Claude Cohen-Tannoudji, Jacques Dupont-Roc, Gilbert Grynberg PHOTONS AND ATOMS INTRODUCTION TO QUANTUM ELECTRODYNAMICS Contents Preface Introduction I CLASSICAL ELECTRODYNAMICS: THE FUNDAMENTAL EQUATIONS AND THE DYNAMICAL VARIABLES Introduction A. The Fundamental Equations in Real Space 1, The Maxwell-Lorentz Equations 2. Some Important Constants of the Motion 3. Potentials—Gauge Invariance B. Electrodynamics in Reciprocal Space 1. The Fourier Spatial Transformation—Notation 2. The Field Equations in Reciprocal Space 3. Longitudinal and Transverse Vector Fields 4, Longitudinal Electric and Magnetic Fields 5. Contribution of the Longitudinal Electric Field to the Total Energy, to the Total Momentum, and to the Total Angular Momentum—a. The Total Energy, b. The Total Momentum, c. The Total Angular Momentum 6. Equations of Motion for the Transverse Fields C. Normal Variables 1. Introduction 2. Definition of the Normal Variables 3. Evolution of the Normal Variables 4, The Expressions for the Physical Observables of the Transverse Field as a Function of the Normal Variables—a. The Energy Htrans of the Transverse Field, b. The Momentum Prrans and the Angular Momentum Strans of the Transverse Field, c. Transverse Electric and Magnetic Fields in Real Space, d. The Transverse Vector Potential A, (r,t) 5. Similarities and Differences between the Normal Variables and the Wave Function of a Spin-1 Particle in Reciprocal Space 6. Periodic Boundary Conditions, Simplified Notation D. Conclusion: Discussion of Various Possible Quantization Schemes 1. Elementary Approach 2. Lagrangian and Hamiltonian Approach Complement Ai — The "Transverse" Delta Function 1. Definition in Reciprocal Space—a, Cartesian Coordinates. Transverse and Longitudinal Components, b. Projection on the Subspace of Transverse Fields 2. The Expression for the Transverse Delta Function in Real Space— a, Regularization of 5; (p). b. Calculation of g(p). c. Evaluation of the wou ul 12 2B 15 7 21 23 23 23 24 26 30 31 33 33 34 36 38Derivatives of g(p). d. Discussion of the Expression for 5;(p) 3. Application to the Evaluation of the Magnetic Field Created by a Magnetization Distribution. Contact Interaction Complement Bi —Angular Momentum of the Electromagnetic Field. Multipole Waves Introduction 1. Contribution of the Longitudinal Electric Field to the Total Angular Momentum 2. Angular Momentum of the Transverse Field—a, Jtrans in Reciprocal Space. b. Jtrans in Terms of Normal Variables, c. Analogy with the Mean Value of the Total Angular Momentum of a Spin-I Particle 3. Set of Vector Functions of k "Adapted" to the Angular Momentum— General Idea. b. Method for Constructing Vector Eigenfunctions for 3? and Jz... Longitudinal Eigenfunctions. d. Transverse Eigenfunctions 4. Application: Multipole Waves in Real Space—a. Evaluation of Some Fourier Transforms, b. Electric Multipole Waves, c. Magnetic Multipole Waves Complement Ct Exercises 1. Hand P as Constants of the Motion 2. Transformation from the Coulomb Gauge to the Lorentz Gauge 3. Cancellation of the Longitudinal Electric Field by the Instantaneous Transverse Field 4, Normal Variables and Retarded Potentials 5. Field Created by a Charged Particle at Its Own Position, Radiation Reaction 6. Field Produced by an Oscillating Electric Dipole 7. Cross-section for Scattering of Radiation by a Classical Blastically Bound Electron TI LAGRANGIAN AND HAMILTONIAN APPROACH TO. ELECTRODYNAMICS. THE STANDARD LAGRANGIAN AND THE, COULOMB GAUGE Introduction A. Review of the Lagrangian and Hamiltonian Formalism 1. Systems Having a Finite Number of Degrees of Freedom— a. Dynamical Variables, the Lagrangian, and the Action, b. Lagrange's Equations, c. Equivalent Lagrangians. d. Conjugate Momenta and the Hamiltonian. e. Change of Dynamical Variables, f. Use of Complex Generalized Coordinates, g. Coordinates, Momenta, and Hamiltonian in Quantum Mechanics. 2. A System with a Continuous Ensemble of Degrees of Freedom— a. Dynamical Variables, b. The Lagrangian. c. Lagrange’s Equations d. Conjugate Momenta and the Hamiltonian. e. Quantization. f. Lagrangian Formalism with Complex Fields, g. Hamiltonian Formalism and 42 55 61 63 64 66 68 1 4 0 81 81 90Quantization with Complex Fields B. The Standard Lagrangian of Classical Electrodynamics 1. The Expression for the Standard Lagrangian—a. The Standard Lagrangian in Real Space, b. The Standard Lagrangian in Reciprocal Space 2. The Derivation of the Classical Electrodynamic Equations from the Standard Lagangian—a. Lagrange's Equation for Particles, b. The Lagrange Equation Relative to the Scalar Potential, c. The Lagrange Equation Relative to the Vector Potential 3. General Properties of the Standard Lagrangian—a. Global Symmetries. b. Gauge Invariance. c. Redundancy of the Dynamical Variables C. Electrodynamics in the Coulomb Gauge 1, Elimination of the Redundant Dynamical Variables from the Standard Lagrangian—a. Elimination of the Scalar Potential, b. The Choice of the Longitudinal Component of the Vector Potential 2. The Lagrangian in the Coulomb Gauge 3. Hamiltonian Formalism—a. Conjugate Particle Momenta, b. Conjugate Momenta for the Field Variables, c. The Hamiltonian in the Coulomb Gauge, d. The Physical Variables 4, Canonical Quantization in the Coulomb Gauge—a. Fundamental Commutation Relations, b. The Importance of Transuersability in the Case of the Electromagnetic Field, c. Creation and Annihilation Operators 5. Conclusion: Some Important Characteristics of Electrodynamics in the Coulomb Gauge—a. The Dynamical Variables Are Independent. b. The Electric Field Is Split into a Coulomb Field and a Transverse Field, c. The Formalism Is Not Manifestly Covariant. d. The Interaction of the Particles with Relativistic Modes Is Not Correctly Described ‘omplement Au — Functional Derivative. Introduction and a Few Applications 1, From a Discrete to a Continuous System. The Limit of Partial Derivatives 2. Functional Derivative 3. Functional Derivative of the Action and the Lagrange Equations 4, Functional Derivative of the Lagrangian for a Continuous System 5. Functional Derivative of the Hamiltonian for a Continuous System Complement Bu —Symmetries of the Lagrangian in the Coulomb Gauge and the Constants of the Motion 1, The Variation of the Action between Two Infinitesimally Close Real Motions 2. Constants of the Motion in a Simple Case 3. Conservation of Energy for the System Charges + Field 4. Conservation of the Total Momentum 5. Conservation of the Total Angular Momentum Complement Cu Electrodynamics in the Presence of an External Field 1. Separation of the External Field 2. The Lagrangian in the Presence of an External Field—a. Introduction of a Lagrangian. b. The Lagrangian in the Coulomb Gauge 100 100 103 105 m1 i 113 115 118 121 126 128 128 130 132 134 136 137 138 139 141 1423. The Hamiltonian in the Presence of an External Field—a. Conjugate Momenta, b. The Hamiltonian. c. Quantization Complement Du —Exercises 1. An Example of a Hamiltonian Different from the Energy 2. From a Discrete to a Continuous System: Introduction of the Lagrangian and Hamiltonian Densities 3. Lagrange’s Equations for the Components of the Electromagnetic Field in Real Space 4, Lagrange’s Equations for the Standard Lagrangian in the Coulomb Gauge 5. Momentum and Angular Momentum of an Arbitrary Field 6. A Lagrangian Using Complex Variables and Linear in Velocity 7. Lagrangian and Hamiltonian Descriptions of the Schrodinger Matter Field 8. Quantization of the Schrodinger Field 9, Schrodinger Equation of a Particle in an Electromagnetic Field: Arbitrariness of Phase and Gauge Invariance Il QUANTUM ELECTRODYNAMICS IN THE COULOMB GAUGE. Introduction A, The General Framework 1, Fundamental Dynamical Variables. Commutation Relations 2. The Operators Associated with the Various Physical Variables of the System 3. State Space B, Time Evolution 1. The Schrodinger Picture 2. The Heisenberg Picture. The Quantized Maxwell-Lorentz Equations—a. The Heisenberg Equations for Particles, b. The Heisenberg Equations for Fields, c. The Advantages of the Heisenberg Point of View C. Observables and States of the Quantized Free Field 1. Review of Various Observables of the Free Field—a. Toral Energy and Total Momentum of the Field, b. The Fields at a Given Point r of Space, c. Observables Corresponding to Photoelectric Measurements 2. Elementary Excitations of the Quantized Free Field. Photons— a. Eigenstates of the Total Energy and the Total Momentum, b. The Interpretation in Terms of Photons, c. Single-Photon States. Propagation 3. Some Properties of the Vacuum—a. Qualitative Discussion, b. Mean Values and Variances of the Vacuum Field, c. Vacuum Fluctuations 4. Quasi-classical States— a. Introducing the Quasi-classical States. b. Characterization of the Quasi-classical States, c. Some Properties of the Quasi-classical States, d. The Translation Operator for a and a* D. The Hamiltonian for the Interaction between Particles and Fields 1. Particle Hamiltonian, Radiation Field Hamiltonian, Interaction Hamiltonian 2. Orders of Magnitude of the Various Interactions Terms for Systems of Bound Particles 143 146 147 150 151 152 154 157 161 167 169 171 im im 175 176 176 176 183 183 186 197 197 1983. Selection Rules 4. Introduction of a Cutoff Complement Au! —The Analysis of Interference Phenomena in the Quantum Theory of Radiation Introduction 1. A Simple Model 2. Interference Phenomena Observable with Single Photodetection Signals — a. The General Case. b. Quasi-classical States, c. Factored States. d. Single-Photon States 3. Interference Phenomena Observable with Double Photodetection Signals—a. Quasi-classical States, b. Single-Photon States, c. Two-Photon States 4, Physical Interpretation in Terms of Interference between Transition Amplitudes 5. Conclusion: The Wave-Particle Duality in the Quantum Theory of Radiation Complement But —Quantum Field Radiated by Classical Sources 1. Assumptions about the Sources 2. Evolution of the Fields in the Heisenberg Picture 3. The Schrodinger Point of View. The Quantum State of the Field at Time ¢ Complement Cur —Commutation Relations for Free Fields at Different Times. Susceptibilities and Correlation Functions of the Fields in the Vacuum Introduction 1, Preliminary Calculations 2. Field Commutators—a. Reduction of the Expressions in Terms of D. b. Explicit Expressions for the Commutators, c. Properties of the Commutators 3. Symmetric Correlation Functions of the Fields in the Vacuum Complement Dur—Exercises 1. Commutators of A, E, , and B in the Coulomb Gauge 2. Hamiltonian of a System of Two Particles with Opposite Charges Coupled to the Electromagnetic Field 3. Commutation Relations for the Total Momentum P with Hp, Hr and Hr 4, Bose-Einstein Distribution 5. Quasi-Probabihty Densities and Characteristic Functions 6. Quadrature Components of a Single-Mode Field. Graphical Representation of the State of the Field 7. Squeezed States of the Radiation Field 8. Generation of Squeezed States by Two-Photon Interactions 9. Quasi-Probability Density of a Squeezed State TV OTHER EQUIVALENT FORMULATIONS OF ELECTRODYNAMICS, Introduction A. How to Get Other Equivalent Formulations of Electrodynamics 199 200 204 205 206 209 221 222 223 227 230 232 233 234 236 2411. Change of Gauge and of Lagrangian 2. Changes of Lagrangian and the Associated Unitary Transformation—a. Changing the Lagrangian. b. The Two Quantum Descriptions. c. The Correspondence between the Two Quantum Descriptions. d. Application to the Electromagnetic Field 3. The General Unitary Transformation. The Equivalence between the Different Formulations of Quantum Electrodynamics B. Simple Examples Dealing with Charges Coupled to an External Field 1. The Lagrangian and Hamiltonian of the System 2. Simple Gauge Change; Gauge Invariance—a, The New Description. b. The Unitary Transformation Relating the Two Descriptions—Gauge Invariance 3. The Goppert-Mayer Transformation—a. The Long-Wavelengih Approximation. b. Gauge Change Giving Rise to the Electric Dipole Interaction, ¢. The Advantages of the New Point of View. d. The Equivalence between the Interaction Hamiltonians A * p and E + r.e. Generalizations 4. A Transformation Which Does Not Reduce to a Change of Lagrangian: The Henneberger Transformation—a. Motivation, b. Determination of the Unitary Transformation. Transforms of the Various Operators, c. Physical Interpretation, d. Generalization to a Quantized Field: The Pauli-Fiers~ Kramers Transformation C. The Power-Zienau-Woolley Transformation: The Multipole Form of the Interaction between Charges and Field 1. Description of the Sources in Terms of a Polarization and a Magnetization Density—a. The Polarization Density Associated with a System of Charges, b. The Displacement, c. Polarization Current and Magnetization Current 2. Changing the Lagrangian—a. The Power-Zienau-Woolley Transformation. b. The New Lagrangian. c. Multipole Expansion of the Interaction between the Charged Particles and the Field 3. The New Conjugate Momenta and the New Hamiltonian—a. The Expressions for These Quantities, b. The Physical Significance of the New Conjugate Momenta, c. The Structure of the New Hamiltonian 4. Quantum Electrodynamics from the New Point of View—a. Quantization. b. The Expressions for the Various Physical Variables 5. The Equivalence of the Two Points of View. A Few Traps to Avoid D. Simplified Form of Equivalence for the Scattering S-Matrix 1. Introduction of the S-Matrix 2. The S-Matrix from Another Point of View, An Examination of the Equivalence 3. Comments on the Use of the Equivalence between the 5-Matrices Complement Av —Elementary Introduction to the Electric Dipole Hamiltonian Introduction 1. The Electric Dipole Hamiltonian for a Localized System of Charges Coupled to an External Field—a. The Unitary Transformation Suggested by 255 256 262 266 266 267 269 275 280 280 286 289 293 296 298 298 300 304 304the Long- Wavelength Approximation, b. The Transformed Hamiltonian. c. The Velocity Operator in the New Representation 2. The Electric Dipole Hamiltonian for a Localized System of Charges. 307 Coupled to Quantized Radiation—a. The Unitary Transformation, b. Transformation of the Physical Variables, c. Polarization Density and Displacement. d. The Hamiltonian in the New Representation 3. Extensions—a. The Case of Two Separated Systems of Charges, b. The 312 Case of a Quantized Field Coupled to Classical Sources Complement Biv —One-Photon and Two-Photon Processes: The Equivalence Between the Interaction Hamiltonians A+ p And E +r Introduction 316 1. Notations. Principles of Calculations 316 2. Calculation of the Transition Amplitudes in the Two Representations: 317 The Interaction Hamiltonian A ¢ p. b. The Interaction Hamiltonian E ¢ r. c. Direct Verification of the Identity of the Two Amplitudes 3. Generalizations—a. Extension to Other Processes, b. Nonresonant 325 Processes Complement Civ —Interaction of Two Localized Systemsof Charges from the Power-Zienau-Woolley Point of View Introduction 328 1. Notation 328 2. The Hamiltonian 329 Complement Dv — The Power-Zienau-Woolley Transformation and the Poincare Gauge Introduction 331 1. The Power-Zienau-Woolley Transformation Considered as a Gauge 331 Change 2. Properties of the Vector Potential in the New Gauge 332 3. The Potentials in the Poincare Gauge 333 Complement Ew—Exercises 336 1. An Example of the Effect Produced by Sudden Variations of the Vector Potential 2. Two-Photon Excitation of the Hydrogen Atom. Approximate Results 338 Obtained with the Hamiltonians A+ p and E +r 3. The Electric Dipole Hamiltonian for an Ion Coupled to an External Field 342 4, Scattering of a Particle by a Potential in the Presence of Laser Radiation 344 5. The Equivalence between the Interaction Hamiltonians A p and Z-VV 349 for the Calculation of Transition Amplitudes 6. Linear Response and Susceptibility. Application to the Calculation of the 352 Radiation from a Dipole 7. Nonresonant Scattering. Direct Verification of the Equality of the 356 Transition Amplitudes Calculated from the Hamiltonians A + p and E +r v ‘RODUCTION TO THE COVARIANT FORMULATION OF QUANTUM ELECTRODYNAMICSIntroduction A. Classical Electrodynamics in the Lorentz Gauge 1. Lagrangian Formalism—a. Covariant Notation. Ordinary Notation. b. Selection of a New Lagrangian for the Field, c. Lagrange Equations for the Field, d. The Subsidiary Condition, e. The Lagrangian Density in Reciprocal Space 1. Hamiltonian Formalism—a. Conjugate Momenta of the Potentials. b. The Hamiltonian of the Field, c, Hamilton-Jacobi Equations for the Free Field 3. Normal Variables of the Classical Field—a, Definition, b. Expansion of the Potential in Normal Variables, c. Form of the Subsidiary Condition for the Free Classical Field. Gauge Arbitrariness, d. Expression of the Field Hamiltonian B. Difficulties Raised by the Quantization of the Free Field 1. Canonical Quantization —a. Canonical Commutation Relations. b. Annihilation and Creation Operators, c. Covariant Commutation Relations between the Free Potentials in the Heisenberg Picture 2. Problems of Physical Interpretation Raised by Covariant Quantization —a, The Form of the Subsidiary Condition in Quantum Theory. k. Problems Raised bv the Construction of State Space C. Covariant Quantization with an Indefinite Metric 1, Indefinite Metric in Hilbert Space 2. Choice of the New Metric for Covariant Quantization 3. Construction of the Physical Kets 4, Mean Values of the Physical Variables in a Physical Ket—a. Mean Values of the Potentials and the Fields, b. Gauge Arbitrariness and Arbitrariness of the Kets Associated with a Physical State, c. Mean Value of the Hamiltonian D. A Simple Example of Interaction: A Quantized Field Coupled to Two Fixed External Charges 1, Hamiltonian for the Problem 2. Energy Shift of the Ground State of the Field. Reinterpretation of Coulomb's Law—a. Perturbative Calculation of the Energy Shift. b. Physical Discussion, Exchange of Scalar Photons between the Two Charges, c. Exact Calculation 3. Some Properties of the New Ground State of the Field—a. The Subsidiary Condition in the Presence of the Interaction. The Physical Character of the ‘New Ground State, b. The Mean Value of the Scalar Potential in the New Ground State of the Field 4, Conclusion and Generalization Complement Av —An Elementary Introduction to the Theory of the Electron- Positron Field Coupled to the Photon Field in the Lorentz Gauge Introduction 1. A Brief Review of the Dirac Equation—a. Dirac Matrices, b. The Dirac Hamiltonian, Charge and Current Density, c. Connection with the 361 364 364 369 371 380 380 383 387 387 390 393 396 400 400 401 405 407 408 408Covariant Notation, d. Energy Spectrum of the Free Particle, e. Negative- Energy States. Hole Theory 2. Quantization of the Dirac Field—a. Second Quantization, b. The Hamiltonian of the Quantized Field. Energy Levels, c. Temporal and Spatial Translations 3. The Interacting Dirac and Maxwell Fields—a. The Hamiltonian of the Total System. The Interaction Hamilionian. b. Heisenberg Equations for the Fields, c. The Form of the Subsidiary Condition in the Presence of Interaction Complement Bv —Justification of the Nonrelativistic Theory in the Coulomb Gauge Starting from Relativistic Quantum Electrodynamics Introduction 1. Transition from the Lorentz, Gauge to the Coulomb Gauge in Relativistic Quantum Bleetrodynamies—a. Transformation on the Scalar Photons Yielding the Coulomb Interaction, b. Effect of the Transformation on the Other Terms of the Hamiltonian in the Lorentz Gauge, c. Subsidiary Condition. Absence of Physical Effects of the Scalar and Longitudinal Photons. d. Conclusion: The Relatiuistic Quantum Electrodynamics Hamiltonian in the Coulomb Gauge 2. The Nonrelativistic Limit in Coulomb Gauge: Justification of the Pauli Hamiltonian for the Patticles—a. The Dominant Term Hy of the Haniltonian in the Nonrelativistic Limit; Rest Mass Energy of the Particles, b. The Effective Hamiltonian inside a Manifold, c. Discussion Complement Cv —Bxercises 1. Other Covariant Lagrangians of the Electromagnetic Field 2. Annihilation and Creation Operators for Scalar Photons: Can One Interchange Their Meanings? 3. Some Properties of the Indefinite Metric 4, Translation Operator for the Creation and Annihilation Operators of a Scalar Photon 5. Lagrangian of the Dirac Field. The Connection between the Phase of the Dirac Field and the Gauge of the Electromagnetic Field 6. The Lagrangian and Hamiltonian of the Coupled Dirac and Maxwell Fields 7. Dirac Field Operators and Charge Density. A Study of Some Commutation. Relations References Index. Index References to Exercises are distinguished by an “e” A Absorption (of photons), 316, 325, 338e, 344e, 348e, 349e Action: after the page number. 414 418 424 425 432 441 443 445 446 449 451 454 457 459for a discrete system, 81 fora field, 92 functional derivative, 128 principle of least action, 79, 81 for a real motion, 134, 152e Adiabatic (switching on), 299 Adjoint (relativistic), 411 Angular momentum, see also Multipole, expansion conservation, 8, 139, 200 contribution of the longitudinal electric field, 20, 45 eigenfunctions for a spin-1 particle, 53 for the field + particle systems, 8, 20, 118, 174, 200 for a general field, 152e for a spinless particle, 137 for a spin-1 particle, 49 of the transverse field, 20, 27, 47 Annihilation and creation operators, see also Expansion in a and a*; Translation operator a, and a, operators, 394, 429 a, and , operators, 391 anticommutation relations, 163e, 414 commutation relations, 121, 171, 391 for electrons and positrons, 414, 433 evolution equation, 179, 217, 249e, 420 for photons, 33, 121, 294 for scalar photons, 381, 391, 443e, 446¢ Antibunching, 211 Anticommutation relations: for a complex field, 98 for the Dirac field, 414, 415, 453e, 454e and positivity of energy, 99, 416, 440, 453e for the Schrodinger field, 99, 162e Antihermiticity, see Scalar potential Antiparticle, 187, 413, 433 Approximation: long wavelength, 202, 269, 275, 304, 342¢ nonrelativistic, 103, 122, 200 Autocorrelation, 229 B Basis: in reciprocal space, 25, 36 of vector functions, 51, 55 Bessel:Bessel functions, 345¢ spherical Bessel functions, 56, 71e Bom expansion, 300 Bose-Einstein distribution, 234e, 238¢ Bosons, 99, 161e, 187 Boundary condition: c Canonical (commutation relations), see also Commutation relations: Quantization (general) for a discrete system, 89, 90, 147e, 155e, 258 fora field, 94, 98, 148e, 158e, 380 Center of mass, 232¢, 342¢ Change, see also Gauge: Lagrangian (general); Transformation of coordinates, 84, 88 of dynamical variables, 86, 260 of quantum representation, 260, 262 Characteristic functions, 236e Charge, see also Density conservation, 7, 12, 108, 368, 411, 416, 421 total, 416 Charge conjugation, 438 Classical electrodynamics: in the Coulomb gauge, 111, 121 in the Lorentz. gauge, 364 in the Power-Zienau-Woolley picture, 286 in real space, 7 in reciprocal space, 11 standard Lagrangian, 100 Coherent state, see Quasi-classical states of the field ‘Commutation relations: canonical commutation relations for an arbitrary field, 94, 98, 148 canonical commutation relations for a discrete system, 89, 147e, 155e, 258 covariant commutation relations, 381, 382, 391 for electromagnetic fields in real space, 120,173, 230 for electromagnetic fields in reciprocal space, 119, 145, 380 of the fields with the energy and the momentum, 233e, 383, 417 for free fields in the Heinsenberg picture, 223, 355e, 382 for the operators a and a, 34, 171, 241e, 391, 394, 443e for the operators a and @ , 391, 395 for the particles, 34, 118, 145, 171 Complex, see Dynamical variables: Fields (in general) ‘Compton: scattering, 198 ¢ Periodic boundary conditionswavelength, 202 Conjugate momenta of the electromagnetic potentials: in the Coulomb gauge, 115, 116, 143 in the Lorentz, gauge, 369 in the Power-Zienau-Woolley representation, 289, 291, 294 Conjugate momenta of the particle coordinate: in the Coulomb gauge, 20, 115, 143 in the Goppert-Mayer representation, 270 in the Henneberger representation, 276 for the matter field, 157e in the Power-Zienau-Woolley representation, 289,290, 293 transformation in a gauge change, 267 ‘Conjugate momentum (general): of a complex generalized coordinate, 88, 96,154e of a discrete generalized coordinate, 83, 147e, 256 of a field, 93, 96, 148e in quantum mechanics, 258, 266 transformation in a change of generalized coordinates, 85 transformation in a change of Lagrangian, 257 Conservation: of angular momentum, 8, 139, 200 of charge, 7, 12, 108, 368, 411, 416, 421 of energy, 8, 61¢, 137, 200 of momentum, 8, 6le, 138, 200, 232e Constant of the motion, 8, 61, 134, 152¢, 200, 370 Contact interaction, 42 Continuous limit (for a discrete system), 126, 147e Convolution product, 11 Correlation function, 181, 191, 227, See also Intensity correlations Correlation time, 191 Coulomb, see also Coulomb gauge; Energy: Scalar photons field, 16, 122, 172, 295 interaction, 18, 122, 330, 401, 426, 435 interaction by exchange of photons, 403 potential, 16, 67e, 172, 407 self-energy, 18, 71¢, 201 Coulomb gauge, see also Hamiltonian (total): Lagrangians for electrodynamics: Transformation definition, 10, 113 electrodynamics in the Coulomb gauge, 10, 113, 121,169,439 relativistic QE.D. in the Coulomb gauge, 424, 431 Counting signals, see Photodetection signals Covariant: commutation relations, 391formulation, 361 notation and equations, 10, 17, 364, 411, 449e Covariant Lagrangians: for classical particles, 106 for coupled electromagnetic and Dirac fields, 451e for the Dirac field, 449¢ for the electromagnetic field (standard Lagrangian), 106, 365 Fermi Lagrangian, 366 interaction Lagrangian, 106, 365 in the Lorentz gauge, 365, 369, 441¢ Creation operator, see Annihilation and creation operators Cross-section, see Scattering ‘Current: density, 7, 101, 115, 410, 419 four-vector, 10, 365, 411 of magnetization, 284 of polarization, 284 Cutoff, 124, 190, 200, 287 D d'Alambertian, 10, 367 Damping (radiative), 71e, 76 Darwin term, 440 Delta function (transverse), 14, 36, 38, 42, 64e, 120, 173, 231¢ Density, see also Quasi-probability density of charge, 7, 101, 309, 410, 419, 434, 454e of current, 7, 101, 115, 410, 419 Hamiltonian, 93, 106, 147e, 158e, 370 Lagrangian, 91, 101, 106, 113, 147e, 157e, 167e, 365, 369, 441¢ of magnetization, 42, 284, 285, 292 of polarization, 281, 292, 308, 329 Diamagnetic energy, 290, 293 Dipole-dipole interaction: electric, 313 magnetic, 43 Dipole moment, see Electric dipole: Magnetic dipole moment Ditac, see also Matter field; Spinors delta function, 94 equation, 408, 449e, 452e Hamiltonian, 410 matrices, 409 Discretization, 31 Dispacement, 282, 291, 292, 308, 310 Dynamical variables: canonically conjugate, 34, 86, 93, 257, 258, 369change of dynamical variables in the Hamiltonian, 86, 260 change of dynamical variables in the Lagrangian, 84 complex dynamical variables, 87, 90 for a discrete system, 81 fora field, 90 redundancy,109, 113, 154e, 157e, 362 E Effective (Hamiltonian), 435, 438 Einstein, 204 Electric dipole: approximation, 270 interaction, 270, 288, 304, 306, 312, 313, 342 moment, 270, 288, 306, 343 self-energy, 312 wave, Tle Electric field, see also Electromagnetic field: Expansion in the Coulomb gauge, 117, 122, 172 longitudinal, 15, 64e, 117, 172, 283 of an oscillating dipole moment, 71e, 353e in the Power-Zienau-Woolley picture, 295 total, 66e, 117, 172, 291, 295, 310, 330, 355e transverse, 21, 24, 27, 32, 64e, 117, 171, 287, 295, 310 Electromagnetic field, see also Expansion in normal variables: External field: Quantization of the electromagnetic field associated with a particle, 68¢ free, 28, 58, 181, 221, 230e, 241e mean value in the indefinite metric, 396 in real space, 7 in reciprocal space, 12 tensor F", 17, 106, 365, 378 Electromagnetic potentials, see also Free (fields, potential): Gauge covariant commutation reactions, 382 definition and gauge transformation, 9 evolution equations, 9, 10, 366, 367 four-vector potential, 10, 364, 376, mean value in the indefinite metric, 396, 406 retarded, 66 Electron, see also Matter field classical radius, 75¢ elastically bound, 74e g-factor, 439 Electron-positron pairs, 123, 413, 417 Elimination:of a dynamical variable, 85, 154e, 157¢ of the scalar potential, 111 Emission (of photons), 344e, 348e, 349¢ Energy, see also Hamiltonian; Self-energy conservation of, 8, 61e, 137, 200 Coulomb energy, 18, 114, 173, 283, 401, 403, 426 of the free field, 183, 378 negative energy states, 413 of the system field + particles, 8, 19, 116 of the transverse field, 26, 31 Equations, see Dirac; Hamilton's equations; Heisenberg: Lagrange's equations: Maxwell equations; Newton-Lorentz equations; Poisson; Schrodinger Equivalence: between the A+ p and E +r pictures, 272, 296, 316, 321, 337e, 356e between the Ae p and Z¢VV pictures, 349e between relativistic Q.E.D. in the Lorentz and the Coulomb gauges, 424 between the various formulations of electrodynamics, 253, 300, 302 Expansion in a and at (or ina and @): of the electric and magnetic fields, 171, 241e of the four-vector potential, 391 of the Hamiltonian and momentum in the Lorentz gauge, 382, 391 of the Hamiltonian and momentum of the transverse field, 172 of the transverse vector potential, 171 Expansion in normal variabl of the electric and magnetic fields, 27, 28, 32 of the four-vector potential, 372, 376 of the Hamiltonian and momentum in the Lorentz gauge, 378, 379 of the transverse field angular momentum, 27, 48 of the transverse field Hamiltonian, 27, 31 of the transverse field momentum, 27, 31 of the transverse vector potential, 29, 31 Extemal field, 141, 172, 178, 180, 198, See also Hamiltonian for particles in an external field: Lagrangians for electrodynamics External sources (for radiation), 24, 219, 314, 370, 372, 400, 418 F Factored states, 207 Fermi: golden rule, 323 Lagrangian, 366 Fermion, 99, 16le, 413,414 Fields (in general), see also Angular momentum: Energy; Hamiltonian (general considerations); Lagrangian (general); Momentum; Quantization (general) complex, 95real, 90 transverse and longitudinal, 13, 37 Fierz, see Pauli-Fierz-Kramers transformation Final, see Initial and final states of a process Fock space, 31, 175 Fourier transform, 11, 12, 15, 56, 97 Four-vector: current, 10, 365, 411 field energy-momentum, 379 potential, 10, 364, 376 Free (fields, potentials), 28, 58, 183, 205, 373, 376, 382,414 Fresnel mirror, 208 Functional derivative, 92, 126 G Gauge, see also Coulomb gauge; Lorentz gauge; Poincare gauge gauge transformation and phase of the matter field, 167e, 449e invariance, 8, 17, 107, 269 transformation, 9, 13, 108, 255, 267, 270, 331, 368, 375, 397 Generalized coordinates: change of, 86, 260 complex, 87, 88 real, 81, 84 Goppert-Mayer transformation, 269, 275, 304 Ground state: of the quantized Dirac field, 417 of the radiation field, 186, 189, 252e, 385, 386, 394 H Hamiltonian (general considerations), see also Effective, (Hamiltonian) with complex dynamical variables, 88, 97, 154e, 157e fora discrete system, 83, 147e for a field, 93, 97, 148¢ Hamiltonian and energy, 83, 136, 146e in quantum theory, 89, 259 transformation of, 258, 261, 263 Hamiltonian of the particles: Dirac Hamiltonian, 410 expression of, 144, 197 Pauli Hamiltonian, 432 physical meaning in various representations, 271, 297 of the quantized Dirac Field, 415 for two particles with opposite charges, 232e for two separated systems of charges, 313, 328 Hamiltonian for particles in an external field: for a Dirac particle, 410electric dipole representation (E + r), 271, 304, 320 Henneberger picture, 277 for an ion, 342¢ for the quantized Dirac field, 419 standard representation (A + p), 144, 198, 266, 317 Hamiltonian for radiation coupled to external sow in the Couilomb gauge, 218 in the electric dipole representation, 314, 353e in the Lorentz gauge, 370, 400, 418 Hamiltonian (total): in the Coulomb gauge, 20, 33, 116, 138, 173, 439 in the Coulomb gauge with external fields, 144, 174,198 of coupled Dirac and Maxwell fields, 419, 431, 451e in the Power-Zienau-Wooley picture, 289, 292, 295, 329 Hamilton's equations: for a discrete system, 83 fora field, 94, 132, 371 Heaviside function, 226 Heisenberg: equation, 89 equations for a and a+ 179, 217, 249, 420 equations for the matter fields, 99, 161e, 420 equations for the particle, 177 picture, 89, 176, 185, 218, 221, 382 relations, 241e, 248 Henneberger transformation, 275, 344e, 349e Hilbert space, 89, 387 Hole theory, 413, Hydrogen atom: Lamb transition, 327 Is-2s two-photon transition, 324, 338¢ I Indefinite metric, see also Scalar potential definition and properties, 387, 391, 445¢ and probabilistic interpretation, 390, 392 Independent variables, 95, 109, 121, 362, See also Redundancy of dynamical variables Initial and final states of a process, 264, 271, 296, 300, 302, 317, 326, 337e Instantaneous, see also Nonlocality Coulomb field and transverse field, 16, 21, 64e, 67e, 122, 291, 292 interactions, 18, 122, 313, 330 Intensity correlations, 186 Intensity of light, 185 Interaction Hamiltonian between particles and radiation: in the Coulomb gauge, 197, 2326 in the electric dipole representation, 271, 307, 312, 315in the Power-Zienau-Woolley representation, 290, 292, 296, 329 in relativistic Q.E.D., 419 Interactions, see Contact interaction; Coulomb: Dipole-dipole interaction; Electric dipole; Instantaneous: Magnetic dipole moment: Quadrupole electric (momentum and interaction): Retarded: Hamiltonian Interference phenomen: with one photon, 208, 210 quantum theory of light interference, 204 with two laser beams, 208, 212 with two photons, 209, 211 Interferences for transition amplitude, 213 Invariance, see also Covariant gauge invariance, 9, 107, 167e, 267 relativistic invariance, 10, 15, 106, 114 translational and rotational, 134, 153e, 200, 370 Jon (interaction Hamiltonian with the radiation field), 342e K Kramers, see Pauli-Fierz-Kramers transformation Kronecker (delta symbol), 94, 148¢ L Lagrange’s equations: with complex dynamical variables, 87, 96, 154e fora discrete system, 82, 129, 147e for the electromagnetic potentials, 104, 142, 150¢, 151¢, 366 fora field, 92, 96, 131, 147e, 150e for a matter field, 157e, 167e, 367, 449¢ for the particles, 103, 142, 15le Lagrangian (general), see also Density, Lagrangian: Functional derivative: Matter field with complex dynamical variables, 87, 95, 154e, 157e of a discrete system, 81, 147e elimination of a redundant dynamical variable, 84, 154e, 157e equivalent Lagrangians, 82, 92, 108, 256 of a field, 91, 95, 147e formalism, 79, 81 linear in velocities, 154e, 157e Lagrangians for electrodynamics, see also Covariant Lagrangians; Standard Lagrangian in the Coulomb gauge, 113, 137 with external fields, 142, 143, 266, 271, 449e in the Power-Zienau-Woolley picture, 287 Lamb: shift, 191 transition, 327 Least-action principle, 79, 81 Light intensity, 185 Linear response, 221, 352¢Linear susceptibility, 221, 352¢ Locality, 12, 14, 15, 21, 103, 291, See also Instantaneous; Nonlocality Localized systems of charges, 281, 304, 307 Longitudinal: basis of longitudinal vector functions, 53 contribution of the longitudinal electric field to the energy, momentum and angular momentum, 17, 19, 20 electric field, 15, 64e, 172, 283 normal variables, 374 photons, 384, 430 vector fields, 13 vector potential, 112, 255 Longitudinal vector potential: in the Coulomb gauge, 16, 113 in the Lorentz gauge, 22 in the Poincare gauge, 332 Lorentz equation, 104, 178, See also Lorentz gauge: Subsidiary condition Lorentz gauge, see also Subsidiary condition classical electrodynamics in the Lorentz. gauge, 364 definition, 9 relativistic Q.E.D. in the Lorentz gauge, 361, 419, 424, 453e M Magnetic dipole moment: interaction, 43, 288 orbital, 288 spin, 44, 197, 439 Magnetic field, 21, 24, 27, 32, 42, 118, 171, See also Expansion Magnetization: current, 284 density, 42, 284, 292 Mass: correction, 69¢ rest mass energy, 432 Matter field: Dirac matter field, 107, 366, 408, 414, 433, 451e, 454e quantization, 98, 161e, 361, 414 Schrodinger matter field, I57e, 161e, 167e Maxwell equations, see also Heinsenberg: Normal variables of the radiation covariant form, 17, 366 for the potentials, 9, 10, 366 quantum Maxwell equations, 179 in real space, 7 in reciprocal space, 12, 21 Mean value in the indefinite metric, 389, 396, 398, 406 Mechanical momentum, 20, 177, 271, 290Mode, 24, 27, 374, See also Normal mode, Normal variables of the radiation: Expansion Momentum, see also Commutation: Expansion in normal variables: Expansion in a and at (orinaand 7) conservation, 8, 61e, 138, 200 contribution of the longitudinal field, 19, 20 of the Dirac field, 451e of the electromagnetic field in the Lorentz. gauge, 370, 379 of a general field, 152e momentum and velocity, 20, 177, 271, 290 for a particle, 20, 177 of the particle + field system, 8, 20, 118, 139, 174, 199 of the Schrodinger field, 158e of the transverse field, 19, 27, 31, 172, 193, 188 Multiphoton amplitudes (calculations in various representations), 316, 325, 338¢, 34e, 3480, 349 Multipole: expansion, 287 waves, 45, 55, 58, 60 N Negative energy states, 413 Negative frequency components, 29, 184, 193,422 Newton-Lorentz equations, 7, 104, 178 Nonrelativistic: approximation, 103, 122, 200 limit, 424, 432, 439 Nonresonant processes, 325, 356¢ Nonlocality, 14, 15, 21, 151e, See also Instantaneous; Locality Norm: in the indefinite metric, 388, 445e, 447¢ negative, 385 Normal mode, 24, 27, 374, See also Normal variables of the radiation: Expanion Normal order, 185, 195, 237e Normal variables of the radiation, see also Expansion in normal variables a, and a, normal variables, 375, 376, 378 analogy with a wavefunction, 30 definition and expression, 23, 25, 29, 371 discretization, 31 evolution equation, 24, 26, 32, 66¢, 219, 371, 372 Lorentz subsidiary condition, 374 quantization, 33, 171 scalar and longitudinal normal variables, 372, 374, 379 transverse normal variables, 25, 29, 374° Observables, see Physical variables Operators in the indefinite metric: adjoint, 388 eigenvalues and eigenfunctions, 389, 445¢ hermitian, 388, 445¢ Order: antinormal, 237e normal, 185, 195, 238¢ iy Parseval-Plancherel identity, 11 Particles see Conjugate momenta of the particle coordinates; Matter field: Hamiltonian for particles in an external field Particle velocities: in the Coulomb gauge, 117, 177 in the Goppert-Mayer approach, 271, 306 in the Henneberger approach, 277 in the Power-Zienau-Woolley approach, 290, 295 Paul exclusion principle, 163¢, 413, 416 Hamiltonian, 432 matrices, 410, 437 Pauli-Fierz-Kramers transformation, 278, 429 Periodic boundary conditions, 31 Phase: of an electromagnetic field mode, 208, 212, 243e of a matter field and gauge invariance, 167e, 449e Photodetection signals, see also Interference phenomena double counting signals, 185, 209, 214 single counting signals, 184, 188, 206, 213 Photon, see also Annihilation and creation operators: Bose-Einstein distribution: Interference phenomena; S-matrix: States of the radiation field; Wave-particle duality as an elementary excitation of the quantized radiation field, 30, 187 longitudinal and scalar photons, 384, 392, 403, 425, 430, 443e, 446e nonexistence of a position operator, 30, 50, 188 photon number operator, 187 single-photon states, 187, 205, 208, 210, 385 transverse photons, 186, 385 wavefunction in reciprocal space, 30 Physical meaning of operators: general, 259, 269 in the Goppert-Mayer approach, 271, 306, 310 in the Henneberger approach, 277, 345¢in the Power-Zienau-Woolley approach, 290, 292 Physical states, 384, 394, 396, 405, 423, 430, 443e, See also Physical meaning of operators: Physical variables; Subsidiary condition Physical variables, see also Angular momentum: Electric field: Energy; Magnetic field: Momentum; Particle velocities: Photodetection signals: Physical meaning of operators; Position operator in classical theory, 257 corresponding operators in various representations, 116, 117, 271, 277, 294, 306, 310 ‘mean value in the indefinite metric, 396 in quantum theory, 259, 296 transformation of the corresponding operators, 260, 263 Planck, 1 Poincare gauge, 331, 333 Poisson: brackets, 86 equation, 10, 345¢ Polarization: current, 284 density, 281, 292, 308, 329 Polarization of the radiation: polarization vector, 25, 376 sum over transverse polarizations, 36 Position operator, see also Photon; Translation operator in the Henneberger approach, 276, 345¢ for the particles, 33, 118, 258 Positive: positive energy slates, 412 positive frequency components, 29, 184, 193,422 Positron, 408, 413 Potential, see Longitudinal vector potential; Scalar potential: Transverse vector potential Power-Zienau-Woolley transformation, 280,286, 328, 331 P-representation, 195, 206, 211, 236e, 251e Processes, see Absorption (of photons); Emission (of photons): Multiphoton (amplitudes (calculations in various representations): Nonresonant processes: Resonant, processes: Scattering: S-matrix Q Quadrupole electric (momentum and interaction), 288 Quantization (general), see also Matter field with anticommutators, 98, 162e, 453e canonical quantization, 34, 89, 258, 380 for a complex field, 98, 99, 161¢ for areal field, 94, 148e second quantization, 414, 439Quantization of the electromagnetic field: canonical quantization in the Coulomb gauge, 119, 144 canonical quantization in the Power-Zienau-Woolley representation, 294 covariant quantization in the Lorentz gauge, 380, 383, 387, 391 elementary approach, 33 methods, 33, 34 Quantum electrodynamics (Q.E.D.): in the Coulomb gauge, 169 in the Power-Zienau-Woolley picture, 293 relativistic QED. in the Coulomb gauge, 424,431 relativistic QE.D. in the Lorentz gauge, 361,419, 424, 453e Quasi-classical states of the field, see also Photodetection signals; Quasi-probability density definition, 192 graphical representation, 242e interferences with, 207, 209 production by external sources, 217, 404 properties, 194, 47e Quasi-probability density: suited to antinormal order, 236e, 250¢ suited to normal order, 195, 206, 211, 236e, 250e R Radiation emitted by an oscillating dipole, 71e, 352¢ Radiation Hamiltonian: eigenstates of, 186 as a function of a and a* 172, 197, 24le, 296, 382 as a function of a and @, 391 as a function of the conjugate variables, 116, 144, 290, 296, 370 as a function of the fields, 18, 312 as a function of the normal variables, 27, 31, 378 in the Lorentz. gauge, 370, 378, 382, 391, 398 physical meaning, 292, 312 Radiation reaction, 68e, 74e Radiative damping, 7le, 76e Raman scattering, 326 Rayleigh scattering, 75e, 198, 326 Reciprocal: half-space, 102 space, 11, 36 Redundancy of dynamical variables, 109, 113, 154e, 157e, 362, See also independent variables Relativistic, see also Covariant; Covariant Lagrangian: Quantum electrodynamics (QED.description of classical particles, 107 Dirac field, 366, 408, 414, 433, 45le, 454e modes, 123 Resonai processes, 316, 326, 349e scattering, 75¢ Retarded, see also Instantaneous field, 21, 310, 330 potential, 66¢ s Scalar photons, 384,392,403,425,430,443e, 446e Scalar potential, see also Expansion in a and a* (or ina and 7), Expansion in normal variables absence of a conjugate momentum with the standard Lagrangian, 109, 362 antihermiticity in the Lorentz gauge, 392 conjugate momentum in the Lorentz gauge, 369 in the Coulomb gauge, 16, 22, 67e climination from the standard Lagrangian, 111 in the Poincare gauge, 333 Scalar product: in a Hilbert space, 387 with the indefinite metric, 387, 395, 445¢ Scattering, see also Compton: Raman scattering; Rayleigh scattering: Thomson scattering: Transition amplitudes cross section, 74e, 346e nonresonant scattering, 356e in presence of radiation, 344e process, 326 resonant scattering, 75e Schrodinger: equation, 89, 157e, 167e, 176, 261, 263 representation, 89, 176, 219 Schrodinger field: Lagrangian and Hamiltonian, 157e, 167e quantization, 161e Schwarzchild, 79 Second quantization, 414 Selection rules, 199, 233¢ Self-energy Coulomb, 18, 71e, 201 dipole, 312 of the transverse polarization, 290, 329 S-matrix:definition, 299, 317 equivalence in different representations, 298, 302, 321, 349e, 356¢ for one- and two-photon processes, 317, 349¢ Sources (classical or external), 24, 217, 314, 370, 372, 400, 418 Spectral density, 191 Spin: magnetic moment, 44, 197, 439 spin-statistics theorem, 99 Spin-1 particle, 49 Spin-orbit interaction, 440 Spinors: Dirac spinors, 409, 412, 433 two-component Pauli spinors, 434 Squeezed states, 245e, 246e, 248, 250 Standard Lagrangian: difficulties for the quantization, 109 expression, 100 symmetries, 105 State space, see also Subsidiary condition in the Coulomb gauge, 175 in the covariant formulation, 385 for scalar photons, 392, 443e States of the radiation field, see also Physical states: Quasi-classical states of the field: Vacuum factored states, 205, 207 graphical representation, 241¢ single-photon states, 187, 205, 208, 210, 385 squeezed states, 243e, 246e, 248e, 250e two-photon states, 211 Subsidiary condition: in classical electrodynamics, 9, 10, 22, 368, 370, 374, 442¢,443¢ in presence of interaction, 406, 421, 430 for the quantum free field, 384, 386, 394 Sudden switching-on of the potential, 264, 336 Symmetries and conservation laws, 134 of the standard Lagrangian, 105 7 ‘Thomson scattering, 75e, 198 ‘Transformation, see also Physical variables; Unitary transformation; entries under Gauge; Hamiltonian; Lagrangian of coordinates and velocities, 85 from the Coulomb gauge to the Lorentz, gauge (or vice versa), 63e, 425 Goppert-Mayer transformation, 269, 304Henneberger transformation, 275, 344e, 349 Pauli-Fierz-Kramers transformation, 278,429 Power-Zienau-Woolley transformation, 280, 287, 328, 331 of the state vector, 261, 263, 268 ‘Transition amplitudes definition and calculation, 176, 271, 316, 337e, 338e, 346¢ identity in different pictures, 264, 269, 273, 297, 316, 321, 349e, 3560 interference between, 213 ‘Transition matrix, 300, 356¢ Transition rate, 323 Translation operator: for the a and a* operators, 195, 308 for the a and @ operators, 404, 425, 446e infinitesimal generators, 163¢, 199, 383, 417 for the momentum of a particle, 305 for the position of a particle, 276 Transverse, see also Expansion; Instantaneous: Nonlocality: Photon basis of transverse vector functions, 25, 37, 53 commutation relation for the transverse field, 119, 223, 230e delta function, 14, 36, 38, 42, 64e, 120, 173, 231¢ displacement, 283, 291, 295, 310 energy, momentum and angular momentum of the transverse field, 18, 19, 47, 48, 174, 312 equations of motion of the transverse field, 21 electric field, 21, 24, 27, 32, 64e, 117, 171, 287, 295, 310 magnetic field, 21, 24, 27, 32, 42, 118, 171 projector onto the subspace of transverse fields, 37 summation over transverse polarizations, 36 vector field, 13, 50 vector potential, 17, 29, 31, 119, 171, 223, 294, 377, 396 ‘Transverse vector potential, see also Expansion: Instantaneous: Nonlocality commutation relations, 119, 223, 230e. conjugate momentum, 115, 289 gauge invariance, 17 itary transformation, see also Translation operator associated with a change of Lagrangian, 260, 262, 296 associated with a gauge transformation, 268, 271 on the Hamiltonian, 262, 276, 304, 343¢ Vacuum, 186, 189, 252e, 385, 386, 394 Vacuum fluctuations, 191, 199, 279Vector potential, see Longitudinal vector potential: Transverse vector potential Velocity, see Particle velocities Ww Wavefunction of the photon, 30, 50, See also Photon Wavelength scale, 202, See also Approximation: Compton Wave-particle duality, 204, 215 Waves: multipole waves, 45, 55 traveling plane waves, 28 Woolley, see Power-Zienau-Woolley transformation Z Zienau, ,see Power-Zienau-Woolley transformationIntroduction The electromagnetic field plays a prominent part in physics. Without ‘going back to Maxwell, one can recall for example that itis from the study of light that the Planck constant and the ideas of wave-particle duality arose for the first time in physics. More recently, the electromagnetic field has appeared as the prototype of quantum gauge fields IC is therefore important to develop a good understanding of the dynamics of the electromagnetic field coupled to charged particles, and in particular of its quantum aspects. To this end, one must explain how the electromagnetic field can be quantized and how the concept of photon arises, One must also specify the observables and the states which describe the various aspects of radiation, and analyze the Hamiltonian which governs the coupled evolution of photons and atoms. It is to the study of these problems that this volume is devoted. “The quantization of the electromagnetic field is the central problem around which the various chapters are organized. Such a quantization requires some caution, owing to the gauge arbitrariness and to the redun- dancy associated with the vector and scalar potentials. As a result, we will treat these problems at several levels of increasing difficulty. In Chapter 1, we begin with the Maxwell-Lorentz. equations which dlescribe the evolution of an ensemble of charged particles coupled to the electromagnetic field and show that a spatial Fourier transformation of the field allows one to see more clearly the actual independent degrees of freedom of the field. We introduce in this way the normal variables which describe the normal vibrational modes of the field in the absence of sources. Quantization then is achieved in an elementary fashion by quan- tizing the harmonic oscillators associated with each normal mode, the normal variables becoming the creation and annihilation operators for a photon. ‘The problem is treated again ina more thorough and rigorous fashion in Chapter HI, starting with the Lagrangian and the Hamiltonian formula- tion of electrodynamics. One such approach allows one to define unam- biguously the canonically conjugate field variables. This provides also a straightforward method of quantization, the canonical quantization: two operators whose commutator equals iA then represent the two correspondPreface The spectacular development of new sources of electromagnetic radia- tion spanning the range of frequencies from rf to the far ultraviolet (lasers, masers, synchrotron sources, etc.) has generated considerable interest in the interaction processes between photons and atoms. New methods have been developed, leading to a more precise understanding of the structure and dynamics of atoms and molecules, to better control of their internal and external degrees of freedom, and also to the realization of novel radiation sources. This explains the growing interest in the low-energy interaction between matter and radiation on the part of an increasing number of researchers drawn from physics, chemistry, and engineering This work is designed (0 provide them with the necessary background (0 understand this area of research, beginning with elementary quantum theory and classical electrodynamics Such a program is actually twofold. One has first to set up the theoretical framework for a quantum description of the dynamics of the total system (electromagnetic field and nonrelativistic charged particles), and to discuss the physical content of the theory and its various possible formulations. This is the subject of the present volume, entitled Photons and Aioms— Introduction t0 Quantum Electrodynamics. One has also 10 describe the interaction processes between radiation and matter (emission, absorption, scattering of photons by atoms, etc.) and to present various theoretical methods which can be used to analyze these processes (per- turbative methods, partial resummations of the perturbation series, master equations, optical Bloch equations, the dressed-atom approach, etc). These ‘questions are examined in another volume entitled Interaction Processes between Photons and Atoms. The objectives of these two volumes are thus clearly distinct, and according to his interests and to his needs, the reader may use one volume, the other, or both. An examination of the topics presented here clearly shows that this book is not organized along the same lines as other works treating {quantum electrodynamics. In fact, the majority of the latter are addressed to an audience of field theorists for whom such ideas as covariance, relativistic invariance, matter fields, and renormalization, to name a few, are considered as fundamentals. On the other hand, most of the books dealing with quantum optics, and in particular with laser optics, treat the XVExv Preface fundamentals of electrodynamics, as well as the problems posed by quantization of radiation, rather succinctly, We have chosen here an approach between these two, since there seems to be a real need for such an intermediate treatment of this subject. ACKNOWLEDGMENTS. This book is an outcome of our teaching and research, which we have worked at over a period of many years at the College de France, at the University P. et M. Curie, and at the Laboratoire de Physique de 'Ecole Normale Superieure. We would like to express our thanks here to our friends and coworkers who have participated in our research and who hhave made us the beneficiaries of their ideas. ‘We want to thank particularly Jean Dalibard, who has been of such great help in the development of the exercises.2 Introduction ing classical conjugate variables. We show nevertheless that such a theoret- ical approach is not directly applicable to the most commonly used Lagrangian, the standard Lagrangian. This is due to the fact that the dynamical variables of this Lagrangian, the vector and scalar potentials, are redundant. The most simple way of resolving this problem, and then quantizing the theory, is to choose the Coulomb gauge. Other possibilities exist, each having theit advantages and disadvantages: these are examined ater in Chapter IV (Poincaré gauge) and Chapter V (Lorentz gauge) Many of the essential aspects of quantum electrodynamics in the Coulomb gauge are discussed in detail in Chapter III. ‘These inelude the quantum equations of motion for the coupled system charges + field: the study of the states and observables of the free quantized field, of the Properties of the vacuum, and of coherent states; and the analysis of interference and wave-particle duality in the quantum theory of radiation. We also examine in detail the properties of the Hamiltonian which describes the coupling between particles and photons. This last subject is treated in more detail in Chapter IV, which is devoted to other equivalent formulations of electrodynamics derived from the Coulomb gauge. We show how it is possible to get other descriptions of electrodynamics, better adapted to this or that type of problem, either by changing the gauge or by adding to the standard Lagrangian in the Coulomb gauge the total derivative of a function of the generalized coordinates of the system, or else by directly performing a unitary transformation on the Coulomb-gauge Hamiltonian. Emphasis is placed ‘on the physical significance the various mathematical operators have in the different representations and on the equivalence of the physical predictions derived from these various formulations. It is here that a satisfactory understanding of the fundamentals of quantum eleetrodynam- ics is essential if one is to avoid faulty interpretations. concerning for example the interaction Hamiltonians A p or Er From the point of view adopted in Chapters I and IV, the symmetry between the four components of the potential four-vector is not main- tained. The corresponding formulations are thus not adaptable to a covariant quantization of the field, These problems are dealt with in Chapter V, which treats the quantization of the field in the Lorentz gauge. We explain the difficulties which arise whenever the four components of the potential are treated as independent variables. We point out also how it is possible to resolve this problem by selecting, using the Lorentz condition, a subspace of physical states from the space of the radiation states We mention finally that, with the exception of the complements of Chapter V, the particles are treated nonrelativistically and are described by Schrédinger wave functions or Pauli spinors. Such an approximation is generally sulficient for the low-energy domain treated here. In addition.Introduction 3 the choice of the Coulomb gauge, which explicitly yields the Coulomb interaction between particles which is predominant at low energy. is very convenient for the study of bound states of charged particles, such as atoms and molecules. This advantage holds also for the other formulations derived from the Coulomb gauge and treated in Chapter IV. A quantum relativistic deseription of particles requires that one consider them as clementary excitations of a relativistic matter field, such as the Dirae field for electrons and positrons, We deal with these problems in two comple. ments in Chapter V. We show in these complements that it is possible to justify the nonrelativistic Hamiltonians used in this volume by considering them as “effective Hamiltonians” acting inside manifolds with a fixed number of particles and derived from the Hamiltonian of relativistic quantum electrodynamics, in which the number of particles, like the number of photons, is indeterminate. This volume consists of five chapters and nineteen complements. The complements have a variety of objectives. They give more precision to the physical or mathematical concepts introduced in the chapter to which they are joined, or they expand the chapter by giving examples of applications, by introducing other points of view, or by taking up problems not studied in the chapter. The last complement in each chapter contains worked exercises. A short, nonexhaustive bibliography is given, either in the form of general references at the end of the chapter or complement, or in the form of more specialized references at the foot of the page. A detailed list Of the books, cited by the author’s name alone in the text, appears at the end of the volume. It is possible to read this volume serially from beginning to end. It is also possible, however, to skip certain chapters and complements in a first study. If one wishes to get a flavor of field quantization in its simplest form, and to understand the particle and wave aspects of radiation and the dynamics of the system field + particles, one can read Chapter I, then Chapter III and its Complement Ay. Reading Complements Ay and Byy can also give one a simple idea of the electric dipole approximation and of the equivalence of the interaction Hamiltonians Ap and Er for the study of one- of two-photon processes. ‘A graduate student or researcher wanting to deepen his understanding of the structure of quantum electrodynamics and of the problems tied to the gauge arbitrariness, should extend his reading to Chapters II, 1V, and V and choose those complements which relate best t0 his needs and his area of interestCHAPTER | Classical Electrodynamics: The Fundamental Equations and the Dynamical Variables The purpose for this first chapter is to review the basic equations of classical electrodynamics and to introduce a set of dynamical variables allowing one to characterize simply the state of the global system field + particles at a given instant The chapter begins (Part A) with a review of the Maxwell—Lorentz equations which describe the joint evolution of the electromagnetic field and of a set of charged particles. Some important results concerning, the constants of motion, the potentials, and gauge invariance are also reviewed. With a view to subsequent developments, notably quantization, one then shows (Part B) that classical electrodynamics has a simpler form in reciprocal space, after a Fourier transformation of the field, Such a transformation allows a simple decomposition of the electromagnetic field into its longitudinal and transverse components. It is then evident that the longitudinal electric field is not a true dynamical variable of the system, since it ean be expressed as a function of the positions of the particles. The following part (Part C) introduces finear combinations of the transverse electric and magnetic fields in reciprocal space which have the important property of evolving independently in the absence of particles and which then describe the normal vibrational modes of the free field. These new dynamical variables, called normal variables. play a centeal role in the theory, since they become, after quantization, the creation and annihilation operators for photons. All the field observables can be ex- pressed as a function of these normal variables (and the particle variables) The chapter ends finally (Part D) with a discussion of the various possible strategies for quantizing the foregoing theory. One simple. eco- nomic method, albeit not very rigorous. consists of quantizing each of the harmonic oscillators” associated with the various normal modes of Vibration of the field. One then gets all the fundamental commutation 56 (Chassical Electrodynamics 1 relations necessary for Chapter III. The problem is approached in a more rigorous manner in Chapter II, beginning with a Lagrangian and Hamilto- nian formulation of electrodynamics. Finally, Complement B, compiles some results relative to the angular momentum of the electromagnetic field and (0 the multipole expansion of the field.AL ‘The Fundamental Equations in Real Space 7 A—THE FUNDAMENTAL EQUATIONS IN REAL SPACE 1. The Maxwell-Lorentz Equations The basic equations are grouped into two sets. First, the Maxwell equations relate the electric field E(r, 1) and the magnetic field Bir, 1) to the charge density o(r, 1) and the current j(r, 1) ve Ben = Leen (A.L.a) v= Bu.) =0 (1b) JoxEeo -5 (AL [vx amo Ane Next, the Newton—Lorer cle a, having mass m,, charge qq, position r,(1), and velocity ¥,(1), under the influence of elecitic and magnetic forces exerted by the fields m, Sy) = q LE.) + 400) x Ber. 0]. (A.2) re ‘The equations (A.2) are valid only for slow, nonrelativistic particles (0, <0). From (A.1.a) and (A.1.d) one can show that pin) + Ve) (a3) Such an equation of continuity expresses the local conservation of the global electric charge, = arp. Oe) ‘The expression of p and j as a function of the particle variables is Ya dtr - 10] 5a) f ptr.) | io = Davo obe el] (A.5.b) One can show that Equations (A.5) satisfy the equation of continuity {A3).8 Classical Electrodynamics 1Az Equations (A.1) and (A.2) form two sets of coupled equations. The evolution of the field depends on the particles through p and j. The motion of the particles depends on the fields E and B. The equations (A.1) are first-order partial differential equations, while the equations (A.2) are second-order ordinary differential equations. It follows that the state of the global system, field + particles, is determined at some instant tq by aiving the fields E and B at all points r of space and the position and velocity r, and ¥, of each particle «: { B(Q,10). BUF, 10). lg). Valo) (A.6) 11 is important to note that in the Maxwell equations (A.1),r is not a dynamical variable (like r,) but a continuous parameter labeling the field variables. 2. Some Important Constants of the Motion Starting with Equations (A.1) and (A.2) and the expressions (A.S) for » and j, one can show (see Exercise 1) that the following functions of E, B, and PLEX. 1) + 2 BUD] (AD 1 & H= Sym +8 Pesta far EGr.1) x Brn) (a.8) I= THO x mvlO + & Jere x (Et) x Bir} (4.9) are constants of the motion, that is, independent of r His the total energy of the global system field + particles, P is the coral ‘momentum, and J the total angular momentum. The fact that these quantities are constants of the motion results from the invariance of the equations of motion with respect to changes in the time origin, the coordinate origin, and the orientation of the coordinate axes. (The connec tion between the constants of the motion and the invariance properties of the Lagrangian of electrodynamics will be analyzed in Complement By). 3. Potentials—Gauge Invariance Equations (A.1.b) and (A.1.c) suggest that the fields F and B can always be written in the form1A3 ‘The Fundamental Equations in Real Space 9 Bo.) = x MD) (A.10.8) Be.) = — Lawn — Vue) (A.10.b) where A is a vector field, called the vector potential, and U a scalar field called the scalar potential. A first advantage in introducing A and U is that the two Maxwell equations (A.1.b) and (A.1.c) are automatically satisfied, Other advantages will appear in the Lagrangian and Hamiltonian formu: lations of electrodynamics (see Chapter II) Substituting (A.10) in Maxwell's equations (A.1.a) and (A.1.4), one gets the equations of motion for A and U ( which are second-order partial differential equations and no longer first- order as in (A.1). Actually, since @7U/@t? does not appear in (A.11.2), this equation is not an equation of motion for U, but rather relates U to A781 at each instant. The state of the field is now fixed by giving A(r. 4) and AA(F, f9)/41 forall r. It follows from (A.10) that E and B are invariants under the following gauge transformation: 1 é@ Aut.) pi) AG) (Atha) ) awn = mn av- awn 5 o09 (A.1L.b) AG. 1) > Aln 1) = AG.) + VED) (4.12.2) UE) + U1) = UO Feo (A.12.b) where F(r, 1) is an arbitrary function of r and 1. There is then a certain redundaney in these potentials, since the same physical fields E and B can be written with many different potentials A and U. This redundancy ean be reduced by the choice of one gauge condition which fixes 9 + A (the value of y x A is already determined by (A.10.a)) The two most commonly used gauges are the Lorentz gauge and the Coulomb gauge. (i) The Lorentz gauge is defined by Vea) (A.13)10 Classical Bletrodynamies 1a3 ‘One can prove that it is always possible to choose in (A.12) a function F such that (A.13) willbe satisfied for A’ and U’. In the Lorentz gauge, the equations (A-11) take a more symmettie form: (A.14.a) (A.14.b) where 0 = 42/c? ar? ~ A is the d’Alembertian operator. This is due to the fact that the Maxwell's equations on one hand and the Lorentz condition on the other are relativistically invariant, that is, they keep the same form after a Lorentz transformation. Using covariant notation, Equations (A.13) and (A.14) can be written (as, with ¥ al evs and Deca = (4.16) with i= (end) where 4 and j* are the four-vectors associated with the potential and the current respectively. (ii) The Coulomb (or radiation) gauge is defined by vA =0 (A17) Equations (A.11) then become (A.18.a) (A.18.b) Equation (A.18.a) is Poisson’s equation for U. The covariance is lost, but other advantages of the Coulomb gauge will be seen in the subsequent chapters.1B. lectrodynamics in Reciprocal Space n B—ELECTRODYNAMICS IN RECIPROCAL SPACE 1. The Fourier Spatial Transformation—Notation, Let €(k, 1) be the Fourier spatial transform of E(r, 1). Then E and & are related through the following equations: sik) = 1 4, ener a 1 Bi) =a a (Qn) |e net (B.1.b) In Table I the notations used for the Fourier transforms of various other physical quantities are shown, Block letters are used for the quantities in real space, and script ones for the same quantities in reciprocal space. Tae Bi, 1) 76K) Bir, 1) + AK) Air Ak 1) Ulin WK) ptr. 1) ++ ak. 1) jt eft) Since E(r, 1) is real, it follows that Sk = &(-kD (B.2) In this treatment one frequently uses the Parseval-Plancherel identity jarrince=ferewrny a where ¥ and @ ate the Fourier transforms of F and G. as well as the fact that the Fourier transform of a product of two functions is proportional to the convolution product of the Fourier transforms of these two functions: 1 en Pr Flr) Gr = Pye F(R) G(R) ce)2 assical Blectrodynamics 1B2 Table II lists some Fourier transforms that are used throughout this book ‘Tabi I psoas Tar Ty ik Finally, to simplify the notation, we write é, in place of dr,(1)/dr, Bin place of E(t, 1)/8%, & in place of 3°6(k, 1)/0?,..., whenever there is ‘no chance of confusion. 2. The Field Equations in Reciprocal Space Since the gradient operator 7 in real space transforms into multiptica- tion by ik in reciprocal space, Maxwell's equations (A.1) in reciprocal space become (B.5.a) (B.5.b) (B.5.4) It is apparent in (B.5) that &(k) and (k) depend only on the values of &(), 40), p(k), and _f{k) at the same point k, Maxwell's equations, which are partial differential equations in real space, become strictly local in reciprocal space, which introduces a great simplification ‘The equation of continuity (A.3) is now written ik-j+p=0 (B.6) ‘The relationships between the fields and potentials become (B.7.a) (B.7.b)1B3 lectrodynamies in Re B the gauge transformation (A.12) (B.8.a) (B.8.b) (B9.a) (B.9.b) (ik 3. Longitudinal and Transverse Vector Fields By definition, a longitudinal vector field V,(r) is a vector field such that Vx Vy =o (B.10.a) which, in reciprocal space, becomes ik x 45k) =0 (B.10.b) ‘A transverse vector field V, (r) is characterized by i Vu) = (B.1L.a) ik + Wk) =0 (B.11.b) ‘Comparison of (B.10.a) and (B.10.b) or (B.11.a) and (B.11.b) shows that the name longitudinal or transverse has a clear geometrical significance in reciprocal space: for a longitudinal vector field, ¥\(k) is parallel to k for all k; for a transverse vector field, ¥, (K) is perpendicular to k for all k Itis important to note that a vector field is longitudinal for transverse] if and only if (B.10) [or (B.11)] are satisfied for all ror all k, For example, in the presence of a point charge at r,,V « Fis, according to (A.1.), 2er0 everywhere except al r,, where the particle is located. In the presence of a charge, E is therefore not a transverse field. This is even more evident in reciprocal space, since k- & is then proportional to €~"*", which is clearly nonvanishing everywhere. Working in reciprocal space allows also a very simple decomposition of all vector fields into longitudinal and transverse components: HUW) = Uk) + HUW) B12)4 Gsssical Electrodynamics 1LB3 At all points k, ¥,(k) is gotten by projection of ¥(k) onto the unit vector x in the direction k: w= kik (B.13) ‘One thus has ¥(K) = K[x 1°) (B.14.a) a te ) (B.14.b) Vy(r) and V, (r) are then gotten by a spatial Fourier transformation of (B14) Remarks @ In reciprocal space, the relationship which exists between a vector field (k) and its longitudinal or transverse components is a local relationship. For example, one can show from (B.14) that /, _ hk vito = © (5, 48) 500 (B15) where i, j = x, y, 2. Each component of ¥7, (k) at point k depends only on the ‘components of ¥(k) at the same point k. By Fourier transformatio', Equation (B.13) then becomes, using (B.4), van = (B16) where oye) am [ete (0, 8,80) + = 6,5) + BITa) 8, (r) is called the “transverse 8-function”. The presence of the last term in (B.17.2) shows that the relationship between V, (e) and V(r) is not local: V. (r) depends on the values V(r’) of ¥ at all other points r’. Note also that the calculation of the last term in (B.17.a) needs special caution at r= 0. The second derivative of 1/r must be calculated using the theory of distributions and contains a term proportional to 6,,8(¢). The calculation, presented in detail in Complement Ay, leads to Sided = 25, He) ane yt (B.17.b)LB Electrodynamics in Reciprocal Space 18 (Gi) The decomposition of a veetor field, arising from a four-vector or from an antisymmetric four-tensor, into longitudinal and transverse components is not ‘elativistically invariant, A vector field that appears transverse in a Lorentzian frame is not necessarily transverse in another Lorentzian frame. Gi) Even though the separation (B.12) introduces nonlocal effects in real space and is no longer relaivstically invariant, itis nonetheless interesting in that it ‘Simplifies the solution of Maxwell's equations. In effect, as will be seen in the following subsections, two of the four Maxwell equations establish only the longitudinal part of the electric and magnetic fields, whereas the other equa- tions give the rate of variation of the transverse fields. Such an approach then allows one to introduce a convenient set of normal variables for the transverse field 4. Longitudinal Electric and Magnetic Fields Return to Maxwell's equations. It is clear now that the first two ‘equations (B.5.a) and (B.5.b) give the longitudinal parts of & and @. The second equation clearly shows that the magnetic field is purely transverse: B= 0=8 (B.18) “The first equation (B.5.a) relates the longitudinal electri field (k) to the charge distribution p(k): 66 = - Loo (B.19) and &(k) appears then as the product of two functions of k whose Fourier transforms are tk) <> per) (B.20.a) (B.20.b) Using (B.4), one then has16 ‘Chassical Flectrodynamies LBs It thus appears that the longitudinal electric field at some time & is the Coulomb field associated with » and calculated as if the density of charge 6 were static and assumed to have its value taken at 1, ie. the instanta- ‘neous Coulomb field It is important to note that this result is independent of the choice of gauge, since it has been derived directly from Maxwell's equations for the field E and B without reference to the potentials. The fact that the longitudinal electric field instantly responds to a change in the distribution of charge does not imply the existence of perturbations traveling with a velocity greater than that of light. Actually, ‘only the total electric field has a physical meaning, and one can show that the transverse field FE, also has an instantaneous component which exactly cancels that of E,, with the result that the total field remains always a purely retarded field. This point will be discussed again later. Consider now the longitudinal parts of (B.5.c) and (B5.a). The two terms of (B.5.c) are transverse. The longitudinal part of (B.5.d) is written (B.22) Taking the scalar product of (B.22) with k, and using (B.19) and the fact that k-4)= k-f. one gets b+ik-j=0 which is just the expression of the conservation of charge (B.6) and thus conveys nothing new. Remarks (From equation (A.10}) or (B78) connecting the electric field to the potentials it follows that - A, (8.240) Ay we 24.0) E, 5, with the result that In the Coulomb gauge, one has A, AsO 4 B= ve (B.28.a) 11 follows that the longitudinal and transverse parts of E are associated. in the Coulomb gauge, with U and A respectively, Comparison of (B.25.a) and (B21) shows that, in the Coulomb gauge, U is nothing more than the Coulomb potential of the charge distribution: ol) if = vena [ay 25 ven aay ler (B.25.b)LBs lectrodynamics in Reciprocal Space a ‘The same result can be gotten directly from Equation (A.18.a). The solution of, this Poisson equation, which tends 10 2ero as |r| -* 20, is nothing more than (B25), (iy Tis clear from (B.8.a) that a gauge transformation does not change A , . 1 follows that she transverse vector potential A , is gauge invariant AL=AL (8.26) (Gi) Maxwell's equations are presented here in two sets: (A.L.a) and (A.1b) five the longitudinal fields, and (A.1.c) and (A.1.6) give the rate of variation of fhe transverse fields ($8.6), This grouping is different from the one used in relativity, where Equations (A.1.b) and (A.1.c) on one hand, and (A.t.a) and {A.14) on the other, are combined in two covariant equations OFap tO Fy tGyFn 20 weve DP (BATA) (B.27.b) where Fy = Ae Ay (B.28) is the electromagnetic field tensor, 4, the potential four-vector, and j, the current four-vector. 5, Contribution of the Longitudinal Electric Field to the Total Energy, to the Total Momentum, and to the Total Angular Momentum One now uses (B.19) for &(k) to evaluate the contribution of the longitudinal electric field to various important physical quantities, a) Tue TorAt. ENERGY “The Parseval—Plancherel identity (B.3) allows one to write Sones Om) = fore ‘E ne then replaces & by 6+ €, and uses €- 6, = 0. This yields (wre <8 faxjamr +4 feriaw?. B20 The first term in (B30) is the contribution Hyg, of the longitudinal electric field to the total energy given in (A.7) Hg = fora} =sferg@ @918 ‘Classical Blectrodynamics LBS while the second, when added to the magnetic energy, gives the contribu- tion Han, Of the fields E, and B: [eal .doP +o? jae) P) BA] (B.31.b) - fertete Inserting the expression (B.19) for & (k) in (B.31.a), one gets Mog [ee ory 2 (B.32) which can finally be written using (B.3) and (B.4) as Brey f [arav newirh os Hygag is nothing more than the Coulomb electrostatic energy of the system of charges. Finally, one calculates Hq, for a system of point charges. For this it is convenient to use the expression a. sk) = ean : a) = Sato os) for the Fourier transform of the charge distribution given in (A.5.a). Substituting (B.34) in (B.32), one gets 7 a (ak fons Hiroe = Veous = 3 Ia + sok lak trom et Secon! (B.35) ‘The first term of (B.35) can be written Deny where 2 ae Pk is the Coulomb self energy of the particle a (in fact, infinite, unless one introduces a cutoff in the integral on &). The second term is nothing more than the Coulomb interaction between pairs of particles (a, 8), so that finallyIBS ‘lectrodynamies in Reciprocal Space 19 In conclusion, one has seen in this subsection that the total energy (A.D) of the system can be written HY bm + Vous + Hoan (6.38) and appears as the sum of three energies: the kinetic energy of the particles (first erm), their Coulomb energy (second term), and the energy Bf the transverse field (third term). As in the preceding subsection, these results are independent of the choice of gauge b) Tue Tora, Momentum ‘One substitutes E, +E, for E in the second term of (A.8). The total ‘momentum of the field appears then as the sum of two contributions, Plog, and Prag: given bY (k) x Ak) Prag = & [87 E\() x BO) = co fers 4; , (B.39.a) Prane = &0 [or Ec) x Ber) = ey [UKEMK) x BUR) (B.39.b) Using (B.19) for &, the relationship (B.7.2) between @ and sf, and the identity ax(bxe as e)b (as be (B.40) ‘one can transform (B.39.a) into 4 far i#t & » (ik x Po ofa kT Be fk x a) = fax ptLol — K(x of) (B.41) ‘The factor in brackets in (B.41) is nothing more than the transverse component of .6/, with the result that Prog, takes the simpler form Prons [apr = fer pAL=YaAde) — (B2) where (A.5.a) has been used for p. As before, this result is independent of the choice of gauge, since A , is gauge invariant [see (B.26)]» ‘Classical Electrodynamics Las Finally, the total momentum P given in (A.8) can be written Ll B+ ay ALO] + Pros (B.43) and is the sum of the particle mechanical momenta mt, the longitudinal field momentum ¥q,A ,(f,). and the momentum of the transverse field, Equation (B.43) suggests that one introduce for each particle the quantity Pe = make + Ge Alt) (B.44) so that P can be written P= Sp. + Pras (B.45) In fact, one can show that in the Coulomb gauge, p, is the conjugate momentum to r, or the generalized momentum of the particle a (see § Chapter 11). One can see then that, in the Coulomb gauge, the difference between the conjugate momentum p, and the mechanical momentum ‘mgt, of the particle a is nothing more than the momentum associated with the longitudinal field of the particle a. Remark Using (B44), the total energy given in (B.38) can be written = She Dh — GAMEIE + Ves + Ham (8.46) One can show that 17 is nothing more than the Hamiltonian of the system in the Coulomb gauge (see §C3, Chapter ID, ©) Tur Tora ANGULAR MOMENTUM Calculations analogous to the foregoing (see also Complement By, §1) show that the total angular momentum J given in (A.9) can be written SHEE & Bt diane (B47) where p, is defined in (B44), and where Syne = % | Pre x [E(0) * Bir)] (B.48) is the angular momentum of the transverse field.1.B6 Blectrodynami in Reciprocal Space a 6. Equations of Motion for the Transverse Fields, ‘One now returns to the second pair of Maxwell's equations (B.5.c) and (BS.d), and one examines the transverse parts of these two equations, which can be written in the form B= -ikx = —ikx & (B.49.a) nick x a4), (B.49.b) The second pair of Maxwell's equations then appear as the dynamical ‘equations giving the rate of variation of the transverse fields # and 6, Tis important to note that the source term appearing in the equation of motion (B49) for &, is j,, and not 7 Since, in real space the relationship between j, and j is not local (see Remark i of §B:3 above), the rate of change of E.,(r, 1) at point r and time ¢ depends on the current je’, 1) at all other points r’ at the same time t. It follows that Ey includes, like E,, instantaneous contributions from the charge distribu- tion. It can be shown (see Exercise 3) that the instantaneous parts of B., and Ey compensate each other exactly, so that the total field E = Ey +E is a purely retarded field. To conclude this section itis useful to reconsider the definition (A.6) of the “state” of the global system field + particles at time f, Since the longitudinal field can, in fact, be expressed totally as a function ofr, [see (B.21)} the state of the system is completely fixed by giving {BU tg), BUR, fo), EAFo)s Elo) } (B.50) for all k and all a, We will see in the next section that it is possible to improve the choice of the dynamical variables characterizing the state of the field. Remark ere eee ae posh t sy he coulis amma 99) 2 bopdoa, Oe oe een af(A30) canbe wen Dee's oo and this becomes in real space(Chassical Electrodynamics LBs ‘This equation is analogous to (A.14) except that one now has A, and j, in place of A and j. Ione takes the longitudinal part of (B.9.b) and uses (B9.a) to eliminate @, once again one gets the conservation of charge (B.6). As with (B.S), the longitudinal part of (B.9.b) gives rise to nothing new. Finally, only (B9.a) remains, and it can be written Rawat rine sd, (B.53) (ince k - sf , = 0). This equation is not sufficient to fix the motion of , and H, This is not a surprising result, since there is a redundancy in the potentials. To find of, and 4, itis necessary to have an additional condition, that is, 0 define the gauge. If one chooses the Coulomb gauge, one makes «f, = 0, and (B53) then gives 4 [see also (A.18.)). If one chooses the Lorentz gauge, the supplementary condition (A.13) in reciprocal space is wicks, Bsa) ‘The pair of equations (B.53) and (B.S4) then forms a system of two first-order equations giving the evolution of f, and %. Other choices of gauge are equally possibleLea Normal Variables 2B (C—NORMAL VARIABLES 1, Introduction In ordinary space the rates of change, E(r) and Br), of the fields E and B at point r depend on the spatial derivatives of E and B and thus on the values of E and B in the neighborhood of r. Maxwell's equations (A.1) are partial differential equations. In going to reciprocal space, one has first of all eliminated 6 (k) which is not really a dynamical variable, since it can be expressed as a function of r,. One has then seen that the rates of change &, (k) and (k) depend ‘only on the values of & (k) and B(k) [and on that of (kat the same point k. Equations (B.49) give a system of two coupled differential equations for each point k Inspection of this linear system (B.49) suggests that one attempt to introduce two linear combinations of €, and @ which evolve indepen- denily of one another, at least for the free field where 4, = 0. 2, Definition of the Normal Variables To begin, one writes Equations (B.49) in the form C1.) C1) ‘One seeks the eigenfunctions for such a system in the case 7, = 0. One then finds from (C.1) that Fe F cx x B) = F ied, F x x B) (€.2) with, mck Kh. C3 ‘One is then led to define, even if 7, # 0, two new variables a(k, 1) and Buk, 2) atk) = Top lak 0 = MKD] C4.) Bik 1) = [ak ) + x x BK] (C40) mwm Classical Flectrodynamies 1c3 where ”(k) is a normalization coefficient which will be chosen later so as to have the simplest and clearest form for the total energy H. Before proceeding farther, it is important to note that @ and B are not in fact independent dynamical variables. The real character of E., and B. which gives rise to equations such as (B.2) for &, and , requires that Bik) = ~ a-ha) «cs Inverting the linear system (C.4) and using (C.5), one then gets Blk. 1) = iN) [atk 9 = (=k) (C.6.a) atk) = he x atk) + x a kD]. (C.6.0) Knowledge of a(k, f) for all the values of k is then equivalent to knowing ,(k,1) and @(k,1). In addition, the a(k, 1) are truly independent variables, since no conditions such as (B.2) exist for a(k, 1). One is able then, for determining the global state of the system, to replace (B.50) with 1 UK, 0). y(t) Fata) } cn Evolution of the Normal Variables From Maxwell's equations (C.1) and the definitions (C.4.a) for a, one gets 4(k, 1) + ioatk. 4 =r hh (C.8) One notes especially that since 6, and @ are related to a by (C6), Equation (C8) is strictly equivalent to Maxwell's equations. It is neverthe- less simpler than Maxwells equations. It resembles the equation of motion of the variable x + i(p/me) of a fictitious harmonic oscillator with cigenfrequency @, driven by a source term, due to the particles, propor- tional to A, (k, 1. When “7, = 0 (the case of the free field), the evolutions of the various normal variables a(k, 1) are completely decoupled. The solution of (C.8) is then a pure harmonic oscillation describing a normal vibrational mode of the free field. This is the reason why the a(k, ) are called “normal variables” I external sources are introduced, that is to say, sources independent of «, the variables a corresponding to different k continue to evolve indepen- dently of one another, each driven by (k, 1) (see. for example, Comple- ment By)1c3 Normal Variables, 8 Finally, if the sources are the particles interacting with the field, the motion of J, depends on a, with the result that the evolutions of the various Variables a(k, ’) are, in general, coupled through the action fof the current , (k,#). Ibis then necessary 10 add to (C:8) the equation of motion of 7, (& 1) [determined from the Newton-Lorentz equation (A.2) fand the definition of the current (A.5)] and to solve this coupled set of equations. To conclude this subsection some new notation is introduced. Since a is (ike &, and @), a transverse vector field, one can, for each value of k, expand a(k, 1) on two unit vectors ¢ and normal to one another and both located in the plane normal to « (Figure 1) crease aww al i cs are cK Sek =0 J ky Figure 1, The transverse polarization vectors e and One thus gets ak, 1) = 02(k. 1) + 6 a,(k. 1) = Deak 2) (C.10) where afk.) = e+ atk.) cy is the component of a along ¢. The set (a,(k, 1)) for all k and e forms a complete set of independent variables for the transverse field, The equa- tion of motion for a,(k, 1) is,% ‘Classical Electrodynamics 14 lk.) + oak, 1) ef (C12) Tr TW where one uses © f= 8-7. 4. The Expressions for the Physical Observables of the Transverse Field as a Function of the Normal Variables Later on one always uses the normal variables a,(k, #) (and the corre- sponding quantum operators) (o characterize the state of the transverse field. Thus it is important to have expressions for the various physical observables of the transverse field as a function of the «,. 4) THE ENERGY Hyung OF THE TRANSVERSE FIELD We substitute in (B.31.b) the expressions (C.6) for &, and @ as a function of a and a* {the more concise notation a? is used for a%(—k, 1) In addition, one respects the ordering between a and a as it arises in the calculation although a and a? are numbers which commute. The reason for doing this is that in quantum electrodynamics, a and a® will be replaced by noncommuting operators. The results obtained in this subsec- tion then remain valid in the quantum case. From (C.6), one finds 65-6, = Sat — a) (a at) Vat ada cat —atsat — a+) CB B= Nat 4a) (a + at) = Vata ta cat tot a_+a) Cc.) with the result that (B.31.b) becomes Hyams = % [Bk Wat sat a + ot] C14) Changing from k to —k in the integral of the second term allows one to replace a_-a* by a-@*, Let us now take for the normalization coeffi- cient (Kk) the value stk) = (c.15) V2% chosen so that in the quantum theory the commutation relations betweenLes Normal Variables 7 the operators corresponding to @, and a are simple. Equation (C.14) then takes the more suggestive form Hane foxy tare tak, 1) + 9,(k 1) 28k). (C.16) It then appears as the sum of the energies of a set of fictitious harmonic oscillators with an oscillator of frequency @ = ck being associated with each pair of vectors k, e (with © normal to k). Such a pair defines a “mode” of the transverse field. b) THE MOMENTUM Pag, AND THE ANGULAR MOMENTUM Jan, OF THE ‘TRANSVERSE FIELD A calculation similar to that above allows one to get from (B.39.b) P, [ nk ane jee YE [atk 9 alk. 1) + alk, 1) 22k, N]. (C17) For the angular momentum Jyaq, of the transverse field given in (B.48), the caleulations are a little more tedious than for Hyjay ANd Pays (86 §2.b of Complement B,). The following result is obtained aol Sm = FY. | dP Rad cas ky ee + at Ete he = ~ By bape hy 2, 3F — Fate AP] (C18) where a,b, c,d = x, ys 0F 2, 9, tensor. /Ak., and eyyq is the antisymn Remark The product E, XB appears in the expressions for Phaq, and Jean In quantum theory E, and B become operators and one can ask if it is not necessary to symmetrize E, XB in the form (E, xB - Bx E,)/2. In fact, E, and B are taken at the Same point in space, and we will soe in Chapter HI (§A.2) that E, (e) and Bie) are commuting observables, so that symmetrization is not required. ©) TRANSVERSE ELFCIRIC AND MAGNETIC FIELDS IN REAL SPACE ‘The expansions of E(t, 1) and B(e, fare gotten by taking the Fourier transforms of (C.6.a) and (C.6.b) [in the integral over k of the last terms of (C.6.a) and (C.6.5) one replaces k by —k}. This then becomes Een =i faky slack nee — ak nee] C19)30 ‘Qhassical Electrodynamics Les ilarities and Differences between the Normal Vi ‘Wave Function of a Spin-1 Pa 5 Consider first the free field. Equation (C.8) can be written ind(k, 1) = fova(k. 1) C31) and appears then as a Schrédinger equation relative to a “ vector wave function” a(k, 1), the corresponding Hamiltonian being diagonal in the reciprocal space with matrix elements /w8(k ~ k’). Equation (C.16) can also be interpreted as the mean value of such a Hamiltonian in the wave function a(k, 1). Likewise, since in quantum mechanics the momentum ‘operator of a particle is diagonal in reciprocal space with matrix elements ‘AKB(K ~ K’), Equation (C.17) can be interpreted as the mean value of the ‘momentum operator in the wave function a(k,f). Finally, one can show (see §2.c, Complement B,) that Equation (C.18), giving the angular ‘momentum of the transverse field, coincides with the mean value in the wave function a(k, t) of J = L + S (where L and § are the usual quantum operators for the orbital angular momentum and spin angular momentum), The first term in the bracket of (C.18) corresponds to L, the second to S. Al the preceding results suggest that one interpret a(k, f) as the wave function in reciprocal space of a particle of spin 1 (*), namely the photon Such an analogy should not, however, be pushed too far. First of all, one can show that the Fourier transform of a(k, f) can not be interpreted as the photon wave function in real space and, more generally, that it is impossible to construct a position operator for the photon. (**) Addition: ally, the equation of motion of « no longer has the form of a Schridinger equation in the presence of sources: it is not homogeneous. Such a result is not surprising. The Schrédinger equation preserves the norm of the wave function and thus the number of particles. Now it is well known that in the presence of sources, photons can be absorbed or emitted. Thus one cannot introduce a Schridinger equation for a single photon in the presence of sources. In fact, the electromagnetic field itself must be quantized, and photons then occur as elementary excitations of the quantized field. We will see in the following chapters that the “wave function”, or more properly the state vector, of the quantized field is a (9) The value of 1 forthe spin is tod to the vector character of a, See Akhiezer and Berestetski, Chapter 1 (07) See for example, E. Wigner and T. D. Newton, Rev. Mod, Phys. 21,400 (1949): M. HL. Pryee, Proc. Roy. Soc., ISA, 62 (1943)Lee Normal Variables 31 vector in a Fock space where the number of photons can vary from zero (vacuum state) to infinity. ‘The preceding analogy can however be of some use. It suggests, for example, that one examine the transverse eigenfunctions of J? and J. in reciprocal space. This leads to the multipolar expansion of the transverse field (see Complement B,), more convenient than the plane-wave expan- sion (given above) in all problems where the angular momentum is important. 6. Periodic Boundary Conditions. Simplified Notation It is common to consider the fields as being contained in a cube of edge L, and satisfying periodic boundary conditions at the sides of the cube. At the end of the calculation one lets L go to infinity. All the physical predictions (cross-sections, transition probabilities, etc.) should certainly be independent of L. The advantage of such a procedure is the replacement of the Fourier integrals by Fourier series. In other words, the integrals on k are replaced by discrete sums over k (C.32) 2am, where n,.,., are integers (positive, negative, or zero). The variables a;{k, 1) are replaced by the discrete variables ay,(1) ail, 1) + (0) (€.33.a) One can even use the more concise notation a, (€.33.b) where the index i designates the set (k,, ¢,). The correspondence between the two types of sum obeys the following rule: fouy mos (2) lk 6) (C4) In summary, the following are the expansions in a, and a of H, Phanss Ay E, and B: (C.35) 1, = SP a, + a at) (€.36) Lala eet + ah em] (C.37)2 Classical Electrodynamics Lee = iD Sula, go — at ee] (C38) B=iD Alan, x 50M — are x se) (C39) with (C40) In these expressions E, indicates summation on all the modes k,e,. It is convenient to note also that when going from the Fourier integral to the Fourier series, the factor (1/27)°/* of Equations (B.1) is replaced by 1/137, This explains why 4, contains L? in place of 2)” [compare (C.21) and (C40), Finally the evolution of a, is governed by & + iw, a = = ; c.4) V2 hn, with Bre Bere,- ie) (C.42) Remark ‘The discrete variable a, has not the same dimensions as the continuous variable 4,{h). More precisely,LDa ‘Various Possible Quantization Schemes 3B D—CONCLUSION: DISCUSSION OF VARIOUS POSSIBLE QUANTIZATION SCHEMES After this rapid survey of classical electrodynamics, we now face the problem of quantization of the theory. Here we will review various possible strategies for quantization which will clarify the motivation and the organization of the following chapters. 1, Elementary Approach The formulation given in this chapter lends itself particularly well to an elementary approach, Indeed, we have shown that the global system (clectromagnetic field + particles) is formally equivalent (0 a set of mutu- ally interacting particles and oscillators. The simplest idea which can then be put forth for quantizing such a system is to quantize the particles and the oscillators in the usual way. With the position r, and with the momentum p,of the particle a we associate operators (*) whose commuta- tor is sh, and we replace the normal variables @, and a of the oscillator i by the well-known annihilation and creation operators @, and a,” with commutator equal {0 1 taken from the quantum theory of the harmonic oscillator: aa, af ay oy Al the physical quantities, which can be expressed as functions of tys Pas «@,, and a, become operators acting in the space of the quantum states of the global system Such an approach is, however, heuristic. Since it does not come from a Lagrangian or Hamiltonian formulation, we do not know ifr, and p, on cone hand or a, and a? on the other (more precisely, 4, = (a, + a,7)/ v2 and p,=i(a? —a,)/ 2] can be thought of as conjugate dynamical variables with respect to a Hamiltonian which has yet to be written Certainly, in this chapter the expression for the total energy of the system hhas been given, but the conditions under which this expression can be considered as the Hamiltonian of the system have not been established It is nevertheless possible to avoid this difficulty. One postulaces the following expression for the Hamiltonian in the Coulomb gauge: a MOP + Veou +E Pay a, + aa?) (D2) (*) To simplify the notation the same symbols are retained to designate the classical variables ofthe particles, and p, and the associated operatorsu ‘Classical Electrodynamics D2 which is nothing more than Equation (B.46) for the total energy (in the Coulomb gauge A= A.) a, and a? being replaced in Equation (C.35) for Han. bY the operators a, and a,’. One also postulates the following commutation relation for r, and p, on one hand: trast) = (ae Pal = 0 Tran] = ity 6 3) where i, j=, 942 (the 8p indicates that the variables of two different particles commute), and for a, and a on the other: { (a, a] = [a?,af] = 0 =, (0.4) The 6,, indicates that the variables of two different modes of the trans verse field commute. One can then show (as will be done in §B.2, Chapter I11) that the Heisenberg equations derived from the Hamiltonian (D.2) ‘and the commutation relations (D.3) and (D.4) lead to good equations of motion, that is to say, to the Maxwell-Lorentz equations between opera- tors The reader ready to accept the foregoing points and wishing to get as quickly and simply as possible to the quantum theory can skip Chapter TI, devoted to the Lagrangian and Hamiltonian formulation of electrodynam- ies, and go directly to Chapter III, which starts from the expressions given in this subsection. 2. Lagrangian and Hamiltonian Approach This approach consists in showing initially that the basic equations of classical electrodynamics, the Maxwell-Lorentz equations, can be thought of as Lagrange’s equations derived variationally from a certain La- grangian, Canonical quantization of the system is then achieved by associating with each pair formed by a “generalized coordinate” and its “canonically conjugate” momentum two operators with commutator if ‘Although more abstract, such an approach offers a number of advan- tages. It allows one to identify which field variables are conjugate (for ‘example, in the Coulomb gauge the vector potential and the transverse electric field) and to obtain the Hamiltonian directly without it being necessary to postulate it. This approach also allows a deeper understand- ing of the problems tied to the choice of gauge. The Coulomb gauge appears then as the most “economical” gauge, allowing one to eliminate ‘most easily the redundant variables in the Lagrangian. Finally, it is well known that two Lagrangians differing only by a total derivative are physically equivalent. It is thus possible to construct many equivalentLD2 ‘Various Possible Quantization Schemes. 38 formulations of quantum electrodynamics and to discuss directly the relations which exist between t Chapter II presents classical electrodynamics and its canonical quanti- zation starting from such a point of view. Equations (D2), (D.3), and (D4) are therein justified in a rigorous fashion. Changes of Lagrangian and Hamiltonian will be treated in Chapter IV along with the various formulations of electrodynamics to which they give rise For certain problems it is important to use a manifestly covariant formulation, This leads one to choose a different gauge from the Coulomb ‘one and complicates the problem of quantization. These questions will be considered in Chapter V. (GENERAL REFERENCES AND ADDITIONAL READINGS Jackson, Feynman et al (Volume 11), Landau and Lifschitz (Volume 11), Messiah (Chapter XXI, §III), Akhiezer and Berestetskii (Chapter 1) Cohen-Tannoudji (§1).36 ‘Chassical Electrodynamics Aul COMPLEMENT A, ‘THE “TRANSVERSE” DELTA FUNCTION The transverse delta function 8(p) allows one to extract from a vector field its transverse component. It is easy to understand why it plays aan important role in electrodynamics in the Coulomb gauge, since the transverse fields are favored in this gauge. The purpose of this comple- ‘ment is to establish the expressions and the properties of this function and to illustrate its use in a simple example. The expression for the transverse delta function is particularly simple in reciprocal space. On the other hand, we will see that calculating its Fourier transform to find 8;:(p) presents certain difficulties which justify the detailed treatment given here 2) CARTESIAN COORDINATES, TRANSVERSE AND LONGITUDINAL ComPoNENTs Two different types of basis vectors will be used in reciprocal space to define vector fields: the Cartesian system (e,) (i= x, y,2), and the system composed of the longitudinal unit vector x = k/k and the two unit transverse vectors ¢ and e’, introduced in (C.9). The Cartesian compo- nents of vector ¥ are denoted by ¥;, One will often have to perform summations on the two transverse polarizations of products of compo- nents of e and e’. Their expressions are as follows, Consider first Sageagtag = [le (ese) + (ee) &) +(e W)(K-e)] — = le x)O6>e) (We +8) Oj — Ky a) Another summation is a x 8) = 0,654 of 6) @ where in writing the right-hand side, we have noted that Kxe=e Kxeowe 8Ad ‘The “Transverse” Delta Function a7 It appears then that (2) is nothing more than the component of the vector product e Xe’ on ¢, X e,. This then becomes E ele x 0), = Deh @ where ¢,, is the antisymmetric tensor. Finally, using (3) and (1), one immediately gets Fe x B(x 8), = 3, Hx, re) 1b) PROJECTION ON THE SUBSPACE OF TRANSVERSE FIELDS The transverse delta function is closely tied to the projection operator on the subspace of transverse veetor fields. In order to see this, consider a vector field V(r) and its Fourier transform ¥(k). In reciprocal spa ¥ .(k) is easily gotten from ¥(k) by projecting ¥(k) onto the plane normal to k at point k: Vik) = e(e » ¥(k)) 6) By projecting on e, and using (1), we obtain 0 Fl) = LO, — 6) V0 Let us denote by + the projection operator acting in the space of vector fields and generating the correspondence between ¥* and ¥7, Vi = atl co) This relation, written between the Cartesian components in reciprocal space, becomes yak) Jo Dagth, k) Hk) 0) where 4;,(K,K) is the matrix element of the operator 4+ in the basis {Ik,e,)). ‘Comparison with (7) shows that this matrix element is equal t0 (10) In seal space, the same relation (8) is written Vuln = [a'r Saye ry vie) ay38 (Chassical Electrodynamies Awd ‘The matrix elements of the operator Ain the basis {|r,e,)} are given by ce a oe Spit MRR) So eK Aine) = foxf {ak ~ le It appears then that the transverse delts- function introduced in (B.17), (12) 3) is tied to the matrix element of A+ in the basis {[r,e,)) by Ait.) = dif — F) to) Remark One can likewise introduce the projector A! on the subspace of longitudinal fields which is the complement to t of 4+ aw a us) 2. The Expression for the Transverse Delta Function in Real Space From the definition (13), it appears that 6, (p) is the Fourier trans- form of a function which does not tend to zero ‘when |k| tends to infinity ‘The transverse delta function then has a singularity at » = 0 which one must carefully characterize. To this end, one regularizes this singularity by truncating the spatial frequencies greater than some bound k,y. One later allows k, to go to infinity, Physically, such a procedure means that one is not interested in variations of the field over infinitesimally short distaness, but rather in the mean field over small but finite regions of space. a) REGULARIZATION oF 5,5 (p) Mathematically, one achieves the eularization by multiplying 8,,~ (kk /2) by ki/2 + kiy), which has magnitude 1 for k< ky, and decreases as 1/K* at infinity (16)Ay ‘The “Transverse” Delta Function 9 ‘This regularized function can be written, taking account of the properties of the Fourier transformation, as 330) ay were ap) (18) b) CALCULATION OF g(6) One frst performs the angular integral on k dk i ok 90) | oms?* | Re |, Gm ip ay ah gy ee Onip + This last integral is easily evaluated by the method of residues Kin gp) Gan? im Res(k = iky) + im Res(k = 0)) eke 4 Ixp Ligy@iby * TR So that finally 1 seo) Cee 20) Outside the neighborhood of the origin (p > 1/ky) 9(p) is equal to 1/4mp, which is indeed the Fourier transform of 1/k?, But as p — 0, 8(@) remains finite and tends to k y/4m. ©) EVALUATION OF THE DERIVATIVES OF g(0) Equation (17) gives the transverse delta function as a function of the second derivative of the function g(p). Since this latter is only a function of the modulus of p, one uses to evaluate its derivatives na) la)40 ‘Classical Electrodynamics Aya oy (1b) which yields by pe) 4 ale) = Du gp) 4 Bits 9 (ate) Tae) = Gate) + A pe o) ‘Then Sap) = 9" +2 ard T Opy pee Pe y Substituting in (17) this gives the expression for 8,:(p) sup) <2! yo) = 2) ? Evaluation of g” and g'/o gives rip) = [2-2-1 + kyo 23a) f= [2 2 (0 ky p+ (23a) 4 yp) =a - eM + ky Al (23) po aap me * d) DISCUSSION OF THE EXPRESSION FOR 6,5 (0) Equations (21), (22), and (23) lead to the following expression for the regularized transverse delta function (17) 1 P3e) | 3440) = (0) + peg PEPE — 3, 04 0) = 10) ail 7 10) 4) where He (0.85 i, e ace >: ufo) = 5G (Pett + oy Je 3) np) = 1 -(' + hyp + hh pe tue 26) The function +, (p) is localized about the origin. At the limit ky > 20 it tends to a point-source distribution centered on 9 = 0. Such a distribution can a prioti be written as a sum of a function &(p) and its derivatives of order 1,2,,.., Which we are going to evaluateAy ‘The “Transverse” Delta Function a ‘Note first that, simply as a result of homogeneity, the integral fe 74f0) = fy en is a number independent of ky. In contrast, the integral of the same function multiplied by a term of degree m in p, is proportional to (1/k,)" and goes to zero when k,, — 20. Thus, in the integral of the product of +,,(p) with a function ¥() expandable about p = 0, only the first term of the series gives a nonzero contribution in the limit k y —> 20 {2p 70) Yeo) = 1, AO) (28) It follows that 2 34) = 1, 610) 09) where he {arr 0) Se eo 0 ‘The angular average of p,p,/p" is 6,,/3, and the radial integral gives, a1) 2 where it is understood that the function 6(p) has an extent of 1/k,y The function m(p) is a regularizing function which becomes 1 for > I/ky and which starts as kjyp'/6 at the origin, with the result that the second term in (24) does not diverge at p = 0. It behaves like a dipole field regularized at the origin. Such a function has properties in three- dimensional space analogous to those of the principal-part function #(1/x). Actually, on integrating over a small volume centered at the origin, the second term of (24) gives zero, although it has in this region a value of the order of k3,. This property arises from the vanishing of the angular average of (3 p, in the same way that the integral of 1 (1/x) is zero asa result of the odd parity of this function2 ‘Classical Electrodynamics AS Finally, the transverse delta function can be written 2 19) (3 0:0; ) 35,5 Ps _ 5, 3) 54, io) + PO ra (3) 50) where (p) is equal to 1 away from the origin and suppresses the divergence at p = 0. Remarks (@) The factor } in (32) can be simply found, Taking the trace on # of (13) gives immediately Ls) = 2 5p) 4) In the expression (33) for 8,1 only the first term contributes to the trace and the factor + of (32) is necessary to satisty (34, (iiy One can ask if the second term of (24) does not give rise. in the limit ky — 2, 10 derivatives of the delta function 8(9). In fact, the dimensional ‘argument already developed for y,(p) applies: for functions of degree m in o, the contribution of the neighborhood ofthe origin tothe integral ofthe product ‘of these functions with the second term of (24) is ofthe order 1/43) and tends to 2e10 when ky > & 3. Application to the Evaluation of the Magnetic Field Created by a Magnetization Distribution, Contact Interaction ‘The magnetic field Bi) created by a magnetization density M(r) is given by Maxwells equation Vx Ber) ir) @s) where Jt) = x May 36) is the current associated with M(e). Substituting (36) in (35) and trans- forming into reciprocal space, this becomes ik x Blk) =— ik x Mk) en This equation allows one to find #(k). Projecting both sides of (37) on k and using the transverse nature of @(k - @ = 0), one gets, 3 LMtk) = Kix MW] = 5M) 8) Atk) =As ‘The “Transverse” Delta Function “ which shows that # is (except for a factor 1/egc?) the transverse compo- nent of . In real space, Equation (38) is written 1 Maa = 5 fer oye- omer «9 Be) Using (33) for the transverse delta function appearing in (39) shows that Br) is a sum of two contributions, The first one, coming from the first term in (33), is simply proportional to the density M(t) taken at point r. The second one, coming from the second term of (33), represents physically the dipolar field created at r by the magnetization density M(t’) at all other points #'. The presence of the regularization function 7 in the second term of (33) makes all the expressions finite, and symmetry arguments then allow one to show that the immediate neighborhood of r does not contribute to the integral on r’ of the product of the second term. of (33) with M,(P) (see §2.d above. Consider now another magnetization density M’(e). The interaction energy of M‘(r) with the field B(r) created by M(r) is W = — Jar Mie): Bee) (40) Equation (39) allows this to be written in a more symmetric form, We zee x far Jor min sye— ey mgr). 4) Using (33) again allows one to separate two contributions in W. The first, [ermo Mn) (42) 3 ee which depends on the magnetization densities M and M’ at the same point r, is called for that reason the contact interaction. The second, 5 far fare Mec ¥ for for eh * x ano [ES Pyle 0), (ery W, = Mir) (43) represents the magnetic dipole-dipole interaction between the two densi- ties. As above, the regularization introduced by and symmetry argu- ‘ments show that the immediate neighborhood of |r — Fr} = 0 does not contribute to W.Classical Electrodynamics AS Remark The foregoing can be applied to the study of the magnetic interaction between the niiclear spin and the electron spin in an atom. One takes as M(e) the magnetization density of the nucleus. This density can be appreciable only in a volume of the order of x}, where 7, characterizes the dimensions of the nucleus, ‘which is taken at the coordinate origin, The integral of Mi) is just the magnetic moment of the nucleus, = forma a One assumes in addition that M/() represents the spin magnetization density of the electron in the state (2), so that Mn n. [veo (45) where 1, is the spin magnetic moment of the electron, Important simplifica: tions appear in the interaction energy Was a result of the different spatial extensions of Mr) and M'(). Indeed, the spatial extent of [Y()|° is of the order of the Bohr radius ay, which is much larger than the nuclear radius. If cone ignores the variation of |Y()|? inside the nucleus, the contact interaction (42) becomes Fa }tno mn 45) that is, taking into account (44) and (45), Wes = ora veo P «a Tn the same way, one can make a multipole expansion of M(r) in (43) and only keep the lowestorder term, which yields Moy 1) (48) Equation (43) then becomes sy | wer? 9) Finally, regrouping (47) and (49) shows that the magnetic interaction energy between the two spins appears as the mean value in the state y(e) of the interaction Hamiltonian Ym 849 Hy, Lee a ~ aba E ene My (30) where ris the position ofthe electron with respect to the nucleus and where itis further understood that the dipole-dipole interaction is regularized at = 0,B.A [Angular Momentum of the Electromagnetic Field 45 COMPLEMENT By ANGULAR MOMENTUM OF THE ELECTROMAGNETIC FIELD, MULTIPOLE WAVES ‘The first motivation for this complement is to establish some results given without proof in Chapter and related to the angular momentum of the electromagnetic field. We first show (1) that the contribution of the longitudinal electric field to the angular momentum of the field can be reexpressed as a function of the particle coordinates r, and the transverse ‘ector potential A ,(r,) and regrouped with the angular momentum of the particles. We also establish (§2) the expression for the angular momentum Of the transverse field as a function of the normal variables a(k). The expression gotten is closely analogous to that giving the mean value for the total angular momentum of a spin-l particle whose vector wave function is precisely a(k) in reciprocal space. The foregoing analogy suggests then that one look for functions a(k) suitable for the angular momentum of the transverse field. More precisely one tries to determine transverse vector functions of k, defined on a sphere of radius ky, which are also eigenfunctions of J? and J,, where J is the fotal angular’ momentum of a spin-1 particle. Instead of coupling in the usual fashion the orbital angular momentum L to the spin angular momentum S of such a particle, we will see in §3 a simpler method for constructing the eigenfunctions of J and J, which give the longitudinal or transverse eigenfunctions directly. When the eigenfunctions thus found are substituted for the normal variables in the expansion of the electric and magnetic fields, one then gets, in real space, electromagnetic waves corresponding to photons with well-defined energy, angular momentum, and parity (§4). The second motivation behind this complement is to give a simple derivation of such multipolar waves, which are well suited to all the problems of atomic or nuclear physics where exchanges of angular momentum between matter and radiation play an important role. 1. Contribution of the Longitudinal Electric Angular Momentum jeld to the Total Let Pre x (Ey x B) w46 Gassical Electrodynamies BA be the contribution of the longitudinal electric field to the total angular momentum of the system field + particles. Replacing B by ¥ x A, (since 7 x Ay = 0) and using the expression for the double vector prod uot to transform E, x (y A.) yields Shes = Fo fof ZF WA, — «Ey a} @ The last term in (2) can be rewritten by moving r to the right of ¥: oy |a’r[~ (Ey 9) x A +E, x A 3) Integrate the first term of (3) by parts. The integrated term gives a surface integral at infinity which vanishes if the fields go to zero sufficiently quickly. In the remaining (erm the quantity y - Ey, appears, and this is /¢ from Maxwells equation (A.l.a). Regrouping the expression so obtained for (3) and the first term of (2), and making use of the fact that E, = —7U, where U is the Coulomb potential, one gets finally Sioog = ‘One can see now that the last two terms in (4) cancel, Integrating them by parts, one gets, be forfy UV At x V)A,, + UW x a} “) Now TEV x V) AL = Ulex PVA Ux AD. 6) The first term of (6) vanishes, since T - A , = 0. The second term of (6) cancels with the last term of (5). Only the first term of (4) remains, which, using (A.5.a) for p, gives aon = Lda te % Ault) a Its clear then that Jog can be written as a function of the coordinates r, of the particles and of the transverse potential A, . This result is gauge independent, since A , is gauge invariant {sec (B.26)Bz Angular Momentum of the Electromagnetic Field a7 ‘One now groups Jiog, with the angular momentum of the particles, Lt X Mate This gives Joos + DF * mt, = Dey * [dy A) + mB) =Dax 6 where p, is defined by (B.44). In the Coulomb gauge (A = A.,). py is the momentum conjugate with r,, or equivalently the canonical momentum of particle «. It appears then that in the Coulomb gauge the difference between the quantities r, x p, and r, X m,é, is just the angular momen- tum associated with the longitudinal electric field of particle a. 2. Angular Momentum of the Transverse Field In this section, we transform the expression for the angular momentum, of the transverse field, [arre x (8, x B) ° Sevan ) Juans IN RECIPROCAL SPACE ‘The calculations at the beginning of §1 above remain valid when one replaces E, by E, throughout. Since y - E , = 0, the integration by parts Cf the term corresponding to the first term of (3) now gives a zero result, ‘and only the terms corresponding to the first term of (2) and the last term of (3) remain: tea fPfdauema reat 4 ‘One now expresses Jaq, a8 a function of the Fourier spatial transforms 6, and of, of E, and A, For this, one wses the Parseval-Plancherel identity and the following table: ‘Tas roi Vvoik8 Ghassieal Electrodynamics B,2 giving the correspondence between the operators multiplication by and gradient with respect to r in real space on one hand and the operators multiplication by k and gradient with respect to k (denoted by V to distinguish it from V ) in reciprocal space on the other. One then gets eae ve fae] St xP), + 8t x a} ay ol 5) Jesu 18 TERNS OF NORMAL VARIABLES ee normal variables a(k): 6, =i Maat) 2» 4 oda +at) (yy Recall that #” is a normalization coefficient given by (C.15) and that a is an abbreviated notation for a(~, 1). Substituting (12) and (13) in (11) and changing k to ~k in certain terms, one gets oe fou fy ac in xP) a, iat x a Dade ik xP) at ie x a — [same terms where a = ai} (a) ‘The contribution of the second line of (14) to the integral is zero. Firstly, the term @_ a is odd in k and its integral vanishes. Then, changing k to =k in the first term of the second line of (14) and integrating it by parts, one sees that the contribution of this term is equal to its opposite and thus vanishes. One has then Jem = 4 fe ‘k { an expression equivalent to Equation (C.18). In all the calculations we have done, the ordering between a and a* has always been respected and is as it appears in the equations, If one neglects Yat tk xr) a, — iat x a -toait as)Bz Angular Momentum of the Electromagnetic Field ” to take this ordering into account, Equation (15) simplifies and becomes Sean = fi jody at(— ik x Pa, — iat x a (16) ¢) ANALOGY WITH THE MEAN VALUE OF THE TOTAL ANGULAR MOMENTUM OF & SPIN-L PARTICLE We will now sce that it is possible to reinterpret Equation (16) fof Iya ‘To do this, set aside for the moment the problem of the angular momen- tum of the electromagnetic field, and consider the quantum mechanics of a spin-1 particle. There are three possible spin states for such a particle, and the wave function representing the state |) of the particle will be a vector wave function with three components. In reciprocal space these components will be called = dak) ay We have taken a spin-state basis (|a)} which is not the set of eigenstates {1 1).10),| + 1)) of S., but the set of Cartesian spin states {lx)sv). [2)) felated t0 it =1y-141>) (14th Iz> =10> (18) ‘One now evaluates the mean value in the state |) of the total angular ‘momentum, J=L+8 a9 for such a particle, where L and § are the orbital and spin angular momenta. Using Table I, L is given in reciprocal space by the operator L=i Px fk = —ifk x (20) Since L does not act on the spin quantum numbers, COIL 16> = A [BHT O20 (~ ik S.1b> which derives from the known action of §. and S.=S,+ iS, on the cigenstates of S, and from equations (18).°Since $ does not act on the quantum numbers k, one gets then from (22) 1) ithe = glosn (ar 19) o and the asymptotic forms of the fields are Ror b= 6 {Lon (tor —1F) nin + FD a (igr 38) yaar ca rare meter Bur.) = s6{t sinftyr 9 5) X,ulp) (i)! +! eM" + ce } The radial part of E decreases as 1/r°, so that at infinity only the wave decreasing as 1/r remains, with the structure of a stationary plane wave, transverse in F-space.« Cassical Electrodynamics Ba ©) MAGNETIC MULTIPOLE Waves. The waves associated with $,,,,/y(K) are called magnetic multipole waves, The corresponding free fields E, and B are given by Euronext = 12 0)? 6, Nuanlt eM + cc im? Fe Nault + ce. (69) Brunt 6 In comparing (69) and (63) one discovers that one has simply inverted E and cB and changed the sign of one of the two fields. All of the conclusions of the previous section can be carried over without any dificulty GENERAL REFERENCES AND FURTHER READING Akhiezer and Berestetskii (Chapter I), Berestetskii Lifshitz, and Pitayevski (Chapter 1), Jackson (Chapter 16), Blatt and Weisskopf (Appendix B).Exercises 6 COMPLEMENT C, EXERCISES, Exercise 1, H and P as constants of the motion, Exercise 2. Transformation from the Coulomb gauge to the Lorentz, gauge. Exercise 3. Cancellation of the longitudinal electric field by the in- stantaneous transverse field Exercise 4, Normal variables and retarded potentials. Exercise 5. Field created by a charged particle at its own position. Radiation reaction. Exercise 6, Field produced by an oscillating electric dipole Exercise 7, Cross-section for scattering of radiation by @ classical elastically bound electron. 1, HAND P AS CONSTANTS OF THE MOTION a) Show that the energy of the system particles + electromagnetic field given by u-Syme + Bleue +eBy Ww is a constant of the motion. b) Derive the same result using the expansion of the transverse field in normal variables. ¢) Show that the total momentum P=Ymy to | Pre xB @ is also a constant of the motion. Solution 44) One calelates on Ho Sma Set fofe ° an bites for dy/ ing he Loans equton and for 38/91 and 98/2" wing Moco equation. at ges, 7 : Hoy tactien + 6 far since ny 04 XB) e-(evep-ti)-e8 w~o] o °a Exercises Ca Substitute (A.Sb) for j into fly Ej. One ges [ori Dee 9 tnd ° Now evalua fone ecw m= [erve xe © ‘which can be transformed into a surface integral whichis zero when the flds are zero at infinity. Using (4. (5). and (6). one gets . an eo. o ') tn ters of noemal variables, the energy is given by a fongeege x font as aatl “ Calculate diH/at, Here dx,/de is given by the Lorentz equation, As a result of charge ‘conservation j is equal 1o “ik -j, Equation (C8) gives 4. One then gets aH a Here Peper 719 Wh frp") fies +f o = he [Sh mtesiya tener o Now, using (B19, we write 4 = ~ipk/rok?. The second term of (9) can then be writen ‘To evaluate the last term of (9) note that J %(—K) = fk). Since A”(—K) = AUK). one can write the last teem a5 Fo Phy mttae = at wpe o = fowar sys fave ay ‘One can see that (10) and (12) cance the fist term of (9) ) Calculate €P/r and use the Maxwell-Lorentz equations to reexpress dd), 9E/It, and 08/2: ar w we o ay Lae Ble + Sas se Biro) vujif[er efi) <2 ewe) uo2 Exercises 8 [Now Equation (A.5.) gives = fein Be Saye mean as which cancels the second term of (Ld), In addition, fora vector field X. xv x) over = 6X Ban) Ee 8 8) + NX) as “The integral over all space of the fst two terms of the righthand side of (16) can be rsformed into a surface integral which is 20 ifthe fel decreases sufciently rapidly at infinity. As forthe last ter, i s 2070 for X ~ B and is equal to Bp/y for N= E, Using (Asa Se ee wo Which cancels the fist term of (14), Finally, Equations (14), (15) and (16) give dP 2, TRANSFORMATION FROM THE COULOMB GAUGE TO THE LORENTZ GAUGE Assume the potentials of and @ are known in the Coulomb gauge. What equation must be satisfied by the gauge function # which trans- forms .of and % into the potentials in the Lorentz gauge? Solution Let of and «be the new potentials The Lorentz condition (A.13) in reciprocal space is # 21" aid "can be wsitten a8 Functions of sf and using (B.). Combining (1) and (B. fone ges ike o keer keys eh FH e ‘The transversality of 7 in the Coulomb gauge gives then Agaee-t % so that, using the expression for 4 in the Coulomb gauge. 1 eae er aprglkme st arte Se “ eis eear that depends om the vost. tis for hs reason that tansormation from the Coulomb gauge tothe Lorn age not equivalent a ange inthe Lagrangian (8 Ream of 983. Chapter ‘Note nally that Equation (3) can be weten or-§ ° in real space. G“a Exercises G3 3. CANCELLATION OF THE LONGITUDINAL ELECTRIC FIELD BY T INSTANTANEOUS TRANSVERSE FIELD A particle with charge g, is located at the origin 0 of the coordinates. In the interval 0 to T the particle is displaced from @ to r,(T) along a path ,(1) (017). Let r be a point distant from the origin (> |r,(1)|, eT). The purpose of this exercise is to prove, starting with Maxwell's equations, that the instantaneous variations of the longitudinal electric field created by the charge q, at r are exactly compensated by the instantaneous component of the transverse electric field produced by the displacement of the particle a) Calculate, as a function of +,(z), the longitudinal electric field E,(r, 1) at point r and time ¢ from charge q,. Show that Ey(r, ¢) can be written E\(r. 0) = E(t, 0) + 0B, where BE, is given by @ power series in |r,(x)|/r. Show that the lowest-order term of this expansion can be expressed as a function of qata(t) and of the transverse delta function 8; (1). b) Find the current j(r, ) associated with the motion of the particle Express the transverse current j, (F,1) at the point of observation F as a function of q,f,(c) and the transverse delta function 6,+ (r ~ r,(1)). Show that to the lowest order in [r,(z)|/r, one can replace 8,5 (r ~ r,(1)) by 8, (2), Write the Maxwell equation giving JE ,(r,1)/8¢ as a function of i. (0,1) and B(r, 1). Begin by ignoring the contribution of B to aE, /at Integrate the equation between 0 and f. Show that the transverse electric field E,,(r, 1) produced by j, (r,t) compensates exactly (to the lowest order in |r,(1)|/r) the field 3E,\(r, 1) found in part a), ©) By eliminating the transverse electric field between the Maxwell equations for the transverse fields, find the equation of motion of the magnetic field B. Show that the source term in this equation can be written in a form which only involves the total current j. Justify the approximation made above of neglecting the contribution of Bo aE , /d¢ ‘over short periods. 14) From 88.4, Ef, the Coulomb field created by the charge gy at time Eun = — Yum) o with a Tae, [ea wense3 Exercises 6 Expand U in powers of [r,(91/r | ° ‘Substituting (3) in (1), ome gets forthe sth component of By(e,1)(F= x 8.2) F.(6.) = Fue) + Ele ” where rae (0, ~ 254) ” umpaion of () ih (17 ges king as oe fa also 600 = A Synnd 1) Following (AS), ie = a8 He ~ 40) o (8.16) then gives ule = Z fae oye — 440.9 = Yoile - 540 a.F.60 ® Since |r ~ #,(2)| is nonzero, 8+ (¢~ rf) varies as |p ~ F,(0]2. which in the lowest fonder in |r f)/P goes as 7", One can write then fo this order Ase = L940 abel ° “The Maxwell equation (B49) in real space is writen weve men tito 8% For ¢< 0, the vectors EB, and j, are zero. If one ignores the contribution of B. the Jmegra of (10) between O-and # gives Eni Lf ar jane ny a), arden f ‘ich taing (9) no acon besos tex aye f paneer 1S stmarco a Comparison of (6) and (12) shows that (1) at each instant cancels 84, (6.1) ) The cimination of €, between the equations (B.49) gives Aas eae Sik -gaen) a6 Exercises Ca since j= In espace, (1) besos (Sa -4)ae0 Since the source ter, which ines ocled at r= 0 the solution af 1) is purely Tetared and has no isantancous tah, D1 Senta oom the ime nena {0,77}, and its permissible to ignore the fst term of the right-hand side of (10) over this interval Lv hen ag 4, NORMAL VARIABLES AND RETARDED POTENTIALS, 4) Consider a set of charged particles producing a current j(r, 1). Integrate the equation of motion of the normal variables of the field, and write the normal variables at time ¢ in the form of an integral of the current between —so and 1. Derive the value of the vector potential of , (k. 1) in reciprocal space. b) Calculate, using the results of a), the potentials in Coulomb gauge in real space. ) By again starting with the results of a) in reciprocal space, find the total electric field &(k, £). Show that the electric field in real space is of the form hae fer rari) Bn) o Ir=rl and that it has no instantaneous Coulomb term. Solution 42) The solution of Equation (C8) giving the motion of atk 1) is ane [armen ep hike ° where @(r) = 1 for r> 0, (7) =O for + <0, and where W(K) = O72 (C29), one gets Starting with Hiki= sic [ diafem fea etki o1 so that, using the fac that je. is seal antkn=4 » owC4 Exercises or 'b) The transverse potential is gen in teal space by that is rom (, Pe [acetras Auea Melee 9e. ‘Now, one can show that 1 fore sine (-¥ (One has just wo perform the integration on the polar angles of k and then the integration on 1) It follows that o Sno aate foe [a hits tat) ® In the Coulomb gauge, A(t.) = A. (f.1) and the scalar potential is the statie Coulomb potential (8.25.6) ve. oma fee — U isan instantaneous term In order that there be 2 finite velocity of propagation for the E-field, a term in A, which compensates the contnibution of (9) to Es required. Note that ‘A. i not purely retarded, since the transverse current j_ which appears in (8) introduces a ‘nonlocal component 6) The electri fied in reciprocal space is given by 4 =~ ik, 19) Equations (9) and (4 then give aus ik ~ ak if 6ST jer. ay ik ik ik af crooner 9 = amt [aeosinonmtr= 9. 0) Substituting (13) in (1), one finds wera ay68 Exercises os which, taking the results of part 6) into account, gives Equation (I) in teal space as required. ‘The electric field E is ust the field devived from tetarded potentials in the Lorentz gauge (*) This proves thatthe instantaneous Coulomb term {the fst term of (11)] has been compen- sated for by aterm arising from j, [the fst term of 13} 5. FIELD CREATED BY A CHARGED PARTICLE AT ITS OWN POSITION, RADIATION REACTION Consider a charged particle with charge g,, mass m,, and bound in the neighborhood of the origin by a force F which derives from an external potential energy Vir): F = — VV. The purpose of this exercise is to calculate, starting with the equation of motion of the normal variables, the field created by this charge at its own position and to get the expression for the “radiation reaction” which results from the interaction of the particle with its own fcld (**) 2) Find the current associated with the particle in real space and reciprocal space. +b) Write the equation of motion of the normal variables a,(k, 1) describing the state of the transverse field. Let a be the order of the spatial dimensions of the region in which the particle is bound about 0. In this exercise only the modes ke of the transverse field for which k < ky with kydtq © 1 will be of interest. Write, to zero order in kyity, the equation of motion of a,(k, 1). Integrate this equation between the initial time fg and 1, and express aj(k,) as a function of a,(k, 4) and the velocity of the particle, ¢,(r), between f, and f What'is the physical significance of the two terms one gets? c) In the mode expansion of the transverse fields E , and B one only considers the contribution of the modes k < ky. Show that in that case fone can take E, (0,1) and B(O, r) as approximations 10 the transverse fields at point r, where the particle is located. Using this approximation (order zero in Kiydy), give, using the results of ), the expression for the transverse fields E , p(0,1) and B,(O, 1) produced by the particle at its ‘own position. Find the sum over the polarizations and the integral over the angles of k. Show that B,(0,1) is zero and that E, »(0,1) can be written E01) = Pera dre — Dby() where is a 6-funetion of width 1/oy = 1/ck y. (1) See Feynman etal, Vol It (Chapter 2, (0°) See Feynman et al, Vol T (Chapter 28), Jackson (Chapter 17)cs Exercises o 4) Integrate the expression giving E, p(0,1) by parts. Assume that 1 = tis very large with respect to 1/ay, and that r, varies slowly on a time interval 1/wy. Show then that E , ,(0, ) is the sum of two contribu- tions, one proportional to #,(¢) and the other to F,(1). e) Write the equation of motion for the particle. Show that one of the two terms arising from the interaction of the particle with its own transverse field can be interpreted as describing a modification 8m, of the mass m,, of the particle. Calculate 6m,. What is the physical origin of im? J) To interpret physically the other term arising from the coupling of the particte with E , »(0, 7), assume that the particle's motion is sinusoidal. Show that this other term then describes a damping of the particle's motion, What is the physical origin of the damping? Justify giving the name radiation reaction to this term. Solution 2) From (Sb) in. = qierate = 600) aa u ot jth = shea tine a) +b) The normal variable a(k,) obeys Equation (C12), which using (C15) and (1. can be written eine en a ike) + ioayth Die = aa To order 40 in ha 2) yields tone can replace e° #4" by 1 inthe last term of 2), Integration of un are Perkin), ) Toon AP 4k) = atk he The frst term of (3) represents afield frely evolving between fy and 1 from the intial state ‘Uh 1a). The second term desribes the field created by the particle between fy and ©) The exponentials ¢7 "appear in the mode expansion of the fields E, and B evaluated at point ey. If the sums on kat linited 10 K = ky its eorect to order 20 in ‘yay 10 approxinate the exponentials by 1, which amounts to identifying the field a, with 1 atthe origin. To get the feds produced by the particle in its own poston, tsices, then to put the las term of (3) and its complex conjugate in the mode expansion of E, (0.1) ard B(O, 1), One gets then forthe -sompanent of Ey p eaane Ff aan0 Exercises CS ‘Summing over the polarizations gives ay Fae = 3-4 ° and the angular integral leads to foo © wh the sasha re ef materi ih) =~ pte “deige— of awe o Ce evans the impr) ver. Tis es Caren. 2 ue whe as a0 = [anes 0 is 4 et inton of Width 1/ey, Fina £0. = satis [dee nun a0) {An analogous calculation can be done for By(O1). The sum on the polarizations then Jeads to an odd function of 8 = k/k whose angular interal i zero. One has then BO. =0 a 4) & double integration of (10) by parts gives, taking into account the Fact that By isan ‘4d function of + and that 8, and By are negligible for + =~ fy, Buon = staf -avokine[ de 9sao] ay From (9), 8y(0) = wy/s. On the other hand, since #, varies slowly on the scale af 1/ey. fone can replace (1 ~ +) by’ F(t) in the lst integral and remove Ft) from the integral, ‘which then is | as'a result of the even party of 8yy(7) and the fact that 1 ~ ty Lay. One et nally 404, as kn Fu =5 en 0 €) Since B,(O,) is zero, the equation of motion of the partele is writen im, BAO) = = HIG) + 4A. + 0 Bao ay ‘The first term describes the effect of the external potential binding the particle near the origin, the second the interaction ofthe particle with its own transverse Held (the interetion ff the particle with it own longitudinal field leads 10 a 2ero et force through symmetry land the last the interaction with te free felt [Ey 8 associated with the ist term of (3) and describes posible incident radiation) ‘Substitute (13 i (14) and bring the lerm 10 the Fef-hand side. One gets then a EU) + 4, Bac(O2) us im, + om FAN = — Vin) + =C6 Exercises n where Bm, is given by ah in, = 6) “The term in, in (13) then deserves a moskfcaton ofthe mass of the particle which can be interpreted as the “electromagnetic inertia” ofthis particle tis possible moreover to relate 8m, to the Coulom energy #f of the particle Indeed, it one introduces the same cuto at Ky inthe integral (B.36) aking ey, Ome gels diky 3pq ot io Bet - pine 17 To within a factor $, tbe energy Brye? associated with 8m, particle is the Coulomb energy of the {) tthe particle undergoes an oscillatory motion of frequency a, one can replace &, by 1, The second term of (15) then describes a frictional force — (424? /Smeye" Pt, which dampens the motion of the particle, This orce is nothing more than the radiation reaction describing the los of energy of a charged paticle undergoing accelerated mation —lss resulting from the radiation which t emits (see Exercise 7) 6. FIELD PRODUCED BY AN OSCILLATING ELECTRIC DIPOLE ‘A microscopic emitting system is made up of a particle with charge ~4, fixed at the origin and a charge q,, undergoing a motion described by £0) = 9 608 0% t 0 One is interested in the field produced by this system at « great distance from the origin, that is, where r= ay (this corresponds in reciprocal space to wave vectors such that kay « 1), 4a) Find the charge and current densities associated with the system as well as their spatial Fourier transforms. Expand the latter to first order in ay, Derive expressions for j(, 1) and p(t, 1) to first order in ay, b) By substituting the expressions so found into Equation (1) of exercise 4, show that the field E(r, r) radiated by the dipole can be written % £05 (ky f ~ 0!) Er) A [Kay + Viay - ¥)) Q Show that E is a sum of three terms in 1/r, 1/r?, and 1/r? respec tively (*) ©) Show that E(, 1) can also be written as an electric dipole wave such as those defined in Complement B, with J = 1 and M = 0, Oz being taken along ay. The spherical Bessel function of order 1 is given by sinx _ cosx J) es @) (*) See Feynman et al, Vol. Il (Chapter 21).n Exercises Sobtion 2 Using (AS) and (AS). one ge ie. = a.0e— ry iireametan a “The expansion oft order in dy of the spatial Fourier ansform of (4) es te fenihs ys — ee ik, 6.2) 1h = lirateM = I = lira) » jk = pliaisetn = ce sb) si Returning to real space, one gets ‘ Vern si = = qyson in Y= 0) oa oy Sieg Fy 30) ) +b) Put (6) and (6) in Equation (1) of Exercised and consider first the contribution of 4. The fist imtegral in” yields, taking into account (6), ay esin(kyr #90) Fai Tree Following Exercise 4, the contribution of j 10 Et) is gotten by diferentating (7) with respect tot, whieh gives 4.463605 hor — 29 1 Tr 5 6 ‘The second nepal in in Equation 1) of ser 4am be writen sing (58) He ferroe— ents, mae ” were ru 0 Using he ienity We A=W -UA= AVE wy allows one to transform (9) ito PE [er dley ag Wie yf) (2 msl ay YH). Tre ollowing Exercise 4, the contribution of » to E(t i gotten by applying ~ V to (12), which gives, using (10), deve, - yp Sotlbor = 0) 13 ar 7 a ‘The sum of (8) and (13) gives Equation (2) as stated6.6 Exercises, n Calculating the double spatial derivative in (13) gives Jestes =o ~ a, Ser eo +] po Slor at estes " a sort = 4) us ‘One gets the well-known thre terms of dipole radiation, behaving respectively as 1/r. 1/7, and 1/r' ©) To make the connection with the multipole waves defined ia Complement By. ii ‘seul to introduce the radial and tangential components of ag with respect toe ra) By = Me = Ay uy) Equation (14) cam then be written in the form = ep hysinikyr — ot) _ c08 (kyr =e] . 2, [zeatie st), 206th ny i ‘Compare this with the results of Complement Bh, The electric field of an electric dipole wane is, using (63) proportional to Kgyy2(®) with J= L and Af'= 0, the Os axis being taken along ay, We now apply Equation (62) of complement By to this particular case. The spherical vector functions Zay(F/#) and Nu(t/r) are derived from the spherical harmonic Yoolt/Pi whose value, to within a constant Tact, “ Using Equations (3) and (8) ofthe complement oe ids, with Nuff) =? Be (18.0) (es i, «18.09 Equation (62) can then be wetten 1 Nawal) x 7 [eevee See]. “91” Exercises C7 Thus to within a multiplicative facoe anda phase one gets (16. Is ako useful to note that (16 is traveling wave, while (19) is the amplitude ofa spherical standing wave resulting from the superposition of an outgoing and an incoming wave, 7. CROSS-SECTION FOR SCATTERING OF RADIATION BY & CLASSICAL ELASTICALLY BOUND ELECTRON. A classical electron with charge q and mass m. elastically bound to the origin by a restoring force —mogr, is set into forced motion by an incident monochromatic wave with frequency « and emits into all space radiation of the same frequency. The purpose of this exercise 1s 10 calculate the total scattering cross-section 9(.) of the electron and to examine its order of magnitude as well as its variation with & (*, 1a) The electron undergoes forced motion along Oz with amplitude a and frequency «: a cos ot © One recalls that the total radiated electric field at a distant point M (OM > 2 = 2nc/a) isin the plane (Oz, OM). is normal to OM, and has an amplitude (**) ga sin8 @ where r is the distance OM and 6 is the angle between Oz and OM. The field B has an amplitude £/c and a direction parallel to OM X E. Find the mean value (over a period 2/«) of the flux of Poynting’s vector #e2E x B through a sphere of very large radius r, and find AW /at, the mean energy radiated per unit time 'b) The interaction of the electron with the field it creates at its own position can be described by a force, called the radiation reaction, whose component along Oz has the value (***) R — Ba) For the forced motion (1), find the mean value (over a petiod 22 /«) of the work done by R on the electron. Compare this result with that from a), What is the physical interpretation? (7) See Jackson (Chapter 17). (2°) These properties can be goten from the tests of Exercise 6 (6) See Exercise $ fora demonstration ofthis result. The interaction ofthe electron with ts own field is also responsible for a change 8m in its mas, which is assumed tobe included fn the mass mused hete7 Exercises B ¢) One writes the radiation reaction in the form B.b) © @ Rey me ® and Aq = e/a. In the absence of incident radiation, the dynamical equation for the electron is written 6) R (which is proportional to the factor r/Ap * 1) can be treated as a perturbation. Find the solutions of (5) of the form e' and show that, to first order in 7/Xp, one has = 0+ O) Give the expression for Yo a8 a function of rp. wy, and c. What does the time 7 = Yq! represent? d) In the presence of an incident field polarized along Oz whose amplitude at the origin is Ecos wt, the dynamical equation for the electron is written Cae = mag 2 + F 4 qBoosor 0 Find the forced oscillatory motion of the electron and, using the results of 4), the energy radiated per unit time into all space by the electron. ¢) Find the energy flux (averaged over one period 27/.) associated with the incident wave, which is assumed to be plane and propagating along Ox. Using the results of d), find the total scattering cross-section (0(.). Express 0() as a function of rj, w, and Y%, J) Assume ow, (Rayleigh scattering). Show that o(w) is then proportional to a power of «, which should be found. 1g) Assume wy © w € ¢/r (Thomson scattering), Show that a(w) is equal to a constant hy Assume finally @ near @, (resonant scattering). Show that the variation of o(w) with — a exhibits a resonance. What is the width of the resonance? What is the value of the cross-section o(,) at resonance?6 Exercises C7 Solution 1a) The ux of the Poymting vector through a sphere of radius is equal to ps off) fr a(t) SR © te soit ition otis re wenger on sd g dot Timo to) 3 is just the energy radiated per unit time, dH//dt, by the oscillating charge 'b) During time dr, the charge is displaced by d= ~ £4, and the work dis” done by R is ‘equal to aw «io From (10) it follows that aw og ‘ Gr ome [a Bree a Let us take the average of (I1) over one period 2=/a, Since 42 isa periodic function of ¢ with period 2n/u, the average of (32)/dr is 22t0. AS for the average of is value, according to (1), is @°s8/2, Finally, aw Be aw wa. wot = 5a F «2 “The mean work per unit time which does onthe oscillating charg is equal. except forthe sign. tothe mean energy which this charge radiates into all space per unit time. The radiation reaction thus deserines for the electron the dissipative phenomenon associsted with the raiative energy loss £6) Replacing z by cin (S) gives, using (3) and Ay = €/4- dea 7 ‘To zeroth order in G/Ao, = up. To fist order, one can replace 9° by eu} on the right-hand side of (13), and 2 — af by (2X + ay) = + 2a4($2 F uy). which gives (with Qo = 03) as If ay is an sf oF optical frequency, fy = Ay and yy 4, The time % = 4" is the decay time of the oscillatory energy due to the radiative energy toss. With a7) 1 the electron Undergoes numerous osiations during decay. 4) Assume = teleg a9 ‘One then gets, substituting (15) in (, 416)er Exercises n “To get the ncan energy rated per uit ie in steady satis necessary to replace a? by {sulin (9), which then gives WW 1g gett 1 G39 wT FO oy ae on i ag fe) The incident flux, is equal 19 4, as} shih ves for the toa erss-section (4) 107 sr, ot a) = =e + “9 ) Wee © oy, the denominator of (19) is ofthe onder of, which gives 8s 7 edgar =. eo “The total scattering cross-section varies as the fourth power of the inideat frequency 1s) The condition = ¢/ which indicates thatthe wavelength of the incident radia: vio’ ver Tange compared 10 shows in view of (14) that of > au /a IF im addition ie very Tage with respect to uy the denominator of (19) reduces to", which shows that PE ea en which is jst the Thomson scattering cross-section fh) Near eevonance, in (17) one can replace aul /of by wa (ay! by ‘fajtto — anys and inthe numerator of by uf, which yields oso, © au) en ono a “The resonance vatiations of of) are those of a Lorentzian with total width at half 40. = 3 ev At resonance, the value of (4) sing (14), ew where 1s the resonant wavelength For optical radiation, Ay is of the order of $ % 10-7 m, whereas r is of the order of 250 10" m, One has then about 16 orders of magnitude difference between the resonance ‘eroseseetion (24) and the Thomson cross-section (21).CHAPTER II Lagrangian and Hamiltonian Approach to Electrodynamics. The Standard Lagrangian and the Coulomb Gauge Classical electrodynamics, which has been presented in Chapter T beginning with the Maxwell-Lorentz equations, can also be derived from ‘a very general variational principle, the principle of least action. In classical mechanics, the principle of least action allows one to select, from all the possible “paths” leading from a given initial state to a given final state, that path which is indeed followed by the system. (See for ‘example, Figure 1, corresponding to a system with one degree of freedom.) One associates with each path a number, called the action, which is the time integral of an important function, the Lagrangian, and one seeks the path for which the action is an extremum, Figure 1. “Paths” connecting the initial state x,,f, with the final state x3, f,. The true path, that one followed by the system, is that for which the action is an extremum. The Lagrangian formulation of classical electrodynamics, using the Principle of least action, was first put forward by Schwarzschild in 1903. It required the generalization of the usual Lagrangian formalism to the case ”80 Lagrangian and Hamiltonian Approach 0 where the dynamical system has an infinite number of degrees of freedom. Indeed, to specify the state of the electromagnetic field, it is necessary to give the values of the fields (or the potentials) at all points in space. Beyond its uncontestable compactness and aesthetic character, this formu- lation has the advantage of introducing all the quantities necessary for canonical quantization without ambiguity (Hamiltonian, momentum). In addition, recent progress in quantum field theory, such as the unification of the electromagnetic and weak interactions, is based on such a formula- tion of the theory In Part A of this chapter, we briefly review the essentials of the Lagrangian formalism in the simple case of a system having a discrete set of degrees of freedom. The generalization to systems having a continuum ‘of degrees of freedom is then examined, as well as some results relevant 10 the use of complex variables in the Lagrangian formalism. It is clear from Chapter I that the equations of electrodynamics have a more transparent structure in reciprocal space. The variables which define the field are then complex. Part B is devoted to a presentation of the standard Lagrangian formu- lation of classical electrodynamies with a nonrelativistic treatment of the particles. One shows that the Lagrange equations associated with a certain Lagrangian (the standard Lagrangian) reduce to the Maxwell-Lorentz equations. In fact, the standard Lagrangian uses potentials to describe the electromagnetic field; this results in redundant degrees of freedom, which causes certain difficulties, particularly when one tries to quantize the theory. One method for resolving these difficulties consists in eliminating certain degrees of freedom such as the scalar potential and in using the Coulomb gauge. Part C is precisely devoted to a presentation of electrodynamics in the Coulomb gauge and to a discussion of several important aspects of this theory. One sees how the application of the Lagrangian formalism in this gauge as well as the Hamiltonian formalism allows one to proceed to a canonical quantization of classical electrodynamics and then to justify the commutation relations introduced heuristically in Chapter IWAL ‘Review of the Lagrangian and Hamiltonian Formalism aL ‘A—REVIEW OF THE LAGRANGIAN AND. HAMILTONIAN FORMALISM We will review here, without going into detailed calculations, the general Lagrangian and Hamiltonian formalism for a finite number (§1) End then for a continuum of degrees of freedom (§2). The reader who is ‘hot familiar with the ideas introduced in §2 is advised to study Exercise 2, dealing with a chain of linear coupled oscillators. The passage to the Continuous limit of such a system renders more plausible the results of §2 (ihe appearance of spatial derivatives in the Lagrangian density, replace: ment of Kronecker symbols by the Dirac delta function, and so on). 1. Systems Having a Finite Number of Degrees of Freedom 1a) DYNAMICAL VARIABLES, THE LAGRANGIAN, AND THE ACTION For a system having N degrees of freedom, giving the N generalized coordinates Xy,....%y and the corresponding velocities %,....%y at given time completely determines the subsequent motion. The 2 quanti fies X,..-%y and Xy,..-58y form an ensemble of dynamical variables. ‘The accelerations £,,..-, y can be expressed at any time as a funetion of these variables. The resulting equations of motion are then second-order differential equations in time. The motion of the system is determined by integrating these equations. Tt is equally possible to specify the motion of the system by means of variational principle. In the Lagrangian approach, one postulates the tristence of a function L(x, 1), called the Lagrangian, which depends fon the coordinates and the velocities (and perhaps explicitly on time) Such that the integral of L between times 1, and f, will be an extremum when x(t) corresponds to the real path of the system between 4 andl [the initial and final coordinates x,(/,) and. x,(t3) are assumed to be Known], The integral Usp. 4f0. dt a is the action, and the corresponding variational principle is called the principle of least action. Tn the mechanics of a point particle, the Lagrangian is equal to the difference between the kinetic energy and the potential energy. In particu- far, for a particle moving in a time independent potential, the Lagrangian does not depend explicitly on time. In the following, we will preserve for fsolated systems this time translation invariance and will not show an txplicit time dependence for the Lagrangian, which will be writtena ‘Lagrangian and Hamiltonian Approach, WAI b) LAGRANGE’s EQUATIONS In the subsection above, two possible approaches to the study of ‘motion have been indicated: one local in time (the equations of motion), and the other global (the principle of least action). We will review briefly how the equations of motion can be derived from the principle of least action, In expressing the stationarity of the action with respect to variations of path 8x,(1) {provided 6x,(t,) = 6x,(t) = 0] about the real path followed by the System, one shows that at each instant the dynamical variables should satisfy on that path relationships which are equivalent to the equations of motion. These are Lagrange’s equations: (A2 The explicit derivation of these equ: Ay, and in a number of books (set chapter) ns can be found in Complement the references at the end of the ©) EQUIVALENT LAGRANGIANS ‘The Lagrangian of a system is not unique. For example, if one adds to the Lagrangian L the total derivative with respect to time of an arbitrary function f depending on the coordinates x, and the time, one gets a new function 1 LO i, 0 = Ley 5.0 + 2 fey. (a3) are eo S + Seta), ta) — F0ehn). 4) (A.4) Since the initial and final positions are fixed, it follows that S and S’ differ only by a constant and thus have the same extremum, Land L’ are then equivalent Lagrangians for the study of the dynamics of the system. The transformation (A.3) then allows one to transform a Lagrangian into ‘an equivalent second Lagrangian,HAA Review of the Lagrangian and Hamiltonian Formalism B Remark ‘The function f must not depend on the velocities ), so that the new Lagrangian, like the original, depends only on the dynamical variables x, and &, and not on the accelerations. d) CONJUGATE MOMENTA AND THE HAMILTONIAN The momentum conjugate to x, is defined as the partial derivative of the Lagrangian with respect to the velocity x; ok mee as a, “The time derivative of p, is gotten by using the Lagrange equation (A2) a. b- Ce) % ‘The simplicity of Equation (A.6) suggests the use of the coordinates and ‘momenta as dynamical variables rather than the coordinates and veloci- ties. It is then preferable to substitute for the Lagrangian another func tion, the Hamiltonian, which is considered as a function of x, and p, and is defined by Hex p) = Lp, ~ b ra) It suffices then {o differentiate (A.7) and to use (A.5) to find that dH is only a function of dx, and dp,, which leads to the following equations, called Hamilton’s equations: (A.8.a) (A.8.b) In order to describe the dynamics of the system, the N Lagrange equa tions (A.2), which are second-order differential equations, have thus been replaced by a system of 2N first-order differential equations (A.8.a) and (AB). In comparison with the Lagrangian formalism introduced earlier, the Hamiltonian formalism presents several advantages. First of all, if the Lagrangian does not depend explicitly on time, H is a constant of the motion which generally corresponds to the energy (*) and thus has a clear (+) Im certain cases one gets Hamiltonians which are diferent from the energy (ee [Exercise 1). This is ever the case hee8 [Lagrangian and Hamiltonian Approach WAL physical significance. Additionally, the coordinates and momenta play a more symmetric role than the coordinates and velocities [this arises for ‘example in the equations of motion (A.8.a) and (A.8.b)). It follows that changes of variable are a priori more flexible than in the framework of the Lagrangian formalism, since one can mix the various coordinates and momenta. Finally, the introduction of the momenta and of the Hamilto- nian is essential for the quantization of the theory. ¢) CHANGE OF DYNAMICAL VARIABLES 4) Change of Generalized Coordinates in the Lagrangian It may be useful, for the solution of a problem, to make a.change of dynamical variables. In the framework of the Lagrangian formalism, only changes of coordinates which substitute the coordinates Xj...., Xy for Xyseee Xy Stich that =H Xv) (A.9) are allowed. The differentiation of (A.9) with respect to time gives the ‘elationship between the old and new velocities. The new Lagrangian is gotten by replacing in L(x,, %,) the x, and x, as functions of X, and X,, and the new action is equal to the old one. The transformation (A.9) does not in general involve the velocities X,, since the accelerations X, would then occur in the Lagrangian, iil) The Special Case Where a Velocity Does Not Appear in the Lagrangian There is nevertheless a case where a transformation of the type (A.9) including the velocities is possible. This is when one of the velocities does not appear explicitly in the Lagrangian. It is then possible to completely eliminate the corresponding degree of freedom and to substitute for the initial Lagrangian a Lagrangian having fewer dynamical variables. Assume, for example, that in the Lagrangian L, the velocity xy does not appear explicitly. The Lagrangian L is then a function of N coordi- nates and N ~ 1 velocities and will be written as L(x), xy. %,). We will show that it is possible to replace this Lagrangian by another Lagrangian L which depends only on the N ~ 1 coordinates x;,..-,Xy-y and the N = 1 velocities %,..+.2y—1. The Lagrange equation relative t0 xy is eb (A.10) This equation allows one to express xy as a function of the V ~ 1 other coordinates x, and the corresponding velocities i, so that xy = Ms) aWAL Review of the Lagrangian and Hamiltonian Formalism 85 If in L one replaces xy by its equivalent (A.11), one gets a function of x, and, which is the Lagrangian £: Esp.) = Loy wer, 5), 4) (A.12) Note first of all that no acceleration appears in £ in spite of the form of the transformation (A.11). That is due, of course, to the fact that 1. does hot depend on xy. To show that Z has all the properties of a Lagrangian, itis sufficient to note that, ifthe action relative to L is an extremum for all independent variations of the coordinates x, and xy, then that is also the case for the action relative to L, in which xy is fixed by (A.11). Remark One can verify directly that the Lagrange equations associated with Z have the expected form, We caleulate, for this purpose, the partial derivatives of L with respect to x, and &, ah ik, th oxy ak (Ata) (A.13.6) ‘When Lagrange’s equation (A.10) relative to is satisfied, Land have the Same partial devvatves with respect tox, and, Lagrange’s equations relative tox, and derived from 1 thus involve those associated with £., Note also that SNovt the las term of (A13b) is eto, the momenta conjugate with x, im Land Fare equal iii) Velocity and Momentum Transformation It is interesting to establish the transformations for the velocities and ‘momenta when one changes variables in the Lagrangian. The equations (A9), allowing one to go from the old coordinates x, to the new coordinates X;, give by differentiation (A.14) with (A.15) We denote by p, and P, the momenta conjugate with x, and X, respectively. One gets (A.16)86 Lagrangian and Hamiltonian Approach MAL that is, on using (A.14), P= Doan (a. ‘The equations (4.14) and (A.17) can be rewritten in matrix form. Denot- ing by A the N x N matrix whose elements are a,,, and by (),(X), (p). and (P) the column vectors whose elements are %,, X,, pj, and Py respectively, one finds = 4%) (A.18.a) (P) = 4%) (A.18.b) ‘The transposed matrix A" is assumed to be invertible, so that.the relation- ship between the momenta can be rewritten by going, as with the veloci- ties, from the new to old variables: A) (P) (A.19) @) Its clear then that the momenta transform like the velocities only if the matrix A is orthogonal iv) Changes of Variables in the Hamiltonian Formalism ‘The changes of variables in the framework of the Hamiltonian formal- ism are inherently broader than the transformations of the form (A.9), which depend only on the coordinates, Certain transformations of the form Xp = G10 XysPa~ Pad (A.20.a) Py iy Xys Ph om Py) (A.20.b) are possible a priori. The equations of motion for the new variables X and P have a form analogous to (A.8) only if certain restrictions are imposed. One can show that the corresponding conditions are {X.x)} =0 (4.21.2) (P,P) =0 (A.21.b) {XP} = 4, (A210 where the Poisson bracket (a, b) of the two functions a and b is defined by Quantities X and P satisfying (A.21) are called canonically conjugateWAL Review of the Lagrangian and Hamiltonian Formalism 87 1) Use oF ComPLEX GENERALIZED COORDINATES i) Introduction Up to this point we have considered only the case where the coordi- ates and velocities afe real quantities. It can be useful to introduce complex quantities for the resolution of certain problems. Consider, for ‘example, a Lagrangian depending on two coordinates x, and x, and their velocities, and introduce the complex variable x= 4, tin) (A.23) v2 ‘The Lagrangian is now a function of X and X* as well as their derivatives with respect to time. We will show below that the Lagrangian and Hamiltonian formalisms can be generalized to complex coordinates X and X* [related to x, and x, by (A.23)], and that all the results obtained above remain valid (with certain amendments to the definitions, notably that for momentum) provided that X and X* are considered formally as independent vari- ables Remark One could consider from the beginning Lagrangian depending on complex. dynamical variables. One has however to keep in mind that the action must be real, since the principle of least action involves finding a minimum of the action, The Lagrangian is thus a real quantity and depends on both X and X*. fil) Lagrange's Equations ‘The passage from variables x, and x to X and X* is linear and can be inverted, One deduces the following relations: eb cL i a xe A a, 71am. ee) and the similar expressions for the derivatives relative to the velocities The Lagrange equations for the complex variables are gotten then by combining the Lagrange equations relative to x, and x3. One then gets two Lagrange equations, one relative to X and the other to X*, which have the normal form (A.2)88 Lagrangian and Hamiltonian Approach WAL iii) Conjugate Momenta Rather than define the momentum conjugate to X by L/@X, we prefer to take (A.25) To justify such a choice, note that the laws of transformation of the partial derivatives (A.24) give (p, + ips) (A.26) Comparison of (A.23) und (A.26) shows then that the real and imaginary parts of the momentum P correspond respectively to the momenta conju gate to the real and imaginary parts of the generalized coordinate X, Which justifies the choice of (A.25). Equations similar 10 (A.24) for velocities show in addition that (a2) In fact, this property is general and is a consequence of the real nature of the Lagrangian. An equivalent definition of P is then aL (A.28) iv) The Hamiltonian The Hamiltonian introduced in (A.7) depends on the quantity 2%, p,. By applying (A.23) and (A.26), one easily sees that i. py + kop, = XP* 4 XP (A.29) It follows that the Hamiltoni variables, is H, expressed as a function of complex H= Xpt+X*P—L (4.30) His clearly real ¥) Change of Complex Variables It is possible as above to imagine a change in complex variables transforming an ensemble of complex coordinates X,.... Xv. X,..., Xf to another ensemble Z;,..-, Zys Zifs+--5 Zg. Hone requires that the momenta (defined in A.25) transform like velocities, there areReview of the Lagrangian and Hamiltonian Formalism 89 certain constraints on the matrix of partial derivatives of the old coordi nates with respect to the new ones. A development analogous to that in §A.Leziii shows that this matrix must be unitary 1s) COORDINATES, MoMENTA, AND HAMILTONIAN IN QUANTUM MECHANICS ‘The various physical quantities of a system become operators in quan- tum mechanics. These operators act in an abstract space called the state space, which has the properties of a Hilbert space. The canonical commu- tation relations between Cartesian components of the position and mo- mentum operators x and p are equal to Liq. Xa) = 0 (A.31.a) [Pa Pal = 9 (A.31.b) [ves Pal = iF By (A31.¢) In quantum mechanics, the state of a system is described by a vector |) of Hilbert space, One can at this stage adopt essentially two points of view, either assuming thatthe state vector is fixed and the operators evolve with time (Heisenberg) or assuming that the operators are fixed and the state vector evolves with time (Schrdinger). Tn the first point of view the evolution of a physical quantity G is described by the Heisenberg equation a ihG=(6H) (4.32) where H is the quantum operator associated with the Hamiltonian, In the case where the operator G corresponds to a coordinate or a momen- tum, the equations derived from (A.52) are the quantum equivalents of Hamilton's equations (A.8). {in the second point of view, the operators are fixed and the evolution of the state vector is determined by Schrisdinger’s equation diyye in |W> = Hiv> (4.33) Mathematically, the correspondence between these two points of vie via a unitary transformation on the state vector. Remark In the case where a velocity does not appear in the initial Lagrangian, the conjugate momentum associated with the corresponding coordinate is ident cally zero. This poses a serious problem for quantization, since it is then0 Lagrangian and Hamiltonian Approach WA impossible to postulate the canonical quantization relation (A.31), One can resolve this problem by climinating from the Lagrangian the coordinate associ- ated with this velocity (see §A.1.ei). The conjugate momenta are then calcu: lated from the new Lagrangian (with a reduced number of dynamical variables), and the canonical quantization relations (A.31) are then imposed ‘The complex variables have been introduced in §A.1.. We now exam ine the canonical commutation relations in this case, Equations (A.23) and (A.26) show that 1X, Pl = 4h, + ixseps + ined (4.34.2) 1X, Pt} = bby + inary — ind (A'34.b) It follows then from (A.31) that Ix, P]=0 (A.35.a) xt, P*]=0 (A.35.b) (Pl (A350) xP) (4.35.4) (the other commutators, between two X or between two P, ate zeto). The definition (A.25), which we have taken for the conjugate momentum, leads to a nonzero commutation relation between the operator X and the adjoint of the operator associated with the conjugate momentum, 2, A System with a Continuous Ensemble of Degrees of Freedom a) DYNAMICAL VARIABLES The state of the system is now determined by a set—no longer discrete, but continuous—of dynamical variables. This extension is necessary in so far as one wishes to study the electromagnetic field, which is defined by its value at all points of space. We thus consider generalized coordinates which depend on a continuous index (denoted by r, a point in three- dimensional space, in anticipation of the application to the electromag- netic field) and a discrete index j (which varies from 1 to N). As with the discrete case, the coordinates 4,(r) and the velocities 4,(r) defined by Ale = EA) (4.36)WA2 Review of the Lagrangian and Hamiltonian Formalism na form an ensemble of dynamical variables for the system. It is important to stress the fact that in the Lagrangian formalism developed below, r is not ‘a dynamical variable but an index (of the same nature as j) 1b) THE LAGRANGIAN ‘The Lagrangian L, which is a function of the dynamical variables ,(r) and 4,(e) (where j and r take on all possible values), can have a very large variety of structures, We assume here that one can write are (a3) where the function & is called the Lagrangian density. 2 is a real function of the coordinates A,(0), the velocities 4,(r) and also the spatial derivatives (denoted 4,4, with 3,= 4,,0,, 4.) whose presence simply shows that the motion of the coordinate A, (r, 1) is coupled to the motion fl a neighboring point in the same way as, in a problem with discrete variables q,, the motion of q, depends on q,_, and 4, (see Exercise 2) It should be clear that these spatial derivatives are not new independent dynamical variables, but rather linear combinations of generalized coordi hates. One can include a priori in the Lagrangian density the spatial derivatives of all orders (remember, though, that only the first-order time derivative is allowed). Now taking into account the later application to the electromagnetic ficld, we will only study Lagrangian densities of the form BA, 4,8). Remarks @) One can imagine an explicit dependence of ~ on the point rand the time ¢ ‘We will not show that, to prevent overburdening the notation, (i) The Lagrangian density that is used in electrodynamics contains spatial derivatives. Such a structure can easily be understood. Maxwell's equations ‘describe the motion of fields coupled from point to point in space, and the absence of spatial derivatives in the Lagrangian density would lead to a theory ‘where the feld evolves independently at each point in space. The fact that the ‘Maxwell equations involve the spatial derivatives of the field requires taking a Lagrangian density that likewise depends on the spatial derivatives. This suggests studying the Lagrangian density in reciprocal space rather than in real Space, since it has been seen in Chapter T that the Maxwell equations are Strictly local in reciprocal space. We will return to this point in Part B of this chapter. ©) LAGRANG#'S EQuaTions In going from the discrete case to the continuous one, most of the equations written in §A.1 remain formally valid, However, certain opera-2 ‘Lagrangian and Hamiltonian Approach Wad tions (the derivative of the Lagrangian with respect to continuous vari ables, for example) are not mathematically obvious, and it will be useful to explain certain points in more deal Note frst of all that the action $ is, as in the diserete case, the time integral of the Lagrangian. By reason of the form postulated for the Lagrangian density, we can then write | a far eeay a (4.38) The principle of least action is of course unchanged: S is an extremum when 4,(r, 1) corresponds to the actual motion of the field between times 1, and To establish the equations of motion, one uses the principle of least action. The same steps can be followed as in the discrete case; one studies the modification of § when the field is varied by the quantity 84,(r, 1) with respect to the path for which is extremal (84 ,(r, 1) being zero at the temporal limits 1, and ¢, of the integral and likewise when |r| tends toward infinity), By stating that Sis extremal one then gets (see Comple- ‘ment A,,) Lagrange’s equations, which can be written in the form bee ew aan Remark Equation (A.39) uses the Lagrangian density 7, and not the Lagrangian I as in the diserete case (A.2). However, itis possible to write (A.39) in a form identical to (A.2). To do this, the notion of “functional derivative” must be introduced (the extension of the idea of the partial derivative in the continuous. ccase), and this is discussed in Complement Ay. The introduction of the Lagrangian density is mathematically convenient in the sense that, since # depends only on a finite set of variables, the use of the partial derivative is perfectly clear ‘The Lagrangian of a continuous system, like the Lagrangian of a discrete system, is not unigue. One can add to the Lagrangian density the time derivative of a function and the divergence of an arbitrary field (but, one which tends to 0 sufficiently quickly at infinity), possibly depending on the generalized coordinates A(t) g + Bplay, n+ Seyler. (4.40) To calculate the new Lagrangian L’, it is necessary to integrate ” over space. The integral of ¥ -f then transforms into a surface integral atWA Review of the Lagrangian and Hani Formalism 3 infinity, which vanishes by hypothesis, and L' differs from L only by the time derivative of some function. L’ is then equivalent to L, d) CONJUGATE MOMENTA AND THE HAMILTONIAN ‘The conjugate momentum is defined in the continuous case by general- izing the equations gotten in the discrete case. For a Lagrangian of the type (A.37), the conjugate momentum associated with the variable A,(e) has a simple form as a function of the Lagrangian density’ ce cafe) ngs) = A.40, Remark I is easy to understand why (A.41) is the generalization to the continuous case of (AS) for the discrete case, To see this we transform the integral (A.37) defining the Lagrangian to a sum over small spatial elements of volume a? centered on the points f(xy Js Zn) Niet tame Ea) — Ata | Sat of 4feas ditto. Ak (aa ‘The final argument in the bracket symbolizes the various quantities which, in the limit a — 0, tend to the partial derivatives 9,4. The conjugate momentum associated with the variable 4,(r,,) can be found, for the Lagrangian Las in the discrete case and is Me) = (A. Cty" eAfee) I: appears then that the conjugate momentum 1,7) defined in (AL) is equal to the limiting value of a “TTV(e,) when a goes to zero. Now the limit of a 4L,/04 {4,) when a goes 16 70 is nothing but the functional derivative of the Lagrangian 92 24,(e) (see Complement Ay, for more details). Equation (A41) can then be writen in the equivalent form mye) = (aay 21) A step analogous to that used in the discrete case allows one to go from the coordinate-velocity pairs of dynamical variables (4,(r), 4,(r)) to another pait made up of the coordinate 4,(r) and its conjugate mome tum IT,(r) and then to introduce the Hamiltonian H and the Hamiltonian density 20 na forsnmiget-ferr ass94 ‘Lagrangian and Hamiltonian Approach TAD # = TMAe) Ay ~ (A.45.b) Remark Its similarly possible to introduce the momentum and the momentum density of the field as well as the angular momentum and the angular momentum density (see Exercise 5) As in the discrete case, the introduction of the Hamiltonian formalism introduces new dynamical equations (Hamilton’s equations). These can be written simply using the Hamiltonian density 2 as (A.46.a) (A.46.b) These equations can be written directly as a function of the Hamiltonian H with the aid of the functional derivative (see Complement A,,). They are then identical to Hamilton's equations in the discrete ease. e) QUANTIZATION As in the discrete case, the fundamental commutation relation is imposed between the operators associated with a coordinate and its conjugate momentum. In the case where the field is expressed as a funetion of its Cartesian coordinates and where the three coordinates are independent dynamical variables, the canonical commutation relations can be written [A,(9, A,(0)] = 0 (A.47.a) [1 1,(0)] = 0 (A.47.) [4,0 140] = if 6, 5 =F) (A.47.0) Remarks ) Im the continuous case, the Dirac distribution &(r ~ r) has replaced the Kronecker symbol ofthe discrete case. This can be understood if one recalls the situation considered in the Remark of the preceding subsection (§A.2.d), where the space was divided into cells of dimension a, In that case, the rules of the discrete case give [et(ea) 1200) = De Sue (4.48)Wad Review of the Lagrangian and Hamiltonian Formalism 5 Since I1,(r) isthe limit of a” 11s (r,) when a goes to zero, it follows that om [ATC] = iF By im (a.49) ,, is 1 i'r, and r, belong to the same cell of volume a’, and 0 otherwise. It ‘appears then that, given as a function of 1 and r), the limiting value of the second term of (A.49) is just the Dirac distribution 8(r, ~ 1). ii) The commutation relations (A.47) are valid only ifthe three components of the fill 4,, 4, and 4, are independent dynamical variables. We will see that in electrodynamics there are situations where this is not the case. Itis impor tant then to identify the truly independent dynamical variables in order to write the commutation relations correctly. ‘The state of a system is described by a vector in state space, and—as in the discrete case—one can treat the dynamics from the Heisenberg point of view or from the Schrodinger point of view. f) LAGRANGIAN FORMALISM WrTHt COMPLEX FIELDS The generalization of the foregoing results to the case of a complex field is particularly important, since in electrodynamics it is often most interesting to study the equations in reciprocal space. Consider now a Lagrangian L and a Lagrangian density # dependent ‘on the complex fields w¥, and their velocities «/,, Since L must be real, must then depend on ./* and .//*, with the result that L= fore Lol, chy Oyely, oly oh *eal*) (A.50) the integration variable being now denoted as k in anticipation of later applications, and a, denoting 4/8, Remark ‘The electrodynamic Lagrangian is simpler than (A.S0). Since Maxwell's equa- tions are strictly local in reciprocal space, so is the Lagrangian density. One then does not get the derivatives 2/2k, in the electrodynamic Lagrangian (on the other hand, one has an explicit dependence on k which arises from the Fourier transform of the spatial derivatives). However, we will retain the form (of the Lagrangian (ASO) for the general considerations in this subsection, the results providing on one hand application to other physical situations (see Exercise 7, where the Schrodinger equation is derived from a variational principle) and allowing on the other hand a clearer comparison with the real-field case.96 Lagrangian and Hamiltonian Approach MA Consider the following linear combinations: fh) + of th) ok) = (Asia off(k) = (A.51.b) which allow the replacement in (A.50) of the complex variables with the real variables /)M(k) and .2//(k). It is then possible to develop the Lagrangian and Hamiltonian formalism previously set forth using these real variables then to restate the equations one gets as a function of ., and #,*, Such a procedure has already been carried out in the diserete case (§A1.f), and we have seen that one gets the same result by considering at once the complex dynamical variables and their complex conjugates as independent dynamical variables. One establishes in this way two Lagrange equations relative to s¥, and dev _ 5, oF y (A.52.a) ee ae") (A.52.b) Geogr e |As for the momentum conjugate with the variable ./,(k), itis defined in a fashion similar to the discrete case: (A.53) This choice assures one, as in the discrete case, that the real and imaginary parts of 7%(k) are clearly the momenta conjugate with #,* and #/. Itis for this reason (see the Remark below) that we have chosen the definition (A.53) in preference to the usual convention [where one does not have the complex conjugate in the right hand member of (A.53)] ‘The fact that . is real shows that (A.53) can finally be written myn) = £ (asa) cathy) which shows that the momentum conjugate with * is T,* Remark ‘The definition (A.53) of the conjugate momentum has another advantage when the fields 9k) are the Fourier transforms of the fields 4,(r). The 7E(k) areWA Review of the Lagrangian and Hamiltonian Formalism ” then the Fourier transforms of the momenta 11(#) conjugate with the variables 4,(8. To understand this result, fist consider going from the coordinates 4, (1) {6 the coordinates 9 (k) as a change of variables to which the results of §A.1.f can be applied. It has been stated there, for discrete variables, that the ‘momenta transform like velocities if the transformation matrix is unitary. This property can be generalized to the continuous case and is clearly fuliled by the Fourier transformation. In addition, the transformation being linear, the coor dinates transform in the same way as the velocities and momenta. It is also possible to give a direct proof of this result by using the definition of the conjugate momentum in terms of the functional derivative (A44). One can write ee [ats i, tay Ripe Paige aL ae (assy ‘Now, by differentiating the relation connecting the flelds in teal space with that in reciprocal space [Chapter I, Equation (B.1)}, one gets aayn et a (A.56) aye) Since Equation (A.53) can also be written mya (As) Hk) it follows that farmee a (A.58) at Wwhich demonstrates that the conjugate momenta are transformed like the variables 8) HAMILTONIAN FORMALISM AND QUANTIZATION WITH COMPLEX FIELDS To find the relationships involving the Hamiltonian density and Hamilton's equations or the canonical commutation relations, itis neces- sary {o start with the expressions found in §§A.2.d and A.2.e for the real fields o/ and 9f/, and to combine them to get the corresponding expressions for the Complex field. This has been done in the discrete case (SA.1.f and A.l.g) and here the results will be given without the intermediate steps. For the Hamiltonian density one finds Eenlt +p aly A.) which generalizes the expression (A.30) relative to the discrete case.98 [Lagrangian and Hamiltonian Approach AD Hamilton's equations are written (A.60.a) (A4.60.b) ‘The canonical commutation relations for the quantized fields are finally [7 k), M(k9] = 0 (A.61.a) LAK), TSK] = i 5, 50K =k) (A.61.b) “The other commutators between 7,(k) and f,(k’) or between 7,(k) and Ta(k’) are zero. As in the discrete case, the field operator and the adjoint of the operator associated with the momentum do not commute, Remarks () In the foregoing, quantization has been accomplished by associating with the dynamical variables and their conjugate momenta operators which satisfy the commutation relations (A.61). In fact, the fundamental requirement with respect to quantum theory involves the quantum equations governing, the evolution of the variables f, and 7T,. These equations, written L ab (0.0) L ) R= FU ‘mean that the Hamiltonian is the generator of time translations. They must have a form analogous to that of the classical equations of, = CHET . (0.6) tt, = - cHtae Such a condition is simply satisfied if one postulates the commutation relations (A.61) between, and 7E,, since these relations imply [= ineHeT: (a.64) (T,, i] = ~ nenits, For certain quadratic Hamiltonians, it is equally possible to satisfy the same requirement by replacing the commurarors (4, 8] = AB ~ BA with the ant commutators [A, B],= AB + BA in (A61): { [a0 Tek) [400.10], 0 (4.65) aq Hh = Ke)WAZ Review of the Lagrangian and Hamiltonian Formalism 9 the other anticommutators being zero, For example, we show in Exercise 8 that the quantization of the Schrddinger equation, considered as the equation of motion of a classical field $(r), can be effected in a coherent fashion either with Commutators or with anticommutators. In both cases, the Heisenberg, equation for the quantum field ¥(r) associated with the classical field ¥(r) Hen = Erve.1) 66) hhas the form of a Schrodinger equation ~Bs even am into) =|— 35 Note that the rules concerning the measurement of physical quantities are ‘unchanged. For example, two physical quantities relative to the quantized field ‘an be measured simultaneously only if the corresponding operators commute, ‘whether the theory is quantized with the commutation relations (A.61) of with the anticommutation relations (A,65). However, it is necessary to mention here that the fields themselves are not necessarily physical variables. Thus, in the example of the quantization of the SchrOdinger equation with anticommutators, fone finds that it is not possible to give physical meaning to the operator Yr) {which has real and imaginary parts) as one can to an electric or magnetic fel, Only the quadratic Hermitian functions of ¥ represent physical quantities, to which one then applies the measurement postulates. For example, q¥*()¥(") Js the operator associated with the charge density at point r. The fact that ¥(") is not a physical variable renders les troublesome certain of its properties —for example, the fact that ¥(F) anticommutes with itself ii) Depending on whether the quantization of the field rests on commutators br anticommutators, the particles associated with the elementary excitations fof the quantized field are bosons or fermions (see for example Exercise 8). ‘When the field is relativistic, a link exists between the “spin” of the field and the statistics of the particles associated with it. Very general considerations (relativistic invariance, causality, positive energy) allow one to show that the {quantization of a relativistic field of integer spin can only occur in a satisfac- tory way (that is, without violating the principles above) if it depends on commutators. In contrast, if the spin is hall-integer, it is necessary to use “nticommutators (*). Thus, the electromagnetic field, which isa vector field and hhas spin 1, must be quantized with commutators, with the result that the particles associated with it are bosons, In contrast, the Dirac field has spin 3 End the particles associated with it (electrons and positrons) are fermions. Complement Ay of Chapter V gives an idea of the connection which exists in this case between the requirement for positive energy and the quantization by anticommutators. (6) This result is known a5 the “spin-statstis” theorem: itis due 10 W. Pauli, Pivs Rev, $8, 716 (1940),100 ‘Lagrangian and Hamiltonian Approach WBA B—THE STANDARD LAGRANGIAN OF CLASSICAL ELECTRODYNAMIC In this part we begin (§B.1) by giving the expression for the Lagrangian ‘most generally used in classical electrodynamics, and which we call “the standard Lagrangian”, We will show then (§B.2) that the Maxwell Lorentz equations arise naturally as the Lagrange equations for one such Lagrangian. Finally (§B.3) we analyze general properties of the standard Lagrangian, namely, symmetry properties, gauge invariance, and redun- daney of the dynamical field variables. 1, The Expression for the Standard Lagrangian a) THE STANDARD LAGRANGIAN IN REAL SPACE ‘The Lagrangian for the system made up of particles interacting with the electromagnetic field is given as a function of the dynamical variables relative to each of the subsystems, The dynamical variables of the particles form a discrete set involving the components of the position r, and of the velocity , for the particles denoted by the index a. For the electromag- netic field, it is the potentials and not the fields which appear as “good’ generalized coordinates in the Lagrangian formalism. This is not surpris- ing, since the equations of motion for the potentials are second order in time, as the Lagrange equations, while the Maxwell equations for the field are first order. At each point r, four generalized coordinates are required, these being the three components 4,(r) of the vector potential A(r) and the scalar potential U(r) and the four corresponding velocities ,(r) and U(e), so that the field dynamical variables are {AG), Ul): Ale), U)) forall (B1) The dynamics of the system particles + electromagnetic field can be derived from the standard Lagrangian L=Yimies 2 {er 'r[E%(r) — ¢? B%(r)] + + Dek AG) — 4, UlQ)) (B.2) the fields E and B being given as a function of the potentials A and U Bir) = = V(r) — Ay (B.3.a) Bir) = V x A(r) (B.3.b)IB. The Standard Lagrangian of Classical Electrodynamics 101 We will show below that this Lagrangian gives back the Maxwell-Lorentz equations, which will justify a posteriori the choice (B.2). This Lagrangian has three terms: the Lagrangian for the particles, Lp [first tetm of (B.2)| the Lagrangian of the electromagnetic field, Ly {second term of (B.2)) and the interaction Lagrangian L, flast term of (B.2)} Lp thy thy (8.4.2) Dhm, a (B.4.b) Laat fone - e BAe) (B.4.6) 1, =S lai, AG) - 4, UO] (B.4.4) By using the charge density p(r) and the current j(r) imtroduced in Chapter I [see Equations (A.5.a) and (A.5.b)], one can then rewrite L, in the form @rLik > AlN) = ple) Ue) (B.4.2) Finally, regrouping (B.4.c) and (B.4.e) leads to the introduction of the Lagrangian density 2. Lee) = 72 [R%E) ~ 2 BAW] +E)“ AW — pH) UI) (B.5.a) and the following form for the standard Lagrangian L =Yhme + fare (B.5.) Note that the interaction term (B.4.) is local; the current density (or the charge density) at point ris muliplied by the vector (or scalar) potential at the same point. In the field Lagrangian (Bc). spatial Gerivatives of the potentials arise through E and B, which expresses a coupling between the field variables from point to point. This coupling is the origin of the propagation of the free fel. b) THE STANDARD LAGRANGIAN IN RECIPROCAL SPACE We have seen in Chapter 1 that the Maxwell equations are much simpler in reciprocal space. In the same way, itis interesting to express the standard Lagrangian as a function of the potentials in reciprocal space.102 Lagrangian and Hlaniltonian Approach mp4 The Parseval-Plancherel equality allows one to rewrite (B.S) in the form Lit 4 0 [as 2 lady P Layimi + #e K[LE) Pe Law P] + + fertjray 109 = pk) uw)] BH) Equation (B.6) suggests choosing as dynamical variables the compo- nents of the potentials in reciprocal space as well as their velocities. However, it is necessary to take several precautions: going from real space to reciprocal space corresponds to a change of variables which transforms real quantities into complex quantities. The new variables then have twice as many degrees of freedom as the old variables. But there are constraint relationships tied to the fact that A(r) and U(®) are real: Ak) = Sk) (B.7.a) ak) = 4*(— k) B.7.b) If the potentials are known in a “reciprocal half space”, they are known everywhere. One is then led to take as independent variables the potentials and their complex conjugates in only half of reciprocal space. The equali- ties 8(— WE) = SM) (ky (B.8.a) P= Wy ol (— ky =/(k) +o 1k) (B.8.b) which follow from (B.7), allow the rewriting of the Lagrangian (B.6) as a function of the field variables in a half space. Denoting by fa2k the integral extended over a haf volume ofthe reciprocal space and by the Lagrangian density in the reciprocal space, one gets aKe (B.9.a) Aw PL + sof Mk) — pthc) Uk) — p(k) UK] (B.9.b) + (7h) f+ / or, again expressing & and 4 as functions of sf and 4, @= ~ of —ikw (B.10.a) Baik x of (B.10.b)WB ‘The Standard Lagrangian of Classical Electrodynamics, 103 cof oh (ky + ikAK) P — Lk x ok) P] + + [7A = 90 (k) +7) F*CK) — p(k) WA) — p(k) WG]. (B.11) This new form, equivalent to the standard Lagrangian, presents certain advantages. First of all, the Lagrangian density is strictly local in k. The derivatives of sf and & with respect to k do not appear (there is no coupling between neighboring points as in real space). Additionally, in (B.9), the contribution of the various modes of the field appear explicitly AAs we shall see below, itis then very easy to separate the contributions of the nonrelativistic modes, or those of the long-wavelength modes, for which an electric dipole approximation is possible. 2. The Derivation of the Classical Electrodynamic Equations from the Standard Lagrangian 4) LAGRANGE'S EQUATION FOR PARTICLES Since the particle variables are discrete, we apply Lagrange’s equation (A.2) to the standard Lagrangian (B.2). One calculates first 32 /9(e.), and aL /att,), fh gue +f Ate) (BID) Gey a DE egies which, using the vector identity VA +B) = (B-V)A 4 (A+ V)B+Bx (Vx A) FA «IV xB) (3.13) becomes eb ey . he iw, GAR, U(r, 1) + 4,(F, VY) Alt. 0 agli, x (VAC. )),. (B14) In addition ke mi + gy Alt (B.15) The Lagrange equation describing the motion of particle a is gotten by differentiating (B.15) with respect to time ae di Ae), = MF) + IS AME + ali, WAM) (B16)104 ‘Lagrangian and Hamiltonian Approach LB and by setting that expression equal to (B.14). One gets finally ani ; e af M29) wyen i) bagi, x (WA Ue 0) BIT.) m8, = 4, Ele) + dei, x Br) (B.17.b) which is the Lorentz equation. b) Tite LAGRANGE EQUATION RELATIVE TO THE SCALAR POTENTIAL For the equations relative to the field, one can use the Lagrangian density in real space or in reciprocal space. Here we take the second ‘option, since it gives the quickest result. Starting with (B.11), one gets To = - toik> (B.18) In addition, since #* does not appear in 2, 0 (B19 cn ‘The Lagrange equation (A.52.b) is then written as = ike bal + ihe] = 2 (B.20) which is finally ik (B.21) and is nothing but one of Maxwell's equations written in reciprocal space [see (B.5.a) of Chapter 1] ©) THE LAGRANGE EQUATION RELATIVE 10 THE VECTOR POTENTIAL Starting with (B.11) for F and using the identity tk x of) (kx of) = [(k x of) x kK] (B.22) one can derive ez tg c[k x(k x AY}, +4, (B.23.a)1B3 ‘The Standard Lagrangian of Classical Electrodynamics 105 In addition solol, + ik; ) (8.23.6) ‘The Lagrange equation relative to «/,* is then = [ik x(k x) + J, (B24) which is finally, using (B.10), ikx @ oa (B.25) ‘One has here another of Maxwells (B.d) of Chapter I]. In conclusion, the application of the principle of least action to the standard Lagrangian has given us on the one hand the Lorentz equation for a particle in an electromagnetic field, and on the other the second pair of Maxwell equations which relates the fields to their sources, (The first pair of Maxwell equations results directly from Equations (B.10) relating the fields & and @ to the potentials «/ and %) equations in reciprocal space [see 3. General Properties of the Standard Lagrangian 2) GlosaL SyMMereies The form of the Lagrangian is invariant under certain geometric transformations: translation and rotation with respect to the system of axes to which the particles and the field are referred. The Lagrangian is also invariant under a change of the time origin. From these invariance properties it is possible to derive expressions for a certain number of Conserved quantities, namely, the momentum, the angular momentum, and the total energy of the system field + particles. (This is done in Complement By,, on the form which the standard Lagrangian takes in the Coulomb gauge.) ‘The standard Lagrangian (B.2) does not transform simply under a Lorentz transformation. Indeed, itis clear that the standard Lagrangian does not treat the particles in a relativistic way. the Lagrangian of the Particles, equal to 5,m,é2/2, being purely Galilean, We are now going to show that the Lagrangian (B.2) can be gotten in the classical low-velocity limit (o/c * 1), starting from a relativistic Lageangian, that is, one with a relativistically invariant action. We note initially that the Lagrangian density of the electromagnetic field is a relativistic scalar field, Indeed, it is a function of the electromag-106 ‘Lagrangian and Hamiltonian Approach mB3 netic tensor field F*" [see Chapter I, Equation (B.28)] of the following form: ‘a DF FM (B.26) °F which is manifestly invariant under a Lorentz transformation. The contri- bution to the action of the Lagrangian density of the free field is written se [u forse on Now, %g on one hand and the volume element drd’r on the other are relativistic invariants. Itis clear then that the action Sq is a relativistic invariant. We will show now thatthe interaction Lagrangian between the particles, and the field contributes equally to the action in a covariant fashion. For this it is sufficient to note that the infinitesimal variation of the action relative to the interaction of particle a with the field arises as the scalar product of the four-vector dx with the four-potential 4, Lydr = Lalde, Me) — dr Ue] (B28) - Ya, det 4, (B.29) as, ds, = Finally, it sufices to transform the Lagrangian of the particles, Lp, to get a relativistic Lagrangian, To this end, we replace Ly by ws -tme fi ‘The infinitesimal variation of the action corresponding to (B,30) is then written, (B.30) dS = Lytdr = — Ym, Lm cdr, (BI) where dr, (B.32) is the proper time of the particle a. Since d7, and dS are relativistic invariants, the Lagrangian (B.30) is also a relativistic Lagrangian. Ad tionally, expansion of (B.30) in powers of #2/c? gives. to within a constant term —E,m,c*, the Lagrangian Ly given in (B.4.b),1LB3 ‘The Standard Lagrangian of Classical Electrodynamics 107 In conclusion, it appears possible to introduce a fully relativistic Lagrangian ES [een eaten + + D LaF Mtr) — 4, Ute] (B.33) which can serve as the basis for classical electrodynamics. However, if one proceeds in this fashion, difficulties arise in quantization of the theory, primarily as a result of the impossibility of constructing a relativistic quantum theory for a fixed number of particles. Remarks ( Teturns out that the correct procedure for constructing a relativistic quan: tum theory involves starting from a classical theory where the particles are described, like radiation, as a relativistic field (Klein-Gordon field, Dirac field, etc.) coupled to the Maxwell field. Then when such a theory is quantized, the particles, indeterminate in number, appear as clementary excitations of the ‘quantum matter field and interact with the photons, which are the elementary excitations of the quantized Maxwell field (see Complement Ay {i)_It is possible to justify the use of the standard nonrelativistic Lagrangian (B.2), and, as a result, of the Hamiltonian in the Coulomb gauge which we will derive below, by starting from relativistic quantum electrodynamics and exam: ining the low-energy limit of this theory. One finds to the lowest order in w/e the dynamics described by (B.2). One also gets the interaction terms tied to the spins of the particles (see Complement By) b) GauGe INVARIANCE The Maxwell-Lorentz theory of electrodynamics is manifestly invariant under a change of gauge, since only the electric and magnetic fields appear in the basic equations. Gauge invariance is less evident for the Lagrangian theory, which uses the potentials as variables to describe the field. It is, thus appropriate to examine the consequences of a gauge change in the Lagrangian formalism, Following Equations (A.12) from Chapter 1, a gauge change is defined by AG) = AUD + VEE, (B.34.a) vin = UEn- FF (B.34,b)108 ‘Lagrangian and Hamiltonian Approach 1B3 F can be an explicit function of r and f, but can also depend on the field variables, which are themselves functions of r and In the transformation (B.34), the Lagrangian of the particles is evi- dently not modified; nor is the Lagrangian of the field, which depends only on the electric and magnetic fields. Only the interaction Lagrangian is changed. The gauge change amounts to adding to the Lagrangian density # of the field given by (B.5.a) the quantity oF L=j VF +0F (B35) which one can write in the form var) + Sn -(vi+2)r (8.36) The fist two terms add to the Lagrangian density a divergence and a time derivative. According to (A.40) this transforms the Lagrangian into an equivalent Lagrangian (see however the Remark below). As for the last term of (B.36), itis zero as a result of charge conservation. It then appears clear that charge conservation is a necessary condition for gauge invari- ance. Remarks (@ There is not total equivalence between the changes in the Lagrangian and the gauge transformations. For example, in (B34), F can depend on A, U, A, and U, which are themselves functions of rand ¢. All transformations which Jeave Maxwell's equations and the fields F and B invariant are gauge transfor ‘mations, On the other hand, itis only when F docs not depend on the velocities Avand U that it also corresponds to a change in the Lagrangian, since otherwise the accelerations A and ( would appear in the Lagrangian. Conversely, the changes in the Lagrangian density defined by (A.40) do not nocessarily corte: spond to a gauge transformation. By comparing (A.40) and (B.36), one sees that for that it is necessary that a function F exist such that (Bata (B.37.b) Now itis not possible in general to satisfy both these conditions, (Gi) Tn the gauge field theories, gauge invariance plays a much more fundamen: tal role, Starting with the fields representing particles, ome requires that the ‘theory be invariant under a local change of phase of the fields. To realize this,uB3 ‘The Standard Lagrangian of Classical Electrodynamics 109 invariance, it appears necessary to introduce a vector field (the clectromagnetic field) coupled to the fled of the particles in such a way that the phase changes in the matter field entail gauge transformations of the vector field (see Exercise 9). One introduces in this way a fundamental relationship between the change of phase of the matter field and the gauge change of the electromagnetic field. ¢) REDUNDANCY OF THE DYNAMICAL VARIABLES In the description of electrodynamics through the standard Lagrangian, the field is described at each point r by the potentials A and U and the corresponding velocities (B.1). Thus, the dynamical variables are eight in number at each point in space. Now the approach to electrodynamics in Chapter I, resting on the Maxwell-Lorentz equations, introduces six degrees of freedom for each point [the three components of the electric and magnetic fields E(r) and B(r)). Besides this, writing the Maxwell equations in reciprocal space allows one to show that the longitudinal components 6)(k) and @\(k) are fixed by algebraic equations [Equations (BS.a) and (BS) from Chapter I]; the evolution of the four other dynamical variables (the transverse electric and magnetic fields) is de- scribed by differential equations which are first order in time [Equations (B.49.a) and (B.49.b) of Chapter I] I is thus evident that in describing the electromagnetic field by the potentials sf and @% one has introduced an overabundance of degrees of freedom. Thus constraint relations must exist between the dynamical field variables We now examine how these constraints appear. An analysis of the Lagrangian density @ in (B.11) shows that % does not appear in this, Lagrangian density. This implies, on one hand, that the conjugate momen- tum associated with the variable @ is identically zero, and on the other hhand, that the Lagrange equation (B.20) associated with % relates % to the other dynamical variables by an algebraic equation. This type of problem has already been considered in §A.1.e. When the velocity associ- ated with a generalized coordinate does not appear in the Lagrangian, this coordinate can be eliminated by expressing it as a function of the other dynamical variables, giving a reduced Lagrangian, Here such a step allows the elimination of the scalar potential #7, and one gets a Lagrangian where only the three components of the vector potential of and their time derivatives appear. One can further reduce the number of degrees of freedom of the electromagnetic field through the choice of gauge. It follows from equations (B.8.a) and (B.26) of Chapter I that a choice of gauge amounts to fixing the longitudinal component of the vector poten tial of, which is otherwise arbitrary. This then leads to a satisfactory Physical situation where the field has at cach point four independent Physical variables which correspond to the two transverse orthogonal110 Lagrangian and Hamiltonian Approach 1.B3 se derivatives ;(k) and /,(k), Remarks {G) 9,04) is taken in a reciprocal half space and satisfies fk) = € (ky (B.38) where € is one of the two (real) transverse vectors. In the other half space we define off Ak) (B.39) ‘where € isthe same vector for k and —k Gi) Since the fields in reciprocal space are complex, one could imagine that the ‘component .9/(k), for example, corresponds to two real degrees of freedom (the real and imaginary pars) In fact, since A— k= AW = 8AM) = 0K) (B40) fone has for the set of points k and —K two real degrees of freedom, ie. one at each point The step described above results in a reduced Lagrangian where the field is described only by four dynamical variables. and the symmetry between the four components of the four-potential in the standard La- ‘grangian is now destroyed. It is of course tempting to try to quantize the theory without going through the reduced Lagrangian, keeping the four components of the four-potential as independent variables. However, such ‘a procedure is impossible if one starts from the standard Lagrangian, since the conjugate momentum 7, associated with ® is identically zero accord- ing to (B.19), Ibis thus impossible to impose upon the operators associated with @ and 7, the canonical commutation relations (A.61.b). The conser- vation of symmetry between the four components of the four-potential is then possible only through the use of another Lagrangian (see Chapter V), ‘The natural step to quantize the theory starting from the standard Lagrangian consists then in eliminating & to get a reduced Lagrangian, and then choosing a gauge by fixing .o/,. The simplest possible choice corresponds to the Coulomb gauge (see ‘Chapter 1, §A.3). One is then naturally Jed to study electrodynamics in the Coulomb gauge. Other choices of gauge corresponding to other values of .f, can of course be considered (examples of this are given in Chapter IV).mca Electrodynamics in the Coulomb Gauge cr (C—ELECTRODYNAMICS IN THE COULOMB GAUGE, In this final part, we will show how one eliminates the redundant dynamical variables in the standard Lagrangian (§C.1). This will lead to the Lagrangian in the Coulomb gauge, the properties of which will be examined in §C.2. We will then pass to the Hamiltonian formalism (§C.3) and to the canonical quantization of the theory (§C.4). Finally we will discuss the important characteristics of this theory (§C.5). 1. Elimination of the Redundant Dynamical Variables from the Standard Lagrangian 4) ELIMINATION OF THE SCALAR POTENTIAL Following the route sketched in §B.3.c, we will use Lagrange’s equation relative to 4 to express the scalar potential as a function of the other dynamical variables and thus get a Lagrangian depending on a smaller number of degrees of freedom. Lagrange’s equation (B.20) allows one to write oy +2] c.1) where A 2 By replacing % with (C.l) in the standard Lagrangian (B.11), one gets a Lagrangian depending on a reduced number of dynamical variables (the components of the veotor potential and the associated velocities) which we still call L 2M x ok «of, | (or odp)| C3) In the same way, one eliminates % from the expressions for all the physical variables which depend on it. Thus, the electric field in reciprocal space (B.10.a) is now written as Bik) = — A) p cay 0 It depends on the field variables (.0/ ,) and the positions of the particles (which appear in p).12 Lagrangian and Hamiltonian Approach met Returning to the Lagrangian (C.3) and grouping the terms, one gets OR At ah) + 3p OD Bead 8 so faneds fort gn fone fern — herd ~ ress] c.5) ; fear A, +f ofS ee 6) THE CHOICE OF THE LONGITUDINAL COMPONENT OF THE VECTOR POTENTIAL The longitudinal component # of appears only in the density 2, arising in the last term of (C.5): By nit, thy alt ~ pot oh ~ palsy c.6) The Lagrange equation for s, derived from (C.6) is written b — kj (c.7) and is just the well-known equation for charge conservation (in reciprocal space). Clearly, this is not an equation of motion for ,, so that wi, can take any value. This last point is even clearer if the equation of charge conservation (C.1) is used to express J, as a function of 4. One then finds that (C.6) can be written Liber + asd = pth, = joy Sten ara le kage a which gives for the Lagrangian (C.5) be feels he Ralf) + ctf +a fe Mk Eleaf ~ po). (C.9) Since two Lagrangians which differ only in a total time derivative of a function of the coordinates are equivalent, it appears that the evolution ofwc2z Electrodynamics in the Coulomb Gauge 13 the system does not depend on the value of ., which appears exclusively jn such a total derivative, fj is not a irue dynamical variable, since its value can be arbitrarily chosen without changing the dynamics of the system, Remark The possiblity of choosing ., arbitrarily is evidently related to gauge invari ‘ance. On changing the gauge, does not change and .#/, beoomes [sce (B8), Chapter I) Ai hy FF «19 (On changing the gauge, only the last term of the Lagrangian (C.9) is changed, and this is @ total derivative with respect to time. ‘One can imagine various possible choices for the longitudinal compo- nent of the vector potential. The simplest choice obviously is vy =0 cay which requires that V - A be zero in the entire real space and thus selects the Coulomb gauge. Starting from this point, unless otherwise stated we will work in the Coulomb gauge, where the vector potential is purely transverse: of =o, cy ‘To simplify the notation, we will henceforth omit the index L 2. The Lagrangi in the Coulomb Gauge ‘The Lagrangian in reciprocal space in the Coulomb gauge derives from (C9): ee pine feng Fo = edd od — AR at Al +p tft, (C.13.b) ‘The dynamical variables of the particles, r, and #,, appear not only in the term S.,m,i2/2 but also in the charge density p and in the current The second terin of (C.13.a) can be transformed into an integral overall Space thanks to the reality condition p(k) ~ p*(—k). One then finds precisely