18.2 A Physical Derivation of Einstein’s Field Equations

In this section we will give a more physical derivation of Einstein’s field equations, the method in
which Einstein himself used to arrive to his famed equations. Before we get to gravity as inter-
preted by general relativity, let us first review some facts about Newtonian physics. In Newtonian
mechanics there are three different contexts in which mass shows up, and therefore, in a sense, three
different types of mass. The first is inertial mass which measures the ability for a body to resist
changes in motion, which is related to Newton’s first law. The other two types of mass are related
to gravity, namely passive gravitational mass and active gravitational mass. The first describes the
reaction of a body by some gravitational field φ, while the second is the mass which gives rise to a
gravitational field. As an exercise the reader will show that these masses are all equivalent, a fact
that can be seen by Galileo’s famed Tower of Pisa experiments.
Crucial to the formulation of a correct theory of gravity is this so-called principle of equivalence.
The basic idea of the principle of equivalence is that inertial forces are physically indistinguish-
able from gravitational forces. This notion came from the Einstein elevator thought experiments.
Imagine an observer in an elevator sitting at rest on the earth. If they were to release a ball, the
ball would, as expected, fall at a rate of acceleration equal to g = 9.81m/s2 . Moreover, if the
elevator was in free fall and the observer released the ball once again, the ball would no longer be
subject to any forces, similar to if the observer was in an intertial frame in outer space (far from
any gravitational sources). On the other hand, if an observer is accelerating at the same rate as
g, the ball will drop just as before. In short, the principle of equivalence tells us that the laws of
physics in an accelerated frame are the same as the laws of physics in a uniform gravitational field
[62]. Lastly, using the mathematical framework we developed in the previous chapter, we can say
that what we experience as gravity, is simply the curvature of space-time, which is caused by the
presence of matter and energy in the background space-time.
For a rather simple guess approach to Einstein’s field equations, recall the Poisson equation
describing the Newtonian gravitation potential:
∇2 φ = 4πGρ (18.13)
If we look for a relativistic generalization of this equation, or rather a tensorial equation, we see
may argue that the gravitational potential of Newtonian theory is to be replaced by the metric, as
it is the metric which encodes the information of the gravitational field distribution. Moreover, the
tensorial generalization of the mass density ρ is, as examined in the previous section, the energy-
momentum tensor Tµν . Now notice that the left hand side of the Poisson equation has a second
ordered differential operator acting on the field. This means in our generalization we would guess
a second order differential operator on the metric. Now recall that in the explicit representation of
the Riemann curvature tensor there are second order derivatives on the metric tensor. So a guess
might be
Rµνρσ ∝ Tµν
Immediately we know this isn’t right as the indices between both sides are not balanced. To get
indices to match we contract the Riemann tensor, yielding the Ricci tensor. Therefore our next
guess would be
Rµν = kTµν
18.2. A PHYSICAL DERIVATION OF EINSTEIN’S FIELD EQUATIONS 447
where k is some constant. Remember now that the energy-momentum tensor must obey the
conservation law: ∇µ Tµν = 0. If one recalls the Bianchi identity,
∇λ Rρσµν + ∇ρ Rσλµν + ∇σ Rλρµν = 0

µ
we can easily see that ∇ Rµν 6= 0. Contracting the Bianchi identity twice gives us
0 = g νσ g µλ (∇λ Rρσµν + ∇ρ Rσλµν + ∇σ Rλρµν ) = ∇µ Rρµ − ∇ρ R + ∇ν Rρν

where we used metric compatibility. Rearranging leads us to conclude that
1
∇µ Rρµ = ∇ρ R
2
From here it is easy to see that in order for the conservation of Tµν to be satisfied we must have
1
Rµν − Rgµν = kTµν
2
With a bit of work, one can show that the constant k = 8πG, which allows us to write out the
correct form of Einstein’s field equations:
1
Gµν ≡ Rµν − Rgµν = 8πGTµν (18.14)
2
Sometimes (18.14) will be written another way. Notice that if we contract both sides of the
Einstein equation we find
1 1
g µν (Rµν − Rgµν ) = R − (4)R = −R = 8πGg µν Tµν = 8πGT
2 2
Therefore, another common way to write the Einstein field equations is
1
Rµν = 8πG(Tµν − T gµν ) (18.15)
2
Historically, Einstein thought the above form of his field equations were incorrect as it allowed
for a non-static universe, a belief which he and much of the rest of the scientific community held
on to. To adjust his equations, Einstein added in the cosmological constant Λ, in which case the
field equations take the form
Gµν − Λgµν = 8πGTµν (18.16)

Later, after Hubble discovered that the universe was indeed expanding, Einstein removed the
extra term, referring to it as his greatest blunder. Within the last 20 years however, researchers
have found that the cosmological constant term can indeed be present in the field equations as
it gives a possible model for the vacuum energy, or more famously known as dark energy, of the
universe causing it to accelerate. We explore the presence of the cosmological constant term briefly
later in this chapter.
Before we move on to the alternative derivation of Einstein’s field equations, let us briefly discuss
the structure of these equations. Why is (18.14) even called a set of equations anyway? The
reason is because we are dealing with two indexed tensors, each of which is symmetric, resulting
in ten different equations. What’s more is, by our construction, the Einstein field equations are
second order differential equations for the metric tensor field gµν . As differential equations, the
field equations are particularly difficult to solve, especially since most are non-linear differential
equations. Much research has gone into solving Einstein’s equations with a certain set of parameters
and physically desirable conditions, sometimes resulting in a completely analytical solution, while
others require the most advanced numerical analysis techniques. Later in this chapter we will
explore two solutions to Einstein’s field equations, both of which are arguably the simplest analytical
solutions.
18.3 Lagrangian Formulation of Einstein’s Field Equations

Much of this text has been devoted to using the Lagrangian formalism to derive import quantities
in string theory. We can use this technique again to determine Einstein’s field equations. Though
this method is rather straightforward, it is rather tedious and can therefore be skipped on the first
time read through.
Finding Einstein’s field equations is a matter of finding the correct Lagrangian density and then
use the principle of least action to derive the equations. Typically we have been working flat space
in which we assume that our Lagrangian density L is a function of a set of fields φi and there first
derivatives ∂µ φi , i.e. L = L(φi , ∂µ φi ). To generalize to curved space, all it takes is replacement
of the partial derivative ∂µ with the covariant derivative ∇µ . The action S may then be written
generally as
Z
S = L(φi , ∇µ φi )dn x (18.17)
Recall that typically construct the Lagrangian density such that it is a Lorentz scalar. As written
√
L might not be a Lorentz scalar, and therefore one often writes L as −g L̂ where L̂ is a scalar [17]
and therefore the action is written as
√
Z
S = dn x −g L̂ (18.18)
In general relativity, our dynamical variable is no longer a field φ, but rather the metric tensor
field gµν . The great mathematician Hilbert showed that the simplest possible choice for a La-
grangian depending on the metric that is also a Lorentz scalar is just the Ricci scalar R, yielding
the Einstein-Hilbert action
√
Z
S= −gRdn x (18.19)
As it happens, the Einstein-Hilbert action is the action which yields Einstein’s field equations.
To see this, let us vary the above action, which gives us three integrals:
√ √ √
Z Z Z
δS = dn x −gg µν δRµν + dn x −gRµν δg µν + dn xRδ −g (18.20)
where we have written R = g µν Rµν . Since we the metric tensor g µν is the dynamical variable
here, we seek terms that strictly have the variation δg µν . Thus, we don’t have to do anymore work
to the second integral as this term is already present. Let’s focus on the more difficult term first,
18.3. LAGRANGIAN FORMULATION OF EINSTEIN’S FIELD EQUATIONS 449
the first integral. Recall that the Ricci tensor is simply the contraction of the Riemann tensor,
which we found to be written as
Rρ µλν = ∂λ Γρ νµ + Γρ λσ Γσνµ − ∂ν Γρ λµ − Γρ νσ Γσλµ

Therefore, when we vary the Ricci tensor, we can think of it as varying the Riemann tensor,
which is done by varying the connection via the arbitrary variation
Γρ νµ → Γρ νµ + δΓρ νµ
The variation δΓρ νµ is actually a tensor, allowing us to take its covariant derivative, yielding
∇λ (δΓρ νµ ) = ∂λ δΓρ νµ + Γρ λσ δΓσνµ − Γσλν δΓρ σµ − Γσλµ δΓρ νσ

From here it is straightforward but tedious to show that the variation of the Riemann tensor is
simply
ρ
δRµλν = ∇λ (δΓρ νµ ) − ∇ν (δΓρ λµ ) (18.21)
This allows us to write the first integral in (18.20) as
√
Z
dn x −gg µν ∇λ (δΓλνµ ) − ∇ν (δΓλλµ )

where we have contracted over ρ and λ to get the right Ricci tensor. Moreover, by metric
compatibility and some minor relabeling of indices, the above becomes
√
Z
dn x −g∇σ g µν (δΓσµν ) − g µσ (δΓλλµ )

As written, we can see that we have the covariant divergence of some vector as it is integrated over
a volume element. By Stoke’s theorem then this integral is simply equal to a boundry contribution
out at infinity, which we are free to set equal to zero. Hence, this integral vanishes and does not
contribute to the overall variation of the Einstein Hilbert action [11]. Let’s move on to the third
integral. As one might recall, we have actually computed a variation similar to this one before,
back when we varied the Polyakov action. To calculate this variation we need the identity
ln(detM ) = T r(lnM )
which is true for any square matrix M . Remembering that the variation acts like a derivative,
the variation of this identity is
1
δ(detM ) = T r(M −1 δM ) (18.22)
detM
As an exercise, the reader will show that this is true for 2 × 2 matrices. Letting g = detM and
M = gµν it follows that
δg = g(g µν δgµν )
since g µν is the inverse of gµν . Now recall that g µν gµν = constant. This means that
δ(g µν gµν ) = δg µν gµν + g µν δgµν = 0 ⇒ g µν δgµν = −δg µν gµν

Therefore,
δg = g(g µν δgµν ) = −g(gµν δg µν )

Again using the fact that the variation acts like a derivative we find that
√ 1 1√
δ −g = − √ δg = − −ggµν δg µν (18.23)
2 −g 2
Altogether then, the variation of the Einstein-Hilbert action is
√
Z
1
δS = dn x −g Rµν − Rgµν δg µν (18.24)
2
By setting δS = 0, the only way this happens for arbitrary δg µν is if the integrand itself is zero,
i.e.
1
Rµν − Rgµν = 0 (18.25)
2
What we have derived is Einstein’s field equations in a vacuum. In order to get Einstein’s general
1
field equations, we begin with the action (which has been normalized ad-hoc) S = 16πG R + SM ,
where SM is some action with energy/matter terms. Following the same procedure as before leads
us to
1 1 1 δSM
(Rµν − Rgµν ) + √ =0 (18.26)
16πG 2 −g δg µν
We have seen the last term before. From (10.103), we recognize the last term as the energy-
momentum tensor,
1 δSM
Tµν = −2 √ (18.27)
−g δg µν
which, as hoped for, gives us Einstein’s field equations
1
Rµν − Rgµν = 8πGTµν (18.28)
2
Now that we have derived Einstein’s field equations two different ways, let us examine our first
solution of them.
18.4 The Schwarzschild Solution

Typically, the best way to study a theory is to examine limiting cases from which one can glean
physical insight. Such is the case for German physicist Karl Schwarzschild, who submitted a solution
to Einstein’s field equations within a year of Einstein publishing his theory in 1915. Schwarzschild’s
solution happens to be one of the simplest yet one of the most important, as it is Schwarzschild’s
solution that led to the study of black holes, the subject of the next chapter. Schwarzschild’s
solution consisted of a model that was both time independent and spherically symmetric. Moreover,
18.4. THE SCHWARZSCHILD SOLUTION 451
Schwarzschild also considered a matter free region, in which case we let the energy-momentum tensor
go to zero. We call Einstein’s field equations the vacuum equations when we let Tµν = 0.
In order to obtain a form of the metric that is both time-independent and describes a spherical
symmetric body, we assume that when we are in asymptotically far away regions the metric should
take the form of flat Minkowski space
ds2 = −dt2 + dr2 + r2 (dθ2 + sin2 θdφ2 )

where we have chosen to write the line element in spherical coordinates since we are considering
a spherially symmetric solution. Now, by time-independence, we expect that the line element
shouldn’t change under the exchange dt → −dt, which means that no cross terms with dt exist. In
other words, we may write the line element as [38]
ds2 = gtt dt2 + gij dxi dxj

Similarly, a spherically symmetric metric is one which does not have a preferred angular direction.
This means that if we were to make the changes dθ → −dθ and dφ → −dφ, the form of the line
element shouldn’t change, leaving us with a metric that is completely diagonal, in which the most
general form is
ds2 = −A(r)dt2 + B(r)dr2 + C(r)r2 dθ2 + D(r)r2 sin2 θdφ2

Due to spherical symmetry however, we quickly recognize that it must be the angular terms take
the form of the usual dΩ2 , implying that C(r) = D(r). Moreover,pby a change of radial coordinates
we can eliminate the coefficient C(r). Notice that if we let ρ = C(r)r it follows that
C(r)r2 (dθ2 + sin2 θdφ2 ) = ρ2 (dθ2 + sin2 θdφ2 )

Also note that
√ r dC √ √

r r dC
dρ = √ dC + Cdr = √ + C dr = C + 1 dr
2 C 2 C dr 2C dr
Squaring both sides yields
−2
1 r dC
dr2 = 1+ dρ2
C 2C dr
Now by defining a new constant B 0 to be
−2
0 1 r dC
B = 1+ B
C 2C dr
we find that Bdr2 = B 0 dρ2 , allowing us to write the general line element, after relabeling ρ → r,
as
ds2 = −A(r)dt2 + B 0 dr2 + r2 (dθ2 + sin2 θdφ2 )

Lastly, remember that we want this line element to look like Minkowski space in the large r limit,
having us choose that A(r) = e2v(r) and B 0 = e2λ(r) , which yields
ds2 = −e2v(r) dt2 + e2λ(r) dr2 + r2 (dθ2 + sin2 θdφ2 )

In the last chapter we calculated the components of the Riemann tensor corresponding to this
metric in a local frame. There we found
" 2 #
r̂ d2 v dv dv dλ −2λ(r)
R t̂r̂t̂ = + − e (18.29)
dr2 dr dr dr
1 dv −2λ
Rt̂ θ̂t̂θ̂ = Rt̂ φ̂t̂φ̂ = − e (18.30)
r dr
1 dλ −2λ 1 − e−2λ
Rr̂ θ̂r̂θ̂ = Rr̂ φ̂r̂φ̂ = e Rθ̂ φ̂θ̂φ̂ = (18.31)
r dr r2
Now remember that the Einstein tensor Gµν consists of the Ricci tensor and the Ricci scalar.
Therefore let us compute the components of the Ricci tensor. First of all
Rt̂t̂ = Rt̂ t̂t̂t̂ + Rr̂ t̂r̂t̂ + Rθ̂ t̂θ̂t̂ + Rφ̂t̂φ̂t̂
" 2 #
d2 v

dv dv dλ 2 dv −2λ(r)
= + − + e (18.32)
dr2 dr dr dr r dr
Similarly, using the symmetry Rt̂ r̂t̂r̂ = −Rr̂ t̂r̂t̂ we have

" 2 #
d2 v dv dv dλ 2 dλ −2λ(r)
Rr̂r̂ = − + − − e (18.33)
dr2 dr dr dr r dr
The last two components of the Ricci tensor are
1 dv −2λ 1 dλ −2λ 1 − e−2λ

Rθ̂θ̂ = Rφ̂φ̂ = − e + e + (18.34)
r dr r dr r2
Now remember that we are considering the vacuum equations, in which Tµν = 0 for each com-
ponent. This means that each component of the Ricci tensor must also vanish. Therefore, we set
each component of the Ricci tensor to zero. Notice that if we subtract (18.33) from (18.32) we find
dv dλ
+ =0⇒v+λ=k
dr dr
where k is some constant. If we transform v → v + k (equivalent to a change in the time
coordinate by t → tek ) we have instead [38]: v + λ = 0 ⇒ λ = −v. From here find by substitution
that
2
d2 λ

dλ 2 dλ
−2 + =0
dr2 dr r dr
To solve this differential equation notice that
18.5. THE FRIEDMANN EQUATIONS AND COSMOLOGY 453
2 !
d2 d2 λ

dλ dλ
(re−2λ ) = −4 e−2λ − 2r 2 e−2λ + 4r e −2λ
dr2 dr dr dr
Setting this equal to zero, dividing by e−2λ and −2r gives us the differential equation we found
above. Thus,
2
d2 −2λ d2 λ dλ 2 dλ
2
(re ) = 2
− 2 + =0
dr dr dr r dr

d −2λ −2λ dλ −2λ
(re )= e − 2r e = constant
dr dr
Now notice that by setting Rφ̂φ̂ = 0 we find
1 dv −2λ 1 dλ −2λ 1 − e−2λ

− e + e + =0
r dr r dr r2
Multiplying by r2 and using v = −λ leads to
dλ d
e−2λ − 2r =1⇒ (re−2λ ) = 1
dr dr
Integrating both sides leads to
re−2λ = r − 2m
where −2m is a carefully chosen integration constant. There is a physical reason for choosing
this to be the integration constant, however we will not go into the details here. Rearranging leads
to
2m 2m 2m −1
e−2λ = 1 − ⇒ e2v = 1 − e2λ = (1 − ) (18.35)
r r r
Going all of the way back to the general form of the line element, we arrive to the Schwarzschild
line element:
2m 2 2m −1 2
ds2 = −(1 − )dt + (1 − ) dr + r2 (dθ2 + sin2 θdφ2 ) (18.36)
r r
We will explore this line element in more detail in the next chapter, however we already see that
at r = 2m something strange happens.
18.5 The Friedmann Equations and Cosmology

Here we will examine a solution to Einstein’s equations that are not in vacuum. As an exercise in
the last chapter, the reader should have worked out the components of the Riemann tensor for the
Robertson-Walker line element. The components of the Ricci scalar in a local frame are then
3ä
Rt̂t̂ = Rr̂ t̂r̂t̂ + Rθ̂ t̂θ̂t̂ + Rφ̂t̂φ̂t̂ = − (18.37)
a

ȧ + k ä
Rîî = 2 + (18.38)
a2 a
where Rîî stands for all of the spatial components of the Ricci tensor as they are all equivalent.
The Ricci scalar is simply the contraction of the Ricci tensor, therefore

µ̂ν̂ t̂t̂ r̂r̂ θ̂ θ̂ φ̂φ̂ ä ȧ + k
R=η Rµ̂ν̂ = η Rt̂t̂ + η Rr̂r̂ + η Rθ̂θ̂ + η Rφ̂φ̂ = 6 + (18.39)
a a2
Using Einstein’s field equations in a local frame (the general metric gµν is replaced by the flat
space metric) with cosmological constant
Gµ̂ν̂ + Ληµ̂ν̂ = 8πGTµ̂ν̂ (18.40)
we find two equations separately from the time components and the spatial components, namely
1
Gt̂t̂ + Ληt̂t̂ = Rt̂t̂ + R − Λ = 8πGTt̂t̂ = 8πGρ
2
1
Gîî + Ληîî = Rîî − R + Λ = 8πGTîî = 8πGP
2
where we have used the energy-momentum tensor describing a perfect fluid. These expressions
lead us to the Friedmann equations

ȧ + k 2ä ȧ + k
3 − Λ = 8πGρ + − Λ = −8πGP (18.41)
a2 a a2
Recalling that the Robertson-Walker line element describes a homogeneous, isotropic expanding
universe (where the rate of expansion is given by a(t)), the reason why this description holds
true is because of the Friedmann equations. These equations are often the basis of the study of
comoslogy, at least from a classical general relativistic view. A typical cosmological analysis might
include changing one of the factors in the Friedmann equations (or adding in other terms that we
have not discussed), each of which might give rise to a different model for the universe. Two very
important models are the de Sitter model and anti-de Sitter model. Simply put, de Sitter space
is an analog to Minkowksi space on a sphere with a non-zero cosmological constant. Specifically,
de Sitter space is the maximally symmetric vacuum solution to Einstein’s field equations with a
positive cosmological constant (or a repulsive cosmological constant. By maximmally symmetric we
mean that the curvature of the manifold we are representing our space-time with is the same every
where an in every direction. That is, if one knows the curvature at one point on the space, they
know the curvature everywhere on the space (some examples include Rn and Sn ) [11]. Conversely,
anti-de Sitter space is the maximally symmetric vacuum solution to Einstein’s field equations that
has a negative cosmological constant.
Both of the models deserve a great deal of study, however we will not cover them in detail in
this text. Rather, as an application, let us consider the de Sitter model in the case of flat space
(k = 0, this is actually not a bad choice of the space-time curvature as recent experiments indicate
our universe is almost perfectly flat). The Friedmann equations then become
18.5. THE FRIEDMANN EQUATIONS AND COSMOLOGY 455
Figure 18.1: A graph of the de Sitter model in flat space, k = 0. Notice that the universe accelerates
at an exponential rate.
3ȧ2 ä ȧ
−Λ=0 2 + 2 −Λ=0
a2 a a
The first expression implies
r
ȧ2 Λ 1 da Λ
= ⇒ =
a2 3 a dt 3
Integrating yields
√Λ
a(t) = Ce 3 t
From figure 18.1, we see that in the de Sitter model using the Friedmann equations, we have a
universe which accelerates at an exponential rate.
Before moving on, let us consider one more cosmological model. This time we set Λ = 0, k = 1
and G = 1. Using the Friedmann equations we find
8π B−a
ȧ2 = −1≡
3a a
8π
where B = 3 . Notice then that if we parameterize a(t) by a = B sin2 τ (t) we find that the first
derivative is
2
B − B sin2 τ cos2 τ

da dτ dτ
= 2B sin τ cos τ ⇒ ȧ2 = 4B 2 sin2 τ cos2 τ = =
dτ dt dt B sin2 τ sin2 τ
Rearranging and square rooting both sides leaves us with
2B sin2 τ dτ = dt
This equation allows to parametrically determine the rate of expanson a(t). From our choice of
a(t), we see that at t = 0, τ = 0 and therefore a(0) = 0. In other words, by tracing back to time
zero, the universe begins with the size of a point, a singularity. This is what we know as the Big
Bang. The above is just one model hinting at the Big Bang, however several realistic cosmological
models all have the common origin of the universe happening with a bang. Now since general
relativity does not behave well when quantum effects are introduced, the study of the singularity
at the origin of the universe cannot be done with relativity alone, a quantum theory of gravity is
required.
This concludes our crash course on general relativity. But before we move on to studying black
holes in more depth is important to dwell a little further on what general relativity gives us. In
Newtonian physics, and even special relativity, when we take away all of the dynamical objects, both
particles and fields, what remains is space-time. This is not the case for general relativistic physics.
By taking away all of the dynamical variables, nothing remains since space and time themselves
are dynamical entities. Another way to think about this is that space-time is to be reinterpreted
as a configuration of the gravitational field, meaning that all physical objects, both particles and
fields no longer live in space-time itself, but on top of one another [50].
If we try to apply this logic to quantum field theory, that the quantum fields cannot live on
a background space-time, we are left with assuming that space-time itself is built out of these
quantum fields. That is, space-time itself is to become quantized. This is exactly what happens in
Loop Quantum Gravity, an alternative approach to a quantum theory of gravity. What this means
is that to incorporate quantum theory in a general relativistic theory, we must abandon the notion
of a continuous space-time, unless we are considering a space-time on a large scale, in which we
may approximate the quantum structure of space-time with a continuous manifold, disregarding
quantum non-commutativity.
18.6 Exercises
1. Show that inertial mass, passive gravitational mass, and active gravitational mass are all equiv-
alent. (Hint: To show that intertial mass and passive gravitational mass are the same start with
two forces F1 and F2 , apply Newton’s second law, then the relation to F = −mp ∇φ. To show that
−GmA
passive and active gravitational masses are the same, start with a potential φ1 = r 1 . The force
on m2 is F2 = −mp2 ∇φ1 . Apply Newton’s third law.)
2. Show that (18.22) holds for 2 × 2 matrices.
3. When one studies cosmology, a typical assumption is to include spatial homogeneity and
isotropy, which allows the metric describing an expanding universe to take the general form
ds2 = dt2 − a2 (t)dσ 2
where
dσ 2 = e2f (r) dr2 + r2 (dθ2 + sin2 θdφ2 )

18.6. EXERCISES 457
Find the components of the Ricci tensor, using the Schwarzschild solution as a guide. Ultimately
the task is to determine the expression for e2f (r) , which will allow us to derive the Robertson-Walker
metric:
a2 (t)dr2
ds2 = −dt2 + + r2 (dθ2 + sin2 θdφ2 )
1 − kr2
Hint: The identity Rijkl = k(gik gjl − gil gjk ) proves useful.
4. Early universe models show that it was dominated by radiation. Using Friedmann’s equations,
work out an expression for the expansion factor of a(t) in a radiation dominated universe. In early
universe models one may approximate ρ = 3P . Moreover, for simplicity let Λ = 0.
Chapter 19
Black Holes in General Relativity
In this chapter we explore the basic properties of three different types of black holes in the context
of general relativity. We have already become familiar with the Schwarzschild solution, the time
independent, spherically symmetric solution to Einstein’s vacuum equations. As we will see, the
Schwarzschild solution does in fact describe what we call a black hole. In summary, the chief aim of
this chapter is to attain an intuitive notion of a black hole as well as a technical definition. Lastly,
we will briefly examine the thermodynamics of black holes and realize the problems of black holes
in general relativity.
19.1 The Schwarzschild Black Hole

Before we discuss why we view the Schwarzschild solution as a black hole, let us first consider a few
properties of the solution. In the last chapter we finished the derivation of the Schwarzschild line
element, which we found to be
dr2

2m
ds2 = − 1 − dt2 + + r2 (dθ2 + sin2 θdφ2 ) (19.1)
r 1 − 2m
r
There are three interesting cases to consider for the value of r.Firstly, as r → ∞, we notice that
the above line element approaches the Minkowski line element, that of flat space. This isn’t too
surprising since by construction we sought a solution to Einstein’s equations that was asymptotically
flat. The second interesting case occurs at r = 2m, the Schwarzschild radius. If we were to put
all of the constants that we normally set equal to 1 in general relativity, we would find that the
Schwarzschild radius is written as r = 2Gmc2 . For ordinary stars this radius lies within the interior
of the star, allowing us, up to approximation, to analyze regions of space-time outside of stars
using the Schwarzschild solution. We won’t discuss the physics of stars in this text, however the
Schwarzschild solution is a good place to start in such an analysis.
The third interesting value is at r = 0 and it is related to r = 2m. As one can immediately see,
at the value r = 2m the line element blows up, indicating there is a singularity of some kind in
the metric. Similarly, at r = 0 the line element also blows up, indicating another singularity
in the metric. There is a difference between these two singularities however. As we will see
shortly, although the metric blows up at r = 2m, it does not imply that this singularity is a
459
460 CHAPTER 19. BLACK HOLES IN GENERAL RELATIVITY
physical singularity in which the curvature is infinite. Rather, the singularity at r = 2m is an

artifact of the particular coordinate system we chose to write the line element in, leading to a
coordinate singularity. Alternatively, one can show that there is a physical singularity admitted by
the Schwarzschild solution. Since scalars hold in all coordinate systems, one can calculate [38]
48m2
Rµνρσ Rµνρσ =
r6
which we see has a singularity at r = 0. This means that at r = 0 the Schwarzschild solution
has a physical singularity at which the curvature is infinite.
The Schwarzschild solution has also been used as the theoretical model behind several exper-
imental tests of general relativity, including gravitational redshift, the precession of perihelion of
Mercury, and the deflection of light rays from a celestial body. These topics are discussed exten-
sively in some of the references for this text in which the reader is pointed to as we will not go into
further detail on such subjects in this text. The reader will however work out the geodesics of the
Schwarzschild solution, a starting point for some of the models used in these experimental tests, as
an exercise.
Let’s now consider why the Schwarzschild solution leads to the notion of a black hole. Specifically,
let us consider the behavior of light-cones as they approach r = 2m. For simplicity, let us consider
light-rays which travel on radial lines, i.e. dθ = dφ = 0. The Schwarzschild line element becomes
dr2

2 2m
ds = − 1 − dt2 +
1 − 2m

r r
Since we are considering light-rays, the space-time interval is null, ds2 = 0, allowing us to solve
for the slope of a light-cone:
−1
dt 2m
=± 1− (19.2)
dr r
dt
Notice that as r → ∞, dr → ±1, recovering the notion of light rays in Minkowski space, as
expected. On the other hand, notice that as r → 2m outgoing light rays satisfy
dt r
= →∞
dr r − 2m
meaning that light cones become narrower on approach to r = 2m. Figure 19.1 depicts this
behavior of light-cones. Let us now consider a radially infalling particle from infinity (very far away
from the location r = 2m) with an initial velocity of zero. Since we are considering a real particle,
we expect that it follow a time-like geodesic, which, as the reader will show in an exercise, satisfies
[14]
2 −1 2
2m dt 2m dt 2m dr
1− =k 1− − 1− =1 (19.3)
r dτ r dτ r dτ
19.1. THE SCHWARZSCHILD BLACK HOLE 461
Figure 19.1: Behavior of light-cones as they approach the event horizon [40].
where τ is the proper time along the worldline of the particle. As the reader will also show,
k = 1 corresponds to a particle with an initial velocity of zero. In this case we find that a radially
infalling particle from infinity with a vanishing intial velocity can be described by
2
2m dt dr 2m
1− =1 = (19.4)
r dτ dτ r
Notice then that if we describe the motion of the particle in Schwarzschild time t we find
r
dt dr/dτ 2m 2m
= =− 1−
dr dt/dτ r r
where we have taken the negative square root of ṙ. Integrating the above from an initial time t0
and distance r0 yields
√ √ √ √
2 3 3 √ √ ( r + 2m)( r0 − 2m)
t − t0 = √ r 2 − r0 + 6m( r − r0 + 2mln √
2
√ √ √ (19.5)
3 2m ( r0 + 2m)( r − 2m)
As r → 2m one can work out that

(t−t0 )
r − 2m = 8me− 2m
Therefore t → ∞ as r → 2m. Remember that t is the Schwarzschild time, representing the proper
time of a distant observer, not the proper time of the infalling particle. This means that, according
to a distant observer, as an object as described above approaches the Schwarzschild radius, the
object will continue to get slower and slower the closer it gets to r = 2m. In fact, as seen by an
outside observer, an infalling object will never actually reach r = 2m. This is not the case, however,
from the frame of the falling object. Consider again (19.3) with k = 1, however this time consider
dt
the case for large r. This means we may approximate dτ ≈ 1, or, asymptotically, t ≈ τ . Therefore
we are considering the proper time of infalling particle. Moreover, in this region we still find
p
ṙ = − 2m/r
. Integrating this leads to

2 3 3
τ − τ0 = √ (r02 − r 2 )
3 2m
Based on this expression, we don’t find any indication that there is singular behavior at r = 2m
as the infalling object, in its own frame, falls all of the way to r = 0 in finite time. Simply put,
something strange is going on at r = 2m, and to investigate it we must remove the coordinate
singularity at r = 2m.
19.2 Eddington-Finklestein Coordinates

To examine more of what happens at r = 2m we seek to eliminate the coordinate singularity. To
do this we introduce the so-called tortoise coordinate
r
r∗ = r + 2mln −1 (19.6)
2m
as well as two null coordinates
u = t − r∗ v = t + r∗ (19.7)
Notice then that
2m 1 dr
dr∗ = dr + dr =
1 − 2m

(r/2m − 1) 2m r
Then where dt = dv − dr∗ , it is a simple matter to work out
2dvdr dr2
dt2 = dv 2 − + 2
1 − 2m

r 1 − 2m
r
Subsituting this into the Schwarzschild solution leads to the Schwarzschild line element in
Eddington-Finklestein coordinates

2m
ds2 = − 1 − dv 2 + 2dvdr + r2 (dθ2 + sin2 θdφ2 ) (19.8)
r
Written in this form we notice that the line element no longer exhibits a singularity at r = 2m,
proving that the singularity at r = 2m is indeed a coordinate singularity and not a physical
singularity. Moreover, we still notice that the singularity at r = 0 remains, which makes sense as
this is a physical singularity.
To analyze this line element, let us again consider the case of radial paths taken by light rays.
In this case we find
19.2. EDDINGTON-FINKLESTEIN COORDINATES 463

2m dv 2
− 1− dv 2 + 2dvdr = 0 ⇒ =
1 − 2m

r dr r
Integrating yields
v(r) = 2(r + 2mln(r − 2m)) + const

To analyze this expression, let’s start by considering the region r > 2m. As r increases so does
v(r), which describes the behavior for radial light rays that are outgoing [14]. In the region r < 2m
as r increases v(r) decreases, thereby describing light rays that are in going. What’s more is, in
Eddington-Finklestein coordinates light-cones no longer grow narrower as they approach r = 2m.
Rather the time and radial coordinates reverse their character inside r = 2m, indicating that as
light-cones approach r = 2m, they begin to tilt. Figure 19.2 exemplifies this behavior.
What have we shown? Using Eddington-Finklestein coordinates has allowed us to see that
r = 2m is indeed a coordinate singularity and not an actually singularity of the system. More
intriguing is that the surface at r = 2m is a surface from which nothing can escape. It is certainly
possible, as seen by the figure, light future directed time-like and null curves may cross from r > 2m
to r < 2m safely, but it is not possible to go the other way. Notice that at r = 2m the light-cones
are tipped in such a way that light is stationary, and in the region r < 2m the light-cones are
tipped in such a way that not even light itself can escape out to the region of r > 2m. What this
means physically then is, since even light cannot escape, any event taking place inside the region
r < 2m cannot be seen by an outside observer. An outside observer may send information inside
the black hole, but a return signal could never be reached to the outside world. This boundary is
what is called the event horizon, as it is the boundary of all possible events which can be observed,
in principle, by some outside observer. Put another way, an event horizon is a hypersurface that
separates space-time points that are connected to infinity by time-like paths and from those that
are not.
It is here where we realize the popularized albeit accurate definition of a black hole. A black
hole is a region of space-time where gravity is so strong that not even light can escape. But, as
seen above, this region is characterized by the event horizon rather than the singularity. Therefore,
black holes are defined by their event horizons and the not the singularities that lay behind them.
This is not to say there is nothing interesting about the singularities admitted by black hole solu-
tions. In fact, Stephen Hawking and Roger Penrose worked out their famous singularity theorems,
which essentially say that once the gravitational collapse of star begins, it will inevitably lead to
a singularity, outlining the general properties of black holes and singularities (including the singu-
larity at the Big Bang). Penrose also came up with the cosmic censorship hypothesis which states
that every singularity spawned by a black hole must be shrouded by an event horizon, i.e. naked
singularities cannot exist. Though it has yet to be proven, one can imagine why we would desire
for cosmic censorship to be true. If a naked singularity did exist, then an external observer could
‘see’ the singularity, or, put another way, an external observer could see infinity. This would be the
interpretation in the context of general relativity; with a modified theory of gravity, this might not
be the case. Moreover, one consequence of the cosmic censorship hypothesis is that classical black
holes may never shrink, rather they may only get larger, leading us to Hawking’s Area Theorem
which states that the area of a future event horizon is non-decreasing (including the case where
two or more black holes coalesce). As we will see later on, this theorem no longer holds true in
semi-classical gravity, where we take quantum mechanical effects into account.
Figure 19.2: As light-cones approach the event horizon at r = 2m, they begin to tilt. This is
characterized by Eddington-Finklestein coordinates.
19.3 Rindler Space-Time

We can also examine the region near the event horion of a black hole using another set of coordinates.
To do this we replace the coordinate r with the coordinate ρ which is defined by [56]
Z r Z r − 12 r
p 2m p r
ρ= grr (r0 )dr0 = 1− 0
dr = r(r − 2m) + 2msinh −1
−1 (19.9)
2m 2m r 2m
The line element in terms of proper time can be recast as

2 2m
dτ = 1 − dt2 − dρ2 − r(ρ)2 (dθ2 + sin2 θdφ2 ) (19.10)
r(ρ)
When we are near the horizon second term may be approximated by
r r
r r − 2m p
≈ 2m − 1 = 2m = 2m(r − 2m)
2m 2m
19.3. RINDLER SPACE-TIME 465
p
From which we see that ρ may be approximated ρ ≈ 2 2m(r − 2m), and therefore the line
element near the horizon becomes
2
dt
dτ 2 ≈ ρ2 − dρ2 − r(ρ)2 (dθ2 + sin2 θdφ2 )
4m
To make things suppose we are interested in a region where θ ≈ 0, we may replace the angular
coordinates by the following Cartesian coordinates:
X = 2mθ cos φ Y = 2mθ sin φ (19.11)

To make things neater, by defining w ≡ t/4m we may write the line element near the horizon as
dτ 2 = ρ2 dw2 − dρ2 − dX 2 − dY 2 (19.12)

Lastly, if we further define T = ρsinhw and Z = ρcoshw, we find
dτ 2 = dT 2 − dZ 2 − dX 2 − dY 2 (19.13)
From here we realize that when considering small angular regions that are close to the event
horizon of a large black hole look like regions of Minkowski space-time, i.e. the horizon for a large
black hole is locally indistinguishable from flat space-time. We call this Rindler space-time. Rindler
space-time describes uniformly accelerated observers in Minkowski space. A Rindler observer is an
observer in Rindler space-time that moves along a path of constant acceleration [11]. To see this
schematically, figure 19.3 is helpful.
Figure 19.3: A portrait of Rindler space-time, i.e. Minkowski space in Rindler coordinates. A
Rindler observer is an observer who moves along a path of constant acceleration. Here region I is
a region accesible to a Rindler observer moving in the +x direction.
19.4 Kruskal-Szekeres Coordinates

In our analysis of the Schwarzschild black hole we have only been able to really consider regions of
space-time near the horizon at r = 2m. We have yet to truly examine the interior of a black hole,
to investigate the region of space-time that is r < 2m. To examine the geometry in this region we
use Kruskal-Szekeres coordinates u, v which depend on which region we consider. For r > 2m we
have [38]
r r
r/4m r t r/4m r t
u=e − 1cosh v=e − 1sinh (19.14)
2m 4m 2m 4m
and for r < 2m we instead have
r r
r/4m r t r/4m r t
u=e 1− sinh v=e 1− cosh (19.15)
2m 4m 2m 4m
It’s a simple matter to show that in Kruskal coordinates the line element becomes
32m3 −r/2m
ds2 = e (du2 − dv 2 ) + r2 (dθ2 + sin2 θdφ2 ) (19.16)
r
The picture to go along with these coordinates is given in figure 19.4. From this figure we can
extract some interesting details. First we note that the region inside the event horizon, r < 2m is
characterized by v > |u|. Moreover, we notice the line u = ±v corresponds to t → ±∞. One of the
more outstanding features of the Kruskal-Szekeres black hole is that all radial light-like geodesics
look like straight lines drawn at a 45 degree angle when drawn in the diagram:
2
32m3 −r/2m

2 du
ds = 0 = e (du2 − dv 2 ) ⇒ =1
r dv
Moreover, all of the time-like world lines, paths taken by massive bodies, will at every point have
2
a slope (dv/du) > 1. This means that light-cones in the Kruskal-Szekeres diagram will always be
at 45 degree angles (since c = 1), and therefore look like light-cones in Minkowski space-time.
Perhaps the most interesting feature of the Kruskal-Szekeres diagram is the fact that there are
other regions to the diagram which show some interesting time-reverse symmetry. It is first helpful
to realize that we have the following relations with coordinates u and v:
r
u2 − v 2 = − 1 er/2m (19.17)
2m
from which we find u2 − v 2 → 0 as r → 2m and v 2 − u2 → 1 as r → 0. Let’s now think about
what we have done with the Kruskal-Szekeres coordinates. The transformation from Schwarzschild
coordinates to Kruskal-Szekeres coordinates is initially defined for r > 2m and −∞ < t < ∞. But
notice (19.17) is quadratic in both u and v, and that one value of r will determine two hyper-
surfaces. From the figure, as shown in two dimensions, there are two hyperbolae at r = 0, two
singularities termed the past singularity and future singularity. The future singularity is the one
which an infalling observer would experience and is hence unavoidable. On the other hand, the
past singularity is time-reversed, often called a white hole. A white hole, though a common feature
in science fiction, is typically assumed to be unphysical as it would allow for a visible singularity,
which, according to cosmic censorship, cannot happen.
19.4. KRUSKAL-SZEKERES COORDINATES 467
Figure 19.4: A diagram of the Schwarzschild black hole in Kruskal-Szekeres coordinates.Notice that
region I 0 is the so-called ‘white hole’, and O0 is a ‘parallel universe’, geometrically equivalent to O,
however not physically equivalent.
Even more surprising is that there is another region, O0 which is geometrically identical to
the asymptotically flat exterior Schwarzshild solution in region O. But these two regions, though
geometrically equivalent, are not the same physical space-time. That is, the regions O and O0
represent different physical worlds, parallel worlds if you will. What’s more is these two worlds can
be connected via a complicated topology. Recall that each point in the diagram actually represents
a 2-sphere. Now consider the case when v = 0, which can be thought of as Kruskal-Szekeres time.
At this point the Kruskal diagram can be viewed as being constructed from two asymptotically
flat Schwarzschild space-times joined together at a ’throat’ at r = 2m. As v increases, this same
picture holds however the throat gets narrower, in which the parallel universes are connected at
some region r < 2m. Once v = 1 the throat pinches off completely and the two universes touch
at the singularity r = 0, and as v gets even larger the universes are completely separate from one
another. Figure 19.5 gives a schematic view of this evolution. The connection between the parallel
universes is often called a wormhole or the Einstein-Rosen bridge. Although there is a bridge
connecting the two parallel worlds, one can work out that to make it from one universe to the next
one would have to travel faster than the speed of light, the cosmic speed limit of massive particles.
Nonetheless, wormholes are still being researched as a possible method of interstellar travel, and,
even more provocative, as a method for time travel.
Figure 19.5: A representative picture of the Einstein-Rosen bridge. In order to successfully traverse
a wormhole one must move faster than the speed of light (at least in our classical model) [4]
19.5 Conformal Compactification and Penrose Diagrams

Our analysis has allowed us to consider regions of space-time inside and outside of the event horizon,
but we still don’t know how space-time points at infinity are influenced by the presence of a
Schwarzschild black hole. We would like to consider the global structure of space-time in the
presence of a black hole. As we will see, a process called conformal compactification will allow us to
investigate this structure and lead us to another more elegant view of space-time through a picture
called a Penrose diagram.
Before we consider the case of the Schwarzschild black hole, let us consider the simpler Minkowski
space-time. Back when we discussed the covariant quantization of the bosonic string we introduced
the notion of conformally related metrics. What we saw was we may relate a metric gab with
another metric ḡab via the conformal relation
ḡab = Ω2 gab (19.18)
where Ω is the conformal factor. Now remember what we are trying to do in general: to examine
the global structure of a space-time. The way we do this is to bring in points from infinity to
a finite position; we aim to compactify our space-time so that we may investigate the structure
of the space-time at infinity on a finite diagram. The conformal factor does this precisely, it will
transform our original metric into a metric which has brought in points at infinity. Most of all, as
we will see, the process of conformal compactification will allow us to study the causal structure
of space-time. This is because, as one can show, null geodesics of conformally related metrics are
the same. The null geodesics themselves determine light-cones which in turn defines the causal
structure of a specific space-time.
The basic idea behind bringing in points from infinity is to use coordinate transformations
involving functions like tan−1 x, which takes the infinite interval (−∞, ∞) to the finite interval
(−π/2, π/2). To study this process for Minkowski space, consider Minkowski space in spherical
polar coordinates:
ds2 = −dt2 + dr2 + r2 (dθ2 + sin2 θdφ2 )

19.5. CONFORMAL COMPACTIFICATION AND PENROSE DIAGRAMS 469
where we have −∞ < t < ∞ and 0 ≤ r < ∞. To get our coordinate ranges to go to finite
coordinate ranges, we first introduce the null coordinates
v =t+r u=t−r (19.19)
from which we see that −∞ < u, v < ∞ and by definition u ≤ v. Using these null coordinates,
the Minkowski line element changes to
1 1
ds2 = − (dudv + dvdu) + (v − u)2 (dθ2 + sin2 θdφ2 ) (19.20)
2 4
In this the space we would have that each point is represented by a 2-sphere with a radius of
r = 1/2(v − u). Now, to bring in the points from infinity we define coordinates U, V as
U = tan−1 u V = tan−1 v (19.21)
which now have the range −π/2 < U, V < π/2 and U ≤ V . It is straight forward to show that
the line element in these coordinates changes to
1 2
ds2 = sec U sec2 V −2(dU dV + dV dU ) + sin2 (V − U )(dθ2 + sin2 θdφ2 )

(19.22)
4
Let us now transform back to a time coordinate T = U + V and a radial coordinate R = V − U ,

now with the ranges 0 ≤ R < π and |T | + R < π [17]. The line element can now be written as
ds2 = Ω−2 (T, R)(−dT 2 + dR2 + sin2 R(dθ2 + sin2 θdφ2 )) (19.23)
where
Ω(T, R) = 2 cos U cos V = cos T + cos R (19.24)
2
¯
From here we find that the Minkowski line element is conformally related to the line element ds
where
¯ 2 = Ω2 (T, R)ds2 = −dT 2 + dR2 + sin2 R(dθ2 + sin2 θdφ2 )

ds (19.25)
Figure 19.6: The conformal diagram of the Einstein static universe. [17]
This line element is formally known as the Einstein static universe. One can prove that a
manifold endowed with this line element (and thereby metric) is R × S3 , which is topologically
equivalent to a cylinder. Figure 19.6 presents the conformal diagram of the space. What’s more is
from this diagram we can further subdivide the conformal infinity into a few regions which yield
the causal structure of Minkowski space. These regions are as follows:
i+ (T = π, R = 0) i0 (T = 0, R = π) i− (T = −π, R = 0)
I + (T = π − R, 0 < R < π) I − (T = −π + R, 0 < R < π)
where we call i+ future time-like infinity, i0 spatial infinity, i− past time-like infinity, I + future
null infinity, and I − past null infinity. With these definitions we may draw the Penrose diagram
for Minkowski space-time, as seen in figure 19.7. A number of important features can be extracted
from this diagram. First, all radial null geodesics are at 45 degrees, yielding the expected geometry
of a light-cone in Minkowski space-time. Moreover, all time-like geodesics start at i− and end at
i+ ; all null geodesics begin at I − and end at I − , and all space-like geodesics begin and end at
i0 . It is also important to point out that time-like curves that are not geodesics may also approach
future null infinity as long the curves are ‘asymptotically null’. Why do we say Penrose diagrams
yield the causal structure of space-time? This is because in a Penrose diagram we can determine
whether the light cones (both past and future) from two distinct points will ever intersect, giving
us the ability to find whether two events will be causally related.
19.5. CONFORMAL COMPACTIFICATION AND PENROSE DIAGRAMS 471
Figure 19.7: The Penrose diagram for Minkowski space.[17]
We can now come up with the Penrose diagram for the Schwarzschild black hole using this process
of conformal compactification. Starting with the Schwarzschild line element in Kruskal-Szekeres
coordinates from (19.16), and rewriting it using
r
u0 v 0 = − 1 er/2m (19.26)
2m
yields
16m3 −r/2m
ds2 = − e (du0 dv 0 + dv 0 du0 ) + r2 (dθ2 + sin2 θdφ2 ) (19.27)
r
To bring points in from infinity to a finite position we introduce the coordinates
u0 v0

u00 = tan−1 √ v 00 = tan−1 √ (19.28)
2m 2m
where −π/2 < u00 , v 00 < π/2 and −π < u00 + v 00 < π. From here it is easy to see that, when
considering constant angular coordinates, the above metric is conformally related to Minkowski
space. Moreover, in these coordinates the singularities at r = 0 a connected from time-like infinity
in one region to the time-like infinity in the other region. Figure 19.8 gives the Penrose diagram
for Schwarzschild black hole.
Figure 19.8: The Penrose diagram for the Schwarzschild black hole. Notice that as one passes
through the event horizon at r = 2m their future is to hit the singularity at r = 0; they cannot
escape. Moreover, note that regions I and III are actually separate universes, however we are unable
to communicate with them according to classical general relativity. [4]
19.6 Charged Black Holes

Although the Schwarzschild black hole is of tremendous importance, it is a rather unrealistic type
of black hole. Since it is believed that black holes happen from stars collapsing in on themselves,
it is believed that physical black holes may also carry charge and angular momentum. This notion
is actually captured by the no-hair theorem which states that all classical black holes (those we
consider without quantum effects) can be entirely characterized by their mass, charge, and spin.
In this section we will qualitatively review charged black holes, also known as Reissner–Nordström
black holes, while in the next section we will briefly consider spinning black holes.
The Reissner–Nordström solution is that pertaining to a charged black hole, by charge we mean
both electric and magnetic charge (that is correct, magnetic charge as in magnetic monopoles,
the solution still holds even without magnetic charge however). The approach to the solution is
similar to before with the Schwarzschild solution, i.e. looking for a static, spherically symmetric,
and asymptotic solution, this time to the Einstein-Maxwell field equations
Gµν = 8πTµν (19.29)

This time however the energy-momentum tensor is the Maxwell energy-momentum tensor [14]
1 1
Tµν = (−g ρσ Fµρ Fνσ + gµν Fρσ F ρσ ) (19.30)
4π 4
where Fµν is the usual Maxwell tensor (see chapter 3). We won’t go through the details of
the calculation here as it is quite lengthy. The interested reader should refer to D’Inverno’s text
Introducing Einstein’s Relativity, as it provides a more complete introduction to the derivation for
19.6. CHARGED BLACK HOLES 473
the case without magnetic charges. With both electric and magnetic charges Q and P , one can
work out the Reissner–Nordström to be [11]
−1
2m (Q2 + P 2 ) 2m (Q2 + P 2 )

2 2
ds = − 1 − + dt + 1 − + dr2 + r2 (dθ2 + sin2 θdφ2 )
r r2 r r2
(19.31)
Notice that if Q = P = 0, we recover the Schwarzschild solution. Moreover, we recognize that
the way in which the charge enters the line element is identical to that of a point charge, allowing
us to interpret the Reissner–Nordström solution as describing a charged massive point, albeit one
with very interesting properties. Just as we saw in the Schwarzschild solution, we notice that when
the metric component g rr = 0 we have singularities cropping up in our solution, along with the
usual curvature singularity at r = 0. As one might expect, the singularities correlated with g rr = 0
are not physical singularities but instead give us the location of the event horizon. Looking at this
we find
2m (Q2 + P 2 ) p
∆≡1− + 2
= 0 ⇒ r± = m ± m2 − (Q2 + P 2 ) (19.32)
r r
We immediately see that there are three cases for the location of the event horizon, and therefore,
in a sense, three possible types of static, charged black hole. First consider the case when m2 <
Q2 + P 2 , then ∆ > 0. In this case the time coordinate t is always time-like, while r is always space-
like. What is most interesting about this case is that the singularity at r = 0 is now time-like,
which means that the singularity is not hidden by an event horizon, yielding a naked singularity.
As noted by cosmic censorship, this case is believed to be related to a non-physical black hole.
Moreover, the naked singularity arises from the condition that m2 < Q2 + P 2 which suggests that
the total energy is less than the contribution from the electromagnetic fields alone. This would
mean that the mass of the matter carrying charge is negative, which is again unphysical. So even
without cosmic censorship, we have an unrealistic case. If one works out the Penrose diagram for
this case, they would find a diagram similar to that of Minkowski space, except now an observable
singularity at r = 0.
The second case of interest is when m2 > Q2 + P 2 is one that is physical. We see that ∆ is
positive at both large r and small r, but negative in between the two points r± , each of which has
a corresponding coordinate singularity. Therefore in this case we find two event horizons, each at
r = r± . The two event horizons adds in some interesting possibilities for an infalling observer. For
an infalling observer, the approach is at first similar to the approach of an observer falling into
a Schwarzschild black hole, as r = r− is just like the Schwarzschild radius. At r = r+ the time
the coordinate r switches from being space-like to time-like as the observer continues to fall in the
direction of decreasing r. Continuing in this direction, the infalling observer eventually hits r = r−
where r switches back to being space-like, and r = 0 is now time-like. This means though that the
infalling observer no longer has to hit the singularity since r = 0 is not necessarily in the observer’s
future. Rather, the observer has two options, continue to the singularity at r = 0 and die, or move
on to r = r− , and then to r = r+ . At this point the observer can decide to go back into a black hole
(this time a different one before), or continue on only to live in another part of the universe. To
track an infalling observer’s path, see figure 19.9, which gives the Penrose diagram for this scenario.
Figure 19.9: The Penrose diagram for the charged black hole with m2 > Q2 + P 2 . [26]
The final case we consider is when m2 = Q2 + P 2 . This solution is known as the extreme
Reissner–Nordström solution. As we will see in the next chapter, this case is of particular interest
in string theory. One can work out that the singularity at r = 0 is time-like and is not shrouded by
an event horizon, therefore yielding another naked singularity. This time however we cannot throw
away this solution as being unphysical on grounds other than the cosmic censorship hypothesis.
The extremal black hole is also an unstable solution as perturbing the mass only slightly would
result in one of the other two cases.
19.7 Rotating Black Holes

Since stars rotate, it is assumed that the collapse of a spinning star will yield a rotating black hole.
Therefore, the most realistic type of black hole is a charged, spinning black hole, which is formally
known as the Kerr-Newman metric. Here we will make things simpler by assuming there is no
charge, leaving us with just the Kerr metric. Unlike the Schwarzschild and Reissner–Nordström
black hole solutions, the Kerr black hole solution is rather difficult to derive (especially since the
Kerr solution was found in 1963). D’Inverno, using the method of null tetrads, outlines one possible
route of derivation, however we will not cover the tedious derivation here. After much work, one
finds that one form of the Kerr line element is [11]
19.7. ROTATING BLACK HOLES 475
2mar sin2 θ ρ2 2

2 2mr
ds = − 1 − 2 dt2 − (dtdφ + dφdt) + dr
ρ ρ2 ∆
sin2 θ 2
+ρ2 dθ2 + (r + a2 )2 − a2 ∆ sin2 θ dφ2

(19.33)
ρ2
where
∆(r) = r2 − 2mr + a2 ρ2 (r, θ) = r2 + a2 cos2 θ (19.34)

One can also work out that a = J/m, the angular momentum per unit mass. Once again, we
can find the positions of the event horizons by considering when g rr = 0, yielding
∆
= 0 ⇒ ∆ = r2 − 2mr + a2 = 0
ρ2
We again have three possibilities: m > a, m = a, and m < a. When m = a we obtain another
extremal black hole that is unstable, and when m < a we arrive to another unphysical naked
singularity. These results don’t reveal anything too interesting so we will only focus on the case
when m > a. We still have two event horizons given at the positions
p
r± = m ± m2 − a2 (19.35)
The Kerr black hole is said to be stationary because it always rotates in the same direction, but
it is not a static solution since it is rotating in the first place. Moreover, one can work out the norm
of the Killing vectors associated with this metric to be [11]
1
K µ Kµ = − (∆ − a2 sin2 θ)
ρ2
This means that at the outer horizon, r = r+ , ∆ = 0 we find
a2
K µ Kµ = sin2 θ ≥ 0
ρ2
which means that the Killing vector is space-like at the outer horizon unless positioned at the
north and south poles (θ = 0, π), in which case it is null. We call the set of points where K µ Kµ = 0
the stationary limit surface or the ergosurface, whose name will become clear momentarily. The
stationary surface is given by set of points satisfying
(r − m)2 = m2 − a2 cos2 θ
while the outer event horizon is given by the set of points satisfying
(r+ − m)2 = m2 − a2
It follows then that there is a region between the stationary limit surface and the outer event
horizon called the ergosphere. Figure 19.10 exemplifies the horizon structure of the Kerr black hole.
Inside this region, an observer must move in the direction of rotation of the black hole (φ direction),
which we now realize the name ‘stationary limit surface’. As an exercise the reader will show that
the ergosphere is a region which can cause frame dragging.
Figure 19.10: A depiction of the ergosphere of a Kerr black hole.
The curvature singularity at ρ = 0 is also an interesting feature of the Kerr black hole. Since
ρ2 = r2 + a2 cos2 θ, it can only vanish when both quantities are zero or when r = 0 and θ = π/2.
Now since r = 0 is not a point in space but rather a disk, the set of points corresponding to r = 0
and θ = π/2 is actually the ring at the edge of this disk. Therefore, the solid singularity in the
Schwarzschild solution is spread out over a ring due to the rotating nature of the black hole. One
can show that an observer going into the ring can actually exit into a different asymptotically
flat space-time. Perhaps even more interesting is in the region close to the ring singularity. If we
consider paths that go around in φ, while keeping t and θ constant and r a small negative value,
we may approximate the line element by

2m
ds2 ≈ a2 1 + dφ2
r
These paths are actually closed, and therefore represent closed time-like curves, meaning that
one could, in principle, interact with their past selves, and even try to start a paradox. Before all
of the eager science fiction writers jump out of their seat, it should be mentioned that although the
Kerr solution exhibits several interesting properties, it is unlikely that the gravitational collapse of
a star will exactly lead to this bizarre space-times. Nonetheless, it is fun to imagine.
19.8 The Unruh and Hawking Effects

We now turn from classical general relativity and move on to considering semi-classical general
relativity, where we introduce quantum field theory in a gravitational background. Semi-classical
gravity is a subject in itself which requires much study, revealing that in order to reconcile both
quantum mechanics and general relativity, one needs a quantum theory of gravity. There numerous
good texts on the subject of quantum field theory in curved space-time, the standard being Birrell
19.8. THE UNRUH AND HAWKING EFFECTS 477
and Davies, although a thorough introduction aimed for students with a minimal background in
relativity is presented by Mukhanov’s and Winitzki’s Quantum Effects in Gravity. The interested
reader should refer to these texts for more details on this subject.
In this section we aim to provide a qualitative view of Hawking radiation, which in summary
states that black holes can radiate. But if black holes radiate, it would follow that black holes have
a temperature, casting aside the no-hair conjecture from classical relativity. Before we get to the
details and consequences of this result, let us first consider the analogous problem in flat space-time
known as the Unruh effect. In short, the Unruh effect states that an accelerating observer in a
traditional Minkowski vacuum state will observe a thermal spectrum of particles. The basic idea
is rather simple, lying in the fact that observers with different notions of positive and negative
frequency modes of creation and annihilation operators will observe different particle content in
any given state. How one goes about proving the Unruh effect is to start with a Rindler observer
(uniformly accelerating observer in Minkowski space) and then work through to find the modes
related to a Minkowski observer and a Rindler observer. The result is that there is a ’vacuum’
state as seen by each observer, and the observer moving with uniform acceleration through the
Minkowski vacuum will observe a thermal specturm of particles at a temperature T of [11]
a
T = (19.36)
2π
Alternatively, inertial observers would observe the vacuum state with no thermal spectrum of
particles. The reason why Rindler observers can detect particles is a subtle one. In order to
detect particles in the first place, one must have a particle detector with them. If the detector is
then maintained at a constant acceleration (necessary to stay with the accelerating observer), the
observer must do work to keep the detector accelerating. An inertial observer would see that when
the detector detects a particle, it did not actually come from the background energy-momentum
tensor but instead from the energy put into the detector to maintain its acceleration. All in all, the
Unruh effect explains that the notion of ‘vacuum’ and ‘particles’ are observer dependent.
A similar derivation can be done to prove that black holes radiate, known as the Hawking effect.
We will avoid the complicated derivation and instead use the slick proof given in Carroll. First, let
us consider a Schwarzschild black hole. From the Killing vectors of associated with any metric, one
can define a quantity called the surface gravity κ given by
p
κ=Va= ∇µ V ∇µ V (19.37)
where V is the magnitude of the Killing field
p
V = −Kµ K µ (19.38)
and a is the magnitude of the 4-acceleration aµ = U σ ∇σ U µ , which also equals aµ = ∇µ lnV
aµ aµ = V −1 ∇µ V ∇µ V
p p
a= (19.39)
In the case of the Schwarzschild line element, the V associated with the time-like Killing field
K t is
s s
p p 2m 2m
V = −Kt K t = gtt (−K t K t ) = (−1)(−1) 1 − = 1−
r r
where we used the fact that in a static, asymptotically flat space-time the time translational
Killing vector K = ∂t can be normalized by setting Kµ K µ (r → ∞) = −1 [11]. From here one can
work out the magnitude of the 4-acceleration a to be
−1/2 r !
2m d 2m m m
aµ = ∇µ lnV = 1− 1− ∇µ r = 2m
∇µ r ⇒ a = 2m

r dr r r2 1− r r2 1− r
It follows then that the surface gravity κ is simply

m
κ=Va= (19.40)
r2
Moreover, the surface gravity at the position of the event horizon of a Schwarzschild black hole
r = 2m is κ = 1/4m. With the surface gravity in hand, let us consider a static observer at a radius
r1 > 2m outside of a Schwarzschild black hole. As seen from our expression for a, we find that
an observer close but still outside of the event horizon will have an acceleration that a1 1/2m.
It follows then that since the length and time scales in this region may be set by a−1
1 , space-time
looks almost flat. We then make the critical assumption that the quantum state in consideration
looks like the Minkowski vacuum. It follows then from the Unruh effect that the static observer
looks just like a constant-acceleration observer in flat space-time, detecting Unruh radiation at a
temperature T1 = a1 /2π.
Let us now consider a second observer, but this time one that is very far away from the event
horizon at r = 2m. Now the curvature of the black hole will have an effect on the observed radiation
that will render the Unruh effect useless. However, the radiation near the horizon will propagate
to the distant observer with some gravitational redshift. One can show that this thermal radiation
has a redshifted temperature of
V1 V1 a1
T2 = T1 =
V2 V2 2π
If we take the limit that the distant observer is at infinity, it follows from (19.38) that V2 → 1,
and therefore the observed temperature is
V1 a1 κ
T = lim = (19.41)
r1 →2m 2π 2π
where in the present case κ = 1/4m. In summary, observers far from a black hole will see a flux of
thermal radiation emitted from the black hole at a temperature proportional to the surface gravity.
This result is known as the Hawking effect, and the radiation emitted is known as Hawking radiation.
It is important to point out that the Hawking effect does not just hold for the Schwarzschild black
hole, but holds in general. For a complete and detailed derivation of both the Unruh and Hawking
effects, see Mukhanov and Winitzki.
The standard picture of Hawking radiation is typically done with Feynman diagrams. From
our brief study of quantum field theory, we know that vacuum fluctuations can lead the creation
of virtual pairs of particles, a particle and antiparticle pair. In most cases, virtual pairs always
annihilate; the particle is always annihilated by its antiparticle partner. But let us suppose the
case that these vacuum fluctuations happen near an event horizon. Occasionally one member of
the virtual pair will fall into the black hole, while its partner escapes off to infinity. It is the escape
19.9. BLACK HOLE THERMODYNAMICS AND BEYOND 479
of these virtual particles that an external observer would see as Hawking radiation. Figure 19.11
depicts this scenario using an electron-positron pair.
Figure 19.11: The standard picture of Hawking radiation drawn with Feynman diagrams. Notice
that some particles part of a virtual pair may be lost in the event horizon, while the partner retreats
to infinity.
19.9 Black Hole Thermodynamics and Beyond

With an expression of the temperature of a black hole in hand, let us compute its corresponding
entropy. For simplicity, let us again consider the case of a Schwarzschild black hole, which has a
temperature of
1
T = (19.42)
8πm
Using units where c = 1, we have that the total energy is E = m. Then, by the first law of
thermodynamics dE = T dS we find
1
dm = dS
8πm
Integrating yields to the entropy
S = 4πm2
The Schwarzschild radius is r = 2m, which means that the surface area of the black hole is
4π(4m2 ). Using this allows us to write the entropy of a black hole in general as
A
SBH = (19.43)
4
The above is an exact expression for the entropy of a black hole. This means we may apply
it to any of the black hole solutions we have discussed. For example, in the case of the extremal
Reissner–Nordström black hole the entropy is calculated by
2
A 4πr+
S= = = πm2
4 4
which is proportional to the entropy of the Schwarzschild black hole. This relation similarly
holds for the Kerr solution.
It is important to point out that before Hawking derived his famous result, a young student
of John Wheeler, Jacob Bekenstein, first pondered the potential violation of the second law of
thermodynamics. His argument went something as follows. Suppose we have some observer orbiting
a black hole and they drop in a container with some amount of entropy into the black hole. This
would mean that the black hole of the exterior world would decrease. Moreover, according to
classical general relativity, an exterior observer uses the no-hair conjecture to reason that the
entropy on the interior of the black hole, after settling down to equilibrium, may have vanished. In
short, there is no way an external observer can determine the interior entropy of the black hole. It
is possible to conceive then that the black hole has no interior entropy, which means that, since the
exterior entropy was lowered, the entropy of entire universe was lowered, violating the second law
of thermodynamics. Bekenstein’s thought experiment seemed so outrageous that Hawking himself
didn’t believe him. It wasn’t until 1973 when Hawking published his famous result, finding that
black holes indeed have entropy and can actually radiate, contrary to the classical belief that black
holes are only characterized by mass, charge, and spin.
It is reasonable to question, since according to semi-classical gravity black holes have a temper-
ature and associated entropy, whether one can come up with thermodynamic laws for black holes.
This is in fact possible to do, and it results in the laws of black hole mechanics, which are cast
in direct analogy with the familiar four laws of thermodynamics. Below we list the four laws of
thermodynamics and the black hole analog:
Zeroth Law: In thermodynamics the zeroth law states that a system that is in thermal equi-
librium has a constant temperature throughout the system. The analogous statement in black hole
mechanics is that stationary black holes have constant surface gravity on the entire horizon.
First Law: The first law of thermodynamics is given by the identity
dE = T dS − pdV
If we make the identifications E = m, T = κ/2π, and S = A/4, one can work out the first law
of black hole mechanics [11]
19.9. BLACK HOLE THERMODYNAMICS AND BEYOND 481
κ
dm = dA + ΩH dJ (19.44)
8πm
where J = ma is the angular momentum, and ΩH is the angular velocity of the horizon. In
the case of a non-rotating black hole the above expression simplifies to what we used above in
computing the entropy of the Schwarzschild black hole.
Second Law: The second law of thermodynamics is a statement that the entropy of the entire
system and surroundings never decrease. In black hole mechanics, the analogous statement is that
the area of a horizon may never decrease, δ ≥ 0. Moreover, from Hawking’s area theorem (if two
black holes coalesce, the area of the final event horizon is greater than the sum of the areas of
the initial horizons), we see that the second law of black hole mechanics is slightly stronger than
the second law of thermodynamics. In thermodynamics it is possible for one to transfer entropy
from one system to another (only the total entropy cannot descrease). This is not the case for
black holes; one cannot transfer the area from one black hole to another since black holes cannot
bifurcate. Taking into account the entropy of black holes, Bekenstein has proposed the generalized
second law of thermodynamics
δStotal = δSmatter + δSBH ≥ 0 (19.45)

As an exercise, the reader will show that this result holds for a simple example.
Third Law: The third law of thermodynamics states that it is impossible to achieve absolute
zero, T = 0, for any physical process, or the entropy must go to zero as the temperature goes
to zero. For black holes there isn’t a perfect analogy because one can have the surface gravity
go to zero (this is the case for extremal black holes), which don’t have vanishing area. This isn’t
too crucial as the third law in thermodynamics doesn’t work perfectly either since some physical
systems can be shown to violate it.
It is also important to point out that black holes cannot be in a stable equilibrium with an
infinite heat reservoir, contrary to ordinary thermodynamical systems. The reason is because black
holes have a negative heat capacity
∂E ∂ ∂ 1
CBH = = m= (8πT )−1 = − <0 (19.46)
∂T ∂T ∂T 8πT 2
This means that black holes become colder when they absorb energy. A black hole in contact
with an infinite thermal reservoir with the temperature T < TBH would emit radiation and become
hotter. Alternatively, a black hole placed in a reservoir with T > TBH will absorb radiation and
get colder [40].
Hawking’s work didn’t just lead physicists to black hole thermodynamics, but to another far
reaching consequence. Hawking found that not only do black holes radiate, but they also evaporate!
As Hawking radiation leaves the black hole, the black hole loses energy, and therefore shrinks in
mass (using the standard picture of virtual pairs, one can imagine that the infalling virtual particle
adds negative energy to the black hole). As the mass shrinks, the surface gravity increases which in
turn increases the temperature. An increase in temperature only increases the emission of thermal
radiation, and we find a runaway process resulting in the total evaporation of the black hole in a
finite amount of time (albeit a very long time). Black hole evaporation causes a serious problem for
quantum mechanics however. Suppose we were to assemble two different quantum states, collapse
them into black holes of the same type, we would find that the black holes evaporate into two
indistinguishable clouds of Hawking radiation. In short, we have lost information to what the initial
state of each black hole was, leading to the information loss paradox.How does this violate quantum
mechanics? Ordinary quantum mechanics is governed by unitary evolution which basically states
that the information required to specify a state at early times is equal to that needed to specify a
state at later times. A more precise way to think about it is that the radiation the black hole emits
has no correlations and is therefore in a mixed state as characterized by a density matrix. But
a black hole can be constructed from a pure quantum state. Therefore, Hawking suggested that
pure states could evolve into mixed states, violating unitary evolution. Somehow when we mix the
quantum field theory and general relativity unitarity is violated. Much recent, and not so recent,
research has been devoted to find a solution to this problem. As we will see in the next chapter,
string theory offers some insight into resolving this paradox. For a qualitative overview of finding
such a solution, the reader is urged to consider Susskind’s popular science book The Black Hole
War.
19.10 Exercises
1. Here the reader will find the geodesics in Schwarzschild space-time, an important starting place
for some of the models used in experimental tests of general relativity. One way to find the geodesics
is to take the variation of ds:
Z " −1 #
ds2
Z Z
2m 2 2m
δ ds = δ ds = δ − 1− ṫ + 1 − ṙ2 + r2 (θ̇2 + sin2 θφ̇2 )
ds2 r r
where ṫ = dt/ds and so forth. We then define the Lagrangian L as

−1
2m 2 2m
L=− 1− ṫ + 1 − ṙ2 + r2 (θ̇2 + sin2 θφ̇2 )
r r
Using the Euler-Lagrange equations

d ∂L ∂L
− =0
ds ∂ ẋa ∂xa
one can find the geodesic equations.
2. Show that (19.3) holds for a radially infalling particle. Reason that k = 1 corresponds to
particle having a vanishing initial velocity at r → ∞.
3. Consider the classical argument for a black hole. Start with a particle with both kinetic energy
from the particles motion, and a potential energy felt from the black hole. Define the escape velocity
to be the velocity of the object at the surface which just enables the particle to escape to infinity.
Solve for this escape velocity. Moreover, consider the case for light and solve for the Schwarzschild
radius. (Do not set G or c to 1).
4. Here the reader will show that the rotational nature of the Kerr black hole leads to the notion
of frame dragging. Consider a photon emitted in the φ direction at some radius r in the equatorial
plane (θ = π/2) of a Kerr black hole. The instant the photon is emitted its momentum has no
components in the r or θ directions, and therefore satisfies
ds2 = 0 = gtt dt2 + gtφ (dtdφ + dφdt) + gφφ dφ2

Solve for dφ/dt and evalute this quantity at the stationary surface. Discuss the meaning of the
two solutions. What does this for mean massive particles in the ergosphere?
5. Compute the exact entropy S and T of a charged black hole. Take the limit of an extremal
black hole. Show that this implies the third law of black hole mechanics does not hold in all cases.
6. The aim of this exercise is to show that generalized second law of thermodynamics holds. The
example we choose is one that can be found in Bekenstein, 1973. Consider a harmonic oscillator
composed of two particles each with rest mass m/2 and connected by spring with a spring constant
K. Let us imagine that this oscillator is enclosed in a spherical box maintained at some temperature
T . For simplicity we assume that the harmonic oscillator is non-relativistic (T m). Let ω be the
vibrational frequency of the oscillator. Let kb = 1.
(a) Using the standard definition of the probabilty of a system in its nth state
1 −βEn
pn = e
Z
P
where Z is the partition function, show that the average energy hEi = n En pn can also be
written as
∂
hEi = T 2 (lnZ)
∂T
Calculate the average energy of the harmonic oscillator.
(b) Starting from the Shannon definition of entropy

X
S=− pn lnpn
n
Show that
∂
S= (T lnZ)
∂T
Derive the entropy of the harmonic oscillator.
(c) Now suppose the spherical box goes down to a Kerr black hole. As Bekenstein worked out,
for a spherical particle of rest mass µ and proper radius b, the minimum increase in rationalized
area (∆α)min is
(∆α)min = 2µb
This expression gives the minimum possible increase in black hole area that results if a given
particle is added to a Kerr black hole. Bekenstein also showed that the entropy of a black hole is
given by SBH = (1/2)ln2~−1 α. Using these two results, argue that
δSBH ≥ µb~−1 ln2

where b is the outer radius of the box, and µ is the total rest mass of the box. Since b must be
at least as large as half of the mean value hyi of the separation of the two masses y, and, by the
virial theorem, 1/2hEi = 1/2K(∆y)2 , also show that
b > hEi1/2 m−1/2 ω −1

(Hint: Note that hyi > ∆y and don’t forget about the reduced mass of the oscillator.)
(d) Since the box itself must have mass, notice that µ > m + hEi, show that
δSBH > hEi1/2 m−1/2 (~ω)−1 (m + hEi)ln2
(e) Since the only contribution to the entropy in the box is the one found in part (a), one can
reason that the change in entropy corresponding to matter (sometimes called the common entropy)
is given by δSspring = −Sspring . Putting everything together, show that

−1/2 1 −1
δ(SBH + Sspring ) > ξ (1 + ξ) x
+ (e − 1) ln2 − x(ex − 1)−1 + ln(1 − e−x )
2
where we have introduced x = ~ω/T and ξ = mhEi−1 .
(f ) Finally, show that the generalized second law holds true, i.e. δ(SBH + Sspring ) > 0. To do
this find the minimum of the expression found in part (e) and show that it is positive for all x and
all ξ. (Hint: Recall that the oscillator is non-relativistic, and therefore ξ 1. This will mean that
the minimum position xm of the solution to part (e) satisfies xm 2ln2.)
Chapter 20
Black Holes in String Theory
In this chapter we aim to resolve some of the issues that were brought up in the last chapter. We
found that black holes emit thermal radiation and thereby had an associated entropy, yielding us
the Bekenstein-Hawking entropy formula. This was motivated as an analog to thermodynamics
and hence the entropy we found was a thermodynamic identity. However, as noted before, the
more pragmatic view of entropy is from the microcanonical ensemble using Boltzmann’s formula
for entropy. Hence, we aim to find a statistical derivation of the entropy of a black hole, and check
to see if it matches the Hawking formula. This in fact is the case, as Strominger and Vafa have
shown initially, and many others since. The second issue we wish to resolve is the information
paradox. This problem is still being researched, however many physicists agree (including Hawking
himself) that unitarity cannot be violated, though the way in which it is not violated is an open
question.
In a sense this chapter is incomplete. Not because the information is not available, but rather
because of the difficulty of the material. To fully understand the calculations involving the entropy
of black holes, both from a thermodynamic and statistical mechanics standpoint, one should be well
versed in supergravity, a topic which goes beyond the scope of this book. Similarly, the resolution
of the information paradox requires a bit of quantum information theory, which we will not focus on
here. Therefore, in this chapter we choose to provide a more qualitative analysis on both of these
issues, and will present them as such (mostly). At times we will motivate some of the derivations,
choosing to not go through all of the messy details, while at other times we will simply quote well
known results. Nonetheless, the results, and perceived consequences of the results, remain to be
intriguing. For more details on black holes in string theory, the reader is pointed to the references,
specifically Becker, Becker, and Schwarz, and Susskind and Lindesay. Before we get to discuss these
two outstanding problems however, let us first briefly explore the higher dimensional generalization
of specific black hole solutions.
20.1 Black Holes in Higher Dimensions

The study of black holes in higher dimensions is almost an entire field in itself. Here we will
only consider what we need, avoiding the mess of more complicated solutions. Let us start with
the simplest of black holes. When we solved the Schwarzschild solution in space-time dimension
D = 4, we found it by solving Einstein’s vacuum equations, with some underlying assumptions,
485
486 CHAPTER 20. BLACK HOLES IN STRING THEORY
most notably spherical symmetry. If attempt to generalize the Schwarzschild solution to higher
dimensions, we must be careful when we say ‘spherically symmetry’. That is, we must choose an
appropriate symmetry for the space-time metric, which exhibits the same notion of ‘spherically
symmetric’, although picturing it as such is an entire task in itself. Let us write the Schwarzschild
line element using the notation common to the literature:
ds2 = −e2Φ(r) dt2 + e2∆(r) dr2 + r2 dΩ22 (20.1)

where we have used the common definition
dΩ22 = dθ2 + sin2 θdφ2 (20.2)

which we recognize as the line element of a unit 2-sphere. In a sense, all of the spherical symmetry
is wrapped up in this portion of the line element, and therefore when we generalize the Schwarzschild
line element to arbitrary dimension, all we really need to change is what type sphere we consider in
higher dimensions. This isn’t exactly right as we do have to generalize the Schwarzschild radius (the
position of the horizon) to higher dimensions since it includes constants such as the gravitational
constant, which we know has a higher dimensional generalization. Nonetheless, to extend a static
spherically symmetric geometry to an arbitrary dimension D = d + 1, the metric takes the assumed
form [56]
ds2 = −e2Φ dt2 + e2∆ dr2 + r2 dΩ2D−2
= −e2Φ dt2 + e2∆ dr2 + r2 (dθ12 + sin2 θ1 dθ22 + ... + sin2 θ1 ... sin2 θd−2 dθd−1
2
) (20.3)
We won’t go through the gory details here, but one way of realizing the ‘spherical’ part of the
line element takes the above form is to start with cartesian coordinates xi in an n-dimensional
Euclidean space where
x1 = r cos θ1 x2 = r sin θ1 cos θ2 ...xn−1 = r sin θ1 ... sin θn−2 cos θn−1 xn = r sin θ1 ... sin θn−2 sin θn−1
with
r2 = x2n + x2n−1 + ... + x22 + x21

and θ1 , θ2 , ...θn−2 ∈ [0, π], θn−1 ∈ [0, 2π].
Moving along, with the above form of the line element, using tetrad methods one can show that
Einstein’s equation Gr̂ r̂ takes the form
dΦ e−2∆ dΦ d∆ e−2∆

r̂ (D − 2)(D − 3) −2∆
G r̂ = − −(D − 2) + (1 − e ) = − −(D − 2)( + ) =0
dr r 2r2 dr dr r
(20.4)
⇒ Φ = −∆
The reason we choose to solve out all Einstein’s equations instead of stopping with the compo-
nents of the Ricci tensor is because it is from the Einstein equation Gt̂ t̂ we can define the higher
20.1. BLACK HOLES IN HIGHER DIMENSIONS 487
dimensional Schwarzschild radius. If we assume for now that we are not considering Einstein’s
vacuum equation but instead the case for ideal, pressureless matter, we would find
(D − 2) d h (D−3) −2∆
i
Gt̂ t̂ = − r (1 − e ) = −κρ
2rD−2 dr
where κ is some constant to be determined and ρ is a mass density. Rearranging and integrating
both sides with respect to r, one can show [56]
Z r
2κ 2κ m
(1 − e−2∆ )rD−3 = ρ(r0 )r0D−2 dr0 =
D−2 0 (D − 2) ΩD−2
where we have used the expression for the volume of a unit D − 2-sphere:
2π (D−1)/2
ΩD−2 = (20.5)
Γ[(D − 1)/2]
If we then define the higher dimensional Schwazrschild radius as
16π(D − 3)Gm
RsD−3 = (20.6)
(D − 2)ΩD−2
where we used κ = 8π(D − 3)G (remember that in D = 4, κ = 8πG) we find from the above that
D−3
−2∆ Rs
e =1− = e2Φ
r
which allows to write line element for the Schwarzschild black hole in dimension D = d + 1 as
" D−3 # " D−3 #−1
2 Rs 2 Rs
ds = − 1 − dt + 1 − dr2 + r2 dΩ2D−2 (20.7)
r r
√
We check to see if this gives us the right form of the line element in D = 4. Using Γ(3/2) = π/2,
we find that Rs = 2Gm, yielding the familiar line element of the Schwarzschild solution. As another
check, the reader will show in an exercise that this solution still obeys the Bekenstein-Hawking
entropy relation
A (2Gm)(D−2)/(D−3) ΩD−2
S= = (20.8)
4G 4G
where we have put in explicit units of G.
In a similar way, the Reissner–Nordström solution may generalized to higher dimensions. For
the purposes of this chapter however, we will only consider the five dimensional extremal Reiss-
ner–Nordström solution, which can be shown to take the form [5]
r 2 2 r 2 −2
0 0
ds2 = − 1 − dt2 + 1 − dr2 + r2 dΩ23 (20.9)
r r
p
where r0 = mG. To make this line element more tractable, define r̃ = r2 − r02 . Notice then
2 r 2 2
r̃ 0 r̃
= 1− dr2 = dr̃2
r r r
The line element then becomes
2 !2 2 !−2 2 −1
r̃2 r̃2

2 r̃ 2 r̃
ds = − dt + dr̃2 +(r̃2 +r02 )dΩ23 = − dt2
+ dr̃2 +(r̃2 +r02 )dΩ23
r r r̃2 + r02 r̃2 + r02
−2
r̃2 + r02
2
r̃ + r02

2
=− dt + dr̃2 + (r̃2 + r02 )dΩ23
r̃2 r̃2
r 2 −2 r 2
0 0
=− 1+ dt2 + 1 + (dr2 + r2 dΩ23 ) (20.10)
r r
where in the last line we chose to relabel r̃ as r. In this form it is easy to see that the position
of the horizon is at r = 0, and the area of the horizon is simply
A = Ω3 r03 = 2π 2 r03
So far we have only considered the ’simple’ black hole solutions. One can actually work out
a higher dimensional generalization of rotating black holes, however the derivation becomes sig-
nificantly more difficult. Part of this is because the line element of the solution depends on if
one is considering an even or odd space-time dimension. Nonetheless, these solutions do exist,
called the Myers-Perry black holes. Other higher dimensional ‘black’ objects also exist, such as the
black string, and, in string theory, the black brane. The black string solution is actually a simple
construction for which the line element takes the general form
ds2 = gµν dxµ dxν + dz 2

where µ, ν only run over indices in dimension D = 4. Moreover, we also assume that the Riemann
tensor has only D = 4 components, which means that solutions to the four dimensional Einstein
equations are automatically solutions to five dimensional Einstein equations. This allows one to
extend the Schwarzschild into an extra dimension giving us the line element for the black string
[23]:
−1
2m 2m
ds2 = − 1 − dt2 + 1 − dr2 + r2 dΩ22 + dz 2 (20.11)
r r
which we can further study to show that the classical notions of ‘horizon’ and ‘spherical symme-
try’ don’t exactly hold in the higher dimensional generalization.
Much has gone into researching the properties of black holes in higher dimensions as it leads to
interesting insights in the topology of event horizons, and even black hole singularities (it is known
that the singularity theorems of Hawking and Penrose don’t necessarily extend nicely in higher
dimensions). We won’t go into further detail on this subject, but the reader is pointed to the list
of references for more information on this fascinating subject.
20.2. ENTROPY OF THE SCHWARZSCHILD BLACK HOLE IN D=D+1 489
20.2 Entropy of the Schwarzschild Black Hole in D=d+1

In this section we will show the general method of calculating the entropy of black holes using string
theoretic approximations. We choose to work with the simplest case, the Schwarzschild black hole,
as it will lend us insight into the method of calculation, as well as the string theoretic view of black
holes. Our approach will follow closely to Susskind’s.
In general, the first step of calculating the entropy of a black hole in string theory is to find
some type of control parameter that can be adiabatically varied. Why? By the adiabatic theorem,
entropy is an adiabatic invariant, and therefore whatever control parameter of the black hole we
choose to vary adiabatically will leave the entropy invariant. If we model a black hole using string
theory, we would say that the black hole is a highly complicated collection of various string states.
These strings are assumed to be interacting, and the strength of string interactions is governed by
the string coupling constant g (as g → 0 the weaker the interactions between strings until finally
at g = 0 the the strings are viewed to be non-interacting). Since black holes are therefore possibly
highly complex objects in terms of string states, it would prove useful for us to morph our black
hole into a more tractable object. The way we do this is by adiabatically varying the string coupling
constant g to zero, making it so the black hole is far less complicated, turning it into a collection
of free strings, however retaining the entropy of the original black hole.
Therefore, let us start with a Schwarzschild black hole in any dimension D = d + 1 with mass
M0 and a string coupling g0 . For simplicity, we will use light-cone coordinates and our results from
string thermodynamics in chapter 16 to aid in this calculation. Before we move to estimate the
entropy of our black hole let us start off with some important assumptions [56]:
- In the limit that the string coupling goes to zero, g → 0, the black hole string states evolve
into a single excited string, rather than a collection of non-interacting strings.
- Since entropy is an adiabatic invariant, by adiabatically varying the string coupling constant to
zero means that the final single excited string will have the same entropy as the initial black hole.
- As we calculated in chapter 16, the entropy of a highly excited string is (16.61)
√
S = 4πk α0 E ≈ m`s (20.12)
√ where we used that the energy of a string is approximately the mass of the string, E ≈ m, and
α0 ≈ `s where `s is the characteristic string length.
- At some point as g → 0, the black hole will make its transition to a string, at which point the
radius of the event horizon is on the order of the string length scale. What this assumption really
says is that, no matter how big the black hole may be, as we decrease the string coupling constant
there is a point where the gravitational constant G is too weak to matter. This is the point where
we may model our black hole as a string.
It turns out that the characteristic string length may be related to the Planck length in the
following way [56]
D−2
g 2 `sD−2 = `P (20.13)
where `D−2
P is given in (1.15). From the above we recognize that as g → 0 `s → ∞ in Planck units.
What this tells us is that at some point the string length will actually exceed the Schwarzschild
radius of the black hole, further pointing to the transition of black hole to a single excited string.
Using (20.6), notice that for our black hole of mass M0 the Schwarzschild radius is approximately
Rs ≈ (M0 G)1/(D−3)
where G is Newton’s gravitational constant in D space-time dimensions. Then, using the fact
that G ≈ g 2 `D−2
s , we find
Rs
≈ (`s M0 g02 )1/(D−3) (20.14)
`s
where g0 is our initial coupling constant. Evidently, for a large enough black hole, the radius
of the event horizon will be much larger than the characteristic string length. Let us now start to
decrease g. Let us also assume the general case where the mass of the black hole varies during this
adiabatic process of tending g → 0. We will denote this mass as a function of the coupling constant,
M = M (g), in which we readily see M0 = M (g0 ). From (20.8) we know that the entropy of the
black hole goes like m`P . Therefore, as long as the object we consider is a black hole, we have that
M (g)`P = constant. Using `P ≈ `s g 2/(D−3) we find
1/(D−2)
g02

M (g) = M0 (20.15)
g2
As an exercise, the reader will show that the ratio of the g dependent radius and string scale
goes to unity when
M (g)`D−2
P ≈ `D−3
s (20.16)
which also gives M (g)`s ≈ 1/g 2 . Using this (20.15) and G ≈ g 2 `D−2
s we find
1/(D−2) 1/(D−2)
g02

G0 M (g)`s 1/(D−2)
M (g)`s = M0 `s ≈ M0 `s ⇒ ≈ M0 G0
g2 2
(`s g )1/(D−2) (M (g)`s )1/(D−2)
D−2
1/(D−2) 1/(D−3)
⇒ (M (g)`s )(D−3)/(D−2) ≈ M0 G0 ⇒ M (g)`s ≈ M0D−3 G0 (20.17)
Based on our starting assumptions, we have found that a black hole of mass M0 will evolve into a
single free string satisfying (20.17). Comparing this to the entropy one may compute using (20.12)
we find that the entropy is
D−2
1/(D−3)
S ≈ M0D−3 G0 (20.18)
If we compare this result to the entropy of the Schwarzschild black hole as given in (20.8), we
find that the two are in agreement! What this calculation signals is that the Bekenstein-Hawking
entropy of a black hole (at least a Schwarzschild black hole) agrees with the microscopic entropy
of a free string. Therefore, the famous thermodynamic Hawking entropy formula does indeed have
a statistical mechanical analog. An interesting side note is that since the black hole evolves into a
highly excited string with an entropy given by (20.8), it follows from (16.62) that the temperature
of the horizon right before it becomes a string is equal to the Hagedorn temperature TH ≈ 1/`s ,
which we know from before as the maximum temperature a string can achieve. More precisely
20.3. MICROSCOPIC ENTROPY OF AN EXTREMAL BLACK HOLE 491
is that the temperature of the so-called stretched horizon of the black hole, an effective surface
about one Planck length from the horizon we are used to, has a temperature equal to the Hagedorn
temperature. The above result is indeed fascinating, and it does show the method string theorists
use to calculate the entropy of a black hole, but no where in our derivation did we consider fermions.
Rather, we used the results of a highly excited bosonic string. It is possible to include fermion states
using supersymmetry and hence superstring theory. This is the topic of the next section, which we
will now move to.
20.3 Microscopic Entropy of an Extremal Black Hole

Here we will give an equivalent albeit far more involved calculation of the entropy of a non-rotating,
extremal black hole in five space-time dimensions. This is the simplest non-trivial example one may
consider, and it was the black hole Strominger and Vafa considered in their foundational paper
Microscopic Origin of the Bekenstein-Hawking Entropy. The approach we will take is in similar
spirit as their paper, however ours will be far more qualitative, matching the derivation given in
Becker and Schwarz.
The starting point of our derivation begins with a D = 10 type IIB closed superstring theory. The
black hole solution itself actually comes from a regime in where our string theory is approximated
by a type of supergravity theory known as type IIB supergravity. Moreover, for simplicity, we
choose to curl up five of the spatial dimensions into circles, which we denote by x5 , x6 , x7 , x8 , and
x9 . The black hole itself is a spherically symmetric, non-rotating, 3-charge, extremal in the five-
dimensional uncompactified effective space-time, M 5 , defined by coordinates x0 , x1 , x2 , x3 , and x4 .
This uncompactified space is actually realized as a 5-torus, T 5 = T 4 × S 1 . One may derive the line
element of this particular black hole solution, yielding [5]:
ds2 = −λ−2/3 dt2 + λ1/3 (dr2 + r2 dΩ23 ) (20.19)
with
3 r 2
i
Y
λ= 1+ (20.20)
i=1
r
Notice that when r1 = r2 = r3 = r0 we have

r 2 3
0
λ= 1+
r
, which upon substitution we obtain the line element given in (20.10), showing that this solution
does correspond to an extremal black hole. Moreover, just as we saw from (20.10), the position of
the horizon is located at r = 0 with an area of A = 2π 2 r1 r2 r3 .
It should be pointedout that when we say ‘3-charge black hole’ we mean that the black hole
carries three different electric charges with respect to three different Maxwell-like gauge fields that
live on M 5 . We denote these charges by Q1 , Q5 and N . Each of these charges actually come
from D-branes wrapping around the compact dimensions, except for N , which corresponds to a
momentum quantum number. Lastly, the reason why one uses an extremal supersymmetric black
hole is because in D = 5, most of the original supersymmetry found in D = 10 Minkowski space-

time from the type IIB theory remains [66]. This is crucial to the calculation, however since we
won’t go through the hairy details, we may take this property for granted.
Moving on, if we let (2π)4 V be the volume of the 4-torus T 4 making up part of M 5 , and R is the
radius S 1 , one can out that each of the radii ri correspond to a specific mass Mi in the following
way
g 2 `8s
ri2 = Mi (20.21)
RV
Why are there three distinct masses? It turns out that each of the distinct charges have an
associated ‘mass’, which one may calculate to be [5]
Q1 R Q5 RV N
M1 = M2 = M3 = (20.22)
g`2s g`6s R
It does hold however that the total mass of the black hole is simply the sum of each of these
masses, i.e. M = M1 + M2 + M3 . From here we may calculate the entropy using the Bekenstein-
Hawking formula in five dimensions:
r r r
A 2π 2 g 2 `8s g 2 `8s g 2 `8s 2πg`4s p
S= = M 1 M 2 M 3 = √ M1 M2 M3
4G5 4G5 RV RV RV RV
where we used G5 = G10 /((2π)5 RV ), with G10 = 8π 6 g 2 `8s , which one can get by massaging
(1.16) a little. Using (20.22) we find that the entropy according to the Bekenstein-Hawking entropy
formula is
s
2πg`4s p 2πg`4s Q1 Q5 N RV p
S= √ M1 M2 M3 = √ = 2π Q1 Q5 N (20.23)
RV RV g 2 `8s
The trick now is to regain this same expression for the entropy from the microscopic level.
We begin with a black hole, at zero string coupling, constructed using a type IIB superstring
theory with the same set of compactified coordinates curled up into circles. We still have the charges
Q1 and Q5 , in which we interpret has D-brane wrapping. That is, just as we saw in section 12.6, the
charges Q1 and Q5 are generated by D-branes wrapping themselves around the compact dimensions.
Specifically, the charge Q1 is generated by wrapping a number Q1 of D1-branes around the circle
x5 , while the charge Q5 is generated by wrapping a number Q5 of D5-branes around the five circles,
which, since a D5-brane is five dimensional, it wraps around the entire compactified space. What’s
more is in this configuration we make it so all of the D-branes are coincident, a configuration which
cannot be built in different ways without losing supersymmetry [66]. But remember the point of
this calculation is to calculate the entropy of the black hole by counting its possible states, its
possible configurations. Put another way, we must explain how this black hole can be constructed
in many possible ways. Presently, how to do this is not obvious, especially since our black hole can
only take on a few different configurations.
Luckily we still have one more charge, N . In our brane construction, N is the momentum
quantum number. If we consider the circle x5 , we find that the momentum around this circle is
equal to p5 = N/R, where R is the radius of the circle. This momentum does not correspond to the
momentum of the D-branes however since they are translationally invariant along the x5 direction.
This momentum is actually carried by the open strings attached to the D-branes. It is in this
20.3. MICROSCOPIC ENTROPY OF AN EXTREMAL BLACK HOLE 493
way that gives us a plethora of possible different states: there are many types of strings stretching
between the Q1 D1-branes and the Q5 D5-branes, each giving a different microstate. Since these
various states are all related to N , we aim to partition N . Before we do this however, it behooves
us to consider a few more properties of our system of coincident D1-D5 branes.
First of all, our system is a bound state. This means that strings of type (1, 1) (stretching from
a D1-brane to a D1-brane), and of type (5, 5) become massive and will not become excited in our
configuration. But since we only care about excited strings (as these are the strings that arise as we
send the coupling constant to zero), we will drop the (1, 1) and (5, 5) strings from counting. Second,
one can work out that the total number of bosonic ground states for the (1, 5) and (5, 1) strings
is four. It follows then by supersymmetry that there are four fermionic ground states, yielding a
total of eight ground states. Lastly, it is possible that the Q1 D1-branes join to form one D1-brane
wrapped Q1 times around the circle x5 , and similarly the Q5 D5-branes may form a single D5-brane
that wraps itself around the entire compactified space. In this configuration however, the charges
do not change [66].
Let us further examine this third property by considering a D1-brane wrapping Q1 times around
the circle x5 , and also consider a (1, 1) string that moves along this D1-brane. The string is
quantized as it moves along this D1-brane in the following sense: the circle x5 has become Q1 times
longer, meaning that the string must travel a distance of (2πR)Q1 to return to its original position
on the D1-brane; meaning that the momentum of the string is quantized by units of 1/Q1 R.
Remember however for our counting argument we are considering strings which stretch between
D1-branes and D5-branes. Let us also assume that the D5-branes are wrapped in the same sense as
described above. Moreover, let us assume for now that the charges Q1 and Q5 are relatively prime.
For a (1, 5) string it takes Q1 turns before its first endpoint returns to its initial position, but, since
Q1 and Q5 are relatively prime, the second endpoint has not returned to its original position (it
takes Q5 turns for this to happen). It follows then that it takes Q1 Q5 turns for both endpoints to
return to their initial position. The end result is that the momentum of the (1, 5) and (5, 1) strings
is quantized in units of 1/(Q1 Q5 R). This result approximately holds even if the charges are not
relatively prime. Take for example Q1 = Q5 = 100, and Q01 = 99 and Q5 = 100. It’s easy to see
that 1/(Q01 Q5 R) ≈ 1/(Q1 Q5 R). Therefore, for large enough Q1 and Q2 , the momentum along the
circle x5 is quantized suggestively as
N Q1 Q5
p5 = (20.24)
Q1 Q5 R
Earlier we said we seek to partition N , however now we realize it is more accurate to partition
the number N Q1 Q5 . In this configuration we have one D1-brane and one D5-brane, which means
there is one kind of string stretching between the D-branes. Also remember that we have 4 bosonic
and 4 fermionic ground states (b = f = 4). Therefore, using (16.58) with Boltzmann’s formula for
entropy we arrive to the entropy of the black hole
r
N Q1 Q5 3 p
S = lnp(N Q1 Q5 ; 4, 4) ≈ 2π (4) = 2π Q1 Q5 N (20.25)
6 2
which matches the entropy given in (20.23)! What we have shown is that the Bekenstein-Hawking
entropy matches the statistical mechanical derivation of black hole entropy in string theory. It
should be noted that since (16.63) is an approximation, the above entropy formula is also an
approximation. Nonetheless, this result is imperative to the validity of string theory, and also
validates the Bekenstein-Hawking entropy.
Since Strominger’s and Vafa’s seminal paper, the entropy of several other black holes have been
calculated, including rotating black holes. One type of black hole that is particularly interesting is
the non-extremal black hole in five space-time dimensions. Since the black hole is not extremal, it
has a finite, non-zero temperature and will decay by the emission of Hawking radiation. Moreover,
in the D-brane picture this D-brane instability can be fully realized. The decay of the black hole
may be interepreted as the collision of a brane and an anti-brane. When the brane and anti-brane
collide, the world volume contains a tachyonic state arising from the lowest mode of the open string
that connects the brane and anti-brane. This tachyon signals an instability in the system, resulting
in the emission of closed string states, which are then observed as Hawking radiation.
Over the last few sections we have focused on resolving the issue of whether there is a statistical
derivation for the entropy of a black hole, and if it matches the thermodynamic entropy as given
by the Bekenstein-Hawking entropy formula. As we have seen by now, this problem can indeed be
solved in string theory, and the microscopic view of black hole entropy matches the thermodynamic
entropy as hoped for. Although we mentioned that the Hawking radiation may be viewed in a
particular case as closed string states, we still have yet to resolve a deeper problem, the information
paradox. That is, it remains to be seen whether the quantum information stored in a system is
conserved. We move to begin to understand this problem now.
20.4 The Laws of Nature and Black Hole Complementarity

Hawking, after discovering that not only black holes emit thermal radiation but also eventually
evaporate, argued that information is not conserved, violating one of the assumed laws of quantum
mechanics. Hawking’s declaration led t’Hooft and Susskind to enter, as Susskind refers to it, the
black hole war in order to resolve this information loss paradox. It wasn’t until in 1993 when
Susskind, Thorlacius, and Uglum came up with a resolution commonly referred to now as black
hole complementarity. In order to have a fair grasp of the postulates of black hole complementarity,
let us first review what many physicists feel are some of the fundamental laws of nature.
The first of these laws is the conservation of information. Essentially what this says is the
information of an ‘event’ must be conserved. As we noted in chapter 16, one way of characterizing
information is through Von Neumann entropy, which is defined in terms of a density matrix
S = −T r(ρlnρ) (20.26)
When one studies quantum information theory, one can define various types of entropy depending
on the structure of a system. If we assume we have a large system composed of several subsystems,
one defines the so-called coarse-grained entropy or thermal entropy of the composite system to be
the sum of the entropies of the smaller subsystems:
X
ST hermal = Si (20.27)
i
If we let the subsystems to interact, or entangle with one another, the subsystem entropies Si
become non-zero, and therefore the thermal entropy is also non-zero. In this sense the subsystem
entropies, sometimes called fine-grained entropy or entanglement entropy is a measure of how much
the subsystems interact with each other [56]. Using these definitions, one can consider a quantum
system that exchanges information, and shows that information must be conserved.
20.4. THE LAWS OF NATURE AND BLACK HOLE COMPLEMENTARITY 495
Another way to think about quantum information is to apply unitarity evolution of quantum
mechanics. What this says is suppose we have some initial state |ψin i, then the evolution of this
state is governed by a unitary matrix S, such that the final state is given by
|ψout i = S|ψin i
From here we see the alternative way of stating the principle of information conservation, and it
is through the unitarity of S. Recall that a unitary matrix satisfies U U † = I. This means that in
principle we should be able calculate
|ψin i = S † |ψout i
We can now realize the statement of information conservation: the information needed to specify
a state at early times is equal to the information needed to specify a state at later times. Hawking
in 1976 argued that black hole evaporation seems to violate this tenet, claiming that systems that
might have started in pure states would end up in mixed states, and hence a loss of information.
Moreover, the purity of the state wouldn’t be restored if the black hole evaporated.
Another one of nature’s laws also comes from quantum mechanics and it is known as the Quantum
Xerox Principle, or sometimes referred to as the no-cloning principle. In short, this principle states
that one cannot duplicate an exact copy of a quantum system. That is, a device which allows the
exact replication of a quantum system cannot physically exist. For example, suppose we had a such
device and sent in a spin up state, the result would be two spin up states, the copy and the original,
i.e.
| ↑i → | ↑i| ↑i
Similarly, we might copy a spin down state
| ↓i → | ↓i| ↓i
Now suppose we considered a state with a polarization along the x-axis, a spin up x-state in the
z basis
1
√ (| ↑i + | ↓i)
2
Now if we had a quantum xerox machine, we would find
1 1 1
√ (| ↑i + | ↓i) → √ (| ↑i + | ↓i) √ (| ↑i + | ↓i)
2 2 2
which violates not only how a general superposition state evolves (quantum states evolve linearly
in their basis states), one can also show that with a set of measurements of these states leads to
a violation of the Heisenberg uncertainty principle, one of the fundamental principles of quantum
mechanics.
The third law of nature that most physicsts assume to be true is the equivalence principle, which
we reviewed back in chapter 18. In essence, the equivalence principle states that a gravitational
field is locally equivalent to an inertial frame. Put another way, it says that a freely falling observer
will not experience gravitational affects aside from tidal forces. This principle is observed for an
infalling observer into a black hole. As the infalling observer falls through the event horizon, despite
the strange affects an external observer sees, the infalling observer passes straight through without
noticing anything different. This principle allows us to approximate the horizon of a significantly
large black hole with flat Minkowski space. Sometimes researchers say there is no ‘drama’ at the
horizon as experienced by the infalling observer.
In order to resolve the information paradox many researchers proposed several possible resolu-
tions. One solution was that the horizon of a black hole is not physically penetrable. That is,
from the point of view from an infalling observer everything that hits the horizon simply bounces
back, never entering the black hole, and hence no loss of information. But if this were the case
then an infalling observer would also hit this impenetrable boundary, a ’brickwall’, just above the
horizon and would also be flung back out. The problem with this notion is obvious: it violates the
equivalence principle. As mentioned above, the near horizon region of a Schwarzschild black hole
can be approximated by flat space, and therefore any disturbance to a freely falling system would
violate the equivalence principle.
A second conceived resolution was that a black hole could never fully evaporate and that all of
the information would be returned to the outside in the form of Hawking radiation. Indeed this
solves the information problem, but another law of nature is violated. By the equivalence principle,
information will also pass freely through the event horizon. But we now see that the quantum
xerox principle is violated. Since the black hole doesn’t fully evaporate, information remains inside
the uevaporated black hole. But if all of the information is to be returned to the outside in the
form of thermal radiation, there must now be duplications of some of the information, or a copy of
quantum states, thereby violating the no-cloning principle. For a long time researchers were unable
to reconcile the information paradox without violating some other law of nature.
It wasn’t until 1993 when a paper came out which solved the information paradox proposed by
Hawking without violating any of the laws of nature described above. The solution has become
known as black hole complementarity. In essence, black hole complementarity says that at the end of
evaporation of a black hole, all of the information is carried by Hawking radiation, and no observer
witnesses the violation of a law of nature. The contemporary view of black hole complementarity
can be summarized by four postulates [58]:
1. The process of formation and evaporation of a black hole seen by a distant observer can be
described entirely within the context of standard quantum mechanics. That is, there is a unitary
matrix which describes the evolution from infalling matter to outgoing Hawking radiation. In short,
unitarity is not violated.
2. The physics outside of the stretched horizon of a massive black hole can be described to a
good approximation by a set of semi-classical field equations. Semi-classical physics here means
the marriage of quantum field theory and general relativity which led Hawking to his remarkable
discovery.
3. To a distant observer, a black hole appears to be a quantum system with discrete energy levels.
4. A freely falling observer experiences nothing out of the ordinary as they the cross the event
horizon of a significantly large black hole. Therefore, the equivalence principle is upheld.
20.5. UV/IR CONNECTION AND BOUNDS ON ENTROPY 497
All in all, black hole complementarity tells us that an infalling observer experiences nothing
dramatic as they cross the horizon and hence the equivalence principle is respected. Moreover, there
is no obvious contradiction for an external observer since the infalling observer cannot send signals
from behind the horizon (no duplication of information, and information is seen to be conserved
since it comes out in the form of Hawking radiation). There is however a potential contradiction
for the infalling observer which deserves explanation.
Consider a large black hole and an infalling system A. A contains some information and passes
through the horizon with no incident, respecting the equivalence principle. Moreover, suppose an
observer B who hovers above the black hole monitoring the outgoing Hawking radiation. Photons
recorded by B encode the information A carried. So far there is no contradiction. Now suppose
after B has collected some information about A it follows suit an jumps into the black hole. We
quickly realize that B acts like a mirror outside of the black hole horizon reflecting back in the
Hawking radiation. This means that A can acquire two copies of original information and we can
imagine A sending a signal to B discovering duplicate information at some point C inside the event
horizon, thereby violating the quantum xerox principle, and hence black hole complementarity is
not self-consistent.
Luckily, this thought experiment fails. Susskind, among others, showed that for A to send a
signal to B with duplicate information before B were to hit the singularity, the energy carried by
A would have to be be several orders of magnitude larger than the mass of the black hole, in which
case A could not fit inside of the black hole, and black hole complementarity is safe.
Rather recently, black hole complementarity is being challenged. In 2012, Almheiri, Marolf,
Polchinski, and Sully (AMPS) came out with a paper arguing that the postulates of black hole
complementarity could not all simultaneously be true. In summary, AMPS has claimed that once
a black hole has radiated more than half of its initial entropy (the time this occurs at is called the
Page time), the horizon is replaced by a ‘firewall’ at which an infalling observer burns up, violating
the equivalence principle and therefore black hole complementarity. Susskind himself seems to agree
with this new theory, albeit cautiously and disagrees on some of the details (Susskind has argued
that Hawking radiation causes the singularity to move toward the horizon an eventually intersect
with the horizon at the Page time leading to a firewall) [58]. Other researchers that are candidates
of fuzzballs (one type of string theoretic black hole) disagree with the theory firewalls, and have
also refined black hole complementarity to include fuzzball complementarity.
At this point in time, many physicists have felt that there is a solution to the information paradox,
one which says information is conserved. However, now it may turn out that the leading solution
for nearly two decades may be replaced by another solution which seemingly violates another law
of nature. Regardless, the search for solving the problem of information starting back in 1976 has
led to some remarkable proposals with surprising consequences, and will most likely lead to many
more.
20.5 UV/IR Connection and Bounds on Entropy

Imagine that we have a proton falling into a large Schwarzschild black hole. The falling proton
carries with it some information: charge, particle type, momentum, etc. What is interesting is
how the proton looks using the principles of black hole complementarity. From the perspective
of an outside observer, as the proton falls to the black hole, it is falling into an increasingly hot
region since as the closer it gets to the horizon, the more concentrated the thermal radiation is.
Susskind uses the analogy that the proton is like a tiny piece of ice that is heading into a very hot
tub of water. Just as ice would melt and diffuse in the hot water, the proton is ’smeared’ over the
event horizon, including the information the proton carried with it. By black hole complementarity
however, an infalling observer wouldn’t see anything strange about the infalling proton; it wouldn’t
see any diffusion of information.
In order to reconcile the two descriptions as given by the infalling observer and external observer,
one must consider the time delaying effects for the external observer. We can imagine that the
infalling observer has a clock which counts down the time for the proton as it reaches the horizon
(the proper time of the proton), as it passes through the horizon, the clock reaches τ = 0. On the
contrary, Schwarzschild time, the time of the outside observer, would tend to infinity as the proton
approached the horizon; the external observer would have to wait an infinite amount of time before
the proton were to pass through the event horizon. Susskind has actually shown that in order to
observe the proton before it were to cross the horizon, one must do it in a time that is exponentially
small as the Schwarzschild time heads off to infinity. The consequence of this is that the region
over which the proton is localized grows at the approximate rate [56]
`2P t/4M G
∆X ≈ e (20.28)
MG
where M is the mass of the black hole. Therefore, black hole complementarity tells us that the
proton will spread over the horizon exponentially fast (t → ∞) as one tries to observe the proton
before it crosses the horizon. This simple example is the most basic form of the UV/IR connection.
As we see, this connection is deeply related to black hole complementarity, and is actually related
to one of the most intriguing ideas to be proposed by string theory, the notion of holography. In
order to further understand holography and the so-called holographic principle, let us first motivate
some of its basic elements.
In this analysis we follow Susskind and Lindesay. Consider a large spherical region of space
Γ. Further consider that this space is full of states which describe some physical system, while
the region outside of Γ is empty. The prime objective is to determine the dimensionality of this
state-space. For simplicity, let us suppose we are dealing with a lattice of discrete spins. Let the
lattice spacing be a and imagine that the lattice fills the entire volume V of the space Γ. It follows
that the total number of spins contained in this volume is V /a3 . A simple combinatorial argument
reveals that the total number of orthogonal states in Γ is
3
N = 2V /a
Since one can thermodynamically relate the total number of states N to the entropy S of the
system as N ≈ expS, it follows that for this system the entropy S is
V
S= ln2 (20.29)
a3
What’s more is that since this entropy is related to the total number of orthogonal states in Γ,
this is the maximum possible entropy. We have therefore found a bound on the entropy of this
system, an it is proportional to the volume of the space. This is in fact the case in general; the
maximum possible entropy for a system confined in a space as described here is proportional to its
volume. This is not the case, however, if we include gravity.
20.6. ADS/CFT CORRESPONDENCE AND HOLOGRAPHY 499
Let us consider a system including gravity in four-dimensional space-time, and again focus on
a spherical region Γ, with a boundary ∂Γ. We denote the area of the boundary by A. It follows
from our previous studies, or perhaps intuitively, that the total mass of a thermodynamic system
with entropy S cannot exceed the mass of a black hole of the same area A, otherwise it would be
large than the region Γ. If we then imagine a spherically symmetric shell of matter that undergoes
gravitational collapse with just the right amount energy such that if forms a black hole, it will
have an entropy satisfying the Bekenstein-Hawking entropy formula, S = A/4G. Moreover, by the
second law of thermodynamics, we know that the original entropy inside of Γ must be less than or
at most equal to A/4G, indicating that the maximum entropy for a region of space when we include
gravity is proportional to its surface area.
This is somewhat of a peculiar result as the system we considered before, any system that does
not include gravity, has its maximum entropy proportional to the volume of the space, dissimilar to
a thermodynamic system which includes gravity. It is here where we begin to understand the notion
of holography. Similar to optics, in gravitational physics, a system that extends over a macroscopic
region of space is said to be holographic if all of its physics can be represented on the boundary of
the region. Using the above as an example, we see that all of the necessary physics in describing
the entropy and hence information of a system which includes gravity can be represented on the
surface, or the boundary, of the space containing the system, instead of the entire volume of the
space. We note that so far we have not mentioned string theory at all, pointing to the fact that
the holographic principle is a general feature of gravitational physics. In the next section, we will
explore the holographical principle in the context of string theory, as it is through string theory
that the holographic principle is best understood so far.
20.6 AdS/CFT Correspondence and Holography

Holography is best exemplified using the geometry of Anti-de Sitter space-time, so let us briefly
explore this model of space-time. Why is this the case? The holographic principle tells us that
the physics of a space can be described by a hologram on the boundary of that space. But it
doesn’t typically make sense to consider a ball-like region with a boundary in general relativity
except for the case of Anti-de Sitter space, which we described as a maximally symmetric space-
time endowed with a negative cosmological constant. This makes AdS space the natural framework
for gravitational holography.
A good place to start to examine the geometry of the space-time is the line element, which in
n + 1 dimensions can be worked out to be [66]
" 2 #
2 2 1 + r2 2 4dξ i dξ i
ds = R − dt + (20.30)
1 − r2 (1 − r2 )2
where R is the radius of the curvature of the space, r2 = ξ i ξ i with i = 1, 2, ...n. We say that
this line element is the line element describing AdSn+1 space. We can readily see that at any fixed
time (dt = 0), the distance from any point to the boundary r → 1 is infinite. An interesting note
however is that it does not take an infinite amount of time for light rays to reach this boundary.
To see this, let us assume that a light ray is traveling in the ξ 1 direction. Using ds2 we see that
1 + r2 2dξ 1 1 dξ 1
dt = ⇒ dt =
1 − r2 1 − r2 2 1 + (ξ 1 )2
Integrating both sides and defining that ξ 1 = 0 at t = 0, we find that xi1 = tan(t/2). This means
that light rays may reach ξ 1 = 1 and thus r → 1 can be reached within a finite time of t = π/2.
We can also examine the conformal properties of AdS space by considering the conformally
related metric
2
(1 − r2 )2 2 1 + r2

02
ds = ds = − dt2 + dξ i dξ i (20.31)
4R2 2
The line element ds02 actually describes a spatial n-dimensional ball with a single infinite time
coordinate. This is why in many texts one will find the author viewing AdS space in n + 1 space as
an n-dimensional sphere with a single time direction, because AdS space is conformally related to
such a space. Moreover, notice that near r = 1 we can approximate this conformally related metric
as
ds02 ≈ −dt2 + dξ i dξ i (20.32)
and the conformal boundary takes the form R × S n−1 , where R is for the single infinite time
direction.
How are we able to recognize that AdS does in fact include gravitational holography? The way
one sees it using string theory is to consider AdS5 space-time, in which the boundary is R × S 3 .
We won’t go through the details here, but researchers have shown that there is a correspondence
between AdS space in five space-time dimensions and a conformal field theory which includes a
super Yang-Mills theory, a theory of particle interactions that includes supersymmetry. What the
correspondence shows is that there is a duality, similar to T-duality, between 5-D gravity and super
Yang-Mills theory defined on the boundary of AdS space in five dimensions. In short, researchers
have shown that a gauge theory that is dual to a type IIB superstring background lives on the
boundary of AdS5 . This correspondence is commonly referred to as AdS/CFT correspondence,
and is still a hot research topic. As an exercise, the reader will look at a specific example of this
correspondence.
It should be pointed out that superstring theory assumes that the dimension of space-time is
D = 10. This means that five of the spatial dimensions must be compactified in order to get back to
AdS5 . To do this, the five extra dimensions are compactified so small that they may be effectively
ignored and the ten dimensional space-time is then effectively treated as a five dimensional ‘bulk’,
described by AdS5 , while the boundary of the space has three spatial dimensions and a single
time dimension. With this model we can now realize a rather exciting result: the world we are
familiar with, a world thought to be a four dimensional manifold, is really the surface of an effective
five-dimensional universe! That is, all of the physics we experience, such as the interactions of
elementary particles, occur on the boundary of the ‘bulk’. Moreover, we experience gravity on this
boundary, but gravity is not limited to this boundary. Rather it may propagate into the bulk, in
which case it becomes the only interaction and is described by supergravity theories. Although
we have discussed the AdS/CFT correspondence qualitatively, the possible consequences of it are
without a doubt intriguing and seem a bit science fictional.
20.7. EXERCISES 501
20.7 Exercises
1. Using (20.6), show that the Schwarzschild solution in arbitrary dimension, as given in (20.7),
obeys the Bekenstein-Hawking entropy formula given in (20.8).
2. In D = 5, it is possible for a three charge black hole to rotate and also be supersymmetric.
One can show that the line element for a supersymmetric black hole in five dimensions is given by
a a 2
ds2 = −λ−2/3 dt − 2 sin2 θdφ + 2 cos2 θdψ + λ1/3 (dr2 + r2 dΩ23 )
r r
where λ is defined in (20.20), and the angular momentum of the black hole is given by J =
(πa)/(4G(5) ).
(a) First show that when the angular momentum is zero, the line element above reduces to that
in (20.19), which is a non-rotating three charge supersymmetric black hole.
(b) Just as for the black hole in (20.19), the horizon for this black hole is also located at r = 0.
Show that in the near horizon limit r ≈ 0 and at constant t, the line element reduces to
ds2 = R2 dθ2 + R2 (cos θ sin θ)2 (dθ + dψ)2 + (R2 − (a/R2 )2 )(cos2 θdψ − sin2 θdφ)2
where R2 = (r1 r2 r3 )2/3 . (Hint: dΩ23 = dθ2 + sin2 θdφ2 + cos2 θdψ 2 ).
(c) Compute the entropy of this black hole in terms of three charges Q1 , Q5 , and N , and the
angular momentum J, using the Hawking entropy formula in five dimensions. Hint: The simplest
way to compute the area of the horizon A is to define a convenient set of orthonormal 1-forms
{σ 1 , σ 2 , σ 3 }, and compute the area using
Z
A = σ1 ∧ σ2 ∧ σ3
Keep in mind that 0 ≤ θ ≤ π/2, and 0 ≤ φ, ψ ≤ 2π. Computing the entropy in this way is the
first step for one to compare the Hawking formula with the microscopic derivation, as done for the
non-rotating three charge black hole.
3. To see how big a typical system must be in order to reach the maximum entropy, consider
thermal radiation at 1000K, corresponding to photons with a wavelength approximately 10−5 cm.
The number of photons N in a volume of radius R is also approxiately given by [56] N ≈ V /λ3 .
Since the entropy is proportional to the number of photons, show that the entropy of this system
is approximately
S ∝ (R(cm) × 105 )3
. Compare this to the maximum entropy calculated using the holographic limit Smax ≈ (R/`P )2 .
4. One way to see the AdS/CFT correspondence is by considering the entropy of black holes in
AdS5 space-time. The line element of a Schwarzcshild black hole in this space-time is given by [66]
−1
r2 r2 r2 r02

ds = − 1 + 2 + 02
2 2
dt + 1 + 2 − 2 dr2 + r2 dΩ23
R r R r
where r02 ≈ G(5) M . (a) Solve for the position of the event horizon r+ .
(b) Using the previous result, show that the entropy of this black hole is
1 2 2 3
SBH = π N TH (2π 2 R3 )
2
where the temperature TH = (R2 + 2r+ 2
)/(2πr+ R2 ) and one may use the following expression
for the five dimensional gravitational constant G(5) = πR3 /2N 2 .
(c) Show that for fixed large r, the line element for the Schwarzschild black hole in AdS5 asymp-
totically approaches AdS5 . Argue from this line element that the boundary is R × S 3 , allowing us
to set the stage for the correspondence. On this boundary, one can calculate the finite entropy of
a SU(N) Yang-Mills theory to be
2 2 2 3
SY M = π N TH (2π 2 R3 )
3
Compare this result with the entropy calculated in part (b). Although we have swept some
important details under the rug, this exercise indicates that the physics of the bulk can be described
by the physics on the boundary of that bulk.
Chapter 21
Alternative Approaches toward

Quantum Gravity
String theory is by all means not the sole theory of quantum gravity. Despite its popular and
obvious success, string theory is but one or many roads toward a quantum theory of gravity. In
this chapter we explore some of the critiques other researchers have given string theory and some
of the alternatives to string theory.
21.1 A Critique of String Theory

By now we have seen the successes of string theory. We have found that the 20 or so parameters in
the Standard Model may be substituted with one single fundamental parameter, an obvious success.
We have observed first hand that gravity does not have to put into the theory by hand, rather it
arises in a natural way, making string theory not only a potential quantum theory of gravity, but
a unifying field theory as well, making it unique in this respect. Lastly, but certainly not least, we
have shown that string theory allows us to find the microscopic origins of black hole entropy, and
matches the famous Bekenstein-Hawking formula. No doubt each of these strengths are tremendous
feats, and it is these feats which allow string theory to be bolstered as the single unifying theory.
This is not true however. String theory is not the single unifying theory of everything. Rather it
is one potential theory among several others, and it is the pitfalls of string theory that have left
many physicists weary.
We have already discussed the major shortcoming of string theory. Namely, its failure to produce
any verifiable prediction. There have not been any real predictions that string theory can propose
for experimentalists. In short, string theory, as of yet, is an untestable theory. As scientists, this
should cause some concern, as a good scientific theory is one which has predictive power, not one
that appeals only to pure, albeit elegant, mathematics. As we have discussed in chapter one,
some feel as though the conventional line of thinking for experiments might be the issue; that the
paradigmatic particle accelerators will not help us probe the short scale properties of the universe
inherent in string theory, as well as any other theory of quantum gravity. Others feel that 21st
century physics is to go into a ‘post-modern’ world, where testing physical theories that are difficult
to do experimentally will be done by alternative means (such as checking the consistency with other
503
504 CHAPTER 21. ALTERNATIVE APPROACHES TOWARD QUANTUM GRAVITY
models). All in all, the greatest hurdle of string theory is its inability to produce any real testable
prediction. This is a serious issue and should be taken honestly.
Another pitfall of string theory is the so-called ‘landscape problem’. We briefly reviewed this
aspect of string theory during our discussion on string theory dualities, and string theory models of
particle physics. Basically the issue of the landscape problem is that string theory does not predict
one type of universe or even a few universes from which we may choose our own from, but rather a
whole plethora of universes. Therefore, if there happens to be an accurate string description of our
world, it lies among many, many others, making it virtually impossible to pick the right one. String
theorists are hopeful that the number of possible universes will either become severely restricted,
or our world will become obvious amongst the rest. How this will achieved exactly remains to be
seen. It is important to remind ourselves of the proposed purpose of string theory: to be both a
quantum theory of gravity, and a unifying theory of all the forces, particles, and fields. This was
the proposal in 1985 after the first string revolution. During this time a unifying theory meant
finding a theory which contains both the standard model and general relativity. String theory, and
not for lack of trying, so far has been unable to give a definitive computation of standard model
parameters we are familiar with. That is, string theory has been unable to unify the forces in the
world as we know it. If string theory is to make good on its promise, it should lead us to a unifying
theory that we can recognize as one which describes our world. Simply put, string theory fails to
describe the world as we see it. For some this isn’t too big of an issue, but it is an important one
to be considered.
The final shortcoming of string theory we consider has to do with its quantization procedure.
By now we are familiar with the unifying theory agenda: the tenets of general relativity must be
reconciled with those of the standard model. There are obvious differences with these two models,
the most important being that the standard model assumes a fixed space-time background, while
in general relativity space-time is dynamical. Part of this issue arose because of the historical
fissure between particle physicists and relativists, where many particle physicists felt that it was
unnecessary to learn the tenets and jargon of general relativity. In a sense, string theory can be
viewed as a theory beyond the standard model (however one with obvious far reaching goals and
consequences), and, for the most part, assumed a fixed space-time background. But remember the
process of quantization we developed in this text: it is the dynamical variables which are to be ele-
vated to quantum operators which satisfy some set of commutation relations. In ordinary quantum
mechanics, the dynamical variables position x and momentum p become quantum operators, lead-
ing to a quantized theory. If we were to take this same attitude seriously, along with the important
feature of general relativity, it should be expected that space and time themselves should become
quantized. Since space and time are dynamical entities, by basic canonical quantization, space and
time should be elevated to quantum operators, revealing the quantum geometry of space-time. In
this text, we have yet to include any discussion on this, which is frankly due to the fact that most of
string theory assumes a fixed background. As we will see momentarily, this is the striking difference
between string theory and the other potential quantum gravity theories.
There have been some attempts to find a fully background independent formulation of string
theory; most notably is string field theory, M-theory, and holography. Despite these attempts
however, a background independent string theory has not been properly understood. If string
theory is to connect to this important quantization procedure, and hence the other approaches to
quantum gravity, this issue must be resolved.
This critical analysis has not been meant to belittle all that we have done, or push aside string
theory. Rather it has been meant to give an honest view of the progress of string theory. By
21.2. TWISTOR THEORY 505
all means, string theory is an outstanding intellectual achievment, one that has made important
contributions to mathematical physics as well as fields in pure mathematics. And when one weighs
what string theory aims to do against its outstanding problems, this theory should not be abandoned
as it stands presently. All in all, string theory has had many successes, along with several pitfalls,
some of which are more serious than others. For string theory to be the theory it holds itself up as,
these shortcomings must be reconciled.
21.2 Twistor Theory

Consider a bucket fall of water that hangs by cord from a rail. Imagine that the bucket has been
wound such that the cord is tightly twisted. When the bucket is released it begins to spin and two
things happen. First the bucket rotates with respect to us and the water inside remains still; the
surface of the water is flat. Eventually, the motion of the bucket is transmitted to the water and
the water and bucket rotate together; the surface of the water becomes concave, as caused by the
rotation. The question is, rotation with respect to what?
This famous experiment known as Newton’s Bucket has raised questions on the nature of space
and time more than once. Newton famously answered this question by saying that the water rotates
with respect to absolute space, proving its existence. Remember that Newton thought that if one
removed all matter in the universe, leaving us with empty space, what would remain are the rigid
structures of absolute space and absolute time. In a similar vein, motivated by Ernst Mach, Einstein
asked a similar question: What would it mean to say a planet is rotating in space if there is nothing
else in the universe against which the rotation can be measured? Einstein’s response, key to his
formulation of general relativity, was that the rotation is with respect to a local dynamical entity,
namely, the gravitational field. This insight, uprooting the longheld Newtonian perspective, left
us with our current understanding of space and time, that they are dynamical, but in the unified
framework of space-time, are rigid and absolute.
Roger Penrose, in similar spirit, asked this same question, however in a more contemporary way.
One of the more fascinating experiments in quantum mechanics is the Stern-Gerlach experiment,
which showed that elementary particles have the curious property called spin. For instance, the
electron can ‘spin’ in two different orthgonal states: spin up and spin down. What Penrose asked
was what is the meaning of the spin of an electron if the universe is otherwise completely empty?
If there is only one electron present in the universe, what does it mean to say that the electron is
spin up or spin down? What is the difference between spin up and spin down? Penrose argued
that if there is such a distinction in quantum theory set in empty space, it would seem that instead
of living in some pre-determined background space, the electron would create its own space, or,
going further, its own space-time. Based on this thought, Penrose then asked whether space itself
could be built out of electrons, or rather more generally, the spin of particles. More precisely, since
spinors are the mathematical tools which describe the spin of a particle, Penrose wondered if space
could be constructed out of spinors.
To answer these questions, Penrose began adding spinors together in a quantum mechanical way
creating so-called spin networks. An illustration of a spin network is given in figure 21.1. As shown
in the illustration, a spin network is a type of abstract graph. However the graph does not live in
some other space, it is space. Penrose found with his spin networks that ‘direction’ comes from
the relationships between other spinors. A rather fantastic result was that spin networks actually
generated properties of angular directions in a three dimensional space. What’s more is large spin
networks can be joined together, only to reveal that the individual networks would not cohere.
Rather they could be patched together in a way similar to a manifold in general relativity. It was
in this way that Penrose saw space could be discretized.
Figure 21.1: A representation of a spin network. The first picture is that of a simple spin network.
The second picture is that of two spin networks interacting with each other.
As far as they got him, Penrose realized that spin networks themselves were not the road to
quantum geometry. What he found was that spin networks were not rich enough to accomplish
his goal because spinors themselves are not rich enough. In a very basic sense, a spinor is a
stationary object that just spins in place; it is not dynamical in any other way. But since Penrose
wanted to construct a quantum space-time that was dynamical, he sought to include objects that
combined angular momentum (or spin, in a sense), and linear momentum. In 1967 Penrose did in
fact accomplish with coming up with such an object; it is called a twistor, which lives in a four
dimensional complex space known as twistor space.
Twistors are actually related to null lines, albeit in a fairly complicated way. Using twistors,
Penrose was able to come up with a mathematical formalism for the quantum geometry of space-
time. What is interesting is that Penrose found that the fundamental object in space-time time is
not a point, but a null line, i.e. the paths followed by light rays and light-cones seem to be the
most fundamental object of space-time. We saw hints of this when studying relativity theory since
it is the light-cone which determines causality. Moreover, a the conventional point in space-time
was replaced by a twistor theoretic interpretation: a point in space-time can be thought of as the
intersection of null twistors, or equivalently, light rays. Since points in space-time are defined in
this way, by a collection of null twistors that pass through it, a point becomes a non-local object,
making space-time itself appearing non-local. The complementary picture in twistor space is that
a point in twistor space corresponds to a complex ‘twisting’ collection of null lines in space-time.
Twistor theory to many is not viewed as an established theory of quantum gravity, but twistor
theory is important nonetheless as the techniques for recognizing quantum geometry is rather similar
to the more modern agendas. Regardless, though twistor theory does not enjoy the same popular
acceptance as other potential theories, it is still a field of research, mostly in pure mathematics.
Penrose, along with others, have been able to describe some particles and even space-time curvature,
and hence gravity, with twistors and twistor networks, though nothing conclusive has come out of
it yet. Lastly, despite its lack of popular success, researchers in both string theory and Loop
Quantum Gravity have begun to implementing twistors, realizing both their mathematical and
physical importance, indicating that Penrose’s twistors and initial attempts at coming up with a
21.3. LOOP QUANTUM GRAVITY 507
quantum theory of gravity may play a decisive role after all.
21.3 Loop Quantum Gravity

Loop Quantum Gravity (LQG) is perhaps the next most popular approach to a quantum theory of
gravity, second to string theory. LQG takes the same position as twistor theory, namely, space-time
itself should be quantized. This is related to the fact that from the relativists point of view space
and time are dynamical objects, and should therefore become quantized. Therefore space-time is to
become granularized. LQG actually begins with a more direct way of approaching the quantization
of Einstein’s theory of gravity: formulate using Hamiltonian theory and apply the usual procedures
of canonical quantization. This is perhaps the most direct way of doing this, but is rather difficult
since Hamiltonian general relativity has a non-polynomial structure, and becuase general relativity
is generally covariant (meaning that a specific set of coordinates have no significance over any other
set of coordinates). It wasn’t until 1986 when Ashtekar found the suitable choice of variables, now
known as Ashtekar variables. Ted Jacobson and Lee Smolin later realized that the Wheeler-DeWitt
equation familiar to quantum gravity admitted solutions when written in the Ashtekar formalism.
Not long after, in 1994, Carlo Rovelli and Lee Smolin showed that the quantum operators of the
theory that were associated to area and volume had a discrete spectrum, meaning the geometry
was quantized.
The quantum geometry developed in LQG actually makes heavy use of the spin networks de-
veloped by Penrose. The crucial difference with LQG is a choice of a noncanonical algebra that is
based on the holonomies (Wilson loops) of the gravitational connection. Holonomies become far
more crucial in LQG than in string theory, as it is the holonomy which becomes a quantum operator
that creates so-called ‘loop states’. In LQG it is believed that space-time itself is formed by loop-like
states, and it is therefore the position of these loops that is relevant, not some pre-determined back-
ground, making LQG a background independent theory. This notion actually stems from quantum
space as described by the same spin networks Penrose developed several decades earlier. Similar to
twistor theory, in LQG, physical space is interpreted as the quantum superposition of these spin
networks.
Where twistor theory and LQG differ is the construction of quantum space-time. Unlike twistor,
LQG uses the path integral formulation of general relativity as motivation. In general relativity,
similar to string theory, path integrals yield transitions between geometries. The point of the path
integral formulation of general relativity is to determine the probability of the transition between
these geometries. Similarly, using the tools of LQG one can define a path integral, which leads to
the interesting structure known as a spinfoam. Simply put, a spinfoam is a geometric structure
built from vertices, edges, and polygonal faces. If one has studied combinatorial topology, such
structures are obtained by the ’gluing’ of points, line segments, triangles, and the higher dimensional
analogs, simplexes. Such entities also go by the name simplicial complexes. Just for reference, a
3-simplex is more commonly known as a tetrahedron. The spinfoam approach actually require a
slightly older branch of mathematics known as Regge calculus. In Regge calculus one approximates
the curved space-time of general relativity with flat simplexes, discretizing space-time. Originally
Regge calculus was introduced for the purposes of numerical relativity, however the techniques have
found their way into quantum gravity.
What is interesting is that a ‘slice’ of a spinfoam is actually a spin network, similar to how a
time slice of space-time gives the three dimensional space of the space-time manifold. In this sense,
spinfoams seem to construct quantum space-time. All in all, spin networks represent quantum space
while spinfoams represent quantum space-time.
LQG, unlike string theory does not in any way aim to be a theory of everything. Rather it only
focuses on developing a quantum theory of gravity, starting with the assumptions that individually
ordinary quantum mechanics and general relativity are correct. The result is that the quantization
strategy of LQG is based around background independence, a pitfall of string theory. Moreover,
LQG does not require additional dimensions or supersymmetry to work, although the theory does
not forbid extra dimensions or SUSY either. Instead quantum geometry, and hence quantum
gravity seems to emerge in a natural way in LQG, making it a decent candidate for quantum
gravity. LQG has had other successes as well, namely the statistical derivation of the Bekenstein-
Hawking entropy formula, and the understanding of the Big Bang singularity. When compared,
loop quantum cosmology seems to be more reasonable than string cosmology. Nonetheless, LQG
does face the same great shortcoming that string theory faces, and that is a lack in predictive power.
So far LQG has not made any realistic predictions, which is absolutely crucial in model of physics.
A conceptual, or perhaps philosophical, problem has also arisen from LQG. It is the problem
of time. This problem can already be seen with the two pillars of modern physics. In quantum
mechanics time takes on the role of Newtonian time: it is fixed in nature, running from negative
infinity to positive infinity. In general relativity however, time is flexible; dynamic. If one seeks
a theory of quantum gravity, the different perceptions of time must be reconciled. What ends up
happening in LQG is the transition amplitudes given using spinfoams do not explicitly depend
on time, a seemingly odd result. The issue is that in quantum gravity the notion of space-time
disappears, similar to how the idea of a trajectory of a particle in quantum mechanics disappears.
Since there is no background space-time, there is no ‘time’ along which everything flows. LQG
is completely well defined without making use of the notion of time, and therefore the world in
which we live can be understood without referring to time. The idea of physics without time to
many might be a frightening thought, and does not have a place in science, but time and time
again physics has altered our pre-conceived illusions of reality. It is without a doubt that quantum
gravity will do this again.
21.4 Causal Dynamical Triangulation
Similar to both twistor theory and LQG is the third more popular approach to quantum grav-
ity: Causal Dynamical Triangulation (CDT). Invented by Loll, Ambjorn, and Jurkiewicz, CDT is
another background independent approach to quantizing space-time itself. To summarize briefly,
CDT is really nothing more than a modification of Regge calculus. It also approximates space-time
with simplexes through a process called triangulation. Each individual simplex becomes a building
block of space-time, but CDT adds one crucial ingredient: preservation of causality. The way this is
done is that the edges of 4-simplexes (pentachoron, see figure 21.2) that are joined together during
the triangulation process must have the time direction going in the same direction, allowing the
simplicial manifold evolve in an ordely fashion, and allowing models of curved space-time to emerge.
21.5. FINAL REMARKS 509
Figure 21.2: The ‘vertex first’ projection of the pentachoron. [26]
An obvious strength of CDT, similar to LQG and twistor theory, is that it begins with a min-
imal set of underlying assumptions. Basically that quantum mechanics and general relativity are
accurate. CDT also does not require any adjustable factors, unlike string theory (1 is still more
than 0!). An obvious disadvantage to CDT is the lack of experimental evidence, just like the other
approaches. Recently however it was shown that CDT simulated cosmological models may in fact
be accurate descriptions of our own universe! Some researchers have found reason to believe that
the shape of our universe is that of a dodecahedron (figure 21.3), or for those who are familiar with
Dungeons and Dragons, a D-12. This polyhedron is also one of the Platonic solids and, ironically,
was thought to represent the universe according to the Greeks.
Figure 21.3: The dodecahedron, a proposed model for the shape of the universe. [26]
21.5 Final Remarks

We have come along way. In this text we started with classical strings and went all of the way to
superstrings, to D-branes, and lastly, to higher dimensional black holes. We examined several facets
of string theory, including its most cherished success, as well as some of its more puzzling issues. As
we have seen in this chapter, there is more than one road to a quantum theory of gravity, but it is
string theory which aims to be more than the other approaches. String theory tries to solve both of
the outstanding problems of modern theoretical physics simultaneously. And it does so by altering
our previous notions of physical reality, and replaces them with a far richer and complicated universe
that seems only to exist in human imagination. As Ed Witten once remarked, ‘string theory is a
theory of the 21st century which happened to be discovered in the 20th century’. Witten may be
right in this respect, but string theory has found tremendous popular success amongst the other
potential theories of quantum gravity.
Does this mean that the alternative theories are wrong? Definitely not. It is the author’s belief
that the other theories, if not correct in what they seek to describe, will have some meaningful
contribution to string theory. Or, it might even be the other way; that string theory will likely
make some contribution to the other theories of quantum gravity. By and large, taking a more
conservative stance, string theory appears to be on the right track. However it will need some other
insight, perhaps that of twistor theory, LQG, or even CDT, to become an accomplished theory of
quantum gravity. In all likelihood, it will be a unification of string theory and the other approaches
which will yield the accurate description of quantum gravity, and even a theory of everything.
Lastly, with all of this talk on the shortcomings of string theory, notably the lack of predictive
power, a banal question is often raised: what if string theory is wrong? What if an experiment
comes out showing explicitly that the universe is not higher dimensional, or that supersymmetry
does not actually exist in our world, what is to become of string theory? Like all theories, string
theory would have to be reimagined in some way; it would have to be reworked to see if the theory
can accomplish its original goals without these ‘requirements’. If this is not possible, if string theory
cannot be mended, then as good scientists we must take it upon ourselves to move on. This would
not mean however that the efforts of string theory would have been fruitless. As we have already
mentioned, string theory has had tremendous influence on other fields of mathematical physics and
several branches of pure mathematics. Either way, whether string theory is found to be true or not,
our perspective on physical reality will likely never be the same.
Appendix A
Van der Waerden Notation for

Weyl Spinors
A.1 Lorentz Invariants

In this appendix we will explore some of the details of the van der Waerden notation of Weyl spinors
which is used heavily in the superspace formalism of supersymmetry. As mentioned in chapter 14,
our motivation for Lorentz invariants is the Einstein summation convention, that is, matching a
lowered index with an upper index. As developed in chapter 14, we have eight options
η a χa χa η a η̄ȧ χ̄ȧ χ̄ȧ η̄ ȧ (A.1)
χa χa χ̄ȧ χ̄ȧ η a ηa η̄ȧ η̄ ȧ (A.2)

a a ȧ ȧ
A brief calculation shows that η χa = χ ηa and η̄ȧ χ̄ = χȧ η , thereby leaving us with six Lorentz
invariants, the same amount we had found before using van der Waerden notation. These actually
match the Lorentz invariants we had been working with before introducing the dotted notation.
For example, consider χa χa . We must represent χa in terms of lower indices so we must use (iσ 2 )
to lower indices. Therefore, in exactly the same way as the metric is used to raise and lower indices
of vector and tensor components, we may write
χa χa = (iσ 2 )ab χb χa = χb (iσ 2 )ab χa

Had we chosen to write this in matrix form we would make the identification
χa χa = χb (−iσ 2 )ba χa = χT (−iσ 2 )χ

Similarly, one can show that
1 1
χa χb = − (iσ 2 )ab χ · χ χ̄ȧ χ̄ḃ = (iσ 2 )ȧḃ χ̄ · χ̄ (A.3)
2 2
There is a crucial notational convention which must be pointed out. A dot product between
unbarred spinors corresponds to undotted indices written diagonally downward (i.e., η · χ = η a χa ),
511
512 APPENDIX A. VAN DER WAERDEN NOTATION FOR WEYL SPINORS
while a dot product between barred spinors corresponds to the dotted indices written diagonally
upward (i.e., η̄ · χ̄ = η̄ȧ · χ̄ȧ ) [30]. Using this notation, we have the Lorentz invariants
η a χa = η · χ χ̄ȧ η̄ ȧ = χ̄ · η̄ χa χa = χ · χ (A.4)
χ̄ȧ χ̄ȧ = χ · χ η a ηa = η · η η̄ȧ η̄ ȧ = η̄ · η̄ (A.5)

This notation is certainly annoying at first, however, with practice, using it becomes second
nature.
A.2 Index Notation for the Pauli Matrices

Before we calculate some of the important results, we must introduce index notation for ±iσ 2 and
for σ µ . When we write the matrices ±iσ 2 , it makes some calculations in SUSY awkward, for that
reason we introduce two definitions
(iσ 2 )ab ≡ ab (iσ 2 )ȧḃ ≡ ȧḃ (A.6)
(−iσ 2 )ab ≡ ab (−iσ 2 )ȧḃ ≡ ȧḃ (A.7)
Immediately we see that 12 = 1̇2̇ = 1, 21 2̇1̇ = −1, 12 = 1̇2̇ = −1, and 21 = 2̇1̇ = 1, from
which we immediately see that is antisymmetric in both its dotted and undotted indices, both
raised and lowered. Keeping in mind that (±iσ 2 ) is used to raise and lower undotted and dotted
indices we see that we may write expressions like
χa = ab χb χ̄ȧ = ȧḃ χ̄ḃ

Moreover, using the notation to write the Pauli matrices we also find that
1 1 1 1 ȧḃ
χa χb = ab χ · χ χ̄ȧ χ̄ḃ = − ab χ̄ · χ̄ χa χb = − ab χ · χ χ̄ȧ χ̄ḃ = χ̄ · χ̄ (A.8)
2 2 2 2
Lastly, since −iσ 2 is the inverse of iσ 2 we may write the pair together to obtain the Kronecker
delta. This means that
ab bc = δ ac ȧḃ ḃċ = δ ȧċ (A.9)
Now let’s move on to the van de Waerden notation for expressions with σ µ and σ̄ µ . Remember
back in chapter 14 where we found that σ µ i∂µ η transforms like a left-chiral spinor and σ̄ µ i∂µ χ
transforms like a right-chiral spinor. What this means then, using the van der Waerden notation
introduced back in chapter 14, we have that the indices of σ̄ µ are written as
(σ̄ µ )ȧb (A.10)

while the indices of σ µ are written as
A.2. INDEX NOTATION FOR THE PAULI MATRICES 513
(σ µ )aḃ (A.11)
The reason for this notation comes from our definitions earlier on. Remember that we had
assigned a lower undotted index to χ, while assigning an upper dotted index to quantities that
transform as right-chiral spinors. Therefore, we see that σ̄µ converts a lower undotted index into
an upper dotted index. Hence, the mixed dotted and undotted notation of (A.10). As a warning
to the reader, it is important to keep in mind the order of the dotted and undotted indices in the
mixed index expression. On the hand, σ µ converts an upper dotted index into a lower dotted index,
hence the form of (A.11).
Recall the identity σ 2 σ µ σ 2 = σ̄ µT , which means that
iσ 2 σ µ iσ 2 = −σ̄ µT
Or, using the notation just developed,
˙
iσ 2 σ µ iσ 2 = ca (σ µ )aḃ ḃd
Matching indices yields,
˙ ˙
ca (σ µ )aḃ ḃd = −(σ̄ µT )cd (A.12)
Moreover, since the transpose of a matrix effectively switches the order of the indices, and since
is antisymmetric, we also have
˙ ˙
ca (σ µ )aḃ dḃ = (σ̄ µT )dc
The inverted version of this formula is given by
cb d˙ȧ (σ̄ µ )ȧḃ = (σ µ )cd˙ (A.13)

Using this notation we see that we may write expressions like
χ̄σ̄ µ λ = χ̄ȧ (σ̄ µ )ȧb λb = ȧċ χ̄ċ (σ̄ µ )ȧb bd λd

for left-chiral spinors χ and λ. Moreover, if we contract the indices of with σ̄ µ The above just
becomes
ċȧ db (σ̄ µ )ȧb χ̄ċ λd = (σ µ )dċ χ̄ċ λd = −λd (σ µ )dċ χ̄ċ = −λσ µ χ̄
For practice, consider the expression (λσ µ χ̄)(λσ ν χ̄). Using index notation this simply becomes
˙ ˙
(λσ µ χ̄)(λσ ν χ̄) = λa (σ µ )aḃ χ̄ḃ λc (σ ν )cd˙χ̄d = −(σ µ )aḃ (σ ν )cd˙λa λc χ̄ḃ χ̄d
Then, making use of (A.8), this just becomes
1 µ ˙ 1 ˙
= (σ )aḃ (σ ν )cd˙ac ḃd λ · λχ̄ · χ̄ = ca dḃ (σ µ )aḃ (σ ν )cd˙λ · λχ̄ · χ̄
4 4
1 µ dc ˙
= (σ̄ ) (σ ν )cd˙λ · λχ̄ · χ̄
4
˙
A brief exercise shows that (σ̄ µ )dc (σ ν )cd˙ = T r(σ̄ µ σ ν ) = 2ηµν , yielding
1 µν
(λσ µ χ̄)(λσ ν χ̄) =
η λ · λχ̄ · χ̄
2
Now that we have all the notation we need, let’s move on to working out some of the important
results found in chapter 14, this time using van der Waerden notation.
A.3 The SUSY Algebra in van der Waerden Notation

The goal of this section is to work out the details of (14.134) and (14.135), giving us more practice
with van der Waerden notation, while yielding the results we worked so hard to get back in chapter
14. Rather than doing all of the work, we will leave the derivation for (14.134) as an exercise for
the reader. Instead, we will prove (14.135) relying on another exercise the reader is supposed to
complete back in chapter 14. To compute these commutators, we follow closely to the work done
in back in chapter 14 section 14.10. Our goal is to, from (14.79), get to an expression of the form
1
− [[ζ · Q, ω µν Jµν ], φ] = δω δζ φ − δζ δω φ (A.14)
2
Our first goal is to first write a simple transformation δω , for a left-chiral spinor when subject to
a Lorentz transformation with a parmater ω. Motivated by (14.76) to write the transformation as
[30]
i
χ0 = exp( ω µν σµν )χ (A.15)
2
where
i
σµν ≡ (σµ σ̄ν − σν σ̄µ )
4
Therefore, we see that
δω δζ φ† = δω (χ̄ · ζ̄) = (δω χ̄ȧ )ζ̄ ȧ

where we used the dotted index notation for spinor dot products. Now, using (A.15), its a simple
task for one to show that
1 µν
ω (iσµν χ + xν ∂µ χ − xµ ∂ν χ)
δω χ =
2
Taking the Hermitian conjugate of this expression, we gain the variation we require to finish our
above calculation:
1 µν
δω χ̄ = ω (−iχ̄σ̄µν + xν ∂µ χ̄ − xµ ∂ν χ̄)
2
where
i
σ̄µν ≡ (σ̄µ σν − σ̄ν σµ )
4
and note that (σµν )† = (σ̄µν ). If we put in explicit indices, we notice that δω χ̄ becomes
A.3. THE SUSY ALGEBRA IN VAN DER WAERDEN NOTATION 515
1 µν
δω χ̄ȧ = ω [−iχ̄ḃ (σ̄µν )ḃ ȧ + xν ∂µ χ̄ȧ − xµ ∂ν χ̄ȧ ]
2
Therefore,
1 µν
δ ω δ ζ φ† = ω [−iχ̄ḃ (σ̄µν )ḃ ȧ ζ̄ ȧ + xν ∂µ χ̄ȧ ζ̄ ȧ − xµ ∂ν χ̄ȧ ζ̄ ȧ ]
2
Similarly,
1 µν
δ ζ δ ω φ† = ω δζ (xν ∂µ φ† − xµ ∂ν φ† )
2
1 µν
ω (xν ∂ χ̄ȧ ζ̄ ȧ − xµ ∂ν χ̄ȧ ζ̄ ȧ )
=
2
Then, taking the complex conjugate of (A.14), we have
1 1
− [[Q̄ · ζ̄, ω µν Jµν ], φ† ] = − [[Q̄ȧ ζ̄ ȧ , ω µν Jµν ], φ† ] = (δω δζ − δζ δω )φ†
2 2
i
= − ω µν χ̄ḃ (σ̄µν )ḃ ȧ ζ̄ ȧ
2
And therefore
[[Q̄ȧ , Jµν ], φ† ] = iχ̄ḃ (σ̄µν )ḃ ȧ

Now recall (14.66),
±[ · Q, φ] = −iδφ
Moreover, since, δφ = ζ · χ, we have that [ζ · Q, φ] = −iζ · χ, which means for us
[Q̄ · ζ̄, φ† ] = −iχ̄ · ζ̄

Yielding
[Q̄ḃ , φ† ] = −iχ̄ḃ
Making this substitution, we finally arrive to the commutator we have been looking for:
[Q̄ȧ , Jµν ] = −Q̄ḃ (σ̄ µν )ḃ ȧ (A.16)

To obtain the second commutator, we may simply raise indices using . That is
˙
ȧċ [Q̄ċ , Jµν ] = −ḃd˙Q̄d (σ̄ µν )ḃ ȧ
Then, multiplying both sides by ėȧ we have
˙
ėȧ ȧċ [Q̄ċ , Jµν ] = −ėȧ ḃd˙Q̄d (σ̄ µν )ḃ ȧ
˙
⇒ δ ė ċ [Q̄ċ , Jµν ] = −ėȧ ḃd˙Q̄d (σ̄ µν )ḃ ȧ
⇒ [Q̄ė , Jµν ] = ėȧ Q̄ȧ (σ̄ µν )ḃ ȧ

Then, using the identity ċḃ d˙ȧ (σ̄ µν )ȧḃ = (σ̄ µν )ċ d˙ we find the second commutator we were looking
for:
˙ ˙
[Q̄ė , Jµν ] = ėȧ d˙ḃ Q̄d (σ̄ µν )ḃ ȧ = (σ̄ µν )ė d˙Q̄d (A.17)
Appendix B
Grassmann Variables and

Grassmann Integration
In quantum field theory, and supersymmetry, one makes heavy use of anticommuting numbers, or
Grassmann variables. This is because when dealing with fermions, the operator algebra satisfies
anticommutation relations rather than commutation relations. Moreover, from ordinary quantum
mechanics, a consequence of the Pauli exclusion principle is that fermions are represented math-
ematically as anticommuting numbers. In this appendix, we will explore the basic properties of
Grassmann numbers as well as Grassmann differentiation and integration.
Let η be such an anticommuting variable. This means that η anticommutes with itself,
{η, η} = 0 (B.1)
From here we find that

η 2 + η 2 = 0 ⇒ η 2 = −η 2 ⇒ η 2 = 0
Moreover, if x is an ordinary commuting variable, the derivative operator satisfies

d
,x = 1 (B.2)
dx
Let us similarly define a derivative operator for anticommuting variables. In an analogous way,

d
,η = 1 (B.3)
dη
For any function f (η), the power series expansion terminates after the linear term, since, as we
have already established that η 2 = 0. Thus, the power series expansion of f (η) must be of the form
f (η) = a + bη
. Moreover, since this is the entire expansion and η 2 = 0, we also have
d2 f
=0
dη 2
517
518 APPENDIX B. GRASSMANN VARIABLES AND GRASSMANN INTEGRATION
which yields
d d
, =0 (B.4)
dη dη
Note that if constants a, b are ordinary commuting variables, then
df
=b
dη
But, if a, b are Grassmann variables, we have
df
= −b
dη
When we get to the superspace formalism of SUSY, we have the Grassmann variables θ1 , θ2 , θ̄1̇ ,
and θ̄2̇ , where we are using the usual van der Waerden notation. As far as differentiation goes, we
define that
∂θ1
=1
∂θ1
and
∂θ2 ∂ θ̄ ∂ θ̄
= 1̇ = 2̇ = 0
∂θ1 ∂θ1 ∂θ1
Moreover, using van der Waerden notation, we have that θ1 = θ2 , θ2 = −θ1 , θ̄1̇ = θ̄2̇ and
θ̄2̇ = −θ̄1̇ . All this means is that we have to be careful when it comes to differentiating Grassmann
variables. For example,
∂θ1
=0
∂θ1
but
∂θ2
= −1
∂θ1
A calculation which comes up often is computing the derivative of a spinor dot product. Let’s
compute them in general. Consider for example
∂
θ·θ
∂θa
Using the dot product notation defined in chapter 14, this simply becomes
∂ ∂
θ · θ = a bc θb θc = bc δ b a θc − bc θb δ c a
∂θa ∂θ
where we have used the product rule and kept in mind that, depending on the component, the
derivative of the Grassmann variable with respect to another Grassmann variable is either one or
zero, hence the Kronecker delta functions. With simple algebraic manipulation, the above becomes
519
= ac θc − ba θb = ac θc + θab θb = 2θa

Therefore,
∂
θ · θ = 2θa (B.5)
∂θa
Similarly,
∂
θ · θ = −2θa (B.6)
∂θa
For anticommuting numbers we introduce integration by making note of the fact that we desire
translationally invariant integration. If we integrate a function of a real variable over the entire
number line, that integral is translation invariant. That is, if a is some real constant, then it must
be true that Z ∞ Z ∞
f (x)dx = f (x + a)dx (B.7)
−∞ −∞
Now consider our function of a Grassmann variable, f (η). We also desire this same translation
invariance, so we require Z Z
dηf (η) = dηf (η + c) (B.8)
where c is some constant anticommuting number. Using the same Taylor series approximation we
did before, we have that
Z Z Z Z Z
dηf (η) = dη(a + bη) = dη(a + b(η + c)) = (a + bc) dη + b dηη
Therefore, in order for this integral to be translation invariant, the integral cannot depend on c,
allowing us to conclude Z
dη ≡ 0 (B.9)
Moreover, we normalize the Grassmann integral in the following way [37]:

Z
dηη ≡ 1 (B.10)
Note that Grassmann integration is not anti-differentiation. In fact, integration of anticommut-

ing variables behaves just like differentiation. Moreover, due to the anticommuting nature of the
variables, we must keep track of the order of the variables when we integrate. For example, notice
Z Z Z
dη1 dη2 η1 η2 = − dη1 dη2 η2 η1 = − dη2 dη1 η1 η2 = −1
For the Grassmann variables used in the superspace approach to SUSY, we make the definition
that
d2 θ ≡ dθ1 dθ2 (B.11)

Notice then
520 APPENDIX B. GRASSMANN VARIABLES AND GRASSMANN INTEGRATION
Z Z Z
d2 θθ · θ = d2 θ(θ1 θ1 + θ2 θ2 ) = 2 d2 θθ2 θ1 = 2
Therefore,
Z
1
d2 θθ · θ = 1 (B.12)
2
We also make the definition [30]
d2 θ̄ ≡ −dθ̄1̇ dθ̄2̇ (B.13)

With this definition, one can easily verify that
Z
1
d2 θ̄θ̄ · θ̄ = 1 (B.14)
2
Putting both definitions (B.11) and (B.13) together, we also have
Z
1
d2 θ̄d2 θ(θ · θ)(θ̄ · θ̄) = 1 (B.15)
4
As mentioned before, the crucial point is that Grassmann integration isn’t based on the usual
Riemann sums one might expect from real variable calculus. Rather, it is more of an abstract
notion, behaving more like differentiation than integration. Nonetheless, Grassmann variables and
Grassmann calculus are essential for understanding calculations involving fermionic variables, par-
ticularly those in SUSY, and superstring theory.
Appendix C
Solutions to Exercises
Chapter 2 Solutions
#1 Consider c2 t2 − x2 = c2 t02 − x02 . Plugging (2.4) into the RHS, we find
c2 t2 − x2 = c2 (D2 − B 2 )t2 − (A2 − C 2 )x2 + 2(CD − AB)ctx
Comparing each term we find that it must be
CD − AB = 0 ⇒ CD = AB D 2 − B 1 = A2 − C 2 = 1
Notice then that we may make the identification A = D = coshφ, and B = C = −sinhφ, where
we have chosen the minus sign because we think of Lorentz transformations as hyperbolic rotations.
With these identifications, the result in (2.6) follows.
#2 The pole-in-barn paradox concerns itself with the length contraction of a pole as it enters a
barn that in the rest frame is shorter than the length of the pole. The paradox makes itself manifest
when taking into account the perspectives of various observers. It is resolved by the fact that the
barn doors cannot be closed simultaneously.
#3 Swapping indices, we find
Qij Rij = Qji Rji = −Qij Rij
The only way this relation holds is if Qij Rij = 0. Another approach would be to write out each
tensor in its symmetric and antisymmetric parts and relabel indices when convenient.
0
#4 (a) Using Λν µ0 = ∂xν /∂xµ , we find
 
ν ∂x/∂r ∂y/∂r ∂z/∂r
Λ µ0 =  ∂x/∂θ ∂y/∂rθ ∂z/∂θ 
∂x/∂φ ∂y/∂φ ∂z/∂φ
521
522 APPENDIX C. SOLUTIONS TO EXERCISES
 
sin θ cos φ sin θ sin φ cos θ
=  r cos θ cos φ r cos θ sin φ −r sin θ
−r sin θ sin φ r sin θ cos φ 0
(b) To work out the basis vectors, recall that basis vectors transform as eµ0 = Λν µ0 eν . Using this
we find:
er = Λν r eν = Λxr ex + Λy r ey + Λz r ez
= sin θ cos φex + sin θ sin φey + cos θez

Similarly,
eθ = r cos θ cos φex + r cos θ sin φey − r sin θez
eφ = −r sin θ sin φex + r sin θ cos φey
#5 It helps to note that the metric is symmetric. Therefore terms with mixed differentials may
be split equally. For example,
2e2β dudr = e2β dudr + e2β drdu = gur dudr + gru drdu
Keeping this in mind we find
 f 2β 
re − g 2 r2 e2α e2β gr2 e2α 0
 e2β 0 0 0 
[gab ] =  
 gr2 e2α 0 −r2 e2α 0 
0 0 0 −r2 e−2α sin2 θ
#6 We use the metric given to raise and lower indices:
Vi = gij V j ⇒ V1 = g11 V 1 + g12 V 2 = g11 V 1 = (1)(1) = 1

Similarly, V2 = g22 V 2 = r1 . Thus Vi = (1, r2 ). Moreover,
W i = g ij Wj → W i = (0, 1/r2 )
For the dot product, we again use the metric:
V · W = gij V i W j = g11 V 1 W 1 + g22 V 2 W 2 = 1
#7 (a) Using the Minkowski metric in light-cone coordinates, the dot product of x with itself is:
η̂µν xµ xν = −x− x+ − x+ x− + x2 x2 + x3 x3
Alternatively,
523
x · x = xµ xµ = x+ x+ + x− x− + x2 x2 + x3 x3
Comparing both expressions yields x+ = −x− and x− = −x+ .
(b) By analogy with x+ and x− , we have the momentum in light-cone coordinates:
1 1
p+ = √ (p0 + p1 ) p− = √ (p0 − p1 )
2 2
Just as done above, one can show that −p+ = p− and p+ = −p− . The dot product p · p follows
similarly as well.
Chapter 3 Solutions
#1 Using the general rule that L = T − V , in terms of generalized coordinates q, we find
m 2 1 2
L(x, ẋ) = ẋ − kx
2 2
Using the Euler-Lagrange equations,
dL d dL
= −kx = = mẍ ⇒ −kx = ma
dx dt dẋ
which we recognize as Hooke’s Law.
#2 (a) Using the definition of the functional derivative, we find the same equations of motion
as we did in question 1.
(b) Rewriting the action with delta functions, we have

p Z h p i
S[q] = 3 q(1) + cos(q(2)) = 3δ(t − 1) q(t) + δ(t − 2) cos(q(t)) dt
Therefore,
δS[q] 3
=p δ(t − 1) − δ(t − 2) sin(q(2))
δq(t) q(1)
#3 First consider Gauss’ law in differential form:
~ = ∂Ex + ∂Ey + ∂Ez = 1 ρ

~ ·E
∇
∂x ∂y ∂z c
But using the matrix representation of F µν
∂Ex ∂Ey ∂Ez ∂F 0x ∂F 0y ∂F 0z

+ + = + + = ∂ν F 0ν
∂x ∂y ∂z ∂x ∂y ∂z
Noting that ρ = j 0 , we find that Gauss’ law can recast as
1 0
∂ν F 0ν =
j
c
If we now consider the spatial components of F µν we obtain the components for the other
Maxwell equation:
1 ∂Ex ∂Bz ∂By 1

∂ν F xν = − + − = jx
c ∂t ∂y ∂z c
1 ∂Ey ∂Bz ∂Bx 1
∂ν F yν = − − + = jy
c ∂t ∂x ∂z c
1 ∂Ez ∂By ∂Bx 1
∂ν F zν = − + − = jx
c ∂t ∂x ∂y c
0
#4 Fµν → Fµν = ∂µ A0ν − ∂ν A0µ
= ∂µ (Aν + ∂ν ) − ∂ν (Aµ + ∂µ ) = Fµν + (∂µ ∂ν − ∂ν ∂µ )
= Fµν
#5 To find the given equations, we compute the functional derivative with respect to the time-
independent parameter a and set it equal to zero:
δS ∂L d ∂L ∂L
= − =
δa ∂a dt ∂ ȧ ∂a
1 2
=− ẋ − m2 = 0
2a2
which yields (3.72).
Chapter 4 Solutions
#1 Notice
 2
3
X 3
X 3
X
(x · x)(y · y) − (x · y)2 = x2i yj2 −  xi yj 
i=1 j=1 i,j=1
 
3
X X3 3
X X X
= x2i yj2 −  x2i yi2 + 2 xi yi xj yj  = x2i yj2 − 2 xi yi xj yj
i,j=1 i=1 i<j i6=j i<j
Alternatively,
||x × y||2 = (x × y) · (x × y) = (x2 y3 − x3 y2 )2 + (x3 y1 − x1 y3 )2 + (x1 y2 − x2 y1 )2

Expanding the above expression and comparing to our previous computation we find equality.
525
#2 (a) Using the change in variable theorem from vector analysis we find
i
∂ξ ˜1 ˜2
1 2
dξ dξ ≡ |detM |dξ˜1 dξ˜2

dξ dξ =det

∂ ξ˜j
Similarly,
dξ˜1 dξ˜2 = |detM̃ |dξ 1 dξ 2

Note that |detM ||detM̃ | = 1. To show the area given in (4.7) is reparameterization invariant,
we first not that the metric is inherently reparameterization invariant; therefore,
˜p ˜q
˜ ξ˜p dξ˜q = g̃pq ∂ ξ ∂ ξ
gij (ξ)dξ i dξ j = g̃pq (ξ)d
∂ξ i ∂ξ j
≡ g̃pq M̃pi M̃qj = (M̃ )Tip g̃pq M̃qj

Taking the determinant of both sides yields
√ p
g = (detM̃ T )g̃(detM̃ ) = g̃(detM̃ )2 ⇒ g= g̃|detM |
Altogether we find
2√
Z Z Z
1 2
A = dξ dξ g = dxi dxi |detM | g̃|detM̃ | = dξ˜1 dξ˜2 g̃ = Ã
˜ ˜
1
p p
(b) Since we may write (4.12) in terms of an induced metric, as shown in (4.16), the Nambu-Goto
action is manifestly reparameterization invariant.
#3 Using the definitions of Pµτ and Pµσ , and (4.27) and (4.28), we find
∂ τ ∂ σ
P + P
∂τ µ ∂σ µ
   
2 2
~ ∂X
~ ~
∂X
− c2 Xµ0 
∂X

~ ∂X ~ ~ Ẋµ −
T ∂ 
 ∂X
∂σ ∂t Xµ0 − ∂∂σ X
Ẋµ 
 T ∂ 
 ∂σ ∂t ∂t 
=−  s −  s 
c ∂τ 

~ ∂X~
2

~
2

~
2
 c ∂σ 
~ ∂X
~
2

~
2

~
2

∂X ∂X 2 ∂X   ∂X ∂X 2 ∂X 
∂σ ∂t − ∂t − c ∂σ ∂σ ∂t − ∂t − c ∂σ
Due to the symmetry of mixed partial derivatives, it is easy to see that the sum opf each term
vanishes, thereby yielding the equations of motion.
#4 (a) Recall from a basic class on vector analysis, for any vector ~u, its component orthogonal
to a unit vector ~n is
~u⊥ = ~u − (~u · ~n)~n

In our case,
!
~
∂X ~ ∂X
∂X ~ ~
∂X
~v⊥ = − ·
∂t ∂t ∂s ∂s
Or, with our parameterization of σ
~
∂X
~v⊥ =
∂t
(b) Then, using the static gauge,

!2
2 ~
∂X
2
Ẋ = −c +
∂t
!2
~
∂X
0 2
(X ) =
∂σ
∂X~ ∂X~
Ẋ · X 0 = ·
∂t ∂σ
These relations allow us to simplify the term under the square root in the Nambu-Goto action,
!2  !2  !2
0
2 2 0
2 ~
∂X ∂X~ ∂ ~
X ∂ ~
X
Ẋ · X − Ẋ X = · + c2 − 
∂t ∂σ ∂t ∂t
 !2 !2 
2 ~ ~ ~
ds  ∂X · ∂X ∂X
= + c2 − 
dσ ∂t ∂s ∂t
Making use of our choice of σ parameterization, we find that

 !2 
2 ~ 2
0 2 2 0 2 ds ∂ X  = ds

c2 −
2 2

Ẋ · X − Ẋ X = c − ~v⊥
dσ ∂t dσ
Resulting in r
2
~v⊥
ds
q
0
2 2 0 2
Ẋ · X − Ẋ (X ) = c 1−
dσ c2
The action given in (4.31) follows quickly from here.
#5 To do this explicitly, one must show the wave equation is satisfied for each component of
f~ and ~g . For each component, let ξ = ct + σ and η = ct − σ. From here, using the chain rule, it
is easy to show that each component of f~ and ~g satisfies the wave equation, and therefore (4.56)
satisfies (4.54).
Chapter 5 Solutions
527
#1 Starting from (5.11), apply (5.13) and (5.15) to obtain P µτ an P µσ .
#2 Taking the time derivative of (5.34) yields

√ √ Xn
Ẋ µ = 2α0 + i(−i) 2α0 αµ e−inτ cos(nσ)
n n
n6=0
√ X
= 2α0 αnµ cos(nσ)e−inτ
n∈Z
0
A similar calculation yields X µ . Summing both results yields (5.37).
#3 Let us consider an open string with fixed boundary points. Since the endpoints are fixed,
we use Dirichlet boundary conditions
∂X µ

=0
∂τ σ=π,0
Taking τ derivatives of the left and right moving wave equations and examining at the point σ = 0,
we find that r
µ α0 µ α0 X µ −inτ
ẊL (τ, σ = 0) = p̃ + α̃n e
2 2
n6=0
and r
µ α0 α0 X µ −inτ
ẊR (τ, σ = 0) = pµ + αn e
2 2
n6=0
Then, using Dirichlet boundary conditions,

r
α0 X µ
ẊLµ µ 0 µ µ
(αn + α̃nµ )e−inτ = 0

+ ẊR = α (p + p̃ ) +
σ=0

σ=0 2
n6=0
which is only satisfied if p̃µ = −pµ and α̃nµ = αnµ . Notice that we have only considered one fixed
endpoint. Therefore, if pµ must also satisfy (5.41), we are led to conclude that the total momentum
of the string pµ must vanish.
If both endpoints of the string are fixed, we may determine the length of the string. Assuming
both endpoints of the string are fixed at the boundary points σ = 0, π, then we have by Dirichlet
boundary conditions
ẊLµ (τ, σ = π) + ẊR
µ
(τ, σ = π) = 0
Therefore, r
α0 α0 X µ −in(τ +π)
ẊLµ (τ, σ = π) = p̃µ + α̃n e
2 2
n6=0
r
µ α0 µ α0 X µ −in(τ −π)
ẊR (τ, σ = 0) = p + αn e
2 2
n6=0
Using the fact that for fixed boundary points p̃µ = −pµ , we find that
√ √ X αµ
X µ (τ, σ) = xµ0 + 2 α0 pµ σ − 2α0 n −inτ
e sin(nσ)
n
n6==0
The length of the string is then

√
∆X µ = X µ (τ, π) − X µ (τ, 0) = 2α0 pµ π = wπ
√
where w = 2α0 pµ is called the winding term.
#4 Taking the time derivative of X µ gives
√
r
µ α0 X e−inτ µ inσ
µ
Ẋ = 2α0 α0 + i (−in) (αn e + α̃nµ e−inσ )
2 n
n6=0
√
r
α0 X −inτ µ inσ
2α0 α0µ + e (αn e + α̃nµ e−inσ )
2
n6=0
Also,
r
µ0 α0 X −inτ µ inσ
X =− e (αn e − α̃nµ e−inσ )
2
n6=0
Adding and subtracting these two equations yield the desired result.
#5 Using β = 1 for closed strings, apply the same procedure as done in the section on solving
the wave equation in the light-cone gauge. The calculation is nearly identical.
Chapter 6 Solutions
#1 We use the definition of the Noether current as derived in the chapter. When there are
multiple fields to consider (such as the present case), we must find the current due to each field and
sum up the result. Therefore, the spatial part is
∂L ∂L
Jk = δψ + (δψ)∗
∂(∂k ψ) ∂(∂k ψ ∗ )
Here δψ = iψ, resulting in
i~2
Jk = (ψ∂k ψ ∗ − ψ ∗ ∂k ψ)
2m
Putting all of the spatial components together yields the desired result. Similarly for the time
component J 0 . This might not seem like the best procedure. There is in fact an alternative method
in which one takes the governing equation, and rearranges it such that one may use the conservation
of current equation. The trick involves multiplying the governing equation by its conjugate field,
and subtracting this result from the complex conjugate of the entire expression. As an example,
consider the Schrödinger equation in one dimension:
529
∂ψ 1 ∂2
i~ =− ψ + V (x)ψ
∂t 2 ∂x2
Multiply both sides by ψ ∗ . Then, take the complex conjugate of the resulting expression and
subtract it, resulting in
∂ψ ∗ ∂ψ ∗

∂ψ ∂ 1 ∂ ∂ψ
i~ ψ ∗ +ψ = i~ (ψ ∗ ψ) = − ψ∗ −ψ
∂t ∂t ∂t 2 ∂x ∂x ∂x
Then, using the conservation of current equation ∂µ J µ = 0 we identify the spatial and time
components of the Noether current:
∂ψ ∗

0 ∗ k i~ ∗ ∂ψ
J = ~ψ ψ J = ψ −ψ
2 ∂x ∂x
Sometimes this is the more pragmatic procedure in finding the Noether current.
#2 Let us reconsider (6.24). The integral is actually a flux of the current across a curve where
τ is constant. To generalize this integral, consider an infinitesimal (dτ, dσ) along an oriented closed
curve Γ that encloses a simply connected region R of the world-sheet (figure 6.1). We have that
(dτ, dσ) is parallel to the tangent to the infinitesimal segment, indicating that (dσ, −dτ ) is normal
to the segment. We then define the infinitesimal flux of current across the segment as
(Pµτ , Pµσ ) · (dσ, −dτ ) = Pµτ dσ − Pµσ dτ
The outgoing flux across the curve Γ is then

I
pµ (Γ) = (Pµτ dσ − Pµσ dτ )
Γ
If we use the two-dimensional divergence theorem, we find that the flux becomes
Z τ
∂Pµ ∂P σ µ

pµ (Γ) = + dτ dσ = 0
R ∂τ ∂σ
where we used the fact that we are dealing with conserved currents Pµa , thereby satisfying the
conservation of current equation (6.23). The integral we performed showed that the flux of current
across a closed contractible curve on the world-sheet vanishes [72].
We may generalize (6.27) even further. Pick any arbitrary curve γ, starting and ending at the
boundary points of the world-sheet. Let us then define
Z
pµ (γ) = (Pµτ dσ − Pµσ dτ )
γ
Moreover, consider curves ξ,α, and β, where ξ is a curve of constant τ , and α and β are paths
oriented in such a way that we may define a closed contractible curve Γ as Γ = ξ − β − γ + α. For
an illustration, refer to figure (8.2). Since the curve is contractible we have that
Z Z Z Z Z
pµ (Γ) = (Pµτ dσ − Pµσ dτ ) = − + − (Pµτ dσ − Pµσ dτ ) = 0
Γ ξ γ α β
By construction, curves α and β are curves where dσ is constant. Using this fact and that Pµσ
vanishes at the boundary points leaves us with
Z Z Z
(Pµτ dσ − Pµσ dτ ) = (Pµτ dσ − Pµσ dτ ) = Pµτ dσ = pµ
γ ξ ξ
Altogether we have that Z

pµ = (Pµτ dσ − Pµσ dτ )
γ
for any arbitrary curve γ. We may conclude then that string charge conservation (conservation
of momentum) is an integral of the current over any curve on the world-sheet.
Chapter 7 Solutions
#1 Rewrite X̂ and P̂ in terms of a and a† . Taking the commutator of X̂ and P̂ yields:

i~ i~
(a + a† ), (a† − a) = 2[a, a† ]

i~ = [X̂, P̂ ] =
2 2
⇒ [a, a† ] = 1
√ √
#2 Given a|ni = n|n − 1i and a† |ni = n + 1|n + 1i, we note
hn|a† |ni ∝ hn|n + 1i = 0 hn|a|ni ∝ hn|n − 1i = 0

†
Using this we see that hXi ∝ h(a + a )i = 0, and similarly hP i = 0, leaving us with
∆X 2 = hX 2 i ∆P 2 = hP 2 i
Working out hX 2 i,
~ ~ ~
hX 2 i = hn| a2 + (a† )2 aa† + a† a |ni = hn|aa† + a† a|ni =

(2n + 1)
2mω 2mω 2mω
Similarly,
mω~
hP 2 i = (2n + 1)
2
~ ~
⇒ ∆X∆P = (2n + 1) ≥
2 2
Notice for n = 0 we have equality.
#3 (a) Starting with a|0i = 0 we project onto the x-basis to find

mω ~ d
a|0i → x+ ψ0 (x) = 0
2~ 2mω dx
1/2 √
To make the expression simpler, let y = (mω/~) x, then a = (1/ 2)(y + d/dy) reducing our
above differential equation to
531

d
y+ ψ0 (y) = 0
dy
which has the solution
2
mω
ψ0 (y) = A0 e−y /2
⇒ ψ0 (x) = A0 exp − x2
2~
Upon normalization, we find the desired result.
(b) Notice
a† √ a† (a† )n
|ni = √ |n − 1i = f raca† n √ |n − 2i... = √ |0i
n n−1 n!
√
(c) Note that a† = (1/ 2)(y − d/dy). Using this we project part (b) onto the x basis and rewrite
in terms of y, yielding
r ! n
~ 1 1 d mω 1/4 −y2 /2
hx|ni = ψn x = y =√ √ y− e
mω n! 2 dy π~
Using a table of functions, one can show

n
1 d 2
Hn (y) = √ y− e−y /2
2 dy
Thus leaving us with
1/4
mωx2

mω mω 1/2
ψn (x) = exp − Hn x
π~22n (n!)2 2~ ~
#4 Using our result from (a) of the previous exercise we notice that by even and odd arguments
hXi = hP i = 0. But,
Z ∞
mωx2
1/2
2 mω

2
hX i = x exp − dx
−∞ ~π ~
Let mω/~ = α, then the above becomes
r Z ∞ r r √
α d α d π π ~
− exp(−αx2 )dx = − =
π dα −∞ π dα α 2 mω
A similar relation holds for hP 2 i. Ultimately we find ∆X∆P = ~/2, just as before.
#5 These calculations follow exactly as shown in the text.
#6 (a) Consider the case when i = j, then the anticommutation relation becomes
2(M i )2 = 2I
Operating on an eigenvector in the M i basis, we find that
M i M i |mi = |mi ⇒ λm = ±1
(b) Using i 6= j leads to
M i M j = −M j M i ⇒ T r(M i M j M i ) = −T r(M j M i M i )
Using the cyclic rule for the trace operator we find
T r(M j M i M i ) = −T r(M j M i M i ) ⇒ T r(M j ) = −T r(M j ) ⇒ T r(M j ) = 0

Lastly, note that since M i has eigenvalues of ±1 and T r(M i ) = 0, we can never have matrices
that are odd dimensional since the sum of the eigenvalues must sum to zero.
#7 (a) Using the ‘mostly minus’ convention of the Minkowski metric and considering only the
time components, it’s easy to see that the ‘factorized’ Einstein relation is
p0 p0 − m2 c2 = (p0 + mc)(p0 − mc) = 0
(b) Multiplying out the RHS yields
β k γ λ pk pλ − mc(β k − γ k )pk − m2 c2
Since we don’t want linear terms in pk we see that we must have β k = γ k . Thus,
pµ pµ = γ k γ λ pk pλ
(c) Writing out everything in component form yields
(p0 )2 − (p1 )2 − (p2 )2 − (p3 )2 = (γ 0 )2 (p0 )2 + (pγ 1 )2 (p1 )2 + (γ 2 )2 (p2 )2 + (γ 3 )2 (p3 )2
+(γ 0 γ 1 + γ 1 γ 0 )p0 p1 + (γ 0 γ 2 + γ 2 γ 0 )p0 p2 + (γ 0 γ 3 + γ 3 γ 0 )p0 p3
+(γ 1 γ 2 + γ 2 γ 1 )p1 p2 + (γ 1 γ 3 + γ 3 γ 0 )p1 p3 + (γ 2 γ 3 + γ 3 γ 2 )p2 p3

From here, in order for sides of the equation to be satisfied we notice we must insist the γ’s
satisfy the Dirac algebra
{γ µ , γ ν } = 2η µν
(d) Altogether we have that the relativistic energy-momentum relation factors as
(pµ pµ − m2 c2 ) = (γ k pk + mc)(γ λ pλ − mc) = 0

The convention is to take
533
(γ µ pµ − mc) = 0
as the Dirac equation. Let pµ → i~∂µ arriving to
i~γ µ ∂µ ψ − mcψ = 0
This exercise was taken almost directly from Griffiths’ Introduction to Elementary Particles.
The reader with further interest in this subject is urged to examine his text as it gives thorough
discussion on the subject.
#8 (a) From the Dirac equation, we let µ = 0, ∂0 = ∂/∂(ct) (putting c back in explicitly),
yielding
i~ 0 ∂ψ
γ − mcψ = 0
c ∂t
Using the Bjorken-Drell convention for the gamma matrices we find the above becomes
imc2 ψA

1 0 ∂ψA /∂t
=−
0 −1 ∂ψB /∂t ~ ψB
∂ψA imc2 ∂ψB imc2

⇒ =− ψA − =− ψB
∂t ~ ∂t ~
Solving each differential equation gives
2 2
ψA (t) = e−i(mc /~)t
ψA (0) ψB (t) = ei(mc /~)t
ψB (0)
leading to the desired expressions. As ordered in the text, the solutions describe a spin-up
electron, a spin-down electron; a spin-down positron, and a spin-up positron.
(b) Using our plane wave solution we find ∂µ ψ = −ikµ ψ, and therefore by substitution
(~γ µ kµ − mc)u = 0
(c) Using γ µ kµ = γ 0 k 0 − ~γ · ~k we find

" 0 #
−~k · ~σ

1 0 0 ~σ k
µ
γ kµ = k 0
− ~k · = ~
0 −1 −~σ 0 k · ~σ −k 0
Therefore,
" #
(~k 0 − mc)uA − ~~k · ~σ uB
(~γ µ kµ − mc)u = ~
~k · ~σ uA − (~k 0 + mc)uB
For this to equal zero we must have each component equal to zero, i.e.
~k · ~σ ~k · ~σ
uA = uB uB = uA
k 0 − mc/~ k 0 + mc/~
(~k · ~σ )2
⇒ uA = uA
(k 0 )2 − (mc/~)2
Working out the matrix algebra, one can show

~k · ~σ = kz (kx − iky
(kx + iky ) −kz
and therefore it is easy to show that (~k · ~σ )2 = ~k 2 , leading to our desired result. Lastly, for our
result to be true, we require that
~k 2 = (k 0 )2 − (mc/~)2 ⇒ k 2 = k µ kµ = (mc/~)2
which we recognize that it must be k µ = ±pµ /~.
(d) Using
~k · ~σ ~k · ~σ
uA = uB uB = uA
k0 − mc/~ k0 + mc/~
one sees that for uA as given, we find

p~ · ~σ 1 c pz
uB = 0 =
p + mc 0 E + mc2 px + ipy
Similarly, choose

0 1 0
uA = uB = uB =
1 0 1
The first two solutions with chosen uA are “particle” solutions; the last two with chosen uB
are “antiparticle” solutions. This exercise was also taken nearly directly from Griffiths text on
elementary particles. The reader is therefore urged to review his text for much more information
on the subject.
Chapter 8 Solutions
#1 (a) As suggested, notice that since Ω is not time dependent in the Schrödinger picture
d d
hΩi = hψ|Ω|ψi = hψ̇|Ω|ψi + hψ|Ω|ψ̇i
dt dt
The Schrödinger equation is given by
i i
|ψ̇i = − H|ψi hψ̇| = hψ|H
~ ~
d i
⇒ hΩi = − h[Ω, H]i
dt ~
(b) We have
535
[Ω, ΛΘ] = Ω(ΛΘ) − (ΛΘ)Ω
= (ΩΛ)Θ − (ΛΩ)Θ + (ΛΩ)Θ − Λ(ΘΩ) = Λ[Ω, Θ] + [Ω, Λ]Θ

The second identity follows similarly.
(c) Given the Hamiltonian, we know X commutes with itself, so X also commutes with any other
powers of X. Therefore, using Ehrenfest’s theorem we find
i 1 i 1
hẊi = − h[X, P 2 ]i = − h(2i~P )i
~ 2m ~ 2m
hP i
⇒ hẊi =
m
For hṖ i we have
i
hṖ i = − h[P, V0 sin(2πX/a)]i
~
Let’s work out the commutator:
∞ 2n+1
X (−1)n 2π
[P, V0 sin(2πX/a)] = V0 [P, X 2n+1 ]
n=0
(2n + 1)! a
Notice
[P, X 2 ] = −2i~X [P, X 3 ] = [P, XX 2 ] = −3i~X 2 ⇒ [P, X n ] = −ni~X n−1 ⇒ [P, X 2n+1 ] = −i~(2n+1)X 2n
Thus,
∞ 2n
(−1)n 2π

2π X
2n 2π
[P, V0 sin(2πX/a)] = −i~ V0 X = −i~ V0 cos(2πX/a)
a n=0
2n! a a

2π
⇒ hṖ i = − V0 hcos(2πX/a)i
a
We notice that under the given translation hṖ i can never be zero, and is therefore not conserved.
(d) Just as before,
hP i
hẊi =
m
Then, writing out V (X) in terms of a power series (we assume analytic potentials), we find

dV
hṖ i = −
dX
Ehrenfest’s theorem allows us to conclude that there is a classical correspondence between quan-
tum mechanics and classical mechanics (when the spatial extent of the wavefunction is negligible,
Ehrenfest’s theorem yields the classical limit). Moreover, notice that if we have an operator that
commutes with the Hamiltonian, we may find certain conserved quantities.
#2 (a) Notice for mutiple pi , qi we have that the total time derivative of ω(q1 , q2 , ..., qn ; p1 , p2 , ..., pn )
is simply
dω X ∂ω ∂ω
= q̇i + ṗi
dt i
∂qi ∂pi
Using Hamilton’s equations we reach the desired result.
By plugging in qi and pi directly into the Poisson bracket we immediately find
q̇i = {qi , H} ṗi = {pi , H}

We realize that Ehrenfest’s theorem yields a quantum correspondence to the Hamilton equations.
(c) Again, by direct substitution the result quickly follows. Notice the similarity to the canonical
commutation relations. This leads us to realize that we choose our canonical commutations based
on the corresponding classical statement with Poisson brackets. In a sense, when we quantize
our theory, we elevate our dynamical variables to operators, and replace the Poisson bracket with
commutators.
#3 Using (8.25) and (8.51) both (8.52) and (8.53) follow immediately.
#4 Notice
(∂ µ ∂µ − m2 )φ(x+ , x− , xI ) = (∂ + ∂+ + ∂ − ∂− + ∂ I ∂I − m2 )φ(x+ , x− , xI ) = 0
Then using the Minkowski metric in light-cone coordinates the above becomes

∂ ∂ ∂ ∂
−2 + − − − m φ(x+ , x− , xI ) = 0
2
∂x ∂x ∂xI ∂xI
Using the hint, our Fourier transform is
dp+ dD−2 pI
Z Z
φ(x+ , x− , xI ) = exp(−ix− p+ + ixI · pI )φ(x+ , p+ , pI )
2π (2π)D−2
Substituting this into our above expression

∂
−2 + (−ip ) − p p − m φ(x+ , p+ , pI ) = 0
+ I I 2
∂x

∂ 1
⇒ i + − + (p p + m ) φ(x+ , p+ , pI ) = 0
I I 2
∂x 2p
537
Notice now that although we started with a second order differential equation for a scalar field,
we have found a first order differential equation in light-cone time, allowing us to realize a corre-
spondence with the Schrödinger equation. Lastly, using x+ = p+ τ /m2 we gain the desired result.
#5 Using the Minkowski metric in light-cone coordinates, each of the commutators fall out
quickly.
Chapter 9 Solutions
#1 Notice
[(Ẋ I + X 0I ), (Ẋ J − X 0J )] = [X 0I , Ẋ J ] − [Ẋ I , X 0J ]
d d
= 2πα0 iη IJ δ(σ − σ 0 ) − 2πα0 iη IJ δ(σ − σ 0 ) = 0
dσ dσ
Similarly for the other sign.
#2 Since
Z π π
1
cos(nσ) = sin(nσ) = 0
0 n 0
the result follows quickly.
#3 (a) We have
[L⊥ ⊥ ⊥ ⊥ ⊥ ⊥ ⊥ ⊥ ⊥
n , aLm + bLk ] = Ln (aLm + bLk ) − (aLm + bLk )Ln
= a[L⊥ ⊥ ⊥ ⊥
n , Lm ] + b[Ln , Lk ]
(b) To check both conditions we simply apply the definition of the Poisson bracket. Notice
X ∂ω ∂λ ∂ω ∂λ
X ∂ω ∂λ ∂ω ∂λ

{ω, λ} = − =− −
i
∂qi ∂pi ∂pi ∂qi i
∂pi ∂qi ∂qi ∂pi
= −{λ, ω}
And
X ∂ω ∂(aλ + bσ) ∂ω ∂(aλ + bσ)

{ω, aλ + bσ} = −
i

X ∂ω ∂λ
∂ω ∂λ X ∂ω ∂σ ∂ω ∂σ
a − +b −
i
∂qi ∂pi ∂pi ∂qi i
= a{ω, λ} + b{ω, σ}
Showing that the Poisson bracket forms a Lie algebra is a first step in realizing the rich geometric
structure inherent in Hamiltonian mechanics. For those who go on to study the Hamiltonian
formalism of classical mechanics will become familiar with these notions as well as symplectic
geometry.
#4 Using the Virasoro algebra in (9.156) we find
[L⊥ ⊥ ⊥
1 , L0 ] = L1 [L⊥ ⊥ ⊥
0 , L−1 ] = L−1 [L⊥ ⊥
1 , L−1 ] = 2L0
Since the commutation relations satisfy the Virasoro algebra, they form a Virasoro subalgebra.
Further notice that they do not have a central term. To calculate their action on the vacuum,
write each operator in terms of their general expansion. Since the Virasoro operators are normal
ordered, there will always be an annihilation operator acting on the vacuum, and will therefore
vanish. Thus,
L⊥ ⊥ ⊥
0 |0i = L1 |0i = L−1 |0i = 0
Chapter 10 Solutions
#1 By N being non-negative, we mean that it’s eigenvalues are non-negative. Using the ‘mostly
plus’ convention of the Minkowski metric, we find
∞ ∞ ∞
!
† †
X X X
µ †
N= n(an ) anµ = −an0 an0 + ani ani
n=1 n i=1
Since the two time-like oscillators carry a minus sign, [an0 , a†n0 ] = −1. Hence
[−na†n0 an0 , a†n0 ] = +na†n0

Therefore the time-like oscillators also give positive contributions to N (the spatial oscillators
do manifestly). Altogether than, N is non-negative.
#2 H|ψi = (L0 − 1)|ψi = 0.
#3 Using the definition of the inverse of a square matrix, the desired result quickly falls out.
#4 (a) Breaking up the problem into two terms we find

T√ T√
− −hhαβ (∂α X µ ∂β bν )ηµν − −hhαβ (∂α bµ ∂β X ν )ηµν
2 2
Exchanging µ and ν in the first term, and exchanging α and β in the second term, and using
the symmetry of the metric in each yields the desired result.
(b) Rewriting δX µ = ω µν X ν using the Minkowski metric, one finds the desired upon substitution
and using the antisymmetry of ωµν .
539
#1 (a) We must show that the four axioms are satisfied. (i) Given any a, b ∈ Z, a + b ∈ Z, hence
we have closure. (ii) For any a, b, c ∈ Z, a + (b + c) = (a + b) + c, thus we have associativity. (iii)
0 ∈ Z, which is the additive identity element, hence, for every a ∈ Z, there exists a unique identity
element, 0. (iv) For every a ∈ Z there exists an additive inverse, a−1 = −a such that a + (−a) = 0.
Since all four axioms are satisfied, we have a group.
(b) The solution to this problem was motivated by [70]. Consider ψ 0 (~x, t) = ψ(~x0 , t). Notice
then,
∂ 0 ∂ X ∂x0j ∂ X ∂
ψ (~x, t) = ψ(~x0 , t) = 0 ψ(~
x 0
, t) = Rji 0 ψ(~x0 , t)
∂xi ∂xi j
∂xi ∂xj j
∂xj

X ∂ ∂ X ∂ ∂ X ∂ ∂
ψ(~x0 , t) = Rji Rki 0 0 ψ(~x0 , t) = x0 , t)
0 ∂x0 ψ(~
i
∂xi ∂xi ∂xj ∂xk i
∂xi i
i,j,k
where we used
X
Rji Rki = δjk
i
In a similar way it’s easy to show that ~x02 = ~x2 , thereby leaving the potential V (r) invariant
under the rotation. Altogether, the differential equation, i.e. the Schrödinger equation in this form
is invariant under three dimensional rotations.
(c) We need to show that the four conditions are satisfied. (i) Let R1 , R2 ∈ SO(3). It follows then
that R3 = R1 R2 ∈ SO(3). To prove this we must show that R3 is orthogonal and has determinant
one. Since there exist R1−1 and R2−1 , then
R2−1 R1−1 = (R1 R2 )−1 = R3−1

since R2−1 R1−1 ∈ SO(3), it follows that R3−1 ∈ SO(3). Moreover,
det(R3 ) = det(R1 R2 ) = det(R1 )det(R2 ) = 1

Therefore we have closure. (ii) − (iv) We get associativity basically for free since square matrices
are associative under matrix multiplication; I ∈ SO(3), and by definition we have an inverse.
Therefore we see that the set of rotation matrices form a group under matrix multiplication. Lastly,
SO(3) is an example of a non-Abelian group, R1 R2 6= R2 R1 . Think about this intuitively: rotating
an object 90 degrees away from you and then 90 degrees to the left is not the same as doing the
operations the other way.
#2 (a) From the transform we have

 
cos α sin α 0
Rz = − sin α cos α 0
0 0 1
(b) For rotation angle α/n with large n, we may approximate our cos α and sin α with first terms
power series, yielding
 
1 α/n 0
Rz ≈ −α/n 1 0 + O(α2 /n2 )
0 0 1
which is easily written as the desired result.
(c) Defining the matrix T as

 
0 1 0
T = −1 0 0
0 0 0
we find that
h α in
(R(α/n))n = I + T
n
Taking the limit n → ∞ we use the given identity to obtain
∞
n
X αn n
lim (R(α/n)) = exp(αT ) = T
n→∞
n=0
n!
∞ ∞
X α2n 2n X α2n+1
=I+ T + T 2n+1
n=1
2n! n=0
(2n + 1)!
It’s easy to check that
 
1 0 0
T 2n = (−1)n 0 1 0 T 2n+1 = (−1)n T
0 0 0
Upon substitution we attain the desired result.
(d) We know from our studies of ordinary quantum mechanics that
[L1 , L2 ] = iL3 [L2 , L3 ] = iL1 [L3 , L1 ] = iL2

Or, more compactly,
[Li , Lj ] = iijk Lk ⇒ ck ij = iijk

Lastly, considering the Jacobi identity
[[A, B], C] + [[B, C], A] + [[C, A], B] = 0

and using A = Li , B = Lj , and C = Lk we have
[[Li , Lj ], Lk ] + [[Lj , Lk ], Li ] + [[Lk , Li ], Lj ] = 0

541
⇒ cmij [Lm , Lk ] + cmjk [Lm , Li ] + cmki [Lm , Lj ] = 0

From here the result s obtained. Moreover, substituting ck ij = iijk into the above gives us the
identity with the Levi-Civita tensor.
#3 Considering our Minkowski line element after a Wick rotation, which we denote as gαβ =
Diag(1, 1), we may obtain hαβ easily by computing each individual component:
hzz = ∂z τ ∂z τ gτ τ + 2∂z τ ∂z σgτ σ + ∂z σ∂z σgσσ

1 1 1 1
= ∂z τ ∂z τ + ∂z σ∂z σ = + =0
2 2 2i 2i
Similarly hz̄z̄ = 0. But,

1 1 1
hzz̄ = ∂z τ ∂z̄ τ + ∂z σ∂z̄ σ = + = = hz̄z
4 4 2
giving us our desired result.
#4 Using `¯n = −z̄ n+1 ∂,

¯ we see that
[`¯m , `¯n ]f = (`¯m `¯n − `¯n `¯m )f
¯ = (m − n)`¯m+n f
= (m − n)(−z̄ m+n+1 ∂)f
which proves what we wanted to show. An identical calculation holds for (11.66).
#5 Use
∞
¯ µ i X µ n−1
∂X (z̄) = − α̃ z̄
2 −∞ n
By Cauchy’s integral formula we find

I
1 ¯ µ (z̄)
α̃nµ = z̄ n ∂X
π γ
From here, apply the same procedure and use the same contour as done for (11.122).
#1 (a) Using
1 a
X a (τ, σ) = x̄a +
(f (τ + σ) − f a (τ − σ))
2
upon ubstituting in our summation expression for f a (u) and using some trig identities we arrive
to the desired result.
(b) Use (12.17) to rewrite X a (τ, σ) in terms of oscillators.
#2 (a) This exercise is identical to the previous one except modifying the previous example with
half-integer moding. (b) Substitute (12.54) into (12.55), leading to the result.
#3 This exercise is part of one borrowed from Zwiebach [72]. Notice

2
dV 1 φ
= φ −1
dφ 2α0 φ20
Setting this equal to zero we find that the critical points are φ∗ = 0, ±φ0 . The second derivate
is simply
d2 V
2
1 2φ φ
= φ + − 1
dφ2 2α0 φ20 φ20
We get the mass terms by plugging in our critical points from which we find
1 1
m(0) = − m(±φ0 ) =
2α0 α0
#1 (a) Use the note that only terms linear in arguments u and v contribute to the integral.
(b) For the first commutator, take the σ 0 derivative of (13.23), and then add ‘zero’ for free to
(13.22). Integrate (13.29) over σ ∈ [0, 2π] to get (13.30). Lastly, use (13.20), (13.21), and (13.30)
to attain (13.31).
(c) Use (13.37) and (13.38).
#2 Simply substitute in the gauge transformation into the equation and work out the algebra.
Alternatively, one can show that for our defined U and for any function G we may verify

∇ ~ − q∇χ U G = U ∇ − q A ~ G
− qA
i i
Using this makes the proof slightly easier, albeit not as straightforward.
#3 For U, V ∈ U (1) note that we have closure: U V = eiφ eiθ = ei(φ+θ ∈ U (1). Moreover, we
have associativity by the associativity of real numbers:
U (V W ) = eiφ (ei(θ+α) ) = ei(φ+θ) ei(α) = (U V )W

There obviously exists an identity element, that for which the angle of rotation is zero. Lastly,
the inverse is merely the complex conjugate of U .
#4 (a) Using (13.119) and following the analogous calculation for open strings stretched between
D-branes of different dimensionality yields the desired result.
543
(b) Recall that 1/2α0 M 2 = N ⊥ + Ñ ⊥ + constant. This time,

∞
1 0 i i X i i 1 X
L̃⊥
0 = αpp + α̃p α̃−p + α̃k/2 α̃−k/2
4 p=1
2
n∈Zodd
Upon normal ordering we find

∞
1 0 i i X i i 23 1 X 1 1
L̃⊥
0 = αpp + α̃−p α̃p − + α̃−k/2 α̃k/2 + ·
4 p=1
24 2 4 12
n∈Zodd
1 0 i i 15
≡ α p p + Ñ ⊥ −
4 16
A similar result holds for L⊥
0 . Altogether then gives
1 0 2 15
α M = N ⊥ + Ñ ⊥ −
2 8
#1 (a) The Dirac equation becomes

η η η
γ µ Pµ = γ 0 P0 − ~γ · P~ =m
χ χ χ

0 I 0 −~σ η η
⇒ E− ·P =m
I 0 ~σ 0 χ χ
Working out the matrix algebra, the result readily follows. To rewrite the Dirac Lagrangian first
note that
L = ψ̄γ µ Pµ ψ − mψ̄ψ = ψ † γ 0 γ µ (i∂µ )ψ − mψ † γ 0 ψ

Also recall
σ̄ µ

0
γµ =
σµ 0
With some matrix algebra we arrive to the result.
(b) First note that σ̄ µ i∂µ χ transforms as a right chiral spinor. Also recall that χ† η and η † χ are
Lorentz invariants. It follows then
χ† σ̄ µ i∂µ χ (σ̄ µ i∂µ χ)† χ = −i(∂µ χ† )σ̄ µ χ

are Lorentz invariants. Lastly, by integration by parts,
Z Z
d4 x(χ† σ̄ µ i∂µ χ) = − d4 xi∂µ χ† σ̄ µ χ
⇒ χ† σ̄ µ i∂µ χ = −i(∂µ χ† )σ̄ µ χ
#2 Using
2 †T
η iσ χ
ψM = =
χ χ
we find
ψ̄M ψM = (χT (−iσ 2 )χ + χ† iσ 2 χ†T = χ · χ + χ̄ · χ̄

Then,
ψ̄M ψM + ψ̄M γ 5 ψM = 2χ̄ · χ̄
1
ψ̄M IψM + ψ̄M γ 5 ψM = ψ̄M PR ψM

⇒ χ̄ · χ̄ =
2
Similarly for χ · χ. These results are easily generalized to include the other left chiral spinor λ.
#3 Using σ µ σ̄ ν + σ ν σ̄ µ = 2η µν I, notice
σ µ σ̄ ν ∂µ ∂ν χ = (2η µν I − σ ν σ̄ µ )∂µ ∂ν χ
Relabeling indices in the second term of the RHS and rearranging terms yields,
2σ µ σ̄ ν ∂µ ∂ν χ = 2∂ µ ∂µ χ
implying the result.
#4 Using the antisymmetry of coefficient ω µν we find

1
φ(xµ + ω µν xν ) ≈ φ(x) + ω µν xν ∂µ φ(x) = φ(x) + (ω µν − ω νµ )xν ∂µ φ(x)
2
1
= φ(x) + ω µν (xν ∂µ − xµ ∂ν )φ(x)
2
Using ±[ · Q, φ] = −iδφ we find
1 i
−[ ω µν Jµν , φ] = − ω µν (xν ∂µ − xµ ∂ν )φ
2 2
which yields the desired result.
#5 δβ δα φ − δα δβ φ =
α · Qβ · Qφ − β · Qα · Qφ − φα · Qβ · Q + φβ · Qα · Q
= [α · Q, β · Q]φ − φ[α · Q, β · Q]
545
giving us our result. Alternatively, by applying the Jacobi identity, the result immediately
follows.
#6 First we consider a transformation of a translation with a parameter a and a second trans-

formation with a Lorentz transformation with parameter ω µν . Then
1
δω δa φ − δa δω φ = δω (aλ ∂λ φ) − ω µν δa (xν ∂µ φ − xµ ∂ν φ)
2
1 λ µν 1
= a ω ∂λ (xν ∂µ φ − xµ ∂ν φ) − aλ ω µν (xν ∂µ ∂λ φ − xµ ∂ν ∂λ φ)
2 2
1 λ µν
= a ω (ηνλ ∂µ φ − ηλµ ∂ν φ)
2
i λ µν
= a ω (ηνλ [Pµ , φ] − ηλµ [Pν , φ])
2
which is equal to

1
aλ Pλ , − ω µν Jµν , φ
2
Therefore,
iηνλ [Pµ , φ] − iηλµ [Pν , φ] = [[Pλ , −Jµν ], φ]

which yields the desired result.
#7 We will show the case for the auxiliary field F : δβ δφ F = δβ (−iζ † σ̄ µ ∂µ χ)
= −iζ † σ̄ µ σ ν σ 2 β † ∂µ ∂ν φ − i(ζ † σ̄ µ β)∂µ F

Using our result from exercise (3), the first term above is simply
−iζ † σ 2 β † φ = −ζ̄ · β̄φ

Therefore,
δβ δφ F = −ζ̄ · β̄φ − i(ζ † σ̄ µ β)∂µ F

Similarly, it’s easy to show that
δζ δβ = −β̄ · ζφ − i(β † σ̄ µ φ)∂µ F

Taking the difference of these two results and using ζ̄ · β̄ = β̄ · ζ̄ yields the result.
#8 Applying each equation immediately yields the result.
#9 Let’s see why the other anticommutator is zero. Consider an arbitrary function S = S(x, θ, θ̄)
and have the anticommutator act on S:

∂ ḃ ∂ ¯ḃ 1 µ ḃc ¯ ḃ 1 µ ḃc ∂
, Q̄ S = ȧ i∂ − (σ̄ ) θc ∂µ S + [i∂ − (σ̄ ) θc ∂µ S
∂ θ̄ ȧ ∂ θ̄ 2 2 ∂ θ̄ȧ

∂ ¯ḃ ¯ḃ ∂ 1 µ ḃc ∂ ∂
= i ȧ ∂ + i∂ S − (σ̄ ) (θc ∂µ ) + θc ∂µ ȧ S
∂ θ̄ ∂ θ̄ȧ 2 ∂ θ̄ȧ ∂ θ̄
For the first term we can move our derivatives through each other, picking up a minus sign as we
are dealing with Grassmann variables, hence the first term anticommutes and vanishes. Similarly,
since
∂
θc = 0
∂ θ̄ȧ
the second term also anticommutes; hence the entire anticommutator vanishes. Note however
for the other anticommutator we have that Qa depends on θ̄ḃ , and therefore the simillar second
term does not anticommute, yielding an overall non-zero result.
#10 (a) Let’s raise the indices of D̄ȧ . Using notation we have that
∂ i
D̄ȧ = ȧḃ D̄ḃ = ȧḃ − ȧḃ θc (σ µ )cḃ ∂µ
∂ θ̄ ḃ 2
Let θc = cb θb , then using ȧḃ ∂¯ḃ = −∂¯ȧ and cb = −bc we find
∂ i ∂ i
D̄ȧ = − + bc ȧḃ (σ µ )cḃ θb ∂µ = − ȧ + (σ̄ µ )ȧb θb ∂µ
∂ θ̄ ȧ 2 ∂ θ̄ 2
˙ ˙
where we used ca dḃ (σ̄ µ )aḃ = (σ̄ µ )dc .
(b) The first four anticommutators go to zero immediately by applying our arguments for exercise
(9). The fifth anticommutator is the only nonzero one:
iθc µ

∂ i µ ḃ ∂
{Da , D̄ḃ } = − (σ )aḃ θ̄ ∂µ , − (σ )cḃ ∂µ
∂θa 2 ∂ θ̄ḃ 2

∂ ∂ i ∂ i ∂ 1n µ o
, − , θc (σ µ )cḃ ∂µ − (σ µ )aḃ θ̄ḃ ∂µ , − (σ )aḃ θ̄ḃ ∂µ , θc (σ µ )cḃ ∂µ
∂θa ∂ θ̄ḃ 2 ∂θa 2 ∂ θ̄ḃ 4
The first and last term vanish since they anticommute, but the third and second term contribute.
Consider the second term:

i ∂ c µ i µ ∂ c c ∂
− , θ (σ )cḃ ∂µ = − (σ )cḃ (θ ∂µ ) + θ ∂µ a
2 ∂θa 2 ∂θa ∂θ

i ∂ ∂ i i
= − (σ µ )aḃ δ c a ∂µ − θc a ∂µ + θc ∂µ a = − (σ µ )aḃ δac ∂µ = − (σ µ )aḃ ∂µ
2 ∂θ ∂θ 2 2
We attain an identical term from the third term as well, yielding
{Da , D̄ḃ } = −i(σ µ )aḃ ∂µ

547
#1 Plugging in the matrix representation given in (15.3), the result readily follows.
#2 This exercise follows almost immediately by applying the discussion leading up to (15.12).
#3 Using (15.21),
1
LB + δLB = ∂α (X µ + β ∂β X µ )∂ α (Xµ + β ∂β Xµ )
2
1
⇒ δLB = (∂α β X µ ∂ α Xµ + ∂α X µ ∂ α β ∂β Xµ )
2
Using the metric to raise and lower indices and relabeling in the second term yields the desired
result.
#4 (a) Using the Ramond boundary conditions the field is periodic and may be expanded using
integral modes, yielding (15.51). Alternatively, the Neveu-Schwarz boundary conditions change
sign when σ → σ + 2π, and therefore the field must be expanded with fractional modes.
(b) Following the same procedure leading to (15.58), (15.63) follows.
#5 (a) Using (15.79) and noting that a minus one appears for every fermionic oscillator in the
state, the result follows.
(b) Note that the |Ra i states are created from an even number of creation operators while the
|R̄a i are created from an odd number of creation operators. By the GSO operator then, it follows
that the eight states of |Ra i and |R̄a i are fermionic and bosonic respectively.
(c) For α0 M 2 = 2 we have the (−1)F = −1 fermionic states:
I I J
{α−2 , α−1 α−1 , dI−1 dJ−1 }|Ra i I
{α−1 dJ−1 , dI−2 }|R̄a i
and we have the bosonic states:
I I J
{α−2 , α−1 α−1 , dI−1 dJ−1 }|R̄a i I
{α−1 dJ−1 , dI−2 }|Ra i
#6 (a) We use the fact that α0 M 2 = N ⊥ with no ‘offset’. The overall factor of 16 occurs because
each combination of oscillators gives rise to 16 states by acting on each of the ground states [72].
#1 ρ is trivially hermitian. For part (b) we have

X X
ρ2 = pi pj δij |ψi ihψj | = p2i |ψi ihψi |
i,j i
Taking the trace leaves a sum over p2i . In a mixed ensemble, each of the pi is less than one,
yielding the desired result.
#2 We have
∞ X
∞ ∞
Y mn
Y 1
Z= (exp (−~ω0 n/kT )) =
n=1 mn =0 n=1
1 − exp (−~ω0 n/kT )
where we applied the definition of a geometric series.
#3 For this exercise the hints bear the most help. From these hints and carrying through the
induction proof the result follows.
#4 This exercise is motivated by [72], therefore for a more in depth solution see the references.
We start from (16.67), using spherical coordinates, we temporaily let ~u2 = u2 and thus dd ~u ≈
ud−1 du. Therefore the integrand becomes
p
≈ ud−1 exp(−βm 1 + u2 )
Note that the integrand vanishes at 0 and at ∞, however peaks somewhere in between these
values. To find this maximum, we calculate the derivative of the integrand and set it equal to zero:
d d−1 p βmu d−1 p p

(u exp(−βm 1 + u2 )) = − √ u exp(−βm 1 + u2 )+(d−1)ud−2 exp(−βm 1 + u2 ) = 0
du 1+u 2
d−1 u2
⇒ =√
βm 1 + u2
Since βm 1, the LHS is small, implying that u2 1, allowing us to ignore the square root in
the denominator, leaving us with
d−1
u2 ≈ 1
βm
This allows use to the expand the square root in (16.67). Putting back in the vector notation
we find
dd ~u
Z
2 d −βm 1 2
Z(m ) ≈ V m e exp − βm~u
(2π)d 2
which we recognize is just a Gaussian integral. Upon integrating, we find our desired result.
#5 Since lnZstr ≈ −ln(β − βH ),

∂lnZ 1 TH
Estr ≈ − ≈ ≈ kTH
∂β β − βH TH − T
549
#1 Using
er = sin θ cos φex + sin θ sin φey + cos θez eθ = r cos θ cos φex + r cos θ sin φey − r sin θez
eφ = −r sin θ sin φex + r sin θ cos φey

the result follows quickly.
#2 Using the definition of the covariant derivative we have
∇c gab = ∂c gab − gdb Γd ac − gad Γd bc

Substitute in (17.21) and collect like terms. Moreover, use gab g bc = δa c frequently. With some
tedious algebra, the result follows.
#3 From the line element of a 2-sphere with radius a we find
gθθ = a2 gφφ = a2 sin2 θ g θθ = 1/a2 g φφ = 1/a2 sin2 θ

All other components are zero. The only nonzero derivative of these components is
∂gφφ
= 2a2 sin θ cos θ
∂θ
To find the Christoffel symbols, we use (17.21). The first non-zero term occurs when σ = ρ = θ.
Then,
1 θθ
Γθ νµ = g [∂µ gνθ + ∂ν gθµ − ∂θ gµν ]
2
1 ∂gφφ
⇒ Γθ φφ = − g θθ = − sin θ cos θ
2 ∂θ
One can similarly show that the only other nonzero Christoffel symbols are
Γφφθ = Γφθφ = cot θ
#4 We have that U a Ua = constant and ∇a Ub − ∇b Ua = 0. Therefore
0 = ∇b (U a Ua ) = U a ∇b Ua + Ua ∇b U a
Using our second relation above on the second term, we find
U a ∇b Ua + Ua ∇b U a = U a ∇b Ua + Ua ∇b U a = gbc U a ∇a U c + Ua ∇b U a
where we used metric compatibility,
= gbc U a ∇a U c + g ac Ua ∇c Ub = gbc U a ∇a U c + g ac gae U e ∇c gbd U d

= gbc U a ∇a U c + gbd U c ∇c U d = gbc (U a ∇a U c + U a ∇a U c ) = 2gbc U a ∇a U c

which implies the desired result.
#5 Apply Killing’s equation and look for expressions similar to (17.24).
#6 (a) Applying (17.58) one finds
∗dr = r2 sin θdθ ∧ dφ ∗ rdθ = −r sin θdr ∧ dφ ∗ r sin θdφ = rdr ∧ dθ
∗(dr ∧ rdθ) = r sin θdφ ∗ (dr ∧ r sin θdφ) = −rdθ ∗ (rdθ ∧ r sin θdφ) = dr
(b) Note that the basis one forms are given by {dr, rdθ, r sin θdφ}. Then apply the definitions of
the divergence and curl to F = fr dr + fθ (rdθ) + fφ (r sin θdφ).
#7 This procedure is nearly identical to the computation to find the components of the Riemann
tensor for the Schwarzschild solution. Using Cartan’s first structure equation, one should find
ȧ r̂ ȧ θ̂ ȧ
Γr̂ t̂ = ω Γθ̂ t̂ =
ω Γφ̂t̂ = ω φ̂
a a a
√ √
1 − kr2 θ̂ 1 − kr2 φ̂ cot θ φ̂
Γθ̂ r̂ = ω Γφ̂r̂ = ω Γφ̂θ̂ = ω
ra ra ra
Using Cartan’s second structure equation yields the final result.
#1 Consider the motion of two bodies moving in a gravitational field. By Newton’s second law,
the “inertial mass” for each is given by
F1 = mI1 a1 F2 = mI2 a2
Moreover, given the gravitational field φ, related to “passive mass” we find
mp2
F1 = mI1 a1 = −mp1 ∇φ F2 = mI2 a2 = −mp2 ∇φ ⇒ a2 = − ∇φ
mI1
Experiment says a1 = a2 = g, and thus g equals the above, and similarly for mp1 and mI1 .
Equating the two yields
mp1 mp2
=
mI1 mI2
Since m1 and m2 are arbitrary, this leads to mI1 = mp1 . For “active mass”, consider
Gma1
φ1 = −
r
551
Then, if we only consider the radial coordinate for simplicity we find
Gma1 mp2
F2 = −mp2 ∇φ1 = −
r2
Finding the similar equation for F1 and applying Newton’s thrid law we find ma = mp , yielding
the desired result.
#2 This problem was proven previously.
#3 Using the Schwarzschild solution as a guide, in the coordinate basis we have that Rrr is
" 2 #
d2 ν dν dν dλ 2 dλ
Rrr = − + − −
dr2 dr dr dr r dr
Examining our given metric, we let λ → f (r) and ν → 0, leaving us with
2 df
Rrr =
r dr
Using Rj` = 2Kgj` we find
2 df
Rrr = = 2Ke2f ⇒ e−2f df = Krdr
r dr
Integrating yields
1 Kr2 1
− e−2f = + C ⇒ e2f =
2 2 C − Kr2
To find C, we use Rθθ , which in our present case is simply
df
Rθθ = 1 − e−2f + re−2f = 1 − C + 2Kr2
dr
Again using Rj` = 2Kgj` with gθθ = r2 , we are led to C = 1. Putting everything together leads
to the familiar form of the Robertson-Walker line element.
#4 For ρ = 3P and Λ = 0, the Friedmann equations become

3 2ä 1 8πρ
(k + ȧ2 ) = 8πρ + 2 (k + ȧ2 ) = −
a2 a a 3
2ä 1 1 1
⇒ + 2 (k + ȧ2 ) = − (k + ȧ2 ) ⇒ ä + (k + ȧ2 ) = 0
a a a a
We may neglect k/a in this model, so
ȧ2 d
ä + = 0 ⇒ aä + a2 = (aȧ) = 0 ⇒ aȧ = C
a dt
√
Upon integrating we find a(t) ∝ t.
∂L 2m

#1 Note ∂ ṫ
=2 1− r ṫ

d ∂L 4m 2m
⇒ = 2 ṙṫ + 2 1 − ẗ
ds ∂ ṫ r r
Since L has no explicit time dependence, the first geodesice equation is
2m
ẗ + ṙṫ = 0
r(r − 2m)
Following this same process yields the other three geodesic equations:
2 2
θ̈ + ṙθ̇ − sin θ cos θφ̇2 = 0 φ̈ + 2 cot θθ̇φ̇ + ṙφ̇ = 0
r r
m 2 mṙ2
r̈ + (r − 2m)ṫ − − (r − 2m)(θ̇2 + sin2 θφ̇2 ) = 0
r3 r(r − 2m)
#2 Considering when θ = π/2 and at constant φ, we divide the remaining line element by dτ 2
yielding
−1
2m 2 2m
1− ṫ − 1 − ṙ2 = 1
r r
From here, if we consider when ṙ = 0 we attain the other expression. If we let k = 1 we find
−1
2m
1− = ṫ
r
Upon substitution and taking the limit as r → ∞, noting that at this distance we have our initial
velocity, we find the desired result.
#3 By energy conservation we have

1 GmM
mv 2 − E=
2 r
The escape velocity corresponds to when E = 0 (v → 0 as r → ∞), leading to
2 2GM
vescape =
R
In the case of light,
2GM
c2 =
R
which leads to the Schwarzschild radius.
#4 Using
a2 2ma2

2m 2 4ma
0= 1− dt + dtdφ − 1 + 2 + r2 dφ2
r r r r3
553
Consider when r = 2m we find

2
dφ a dφ dφ a dφ
− 2
⇒ = =0
dt 3m dt dt 3m2 dt
This second equation tells us that any massive particle will always move in the direction of the
black hole’s rotation.
p
#5 Let’s be conservative and set P = 0. Then, r± = m ± m2 − Q2 . Let’s calculate the
temperature T as seen from infinity
p
2m 1 − Q2 /m2
T (∞) = p
4πm2 (1 + 1 − Q2 /m2 )
Moreover,
A 2 2
= 4πr+ S=
/4 = πr+
4
In the extreme limit, m = Q and T (∞) = 0 while S = πm2 6= 0, violating the third law.
1
P P
#6 (a) hEi = n En pn = Z n En exp(−βEn )
1 ∂Z ∂ ∂
=− = − (lnZ) = T 2 (lnZ)
Z ∂β ∂β ∂T
(b)
X 1 X
X 1 1
S=− pn lnpn = − exp(−βEn )ln exp(−βEn ) = exp(−βEn )(βEn + lnZ)
n n
Z Z n
Z
∂
= β(hEi + lnZ) =
(T lnZ)
∂T
For the harmonic oscillator we have En = (n + 1/2)~ω. Plugging this into Z and using the
definition of a geometric series we have
exp(−β~ω/2)
Z=
1 − exp(−β~ω)
Then
β~ω
lnZ = − − ln(1 − exp(−β~ω))
2
~ω exp(−β~ω)
⇒ S = −ln(1 − exp(−β~ω)) +
T 1 − exp(−β~ω)
(c) δSBH = 21 ln(2)~−1 δα ≥ 12 ln(2)~−1 δαmin = µb~−1 ln2

For the second part note that the reduced mass of the oscillator µ = m/4, thus the “spring
constant” is k = mω 2 /4. By the Virial theorem we have
1
hEi = mω 2 (δy)2 ⇒ 2hEi1/2 m−1/2 ω −1 = δy < 2b
4
(d) Using µ > m + hEi and part (c) the desired result follows quickly.
(e) Using parts (b) and (d) yield the sought expression for δ(SBH + Sc ).
(f ) To show δ(SBH + Sc ) > 0 take the derivative of the expression found in (e) and compute it’s
minimum x, which is found to be
xmin = ξ 1/2 (1 + ξ)ln2

Therefore,

1
δ(SBH + Sc ) = xmin + ln(1 − exp(−xmin ))
xmin 2
For a non-relativistic oscillator, ξ 1 so xm 2ln2 and thus xmin > 0 and hence δ(SBH +Sc ) >
0 for all x and ξ.
#1 This problem takes a bit of massaging with the gamma function in higher dimensions. It
helps to prove a couple of lower-dimensional cases and then generalize.
#2 (a) When J = 0, a = 0, reducing to the expected line element.
(b) Using the hint

a a
ds2 = −λ−2/3 (dt − sin2 θdφ + 2 cos2 θdψ)2 + λ1/3 (dr2 + r2 dΩ23 )
r2 r
Note that in the near horizon limit
r r r 2
1 2 3 −2/3 r2
λ≈ ⇒ λ =
r3 R2
Then using constant t, we are left with the desired line element.
(c) From the line element, our basis one forms are
p
σ1 = Rdθ σ2 = R cos θ sin θ(dθ + dψ) σ3 = R2 − (a/R2 )2 (cos2 θdψ − sin2 θdφ)
Then
Z p Z p
A= σ1 ∧ σ2 ∧ σ3 = R 2 R2 − (a/R2 )2 cos θ sin θdθ ∧ dφ ∧ dψ = 2π 2 (r1 r2 r3 )2 − a2
555

A p
S= = 2π Q1 Q2 − J 2
4G5
#3 Upon comparison, we see that R ≈ 1051 cm, several orders of magnitude larger than the size
of the observable universe 1028 cm.
#4 (a) The position of the event horizon occurs when
r2 r02

1+ 2 − 2 =0
R R
Solving for r2 yields
r !
2 R2 4r02
r+ = 1+ 2 −1
2 R
(b) Using TH ≈ r+ /πR2 and noting we are considering a 3-sphere, meaning AHor = 2π 2 r+
3
, the
result quickly follows.
(c) For large r we find
r2
ds2 ≈ (−dt2 + R2 dΩ23 )
R2
The boundary takes R × S 3 because our time direction is given by R and we are dealing with
3-spheres. Comparing to part (b) we find
3
SBH = SY M
4
Bibliography
[1] A. Almheiri, D. Marolf, J. Polchinski, J. Sully. “Black Holes: Complementarity or Firewalls?”,

hep-th/1207.3123.
[2] J. Ambjorn, J. Jurkiewicz, R. Loll, “Causal Dynamical Triangulations and the Quest for
Quantum Gravity”, hep-th/1004.0352.
[3] Andrews, George E. “Generating Functions.” Number Theory. Philadelphia: Saunders, 1971.
160-67. Print.
[4] J. J. Atick and E. Witten, Nucl. Phys. B 310, 291 (1988).
[5] Bain, J. “Black Holes.” Lecture. Black Holes 2. Web. 24 Feb. 2013.
http://ls.poly.edu/ jbain/philrel/philrellectures/14.BlackHoles2.pdf.
[6] Becker, Katrin, Melanie Becker, and John H. Schwarz. String Theory and M-theory: A
Modern Introduction. Cambridge: Cambridge UP, 2007. Print.
[7] J. D. Beckenstein. Phys. Rev. D 7, 8 (1973).
[10] Butkov, Eugene. “The Stretched String. Wave Equation.” Mathematical Physics. New
York: Addison Wesley, 1968. 287-91. Print.
[11] Carroll, Sean. Joe Polchinski on Black Holes, Complementarity, and Firewalls. Discover,
2012.
http://blogs.discovermagazine.com/cosmicvariance/2012/09/27/guest-post-joe-polchinski-on-
blackholes-complementarity-and-firewalls/
[12] Carroll, Sean. Spacetime and Geometry: An Introduction to General Relativity.

San Francisco: Addison Wesley, 2004. Print.
557
[13] B. Carter, “4 Laws of Black Hole Mechanics”. Communications in Mathematical Physics,

31, 161-170 (1973).
[14] B. D. Chowdhury, A. Puhm, “Is Alice Burning or Fuzzing?”, hep-th/1208.2026.
[15] D’Inverno, Ray. Introducing Einstein’s Relativity. Oxford [England: Clarendon, 1992.
Print.
[16] Davies, Paul C., and Julian Brown, eds. Superstrings : A Theory of Everything? N.p.:
Cambridge UP, 1992. Print.
[17] S. Doplicher, K. Fredenhagen, J. Roberts. “The Quantum Structure of Space-time at the

Planck Scale and Quantum Fields”, hep-th/0303037.
[18] Felinska, Ewa. ”Carter-Penrose Diagrams and Black Holes.” Carter-Penrose Diagrams and
Black Holes. University of Wroclaw. 05 Apr. 2010. Lecture.
[19] S. Ferrara, R. Kallosh, A. Strominger. Phys. Rev. D 52, 10 (1995).
[20] Freedman, Daniel Z., and Proeyen Antoine. Van. “Connections and Covariant Derivatives.”
Supergravity. New York: Cambridge UP, 2012. 154-63. Print.
[21] Gasperini, Maurizio. Elements of String Cosmology. Cambridge: Cambridge UP, 2007.
Print.
[22] Giddings, Steven. “Black Holes, Quantum Information, and Unitary Evolution”, arXiv:1201.1037v3
[hep-th].
[23] Greene, B. The Elegant Universe: Superstrings, Hidden Dimensions, and the Quest for the
Ultimate Theory. New York: W.W. Norton, 1999. Print.
[24] Greene, B. The Fabric of the Cosmos: Space, Time, and the Texture of Reality. New York:
A.A. Knopf, 2004. Print.
[25] Green, M., John H. Schwarz, and E. Witten. Superstring Theory. Vol. I. Cambridge:
[26] Green, M., John H. Schwarz, and E. Witten. Superstring Theory. Vol. II. Cambridge:
[27] Gregory, Ruth. “The Gregory-Laflamme Instability.” Black Holes in Higher Dimensions.
Ed. Gary T. Horowitz. Cambridge: Cambridge UP, 2012. 29-33. Print.
[28] Griffiths, David. “Quantum Electrodynamics.” Introduction to Elementary Particles. Wein-

heim: Wiley- VCH, 2008. 225-35. Print.
559
[29] Hamilton, Andrew. “Penrose Diagrams.” Lecture. Penrose Diagrams. National Science
Foundation, 14 Jan. 2013. Web. 24 Feb. 2013. ¡http://jila.colorado.edu/ ajsh/insidebh/penrose.html¿.
[30] Hatfield, Brian F. Quantum Field Theory of Point Particles and Strings. Reading: Addison-
Wesley, 1998. Print.
[31] Hawking, S. W., and Roger Penrose. The Nature of Space and Time. Princeton, NJ:
Princeton UP, 1996. Print.
[32] Hawking, Stephen. A Brief History of Time. New York: Bantam, 1996. Print.
[33] S. W. Hawking. Phys. Rev. D 13, 2 (1976).
[34] S. Hossenfelder, “Minimal Length Scale Scenarios for Quantum Gravity”, hep-th/1203.6191.
[35] Krantz, Steven G. Complex Variables: A Physical Approach with Applications and MAT-
LAB. Boca Raton: Chapman & Hall/CRC, 2008. Print.
[36] Labelle, Patrick. Supersymmetry Demystified. New York: McGraw-Hill, 2010. Print.
[37] Loll, Renate. “Quantum Origins of Space and Time.” Lecture. Quantum Origins of Space
and Time. Perimeter Institute, Waterloo. 27 Dec. 2012. Renate Loll on the Quantum
Origins of Space and Time. Youtube, 24 June 2011. Web. 27 Dec. 2012.
http://www.youtube.com/watch?v=fv2gBjQ8xIo&list=PLF244E3ED0BCDAF23
[38] D. A. Lowe, J. Polchinski, L. Susskind, L. Thorlacius, J. Uglum. Phys. Rev. D 52, 12

(1995).
[39] Maggiore, Michele. A Modern Introduction to Quantum Field Theory. Oxford: Oxford UP,
2005. Print.
[40] A. Maloney, A. Strominger, X. Yin, “S-brane Thermodynamics”, hep-th/0302146.
[41] Markopoulou-Kalamara, Fotini. “Quantum Gravity and the Problem of Time.” Lecture.
FQXI Quantum Gravity. Azores, Ponte Del Gada. 27 Dec. 2012. Fotini Markopoulou-
Kalamara: “Quantum Gravity”, FQXi Azores Conference 2009. FQXI, 8 July 2011. Web.
27 Dec. 2012.
http://www.youtube.com/watch?v=vOrvVTFKDUA&playnext=1&list=PLF244E3ED0BCDAF23&
feature=results main
[42] McIntyre, David H., Corinne A. Manogue, and Janet Tate. Quantum Mechanics: A
Paradigms Approach. Boston: Pearson, 2012. Print.
[43] McMahon, David. “Quantizing the Gravitational Field.” String Theory Demystified. New
York: McGraw Hill, 2009. 8-10. Print.
[44] McMahon, David. Relativity Demystified. New York: McGraw-Hill, 2006. Print.
[45] McMahon, David. “Scalar Fields.” Quantum Field Theory Demystified. New York, NY:
McGraw-Hill, 2008. N. pag. Print.
[46] Mukhanov, V.F, Winitzki, S. “Thermodynamics of Black Holes.”Introduction to Quantum

Effects in Gravity. New York: Cambridge UP, 2010. 120-24. Print.
[47] Myers, Robert C. “Myers-Perry Black Holes.” Black Holes in Higher Dimensions. Ed. Gary
T. Horowitz. Cambridge: Cambridge UP, 2012. 101+. Print.
[48] Y. Okawa, L. Rastelli, B. Zwiebach, “Analytic Solutions for Tachyon Condensation with
General Projectors,”hep-th/0611110.
[49] Peat, F. David. Superstrings and the Search for the Theory of Everything. Chicago:
Contemporary, 1988. N. pag. Print.
[50] Penrose, Roger. “Status of Loop Quantum Gravity.”The Road to Reality: A Complete Guide
to the Laws of the Universe. New York: A.A. Knopf, 2005. 952-55. Print.
[51] Polchinski, Joseph Gerard. String Theory. Vol. I. Cambridge, UK: Cambridge UP, 1998.
Print.
[52] Polchinski, Joseph Gerard. String Theory. Vol. II. Cambridge, UK: Cambridge UP, 1998.
Print.
[53] J. Polchinski, A. Strominger. Phys. Rev. D 50, 12 (1994).
[54] R. Gambini, J. Pullin. “Spin Foams”. A First Course in Loop Quantum Gravity. New
York: Oxford UP, 2011. 149-56. Print.
[55] C. Rovelli, “A Critical Look at Strings”, hep-th/1108.0868.
[56] Rovelli, Carlo. “General Ideas and the Heuristic Picture”. Quantum Gravity. Cambridge,
UK: Cambridge UP, 2004. 3-33. Print.
[57] B. Sathiapalan, Phys. Rev. D 35, 3277 (1987); Y. I. Kogan, JETP Lett. 45, 709 (1987)
[Pisma Zh. Eksp. Teor. Fiz. 45, 556 (1987)]; K. H. O’Brien and C. I. Tan, Phys. Rev. D
36, 1184 (1987).
[58] Schroeder, Daniel V. “Boltzmann Statistics”. An Introduction to Thermal Physics. San

Francisco, CA: Addison Wesley, 2000. 220+. Print.
[59] Shankar, Ramamurti. ”The Continuity Equation for Probability.” Principles of Quantum
Mechanics. 2nd ed. New York: Plenum, 1994. 164-67. Print.
561
[60] E. Silverstein, “Singularities and Closed String Tachyons”, hep-th/0602230.
[61] A. Strominger and C. Vafa, hep-th/9601029 (1996).
[62] Susskind, Leonard, and James Lindesay. “The Laws of Nature”. An Introduction to
Black Holes, Information and the String Theory Revolution: The Holographic Universe. Hackensack,
NJ: World Scientific, 2005. 69+. Print.
[63] Susskind, Leonard. The Black Hole War. New York: Little, Brown and, 2008. Print.
[64] L. Susskind, “Singularities, Firewalls, and Complementarity”, hep-th/1208.3445.
[65] L. Susskind. “String Theory and the Principle of Black Hole Complementarity”, Phys. Lett.
71, 15 (1993).
[66] Susskind, Leonard. “The Theoretical Minimum.”Lecture. The Theoretical Minimum. Stanford
University. Summer 2011. The Complete Leonard Susskind Lectures. Ted Young, 11 Jan.
2011. Web. 1 Jan. 2013.
http://tedyoung.me/2011/01/22/leonard-susskind-lectures/
[67] L. Susskind, L. Thorlacius, J. Uglum. Phys. Rev. D 48, 8 (1993).
[68] Szekeres, Peter. A Course in Modern Mathematical Physics: Groups, Hilbert Space, and
Differential Geometry. Cambridge, UK: Cambridge UP, 2004. Print.
[69] t’Hooft, Gerard. “Black Hole Interpretation of String Theory”. Nuclear Phys. B 335 1
(1990).
[70] t’Hooft, Gerard. “Introduction to Lie Groups in Physics.” Lecture. Introduction to Lie
Groups in Physics. Institute for Theoretical Physics Utrecht University. Feb. 2013, Web.
[71] Wannier, Gregory H. “Statistical Justification of the Second Law.” Statistical Physics. New
York: J. Wiley, 1966. 83-88. Print.
[72] Zwiebach, Barton. A First Course in String Theory. New York: Cambridge UP, 2004.
Print.
Index
2-dimensional conformal group, 234 Calabi-Yau manifold, 307

calculus of residues, 181
Schrödinger picture, 148 canonical ensemble, 387
canonical quantization, 163
abelian group, 226 canonical transformations, 278
absolute space, 33 Casimir operator, 339
absolute time, 33 Cauchy’s integral formula, 237
active gravitational mass, 446 Cauchy’s integral theorem, 244
adiabatic theorem, 489
Causal Dynamical Triangulation, 508
admissible states, 207
central charge, 251
AdS/CFT correspondence, 500
central extension, 184
affine connection, 410
Chan-Paton indices, 265
Aharonov-Bohm effect, 295
charge conjugation, 321
algebra, 228
chiral, 230
analytic, 232
chiral multiplet, 344
analytic continuation, 180
chirality, 316
anthropic, 310
Christoffel symbols, 410
anti-D-branes, 274
Clifford algebra, 358
anti-de Sitter model, 454
closed, 302
antiholomorphic, 232
Ashtekar variables, 507 closed bosonic string, 192
auxiliary field, 337 closed string state space, 286
closure, 225
Bianchi identity, 419, 447 coarse-grained entropy, 494
Big Bang, 456 coherence, 391
binary operation, 225 commutation coefficients, 432
black brane, 488 commutator, 408
black hole, 463 compact, 301
black hole complementarity, 494, 496 Compactification, 23
black string, 488 compactification, 200
Boltzmann’s constant, 387 complex manifolds, 302
Born-Infeld theory, 300 complex Riemannian manifold, 302
bosonic string theory, 200 component fields, 365
BRST charge, 248 condensate, 196
BRST current, 251 conformal compactification, 468
BRST invariant, 249 conformal dimension, 241
BRST operator, 248 conformal field, 241
BRST quantization, 203, 225, 247 conformal field theory, 225
562
INDEX 563
conformal gauge, 358 Einstein-Cartan theory, 412

conformal group, 232 Einstein-Hilbert action, 448
conformal metric, 216 Einstein-Rosen bridge, 467
conformal transformation, 229 electroweak theory, 265
conformal vacuum, 247 energy-momentum tensor, 217, 236, 362, 443
conformal weights, 241 entanglement entropy, 494
conformally flat, 216 equivalence principle, 407, 495
congruence, 420 ergosphere, 475
coordinate basis, 408 ergosurface, 475
coordinate patches, 300 Euclidean quantum field theory, 230
cosmic censorship hypothesis, 463 Euler characteristic, 301
cosmic strings, 26 event horizon, 463
cosmological constant, 447 exterior derivative, 302, 426
covariant derivative, 409, 410 exterior product, 426
covering, 300 extra dimensions, 191
covering space, 279 extrinsic curvature, 416
CPT theorem, 322
currents, 330 Faddeev-Popov ghosts, 251
curvature 1-forms, 434 Faddeev-Popov procedure, 221
curvature tensor, 414 fermion number, 373
fermionic oscillators, 396
D-brane, 26 Feynman Diagrams, 19
D-brane decay, 194, 273 Fierz identity, 365
D-branes, 290 fine-grained entropy, 494
D-instanton, 272 flux compactifications, 310
de Sitter model, 454 Fock space, 193
decoherence, 391 four-vectors, 35
density of states, 404 frame dragging, 475
density operator, 389 Friedmann equations, 454
diffeomorphisms, 232, 302 functional derivative, 217
differentiable manifold, 407 fundamental domain, 303
differential forms, 426 future null infinity, 470
dilaton, 197 future time-like infinity, 470
Dirac algebra, 358 fuzzball complementarity, 497
Dirac equation, 315 fuzzballs, 497
Dirac spinor, 315
directional covariant derivative, 417 G-parity, 373
Dirichlet boundary conditions, 258 Galilean relativity, 29
double strike, 305 gamma function, 180
Dp-branes, 257 gauge transformations, 293
duality symmetries, 277 generalized second law of thermodynamics, 481
dust, 444 generating functions, 375
genus, 301
Eddington-Finklestein coordinates, 462 geodesic, 417
effective field theory, 314 geodesic deviation, 420
Ehrenfest’s Theorem, 163 ghost fields, 248
Einstein static universe, 470 ghost momenta, 248
564 INDEX
graded Lie algebra, 328 Kalb-Ramond, 197

Grassmann coordinates, 347, 364 Kalb-Ramond field, 271
Grassmann differentiation, 518 Kappa symmetry, 380
Grassmann integration, 519 Kerr metric, 474
Grassmann numbers, 333 Killing vectors, 415, 422
Grassmann variables, 320, 517 Klein-Gordon Lagrangian, 273
graviton, 20 Kruskal-Szekeres coordinates, 466
Green-Schwarz formalism, 357, 380
group, 225 landscape, 277, 310
group dimension, 226 Laurent expansion, 237
GSO projection, 375, 378 left-chiral spinor, 316, 333
level matching, 188
Hagedorn temperature, 399, 405 Lie algebra, 159, 184, 225, 409
Hawking effect, 477 Lie bracket, 408
Hawking radiation, 477, 478, 494 Lie derivative, 414
Hawking’s Area theorem, 463 Lie groups, 225
Heisenberg picture, 148 light-cone, 34
helicity, 341 light-cone gauge, 148
Helmholtz free energy, 386 line element, 302
Hermitian manifold, 302 loop integral, 19
heterotic superstring theory, 381 Loop Quantum Gravity, 456, 507
hiearchy problem, 22, 314 Lorentz charges, 158, 191
Higg’s field, 270 Lorentz covariant quantization, 203
Hodge duality, 428 Lorentz current, 219
Hodge star operator, 428 Lorentz generators, 159
holography, 498 Lorentz group, 228
holomorphic, 232 Lorentz transformation, 218
holonomic basis, 432 Lorentz transformations, 32, 316
holonomies, 293
holonomy, 296 M-theory, 191, 300, 381
homeomorphic, 280 macrostate, 386
homeomorphism, 301 Majorana spinors, 323
homomorphism, 226 manifold, 300
hyperplanes, 257 mass-squared operator, 193, 206
matrix representation, 226
inertial mass, 446 maximally symmetric, 454
inflation, 310 Maxwell relation, 386
information loss paradox, 482 Maxwell’s equations, 277
internal energy, 386 metric, 302
internal symmetry, 330 metric compatiabilty, 412
intrinsic curvature, 416 microstate, 387
invariant subspace, 226 Minimal Supersymmetric Standard Model, 342
irreducible, 226 Minkowski space, 35
mircocanonical ensemble, 386
Jacobi identity, 184 mirror symmetry, 306, 308
mixed state, 390
Kähler manifold, 302 moduli, 288
INDEX 565
moduli space, 288 pentachoron, 508

moduli stabilization, 310 perfect fluid, 444
multiverse, 266 photino, 344
Myers-Perry black holes, 488 physical state, 206
Planck length, 21
N=1 supersymmetry, 344 Planck mass, 21
naked singularities, 463 Planck time, 21
naked singularity, 474 Poincaré group, 228
Nambu-Goto action, 204 Poincaré transformations, 218
negative norm, 206 Poisson bracket, 164
Neumann boundary conditions, 258 Polyakov action, 214, 358
Neveu-Schwarz sector, 368 positive definite, 302
Newton’s bucket, 505 primary field, 241, 247
nilpotent, 248 principle of equivalence, 446
no-cloning principle, 495 propagator, 238
no-hair theorem, 472 pseudo-Riemannian manifold, 302
non-abelian, 226 pure state, 389
non-holonomic basis, 432
noncommutative geometry, 264 quantizing the relativistic point particle, 147
normal-ordered, 178 Quantum Chromodynamics, 18
null state, 250 quantum information theory, 391
null tetrads, 474
number operator, 188 Ramond sector, 367
number theory, 393 Ramond-Neveu-Schwarz formalism, 357
reducible, 226
occupation numbers, 393 Regge calculus, 507
operator product expansion, 240 Reissner–Nordström, 472
orbifolds, 303 renormalization, 20
orientifolds, 306 representation, 226
orthogonal group, 228 Ricci scalar, 419, 452
Ricci tensor, 419, 452
p-branes, 257 Riemann surface, 220
Page time, 497 Riemann tensor, 414
parallel D-branes, 292 right-chiral spinor, 316
parallel Dp-branes, 262 Rindler observer, 465
parallel transport, 297, 416, 417 Rindler space-time, 465
parity, 322 RNS superstrings, 378
partially mixed state, 391
partition, 393, 398 s-branes, 405
partition function, 388 S-duality, 300
passive gravitational mass, 446 scale invariant, 230
past null infinity, 470 Schwarzschild line element, 436
past time-like infinity, 470 Schwarzschild radius, 459
path integrals, 220 Schwarzschild solution, 450, 459, 485
Pauli matrices, 315 sectors, 262
Pauli-Lubanski, 339 selectron, 344
Penrose diagram, 468 self-dual radius, 288
566 INDEX
simple harmonic oscillator, 278 tachyons, 33

simplexes, 507 tangent bundle, 37
simply-connected, 279 tangent space, 407
singularity theorems, 463 tetrad, 433
space-time interval, 35 Theodor Kaluza, 24
sparticles, 314 thermal entropy, 494
spatial infinity, 470 time-ordered, 238
special conformal transformations, 233 topological property, 280
special orthogonal group, 228 torsion tensor, 412
special relativity, 30, 33 tortoise coordinate, 462
special unitary group, 229 translations, 233
spin network, 505 transverse Virasoro operators, 178
spinfoam, 507 twist, 198
spurious state, 208 twistor, 506
Standard Model, 27 twistor networks, 506
Standard model, 314 twistor space, 506
state space, 193, 195 type IIA superstring theory, 271, 380
Stoke’s theorem, 296 type IIB superstring theory, 271, 380
stress, 444
stretched horizon, 491 U-duality, 300
stretched strings, 262 unitary groups, 295
string charge, 270 unoriented strings, 198
string cosmology, 194, 275 Unruh effect, 477
string field theory, 195
vacuum energy, 447
string geometry, 277 vacuum equations, 451
string length, 240 van de Waerden, 345
structure constants, 227 van der Waerden notation, 511
structure equations, 434 vector field, 408
super-Virasoro algebra, 370 Verma module, 247
super-Virasoro operators, 369 vielbein, 433
supercharges, 332 vierbein, 433
supercovariant derivatives, 354 Virasoro algebra, 183, 184, 242
supercurrent, 362 Virasoro descendents, 207
superfield, 348 Virasoro operator, 205
supergravity, 19, 314, 491 volume form, 429
supermultiplets, 343 von Neumann entropy, 391
superpartner, 314
superspace, 331, 345, 347 wedge product, 426
superstring theory, 313 Weyl equation, 337
superstrings, 357 Weyl spinor, 316
supersymmetry, 18, 313 Weyl spinors, 511
supersymmetry breaking, 342 Weyl transformation, 216
surface gravity, 477 white hole, 466
Wick expansion, 238
T-duality, 277, 288 Wick rotation, 230
tachyon, 194, 273 Wick’s theorem, 241
INDEX 567
Wilson line, 296, 297

winding number, 279
Witt algebra, 183
world-volume, 257
worldline, 33
wormhole, 467
Yang-Mills theories, 265

Yang-Mills theory, 297
zeta functions, 180

A Detailed Introduction To String Theory

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Detailed Introduction To String Theory

Uploaded by

Copyright:

Available Formats

A Detailed Introduction to String Theory

I Fundamentals and the Bosonic String 15

2 Special Relativity and Light Cone Coordinates 29

3 Non-Relativistic Strings and the Relativistic Point Particle 51

4 Classical Relativistic Strings 73

5 Relativistic Strings and Mode Expansions 93

6 Charges, Currents, and Symmetries 107

7 A Crash Course on Quantum Field Theory 117

8 Quantizing the Relativistic Point Particle 147

9 Light-Cone Quantization of the String 165

9.5 Lorentz Generators and the Dimension of Space-Time . . . . . . . . . . . . . . . . . 190

10 Lorentz Covariant Quantization 203

II Advanced Topics and Superstrings 223

13 T-Duality, Symmetries, and Compactification 277

13.11 Calabi-Yau Manifolds and Mirror Symmetry . . . . . . . . . . . . . . . . . . . . . . 307

14 A Crash Course in Supersymmetry 313

15 An Introduction to Superstrings 357

III General Relativity and String Theory 383

16 The Thermodynamics of Strings 385

17 Elements of Differential Geometry 407

18 A Crash Course on General Relativity 443

19 Black Holes in General Relativity 459

20 Black Holes in String Theory 485

21 Alternative Approaches toward Quantum Gravity 503

A Van der Waerden Notation for Weyl Spinors 511

B Grassmann Variables and Grassmann Integration 517

C Solutions to Exercises 521

Fundamentals and the Bosonic

1.1 A Brief History of Modern Physics and the Mark of

1.2 A Qualitative Look at Renormalization Theory

1.3 Fundamental Units and the Planckian System

1.4 Planck Length in Arbitrary Dimensions

∇2 V (4) = 4πGρm (1.11)

∇2 V (D) = 4πG(D) ρm (1.12)

Multiplying by one in an interesting way gives

The above statement can be generalized to D space-time dimensions (Zwiebach, 63):

1.5 Compactification and Large Extra Dimensions

P1 ∼ P2 ←→ x(P1 ) = x(P2 ) + 2πRn, n ∈ Z (1.17)

If we consider a 5-dimensional universe, the circumference of the compact dimension is approx-

1.6 Verifying String Theory

Special Relativity and Light Cone

2.1 Special Relativity and Einstein’s Postulates

x0 = Ax + Bct ct0 = Cx + Dct (2.4)

x0 = x cosh(φ) − ct sinh(φ) ct0 = −x sinh(φ) + ct cosh(φ) (2.7)

x0 = 0 = x cosh(φ) − ct sinh(φ) = vt cosh(φ) − ct sinh(φ) = t(v cosh(φ) − c sinh(φ))

Focusing on the second expression in (2.7), we find that

x0 = cosh(φ)(x − vt) (2.10)

2.2 Space and Time, and Space-time

(∆S)2 = −(c∆t)2 + (∆x)2 + (∆y)2 + (∆z)2 (2.15)

(∆S)2 = −(c∆t0 )2 + (∆x0 )2 + (∆y 0 )2 + (∆z 0 )2 (2.16)

2.3 A Quick Glance at Four-Vectors

the famous Einstein relation for energy.

2.4 Vectors, Dual Vectors, and Tensors

ω(aU + bV ) = aω(U ) + bω(V ) (2.30)

(aω + bη)(V ) = aω(V ) + bη(V ) (2.31)

T : Tp∗ × ... × Tp∗ × Tp × ... × Tp → R (2.36)

T ⊗ S(ω 1 , ..., ω k , ..., ω k+m , V 1 , ..., V ` , ...V `+n )

eµ1 ⊗ ... ⊗ eµk ⊗ θν1 ... ⊗ θν` (2.38)

T = T µ1 ...µkν1 ...ν` eµ1 ⊗ ... ⊗ eµk ⊗ θν1 ... ⊗ θν` (2.39)

Rijk = S ijk + T ijk

2.5 Coordinate Transformations

x = r cos θ y = r sin θ (2.47)

er = Λxr ex + Λy r ey = cos θex + sin θey

The transformations A ~0 → A ~ + ∇,

where = (x) is a function of spacetime coordinates, and where we write A0 = φ. Furthermore,