You are on page 1of 446
p elie arieran Field /Iheory From Operators to Path Integrals Te eur) een co) Compactified 2D space Kerson Huang QUANTUM FIELD THEORY From Operators to Path Integrals Kerson Huang Massachusetts Institute of Technology Cambridge, Massachusetts A Wiley-Interscience Publication JOHN WILEY & SONS, INC. New York = Chichester = Weinheim = Brisbane = Singapore = Toronto This text is printed on acid-free paper. @ Copyright © 1998 by John Wiley & Sons, Inc. All rights reserved. Published simultaneously in Canada. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate pet-copy fee to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA (01923, (978) 750-8400, fax (978) 750-4744, Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 605 Third Avenue, New York, NY 10158-0012, (212) 850-601, fax (212) 850-6008, E-Mail: PERMREQ @ WILEY.COM. Library of Congress Cataloging in Publication Data: Huang, Kerson, 1928— Quantum field theory / by Kerson Huang. pcm Includes bibliographical references and index. ISBN 0-471-14120-8 (cloth : alk. paper) 1. Quantum field theory. I. Title QC174.45,H82 1998 530.143—de 21 98-13437 Printed in the United States of America 10987654321 To Rosemary Contents Preface xv Acknowledgment xvii 1. Introducing Quantum Fields 1 1.1, The Classical String, 1 1.2. The Quantum String, 5 1.3. Second Quantization, 7 1.4, Creation and Annihilation Operators, 10 1.5. Bose and Fermi Statistics, 12 Problems, 13 References, 16 . Scalar Fields 17 2.1. Klein-Gordon Equation, 17 Real Scalar Field, 18 Energy and Momentum, 20 Particle Spectrum, 22 Continuum Normalization, 22 Complex Scalar Field, 24 Charge and Antiparticle, 25 .8. Microcausality, 27 2.9. The Feynman Propagator, 28 2.10. The Wave Functional, 31 2.11. Functional Operations, 32 2.12. Vacuum Wave Functional, 33 2.13. The * Theory, 35 Problems, 37 nv vii oe 4 wv a . Dirac Equation Contents Relativistic Fields 40 3.1. Lorentz Transformations, 40 Minimal Representation: SL(2C), 43 3.3. The Poincaré Group, 44 3.4. Scalar, Vector, and Spinor Fields, 46 Relativistic Quantum Fields, 48 3.6. One-Particle States, 50 Problems, 52 Reference, 53 Canonical Formalism 54 4.1. Principle of Stationary Action, 54 4.2. Noether’s Theorem, 57 4.3. Translational Invariance, 58 4.4. Lorentz Invariance, 61 4.5. Symmetrized Energy-Momentum Tensor, 63 4.6. Gauge Invariance, 63 Problems, 65 Reference, 67 Electromagnetic Field 68 5.1. Maxwell’s Equations, 68 5.2. Covariance of the Classical Theory, 70 5.3. Canonical Formalism, 72 5.4. Quantization in Coulomb Gauge, 75 5.5. Spin Angular Momentum, 78 5.6. Intrinsic Parity, 80 5.7. Transverse Propagator, 81 5.8. Vacuum Fluctuations, 82 5.9. The Casimir Effect, 85 5.10. The Gauge Principle, 88 Problems, 90 References, 93 94 6.1. Dirac Algebra, 94 6.2. Wave Functions and Current Density, 98 6.3. Plane Waves, 99 6.4. Lorentz Transformations, 102 6.5. Interpretation of Dirac Matrices, 106 6.6. External Electromagnetic Field, 107 6.7. Nonrelativistic Limit, 109 6.8. Thomas Precession, 111 6.9. Hole Theory, 114 7 - a 6.10. Charge Conjugation, 117 6.11. Massless Particles, 118 Problems, 119 References, 122 The Dirac Field 7.1. Quantization of the Dirac Field, 123 7.2. Feynman Propagator, 126 7.3. Normal Ordering, 128 7.4. Electromagnetic Interactions, 129 7.5. Isospin, 130 7.6. Parity, 132 7.7. Charge Conjugation, 133 7.8. Time Reversal, 134 Problems, 136 Reference, 137 . Dynamics of Interacting Fields 8.1. Time Evolution, 138 8.2. Interaction Picture, 139 8.3. Adiabatic Switching, 142 8.4. Correlation Functions in the Interaction Picture, 144 8.5. S Matrix and Scattering, 147 8.6. Scattering Cross Section, 148 8.7. Potential Scattering, 151 8.8. Adiabatic Theorem, 154 Problems, 158 References, 159 Feynman Graphs 9.1. Perturbation Theory, 160 9.2. Time-Ordered and Normal Products, 161 9.3. Wick’s Theorem, 162 9.4. Feynman Rules for Scalar Theory, 165 9.5. Types of Feynman Graphs, 171 9.5.1. Vacuum Graph, 171 9.5.2. Self-Energy Graph, 171 9.5.3. Connected Graph, 172 9.6. Wick Rotation, 172 9.7. Regularization Schemes, 173 9.8. Linked-Cluster Theorem, 175 9.9. Vacuum Graphs, 175 Problems, 176 Reference, 181 Contents ix 123 138 160 x 10. 12. ae Contents Vacuum Correlation Functions 182 10.1. Feynman Rules, 182 10.2. Reduction Formula, 185 10.3. The Generating Functional, 188 10.4. Connected Correlation Functions, 189 10.5. Lehmann Representation, 190 10.6. Dyson-Schwinger Equations, 193 10.7. Bound States, 197 10.8. Bethe-Salpeter Equation, 200 Problems, 201 References, 203 |. Quantum Electrodynamics 204 11.1, Interaction Hamiltonian, 204 11.2. Photon Propagator, 207 11.3. Feynman Graphs, 210 11.4. Feynman Rules, 215 11.5. Properties of Feynman Graphs, 217 Problems, 219 References, 219 Processes in Quantum Electrodynamics 220 12.1, Compton Scattering, 220 . Electromagnetic Form Factors, 224 . Anomalous Magnetic Moment, 229 12.4. Charge Distribution, 232 Problems, 234 References, 235 Perturbative Renormalization 236 13.1. Primitive Divergences in QED, 236 13.2. Electron Self-Energy, 237 Vacuum Polarization, 243 Running Coupling Constant, 247 Full Vertex, 248 Ward Identity, 249 Renormalization to Second Order, 251 Renormalization to All Orders, 252 9. Callan-Symanzik Equation, 257 13.10. Triviality, 259 Problems, 260 References, 263 Contents xi 14, Path Integrals 264 14.1. Path Integrals in Quantum Mechanics, 264 14.2. Imaginary Time, 268 143, Path Integrals in Quantum Field Theory, 269 144, Euclidean Space-Time, 271 14.5. Vacuum Amplitudes, 272 14.6. Statistical Mechanics, 274 14.7. Gaussian Integrals, 276 14,8, Perturbation Theory, 280 14.9. The Loop Expansion, 283 14.10. Boson and Fermion Loops, 284 14.11. Grassmann Variables, 287 ae 16. Problems, 290 References, 293 Broken Symmetry 294 (Ea Be 15.2. 15.3. 15.4. 15.5. 15.6. 15.7. 15.8 Why Broken Symmetry, 294 Ferromagnetism, 296 Spin Waves, 300 Breaking Gauge Invariance, 303 Superfluidity, 306 Ginsburg-Landau Theory, 310 Effective Action, 312 Effective Potential, 314 Problems, 315 References, 318 Renormalization 320 16.1. 16.2. 16.3. 16.4. 16.5. 16.6. 16.7. 16.8. 16.9. The Cutoff as Scale Parameter, 320 Momentum Space RG, 324 16.2.1. Designating Fast and Slow Modes, 325 16.2.2, Coarse Graining, 326 16.2.3. Rescaling, 326 Real Space RG, 327 163.1. Making Blocks, 328 16.3.2. Coarse Graining, 329 16.3.3. Rescaling, 330 Renormalization of Correlation Functions, 330 Relevant and Irrelevant Parameters, 331 The Free Field, 332 IR Fixed Point and Phase Transition, 334 Crossover, 336 Relation with Perturbative Renormalization, 337 xii Contents 16.10. Why Correct Theories are Beautiful, 339 Problems, 341 References, 342 17. The Gaussian Fixed Point 343 17.1. Stability of the Free Field, 343 17.2. General Scalar Field, 344 17.3. Feynman Graphs, 345 17.4, Wegner-Houghton Formula, 347 17.5. Renormalized Couplings, 350 17.6. The RG Matrix, 352 17.7. Nontriviality and Asymptotic Freedom, 356 17.8. The Case d= 2, 357 Problems, 359 References, 359 18. In Two Dimensions 360 18.1. Absence of Long-Range Order, 360 18.2. Topological Order, 361 18.3. XY Model, 363 18.4. Kosterlitz—Thouless Transition, 367 18.5. Vortex Model, 368 18.6, 2D Superfluidity, 370 18.7. RG Trajectories, 372 18.8. Universal Jump of Superfluid Density, 376 Problems, 378 References, 378 19. Topological Excitations 380 19.1. Topological Soliton, 380 19.2, Instanton and Tunneling, 383 19.3. Depinning of Charge Density Waves, 385 19.4, Nonlinear Sigma Model, 389 19.5. The Skyrmion, 391 19.6. The Hopf Invariant, 395 19.7. Fractional Spin, 398 19.8. Monopoles, Vortices, and Anomalies, 401 Problems, 403 References, 405 A. Background Material 406 A.1. Notation, 406 A.2. Classical Mechanics, 408 A.3. Quantum Mechanics, 408 B. Linear Response References, 418 Index Contents xiii 412 419 Preface Quantum field theory, the quantum mechanics of continuous systems, arose at the beginning of the quantum era, in the problem of blackbody radiation. It became ful- ly developed in quantum electrodynamics, the most successful theory in physics. Since that time, it has been united with statistical mechanics through Feynman's path integral, and its domain has been expanded to cover particle physics, con- densed-matter physics, astrophysics, and wherever path integrals are spoken. This book is a textbook on the subject, aimed at readers conversant with what is usually called “advanced quantum mechanics,” the equivalent of a first-year gradu- ate course. Previous exposure to the Dirac equation and “second quantization” would be very helpful, but not absolutely necessary. The mathematical level is not higher than what is required in advanced quantum mechanics; but a degree of matu- rity is assumed. In physics, a continuous system is one that appears to be so at long wavelengths or low frequencies. To model it as mathematically continuous, one runs into diffi- culties, in that the high-frequency modes often give rise to infinities. The usual pro- cedure is to start with a discrete version, by discarding the high-frequency modes beyond some cutoff, and then try to approach the continuum limit, through a process called renormalization. Renormalization is a relatively new concept, but its workings were already evi- dent in classical physics. At the beginning of the atomic era, Boltzmann noted that classical equipartition of energy presents conceptual difficulties, when one serious- ly considers the atomic structure of matter. Since atoms are expected to contain smaller subunits, which in turn should composed of even smaller subunits, and so ad infinitum, and each degree of freedom contributes equally to the thermal energy of a substance, the specific heat of matter would be infinite.. The origin of this di- vergence lies in the extrapolation of known physical laws into the high-frequency domain, a characteristic shared by the infinities in quantum field theory. Boltzmann's “paradox,” however, matters not a whit when it comes to practical calculations, as evidenced by the great success of classical physics. The reason is that most equations of macroscopic physics, such as those in thermodynamics and xv Xvi Preface hydrodynamics, make no explicit reference to atoms, but depend on coefficients like the specific heat, which can be obtained from experiments. From a modern per- spective, we say that such theories are “renormalizable,” in that the microstructure can be absorbed into measurable quantities. One goal of this book is to explain what renormalization is, how it works, and what makes some systems appear “renormalizable” and others not. We follow the historical route, discovering it in quantum electrodynamics through necessity, and then realizing its physical meaning through Wilson’s path-integral formulation. This book, then, starts with a thorough introduction to the usual operator for- malism, including Feynman graphs, from Chapters 1-10. This is followed by Chap- ters 11~14 on quantum electrodynamics, which illustrates how to do practical calcu- lations, and includes a complete discussion of perturbative renormalization. The last part, Chapters 15-19, introduces the Feynman path integral, and discusses “mod- ern” subjects, including the physical approach to renormalization, spontaneous symmetry breaking, and topological excitations. I have entirely omitted non- Abelian gauge fields and the standard model of particle physics, because these sub- jects are discussed in another book: K. Huang, Quarks, Leptons, and Gauge Field, 2nd ed. (World Scientific, Singapore, 1992). Ihave chosen to introduce path integrals only after the canonical approach is fully developed and applied. Others might want them discussed earlier. To accom- modate different tastes, I have tried to make each chapter self-contained in as much as possible, so that a knowledgeable reader can pick and skip. There is definitely a change in flavor when quantum field theory is conveyed through the path integral. Apart from the union with statistical mechanics, which immeasurably enriches the subject, it liberates our imagination by making it possi- ble to contemplate virtual but fantastic deformations, such as altering the structure of space-time. I am reminded of the classification of things as “gray” or “green” by Freeman Dyson, in his book Disturbing the Universe (Harper & Row, New York, 1979). He classified physics gray (and I suppose that included quantum field theo- ry,) as opposed to things green, such as poems and horse manure. Ina private letter dated August 3, 1983, Dyson wrote, “Everyone has to make his own choice of what to call gray and green. I took my choice from Goethe: Grau, tenerer freund, ist alle Theorie, Und griin des Lebens Goldner Baum. Dear friend, all theory is grey, ‘And green is the golden tree of life. I must admit that Hilbert space does seem a bit dreary at times; but, with Feynman’s path integral, quantum field theory has surely turned green. Kerson Huanc December, 1997 Marblehead, Massachusetts Acknowledgments T wish to thank my colleagues and students for helping me to learn the stuff in this book over the years, in particular Herman Feshbach, from whom I took the first course on the subject. I thank Jeffrey Goldstone, Roman Jackiw, and Ken Johnson, who can usually be relied on to provide correct and simple answers to difficult ques- tions; and Patrick Lee, who taught a course with me on the subject, thereby broad- ening my horizon. Last but not least, I thank my editor Greg Franklin for his under- standing and support. K.H. xvii QUANTUM FIELD THEORY CHAPTER ONE + Introducing Quantum Fields 1.1 THE CLASSICAL STRING We obtain a quantum field by quantizing a classical field, of which the simplest ex- ample is the classical string. To be on firm mathematical grounds, we define the latter as the long-wavelength limit of a discrete chain. Consider N + 2 masses de- scribed by the classical Lagrangian NA Ua. d= > [FB ZO- su? a FO where m is the mass and « a force constant. The coordinate q,(¢) represents the later- al displacement of the jth mass along a one-dimensional chain. We impose fixed- endpoint boundary conditions, by setting Gol!) = Gnei() = 0 (1.2) The equations of motion for the N remaining movable masses are then mg,—Kqu-24+9,-)=0 CG +N) (3) The normal modes have the form g(t) = cos(wt) sin( jp) (4) To satisfy the boundary conditions, choose p to have one of the N possible values ™ N+1 Pa= (m=1,...,N) (1.5) 2 Introducing Quantum Fields ‘Substituting this into the equations of motion, we obtain N independent normal fre- quencies «,: (n=1-+-N) (1.6) wt= asin where: a7 This is a cutoff frequency, for the modes with n > N merely repeat the lower ones. For N= 4, for example, the independent modes correspond to n = 1,2,3,4. The case n= Sis trivial, since p = 7, and hence qt) = 0 by (1.4). The case n = 6 is the same as that for n= 4, since wg = a,, and sin( jpg) = -sin(jp,). When Nis large, and we are not interested in the behavior near the endpoints, it is convenient to use periodic boundary conditions: en) = Ghd) (1.8) In this case the normal modes are gf)= er“ (19) For N even, the boundary conditions can be satisfied by putting 2nn N = (n=081,...# 3) (1.10) The corresponding normal frequencies are o2= a3sin( 7) (1.11) Compared to the fixed-end case, the spacing between normal frequencies is now doubled; but each frequency is twofold degenerate, and the number of normal modes remains the same. A comparison of the two cases for N = 8 is shown in Fig, 1.1. The equilibrium distance a between masses does not explicitly appear in the Lagrangian; it merely supplies a length scale for physical distances. For example, it appears in the definition of the distance of a mass from an end of the chain: .N) (1.12) The total length of the chain is then defined as 1.1 The Classical String = 3 Fixed-end Periodic @p ®o o4 N -Ne2 0 ne Mode number Mode number Figure 1.1 Normal modes of the classical chain for fixed-end and periodic boundary conditions. R=Na (1.13) In the continuum limit a>0 N+@ (R=Na fixed) (1.4) the discrete chain approaches a continuous string, and the coordinate approaches a classical field defined by qx, 1) = gi) (1.15) The Lagrangian in the continuum limit can be obtained by making the replacements , 1) 2 mia af A] 1 ad 7 fe (1.16) Assuming that the mass density p and string tension «approach finite limits m p= (7) o= Ka (1.18) we obtain the limit Lagrangian 4 Introducing Quantum Fields Lolly (3587) am This leads to the equation of motion F(x, t) 1 Pq@s _ a et ae 0 (1.20) which is a wave equation, with propagation velocity (1.21) The general solutions are the real and imaginary parts of ge, 1) = eteen (1.22) with a linear dispersion law w=ck (1.23) For fixed-end boundary conditions q(0, t) = q(R, )=0 (1.24) the normal modes of the continuous string are Gn(% t) = COS(pt) sin(k,x) (1.25) with w, = ck,, and (1.26) The normal frequencies «, are the same as those for the discrete chain for n/N < 1, as given in (1.6). However, the number of modes of the continuum string is infinite, and only the first N modes have correspondence with those of the discrete string. This is illustrated in Fig. 1.2 for N = 4. Thus, there is a cutoff frequency oy= a (1.27) This is of the same order, but not same as the maximum frequency defined earlier, y= 2cla, for «, is based on a linear dispersion law. The continuum model is an ac- curate representation of the discrete chain only for @ < a. 1.2 The Quantum String 5 (aoe AS LAA, tS B Lee 8 Soe Figure 1.2. Normal modes of a discrete chain of four masses, compared with those of a continuous string. The former repeat themselves after the first four modes. (After J.C. Slater and N. H. Frank, Me- chanics, McGraw-Hill, New York, 1947.) For periodic boundary conditions (0, 1) = 4(R, t) (1.28) the allowed wave numbers are ky = (n=0,41,42,...) (1.29) We obtain the cutoff frequency «, by setting n = N/2. The high-frequency cutoff is a theoretical necessity. Without it, the specific heat of the string will diverge, since each normal mode contributes an amount kT. The value of the cutoff cannot be determined from the long-wavelength effective theory, because only the combination c = aa/7 occurs. Absorbing the cutoff into measurable parameters, as done in (1.17), is called renormalization. A theory for which this can be done is said to be renormalizable. Nonrenormalizable systems exhibit behavior that is sensitive to details on an atomic scale, Such behavior would appear to be random on a macroscopic scale, as in the propagation of cracks in materials, and the nucleation of raindrops. 1.2. THE QUANTUM STRING We now quantize the classical chain, to obtain a quantum field in the continuum limit. The Hamiltonian of the classical discrete chain is given by 6 Introducing Quantum Fields pp Hp.) [FE + Fao? (1.30) where pj = md. The system can be quantized by replacing p; and q, by Hermitian operators satisfying the commutation relations [Pp 4] = “15 (3th We impose periodic boundary conditions, and expand these operators in Fourier se- ries: 1 om - mui D> DN Da Oe 1 8 aoe >, Perms (1.32) where P,, and Q,, are operators satisfying (PE, Qn) = 18m PL=P., 2-2, (1.33) The system is reduced to a sum of independent harmonic oscillators: ve 1 t 1 20r = ¥ [Lopes HD, [ag Piet ymei0t| 4K ™m 2a A ina 7 o- = sim v7) (1.34) The eigenvalues are labeled by a set of occupation numbers {a,}: N2 >, nly + 5) (1.38) 2 where @, = 0,1,2, .... The frequency «, is taken to be the positive root of w2, since His positive-definite. In the continuum limit (1.14) the Hamiltonian becomes _f® 1 @ [ q(x, t) \? Heont = I a 55m n+ (Ae) | (1.36) where, with x= ja, 1.3. Second Quantization 7 — PO P(x, t= a =p 2D ms (1.37) The quantum field g(x, f) and its canonical conjugate p(x, 1) satisfy the equal-time commutation relation [PG 0, g@', 0) = -16(x - x") (1.38) Just as in the classical case, we have to introduce a cutoff frequency «,,. General properties of the quantum field will be discussed more fully in Chapter 2. 1.3. SECOND QUANTIZATION Another way to obtain a quantum field is to consider a collection of identical parti- cles in quantum mechanics. In this case, the quantum field is an equivalent descrip- tion of the system. Identical particles are defined by a Hamiltonian that is (1) invari- ant under a permutation of the particle coordinates and (2) has the same form for any number of particles. The quantized-field description is called “second quantiza- tion” for historical reasons, but quantization was actually done only once. Let Hy be the Hilbert space of a system of N identical nonrelativistic particles. The union of all Hy is called the Fock space: F=U Hy (1.39) The subspace with N’= 0 contains the vacuum state as its only member. We assume that N is the eigenvalues of a “number operator” Nop, which commutes with the Hamiltonian. It is natural to introduce operators on Fock space that connect sub- spaces of different N. An elementary operator of this kind creates or annihilates one particle at a point in space. Such an operator is a quantum field operator, since it is a spatial function. This is why a quantum-mechanical many-particle system auto- matically gives rise to a quantum field. For definiteness, consider N nonrelativistic particles in three spatial dimen- sions, with coordinates {r,,... , ry}. The Hamiltonian is Mez V2 +H... 8y) (1.40) where V? is the Laplacian with respect to r, and where V is a symmetric function of its arguments. The eigenfunctions ¥, are defined by AY, (0), 0.5 ty) = E(t 5 By) (1.41) 8 Introducing Quantum Fields For Bose or Fermi statistics, V,, is respectively symmetric or antisymmetric under an interchange of any two coordinates r, and r). The particles are called bosons or fermions, respectively. We now describe the equivalent quantum field theory, and justify it later. Let Urr) be the Schrédinger-picture operator that annihilates one particle at r. Its Her- mitian conjugate '(r) will create one particle at r. They are defined through the commutation relations (An), Wor]. = Br— 4’) (YO), Wr’). = 0 (1.42) where [4,B], = AB + BA, with the plus sign corresponding to bosons and the minus sign to fermions. The Fock-space Hamiltonian is defined in such a manner that it re- duces to (1.40) in the N-particle subspace. ‘A general N-particle Hamiltonian has the structure H=D fl) + 3 att» m+ Mo rte (1.43) where the functions g, h, and so on are symmetric functions of their arguments. The first term is a “one-particle operator,” a sum of operators of the form f(r), which act on one particle only. The second term is a “two-particle operator,” a sum of opera- tors of the form g(r, r), over all distinct pairs. Generally, an “n-particle operator” is a sum of operators that depend only on a set of n coordinates. To construct the Hamiltonian on Fock space, we associate an n-particle operator with an operator on Fock space, with the following correspondences: Ysa > far vreau) Yateany > 5 far din vendors i sp [ars Pra drs Uh Whaat iJek (1.44) where for brevity we have written y = Ur), 212 = a(t, ¥), and so on. As an example, suppose the potential in (1.40) is a sum of two-body potentials: Vir .. 5 ty) => v(t 5) (1.45) % Then the corresponding Fock-space Hamiltonian, also denoted H, takes the form 1.3 Second Quantization = n= fennvrne 2m 1 + Fen Prove nwrao(r, rdMrMe) (1.46) The particle number is the eigenvalue of the number operator, defined as Nop = far Worry (1.47) By using (2.18), we can verify the relations Nop-#f] = 0 [YAr),Nop] = Yr) [Y'@).Nop] = Wir) (1.48) These imply that the action of {(r) on a eigenstate of Ny is to decrease its eigenval- ue by 1, while that of y'(1r) is to increase it by 1. Thus yr) is an annihilation opera- tor, while yt(r) is a creation operator. The vacuum state |0) is defined as the eigen- state of N,, with eigenvalue zero. It is annihilated by all annihilation operators: ww(r)/0) = 0 (1.49) By applying y(r) to the vacuum state repeatedly, it is easy to show that the eigen- values of N,, are nonnegative integers. To demonstrate that the quantum field is equivalent to the many-particle sys- tem, consider a complete set of states |E,N) of the quantum field, which are simulta- neous eigenstate of H and Noy: HEN)= EJB) NoglE.N) = ME,N) We define the N-particle wave function ‘V(r,,.. .,ry) corresponding to |E,N) by 1 Welt. tW) = eel) Wee IEN) (1.50) which has the correct symmetry with respect to particle permutation. It tells us that the probability amplitude for finding N particles at the positions r,,..., ry can be found by annihilating the particles at the respective locations from the state |E,N), and evaluating the overlap between the resulting state and the vacuum state. We leave it as an exercise to show that this wave function satisfies the N-particle Schrédinger equation (1.41). (See Problem 1.3.) 10 Introducing Quantum Fields 1.4 CREATION AND ANNIHILATION OPERATORS The field operator y(r) annihilates a particle at r. That is, it annihilates a particle whose wave function is a 6 function. Since the latter can be written as a linear su- perposition of a complete set of wave functions, we can express Y(r) as a linear su- perposition of operators that annihilate particles with specific types of wave func- tions. Suppose that u(r) is a member of a complete orthonormal set of single-particle wave functions: fr uteuce) = bux > uduter’)= 8-1’) 7 An example of such a set is plane waves: ul) = yee" (1.51) ‘We can expand the field operators with respect to such a basis: vee) = 3, wlan, wo)= > ut(rjal The coefficient a, and aj are operators that satisfy the commutation relations [ay abe = Bix [aps ag]. = 0 (1.52) where the + sign is for bosons and the — sign is for fermions. These relations follow from (1.42) and the orthonormality of the functions u,(r). It follows from the commutation relations that, for each , the eigenvalues of a,'a, are integers n,, called the “occupation number of the single-particle state k”: ata\n) = nln) (nlm) = Bim (1.53) where we have omitted the label k for brevity. The allowed values of the occupation number are given by nal %b2..-4% Bose statistics) 0,1 (Fermi statistics) 1.4 Creation and Annihilation Operators. 11 The actions of a and at have the following results: a|n) = Vn|n — 1) a‘|n) = V1 £n)n+ 1) (1.54) where the +sign corresponds respectively to Bose (+) and Fermi (~) statistics, which show that a annihilates a particle in the state with wave function u(r), and a‘ creates such a particle. We leave it as an exercise to derive these basic results. (See Problem 1.2) The state |0) corresponding to n = 0 is the vacuum state, which satisfies ald)=0 (1.55) We assume that it is normalizable: (0/0) = 1 (1.56) Obviously all other states can be obtained by creating particles from the vacuum: (1.57) ‘We can simultaneously diagonalize aja, for all k. The eigenstates are then la- beled by a set of occupation numbers {1, m,, .. . }, and they constitute a basis for the Fock space. The total number of particles present is N= 2,1. We have alaylto, «5 My.) = Meloy «= 5 Mis») Alto, «5 Mey) = (DV, «4-1, allng, «- - 5 My ---)=(CLVTE ml, -- 5m +1...) where -{S (Bose statistics). (1:58) 2p ®), with periodic boundary conditions. Expand the field in terms of annihilation operators a, for free-par- ticle states of momentum k, and show that kK 1 HD ay alent 2A 2, Raha =S dr e&ty(r). 1.5. Consider a system of N nonrelativistic electrons and N positive ions with Coulomb in- teractions, enclosed in a periodic box of volume (2. The Hamiltonian is given by & pt 7 & @ ee 4P Pi i H=> Piay -> > yt 2 2m * 2 Ae Tn R) 2 IRR)” 2M The ions are heavy. Hence consider R, to be fixed numbers, neglect P,, and drop the last two terms, (a) Label single-electron states by momentum k and spin s, designated collectively as a= {k,s}. The corresponding wave function is Problems 15 ef) et) (b) To go to the quantized-field representation, replace one- and two-particle operators by the rules x 2ke) > (one) ahag Dae FD aap MaBolnayad & By (© Define Fourier transforms: 2 Bur 4tre? Sexe AME ety R (is F-R OO F-R ¢) = 8 Janene Ane? [ks — Kil Buy 8 ee aklts + ky — ka — ky) where 5, is the Kronecker delta. (@ Obtain the Hamiltonian in quantized-field form: re ne? 1 He DF Mets ED Ds Cn tee) Apes 2 Pak 55" Ane? 1 SS aR! a XS Bae (©) Show that the second term gives, for small k, 2re* 2me?N(N - 1 Be LL ote ney which is divergent at k = 0. Show that the divergent term proportional to N? is can- celed by the k = 0 limit of the third term. The O(V) term above remains divergent. The source of this divergence is the periodic boundary conditions, by which the set of coordinates {r},..., ry} is being repeated an infinite number of times in space. ‘Consequently, the Coulomb energy of an electron diverges, due to long-range inter- actions with an infinite number of distant copies. (f) itis clear that this is a mathematical artifact, and to avoid it we should treat the ions dynamically; but we do not wish to add that complication. The expedient way out is to simply leave out the k = 0 contribution in both the second and third terms. Hav- 16 Introducing Quantum Fields ing done this, we can ignore the third term altogether, because it sums to zero under the assumption thatthe fons are uniformly distributed in space, Thus we take as ef fective Hamiltonian ae YF tes Ye (Apri sg bes) Mp sAgs” fe Sat Keo This describes electrons immersed in a uniform positively charged background that, makes the whole system electrically neutral. 1.6 Imposing periodic boundary conditions means filling infinite space with identical cells that contain copies of our system. This problem illustrates the effect of long-range inter- actions among the cells. Consider a unit point charge at the center r= 0 of a cubic cell of volume Z?, which contains a uniform negative charge density, so that the total charge in the cell is zero. Impose periodic boundary conditions, and calculate the potential in the neighborhood of the unit point charge. (a). Show that the potential is given by 4nw et 2m, os i O° Be B by showing vere =4ed 8(r—nL)— * (b) Forr/L <1, show? Yn)= 1 5 7 +0) = 2.837297 «+ - REFERENCES 1, K. Huang, Statistical Mechanics, 2nd ed., Wiley, New York, 1987, Appendix. 2. K, Huang and C. N. Yang, Phys. Rev. 105, 767 (1957). 3. Liischer, Commun, Math. Phys. 105, 153 (1986). 4, C.N. Yang, Chin, J: Phys. 25, 80 (1987). 2Huang and Yang (2] first calculated c, but gave an incorrect value, 2.37. The correct value was ob- tained by Liischer [3]. An elementary derivation was given by Yang [4]. Scalar Fields 2.1 | KLEIN-GORDON EQUATION A fast way to go from classical mechanics to quantum mechanics is to replace the energy and momentum of a particle by operators, according to the prescription Eait a p>-iV (2.1) Making the replacements in the nonrelativistic relation E = p?/2m, and applying the result to the wave function, we obtain the Schrédinger equation: 1 a > Vr, ) =i Wr, 1 2.2) Im YE =i we (2.2) Of course, this is not covariant under Lorentz transformations. For a covariant equa- tion, we use the same trick on the relativistic relation E? = p? + m?. The result is the Klein—Gordon equation (CP + m?) W(x) =0 (2.3) where x stands for x# = (f, x), and a P=09,-3-V? (2.4) in units with c = 1. Assuming that y is invariant under Lorentz transformations, we have a covariant equation—one that has the same form in all Lorentz frames. What is not clear is how to interpret yx). By analogy with the Schrédinger equation, we might interpret ¢{x) to be the 17 18 Scalar Fields wave function of a relativistic particle. That would require the existence of a 4-vec- tor probability current density j#, which should be conserved: 4,, j# = 0. However, the obvious choice jy = #*ys is untenable, for ¥*w is Lorentz-invariant by assump- tion, and cannot be part of a 4-vector. As in the case of the Schrddinger equation, we can construct a conserved cur- rent as follows. Multiply (2.3) from the left by y* to obtain (YA OW) — (YAY d,s) + mys = 0 (2.5) Subtracting this from its complex conjugate leads to 4, j= 0 (2.6) where JE = Yon — Yroey (2.7) However, the time component ay Oe jo = yr—— — ——_, Pea a @8) is not positive-definite, and therefore cannot be a probability density. The root of the difficulty lies in the second time derivative in the Klein—-Gordon equation. As we shall see, this leads to negative frequencies corresponding to an- tiparticles. The relativistic kinematics makes it impossible to have a one-particle theory, We shall regard 4(x) not as a wave function but as a classical wave field, and as such should be quantized. 2.2 REAL SCALAR FIELD Consider a real scalar field ¢(r, ‘), which is invariant under Lorentz transforma- tions. The current j# vanishes identically in this case. We enclose the system in a large periodic box of volume 2, and expand the field in a Fourier series: d= va > aude" (2.9) where GO = G4) (2.10) because the field is real. Assuming that 4(r, £) satisfies the Klein-Gordon equation, we have 2.2 RealScalar Field == 19 Gxt fq, =0 (2.11) where , oR =k +m? (2.12) The system is equivalent to a collection of harmonic oscillators, and may be quan- tized by imposing the commutation relations GLO), gu-(0)] = uae (4x(0), 4ue(0)] = 0 (2.13) where qi(0) is the hermitian conjugate of q,(0). This fixes the normalization of the field, left arbitrary by the Klein-Gordon equation. Since the Klein-Gordon equation is invariant under time translation, the origin of time is arbitrary. The commutations relations in fact hold at any time f: GO, WO] = Sue [als ae (= 0 (2.14) which are equivalent to de, 0, 60°, 0] = Fr’) [h(r, 1), dr’, N= 0 (2.15) These are called equal-time commutators, which serve as initial conditions for the equation of motion. The unequal-time commutators must be calculated from the so- lutions, and contain dynamical information. In the present free-particle case, the equation of motion (2.11) is trivial to solve. For given wave number k there are two frequencies +, with Oy = +V + (2.16) Taking into account the reality property (2.10), we write the solution in the form 1 rte = aye + atyelow ; a = Foeelne™ + abel] 17) where a, and aj are operators, with commutation relations determined by (2.14): Lays aL = Ba [ax ay] =0 (2.18) 20 Scalar Fields The normalization factor (2w)~'? in (2.17) is chosen to make the commutators sim- ple. We recognize that a, is an annihilation operator, and aj a creation operator for a boson, as introduced in Section 1.4. The time-dependent quantized-field operator can be represented in the form H.-S agian talemreny 19) x Vie, The positive-frequency part (the first term) annihilates a particle, and the negative- frequency part creates a particle. The negative-frequency part is absent in a nonrela- tivistic field, because the kinematic relation E = p?/2m allows only one sign of the frequency. 2.3 ENERGY AND MOMENTUM Analogy with the harmonic oscillator suggests that the Hamiltonian of the free scalar field should be 1 H= 7 dul? + lq?) (2.20) In terms of the field (x) = (r, 2), it has the form = [ar 2f(x) e)=5](Fe) +19 ¢r +e | (2.21) where (x) is called the Hamiltonian density. The Lagrangian of the system can be obtained through the general relation L(q, -q) = pq —L(p, 4): L= fers (2.22) where £(x) is the Lagrangian density given by £@)=+| (4) - -|Vor- mae] 3 edad i 223) The last form shows that the Lagrangian density is Lorentz-invariant. In contrast, the Hamiltonian density, which is an energy density, cannot be invariant. For this reason, relativistic theories are usually specified via the Lagrangian density. In terms of creation and annihilation operators, the Hamiltonian takes the diag- onalized form 23 Energy and Momentum = 21 H=> wala, + +) (2.24) © The zero-point energy diverges unless there is a cutoff; but the cutoff has no physi- cal relevance, since the energy of any state relative to that of the vacuum is indepen- dent of it. The energy of a particle is given by 0, = Vie +m? (2.25) where k is its momentum and m is the rest mass. Accordingly, the total momentum operator is P=> kaja, (2.26) f According to the general principles of quantum mechanics, the Hamiltonian is the generator of time evolution, through the Heisenberg equation of motion: aH, 6¢¢, 9) AED 227) ‘The formal solution yields Hr, t) = "g(r, Oe (2.28) For consistency, we must show that this is consistent with the Klein-Gordon equa- tion, which we used to arrive at the solution (2.19). Substituting (r, 0) from (2.19) into (2.28), we obtain dr, = ae agent ale Te ttte (2.29) For the free-field Hamiltonian, we have (see Problem 2.1) eget = queria (2.30) This demonstrates that (2.28) is the same as (2.19). Again, according to general principles, the momentum operator P should gen- erate spatial translations: =P, Hr, )] = V dtr, 2) (2.31) with formal solution Hr, ) =e? *H(0, DeP* (2.32) Asa straightforward calculation shows, this is the same as (2.19). 22 Scalar Fields 2.4 PARTICLE SPECTRUM The vacuum state |0) is the state of lowest energy, defined ty a(0)=0 (all k) (2.33) and normalized to (0|0) = 1 (2.34) A one-particle state is defined by Ip) = afl0) (2.35) Using the commutation relations, we find aylp) = 444310) = (ayay + Byp) |0) = dypl0) (236) The field operator has nonvanishing matrix elements only between a one-particle state and the vacuum: loro) <0| !p) = —=— (2.37) Oo = aa This is the wave function of a particle of momentum p, normalized to a density (2e,0)". By successively creating particles from the vacuum, we can build a com- plete set of states: Vacuum: — |0) 1-particle states: Ip) = a,*|0) 2-particle states: [PiP2) = ap, "ap, "10) 2.5 CONTINUUM NORMALIZATION In the limit © — ©, the allowed values of k approach a continuum, and we can make the replacements I &k azo! Qn 2.5 Continuum Normalization — 23 DB —(277H(k-k’) (2.38) We define continuum versions of the annihilation and creation operators by putting atk) = Oa, at(k) = Oat (2.39) The commutators then have limiting forms: [a(ho, a%(k’)] = 22P H(k -k’) [a(k), a(k’)] = 0 (2.40) The field operator can be represented as a Fourier integral: Pk 1 = kT-0K) + qt(Ik)e ler-ox0 (rr, 1) J On? Vie [a(k)e" +al(kje“ ) (2.41) As before, the vacuum state |0) is defined by a(k)|0) = 0 with (00) = 1, and a one-particle state is defined by Ip) = at(p)l0) (2.42) Using the commutation relations, we find a(k)|p) = (277)8°(p — k)/0) (2.43) The single-particle wave function is elorep) 0 = 2.44) (ovb0anw) = Fae 4) with a particle density (2w,)'. The normalization is such that the number of parti- cles in volume element d*r is the Lorentz-invariant combination d°r/(2w,). The Hamiltonian and total momentum now take the forms =(e Gap one toatl) Bk P= Jeeps ko") (2.45) The choice between discrete or continuum normalization is a matter of nota- tion, since we always regard as large but finite in intermediate steps of calcula- 24 Scalar Fields tions. The limit (2 — © is taken only in final answers. This is done to avoid irrele- vant concerns about mathematical rigor, such as how to define the Hilbert space when the dimensionality is noncountably infinite. The continuum normalization merely anticipates this limit 2.6 COMPLEX SCALAR FIELD ‘A complex scalar field is just two real scalar fields constituting the real and imagi- nary parts. What is new is a symmetry between the two fields, and this leads to a conserved current. In physical terms, a complex field can have electric charge, whereas a real field must be neutral. ‘We denote the complex scalar field by y(r, 1), and decompose it into real and imaginary parts in the form 1 = qld + is) (2.46) The Lagrangian density is taken to be L(x) = HU ()A,. Wee) — mp) 2 DL (2) 52) — mbox) h(x] (2.47) The normalization factor 1/V2 in (2.46) is chosen in order that ¢ has the same nor- malization as the real scalar field discussed previously. To quantize the system, we impose the commutation relations Tbe, , H(e", )) = 8,F0-1') [hr 1), (0, 0] = 0 (2.48) The complex field yr, 1) becomes a non-Hermitian operator satisfying Wr, ), WO", O)= 8-2’) (Hr, 9, He’, 1 = LH, 0, Wr’, O1= Lr, ), Wr’, D]=0 (2.49) Thus, the canonical conjugate to wis y*. In accordance with (2.19), we can expand ¢; in terms of annihilation and cre- ation operators: 1 Tage anor + ahewerond) (j= 1,2) (2.50) V20, 605.0" Se 2.7. Charge and Antiparticle 25 with the commutation relations [ats afi] = 8B. [ans a']= 0 (2.51) In the complex representation, we have 1 1 Ur, t)= Var Via, OH ellen) + oifeiter-o4) We )= Wad Vim (ble r-o10 + celeron (2.52) where 1 b= lain tn) - - = alan —idy,) (2.53) with commutation relations [bus O53] = Ske Lew ¢B] = xp [bus Bp] = [us Cp) = [us ep] = 0 (2.54) The total energy and momentum can be expressed as follows: H=> ox(ahyay, + aban) = >, oy(bhdy + chew) v. i P=S ghklahon+aand= F kbli tle) 58) There are two type of quanta, which can be designated either as a, and a, quanta, or as b and c quanta. The energy and the momentum do not distinguish between these descriptions. We shall see, however, that only the b and c quanta have definite charge. 2.7 CHARGE AND ANTIPARTICLE The current density for the complex scalar field is given by IH= Wonk — ROW = 3 (dod — br) (2.56) 26 Scalar Fields which satisfies the conservation law 4, j# = 0, or a jo - ah tVi-0 (2.57) Integrating both sides over the spatial volume, we obtain a d By. yyy = 4 (gap) = fa at (x) a fa ‘xf(x) = 0 (2.58) or do a7 (2.59) where Q is the total charge operator 0= faxPe) = TeatheeT => @hah-ahaw) K => Ox - chew) (2.60) 7 ‘As we can see, a b quantum carries one unit of positive charge, and a c quantum car- ries one unit of negative charge. The aj, a) quanta, which are linear combinations of those of b and c, do not have definite charge. By convention, we refer to a c quan- tum as an “antiparticle.” Thus, the positive-frequency part of annihilates a parti- cle, and its negative-frequency part creates an antiparticle. Similarly, yt either cre- ates a particle or annihilates an antiparticle. In light of this, we can say that for the real field, the particle is its own antiparticle. The term “charge” is used in a generic sense, and does not necessarily mean the electric charge, since we have not turned on the electromagnetic coupling. The unit of charge is arbitrary, because j* is defined only up to a multiplicative constant. It is straightforward to verify that charge is conserved: (0, H]=0 (2.61) This implies that the number of particles NV, minus the number of antiparticles N_ is a constant of the motion. In the free-field case, this conservation law is trivial, since N, are separately conserved. It becomes significant when, in the presence of inter- actions, N, are no longer conserved. In that case, N, — N_is still conserved as long as (2.61) holds. 2.8 Microcausality 27 2.8 MICROCAUSALITY A classical signal propagating according to the Klein-Gordon equation has a group velocity = 9% Ik! Perour a * Vitam (2.62) which never exceeds 1. This means that events at two space-time points lying out- side of each other's light cone (or separated by a spacelike interval) cannot influence each other. In the quantum theory, this means that two field operators at points sep- arated by a spacelike interval must commute with each other: [d), 6@']}=0 if (e-x'P <0 (2.63) This condition is called microcausality. We must verify that our quantized field the- ory satisfies this condition. To compute the commutator in (2.63), note that it is a c-number,! and at fixed x’ it satisfies the Klein-Gordon equation, because ¢(x) does. The initial condition at Xo = x5 is the equal-time commutator (2.15), which is a c-number. Therefore the commutator remains a c-number at all times, and we can equate it with its vacuum expectation value: [4(x), $0)] = (OGG), (110) = AG -y) (2.64) This defines a Lorentz-invariant correlation function A(x — y), which depends on x — y, and not on x and y separately, because of the translational invariance of the vacu- uum state. (See Problem 2.1.) We use the expansion (2.41) to obtain ar oyjoy = { —22E— eater Ike, DON) = | So, Pk wy (odode, 010) = [ aay ewer (265) Subtracting one from the other, we have Pk sind) ee Ga¥ a (2.66) A(x) = -i(O1L4(r, 0), (0110) = - 4A c-number is “classical” number, a multiple of the identity operator. 28 Scalar Fields Since A(x) is Lorentz-invariant, it can depend only on the invariant e-x : (2.67) If.x is spacelike, for which x? < 0, we can put = 0. By (2.66) this gives A(x) = 0. = The proof of microcausality depends on the initial condition from the commu- tator in (2.15), which quantizes the system according to Bose statistics. Had we used Fermi statistics by replacing commutator with anticommutator, microcausality would have been violated. The particles here have spin 0, since there does not exist discrete degrees of freedom corresponding to spin. Our result is partial demonstra- tion of the spin-statistics connection, which states that particles with integer spin are bosons, while those with half-integer spin are fermions. The second half of the statement will be shown in Chapter 6 on the Dirac field. 2.9 THE FEYNMAN PROPAGATOR The propagation of a free particle in the vacuum can be described by the correlation function AO (Xp — x1) = =H) ()10) (2.68) in which y*(x,) creates a particle from the vacuum at x, which is annihilated by Wx) at x2. This makes sense physically when , > f,. Similarly, the correlation func- tion Ae, =x) = =O] Ce) O2)10) (2.69) describes the propagation of a test antiparticle from x, to x,, and is physically mean- ingful when f, > 4. To obtain a correlation function that has physical meaning, we use either A‘ or AO depending on the sign of the relative time. The result is the Feynman propagator, or causal propagator: Mex — x1) = HKO| THO) WCx,)/0) (2.70) where the time-ordering operator T rearranges the operators, if necessary, such that the operators stand in such order that time increases from right to left: A(ty)B(t) if >t, ae eas if ney ee If 4 > f,, the Feynman propagator describes the propagation of a particle from x, to %; if fy < ty, it describes the propagation of an antiparticle from x, to x;. This is the 2.9 The Feynman Propagator 29 basis of Feynman's famous remark: “An antiparticle is a particle traveling back- wards in time.” To calculate the propagator, we start with the expression ae OldAx)yit(0)|0) if x°>0 Arle) = -{ (Old (O)ykx))0) if x°<0 7) and insert a complete set of states between the operators. Since the field operator connects the vacuum to one-particle states only, we have (OlWCx)Ik) {kly*(0)0) if, 29> 0 (2.73) A) =} Oop { (OW COVK) (KiyoOI0) if x°>0 Using Wx) = e”*yA0)e*”*, and changing the integration variable from k to —K in the lower formula, we obtain Ap(x) = =i f ei YA0)|0)/Pe* reread (2.74) Quy " Using the following integral representation cd ike ioe = 2 f ee ds Gooltm (1) (2.75) we obtain ArQ) = fe On 2 lOWOOP Se (n> 0°) (2.76) From (2.44) we have Koimonone = = em os Therefore = &y)= |S tn) 278) Operating on both sides of this equation by (? + m? gives (CP + mi?) p(x) = 54x) (2.79) This shows that the Feyriman propagator is a Green's function of the Klein-Gordon equation. 30 Scalar Fields The Fourier transform of the Feynman propagator is 5 1 : \0- Bate 2.80) which has poles at k= VE (2.81) corresponding to a particle or antiparticle of mass m. The residue 2c4{(0|y(0)|0)? = 1 reflects the normalization of the wave function. We may view the propagator as the propagation amplitude of a virtual particle of 4-momentum k“. The virtual parti- cle, whose squared mass #? ranges between —~ and ©, becomes a real particle when it “goes on mass shell,” at (2 = m?. To obtain Ay(x) explicitly, we integrate over the angles of k in (2.74) to obtain 2 sin kr Ais) = G5 al a= Oe (2.82) By Lorentz invariance A,(x) can depend only on sex=P-P 2.83) For s > 0, we can put r = 0 to obtain the representation a@)= of “ae envi" _ mV) (6>0) 2.84) 4 wy 8aVs For s <0, we put = 0 to obtain R sink V=s dee= ef eS — = K(mV=) — (s<0) (2.88) where H{" and K, are Bessel functions. In the timelike region s > 0 the function de- scribes an outgoing wave for large s. This corresponds to the in prescription in (2.80). The in prescription would have yielded an incoming wave. In the spacelike region s <0 it damps exponentially for large |s|. On the light cone s = 0 there is a delta-function singularity not covered by the preceding formulas: Jim Axo) =F Tate) (2.86) 2.10 The Wave Functional 31 2.10 THE WAVE FUNCTIONAL In quantum mechanics, the coordinate representation is defined by basis states |r) satisfying Toplt) = r[r) (2.87) with the basic commutator realized through the replacements Top (c-number) Pop > -iV (2.88) A state |A) is represented by the wave function. Walt) = (r/A) (2.89) and inner products are defined by (AB) = [aru ste) vale) (2.90) In the analogous field representation in quantum field theory, we diagonalize the field operator, thus representing it by its eigenvalue, which is a c-number func- tion, For a real scalar field ¢,,(r) at a fixed time r= 0, we denote its eigenstates by \): dop(H)IP) = H1)|4) (2.91) where the eigenvalue (1) is a real-valued function of r. The commutation relations (2.15) are realized through the replacements dop(t) > a (c-number function) ae al) 5 92) where 8/84(r) is denotes the functional derivative with respect to the value of the function gat r. A state |A) is represented in the field representation by the wave functional Wyld] = (pla) (2.93) which is a complex-valued function whose argument is a function; that is, its value depends on the form of the function. Inner products between wave functional are functional integrals: 32 Scalar Fields (He ¥0)= [D6 VIL6oI1 2.94) where Dé denotes the measure on the space of functions. Writing the Hamiltonian in the field representation, we have the Schrédinger equation for the wave function- al: IV, #] zl gag VHP +more] WEA a 295) 2.11 FUNCTIONAL OPERATIONS We digress on functional operations on a functional F[¢]. First, the functional de- rivative 5F[4]/6¢(x) is defined as follows. We make a small change d—> # + 8¢, where the function 8¢(x) is different from zero only in the neighborhood of x. Then the functional derivative is given by SIG] _ |, [ Fld + 66] -FId] B(x) “ke 5px) ] 2.96) To calculate any functional derivative, we need the elementary functional derivative 5(x)/8(y), which is obviously proportional to &(x — y). To determine the propor- tionality constant, we replace the continuous space of x by a lattice of spacing a, and denote by ¢; the value of the function on site j. Clearly, 9b, by og se = 88 297) Inthe continuum limit Y safer (2.98) 5 5, > ade —y) we have Sd(x) = -. (2.99) Bao) ED 2.99) With this formula, we can calculate a general functional derivative. As illustration, take FT] = J d?y{V h(y)?. Then 2.12 Vacuum Wave Functional 33 FAS », In either of the preceding methods, the integral by itself may not have a contin- uum limit; but matrix elements of the form (2.103) usually has a definite continuum limit. 2.12 VACUUM WAVE FUNCTIONAL We now calculate the wave functional for the free vacuum state. First let us express the annihilation operator a(k) in the field representation. From (2.41) we have, at t=0, Jarre de) a [a(k) + at(k)] 7" Jar erste) =-1 tea -at()] (2.104) 34 Scalar Fields Solving for a(k), we obtain a(k)= Vaal e™*[0,4(0) + 16(0)] (2.105) Now we write ey = VIE + ne ht = VV im eet (2.106) so that a(k) = Tix. farlo@v=F +m + id(e* _ vag lO + md(n)] + iG(n) fer (2.107) The last step is obtained by expanding V—V? + m in a power series in V2, per- forming a partial integration in every term, and summing the series again. The sur- face integrals generated in the partial integrations vanish as a result of periodic boundary conditions. Replacing id(r) by 5/5¢(r), we obtain 1 atk) x far em [Vivre Fm b(n) + wel (2.108) The virtue of this representation is that the Fourier coefficient in the integrand is in- dependent of k. The wave functional of the free vacuum satisfies the equation a(k)Vo[ 4] = 0 (2.109) Thus, it must be annihilated by the Fourier coefficient in (2.108): é [v=o FH oH + sas Moldl=0 2.110) The solution to this equation is Wolw] = C exp[-+ far $n) V=V7 = Fd) (2.111) where C is a normalization constant. This gives the probability amplitude that the field has the functional form ¢(r) in the vacuum state. The relative probability for the field to have a functional form lying in the neighborhood of ¢ in the volume el- ement D¢ of function space is 213 The ¢*Theory 35 IWlelPD¢ The most probable form is ¢ = 0, and deviations from it occur with a Gaussianlike distribution. ‘The exponent in (2.111) can be rewritten in different forms. Introducing the Fourier transform Hk) = J Br (nr) (2.112) we can write Jar deyV-V FHP on) = [ee am VE Fre do? = [a'r ar $Kee—r') dee") 2.113) where Kw) = [eee = eer a me (2.114) For a complex scalar field, there are now two coordinates, which can be taken as either {¢,, 2} or {y, y+}. Inner products of wave functionals now take the form (Ya Va) = fos, Do2V5ltos boVal 2. $2] (2.115) or equivalently (Va, Va) = Jowywau, wel w] (2.116) The complex measure is defined in terms of the real and imaginary parts: Di Di = Dd, Dd (2.117) The vacuum wave functional for the free complex field is just the product of those for the two independent real fields. Reexpressing the result in terms of the complex field, we have Wold] = Cexp[-f a'r vA(V-V? + mF U)] (2.118) 2.13 THE ¢* THEORY As the simplest example of an interacting field theory, consider the Lagrangian den- sity of the so-called $* theory: 36 Scalar Fields L(x) = 5 bd, — 3? P - gdb* (2.119) where g > 0. The quartic term makes the equation of motion nonlinear: (P+ m)p+ge=0 (2.120) The Hamiltonian density takes the form He) =3](F) 1797] + Moe) 1 V(b) = ym dx) + gb") (2.121) which suggests that V((x)) is a potential. To quantize the theory, we impose the equal-time commutators (2.15), which can be satisfied by taking as initial condition = Ly jane + ate or, 0) vad Valance + ale * ety i, S Vad 7a oan + a") (2.122) where the creation and annihilation operators aj, a, are defined by the commutation relations [ay ah] = Bx [a,, ay] = 0 (2.123) The equation of motion is not soluble unless g = 0. We can always write, as a formal solution, (r,t) = edn, Oe (2.124) but this is not simple unless g = 0. To see the effects of the interactions, separate the Hamiltonian into a “free” term and an “interaction” term, at some arbitrary time t = 0: H=Ho+ Hix (2.125) where Jarno, 0 +1V 4, OF] Problems 37 Hige= g[°rd%(r, 0) (2.126 In terms of the creation and annihilation operators we have Hy=Co+ >, wyalay c Hoy = & + Sxl He tH AMO Gt saya} +a,Nal+asyal +a,) where Cy is an irrelevant zero-point energy and 6, denotes the Kronecker 5. We use the shorthand 0; = a,, 4) = ay, and so on. The interaction Hamiltonian Hix, de- scribes four-particle processes that conserve momentum. Substitution of this expan- sion into (2.124) generates a complicated series for the time-dependent field opera- tor. We shall learn how to organize such terms in a systematic manner in Chapter 9. PROBLEMS 2.1 Space-Time Translation Consider a free scalar field (x), which can be expanded in terms of the annihilation operators a,. This problem illustrates the fact that the 4-mo- mentum P+ = 3,é4a.a, is the generator of space-time translations. (a) Asa useful tool show that, for two operators 4 and B, Be“ =B+ (4, B]+ 314,14, B+ 5 (b) Use this formula to show Page = ayes and the infinitesimal form [P#, ay] = kay, (©) Establish that P# is the generator of space-time translations by showing (PH, (x)] = thx) (@)_ Let |K) be an eigenstate of P#, satisfying PH\K) = KK) . Show that this state is translationally invariant: (KI) PODIK) = (Kb ~») AOI) 2.2. Charge Conjugation The designation of particle and antiparticle is a matter of con- vention, and we can freely reverse the labels. More specifically, for a complex scalar 38 Scalar Fields field (x), construct an operator C’that takes by to c, and vice versa, and commutes with Hamiltonian CYQ)C = Yi(x) [4, C]=0 cc=1 Cel The operation C is called charge conjugation, or particle-antiparticle conjugation. 2.3 Lorentz Invariance We have calculated the function A(x) = (O\[¢(x), #(0)}|0) in (2.66), but not in a manifestly Lorentz-invariant form. Show that it can be put into the desired form {4 A(x) = anil @ay et8(12 — melo) 2.4 Spin and Statistics | Quantize the real scalar field according to Fermi statistics; in oth- er words replace the commutators in (2.15) by anticommutators. Show that this will vio- late microcausality. 2.8 External Source Consider a real scalar field ¢(x)\coupled to an external source func- tion J(x), with Lagrangian density L)= FHbIG+ amide + Jb (a) Obtain the Hamiltonian in terms of the creation and annihilation operators a4, aj, for plane-wave states. (b) Suppose that the source is static, that is, that J(x,t) is independent of t. Using per- turbation theory, show that there is no scattering from the fixed source to second order. (©) Show that there is no scattering at all, to any order. (Hint: Show that a linear canon- ical transformation of ay, reduces the Hamiltonian to the source-free case.) 2.6 Level Shift Suppose that the external source in the previous problem consists of a sin- gle static point source: J(x) = g%(r). (a) Calculate the change in the energy of the vacuum state to order g?. The result will be a divergent integral. Cut it off at a large momentum A. This illustrates a proto- type of divergence in quantum field theory. (b) Show that all levels of the system shift by the same amount and therefore that the divergence in this case has no physical relevance. 2.7 Yukawa Potential Continuing with the last two problems, suppose the source function U(x) consists of two static point sources located at ry,r2: Jr, = g(r ~ 11) + Hr ~r2)] Treating g as a perturbation, calculate the change in the vacuum energy to second order in g, and show that there is an attractive potential between the two point sources: 28 Problems 39 where R = |r, ~ r9|. This is the Yukawa potential, originally proposed as the potential be- tween two nucleons due to interactions with scalar mesons. ‘Vacuum Fluctuations Consider a free scalar field in a large periodic box of volume . Let the Fourier transform be denoted @) ) © @ de = Fz Jarre) Show that the vacuum expectation value of G(k) is zero. By expanding the field in terms of creation and annihilation operators, show that the mean-square fluctuation of the Fourier transform is given by 5 1 (OP W00) = Saree ‘The mean-square average can be expressed in the field representation as. foowaovates (016(&)0) = foovae where V4{ ¢] is the wave functional ofproduce the last result from this formula. Calculate the mean-square fluctuation {0|¢7(x)|0) in coordinate space. The result is divergent because of the high-momentum modes. Exhibit its dependence on the cutoff momentum A. Relativistic Fields 3.1 LORENTZ TRANSFORMATIONS Relativistic quantum fields can be classified according to the way they transform under Lorentz transformations. More specifically, they transform according to irre- ducible representations of the Lorentz group. The different representations give rise to particles with different values of the spin angular momentum. ‘According to the principle of special relativity, the laws of physics should be covariant with respect to Lorentz transformations; that is, they should have the same forms in all reference frames connected by Lorentz transformations. The simplest Lorentz transformation is a “boost” of the reference frame with velocity v along some axis, say, the x axis: GB.) This may be supplemented by a rotation of the coordinate system, say, about the z axis through an angle 6: x’ =x cos6+y sind y' =-x sind + y cosd (3.2) Defining a boost “angle” by coshg = 1/V1— 22 sinhd = V1 —0 (3.3) 40 3.1 Lorentz Transformations 41 we can write the matrices of these transformations as follows: cosh -sinh 0 0 ; ~sinh ¢ 0 0 Lorentz boost: $ coshd 1 oO}. 0 001 1 0 00 Rotation, [2 C088 sind 0 G4) 0 -sin@d cos@ 0 0 0 oo. The inverses of these matrices can be obtained by reversing the signs of # and 6. The rotation matrices are orthogonal matrices, while the Lorentz boosts are not, be- cause the invariant form — x? for the Lorentz boost is not positive-definite. The angles of rotation are not additive, unless the rotations are all made about the same axis. Similarly, the velocities of successive Lorentz boosts are not additive, unless the boosts are all made along the same direction. We use a relativistic notation in which the coordinate 4-vector is denoted by x= (1, r) and the metric tensor is diagonal: @.5) -1 A general Lorentz transformation is a linear transformation A on x that leaves 2? = P — r? invariant: xH= AM XY (3.6) with the requirement Suv AMaA"s = Sap @B.7) which ensures the invariance of x*. In shorthand, we write the transformation in the form x=Ax (3.8) The transformations above form the continuous Lorentz group, which is character- ized by six parameters: three velocity components and three angles of rotation. As we can see from (3.4), they are represented by matrices with determinant +1. In contrast, the discrete transformations 42 Relativistic Fields Spatial reflection: G9) Time reversal: have determinant —1. These discrete elements together with the continuous Lorentz transformations form the general Lorentz group. We shall reserve the name “Lorentz transformation” for the continuous Lorentz transformations. Any element of the Lorentz group can be built up from infinitesimal ones, with the general form At, = gt, + of, (3.10) We write in shorthand A=l+o G.I) Lorentz transformations generally do not commute with one other; but the infinites- imal transformations do, because their commutators are of second-order smallness: (1+ @)(1 + @) = 1+ @ + @ + Ow?) (@.12) Thus, group multiplication is equivalent to addition of the ws. An infinitesimal transformation of the coordinate system, characterized by boosts with velocities v/ along the x/ axes, and rotations of angles 6* about the x* axes, is described by the tensor 0 vt =P 0B 8 0 6.13) # 0 By raising the lower index, we obtain an antisymmetric tensor of 2 8 = ms co gireon,=| 2 0 3.14) 2 8 0 -6 v7 -P 6! 0 whose elements can be summarized as follows: = ke -ol = coll =e! = —eb'ge (3.15) 3.2. Minimal Representation: SL(2C) 43 3.2. MINIMAL REPRESENTATION: SL(2C) It is well known that the smallest faithful representation of the rotation group is SU(2), the group of 2 x 2 unitary matrices of unit determinant. For the Lorentz group, the minimal representation is SL(2C), the linear group of 2 x 2 complex ma- trices of unit determinant. To see this, let us organize the coordinates into a 2 x 2 complex matrix: xerr@n=(CY, ae ) G.16) where o* are the Pauli matrices, with the following properties: fo, 08} = by o'o? = io? (and cyclic permutations) [o', 02] =2i03 (and cyclic permutations) @.17) ‘We see that det ¥=x? (3.18) A Lorentz transformation that takes X into X’ can be represented by the operation X" = L(A)XL"(A) (3.19) where L(A) is a2 x 2 complex matrix and L"(A) its Hermitian conjugate. Taking the determinant of both sides, we have det X" = det X|det L(A)? (3.20) To preserve x2, we must have det X” = det X, and hence det L(A) = #1 21) Consequently det A = +1. This is a more formal proof of a result stated earlier. The matrices L(A) with det L = | constitute the group SL(2C). Any 2 * 2 can be represented in the form (3.22) 4+@0-( A+B; a) By +iBy A-By where A and B, are complex numbers. The determinant of the preceding equation is 44 Relativistic Fields A? ~3,B}. Hence L(A) is a matrix of this form, with 4? -2,B} = 1. We leave it as exercises to show that a pure boost and pure rotation are represented by the follow- ing: : a: = ertob? = cosh? — (fo) sinh? Boost along: L(A) = e##? = cosh — (ra) sinb Rotation about: L(A) = elit 282 = cos +i(fro) sin 3.23) where ft is a unit vector, @ is the boost angle defined in (3.3), and @ is the rotation angle. 3.3 THE POINCARE GROUP The laws of physics should be covariant with respect to space-time translations as well as Lorentz transformations. These transformations combined constitute the in- homogeneous Lorentz group, or the Poincaré group. The transformation law is as follows: x's al + ABx” (3.24) where a is a 4-vector. The infinitesimal version has the form XH xh + Ql + aot, x” (3.25) which contains 10 independent parameters: a” and w#” = -w", We can realize the Poincaré group on the space of functions /{x), through the transformation Se!) =fe+ a+ wx) =I) + ay d,f G0) + OF, x”, I) = [P+ ad 3 py XH” — x” NFR) (3.26) where we have use the fact that w,,,, is antisymmetric. We can rewrite Se" () —iaMPH + yHaoht) ie) (3.27) which defines the generators 3.3 ThePoincaréGroup 45 Me? = x#P¥ — x¥Pe (3.28) Of these, 10 are independent operators, constituting the Lie algebra of the Poincaré group. An arbitrary element of the Poincaré group can be written in the form exp (ia, P# —iw,,,MH") (3.29) where a and wo” represent 10 real independent parameters. From (3.28) we obtain the commutator be PY igh (3.30) Although derived from an explicit representation, we consider the preceding equa- tions as abstract algebraic relations. Such a procedure is analogous to obtaining the Lie algebra [J/, J*] = ie/*J! for angular momentum from the special representation -ir x V.. As abstract relations, the Lie algebra admits half-integer representations. The Lie algebra of the Poincaré group consists of the following commutators: [Me, M28] = —~i(gua 8 — graiyub + gyB\ua — gubyra) [M, P"]= i(gt*P'- g”P#) [P4, P*]=0 31) which can be obtained through a straightforward calculation. In physical terms, the four generators PH = (H, P', P?, P?) (3.32) make up the total 4-momentum operator, and P° = H is the Hamiltonian. The six in- dependent components of M“” are generalized angular momentum operators made up of the angular momentum J and the Lorentz boost K: Mi = eiktyt M%=Ki (3.33) We can recast the Poincaré algebra as follows. The last two lines in (3.31) are equiv- alent to [P’, P= [P), H] = [, H]=0 [), PX] =-ie/t!P! [k), H] =-iP! [B, P)] = 18, (3.34) 46 Relativistic Fields These relations all involve the inhomogeneous part of the group. The first equation above expresses the independence of the spatial translations among themselves, of spatial and time translations, and of rotations and time translations. The second equation is what one can deduce from J = -ir x V and P = iV. The other equa- tions above describe how energy and momentum change under a Lorentz. boost. In addition to these, we obtain from (3.31) a closed set of commutation relations among angular momentum and boost operators: VJ = ie! [K!, K4]=-iet¥y! WKN =iel'K! (3.35) These form the Lie algebra of the Lorentz group. 3.4 SCALAR, VECTOR, AND SPINOR FIELDS In quantum mechanics, the wave functions in a central potential can be classified according to orbital angular momenta, which correspond to irreducible representa- tions of the rotation group, with possible dimensions 2/ + 1, (/ = 0,1,2, ...). Ina similar way, relativistic fields transform according to irreducible representations of the Lorentz group, which have definite dimensions. Accordingly, a relativistic field has a definite number of components, related to the spin angular momentum of the field. The simplest relativistic field is a scalar field, which may have more than one component, but each component x) must be invariant under Lorentz transforma- tions: '(x') = @) (3.36) This says that the transformed field called ¢’, at the transformed coordinate x’, is the same as the original field called ¢, at the old coordinate x. It expresses the fact that x’ and x are different labels that we use for the same physical point, and the scalar field is unaffected by this; but for us the functional form of the field must change: '@)= GA'x) (3.37) ‘As we shall see, the spin of a scalar field is zero. A vector field, such as the electromagnetic field 4“(x), is affected by a change in the coordinate system, since by definition its four components transform among themselves like x", The transformation law is 3.4 Scalar, Vector, and Spinor Fields 47 A’#(x!) = AH AMX) (w= 0,1,2,3) (3.38) The spin of a vector field is 1. This will be demonstrated in Section 5.5. In general, a tensor field of rank n transforms like a product of n x“ terms, and corresponds to spin n. For example, the gravitational field is a symmetric tensor of rank 2. There are “half-integer” representations, analogous to those for the rotation group. The latter are representations of SU(2), which generalizes to SL(2C) in the present case. To accommodate space-time reflections, we have to include two copies of SL(2C), so that they transform into each other under a reflection. Accord- ingly, the minimal representation space is spanned by a four-component complex field, called the Dirac spinor field y(x), which transforms according to Wi’) = SCA) (r= 1,2,3,4) (3.39) where S(A) is a 4 x 4 complex matrix, discussed in more detail in Chapter 6. The spin of a spinor field is +. In general, a field forming a K-dimensional irreducible representation of the Lorentz group has K components: bx) (a= 1,2,...,K) (3.40) which transform under a Lorentz transformation A according to Pal") = San Mb) 3.41) For an infinitesimal transformation A = 1 + @, we can put S(A) in the form Sas = Sap + $k (3.42) this defines the coefficients 3/4”, which, as we will show, constitute the spin matrix. Under an infinitesimal Lorentz transformation, then, a general field transforms according to hale’) = dalx) + $y Xhs P40) (3.43) The change in the functional form of the field can be found by writing bale!) = bale + ex) = by(X) + Oy X”Iuha(X) = bi x) — Oy Axd” - x9, )b4 (x) (3.44) Thus 48 Relativistic Fields PAC) = ha!) + F Oy fXtd” — XH) a(x) (3.45) Substituting {(x’) from (3.42), we obtain bal) = Pax) + 3 Oy sL HO” — x” Eap + Zio bs(x) (3.46) This identifies the K x K matrices ¥#” = -3" as spin matrices, since they are added to the generalized orbital angular momentum. The spin matrix for a scalar field is obviously zero. For the vector field, we can find it from its transformation law under an infinitesimal Lorentz transformation AL(a!) = Aah) + ap AMG) G47) Putting wag = 2,28, We obtain for the vector field Xap Sap 8hBa (3.48) As we shall show in Section 5.5, this gives spin 1. The case of the spinor field will be discussed in Chapter 6, and is included in the following summary for reference: Scalar field: 2"”=0 Vector field: X45=ghes—geen Spinor field: a= sor YY es (3.49) where are the 4 x 4 Dirac matrices defined in Chapter 6. 3.5 RELATIVISTIC QUANTUM FIELDS Since quantum fields are operators that act on a Hilbert space, we can represent Lorentz transformations by transformations on the Hilbert space. Recall that a Lorentz transformation changes the functional form of a classical field: Pale} bale) (3.50) In the quantized version, this means that the operator ¢, attached to point x is re- placed by ¢/. Since @, and ¢/ act on the same Hilbert space, the transformation is a mapping of the Hilbert space into itself. Since ¢, and ¢/ are physically equivalent, the transformation must be unitary. Thus, there should exist a unitary operator U(A) on the Hilbert space, corresponding to the Lorentz transformation A, such that Pale) = UA) b(X)U A) @.51) 3.5 Relativistic Quantum Fields 49 The fact that the transformation is unitary means UN(A) = UA) (3.52) From the definition of the primed fields 5(x) = Sasds(x), we obtain the condition Udslx)U*' = Sans A'x) (3.53) The set of operators U(A) forms an infinite-dimensional unitary representation of the Lorentz group. In contrast to this, the finite-dimensional representations of the Lorentz group are nonunitary. As examples, we have Scalar field: Ud(x)U“! = (Ax) Vector field: UAM(x)U-! = Ag A*(A“!x) Spinor field: Uy,(x)U- = S,,,(A“'x) (3.54) We can immediately extend this consideration to Poincaré transformations: Uebs ()U"! = Saab A“e- a) (3.55) For infinitesimal Poincaré transformations, U must be in the neighborhood of the identity operator, and linear in the parameters of the Poincaré group: U=1-iaupe+ ad (3.56) This defines the Hermitian operators P and M«”, which represent the generators of the Poincaré group on the Hilbert space. In contrast, the generators denoted by the same symbols in (3.28) are finite matrices, generally non-Hermitian. Substituting (3.56) into (3.55), we obtain ( -iaP + you bse(1 +iaP - zoM) = (5 + 73s) Oana) (3.57) which is written in an obvious abbreviated notation. Expanding both sides to first order in a“ and w”, and equating their coefficients, we obtain HTPH, ba(x)] = bax) HME”, ba(x)] = (x — xO") hale) + Eas ho(x) (3.58) This shows that P* is the 4-momentum operator, since it generates space-time 50 Relativistic Fields translations, and Mé” is a generalized angular momentum operator, since it gener- ates space-time rotations. The spin matrix Sy, induces a mixing of the field compo- nents undergoing space-time rotation. The generators P# and M+” can be constructed explicitly from the field opera- tors @,. Rather than doing this on a case-by-case basis, we shall do it via a unified approach in the next chapter. 3.6 ONE-PARTICLE STATES A one-particle state is an eigenstate of P#, with energy eigenvalue E > 0, and mo- mentum eigenvalue p, such that the invariant mass squared Ps Ep? =m? (3.59) is a fixed number. Such a state corresponds to a particle of mass m. The P? of any state generally lies in a continuum, but those of one-particle states form a discrete set. If there are no massless particles, the invariant-mass spectrum of a field theory consists of the vacuum value 0 as a lower bound, a discrete set of particle masses, and a continuum separated from the vacuum value by a finite gap. The continuum corresponds to states containing two or more particles, whose masses occur within the gap. There can be particles whose mass occurs in the continuum, but only if these particles are stable against decay, due to selection rules. The gap vanishes when there are massless particles, such as the photon. In this case, there is a discrete mass in the continuum corresponding to the electron, which cannot decay into pho- tons because of conservation laws. The one-particle states of a free field can be generated by applying creation op- erators a, to the vacuum state. If we do this for a nonfree field, we will not get one- particle states, because we will not get eigenstates of P+. Instead, we will have a mixture of states involving interacting particles. Nevertheless, we can discuss prop- erties of a one-particle state through general considerations, without constructing it explicitly. We confine our attention to massive particles, with m > 0. There exists a Lorentz frame in which p = 0, called the rest frame. The spin operator S of the one- particle state is defined as the total angular momentum in the rest frame. The eigen- value of S? has the form S(S+ 1), where Sis called the spin of the particle. The pro- jection s of S along the momentum of the particle is called the helicity, which for m> Ohas 2S + | possible values S, S—1,...,—S. We can label a one-particle state by momentum p and helicity s: ||-particle state) = |p, s) (3.60) The parameters m and S are suppressed, because they are constants. For m > 0a one-particle state in the rest frame is denoted by |s), and we can ob- tain |p, s) from |s) through a Lorentz boost L(p): 3.6 One-Particle States 51 Ip, s) = UL(p))\s) (3.61) Applying a Lorentz transformation A to both sides, we obtain U(A)Ip, s) = UA)UL(p))\s) (3.62) Now insert in front of the right side the identity operator in the form 1= UL(p)U'L(p)) 3.63) and regroup the factors in the following manner: U(A)p, s) = UL@U"(LP) U(AUL(p))IIs) G.64) By the group property, the operator within the square brackets can be rewritten as. UL@)UAUL(p)) = UL-(A p)AL(p)) (3.65) which represents a pure rotation called the Wigner rotation: R(A, p) = L-!(Ap)AL(p) (3.66) It boosts a particle from rest to momentum p, makes a Lorentz transformation A, and then brings the particle back to rest. The operation has no effect on the state vector except possibly multiplying it by a phase factor, which represents a rotation. Thus, the general effect of a Lorentz transformation on a one-particle state is to boost the momentum, and rotate the spin by a Wigner rotation: U(A)Ip, s) = UL(Ap))U(R)|s) (3.67) For a more explicit representation of R, we insert a complete set of helicity states to obtain UAylp, s) = 5, ULCAR))s") s"|U)s) = > Du(R)iAp,s’) (3.68) where DJ..(R) are the rotation coefficients. An example of the Wigner rotation is the Thomas precession discussed in Section 6.8. Massless particles are special, in that there is no rest frame. A massless particle of spin S can have the values +S only. We shall explicitly demonstrate this for pho- tons in Section 5.5. A general proof may be found in books on representations of the Lorentz group (see, e.g., Tung [1]). 52 Relativistic Fields PROBLEMS 3.1 Verify the Poincaré algebra (3.31). 3.2. Verify the spin matrix (3.48) for the vector field. 3.3. Verify that the SZ(2C) matrices L(A) given in (3.23) correctly represent Lorentz trans- formations. It is necessary to verify them only for infinitesimal transformations. 3.4. Show the following identity, which is usefull when working with SL(2C) matrices: (0: Ao: B)=A-Btio- AB where the components of o-are Pauli matrices and the components of A and B are num- bers. 3.5 Consider two infinitesimal successive infinitesimal Lorentz boosts with angles g, and >. Show that the result is equivalent to a boost #; + > plus a rotation + ¢, x gs. Here, @=¥ tanhr'», where v is the velocity of the boost. Lorentz boosts. 3.6 (a) Under the action of a Lorentz boost with velocity v, a 4-momentum p# is trans- formed to p', show a PO = XPo- VP) vepty( >~ We) where y= (1-7), (b) Writing p"# = A+,p’, obtain the transformation matrix ec % Ms ” = Vor yt aj-& =f UR (©) Let L(p) be the transformation matrix corresponding to a Lorentz boost that trans- forms the rest frame of a particle of mass m into a frame in which the particle has momentum p and energy £. Show Em woy-| Pe et a- Pe m mE + m) 3.7 (a) Consider a particle of mass m and helicity s, moving with momentum p along the z axis. Make a Lorentz boost of velocity v along the x axis, so that p> p'. Find the rotation matrix Ri for the Wigner rotation. (b) Show that for an ultrarelativistic particle iP opie Reh Reference 53 That is, the Wigner rotation is the same as that taking the initial velocity p/E to the final velocity p'/E'. This shows that the helicity of a massless particle such as the photon is Lorentz-invariant. REFERENCE 1. W.K. Tung, Group Theory in Physics, World Scientific, Singapore, 1985, Section 10.4.4. Canonical Formalism 4.1 PRINCIPLE OF STATIONARY ACTION The equations of motion for a classical field can be derived from a Lagrangian through the principle of stationary action. This approach gives a unified treatment of topics discussed previously through special examples, It also makes clear that symmetries of the system give rise to conservation laws. Consider a set of classical fields collectively denoted by (x): Hx) = {h1(2), ---, bx} 4.1) We denote their space-time derivatives by lx) = 9,900) (4.2) ‘The Lagrangian density is assumed to depend on the fields and their first deriva- tives: L(x) = L(4(x), bul) (43) This will ensure that the equations of motion are second-order differential equations in the time, as in classical mechanics. We assume that, unless external fields are ex- plicitly introduced, space-time is homogeneous. Thus, £(x) depends on x not explic- itly, but only implicitly through q(x) and $,(x). We consider only local field theo- ries, for which the Lagrangian density at x depends only on properties of the field at x. Nonlocal terms of the form Je'rb.00K ale - 4.0) a4) are ruled out, unless K,,(x — y) * &(x —y). 54 4.1 Principle of Stationary Action 55, The classical action of the system is = | dtxei 4. SI] [i L(x) 4.5) where 0 is the space-time volume, which eventually goes to infinity. We impose definite boundary conditions on the surface of 1, say, # = 0. The principle of sta- tionary action, which is a generalization of that in classical mechanics, states the following: Suppose that (x) is a solution to the equations of motion. If we vary its functional form by adding an arbitrary infinitesimal function 6¢(x) that preserves the bound- ary condition: H&) > Hx) + Sx) 6b(x) = 0 on boundary of O then the variation of the action will be of second-order smallness: 8S = Sb + 86] - ST] = 0 This means that S[@] is at an extremum when q(x) is a solution to the equation of motion, To find the equation of motion according to this principle, let us calculate the variation of the Lagrangian density: or spoe a0 (4.6) Using the fact that 5, = (4,8) = 9,59) 472 we get a a&L=|— -4, 5h + 4, —— 5. 4.8) [aaa] Pera(ae) a8 The last term is a total 4-divergence. It vanishes when integrated over the space-time volume, since it then becomes a surface integral by Gauss’ theorem, and 5d = 0 on the surface. Thus eleSAg er 56 Canonical Formalism Since 5¢(x) is arbitrary, its coefficient must vanish. We thus obtain the equation of motion (4.10) where a 7 = — 4.11 a, G11) The canonical conjugate to (x) is defined by analogy with classical mechan- ics: aw max) = W(x) = a (4.12) where a dot denotes partial time derivative. The Hamiltonian density is defined by H(x) = H( mx), Hx), V d(x) = 7b L(x) (4.13) where d(x) should be re-expressed in terms of m(x) according to (4.12). The Hamil- tonian is given by H= faal00) (4.14) To quantize the system, we replace the field and its canonical conjugate by op- erators, which are defined by the equal-time commutators i[m(x, 1), HY, Dl. = RX -y) Lax, ), my, De = (4, 0, Hy, D].=0 (4.15) where for bosons we use the commutator [4, BL = [4, B] = AB-BA (4.16) and for fermions we use the anticommutator [A, B], = (4, B} = AB+BA (4.17) If a(x, t) = 0, as is the case for the electromagnetic field, then the field conjugate to a(x, 2) is not an independent dynamical variable, and should not be independently quantized. 4.2 Noether’s Theorem 57 4.2 NOETHER’S THEOREM A transformation (x) > (x) + 5d(x) is called a symmetry transformation of the system if it changes the Lagrangian density only by the addition of a 4-divergence. As we have seen, this does not change the equations of motion. More specifically, the change must be of the form 8£(x) = 4,,//(x) for arbitrary (x), regardless of whether it obeys the equation of motion. If the symmetry transformation is continu- ous, then there is an associated conserved current density. The formal statement is as follows. i NOETHER’S THEOREM [1] /f, under a continuous infinitesimal transfor- mation Hx) > Hx) + SGX) the change in the Lagrangian density is found to be of the form BL(x) = YW,0) without using the equations of motion, then there exists a current density JER) = (0) h(x) — Wer) (4.18) which, for fields obeying the equations of motion, satisfies In JMx) = 0 (4.19) Proof. We calculate the 8£(x) when the field changes by 54, using the equa- tion of motion, but without assuming that 8 comes from a symmetry transforma- tion: “ow a a= wo? 06, Oe woh d,(56) = a,(m*5$) + (35 . a.n)50 = 4,75) where the equation of motion was used in the last step. Specializing the preceding to symmetry transformations, we equate it with d,,/7* to obtain - 3,18 — W) =0 . 58 Canonical Formalism The conserved current j# is called a Noether current, and is determined only up to an arbitrary normalization factor. Noether’s theorem was proved for classical fields, and one usually extends it to quantum theory by replacing the fields in j# by the corresponding quantized fields. This does not always give a conserved current in the quantum theory. When the quantum current so obtained fails to be conserved, the nonzero divergence 4, is called a “current anomaly.” Some examples of this are discussed in Section 19.8. 4.3 TRANSLATIONAL INVARIANCE ‘An important symmetry for any system is Poincaré invariance, which is called a “space-time symmetry,” because it is associated with the transformation of the field under a change in the coordinate system. We discuss this symmetry by breaking up the Poincaré group into the translation and Lorentz subgroups. Invariance under the translation group should give rise to four independent Noether currents corresponding to the four possible space-time translations. Con- sider an infinitesimal translation who xt at (4.20) under which each component of (x) independently transforms according to $'(x + a) = G(x). The functional form of the field changes according to (x) = Px — a) = Hx) — a, 0" b(x) (4.21) We shall choose a,, to have only one nonvanishing component, say, a, and take a= 0,1,2,3 in turn. Thus, the change in functional form of the field is SG(x) = HHx) (4.22) where we have dropped the proportionality constant —a,,, because it enters all subse- quent formulas only through the overall normalization of the Noether current. Dif- ferentiating the above with respect to x“ gives 5,2) = Fh, (x) (4.23) The statement of translational invariance is that £(x) does not depend on x ex- plicitly. Hence L(x) = I b+ = #6, = PL(x) a oh %, = d,(g**L(x)) (4.24) 43 Translational Invariance 59 from which we read off Weer) = gh£(x) (4.25) The four corresponding Noether currents are denoted by 7.“, where a labels the direction of translation: T B(x) = 1d P(x) — ghAL(x) (4.26) They satisfy the conservation law 4,T (0) = 0 (4.27) This is called the canonical energy-momentum tensor. It is generally not a symmet- ric tensor. The subscript “c,” which stands for “canonical,” distinguishes it from a symmetrized version to be discussed later. The conservation law can be rewritten in the form a Gale =0 (4.28) 2 rms ot Integrating both sides over all space, and assuming that surface contributions van- ish, we obtain d arn (4.29) where Pox faxree (4.30) is the total 4-momentum. Thus, 7% is the energy density and T* the kth compo- nent of the momentum density. Their conservation laws are given respectively by the time and spatial components of (4.28): a ot Zry+ TH=0 (431) It is clear that 7 is the kth component of the energy current and 7 is the kth com- ponent of the current of the “jth component of momentum.” The latter is called the stress-energy tensor. These components are displayed in the following matrix: 60 Canonical Formalism Te Tol re 78s THa=| 73 (4.32) The identification of T as the energy density is consistent with the definition of the Hamiltonian density in (4.13). The explicit expressions for the total energy and momentum are P= fa®xfme)'4) — £09] = [d*eale0) P= faremoatete) (4.33) where a dot denotes time derivative and (x) is the Hamiltonian density. We go over to the quantum theory by replacing (x) and (x) by the appropriate operators. To show that P* generates spatial translations, we calculate the commutator of (x) with the total momentum operator Pea [aby my) AO(y) (434) This does not depend on yo. We are therefore free to choose yp = xo =f, to take ad- vantage of the simplicity of the equal-time commutators. Thus, we have 1% dex a1 form. 9°, oe] fay 80 - yee «e (435) To show that P° generates time translations, we calculate [P®, bx, 1] = J Ay {Lmty, dy. Ds HX, D] - [L(Y 1), Hx, D)} = bls, + faryt AY, HY, D, Hx, (L(y. ), OO, D]} (4.36) The integral in the second term identically vanishes. Proof. Use the representation = i8/51 to write SOY) _ ans, dy) Bate) aagay [0(y), 6@)] = Using = 46/46, we can calculate the integral as follows: L(x) Ad(x) _ x [ses amx) ma | AL(y) SHY) AL(y)] _ fa] By dmx) ae 44 Lorentz Invariance 61 Combining the preceding results, we have i[P?, $x] = P(x) (4.37) This shows that the total 4-momentum generates space-time translations. If F is a function of ¢ and its derivatives, then i[P*, F] = ®F (4.38) Choosing, in particular, F = P#, we have i[P*, P#] = 9°P#. The right side vanishes because P# is independent of space by construction, and independent of time by 4-momentum conservation. Thus [P*, P4]=0 (4.39) which is part of the Poincaré algebra discussed in Chapter 3. It is realized here in terms of the field operators. 4.4 LORENTZ INVARIANCE Consider an infinitesimal Lorentz transformation in the direction labeled by {a,8}. For example, a rotation about the x? axis corresponds to {1,2}, or a boost along the x! axis corresponds to {0,1}. Under the transformation, the functional form of the field changes by 5¢(x), as given by (3.46). We have, up to a multiplicative constant, B(x) = (x4 — xa + BP) hx) (4.40) where 52 is the spin matrix, antisymmetric in {a, 8}. Differentiating 5 with re- spect to x", we find 5d,(2t) = [x%9 — xP + BP] h(x) + (geGF - ghb%) (4.41) Using the fact that £(x) has no explicit x dependence, we have a“ a 8L= 9g apo 3g, 7% = (9 — Pa) + apne + LB, + gh mg" (4.42) The statement of Lorentz invariance is that the preceding is a 4-divergence. The first term is of the desired form, for it can be rewritten (x) — %(x8L) = d,(ghPx2L — g#x6C), Therefore the rest must vanish: SpE + ALO, + 1G? — ag =0 (4.43) 62 — Canonical Formalism With this, we have AL = d,(grPxel — grexPL) (4.44) which gives We(x) = (ghPxa — gh )L(x) (4.45) This leads to six independent Noether currents labeled by {@, 8}: MESB = ar c2® — xP + 3) bfx) — (ghPx — ger )L(x) (4.46) which can be written in the form MBF (x) = x°T#A(x) — xP TH(x) + THLE Hx) (4.47) where T#” is the canonical energy-momentum tensor. It satisfies the conservation law 8, M#°(x) = 0 (4.48) and is called the canonical angular momentum tensor. The conservation law gives rise to six constants of the motion, the Lorentz boost W= [ae Mm) = [asf He) - TEI] (49) and the angular momentum STe= sell dx MON(x) = $e/#[dPxfxt 19x) —x/T MQ) + mBIGG)] (4.50) where the first two terms represent orbital angular momentum and the last term is the intrinsic spin. It is straightforward to verify that K’ and J* generate Lorentz transformations by showing i[Jary moe, 0, des. 0]= 549 as where 5¢ is as given by (4.40). It can also be verified that the commutation rela- tions among the operators P# and M22 realize the Poincaré algebra (3.31). 4.5 Symmetrized Energy-Momentum Tensor 63 4.5 SYMMETRIZED ENERGY-MOMENTUM TENSOR The canonical energy-momentum tensor 7 #“ is not unique, because the Lagrangian density is defined only up to a 4-divergence. We can replace it by any tensor of the form T= TH+ LOX ue (4.52) where X;yq is antisymmetric in A: X ua = XH (4.53) The antisymmetry is a sufficient condition that the conservation law be unchanged: 9,T#* = 6,T#* + 4$3,9,Xsyar= 0 (4.54) A possible change in the total 4-momentum is fareroe— fareroe= 5 [atea, x0" (4.55) which vanishes for the following reasons: (1) the term with A = 0 vanishes because Xa = () by antisymmetry and (2) the terms from A = k vanish because they give a surface integral. From a physical point of view, therefore, 7 is equivalent to T¢*, because they give the same total 4-momentum, The fact that 7#“ is not guaranteed to be a symmetric tensor poses a problem, if it is to be used as a source of the gravitational field. We can, however, replace it with a equivalent symmetric tensor 7#*. The condition for symmetry is Twa Taw = [Tea Ten] + $9,(XMa— Yow) = 0) (4,56) The term in brackets can be rewritten using the condition (4.43) for Lorentz invari- ance: Tes Tee = mip — nig =p =H IA =A m'3Heg) — (35 - ames (437) where the last term vanishes by the equation of motion. Substituting this result into (4.56), we obtain the condition XAwa — ran = 2DPSueds (4.58) 64 Canonical Formalism a solution to which is Xue = (AS He — AEM — Se)” (4.59) This term is needed only when the spin is nonzero. Corresponding to the symmetrized energy-momentum tensor, we can define a new angular momentum tensor: MueB == xa THB — x6 THe (4.60) This is related to the canonical angular momentum tensor through MuoB = MusB + 29, (xAXMM8 — xPX AHA) (4.61) It is easy to show that M+? is conserved, and that it preserves the definition of the boost and the angular momentum: 4,MHe8 = 0 f. PxMoo8 = f. d2xM 008 (4.62) From a physical point of view, therefore, M*“8 and M4? are equivalent. 4.6 GAUGE INVARIANCE In contrast to space-time symmetries, there are internal symmetries, which are x-in- dependent transformations of the field that leaves the Lagrangian density invariant. ‘A simple example is a change of phase in a complex scalar field: x) > EF plx) We) > eye) (4.63) This is called a global gauge transformation, where the label “global” refers to the fact that @ is independent of x. Invariance with respect to it means that the La- grangian density is independent of the phase. This is true for the free field, in which the fields appear in the combination y"(x)ix) or y'(x)4,1Mx). The infinitesimal form of the transformation is Sx) = Wx) Sy") = We) (4.64) where we left out a proportionality constant ia. In terms of the real and imaginary parts defined by Problems 65 He) = Fy ldale) + ibst0)] Va) = Fylh e609] 465) the transformation is a rotation in internal space: /(&) = d(x) cosa + py(x) sina 2(x) = —d;(x) sina + (x) cosa (4.66) The Noether current is just the conserved current mentioned in Chapter 2: a ¥ © 45,0) = yore — yr oy 2( doh, — bho) (4.67) More generally, internal symmetries are linear transformations for a multicompo- nent field $,(x), (a= 1, ...,K), of the form a(x) = Carbs) (4.68) where C,, are elements of a K x K constant matrix. If the matrix belongs to a K-di- mensional representation of some group G, we call G an internal symmetry group. The group can be continuous or discrete. In the previous example G = U(1), the uni- tary group of dimension 1. Physical examples of conserved charges are © Electric charge = positive minus negative charge © Baryon number = number of protons minus number of antiprotons Electron number = number of electrons minus number of positrons A important case is isospin, which is discussed in Section 7.5. PROBLEMS 4.1 Consider a field (x) with Lagrangian density £o(x) +.£,(x), where the first term has a certain symmetry, while the second term does not. That is, under a transformation 44x) > Hx) + 640), - Lox) = HW, (x), 66 Canonical Formalism whereas £,(x) cannot be put into this form. If £,(x) were absent, the system would have a conserved Noether current, Show that, in the presence of L,(x), the divergence of the would-be Noether current is 82; (x). 4.2 A condition for Lorentz invariance is (4.43). For scalar fields, for which 2° = 0, what restriction does this place on the Lagrangian density? 43. Nonrelativistic System A nonrelativistic many-particle system has a second-quan- tized Hamiltonian H= fox V00(-a5 vit )400 where jis the chemical potential. Usually one assumes the commutation relation [Y(x), W(y)] = (x —y). We want to see whether this is consistent with the canonical formal- ism (a) Find the equation of motion using i= [W, H]. (b) Regard H as a classical Hamiltonian, Show that the corresponding Lagrangian den- sity is ca inragiars (3, 2-H) o Work out the equation of motion using the canonical formalism. (©) Show that the usual commutation relation [¥(x), ¥#(y)] = &(x— y) is canonical. (@) Work out the Noether current associated with space-time translational invariance, and global gauge invariance. 4.4. Field Representation Since y(r) and iy*(r) are canonical conjugates in the nonrela- tivistic system, it would be awkward to introduce the field representation by diagonaliz~ ing Yd). Show that in this case we can put 1 6 e)= |" a | 6 ban) veo al - where g(r) is a c-number function, 4.5. First-Order Lagrangian The nonrelativistic Lagrangian in the last problem differs from a relativistic one, in that it is first-order instead of second-order in the time deriva tive. The Dirac field discussed in Chapter 7 also has a first-order Lagrangian. To fully explore the consistency of the canonical formalism, let us strip the problem down to bare essentials, and consider a classical system with two coordinates a and 6, which are like iy* and y. Take the Lagrangian to be L(a, b, b)=ab-Va, b) The canonical rule says that a has no canonical conjugate, It is the conjugate to b. Is this, completely consistent with the Lagrangian and Hamiltonian equations of motion? (a) Find the Lagrangian equations for motion fora and. Reference 67 (b) Find canonical momenta and the Hamilton equations of motion, Check that they are the same as the Lagrangian ones. (©) Itis thus completely consistent to regard a and b as canonically conjugate. To quan- tize the system, impose (4, 6], =—i. REFERENCE 1. E. Noether, Nachr. kgl. Wiss. Géttingen 235 (1918); E. L. Hill, Rev. Mod. Phys, 23, 253 951). Electromagnetic Field 5.1 MAXWELL EQUATIONS The classical electromagnetic field is described by two 3-vector fields, the electric field E(x, t) and the magnetic field B(x, t), which obey Maxwell’s equations. In ra- tionalized units with c = 1, they read VE=p VB=0 (6.1) where p(r, 1) and j(r, £) are respectively the external charge density and external cur- rent density, which must satisfy the continuity equation 4 +Z (5.2) The second and third equations are solved by introducing the vector potential A(r, #) and scalar potential (r, 1): B=VxA E=-Vg-— (5.3) whereupon the remaining two equations become 68 5.1 Maxwell’s Equations 69. $ 2 j At ( -v)A i-v(va+ Ze) 6.4) The potentials are determined only up to a local gauge transformation, which in- volves an arbitrary function x(r, ‘) ASAt+VX Ox b> 6- 63) The Lorentz gauge corresponds to the condition V-At # =0 — (Lorentz gauge) (5.6) In this gauge, both potentials satisfy the wave equation: é 2 (a> 7-3 & 2 = (-W)e-e 6.7) The symmetric appearance of these equations is sometimes convenient, but it actu- ally obscures the physics. These equations seem to indicate that there are four inde- pendent propagating modes, but actually there only two—the transverse compo- nents of A. This can be shown by going to the Coulomb gauge. In Coulomb gauge (or radiation gauge), A is purely transverse: V-A=0 — (Coulomb gauge) (5.8) The equations for the potentials become (re V?b=-p (3.9) where j, is the transverse current density 70 Electromagnetic Field in-j-2V¢ 5.10) in=i-G 6. which satisfies V-j; = 0. In this gauge, A describes transverse electromagnetic’radi- ation, whose source is the transverse current density, while describes the instanta- neous Coulomb interaction between charges. The potential between two unit charges located at r, and r, is given by _ i Aqir,— 1) or) (3.11) To show that we can always impose the Coulomb gauge, suppose V-A =/. To go to Coulomb gauge, we make the gauge transformation A > A+ V x, with y satisfying Vx =-f. The solution corresponds to the statement that x is the electrostatic po- tential due to the charge distribution V-A. We are using rationalized instead of unrationalized units. The difference be- tween these systems arises from the normalization convention for the free fields, and is tabulated as follows: Rationalized __Unrationalized 4-Current je 4aje Coulomb’s law gir Paar Energy density (E2 + B’y72 (E2+ B87 Field operator A 4nA. 5.2. COVARIANCE OF THE CLASSICAL THEORY ‘We postulate that the potentials form a 4-vector A= ($A) (5.12) and this determines how Maxwell's equations transform under a Lorentz transfor- mation. Since we always impose a gauge condition, a Lorentz transformation must be accompanied by a gauge transformation Ab —> AB Ay (5.13) in order to maintain the gauge condition. Under an infinitesimal Lorentz transfor- mation, therefore, the vector potential transforms according to A'#= AH wp” — dy (6.14) a 5.2 Covariance of the Classical Theory 71 where y is such that 4'# satisfies the gauge condition. The Lorentz gauge 0,,4# = 0 is covariant, and the equations of motion take the form (5.7), which are manifestly covariant ceae (6.15) where jis the conserved 4-vector current density (D4, jn=0 (5.16) In this gauge, however, the physical degrees of freedom are not manifest. In Coulomb gauge, where physical degree freedom are made explicit, the equa- tions of motion (5.9) are not manifestly covariant; but they actually are, because there always exists a gauge transformation to maintain the appearance of the equa- tions in all Lorentz frames. One has to choose between manifest covariance with Lorentz gauge, or manifest transversality with Coulomb gauge, and we choose the latter. The electric and magnetic fields are components of the antisymmetric field tensor Fr gud” — AH (5.17) which is gauge-invariant. The dual field tensor is defined as For = Len Fag (5.18) In terms of the electric and magnetic fields, we have PO=E Pia —eitpt =~4el/P) (5.19) The components of the field tensor and its dual can be displayed as matrices: 0 -f -P -B pow( BOO, OR (5.20) BB Bl oO 0 Bl -B 8 Fur= B 0B -B (5.21) - 0 OE! B EF -E 0 2 Electromagnetic Field We see that F*” is obtainable from F#” through the duality transformation {E, B} > {B,-E} (5.22) From the field tensors we can form two independent Lorentz invariants: 4 F!F,,= 5 (BP-E*) (scalar) + PMR, =-BE (pseudoscalar) (5.23) In terms of the field tensors, Maxwell's equations read 6,FHY = (5.24) which are gauge-invariant and Lorentz-covariant, and are invariant under the duality transformation when j” = 0. Since F#” = —F, the first equation is consistent only if ,j" = 0. The second equation is identically satisfied by putting F“”= d¥A— PAM. 5.3. CANONICAL FORMALISM ‘The Lagrangian density of the free electromagnetic field is E? — B?) (5.25) PYr,, Apart from an overall factor, this is uniquely determined by the requirement that it be Lorentz- and gauge-invariant, and does not contain higher derivatives of 4¥ than first derivatives. The minus sign in front is chosen to give a positive energy density for the free field, and the factor + sets the normalization of the fields. To obtain the equations of motion from the action principle, we must use the potential 4” as the field variable. The Lagrangian density then reads = AMAA, (5.26) from which we obtain a m= =-Fe 5.27 0,4) oe The equation of motion is 5.3. Canonical Formalism 73 a o,meY — oA =0 (5.28) Since the last term on the left side is zero, we have 0,FH*= 0 (5.29) The use of the potential makes d,, 74” = 0. Thus, we correctly recover Maxwell's equations for free fields. The canonical conjugate to 4” is 7 -Fe (5.30) which vanishes identically for v= 0, indicating that 4° is not a dynamical variable. The dynamical fields are 4‘, with canonical conjugate ~F° = E*, However, the lon- gitudinal part of A has no physical significance, because it can be changed at will through a gauge transformation (see Table 5.1.) The only dynamical degrees of free- dom are the two transverse components of A, and we can go to the Coulomb gauge to make this explicit. In Coulomb gauge 4° satisfies the Poisson equation, and is de- termined by the external charges. The canonical energy-momentum tensor is, according to (4.26), TES = He, — ghOL = FHA, — geal (531) which can be rewritten using the equation of motion: TE, = FMF ay BEL 0(F HAG) (5.32) The last term is not symmetric in u and a, and not gauge-invariant. However, it is a total 4-divergence antisymmetric in 1 and v, and is conserved because 4,,0,(FH¥A*) = 0. As discussed in Section 4.5, such a term has no effect on the conservation law and the definitions of total energy and total momentum, and may therefore be omit- ted. Thus we take as energy-momentum tensor the symmetric and gauge-invariant tensor Th =-FYF ay — Bah (5.33) TABLE 5.1 Fields and Canonical Conjugates Field Canonical Conjugate Remark a 0 Not dynamical variable A - Only transverse part physical 74 Electromagnetic Field which satisfies The trace of the tensor vanishes: It is now straightforward to obtain the Hamiltonian density: H= 1% =—F%Fy,—£ = + (BE? +B?) The momentum density (the Poynting vector) is Sk = [0 = FUP = eHiEB) The total Hamiltonian H and total momentum P are given by H=t Jarre +B?) P= fa Sr Ex B The conservation of energy and momentum correspond to the statements 2 H+TS=0 Sst+apt=0 where Dk = (EVES + BIBS) + 5 8y(E? + B®) is the stress tensor. According to (4.47), the generalized angular momentum tensor is Moab = xaTuB— x8 Tua which satisfies the conservation law 6,Mu2b = 0 (5.34) (6.35) (5.36) (537) (5.38) (6.39) (5.40) (5.41) (5.42) It follows that the total angular momentum J and the Lorentz boost K are given by 5.4 Quantization in Coulomb Gauge 75 I= fatrrx (EB) K=®-rH (5.43) 5.4 QUANTIZATION IN COULOMB GAUGE To quantize the electromagnetic field, we must first eliminate all unphysical de- grees of freedom by fixing the gauge, and in the following we shall use Coulomb gauge.' In the absence of external charges, we can set 4° = 0, and write the Hamil- tonian in the form t far (E+|Vx AP) (V-A=0) (5.44) The canonical conjugate to A is -E = 2 A/ét, and we would normally impose the equal-time commutation relation [(r, #),A"(r’, #)] = 18,5°(r - r’). But this is incor- rect here, because the right side is not consistent with V-A = 0, nor with one of Maxwell’s equations V-E = 0. We therefore replace 5,,5°(r — r’) by its transverse projection, and take [Pr 0, AM’, 0) = Bh (r-r’) (5.45) where the transverse delta function 8 1,(r— r’) is defined by &k Kk an0)= | Sap (5 je et (5.46) and satisfies aIBT(r) = T(r) = 0 (5.47) ‘A complete set of solutions to Maxwell’s equations in a periodic box of volume are the transverse plane waves (ker a = [kl where €(k) is a unit polarization vector normal to k. For each k, there are two inde- pendent polarization vectors €,(k) and €,(k); and €,(k), €(k), k together form a right-handed coordinate system: ke,(k) = ke,(k) = 0 'For quantization in other gauges, see Huang [1]. 76 Electromagnetic Field €,(k)€(k) = 5, k i(k) * €,(k) = 610) * el Ty (5.48) Having chosen €;(k), €,(k), there is still arbitrariness in the choice of €,(-k), €,(-k). By convention, we choose (k= x(k) €,(-k) = -€,(k) (5.49) as illustrated in Fig. 5.1. The following sum over polarizations results in the trans- verse projection operator (see Problem 5.2): Kei md (5.50) 2 Ti(k) = Y(W)ef(k) = 5, — A We now expand the field in terms of the transverse plane waves: 2 el([a(kye™* + a,1(Ke*"] 1 409-2 at 4,0) | Yettolatve*— aoe] 651) The commutation relations (5.45) are satisfied by imposing the commutation rela- tions E(k) Ex(k) E,(-k) E,{k) Figure 5.1 Polarization vectors of an electromagnetic wave. 54 Quantization in Coulomb Gauge 77 {4,00}, @7()] = 845kp [4,(),a,(p)] = 0 (5.52) where a,(k) is the annihilation operator of a photon—a field quantum of momentum k and linear polarization s, In the free-field case, the time-dependent operator A(t, t) is simply obtained by replacing (Kr by (kr — af) in the exponents in (5.51), because eltia(p)e = a,(p)e ">" (5.53) The Hamiltonian and the total momentum operator of the electromagnetic field are given respectively by t 2 (kl[a3d04,0) + +] P= kal(ka,(k) (5.54) iG These equations show that photons are boson with energy-momentum relation @, = |k|. The vacuum state |0) is the state with no photons. All other states of the system can be generated by applying creation operators repeatedly to the vacuum state. In the limit Q — , the expansion (5.51) becomes a Fourier integral: ak 2 Air, 0) = ospsvaas 2 Ath s)[a(k, s)e®*™ + at(k, s)e*™] (5.55) where we write e(k, s) = €,(k) for consistency in notation. The continuum form of the annihilation operator is a(k, s) = VQa,(k) (5.56) which obeys the commutation relations [a(k, s), at(k’, s')] = 2m)°5,.5(k — k') (8.57) The Hamiltonian and total momentum now take the forms H=~ fee Japs ila ak, 5) P= S j= ea ark sdall) (5.58) 78 Electromagnetic Field 5.5 SPINANGULAR IMENTUM According to (5.43), the angular momentum density is u=rx[Ex(V xA)] (5.59) We define the spin density to be the part that is independent of the origin of r. To find it, let us first rewrite the preceding in component form: uf = etkelinennay/B!9,A9 (5.60) Now combine the last two € symbols according to the rule etimgnam = 5, Big — BigBin (5.61) We then obtain ui = eW'Q/Et9, A? — VEG, A*) = eV IEIA, AY — dy(x/E"AS) + xI(d, EB") Ak + EA] (5.62) The factor »/ in the first term cannot be removed by manipulations involving 4,, be- cause j # k. The second term is a total 3-divergence, and can be ignored. The third term vanishes because V-E = 0. The last term is independent of r, and is identified as the spin density: s=EXA (5.63) This is the spatial part of the tensor sh? = AS ae (5.64) where 223 is the spin matrix given in (3.49). The spin angular momentum is given by s=arExa (5.65) Using the expansion (5.51), we obtain S= d c{an(boy'ay(k) - af Qax(k)] (5.66) where k = k/|k|. To diagonalize this, we make a linear transform to circularly polar- ized photons. 5.5 Spin Angular Momentum 719 The polarization vectors for circularly polarized photons are 1 > AG tie) (5.67) where the label k has been suppressed. As one can easily verify, the plane wave Refe,ci0%] (5.68) represents a traveling wave whose polarization vector rotates in a right-handed sense about k. This is called a lefi-circularly polarized wave, because an observer facing the incoming wave would see the polarization rotating to the left. Similarly, € corresponds to a right-circularly polarized wave. The annihilation operators for a circularly polarized photons are given by Left-circularly polarized: a,(k) = Slat ~ia(k)) Right-cireularly polarized: a.(k)= Zefayh)+iasQ] (5.69) The commutation relations are (a.(86), @2(p)] = Bap [a,(k),a(p)] = [a.(k), a%(p)] = 0 (5.70) In terms of these, the spin operator becomes diagonal: S= 2 Kc[a,(e)'a, (06) ~ a1(k)a.(49)] (3.71) This shows that the photon has spin 1, but there are only two helicity states. The he- licity +1 corresponds to left-circular polarization, and —I corresponds to right-circu- lar polarization: a,(k) annihilates helicity state + 1 (6.72) In terms of circular polarization, the field operator has the expansion 1 oo ler A(r, 0) 2 Viati {[e.(k)a,(k) + €(k)a_(k) Je’ + [e#(k)al(k) + €*(k)at(k)] er} (5.73) In the convention (5.49) the sense of the circular polarizations remains unchanged when k > -k: 80 Electromagnetic Field €.(k) = ie,(k) €(-k) = -i€(k) (5.74) 5.6 INTRINSIC PARITY Let us make a coordinate transformation r — r’, with the transformation law 3 => pitr’t (5.75) fi Since A* is a vector field, this induces the unitary transformation U according to (3.54): 3 UAi(n)U"! = Spite’) (5.76) A where 4/(r) = 4(r, 0). For spatial reflection r’ = —r, we denote the unitary transfor- mation by ?: PAM EYP! = Ar) (5.77) This establishes the fact that the electromagnetic field has odd intrinsic parity. To investigate how photon states transform, we substitute into the preceding the expansion (5.51). Using the abbreviation 2 a(k) = 26004,(8) (5.78) we have PAP! = > Fggn PAVE e+ Pat(k)P tet] = 2) aa Fla eB + at eT] => ca Fla We + atte] (5.79) © where the last relation is obtained by changing the summation variable from k to ~k. Thus Pa(kyP"' = -a(-k) (5.80) 5.7 Transverse Propagator 81 which gives Pa,(k)P-! = —a,(-k) Pa,(k)P"' = a,(-k) (5.81) In terms of circular polarization, we obtain Pa,(k)P"' = —a-(-k) (5.82) A one-photon state of momentum k, linear polarization s, is defined by Ik, s) =a 3()/0) (5.83) States with circular polarization are given by Ik, +) = aJ()/0) (5.84) which are linearly superpositions of states with linear polarizations: 1 4+) = =s5[|k,1) +] e Ik, +) Valk, ) + i]k,2)] (5.85) Assuming that the vacuum state is invariant under reflection, we have Ik, +) = Pa}(k)P|0) = —|k, *) (5.86) Thus, under spatial reflection, left and right are interchanged, and the state vector changes sign. 5.7 TRANSVERSE PROPAGATOR We now calculate the photon propagator in Coulomb gauge: Di(x) = -i(O|TA‘(x)4/(0)10) (5.87) where the subscript “T” reminds us that the field is transverse: #4*= 0, Expanding the field in creation and annihilation operators, we have i d?kd?k! f Ola'(k, s)ait(k’, s‘)]0e** (vy > 0) Died= a) Vaeat {tie "che >) 688) 82 Electromagnetic Field where o=ky (5.89) and we use the abbreviation a'(k, s) = e(k, s)a(k, s) (5.90) The vacuum expectation values are easily calculated: (Ola‘(k, s)a"(k’, s°)]0) = (Ola’(k’, s")a't(k, s)|0) = (277)36,,-5(k — k’)e(k, s)e(k’, s’) (5.91) Therefore Di) =- @ a je EE eet i(k) (5.92) where /! is defined in (5.50). This can be rewritten as a four-dimensional Fourier in- tegral, with the help of the identity eo ea . sal Mam IM) (5.93) 20 The final form is elke Dil) = lee B+in Mk) (5.94) where 42 = ko? |k?. The Fourier transform is Bim = Cad ) (5.95) Fral® ae This is not Lorentz-covariant, for it is in Coulomb gauge. To prove that the quan- tized field theory is covariant, we should exhibit the gauge transformation that will maintain the form of the transverse propagator under Lorentz transformations. However, this is unnecessary, as we shall show in Chapter 11. The point is that non- covariant part of the propagator is physically irrelevant, because, owing to current conservation, it does not contribute to the scattering amplitude. 5.8 VACUUM FLUCTUATIONS The vacuum state is neither an eigenstate of E nor B, since these operators annihi- late or create photons singly. Although the fields average to zero, their mean-square 5.8 Vacuum Fluctuations 83 fluctuations are large. This can be shown via direct calculation, as in Problem 2.8. We can also demonstrate it through the following argument. The energy density in the vacuum state is + (0|E? + B?|0) = (O|E?|0) (5.96) for the free-field theory is invariant under the duality transformation. Equating this with the zero-point energy per unit volume in (5.54), we have 1 1 (OIE) = 55D = Gop [HRI 697) which diverges because of the short-wavelength modes. This divergence is harm- less, since only energy differences have physical significance; but the long-wave- length part of the fluctuations gives rise to observable effects, including the Casimir effect. We illustrate the essence of the Casimir effect in a simple one-dimensional ex- ample, leaving for the next section a more detailed treatment. Consider the modes of a harmonic oscillator in a box of length L. The zero-point energy is E(L)= +> flo) o=— = (n=1,2,...%) (5.98) where we have introduced a cutoff function (w), with the properties A0)=1 Ko) =>0 (5.99) There is a cutoff frequency «,, above which /(w) decreases rapidly to zero, and we take the limit w, — © eventually. Suppose that a partition is inserted, such that nor- mal modes are required to have a node at the wall. The modes near the cutoff fre- quency are hardly affected, because their wavelengths are vanishingly small. There- fore, there are now fewer normal modes below the cutoff, as illustrated in Fig. 5.2, and the zero-point energy decreases. For definiteness, choose the cutoff function to be Slo) = ewe (5.100) The zero-point energy for a box without partitions can be easily calculated, with the result 84 Electromagnetic Field Figure 5.2. When a wall is inserted into a box, those normal modes that do not have a node at the wall are suppressed. Consequently, the number of modes below a fixed frequency decreases, and the zero- point energy is lowered. 7 1 Loe 7 ie ne ae > BL sinh(mlaL) = Qn? Dap POW) — 101) Eg(L) = Now insert two partitions centered about the midpoint, separated by distance a. The box is divided into three compartments—one with length a and the others with length (L - a)/2—and the zero-point energy becomes Lez 7 L-a 7 = F(a) + 2p — SE - ; B(@)= Ela) + 2Eo>~ 8 Fe ~ GLa) ~ Tha (5.102) In the limit L — &, the attractive force between the walls is given by dE@)_ Ga (4a? 6.103) which is independent of the cutoff. 5.9 TheCasimir Effect 85 5.9 THE CASIMIR EFFECT We now calculate the force between two metallic plates in the electrodynamic vacu- um. The first task is to obtain the normal modes of the electromagnetic field in a perfectly conducting box of size a x b x c. We choose one corner of the box as ori- gin, and use Coulomb gauge. On each face of the box, the boundary condition is Ej=0 B,=0 (8.104) where the subscripts || and + denote respectively the tangential and normal compo- nents. We put B= V x A, E=-A, to obtain Aj=0 (VxA)L= (5.105) On the y-z plane, for example, the boundary conditions are A,=A,=0 4,4,-6.4,=0 The first says that A is normal to the surface, and therefore the second condition is automatically satisfied. We must, however, satisfy the gauge condition 4,4, + 8,4, + 4,4, =0 (5.106) which leads to A (5.107) Thus, the boundary conditions in Coulomb gauge are Ay=0 (5.108) For A,, for example, the conditions are [44s ¥, 2) =0 = [84:0 ¥, Zena = 0 Axx, ¥, 0) = AxQ y, a) = 0 AG Y, 0) = ALG y, €-) = 0 (5.109) A complete set of solutions to the wave equation is given by A,= + cos(k,x) sin(ky) sin(kz) 86 Electromagnetic Field in(k,x) cos(k,y) sin(k,z) in(k,x) sin(k,y) cos(kz) (5.110) where mn, f= om b marc (5.111) with n,= 1,2,..., %. The frequency is given by «y= VET +R (6.112) If all three components of k are nonzero, there are two independent solutions corre- sponding to the + signs in 4,. If any component of k vanishes, there is only one so- lution, For example, if &, = 0, then 4, = 4, = 0, and the + sign does not make any dif- ference. We can now obtain the zero-point energy: Eo(a,b,c)= 4 Y Vi +R F(VE +) (5.113) oe +S Vi+B+BRVE +R +R) (5.114) Keoliyskz where F(k) is a cutoff function. Consider now a large cubicle box of edge L, which is divided into three com- partments as shown in Fig. 5.3, with two parallel metallic plates inserted normal to the x axis, separated by a distance a, symmetric about the midpoint. The zero-point energy is the sum of those of the compartments. That of the middle compartment, of dimensions L x L x a, with L > ©, is given by L Figure 5.3 Two metallic plates separated by distance a in the electrodynamic vacuum, which is repre~ sented by a cube of edge L > =. 5.9 TheCasimir Effect 87 oe) u@= Pe LL) “x = > sh oie Fb) + 25, (5.115) where oo [raver 2) (5.116) We can rewrite the n-sum using the Euler-MacLaurin formula [2] > - 1 Br cng) Ba girs Sate [am G(n) + 5 GO)~F-G'O)- FG" +++» 6.117) where B, = 4, By =—ss. Using G’(0) = 0, G'""(0) = -4, and the fact that all higher derivatives vanish at n = 0, we obtain Som -[a Gin) + + [ ovor( 22) - = a This leads to n Ut@)= 2 a+ - es | (5.118) where C= afar [aeevere Fe V4mlo 4 1 C= ge [deere (5.119) The zero-point energy in the box in Fig. 5.3 is given by E(a) = Ula) + 2U(L - a2) = ela +20,- +O) mal This gives an attractive force per unit area between the plates: 88 Electromagnetic Field 1 @)_ “Pa 20a" oe or, in practical units, whe 01 f= aa - “2 dyn/om? (5.121) where a is in micrometers. Figure 5.4 compares this result and early measurements [3], with reasonable agreement. More recent measurements of a similar force be- tween a plate and a sphere have achieved much greater experimental accuracy [4]. 5.10 THE GAUGE PRINCIPLE We now discuss how the electromagnetic field should be coupled to charged fields. A nonrelativistic charged particle obeys the Schrédinger equation 1 (V +ieAP +e] iar, 1) cue. (5.122) 0.10 0.05 0.01 0 0.5 1.0 1.5 2.0 d (micron) Figure 5.4 The Casimir attractive force between two metallic plates in the vacuum: x—chromium steel, o—chromium; solid line—theory. [Data from M. J. Sparnay, Physica 24, 751 (1958).] 5.10 The Gauge Principle 89 where A, ¢ are respectively the vector and scalar potentials of an external electro- magnetic field and e is the charge of the particle. We can derive the form of the in- teraction as follows. In the absence of external fields, the Schrédinger equation is invariant under a global gauge transformation Ur, )—> eur, t) (8.123) where « is an arbitrary real constant. The invariance depends on the fact that dy) transforms in the same manner as y. If we make a local gauge transformation, with dependent on r, f, this condition will not hold, for we have Hy Cae i( Hoy] (5.124) To make the equation invariant, we must cancel the terms involving de. This can done by introducing the fields 4 = (A,¢) through the replacement 2 ys) {2 + ieanca)] 4a) (6.125) 2 yo [© + ier] 100 ; a, X, The Schrédinger equation is now invariant under the local gauge transformation AMX) > AMR) + dF Y(x) Wx) > expl-iex) W(x) (5.126) The quantity Deft) = [a + ieA*(x) Wx) (5.127) is called the covariant derivative, A» is called the gauge field, and the recipe for re- placing 4 by Dé is called the gauge principle. ‘Actually, the gauge principle works only for a fully relativistic theory. For the nonrelativistic Schrédinger equation, it fails to produce magnetic moment terms of the form —p-V x A, which has to be put in by hand, with 2 arbitrary. In the rela- tivistic Dirac equation discussed in the next chapter, the gauge principle gives the full electromagnetic interaction of the electron, with a completely determined mag- netic moment. ‘As a relativistic example, consider the complex scalar field with Lagrangian density Lox) = Bey ep mn? rp (5.128) which is invariant under the global gauge transformation Wax) > etx) (5.129) 90 Electromagnetic Field where w is a constant; but it is not invariant when w depends on x. To extend the symmetry to local gauge invariance, we make the replacement Di = 9 + iedH(x) (6.130) where e is the electric charge. The Lagrangian density is generalized to L(t) =-4 FHF, + (Dey) "Dey (6.131) which is invariant under the local gauge transformation G(x) > eR POR) AM(x) > AMX) + dEX(x) (5.132) where x(x) is an arbitrary space-time function. The Lagrangian density of the free electromagnetic field is included to make the system self-contained dynamically. PROBLEMS 5.1 The Lagrangian density for the electromagnetic field in the presence of an external cur- rent density j# is L=-3 FOR, JHA, ‘What is the condition on j# for this to be gauge-invariant? Consider the symmetric tensor u 5. 2 T=) eta 4 (@) Show that #7! = kT! =0, and T= 2. (b) Using the preceding conditions show the statement in (5.50): Kd Hb) = 8, Ts 5.3. Verify that the field operators (5.51) satisfy the commutation relations (5.45). Show, in particular, that the transverse delta function arises from the transversality of the polar- ization vectors expressed by (5.50) . Rotations Apply the transformation law (5.76) to rotations. In particular, let R be a r0- tations of the coordinate system about z axis through g, and ¢ be that about the x axis through 7. 5. = Problems = 91 x=x' cos p+y' sing xx’ R:y y=-x'sin 9+ y'c0s @ Ly yay za-z' ‘Show that the creation operators transform as indicated in the following: Ral(k)R* = e¥al(k) Eal(K)E" = al) Ral(lQR* = elval-k) gale '= al) Ral(k)R" = el¥al(k) gale? Rat(-K)R! = eieal-k) al(-K)E = 5.5 Two-Photon States [5] We can obtain interesting information about a state of two photons by examining its behavior under rotations and reflection. Consider two photons with momenta k and ~k. There are four independent states of polarization, which can be classified according to circular polarizations: I++) = ala!) 1) =al(ia!(-t0)0) F) = al(kyalt9)0) +) = al(keat(-4))0) (a) Verify that, in terms of states with linear polarization, = [a{(d0a{(-k) - a}(00a}(-K)] (0) A+) = [al doiaito) + a}(ka}(-¥)] [0) H+) = [al QoatK) + a}(k)ah-Ks) + ia} (k)a}(-A0) ~ ia}(tepa}(-K)] |0) I) = [al Wal (A) + a}(he)a}(H) — ia}()a}(Ao) + ta}(k)a{(-k)] |0) HH) +S) From this, note that the polarization of the two photons are correlated: © Inthe state |++)+|-—) the planes are parallel. © In the state |++)-—) the planes are orthogonal. In the states |+-) and |-+), the planes have equal probability of being parallel or orthogonal. Work out the transformation laws for the four polarization states under R,&P, using results of the last problem, and the fact that the vacuum state is invariant. Verify the results summarized in the table of eigenvalues (listed whenever the state in ques- tion is an eigenstate of the operation indicated): (b) Byte) KH) I) EY) 1 1 ete ete él 1 Pol -1 1 o4 92 Electromagnetic Field (©) From the preceding table, verify the following quantum numbers for a two-photon state: © The only state with odd parity is |+)~). There are three states with even par- ity: }+4)+1-), |+-), and [-+). For odd total angular momentum J= 1,3,5,... ., the only possible states are |+~) and |-+). The reason is as follows. The other two states are both eigenstates of R and € with eigenvalue 1. However, an initial state that is an eigenstate of R with eigenvalue 1 must have the rotation properties of the spherical harmonic Y/°, and therefore changes sign under € for J = 1,3,5,. . © For total angular momentum J = 0,1, the only possible states are |++) +|-~) and [++}1--), because the other two states have spin projections + 2 along the z axis, values that are too large for J= 0,1. (@)_ Verify that a two-photon state cannot have J= This gives Yang's selection rule [6]: A spin J particle cannot decay into two photons, For example, just by observing that the 7 meson decays into two photons, we can conclude that its spin cannot be 1. (Its, in fact, a spin 0 particle.) 5.6 Dirac Monopole A magnetic monopole has a magnetic field Baie = gf/r?, with total magnetic flux 47 g. Accommodate such a magnetic field into Maxwell’s equations in the following manner. To keep V-B = 0, postulate that there is a return flux -47 g con- centrated in an infinitely thin string attached to the monopole. The vector potential then consists of a part due to the monopole, and a part due to the string: A= Apa + Asring where A,oie is any vector potential that satisfies VX pote = Byring, and is, of course, de- termined only up to a gauge transformation, (a) Give one solution for Apo. (b) The shape of the string can be changed through a gauge transformation. For a straight-line string leading from the monopole to infinity, show that the vector po- tential of the string is of the pure-gauge form Ascing = 28 V8 where @ is the azimuthal angle around the string. (©) Consider a quantum-mechanical particle of electric charge e in the field of the monopole, with wave function y. Show that the string can be transformed away through a gauge transformation Ym erieety ()_ Since w has to be single-value, the coefficient of 8 in the exponent must be an inte~ ger n, and thus ge=ni2 References 93 ‘This is the Dirac quantization condition. The mere possibility that a monopole can exist quantizes the electric charge. (©) Show that the total angular momentum of the system consisting of a charge e and a monopole g points from the charge to the monopole, and has the magnitude ge. Ob- tain the Dirac quantization condition by quantizing the angular momentum. 5.7, Cutoff Functions (a) Calculate the vacuum energy (5.98) for a one-dimensional system using a sharp cutoff, which corresponds to f(w) = (0, - @), and show Lot 3 E(a)= [5 -=F* (harp cutoff) Since this is independent of a, there will be no force between inserted walls. (b) Show, on the other hand, that any continuous cutoff function will have a nonzero cutoff-independent force. To do this, write cao 50S) Since the argument of f approaches a continuous variable in the limit w, —> 2, we can approximate the sum by an integral, using the Euler-MacLaurin formula 6.117): Se el EelL)= Henna aE - Let nae 2 = 2 Jarygy- se + 00% The cutoff-independent term is the same as that in (5.101). REFERENCES 1. K. Huang, Quarks, Leptons, and Gauge Fields, 2nd ed., World Scientific, Singapore, 1992, Chapter 8. 2. M. Abramowitz and I. A, Stegun, Eds., Handbook of Mathematical Functions, National Bureau of Standards, Washington, DC, 1964, p. 806. M. J. Sparney, Physica 24, 751 (1958). S. K, Lamoreaux, Phys. Rev. Lett. 78, 5 (1997). . C.N. Yang, Phys, Rev, 77, 242 (1950). . C. N. Yang, Phys. Rev. 77, 242 (1950); L. D. Landau, Dokl. Akad. Nauk (USSR) 60, 207 (1948), away Dirac Equation 6.1 DIRAC ALGEBRA A relativistic wave equation must treat space and time on the same footing. The Klein-Gordon equation does that, but it involves second time derivatives, a feature responsible for its failure as a one-particle equation. Dirac tries to remedy this by proposing a first-order differential equation. To obtain a equation for the wave func- tion ythat is linear in the space-time derivatives 2), Dirac writes (éy#d,, — m) Wx) = 0 (6.1) where the +" are numerical coefficients, so far undetermined. To satisfy the rela- tivistic kinematics, yx) must also satisfy the Klein-Gordon equation. Multiplying from the left by (yd, +m), we have 0= (ya, + m)(éy*9,, — m2) = rtd dy + MyM) = LECH Yt YG ud, +m] Wx) (6.2) This reduces to the Klein~Gordon equation (CP + my Wx) =0 (6.3) if and only if HY + P= 2h" (64) This algebraic relation defines four objects “, which anticommute with one anoth- er, with 94 6.1 Dirac Algebra 95. oy OP =-1 (6.5) Clearly +“ cannot be numbers. They can be represented by matrices, called Dirac matrices. According to (6.4) *y” should be a Hermitian matrix. Thus “ is either Her- mitian or anti-Hermitian. Putting « # v and taking the trace of both sides in (6.4), we obtain Try=0 (w= 0,1,2,3) (6.6) This condition immediately rules out matrices of odd dimension. It also rules out dimension 2, for there are only three independent traceless 2x2 matrices—the Pauli matrices. Therefore, the dimension must be at least 4. That a 4x4 representation ex- ists can be shown by explicit construction. Define the following 4x4 Hermitian matrices: “(2 &( 3) © where 1 stands for the 2x2 unit matrix, and o* are the 2x2 Pauli matrices: ol 04 1 0 = - = a i 4) o fr 3) o (0 4) (68) ‘We shall not use different notations for 2x2 and 4x4 matrices, since the context usu- ally makes the meaning clear. It follows from the definitions that (ap = p= {at B}=0 (6.9) A standard representation for the Dirac matrices is y= =(6 if] (6.10) 0 -l 0 ot k= sgt = yt= Pat (. a ) (6.11) ‘The matrix y’ is Hermitian with (y°)? = 1, and y*is anti-Hermitian, with (y= -1: Y= OPFHI 96 Dirac Equation ayy OAP=1 6.12) From these we can show Aone = 6.13) The representation given here is not unique. A unitary transformation SS! gives an equally acceptable set of matrices, since such a transformation obviously pre- serves (6.4). The and their products, together with the unit matrix, generate a set of 16 in- dependent 44 matrices, in terms of which any 4x4 matrix can be expanded. We in- troduce special symbols for some of their products: ys= ivy yy on = Sty) (6.14) The “fifth” Dirac matrix ys is Hermitian, with square 1, and anticommutes with all four ys (95)" = Ys (ys =1 {YY} =0 (6.15) In our standard representation it has the form oe G 0) (6.16) The generalized Pauli matrices o#” = -c#” have six independent members: of = iat ot = ett (6.17) where o* denotes the matrix of 2x2 blocks made up of Pauli matrices along the di- agonal. It is straightforward to show that of = ys Ports = oH (6.18) A complete set of 16 independent 4x4 matrices I’, is given in Table 6.1. By de- finition, we take I'y = 1. All the I, are traceless except for To: 6.1 Dirac Algebra 97 TABLE 6.1 Matrices of Dirac Algebra r, Number 1 1 ~ 4 ys 4 on 6 Ys 1 Total 16 TY, =0 (n#0) (6.19) The set is closed under multiplication and commutation, and is called Dirac alge- bra. The commutators are given in Table 6.2. An arbitrary 4x4 matrix M can be expanded in the form (6.20) where Tr(MT,) oT? (6.21) TABLE 6.2 Commutators of Dirac Algebra [1 A] = 274 -2e" 98 Dirac Equation 6.2 WAVE FUNCTIONS AND CURRENT DENSITY Rewriting the Dirac equation in a 3-vector notation, we have Cia V+ Am) ye) =: (622) This looks like a single-particle wave equation with Hamiltonian H=a-p+ Bm (6.23) where p is the momentum operator. The wave function (x) is a four-component column vector called a Dirac spinor: dy =| % wl a (6.24) Ws where yf, are complex numbers. The complex conjugate is the column vector w-( i (625) ui and there are other types of conjugates: Hermitian adjoint: wt=(WF UF wf Wh) Pauli adjoint: P= wrP=(Wh vt -W -Wa) (6.26) The Hermitian conjugate of (6.1) reads =i") yt — mypt(x) = 0 (6.27) Now write yt = 7, and use 7°“t7? = y* to obtain the equation for the Pauli ad- joint: (4,Diy" + mib=0 (6.28) Another way of writing this is Wiy"d, + m)=0 (6.29) 63 Plane Waves 99 where the overhead arrow on d,, indicates that it acts to the left. The conserved density current is given by aye (6.30) It is easy to see, with the help of (6.1) and (6.29), that SF" = (Wd + Pye aw =-mibb+ mpb=0 (6.31) Note that ° is positive-definite: P= b= Why + Po¥ dr + Ws ts + Yate (6.32) As opposed to Us = Wy + Yaa — Ws ls — Ua (6.33) The current j# can therefore serve as a particle current density. As we shall see, however, the Dirac equation fails to qualify as a single-particle equation for a differ- ent reason; namely, the energy spectrum is not bounded from below. As we shall discuss in Section 6.9, the remedy is a redefinition of the vacuum state known as “hole theory,” which makes the system a many-particle system. With this modifica- tion, j° will become an operator, whose expectation values are no longer positive- definite, but can be interpreted as charge density. 6.3 PLANE WAVES Plane-wave solutions to the Dirac equation can be constructed by putting Wx) = e-P*u(p) (6.34) where p“ = (p°, p), and u(p) is a column vector called a Dirac spinor: 4 (Pp) w= | 20) (635) u4(P), Since U(x) satisfies the Klein-Gordon equation, we have pot —p?— mi? =0 100 Dirac Equation For given momentum, there are two roots for the energy p°, with opposite signs: Do=tE (6.36) where £ is defined as the positive quantity E=+Vp tm (637) The Dirac equation now takes the form (p-m)u(p) = 0 (6.38) where p is a 4x4 matrix defined by B= ¥DL= Pp? 7k (6.39) It has the property Bat PA =2pq (6.40) which follows from (6.4). To find explicit solutions, we note that (p—m)\(h + m)= p?- m2 =0 (641) Thus, each column of the matrix (~ + m) satisfies the Dirac equation. The explicit form of the matrix is m+p? 0 pp 0 +p? " 3 Bem=! os m mye a (6.42) Ps P 0 m-p® where pi= pip (6.43) The number of independent columns can be found by letting p*—> 0, since the ma- trix is a continuous function of p*. In that limit p° = +m, and the matrix becomes proportional to 1000 0100 0000 ° 0000 9 0000) frre 001 0] frp 0, while columns 3 and 4 are independent for p® < 0. The independent solutions are then columns | and 2 of (6.42) for p® = E, and columns 3 and 4 for p° = —E. We designate them as u(p, s). The explicit solutions for p° = E are 1 0 0 1 -c| - P- up, 1)= Cl oe up, 2)= C] (6.44) Ps oa mt+E mt+E The solutions for p? = -E are * P m+E mt+E =P Pe = 4)= 4 wp.3)=C] P= | ip, )=c| A (643) 1 0 0 1 where E=+Vp+m m+E c- (6.46) For a given p, these solutions form an orthogonal set: E WN(p, s)u(p, 8’) = 783s (6.47) For a given energy, the wave functions above resemble those of a nonrelativistic par- ticle of spin +, and it is natural to regard s as a spin label. We shall see that this is a correct interpretation. Taking the Hermitian conjugate of (6.38), we have ul(p, spt —m)=0 (6.48) Multiplying the equation from the right by 7, and using the identity Poy =p (6.49) 102 Dirac Equation we find the Pauli-adjoint equation u(p, s\(p~m) =0 (6.50) Multiplying (6.48) from the right by u, and (6.38) from the left by u‘, and writing out p more explicitly, we have ul(yp? — py —m)u=0 ul(yp? + y'p,—m)u = 0 (651) Adding the two equations leads to the relation 0 ty= in 6.52 wu = iw (6.52) We can restate the orthonormality of the solutions in the form U(p, s)u(P, s')= + dy (6.53) where the plus sign applies for the positive-energy solutions, corresponding to s = 1,2, and the minus sign is used for the negative-energy solution with s = 3,4. 6.4 LORENTZ TRANSFORMATIONS Under a Lorentz transformation x’ = Ax, the Dirac equation in the new frame reads (ey, — mV’) = 0 (6.54) Note that remains unchanged, because it is just a numerical matrix, We relate the new wave function to the old through a linear unitary transformation: W(e')= Six) SIS=1 (6.55) where Sis a nonsingular 44 unitary matrix. To demonstrate Lorentz covariance, we shall show that there exists a nonsingular transformation on -* that will restore the Dirac equation to the old form. Putting 4, = A¥.d, and multiplying the equation by ‘S"' from the left, we obtain S-(ey#AZ A, — m)SWUx) = 0 which reduces to the original equation if (S"'yS)A¥ = y” or S1yS= Mey” (6.56) 64 Lorentz Transformations 103 The existence of S will be demonstrated by explicit construction. It suffices to consider an infinitesimal Lorentz transformation XH = xt + ot?” (6.57) where « contains six infinitesimal parameters, the three rotations @ and the three boosts of the coordinate frame: co = af = yk aii = y + ¥. The x mesons were ob- served after being stopped in matter, and found to be longitudinally polarized. The preceding result shows that they had the same longitudinal polarization at the mo- ment of decay. 6.7 NONRELATIVISTIC LIMIT We shall study the nonrelativistic limit of (6.82), by first putting it in second-order form, Multiplying it from the left by #°D, + m, we obtain (YD D, + mPyb=0 (6.93) Writing the first term as half the symmetric part plus the antisymmetric part with respect to the labels z and v, we can show PYD,D, = gD Dy + 2 HY Dy Dy] (6.94) A straightforward calculation gives [D,, D,) = ieF (6.95) ‘We thus arrive at the second-order equation (>, + Sot ae + m) u=0 (6.96) Consider a stationary solution of energy E, with di/ot = iE . We can rewrite the equation in the form [(p- eA)? - eo B+ iea: E-mJy= (E- eh)? (6.97) where we have used the relation to"F,,=—-o- B+ ia E (6.98) The equation displays a magnetic-moment term @ - B, with electric-moment term a: E generated by the moving magnetic moment. In the nonrelativistic limit the components y, and y, are small, and it is conve- nient to rewrite the above in two-component form by putting Ue (3) (6.99) where y and é are two-component column vectors. Substituting this into the Dirac ‘equation (6.82), we obtain the coupled equations _E-eb-myx-o- (p—eA)E=0 (E-eb + m)E-o- p—eA)y=0 (6.100) 110 Dirac Equation Solving for the “small” component & we have go: (p-eA)x E-eb+m say) which shows that it is of order |p|/E compared to the “large” component x. ‘The second-order equation can be rewritten in the block form (meet 1B) 2) -2-e0r( 2) (6.102) ieoE ieoE where a= p — eA. We write the equation for x, and eliminate ¢ with the help of (o- E\(o- 7) [= -co- Bene ie E-eb+m |r e-core (6.103) which is an exact equation. We go to the nonrelativistic limit by putting E=mte (6.104) and assume € 0: p*=(E, p) (6.132) In 3-vector form, the equations read. (@: p+ Bm)u(p, s) = Eu(p, s) (a: p + Bm)o(p, s) = -Ee(p, s) (6.133) Note that the energy of o(p, s) is still negative, for all we did was reverse the mo- mentum, and write its energy as —E (with E > 0). The orthonormality of the solutions is expressed by the relations 116 Dirac Equation U(P, S)u(P, 5’) = de Up, sp, 8") = 3 Up, s)u(p, s')=0 (6.134) which are equivalent to E ul(p, s)u(p, s') = 7 Bae E it = 5. MP, SAP, 8') = TBs ul(p, s)o-p, s’) = 0 (6.135) The completeness of the solution is stated as Sup (6.136) 2 DLP, s)us*(p, 8) + op, s)o#(p, 8)] = & where a and b are spinor indices. This is equivalent to the matrix equation 2 > lp, s)u(p, 5) - o(p, Jap, 5)] = 1 (6.137) a The terms above are respectively projection operators onto positive-energy and neg- ative-energy states: A.(p)= up. site, s)= =P 2 - A.@)=-Sa. yap, 9)= =F (6.138) sl m which have the properties [APP= ALP) Alp) +A.(@)=1 (6.139) Note that the 4-vector p* in /p is defined to have positive time component p = The Dirac equation cannot be a one-particle equation, but it furnishes a finite- dimensional representation of the full Lorentz group. As such, it provides a com- plete set of one-particle wave functions, in terms of which we can analyze the oper- ator of a spin-+ field, as we shall do in the next chapter. 6.10 Charge Conjugation 117 6.10 CHARGE CONJUGATION An antiparticle should have opposite charge to a particle, since it represents the ab- sence of a particle in the negative-energy sea. This is intuitively obvious; but let us make certain that the formalism gives this result. In the presence of an external elec- tromagnetic field A¥(x), the Dirac equation is as given by (6.82). We denote the wave function as (x) for positive-energy plane wave states, and y(x) for negative- energy plane-wave states: Ux) =e PXy(p, 5) YE(x) = EP *(p, 5) (6.140) where E = + Vp? + m2, Then (6.82) can be rewritten Liy"(6,. + ieA,,) — m] Wx) =0 liv"(6,.—ieA,) — m] YF) = 0 (6.141) which show that the charge indeed has opposite signs for particle and antiparticle. The two equations above can transformed into each other through “charge conjuga- tion,” or “particle-antiparticle conjugation.” To change the sign of the coupling term in the first equation, we take the complex conjugate: Liy*#(d,, — ied.) — m] YA(x) = 0 (6.142) We then make a unitary transformation to bring it to the form of the second equa- tion. Thus YO) = mY) (6.143) where 7) is a 4x4 matrix with the properties wel (6.144) TW nyn= The solution is, in our standard representation of the Dirac matrices, n= (6.145) (where 7? is the second Dirac matrix). In terms of the spinors, charge conjugation corresponds to the transformation v(p, s) = mu*(p, s) (6.146) 118 Dirac Equation Since {7, P} = 0. This shows that particles and antiparticles have opposite parity. Like the time reversal discussed earlier, the charge conjugation here is an oper- ation on Dirac wave functions, and not on physical states, which are defined in quantum field theory. The operation is relevant because we expand the quantum field operators in terms of Dirac wave functions. 6.11 MASSLESS PARTICLES For a massless Dirac particle, with m = 0, the equation for the Dirac spinor reduces to pu(p) = 0, or a@pu(p) = pou(p) (6.147) where po=tE E=|p| (6.148) Since [a, ys] = 0, we can diagonalize +s, whose eigenvalue +1 is called “chirality.” The solution with chirality + | is called “right-handed,” denoted up; one with chiral- ity -1 is called “left-handed,” denoted u,: ‘Ystie(P) = Ua(P) ‘Y¥sui(P) = —u(P) (6.149) Using the relation yWa=o (6.150) we have o: pu(p)= o ‘ysu(p) (6.151) which states that the helicity op is the chirality time the sign of the energy. Thus, for a right-handed particle, the helicity is correlated with the sign of the energy, and for a left-handed particle it is anti-correlated. For a given momentum p, the four in- dependent solutions are uc(p, s), where C = R, L denotes chirality and s = + 1 de- notes helicity. Explicit solutions can be obtained from (6.45) by putting m = 0; but obviously we cannot normalize them according to (13.105). Instead, we put udp, S)uicAp, 8") = Beco 2E (6.152) Itis easy to show Yic(P, s)uic(P, 5) = 0, it follows that for, since {5, 7} =0, it fol- Problems 119. lows that # and u have opposite chirality. The one-particle states |p) have the proper- ties (Pip) =280n)9@-P) Joie w= (6.133) We must, of course, define the vacuum using hole theory. In analogy with the mas- sive case, we define antiparticle spinors: Vc(P, 5) = uc(-P, =5) (6.154) For a given p, the independent solutions can be taken to be ug(P, 1), 2R(P, -1), u,(p,—1), oL(p, 1). Thus, a right-handed particle is a right-handed screw, and a left- handed particle is a left-handed screw. The correlation between handedness and he- licity is reversed for antiparticles. PROBLEMS 6.1 Lorentz Boost (a) The transformation matrix for an infinitesimal Lorentz transformation is of the form S= 1 + iR, where R satisfies 6 60). Review the argument leading to the form R= Cw,,0#, and show that (b) Using the identity +P {oy = o#, show that Srp =S With this, verify the transformation law for i given in (6.72). (©) Obtain the free-particle solutions u(p, s) to the Dirac equation by applying a Lorentz boost to the solutions in the rest frame: up, s) = [eos £ + p-esinh =p where $= tanhr'y, and b= euoS SoHs 120 62 63 64 65 Dirac Equation Intrinsic Parity Show (6.73) that u(p, s) (s=1,2) uno Mio) Gade and therefore Yup, s)=u(p,s) — (s=1,2) YA-p, 8) =-2(p,8) (8= 1,2) These relations indicate that particles and antiparticles have opposite intrinsic parity. PauliTerm The Dirac equation describes a particle with g= 2. Physical particles have g factors different from 2 because of interactions, which give rise to an “anomalous” magnetic moment. The electron acquire the anomalous moment through interactions with the quantized electromagnetic field. That for the proton and neutron are dominated by the strong interactions. Suppose that the g factor is 2 + «. Show that this can be ac- commodated by taking the Dirac equation in external electromagnetic field to be [may + ieA,)- Set Fae n| Ua)=0 The extra term is called the “Pauli term.” For the proton and the neutron, the experimen- tal values are xp = 1.79, ky =—1.91, respectively. Chiral Current The chiral current density is defined by F8) = Vey ysl) Using the Dirac equation, show that 4,58 = 2mibys The chiral current becomes conserved in the massless limit m — 0. In quantum field theory with electromagnetic interactions turned on, this property is destroyed by the ax- ial anomaly [3]. Zitterbewegung The zitterbewegung [4] is a kinematic property of the spin-+ repre~ sentation of the Lorentz group, the “clockwork” of the Dirac equation. To exhibit this motion, construct a wave packet for a Dirac particle: &p Qay te.) = [SP errtns(pye® + wipe where £ = +V/p? + m?, and w,(p) are linear combinations of Dirac spinors with positive (negative) energies #£. Calculate the expectation value of the velocity (v) = f drueah Show that &p ny (Vv) = Vo + 2Re f (wt anw_jere* References 121 where vo = (27)f p(w! aw. + w! anw_) . Integrate this to obtain the average position (1) = to + vot + inf The last term is the zitterbewegung, which arises from an interference between positive- and negative-energy states. On dimensions grounds, we can conclude that the amplitude of this oscillatory motion is of the order of the Compton wavelength I/m, and therefore unobservable. In the hole theory, when all negative energy state are filled, the zitterbe- wegung becomes part of the vacuum fluctuations of the Dirac field, for it can happen only when holes are momentarily created as a result of fluctuations. 6.6 Gordon Decomposition (a) From the definition of the Dirac matrices, show that, yey = gh iow (b) Multiply the equation (iy“a,, — m)ys= 0 from the left by Jy", and use the identity to © rewrite the result in the form = Lo = Trty= 5 (Mond) — (OD + aor) with spatial components = Fp CVD HT 1+ T «Gow +5 Can} This is the Gordon decomposition, which splits the current density into a “convec- tion” part, plus contributions from the spin. It suggests that the spin is the orbital angular momentum of the zitterbewegung. Let u,= u(p,, s,), = 1,2), be two Dirac spinors. Let Prem pit + pt k= pt-pt Show that Beye, = HP" + oh uy 6.7 Massless Particles Consider massless Dirac particles. (a) Show iic(p, s)uc(P, s) = 0. (b) Show that the projection operators A,(p) for positive and negative energies have the properties - 122 Dirac Equation Aup)=Etep AL(p) = EA.(p) Ad(p) + A(p)=E REFERENCES 1. K.M. Case, Phys. Rev. 106, 177 (1957). 2. RL. Garwin, L. M. Lederman, and M. Weinrich, Phys. Rev. 105, 1415 (1957). 3. K. Huang, Quark, Leptons, and Gauge Fields, 2nd ed. World Scientific, Singapore, 1992, Chapter 11. 4, K. Huang, Am. J. Phys. 20,479 (1952), The Dirac Field 7.1. QUANTIZATION OF THE DIRAC FIELD In hole theory, the Dirac equation describes a many-fermion system, and thus the Dirac “wave function” x) should be regarded as a classical field to be quantized according to Fermi statistics. To carry out the quantization in the canonical formal- ism, we take as classical Lagrangian density L(x) = Wayliy"d,,— m) Wx) (7.1) where (x) is a four-component spinor and the independent field variables are the components (x). We note that £(x) is Lorentz-invariant, and globally gauge-in- variant. This is a first-order Lagrangian density, involving first instead of second derivatives with respect to time. We have illustrated the-self-consistency of the canonical formalism in this case in Problem 4.5. Therefore, following strict canoni- cal procedures, we calculate a“ ae mt = Fado =i Ya ou, (7.2) ait le The equation of motion is id,dne + mp=0 (73) which correctly gives the Dirac equation in Hermitian-conjugate form. The La- grangian density vanishes for fields satisfying the equation of motion: £(x)=0 (for fields satisfying equation of motion) (74) The canonical conjugate to y, is i, since 123 124 The Dirac Field T= 7a = iE (7.5) The Lagrangian does not depend on y,*, and therefore y/# has no conjugate. One must resist the temptation to make the Lagrangian more “symmetric” by replacing ipy“d,,p with (i/2)[y“(9,.W) + (4, YW). This would be akin to “adding feet when drawing a snake,” as a Chinese saying goes. The canonical quantization rules lead to the following anticomutation relations: (Uelet, ), WLC", D} = 55% = 1") {Walt , y(t", )} = 0 (76) where a and b denote spinor indices. The anticommutators serve as initial condi- tions for the Dirac equation. They also fix the normalization left arbitrary in the Dirac equation. The Lagrangian density is invariant under the global gauge transformation y—> ew, where w is a constant. The associated Noether current is FER) = Hx) 7) which is conserved: Gy iM(x) = 0 (7.8) The canonical energy-momentum tensor, which is associated with translational invariance, is given by (79) with conservation law ,.T#%(x) = 0 (7.10) The energy and momentum densities are respectively TQ = WiPw= YCia-V + Bm) To = Yrioy (7.11) When integrate over space, they give the Hamiltonian H and total momentum P: H= f. @r Wi(r, )(-ia-V + Bm)Kr, t) P ifar WEY ur, (7.12) The generalized angular momentum tensor is 7.1 Quantization of the Dirac Field 125 MEA) = YRDLix* — xP yd) + 5 oP) (7.13) and the angular momentum and the boost operators are respectively i = eit dir MH = elf aProtlecta!—40) - Fo] y , 1 Kiz [ar Mo = ifr Pai — IP + ye) uv (7.14) From the angular momentum, we can read off the spin operator: (7.15) in agreement with what we found in the last chapter. The one-particle solutions obtained in Chapter 6 constitute a basis in terms of which the field operators may be expanded. We normalize the wave functions in a large periodic box of volume @, and write 7 We D> planet, 8) + byl, 8] eV 08, Wr, )= —" fa,,teHP-Eut(p, s) + bp,e-rt(p, s)] (7.16) 2 ne ul(p, s) + bye" rK(p, s where E,=+Vpetm (7.17) D The factor Vmi/E, appears because the Dirac spinors are normalized according to (6.133) and (6.134): Hp, utp, 3") = Fw", SMB. 8) = Bu z > , m ” 3(P, 8)Ap, 5!) =F oP, sp, 5") = 'p 2(p, s)u(p, s’)= 0 (7.18) With this factor taken out, we have simple anticommutation rules {aps ape} = {Booby} = Sav Bop {apedp's'} = {bps bps} = (apspsr} = 0 (7.19) 126 The Dirac Field which lead to the interpretation that ap, annihilates a particle whose wave function is u(p, s) and bp, annihilates an antiparticle whose wave function is o'(p, s). Hole theory is implemented through the statement y,!0) = by{0)=0 for all p,s (7.20) This implies that there are neither particles nor antiparticles in the vacuum state 0). In terms of the annihilation and creation operators we have H= SE jlaitys— dyed fe) =D, Ea hetys + bfsbys— 1) e e P= DY Plahdp, — bp.b}.) =D Plapuays + bf.bpe) (721) e fe If we had not used hole theory, b,, would be creation instead of annihilation opera- tor, and 5,,bp,t would have eigenvalues 0,1. Consequently, the Hamiltonian would not be bounded from below. The sign reversal that makes the Hamiltonian positive- definite, of course, depends on the fact that we quantized the system according to Fermi statistics. The charge operator is given by o= far typu= arr vy = > @htys — bpsbfe + 1) (7.22) which shows that particles and antiparticles have opposite charge. The minus sign above arises through rewriting b,.bj, as —b{,bp, + 1. This is dictated by the fact that b{,bp, has positive eigenvalues in hole theory. The normalization of Q is arbitrary, for the magnitude of the charge is determined only when there is interaction with the electromagnetic field. 7.2 FEYNMAN PROPAGATOR The Feynman propagator for the Dirac field is a 4x4 matrix iSp(x) = (O|TYAx)(O)|0) (7.23) where the time-ordering operator T is defined to include a sign change when two fermion operators A and B are interchanged: _fAt)B) if, >t ae) = [nny thee (724) 7.2. Feynman Propagator 127 ‘The propagator can be calculated straightforwardly, using the expansions (7.16). Let x= (t,), For t> 0, the only contribution comes from terms in the expansion of the form aat: [Se)ap = (OlWel, 1)Us(0)10) = ae ett) u(P, s)us(P, S) (7.25) For ¢ <0 we need only keep terms of the form bbt: Sj 1 P . 7 SHC as = — OO, N10)=—H S FreHPPY nD, iP.) (7.26) v Fp a The sum over spin states results in the projection operators given in (6.137). Sup- pressing the spinor indices and going to the limit © — ©, we have L_f(mtpeen — (¢>0) ison= [55 Gn) 26, al (m— per) (<0) Me We can make the replacement m+ pam PE,— yipt> m+ ip Yt (7.28) because this operator acts on the exponential factor. For ¢< 0, make the change of variables p—> —p. Then we have 1 iyo — ht) ertEpllgloe iS-(x) = s&s zs (" + iy? vere pli (7.29) Now use the representation e POF iE, othe el Seren rere CED where 7 — 0°. Then, the operator id/4t in the previous formula can be replaced by Po. The final result is ptm Pips ptm _ 7: 50) = | ioreo eer (731) We leave it as an exercise to show that Sex) = (ey#9,, + m)AelX) (7.32) where A,(x) is the Feynman propagator for a free scalar field of mass m. 128 ‘The Dirac Field The Fourier transform of the propagator is a pP-m+in p-mt+in 5,(p) = (7.33) where the right side is the inverse of a 44 matrix. The 4-vector p is arbitrary, with either p® > 0 or p® > 0. For p> 0, we have, according to (6.137), 2 (ptm mo st u(p, sju(p,s) (p> 0) (7.34) For p? > 0, let us define g# = —p#, Then according to (6.137), we obtain 2 Cem) I) Sq, 554.8) (p90) (735) This shows that the residue at the mass-shell pole at p? = m? contains the wave func- tions of an electron of momentum p, or those of a positron of momentum q = -p. 7.3 NORMAL ORDERING Both H and Q contain divergent contributions from the zero-point energy and charge of the vacuum state. These terms have no physical relevance since energy and charge are measured relative to those of the vacuum state. They can be eliminat- ed by redefining the reference points, and this can be achieved by arranging the or- der of operators appropriately. We first introduce the notion of normal ordering. Suppose that O is a product of creation and annihilation operators. The corresponding normal product : O : is defined as that obtained from O by rearranging the order of the factors, if necessary, such that all creation operators stand to the left of all annihilation operators. In the rearrangement process, an interchange of two fermion operators gives rise to a fac- tor —1. As an example: (7.36) Normal ordering can be naturally extended to a sum of products: 20, + O,:=:0,:+:0,: (731) We now redefine the Hamiltonian and the current as Pri, -ia-V + Bm), 1) : 7.4 Electromagnetic Interactions 129 FE) = yx) = (7.38) It is clear that these operators give zero when operating on the vacuum state, be- cause annihilation operators stand to the far right. This is just a formal way of stat- ing that the zero-point energy and currents are to be omitted. As the notation is somewhat cumbersome, we shall not explicit indicate normal ordering unless neces- sary. 7.4. ELECTROMAGNETIC INTERACTIONS We consider systems of interacting fields with a Lagrangian density consisting of the sum of the free Lagrangian densities of the participating fields, plus an interac- tion Lagrangian density that couple the fields together. This is not the most general case conceivable, but it is what we can handle mathematically. We illustrate the types of interactions commonly encountered, Consider a Dirac field, a complex scalar field, and the electromagnetic field, which have free Lagrangian densities given by Loirac = Wiy"d,, — mip Lecatar = HO*3,6 — PG Lem =—3 FHF yy (7.39) According to the gauge principle, the matter fields can be coupled to the electro- magnetic (em) field by replacing 4, by the covariant derivative DH = 4, + ieAMx) (7.40) where e is the electric charge. Assuming that both the Dirac field and the scalar field have the same charge e, the electrodynamic Lagrangian density is LA POE, + HiyrD,— mW [D*SHD,b-85*S = Lam + Loine + Lae + Li an where Ling = (Gh + PVA, ja iebey Je= ie b*($) — (Hd*)d] + Pb*h AM (7.42) The matter fields are coupled through conserved currents, which are the Noether 130 The Dirac Field currents associated with global gauge invariance. For the scalar field, the current has an e? term proportional to 4*. This becomes a mass term for the photon when *¢ develops a vacuum expectation value, in spontaneous symmetry breaking. (See Problem 15.5.) The electromagnetic field couples to all charged fields through the gauge prin- ciple, and is universal in this sense. The vacuum fluctuations of the electromagnetic field include the momentary creation of virtual particle-antiparticle pairs and their subsequent annihilation. The temporary charge separation makes the vacuum into a dielectriclike medium, and all charged fields of the world participate in this “vacu- um polarization,” as their contributions being determined solely by charge and mass. 7.5 ISOSPIN The Dirac field can be used in a phenomenological description of protons and neu- trons, which are really made of quarks. The effective theory is useful in describing the “charge-independent” pion-nucleon interactions at low energies. It is based on the fact that proton and neutron are almost identical, and so are the three mesons, and the strong nuclear forces respect the identities. By ignoring the electromagnetic and weak interactions, we can regard the proton and neutron as different states of a particle called the nucleon, and the a mesons as different states of the pion. The nucleon field is represented by a two-component Dirac field _ tice) UG) = ( so} (7.43) where i= I corresponds to proton, and i= 2 to neutron. Each y is a four-component Dirac spinor field. Writing out all the indices, we have eight complex fields yi2(), with a=1,...,4 andi= 1, 2. By analogy with spin angular momentum, we define the isospin 7/2 as generators of rotations in the two-dimensional internal space spanned by y, and ys: w(t) a) m6 2 om The proton and neutron states are eigenstates of 73/2 with respective eigenvalues ++ and—3: w=(o) b=(;) 45) which can be created from the vacuum by applying (x). We use a shorthand nota- tion in which the spinor and internal indices are suppressed. For example, 75 isospin 131 DP TY = Vil P aol Miley (7.46) where a, b are summed from | to 4 and i, j are summed from | to 2. More generally we define isospin as an internal symmetry group whose gener- ators >/ obey angular momentum commutation relations (Lie algebra): Un T= icy Thus one can simultaneously diagonalize 7? = (+ 1) and J, and denote isospin eigenstates by |/, /;). The nucleon belongs to the fundamental representation with I= 4, in which [= 7/2. The overhead arrow denotes a vector in isospin space, which has three components because that is the number of generators of the group. The pion field has /= 1, and is described by a three-component real field iQ) $x)=| bx) (7.47) 30) This the “adjoint representation” of the group, which has the same dimension as the number of generators, and in which the generators are represented by matrices 7" taken directly from Lie algebra: (Tim = ie" (7.48) Experimental evidence dictates that (x) be pseudoscalar, that is, that it change sign under spatial reflection. We note that J; is not diagonal. The physical pion fields, which are eigenstates of P, are related to ¢, through 2) = elds) +idsC0] 10) = Tpldi0-is00] P(x) = s(x) (7.49) These operators create states with J; = +1, —1, 0 respectively, when they operate on the vacuum state, “Charge independence” in the pion-nucleon system means that the interaction conserves isospin. A Lorentz-invariant effective Lagrangian density, known as “pseudoscalar coupling,” is given by L(x) = Hiya, — M+ [Hb H— mS GB] + Cy W'S (7.50) The vector notation makes manifest the rotational invariance in isospin space. A competing model is the “pseudovector coupling” model, with 132 The Dirac Field L(x) = Wey"d,,— M+ 3[4,8- 4,-m°$-B) + 2's FW) - 3,6 (7.51) Some consequences of isospin invariance are explored in Problem 7.4. 7.6 PARITY We discuss the discrete symmetries, using as an example the electromagnetic cou- pling as contained in L(x) =-3 FWP + iy"(d,, + ieA,,) — m] (7.52) Under a Lorentz transformation x —> Ax, the field operators ¢,(x) undergoe a uni- tary transformation U given by (3.54): Udalx)U"' = Sapbs(A'®) (7.53) where S,,. This can be extended to spatial reflection x —> x, t—> 1, for which the uni- tary operator U is denoted by ®. For the Dirac field, we have S = according to (6.69) and (6.70), and thus Pr, HP! = Pwr, t) Pur, HP! =W4,jP (7.54) Since A* transforms like a vector, PAE, P| =A, t) PAK, P= Ar, t) (7.55) Thus we have PL(r, NP! = L(-4, t) (7.56) which show that the Lagrangian L = f d°x £(x) is invariant. From the expansion (7.16) at ¢= 0: We) => | Gp lane Fup, 5) + erp, 3)] 31) Ps Ey we have Pure'=S. Fe [Pay Pe Tu(p, s) + PLP? (p, s)] a VE, 7.7 Charge Conjugation 133 mn PO > | Fp net PD.s)+ heh PAp. 3] (58) Using the relations (Problem 6.2) Pulp, s) = u(p, s) Pup, s)=—(P, 5) (7.59) we obtain the statement that particles and antiparticles in Dirac theory have oppo- site intrinsic parity: Pay?! =a», PbyP =—b ps (7.60) The transformation P may be accompanied by a rotation in spin space with respect to the index s, as is clear from (7.58); but we leave it out for simplicity. 7.1 CHARGE CONJUGATION Charge conjugation, or particle-antiparticle conjugation, is defined as a unitary op- eration C on the Hilbert space that interchanges particle and antiparticle, and revers- es the sign of the electromagnetic field: Cap,C-1 = byg CbpsC-! = ap, CAR(R)C = ~AK(x) (7.61) The transformation of A%x) is not specified independently, because in Coulomb gauge it is not an independent field. It is clear that £(x) is invariant under this trans- formation, because the free-field Lagrangian densities are invariant, and the Dirac field is coupled to the electromagnetic field through the current density, which changes sign. To find how the Dirac field operator transforms, let us compare the following expansions: HO= Ep, ane. 9) + Be *0, 9] vo=> ne [ate Tu*(p, s) + bp,e?*a*(p, s)] (7.62) a V OE, 134 ‘The Dirac Field The expansion coefficients satisfy (6.146): 2(p, s)= mu*(p, s) (7.63) where 1 = iy? is areal 4x4 matrix. Therefore + [TR WO) => | op Une up, 3) + ape", 9) (7.64) w V OB, which shows Cur)C! = nb) (7.65) Note that the Dirac wave functions undergo complex conjugation, which is a nonlinear operation, because (Au)* = A*u*. The field operator, however, undergoes a linear transformation, because C(AW)C"! = A Ci C-!, The difference can be traced to the fact that in the Dirac equation we have to change the sign of the coupling to an external electromagnetic field, whereas in the field theory, the electromagnetic field is part of the system, and changes sign under charge conjugation. 7.8 TIME REVERSAL Time reversal is the operation of interchanging past and future, represented by a op- erator T on Hilbert space. Suppose that V, is a member of a complete set of state in Hilbert space, where a stands for quantum numbers, such as momentum p and spin projection s on a fixed axis. The time-reversed state TV, must be a member of the same set: TV,=Vz (7.66) where @ are the time-reversed quantum numbers, defined by correspondence with classical mechanics: (7.67) and the helicity is invariant. The basic property of T is (TV, TVs) = (Us, Va) (7.68) that is, it interchanges initial and final states. This can be rewritten (TV, TVy) = (Yas Vo)* (7.69) Replacing ¥, by AW;, where A is a complex number, we have 7.8 Time Reversal 135 (T¥,, TAV;)) = Va, Vo)* (7.70) Therefore TAY, = TV, (1.7) Thus, when acting on a number, 7 takes its complex conjugate. This makes T non- linear. More specifically, it is called an. “antilinear” operator. A general representa- tion of T is complex conjugation followed by a unitary transformation: T=Us T=Us (7.72) where it is assumed that U commutes with complex conjugation. For the Schrédinger equation wv rae (7.73) time reversal means H(IY)= ADD (7.74) The system is invariant under time reversal if the time-reversed equation is equiva- lent to the original. Taking the complex conjugate, we have HUY) r; (7.75) HUY, Thus, the system is invariant under time reversal if the Hamiltonian is real: H=H* (7.76) which implies that the Lagrangian must be real. Without going through all the details, we can conclude that TAM)T-! =-AK(r) TWe)T = yrystr) (1.77) The first equation follows from the requirement that 4* transform like the current density, which must change sign, because classically it is a velocity. The second fol- lows from the fact that +s is the transformation that preserves the Dirac equation 136 The Dirac Field under time reversal, as shown in (6.70). It is straightforward to verify that the La- grangian is invariant, if the charge ¢ is real. There is a theorem known as the PCT theorem, which states that a local field theory that is Lorentz invariant is automatically invariant under the product PCT, even though it may be separately invariant under PC, separately, We refer the read- er elsewhere [1] for proof. PROBLEMS 7.1 Energy-Momentum Tensor The canonical energy-momentum tensor 7. for the Dirac field is not symmetric in wa. According to Section 4.5, we can construct an equivalent symmetric tensor TH" = T#2 + £,X™<, Find X™#2, 7.2 Propagator Show that the propagator for the Dirac field is related to that of the scalar field through Selx) = (iy#d,, + mg) 7.3. Neutrinos Neutrinos are massless Dirac particles. Using the convention for wave function given in Section 6.11, expand the field operator in terms of annihilation and creation operators. 7.4 Isospin Transformations (a) Show that under an infinitesimal isospin transformation, the nucleon field, and the pion field transform according to $ -b+8~S where the components of ~w are arbitrary infinitesimal real parameters. (b) Let = yl 7, where ¥ is a 4x4 Dirac matrix. Show that / transforms like a vector in isospin space: Vs0+ ax 7.5 Pion-Nucleon Scattering As far as isospin properties are concerned, the pion and nu- cleon states can be labeled by / and J Im')= 11, 1) I) =|1,=1) |7?)=|1, 0) } +) In)=14,-1\) (a) A state containing a pion and a nucleon is a direct product in isospin space, as, for example, |a"n) =|1, 1) x [3 . However, this is not an eigenstate of total isospin and thus not an eigenstate of the Hamiltonian of the system. Show that eigenstates of the isospin are the following: Reference 137 13, 3)= a) $.4)= [Fear — [Son 5,-3)= fina) — [3 13,-3)= /3lam®)—/3lpar) (b) For interactions that conserve isospin, the pion-nucleon scattering amplitude de- pends only on total isospin and not on J; (for the same reason that atomic energy levels are independent of the magnetic quantum number, i.e., the Wigner-Eckart theorem.) Denote the transition amplitudes by ay and a, and show that they have the form (pa |Tipm) = asa (pr Tp) (na?|Tlpar) 5a + Fay2 3 asa 3 ay2 where T is the transition operator. The corresponding scattering cross sections are proportional to the squares of these amplitudes. (©) Neglect ay,2 compared to aya, and show that pion-nucleon scattering cross sections bear the ratio Opa): opr) : (nm > par) =9 1:2 This is verified experimentally at low energies. The reason that ay72 dominates is the resonance scattering 7+ N—> A—> m+ N, where A is a particle of spin 3, isopin 3, with mass 1232 MeV, known at one time as the “33 resonance.” REFERENCE 1. R. Streater and A. Wightman, PCT, Spin and Statistics, and All That, Benjamin, New York, 1964. Dynamics of Interacting Fields 8.1 TIME EVOLUTION The dynamics of a quantum mechanical system is governed by the Hamiltonian H, which generates time translations. One may view the time development from differ- ent perspectives. In the Schrddinger picture, one regards the operators O, as time- independent objects, and the state vector V, changes with time according to the Schrédinger equation VO ie = HY (8.1) Assuming that H is time-independent, we have the formal solution WO) = V0) (8.2) The matrix element of an operator O, evolves in time according to (PO IV(D) = (®,(O)le"#O,0"|¥ (0) (8.3) The subscript “s” identifies states and operators in the Schrédinger picture. In the Heisenberg picture, the state is assumed to be constant in time, but the operators evolve. The matrix elements of an operator must be independent of the picture, and this requirement relates the Heisenberg picture to the Schrodinger pic- ture: ¥,= V0) O,(t) = Oe" (8.4) 138